[PATCH v23 0/9] arm64: add kdump support

[Pratyush Anand]

Hi Takahiro,

On 01/08/2016:01:37:26 PM, AKASHI Takahiro wrote:
> Pratyush,
> 
> On Fri, Jul 29, 2016 at 10:43:00PM +0530, Pratyush Anand wrote:
> > Hi Ruslan,
> > 
> > On 29/07/2016:07:26:03 PM, Ruslan Bilovol wrote:
> > > Hi,
> > > 
> > > On Tue, Jul 26, 2016 at 11:09 AM, AKASHI Takahiro
> > > <takahiro.akashi at linaro.org> wrote:
> > > > On Tue, Jul 26, 2016 at 04:56:51PM +0900, AKASHI Takahiro wrote:
> > > >> This patch series adds kdump support on arm64.
> > > >>
> > > >> To load a crash-dump kernel to the systems, a series of patches to
> > > >> kexec-tools, which have not yet been merged upstream, are needed.
> > > >> Please pick Geoff's kexec patches [1] and my kdump patches [2].
> > > >>
> > > >> To examine vmcore (/proc/vmcore) on a crash-dump kernel, you can use
> > > >>   - crash utility (coming v7.1.6 or later) [3]
> > > >>     (Necessary patches have already been queued in the master.)
> > > >>
> > > >> [1]  http://lists.infradead.org/pipermail/kexec/2016-July/016532.html
> > > >> [2]  TBD
> > > >
> > > > See my kexec-tools patches in:
> > > > http://lists.infradead.org/pipermail/kexec/2016-July/016642.html
> > > 
> > > I tired this patch series with mentioned Geoff's kexec patches and
> > > your kdump patches, but unfortunately it doesn't work for me.
> > > 
> > > While I can successfully load and kexec the kernel ('kexec --load' +
> > > 'kexec -e'), same combination of kernel+ramdisk+dtb doesn't boot
> > > if I do kdump combination ('kexec -p' + sysrq crash), there is no
> > > any output after sysrq crash trace.
> > > Unfortunately I don't have any debugger here for deep investigation
> > > of this issue.
> > 
> > Which type of image do you have? I did face issue with binary image which I
> > fixed in my branch, while elf would work fine for kdump as well.
> 
> Can you describe in more details, please?
> I don't see any problems with either vmlinux or image
> in my environment.

See kexec/crashdump-elf.c:FUNC()
We have:
223                 phdr->p_vaddr = phys_to_virt(elf_info, mstart);
Now, if we do not have page_offset then we will not have correct p_vaddr, and
then vmcore-dmesg/vmcore-dmesg.c:vaddr_to_offset() fails with

No program header covering vaddr 0xfffffc0008c312f0found kexec bug?

Fedora/RHEL kexec-tools saves vmcore only when dmesg save is OK.

I have updated [1] with the modification where --page-offset addition has been
moved to kdump binary image support patch.

[1] https://github.com/pratyushanand/kexec-tools.git : upstream_arm64_devel

~Pratyush