[Crash-utility] [PATCH v3 0/9] teach crash to work with "live" ramdump

[Oleg Nesterov]

n 05/04, Dave Anderson wrote:
>
> Hi Oleg,
>
> The v3 patchset has been queued for crash-7.1.6:
>
>   https://github.com/crash-utility/crash/commit/89ed9d0a7f7da4578294a492c1ad857244ce7352
>
> I added some documentation in help.c for "crash -h", and in the crash.8 man page.
> Also, I changed ACTIVE() to LOCAL_ACTIVE() in the x86_64_calc_phys_base() function,
> so phys_base will default to 0 instead of being based upon the host kernel's.

Great, thanks!

> So the next stage should be support for creation of a permanent ELF vmcore from
> one of these QEMU ramdumps with "crash -o".  AFAICT, it should simply be a matter
> of adding this to ramdump_to_elf():
>
> --- crash-7.1.5/ramdump.c.orig
> +++ crash-7.1.5/ramdump.c
> @@ -186,6 +186,8 @@ char *ramdump_to_elf(void)
>  		e_machine = EM_AARCH64;
>  	else if (machine_type("MIPS"))
>  		e_machine = EM_MIPS;
> +	else if (machine_type("X86_64"))
> +		e_machine = EM_X86_64;
>  	else
>  		error(FATAL, "ramdump: unsupported machine type: %s\n",
>  			MACHINE_TYPE);

Works for me ;)

Tested with and without the "-o" option.

	$ ./crash ../VMLINUX /tmp/MEM at 0
	...
	$ ./crash ../VMLINUX /tmp/MEM at 0 -o /tmp/dump
	...
	$ ./crash ../VMLINUX /tmp/dump

everything seems to work.


> It would be best if it were with a crashed guest
> image.

yes, search-for-panicked-task works too.

> but I will need a pointer to a ramdump image that
> I can work with.

so everything seems to work, but if you want to play with my ramdump too you
can find the following files in http://people.redhat.com/onestero/for_dave/

	VMLINUX				# well, vmlinux ;)
	MEM				# RAM dump before panic
	MEM_AFTER_CRASH			# RAM dump after init (bash with pid==1) exits

Oleg.