linux.git/tools/perf, branch v6.18.21

perf synthetic-events: Fix stale build ID in module MMAP2 records

2026-03-19T15:08:20+00:00

[ Upstream commit 35b16a7a2c4fc458304447128b86514ce9f70f3c ]

perf_event__synthesize_modules() allocates a single union perf_event and
reuses it across every kernel module callback.

After the first module is processed, perf_record_mmap2__read_build_id()
sets PERF_RECORD_MISC_MMAP_BUILD_ID in header.misc and writes that
module's build ID into the event.

On subsequent iterations the callback overwrites start, len, pid, and
filename for the next module but never clears the stale build ID fields
or the MMAP_BUILD_ID flag.

When perf_record_mmap2__read_build_id() runs for the second module it
sees the flag, reads the stale build ID into a dso_id, and
__dso__improve_id() permanently poisons the DSO with the wrong build ID.

Every module after the first therefore receives the first module's build
ID in its MMAP2 record.

On a system with the sunrpc and nfsd modules loaded, this causes perf
script and perf report to show [unknown] for all module symbols.

The latent bug has existed since commit d9f2ecbc5e47fca7 ("perf dso:
Move build_id to dso_id") introduced the PERF_RECORD_MISC_MMAP_BUILD_ID
check in perf_record_mmap2__read_build_id().

Commit 53b00ff358dc75b1 ("perf record: Make --buildid-mmap the default")
then exposed it to all users by making the MMAP2-with-build-ID path the
default.  Both commits were merged in the same series.

Clear the MMAP_BUILD_ID flag and zero the build_id union before each
call to perf_record_mmap2__read_build_id() so that every module starts
with a clean slate.

Fixes: d9f2ecbc5e47fca7 ("perf dso: Move build_id to dso_id")
Reviewed-by: Ian Rogers 
Signed-off-by: Chuck Lever 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ian Rogers 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf ftrace: Fix hashmap__new() error checking

2026-03-19T15:08:18+00:00

[ Upstream commit be34705aa527872e5ce83927b7bc9307ba8095ca ]

The hashmap__new() function never returns NULL, it returns error
pointers. Fix the error checking to match.

Additionally, set ftrace->profile_hash to NULL on error, and return the
exact error code from hashmap__new().

Fixes: 0f223813edd051a5 ("perf ftrace: Add 'profile' command")
Suggested-by: Ian Rogers 
Signed-off-by: Chen Ni 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf annotate: Fix hashmap__new() error checking

2026-03-19T15:08:18+00:00

[ Upstream commit bf29cb3641b80bac759c3332b02e0b270e16bf94 ]

The hashmap__new() function never returns NULL, it returns error
pointers. Fix the error checking to match.

Additionally, set src->samples to NULL to prevent any later code from
accidentally using the error pointer.

Fixes: d3e7cad6f36d9e80 ("perf annotate: Add a hashmap for symbol histogram")
Reviewed-by: Ian Rogers 
Signed-off-by: Chen Ni 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Tianyou Li 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf disasm: Fix off-by-one bug in outside check

2026-03-19T15:08:13+00:00

[ Upstream commit b3ce769203a99d6f3c6d6269ec09232a8c5da422 ]

If a branch target points to one past the end of a function, the branch
should be treated as a branch to another function.

This can happen e.g. with a tail call to a function that is laid out
immediately after the caller.

Fixes: 751b1783da784299 ("perf annotate: Mark jumps to outher functions with the call arrow")
Reviewed-by: Ian Rogers 
Signed-off-by: Peter Collingbourne 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Bill Wendling 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Justin Stitt 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Nathan Chancellor 
Cc: Nick Desaulniers 
Cc: Peter Zijlstra 
Link: https://linux-review.googlesource.com/id/Ide471112e82d68177e0faf08ca411d9fcf0a7bdf
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf test: Fix test case perftool-testsuite_report for s390

2026-03-04T12:19:27+00:00

[ Upstream commit 3d012b8614ee020666f3dd15af9f65dc487e3f5f ]

Test case perftool-testsuite_report fails on s390 for some time
now.

Root cause is a time out which is too tight for large s390 machines.
The time out value addr2line_timeout_ms is per default set to 1 second.

This is the maximum time the function read_addr2line_record() waits for
a reply from the forked off tool addr2line, which is started as a child
in interactive mode.

It reads stdin (an address in hexadecimal) and replies on stdout with
function name, file name and line number. This might take more than one
second.

However one second is not always enough and the reply from addr2line
tool is not received. Function read_addr2line_record() fails and emits
a warning, which is not expected by the test case. It fails.

Output before:

 # perf test -F 133
 -- [ PASS ] -- perf_report :: setup :: prepare the perf.data file
 ==================
 [ perf record: Woken up 1 times to write data ]
 [ perf record: Captured and wrote 0.087 MB \
	/tmp/perftool-testsuite_report.FHz/perf_report/perf.data.1 \
	(207 samples) ]
 ==================
 -- [ PASS ] -- perf_report :: setup :: prepare the perf.data.1 file
 ## [ PASS ] ## perf_report :: setup SUMMARY
 -- [ SKIP ] -- perf_report :: test_basic :: help message :: testcase skipped
 Line did not match any pattern: "cmd__addr2line /usr/lib/debug/lib/modules/
 	6.19.0-20260205.rc8.git366.9845cf73f7db.300.fc43.s390x+next/
	vmlinux: could not read first record"
 Line did not match any pattern: "cmd__addr2line /usr/lib/debug/lib/modules/
	6.19.0-20260205.rc8.git366.9845cf73f7db.300.fc43.s390x+next/
	vmlinux: could not read first record"
 -- [ FAIL ] -- perf_report :: test_basic :: basic execution
	(output regexp parsing)
 ....
 133: perftool-testsuite_report      : FAILED!

Output after:

 # ./perf test -F 133
 -- [ PASS ] -- perf_report :: setup :: prepare the perf.data file
 ==================
 [ perf record: Woken up 1 times to write data ]
 [ perf record: Captured and wrote 0.087 MB \
	 /tmp/perftool-testsuite_report.Mlp/perf_report/perf.data.1
	 (188 samples) ]
 ==================
 -- [ PASS ] -- perf_report :: setup :: prepare the perf.data.1 file
 ## [ PASS ] ## perf_report :: setup SUMMARY
 -- [ SKIP ] -- perf_report :: test_basic :: help message :: testcase skipped
 -- [ PASS ] -- perf_report :: test_basic :: basic execution
 -- [ PASS ] -- perf_report :: test_basic :: number of samples
 -- [ PASS ] -- perf_report :: test_basic :: header
 -- [ PASS ] -- perf_report :: test_basic :: header timestamp
 -- [ PASS ] -- perf_report :: test_basic :: show CPU utilization
 -- [ PASS ] -- perf_report :: test_basic :: pid
 -- [ PASS ] -- perf_report :: test_basic :: non-existing symbol
 -- [ PASS ] -- perf_report :: test_basic :: symbol filter
 -- [ PASS ] -- perf_report :: test_basic :: latency header
 -- [ PASS ] -- perf_report :: test_basic :: default report for latency profile
 -- [ PASS ] -- perf_report :: test_basic :: latency report for latency profile
 -- [ PASS ] -- perf_report :: test_basic :: parallelism histogram
 ## [ PASS ] ## perf_report :: test_basic SUMMARY
 133: perftool-testsuite_report      : Ok
 #

Fixes: 257046a36750a6db ("perf srcline: Fallback between addr2line implementations")
Reviewed-by: Jan Polensky 
Signed-off-by: Thomas Richter 
Cc: Alexander Gordeev 
Cc: Heiko Carstens 
Cc: Ian Rogers 
Cc: linux-s390@vger.kernel.org
Cc: Namhyung Kim 
Cc: Sumanth Korikkar 
Cc: Vasily Gorbik 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf build: Remove NO_LIBCAP that controls nothing

2026-03-04T12:19:27+00:00

[ Upstream commit 169343cc8ff2bd59758760d867bd26adae866a2b ]

Using libcap was removed in commit e25ebda78e230283 ("perf cap: Tidy up
and improve capability testing") and improve capability testing"),
however, some build documentation and a use of the NO_LIBCAP=1 were
lingering.

Remove these left over bits.

Fixes: e25ebda78e230283 ("perf cap: Tidy up and improve capability testing")
Signed-off-by: Ian Rogers 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ian Rogers 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf vendor events amd: Fix Zen 5 MAB allocation events

2026-03-04T12:19:26+00:00

[ Upstream commit 76b2cf07a6d2a836108f9c2486d76599f7adf6e8 ]

The unit masks for PMCx041 vary across different generations of Zen
processors.

Fix the Zen 5 events based on PMCx041 as they incorrectly use the same
unit masks as that of Zen 4.

Fixes: 45c072f2537ab07b ("perf vendor events amd: Add Zen 5 core events")
Reported-by: Suyash Mahar 
Reviewed-by: Ian Rogers 
Signed-off-by: Sandipan Das 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ananth Narayan 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Ravi Bangoria 
Cc: Sandipan Das 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf annotate: Fix BUILD_NONDISTRO=1 missing args->ms conversions to pointer

2026-03-04T12:19:26+00:00

[ Upstream commit dda5f926a1006c735b00ed5c27291fce64236656 ]

Fix a few missing conversions to pointer in the usage of 'struct
annotate_args' 'ms' member in symbol__disassemble_bpf_libbfd().

Fixes: 00419892bac28bf1 ("perf annotate: Fix args leak of map_symbol")
Reviewed-by: Ian Rogers 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf annotate: Fix memcpy size in arch__grow_instructions()

2026-03-04T12:19:26+00:00

[ Upstream commit f0d98c78f8bf73ce2a9b7793f66cda240fa9ab10 ]

The memcpy() in arch__grow_instructions() is copying the wrong number of
bytes when growing from a non-allocated table.

It should copy arch->nr_instructions * sizeof(struct ins) bytes, not
just arch->nr_instructions bytes.

This bug causes data corruption as only a partial copy of the
instruction table is made, leading to garbage data in most entries and
potential crashes

Fixes: 2a1ff812c40be982 ("perf annotate: Introduce alternative method of keeping instructions table")
Reviewed-by: Ian Rogers 
Signed-off-by: Suchit Karunakaran 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf tests sched: Avoid error in cleanup on loaded machines

2026-03-04T12:19:25+00:00

[ Upstream commit c5e47e4d00fbc15f2390bb6ed8d9c21836363291 ]

The stop_noploops function will kill the noploop processes that are
running for 10 seconds.

On a loaded machine they may have already terminated meaning the kill
will return an error of no such process.

This doesn't matter and so ignore the error to avoid the test
terminating in the cleanup.

Fixes: 0e22c5ca44e68798 ("perf test: Add sched latency and script shell tests")
Signed-off-by: Ian Rogers 
Tested-by: Arnaldo Carvalho de Melo 
Cc: Adrian Hunter 
Cc: Alexander Shishkin 
Cc: Ian Rogers 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin