Hello,
I've found a problem in perfmon patch in kernel 2.6.18 - the short
specification is that with "--follow-all", perfmon crashes when during
measuring, another application is run. I can send more info (the kernel
messages are below), but from reading the changelogs, I guess that this
problem has already been fixed in newer versions of perfmon.
However, the question is, can I use the newest perfmon version (I guess
for 2.6.21) with 2.6.18 ? Or, is there an easy way to backport only this
particular fix ?
Thanks,
Tomas
kernel BUG at perfmon/perfmon_ctxsw.c:99!
invalid opcode: 0000 [#1]
SMP
Modules linked in: ipv6 button ac battery dm_snapshot dm_mirror dm_mod loop
parnCPU: 0
EIP: 0060:[<c01bdc7a>] Not tainted VLI
EFLAGS: 00010046 (2.6.18.8vanillapfmon #1)
EIP is at __pfm_ctxsw+0x154/0x3d6
eax: dfb14aa0 ebx: dfe08710 ecx: 000001d2 edx: 0000004f
esi: 00000000 edi: dd826000 ebp: dfe08550 esp: dd1dfe1c
ds: 007b es: 007b ss: 0068
Process cc1 (pid: 6612, ti=dd1de000 task=dfe08550 task.ti=dd1de000)
Stack: d5664819 dfb14aa0 d567dde2 0000004f 00000000 0000000f dd1de000
00000000
dfe08710 dfe08550 c14019e0 dfb14aa0 c0102e95 00000000 dfe08710
dfb14aa0
00000000 c0103025 dfe08550 dfb14c60 c14019e0 dfb14aa0 dfb36e40
dfb14aa0
Call Trace:
[<c0102e95>] __switch_to_xtra+0x61/0x91
[<c0103025>] __switch_to+0x160/0x16b
[<c0289de2>] schedule+0x84e/0x8fe
[<c0122aac>] __do_softirq+0x5a/0xbb
[<c01461c2>] free_pages_bulk+0x1b/0x1eb
[<c0119d70>] __cond_resched+0x16/0x34
[<c028a400>] cond_resched+0x2a/0x31
[<c014cb8b>] unmap_vmas+0x3a8/0x4af
[<c014f0a1>] exit_mmap+0x6a/0xd7
[<c011c48b>] mmput+0x20/0x76
[<c0120c79>] do_exit+0x193/0x71b
[<c011672a>] do_page_fault+0x18a/0x481
[<c0121277>] sys_exit_group+0x0/0xd
[<c0103cf9>] sysenter_past_esp+0x56/0x79
Code: 81 e2 00 e0 ff ff 8b 52 10 03 04 95 c0 de 32 c0 c7 00 00 00 00 00
b0 01 8
EIP: [<c01bdc7a>] __pfm_ctxsw+0x154/0x3d6 SS:ESP 0068:dd1dfe1c
_______________________________________________
perfmon mailing list
[email protected]
http://www.hpl.hp.com/hosted/linux/mail-archives/perfmon/
