Re: [PATCH v4 3.0-rc2-tip 20/22] 20: perf: perf interface for uprobes

[Josh Stone]

On 06/07/2011 06:02 AM, Srikar Dronamraju wrote:
> Enhances perf probe to user space executables and libraries.
> Provides very basic support for uprobes.

Hi Srikar,

This seems to have an issue with multiple active uprobes, whereas the v3
patchset handled this fine.  I haven't tracked down the exact code
difference yet, but here's an example transcript of what I'm seeing:

# perf probe -l
  probe_zsh:main       (on /bin/zsh:0x000000000000e3f0)
  probe_zsh:zalloc     (on /bin/zsh:0x0000000000051120)
  probe_zsh:zfree      (on /bin/zsh:0x0000000000051c70)
# perf stat -e probe_zsh:main zsh -c true

 Performance counter stats for 'zsh -c true':

                 1 probe_zsh:main

       0.029387785 seconds time elapsed

# perf stat -e probe_zsh:zalloc zsh -c true

 Performance counter stats for 'zsh -c true':

               605 probe_zsh:zalloc

       0.043836002 seconds time elapsed

# perf stat -e probe_zsh:zfree zsh -c true

 Performance counter stats for 'zsh -c true':

                36 probe_zsh:zfree

       0.029445890 seconds time elapsed

# perf stat -e probe_zsh:* zsh -c true

 Performance counter stats for 'zsh -c true':

                 0 probe_zsh:zalloc
                 1 probe_zsh:main
                 0 probe_zsh:zfree

       0.030912587 seconds time elapsed

# perf stat -e probe_zsh:z* zsh -c true

 Performance counter stats for 'zsh -c true':

               605 probe_zsh:zalloc
                 0 probe_zsh:zfree

       0.043774671 seconds time elapsed


It seems like among the selected probes, only one with the lowest offset
ever gets hit.  Any ideas?

Thanks,
Josh
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/