Skip to content

L
linux
Switch branch/tag
  1. 23 Mar, 2018 1 commit
  3. 21 Mar, 2018 10 commits
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Mark jumps to outher functions with the call arrow · 751b1783
      Arnaldo Carvalho de Melo authored 
      Things like this in _cpp_lex_token (gcc's cc1 program):

     cpp_named_operator2name@@Base+0xa72

Point to a place that is after the cpp_named_operator2name boundaries,
i.e.  in the ELF symbol table for cc1 cpp_named_operator2name is marked
as being 32-bytes long, but it in fact is much larger than that, so we
seem to need a symbols__find() routine that looks for >= current->start
and  < next_symbol->start, possibly just for C++ objects?

For now lets just make some progress by marking jumps to outside the
current function as call like.

Actual navigation will come next, with further understanding of how the
symbol searching and disassembly should be done.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-aiys0a0bsgm3e00hbi6fg7yy@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      751b1783
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Pass function descriptor to its instruction parsing routines · 85a84e4f
      Arnaldo Carvalho de Melo authored 
      We need that to figure out if jumps have targets in a different
function.

E.g. _cpp_lex_token(), in /usr/libexec/gcc/x86_64-redhat-linux/5.3.1/cc1
has a line like this:

  jne    c469be <cpp_named_operator2name@@Base+0xa72>

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-ris0ioziyp469pofpzix2atb@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      85a84e4f
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: No need to calculate notes->start twice · 425859ff
      Arnaldo Carvalho de Melo authored 
      Since we already set notes->start to map__rip_2objdump(map, sym->start)
in symbol__annotate2(), no need to calculate that address again in
symbol__calc_lines(), just use notes->start.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-ycxlg8mm5ueuj21w6gi62l7g@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      425859ff
    • Arnaldo Carvalho de Melo's avatar
      perf annotate browser: Add 'P' hotkey to dump annotation to file · d9bd7665
      Arnaldo Carvalho de Melo authored 
      Just like we have in the histograms browser used as the main screen for
'perf top --tui' and 'perf report --tui', to print the current
annotation to a file with a named composed by the symbol name and the
".annotation" suffix.

Here is one example of pressing 'A' on 'perf top' to live annotate a
kernel function and then press 'P' to dump that annotation, the
resulting file:

  # cat _raw_spin_lock_irqsave.annotation
  _raw_spin_lock_irqsave() /proc/kcore
  Event: cycles:ppp

    7.14        nop
   21.43        push   %rbx
    7.14        pushfq
                pop    %rax
                nop
                mov    %rax,%rbx
                cli
                nop
                xor    %eax,%eax
                mov    $0x1,%edx
   64.29        lock   cmpxchg %edx,(%rdi)
                test   %eax,%eax
              ↓ jne    2b
                mov    %rbx,%rax
                pop    %rbx
              ← retq
          2b:   mov    %eax,%esi
              → callq  queued_spin_lock_slowpath
                mov    %rbx,%rax
                pop    %rbx
              ← retq
  #

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-zzmnrwugb5vtk7bvg0rbx150@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      d9bd7665
    • Arnaldo Carvalho de Melo's avatar
      perf report: Introduce --ignore-vmlinux command line option · 91340c51
      Arnaldo Carvalho de Melo authored 
      We've had this in 'perf top' for quite a while, useful if one wishes
to force using /proc/kcore to do annotation using the patched kernel
instead of the ELF image it started from, aka vmlinux.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-ircpvox4wzsv7gasrpb28fw9@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      91340c51
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Introduce --ignore-vmlinux command line option · be316409
      Arnaldo Carvalho de Melo authored 
      This is already present in 'perf top', albeit undocumented (will fix),
and is useful to use /proc/kcore instead of vmlinux and then get what is
really in place, not what the kernel starts with, before alternatives,
ftrace .text patching, etc, see the differences:

  # perf annotate --stdio2 _raw_spin_lock_irqsave
  _raw_spin_lock_irqsave() /lib/modules/4.16.0-rc4/build/vmlinux
  Event: anon group { cycles, instructions }

    0.00   3.17      → callq  __fentry__
    0.00   7.94        push   %rbx
    7.69  36.51      → callq  __page_file_index
                       mov    %rax,%rbx
    7.69   3.17      → callq  *ffffffff82225cd0
                       xor    %eax,%eax
                       mov    $0x1,%edx
   80.77  49.21        lock   cmpxchg %edx,(%rdi)
                       test   %eax,%eax
                     ↓ jne    2b
    3.85   0.00        mov    %rbx,%rax
                       pop    %rbx
                     ← retq
                 2b:   mov    %eax,%esi
                     → callq  queued_spin_lock_slowpath
                       mov    %rbx,%rax
                       pop    %rbx
                     ← retq
  [root@jouet ~]# perf annotate --ignore-vmlinux --stdio2 _raw_spin_lock_irqsave
  _raw_spin_lock_irqsave() /proc/kcore
  Event: anon group { cycles, instructions }

    0.00   3.17        nop
    0.00   7.94        push   %rbx
    0.00  23.81        pushfq
    7.69  12.70        pop    %rax
                       nop
                       mov    %rax,%rbx
    7.69   3.17        cli
                       nop
                       xor    %eax,%eax
                       mov    $0x1,%edx
   80.77  49.21        lock   cmpxchg %edx,(%rdi)
                       test   %eax,%eax
                     ↓ jne    2b
    3.85   0.00        mov    %rbx,%rax
                       pop    %rbx
                     ← retq
                 2b:   mov    %eax,%esi
                     → callq  *ffffffff820e96b0
                       mov    %rbx,%rax
                       pop    %rbx
                     ← retq
  #

Diff of the output of those commands:

  # perf annotate --stdio2 _raw_spin_lock_irqsave > /tmp/vmlinux
  # perf annotate --ignore-vmlinux --stdio2 _raw_spin_lock_irqsave > /tmp/kcore
  # diff -y /tmp/vmlinux /tmp/kcore
  _raw_spin_lock_irqsave() vmlinux             | _raw_spin_lock_irqsave() /proc/kcore
  Event: anon group { cycles, instructions }     Event: anon group { cycles, instructions }

   0.00  3.17  → callq __fentry__              |  0.00  3.17     nop
   0.00  7.94    push  %rbx                       0.00  7.94     push  %rbx
   7.69 36.51  → callq __page_file_index       |  0.00 23.81     pushfq
                                               >  7.69 12.70     pop   %rax
                                               >                 nop
                 mov   %rax,%rbx                                 mov   %rax,%rbx
   7.69  3.17  → callq *ffffffff82225cd0       |  7.69  3.17     cli
                                               >                 nop
                 xor   %eax,%eax                                 xor   %eax,%eax
                 mov   $0x1,%edx                                 mov   $0x1,%edx
  80.77 49.21    lock  cmpxchg %edx,(%rdi)       80.77 49.21     lock  cmpxchg %edx,(%rdi)
                 test  %eax,%eax                                 test  %eax,%eax
               ↓ jne   2b                                      ↓ jne   2b
   3.85  0.00    mov   %rbx,%rax                  3.85  0.00     mov   %rbx,%rax
                 pop   %rbx                                      pop   %rbx
               ← retq                                          ← retq
            2b:  mov   %eax,%esi                            2b:  mov   %eax,%esi
               → callq queued_spin_lock_slowpath|              → callq *ffffffff820e96b0
                 mov   %rbx,%rax                                 mov   %rbx,%rax
                 pop   %rbx                                      pop   %rbx
               ← retq                                          ← retq
  #

This should be further streamlined by doing both annotations and
allowing the TUI to toggle initial/current, and show the patched
instructions in a slightly different color.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-wz8d269hxkcwaczr0r4rhyjg@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      be316409
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Add function header to --stdio2 · 864298f2
      Arnaldo Carvalho de Melo authored 
        # perf annotate --stdio2 _raw_spin_lock_irqsave
  _raw_spin_lock_irqsave() /lib/modules/4.16.0-rc4/build/vmlinux
  Event: anon group { cycles, instructions }

    0.00   3.17      → callq  __fentry__
    0.00   7.94        push   %rbx
    7.69  36.51      → callq  __page_file_index
                       mov    %rax,%rbx
    7.69   3.17      → callq  *ffffffff82225cd0
                       xor    %eax,%eax
                       mov    $0x1,%edx
   80.77  49.21        lock   cmpxchg %edx,(%rdi)
                       test   %eax,%eax
                     ↓ jne    2b
    3.85   0.00        mov    %rbx,%rax
                       pop    %rbx
                     ← retq
                 2b:   mov    %eax,%esi
                     → callq  queued_spin_lock_slowpath
                       mov    %rbx,%rax
                       pop    %rbx
                     ← retq
  #

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-i86yfyzl8m194ioxgj1jo32f@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      864298f2
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Use the default annotation options for --stdio2 · 35632892
      Arnaldo Carvalho de Melo authored 
      With an empty '[annotate]' section in ~/.perfconfig:

  # perf record -a --all-kernel -e '{cycles,instructions}:P' sleep 5
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 2.243 MB perf.data (5513 samples) ]
  # perf annotate --stdio2 _raw_spin_lock | head -20

                     Disassembly of section .text:

                     ffffffff81868790 <_raw_spin_lock>:
                     _raw_spin_lock():
                     EXPORT_SYMBOL(_raw_spin_trylock_bh);
                     #endif

                     #ifndef CONFIG_INLINE_SPIN_LOCK
                     void __lockfunc _raw_spin_lock(raw_spinlock_t *lock)
                     {
                     → callq  __fentry__
                     atomic_cmpxchg():
                             return xadd(&v->counter, -i);
                     }

                     static __always_inline int atomic_cmpxchg(atomic_t *v, int old, int new)
                     {
  # perf annotate --stdio2 _raw_spin_lock | head -20
                     → callq  __fentry__
                       xor    %eax,%eax
                       mov    $0x1,%edx
   87.50 100.00        lock   cmpxchg %edx,(%rdi)
    6.25   0.00        test   %eax,%eax
                     ↓ jne    16
    6.25   0.00        repz   retq
                 16:   mov    %eax,%esi
                     ↑ jmpq   ffffffff810e96b0 <queued_spin_lock_slowpath>
  #
  # cat ~/.perfconfig
  [annotate]

    hide_src_code = false
    show_linenr = true
  # perf annotate --stdio2 _raw_spin_lock | head -20

                 3   Disassembly of section .text:

                 5   ffffffff81868790 <_raw_spin_lock>:
                 6   _raw_spin_lock():
                 143 EXPORT_SYMBOL(_raw_spin_trylock_bh);
                 144 #endif

                 146 #ifndef CONFIG_INLINE_SPIN_LOCK
                 147 void __lockfunc _raw_spin_lock(raw_spinlock_t *lock)
                 148 {
                     → callq  __fentry__
                 150 atomic_cmpxchg():
                 187         return xadd(&v->counter, -i);
                 188 }

                 190 static __always_inline int atomic_cmpxchg(atomic_t *v, int old, int new)
                 191 {
  #
  # cat ~/.perfconfig
  [annotate]

    hide_src_code = true
    show_total_period = true
  # perf annotate --stdio2 _raw_spin_lock | head -20
                               → callq  __fentry__
                                 xor    %eax,%eax
                                 mov    $0x1,%edx
      1411316      152339        lock   cmpxchg %edx,(%rdi)
       344694           0        test   %eax,%eax
                               ↓ jne    16
        80806           0        repz   retq
                           16:   mov    %eax,%esi
                               ↑ jmpq   ffffffff810e96b0 <queued_spin_lock_slowpath>
  #

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-nu4rxg5zkdtgs1b2gc40p7v7@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      35632892
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Move the default annotate options to the library · 7f0b6fde
      Arnaldo Carvalho de Melo authored 
      One more thing that goes from the TUI code to be used more widely,
for instance it'll affect the default options used by:

  perf annotate --stdio2

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-0nsz0dm0akdbo30vgja2a10e@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      7f0b6fde
    • Arnaldo Carvalho de Melo's avatar
      perf annotate: Introduce the --stdio2 output mode · befd2a38
      Arnaldo Carvalho de Melo authored 
      This uses the TUI augmented formatting routines, modulo interactivity.

  # perf annotate --ignore-vmlinux --stdio2 _raw_spin_lock_irqsave
  _raw_spin_lock_irqsave() /proc/kcore
  Event: cycles:ppp

  Percent

              Disassembly of section load0:

              ffffffff9a8734b0 <load0>:
                nop
                push   %rbx
   50.00        pushfq
                pop    %rax
                nop
                mov    %rax,%rbx
                cli
                nop
                xor    %eax,%eax
                mov    $0x1,%edx
   50.00        lock   cmpxchg %edx,(%rdi)
                test   %eax,%eax
              ↓ jne    2b
                mov    %rbx,%rax
                pop    %rbx
              ← retq
          2b:   mov    %eax,%esi
              → callq  queued_spin_lock_slowpath
                mov    %rbx,%rax
                pop    %rbx
              ← retq
Tested-by: default avatarJin Yao <yao.jin@linux.intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-6cte5o8z84mbivbvqlg14uh1@git.kernel.orgSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
      befd2a38
  5. 20 Mar, 2018 29 commits
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7