kdump not writing any vmcore

[Neil Horman]

On Fri, Feb 12, 2010 at 03:52:01PM +0100, Gallus wrote:
> PS. I can add some overall information about the box:
> 
> a) Modules loaded:
> kernel: Modules linked in: dm_mirror dm_multipath dm_mod video sbs
> backlight i2c_ec i2c_core button battery asus_acpi ac parport_pc lp
> parport serio_raw ide_cd bnx2 cdrom pcspkr cciss sd_mod scsi_mod ext3
> jbd uhci_hcd ohci_hcd ehci_hcd
> 
> b) Some Oops that occurred (I'm not sure whether it was in crash
> kernel or in normal one):
> kernel: list_add corruption. prev->next should be e70ee78c, but was 5474646e
> kernel: BUG: unable to handle kernel NULL pointer dereference at
> virtual address 00000084 kernel:  printing eip:
> kernel: c06097b2
> kernel: *pde = 1cf15001
> Oops: 0000 [#1]
> kernel: SMP
> kernel: last sysfs file: /devices/pci0000:00/0000:00:00.0/class
> kernel: CPU:    0
> kernel: EIP:    0060:[<c06097b2>]    Tainted: G      VLI
> kernel: EFLAGS: 00210046   (2.6.18-92.el5PAE #1)
> kernel: EIP is at do_page_fault+0x1d1/0x5d3
> kernel: eax: d2ef6000   ebx: 00000000   ecx: d2ef6064   edx: 0000000d
> kernel: esi: 00000000   edi: d2ef6094   ebp: 00000000   esp: d2ef6044
> 
> c) hardware:
> Quad-Core AMD Opteron(tm) Processor 8356, model 2, stepping 3
> 
> Gallus
> 

This helps not one bit.  Theres not enough info here to determine what the cause
(or even the location) of this corruption is.  All it tells us is that a list in
the kernel was corrupted.  If that happened in the production kernel it would
cause kdump to start.  If it happened in the kdump kernel, it would explain why
you never got a vmcore.  But without the call stack, I can't even tell you what
list this might have been.  If we knew that, we might be able to dredge bugzilla
to search for simmilar bugs, and if they were fixed.

Currently, about all I can tell you from this is that 2.6.18-92 is a bit of an
old kernel.  If you're working with RHEL, I'd update to whatever the latest
release is (I don't recall the exact number, -128 perhaps).  I know that since
that kernel I have fixed a few quad core opteron specific bugs (although those
were primarily hangs in kdump, not oopses).  Worth a shot though.

Better still, open a support issue with Red Hat, and this bug will most likely
find its way to me, then we can get you a fix in a more timely manner :)

Neil

> On 12 February 2010 15:42, Gallus <gall.cwpl at gmail.com> wrote:
> > On 11 February 2010 19:06, Neil Horman <nhorman at redhat.com> wrote:
> >> On Thu, Feb 11, 2010 at 05:26:50PM +0100, Gallus wrote:
> >>
> >> > Does someone have any experience with such problems? Is there
> >> > something left to try out?
> >> >
> >> Yes, this is what I fix all the time.  despite your comments above, a console
> >> log would still be helpful.  It will give us the bet pointer to whats going
> >> wrong.
> >> Neil
> >
> >
> > Can you share some thoughts about such problems? Any suggestion may be helpful.
> >
> > Gallus
> >