Pull perf tools fixes from Arnaldo Carvalho de Melo:
- Fix tools/ quiet build Makefile infrastructure that was broken when
working on tools/perf/ without testing on other tools/ living
utilities.
* tag 'perf-tools-fixes-for-v6.14-2-2025-02-25' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools:
tools: Remove redundant quiet setup
tools: Unify top-level quiet infrastructure
Pull perf-tools updates from Namhyung Kim:
"There are a lot of changes in the perf tools in this cycle.
build:
- Use generic syscall table to generate syscall numbers on supported
archs
- This also enables to get rid of libaudit which was used for syscall
numbers
- Remove python2 support as it's deprecated for years
- Fix issues on static build with libzstd
perf record:
- Intel-PT supports "aux-action" config term to pause or resume
tracing in the aux-buffer. Users can start the intel_pt event as
"started-paused" and configure other events to control the Intel-PT
tracing:
# perf record --kcore -e intel_pt/aux-action=start-paused/ \
-e syscalls:sys_enter_newuname/aux-action=resume/ \
-e syscalls:sys_exit_newuname/aux-action=pause/ -- uname
This requires kernel support (which was added in v6.13)
perf lock:
- 'perf lock contention' command has an ability to symbolize locks in
dynamically allocated objects using slab cache name when it runs
with BPF. Those dynamic locks would have "&" prefix in the name to
distinguish them from ordinary (static) locks
# perf lock con -abl -E 5 sleep 1
contended total wait max wait avg wait address symbol
2 1.95 us 1.77 us 975 ns ffff9d5e852d3498 &task_struct (mutex)
1 1.18 us 1.18 us 1.18 us ffff9d5e852d3538 &task_struct (mutex)
4 1.12 us 354 ns 279 ns ffff9d5e841ca800 &kmalloc-cg-512 (mutex)
2 859 ns 617 ns 429 ns ffffffffa41c3620 delayed_uprobe_lock (mutex)
3 691 ns 388 ns 230 ns ffffffffa41c0940 pack_mutex (mutex)
This also requires kernel/BPF support (which was added in v6.13)
perf ftrace:
- 'perf ftrace latency' command gets a couple of options to support
linear buckets instead of exponential. Also it's possible to
specify max and min latency for the linear buckets:
# perf ftrace latency -abn -T switch_mm_irqs_off --bucket-range=100 \
--min-latency=200 --max-latency=800 -- sleep 1
# DURATION | COUNT | GRAPH |
0 - 200 ns | 186 | ### |
200 - 300 ns | 256 | ##### |
300 - 400 ns | 364 | ####### |
400 - 500 ns | 223 | #### |
500 - 600 ns | 111 | ## |
600 - 700 ns | 41 | |
700 - 800 ns | 141 | ## |
800 - ... ns | 169 | ### |
# statistics (in nsec)
total time: 2162212
avg time: 967
max time: 16817
min time: 132
count: 2236
- As you can see in the above example, it nows shows the statistics
at the end so that users can see the avg/max/min latencies easily
- 'perf ftrace profile' command has --graph-opts option like 'perf
ftrace trace' so that it can control the tracing behaviors in the
same way. For example, it can limit the function call depth or
threshold
perf script:
- Improve physical memory resolution in 'mem-phys-addr' script by
parsing /proc/iomem file
# perf script mem-phys-addr -- find /
...
Event: mem_inst_retired.all_loads:P
Memory type count percentage
---------------------------------------- ---------- ----------
100000000-85f7fffff : System RAM 8929 69.7
547600000-54785d23f : Kernel data 1240 9.7
546a00000-5474bdfff : Kernel rodata 490 3.8
5480ce000-5485fffff : Kernel bss 121 0.9
0-fff : Reserved 3860 30.1
100000-89c01fff : System RAM 18 0.1
8a22c000-8df6efff : System RAM 5 0.0
Others:
- 'perf test' gets --runs-per-test option to run the test cases
repeatedly. This would be helpful to see if it's flaky
- Add 'parse_events' method to Python perf extension module, so that
users can use the same event parsing logic in the python code. One
more step towards implementing perf tools in Python. :)
- Support opening tracepoint events without libtraceevent. This will
be helpful if it won't use the tracing data like in 'perf stat'
- Update ARM Neoverse N2/V2 JSON events and metrics"
* tag 'perf-tools-for-v6.14-2025-01-21' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools: (176 commits)
perf test: Update event_groups test to use instructions
perf bench: Fix undefined behavior in cmpworker()
perf annotate: Prefer passing evsel to evsel->core.idx
perf lock: Rename fields in lock_type_table
perf lock: Add percpu-rwsem for type filter
perf lock: Fix parse_lock_type which only retrieve one lock flag
perf lock: Fix return code for functions in __cmd_contention
perf hist: Fix width calculation in hpp__fmt()
perf hist: Fix bogus profiles when filters are enabled
perf hist: Deduplicate cmp/sort/collapse code
perf test: Improve verbose documentation
perf test: Add a runs-per-test flag
perf test: Fix parallel/sequential option documentation
perf test: Send list output to stdout rather than stderr
perf test: Rename functions and variables for better clarity
perf tools: Expose quiet/verbose variables in Makefile.perf
perf config: Add a function to set one variable in .perfconfig
perf test perftool_testsuite: Return correct value for skipping
perf test perftool_testsuite: Add missing description
perf test record+probe_libc_inet_pton: Make test resilient
...
Some versions of kernel were stripping out '.llvm.<hash>' suffix from
kerne symbols (produced by Clang LTO compilation) from function names
reported in available_filter_functions, while kallsyms reported full
original name. This confuses libbpf's multi-kprobe logic of finding all
matching kernel functions for specified user glob pattern by joining
available_filter_functions and kallsyms contents, because joining by
full symbol name won't work for symbols containing '.llvm.<hash>' suffix.
This was eventually fixed by [0] in the kernel, but we'd like to not
regress multi-kprobe experience and add a work around for this bug on
libbpf side, stripping kallsym's name if it matches user pattern and
contains '.llvm.' suffix.
[0] fb6a421fb6 ("kallsyms: Match symbols exactly with CONFIG_LTO_CLANG")
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Link: https://lore.kernel.org/bpf/20250117003957.179331-1-andrii@kernel.org
When redirecting the split BTF to the vmlinux base BTF, we need to mark
the distilled base struct/union members of split BTF structs/unions in
id_map with BTF_IS_EMBEDDED. This indicates that these types must match
both name and size later. Therefore, we need to traverse the entire
split BTF, which involves traversing type IDs from nr_dist_base_types to
nr_types. However, the current implementation uses an incorrect
traversal end type ID, so let's correct it.
Fixes: 19e00c897d ("libbpf: Split BTF relocation")
Signed-off-by: Pu Lehui <pulehui@huawei.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20250115100241.4171581-3-pulehui@huaweicloud.com
Jordan reported an issue in Meta production environment where func
try_to_wake_up() is renamed to try_to_wake_up.llvm.<hash>() by clang
compiler at lto mode. The original 'kprobe/try_to_wake_up' does not
work any more since try_to_wake_up() does not match the actual func
name in /proc/kallsyms.
There are a couple of ways to resolve this issue. For example, in
attach_kprobe(), we could do lookup in /proc/kallsyms so try_to_wake_up()
can be replaced by try_to_wake_up.llvm.<hach>(). Or we can force users
to use bpf_program__attach_kprobe() where they need to lookup
/proc/kallsyms to find out try_to_wake_up.llvm.<hach>(). But these two
approaches requires extra work by either libbpf or user.
Luckily, suggested by Andrii, multi kprobe already supports wildcard ('*')
for symbol matching. In the above example, 'try_to_wake_up*' can match
to try_to_wake_up() or try_to_wake_up.llvm.<hash>() and this allows
bpf prog works for different kernels as some kernels may have
try_to_wake_up() and some others may have try_to_wake_up.llvm.<hash>().
The original intention is to kprobe try_to_wake_up() only, so an optional
field unique_match is added to struct bpf_kprobe_multi_opts. If the
field is set to true, the number of matched functions must be one.
Otherwise, the attachment will fail. In the above case, multi kprobe
with 'try_to_wake_up*' and unique_match preserves user functionality.
Reported-by: Jordan Rome <linux@jordanrome.com>
Suggested-by: Andrii Nakryiko <andrii@kernel.org>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20250109174023.3368432-1-yonghong.song@linux.dev
Starting from 105ff5339f ("mm/memfd: add MFD_NOEXEC_SEAL and
MFD_EXEC") and until 1717449b44 ("memfd: drop warning for missing
exec-related flags"), the kernel would print a warning if neither
MFD_NOEXEC_SEAL nor MFD_EXEC is set in memfd_create().
If libbpf runs on on a kernel between these two commits (eg. on an
improperly backported system), it'll trigger this warning.
To avoid this warning (and also be more secure), explicitly set
MFD_NOEXEC_SEAL. But since libbpf can be run on potentially very old
kernels, leave a fallback for kernels without MFD_NOEXEC_SEAL support.
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Link: https://lore.kernel.org/r/6e62c2421ad7eb1da49cbf16da95aaaa7f94d394.1735594195.git.dxu@dxuuu.xyz
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The new_fd and add_fd functions correspond to the original new and
add_file functions, but accept an FD instead of a file name. This
gives API consumers the option of using anonymous files/memfds to
avoid writing ELFs to disk.
This new API will be useful for performing linking as part of
bpftrace's JIT compilation.
The add_buf function is a convenience wrapper that does the work of
creating a memfd for the caller.
Signed-off-by: Alastair Robertson <ajor@meta.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20241211164030.573042-3-ajor@meta.com
Move the filename arguments and file-descriptor handling from
init_output_elf() and linker_load_obj_file() and instead handle them
at the top-level in bpf_linker__new() and bpf_linker__add_file().
This will allow the inner functions to be shared with a new,
non-filename-based, API in the next commit.
Signed-off-by: Alastair Robertson <ajor@meta.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20241211164030.573042-2-ajor@meta.com
Since the linked fixes: commit, specifying a CPU on hybrid platforms
results in an error because Perf tries to open an extended type event
on "any" CPU which isn't valid. Extended type events can only be opened
on CPUs that match the type.
Before (working):
$ perf record --cpu 1 -- true
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 2.385 MB perf.data (7 samples) ]
After (not working):
$ perf record -C 1 -- true
WARNING: A requested CPU in '1' is not supported by PMU 'cpu_atom' (CPUs 16-27) for event 'cycles:P'
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (cpu_atom/cycles:P/).
/bin/dmesg | grep -i perf may provide additional information.
(Ignore the warning message, that's expected and not particularly
relevant to this issue).
This is because perf_cpu_map__intersect() of the user specified CPU (1)
and one of the PMU's CPUs (16-27) correctly results in an empty (NULL)
CPU map. However for the purposes of opening an event, libperf converts
empty CPU maps into an any CPU (-1) which the kernel rejects.
Fix it by deleting evsels with empty CPU maps in the specific case where
user requested CPU maps are evaluated.
Fixes: 251aa04024 ("perf parse-events: Wildcard most "numeric" events")
Reviewed-by: Ian Rogers <irogers@google.com>
Tested-by: Thomas Falcon <thomas.falcon@intel.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: https://lore.kernel.org/r/20241114160450.295844-2-james.clark@linaro.org
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The perf_cpu_map__merge() function has two arguments, 'orig' and
'other'. The function definition might cause confusion as it could give
the impression that the CPU maps in the two arguments are copied into a
new allocated structure, which is then returned as the result.
The purpose of the function is to merge the CPU map 'other' into the CPU
map 'orig'. This commit changes the 'orig' argument to a pointer to
pointer, so the new result will be updated into 'orig'.
The return value is changed to an int type, as an error number or 0 for
success.
Update callers and tests for the new function definition.
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Leo Yan <leo.yan@arm.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20241107125308.41226-2-leo.yan@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Libelf functions do not set errno on failure. Instead, it relies on its
internal _elf_errno value, that can be retrieved via elf_errno (or the
corresponding message via elf_errmsg()). From "man libelf":
If a libelf function encounters an error it will set an internal
error code that can be retrieved with elf_errno. Each thread
maintains its own separate error code. The meaning of each error
code can be determined with elf_errmsg, which returns a string
describing the error.
As a consequence, libbpf should not return -errno when a function from
libelf fails, because an empty value will not be interpreted as an error
and won't prevent the program to stop. This is visible in
bpf_linker__add_file(), for example, where we call a succession of
functions that rely on libelf:
err = err ?: linker_load_obj_file(linker, filename, opts, &obj);
err = err ?: linker_append_sec_data(linker, &obj);
err = err ?: linker_append_elf_syms(linker, &obj);
err = err ?: linker_append_elf_relos(linker, &obj);
err = err ?: linker_append_btf(linker, &obj);
err = err ?: linker_append_btf_ext(linker, &obj);
If the object file that we try to process is not, in fact, a correct
object file, linker_load_obj_file() may fail with errno not being set,
and return 0. In this case we attempt to run linker_append_elf_sysms()
and may segfault.
This can happen (and was discovered) with bpftool:
$ bpftool gen object output.o sample_ret0.bpf.c
libbpf: failed to get ELF header for sample_ret0.bpf.c: invalid `Elf' handle
zsh: segmentation fault (core dumped) bpftool gen object output.o sample_ret0.bpf.c
Fix the issue by returning a non-null error code (-EINVAL) when libelf
functions fail.
Fixes: faf6ed321c ("libbpf: Add BPF static linker APIs")
Signed-off-by: Quentin Monnet <qmo@kernel.org>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20241205135942.65262-1-qmo@kernel.org
When running `bpftool` on a kernel module installed in `/lib/modules...`,
this error is encountered if the user does not specify `--base-btf` to
point to a valid base BTF (e.g. usually in `/sys/kernel/btf/vmlinux`).
However, looking at the debug output to determine the cause of the error
simply says `Invalid BTF string section`, which does not point to the
actual source of the error. This just improves that debug message to tell
users what happened.
Signed-off-by: Ben Olson <matthew.olson@intel.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/Z0YqzQ5lNz7obQG7@bolson-desk
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
USDT ELF note optionally can record an offset of .stapsdt.base, which is
used to make adjustments to USDT target attach address. Currently,
libbpf will do this address adjustment unconditionally if it finds
.stapsdt.base ELF section in target binary. But there is a corner case
where .stapsdt.base ELF section is present, but specific USDT note
doesn't reference it. In such case, libbpf will basically just add base
address and end up with absolutely incorrect USDT target address.
This adjustment has to be done only if both .stapsdt.sema section is
present and USDT note is recording a reference to it.
Fixes: 74cc6311ce ("libbpf: Add USDT notes parsing and resolution logic")
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20241121224558.796110-1-andrii@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
