Re: perf_evlist__filter_pollfd() in trace__run()

[Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>]

Arnaldo Carvalho de Melo [acme@kernel.org] wrote:
| Em Thu, Jan 29, 2015 at 03:55:22PM -0800, Sukadev Bhattiprolu escreveu:
| > Should there be another check to before reading the mmap again ?
| 
| Possibly, checking, but a similar algorithm should be in place for
| 'record', do you see any problems there? I.e. with 'perf record sleep
| 1'?

No, I don't see the failure with perf record in several hundred attempts.
(both the distro and latest perf-core versions work ok).

| 
| - Arnaldo
| 
| > I must add that I don't get the SIGSEGV on recent perf-core and the
| > system where we get the crash, first runs into the following
| > errors that we are still looking into (maybe related to "ppc64le"
| > architecture).

With recent perf core I did not get the SIGSEGV (or the syscall translation
failures below) on a different system.

But on this system, a power KVM guest, I get both errors even with latest 
perf core (commit c52686f9f88).

| > 
| > 	Problems reading syscall 45 information
| > 	Problems reading syscall 5 information
| > 	Problems reading syscall 5 information
| > 	Problems reading syscall 108 information
| > 	Problems reading syscall 108 information
| > 	Problems reading syscall 90 information
| > 	Problems reading syscall 90 information
| > 	Problems reading syscall 6 information
| > 
| > Unlike the SIGSEGV, these errors occur always.

Most likely this syscall number translation is due to the

https://fedorahosted.org/audit/browser/trunk/lib/machinetab.h 

missing this change set 1013, which added support for ppc64le.

Sukadev