When injecting events from a guest perf.data file, the events will have
separate sample ID numbers. These ID numbers can then be used to determine
which machine an event belongs to. To facilitate that, add machine_pid and
vcpu to id_index records. For backward compatibility, these are added at
the end of the record, and the length of the record is used to determine
if they are present or not.
Note, this is needed because the events from a guest perf.data file contain
the pid/tid of the process running at that time inside the VM not the
pid/tid of the (QEMU) hypervisor thread. So a way is needed to relate
guest events back to the guest machine and VCPU, and using sample ID
numbers for that is relatively simple and convenient.
Signed-off-by: Adrian Hunter <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
In preparation for recording sideband events in a virtual machine guest so
that they can be injected into a host perf.data file.
This is needed to enable injecting events after the initial synthesized
user events (that have an all zero id sample) but before regular events.
Committer notes:
Add entry about PERF_RECORD_FINISHED_INIT to
tools/perf/Documentation/perf.data-file-format.txt.
Committer testing:
Before:
# perf report -D | grep FINISHED
0 0x5910 [0x8]: PERF_RECORD_FINISHED_ROUND
FINISHED_ROUND events: 1 ( 0.5%)
#
After:
# perf record -- sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.020 MB perf.data (7 samples) ]
# perf report -D | grep FINISHED
0 0x5068 [0x8]: PERF_RECORD_FINISHED_INIT: unhandled!
0 0x5390 [0x8]: PERF_RECORD_FINISHED_ROUND
FINISHED_ROUND events: 1 ( 0.5%)
FINISHED_INIT events: 1 ( 0.5%)
#
Signed-off-by: Adrian Hunter <[email protected]>
Acked-by: Ian Rogers <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Uncore events require a CPU i.e. it cannot be -1.
The evsel system_wide flag is intended for events that should be on every
CPU, which does not make sense for uncore events because uncore events do
not map one-to-one with CPUs.
These 2 requirements are not exactly the same, so introduce a new flag
'requires_cpu' for the uncore case.
Signed-off-by: Adrian Hunter <[email protected]>
Acked-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Alexey Bayduraev <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Leo Yan <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Pull perf tools updates from Arnaldo Carvalho de Melo:
"New features:
perf ftrace:
- Add -n/--use-nsec option to the 'latency' subcommand.
Default: usecs:
$ sudo perf ftrace latency -T dput -a sleep 1
# DURATION | COUNT | GRAPH |
0 - 1 us | 2098375 | ############################# |
1 - 2 us | 61 | |
2 - 4 us | 33 | |
4 - 8 us | 13 | |
8 - 16 us | 124 | |
16 - 32 us | 123 | |
32 - 64 us | 1 | |
64 - 128 us | 0 | |
128 - 256 us | 1 | |
256 - 512 us | 0 | |
Better granularity with nsec:
$ sudo perf ftrace latency -T dput -a -n sleep 1
# DURATION | COUNT | GRAPH |
0 - 1 us | 0 | |
1 - 2 ns | 0 | |
2 - 4 ns | 0 | |
4 - 8 ns | 0 | |
8 - 16 ns | 0 | |
16 - 32 ns | 0 | |
32 - 64 ns | 0 | |
64 - 128 ns | 1163434 | ############## |
128 - 256 ns | 914102 | ############# |
256 - 512 ns | 884 | |
512 - 1024 ns | 613 | |
1 - 2 us | 31 | |
2 - 4 us | 17 | |
4 - 8 us | 7 | |
8 - 16 us | 123 | |
16 - 32 us | 83 | |
perf lock:
- Add -c/--combine-locks option to merge lock instances in the same
class into a single entry.
# perf lock report -c
Name acquired contended avg wait(ns) total wait(ns) max wait(ns) min wait(ns)
rcu_read_lock 251225 0 0 0 0 0
hrtimer_bases.lock 39450 0 0 0 0 0
&sb->s_type->i_l... 10301 1 662 662 662 662
ptlock_ptr(page) 10173 2 701 1402 760 642
&(ei->i_block_re... 8732 0 0 0 0 0
&xa->xa_lock 8088 0 0 0 0 0
&base->lock 6705 0 0 0 0 0
&p->pi_lock 5549 0 0 0 0 0
&dentry->d_lockr... 5010 4 1274 5097 1844 789
&ep->lock 3958 0 0 0 0 0
- Add -F/--field option to customize the list of fields to output:
$ perf lock report -F contended,wait_max -k avg_wait
Name contended max wait(ns) avg wait(ns)
slock-AF_INET6 1 23543 23543
&lruvec->lru_lock 5 18317 11254
slock-AF_INET6 1 10379 10379
rcu_node_1 1 2104 2104
&dentry->d_lockr... 1 1844 1844
&dentry->d_lockr... 1 1672 1672
&newf->file_lock 15 2279 1025
&dentry->d_lockr... 1 792 792
- Add --synth=no option for record, as there is no need to symbolize,
lock names comes from the tracepoints.
perf record:
- Threaded recording, opt-in, via the new --threads command line
option.
- Improve AMD IBS (Instruction-Based Sampling) error handling
messages.
perf script:
- Add 'brstackinsnlen' field (use it with -F) for branch stacks.
- Output branch sample type in 'perf script'.
perf report:
- Add "addr_from" and "addr_to" sort dimensions.
- Print branch stack entry type in 'perf report --dump-raw-trace'
- Fix symbolization for chrooted workloads.
Hardware tracing:
Intel PT:
- Add CFE (Control Flow Event) and EVD (Event Data) packets support.
- Add MODE.Exec IFLAG bit support.
Explanation about these features from the "Intel® 64 and IA-32
architectures software developer’s manual combined volumes: 1, 2A,
2B, 2C, 2D, 3A, 3B, 3C, 3D, and 4" PDF at:
https://cdrdv2.intel.com/v1/dl/getContent/671200
At page 3951:
"32.2.4
Event Trace is a capability that exposes details about the
asynchronous events, when they are generated, and when their
corresponding software event handler completes execution. These
include:
o Interrupts, including NMI and SMI, including the interrupt
vector when defined.
o Faults, exceptions including the fault vector.
- Page faults additionally include the page fault address,
when in context.
o Event handler returns, including IRET and RSM.
o VM exits and VM entries.¹
- VM exits include the values written to the “exit reason”
and “exit qualification” VMCS fields. INIT and SIPI events.
o TSX aborts, including the abort status returned for the RTM
instructions.
o Shutdown.
Additionally, it provides indication of the status of the
Interrupt Flag (IF), to indicate when interrupts are masked"
ARM CoreSight:
- Use advertised caps/min_interval as default sample_period on ARM
spe.
- Update deduction of TRCCONFIGR register for branch broadcast on
ARM's CoreSight ETM.
Vendor Events (JSON):
Intel:
- Update events and metrics for: Alderlake, Broadwell, Broadwell DE,
BroadwellX, CascadelakeX, Elkhartlake, Bonnell, Goldmont,
GoldmontPlus, Westmere EP-DP, Haswell, HaswellX, Icelake, IcelakeX,
Ivybridge, Ivytown, Jaketown, Knights Landing, Nehalem EP,
Sandybridge, Silvermont, Skylake, Skylake Server, SkylakeX,
Tigerlake, TremontX, Westmere EP-SP, and Westmere EX.
ARM:
- Add support for HiSilicon CPA PMU aliasing.
perf stat:
- Fix forked applications enablement of counters.
- The 'slots' should only be printed on a different order than the
one specified on the command line when 'topdown' events are
present, fix it.
Miscellaneous:
- Sync msr-index, cpufeatures header files with the kernel sources.
- Stop using some deprecated libbpf APIs in 'perf trace'.
- Fix some spelling mistakes.
- Refactor the maps pointers usage to pave the way for using refcount
debugging.
- Only offer the --tui option on perf top, report and annotate when
perf was built with libslang.
- Don't mention --to-ctf in 'perf data --help' when not linking with
the required library, libbabeltrace.
- Use ARRAY_SIZE() instead of ad hoc equivalent, spotted by
array_size.cocci.
- Enhance the matching of sub-commands abbreviations:
'perf c2c rec' -> 'perf c2c record'
'perf c2c recport -> error
- Set build-id using build-id header on new mmap records.
- Fix generation of 'perf --version' string.
perf test:
- Add test for the arm_spe event.
- Add test to check unwinding using fame-pointer (fp) mode on arm64.
- Make metric testing more robust in 'perf test'.
- Add error message for unsupported branch stack cases.
libperf:
- Add API for allocating new thread map array.
- Fix typo in perf_evlist__open() failure error messages in libperf
tests.
perf c2c:
- Replace bitmap_weight() with bitmap_empty() where appropriate"
* tag 'perf-tools-for-v5.18-2022-03-26' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux: (143 commits)
perf evsel: Improve AMD IBS (Instruction-Based Sampling) error handling messages
perf python: Add perf_env stubs that will be needed in evsel__open_strerror()
perf tools: Enhance the matching of sub-commands abbreviations
libperf tests: Fix typo in perf_evlist__open() failure error messages
tools arm64: Import cputype.h
perf lock: Add -F/--field option to control output
perf lock: Extend struct lock_key to have print function
perf lock: Add --synth=no option for record
tools headers cpufeatures: Sync with the kernel sources
tools headers cpufeatures: Sync with the kernel sources
perf stat: Fix forked applications enablement of counters
tools arch x86: Sync the msr-index.h copy with the kernel sources
perf evsel: Make evsel__env() always return a valid env
perf build-id: Fix spelling mistake "Cant" -> "Can't"
perf header: Fix spelling mistake "could't" -> "couldn't"
perf script: Add 'brstackinsnlen' for branch stacks
perf parse-events: Move slots only with topdown
perf ftrace latency: Update documentation
perf ftrace latency: Add -n/--use-nsec option
perf tools: Fix version kernel tag
...
Pull flexible-array transformations from Gustavo Silva:
"Treewide patch that replaces zero-length arrays with flexible-array
members.
This has been baking in linux-next for a whole development cycle"
* tag 'flexible-array-transformations-5.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gustavoars/linux:
treewide: Replace zero-length arrays with flexible-array members
The existing API perf_thread_map__new_dummy() allocates new thread map
for one thread. I couldn't find a way to reallocate the map with more
threads, or to allocate a new map for more than one thread.
Having multiple threads in a thread map is essential for some use cases.
That's why a new API is proposed, which allocates a new thread map for
given number of threads: perf_thread_map__new_array()
Signed-off-by: Tzvetomir Stoyanov (VMware) <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Ian Rogers <[email protected]>
Link: https://lore.kernel.org/linux-perf-users/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
The "int thread" input arguments of some perf_thead_map APIs are index
of the thread in the thread map.
In order to avoid confusion and to make the APIs consistent with
perf_cpu_map APIs, those arguments are renamed to "int idx".
Suggested-by: Arnaldo Carvalho de Melo <[email protected]>
Signed-off-by: Tzvetomir Stoyanov (VMware) <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Ian Rogers <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
There is a regular need in the kernel to provide a way to declare
having a dynamically sized set of trailing elements in a structure.
Kernel code should always use “flexible array members”[1] for these
cases. The older style of one-element or zero-length arrays should
no longer be used[2].
This code was transformed with the help of Coccinelle:
(next-20220214$ spatch --jobs $(getconf _NPROCESSORS_ONLN) --sp-file script.cocci --include-headers --dir . > output.patch)
@@
identifier S, member, array;
type T1, T2;
@@
struct S {
...
T1 member;
T2 array[
- 0
];
};
UAPI and wireless changes were intentionally excluded from this patch
and will be sent out separately.
[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.16/process/deprecated.html#zero-length-and-one-element-arrays
Link: https://github.com/KSPP/linux/issues/78
Reviewed-by: Kees Cook <[email protected]>
Signed-off-by: Gustavo A. R. Silva <[email protected]>
Tzvetomir Stoyanov reported an issue with using macro
perf_cpu_map__for_each_cpu using private perf_cpu object.
The issue is caused by recent change that wrapped cpu in struct perf_cpu
to distinguish it from cpu indexes. We need to make struct perf_cpu
public.
Add a simple test for using the perf_cpu_map__for_each_cpu macro.
Fixes: 6d18804b96 ("perf cpumap: Give CPUs their own type")
Reported-by: Tzvetomir Stoyanov (VMware) <[email protected]>
Signed-off-by: Jiri Olsa <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
libperf's verbose printing checks the -v option every time the macro _T_ START
is called.
Since there are currently four libperf tests registered, the macro _T_ START is
called four times, but verbose printing after the second time is not output.
Resets the index of the element processed by getopt() and fix verbose printing
so that it prints in all tests.
Signed-off-by: Shunsuke Nakamura <[email protected]>
Acked-by: Rob Herring <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Move evsel::leader to perf_evsel::leader, so we can move the group
interface to libperf.
Also add several evsel helpers to ease up the transition:
struct evsel *evsel__leader(struct evsel *evsel);
- get leader evsel
bool evsel__has_leader(struct evsel *evsel, struct evsel *leader);
- true if evsel has leader as leader
bool evsel__is_leader(struct evsel *evsel);
- true if evsel is itw own leader
void evsel__set_leader(struct evsel *evsel, struct evsel *leader);
- set leader for evsel
Committer notes:
Fix this when building with 'make BUILD_BPF_SKEL=1'
tools/perf/util/bpf_counter.c
- if (evsel->leader->core.nr_members > 1) {
+ if (evsel->core.leader->nr_members > 1) {
Signed-off-by: Jiri Olsa <[email protected]>
Requested-by: Shunsuke Nakamura <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Michael Petlan <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Move evsel::idx to perf_evsel::idx, so we can move the group interface
to libperf.
Committer notes:
Fixup evsel->idx usage in tools/perf/util/bpf_counter_cgroup.c, that
appeared in my tree in my local tree.
Also fixed up these:
$ find tools/perf/ -name "*.[ch]" | xargs grep 'evsel->idx'
tools/perf/ui/gtk/annotate.c: evsel->idx + i);
tools/perf/ui/gtk/annotate.c: evsel->idx);
$
That running 'make -C tools/perf build-test' caught.
Signed-off-by: Jiri Olsa <[email protected]>
Requested-by: Shunsuke Nakamura <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Michael Petlan <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Commit d110162caf ("perf tsc: Support cap_user_time_short for
event TIME_CONV") supports the extended parameters for event TIME_CONV,
but it broke the backwards compatibility, so any perf data file with old
event format fails to convert timestamp.
This patch introduces a helper event_contains() to check if an event
contains a specific member or not. For the backwards-compatibility, if
the event size confirms the extended parameters are supported in the
event TIME_CONV, then copies these parameters.
Committer notes:
To make this compiler backwards compatible add this patch:
- struct perf_tsc_conversion tc = { 0 };
+ struct perf_tsc_conversion tc = { .time_shift = 0, };
Fixes: d110162caf ("perf tsc: Support cap_user_time_short for event TIME_CONV")
Signed-off-by: Leo Yan <[email protected]>
Acked-by: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Gustavo A. R. Silva <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Steve MacLean <[email protected]>
Cc: Yonatan Goldschmidt <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
xyarray__entry() is missing any bounds checking yet often the x and y
parameters come from external callers. Add bounds checks and an
unchecked __xyarray__entry().
Committer notes:
Make the 'x' and 'y' arguments to the new xyarray__entry() that does
bounds check to be of type 'size_t', so that we cover also the case
where 'x' and 'y' could be negative, which is needed anyway as having
them as 'int' breaks the build with:
/home/acme/git/perf/tools/lib/perf/include/internal/xyarray.h: In function ‘xyarray__entry’:
/home/acme/git/perf/tools/lib/perf/include/internal/xyarray.h:28:8: error: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Werror=sign-compare]
28 | if (x >= xy->max_x || y >= xy->max_y)
| ^~
/home/acme/git/perf/tools/lib/perf/include/internal/xyarray.h:28:26: error: comparison of integer expressions of different signedness: ‘int’ and ‘size_t’ {aka ‘long unsigned int’} [-Werror=sign-compare]
28 | if (x >= xy->max_x || y >= xy->max_y)
| ^~
cc1: all warnings being treated as errors
Signed-off-by: Rob Herring <[email protected]>
Suggested-by: Arnaldo Carvalho de Melo <[email protected]>
Suggested-by: Namhyung Kim <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
In order to support usersapce access, an event must be mmapped. While
there's already mmap support for evlist, the usecase is a bit different
than the self monitoring with userspace access. So let's add new
perf_evsel__mmap()/perf_evsel_munmap() functions to mmap/munmap an
evsel. This allows implementing userspace access as a fastpath for
perf_evsel__read().
The mmapped address is returned by perf_evsel__mmap_base() which
primarily for users/tests to check if userspace access is enabled.
Signed-off-by: Rob Herring <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Itaru Kitayama <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Will Deacon <[email protected]>
Link: http://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
We do not store size with build ids in perf data, but there's enough
space to do it. Adding misc bit PERF_RECORD_MISC_BUILD_ID_SIZE to mark
build id event with size.
With this fix the dso with md5 build id will have correct build id data
and will be usable for debuginfod processing if needed (coming in
following patches).
Committer notes:
Use %zu with size_t to fix this error on 32-bit arches:
util/header.c: In function '__event_process_build_id':
util/header.c:2105:3: error: format '%lu' expects argument of type 'long unsigned int', but argument 6 has type 'size_t' [-Werror=format=]
pr_debug("build id event received for %s: %s [%lu]\n",
^
Signed-off-by: Jiri Olsa <[email protected]>
Acked-by: Ian Rogers <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Adrian Hunter