linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Adrian Hunter <adrian.hunter@intel.com>
To: Jiri Olsa <jolsa@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>,
	Peter Zijlstra <a.p.zijlstra@chello.nl>,
	linux-kernel@vger.kernel.org, David Ahern <dsahern@gmail.com>,
	Frederic Weisbecker <fweisbec@gmail.com>,
	Mike Galbraith <efault@gmx.de>, Namhyung Kim <namhyung@gmail.com>,
	Paul Mackerras <paulus@samba.org>,
	Stephane Eranian <eranian@google.com>
Subject: Re: [PATCH V5 3/9] perf tools: workaround objdump difficulties with kcore
Date: Wed, 09 Oct 2013 13:38:04 +0300	[thread overview]
Message-ID: <5255320C.9020009@intel.com> (raw)
In-Reply-To: <20131009101249.GC1005@krava.redhat.com>

On 09/10/13 13:12, Jiri Olsa wrote:
> On Wed, Oct 09, 2013 at 10:33:25AM +0300, Adrian Hunter wrote:
>> On 08/10/13 17:02, Jiri Olsa wrote:
>>> On Tue, Oct 08, 2013 at 11:45:50AM +0300, Adrian Hunter wrote:
>>>> objdump fails to annotate module symbols when looking
>>>> at kcore.  Workaround this by extracting object code
>>>> from kcore and putting it in a temporary file for
>>>> objdump to use instead.  The temporary file is created
>>>> to look like kcore but contains only the function
>>>> being disassembled.
>>>
>>> Excited to ses this one, but looks like I'm hitting some
>>> issue.  All annotation starts for me like this:
>>>
>>>                                                                                                        ▒
>>>        │              Disassembly of section load0:                                                            ▒
>>>        │                                                                                                       ▒
>>>        │              ffffffff815eee80 <load0>:                                                                ◆
>>>   9.33 │ffffffff815eee80:   data32 data32 data32 xchg %ax,%ax                                                  
>>>
>>>
>>> which does not seem right
>>
>> Can you tell me the commits of the kernel and perf tools you
>> were using, plus the commands and what symbol it was?
> 
> kernel: 3.9.10-100.fc17.x86_64
> perf:   latest acme's perf/core (06de626 perf evlist: Fix perf_evlist__mmap_read event overflow )
>         plus your V5 patches
> 
> commands:
>   sudo ./perf record -e cycles:k -a 
>   sudo ./perf report
> 
> ---
> Samples: 2K of event 'cycles:k', Event count (approx.): 445188286                                               
>  14.73%          swapper  [kernel.kallsyms]   [k] intel_idle                                                   ◆
>   3.19%                X  [kernel.kallsyms]   [k] smp_call_function_many                                       ▒
>   1.58%                X  [kernel.kallsyms]   [k] i915_gem_write_fence__ipi                                    ▒
>   1.58%          swapper  [kernel.kallsyms]   [k] iwl_trans_pcie_read32                                        ▒
> 
> 
> annotation of 1st 4 symbols:
> 
> ---
> intel_idle  /proc/kcore                                                                                         
>        │                                                                                                       ▒
>        │                                                                                                       ▒
>        │                                                                                                       ▒
>        │     Disassembly of section load0:                                                                     ▒
>        │                                                                                                       ▒
>        │     ffffffff8135f490 <load0>:                                                                         ▒
>   1.18 │       data32 data32 data32 xchg %ax,%ax                                                               ▒
> 
> 
> ---
> smp_call_function_many  /proc/kcore                                                                             
>        │                                                                                                       ◆
>        │                                                                                                       ▒
>        │                                                                                                       ▒
>        │     Disassembly of section load0:                                                                     ▒
>        │                                                                                                       ▒
>        │     ffffffff810bc270 <load0>:                                                                         ▒
>        │       data32 data32 data32 xchg %ax,%ax                                                               ▒
> 
> ---
> i915_gem_write_fence__ipi  /proc/kcore                                                                          
>        │
>        │
>        │
>        │    Disassembly of section load0:
>        │
>        │    ffffffffa0086630 <load0>:
>        │      data32 data32 data32 xchg %ax,%ax
> 
> ---
> iwl_trans_pcie_read32  /proc/kcore                                                                              
>        │
>        │
>        │
>        │    Disassembly of section load0:
>        │
>        │    ffffffffa0414a50 <load0>:
>        │      data32 data32 data32 xchg %ax,%ax
> 
> 
> the rest of the instruction decode differs.. just the first
> line is same for all
> 
> addresses seem ok:
> 
> [jolsa@krava perf]$ egrep 'ffffffff8135f490|ffffffff810bc270|ffffffffa0086630|ffffffffa0414a50' /proc/kallsyms 
> ffffffff810bc270 T smp_call_function_many
> ffffffff8135f490 t intel_idle
> ffffffffa0414a50 t iwl_trans_pcie_read32        [iwlwifi]
> ffffffffa0086630 t i915_gem_write_fence__ipi    [i915]
> 
> so.. the name of the section, name of the <function> plus the first
> instruction decode seem wrong.. I can see that in every symbol I
> annotate in the report and in annotate command as well.

If you use the --asm-raw option you can see the bytes:

	66 66 66 90 

That looks like a "nop" e.g. K8_NOP4 in arch/x86/include/asm/nops.h

	/*
	 * Define nops for use with alternative() and for tracing.
	 *
	 * *_NOP5_ATOMIC must be a single instruction.
	 */

	#define NOP_DS_PREFIX 0x3e

	/* generic versions from gas
	   1: nop
	   the following instructions are NOT nops in 64-bit mode,
	   for 64-bit mode use K8 or P6 nops instead
	   2: movl %esi,%esi
	   3: leal 0x00(%esi),%esi
	   4: leal 0x00(,%esi,1),%esi
	   6: leal 0x00000000(%esi),%esi
	   7: leal 0x00000000(,%esi,1),%esi
	*/
	#define GENERIC_NOP1 0x90
	#define GENERIC_NOP2 0x89,0xf6
	#define GENERIC_NOP3 0x8d,0x76,0x00
	#define GENERIC_NOP4 0x8d,0x74,0x26,0x00
	#define GENERIC_NOP5 GENERIC_NOP1,GENERIC_NOP4
	#define GENERIC_NOP6 0x8d,0xb6,0x00,0x00,0x00,0x00
	#define GENERIC_NOP7 0x8d,0xb4,0x26,0x00,0x00,0x00,0x00
	#define GENERIC_NOP8 GENERIC_NOP1,GENERIC_NOP7
	#define GENERIC_NOP5_ATOMIC NOP_DS_PREFIX,GENERIC_NOP4

	/* Opteron 64bit nops
	   1: nop
	   2: osp nop
	   3: osp osp nop
	   4: osp osp osp nop
	*/
	#define K8_NOP1 GENERIC_NOP1
	#define K8_NOP2 0x66,K8_NOP1
	#define K8_NOP3 0x66,K8_NOP2
	#define K8_NOP4 0x66,K8_NOP3
	#define K8_NOP5 K8_NOP3,K8_NOP2
	#define K8_NOP6 K8_NOP3,K8_NOP3
	#define K8_NOP7 K8_NOP4,K8_NOP3
	#define K8_NOP8 K8_NOP4,K8_NOP4
	#define K8_NOP5_ATOMIC 0x66,K8_NOP4

I think what you see is correct.


  reply	other threads:[~2013-10-09 10:39 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-10-08  8:45 [PATCH V5 0/9] perf tools: kcore improvements Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 1/9] perf tools: make a separate function to parse /proc/modules Adrian Hunter
2013-10-15  5:31   ` [tip:perf/core] perf symbols: Make a separate function to parse / proc/modules tip-bot for Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 2/9] perf tools: validate kcore module addresses Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 3/9] perf tools: workaround objdump difficulties with kcore Adrian Hunter
2013-10-08 14:02   ` Jiri Olsa
2013-10-09  7:33     ` Adrian Hunter
2013-10-09 10:12       ` Jiri Olsa
2013-10-09 10:38         ` Adrian Hunter [this message]
2013-10-09 12:16           ` Jiri Olsa
2013-10-09 12:43             ` Adrian Hunter
2013-10-08 15:56   ` Arnaldo Carvalho de Melo
2013-10-08  8:45 ` [PATCH V5 4/9] perf tools: add map__find_other_map_symbol() Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 5/9] perf tools: fix annotate_browser__callq() Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 6/9] perf tools: find kcore symbols on other maps Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 7/9] perf tools: add copyfile_mode() Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 8/9] perf buildid-cache: add ability to add kcore to the cache Adrian Hunter
2013-10-08  8:45 ` [PATCH V5 9/9] perf tools: add ability to find kcore in build-id cache Adrian Hunter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=5255320C.9020009@intel.com \
    --to=adrian.hunter@intel.com \
    --cc=a.p.zijlstra@chello.nl \
    --cc=acme@ghostprotocols.net \
    --cc=dsahern@gmail.com \
    --cc=efault@gmx.de \
    --cc=eranian@google.com \
    --cc=fweisbec@gmail.com \
    --cc=jolsa@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=namhyung@gmail.com \
    --cc=paulus@samba.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).