Currently, distros like Fedora/RHEL have enabled CONFIG_IMA_KEXEC by
default. This makes kexec/kdump kernel always fail to boot up on SME
platform because of a code bug. By debugging, the root cause is found
out and bug is fixed with this patchset.

Changelog:
v2->v3:
=======
- Add how the miscaculation is caused into patch 2 log according to
  Tom's suggestion.
- Add Tom's tag.

v1->v2:
=======
- Add patch 1 to clean up the confusing local varibale naming because
  people may mix up the local variable 'size' with the passed in parameter
  in function early_memremap_is_setup_data(). Suggested by Dave and Tom
  during v1 patch reviewing.

Baoquan He (2):
  x86/mm: rename the confusing local variable in
    early_memremap_is_setup_data()
  x86/mm/sme: fix the kdump kernel breakage on SME system when
    CONFIG_IMA_KEXEC=y

 arch/x86/mm/ioremap.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

-- 
2.41.0

On Wed, 11 Sep 2024 16:16:13 +0800 Baoquan He <bhe@redhat.com> wrote:

> Currently, distros like Fedora/RHEL have enabled CONFIG_IMA_KEXEC by
> default. This makes kexec/kdump kernel always fail to boot up on SME
> platform because of a code bug. By debugging, the root cause is found
> out and bug is fixed with this patchset.

[1/1] is a cleanup.  [2/2] fixes a bug which appears to go all the way
back to 5.10.  The bugfix patch has a dependency on the cleanup, which
is unfortunate.

We could add the Fixes: to [1/1] and add cc:stable to both patches so
they get backported into -stable kernels together.  But I think it's
nicer to just concentrate on the single bugfix patch (with Fixes: and
cc:stable) and do the cleanup later, in the usual fashion.

So can I suggest a resend please?

On Tue, Oct 29, 2024 at 06:23:06PM -0700, Andrew Morton wrote:
> So can I suggest a resend please?

Andrew, you can ignore. It is on my radar now.

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

On 10/29/24 at 06:23pm, Andrew Morton wrote:
> On Wed, 11 Sep 2024 16:16:13 +0800 Baoquan He <bhe@redhat.com> wrote:
> 
> > Currently, distros like Fedora/RHEL have enabled CONFIG_IMA_KEXEC by
> > default. This makes kexec/kdump kernel always fail to boot up on SME
> > platform because of a code bug. By debugging, the root cause is found
> > out and bug is fixed with this patchset.
> 
> [1/1] is a cleanup.  [2/2] fixes a bug which appears to go all the way
> back to 5.10.  The bugfix patch has a dependency on the cleanup, which
> is unfortunate.
> 
> We could add the Fixes: to [1/1] and add cc:stable to both patches so
> they get backported into -stable kernels together.  But I think it's
> nicer to just concentrate on the single bugfix patch (with Fixes: and
> cc:stable) and do the cleanup later, in the usual fashion.

Totally agree, thanks a lot.

> 
> So can I suggest a resend please?

Will send a standalone patch to fix the bug, and send clean up patch
after the fix patch is settled.

Hi,

On 09/11/24 at 04:16pm, Baoquan He wrote:
> Currently, distros like Fedora/RHEL have enabled CONFIG_IMA_KEXEC by
> default. This makes kexec/kdump kernel always fail to boot up on SME
> platform because of a code bug. By debugging, the root cause is found
> out and bug is fixed with this patchset.

PING.

Can this be added into 6.12 so that SME system is available? Please
tell if there's any concern or comment.

Thanks
Baoquan

On 09/30/24 at 10:59am, Baoquan He wrote:
> Hi,
> 
> On 09/11/24 at 04:16pm, Baoquan He wrote:
> > Currently, distros like Fedora/RHEL have enabled CONFIG_IMA_KEXEC by
> > default. This makes kexec/kdump kernel always fail to boot up on SME
> > platform because of a code bug. By debugging, the root cause is found
> > out and bug is fixed with this patchset.
> 
> PING.
> 
> Can this be added into 6.12 so that SME system is available? Please
> tell if there's any concern or comment.

Ping again.