Re: 2.6.21-rc4-rt0-kdump (was: Re: [patch] setup_boot_APIC_clock() irq-enable fix)

[Vivek Goyal <vgoyal@in.ibm.com>]

On Thu, Mar 22, 2007 at 02:27:25PM +0100, Michal Piotrowski wrote:
> Michal Piotrowski napisał(a):
> > On 22/03/07, Ingo Molnar <mingo@elte.hu> wrote:
> >>
> >> * Michal Piotrowski <michal.k.k.piotrowski@gmail.com> wrote:
> >>
> >> > Hi Ingo,
> >>
> >> > 2.6.21-rc4-rt0
> >>
> >> > BUG: at kernel/fork.c:1033 copy_process()
> >>
> >> thanks Michal - this is a real bug that affects upstream too. Find the
> >> fix below - i've test-booted it and it fixes the warning.
> > 
> > Problem is fixed, thanks.
> 
> BTW. It seems that nobody uses -rt as a crash dump kernel ;)
> 
> BUG: unable to handle kernel paging request at virtual address f7ebf8c4
>  printing eip:
> c1610192
> *pde = 00000000
> stopped custom tracer.
> Oops: 0000 [#1]
> PREEMPT 
> Modules linked in:
> CPU:    0
> EIP:    0060:[<c1610192>]    Not tainted VLI
> EFLAGS: 00010206   (2.6.21-rc4-rt0-kdump #3)
> EIP is at copy_oldmem_page+0x4a/0xd0
> eax: 000008c4   ebx: f7ebf000   ecx: 00000100   edx: 00000246
> esi: f7ebf8c4   edi: c4c520fc   ebp: c4d54e30   esp: c4d54e18
> ds: 007b   es: 007b   fs: 00d8  gs: 0000  ss: 0068  preempt:00000001
> Process swapper (pid: 1, ti=c4d54000 task=c4d52c20 task.ti=c4d54000)
> Stack: c17ab7e0 c183f982 c1969658 00000400 00000400 00037ebf c4d54e5c c16af187 
>        00037ebf c4c520fc 00000400 000008c4 00000000 00000000 c4c696e0 00000400 
>        c4c520fc c4d54f94 c19a9cfd c4c520fc 00000400 c4d54f78 00000000 c1840996 
> Call Trace:
>  [<c16af187>] read_from_oldmem+0x73/0x98
>  [<c19a9cfd>] vmcore_init+0x26c/0xab7
>  [<c199979b>] init+0xaa/0x287
>  [<c16044eb>] kernel_thread_helper+0x7/0x10
>  =======================
> 
> l *copy_oldmem_page+0x4a/0xd0
> 0xc1610148 is in copy_oldmem_page (arch/i386/kernel/crash_dump.c:35).
> 30       * copying the data to a pre-allocated kernel page and then copying to user
> 31       * space in non-atomic context.
> 32       */
> 33      ssize_t copy_oldmem_page(unsigned long pfn, char *buf,
> 34                                     size_t csize, unsigned long offset, int userbuf)
> 35      {
> 36              void  *vaddr;
> 37
> 38              if (!csize)
> 39                      return 0;
> 

Can you please paste the disassembly of copy_oldmem_page() on your system.
Not sure from where this faulting address 0xf7ebf8c4 is coming. We are still
in vmcore_init(), so we should be copying the data to kernel buffers only.
This looks like a valid kernel address.

Can you also put some printk() here to find out from where 0xf7ebf8c4 has
come? It does not look like a fixed kernel virutual address returned by
kmap_atomic_pfn(). Then is it passed by kernel as a parameter to
copy_oldmem_page()?

Thanks
Vivek