Avoid a race causing corruption when snapshots of the same origin have
different chunk sizes by sorting the internal list of snapshots by chunk
size, largest first.
https://bugzilla.redhat.com/show_bug.cgi?id=182659
For example, let's have two snapshots with different chunk sizes. The
first snapshot (1) has small chunk size and the second snapshot (2) has
large chunk size. Let's have chunks A, B, C in these snapshots:
snapshot1: ====A==== ====B====
snapshot2: ==========C==========
(Chunk size is a power of 2. Chunks are aligned.)
A write to the origin at a position within A and C comes along. It
triggers reallocation of A, then reallocation of C and links them
together using A as the 'primary' exception.
Then another write to the origin comes along at a position within B and
C. It creates pending exception for B. C already has a reallocation in
progress and it already has a primary exception (A), so nothing is done
to it: B and C are not linked.
If the reallocation of B finishes before the reallocation of C, because
there is no link with the pending exception for C it does not know to
wait for it and, the second write is dispatched to the origin and causes
data corruption in the chunk C in snapshot2.
To avoid this situation, we maintain snapshots sorted in descending
order of chunk size. This leads to a guaranteed ordering on the links
between the pending exceptions and avoids the problem explained above -
both A and B now get linked to C.
Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
sata_mv: Prevent PIO commands to be defered too long if traffic in progress.
pata_sc1200: Fix crash on boot
libata: fix internal command failure handling
libata: fix PMP initialization
sata_nv: make sure link is brough up online when skipping hardreset
ahci / atiixp / pci quirks: rename AMD SB900 into Hudson-2
ahci: Add the AHCI controller Linux Device ID for NVIDIA chipsets.
pata_via: extend the rev_max for VT6330
I'm seeing an oops condition when kvm-intel and kvm-amd are modprobe'd
during boot (say on an Intel system) and then rmmod'd:
# modprobe kvm-intel
kvm_init()
kvm_init_debug()
kvm_arch_init() <-- stores debugfs dentries internally
(success, etc)
# modprobe kvm-amd
kvm_init()
kvm_init_debug() <-- second initialization clobbers kvm's
internal pointers to dentries
kvm_arch_init()
kvm_exit_debug() <-- and frees them
# rmmod kvm-intel
kvm_exit()
kvm_exit_debug() <-- double free of debugfs files!
*BOOM*
If execution gets to the end of kvm_init(), then the calling module has been
established as the kvm provider. Move the debugfs initialization to the end of
the function, and remove the now-unnecessary call to kvm_exit_debug() from the
error path. That way we avoid trampling on the debugfs entries and freeing
them twice.
Cc: stable@kernel.org
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
On a 32 bits compile, commit 3da0dd433d
introduced the following warnings:
arch/x86/kvm/mmu.c: In function ‘kvm_set_pte_rmapp’:
arch/x86/kvm/mmu.c:770: warning: cast to pointer from integer of different size
arch/x86/kvm/mmu.c: In function ‘kvm_set_spte_hva’:
arch/x86/kvm/mmu.c:849: warning: cast from pointer to integer of different size
The following patch uses 'unsigned long' instead of u64 to match the
pointer size on both arches.
Signed-off-by: Frederik Deweerdt <frederik.deweerdt@xprog.eu>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
hrtimer->base can be temporarily NULL due to racing hrtimer_start.
See switch_hrtimer_base/lock_hrtimer_base.
Use hrtimer_get_remaining which is robust against it.
CC: stable@kernel.org
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Create an inline function to extract the pnode from a global
physical address and then convert the broadcast assist unit to
use the newly created uv_gpa_to_pnode function.
The open-coded code was wrong as well - it might explain a
few of our unexplained bau hangs.
Signed-off-by: Robin Holt <holt@sgi.com>
Acked-by: Cliff Whickman <cpw@sgi.com>
Cc: linux-mm@kvack.org
Cc: Jack Steiner <steiner@sgi.com>
LKML-Reference: <20091016112920.GZ8903@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Prefix global/setup routines with "mcheck_" thus differentiating
from the internal facilities prefixed with "mce_". Also, prefix
the per cpu calls with mcheck_cpu and rename them to reflect the
MCE setup hierarchy of calls better.
There should be no functionality change resulting from this
patch.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Andi Kleen <andi@firstfloor.org>
LKML-Reference: <1255689093-26921-1-git-send-email-borislav.petkov@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Use excl_link when non NCQ commands are defered, to be sure they are processed
as soon as outstanding commands are completed. It prevents some commands to be
defered indifinitely when using a port multiplier.
Signed-off-by: Gwendal Grignou <gwendal@google.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The SC1200 needs a NULL terminator or it may cause a crash on boot.
Bug #14227
Also correct a bogus comment as the driver had serializing added so can run
dual port.
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
When an internal command fails, it should be failed directly without
invoking EH. In the original implemetation, this was accomplished by
letting internal command bypass failure handling in ata_qc_complete().
However, later changes added post-successful-completion handling to
that code path and the success path is no longer adequate as internal
command failure path. One of the visible problems is that internal
command failure due to timeout or other freeze conditions would
spuriously trigger WARN_ON_ONCE() in the success path.
This patch updates failure path such that internal command failure
handling is contained there.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@kernel.org
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Commit 842faa6c1a fixed error handling
during attach by not committing detected device class to dev->class
while attaching a new device. However, this change missed the PMP
class check in the configuration loop causing a new PMP device to go
through ata_dev_configure() as if it were an ATA or ATAPI device.
As PMP device doesn't have a regular IDENTIFY data, this makes
ata_dev_configure() tries to configure a PMP device using an invalid
data. For the most part, it wasn't too harmful and went unnoticed but
this ends up clearing dev->flags which may have ATA_DFLAG_AN set by
sata_pmp_attach(). This means that SATA_PMP_FEAT_NOTIFY ends up being
disabled on PMPs and on PMPs which honor the flag breaks hotplug
support.
This problem was discovered and reported by Ethan Hsiao.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Ethan Hsiao <ethanhsiao@jmicron.com>
Cc: stable@kernel.org
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
prereset doesn't bring link online if hardreset is about to happen and
nv_hardreset() may skip if conditions are not right so softreset may
be entered with non-working link status if the system firmware didn't
bring it up before entering OS code which can happen during resume.
This patch makes nv_hardreset() to bring up the link if it's skipping
reset.
This bug was reported by frodone@gmail.com in the following bug entry.
http://bugzilla.kernel.org/show_bug.cgi?id=14329
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: frodone@gmail.com
Cc: stable@kernel.org
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Fix the VT6330 issue, it's because the rev_max of VT6330 exceeds 0x2f.
The VT6415 and VT6330 share the same device ID.
Signed-off-by: Joseph Chan <josephchan@via.com.tw>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This code was added for some ancient SH-4 solution engines with peculiar
boot ROMs that did silly things to the UBC MSTP bits. None of these have
been in the wild for years, and these days the clock framework wraps up
the MSTP bits, meaning that the UBC code is one of the few interfaces
that is stomping MSTP bits underneath the clock framework. At this point
the risks far outweigh any benefit this code provided, so just kill it
off.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
This enables SCHED_MC support for SH-X3 multi-cores. Presently this is
just a simple wrapper around the possible map, but this allows for
tying in support for some of the more exotic NUMA clusters where we can
actually do something with the topology.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
In the case where need_resched() is set in between the cpu_idle() and
pm_idle() calls we were missing an else case for just re-enabling local
IRQs and bailing out. This was noticed by the irqs_disabled() warning,
even though IRQs were being re-enabled elsewhere.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
We released the first version of perf with 0.0.1 in v2.6.31,
time to double our version number to 0.0.2 ;-)
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Original radeon didn't have a connector table in the
bios. Check for the CRT table and if we have one,
add a VGA connector.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Need to check the return type for the quirk function
to decide whether we add the connectors and encoders.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
This follows the x86 change and moves check_pgt_cache() up under the
!need_resched() tight loop, rather than simply calling in to it when
exiting idle.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
This does a bit of chainsawing of the idle loop code to get light sleep
working on SMP. Previously this was forcing secondary CPUs in to sleep
mode with them not coming back if they didn't have their own local
timers. Given that we use clockevents broadcasting by default, the CPU
managing the clockevents can't have IRQs disabled before entering its
sleep state.
This unfortunately leaves us with the age-old need_resched() race in
between local_irq_enable() and cpu_sleep(), but at present this is
unavoidable. After some more experimentation it may be possible to layer
on SR.BL bit manipulation over top of this scheme to inhibit the race
condition, but given the current potential for missing wakeups, this is
left as a future exercise.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
When requeuing tasks from one futex to another, the reference held
by the requeued task to the original futex location needs to be
dropped eventually.
Dropping the reference may ultimately lead to a call to
"iput_final" and subsequently call into filesystem- specific code -
which may be non-atomic.
It is therefore safer to defer this drop operation until after the
futex_hash_bucket spinlock has been dropped.
Originally-From: Helge Bahmann <hcb@chaoticmind.net>
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
Cc: <stable@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Dinakar Guniguntala <dino@in.ibm.com>
Cc: John Stultz <johnstul@linux.vnet.ibm.com>
Cc: Sven-Thorsten Dietrich <sdietrich@novell.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <4AD7A298.5040802@us.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
All of the secondary CPUs are forced in to light sleep mode, but we were
missing the same initialization for the boot CPU. This resulted in
inconsistent sleep modes depending on which CPU we were on, confusing the
idle loop when not polling.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
The MCE initialization code explicitly says it doesn't handle
asymmetric configurations where different CPUs support different
numbers of MCE banks, and it prints a big warning in that case.
Therefore, printing the "mce: CPU supports <x> MCE banks"
message into the kernel log for every CPU is pure redundancy
that clutters the log significantly for systems with lots of
CPUs.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
LKML-Reference: <adaeip473qt.fsf@cisco.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
A few parts of the uv_hub_info structure are initialized
incorrectly.
- n_val is being loaded with m_val.
- gpa_mask is initialized with a bytes instead of an unsigned long.
- Handle the case where none of the alias registers are used.
Lastly I converted the bau over to using the uv_hub_info->m_val
which is the correct value.
Without this patch, booting a large configuration hits a
problem where the upper bits of the gnode affect the pnode
and the bau will not operate.
Signed-off-by: Robin Holt <holt@sgi.com>
Acked-by: Jack Steiner <steiner@sgi.com>
Cc: Cliff Whickman <cpw@sgi.com>
Cc: stable@kernel.org
LKML-Reference: <20091015224946.396355000@alcatraz.americas.sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The open function got the BKL via the big push down. Replace it by
preempt_enable/disable as this is sufficient for an UP machine.
The ioctl can be unlocked because there is no functionality which
requires serialization. The usage by multiple callers is broken with
and without the BKL due to the local static variable addr.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
async_syndrome_val check the P and Q blocks used for RAID6
calculations.
With DDF raid6, some of the data blocks might be NULL, so
this needs to be handled in the same way that async_gen_syndrome
handles it.
As async_syndrome_val calls async_xor, also enhance async_xor
to detect and skip NULL blocks in the list.
Signed-off-by: NeilBrown <neilb@suse.de>
md/raid6 passes a list of 'struct page *' to the async_tx routines,
which then either DMA map them for offload, or take the page_address
for CPU based calculations.
For RAID6 we sometime leave 'blanks' in the list of pages.
For CPU based calcs, we want to treat theses as a page of zeros.
For offloaded calculations, we simply don't pass a page to the
hardware.
Currently the 'blanks' are encoded as a pointer to
raid6_empty_zero_page. This is a 4096 byte memory region, not a
'struct page'. This is mostly handled correctly but is rather ugly.
So change the code to pass and expect a NULL pointer for the blanks.
When taking page_address of a page, we need to check for a NULL and
in that case use raid6_empty_zero_page.
Signed-off-by: NeilBrown <neilb@suse.de>
Add code to handle the cache disabled case. Fixes breakage introduced by
37443ef3f0 ("sh: Migrate SH-4 cacheflush
ops to function pointers."). Without this patch configuring caches off
with CONFIG_CACHE_OFF=y makes kfr2r09 and migo-r lock up in fbdev
deferred io or early user space.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
When a raid5 (or raid6) array is being reshaped to have fewer devices,
conf->raid_disks is the latter and hence smaller number of devices.
However sometimes we want to use a number which is the total number of
currently required devices - the larger of the 'old' and 'new' sizes.
Before we implemented reducing the number of devices, this was always
'new' i.e. ->raid_disks.
Now we need max(raid_disks, previous_raid_disks) in those places.
This particularly affects assembling an array that was shutdown while
in the middle of a reshape to fewer devices.
md.c needs a similar fix when interpreting the md metadata.
Signed-off-by: NeilBrown <neilb@suse.de>
The percpu conversion allowed a straightforward handoff of stripe
processing to the async subsytem that initially showed some modest gains
(+4%). However, this model is too simplistic and leads to stripes
bouncing between raid5d and the async thread pool for every invocation
of handle_stripe(). As reported by Holger this can fall into a
pathological situation severely impacting throughput (6x performance
loss).
By downleveling the parallelism to raid_run_ops the pathological
stripe_head bouncing is eliminated. This version still exhibits an
average 11% throughput loss for:
mdadm --create /dev/md0 /dev/sd[b-q] -n 16 -l 6
echo 1024 > /sys/block/md0/md/stripe_cache_size
dd if=/dev/zero of=/dev/md0 bs=1024k count=2048
...but the results are at least stable and can be used as a base for
further multicore experimentation.
Reported-by: Holger Kiehl <Holger.Kiehl@dwd.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
drivers/md/unroll.pl replaced by awk script to drop build-time
dependency on perl
Signed-off-by: Vladimir Dronnikov <dronnikov@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Older binutils breaks if ASSERT() is used without a sink
for the output.
For example 2.14.90.0.6 is known to be broken, the link
fails with:
LD .tmp_vmlinux1
ld:arch/x86/kernel/vmlinux.lds:678: parse error
Document this quirk in all three files that use it.
See: http://marc.info/?l=linux-kbuild&m=124930110427870&w=2
See[2]: d2ba8b2 ("x86: Fix assert syntax in vmlinux.lds.S")
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Roland McGrath <roland@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
LKML-Reference: <4AD6523D.5030909@zytor.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Presently The SH-4 cache flushing code uses flush_cache_4096() for most
of the real flushing work, which breaks down to a fixed 4096 unroll and
increment. Not only is this sub-optimal for larger page sizes, it's also
uncovered a bug in sh4_flush_dcache_page() when large page sizes are used
and we have no cache aliases -- resulting in only a part of the page's
D-cache lines being written back.
Signed-off-by: Valentin Sitdikov <valentin.sitdikov@siemens.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
and replace with vfs_fsync which is much neater (but wasn't exported,
or even in existence at the time the code was written).
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NeilBrown <neilb@suse.de>
Both raid1 and raid10 create a mempool during startup.
If the 'alloc' function for this mempool fails, unplug_slaves
is called.
If that happens when the pool is being initialised, unplug_slaves
will try to use the 'conf' structure that isn't filled in yet, and
badness will happen.
So ensure that unplug_slaves doesn't get called unless we know
that the conf structure if fully initialised.
Signed-off-by: NeilBrown <neilb@suse.de>
Deallocating a raid5_conf_t structure requires taking 'device_lock'.
Ensure it is initialized before it is used, i.e. initialize the lock
before attempting any further initializations that might fail.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
During 'check' of a raid1 or raid10 it is possible for the management
thread to spend a lot of time running 'memcmp' on blocks from
different devices, so make sure the thread has a chance to schedule.
raid5d already has a cond_resched (in process_stripe).
Reported-By: Lee Howard <faxguy@howardsilvan.com>
Signed-off-by: NeilBrown <neilb@suse.de>