[Crash-utility] Re: Problem with using crash 4.0-2.21 on ppc

[Dave Anderson]

Rachita Kothiyal wrote:

> Hi
>
> I tried crash 4.0-2.21 on a 2.6.16-rc4 kernel on a ppc64
> machine(Power5). Trying to analyse vmcore(from kdump) gave me
> the following error:
>
> crash: pglist_data.node_mem_map structure member does not exist.
> crash: certain memory-related commands will fail or display invalid data

This is to be expected on a CONFIG_SPARSEMEM kernel:

> If pglist_data.node_mem_map structure member does not exist, as in a
> ppc64 kernel built with CONFIG_SPARSEMEM, print an init-time warning
> message instead of failing with "crash: invalid structure member
> offset: pglist_data.node_mem_map" message.  (haren at us.ibm.com,
> anderson at redhat.com)


>
> crash: cannot determine idle task addresses from init_tasks[] or runqueues[]
> crash: cannot resolve "init_task_union"
>

This happens in get_idle_threads() when perusing the runqueues array,
where each per-cpu runqueue data structure contains a pointer to the
idle (swapper) task for that CPU.  Now, this process requires that the
per-cpu address manipulations are working correctly in order to find the
each cpu's runqueue data structure.  It looks like the ppc64 change
for per-cpu data accesses is suspect here:

> Fix to recognize post-2.6.15 ppc64 kernels moving the per_cpu_offsets
> to the "paca" structure.  Without this patch, crash fails with the
> following error messages: "crash: cannot determine idle task addresses
> from init_tasks[] or runqueues[]" and "crash: cannot resolve
> init_task_union".  (pbadari at us.ibm.com)

>
> But I was able to run it ok on a live system.
>

Same kernel?  I have no idea why there would be a difference
between live and vmcore.

Dave