There can be a significant gap in memset/memcpy performance depending
on the size of the region being operated on.
With chunk-size=4kb:
$ echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
$ perf bench mem memset -p 4kb -k 4kb -s 4gb -l 10 -f x86-64-stosq
# Running 'mem/memset' benchmark:
# function 'x86-64-stosq' (movsq-based memset() in arch/x86/lib/memset_64.S)
# Copying 4gb bytes ...
13.011655 GB/sec
With chunk-size=1gb:
$ echo madvise > /sys/kernel/mm/transparent_hugepage/enabled
$ perf bench mem memset -p 4kb -k 1gb -s 4gb -l 10 -f x86-64-stosq
# Running 'mem/memset' benchmark:
# function 'x86-64-stosq' (movsq-based memset() in arch/x86/lib/memset_64.S)
# Copying 4gb bytes ...
21.936355 GB/sec
So, allow the user to specify the chunk-size.
The default value is identical to the total size of the region, which
preserves current behaviour.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Raghavendra K T <raghavendra.kt@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Page sizes that can be selected: 4KB, 2MB, 1GB.
Both the reservation and node from which hugepages are allocated
from are expected to be addressed by the user.
An example of page-size selection:
$ perf bench mem memset -s 4gb -p 2mb
# Running 'mem/memset' benchmark:
# function 'default' (Default memset() provided by glibc)
# Copying 4gb bytes ...
14.919194 GB/sec
# function 'x86-64-unrolled' (unrolled memset() in arch/x86/lib/memset_64.S)
# Copying 4gb bytes ...
11.514503 GB/sec
# function 'x86-64-stosq' (movsq-based memset() in arch/x86/lib/memset_64.S)
# Copying 4gb bytes ...
12.600568 GB/sec
Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Raghavendra K T <raghavendra.kt@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When not running as root and with higher perf event paranoia values
the perf record LBR tests could fail rather than skipping the
problematic tests.
Add the sensitivity to the test and confirm it passes with paranoia
values from -1 to 2.
Committer testing:
Testing with '$ perf test -vv lbr', i.e. as non root, and then comparing
the output shows the mentioned errors before this patch:
acme@x1:~$ grep -m1 "model name" /proc/cpuinfo
model name : 13th Gen Intel(R) Core(TM) i7-1365U
acme@x1:~$
Before:
132: perf record LBR tests : Skip
After:
132: perf record LBR tests : Ok
Fixes: 32559b99e0 ("perf test: Add set of perf record LBR tests")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Chun-Tse Shao <ctshao@google.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.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>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The perf record testcase fails on systems with more than 1K CPUs.
Testcase: perf test -vv "PERF_RECORD_* events & perf_sample fields"
PERF_RECORD_* events & perf_sample fields :
--- start ---
test child forked, pid 272482
sched_getaffinity: Invalid argument
sched__get_first_possible_cpu: Invalid argument
test child finished with -1
---- end ----
PERF_RECORD_* events & perf_sample fields: FAILED!
sched__get_first_possible_cpu uses "sched_getaffinity" to get the
cpumask and this call is returning EINVAL (Invalid argument).
This happens because the default mask size in glibc is 1024.
To overcome this 1024 CPUs mask size limitation of cpu_set_t, change the
mask size using the CPU_*_S macros ie, use CPU_ALLOC to allocate
cpumask, CPU_ALLOC_SIZE for size.
Same fix needed for mask which is used to setaffinity so that mask size
is large enough to represent number of possible CPU's in the system.
Reported-by: Tejas Manhas <tejas05@linux.ibm.com>
Signed-off-by: Athira Rajeev <atrajeev@linux.ibm.com>
Tested-by: Ian Rogers <irogers@google.com>
Tested-by: Tejas Manhas <tejas05@linux.ibm.com>
Tested-by: Venkat Rao Bagalkote <venkat88@linux.ibm.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Aditya Bodkhe <Aditya.Bodkhe1@ibm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hari Bathini <hbathini@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Utilizes the previous is_breg_access_indirect function to determine if
the register + offset stores the variable itself or the struct it points
to, save the information in die_var_type.is_reg_var_addr.
Since we are storing the real types in the stack state, we need to do a
type dereference when is_reg_var_addr is set to false for stack/frame
registers.
For other gp registers, skip the variable when the register is a pointer
to the type. If we want to accept these variables, we might also utilize
is_reg_var_addr in a different way, we need to mark that register as a
pointer to the type.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Zecheng Li <zecheng@google.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: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xu Liu <xliuprof@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Introduces the function is_breg_access_indirect to determine whether a
memory access involving a DW_OP_breg* operation refers to the variable's
value directly or requires dereferencing the variable's type as a
pointer based on the DWARF expression.
Previously, all breg based accesses were assumed to directly access the
variable's value (is_pointer = false).
The is_breg_access_indirect function handles three cases:
1. Base register + offset only: (e.g., DW_OP_breg7 RSP+88) The
calculated address is the location of the variable. The access is
direct, so no type dereference is needed. Returns false.
2. Base register + offset, followed by other operations ending in
DW_OP_stack_value, including DW_OP_deref: (e.g., DW_OP_breg*,
DW_OP_deref, DW_OP_stack_value) The DWARF expression computes the
variable's value, but that value requires a dereference. The memory
access is fetching that value, so no type dereference is needed.
Returns false.
3. Base register + offset, followed only by DW_OP_stack_value: (e.g.,
DW_OP_breg13 R13+256, DW_OP_stack_value) This indicates the value at
the base + offset is the variable's value. Since this value is being
used as an address in the memory access, the variable's type is
treated as a pointer and requires a type dereference. Returns true.
The is_pointer argument passed to match_var_offset is now set by
is_breg_access_indirect for breg accesses.
There are more complex expressions that includes multiple operations and
may require additional handling, such as DW_OP_deref without a
DW_OP_stack_value, or including multiple base registers. They are less
common in the Linux kernel dwarf and are skipped in check_allowed_ops.
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Zecheng Li <zecheng@google.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: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xu Liu <xliuprof@google.com>
Link: https://lore.kernel.org/r/CAJUgMyK2wTiEZB__dtgCELmaNGFWhG1j0g9rv_C=cLD6Zq4_5w@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Current code skips to parse events after generating data source. The
reason is the data source packets have cache and snooping related info,
the afterwards event packets might contain duplicate info.
This commit changes to continue parsing the events after data source
analysis. If data source does not give out memory level and snooping
types, then the event info is used to synthesize the related fields.
As a result, both the peer snoop option ('-d peer') and hitm options
('-d tot/lcl/rmt') are supported by Arm SPE in 'perf c2c'.
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Leo Yan <leo.yan@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ali Saidi <alisaidi@amazon.com>
Cc: German Gomez <german.gomez@arm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Since FEAT_SPEv1p4, Arm SPE provides two extra events: "Cache data
modified" and "Data snooped".
Set the snoop mode as:
- If both the "Cache data modified" event and the "Data snooped" event
are set, which indicates a load operation that snooped from a outside
cache and hit a modified copy, set the HITM flag to inspect false
sharing.
- If the snooped event bit is not set, and the snooped event has been
supported by the hardware, set as NONE mode (no snoop operation).
- If the snooped event bit is not set, and the event is not supported or
absent the events info in the meta data, set as NA mode (not
available).
Don't set any mode for only "Cache data modified" event, as it hits a
local modified copy.
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Leo Yan <leo.yan@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ali Saidi <alisaidi@amazon.com>
Cc: German Gomez <german.gomez@arm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
__maps__fixup_overlap_and_insert may split or directly insert a map,
when doing this the map may need to have a kmap set up for the sake of
the kmaps. The missing kmap set up fails the check_invariants test in
maps, later "Internal error" reports from map__kmap and ultimately
causes segfaults.
Similar fixes were added in commit e0e4e0b8b7 ("perf maps: Add
missing map__set_kmap_maps() when replacing a kernel map") and commit
25d9c0301d ("perf maps: Set the kmaps for newly created/added kernel
maps") but they missed cases. To try to reduce the risk of this,
update the kmap directly following any manual insert. This identified
another problem in maps__copy_from.
Fixes: e0e4e0b8b7 ("perf maps: Add missing map__set_kmap_maps() when replacing a kernel map")
Fixes: 25d9c0301d ("perf maps: Set the kmaps for newly created/added kernel maps")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
When "perf lock con" is run in a live mode, with no data file, a host
environment must be provided. Testing missed this as a failing assert
was creating the 1 line of expected stderr output.
$ sudo perf lock con -ab true
perf: util/session.c:195: __perf_session__new: Assertion `host_env != NULL' failed.
Aborted
Fixes: 525a599bad ("perf env: Remove global perf_env")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The function parse_events__sort_events_and_fix_groups is needed to fix
uncore events like:
```
$ perf stat -e '{data_read,data_write}' ...
```
so that the multiple uncore PMUs have a group each of data_read and
data_write events.
The same function will perform architecture sorting and group fixing,
in particular for Intel topdown/perf-metric events. Grouping multiple
perf metric events together causes perf_event_open to fail as the
group can only support one. This means command lines like:
```
$ perf stat -e 'slots,slots' ...
```
fail as the slots events are forced into a group together to try to
satisfy the perf-metric event constraints.
As the user may know better than
parse_events__sort_events_and_fix_groups add a 'X' modifier to skip
its regrouping behavior. This allows the following to succeed rather
than fail on the second slots event being opened:
```
$ perf stat -e 'slots,slots:X' -a sleep 1
Performance counter stats for 'system wide':
6,834,154,071 cpu_core/slots/ (50.13%)
5,548,629,453 cpu_core/slots/X (49.87%)
1.002634606 seconds time elapsed
```
Closes: https://lore.kernel.org/lkml/20250822082233.1850417-1-dapeng1.mi@linux.intel.com/
Reported-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Reported-by: Xudong Hao <xudong.hao@intel.com>
Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ingo Molnar <mingo@redhat.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>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Cc: Yoshihiro Furudera <fj5100bi@fujitsu.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The lzma_is_compressed and gzip_is_compressed functions are declared
to return a "bool" type, but in case of an error (e.g., file open
failure), they incorrectly returned -1.
A bool type is a boolean value that is either true or false.
Returning -1 for a bool return type can lead to unexpected behavior
and may violate strict type-checking in some compilers.
Fix the return value to be false in error cases, ensuring the function
adheres to its declared return type improves for preventing potential
bugs related to type mismatch.
Fixes: 4b57fd44b6 ("perf tools: Add lzma_is_compressed function")
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Yunseong Kim <ysk@kzalloc.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephen Brennan <stephen.s.brennan@oracle.com>
Link: https://lore.kernel.org/r/20250822162506.316844-3-ysk@kzalloc.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The second title line which shows symbol and DSO name is broken after
moving to another function at 'callq' instruction.
The ui_browser__show_title() is used for the first line which shows
global sample count and event name so it doesn't change across the
functions.
What it needs after processing 'call' instruction is to update the
second line onlly. Add a comment and call appropriate function.
You can verify the change by pressing ENTER on a 'call' instruction and
then ESC.
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
