[Crash-utility] Crash issue when loading vmcore

Dave Anderson anderson at redhat.com
Mon Aug 23 21:25:38 UTC 2010


----- "Paul-Kenji Cahier Furuya" <pkc at f1-photo.com> wrote:

> On 08/23/2010 22:51, Dave Anderson wrote:
> > So that doesn't make any sense unless the vmlinux file and the
> vmlinux that
> > was running on the crashed kernel are not the same kernels.  Are you using
> > a different kernel as the secondary kdump kernel?
> 
> Just checked kdump's config and it says:
> #     If these are not set, kdump-config will try to use the current
> #     and initrd if it is relocatable.
> And I did not set those variables.
> 
> However I checked and found out the vmlinuz(bzImage, 7.7MB extracted)
> being run seems to be stripped, while the vmlinux from the kernel 
> directory(124MB) is not.
> 
> Could this affect the result? Is there any way to deal properly with 
> that situation?(I am using my own kernel builds, so I do not have any
> "debug kernel" packages)

It appears that the kdump configuration should be using the
same kernel as the crashed kernel, but would relocate it
when it gets run as the kdump kernel.  But that does not
explain the discrepancy between the symbol values listed
by the "VMCOREINFO" data and that of the vmlinux file that
you are using.

The vmlinuz file (with a "z" at the end) is useless for crash.
Crash needs the debuginfo-full vmlinux file that was created by
compiling the kernel with -g, and which is located at the topmost
directory in the kernel source build tree.

In any case, it would be trivial to figure this out if you could
log into the the live system and try to run crash there -- or even
simpler -- run "cat /proc/kallsyms" on that live system.  Other than
that, I don't know what else to suggest at this point.

Dave




More information about the Crash-utility mailing list