提交 · e049d4a3fa194c8aa0d3ca29a9b11b32387ca6e3 · openanolis / cloud-kernel

27 2月, 2016 4 次提交

perf hists: Fix dynamic entry display in hierarchy · e049d4a3

由 Namhyung Kim 提交于 2月 27, 2016

When dynamic sort key is used it might not show pretty printed output.
This is because the trace output was not set only for the first dynamic
sort key.  During hierarchy_insert_entry() it missed to pass the
trace_output to dynamic entries.  Also even if it did, only first entry
will have it.  Subsequent entries might set it during collapsing stage
but it's not guaranteed.

Before:

  $ perf report --hierarchy --stdio -s ptr,bytes_req,gfp_flags -g none
  #
  #       Overhead  ptr / bytes_req / gfp_flags
  # ..............  ..........................................
  #
      37.50%        0xffff8803f7669400
         37.50%        448
            37.50%        66080
      10.42%        0xffff8803f766be00
          8.33%        96
             8.33%        66080
          2.08%        512
             2.08%        67280

After:

  #
  #       Overhead  ptr / bytes_req / gfp_flags
  # ..............  ..........................................
  #
      37.50%        0xffff8803f7669400
         37.50%        448
            37.50%        GFP_ATOMIC|GFP_NOWARN|GFP_NOMEMALLOC
      10.42%        0xffff8803f766be00
          8.33%        96
             8.33%        GFP_ATOMIC|GFP_NOWARN|GFP_NOMEMALLOC
          2.08%        512
             2.08%        GFP_KERNEL|GFP_NOWARN|GFP_REPEAT|GFP
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Tested-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: NJiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456512767-1164-4-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

e049d4a3

perf report: Left align dynamic entries in hierarchy · cb1fab91

由 Namhyung Kim 提交于 2月 27, 2016

The dynamic entries are right-aligned unlike other entries since it
usually has numeric value.  But for the hierarchy mode, left alignment
is more appropriate IMHO.  Also trim spaces on the left so that we can
easily identify the hierarchy.

Before:

  $ perf report --hierarchy -i perf.data.kmem -s gfp_flags,ptr,bytes_req --stdio -g none
  ...
  #
  #       Overhead                                        gfp_flags /                ptr /          bytes_req
  # ..............  .................................................................................................
  #
      91.67%                   GFP_ATOMIC|GFP_NOWARN|GFP_NOMEMALLOC
         37.50%        0xffff8803f7669400
            37.50%                       448
          8.33%        0xffff8803f766be00
             8.33%                        96
          4.17%        0xffff8800d156dc00
             4.17%                       704

After:

  #       Overhead  gfp_flags / ptr / bytes_req
  # ..............  ....................................
  #
      91.67%        GFP_ATOMIC|GFP_NOWARN|GFP_NOMEMALLOC
         37.50%        0xffff8803f7669400
            37.50%        448
          8.33%        0xffff8803f766be00
             8.33%        96
          4.17%        0xffff8800d156dc00
             4.17%        704
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Tested-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: NJiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456512767-1164-3-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

cb1fab91

perf report: Fix indentation of dynamic entries in hierarchy · d3a72fd8

由 Namhyung Kim 提交于 2月 27, 2016

When dynamic entries are used in the hierarchy mode with multiple
events, the output might not be aligned properly.  In the hierarchy
mode, the each sort column is indented using total number of sort keys.
So it keeps track of number of sort keys when adding them.  However
a dynamic sort key can be added more than once when multiple events have
same field names.  This results in unnecessarily long indentation in the
output.

For example perf kmem records following events:

  $ perf evlist --trace-fields -i perf.data.kmem
  kmem:kmalloc: trace_fields: call_site,ptr,bytes_req,bytes_alloc,gfp_flags
  kmem:kmalloc_node: trace_fields: call_site,ptr,bytes_req,bytes_alloc,gfp_flags,node
  kmem:kfree: trace_fields: call_site,ptr
  kmem:kmem_cache_alloc: trace_fields: call_site,ptr,bytes_req,bytes_alloc,gfp_flags
  kmem:kmem_cache_alloc_node: trace_fields: call_site,ptr,bytes_req,bytes_alloc,gfp_flags,node
  kmem:kmem_cache_free: trace_fields: call_site,ptr
  kmem:mm_page_alloc: trace_fields: page,order,gfp_flags,migratetype
  kmem:mm_page_free: trace_fields: page,order

As you can see, many field names shared between kmem events.  So adding
'ptr' dynamic sort key alone will set nr_sort_keys to 6.  And this adds
many unnecessary spaces between columns.

Before:

  $ perf report -i perf.data.kmem --hierarchy -s ptr -g none --stdio
  ...
  #                Overhead                 ptr
  # .......................  ...................................
  #
      99.89%                 0xffff8803ffb79720
       0.06%                 0xffff8803d228a000
       0.03%                 0xffff8803f7678f00
       0.00%                 0xffff880401dc5280
       0.00%                 0xffff880406172380
       0.00%                 0xffff8803ffac3a00
       0.00%                 0xffff8803ffac1600

After:

  # Overhead                 ptr
  # ........  ....................
  #
      99.89%  0xffff8803ffb79720
       0.06%  0xffff8803d228a000
       0.03%  0xffff8803f7678f00
       0.00%  0xffff880401dc5280
       0.00%  0xffff880406172380
       0.00%  0xffff8803ffac3a00
       0.00%  0xffff8803ffac1600
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Tested-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: NJiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456512767-1164-2-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

d3a72fd8

perf hists: Fix comparing of dynamic entries · 84b6ee8e

由 Namhyung Kim 提交于 2月 27, 2016

When hist_entry__cmp() and hist_entry__collapse() are called, they
should check if the dynamic entry is comparing matching hists only.

Otherwise it might access different hists resulting in incorrect output.
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Acked-by: NJiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456512767-1164-1-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

84b6ee8e

26 2月, 2016 6 次提交

perf report: Show message for percent limit on gtk · 2ddda792

由 Namhyung Kim 提交于 2月 26, 2016

Like the stdio, it should show messages about omitted hierarchy
entries.  Please refer the previous commit for more details.
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456488800-28124-5-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

2ddda792

perf hists browser: Show message for percent limit · 79dded87

由 Namhyung Kim 提交于 2月 26, 2016

Like the stdio, it should show messages about omitted hierarchy entries.
Please refer the previous commit for more details.

As it needs to check an entry is omitted or not multiple times, add the
has_no_entry field in the hist entry.
Suggested-and-Tested-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456488800-28124-4-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

79dded87

perf hists browser: Cleanup hist_browser__update_percent_limit() · 201fde73

由 Namhyung Kim 提交于 2月 26, 2016

The previous patch introduced __rb_hierarchy_next() function with
various move direction like HMD_FORCE_CHILD but missed to change using
it some place.
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456488800-28124-3-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

201fde73

perf report: Show message for percent limit on stdio · bd4abd39

由 Namhyung Kim 提交于 2月 26, 2016

When the hierarchy mode is used, some entries might be omiited due to a
percent limit or filter.  In this case the output hierarchy is different
than other entries.  Add an informative message to users about this.

For example, when 4% of percent limit is applied:

Before:
  #       Overhead  Command / Shared Object / Symbol
  # ..............  ..........................................
  #
      49.09%        swapper
         48.67%        [kernel.vmlinux]
            34.42%        [k] intel_idle
      11.51%        firefox
          8.87%        libpthread-2.22.so
             6.60%        [.] __GI___libc_recvmsg
      10.49%        gnome-shell
          4.74%        libc-2.22.so
      10.08%        Xorg
          6.11%        libc-2.22.so
             5.27%        [.] __memcpy_sse2_unaligned
       6.15%        perf

Note that, gnome-shell/libc has no symbols and perf has no dso/symbols.
With that patch the output will look like below:

After:

  #       Overhead  Command / Shared Object / Symbol
  # ..............  ..........................................
  #
      49.09%        swapper
         48.67%        [kernel.vmlinux]
            34.42%        [k] intel_idle
      11.51%        firefox
          8.87%        libpthread-2.22.so
             6.60%        [.] __GI___libc_recvmsg
      10.49%        gnome-shell
          4.74%        libc-2.22.so
                          no entry >= 4.00%
      10.08%        Xorg
          6.11%        libc-2.22.so
             5.27%        [.] __memcpy_sse2_unaligned
       6.15%        perf
                       no entry >= 4.00%
Suggested-and-Tested-by: NArnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456488800-28124-2-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

bd4abd39

perf hists: Add more helper functions for the hierarchy mode · a7b5895b

由 Namhyung Kim 提交于 2月 26, 2016

The hists__overhead_width() is to calculate width occupied by the
overhead (and others) columns before the sort columns.

The hist_entry__has_hiearchy_children() is to check whether an entry has
lower entries (children) in the hierarchy to be shown in the output.
This means the children should not be filtered out and above the percent
limit.

These two functions will be used to show information when all children
of an entry is omitted by the percent limit (or filter).
Signed-off-by: NNamhyung Kim <namhyung@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456488800-28124-1-git-send-email-namhyung@kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

a7b5895b

perf script: Remove duplicated code and needless script_spec__findnew() · 8560bae0

由 Taeung Song 提交于 2月 26, 2016

script_spec_register() called two functions: script_spec__find() and
script_spec__findnew(). But this way script_spec__find() gets called
two times, directly and via script_spec__findnew().

So remove script_spec__findnew() and make script_spec_register() only
call once script_spec__find().
Signed-off-by: NTaeung Song <treeze.taeung@gmail.com>
Acked-by: NJiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1456413190-12378-1-git-send-email-treeze.taeung@gmail.comSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

8560bae0

25 2月, 2016 23 次提交

perf script: Exception handling when the print fmt is empty · 8579aca3

由 Taeung Song 提交于 2月 26, 2016

After collecting samples for events 'syscalls:', perf-script with python
script doesn't occasionally work generating a segmentation fault.

The reason is that the print fmt is empty and a value of
event->print_fmt.args is NULL, so dereferencing the null pointer results
in a segmentation fault i.e.:

    # perf record -e syscalls:*
    # perf script -g python
    # perf script -s perf-script.py

    in trace_begin
    syscalls__sys_enter_brk  3 79841.832099154  3777 test.sh  syscall_nr=12, brk=0

    ... (omitted) ...

    Segmentation fault (core dumped)

For example, a format of sys_enter_getuid() hasn't
print fmt as below.

    # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_getuid/format
    name: sys_enter_getuid
    ID: 188
    format:
            field:unsigned short common_type;         offset:0; size:2; signed:0;
            field:unsigned char common_flags;         offset:2; size:1; signed:0;
            field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
            field:int common_pid;                     offset:4; size:4; signed:1;
            field:int syscall_nr;                     offset:8; size:4; signed:1;

    print fmt: ""

So add exception handling to avoid this problem.
Signed-off-by: NTaeung Song <treeze.taeung@gmail.com>
Tested-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1456413179-12331-1-git-send-email-treeze.taeung@gmail.comSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

8579aca3

perf tools: Fix parsing of pmu events with empty list of modifiers · bb109acc

由 Arnaldo Carvalho de Melo 提交于 2月 25, 2016

In 1d55e8ef ("perf tools: Introduce opt_event_config nonterminal") I
removed the unconditional "'/' '/'" for pmu events such as
"intel_pt//" but forgot to use opt_event_config where it expected some
event_config, oops. Fix it.

Noticed when trying to use:

  # perf record -e intel_pt// -a sleep 1
  event syntax error: 'intel_pt//'
                               \___ parser error
  Run 'perf list' for a list of valid events

   Usage: perf record [<options>] [<command>]
      or: perf record [<options>] -- <command> [<options>]

      -e, --event <event>   event selector. use 'perf list' to list available events
  #

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: 1d55e8ef ("perf tools: Introduce opt_event_config nonterminal")
Signed-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

bb109acc

perf jvmti: improve error message in Makefile · 7e9551bc

由 Stephane Eranian 提交于 2月 25, 2016

This patch improves the error message given by jvmti Makefile when the
alternatives command cannot be found. It now suggests the user locates
the root of their Java installation and pass it with JDIR=
Signed-off-by: NStephane Eranian <eranian@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1456378056-18812-1-git-send-email-eranian@google.comSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

7e9551bc

perf tools: Use asprintf() for simple string formatting/allocation · 5104ffb2

由 Arnaldo Carvalho de Melo 提交于 2月 25, 2016

No need to use strbuf there, its just a simple alloc+formatting, which
asprintf does just fine.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/n/tip-6q6cxfhk8c8ypg3tfpo0i2iy@git.kernel.orgSigned-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>

5104ffb2

Merge tag 'perf-core-for-mingo-20160224' of... · 06466212

由 Ingo Molnar 提交于 2月 25, 2016

Merge tag 'perf-core-for-mingo-20160224' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements from Arnaldo Carvalho de Melo:

User visible changes:

  - Hierarchy histogram mode for 'perf top' and 'perf report', showing multiple
    levels, one per --sort entry: (Namhyung Kim)

    On a mostly idle system:

    # perf top --hierarchy -s comm,dso

    Then expand some levels and use 'P' to take a snapshot:

    # cat perf.hist.0
    -  92.32%         perf
          58.20%         perf
          22.29%         libc-2.22.so
           5.97%         [kernel]
           4.18%         libelf-0.165.so
           1.69%         [unknown]
    -   4.71%         qemu-system-x86
           3.10%         [kernel]
           1.60%         qemu-system-x86_64 (deleted)
    +   2.97%         swapper
    #

  - Check availability of memory events in 'perf mem': (Jiri Olsa)

    On a Intel Broadwell machine:

    # perf mem record -e list
    ldlat-loads : available
    ldlat-stores: available
    #

  - Decode data_src values (e.g. perf.data files generated by 'perf mem record')
    in 'perf script': (Jiri Olsa)

    # perf script
      perf 693 [1] 4.088652: 1 cpu/mem-loads,ldlat=30/P: ffff88007d0b0f40 68100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK No <SNIP>
                                                                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  - Print bpf-output events in 'perf script': (Wang Nan).

    # perf record -e bpf-output/no-inherit,name=evt/ -e ./test_bpf_output_3.c/map:channel.event=evt/ usleep 1000
    # perf script
       usleep  4882 21384.532523:   evt:  ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
        BPF output: 0000: 52 61 69 73 65 20 61 20  Raise a
                    0008: 42 50 46 20 65 76 65 6e  BPF even
                    0010: 74 21 00 00              t!..
        BPF string: "Raise a BPF event!"
    #
Signed-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NIngo Molnar <mingo@kernel.org>

06466212

perf top: Add --hierarchy option · c92fcfde