mirror of
https://github.com/torvalds/linux.git
synced 2024-11-24 21:21:41 +00:00
891e8abed5
- Use BPF + BTF to collect and pretty print syscall and tracepoint arguments in
'perf trace', done as an GSoC activity.
- Data-type profiling improvements:
- Cache debuginfo to speed up data type resolution.
- Add the 'typecln' sort order, to show which cacheline in a target is hot or
cold. The following shows members in the cfs_rq's first cache line:
$ perf report -s type,typecln,typeoff -H
...
- 2.67% struct cfs_rq
+ 1.23% struct cfs_rq: cache-line 2
+ 0.57% struct cfs_rq: cache-line 4
+ 0.46% struct cfs_rq: cache-line 6
- 0.41% struct cfs_rq: cache-line 0
0.39% struct cfs_rq +0x14 (h_nr_running)
0.02% struct cfs_rq +0x38 (tasks_timeline.rb_leftmost)
- When a typedef resolves to a unnamed struct, use the typedef name.
- When a struct has just one basic type field (int, etc), resolve the type
sort order to the name of the struct, not the type of the field.
- Support type folding/unfolding in the data-type annotation TUI.
- Fix bitfields offsets and sizes.
- Initial support for PowerPC, using libcapstone and the usual objdump
disassembly parsing routines.
- Add support for disassembling and addr2line using the LLVM libraries,
speeding up those operations.
- Support --addr2line option in 'perf script' as with other tools.
- Intel branch counters (LBR event logging) support, only available in recent
Intel processors, for instance, the new "brcntr" field can be asked from
'perf script' to print the information collected from this feature:
$ perf script -F +brstackinsn,+brcntr
# Branch counter abbr list:
# branch-instructions:ppp = A
# branch-misses = B
# '-' No event occurs
# '+' Event occurrences may be lost due to branch counter saturated
tchain_edit 332203 3366329.405674: 53030 branch-instructions:ppp: 401781 f3+0x2c (home/sdp/test/tchain_edit)
f3+31:
0000000000401774 insn: eb 04 br_cntr: AA # PRED 5 cycles [5]
000000000040177a insn: 81 7d fc 0f 27 00 00
0000000000401781 insn: 7e e3 br_cntr: A # PRED 1 cycles [6] 2.00 IPC
0000000000401766 insn: 8b 45 fc
0000000000401769 insn: 83 e0 01
000000000040176c insn: 85 c0
000000000040176e insn: 74 06 br_cntr: A # PRED 1 cycles [7] 4.00 IPC
0000000000401776 insn: 83 45 fc 01
000000000040177a insn: 81 7d fc 0f 27 00 00
0000000000401781 insn: 7e e3 br_cntr: A # PRED 7 cycles [14] 0.43 IPC
- Support Timed PEBS (Precise Event-Based Sampling), a recent hardware feature
in Intel processors.
- Add 'perf ftrace profile' subcommand, using ftrace's function-graph tracer so
that users can see the total, average, max execution time as well as the
number of invocations easily, for instance:
$ sudo perf ftrace profile -G __x64_sys_perf_event_open -- \
perf stat -e cycles -C1 true 2> /dev/null | head
# Total (us) Avg (us) Max (us) Count Function
65.611 65.611 65.611 1 __x64_sys_perf_event_open
30.527 30.527 30.527 1 anon_inode_getfile
30.260 30.260 30.260 1 __anon_inode_getfile
29.700 29.700 29.700 1 alloc_file_pseudo
17.578 17.578 17.578 1 d_alloc_pseudo
17.382 17.382 17.382 1 __d_alloc
16.738 16.738 16.738 1 kmem_cache_alloc_lru
15.686 15.686 15.686 1 perf_event_alloc
14.012 7.006 11.264 2 obj_cgroup_charge
#
- 'perf sched timehist' improvements, including the addition of priority
showing/filtering command line options.
- Varios improvements to the 'perf probe', including 'perf test' regression
testings.
- Introduce the 'perf check', initially to check if some feature is in place,
using it in 'perf test'.
- Various fixes for 32-bit systems.
- Address more leak sanitizer failures.
- Fix memory leaks (LBR, disasm lock ops, etc).
- More reference counting fixes (branch_info, etc).
- Constify 'struct perf_tool' parameters to improve code generation and reduce
the chances of having its internals changed, which isn't expected.
- More constifications in various other places.
- Add more build tests, including for JEVENTS.
- Add more 'perf test' entries ('perf record LBR', pipe/inject, --setup-filter,
'perf ftrace', 'cgroup sampling', etc).
- Inject build ids for all entries in a call chain in 'perf inject', not just
for the main sample.
- Improve the BPF based sample filter, allowing root to setup filters in bpffs
that then can be used by non-root users.
- Allow filtering by cgroups with the BPF based sample filter.
- Allow a more compact way for 'perf mem report' using the -T/--type-profile and
also provide a --sort option similar to the one in 'perf report', 'perf top',
to setup the sort order manually.
- Fix --group behavior in 'perf annotate' when leader has no samples, where it
was not showing anything even when other events in the group had samples.
- Fix spinlock and rwlock accounting in 'perf lock contention'
- Fix libsubcmd fixdep Makefile dependencies.
- Improve 'perf ftrace' error message when ftrace isn't available.
- Update various Intel JSON vendor event files.
- ARM64 CoreSight hardware tracing infrastructure improvements, mostly not
visible to users.
- Update power10 JSON events.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQR2GiIUctdOfX2qHhGyPKLppCJ+JwUCZuwxgwAKCRCyPKLppCJ+
JxfHAQCrgSD4itg4HA7znUoYBEGL73NisJT2Juq0lyDK2gniOQD+Mln6isvRnMag
k7BFXvgHj/LDQdOznkG2pojSFJcSgQo=
=kazH
-----END PGP SIGNATURE-----
Merge tag 'perf-tools-for-v6.12-1-2024-09-19' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools
Pull perf tools updates from Arnaldo Carvalho de Melo:
- Use BPF + BTF to collect and pretty print syscall and tracepoint
arguments in 'perf trace', done as an GSoC activity
- Data-type profiling improvements:
- Cache debuginfo to speed up data type resolution
- Add the 'typecln' sort order, to show which cacheline in a target
is hot or cold. The following shows members in the cfs_rq's first
cache line:
$ perf report -s type,typecln,typeoff -H
...
- 2.67% struct cfs_rq
+ 1.23% struct cfs_rq: cache-line 2
+ 0.57% struct cfs_rq: cache-line 4
+ 0.46% struct cfs_rq: cache-line 6
- 0.41% struct cfs_rq: cache-line 0
0.39% struct cfs_rq +0x14 (h_nr_running)
0.02% struct cfs_rq +0x38 (tasks_timeline.rb_leftmost)
- When a typedef resolves to a unnamed struct, use the typedef name
- When a struct has just one basic type field (int, etc), resolve
the type sort order to the name of the struct, not the type of
the field
- Support type folding/unfolding in the data-type annotation TUI
- Fix bitfields offsets and sizes
- Initial support for PowerPC, using libcapstone and the usual
objdump disassembly parsing routines
- Add support for disassembling and addr2line using the LLVM libraries,
speeding up those operations
- Support --addr2line option in 'perf script' as with other tools
- Intel branch counters (LBR event logging) support, only available in
recent Intel processors, for instance, the new "brcntr" field can be
asked from 'perf script' to print the information collected from this
feature:
$ perf script -F +brstackinsn,+brcntr
# Branch counter abbr list:
# branch-instructions:ppp = A
# branch-misses = B
# '-' No event occurs
# '+' Event occurrences may be lost due to branch counter saturated
tchain_edit 332203 3366329.405674: 53030 branch-instructions:ppp: 401781 f3+0x2c (home/sdp/test/tchain_edit)
f3+31:
0000000000401774 insn: eb 04 br_cntr: AA # PRED 5 cycles [5]
000000000040177a insn: 81 7d fc 0f 27 00 00
0000000000401781 insn: 7e e3 br_cntr: A # PRED 1 cycles [6] 2.00 IPC
0000000000401766 insn: 8b 45 fc
0000000000401769 insn: 83 e0 01
000000000040176c insn: 85 c0
000000000040176e insn: 74 06 br_cntr: A # PRED 1 cycles [7] 4.00 IPC
0000000000401776 insn: 83 45 fc 01
000000000040177a insn: 81 7d fc 0f 27 00 00
0000000000401781 insn: 7e e3 br_cntr: A # PRED 7 cycles [14] 0.43 IPC
- Support Timed PEBS (Precise Event-Based Sampling), a recent hardware
feature in Intel processors
- Add 'perf ftrace profile' subcommand, using ftrace's function-graph
tracer so that users can see the total, average, max execution time
as well as the number of invocations easily, for instance:
$ sudo perf ftrace profile -G __x64_sys_perf_event_open -- \
perf stat -e cycles -C1 true 2> /dev/null | head
# Total (us) Avg (us) Max (us) Count Function
65.611 65.611 65.611 1 __x64_sys_perf_event_open
30.527 30.527 30.527 1 anon_inode_getfile
30.260 30.260 30.260 1 __anon_inode_getfile
29.700 29.700 29.700 1 alloc_file_pseudo
17.578 17.578 17.578 1 d_alloc_pseudo
17.382 17.382 17.382 1 __d_alloc
16.738 16.738 16.738 1 kmem_cache_alloc_lru
15.686 15.686 15.686 1 perf_event_alloc
14.012 7.006 11.264 2 obj_cgroup_charge
- 'perf sched timehist' improvements, including the addition of
priority showing/filtering command line options
- Varios improvements to the 'perf probe', including 'perf test'
regression testings
- Introduce the 'perf check', initially to check if some feature is
in place, using it in 'perf test'
- Various fixes for 32-bit systems
- Address more leak sanitizer failures
- Fix memory leaks (LBR, disasm lock ops, etc)
- More reference counting fixes (branch_info, etc)
- Constify 'struct perf_tool' parameters to improve code generation
and reduce the chances of having its internals changed, which isn't
expected
- More constifications in various other places
- Add more build tests, including for JEVENTS
- Add more 'perf test' entries ('perf record LBR', pipe/inject,
--setup-filter, 'perf ftrace', 'cgroup sampling', etc)
- Inject build ids for all entries in a call chain in 'perf inject',
not just for the main sample
- Improve the BPF based sample filter, allowing root to setup filters
in bpffs that then can be used by non-root users
- Allow filtering by cgroups with the BPF based sample filter
- Allow a more compact way for 'perf mem report' using the
-T/--type-profile and also provide a --sort option similar to the one
in 'perf report', 'perf top', to setup the sort order manually
- Fix --group behavior in 'perf annotate' when leader has no samples,
where it was not showing anything even when other events in the group
had samples
- Fix spinlock and rwlock accounting in 'perf lock contention'
- Fix libsubcmd fixdep Makefile dependencies
- Improve 'perf ftrace' error message when ftrace isn't available
- Update various Intel JSON vendor event files
- ARM64 CoreSight hardware tracing infrastructure improvements, mostly
not visible to users
- Update power10 JSON events
* tag 'perf-tools-for-v6.12-1-2024-09-19' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools: (310 commits)
perf trace: Mark the 'head' arg in the set_robust_list syscall as coming from user space
perf trace: Mark the 'rseq' arg in the rseq syscall as coming from user space
perf env: Find correct branch counter info on hybrid
perf evlist: Print hint for group
tools: Drop nonsensical -O6
perf pmu: To info add event_type_desc
perf evsel: Add accessor for tool_event
perf pmus: Fake PMU clean up
perf list: Avoid potential out of bounds memory read
perf help: Fix a typo ("bellow")
perf ftrace: Detect whether ftrace is enabled on system
perf test shell probe_vfs_getname: Remove extraneous '=' from probe line number regex
perf build: Require at least clang 16.0.6 to build BPF skeletons
perf trace: If a syscall arg is marked as 'const', assume it is coming _from_ userspace
perf parse-events: Remove duplicated include in parse-events.c
perf callchain: Allow symbols to be optional when resolving a callchain
perf inject: Lazy build-id mmap2 event insertion
perf inject: Add new mmap2-buildid-all option
perf inject: Fix build ID injection
perf annotate-data: Add pr_debug_scope()
...
|
||
|---|---|---|
| .. | ||
| api | ||
| bpf | ||
| perf | ||
| subcmd | ||
| symbol | ||
| thermal | ||
| argv_split.c | ||
| bitmap.c | ||
| ctype.c | ||
| find_bit.c | ||
| hweight.c | ||
| list_sort.c | ||
| rbtree.c | ||
| slab.c | ||
| str_error_r.c | ||
| string.c | ||
| vsprintf.c | ||
| zalloc.c | ||