Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Dave Young]

> [..]
> 
> I tested on a prototype system with 231 entries in the map with good results.
> Everything succeeds when using kexec to initiate a fast reboot. For crash, it
> works with and without --pass-memmap-cmdline when using noefi. I hit the
> following panic when initiating a crash leaving EFI enabled in the crash kernel:
> 
>  ? __unmap_pmd_range+0x77/0x190
>  unmap_pmd_range+0xcf/0x1c0
>  populate_pgd+0x16d/0x250
>  __cpa_process_fault+0x15/0xb0
>  __change_page_attr+0x15e/0x2a0
>  __change_page_attr_set_clr+0x59/0xc0
>  kernel_map_pages_in_pgd+0x7a/0xb0
>  __map_region+0x46/0x64
>  ? early_idt_handlers+0x117/0x120
>  efi_map_region_fixed+0xd/0xf
>  efi_enter_virtual_mode+0x4c/0x476
>  ? early_idt_handlers+0x117/0x120
>  ? early_idt_handlers+0x117/0x120
>  start_kernel+0x2e5/0x376
>  ? repair_env_string+0x5b/0x5b
>  ? memblock_reserve+0x49/0x4e
>  x86_64_start_reservations+0x2a/0x2c
>  x86_64_start_kernel+0x19f/0x1ae
> 
> However, I was able to reproduce this panic using an earlier version of
> kexec-tools, so I believe it is unrelated to this patch.

Can you test with matt's tree to see if it works?
If it still happens please post the full log.

Thanks
Dave

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Matt Fleming]

On Fri, 14 Mar, at 10:47:26AM, Dave Young wrote:
> 
> Can you test with matt's tree to see if it works?
> If it still happens please post the full log.

So that'd be the 'next' branch at,

  git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/efi.git

which contains Borislav's fixes for the EFI memmap code.

-- 
Matt Fleming, Intel Open Source Technology Center

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Linn Crosetto]

On Fri, Mar 14, 2014 at 08:26:15AM +0000, Matt Fleming wrote:
> On Fri, 14 Mar, at 10:47:26AM, Dave Young wrote:
> > 
> > Can you test with matt's tree to see if it works?
> > If it still happens please post the full log.
> 
> So that'd be the 'next' branch at,
> 
>   git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/efi.git
> 
> which contains Borislav's fixes for the EFI memmap code.

I believe I included those fixes when testing with 3.14 rc6, adding:

https://lkml.org/lkml/2014/3/8/209
92b71d8
52d201e
a52f515
b660133
e91925d

Since this testing was done on a prototype, I will send you the log in a
separate email. I tried efi.git/next (with the addition of 4ce7a86), and hit the
same panic.

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Borislav Petkov]

On Fri, Mar 14, 2014 at 02:27:06PM -0600, Linn Crosetto wrote:
> Since this testing was done on a prototype, I will send you the log in
> a separate email. I tried efi.git/next (with the addition of 4ce7a86),
> and hit the same panic.

Well, if the stack trace is the same:

>  ? __unmap_pmd_range+0x77/0x190
>  unmap_pmd_range+0xcf/0x1c0
>  populate_pgd+0x16d/0x250
>  __cpa_process_fault+0x15/0xb0
...

it looks like populate_pgd() fails and unmaps the whole range. You could
add some debug printks to find out exactly why populate_pgd() fails.
Also, please try to catch full dmesg and the whole oops.

Btw, Matt, your whole efi/next stuff is already in tip, right? Because
if so, Linn could simply test latest tip/master.

Thanks.

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Borislav Petkov]

On Sat, Mar 15, 2014 at 03:26:25PM +0100, Borislav Petkov wrote:
> Also, please try to catch full dmesg and the whole oops.

Ok, nevermind, I've got your mail with full dmesg in my other mbox; I'll
take a look next week.

Thanks.

Re: [PATCH v3 4/4] x86: Pass memory range via E820 for kdump

[Matt Fleming]

On Sat, 15 Mar, at 03:26:25PM, Borislav Petkov wrote:
> 
> Btw, Matt, your whole efi/next stuff is already in tip, right? Because
> if so, Linn could simply test latest tip/master.

Yep, everything is in tip/master.

-- 
Matt Fleming, Intel Open Source Technology Center