[PATCH 0/4] crash utility: add ARM crashdump support

[Dave Anderson]

----- "Mika Westerberg" <ext-mika.1.westerberg at nokia.com> wrote:

> Hello Dave,
> 
> First of all, thanks for the great tool! It is the best post-mortem analysis
> tool I've ever used.
> 
> This series implements ARM support for the crash utility. Current implementation
> provides following:
> 
> 	o Virtual to physical address translation through page tables
> 	o Stack unwinding support using ARM unwind tables
> 	o And some more to get minimal support for ARM architecture
> 
> There are limitations, however:
> 
> 	o Only one CPU is supported (I don't have access to ARM SMP hardware)
> 	o No support for stack unwinding with CONFIG_FRAMEPOINTER
> 	o Only diskdump format is supported (generated by makedumpfile)
> 
> These apply on top of crash 5.0.5 sources. Please let me know, if there is
> something that needs to be changed. It would be great if you would consider
> taking these in upstream crash.

Hi Mike,

Actually I have a problem with this -- although I think it's a good one for you...

As I understand it, this patch-set will run on an ARM host against either
an ARM kdump-ELF or makedumpfile-generated dumpfile, and presumably on a
live ARM system.

As it turns out, over the last couple of weeks, I've been going back-and-forth
with another ARM-support patch-set created by Jan Karlsson and Thomas Fange of 
Sony Ericcson.  However, their patch set is designed to be built on and run
on a 32-bit x86 host, although I think that with a very minor tweak, it could
also could be built and run on an ARM host.  For that reason, it doesn't support
running on a live ARM system, but rather only for ARM dumpfiles running
on an x86 host (or x86_64 given its capability of running 32-bit binaries).

In any case, I'm more than happy to fold in ARM support, but I don't know what
to do in this case.

I wonder if it would it be possible for you, Jan and Thomas to somehow collaborate
on this effort?  It seems that both sides would benefit from the work of the other
side.  I've added them to the cc list.

This should be interesting...

Dave 

 
> With these patches, the crash utility can analyze ARM crashdumps pretty well (at
> least in my testing environment). What is missing, can be implemented later on
> when we have kernel[1] support and users.
> 
> After this, all necessary userspace components should be ready for ARM
> architecture (kexec-tools, makedumpfile and crash utility).
> 
> I've tested this on OMAP3 based platforms with 2.6.32 and latest mainline
> kernels (with the kdump patches).
> 
> Regards,
> MW
> 
> [1]
> Note that kernel crashdump support is not yet in mainline but I'm hoping that it
> will be merged eventually (patches are in RMK's patch tracker already).
> 
> Mika Westerberg (4):
>   configure: add support for ARM targets
>   crash: add support for ARM kernel image
>   crash: add ARM diskdump support
>   crash: add ARM crashdump support
> 
>  Makefile            |   12 +-
>  arm.c               | 1188
> +++++++++++++++++++++++++++++++++++++++++++++++++++
>  configure.c         |   14 +
>  defs.h              |  131 ++++++
>  diskdump.c          |   50 ++-
>  kernel.c            |    2 +-
>  lkcd_vmdump_v2_v3.h |    4 +-
>  symbols.c           |    8 +-
>  unwind_arm.c        |  702 ++++++++++++++++++++++++++++++
>  9 files changed, 2102 insertions(+), 9 deletions(-)
>  create mode 100644 arm.c
>  create mode 100644 unwind_arm.c