Re: oops, 2.6.15.4 + qla1280+md+xfs+quota (fwd)

[Chris Stromsoe]

On Fri, 17 Feb 2006, Nathan Scott wrote:
> On Thu, Feb 16, 2006 at 10:30:34AM -0800, Chris Stromsoe wrote:
> > I sent this to lkml yesterday.  Short description of the problem -- xfs 
> > on 2.6.15.4 without quota enabled is ok.  When I try to mount with 
> > "quota", the entire machine hangs.  Details are below.
>
> Could you try without 4Kstacks just for grins?  Thanks. There was a 
> stack abuse accidentally introduced in the XFS quota mount path a month 
> or two back - I have a fix for that pending so that may just be it.

I rebuilt with the same config, but without 4K stacks.  I have the xfs 
partition mounted with quota enabled, and everything looks ok.  I don't 
have any content on the partition yet, but this is further than I got 
yesterday.  If you have patches to fix the 4K/8K issue, I'm willing to 
test them to see if they do fix the problem with 4K stacks.

> EIP is at a whacky spot, and your stack trace is all messed up - not 
> sure how we'd get into that situation on a mount with quotas, unless 
> something really bizzare happened (like a stack overflow).

All of the stack traces I got back while testing were different, and none 
made much sense.


-Chris

> cheers.
>
> > ---------- Forwarded message ----------
> > Date: Thu, 16 Feb 2006 03:10:00 -0800 (PST)
> > From: Chris Stromsoe <cbs@xxxxxxxxxxxx>
> > To: linux-kernel@xxxxxxxxxxxxxxx
> > Subject: oops, 2.6.15.4 + qla1280+md+xfs+quota
> >
> > I have a 2.6.15.4 system that was running fine.  Internal drives are using the
> > Fusion MPT driver.  I have two external JBODs with 12 disks each. Each JBOD has
> > two channels, 6 disks per channel, and each channel is connected to a QLogic
> > ISP 10160 controller.
> >
> > Each of the JBODs is built as an md raid5 (md1 and md2).  Both raid5s are
> > mirrored (md3).  I had an ext3 filesystem running fine.
> >
> > I unmounted md3, ran "mkfs.xfs /dev/md3 && mount /dev/md3" and the system hung.
> > md is built in, qla1280 and xfs are both modules.  mount options in fstab are
> > rw, quota, noatime -- possibly also noexec and nosuid, but I don't think they
> > were enabled.  XFS is built with XFS quota support; I've tried also with and
> > without generic quota support.  4K stacks are enabled, if that matters.  Full
> > .config and dmesg output are at http://hashbrown.cts.ucla.edu/pub/oops-200602/
> > (config is with generic quota, dmesg has xfs loaded as a module and /dev/md3
> > mounted without quota set).
> >
> > I added "defaults" and removed "quota" from fstab for /dev/md3 and was able to
> > mount the partition normally.
> >
> > It's not responsive to sysrq on the serial console. The last output (without
> > generic quota) was:
> >
> > XFS mounting filesystem md3
> > XFS quotacheck md3: Please wait.
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > c0114f5a
> > *pde = 00000000
> > Oops: 0000 [#1]
> > SMP
> > Modules linked in: xfs e1000 bonding qla1280 st
> > CPU:    0
> > EIP:    0060:[<c0114f5a>]    Not tainted VLI
> > EFLAGS: 00010006   (2.6.15.4)
> > EIP is at do_page_fault+0x93/0x545
> > eax: eb7ac000   ebx: 00000001   ecx: 0000007b   edx: 00100100
> > esi: 000001ff   edi: c0114ec7   ebp: eb7ac0e8   esp: eb7ac094
> > ds: 007b   es: 007b   ss: 0068
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel NULL pointer dereference at virtual address 00000078
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address bebaee76
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address bebaee76
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> > Unable to handle kernel paging request at virtual address 00100178
> >   printing eip:
> >
> >
> > As soon as I mount the partition with quota in fstab, the machine hangs. The
> > oops is not always the same.  During one test run, I saw a number of "BUG: soft
> > lockup detected on CPU#0" errors.  The above oops is the only one I managed to
> > capture.
> >
> >
> >
> > -Chris
> > -
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at  http://www.tux.org/lkml/
>
> --
> Nathan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/