From: Sami F. <saf...@sa...> - 2007-09-27 17:55:58
|
On Thu, Sep 27, 2007 at 19:24:57 +0200, Philippe Elie wrote: > On Sat, 22 Sep 2007 at 14:57 +0000, Sami Farin wrote: > > > is it supposed to work? > > my system freezes inside one second of starting oprofiled. > > 2.6.22.6 SMP on Pentium D. > > no kernel logs. I have to power cycle system.. > > I installed a 64 bits distro and can't reproduce that, stack trace in > userspace are weird due to -fomit-frame-pointer but that"s expected. This is kind of annoying to debug.. I try with one event and 100000 sample rate inside a couple of day and tell you how it went. > > with CALLGRAPH=0 it's a bit happier, but I got one Oops with it. > > I also (accidentally) sent email to LKML without Cc'ing this list. > > > > ~root/.oprofile/daemonrc: > > > > SESSION_DIR=/var/lib/oprofile > > CHOSEN_EVENTS_0=GLOBAL_POWER_EVENTS:45000:1:1:1 > > CHOSEN_EVENTS_1=BRANCH_RETIRED:45000:5:1:1 > > CHOSEN_EVENTS_2=BRANCH_RETIRED:45000:10:1:1 > > CHOSEN_EVENTS_3=FSB_DATA_ACTIVITY:45000:3:1:1 > > > CALLGRAPH=1 > > KERNEL_RANGE=ffffffff80200000,ffffffffff600c00 Should end range be "_end" in System.map? Now it's ffffffff808e05f8. But it crashes either way. > Your sampling rate seems high, do you freeze with only one events > sets GLOBAL_POWER_EVENTS:100000:1:1:1 ? > > > Oops with CALLGRAPH=0 > > > > 2007-09-22 13:53:32.527237777 <1>[ 3372.390188] Unable to handle kernel NULL pointer dereference at 0000000000000650 RIP: > > 2007-09-22 13:53:32.527245948 <1>[ 3372.390195] [<ffffffff80652f44>] _spin_lock+0x4/0x20 > > > 2007-09-22 13:53:32.527391990 <4>[ 3372.390484] [<ffffffff8831abd4>] :oprofile:sync_buffer+0xf4/0x480 > > look like a NULL task struct during a sync_buffer(). It's perhaps related > to your sampling rate too. Do you mean if sampling rate is "small" oprofile is supposed to crash? I don't hope so. my /var/lib/oprofile is tmpfs.. have you tried oprofile with tmpfs? > Send me privately your .config too please -- Do what you love because life is too short for anything else. |