Add a --workload-ctl=fifo:ctl-fifo[,ack-fifo] option for 'perf test
-w'. When set, run_workload() opens the named FIFO, writes enable before
invoking the builtin workload, writes disable before returning, and
waits for ack responses when an ack FIFO is provided to ensure that the
workload doesn't run until the events are enabled.
This can be used to limit the scope of the recording to only the
workload execution and avoid recording Perf setup and teardown code if
Perf record is started with events disabled (-D 1).
Assisted-by: Codex:GPT-5.5
Signed-off-by: James Clark <james.clark@linaro.org>
Cc: Amir Ayupov <aaupov@meta.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Leo Yan <leo.yan@arm.com>
Cc: Mike Leach <mike.leach@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paschalis Mpeis <Paschalis.Mpeis@arm.com>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
PE_CONTEXT elements update the context ID and exception level, but the
decoder may still have prior packets cached for frontend processing.
Updating the context immediately in the decoder backend can make those
cached packets get consumed with the wrong thread or EL state.
Add a CS_ETM_CONTEXT packet carrying the TID and EL to the frontend,
this keeps context changes ordered with the rest of the packet stream
and avoids mismatches when synthesizing samples from cached packets.
Separate the memory access function into one for the frontend and one
for decoding. The frontend also needs memory access to attach the
instruction to samples. Because the frontend does memory access for
both previous and current packets, change all the frontend memory access
function signatures to take both a tidq and packet. But backend always
uses the current backend EL and thread from the tidq.
Treat context packets as a boundary for branch sample generation and
remove tidq->prev_packet_thread because it's not possible to branch to a
different thread, so only tracking the current thread is required for
sample generation.
Fixes: e573e978fb ("perf cs-etm: Inject capabilitity for CoreSight traces")
Reported-by: Amir Ayupov <aaupov@meta.com>
Closes: https://lore.kernel.org/linux-perf-users/20260515021135.1729028-1-aaupov@meta.com/
Co-authored-by: James Clark <james.clark@linaro.org>
Signed-off-by: Leo Yan <leo.yan@arm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-doc@vger.kernel.org
Cc: Mike Leach <mike.leach@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paschalis Mpeis <Paschalis.Mpeis@arm.com>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Signed-off-by: James Clark <james.clark@linaro.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Replace snprintf with scnprintf in buffer offset calculations to
ensure the 'used' count will not exceed the "len".
The current logic in perf_pmu__for_each_event uses an unconditional
+ 1 increment to buf_used to account for null terminators. This can
cause a stack buffer overflow in the subsequent scnprintf call.
When the local stack buffer buf (1024 bytes) is full, buf_used can
reach 1025. This causes the subsequent remaining space calculation
sizeof(buf) - buf_used to underflow.
Use sub_non_neg() to see if space actually existed, and only
increment the offset if remaining space is present.
Changes includes:
- Use sub_non_neg to check if space exists
- Replacing snprintf with scnprintf to ensure the return value
reflects the actual bytes written into the buffer.
- Only increment buf_used by 1 if space exists
- If a parameterized event uses a built-in perf keyword for its
parameter name (eg, config=?), the lexer parses it as a predefined
term token, which sets term->config to NULL. Add check to use
parse_events__term_type_str() if term->config is NULL.
Signed-off-by: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hari Bathini <hbathini@linux.vnet.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Shivani Nittor <shivani@linux.ibm.com>
Cc: Tanushree.Shah@ibm.com
Cc: Tejas.Manhas1@ibm.com
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The format_alias() function in util/pmu.c has a check to
detect whether the event has parameterized field ( =? ).
The string alias->terms contains the event and if the event
has user configurable parameter, there will be presence of
sub string "=?" in the alias->terms.
Snippet of code:
/* Paramemterized events have the parameters shown. */
if (strstr(alias->terms, "=?")) {
/* No parameters. */
snprintf(buf, len, "%.*s/%s/", (int)pmu_name_len, pmu->name, alias->name);
if "strstr" contains the substring, it returns a pointer
and hence enters the above check which is not the expected
check. And hence "perf list" doesn't have the parameterized
fields in the result.
Fix this check to use:
if (!strstr(alias->terms, "=?")) {
With this change, perf list shows the events correctly with
the strings showing parameters.
Before the fix:
# ./perf list|grep -w PM_PAU_CYC
hv_24x7/PM_PAU_CYC/ [Kernel PMU event]
With this fix:
# ./perf list|grep -w PM_PAU_CYC
hv_24x7/PM_PAU_CYC,chip=?/ [Kernel PMU event]
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Athira Rajeev <atrajeev@linux.ibm.com>
Tested-by: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hari Bathini <hbathini@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Shivani Nittor <shivani@linux.ibm.com>
Cc: Tanushree.Shah@ibm.com
Cc: Tejas.Manhas1@ibm.com
Cc: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When samples are skipped due to time filtering in process_sample_event(),
the early return path bypasses addr_location__exit(), causing memory leaks
of thread, map, and maps references acquired by machine__resolve().
These references must be released through addr_location__exit() before
returning.
Fixes: 8e746e95c3 ("perf data: Allow filtering conversion by time range")
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Tanushree Shah <tshah@linux.ibm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Athira Rajeev <atrajeev@linux.ibm.com>
Cc: Derek Foreman <derek.foreman@collabora.com>
Cc: Hari Bathini <hbathini@linux.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Shivani.Nittor@ibm.com
Cc: Tanushree.Shah@ibm.com
Cc: Tejas.Manhas1@ibm.com
Cc: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
On s390 the Python extension build fails as follows when using a linker
that is configured to treat text relocations (TEXTREL) in shared
libraries as error by default:
GEN python/perf.cpython-314-s390x-linux-gnu.so
/usr/bin/ld.bfd: error: read-only segment has dynamic relocations
This occurrs because util/llvm-c-helpers.o is erroneously built from
util/llvm-c-helpers.cpp without compiler option -fPIC but linked into
the shared library (via libperf-util.a(perf-util-in.o)).
On s390, object files must be compiled as position-indepedent code (PIC)
in order to be linked into shared libraries. Commit a9a3f1d18a ("perf
s390: Always build with -fPIC") added compiler option -fPIC to CFLAGS
for s390, which is used in C compiles. Add -fPIC to CXXFLAGS for s390
as well, so that it is also used in C++ compiles.
Fixes: a9a3f1d18a ("perf s390: Always build with -fPIC")
Reported-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Jens Remus <jremus@linux.ibm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Bill Wendling <morbo@google.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Hendrik Brueckner <brueckner@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jan Polensky <japo@linux.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Justin Stitt <justinstitt@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <nick.desaulniers+lkml@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Make util/setup.py respect the verbose build flag (V=1) by conditionally
passing --quiet only when not in verbose mode.
This eases debugging of Python extension compilation issues and aligns
with the existing perf build system behavior.
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Jens Remus <jremus@linux.ibm.com>
Tested-by: Jan Polensky <japo@linux.ibm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
On risc-v, pref probe generates an invalid syntax for a named register in
a kprobe.
$ perf probe --debug verbose --add "n_tty_write tty"
...
Writing event: p:probe/n_tty_write _text+8922528 tty=%"%a0":x64
Failed to write event: Invalid argument
The problem is the combination of
#define REG_DWARFNUM_NAME(reg, idx) [idx] = "%" #reg
and entries such as
REG_DWARFNUM_NAME("%a0", 10)
where #reg will escape the quotes of the first macro parameter.
Update the macro definition to produce the correct syntax for a named
register in a kprobe, i.e. the unquoted register name with only one
leading %.
Fixes: a90c451918 ("perf riscv: Remove dwarf-regs.c and add dwarf-regs-table.h")
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Cc: Ian Rogers <irogers@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
dwarf-regs.c includes an arch-specific dwarf-regs-table.h for several
architectures. This pulls in different definitions of REG_DWARFNUM_NAME
and causes compiler warnings for W=1 builds.
In file included from util/dwarf-regs.c:23:
.../dwarf-regs-table.h:5: error: "REG_DWARFNUM_NAME" redefined [-Werror]
#define REG_DWARFNUM_NAME(reg, idx) [idx] = reg
Undefine REG_DWARFNUM_NAME before each new definition.
Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Cc: Ian Rogers <irogers@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add a bullet point to the libperf ABI TODO explaining the code
simplification benefit of widening struct perf_cpu.cpu from int16_t
to int: the narrow type forces defensive truncation checks at every
boundary where wider CPU indices are narrowed, and values > 32767
silently wrap to negative numbers (two's complement), bypassing
bounds validation without them.
Acked-by: Ian Rogers <irogers@google.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
build_id__snprintf() and hwmon_pmu__read_events() accumulate formatted
output via snprintf(), which returns the would-have-been-written count
on truncation. In build_id__snprintf(), this inflates the return
value beyond the buffer size. In hwmon_pmu__read_events(), len
overshoots out_buf_len and the next 'out_buf_len - len' underflows.
Switch both to scnprintf() which returns actual bytes written.
In build_id__snprintf(), also tighten the loop guard from
'offs < bf_size' to 'offs + 1 < bf_size': since scnprintf() returns
at most size-1, offs never reaches bf_size, and the original condition
would spin doing zero-byte writes once the buffer fills.
Fixes: fccaaf6fbb ("perf build-id: Change sprintf functions to snprintf")
Fixes: 53cc0b351e ("perf hwmon_pmu: Add a tool PMU exposing events from hwmon in sysfs")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
hists__scnprintf_title() accumulates formatted output into a buffer
using scnprintf() for all filter clauses except the UID filter, which
uses snprintf(). If the buffer fills up and snprintf() returns more
than the remaining space, printed exceeds size and the next 'size -
printed' underflows, causing later scnprintf() calls to write past
the buffer.
Switch the UID filter clause to scnprintf() to match the rest of the
function.
Fixes: 25c312dbf8 ("perf hists: Move hists__scnprintf_title() away from the TUI code")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Both functions accumulate formatted output via ret += snprintf(buf + ret,
size - ret, ...). If the buffer is too small and snprintf() returns more
than the remaining space, ret exceeds size and the next 'size - ret'
underflows, causing snprintf() to write past the buffer end.
Switch to scnprintf() which returns the actual number of bytes written,
making the accumulation safe.
Fixes: 7b612e291a ("perf tools: Synthesize PERF_RECORD_* for loaded BPF programs")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Song Liu <song@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
open() calls in dso.c and symbol-elf.c omit O_CLOEXEC, which leaks
file descriptors to child processes spawned during symbol resolution
(e.g., addr2line, objdump). This can exhaust the fd limit during
long profiling sessions or when processing many DSOs.
Add O_CLOEXEC to all open() calls in both files (12 call sites).
Fixes: cdd059d731 ("perf tools: Move dso_* related functions into dso object")
Fixes: e5a1845fc0 ("perf symbols: Split out util/symbol-elf.c")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
timehist_print_idlehist_callchain() calls rb_first_cached() on
sorted_root, but the sort function (callchain_param.sort) populates it
via rb_insert_color() on the plain rb_root member — not the cached
variant. This means rb_leftmost is never set, so rb_first_cached()
always returns NULL and the entire callchain summary is silently
dropped from --idle-hist output.
The original code in ba957ebb54 ("perf sched timehist: Show
callchains for idle stat") was correct — it used struct rb_root and
rb_first(). The bug was introduced when sorted_root was converted to
rb_root_cached without converting the sort insertion path to use
rb_insert_color_cached().
Use rb_first(&root->rb_root) to match how the tree was populated.
Fixes: cb4c13a513 ("perf sched: Use cached rbtrees")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
timehist_skip_sample() reads prio from untrusted tracepoint data via
perf_sample__intval(sample, "prev_prio") without bounds validation.
A crafted perf.data with prev_prio >= MAX_PRIO (140) causes test_bit()
to read past the end of the prio_bitmap, which is only MAX_PRIO bits.
Add a prio >= 0 guard before the test_bit() call and skip out-of-range
values (>= MAX_PRIO) that can never match the user's filter set.
The original prio != -1 already let all negatives other than -1 through
(after an undefined-behavior bitmap read); the new prio >= 0 guard
preserves that pass-through behavior — negative means "no priority
info", so the event is shown unfiltered — while fixing the OOB.
Values >= MAX_PRIO are skipped because they cannot be represented in
the filter bitmap.
Fixes: 9b3a48bbe2 ("perf sched timehist: Add --prio option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Yang Jihong <yangjihong@bytedance.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Two call sites free bitmaps without NULLing the pointer, risking
double-free if the structure is reused or cleanup is called twice:
- mmap__munmap(): map->affinity_mask.bits
- record__mmap_cpu_mask_free(): mask->bits
Set each pointer to NULL after bitmap_free().
Fixes: 8384a2600c ("perf record: Adapt affinity to machines with #CPUs > 1K")
Fixes: f466e5ed6c ("perf record: Extend --threads command line option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Alexey Bayduraev <alexey.v.bayduraev@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
machine__resolve() accesses env->cpu[al->cpu].socket_id after checking
al->cpu >= 0 and env->cpu != NULL, but without validating al->cpu
against env->nr_cpus_avail. Since al->cpu comes from the untrusted
perf.data sample, a crafted file with a large CPU index causes an
out-of-bounds heap read.
Use perf_env__get_cpu_topology() which validates both NULL and bounds.
Also bounds-check al->cpu before the cast to struct perf_cpu (int16_t):
without this, values like 65536 silently truncate to 0, bypassing the
accessor's internal check and returning CPU 0's topology.
Fixes: 0c4c4debb0 ("perf tools: Add processor socket info to hist_entry and addr_location")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
cpu_map__snprint() accumulates snprintf() return values in ret.
snprintf() returns the number of characters that *would have been
written* on truncation, not the actual count. When a fragmented CPU
list exceeds the buffer, ret grows past size, causing `size - ret` to
underflow (both are size_t), and subsequent snprintf() calls write
past the end of the caller's stack buffer.
Switch to scnprintf() which returns the actual number of characters
written, making ret accumulation safe by construction.
Fixes: a24020e6b7 ("perf tools: Change cpu_map__fprintf output")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
get_max_num() reads a sysfs file (cpu/possible, cpu/present, or
node/possible) and scans backward from the end to find the last
number. If the file is empty, filename__read_str() returns num == 0.
The loop `while (--num)` decrements the size_t from 0 to SIZE_MAX,
reading backward across the heap until a comma or hyphen is found
or unmapped memory is hit.
Add an early return for empty files before the backward scan.
Fixes: 7780c25bae ("perf tools: Allow ability to map cpus to nodes easily")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
struct perf_cpu.cpu is int16_t, limiting perf to 32767 CPUs. This is
part of the libperf ABI (returned by value from perf_cpu_map__cpu() and
friends), so widening it requires an ABI bump.
Add a comment on the struct definition noting this, and create a TODO
file to collect future ABI changes so they can be batched into a single
version bump.
Cc: Ian Rogers <irogers@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
free_idle_threads() relies on the thread priv destructor (free()) to
clean up idle_thread_runtime structs. But free() doesn't walk the
callchain_cursor linked list or the callchain_root tree allocated
by callchain_cursor__copy() and callchain_append() during --idle-hist
processing. Every idle thread with callchain data leaks these nodes.
Introduce callchain_cursor_cleanup() to free the cursor's linked list
of callchain_cursor_node entries, and call it together with
free_callchain() in free_idle_threads() before thread__put().
Fixes: 225b24f569980ac9 ("perf sched timehist: Save callchain when entering idle")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
set_max_cpu_num() assigns the sysfs "possible" CPU count to
max_cpu_num.cpu which is int16_t (struct perf_cpu). On systems
with >32767 possible CPUs the value silently truncates, potentially
wrapping negative. This causes cpunode_map to be underallocated
and subsequent cpu__get_node() calls to read out of bounds.
The matching check for max_present_cpu_num was added by commit
c760174401 ("perf cpumap: Reduce cpu size from int to int16_t")
but max_cpu_num was missed. Add the same INT16_MAX guard.
Fixes: c760174401 ("perf cpumap: Reduce cpu size from int to int16_t")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
cpu2y() indexes topology_map[cpu] without bounds checking. The array
is allocated with nr_cpus entries (from env->nr_cpus_online), but
callers pass sample CPU values from perf.data which can exceed that
size with cross-machine recordings.
Track the topology_map allocation size and bounds-check the CPU
argument in cpu2y() before indexing. Out-of-bounds CPUs fall back
to the identity mapping (cpu2slot(cpu)), which is the same behavior
as when no topology is available.
Fixes: c507999790 ("perf timechart: Add support for topology")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Stanislav Fomichev <stfomichev@yandex-team.ru>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
he__get_c2c_hists() assigns c2c_he->hists before calling
c2c_hists__init(). If init fails, the error path calls free(hists)
but leaves c2c_he->hists pointing to freed memory. On teardown,
c2c_he_free() finds the non-NULL pointer and calls
hists__delete_entries() on it, causing a use-after-free.
Set c2c_he->hists to NULL before freeing so teardown skips the
already-freed allocation.
Fixes: b2252ae67b ("perf c2c report: Decode c2c_stats for hist entries")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
process_cpu_topology() in header.c frees env->cpu on old-format
perf.data files that predate topology information, but leaves
nr_cpus_avail set. The six perf_env__get_*_aggr_by_cpu() functions
in builtin-stat.c pass the bounds check but dereference a NULL
env->cpu pointer, crashing on old recordings.
Introduce perf_env__get_cpu_topology() as a safe accessor that
validates env->cpu, cpu.cpu >= 0, and cpu.cpu < nr_cpus_avail in
one place, returning a struct cpu_topology_map pointer or NULL.
Convert all six topology aggregation callbacks to use it.
Fixes: 88031a0de7 ("perf stat: Switch to cpu version of cpu_map__get()")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf_mmap__aio_mmap() sets map->aio.nr_cblocks before allocating the
data array. If calloc() for aiocb or cblocks fails before the data
array is allocated, the return -1 path leads to perf_mmap__aio_munmap()
which loops nr_cblocks times calling perf_mmap__aio_free(). Both
versions of perf_mmap__aio_free() (NUMA and non-NUMA) dereference
map->aio.data[idx] without checking if data is NULL, causing a NULL
pointer dereference.
Add NULL checks for map->aio.data at the top of both
perf_mmap__aio_free() variants so the cleanup path is safe when
allocation fails partway through perf_mmap__aio_mmap().
Fixes: d3d1af6f01 ("perf record: Enable asynchronous trace writing")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
get_new_event() has three issues:
1. The zalloc() result is dereferenced without a NULL check, crashing
on allocation failure.
2. BUG_ON(!task->atoms) kills the process when realloc() fails.
Since perf.data is untrusted input, this should be a graceful error.
3. The realloc pattern assigns directly to task->atoms, losing the old
pointer on failure. task->nr_events is also incremented before the
realloc, leaving corrupted state on failure.
Fix get_new_event() to:
- Check the zalloc() result before dereferencing
- Use a temporary for realloc() to avoid losing the old pointer
- Increment nr_events only after successful realloc
- Return NULL instead of calling BUG_ON on failure
Also fix add_sched_event_wakeup() where zalloc() for wait_sem is
passed to sem_init() without a NULL check.
Update all callers (add_sched_event_run, add_sched_event_wakeup,
add_sched_event_sleep) to handle NULL returns by returning early.
The replay may produce incomplete output on OOM but will not crash.
Fixes: ec156764d4 ("perf sched: Import schedbench.c")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
free_idle_threads() calls thread__delete() directly instead of
thread__put(), bypassing the reference counting lifecycle. Under
REFCNT_CHECKING builds, this leaks the pointer handle since
thread__delete() frees the object without going through the refcount
wrapper.
The idle threads are created via thread__new() (refcount=1) in
get_idle_thread(). Callers get additional references via thread__get()
which they release with thread__put(). free_idle_threads() drops the
base reference — thread__put() is the correct call, matching the
thread__new() acquisition.
Fixes: 49394a2a24 ("perf sched timehist: Introduce timehist command")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
timehist_sched_change_event() sets itr->last_thread to NULL at the end
of idle hist processing without calling thread__put() first. The
thread reference was acquired via thread__get() in timehist_get_thread()
(line 2581), so every idle context switch leaks a thread reference when
--idle-hist is active.
Use thread__zput() to properly release the reference before clearing
the pointer.
Fixes: 5d8f17fb5822 ("perf sched timehist: Add -I/--idle-hist option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
timehist_sched_change_event() uses thread__tid(thread) == 0 to decide
whether to cast thread_runtime to idle_thread_runtime. However, a
crafted perf.data can set common_pid=0 and common_tid=0 (the perf_sample
fields) while prev_pid != 0 (the tracepoint field). is_idle_sample()
returns false (it checks prev_pid for sched_switch), so
timehist_get_thread() goes through machine__findnew_thread() and returns
the machine's TID 0 thread — whose priv data is a regular thread_runtime,
not the larger idle_thread_runtime allocated by init_idle_thread().
The subsequent cast to idle_thread_runtime reads past the thread_runtime
allocation, accessing itr->last_thread, itr->cursor, and itr->callchain
from adjacent heap memory. Writing to itr->last_thread corrupts the
heap; calling thread__put() on the OOB value frees an arbitrary pointer.
Replace the thread__tid() == 0 check with is_idle_sample(), which uses
the tracepoint-specific prev_pid field and correctly identifies whether
the sample originated from an idle thread with idle_thread_runtime priv.
Fixes: 5d8f17fb5822 ("perf sched timehist: Add -I/--idle-hist option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
get_idle_thread() allocates a thread via thread__new() and stores it in
idle_threads[cpu], then calls init_idle_thread() to set up the private
data. If init_idle_thread() fails (e.g. OOM for the idle_thread_runtime
struct), the function returns NULL but leaves the partially initialized
thread in idle_threads[cpu].
On subsequent calls for the same CPU, get_idle_thread() finds a non-NULL
idle_threads[cpu], skips allocation, and returns thread__get() on a
thread that has no priv data. Callers then get a thread whose
thread__priv() returns NULL, leading to unexpected behavior.
Release the thread and reset the slot to NULL on init failure so the
entry doesn't persist in a corrupted state.
Fixes: 49394a2a24 ("perf sched timehist: Introduce timehist command")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
setup_nodes() iterates CPU maps from the perf.data topology header and
uses cpu.cpu directly as an array index into cpu2node[] (allocated with
c2c.cpus_cnt = env->nr_cpus_avail entries) and __set_bit(cpu.cpu, set)
(bitmap also sized to c2c.cpus_cnt).
A crafted perf.data with topology CPU IDs exceeding nr_cpus_avail causes
out-of-bounds heap writes into both the cpu2node array and the per-node
bitmap.
Add a bounds check to skip CPU IDs that fall outside the valid range.
Fixes: 1e181b92a2 ("perf c2c report: Add 'node' sort key")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
c2c_he__set_cpu() passes sample->cpu directly to __set_bit(cpu, cpuset)
after only checking for the (u32)-1 sentinel. The cpuset bitmap is
allocated with c2c.cpus_cnt bits (from env->nr_cpus_avail), so a crafted
perf.data with CPU IDs exceeding that count causes out-of-bounds heap
writes.
c2c_he__set_node() similarly passes the node ID from mem2node__node()
to __set_bit(node, nodeset) after only checking for negative values.
The nodeset bitmap is sized to c2c.nodes_cnt (from env->nr_numa_nodes),
so a node ID exceeding that causes OOB writes.
process_sample_event() indexes c2c.cpu2node[cpu] and
c2c_he->node_stats[node] without bounds checking. Both arrays are
sized to c2c.cpus_cnt and c2c.nodes_cnt respectively.
Add bounds checks in all three paths:
- c2c_he__set_cpu(): return if sample->cpu >= c2c.cpus_cnt
- c2c_he__set_node(): return if node >= c2c.nodes_cnt
- process_sample_event(): clamp cpu to 0 if >= cpus_cnt,
guard node_stats access with bounds check
Fixes: 1e181b92a2 ("perf c2c report: Add 'node' sort key")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Six perf_env__get_*_aggr_by_cpu() functions access env->cpu[cpu.cpu]
after only checking cpu.cpu != -1. env->cpu[] is allocated with
env->nr_cpus_avail entries, so a CPU index from an untrusted perf.data
file that exceeds that count causes an out-of-bounds heap read.
Replace the != -1 guard with >= 0 && < env->nr_cpus_avail in all six
functions. The >= 0 check also catches -1 and any other negative values
that could bypass the old check.
Affected functions:
- perf_env__get_socket_aggr_by_cpu()
- perf_env__get_die_aggr_by_cpu()
- perf_env__get_cache_aggr_by_cpu()
- perf_env__get_cluster_aggr_by_cpu()
- perf_env__get_core_aggr_by_cpu()
- perf_env__get_cpu_aggr_by_cpu()
Fixes: 68d702f7a1 ("perf stat report: Add support to initialize aggr_map from file")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf_mmap__aio_bind() passes the cpu__get_node() return value directly
to an unsigned long variable (node_index). When cpu__get_node() returns
-1 for an unknown CPU, the implicit int-to-unsigned-long conversion
sign-extends it to ULONG_MAX.
This causes bitmap_zalloc(ULONG_MAX + 1) which wraps to
bitmap_zalloc(0), returning a zero-sized allocation. The subsequent
__set_bit(ULONG_MAX, node_mask) then writes massively out of bounds.
Check the return value in a signed temporary before assigning to
node_index, and skip the NUMA binding when the node is unknown.
Fixes: c44a8b44ca ("perf record: Bind the AIO user space buffers to nodes")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
register_pid() has several issues when processing untrusted perf.data:
1. Integer overflow: (pid + 1) * sizeof(struct task_desc *) can wrap
to a small value on 32-bit systems when pid is large (e.g.
0x40000000), causing realloc to return a tiny buffer followed by
out-of-bounds writes in the initialization loop.
2. Heap buffer overflow: strcpy(task->comm, comm) copies the
untrusted comm string into a fixed 20-byte COMM_LEN buffer with
no length check.
3. BUG_ON on allocation failure: perf.data is untrusted input, so
allocation failures should be handled gracefully rather than
killing the process.
4. Realloc of sched->tasks assigned directly back, leaking the old
pointer on failure; nr_tasks incremented before the realloc,
leaving corrupted state on failure.
Cap pid at PID_MAX_LIMIT (4194304, matching the kernel's maximum
on 64-bit), replace strcpy with strlcpy, guard against NULL comm,
replace BUG_ON with NULL returns using safe realloc patterns, and
add NULL checks in callers that dereference the result.
Fixes: ec156764d4 ("perf sched: Import schedbench.c")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
perf_timehist__process_sample() updates sched->max_cpu from the
sample CPU without bounds checking. Later code uses max_cpu + 1 as
an iteration count over arrays allocated with MAX_CPUS entries
(curr_thread, cpu_last_switched). A recording with CPU IDs >= MAX_CPUS
causes out-of-bounds array accesses.
Also cap the env->nr_cpus_online initialization of max_cpu in
perf_sched__timehist(), which could exceed MAX_CPUS on very large
systems.
Add bounds checks before both max_cpu updates, matching the pattern
already used in map_switch_event().
Fixes: 49394a2a24 ("perf sched timehist: Introduce timehist command")
Reviewed-by: David Ahern <dsahern@kernel.org>
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
timehist_get_thread() acquires a thread reference via
machine__findnew_thread() and an idle thread reference via
get_idle_thread() (which calls thread__get()). Two error paths in
the idle_hist block return NULL without releasing these references:
- When get_idle_thread() fails, the thread reference leaks.
- When thread__priv(idle) returns NULL, both idle and thread leak.
Additionally, the idle thread reference acquired on the success path
is never released, leaking a reference on every sample when
--idle-hist is active.
Add thread__put() calls on both error paths and release the idle
reference after use on the success path.
Fixes: 5d8f17fb5822 ("perf sched timehist: Add -I/--idle-hist option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
cpu__get_node() accesses cpunode_map[cpu.cpu] without checking against
max_cpu_num, the allocation size of cpunode_map. Callers such as
builtin-kmem.c:evsel__process_alloc_event() pass sample->cpu from
perf.data events, which may exceed the host's CPU count when analyzing
cross-machine recordings.
Add a bounds check against max_cpu_num before indexing, returning -1
for out-of-range values. This is a central fix that protects all
callers.
Fixes: 86895b480a ("perf stat: Add --per-node agregation support")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
auxtrace.c:filter_cpu() and builtin-script.c:filter_cpu() call
test_bit(cpu, cpu_bitmap) where cpu_bitmap is declared with
MAX_NR_CPUS bits. When the CPU value from a perf.data event is
corrupt or absent (e.g. negative or >= MAX_NR_CPUS), test_bit reads
out of bounds.
Add bounds checks before test_bit(): >= 0 for the int16_t cpu.cpu in
auxtrace (which also covers the -1 sentinel), and < MAX_NR_CPUS for
both sites. Matches the pattern applied in the previous series for
builtin-annotate.c, builtin-diff.c, builtin-report.c, and
builtin-sched.c.
Fixes: 644e0840ad ("perf auxtrace: Add CPU filter support")
Fixes: 5d67be97f8 ("perf report/annotate/script: Add option to specify a CPU range")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
setup_map_cpus() allocates comp_cpus based on
sysconf(_SC_NPROCESSORS_CONF), the host machine's CPU count. But
map_switch_event() indexes comp_cpus using cpus_nr derived from
bitmap_weight(comp_cpus_mask, MAX_CPUS), where comp_cpus_mask is
declared as DECLARE_BITMAP(..., MAX_CPUS) with MAX_CPUS=4096.
When analyzing a perf.data recording from a machine with more CPUs
than the analysis host (e.g. 128-CPU server recording analyzed on an
8-CPU laptop), cpus_nr exceeds the allocation size, causing a heap
buffer overflow.
Also fix a type mismatch: comp_cpus is 'struct perf_cpu *' (2 bytes
per element) but was allocated with sizeof(int) (4 bytes per element).
Allocate comp_cpus with MAX_CPUS entries using the correct element
size, matching the comp_cpus_mask bitmap bounds. Remove the
sysconf(_SC_NPROCESSORS_CONF) initialization of max_cpu — its only
consumer was the comp_cpus allocation, and max_cpu is dynamically
updated from the recording's events during processing. Fix the
non-compact path to use max_cpu.cpu + 1 as cpus_nr, converting from
0-based index to count — sysconf() returned a count which masked
this off-by-one.
Fixes: 99623c628f ("perf sched: Add compact display option")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
latency_runtime_event() passes the return value of
machine__findnew_thread() directly to thread_atoms_search() at line
1216, before checking for NULL at line 1220. thread_atoms_search()
calls pid_cmp() which dereferences the thread pointer via
thread__tid(), causing a NULL pointer dereference if the allocation
fails.
All other callers of thread_atoms_search() in this file
(latency_switch_event, latency_wakeup_event,
latency_migrate_task_event) correctly check for NULL first.
Move the atoms assignment after the NULL check to match the pattern
used by the other callers.
Fixes: b91fc39f4a ("perf machine: Protect the machine->threads with a rwlock")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In latency_switch_event(), after acquiring thread references for
sched_out and sched_in via machine__findnew_thread(), the first
add_sched_out_event() failure path does 'return -1', bypassing the
out_put label that calls thread__put() on both references.
The second and third add_sched_out_event() failures correctly use
'goto out_put'. Fix the first one to match.
Fixes: b91fc39f4a ("perf machine: Protect the machine->threads with a rwlock")
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When PERF_SAMPLE_CPU is absent from a perf.data file, sample->cpu is
initialized to (u32)-1 by evsel__parse_sample(). Five call sites pass
this value directly to test_bit(sample->cpu, cpu_bitmap), reading
massively out of bounds past the DECLARE_BITMAP(..., MAX_NR_CPUS)
allocation of 4096 bits.
Add a sample->cpu >= MAX_NR_CPUS guard before each test_bit() call,
matching the existing safe pattern in builtin-kwork.c. This catches
both the (u32)-1 sentinel and any corrupted CPU value exceeding the
bitmap size.
Fixes: 5d67be97f8 ("perf report/annotate/script: Add option to specify a CPU range")
Cc: Anton Blanchard <anton@samba.org>
Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
update_lock_stat() handles lock contentions that start but never reach a
contention_end event (e.g., locks still held when profiling stops), but
previously treated LOCK_AGGR_CGROUP as a no-op due to missing cgroup
context in userspace.
Fix this by adding a cgroup_id field to struct tstamp_data, recording it
at contention_begin using get_current_cgroup_id() when aggr_mode is
LOCK_AGGR_CGROUP. Capturing it at contention_begin is semantically
correct, the contention cost is incurred by the task that had to wait,
not by whatever task happens to be running at contention_end. It is also
preferable from a performance standpoint, as contention_end runs just
before the task enters the critical section.
Update contention_end to use pelem->cgroup_id instead of calling
get_current_cgroup_id() dynamically, ensuring both complete and
incomplete contention events attribute the wait time to the cgroup at
wait-start time consistently.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Suchit Karunakaran <suchitkarunakaran@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tycho Andersen (AMD) <tycho@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The is_pmu_core function checks if a PMU name corresponds to a core
CPU PMU. However, it currently fails to recognize "default_core" as
a core PMU.
When "default_core" is used, the PMU scanning fallback in pmus.c
scans the "other_pmus" list. This scan is slow and always misses because
"default_core" is a core PMU, leading to unnecessary overhead.
Update is_pmu_core to recognize "default_core" directly. Additionally,
document the different matching approaches (exact name for x86/s390,
sysfs-based cpus file check for ARM/hybrid) to clarify how core PMUs are
classified.
Also, explicitly treat "default_core" as `all_pmus` in `setup_metric_events()`
to preserve the original metric resolution behavior for this pseudo-PMU.
Assisted-by: Gemini-CLI:Google Gemini 3.1 Pro
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The 1.x branch of Babeltrace has been superseded by 2.x in 2020 and has
been unmaintained since 2022, efforts have started to remove it from
popular distributions.
Babeltrace 2.x offers a very similar 'ctf-writer' library that can be used
with minimal changes for the '--to-ctf' feature and has been packaged
since Debian 11 and Fedora 32.
This patch replaces the 'libbabeltrace' build feature with
'babeltrace2-ctf-writer' using pkgconfig detection, adjusts the naming of
the public headers and applies minor API cleanups.
There is no changes to the output ctf traces, the ctf-writer API still
implements version 1.8 of the CTF specification that can be read by
either Babeltrace 1 / 2 or any CTF compliant reader.
Also remove some ifdefs in the cli option parsing to allow printing the
helpful error message with '--to-ctf' when built without babeltrace2.
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Derek Foreman <derek.foreman@collabora.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The -L/--lock-filter option is to specify target locks by name or
address. It's basically for global locks where name or address is known
and fixed. But 'mmap_lock' is a per-process lock so it cannot be used
for the -L option.
$ sudo perf lock con -ab -L mmap_lock
ignore unknown symbol: mmap_lock
libbpf: map 'addr_filter': failed to create: -EINVAL
libbpf: failed to load BPF skeleton 'lock_contention_bpf': -EINVAL
Failed to load lock-contention BPF skeleton
lock contention BPF setup failed
However, it's still a common source of contention especially in a large
process so we want to use it for the -L/--lock-filter option. As there
is check_lock_type() to check mmap_lock at runtime, let's used it to
filter mmap_locks as a special case.
Of course, this only works with -b/--use-bpf option.
$ sudo perf lock con -b -L mmap_lock -- perf bench mem mmap -f demand -t 2
# Running 'mem/mmap' benchmark:
# function 'demand' (Demand loaded mmap())
# Copying 1MB bytes ...
2.679184 GB/sec/thread ( +- 1.78% )
contended total wait max wait avg wait type caller
1 15.22 us 15.22 us 15.22 us rwsem:W __vm_munmap+0x7e
1 7.72 us 7.72 us 7.72 us rwsem:R lock_mm_and_find_vma+0x97
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suchit Karunakaran <suchitkarunakaran@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
