[PATCH v5 13/20] x86/kexec: Mark relocate_kernel page as ROX instead of RWX

[Ning, Hongyu]

On 2024/12/12 18:13, David Woodhouse wrote:
> On Thu, 2024-12-12 at 11:03 +0800, Ning, Hongyu wrote:
>>
>> Hi David,
>>
>> I've hit some kdump/kexec regression issue for guest kernel in KVM/QEMU
>> based VM and reported in https://bugzilla.kernel.org/show_bug.cgi?id=219592.
>>
>> based on further git bisect, it seems to be related with this commit,
>> would you help to take a look?
> 
> Thanks for the report; I'll take a look. Please could you share your
> kernel .config?
> 

kernel config updated in the bugzilla 
https://bugzilla.kernel.org/show_bug.cgi?id=219592

> Also, you say that this is in QEMU running on an IA64 host. Is that
> true, or did you mean x86_64 host? Are you using OVMF or SeaBIOS as the
> QEMU firmware?
> 

You're right, it's x86_64 host, I miss-selected it in bugzilla.
I'm using OVMF as the QEMU firmware.

> In the short term, I think that just reverting the 'offending' commit
> should be OK. I'd *prefer* not to leave the page RWX for the whole time
> period that the image is loaded, but that's how it's been on i386 for
> ever anyway.

And your latest patch 
https://lore.kernel.org/kexec/9c68688625f409104b16164da30aa6d3eb494e5d.camel@infradead.org/ 
could fix this issue now.