2019-05-27 08:55:05 +02:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:41 -06:00
|
|
|
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 17:04:11 +09:00
|
|
|
|
2017-02-08 18:51:30 +01:00
|
|
|
|
2011-05-27 09:47:43 -04:00
|
|
|
|
2024-10-01 15:35:57 -04:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-11-21 12:04:43 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-11-21 12:04:43 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-11-21 12:04:43 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-11-21 12:04:43 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
|
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
2022-06-16 17:45:50 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-09-05 07:34:23 +05:30
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2014-02-07 00:41:39 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-09-05 07:34:23 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
2014-06-25 15:27:36 +02:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
2020-08-23 17:36:59 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2011-05-27 11:16:33 +00:00
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:39 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-05-27 11:16:33 +00:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:14 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:35 -06:00
|
|
|
|
|
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:35 -06:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-04-18 10:11:51 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2014-03-05 19:43:49 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:35 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-08-31 18:09:24 -04:00
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:24 -04:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:35 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-04-24 16:21:19 -05:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-08-20 15:10:58 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2007-12-13 12:43:35 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
2019-01-29 09:33:07 +01:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2014-06-11 12:09:58 +03:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:33 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:50 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
2021-10-07 13:28:39 -07:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2022-06-16 17:45:56 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-11-23 15:29:21 -08:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2017-03-09 13:46:58 +02:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2010-11-23 15:29:21 -08:00
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2022-06-16 17:45:56 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:50 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
|
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-09-05 07:34:23 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-09-05 07:34:23 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:54 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:54 -05:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-09-05 07:34:23 +05:30
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2019-01-29 09:33:07 +01:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:15 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2022-04-07 19:13:12 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-09-03 00:00:39 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-11-11 16:34:31 -06:00
|
|
|
|
2009-05-13 17:57:43 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-09-03 00:00:39 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-03-09 13:46:58 +02:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-05-13 17:57:39 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2009-05-13 17:57:39 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2009-05-13 17:57:39 -05:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2020-10-23 18:33:14 +02:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2011-05-27 11:16:33 +00:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-03-05 19:43:49 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-04-27 10:30:13 +02:00
|
|
|
|
2015-07-15 10:55:39 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-03-05 19:43:49 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-12-17 12:10:28 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:20 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-12-17 12:10:28 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-04-18 10:11:52 -05:00
|
|
|
|
|
|
|
|
|
2019-01-29 09:33:07 +01:00
|
|
|
|
2008-04-18 10:11:52 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:22 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:22 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-04-18 10:11:52 -05:00
|
|
|
|
2007-06-14 22:13:17 +09:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-04-18 10:11:52 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:42 -05:00
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-29 09:33:07 +01:00
|
|
|
|
2019-02-08 13:24:59 -08:00
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
|
|
|
|
|
2015-09-03 19:49:55 +03:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
2015-09-03 19:49:55 +03:00
|
|
|
|
|
|
|
|
|
2008-12-01 12:13:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-09-03 19:49:55 +03:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-01-09 10:21:34 +08:00
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-01-09 10:21:34 +08:00
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-08-25 13:46:37 -07:00
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-01-09 10:21:34 +08:00
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
2009-05-13 17:57:41 -05:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:41 -05:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-02-09 16:39:40 +00:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-02-09 16:39:40 +00:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:45 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-07-24 15:47:45 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-02-09 16:39:40 +00:00
|
|
|
|
2006-07-24 15:47:45 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] iscsi_tcp: fix padding, data digests, and IO at weird offsets
iscsi_tcp calculates padding by using the expected transfer length. This
has the problem where if we have immediate data = no and initial R2T =
yes, and the transfer length ended up needing padding then we send:
1. header
2. padding which should have gone after data
3. data
Besides this bug, we also assume the target will always ask for nice
transfer lengths and the first burst length will always be a nice value.
As far as I can tell form the RFC this is not a requirement. It would be
silly to do this, but if someone did it we will end doing bad things.
Finally the last bug in that bit of code is in our handling of the
recalculation of data digests when we do not send a whole iscsi_buf in
one try. The bug here is that we call crypto_digest_final on a
iscsi_sendpage error, then when we send the rest of the iscsi_buf, we
doiscsi_data_digest_init and this causes the previous data digest to be
lost.
And to make matters worse, some of these bugs are replicated over and
over and over again for immediate data, solicited data and unsolicited
data. So the attached patch made over the iscsi git tree (see
kernel.org/git for details) which I updated today to include the patches
I said I merged, consolidates the sending of data, padding and digests
and calculation of data digests and fixes the above bugs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2006-08-31 18:09:27 -04:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-05-02 19:46:47 -05:00
|
|
|
|
2006-10-16 18:09:41 -04:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:09 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-08-20 15:10:59 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:13 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:06 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:54:04 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:46:05 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:03 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:03 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:17:55 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:17:55 -05:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:17:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-07-26 12:46:48 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:56 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-02-13 13:21:39 +08:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2022-06-16 17:45:56 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:56 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2009-08-20 15:11:03 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2009-08-20 15:11:03 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2022-06-07 16:19:53 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2021-02-06 22:46:00 -06:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2021-02-06 22:46:00 -06:00
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-12-13 12:43:23 -06:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2009-08-20 15:11:03 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2022-06-07 16:19:53 +03:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-08-20 15:11:03 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2006-11-22 14:57:56 +00:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-11-22 14:57:56 +00:00
|
|
|
|
|
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-05-30 00:37:28 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2017-03-09 13:46:58 +02:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-03-05 19:43:49 +02:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2026-01-15 13:03:41 -08:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:47 -05:00
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:32 -06:00
|
|
|
|
2009-05-13 17:57:47 -05:00
|
|
|
|
scsi: libiscsi: Allow sd_shutdown on bad transport
If, for any reason, userland shuts down iscsi transport interfaces
before proper logouts - like when logging in to LUNs manually, without
logging out on server shutdown, or when automated scripts can't
umount/logout from logged LUNs - kernel will hang forever on its
sd_sync_cache() logic, after issuing the SYNCHRONIZE_CACHE cmd to all
still existent paths.
PID: 1 TASK: ffff8801a69b8000 CPU: 1 COMMAND: "systemd-shutdow"
#0 [ffff8801a69c3a30] __schedule at ffffffff8183e9ee
#1 [ffff8801a69c3a80] schedule at ffffffff8183f0d5
#2 [ffff8801a69c3a98] schedule_timeout at ffffffff81842199
#3 [ffff8801a69c3b40] io_schedule_timeout at ffffffff8183e604
#4 [ffff8801a69c3b70] wait_for_completion_io_timeout at ffffffff8183fc6c
#5 [ffff8801a69c3bd0] blk_execute_rq at ffffffff813cfe10
#6 [ffff8801a69c3c88] scsi_execute at ffffffff815c3fc7
#7 [ffff8801a69c3cc8] scsi_execute_req_flags at ffffffff815c60fe
#8 [ffff8801a69c3d30] sd_sync_cache at ffffffff815d37d7
#9 [ffff8801a69c3da8] sd_shutdown at ffffffff815d3c3c
This happens because iscsi_eh_cmd_timed_out(), the transport layer
timeout helper, would tell the queue timeout function (scsi_times_out)
to reset the request timer over and over, until the session state is
back to logged in state. Unfortunately, during server shutdown, this
might never happen again.
Other option would be "not to handle" the issue in the transport
layer. That would trigger the error handler logic, which would also need
the session state to be logged in again.
Best option, for such case, is to tell upper layers that the command was
handled during the transport layer error handler helper, marking it as
DID_NO_CONNECT, which will allow completion and inform about the
problem.
After the session was marked as ISCSI_STATE_FAILED, due to the first
timeout during the server shutdown phase, all subsequent cmds will fail
to be queued, allowing upper logic to fail faster.
Signed-off-by: Rafael David Tinoco <rafael.tinoco@canonical.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-12-07 19:59:13 -02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-08-23 17:36:59 -05:00
|
|
|
|
2007-12-13 12:43:32 -06:00
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2009-05-13 17:57:47 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:32 -06:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:47 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2008-08-19 18:45:26 -05:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
2007-12-13 12:43:26 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-10-16 18:09:39 -04:00
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
2006-10-16 18:09:39 -04:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
2017-10-09 13:33:19 +02:00
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:25 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-08-20 15:10:58 -05:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2008-05-21 15:54:05 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2008-08-17 15:24:43 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2010-12-31 02:22:21 -06:00
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
|
|
|
|
|
2019-01-29 09:33:07 +01:00
|
|
|
|
2021-10-07 13:28:39 -07:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-04-21 15:32:32 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-10-11 16:25:40 -07:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2025-05-09 07:51:14 +02:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2018-07-30 14:40:18 -07:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:57 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2010-12-31 02:22:18 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2025-04-05 10:17:26 +02:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2014-06-25 15:27:36 +02:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:50 -05:00
|
|
|
|
2009-09-05 07:35:33 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:03 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
2009-05-13 17:57:48 -05:00
|
|
|
|
2022-06-16 17:45:50 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2008-05-21 15:54:03 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
2022-06-16 17:45:49 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:50 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2019-12-09 09:34:57 -08:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2012-01-26 21:13:11 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2012-01-26 21:13:11 -06:00
|
|
|
|
|
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2012-01-26 21:13:11 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
scsi: libiscsi: Allow sd_shutdown on bad transport
If, for any reason, userland shuts down iscsi transport interfaces
before proper logouts - like when logging in to LUNs manually, without
logging out on server shutdown, or when automated scripts can't
umount/logout from logged LUNs - kernel will hang forever on its
sd_sync_cache() logic, after issuing the SYNCHRONIZE_CACHE cmd to all
still existent paths.
PID: 1 TASK: ffff8801a69b8000 CPU: 1 COMMAND: "systemd-shutdow"
#0 [ffff8801a69c3a30] __schedule at ffffffff8183e9ee
#1 [ffff8801a69c3a80] schedule at ffffffff8183f0d5
#2 [ffff8801a69c3a98] schedule_timeout at ffffffff81842199
#3 [ffff8801a69c3b40] io_schedule_timeout at ffffffff8183e604
#4 [ffff8801a69c3b70] wait_for_completion_io_timeout at ffffffff8183fc6c
#5 [ffff8801a69c3bd0] blk_execute_rq at ffffffff813cfe10
#6 [ffff8801a69c3c88] scsi_execute at ffffffff815c3fc7
#7 [ffff8801a69c3cc8] scsi_execute_req_flags at ffffffff815c60fe
#8 [ffff8801a69c3d30] sd_sync_cache at ffffffff815d37d7
#9 [ffff8801a69c3da8] sd_shutdown at ffffffff815d3c3c
This happens because iscsi_eh_cmd_timed_out(), the transport layer
timeout helper, would tell the queue timeout function (scsi_times_out)
to reset the request timer over and over, until the session state is
back to logged in state. Unfortunately, during server shutdown, this
might never happen again.
Other option would be "not to handle" the issue in the transport
layer. That would trigger the error handler logic, which would also need
the session state to be logged in again.
Best option, for such case, is to tell upper layers that the command was
handled during the transport layer error handler helper, marking it as
DID_NO_CONNECT, which will allow completion and inform about the
problem.
After the session was marked as ISCSI_STATE_FAILED, due to the first
timeout during the server shutdown phase, all subsequent cmds will fail
to be queued, allowing upper logic to fail faster.
Signed-off-by: Rafael David Tinoco <rafael.tinoco@canonical.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-12-07 19:59:13 -02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
scsi: libiscsi: Allow sd_shutdown on bad transport
If, for any reason, userland shuts down iscsi transport interfaces
before proper logouts - like when logging in to LUNs manually, without
logging out on server shutdown, or when automated scripts can't
umount/logout from logged LUNs - kernel will hang forever on its
sd_sync_cache() logic, after issuing the SYNCHRONIZE_CACHE cmd to all
still existent paths.
PID: 1 TASK: ffff8801a69b8000 CPU: 1 COMMAND: "systemd-shutdow"
#0 [ffff8801a69c3a30] __schedule at ffffffff8183e9ee
#1 [ffff8801a69c3a80] schedule at ffffffff8183f0d5
#2 [ffff8801a69c3a98] schedule_timeout at ffffffff81842199
#3 [ffff8801a69c3b40] io_schedule_timeout at ffffffff8183e604
#4 [ffff8801a69c3b70] wait_for_completion_io_timeout at ffffffff8183fc6c
#5 [ffff8801a69c3bd0] blk_execute_rq at ffffffff813cfe10
#6 [ffff8801a69c3c88] scsi_execute at ffffffff815c3fc7
#7 [ffff8801a69c3cc8] scsi_execute_req_flags at ffffffff815c60fe
#8 [ffff8801a69c3d30] sd_sync_cache at ffffffff815d37d7
#9 [ffff8801a69c3da8] sd_shutdown at ffffffff815d3c3c
This happens because iscsi_eh_cmd_timed_out(), the transport layer
timeout helper, would tell the queue timeout function (scsi_times_out)
to reset the request timer over and over, until the session state is
back to logged in state. Unfortunately, during server shutdown, this
might never happen again.
Other option would be "not to handle" the issue in the transport
layer. That would trigger the error handler logic, which would also need
the session state to be logged in again.
Best option, for such case, is to tell upper layers that the command was
handled during the transport layer error handler helper, marking it as
DID_NO_CONNECT, which will allow completion and inform about the
problem.
After the session was marked as ISCSI_STATE_FAILED, due to the first
timeout during the server shutdown phase, all subsequent cmds will fail
to be queued, allowing upper logic to fail faster.
Signed-off-by: Rafael David Tinoco <rafael.tinoco@canonical.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-12-07 19:59:13 -02:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2010-02-10 16:51:45 -06:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2020-11-06 11:33:17 -08:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
2009-06-15 22:11:08 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2019-12-09 09:34:57 -08:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-10-18 13:29:50 -07:00
|
|
|
|
scsi: libiscsi: Allow sd_shutdown on bad transport
If, for any reason, userland shuts down iscsi transport interfaces
before proper logouts - like when logging in to LUNs manually, without
logging out on server shutdown, or when automated scripts can't
umount/logout from logged LUNs - kernel will hang forever on its
sd_sync_cache() logic, after issuing the SYNCHRONIZE_CACHE cmd to all
still existent paths.
PID: 1 TASK: ffff8801a69b8000 CPU: 1 COMMAND: "systemd-shutdow"
#0 [ffff8801a69c3a30] __schedule at ffffffff8183e9ee
#1 [ffff8801a69c3a80] schedule at ffffffff8183f0d5
#2 [ffff8801a69c3a98] schedule_timeout at ffffffff81842199
#3 [ffff8801a69c3b40] io_schedule_timeout at ffffffff8183e604
#4 [ffff8801a69c3b70] wait_for_completion_io_timeout at ffffffff8183fc6c
#5 [ffff8801a69c3bd0] blk_execute_rq at ffffffff813cfe10
#6 [ffff8801a69c3c88] scsi_execute at ffffffff815c3fc7
#7 [ffff8801a69c3cc8] scsi_execute_req_flags at ffffffff815c60fe
#8 [ffff8801a69c3d30] sd_sync_cache at ffffffff815d37d7
#9 [ffff8801a69c3da8] sd_shutdown at ffffffff815d3c3c
This happens because iscsi_eh_cmd_timed_out(), the transport layer
timeout helper, would tell the queue timeout function (scsi_times_out)
to reset the request timer over and over, until the session state is
back to logged in state. Unfortunately, during server shutdown, this
might never happen again.
Other option would be "not to handle" the issue in the transport
layer. That would trigger the error handler logic, which would also need
the session state to be logged in again.
Best option, for such case, is to tell upper layers that the command was
handled during the transport layer error handler helper, marking it as
DID_NO_CONNECT, which will allow completion and inform about the
problem.
After the session was marked as ISCSI_STATE_FAILED, due to the first
timeout during the server shutdown phase, all subsequent cmds will fail
to be queued, allowing upper logic to fail faster.
Signed-off-by: Rafael David Tinoco <rafael.tinoco@canonical.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-12-07 19:59:13 -02:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2017-01-30 13:18:58 +01:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2017-10-11 16:25:40 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2025-05-09 07:51:14 +02:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2008-05-07 20:43:52 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-07 20:43:52 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2008-05-07 20:43:52 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
2009-05-13 17:57:45 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2014-07-12 15:51:51 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-07 20:43:52 -05:00
|
|
|
|
2008-05-08 20:15:34 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
2015-09-03 19:49:55 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:50 -06:00
|
|
|
|
2008-05-07 20:43:52 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
2008-01-31 13:36:50 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:17:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:12 -05:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:17:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
2008-12-02 00:32:05 -06:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:33 -04:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2016-03-30 11:27:08 -07:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2009-05-13 17:57:50 -05:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-08-31 18:09:33 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-08-31 18:09:33 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
2009-05-13 17:57:50 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2022-02-18 11:50:53 -08:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
|
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
|
|
|
|
|
2021-10-04 16:06:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
2021-10-04 16:06:08 -05:00
|
|
|
|
2022-06-16 17:45:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-04 16:06:08 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2016-03-30 11:27:08 -07:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-05-25 13:18:10 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2010-12-31 02:22:18 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:08 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
2020-08-23 17:36:59 -05:00
|
|
|
|
2006-07-24 15:47:22 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
scsi: libiscsi: Fix iscsi_task use after free()
The following bug was reported and debugged by wubo40@huawei.com:
When testing kernel 4.18 version, NULL pointer dereference problem occurs
in iscsi_eh_cmd_timed_out() function.
I think this bug in the upstream is still exists.
The analysis reasons are as follows:
1) For some reason, I/O command did not complete within the timeout
period. The block layer timer works, call scsi_times_out() to handle I/O
timeout logic. At the same time the command just completes.
2) scsi_times_out() call iscsi_eh_cmd_timed_out() to process timeout logic.
Although there is an NULL judgment for the task, the task has not been
released yet now.
3) iscsi_complete_task() calls __iscsi_put_task(). The task reference count
reaches zero, the conditions for free task is met, then
iscsi_free_task() frees the task, and sets sc->SCp.ptr = NULL. After
iscsi_eh_cmd_timed_out() passes the task judgment check, there can still
be NULL dereference scenarios.
CPU0 CPU3
|- scsi_times_out() |-
iscsi_complete_task()
| |
|- iscsi_eh_cmd_timed_out() |-
__iscsi_put_task()
| |
|- task=sc->SCp.ptr, task is not NUL, check passed |-
iscsi_free_task(task)
| |
| |-> sc->SCp.ptr
= NULL
| |
|- task is NULL now, NULL pointer dereference |
| |
\|/ \|/
Calltrace:
[380751.840862] BUG: unable to handle kernel NULL pointer dereference at
0000000000000138
[380751.843709] PGD 0 P4D 0
[380751.844770] Oops: 0000 [#1] SMP PTI
[380751.846283] CPU: 0 PID: 403 Comm: kworker/0:1H Kdump: loaded
Tainted: G
[380751.851467] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
[380751.856521] Workqueue: kblockd blk_mq_timeout_work
[380751.858527] RIP: 0010:iscsi_eh_cmd_timed_out+0x15e/0x2e0 [libiscsi]
[380751.861129] Code: 83 ea 01 48 8d 74 d0 08 48 8b 10 48 8b 4a 50 48 85
c9 74 2c 48 39 d5 74
[380751.868811] RSP: 0018:ffffc1e280a5fd58 EFLAGS: 00010246
[380751.870978] RAX: ffff9fd1e84e15e0 RBX: ffff9fd1e84e6dd0 RCX:
0000000116acc580
[380751.873791] RDX: ffff9fd1f97a9400 RSI: ffff9fd1e84e1800 RDI:
ffff9fd1e4d6d420
[380751.876059] RBP: ffff9fd1e4d49000 R08: 0000000116acc580 R09:
0000000116acc580
[380751.878284] R10: 0000000000000000 R11: 0000000000000000 R12:
ffff9fd1e6e931e8
[380751.880500] R13: ffff9fd1e84e6ee0 R14: 0000000000000010 R15:
0000000000000003
[380751.882687] FS: 0000000000000000(0000) GS:ffff9fd1fac00000(0000)
knlGS:0000000000000000
[380751.885236] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[380751.887059] CR2: 0000000000000138 CR3: 000000011860a001 CR4:
00000000003606f0
[380751.889308] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[380751.891523] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400
[380751.893738] Call Trace:
[380751.894639] scsi_times_out+0x60/0x1c0
[380751.895861] blk_mq_check_expired+0x144/0x200
[380751.897302] ? __switch_to_asm+0x35/0x70
[380751.898551] blk_mq_queue_tag_busy_iter+0x195/0x2e0
[380751.900091] ? __blk_mq_requeue_request+0x100/0x100
[380751.901611] ? __switch_to_asm+0x41/0x70
[380751.902853] ? __blk_mq_requeue_request+0x100/0x100
[380751.904398] blk_mq_timeout_work+0x54/0x130
[380751.905740] process_one_work+0x195/0x390
[380751.907228] worker_thread+0x30/0x390
[380751.908713] ? process_one_work+0x390/0x390
[380751.910350] kthread+0x10d/0x130
[380751.911470] ? kthread_flush_work_fn+0x10/0x10
[380751.913007] ret_from_fork+0x35/0x40
crash> dis -l iscsi_eh_cmd_timed_out+0x15e
xxxxx/drivers/scsi/libiscsi.c: 2062
1970 enum blk_eh_timer_return iscsi_eh_cmd_timed_out(struct scsi_cmnd
*sc)
{
...
1984 spin_lock_bh(&session->frwd_lock);
1985 task = (struct iscsi_task *)sc->SCp.ptr;
1986 if (!task) {
1987 /*
1988 * Raced with completion. Blk layer has taken
ownership
1989 * so let timeout code complete it now.
1990 */
1991 rc = BLK_EH_DONE;
1992 goto done;
1993 }
...
2052 for (i = 0; i < conn->session->cmds_max; i++) {
2053 running_task = conn->session->cmds[i];
2054 if (!running_task->sc || running_task == task ||
2055 running_task->state != ISCSI_TASK_RUNNING)
2056 continue;
2057
2058 /*
2059 * Only check if cmds started before this one have
made
2060 * progress, or this could never fail
2061 */
2062 if (time_after(running_task->sc->jiffies_at_alloc,
2063 task->sc->jiffies_at_alloc)) <---
2064 continue;
2065
...
}
carsh> struct scsi_cmnd ffff9fd1e6e931e8
struct scsi_cmnd {
...
SCp = {
ptr = 0x0, <--- iscsi_task
this_residual = 0,
...
},
}
To prevent this, we take a ref to the cmd under the back (completion) lock
so if the completion side were to call iscsi_complete_task() on the task
while the timer/eh paths are not holding the back_lock it will not be freed
from under us.
Note that this requires the previous patch, "scsi: libiscsi: Drop
taskqueuelock" because bnx2i sleeps in its cleanup_task callout if the cmd
is aborted. If the EH/timer and completion path are racing we don't know
which path will do the last put. The previous patch moved the operations we
needed to do under the forward lock to cleanup_queued_task. Once that has
run we can drop the forward lock for the cmd and bnx2i no longer has to
worry about if the EH, timer or completion path did the ast put and if the
forward lock is held or not since it won't be.
Link: https://lore.kernel.org/r/20210207044608.27585-4-michael.christie@oracle.com
Reported-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-02-06 22:46:02 -06:00
|
|
|
|
2021-05-25 13:18:04 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-05-30 12:57:18 -05:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:04 -05:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-06-16 15:57:09 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-25 15:27:36 +02:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2010-12-31 02:22:18 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:18 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:32 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-06-15 22:11:10 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-11-20 20:22:45 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:05 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2021-05-25 13:18:05 -05:00
|
|
|
|
2010-12-31 02:22:18 -06:00
|
|
|
|
2021-05-25 13:18:05 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2016-03-30 11:26:46 -07:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2010-12-31 02:22:18 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-20 08:02:10 +05:30
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
treewide: kvzalloc() -> kvcalloc()
The kvzalloc() function has a 2-factor argument form, kvcalloc(). This
patch replaces cases of:
kvzalloc(a * b, gfp)
with:
kvcalloc(a * b, gfp)
as well as handling cases of:
kvzalloc(a * b * c, gfp)
with:
kvzalloc(array3_size(a, b, c), gfp)
as it's slightly less ugly than:
kvcalloc(array_size(a, b), c, gfp)
This does, however, attempt to ignore constant size factors like:
kvzalloc(4 * 1024, gfp)
though any constants defined via macros get caught up in the conversion.
Any factors with a sizeof() of "unsigned char", "char", and "u8" were
dropped, since they're redundant.
The Coccinelle script used for this was:
// Fix redundant parens around sizeof().
@@
type TYPE;
expression THING, E;
@@
(
kvzalloc(
- (sizeof(TYPE)) * E
+ sizeof(TYPE) * E
, ...)
|
kvzalloc(
- (sizeof(THING)) * E
+ sizeof(THING) * E
, ...)
)
// Drop single-byte sizes and redundant parens.
@@
expression COUNT;
typedef u8;
typedef __u8;
@@
(
kvzalloc(
- sizeof(u8) * (COUNT)
+ COUNT
, ...)
|
kvzalloc(
- sizeof(__u8) * (COUNT)
+ COUNT
, ...)
|
kvzalloc(
- sizeof(char) * (COUNT)
+ COUNT
, ...)
|
kvzalloc(
- sizeof(unsigned char) * (COUNT)
+ COUNT
, ...)
|
kvzalloc(
- sizeof(u8) * COUNT
+ COUNT
, ...)
|
kvzalloc(
- sizeof(__u8) * COUNT
+ COUNT
, ...)
|
kvzalloc(
- sizeof(char) * COUNT
+ COUNT
, ...)
|
kvzalloc(
- sizeof(unsigned char) * COUNT
+ COUNT
, ...)
)
// 2-factor product with sizeof(type/expression) and identifier or constant.
@@
type TYPE;
expression THING;
identifier COUNT_ID;
constant COUNT_CONST;
@@
(
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * (COUNT_ID)
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * COUNT_ID
+ COUNT_ID, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * (COUNT_CONST)
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * COUNT_CONST
+ COUNT_CONST, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * (COUNT_ID)
+ COUNT_ID, sizeof(THING)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * COUNT_ID
+ COUNT_ID, sizeof(THING)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * (COUNT_CONST)
+ COUNT_CONST, sizeof(THING)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * COUNT_CONST
+ COUNT_CONST, sizeof(THING)
, ...)
)
// 2-factor product, only identifiers.
@@
identifier SIZE, COUNT;
@@
- kvzalloc
+ kvcalloc
(
- SIZE * COUNT
+ COUNT, SIZE
, ...)
// 3-factor product with 1 sizeof(type) or sizeof(expression), with
// redundant parens removed.
@@
expression THING;
identifier STRIDE, COUNT;
type TYPE;
@@
(
kvzalloc(
- sizeof(TYPE) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kvzalloc(
- sizeof(TYPE) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kvzalloc(
- sizeof(TYPE) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kvzalloc(
- sizeof(TYPE) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(TYPE))
, ...)
|
kvzalloc(
- sizeof(THING) * (COUNT) * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kvzalloc(
- sizeof(THING) * (COUNT) * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kvzalloc(
- sizeof(THING) * COUNT * (STRIDE)
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
|
kvzalloc(
- sizeof(THING) * COUNT * STRIDE
+ array3_size(COUNT, STRIDE, sizeof(THING))
, ...)
)
// 3-factor product with 2 sizeof(variable), with redundant parens removed.
@@
expression THING1, THING2;
identifier COUNT;
type TYPE1, TYPE2;
@@
(
kvzalloc(
- sizeof(TYPE1) * sizeof(TYPE2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kvzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(TYPE2))
, ...)
|
kvzalloc(
- sizeof(THING1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kvzalloc(
- sizeof(THING1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(THING1), sizeof(THING2))
, ...)
|
kvzalloc(
- sizeof(TYPE1) * sizeof(THING2) * COUNT
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
|
kvzalloc(
- sizeof(TYPE1) * sizeof(THING2) * (COUNT)
+ array3_size(COUNT, sizeof(TYPE1), sizeof(THING2))
, ...)
)
// 3-factor product, only identifiers, with redundant parens removed.
@@
identifier STRIDE, SIZE, COUNT;
@@
(
kvzalloc(
- (COUNT) * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- COUNT * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- COUNT * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- (COUNT) * (STRIDE) * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- COUNT * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- (COUNT) * STRIDE * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- (COUNT) * (STRIDE) * (SIZE)
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
|
kvzalloc(
- COUNT * STRIDE * SIZE
+ array3_size(COUNT, STRIDE, SIZE)
, ...)
)
// Any remaining multi-factor products, first at least 3-factor products,
// when they're not all constants...
@@
expression E1, E2, E3;
constant C1, C2, C3;
@@
(
kvzalloc(C1 * C2 * C3, ...)
|
kvzalloc(
- (E1) * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
|
kvzalloc(
- (E1) * (E2) * E3
+ array3_size(E1, E2, E3)
, ...)
|
kvzalloc(
- (E1) * (E2) * (E3)
+ array3_size(E1, E2, E3)
, ...)
|
kvzalloc(
- E1 * E2 * E3
+ array3_size(E1, E2, E3)
, ...)
)
// And then all remaining 2 factors products when they're not all constants,
// keeping sizeof() as the second factor argument.
@@
expression THING, E1, E2;
type TYPE;
constant C1, C2, C3;
@@
(
kvzalloc(sizeof(THING) * C2, ...)
|
kvzalloc(sizeof(TYPE) * C2, ...)
|
kvzalloc(C1 * C2 * C3, ...)
|
kvzalloc(C1 * C2, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * (E2)
+ E2, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(TYPE) * E2
+ E2, sizeof(TYPE)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * (E2)
+ E2, sizeof(THING)
, ...)
|
- kvzalloc
+ kvcalloc
(
- sizeof(THING) * E2
+ E2, sizeof(THING)
, ...)
|
- kvzalloc
+ kvcalloc
(
- (E1) * E2
+ E1, E2
, ...)
|
- kvzalloc
+ kvcalloc
(
- (E1) * (E2)
+ E1, E2
, ...)
|
- kvzalloc
+ kvcalloc
(
- E1 * E2
+ E1, E2
, ...)
)
Signed-off-by: Kees Cook <keescook@chromium.org>
2018-06-12 14:04:48 -07:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2009-03-05 14:45:55 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-12-21 14:37:27 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2017-05-17 16:21:54 -04:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:04 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-06-16 10:11:33 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:46:04 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-03-22 12:54:44 -07:00
|
|
|
|
2009-03-05 14:46:04 -06:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
|
|
|
|
|
2022-02-26 17:04:35 -06:00
|
|
|
|
2020-05-05 09:19:08 +08:00
|
|
|
|
2022-02-26 17:04:35 -06:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2009-03-05 14:46:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
2009-03-05 14:46:05 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:27:38 -05:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2022-06-16 17:27:38 -05:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 17:27:38 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-05-21 15:54:00 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-12-22 14:08:27 -08:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2017-12-22 14:08:27 -08:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:17 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-09-22 08:21:22 +05:30
|
|
|
|
2008-05-21 15:54:12 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:04 -06:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-06-16 10:11:33 -05:00
|
|
|
|
2021-02-06 22:46:04 -06:00
|
|
|
|
|
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2007-05-30 12:57:19 -05:00
|
|
|
|
2008-05-21 15:54:01 -05:00
|
|
|
|
2009-09-22 08:21:22 +05:30
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2007-12-13 12:43:38 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:17 -05:00
|
|
|
|
2021-02-06 22:46:04 -06:00
|
|
|
|
2007-07-26 12:46:48 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-09-22 08:21:22 +05:30
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
|
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2021-09-11 21:51:59 +08:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2006-06-28 12:00:27 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:12 -05:00
|
|
|
|
2006-06-28 12:00:27 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:25 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-01-17 13:39:36 -06:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2023-01-17 13:39:36 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2008-09-24 11:46:10 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2017-07-13 09:11:21 -07:00
|
|
|
|
2023-01-17 13:39:36 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-07-13 09:11:21 -07:00
|
|
|
|
2021-05-25 13:18:11 -05:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:50 -05:00
|
|
|
|
2012-01-19 03:06:53 -08:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2006-07-24 15:47:50 -05:00
|
|
|
|
2017-07-13 09:11:21 -07:00
|
|
|
|
2006-07-24 15:47:29 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2023-01-17 13:39:36 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:01 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:01 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:01 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:34 -05:00
|
|
|
|
2022-03-09 20:57:58 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2022-03-09 20:57:58 -05:00
|
|
|
|
2008-05-21 15:54:01 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-27 16:53:29 +05:30
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2017-10-11 16:25:40 -07:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2006-11-22 14:57:56 +00:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-12-02 00:32:09 -06:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:34 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:34 -05:00
|
|
|
|
2022-03-09 20:57:58 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2022-03-09 20:57:58 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2006-05-18 20:31:34 -05:00
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2022-03-09 20:57:58 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-12-22 14:08:27 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-09 20:57:59 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2025-04-05 10:17:26 +02:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2015-06-24 06:41:58 +05:30
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-02-28 17:32:15 -06:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2007-02-28 17:32:15 -06:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2008-12-02 00:32:09 -06:00
|
|
|
|
|
|
|
|
|
2022-03-09 20:57:59 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
|
|
|
|
|
2009-12-21 14:37:28 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-07-26 12:46:48 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2015-06-24 06:41:58 +05:30
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2022-03-09 20:57:59 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:24 -04:00
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:31 -04:00
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-08-31 18:09:24 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:52 -06:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2007-07-26 12:46:48 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-01-31 13:36:53 -06:00
|
|
|
|
|
|
|
|
|
2008-01-31 13:36:43 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-06 22:46:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-02-25 09:41:30 -08:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-04-06 12:17:46 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-04-06 12:17:46 -05:00
|
|
|
|
|
|
|
|
|
2006-05-02 19:46:40 -05:00
|
|
|
|
|
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-02 19:46:40 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-05-02 19:46:40 -05:00
|
|
|
|
|
|
|
|
|
2008-05-21 15:54:09 -05:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
2006-05-02 19:46:40 -05:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-03-09 14:14:51 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-02 19:46:40 -05:00
|
|
|
|
2025-04-05 10:17:26 +02:00
|
|
|
|
2009-05-13 17:57:43 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
2009-03-05 14:45:58 -06:00
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2006-05-30 00:37:20 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2006-05-18 20:31:42 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2009-05-13 17:57:49 -05:00
|
|
|
|
2009-05-13 17:57:46 -05:00
|
|
|
|
2021-05-25 13:18:06 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2007-08-15 01:38:30 -05:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2006-10-16 18:09:39 -04:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:12 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] libiscsi: Reduce locking contention in fast path
Replace the session lock with two locks, a forward lock and
a backwards lock named frwd_lock and back_lock respectively.
The forward lock protects resources that change while sending a
request to the target, such as cmdsn, queued_cmdsn, and allocating
task from the commands' pool with kfifo_out.
The backward lock protects resources that change while processing
a response or in error path, such as cmdsn_exp, cmdsn_max, and
returning tasks to the commands' pool with kfifo_in.
Under a steady state fast-path situation, that is when one
or more processes/threads submit IO to an iscsi device and
a single kernel upcall (e.g softirq) is dealing with processing
of responses without errors, this patch eliminates the contention
between the queuecommand()/request response/scsi_done() flows
associated with iscsi sessions.
Between the forward and the backward locks exists a strict locking
hierarchy. The mutual exclusion zone protected by the forward lock can
enclose the mutual exclusion zone protected by the backward lock but not
vice versa.
For example, in iscsi_conn_teardown or in iscsi_xmit_data when there is
a failure and __iscsi_put_task is called, the backward lock is taken while
the forward lock is still taken. On the other hand, if in the RX path a nop
is to be sent, for example in iscsi_handle_reject or __iscsi_complete_pdu
than the forward lock is released and the backward lock is taken for the
duration of iscsi_send_nopout, later the backward lock is released and the
forward lock is retaken.
libiscsi_tcp uses two kernel fifos the r2t pool and the r2t queue.
The insertion and deletion from these queues didn't corespond to the
assumption taken by the new forward/backwards session locking paradigm.
That is, in iscsi_tcp_clenup_task which belongs to the RX (backwards)
path, r2t is taken out from r2t queue and inserted to the r2t pool.
In iscsi_tcp_get_curr_r2t which belong to the TX (forward) path, r2t
is also inserted to the r2t pool and another r2t is pulled from r2t
queue.
Only in iscsi_tcp_r2t_rsp which is called in the RX path but can requeue
to the TX path, r2t is taken from the r2t pool and inserted to the r2t
queue.
In order to cope with this situation, two spin locks were added,
pool2queue and queue2pool. The former protects extracting from the
r2t pool and inserting to the r2t queue, and the later protects the
extracing from the r2t queue and inserting to the r2t pool.
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
[minor fix up to apply cleanly and compile fix]
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
2014-02-07 00:41:38 -06:00
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
2021-02-06 22:46:06 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-07 19:13:11 -05:00
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-03-22 07:41:30 -04:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-03-22 07:41:30 -04:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-08-08 13:44:29 +03:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-01-26 21:13:10 -06:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2012-01-19 03:06:53 -08:00
|
|
|
|
|
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-08-08 13:44:29 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-11-22 05:28:21 -05:00
|
|
|
|
|
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-12-13 12:43:20 -06:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2009-11-11 16:34:33 -06:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2013-07-22 07:46:10 -04:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-22 07:46:10 -04:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2012-01-19 03:06:53 -08:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2012-01-19 03:06:53 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:16 -05:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2008-05-21 15:54:16 -05:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-22 07:46:09 -04:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-06-20 10:21:26 -07:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
|
|
|
|
|
2014-09-29 13:55:42 -05:00
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2011-02-16 15:04:35 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-12-13 12:43:30 -06:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
2013-07-22 07:46:10 -04:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-22 07:46:10 -04:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
|
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-07-01 05:54:12 -04:00
|
|
|
|
2013-11-22 05:28:21 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2013-11-22 05:28:21 -05:00
|
|
|
|
2006-06-28 12:00:23 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[SCSI] iscsi class, iscsi_tcp, iser, qla4xxx: add netdevname sysfs attr
iSCSI must support software iscsi (iscsi_tcp, iser), hardware iscsi (qla4xxx),
and partial offload (broadcom). To be able to allow each stack or driver
or port (virtual or physical) to be able to log into the same target portal
we use the initiator tuple [[HWADDRESS | NETDEVNAME], INITIATOR_NAME] and
the target tuple [TARGETNAME, CONN_ADDRESS, CONN_PORT] to id a session.
This patch adds the netdev name, which is used by software iscsi when
it binds a session to a netdevice using the SO_BINDTODEVICE sock opt.
It cannot use HWADDRESS because if someone did vlans then the same netdevice
will have the same mac and the initiator,target id will not be unique.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2007-05-30 12:57:25 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
[SCSI] iscsi class, iscsi_tcp, iser, qla4xxx: add netdevname sysfs attr
iSCSI must support software iscsi (iscsi_tcp, iser), hardware iscsi (qla4xxx),
and partial offload (broadcom). To be able to allow each stack or driver
or port (virtual or physical) to be able to log into the same target portal
we use the initiator tuple [[HWADDRESS | NETDEVNAME], INITIATOR_NAME] and
the target tuple [TARGETNAME, CONN_ADDRESS, CONN_PORT] to id a session.
This patch adds the netdev name, which is used by software iscsi when
it binds a session to a netdevice using the SO_BINDTODEVICE sock opt.
It cannot use HWADDRESS because if someone did vlans then the same netdevice
will have the same mac and the initiator,target id will not be unique.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2007-05-30 12:57:25 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
2007-05-30 12:57:13 -05:00
|
|
|
|
2021-02-23 18:00:17 -08:00
|
|
|
|
2007-05-30 12:57:13 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-05-21 15:53:59 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
|
|
|
|
|
[SCSI] iscsi class, iscsi_tcp, iser, qla4xxx: add netdevname sysfs attr
iSCSI must support software iscsi (iscsi_tcp, iser), hardware iscsi (qla4xxx),
and partial offload (broadcom). To be able to allow each stack or driver
or port (virtual or physical) to be able to log into the same target portal
we use the initiator tuple [[HWADDRESS | NETDEVNAME], INITIATOR_NAME] and
the target tuple [TARGETNAME, CONN_ADDRESS, CONN_PORT] to id a session.
This patch adds the netdev name, which is used by software iscsi when
it binds a session to a netdevice using the SO_BINDTODEVICE sock opt.
It cannot use HWADDRESS because if someone did vlans then the same netdevice
will have the same mac and the initiator,target id will not be unique.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2007-05-30 12:57:25 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:13 -05:00
|
|
|
|
2009-05-13 17:57:40 -05:00
|
|
|
|
2007-05-30 12:57:12 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2006-04-06 21:13:41 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|