Kdump for ARM (vmcore generation problem)

Kdump for ARM (vmcore generation problem)

[Indraneel]

Hi,
I'm trying to make Kdump work for ARM. Till now i've managed to get the 
/dev/oldmem interface working for taking dumps. Now i want to get the 
/proc/vmcore interface working.
I'm unclear how to generate the elfcore headers required by routines in 
vmcore.c.

TIA,
Indro

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Kdump for ARM (vmcore generation problem)

[Bernhard Walle]

* Indraneel <indraneel.m@samsung.com> [2007-08-14 15:11]:
> I'm trying to make Kdump work for ARM. Till now i've managed to get the 
> /dev/oldmem interface working for taking dumps. Now i want to get the 
> /proc/vmcore interface working.
> I'm unclear how to generate the elfcore headers required by routines in 
> vmcore.c.

The core headers are generated in kexec-tools (userspace).


Thanks,
   Bernhard

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Kdump for ARM (vmcore generation problem)

[Vivek Goyal]

On Tue, Aug 14, 2007 at 06:41:54PM +0530, Indraneel wrote:
> Hi,
> I'm trying to make Kdump work for ARM. Till now i've managed to get the 
> /dev/oldmem interface working for taking dumps. Now i want to get the 
> /proc/vmcore interface working.
> I'm unclear how to generate the elfcore headers required by routines in 
> vmcore.c.
> 

Hi Indraneel,

kexec-tools in user space generates those headers. Have a loot at
kexec-tools code.

Thanks
Vivek

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Kdump for ARM (vmcore generation problem)

[Indraneel]

Hi Vivek/Bernhard,

Vivek Goyal wrote:
> On Tue, Aug 14, 2007 at 06:41:54PM +0530, Indraneel wrote:
>> Hi,
>> I'm trying to make Kdump work for ARM. Till now i've managed to get the 
>> /dev/oldmem interface working for taking dumps. Now i want to get the 
>> /proc/vmcore interface working.
>> I'm unclear how to generate the elfcore headers required by routines in 
>> vmcore.c.
>>
> 
> Hi Indraneel,
> 
> kexec-tools in user space generates those headers. Have a loot at
> kexec-tools code.

Thanks for the lead. I'm looking into kexec-tools and trying to do 
changes for ARM. I'm unable to interpret one of the comments in 
kexec/crashdump-elf.c which says,
" The separate PT_LOAD program header is required either because the
   kernel is mapped at a different location than the rest of the
   physical memory or because we need to support relocatable kernels.
   Or both as on x86_64. "

Can you please provide some more clarification for this so that i can 
figure out whether a separate PT_LOAD header is required for ARM 
architecture as well (esp. the "kernel mapped to diff. loc. part of the 
comment").

TIA,
Indro

> 
> Thanks
> Vivek
> 
> 
> 


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: Kdump for ARM (vmcore generation problem)

[Bernhard Walle]

* Indraneel <indraneel.m@samsung.com> [2007-08-20 13:51]:
> Hi Vivek/Bernhard,
> 
> Vivek Goyal wrote:
> > On Tue, Aug 14, 2007 at 06:41:54PM +0530, Indraneel wrote:
> >> Hi,
> >> I'm trying to make Kdump work for ARM. Till now i've managed to get the 
> >> /dev/oldmem interface working for taking dumps. Now i want to get the 
> >> /proc/vmcore interface working.
> >> I'm unclear how to generate the elfcore headers required by routines in 
> >> vmcore.c.
> >>
> > 
> > Hi Indraneel,
> > 
> > kexec-tools in user space generates those headers. Have a loot at
> > kexec-tools code.
> 
> Thanks for the lead. I'm looking into kexec-tools and trying to do 
> changes for ARM. I'm unable to interpret one of the comments in 
> kexec/crashdump-elf.c which says,
> " The separate PT_LOAD program header is required either because the
>    kernel is mapped at a different location than the rest of the
>    physical memory or because we need to support relocatable kernels.
>    Or both as on x86_64. "
> 
> Can you please provide some more clarification for this so that i can 
> figure out whether a separate PT_LOAD header is required for ARM 
> architecture as well (esp. the "kernel mapped to diff. loc. part of the 
> comment").

Is the ARM kernel relocatable? If no, then you don't need this.


Thanks,
   Bernhard

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec