kdump: Trouble reading /dev/oldmem from capture kernel

kdump: Trouble reading /dev/oldmem from capture kernel

[Seemanta Dutta]

Greetings kernel hackers,

I am trying to dump the raw memory from the /dev/oldmem device from within
a kexec()'ed capture kernel. But it seems to be that I can only read 4k of
data from /dev/oldmem. I am following the kdump documentation given in:
http://www.mjmwired.net/kernel/Documentation/kdump/. And according to the
documentation in the above link, /dev/oldmem *should* present the raw view
of memory.

Some details about my scenario:
1. Production kernel panics, and I boot into the dump capture kernel
without any issues.

2. Dump capture kernel provides two interfaces to access 'old' kernel
memory (as per kdump dpcumentation): /dev/oldmem and /proc/vmcore. The
latter works perfectly. But I am more interested in the raw memory, rather
than the core file.

My expectation in this scenario was that I would be able to access entire
physical RAM using the /dev/oldmem device. But turns out, I can't.

I would be grateful if anyone pointed out if I was doing something wrong
or missing something.

This is happening over ARM. Could that have to do something with this ?
Not sure if the ARM integration of KDUMP/KEXEC has something to do with
it.

Your help would be greatly appreciated!

regards,
Seemanta

--

Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum

Re: kdump: Trouble reading /dev/oldmem from capture kernel

[Mika Westerberg]

On Fri, Aug 05, 2011 at 02:01:03PM -0700, Seemanta Dutta wrote:
> 
> 2. Dump capture kernel provides two interfaces to access 'old' kernel
> memory (as per kdump dpcumentation): /dev/oldmem and /proc/vmcore. The
> latter works perfectly. But I am more interested in the raw memory, rather
> than the core file.

Is there some specific reason to use /dev/oldmem instead? At least I've
never needed that interface. crash utility works pretty well with
/dev/vmcore.

> My expectation in this scenario was that I would be able to access entire
> physical RAM using the /dev/oldmem device. But turns out, I can't.
> 
> I would be grateful if anyone pointed out if I was doing something wrong
> or missing something.
> 
> This is happening over ARM. Could that have to do something with this ?
> Not sure if the ARM integration of KDUMP/KEXEC has something to do with
> it.

I haven't tried lately but it looks like ARM is missing setting of
'saved_max_pfn' which could explain why it works that way.