Skip to contentSkip to conversation

PEBS (in perf) stopped working from 3.6 -> 3.7

DISCUSSION Steinar H. Gunderson 2 participants 6 messages Jan 5, 2013 – Jan 7, 2013

lore
Share & tools
lore
#0 lore raw

[Please Cc me on any replies; I'm not subscribed to lkml]

Hi,

I recently upgraded from 3.6.5 to 3.7.1 to get around some MM issues that have been bothering me. However, it appears it broke PEBS:

  pannekake:/usr/src/linux-3.7.1# perf record -a -e cycles:pp

Error: sys_perf_event_open() syscall returned with 95 (Operation not supported). /bin/dmesg may provide additional information.

Fatal: No hardware sampling interrupt available. No APIC? If so then you can boot the kernel with the "lapic" boot parameter to force-enable it.

Non-precise tracing works fine. This used to work in 3.6.5, and I used make oldconfig (not that I can find any .config settings that would seem relevant either). I certainly have APIC. There's nothing in dmesg about the error. I'm using perf built from the same kernel tree (ie., 3.7.1).

This is on a dual Xeon E5520 (ie. Westmere), on a Supermicro board.

/* Steinar */

Signature · 2 lines
Chief, Tech:Server
http://www.sesse.net/
#1 lore raw
Quoted 4 lines · Steinar H. Gunderson

[Please Cc me on any replies; I'm not subscribed to lkml]

Hi,

Quoted 8 lines

I recently upgraded from 3.6.5 to 3.7.1 to get around some MM issues that have been bothering me. However, it appears it broke PEBS:

   pannekake:/usr/src/linux-3.7.1# perf record -a -e cycles:pp

Error: sys_perf_event_open() syscall returned with 95 (Operation not supported). /bin/dmesg may provide additional information.

Fatal: No hardware sampling interrupt available. No APIC? If so then you can boot the kernel with the "lapic" boot parameter to force-enable it.

Known problem. Pick one of: update perf to 3.7, add H to the command (-e cycles:ppH) or apply this patch: https://lkml.org/lkml/2012/12/28/384

David

#2 lore raw
Quoted 3 lines · David Ahern

Known problem. Pick one of: update perf to 3.7, add H to the command (-e cycles:ppH) or apply this patch:
https://lkml.org/lkml/2012/12/28/384

Oh, thinking of it, I've actually read about this flamew^Wdiscussion :-)

Thanks!

/* Steinar */

Signature · 1 line
Homepage: http://www.sesse.net/
#3 lore raw
In reply to David Ahern
Quoted 3 lines · David Ahern

Known problem. Pick one of: update perf to 3.7, add H to the command (-e cycles:ppH) or apply this patch:
https://lkml.org/lkml/2012/12/28/384

I spoke too soon. This works for cycles, but not for branch-misses:

pannekake:~> sudo perf record -e branch-misses:ppH -a

Error: sys_perf_event_open() syscall returned with 95 (Operation not supported) for event branch-misses:ppH. /bin/dmesg may provide additional information.

No hardware sampling interrupt available. No APIC? If so then you can boot the kernel with the "lapic" boot parameter to force-enable it.

Why would the two be different?

/* Steinar */

Signature · 1 line
Homepage: http://www.sesse.net/
#4 lore raw
Quoted 7 lines · Steinar H. Gunderson
Quoted 3 lines · David Ahern

Known problem. Pick one of: update perf to 3.7, add H to the command (-e cycles:ppH) or apply this patch:
https://lkml.org/lkml/2012/12/28/384

I spoke too soon. This works for cycles, but not for branch-misses:

Quoted 7 lines

pannekake:~> sudo perf record -e branch-misses:ppH -a

Error: sys_perf_event_open() syscall returned with 95 (Operation not supported) for event branch-misses:ppH. /bin/dmesg may provide additional information.

No hardware sampling interrupt available. No APIC? If so then you can boot the kernel with the "lapic" boot parameter to force-enable it.

Why would the two be different?

I will make a guess that is processor dependent. On an E5540 with 3.4.11-1.fc16.x86_64, 3.6.10-2.fc16.x86_64, and 3.8 I get the same failure message.

But on a E5620, it works fine with 3.4 and 3.7 (no intermediate kernels on that box).

David

#5 lore raw
Quoted 7 lines · David Ahern
Quoted 1 line

Why would the two be different?

I will make a guess that is processor dependent. On an E5540 with 3.4.11-1.fc16.x86_64, 3.6.10-2.fc16.x86_64, and 3.8 I get the same failure message.

But on a E5620, it works fine with 3.4 and 3.7 (no intermediate kernels on that box).

That sounds a bit odd; the CPU can certainly do it with 3.6.

/* Steinar */

Signature · 1 line
Homepage: http://www.sesse.net/