Quentin Monnet says:
====================
This set contains some improvements for bpftool's "visual" program dump
option, which produces the control flow graph in a DOT format. The main
objective is to add support for inline annotations on such graphs, so that
we can have the C source code for the program showing up alongside the
instructions, when available. The last commits also make it possible to
display the line numbers or the bare opcodes in the graph, as supported by
regular program dumps.
v3:
- Fixed formatting of DOT graph: escape spaces, and remove indent that
would cause some unwanted spaces to show up in the resulting graph.
- Don't print line information if the record is empty.
- Add '<' and ' ' to the list of escaped characters for generting the
DOT graph.
- Truncate long file paths, use shorter field names ("line", "col") for
code location information in the graph, add missing separator space.
- Add a commit to return an error if JSON output and CFG are both
required.
- Add a drive-by, clean up commit for bash completion (avoid unnecessary
calls to _bpftool_once_attr()).
v2: Replace fputc(..., stdout) with putchar(...) in dotlabel_puts().
====================
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
In bpftool's bash completion file, function _bpftool_once_attr() is able
to process multiple arguments. There are a few locations where this
function is called multiple times in a row, each time for a single
argument; let's pass all arguments instead to minimize the number of
function calls required for the completion.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-8-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add support for displaying opcodes or/and file references (filepath,
line and column numbers) when dumping the control flow graphs of loaded
BPF programs with bpftool.
The filepaths in the records are absolute. To avoid blocks on the graph
to get too wide, we truncate them when they get too long (but we always
keep the entire file name). In the unlikely case where the resulting
file name is ambiguous, it remains possible to get the full path with a
regular dump (no CFG).
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-7-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When dumping a program, the keywords "opcodes" (for printing the raw
opcodes), "linum" (for displaying the filename, line number, column
number along with the source code), and "visual" (for generating the
control flow graph for translated programs) are mutually exclusive. But
there's no reason why they should be. Let's make it possible to pass
several of them at once. The "file FILE" option, which makes bpftool
output a binary image to a file, remains incompatible with the others.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-6-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
We do not support JSON output for control flow graphs of programs with
bpftool. So far, requiring both the CFG and JSON output would result in
producing a null JSON object. It makes more sense to raise an error
directly when parsing command line arguments and options, so that users
know they won't get any output they might expect.
If JSON is required for the graph, we leave it to Graphviz instead:
# bpftool prog dump xlated <REF> visual | dot -Tjson
Suggested-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-5-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
We support dumping the control flow graph of loaded programs to the DOT
format with bpftool, but so far this feature wouldn't display the source
code lines available through BTF along with the eBPF bytecode. Let's add
support for these annotations, to make it easier to read the graph.
In prog.c, we move the call to dump_xlated_cfg() in order to pass and
use the full struct dump_data, instead of creating a minimal one in
draw_bb_node().
We pass the pointer to this struct down to dump_xlated_for_graph() in
xlated_dumper.c, where most of the logics is added. We deal with BTF
mostly like we do for plain or JSON output, except that we cannot use a
"nr_skip" value to skip a given number of linfo records (we don't
process the BPF instructions linearly, and apart from the root of the
graph we don't know how many records we should skip, so we just store
the last linfo and make sure the new one we find is different before
printing it).
When printing the source instructions to the label of a DOT graph node,
there are a few subtleties to address. We want some special newline
markers, and there are some characters that we must escape. To deal with
them, we introduce a new dedicated function btf_dump_linfo_dotlabel() in
btf_dumper.c. We'll reuse this function in a later commit to format the
filepath, line, and column references as well.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-4-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When dumping the control flow graphs for programs using the 16-byte long
load instruction, we need to skip the second part of this instruction
when looking for the next instruction to process. Otherwise, we end up
printing "BUG_ld_00" from the kernel disassembler in the CFG.
Fixes: efcef17a6d ("tools: bpftool: generate .dot graph from CFG information")
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-3-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The documentation states that when line_info is available when dumping a
program, the source line will be displayed "by default". There is no
notion of "default" here: the line is always displayed if available,
there is no way currently to turn it off.
In the next sentence, the documentation states that if "linum" is used
on the command line, the relevant filename, line, and column will be
displayed "on top of the source line". This is incorrect, as they are
currently displayed on the right side of the source line (or on top of
the eBPF instruction, not the source).
This commit fixes the documentation to address these points.
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/r/20230405132120.59886-2-quentin@isovalent.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
qcom_smd_remove() always returned zero, though that isn't completely
trivial to see. So explain that in a comment and convert to
.remove_new().
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Link: https://lore.kernel.org/r/20230321154039.355098-4-u.kleine-koenig@pengutronix.de
The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.
Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Link: https://lore.kernel.org/r/20230321154039.355098-3-u.kleine-koenig@pengutronix.de
This function returned zero unconditionally. Convert it to return no
value instead. This makes it more obvious what happens in the callers.
One caller is converted to return zero explicitly. The only other caller
(smd_subdev_stop() in drivers/remoteproc/qcom_common.c) already ignored
the return value before.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Link: https://lore.kernel.org/r/20230321154039.355098-2-u.kleine-koenig@pengutronix.de
Commit 9d35dc3006 ("csky: Revert mmu ASID mechanism") removes the only
use of CONFIG_CPU_TLB_SIZE. Since then, this config has no effect and can
be deleted.
Remove the obsolete config CPU_TLB_SIZE.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
before: last 6 bits of PID is used as index to store information about
tasks accessing VMA's.
after: hash_32 is used to take of cases where tasks are created over a
period of time, and thus improve collision probability.
Result:
The patch series overall improves autonuma cost.
Kernbench around more than 5% improvement and system time in mmtest
autonuma showed more than 80% improvement
Link: https://lkml.kernel.org/r/d5a9f75513300caed74e5c8570bba9317b963c2b.1677672277.git.raghavendra.kt@amd.com
Signed-off-by: Raghavendra K T <raghavendra.kt@amd.com>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Cc: Bharata B Rao <bharata@amd.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Disha Talreja <dishaa.talreja@amd.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This helps to ensure that only recently accessed PIDs scan the VMAs.
Current implementation: (idea supported by PeterZ)
1. Accessing PID information is maintained in two windows.
access_pids[1] being newest.
2. Reset old access PID info i.e. access_pid[0] every (4 *
sysctl_numa_balancing_scan_delay) interval after initial scan delay
period expires.
The above interval seemed to be experimentally optimum since it avoids
frequent reset of access info as well as helps clearing the old access
info regularly. The reset logic is implemented in scan path.
Link: https://lkml.kernel.org/r/f7a675f66d1442d048b4216b2baf94515012c405.1677672277.git.raghavendra.kt@amd.com
Signed-off-by: Raghavendra K T <raghavendra.kt@amd.com>
Suggested-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Bharata B Rao <bharata@amd.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Disha Talreja <dishaa.talreja@amd.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
During Numa scanning make sure only relevant vmas of the tasks are
scanned.
Before:
All the tasks of a process participate in scanning the vma even if they
do not access vma in it's lifespan.
Now:
Except cases of first few unconditional scans, if a process do
not touch vma (exluding false positive cases of PID collisions)
tasks no longer scan all vma
Logic used:
1) 6 bits of PID used to mark active bit in vma numab status during
fault to remember PIDs accessing vma. (Thanks Mel)
2) Subsequently in scan path, vma scanning is skipped if current PID
had not accessed vma.
3) First two times we do allow unconditional scan to preserve earlier
behaviour of scanning.
Acknowledgement to Bharata B Rao <bharata@amd.com> for initial patch to
store pid information and Peter Zijlstra <peterz@infradead.org> (Usage of
test and set bit)
Link: https://lkml.kernel.org/r/092f03105c7c1d3450f4636b1ea350407f07640e.1677672277.git.raghavendra.kt@amd.com
Signed-off-by: Raghavendra K T <raghavendra.kt@amd.com>
Suggested-by: Mel Gorman <mgorman@techsingularity.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: Disha Talreja <dishaa.talreja@amd.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Pach series "sched/numa: Enhance vma scanning", v3.
The patchset proposes one of the enhancements to numa vma scanning
suggested by Mel. This is continuation of [3].
Reposting the rebased patchset to akpm mm-unstable tree (March 1)
Existing mechanism of scan period involves, scan period derived from
per-thread stats. Process Adaptive autoNUMA [1] proposed to gather NUMA
fault stats at per-process level to capture aplication behaviour better.
During that course of discussion, Mel proposed several ideas to enhance
current numa balancing. One of the suggestion was below
Track what threads access a VMA. The suggestion was to use an unsigned
long pid_mask and use the lower bits to tag approximately what threads
access a VMA. Skip VMAs that did not trap a fault. This would be
approximate because of PID collisions but would reduce scanning of areas
the thread is not interested in. The above suggestion intends not to
penalize threads that has no interest in the vma, thus reduce scanning
overhead.
V3 changes are mostly based on PeterZ comments (details below in changes)
Summary of patchset:
Current patchset implements:
1. Delay the vma scanning logic for newly created VMA's so that
additional overhead of scanning is not incurred for short lived tasks
(implementation by Mel)
2. Store the information of tasks accessing VMA in 2 windows. It is
regularly cleared in (4*sysctl_numa_balancing_scan_delay) interval.
The above time is derived from experimenting (Suggested by PeterZ) to
balance between frequent clearing vs obsolete access data
3. hash_32 used to encode task index accessing VMA information
4. VMA's acess information is used to skip scanning for the tasks
which had not accessed VMA
Changes since V2:
patch1:
- Renaming of structure, macro to function,
- Add explanation to heuristics
- Adding more details from result (PeterZ)
Patch2:
- Usage of test and set bit (PeterZ)
- Move storing access PID info to numa_migrate_prep()
- Add a note on fainess among tasks allowed to scan
(PeterZ)
Patch3:
- Maintain two windows of access PID information
(PeterZ supported implementation and Gave idea to extend
to N if needed)
Patch4:
- Apply hash_32 function to track VMA accessing PIDs (PeterZ)
Changes since RFC V1:
- Include Mel's vma scan delay patch
- Change the accessing pid store logic (Thanks Mel)
- Fencing structure / code to NUMA_BALANCING (David, Mel)
- Adding clearing access PID logic (Mel)
- Descriptive change log ( Mike Rapoport)
Things to ponder over:
==========================================
- Improvement to clearing accessing PIDs logic (discussed in-detail in
patch3 itself (Done in this patchset by implementing 2 window history)
- Current scan period is not changed in the patchset, so we do see
frequent tries to scan. Relaxing scan period dynamically could improve
results further.
[1] sched/numa: Process Adaptive autoNUMA
Link: https://lore.kernel.org/lkml/20220128052851.17162-1-bharata@amd.com/T/
[2] RFC V1 Link:
https://lore.kernel.org/all/cover.1673610485.git.raghavendra.kt@amd.com/
[3] V2 Link:
https://lore.kernel.org/lkml/cover.1675159422.git.raghavendra.kt@amd.com/
Results:
Summary: Huge autonuma cost reduction seen in mmtest. Kernbench improvement
is more than 5% and huge system time (80%+) improvement from mmtest autonuma.
(dbench had huge std deviation to post)
kernbench
===========
6.2.0-mmunstable-base 6.2.0-mmunstable-patched
Amean user-256 22002.51 ( 0.00%) 22649.95 * -2.94%*
Amean syst-256 10162.78 ( 0.00%) 8214.13 * 19.17%*
Amean elsp-256 160.74 ( 0.00%) 156.92 * 2.38%*
Duration User 66017.43 67959.84
Duration System 30503.15 24657.03
Duration Elapsed 504.61 493.12
6.2.0-mmunstable-base 6.2.0-mmunstable-patched
Ops NUMA alloc hit 1738835089.00 1738780310.00
Ops NUMA alloc local 1738834448.00 1738779711.00
Ops NUMA base-page range updates 477310.00 392566.00
Ops NUMA PTE updates 477310.00 392566.00
Ops NUMA hint faults 96817.00 87555.00
Ops NUMA hint local faults % 10150.00 2192.00
Ops NUMA hint local percent 10.48 2.50
Ops NUMA pages migrated 86660.00 85363.00
Ops AutoNUMA cost 489.07 442.14
autonumabench
===============
6.2.0-mmunstable-base 6.2.0-mmunstable-patched
Amean syst-NUMA01 399.50 ( 0.00%) 52.05 * 86.97%*
Amean syst-NUMA01_THREADLOCAL 0.21 ( 0.00%) 0.22 * -5.41%*
Amean syst-NUMA02 0.80 ( 0.00%) 0.78 * 2.68%*
Amean syst-NUMA02_SMT 0.65 ( 0.00%) 0.68 * -3.95%*
Amean elsp-NUMA01 313.26 ( 0.00%) 313.11 * 0.05%*
Amean elsp-NUMA01_THREADLOCAL 1.06 ( 0.00%) 1.08 * -1.76%*
Amean elsp-NUMA02 3.19 ( 0.00%) 3.24 * -1.52%*
Amean elsp-NUMA02_SMT 3.72 ( 0.00%) 3.61 * 2.92%*
Duration User 396433.47 324835.96
Duration System 2808.70 376.66
Duration Elapsed 2258.61 2258.12
6.2.0-mmunstable-base 6.2.0-mmunstable-patched
Ops NUMA alloc hit 59921806.00 49623489.00
Ops NUMA alloc miss 0.00 0.00
Ops NUMA interleave hit 0.00 0.00
Ops NUMA alloc local 59920880.00 49622594.00
Ops NUMA base-page range updates 152259275.00 50075.00
Ops NUMA PTE updates 152259275.00 50075.00
Ops NUMA PMD updates 0.00 0.00
Ops NUMA hint faults 154660352.00 39014.00
Ops NUMA hint local faults % 138550501.00 23139.00
Ops NUMA hint local percent 89.58 59.31
Ops NUMA pages migrated 8179067.00 14147.00
Ops AutoNUMA cost 774522.98 195.69
This patch (of 4):
Currently whenever a new task is created we wait for
sysctl_numa_balancing_scan_delay to avoid unnessary scanning overhead.
Extend the same logic to new or very short-lived VMAs.
[raghavendra.kt@amd.com: add initialization in vm_area_dup())]
Link: https://lkml.kernel.org/r/cover.1677672277.git.raghavendra.kt@amd.com
Link: https://lkml.kernel.org/r/7a6fbba87c8b51e67efd3e74285bb4cb311a16ca.1677672277.git.raghavendra.kt@amd.com
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Raghavendra K T <raghavendra.kt@amd.com>
Cc: Bharata B Rao <bharata@amd.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Disha Talreja <dishaa.talreja@amd.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Attempt VMA lock-based page fault handling first, and fall back to the
existing mmap_lock-based handling if that fails.
This is the s390 variant of "x86/mm: try VMA lock-based page fault handling
first".
Link: https://lkml.kernel.org/r/20230314132808.1266335-1-hca@linux.ibm.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
vma->lock being part of the vm_area_struct causes performance regression
during page faults because during contention its count and owner fields
are constantly updated and having other parts of vm_area_struct used
during page fault handling next to them causes constant cache line
bouncing. Fix that by moving the lock outside of the vm_area_struct.
All attempts to keep vma->lock inside vm_area_struct in a separate cache
line still produce performance regression especially on NUMA machines.
Smallest regression was achieved when lock is placed in the fourth cache
line but that bloats vm_area_struct to 256 bytes.
Considering performance and memory impact, separate lock looks like the
best option. It increases memory footprint of each VMA but that can be
optimized later if the new size causes issues. Note that after this
change vma_init() does not allocate or initialize vma->lock anymore. A
number of drivers allocate a pseudo VMA on the stack but they never use
the VMA's lock, therefore it does not need to be allocated. The future
drivers which might need the VMA lock should use
vm_area_alloc()/vm_area_free() to allocate the VMA.
Link: https://lkml.kernel.org/r/20230227173632.3292573-34-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
call_rcu() can take a long time when callback offloading is enabled. Its
use in the vm_area_free can cause regressions in the exit path when
multiple VMAs are being freed.
Because exit_mmap() is called only after the last mm user drops its
refcount, the page fault handlers can't be racing with it. Any other
possible user like oom-reaper or process_mrelease are already synchronized
using mmap_lock. Therefore exit_mmap() can free VMAs directly, without
the use of call_rcu().
Expose __vm_area_free() and use it from exit_mmap() to avoid possible
call_rcu() floods and performance regressions caused by it.
Link: https://lkml.kernel.org/r/20230227173632.3292573-33-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Attempt VMA lock-based page fault handling first, and fall back to the
existing mmap_lock-based handling if that fails.
Link: https://lkml.kernel.org/r/20230227173632.3292573-31-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Attempt VMA lock-based page fault handling first, and fall back to the
existing mmap_lock-based handling if that fails.
Link: https://lkml.kernel.org/r/20230227173632.3292573-30-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Add a new CONFIG_PER_VMA_LOCK_STATS config option to dump extra statistics
about handling page fault under VMA lock.
Link: https://lkml.kernel.org/r/20230227173632.3292573-29-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Due to the possibility of handle_userfault dropping mmap_lock, avoid fault
handling under VMA lock and retry holding mmap_lock. This can be handled
more gracefully in the future.
Link: https://lkml.kernel.org/r/20230227173632.3292573-28-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Suggested-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Due to the possibility of do_swap_page dropping mmap_lock, abort fault
handling under VMA lock and retry holding mmap_lock. This can be handled
more gracefully in the future.
Link: https://lkml.kernel.org/r/20230227173632.3292573-27-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Laurent Dufour <laurent.dufour@fr.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
When vma->anon_vma is not set, page fault handler will set it by either
reusing anon_vma of an adjacent VMA if VMAs are compatible or by
allocating a new one. find_mergeable_anon_vma() walks VMA tree to find a
compatible adjacent VMA and that requires not only the faulting VMA to be
stable but also the tree structure and other VMAs inside that tree.
Therefore locking just the faulting VMA is not enough for this search.
Fall back to taking mmap_lock when vma->anon_vma is not set. This
situation happens only on the first page fault and should not affect
overall performance.
Link: https://lkml.kernel.org/r/20230227173632.3292573-25-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Introduce lock_vma_under_rcu function to lookup and lock a VMA during page
fault handling. When VMA is not found, can't be locked or changes after
being locked, the function returns NULL. The lookup is performed under
RCU protection to prevent the found VMA from being destroyed before the
VMA lock is acquired. VMA lock statistics are updated according to the
results. For now only anonymous VMAs can be searched this way. In other
cases the function returns NULL.
Link: https://lkml.kernel.org/r/20230227173632.3292573-24-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Per-vma locking mechanism will search for VMA under RCU protection and
then after locking it, has to ensure it was not removed from the VMA tree
after we found it. To make this check efficient, introduce a
vma->detached flag to mark VMAs which were removed from the VMA tree.
Link: https://lkml.kernel.org/r/20230227173632.3292573-23-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Page fault handlers might need to fire MMU notifications while a new
notifier is being registered. Modify mm_take_all_locks to write-lock all
VMAs and prevent this race with page fault handlers that would hold VMA
locks. VMAs are locked before i_mmap_rwsem and anon_vma to keep the same
locking order as in page fault handlers.
Link: https://lkml.kernel.org/r/20230227173632.3292573-22-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Normally free_pgtables needs to lock affected VMAs except for the case
when VMAs were isolated under VMA write-lock. munmap() does just that,
isolating while holding appropriate locks and then downgrading mmap_lock
and dropping per-VMA locks before freeing page tables. Add a parameter to
free_pgtables for such scenario.
Link: https://lkml.kernel.org/r/20230227173632.3292573-20-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Write-lock VMA as locked before copying it and when copy_vma produces a
new VMA.
Link: https://lkml.kernel.org/r/20230227173632.3292573-18-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: Laurent Dufour <laurent.dufour@fr.ibm.com>
Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Write-lock all VMAs which might be affected by a merge, split, expand or
shrink operations. All these operations use vma_prepare() before making
the modifications, therefore it provides a centralized place to perform
VMA locking.
[surenb@google.com: remove unnecessary vp->vma check in vma_prepare]
Link: https://lkml.kernel.org/r/20230301022720.1380780-1-surenb@google.com
Link: https://lore.kernel.org/r/202302281802.J93Nma7q-lkp@intel.com/
Link: https://lkml.kernel.org/r/20230227173632.3292573-17-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Laurent Dufour <laurent.dufour@fr.ibm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Protect VMA from concurrent page fault handler while collapsing a huge
page. Page fault handler needs a stable PMD to use PTL and relies on
per-VMA lock to prevent concurrent PMD changes. pmdp_collapse_flush(),
set_huge_pmd() and collapse_and_free_pmd() can modify a PMD, which will
not be detected by a page fault handler without proper locking.
Before this patch, page tables can be walked under any one of the
mmap_lock, the mapping lock, and the anon_vma lock; so when khugepaged
unlinks and frees page tables, it must ensure that all of those either are
locked or don't exist. This patch adds a fourth lock under which page
tables can be traversed, and so khugepaged must also lock out that one.
[surenb@google.com: vm_lock/i_mmap_rwsem inversion in retract_page_tables]
Link: https://lkml.kernel.org/r/20230303213250.3555716-1-surenb@google.com
[surenb@google.com: build fix]
Link: https://lkml.kernel.org/r/CAJuCfpFjWhtzRE1X=J+_JjgJzNKhq-=JT8yTBSTHthwp0pqWZw@mail.gmail.com
Link: https://lkml.kernel.org/r/20230227173632.3292573-16-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
vma_prepare() acquires all locks required before VMA modifications. Move
vma_prepare() before vma_adjust_trans_huge() so that VMA is locked before
any modification.
Link: https://lkml.kernel.org/r/20230227173632.3292573-15-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Updates to vm_flags have to be done with VMA marked as being written for
preventing concurrent page faults or other modifications.
Link: https://lkml.kernel.org/r/20230227173632.3292573-14-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Introduce per-VMA locking. The lock implementation relies on a per-vma
and per-mm sequence counters to note exclusive locking:
- read lock - (implemented by vma_start_read) requires the vma
(vm_lock_seq) and mm (mm_lock_seq) sequence counters to differ.
If they match then there must be a vma exclusive lock held somewhere.
- read unlock - (implemented by vma_end_read) is a trivial vma->lock
unlock.
- write lock - (vma_start_write) requires the mmap_lock to be held
exclusively and the current mm counter is assigned to the vma counter.
This will allow multiple vmas to be locked under a single mmap_lock
write lock (e.g. during vma merging). The vma counter is modified
under exclusive vma lock.
- write unlock - (vma_end_write_all) is a batch release of all vma
locks held. It doesn't pair with a specific vma_start_write! It is
done before exclusive mmap_lock is released by incrementing mm
sequence counter (mm_lock_seq).
- write downgrade - if the mmap_lock is downgraded to the read lock, all
vma write locks are released as well (effectivelly same as write
unlock).
Link: https://lkml.kernel.org/r/20230227173632.3292573-13-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Move mmap_lock assert function definitions up so that they can be used by
other mmap_lock routines.
Link: https://lkml.kernel.org/r/20230227173632.3292573-12-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This prepares for page faults handling under VMA lock, looking up VMAs
under protection of an rcu read lock, instead of the usual mmap read lock.
Link: https://lkml.kernel.org/r/20230227173632.3292573-11-surenb@google.com
Signed-off-by: Michel Lespinasse <michel@lespinasse.org>
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Per-VMA locks", v4.
LWN article describing the feature: https://lwn.net/Articles/906852/
Per-vma locks idea that was discussed during SPF [1] discussion at LSF/MM
last year [2], which concluded with suggestion that “a reader/writer
semaphore could be put into the VMA itself; that would have the effect of
using the VMA as a sort of range lock. There would still be contention at
the VMA level, but it would be an improvement.” This patchset implements
this suggested approach.
When handling page faults we lookup the VMA that contains the faulting
page under RCU protection and try to acquire its lock. If that fails we
fall back to using mmap_lock, similar to how SPF handled this situation.
One notable way the implementation deviates from the proposal is the way
VMAs are read-locked. During some of mm updates, multiple VMAs need to be
locked until the end of the update (e.g. vma_merge, split_vma, etc).
Tracking all the locked VMAs, avoiding recursive locks, figuring out when
it's safe to unlock previously locked VMAs would make the code more
complex. So, instead of the usual lock/unlock pattern, the proposed
solution marks a VMA as locked and provides an efficient way to:
1. Identify locked VMAs.
2. Unlock all locked VMAs in bulk.
We also postpone unlocking the locked VMAs until the end of the update,
when we do mmap_write_unlock. Potentially this keeps a VMA locked for
longer than is absolutely necessary but it results in a big reduction of
code complexity.
Read-locking a VMA is done using two sequence numbers - one in the
vm_area_struct and one in the mm_struct. VMA is considered read-locked
when these sequence numbers are equal. To read-lock a VMA we set the
sequence number in vm_area_struct to be equal to the sequence number in
mm_struct. To unlock all VMAs we increment mm_struct's seq number. This
allows for an efficient way to track locked VMAs and to drop the locks on
all VMAs at the end of the update.
The patchset implements per-VMA locking only for anonymous pages which are
not in swap and avoids userfaultfs as their implementation is more
complex. Additional support for file-back page faults, swapped and user
pages can be added incrementally.
Performance benchmarks show similar although slightly smaller benefits as
with SPF patchset (~75% of SPF benefits). Still, with lower complexity
this approach might be more desirable.
Since RFC was posted in September 2022, two separate Google teams outside
of Android evaluated the patchset and confirmed positive results. Here
are the known usecases when per-VMA locks show benefits:
Android:
Apps with high number of threads (~100) launch times improve by up to 20%.
Each thread mmaps several areas upon startup (Stack and Thread-local
storage (TLS), thread signal stack, indirect ref table), which requires
taking mmap_lock in write mode. Page faults take mmap_lock in read mode.
During app launch, both thread creation and page faults establishing the
active workinget are happening in parallel and that causes lock contention
between mm writers and readers even if updates and page faults are
happening in different VMAs. Per-vma locks prevent this contention by
providing more granular lock.
Google Fibers:
We have several dynamically sized thread pools that spawn new threads
under increased load and reduce their number when idling. For example,
Google's in-process scheduling/threading framework, UMCG/Fibers, is backed
by such a thread pool. When idling, only a small number of idle worker
threads are available; when a spike of incoming requests arrive, each
request is handled in its own "fiber", which is a work item posted onto a
UMCG worker thread; quite often these spikes lead to a number of new
threads spawning. Each new thread needs to allocate and register an RSEQ
section on its TLS, then register itself with the kernel as a UMCG worker
thread, and only after that it can be considered by the in-process
UMCG/Fiber scheduler as available to do useful work. In short, during an
incoming workload spike new threads have to be spawned, and they perform
several syscalls (RSEQ registration, UMCG worker registration, memory
allocations) before they can actually start doing useful work. Removing
any bottlenecks on this thread startup path will greatly improve our
services' latencies when faced with request/workload spikes.
At high scale, mmap_lock contention during thread creation and stack page
faults leads to user-visible multi-second serving latencies in a similar
pattern to Android app startup. Per-VMA locking patchset has been run
successfully in limited experiments with user-facing production workloads.
In these experiments, we observed that the peak thread creation rate was
high enough that thread creation is no longer a bottleneck.
TCP zerocopy receive:
From the point of view of TCP zerocopy receive, the per-vma lock patch is
massively beneficial.
In today's implementation, a process with N threads where N - 1 are
performing zerocopy receive and 1 thread is performing madvise() with the
write lock taken (e.g. needs to change vm_flags) will result in all N -1
receive threads blocking until the madvise is done. Conversely, on a busy
process receiving a lot of data, an madvise operation that does need to
take the mmap lock in write mode will need to wait for all of the receives
to be done - a lose:lose proposition. Per-VMA locking _removes_ by
definition this source of contention entirely.
There are other benefits for receive as well, chiefly a reduction in
cacheline bouncing across receiving threads for locking/unlocking the
single mmap lock. On an RPC style synthetic workload with 4KB RPCs:
1a) The find+lock+unlock VMA path in the base case, without the
per-vma lock patchset, is about 0.7% of cycles as measured by perf.
1b) mmap_read_lock + mmap_read_unlock in the base case is about 0.5%
cycles overall - most of this is within the TCP read hotpath (a small
fraction is 'other' usage in the system).
2a) The find+lock+unlock VMA path, with the per-vma patchset and a
trivial patch written to take advantage of it in TCP, is about 0.4% of
cycles (down from 0.7% above)
2b) mmap_read_lock + mmap_read_unlock in the per-vma patchset is <
0.1% cycles and is out of the TCP read hotpath entirely (down from
0.5% before, the remaining usage is the 'other' usage in the system).
So, in addition to entirely removing an onerous source of contention,
it also reduces the CPU cycles of TCP receive zerocopy by about 0.5%+
(compared to overall cycles in perf) for the 'small' RPC scenario.
In https://lkml.kernel.org/r/87fsaqouyd.fsf_-_@stealth, Punit
demonstrated throughput improvements of as much as 188% from this
patchset.
This patch (of 25):
This configuration variable will be used to build the support for VMA
locking during page fault handling.
This is enabled on supported architectures with SMP and MMU set.
The architecture support is needed since the page fault handler is called
from the architecture's page faulting code which needs modifications to
handle faults under VMA lock.
Link: https://lkml.kernel.org/r/20230227173632.3292573-1-surenb@google.com
Link: https://lkml.kernel.org/r/20230227173632.3292573-10-surenb@google.com
Signed-off-by: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Since SQE memory is shared with userspace, we should only be reading it
once. We cannot read it multiple times, particularly when it's read once
for validation and then read again for the actual use.
ublk_ch_uring_cmd() is safe when called as a retry operation, as the
memory backing is stable at that point. But for normal issue, we want
to ensure that we only read ublksrv_io_cmd once. Wrap the function in
a helper that reads the value into an on-stack copy of the struct.
Cc: stable@vger.kernel.org # 6.0+
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Prevent filesystems from doing things which sleep in their map_pages
method. This is in preparation for a pagefault path protected only by
RCU.
Link: https://lkml.kernel.org/r/20230327174515.1811532-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
For the map_pages() method, we need a test that does not sleep. The page
fault handler will continue to call the fault() method where we can sleep
and do the full revalidation there.
Link: https://lkml.kernel.org/r/20230327174515.1811532-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: David Howells <dhowells@redhat.com>
Tested-by: David Howells <dhowells@redhat.com>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Prevent ->map_pages from sleeping", v2.
In preparation for a larger patch series which will handle (some, easy)
page faults protected only by RCU, change the two filesystems which have
sleeping locks to not take them and hold the RCU lock around calls to
->map_page to prevent other filesystems from adding sleeping locks.
This patch (of 3):
XFS doesn't actually need to be holding the XFS_MMAPLOCK_SHARED to do
this. filemap_map_pages() cannot bring new folios into the page cache
and the folio lock is taken during filemap_map_pages() which provides
sufficient protection against a truncation or hole punch.
Link: https://lkml.kernel.org/r/20230327174515.1811532-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20230327174515.1811532-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Since commit ee6d3dd4ed ("driver core: make kobj_type constant.") the
driver core allows the usage of const struct kobj_type.
Take advantage of this to constify the structure definition to prevent
modification at runtime.
These structures were not constified in commit e56397e8c4
("mm/damon/sysfs: make kobj_type structures constant") as they didn't
exist when that patch was written.
Link: https://lkml.kernel.org/r/20230324-b4-kobj_type-damon2-v1-1-48ddbf1c8fcf@weissschuh.net
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Reviewed-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The kernel's default behaviour is to obstruct the allocation of high
virtual address as it handles memory overcommit in a heuristic manner.
Setting the parameter as OVERCOMMIT_ALWAYS, ensures kernel isn't
susceptible to the availability of a platform's physical memory when
denying a memory allocation request.
Link: https://lkml.kernel.org/r/20230323060121.1175830-4-chaitanyas.prakash@arm.com
Signed-off-by: Chaitanya S Prakash <chaitanyas.prakash@arm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Although there is a provision for 52 bit VA on arm64 platform, it remains
unutilised and higher addresses are not allocated. In order to
accommodate 4PB [2^52] virtual address space where supported,
NR_CHUNKS_HIGH is changed accordingly.
Array holding addresses is changed from static allocation to dynamic
allocation to accommodate its voluminous nature which otherwise might
overflow the stack.
Link: https://lkml.kernel.org/r/20230323060121.1175830-3-chaitanyas.prakash@arm.com
Signed-off-by: Chaitanya S Prakash <chaitanyas.prakash@arm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "selftests: Fix virtual address range for arm64", v2.
When the virtual address range selftest is run on arm64 and x86 platforms,
it is observed that both the low and high VA range iterations are skipped
when the MAP_CHUNK_SIZE is set to 16GB. The MAP_CHUNK_SIZE is changed to
1GB to resolve this issue, following which support for arm64 platform is
added by changing the NR_CHUNKS_HIGH for aarch64 to accommodate up to 4PB
of virtual address space allocation requests. Dynamic memory allocation
of array holding addresses is introduced to prevent overflow of the stack.
Finally, the overcommit_policy is set as OVERCOMMIT_ALWAYS to prevent the
kernel from denying a memory allocation request based on a platform's
physical memory availability.
This patch (of 3):
mmap() fails to allocate 16GB virtual space chunk, skipping both low and
high VA range iterations. Hence, reduce MAP_CHUNK_SIZE to 1GB and update
relevant macros as required.
Link: https://lkml.kernel.org/r/20230323060121.1175830-1-chaitanyas.prakash@arm.com
Link: https://lkml.kernel.org/r/20230323060121.1175830-2-chaitanyas.prakash@arm.com
Signed-off-by: Chaitanya S Prakash <chaitanyas.prakash@arm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
