perf annotate: Add and use ins__is_nop()

Likewise, add ins__is_nop() to check if the current instruction is NOP.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.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>
Link: https://lore.kernel.org/r/20240329215812.537846-3-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate: Use ins__is_xxx() if possible

This is to prepare separation of disasm related code.  Use the public
ins API instead of checking the internal data structure.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.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>
Link: https://lore.kernel.org/r/20240329215812.537846-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Use evsel__name_is() helper

Code cleanup, replace strcmp(evsel__name(evsel, {NAME})) with
evsel__name_is() helper.

No functional change.

Committer notes:

Fix this build error:

          trace.syscalls.events.bpf_output = evlist__last(trace.evlist);
  -       assert(evsel__name_is(trace.syscalls.events.bpf_output), "__augmented_syscalls__");
  +       assert(evsel__name_is(trace.syscalls.events.bpf_output, "__augmented_syscalls__"));

Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Yang Jihong <yangjihong@bytedance.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@arm.com>
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>
Link: https://lore.kernel.org/r/20240401062724.1006010-3-yangjihong@bytedance.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf sched timehist: Fix -g/--call-graph option failure

When 'perf sched' enables the call-graph recording, sample_type of dummy
event does not have PERF_SAMPLE_CALLCHAIN, timehist_check_attr() checks
that the evsel does not have a callchain, and set show_callchain to 0.

Currently 'perf sched timehist' only saves callchain when processing the
'sched:sched_switch event', timehist_check_attr() only needs to determine
whether the event has PERF_SAMPLE_CALLCHAIN.

Before:

  # perf sched record -g true
  [ perf record: Woken up 0 times to write data ]
  [ perf record: Captured and wrote 4.153 MB perf.data (7536 samples) ]
  # perf sched timehist
  Samples do not have callchains.
             time    cpu  task name                       wait time  sch delay   run time
                          [tid/pid]                          (msec)     (msec)     (msec)
  --------------- ------  ------------------------------  ---------  ---------  ---------
    147851.826019 [0000]  perf[285035]                        0.000      0.000      0.000
    147851.826029 [0000]  migration/0[15]                     0.000      0.003      0.009
    147851.826063 [0001]  perf[285035]                        0.000      0.000      0.000
    147851.826069 [0001]  migration/1[21]                     0.000      0.003      0.006
  <SNIP>

After:

  # perf sched record -g true
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 2.572 MB perf.data (822 samples) ]
  # perf sched timehist
         time cpu task name        waittime  sch delay  runtime
                    [tid/pid]        (msec)  (msec)    (msec)
  ----------- --- ---------------  --------  --------  -----
  4193.035164 [0] perf[277062]        0.000     0.000   0.000 __traceiter_sched_switch <- __traceiter_sched_switch <- __sched_text_start <- preempt_schedule_common <- __cond_resched <- __wait_for_common <- wait_for_completion
  4193.035174 [0] migration/0[15]     0.000     0.003   0.009 __traceiter_sched_switch <- __traceiter_sched_switch <- __sched_text_start <- smpboot_thread_fn <- kthread <- ret_from_fork
  4193.035207 [1] perf[277062]        0.000     0.000   0.000 __traceiter_sched_switch <- __traceiter_sched_switch <- __sched_text_start <- preempt_schedule_common <- __cond_resched <- __wait_for_common <- wait_for_completion
  4193.035214 [1] migration/1[21]     0.000     0.003   0.007 __traceiter_sched_switch <- __traceiter_sched_switch <- __sched_text_start <- smpboot_thread_fn <- kthread <- ret_from_fork
  <SNIP>

Fixes: 9c95e4ef06572349 ("perf evlist: Add evlist__findnew_tracking_event() helper")
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Yang Jihong <yangjihong@bytedance.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@arm.com>
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: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240401062724.1006010-2-yangjihong@bytedance.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate: Honor output options with --data-type

For data type profiling output, it should be in sync with normal output
so make it display percentage for each field.  Also use coloring scheme
for users to identify fields with big overhead easily.

Users can use --show-total-period or --show-nr-samples to change the
output style like in the normal perf annotate output.

Before:

  $ perf annotate --data-type
  Annotate type: 'struct task_struct' in [kernel.kallsyms] (34 samples):
  ============================================================================
      samples     offset       size  field
           34          0       9792  struct task_struct    {
            2          0         24      struct thread_info       thread_info {
            0          0          8          long unsigned int    flags;
            1          8          8          long unsigned int    syscall_work;
            0         16          4          u32  status;
            1         20          4          u32  cpu;
                                         };

After:

  $ perf annotate --data-type
  Annotate type: 'struct task_struct' in [kernel.kallsyms] (34 samples):
  ============================================================================
   Percent     offset       size  field
    100.00          0       9792  struct task_struct       {
      3.55          0         24      struct thread_info  thread_info {
      0.00          0          8          long unsigned int       flags;
      1.63          8          8          long unsigned int       syscall_work;
      0.00         16          4          u32     status;
      1.91         20          4          u32     cpu;
                                      };

Committer testing:

First collect a suitable perf.data file for use with 'perf annotate --data-type':

  root@number:~# perf mem record -a sleep 1s
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 11.047 MB perf.data (3466 samples) ]
  root@number:~#

Then, before:

  root@number:~# perf annotate --data-type
  Annotate type: 'union ' in /usr/lib64/libc.so.6 (6 samples):
  ============================================================================
      samples     offset       size  field
            6          0         40  union         {
            6          0         40      struct __pthread_mutex_s __data {
            2          0          4          int  __lock;
            0          4          4          unsigned int __count;
            0          8          4          int  __owner;
            1         12          4          unsigned int __nusers;
            2         16          4          int  __kind;
            1         20          2          short int    __spins;
            0         22          2          short int    __elision;
            0         24         16          __pthread_list_t     __list {
            0         24          8              struct __pthread_internal_list*  __prev;
            0         32          8              struct __pthread_internal_list*  __next;
                                             };
                                         };
            0          0          0      char*    __size;
            2          0          8      long int __align;
                                     };
  <SNIP>

And after:

  Annotate type: 'union ' in /usr/lib64/libc.so.6 (6 samples):
  ============================================================================
   Percent     offset       size  field
    100.00          0         40  union    {
    100.00          0         40      struct __pthread_mutex_s    __data {
     31.27          0          4          int     __lock;
      0.00          4          4          unsigned int    __count;
      0.00          8          4          int     __owner;
      7.67         12          4          unsigned int    __nusers;
     53.10         16          4          int     __kind;
      7.96         20          2          short int       __spins;
      0.00         22          2          short int       __elision;
      0.00         24         16          __pthread_list_t        __list {
      0.00         24          8              struct __pthread_internal_list*     __prev;
      0.00         32          8              struct __pthread_internal_list*     __next;
                                          };
                                      };
      0.00          0          0      char*       __size;
     31.27          0          8      long int    __align;
                                  };
  <SNIP>

The lines with percentages >= 7.67 have its percentages red colored.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
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: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240322224313.423181-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate: Get rid of duplicate --group option item

The options array in cmd_annotate() has duplicate --group options.  It
only needs one and let's get rid of the other.

  $ perf annotate -h 2>&1 | grep group
        --group           Show event group information together
        --group           Show event group information together

Fixes: 7ebaf4890f63eb90 ("perf annotate: Support '--group' option")
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
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: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240322224313.423181-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add Kan Liang to MAINTAINERS as a reviewer

Kan has been reviewing patches regularly, add him as a perf tools
reviewer so that people CC him on new patches.

Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Ian Rogers <irogers@google.com>
Acked-by: "Liang, Kan" <kan.liang@linux.intel.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move uapi/linux/vhost.h copy out of the directory used to build perf

It is only used to generate string tables, not to build perf, so move it
to the tools/perf/trace/beauty/include/ hierarchy, that is used just for
scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it, just <linux/vhost.h> coming from
either 'make install_headers' or from the system /usr/include/
directory.

Suggested-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf dso: Reorder members to save space in 'struct dso'

Save 40 bytes and move from 8 to 7 cache lines. Make member dwfl
dependent on being a powerpc build. Squeeze bits of int/enum types
when appropriate. Remove holes/padding by reordering variables.

Before:

  struct dso {
          struct mutex               lock;                 /*     0    40 */
          struct list_head           node;                 /*    40    16 */
          struct rb_node             rb_node __attribute__((__aligned__(8))); /*    56    24 */
          /* --- cacheline 1 boundary (64 bytes) was 16 bytes ago --- */
          struct rb_root *           root;                 /*    80     8 */
          struct rb_root_cached      symbols;              /*    88    16 */
          struct symbol * *          symbol_names;         /*   104     8 */
          size_t                     symbol_names_len;     /*   112     8 */
          struct rb_root_cached      inlined_nodes;        /*   120    16 */
          /* --- cacheline 2 boundary (128 bytes) was 8 bytes ago --- */
          struct rb_root_cached      srclines;             /*   136    16 */
          struct {
                  u64                addr;                 /*   152     8 */
                  struct symbol *    symbol;               /*   160     8 */
          } last_find_result;                              /*   152    16 */
          void *                     a2l;                  /*   168     8 */
          char *                     symsrc_filename;      /*   176     8 */
          unsigned int               a2l_fails;            /*   184     4 */
          enum dso_space_type        kernel;               /*   188     4 */
          /* --- cacheline 3 boundary (192 bytes) --- */
          _Bool                      is_kmod;              /*   192     1 */

          /* XXX 3 bytes hole, try to pack */

          enum dso_swap_type         needs_swap;           /*   196     4 */
          enum dso_binary_type       symtab_type;          /*   200     4 */
          enum dso_binary_type       binary_type;          /*   204     4 */
          enum dso_load_errno        load_errno;           /*   208     4 */
          u8                         adjust_symbols:1;     /*   212: 0  1 */
          u8                         has_build_id:1;       /*   212: 1  1 */
          u8                         header_build_id:1;    /*   212: 2  1 */
          u8                         has_srcline:1;        /*   212: 3  1 */
          u8                         hit:1;                /*   212: 4  1 */
          u8                         annotate_warned:1;    /*   212: 5  1 */
          u8                         auxtrace_warned:1;    /*   212: 6  1 */
          u8                         short_name_allocated:1; /*   212: 7  1 */
          u8                         long_name_allocated:1; /*   213: 0  1 */
          u8                         is_64_bit:1;          /*   213: 1  1 */

          /* XXX 6 bits hole, try to pack */

          _Bool                      sorted_by_name;       /*   214     1 */
          _Bool                      loaded;               /*   215     1 */
          u8                         rel;                  /*   216     1 */

          /* XXX 7 bytes hole, try to pack */

          struct build_id            bid;                  /*   224    32 */
          /* --- cacheline 4 boundary (256 bytes) --- */
          u64                        text_offset;          /*   256     8 */
          u64                        text_end;             /*   264     8 */
          const char  *              short_name;           /*   272     8 */
          const char  *              long_name;            /*   280     8 */
          u16                        long_name_len;        /*   288     2 */
          u16                        short_name_len;       /*   290     2 */

          /* XXX 4 bytes hole, try to pack */

          void *                     dwfl;                 /*   296     8 */
          struct auxtrace_cache *    auxtrace_cache;       /*   304     8 */
          int                        comp;                 /*   312     4 */

          /* XXX 4 bytes hole, try to pack */

          /* --- cacheline 5 boundary (320 bytes) --- */
          struct {
                  struct rb_root     cache;                /*   320     8 */
                  int                fd;                   /*   328     4 */
                  int                status;               /*   332     4 */
                  u32                status_seen;          /*   336     4 */

                  /* XXX 4 bytes hole, try to pack */

                  u64                file_size;            /*   344     8 */
                  struct list_head   open_entry;           /*   352    16 */
                  u64                elf_base_addr;        /*   368     8 */
                  u64                debug_frame_offset;   /*   376     8 */
                  /* --- cacheline 6 boundary (384 bytes) --- */
                  u64                eh_frame_hdr_addr;    /*   384     8 */
                  u64                eh_frame_hdr_offset;  /*   392     8 */
          } data;                                          /*   320    80 */
          struct {
                  u32                id;                   /*   400     4 */
                  u32                sub_id;               /*   404     4 */
                  struct perf_env *  env;                  /*   408     8 */
          } bpf_prog;                                      /*   400    16 */
          union {
                  void *             priv;                 /*   416     8 */
                  u64                db_id;                /*   416     8 */
          };                                               /*   416     8 */
          struct nsinfo *            nsinfo;               /*   424     8 */
          struct dso_id              id;                   /*   432    24 */
          /* --- cacheline 7 boundary (448 bytes) was 8 bytes ago --- */
          refcount_t                 refcnt;               /*   456     4 */
          char                       name[];               /*   460     0 */

          /* size: 464, cachelines: 8, members: 49 */
          /* sum members: 440, holes: 4, sum holes: 18 */
          /* sum bitfield members: 10 bits, bit holes: 1, sum bit holes: 6 bits */
          /* padding: 4 */
          /* forced alignments: 1 */
          /* last cacheline: 16 bytes */
  } __attribute__((__aligned__(8)));

After:

  struct dso {
          struct mutex               lock;                 /*     0    40 */
          struct list_head           node;                 /*    40    16 */
          struct rb_node             rb_node __attribute__((__aligned__(8))); /*    56    24 */
          /* --- cacheline 1 boundary (64 bytes) was 16 bytes ago --- */
          struct rb_root *           root;                 /*    80     8 */
          struct rb_root_cached      symbols;              /*    88    16 */
          struct symbol * *          symbol_names;         /*   104     8 */
          size_t                     symbol_names_len;     /*   112     8 */
          struct rb_root_cached      inlined_nodes;        /*   120    16 */
          /* --- cacheline 2 boundary (128 bytes) was 8 bytes ago --- */
          struct rb_root_cached      srclines;             /*   136    16 */
          struct {
                  u64                addr;                 /*   152     8 */
                  struct symbol *    symbol;               /*   160     8 */
          } last_find_result;                              /*   152    16 */
          struct build_id            bid;                  /*   168    32 */
          /* --- cacheline 3 boundary (192 bytes) was 8 bytes ago --- */
          u64                        text_offset;          /*   200     8 */
          u64                        text_end;             /*   208     8 */
          const char  *              short_name;           /*   216     8 */
          const char  *              long_name;            /*   224     8 */
          void *                     a2l;                  /*   232     8 */
          char *                     symsrc_filename;      /*   240     8 */
          struct nsinfo *            nsinfo;               /*   248     8 */
          /* --- cacheline 4 boundary (256 bytes) --- */
          struct auxtrace_cache *    auxtrace_cache;       /*   256     8 */
          union {
                  void *             priv;                 /*   264     8 */
                  u64                db_id;                /*   264     8 */
          };                                               /*   264     8 */
          struct {
                  struct perf_env *  env;                  /*   272     8 */
                  u32                id;                   /*   280     4 */
                  u32                sub_id;               /*   284     4 */
          } bpf_prog;                                      /*   272    16 */
          struct {
                  struct rb_root     cache;                /*   288     8 */
                  struct list_head   open_entry;           /*   296    16 */
                  u64                file_size;            /*   312     8 */
                  /* --- cacheline 5 boundary (320 bytes) --- */
                  u64                elf_base_addr;        /*   320     8 */
                  u64                debug_frame_offset;   /*   328     8 */
                  u64                eh_frame_hdr_addr;    /*   336     8 */
                  u64                eh_frame_hdr_offset;  /*   344     8 */
                  int                fd;                   /*   352     4 */
                  int                status;               /*   356     4 */
                  u32                status_seen;          /*   360     4 */
          } data;                                          /*   288    80 */

          /* XXX last struct has 4 bytes of padding */

          struct dso_id              id;                   /*   368    24 */
          /* --- cacheline 6 boundary (384 bytes) was 8 bytes ago --- */
          unsigned int               a2l_fails;            /*   392     4 */
          int                        comp;                 /*   396     4 */
          refcount_t                 refcnt;               /*   400     4 */
          enum dso_load_errno        load_errno;           /*   404     4 */
          u16                        long_name_len;        /*   408     2 */
          u16                        short_name_len;       /*   410     2 */
          enum dso_binary_type       symtab_type:8;        /*   412: 0  4 */
          enum dso_binary_type       binary_type:8;        /*   412: 8  4 */
          enum dso_space_type        kernel:2;             /*   412:16  4 */
          enum dso_swap_type         needs_swap:2;         /*   412:18  4 */

          /* Bitfield combined with next fields */

          _Bool                      is_kmod:1;            /*   414: 4  1 */
          u8                         adjust_symbols:1;     /*   414: 5  1 */
          u8                         has_build_id:1;       /*   414: 6  1 */
          u8                         header_build_id:1;    /*   414: 7  1 */
          u8                         has_srcline:1;        /*   415: 0  1 */
          u8                         hit:1;                /*   415: 1  1 */
          u8                         annotate_warned:1;    /*   415: 2  1 */
          u8                         auxtrace_warned:1;    /*   415: 3  1 */
          u8                         short_name_allocated:1; /*   415: 4  1 */
          u8                         long_name_allocated:1; /*   415: 5  1 */
          u8                         is_64_bit:1;          /*   415: 6  1 */

          /* XXX 1 bit hole, try to pack */

          _Bool                      sorted_by_name;       /*   416     1 */
          _Bool                      loaded;               /*   417     1 */
          u8                         rel;                  /*   418     1 */
          char                       name[];               /*   419     0 */

          /* size: 424, cachelines: 7, members: 48 */
          /* sum members: 415 */
          /* sum bitfield members: 31 bits, bit holes: 1, sum bit holes: 1 bits */
          /* padding: 5 */
          /* paddings: 1, sum paddings: 4 */
          /* forced alignments: 1 */
          /* last cacheline: 40 bytes */
  } __attribute__((__aligned__(8)));

Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Ben Gainey <ben.gainey@arm.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Chengen Du <chengen.du@canonical.com>
Cc: Colin Ian King <colin.i.king@gmail.com>
Cc: Ilkka Koskinen <ilkka@os.amperecomputing.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linux.dev>
Cc: Li Dong <lidong@vivo.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Markus Elfring <Markus.Elfring@web.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Song Liu <song@kernel.org>
Cc: Sun Haiyong <sunhaiyong@loongson.cn>
Cc: Yanteng Si <siyanteng@loongson.cn>
Cc: zhaimingbing <zhaimingbing@cmss.chinamobile.com>
Link: https://lore.kernel.org/r/20240321160300.1635121-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf lock contention: Trim backtrace by skipping traceiter functions

The 'perf lock contention' program currently shows the caller of the locks
as __traceiter_contention_begin+0x??. This caller can be ignored, as it is
from the traceiter itself. Instead, it should show the real callers for
the locks.

When fiddling with the --stack-skip parameter, the actual callers for
the locks start to show up. However, just ignore the
__traceiter_contention_begin and the __traceiter_contention_end symbols
so the actual callers will show up.

Before this patch is applied:

sudo perf lock con -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         8      2.33 s       2.28 s     291.18 ms     rwlock:W   __traceiter_contention_begin+0x44
         4      2.33 s       2.28 s     582.35 ms     rwlock:W   __traceiter_contention_begin+0x44
         7    140.30 ms     46.77 ms     20.04 ms     rwlock:W   __traceiter_contention_begin+0x44
         2     63.35 ms     33.76 ms     31.68 ms        mutex   trace_contention_begin+0x84
         2     46.74 ms     46.73 ms     23.37 ms     rwlock:W   __traceiter_contention_begin+0x44
         1     13.54 us     13.54 us     13.54 us        mutex   trace_contention_begin+0x84
         1      3.67 us      3.67 us      3.67 us      rwsem:R   __traceiter_contention_begin+0x44

Before this patch is applied - using --stack-skip 5

sudo perf lock con --stack-skip 5 -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         2      2.24 s       2.24 s       1.12 s      rwlock:W   do_epoll_wait+0x5a0
         4      1.65 s     824.21 ms    412.08 ms     rwlock:W   do_exit+0x338
         2    824.35 ms    824.29 ms    412.17 ms     spinlock   get_signal+0x108
         2    824.14 ms    824.14 ms    412.07 ms     rwlock:W   release_task+0x68
         1     25.22 ms     25.22 ms     25.22 ms        mutex   cgroup_kn_lock_live+0x58
         1     24.71 us     24.71 us     24.71 us     spinlock   do_exit+0x44
         1     22.04 us     22.04 us     22.04 us      rwsem:R   lock_mm_and_find_vma+0xb0

After this patch is applied:

sudo ./perf lock con -a -b -- sleep 3
 contended   total wait     max wait     avg wait         type   caller

         4      4.13 s       2.07 s       1.03 s      rwlock:W   release_task+0x68
         2      2.07 s       2.07 s       1.03 s      rwlock:R   mm_update_next_owner+0x50
         2      2.07 s       2.07 s       1.03 s      rwlock:W   do_exit+0x338
         1     41.56 ms     41.56 ms     41.56 ms        mutex   cgroup_kn_lock_live+0x58
         2     36.12 us     18.83 us     18.06 us     rwlock:W   do_exit+0x338

Signed-off-by: Anne Macedo <retpolanne@posteo.net>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.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: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240319143629.3422590-1-retpolanne@posteo.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Remove info metrics erroneously in TopdownL1

Bug affected server metrics only. This doesn't impact default metrics
but if the TopdownL1 metric group is specified. Passes on the fix in:

  https://github.com/intel/perfmon/commit/b09f0a3953234ec592b4a872b87764c78da05d8b

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-13-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update snowridgex to 1.22

Update events from 1.21 to 1.22 as released in:

  https://github.com/intel/perfmon/commit/ba4f96039f96231b51e3eb69d5a21e2b00f6de5b

Updates various descriptions and removes the event
UNC_IIO_NUM_REQ_FROM_CPU.IRP.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-12-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update skylake to v58

Update events from:

  https://github.com/intel/perfmon/commit/f2e5136e062a91ae554dc40530132e66f9271848

This change didn't increase the version number from v58.

Updates various descriptions.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-11-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update skylakex to 1.33

Update events from 1.32 to 1.33 as released in:

  https://github.com/intel/perfmon/commit/3fe7390dd18496c35ec3a9cf17de0473fd5485cb

Various description updates. Adds the event
OFFCORE_RESPONSE.ALL_READS.L3_HIT.HIT_OTHER_CORE_FWD.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-10-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update sierraforest to 1.02

Update events from 1.01 to 1.02 as released in:

  https://github.com/intel/perfmon/commit/451dd41ae627b56433ad4065bf3632789eb70834

Various description updates. Adds topdown events
TOPDOWN_BAD_SPECULATION.ALL_P, TOPDOWN_BE_BOUND.ALL_P,
TOPDOWN_FE_BOUND.ALL_P and TOPDOWN_RETIRING.ALL_P.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-9-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update sapphirerapids to 1.20

Update events from 1.17 to 1.20 as released in:

  https://github.com/intel/perfmon/commit/6f674057745acf0125395638ca6be36458a59bda

Various description updates. Adds uncore events
UNC_CHA_TOR_INSERTS.IO_ITOMCACHENEAR_LOCAL,
UNC_CHA_TOR_INSERTS.IO_ITOMCACHENEAR_REMOTE,
UNC_CHA_TOR_INSERTS.IO_ITOM_LOCAL, UNC_CHA_TOR_INSERTS.IO_ITOM_REMOTE,
UNC_CHA_TOR_INSERTS.IO_PCIRDCUR_LOCAL,
UNC_CHA_TOR_INSERTS.IO_PCIRDCUR_REMOTE,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_ITOMCACHENEAR_LOCAL,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_ITOMCACHENEAR_REMOTE,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_ITOM_LOCAL,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_ITOM_REMOTE,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_PCIRDCUR_LOCAL,
UNC_CHA_TOR_OCCUPANCY.IO_MISS_PCIRDCUR_REMOTE and removes core events
AMX_OPS_RETIRED.BF16 and AMX_OPS_RETIRED.INT8.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-8-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update meteorlake to 1.08

Update events from 1.07 to 1.08 as released in:

  https://github.com/intel/perfmon/commit/f0f8f3e163d9eb84e6ce8e2108a22cb43b2527e5

Various description updates. Adds topdown, offcore and uncore events
OCR.DEMAND_DATA_RD.L3_HIT, OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_NO_FWD,
OCR.DEMAND_RFO.L3_HIT, OCR.DEMAND_DATA_RD.L3_MISS,
OCR.DEMAND_RFO.L3_MISS, OCR.DEMAND_DATA_RD.ANY_RESPONSE,
OCR.DEMAND_DATA_RD.DRAM, OCR.DEMAND_RFO.ANY_RESPONSE,
OCR.DEMAND_RFO.DRAM, TOPDOWN_BAD_SPECULATION.ALL_P,
TOPDOWN_BE_BOUND.ALL_P, TOPDOWN_FE_BOUND.ALL_P,
TOPDOWN_RETIRING.ALL_P, UNC_ARB_DAT_OCCUPANCY.RD and
UNC_HAC_ARB_COH_TRK_REQUESTS.ALL.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-7-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update lunarlake to 1.01

Update events from 1.00 to 1.01 as released in:

  https://github.com/intel/perfmon/commit/56ab8d837ac566d51a4d8748b6b4b817a22c9b84

Various encoding and description updates. Adds the events
CPU_CLK_UNHALTED.CORE, CPU_CLK_UNHALTED.CORE_P,
CPU_CLK_UNHALTED.REF_TSC_P, CPU_CLK_UNHALTED.THREAD,
MISC_RETIRED.LBR_INSERTS, TOPDOWN_BAD_SPECULATION.ALL_P,
TOPDOWN_BE_BOUND.ALL_P, TOPDOWN_FE_BOUND.ALL_P,
TOPDOWN_RETIRING.ALL_P.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-6-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update icelakex to 1.24

Update events from 1.23 to 1.24 as released in:

  https://github.com/intel/perfmon/commit/d883888ae60882028e387b6fe1ebf683beb693fa

Fixes spelling and descriptions. Adds the uncore events
UNC_CHA_TOR_INSERTS.IO_PCIRDCUR_LOCAL and
UNC_CHA_TOR_INSERTS.IO_PCIRDCUR_REMOTE, while removing
UNC_IIO_NUM_REQ_FROM_CPU.IRP.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-5-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update grandridge to 1.02

Update events from 1.01 to 1.02 as released in:

  https://github.com/intel/perfmon/commit/b2a81e803add1ba0af68a442c975683d226d868c

Fixes spelling and descriptions. Adds topdown events and uncore cache
UNC_CHA_TOR_OCCUPANCY.IA_HIT_DRD_OPT,
UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_OPT,
UNC_CHA_TOR_OCCUPANCY.IA_DRD_OPT.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-4-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update emeraldrapids to 1.06

Update events from 1.03 to 1.96 as released in:

  https://github.com/intel/perfmon/commit/21a8be3ea7918749141db4036fb65a2343cd865d

Fixes spelling and descriptions. Adds cache miss latency events
UNC_CHA_TOR_(INSERTS|OCCUPANCY).IO_(PCIRDCUR|ITOM|ITOMCACHENEAR)_(LOCAL|REMOTE).

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Update cascadelakex to 1.21

Update events from 1.20 to 1.21 as released in:

  https://github.com/intel/perfmon/commit/fcfdba3be8f3be81ad6b509fdebf953ead92dc2c

Largely fixes spelling and descriptions.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Edward Baker <edward.baker@intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Weilin Wang <weilin.wang@intel.com>
Link: https://lore.kernel.org/r/20240321060016.1464787-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Add missing libgen.h header needed for using basename()

This prototype is obtained indirectly, by luck, from some other header
in probe-event.c in most systems, but recently exploded on alpine:edge:

   8    13.39 alpine:edge                   : FAIL gcc version 13.2.1 20240309 (Alpine 13.2.1_git20240309)
    util/probe-event.c: In function 'convert_exec_to_group':
    util/probe-event.c:225:16: error: implicit declaration of function 'basename' [-Werror=implicit-function-declaration]
      225 |         ptr1 = basename(exec_copy);
          |                ^~~~~~~~
    util/probe-event.c:225:14: error: assignment to 'char *' from 'int' makes pointer from integer without a cast [-Werror=int-conversion]
      225 |         ptr1 = basename(exec_copy);
          |              ^
    cc1: all warnings being treated as errors
    make[3]: *** [/git/perf-6.8.0/tools/build/Makefile.build:158: util] Error 2

Fix it by adding the libgen.h header where basename() is prototyped.

Fixes: fb7345bbf7fad9bf ("perf probe: Support basic dwarf-based operations on uprobe events")
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Fix 'newfstatat'/'fstatat' argument pretty printing

There were needless two entries, one for 'newfstatat' and another for
'fstatat', keep just one and pretty print its 'flags' argument using the
fs_at_flags scnprintf that is also used by other FS syscalls such as
'stat', now:

  root@number:~# perf trace -e newfstatat --max-events=5
       0.000 ( 0.010 ms): abrt-dump-jour/1400 newfstatat(dfd: 7, filename: "", statbuf: 0x7fff0d127000, flag: EMPTY_PATH) = 0
       0.020 ( 0.003 ms): abrt-dump-jour/1400 newfstatat(dfd: 9, filename: "", statbuf: 0x55752507b0e8, flag: EMPTY_PATH) = 0
       0.039 ( 0.004 ms): abrt-dump-jour/1400 newfstatat(dfd: 19, filename: "", statbuf: 0x557525061378, flag: EMPTY_PATH) = 0
       0.047 ( 0.003 ms): abrt-dump-jour/1400 newfstatat(dfd: 20, filename: "", statbuf: 0x5575250b8cc8, flag: EMPTY_PATH) = 0
       0.053 ( 0.003 ms): abrt-dump-jour/1400 newfstatat(dfd: 22, filename: "", statbuf: 0x5575250535d8, flag: EMPTY_PATH) = 0
  root@number:~#

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240320193115.811899-6-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Beautify the 'flags' arg of unlinkat

Reusing the fs_at_flags array done for the 'stat' syscall.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240320193115.811899-5-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Introduce faccessat2 flags scnprintf routine

The fsaccessat and fsaccessat2 now have beautifiers for its arguments.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240320193115.811899-4-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Introduce scrape script for the 'statx' syscall 'mask' argument

It was using the first variation on producing a string representation
for a binary flag, one that used the system's stat.h and preprocessor
tricks that had to be updated everytime a new flag was introduced.

Use the more recent scrape script + strarray +
strarray__scnprintf_flags() combo.

  $ tools/perf/trace/beauty/statx_mask.sh
  static const char *statx_mask[] = {
   [ilog2(0x00000001) + 1] = "TYPE",
   [ilog2(0x00000002) + 1] = "MODE",
   [ilog2(0x00000004) + 1] = "NLINK",
   [ilog2(0x00000008) + 1] = "UID",
   [ilog2(0x00000010) + 1] = "GID",
   [ilog2(0x00000020) + 1] = "ATIME",
   [ilog2(0x00000040) + 1] = "MTIME",
   [ilog2(0x00000080) + 1] = "CTIME",
   [ilog2(0x00000100) + 1] = "INO",
   [ilog2(0x00000200) + 1] = "SIZE",
   [ilog2(0x00000400) + 1] = "BLOCKS",
   [ilog2(0x00000800) + 1] = "BTIME",
   [ilog2(0x00001000) + 1] = "MNT_ID",
   [ilog2(0x00002000) + 1] = "DIOALIGN",
   [ilog2(0x00004000) + 1] = "MNT_ID_UNIQUE",
  };
  $

Now we need a copy of uapi/linux/stat.h from tools/include/ in the
scrape only directory tools/perf/trace/beauty/include.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240320193115.811899-3-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Introduce scrape script for various fs syscalls 'flags' arguments

It was using the first variation on producing a string representation
for a binary flag, one that used the system's fcntl.h and preprocessor
tricks that had to be updated everytime a new flag was introduced.

Use the more recent scrape script + strarray + strarray__scnprintf_flags() combo.

  $ tools/perf/trace/beauty/fs_at_flags.sh
  static const char *fs_at_flags[] = {
   [ilog2(0x100) + 1] = "SYMLINK_NOFOLLOW",
   [ilog2(0x200) + 1] = "REMOVEDIR",
   [ilog2(0x400) + 1] = "SYMLINK_FOLLOW",
   [ilog2(0x800) + 1] = "NO_AUTOMOUNT",
   [ilog2(0x1000) + 1] = "EMPTY_PATH",
   [ilog2(0x0000) + 1] = "STATX_SYNC_AS_STAT",
   [ilog2(0x2000) + 1] = "STATX_FORCE_SYNC",
   [ilog2(0x4000) + 1] = "STATX_DONT_SYNC",
   [ilog2(0x8000) + 1] = "RECURSIVE",
   [ilog2(0x80000000) + 1] = "GETATTR_NOSEC",
  };
  $

Now we need a copy of uapi/linux/fcntl.h from tools/include/ in the
scrape only directory tools/perf/trace/beauty/include and will use that
fs_at_flags array for other fs syscalls.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240320193115.811899-2-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Run tests in parallel by default

Switch from running tests sequentially to running in parallel by
default. Change the opt-in '-p' or '--parallel' flag to '-S' or
'--sequential'.

On an 8 core tigerlake an address sanitizer run time changes from:

  326.54user 622.73system 6:59.91elapsed 226%CPU

to:

  973.02user 583.98system 3:01.17elapsed 859%CPU

So over twice as fast, saving 4 minutes.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.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@arm.com>
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>
Link: https://lore.kernel.org/r/20240301174711.2646944-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf help: Lower levenshtein penality for deleting character

The levenshtein penalty for deleting a character was far higher than
subsituting or inserting a character. Lower the penalty to match that
of inserting a character.

Before:

  $ perf recccord
  perf: 'recccord' is not a perf-command. See 'perf --help'.
  $

After:

  $ perf recccord
  perf: 'recccord' is not a perf-command. See 'perf --help'.

  Did you mean this?
          record
  $

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
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>
Link: https://lore.kernel.org/r/20240301201306.2680986-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Suggest inbuilt commands for unknown command

The existing unknown command code looks for perf scripts like
perf-archive.sh and perf-iostat.sh, however, inbuilt commands aren't
suggested. Add the inbuilt commands so they may be suggested too.

Before:

  $ perf reccord
  perf: 'reccord' is not a perf-command. See 'perf --help'.
  $

After:

  $ perf reccord
  perf: 'reccord' is not a perf-command. See 'perf --help'.

  Did you mean this?
          record
  $

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
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>
Link: https://lore.kernel.org/r/20240301201306.2680986-1-irogers@google.com
[ Added some fixes from Ian to problems I noticed while testing ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Read child test 10 times a second rather than 1

Make the perf test output smoother by timing out the poll of the child
process after 100ms rather than 1s.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Disha Goel <disgoel@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
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: Song Liu <songliubraving@fb.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lore.kernel.org/r/20240301074639.2260708-4-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Use a single fd for the child process out/err

Switch from dumping err then out, to a single file descriptor for both
of them. This allows the err and output to be correctly interleaved in
verbose output.

Fixes: b482f5f8e0168f1e ("perf tests: Add option to run tests in parallel")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Disha Goel <disgoel@linux.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lore.kernel.org/r/20240301074639.2260708-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Stat output per thread of just the parent process

Per-thread mode requires either system-wide (-a), a pid (-p) or a tid
(-t).

The stat output tests were using system-wide mode but this is racy when
threads are starting and exiting - something that happens a lot when
running the tests in parallel (perf test -p).

Avoid the race conditions by using pid mode with the pid of the parent
process.

Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Disha Goel <disgoel@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
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: Song Liu <songliubraving@fb.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lore.kernel.org/r/20240301074639.2260708-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Delete session after stopping sideband thread

The session has a header in it which contains a perf env with
bpf_progs. The bpf_progs are accessed by the sideband thread and so
the sideband thread must be stopped before the session is deleted, to
avoid a use after free.  This error was detected by AddressSanitizer
in the following:

  ==2054673==ERROR: AddressSanitizer: heap-use-after-free on address 0x61d000161e00 at pc 0x55769289de54 bp 0x7f9df36d4ab0 sp 0x7f9df36d4aa8
  READ of size 8 at 0x61d000161e00 thread T1
      #0 0x55769289de53 in __perf_env__insert_bpf_prog_info util/env.c:42
      #1 0x55769289dbb1 in perf_env__insert_bpf_prog_info util/env.c:29
      #2 0x557692bbae29 in perf_env__add_bpf_info util/bpf-event.c:483
      #3 0x557692bbb01a in bpf_event__sb_cb util/bpf-event.c:512
      #4 0x5576928b75f4 in perf_evlist__poll_thread util/sideband_evlist.c:68
      #5 0x7f9df96a63eb in start_thread nptl/pthread_create.c:444
      #6 0x7f9df9726a4b in clone3 ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

  0x61d000161e00 is located 384 bytes inside of 2136-byte region [0x61d000161c80,0x61d0001624d8)
  freed by thread T0 here:
      #0 0x7f9dfa6d7288 in __interceptor_free libsanitizer/asan/asan_malloc_linux.cpp:52
      #1 0x557692978d50 in perf_session__delete util/session.c:319
      #2 0x557692673959 in __cmd_record tools/perf/builtin-record.c:2884
      #3 0x55769267a9f0 in cmd_record tools/perf/builtin-record.c:4259
      #4 0x55769286710c in run_builtin tools/perf/perf.c:349
      #5 0x557692867678 in handle_internal_command tools/perf/perf.c:402
      #6 0x557692867a40 in run_argv tools/perf/perf.c:446
      #7 0x557692867fae in main tools/perf/perf.c:562
      #8 0x7f9df96456c9 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:58

Fixes: 657ee5531903339b ("perf evlist: Introduce side band thread")
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Disha Goel <disgoel@linux.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Yicong Yang <yangyicong@hisilicon.com>
Link: https://lore.kernel.org/r/20240301074639.2260708-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add/use PMU reverse lookup from config to name

Add perf_pmu__name_from_config that does a reverse lookup from a
config number to an alias name. The lookup is expensive as the config
is computed for every alias by filling in a perf_event_attr, but this
is only done when verbose output is enabled. The lookup also only
considers config, and not config1, config2 or config3.

An example of the output:

  $ perf stat -vv -e data_read true
  ...
  perf_event_attr:
    type                             24 (uncore_imc_free_running_0)
    size                             136
    config                           0x20ff (data_read)
    sample_type                      IDENTIFIER
    read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
    disabled                         1
    inherit                          1
    exclude_guest                    1
  ...

Committer notes:

Fix the python binding build by adding dummies for not strictly
needed perf_pmu__name_from_config() and perf_pmus__find_by_type().

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-7-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Use pmus to describe type from attribute

When dumping a perf_event_attr, use pmus to find the PMU and its name
by the type number. This allows dynamically added PMUs to be described.

Before:

  $ perf stat -vv -e data_read true
  ...
  perf_event_attr:
    type                             24
    size                             136
    config                           0x20ff
    sample_type                      IDENTIFIER
    read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
    disabled                         1
    inherit                          1
    exclude_guest                    1
  ...

After:

  $ perf stat -vv -e data_read true
  ...
  perf_event_attr:
    type                             24 (uncore_imc_free_running_0)
    size                             136
    config                           0x20ff
    sample_type                      IDENTIFIER
    read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
    disabled                         1
    inherit                          1
    exclude_guest                    1
  ...

However, it also means that when we have a PMU name we prefer it to a
hard coded name:

Before:

  $ perf stat -vv -e faults true
  ...
  perf_event_attr:
    type                             1 (PERF_TYPE_SOFTWARE)
    size                             136
    config                           0x2 (PERF_COUNT_SW_PAGE_FAULTS)
    sample_type                      IDENTIFIER
    read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
    disabled                         1
    inherit                          1
    enable_on_exec                   1
    exclude_guest                    1
  ...

After:

  $ perf stat -vv -e faults true
  ...
  perf_event_attr:
    type                             1 (software)
    size                             136
    config                           0x2 (PERF_COUNT_SW_PAGE_FAULTS)
    sample_type                      IDENTIFIER
    read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
    disabled                         1
    inherit                          1
    enable_on_exec                   1
    exclude_guest                    1
  ...

It feels more consistent to do this, rather than only prefer a PMU
name when a hard coded name isn't available.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-6-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf list: Give more details about raw event encodings

List all the PMUs, not just the first core one, and list real format
specifiers with value ranges.

Before:

  $ perf list
  ...
    rNNN                                               [Raw hardware event descriptor]
    cpu/t1=v1[,t2=v2,t3 ...]/modifier                  [Raw hardware event descriptor]
         [(see 'man perf-list' on how to encode it)]
    mem:<addr>[/len][:access]                          [Hardware breakpoint]
  ...

After:

  $ perf list
  ...
    rNNN                                               [Raw event descriptor]
    cpu/event=0..255,pc,edge,.../modifier              [Raw event descriptor]
         [(see 'man perf-list' or 'man perf-record' on how to encode it)]
    breakpoint//modifier                               [Raw event descriptor]
    cstate_core/event=0..0xffffffffffffffff/modifier   [Raw event descriptor]
    cstate_pkg/event=0..0xffffffffffffffff/modifier    [Raw event descriptor]
    i915/i915_eventid=0..0x1fffff/modifier             [Raw event descriptor]
    intel_bts//modifier                                [Raw event descriptor]
    intel_pt/ptw,event,cyc_thresh=0..15,.../modifier   [Raw event descriptor]
    kprobe/retprobe/modifier                           [Raw event descriptor]
    msr/event=0..0xffffffffffffffff/modifier           [Raw event descriptor]
    power/event=0..255/modifier                        [Raw event descriptor]
    software//modifier                                 [Raw event descriptor]
    tracepoint//modifier                               [Raw event descriptor]
    uncore_arb/event=0..255,edge,inv,.../modifier      [Raw event descriptor]
    uncore_cbox/event=0..255,edge,inv,.../modifier     [Raw event descriptor]
    uncore_clock/event=0..255/modifier                 [Raw event descriptor]
    uncore_imc_free_running/event=0..255,umask=0..255/modifier[Raw event descriptor]
    uprobe/ref_ctr_offset=0..0xffffffff,retprobe/modifier[Raw event descriptor]
    mem:<addr>[/len][:access]                          [Hardware breakpoint]
  ...

With '--details' provide more details on the formats encoding:

  cpu/event=0..255,pc,edge,.../modifier              [Raw event descriptor]
       [(see 'man perf-list' or 'man perf-record' on how to encode it)]
        cpu/event=0..255,pc,edge,offcore_rsp=0..0xffffffffffffffff,ldlat=0..0xffff,inv,
        umask=0..255,frontend=0..0xffffff,cmask=0..255,config=0..0xffffffffffffffff,
        config1=0..0xffffffffffffffff,config2=0..0xffffffffffffffff,config3=0..0xffffffffffffffff,
        name=string,period=number,freq=number,branch_type=(u|k|hv|any|...),time,
        call-graph=(fp|dwarf|lbr),stack-size=number,max-stack=number,nr=number,inherit,no-inherit,
        overwrite,no-overwrite,percore,aux-output,aux-sample-size=number/modifier
  breakpoint//modifier                               [Raw event descriptor]
        breakpoint//modifier
  cstate_core/event=0..0xffffffffffffffff/modifier   [Raw event descriptor]
        cstate_core/event=0..0xffffffffffffffff/modifier
  cstate_pkg/event=0..0xffffffffffffffff/modifier    [Raw event descriptor]
        cstate_pkg/event=0..0xffffffffffffffff/modifier
  i915/i915_eventid=0..0x1fffff/modifier             [Raw event descriptor]
        i915/i915_eventid=0..0x1fffff/modifier
  intel_bts//modifier                                [Raw event descriptor]
        intel_bts//modifier
  intel_pt/ptw,event,cyc_thresh=0..15,.../modifier   [Raw event descriptor]
        intel_pt/ptw,event,cyc_thresh=0..15,pt,notnt,branch,tsc,pwr_evt,fup_on_ptw,cyc,noretcomp,
        mtc,psb_period=0..15,mtc_period=0..15/modifier
  kprobe/retprobe/modifier                           [Raw event descriptor]
        kprobe/retprobe/modifier
  msr/event=0..0xffffffffffffffff/modifier           [Raw event descriptor]
        msr/event=0..0xffffffffffffffff/modifier
  power/event=0..255/modifier                        [Raw event descriptor]
        power/event=0..255/modifier
  software//modifier                                 [Raw event descriptor]
        software//modifier
  tracepoint//modifier                               [Raw event descriptor]
        tracepoint//modifier
  uncore_arb/event=0..255,edge,inv,.../modifier      [Raw event descriptor]
        uncore_arb/event=0..255,edge,inv,umask=0..255,cmask=0..31/modifier
  uncore_cbox/event=0..255,edge,inv,.../modifier     [Raw event descriptor]
        uncore_cbox/event=0..255,edge,inv,umask=0..255,cmask=0..31/modifier
  uncore_clock/event=0..255/modifier                 [Raw event descriptor]
        uncore_clock/event=0..255/modifier
  uncore_imc_free_running/event=0..255,umask=0..255/modifier[Raw event descriptor]
        uncore_imc_free_running/event=0..255,umask=0..255/modifier
  uprobe/ref_ctr_offset=0..0xffffffff,retprobe/modifier[Raw event descriptor]
        uprobe/ref_ctr_offset=0..0xffffffff,retprobe/modifier

Committer notes:

Address this build error in various distros:

  55    58.44 ubuntu:24.04                  : FAIL gcc version 13.2.0 (Ubuntu 13.2.0-17ubuntu2)
    util/pmu.c:1638:70: error: '_Static_assert' with no message is a C2x extension [-Werror,-Wc2x-extensions]
     1638 |         _Static_assert(ARRAY_SIZE(terms) == __PARSE_EVENTS__TERM_TYPE_NR - 6);
          |                                                                             ^
          |                                                                             , ""
    1 error generated.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-5-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf list: Allow wordwrap to wrap on commas

A raw event encoding may be a block with terms separated by commas. If
wrapping such a string it would be useful to break at the commas, so
add this ability to wordwrap.

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-4-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf pmu: Drop "default_core" from alias names

"default_core" is used by jevents.py for json events' PMU name when
none is specified. On x86 the "default_core" is typically the PMU
"cpu". When creating an alias see if the event's PMU name is
"default_core" in which case don't record it. This means in places
like "perf list" the PMU's name will be used in its place.

Before:

$ perf list --details
  ...
  cache:
    l1d.replacement
         [Counts the number of cache lines replaced in L1 data cache]
          default_core/event=0x51,period=0x186a3,umask=0x1/
  ...

After:

$ perf list --details
  ...
  cache:
    l1d.replacement
         [Counts the number of cache lines replaced in L1 data cache. Unit: cpu]
          cpu/event=0x51,period=0x186a3,umask=0x1/
  ...

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf list: Add tracepoint encoding to detailed output

The tracepoint id holds the config value and is probed in determining
what an event is. Add reading of the id so that we can display the
event encoding as:

  $ perf list --details
  ...
    alarmtimer:alarmtimer_cancel                       [Tracepoint event]
          tracepoint/config=0x18c/
  ...

Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Tested-by: Kan Liang <kan.liang@linux.intel.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@arm.com>
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: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Yang Jihong <yangjihong1@huawei.com>
Link: https://lore.kernel.org/r/20240308001915.4060155-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Introduce scrape script for 'clone' syscall 'flags' argument

It was using the first variation on producing a string representation
for a binary flag, one that used the copy of uapi/linux/sched.h with
preprocessor tricks that had to be updated everytime a new flag was
introduced.

Use the more recent scrape script + strarray + strarray__scnprintf_flags() combo.

  $ tools/perf/trace/beauty/clone.sh | head -5
  static const char *clone_flags[] = {
   [ilog2(0x00000100) + 1] = "VM",
   [ilog2(0x00000200) + 1] = "FS",
   [ilog2(0x00000400) + 1] = "FILES",
   [ilog2(0x00000800) + 1] = "SIGHAND",
  $

Now we can move uapi/linux/sched.h from tools/include/, that is used for
building perf to the scrape only directory tools/perf/trace/beauty/include.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/ZfnULIn3XKDq0bpc@x1
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Do not retry for invalid types

In some cases, it was able to find a type or location info (for per-cpu
variable) but cannot match because of invalid offset or missing global
information.  In those cases, it's meaningless to go to the outer scope
and retry because there will be no additional information.

Let's change the return type of find_matching_type() and bail out if it
returns -1 for the cases.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-24-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Add a cache for global variable types

They are often searched by many different places.  Let's add a cache
for them to reduce the duplicate DWARF access.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-23-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Add stack canary type

When the stack protector is enabled, compiler would generate code to
check stack overflow with a special value called 'stack carary' at
runtime.  On x86_64, GCC hard-codes the stack canary as %gs:40.

While there's a definition of fixed_percpu_data in asm/processor.h,
it seems that the header is not included everywhere and many places
it cannot find the type info.  As it's in the well-known location (at
%gs:40), let's add a pseudo stack canary type to handle it specially.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-22-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Handle ADD instructions

There are different patterns for percpu variable access using a constant
value added to the base.

  2aeb:  mov    -0x7da0f7e0(,%rax,8),%r14  # r14 = __per_cpu_offset[cpu]
  2af3:  mov    $0x34740,%rax              # rax = address of runqueues
* 2afa:  add    %rax,%r14                  # r14 = &per_cpu(runqueues, cpu)
  2bfd:  cmpl   $0x0,0x10(%r14)            # cpu_rq(cpu)->has_blocked_load
  2b03:  je     0x2b36

At the first instruction, r14 has the __per_cpu_offset.  And then rax
has an immediate value and then added to r14 to calculate the address of
a per-cpu variable.  So it needs to track the immediate values and ADD
instructions.

Similar but a little different case is to use "this_cpu_off" instead of
"__per_cpu_offset" for the current CPU.  This time the variable address
comes with PC-rel addressing.

  89:  mov     $0x34740,%rax                # rax = address of runqueues
* 90:  add     %gs:0x7f015f60(%rip),%rax    # 19a78  <this_cpu_off>
  98:  incl    0xd8c(%rax)                  # cpu_rq(cpu)->sched_count

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-21-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Support general per-cpu access

This is to support per-cpu variable access often without a matching
DWARF entry.  For some reason, I cannot find debug info of per-cpu
variables sometimes.  They have more complex pattern to calculate the
address of per-cpu variables like below.

  2b7d:  mov    -0x1e0(%rbp),%rax           ; rax = cpu
  2b84:  mov    -0x7da0f7e0(,%rax,8),%rcx   ; rcx = __per_cpu_offset[cpu]
* 2b8c:  mov    0x34870(%rcx),%rax          ; *(__per_cpu_offset[cpu] + 0x34870)

Let's assume the rax register has a number for a CPU at 2b7d.  The next
instruction is to get the per-cpu offset' for that cpu.  The offset
-0x7da0f7e0 is 0xffffffff825f0820 in u64 which is the address of the
'__per_cpu_offset' array in my system.  So it'd get the actual offset
of that CPU's per-cpu region and save it to the rcx register.

Then, at 2b8c, accesses using rcx can be handled same as the global
variable access.  To handle this case, it should check if the offset
of the instruction matches to the address of '__per_cpu_offset'.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-20-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Track instructions with a this-cpu variable

Like global variables, this per-cpu variables should be tracked
correctly.  Factor our get_global_var_type() to handle both global
and per-cpu (for this cpu) variables in the same manner.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-19-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Handle this-cpu variables in kernel

On x86, the kernel gets the current task using the current macro like
below:

  #define current  get_current()

  static __always_inline struct task_struct *get_current(void)
  {
      return this_cpu_read_stable(pcpu_hot.current_task);
  }

So it returns the current_task field of struct pcpu_hot which is the
first member.  On my build, it's located at 0x32940.

  $ nm vmlinux | grep pcpu_hot
  0000000000032940 D pcpu_hot

And the current macro generates the instructions like below:

  mov  %gs:0x32940, %rcx

So the %gs segment register points to the beginning of the per-cpu
region of this cpu and it points the variable with a constant.

Let's update the instruction location info to have a segment register
and handle %gs in kernel to look up a global variable.  Pretend it as
a global variable by changing the register number to DWARF_REG_PC.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-18-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate: Parse x86 segment register location

Add a segment field in the struct annotated_insn_loc and save it for the
segment based addressing like %gs:0x28.  For simplicity it now handles
%gs register only.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-17-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Check register state for type

As instruction tracking updates the type state for each register, check
the final type info for the target register at the given instruction.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-16-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Implement instruction tracking

If it failed to find a variable for the location directly, it might be
due to a missing variable in the source code.  For example, accessing
pointer variables in a chain can result in the case like below:

  struct foo *foo = ...;

  int i = foo->bar->baz;

The DWARF debug information is created for each variable so it'd have
one for 'foo'.  But there's no variable for 'foo->bar' and then it
cannot know the type of 'bar' and 'baz'.

The above source code can be compiled to the follow x86 instructions:

  mov  0x8(%rax), %rcx
  mov  0x4(%rcx), %rdx   <=== PMU sample
  mov  %rdx, -4(%rbp)

Let's say 'foo' is located in the %rax and it has a pointer to struct
foo.  But perf sample is captured in the second instruction and there
is no variable or type info for the %rcx.

It'd be great if compiler could generate debug info for %rcx, but we
should handle it on our side.  So this patch implements the logic to
iterate instructions and update the type table for each location.

As it already collected a list of scopes including the target
instruction, we can use it to construct the type table smartly.

  +----------------  scope[0] subprogram
  |
  | +--------------  scope[1] lexical_block
  | |
  | | +------------  scope[2] inlined_subroutine
  | | |
  | | | +----------  scope[3] inlined_subroutine
  | | | |
  | | | | +--------  scope[4] lexical_block
  | | | | |
  | | | | |     ***  target instruction
  ...

Image the target instruction has 5 scopes, each scope will have its own
variables and parameters.  Then it can start with the innermost scope
(4).  So it'd search the shortest path from the start of scope[4] to
the target address and build a list of basic blocks.  Then it iterates
the basic blocks with the variables in the scope and update the table.
If it finds a type at the target instruction, then returns it.

Otherwise, it moves to the upper scope[3].  Now it'd search the shortest
path from the start of scope[3] to the start of scope[4].  Then connect
it to the existing basic block list.  Then it'd iterate the blocks with
variables for both scopes.  It can repeat this until it finds a type at
the target instruction or reaches to the top scope[0].

As the basic blocks contain the shortest path, it won't worry about
branches and can update the table simply.

The final check will be done by find_matching_type() in the next patch.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-15-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Handle call instructions

When updating instruction states, the call instruction should play a
role since it changes the register states.  For simplicity, mark some
registers as caller-saved registers (should be arch-dependent), and
invalidate them all after a function call.

If the function returns something, the designated register (ret_reg)
will have the type info.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-14-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Handle global variable access

When updating the instruction states, it also needs to handle global
variable accesses.  Same as it does for PC-relative addressing, it can
look up the type by address (if it's defined in the same file), or by
name after finding the symbol by address (for declarations).

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-13-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Add get_global_var_type()

Accessing global variable is common when it tracks execution later.
Factor out the common code into a function for later use.

It adds thread and cpumode to struct data_loc_info to find (global)
symbols if needed.  Also remove var_name as it's retrieved in the
helper function.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-12-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Add update_insn_state()

The update_insn_state() function is to update the type state table after
processing each instruction.  For now, it handles MOV (on x86) insn
to transfer type info from the source location to the target.

The location can be a register or a stack slot.  Check carefully when
memory reference happens and fetch the type correctly.  It basically
ignores write to a memory since it doesn't change the type info.  One
exception is writes to (new) stack slots for register spilling.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-11-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Maintain variable type info

As it collected basic block and variable information in each scope, it
now can build a state table to find matching variable at the location.

The struct type_state is to keep the type info saved in each register
and stack slot.  The update_var_state() updates the table when it finds
variables in the current address.  It expects die_collect_vars() filled
a list of variables with type info and starting address.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-10-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Add debug messages

Add a new debug option "type-profile" to enable the detailed info during
the type analysis especially for instruction tracking.  You can use this
before the command name like 'report' or 'annotate'.

  $ perf --debug type-profile annotate --data-type

Committer testing:

First get some memory events:

  $ perf mem record ls

Then, without data-type profiling debug:

  $ perf annotate --data-type | head
  Annotate type: 'struct rtld_global' in /usr/lib64/ld-linux-x86-64.so.2 (1 samples):
  ============================================================================
      samples     offset       size  field
            1          0       4336  struct rtld_global  {
            0          0          0      struct link_namespaces* _dl_ns;
            0       2560          8      size_t _dl_nns;
            0       2568         40      __rtld_lock_recursive_t _dl_load_lock {
            0       2568         40          pthread_mutex_t mutex {
            0       2568         40              struct __pthread_mutex_s __data {
            0       2568          4                  int __lock;
  $

And with only data-type profiling:

  $ perf --debug type-profile annotate --data-type | head
  -----------------------------------------------------------
  find_data_type_die [1e67] for reg13873052 (PC) offset=0x150e2 in dl_main
  CU die offset: 0x29cd3
  found PC-rel by addr=0x34020 offset=0x20
  -----------------------------------------------------------
  find_data_type_die [2e] for reg12 offset=0 in __GI___readdir64
  CU die offset: 0x137a45
  frame base: cfa=1 fbreg=-1
  found "__futex" in scope=2/2 (die: 0x137ad5) 0(reg12) type=int (die:2a)
  -----------------------------------------------------------
  find_data_type_die [52] for reg5 offset=0 in __memmove_avx_unaligned_erms
  CU die offset: 0x1124ed
  no variable found
  Annotate type: 'struct rtld_global' in /usr/lib64/ld-linux-x86-64.so.2 (1 samples):
  ============================================================================
      samples     offset       size  field
            1          0       4336  struct rtld_global  {
            0          0          0      struct link_namespaces* _dl_ns;
            0       2560          8      size_t _dl_nns;
            0       2568         40      __rtld_lock_recursive_t _dl_load_lock {
            0       2568         40          pthread_mutex_t mutex {
            0       2568         40              struct __pthread_mutex_s __data {
            0       2568          4                  int __lock;
  $

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: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-9-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate: Add annotate_get_basic_blocks()

The annotate_get_basic_blocks() is to find a list of basic blocks from
the source instruction to the destination instruction in a function.

It'll be used to find variables in a scope.  Use BFS (Breadth First
Search) to find a shortest path to carry the variable/register state
minimally.

Also change find_disasm_line() to be used in annotate_get_basic_blocks()
and add 'allow_update' argument to control if it can update the IP.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-8-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf annotate-data: Introduce 'struct data_loc_info'

The find_data_type() needs many information to describe the location of
the data.  Add the new 'struct data_loc_info' to pass those information at
once.

No functional changes intended.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-7-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf map: Add map__objdump_2rip()

Sometimes we want to convert an address in objdump output to
map-relative address to match with a sample data.  Let's add
map__objdump_2rip() for that.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-6-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf dwarf-aux: Add die_find_func_rettype()

The die_find_func_rettype() is to find a debug entry for the given
function name and sets the type information of the return value.  By
convention, it'd return the pointer to the type die (should be the
same as the given mem_die argument) if found, or NULL otherwise.

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: Jiri Olsa <jolsa@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-5-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf dwarf-aux: Handle type transfer for memory access

We want to track type states as instructions are executed.  Each
instruction can access compound types like struct or union and load/
store its members to a different location.

The die_deref_ptr_type() is to find a type of memory access with a
pointer variable.  If it points to a compound type like struct, the
target memory is a member in the struct.  The access will happen with an
offset indicating which member it refers.  Let's follow the DWARF info
to figure out the type of the pointer target.

For example, say we have the following code.

  struct foo {
    int a;
    int b;
  };

  struct foo *p = malloc(sizeof(*p));
  p->b = 0;

The last pointer access should produce x86 asm like below:

  mov  0x0, 4(%rbx)

And we know %rbx register has a pointer to struct foo.  Then offset 4
should return the debug info of member 'b'.

Also variables of compound types can be accessed directly without a
pointer.  The die_get_member_type() is to handle a such case.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
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: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-4-namhyung@kernel.org
[ Check if die_get_real_type() returned NULL ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf dwarf-aux: Add die_collect_vars()

The die_collect_vars() is to find all variable information in the scope
including function parameters.  The struct die_var_type is to save the
type of the variable with the location (reg and offset) as well as where
it's defined in the code (addr).

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
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: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-3-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf dwarf-aux: Remove unused pc argument

It's not used, let's get rid of it.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Reviewed-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: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20240319055115.4063940-2-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cpumap: Use perf_cpu_map__for_each_cpu when possible

Rather than manually iterating the CPU map, use
perf_cpu_map__for_each_cpu(). When possible tidy local variables.

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-9-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf stat: Remove duplicate cpus_map_matched function

Use libperf's perf_cpu_map__equal() that performs the same function.

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-8-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf arm64 header: Remove unnecessary CPU map get and put

In both cases the CPU map is known owned by either the caller or a
PMU.

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-7-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cpumap: Clean up use of perf_cpu_map__has_any_cpu_or_is_empty

Most uses of what was perf_cpu_map__empty but is now
perf_cpu_map__has_any_cpu_or_is_empty want to do something with the
CPU map if it contains CPUs. Replace uses of
perf_cpu_map__has_any_cpu_or_is_empty with other helpers so that CPUs
within the map can be handled.

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-6-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt/intel-bts: Switch perf_cpu_map__has_any_cpu_or_is_empty use

Switch perf_cpu_map__has_any_cpu_or_is_empty() to
perf_cpu_map__is_any_cpu_or_is_empty() as a CPU map may contain CPUs as
well as the dummy event and perf_cpu_map__is_any_cpu_or_is_empty() is a
more correct alternative.

Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-5-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf arm-spe/cs-etm: Directly iterate CPU maps

Rather than iterate all CPUs and see if they are in CPU maps, directly
iterate the CPU map. Similarly make use of the intersect function
taking care for when "any" CPU is specified. Switch
perf_cpu_map__has_any_cpu_or_is_empty() to more appropriate
alternatives.

Reviewed-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-4-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf cpumap: Ensure empty cpumap is NULL from alloc

Potential corner cases could cause a cpumap to be allocated with size
0, but an empty cpumap should be represented as NULL. Add a path in
perf_cpu_map__alloc() to ensure this.

Suggested-by: James Clark <james.clark@arm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Closes: https://lore.kernel.org/lkml/2cd09e7c-eb88-6726-6169-647dcd0a8101@arm.com/
Link: https://lore.kernel.org/r/20240202234057.2085863-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

libperf cpumap: Add any, empty and min helpers

Additional helpers to better replace perf_cpu_map__has_any_cpu_or_is_empty().

Signed-off-by: Ian Rogers <irogers@google.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: André Almeida <andrealmeid@igalia.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Atish Patra <atishp@rivosinc.com>
Cc: Changbin Du <changbin.du@huawei.com>
Cc: Darren Hart <dvhart@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: K Prateek Nayak <kprateek.nayak@amd.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Paran Lee <p4ranlee@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Steinar H. Gunderson <sesse@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will@kernel.org>
Cc: Yang Jihong <yangjihong1@huawei.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Cc: Yanteng Si <siyanteng@loongson.cn>
Link: https://lore.kernel.org/r/20240202234057.2085863-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf build: Fix out of tree build related to installation of sysreg-defs

It seems that a previous modification to sysreg-defs, which corrected
emitting the header to the specified output directory, exposed missing
subdir, prefix variables.

This breaks out of tree builds of perf as the file is now built into the
output directory, but still tries to descend into output directory as a
subdir.

Fixes: a29ee6aea7030786 ("perf build: Ensure sysreg-defs Makefile respects output dir")
Reviewed-by: Oliver Upton <oliver.upton@linux.dev>
Reviewed-by: Tycho Andersen <tycho@tycho.pizza>
Signed-off-by: Ethan Adams <j.ethan.adams@gmail.com>
Tested-by: Tycho Andersen <tycho@tycho.pizza>
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: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240314222012.47193-1-j.ethan.adams@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf auxtrace: Fix multiple use of --itrace option

If the --itrace option is used more than once, the options are
combined, but "i" and "y" (sub-)options can be corrupted because
itrace_do_parse_synth_opts() incorrectly overwrites the period type and
period with default values.

For example, with:

--itrace=i0ns --itrace=e

The processing of "--itrace=e", resets the "i" period from 0 nanoseconds
to the default 100 microseconds.

Fix by performing the default setting of period type and period only if
"i" or "y" are present in the currently processed --itrace value.

Fixes: f6986c95af84ff2a ("perf session: Add instruction tracing options")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240315071334.3478-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script: Show also errors for --insn-trace option

The trace could be misleading if trace errors are not taken into
account, so display them also by adding the itrace "e" option.

Note --call-trace and --call-ret-trace already add the itrace "e"
option.

Fixes: b585ebdb5912cf14 ("perf script: Add --insn-trace for instruction decoding")
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240315071334.3478-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf docs arm_spe: Clarify more SPE requirements related to KPTI

The question of exactly when KPTI needs to be disabled comes up a lot
because it doesn't always need to be done. Add the relevant kernel
function and some examples that describe the behavior.

Also describe the interrupt requirement and that no error message will
be printed if this isn't met.

Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: James Clark <james.clark@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
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>
Link: https://lore.kernel.org/r/20240312132508.423320-1-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools headers: Remove almost unused copy of uapi/stat.h, add few conditional defines

These were used to build perf to provide defines not available in older
distros, but this was back in 2017, nowadays most the distros that are
supported and I have build containers for work using just the system
headers, so ditch them.

For the few that don't have STATX_MNT_ID{_UNIQUE}, or STATX_MNT_DIOALIGN
add them conditionally.

Some of these older distros may not have things that are used in 'perf
trace', but then they also don't have libtraceevent packages, so don't
build 'perf trace'.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240315204835.748716-6-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools headers: Remove now unused copies of uapi/{fcntl,openat2}.h and asm/fcntl.h

These were used to build perf to provide defines not available in older
distros, but this was back in 2017, nowadays all the distros that are
supported and I have build containers for work using just the system
headers, so ditch them.

Some of these older distros may not have things that are used in 'perf
trace', but then they also don't have libtraceevent packages, so don't
build 'perf trace'.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240315204835.748716-5-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Use the system linux/fcntl.h instead of a copy from the kernel

Builds ok all the way back to these older distros:

   1  almalinux:8    : Ok  gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-20) , clang version 16.0.6 (Red Hat 16.0.6-2.module_el8.9.0+3621+df7f7146) flex 2.6.1
   3  alpine:3.15    : Ok  gcc (Alpine 10.3.1_git20211027) 10.3.1 20211027 , Alpine clang version 12.0.1 flex 2.6.4
  15  debian:10      : Ok  gcc (Debian 8.3.0-6) 8.3.0 , Debian clang version 11.0.1-2~deb10u1 flex 2.6.4
  32  opensuse:15.4  : Ok  gcc (SUSE Linux) 7.5.0 , clang version 15.0.7 flex 2.6.4
  23  fedora:35      : Ok  gcc (GCC) 11.3.1 20220421 (Red Hat 11.3.1-3) , clang version 13.0.1 (Fedora 13.0.1-1.fc35) flex 2.6.4
  38  ubuntu:18.04   : Ok  gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0  flex 2.6.4

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240315204835.748716-4-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move prctl.h files (uapi/linux and x86's) copy out of the directory used to build perf

It is used only to generate string tables, not to build perf, so move it
to the tools/perf/trace/beauty/{include,arch}/ hierarchies, that is used
just for scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it, just <linux/usbdevice_fs.h> coming
from either 'make install_headers' or from the system /usr/include/
directory.

Suggested-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/20240315204835.748716-3-acme@kernel.org
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Stop using the copy of uapi/linux/prctl.h

Use the system one, nothing used in that file isn't available in the
supported, active distros.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
To: Ian Rogers <irogers@google.com>
Link: https://lore.kernel.org/lkml/20240315204835.748716-3-acme@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move arch/x86/include/asm/irq_vectors.h copy out of the directory used to build perf

It is used only to generate string tables, not to build perf, so move it
to the tools/perf/trace/beauty/include/ hierarchy, that is used just for
scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it.

Suggested-by: Ian Rogers <irogers@google.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move uapi/sound/asound.h copy out of the directory used to build perf

It is used only to generate string tables, not to build perf, so move it
to the tools/perf/trace/beauty/include/ hierarchy, that is used just for
scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

Suggested-by: Ian Rogers <irogers@google.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move uapi/linux/usbdevice_fs.h copy out of the directory used to build perf

It is mostly used only to generate string tables, not to build perf, so
move it to the tools/perf/trace/beauty/include/ hierarchy, that is used
just for scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it, just <linux/usbdevice_fs.h> coming
from either 'make install_headers' or from the system /usr/include/
directory.

Suggested-by: Ian Rogers <irogers@google.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move uapi/linux/mount.h copy out of the directory used to build perf

It is mostly used only to generate string tables, not to build perf, so
move it to the tools/perf/trace/beauty/include/ hierarchy, that is used
just for scraping.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it, just <linux/mount.h> coming from
either 'make install_headers' or from the system /usr/include/
directory.

Suggested-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Don't include uapi/linux/mount.h, use sys/mount.h instead

The tools/include/uapi/linux/mount.h file is mostly used for scrapping
defines into id->string tables, this is the only place were it is being
directly used, stop doing so.

Define MOUNT_ATTR_RELATIME and MOUNT_ATTR__ATIME if not available in the
system's headers.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Move uapi/linux/fs.h copy out of the directory used to build perf

It is mostly used only to generate string tables, not to build perf, so
move it to the tools/perf/trace/beauty/include/ hierarchy, that is used
just for scraping.

The only case where it was being used to build was in
tools/perf/trace/beauty/sync_file_range.c, because some older systems
doesn't have the SYNC_FILE_RANGE_WRITE_AND_WAIT define, just use the
system's linux/fs.h header instead, defining it if not available.

This is a something that should've have happened, as happened with the
linux/socket.h scrapper, do it now as Ian suggested while doing an
audit/refactor session in the headers used by perf.

No other tools/ living code uses it, just <linux/fs.h> coming from
either 'make install_headers' or from the system /usr/include/
directory.

Suggested-by: Ian Rogers <irogers@google.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/CAP-5=fWZVrpRufO4w-S4EcSi9STXcTAN2ERLwTSN7yrSSA-otQ@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty: Fix dependency of tables using uapi/linux/mount.h

Several such tables were depending on uapi/linux/fs.h, cut and paste
error when they were introduced, fix it.

Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/Ze9vjxv42PN_QGZv@x1
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf c2c: Fix a punctuation

s/dont/don\'t/

Signed-off-by: Bhaskar Chowdhury <unixbhaskar@gmail.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
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>
Link: https://lore.kernel.org/r/20210319232824.742-1-unixbhaskar@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Collect sys_nanosleep first argument

That is a 'struct timespec' passed from userspace to the kernel as we
can see with a system wide syscall tracing:

  root@number:~# perf trace -e nanosleep
       0.000 (10.102 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
      38.924 (10.077 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     100.177 (10.107 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     139.171 (10.063 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     200.603 (10.105 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     239.399 (10.064 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     300.994 (10.096 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     339.584 (10.067 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     401.335 (10.057 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     439.758 (10.166 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     501.814 (10.110 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     539.983 (10.227 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     602.284 (10.199 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     640.208 (10.105 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     702.662 (10.163 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     740.440 (10.107 ms): podman/2195174 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
     802.993 (10.159 ms): podman/9150 nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 10000000 })                   = 0
  ^Croot@number:~# strace -p 9150 -e nanosleep

If we then use the ptrace method to look at that podman process:

  root@number:~# strace -p 9150 -e nanosleep
  strace: Process 9150 attached
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  nanosleep({tv_sec=0, tv_nsec=10000000}, NULL) = 0
  ^Cstrace: Process 9150 detached
  root@number:~#

With some changes we can get something closer to the strace output,
still in system wide mode:

  root@number:~# perf config trace.show_arg_names=false
  root@number:~# perf config trace.show_duration=false
  root@number:~# perf config trace.show_timestamp=false
  root@number:~# perf config trace.show_zeros=true
  root@number:~# perf config trace.args_alignment=0
  root@number:~# perf trace -e nanosleep --max-events=10
  podman/2195174 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/9150 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/2195174 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/9150 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/2195174 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/9150 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/2195174 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/9150 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/2195174 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  podman/9150 nanosleep({ .tv_sec: 0, .tv_nsec: 10000000 }, NULL) = 0
  root@number:~#
  root@number:~# perf config
  trace.show_arg_names=false
  trace.show_duration=false
  trace.show_timestamp=false
  trace.show_zeros=true
  trace.args_alignment=0
  root@number:~# cat ~/.perfconfig
  # this file is auto-generated.
  [trace]
   show_arg_names = false
   show_duration = false
   show_timestamp = false
   show_zeros = true
   args_alignment = 0
  root@number:~#

This will not get reused by any other syscall as nanosleep is the only
one to have as its first argument a 'struct timespec" pointer argument
passed from userspace to the kernel:

  root@number:~# grep timespec /sys/kernel/tracing/events/syscalls/sys_enter_*/format | grep offset:16
  /sys/kernel/tracing/events/syscalls/sys_enter_nanosleep/format: field:struct __kernel_timespec * rqtp; offset:16; size:8; signed:0;
  root@number:~#

BTF based pretty printing will simplify all this, but then lets just get
the low hanging fruits first.

Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/lkml/Zbq72dJRpOlfRWnf@kernel.org/
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge tag 'v6.9-rc-smb3-server-fixes' of git://git.samba.org/ksmbd

Pull smb server updates from Steve French:

 - add support for durable file handles (an important data integrity
   feature)

 - fixes for potential out of bounds issues

 - fix possible null dereference in close

 - getattr fixes

 - trivial typo fix and minor cleanup

* tag 'v6.9-rc-smb3-server-fixes' of git://git.samba.org/ksmbd:
  ksmbd: remove module version
  ksmbd: fix potencial out-of-bounds when buffer offset is invalid
  ksmbd: fix slab-out-of-bounds in smb_strndup_from_utf16()
  ksmbd: Fix spelling mistake "connction" -> "connection"
  ksmbd: fix possible null-deref in smb_lazy_parent_lease_break_close
  ksmbd: add support for durable handles v1/v2
  ksmbd: mark SMB2_SESSION_EXPIRED to session when destroying previous session
  ksmbd: retrieve number of blocks using vfs_getattr in set_file_allocation_info
  ksmbd: replace generic_fillattr with vfs_getattr

Merge tag 'trace-tools-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace

Pull trace tool updates from Steven Rostedt:
 "Tracing:

   - Update makefiles for latency-collector and RTLA, using tools/build/
     makefiles like perf does, inheriting its benefits. For example,
     having a proper way to handle library dependencies.

   - The timerlat tracer has an interface for any tool to use. rtla
     timerlat tool uses this interface dispatching its own threads as
     workload. But, rtla timerlat could also be used for any other
     process. So, add 'rtla timerlat -U' option, allowing the timerlat
     tool to measure the latency of any task using the timerlat tracer
     interface.

  Verification:

   - Update makefiles for verification/rv, using tools/build/ makefiles
     like perf does, inheriting its benefits. For example, having a
     proper way to handle dependencies"

* tag 'trace-tools-v6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
  tools/rtla: Add -U/--user-load option to timerlat
  tools/verification: Use tools/build makefiles on rv
  tools/rtla: Use tools/build makefiles to build rtla
  tools/tracing: Use tools/build makefiles on latency-collector

Merge tag 'docs-6.9-2' of git://git.lwn.net/linux

Pull more documentation updates from Jonathan Corbet:
 "A handful of late-arriving documentation fixes and enhancements"

* tag 'docs-6.9-2' of git://git.lwn.net/linux:
  docs: verify/bisect: remove a level of indenting
  docs: verify/bisect: drop 'v' prefix, EOL aspect, and assorted fixes
  docs: verify/bisect: check taint flag
  docs: verify/bisect: improve install instructions
  docs: handling-regressions.rst: Update regzbot command fixed-by to fix
  docs: *-regressions.rst: Add colon to regzbot commands
  doc: Fix typo in admin-guide/cifs/introduction.rst
  README: Fix spelling

tools/rtla: Add -U/--user-load option to timerlat

The timerlat tracer provides an interface for any application to wait
for the timerlat's periodic wakeup. Currently, rtla timerlat uses it
to dispatch its user-space workload (-u option).

But as the tracer interface is generic, rtla timerlat can also be used
to monitor any workload that uses it. For example, a user might
place their own workload to wait on the tracer interface, and
monitor the results with rtla timerlat.

Add the -U option to rtla timerlat top and hist. With this option, rtla
timerlat will not dispatch its workload but only setting up the
system, waiting for a user to dispatch its workload.

The sample code in this patch is an example of python application
that loops in the timerlat tracer fd.

To use it, dispatch:

 # rtla timerlat -U

In a terminal, then run the python program on another terminal,
specifying the CPU to run it. For example, setting on CPU 1:

 #./timerlat_load.py 1

Then rtla timerlat will start printing the statistics of the
./timerlat_load.py app.

An interesting point is that the "Ret user Timer Latency" value
is the overall response time of the load. The sample load does
a memory copy to exemplify that.

The stop tracing options on rtla timerlat works in this setup
as well, including auto analysis.

Link: https://lkml.kernel.org/r/36e6bcf18fe15c7601048fd4c65aeb193c502cc8.1707229706.git.bristot@kernel.org
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>

tools/verification: Use tools/build makefiles on rv

Use tools/build/ makefiles to build rv, inheriting the benefits of
it. For example, having a proper way to handle dependencies.

Link: https://lkml.kernel.org/r/2a38a8f7b8dc65fa790381ec9ab42fb62beb2e25.1710519524.git.bristot@kernel.org
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@kernel.org>
Cc: John Kacur <jkacur@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>

tools/rtla: Use tools/build makefiles to build rtla

Use tools/build/ makefiles to build rtla, inheriting the benefits of
it. For example, having a proper way to handle dependencies.

rtla is built using perf infra-structure when building inside the
kernel tree.

At this point, rtla diverges from perf in two points: Documentation
and tarball generation/build.

At the documentation level, rtla is one step ahead, placing the
documentation at Documentation/tools/rtla/, using the same build
tools as kernel documentation. The idea is to move perf
documentation to the same scheme and then share the same makefiles.

rtla has a tarball target that the (old) RHEL8 uses. The tarball was
kept using a simple standalone makefile for compatibility. The
standalone makefile shares most of the code, e.g., flags, with
regular buildings.

The tarball method was set as deprecated. If necessary, we can make
a rtla tarball like perf, which includes the entire tools/build.
But this would also require changes in the user side (the directory
structure changes, and probably the deps to build the package).

Inspired on perf and objtool.

Link: https://lkml.kernel.org/r/57563abf2715d22515c0c54a87cff3849eca5d52.1710519524.git.bristot@kernel.org
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@kernel.org>
Cc: John Kacur <jkacur@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>

tools/tracing: Use tools/build makefiles on latency-collector

Use tools/build/ makefiles to build latency-collector, inheriting
the benefits of it. For example: Before this patch, a missing
tracefs/traceevents headers will result in fail like this:

  ~/linux/tools/tracing/latency $ make
  cc -Wall -Wextra -g -O2  -o latency-collector latency-collector.c -lpthread
  latency-collector.c:26:10: fatal error: tracefs.h: No such file or directory
     26 | #include <tracefs.h>
        |          ^~~~~~~~~~~
  compilation terminated.
  make: *** [Makefile:14: latency-collector] Error 1

Which is not that helpful. After this change it reports:

  ~/linux/tools/tracing/latency# make

  Auto-detecting system features:
  ...                           libtraceevent: [ OFF ]
  ...                              libtracefs: [ OFF ]

  libtraceevent is missing. Please install libtraceevent-dev/libtraceevent-devel
  libtracefs is missing. Please install libtracefs-dev/libtracefs-devel
  Makefile.config:29: *** Please, check the errors above..  Stop.

This type of output is common across other tools in tools/ like perf
and objtool.

Link: https://lkml.kernel.org/r/872420b0880b11304e4ba144a0086c6478c5b469.1710519524.git.bristot@kernel.org
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@kernel.org>
Cc: John Kacur <jkacur@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Daniel Bristot de Oliveira <bristot@kernel.org>

Merge tag 'bcachefs-2024-03-19' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs fixes from Kent Overstreet:
 "Assorted bugfixes.

  Most are fixes for simple assertion pops; the most significant fix is
  for a deadlock in recovery when we have to rewrite large numbers of
  btree nodes to fix errors. This was incorrectly running out of the
  same workqueue as the core interior btree update path - we now give it
  its own single threaded workqueue.

  This was visible to users as "bch2_btree_update_start(): error:
  BCH_ERR_journal_reclaim_would_deadlock" - and then recovery hanging"

* tag 'bcachefs-2024-03-19' of https://evilpiepirate.org/git/bcachefs:
  bcachefs: Fix lost wakeup on journal shutdown
  bcachefs; Fix deadlock in bch2_btree_update_start()
  bcachefs: ratelimit errors from async_btree_node_rewrite
  bcachefs: Run check_topology() first
  bcachefs: Improve bch2_fatal_error()
  bcachefs: Fix lost transaction restart error
  bcachefs: Don't corrupt journal keys gap buffer when dropping alloc info
  bcachefs: fix for building in userspace
  bcachefs: bch2_snapshot_is_ancestor() now safe to call in early recovery
  bcachefs: Fix nested transaction restart handling in bch2_bucket_gens_init()
  bcachefs: Improve sysfs internal/btree_updates
  bcachefs: Split out btree_node_rewrite_worker
  bcachefs: Fix locking in bch2_alloc_write_key()
  bcachefs: Avoid extent entry type assertions in .invalid()
  bcachefs: Fix spurious -BCH_ERR_transaction_restart_nested
  bcachefs: Fix check_key_has_snapshot() call
  bcachefs: Change "accounting overran journal reservation" to a warning

Merge tag 'soc-late-6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc

Pull more ARM SoC updates from Arnd Bergmann:
 "These are changes that for some reason ended up not making it into the
  first four branches but that should still make it into 6.9:

   - A rework of the omap clock support that touches both drivers and
     device tree files

   - The reset controller branch changes that had a dependency on late
     bugfixes. Merging them here avoids a backmerge of 6.8-rc5 into the
     drivers branch

   - The RISC-V/starfive, RISC-V/microchip and ARM/Broadcom devicetree
     changes that got delayed and needed some extra time in linux-next
     for wider testing"

* tag 'soc-late-6.9' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (31 commits)
  soc: fsl: dpio: fix kcalloc() argument order
  bus: ts-nbus: Improve error reporting
  bus: ts-nbus: Convert to atomic pwm API
  riscv: dts: starfive: jh7110: Add camera subsystem nodes
  ARM: bcm: stop selecing CONFIG_TICK_ONESHOT
  ARM: dts: omap3: Update clksel clocks to use reg instead of ti,bit-shift
  ARM: dts: am3: Update clksel clocks to use reg instead of ti,bit-shift
  clk: ti: Improve clksel clock bit parsing for reg property
  clk: ti: Handle possible address in the node name
  dt-bindings: pwm: opencores: Add compatible for StarFive JH8100
  dt-bindings: riscv: cpus: reg matches hart ID
  reset: Instantiate reset GPIO controller for shared reset-gpios
  reset: gpio: Add GPIO-based reset controller
  cpufreq: do not open-code of_phandle_args_equal()
  of: Add of_phandle_args_equal() helper
  reset: simple: add support for Sophgo SG2042
  dt-bindings: reset: sophgo: support SG2042
  riscv: dts: microchip: add specific compatible for mpfs pdma
  riscv: dts: microchip: add missing CAN bus clocks
  ARM: brcmstb: Add debug UART entry for 74165
  ...