Page 1 of 1

kernel BUG at mm/filemap.c:127

PostPosted: Sun Sep 28, 2008 3:24 am
by Xaid
Hi,

I have been encountering this kernel oops for a bit now , sometimes on random occasions and sometimes I can reproduce the issue easily. I am running Gentoo Hardened and I can reproduce the issue usually by doing the following steps:

1) startx which runs e17.
2) re-emerge all the e17 libraries.
3) restart enlightenment (without exiting X).

I have also noticed it happening on random occasions after heavy IO. I pasted the dmesg output at http://pastebin.ca/1214299
My kernel config can be found at http://pastebin.ca/1214305

Thanks,
Xaid

Update1: I went and looked at my previous logs and it seems that this issue started happening at around 2.6.23. The dmesg log can be found at http://pastebin.ca/1214303. The log from kernel version 2.6.24 can be found at http://pastebin.ca/1214304

Update2: Changed links to use pastebin.ca.

Re: kernel BUG at mm/filemap.c:127

PostPosted: Tue Sep 30, 2008 8:05 pm
by Xaid
It looks like I'm the only one having this issue :)

Any ideas on what the problem might be? I haven't tried with a vanilla kernel yet but I might do that on the weekend and report back my findings.

Re: kernel BUG at mm/filemap.c:127

PostPosted: Wed Oct 01, 2008 7:02 am
by Oscon
Xaid wrote:It looks like I'm the only one having this issue :)

after heavy IO

Any ideas on what the problem might be? I haven't tried with a vanilla kernel yet but I might do that on the weekend and report back my findings.


Only ideas: :-) : CONFIG_NO_HZ not set, IRQBALANCE=n, CPUIDLE=n, and anticipatory IO scheduler, PCI Acces mode=direct,preemption model=server, and agp=n, use builtin agp support of nvidia binary driver (NvAgp=1), and check your ext* filesystem with fsck.

But, there are only ideas. :-)

Re: kernel BUG at mm/filemap.c:127

PostPosted: Thu Oct 02, 2008 7:08 pm
by PaX Team
Xaid wrote:It looks like I'm the only one having this issue :)

Any ideas on what the problem might be? I haven't tried with a vanilla kernel yet but I might do that on the weekend and report back my findings.
my guess is that if anything then it's vma mirroring related, but my attempts at reproducing it based on what i could gather from your logs didn't succeed, so i'll need a more precise way to reproduce this... also just in case, can you reproduce it with the latest PaX test patch itself?