Hi,

Please consider these AMD vIOMMU fixes for 11.0-rc2.

This series fixes a bug in the current page walker implementation.
When triggered, the issue leads to extremely slow boot or hangs and missing
devices in the guest, as reported in:

https://lore.kernel.org/qemu-devel/20260225145831.28275-3-qemu@demindiro.com/

Thank you,
Alejandro

---
Also available on github:

The following changes since commit dd3ee9c63d96b223c8d41bf5d65392eb898be626:

  Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2026-03-30 13:54:56 +0100)

are available in the Git repository at:

  https://github.com/aljimenezb/qemu.git tags/amdvi-11.0-rc2-fixes-20260330

for you to fetch changes up to 9728f609a871606b272fc37226f5af30df7ac51a:

  amd_iommu: Reject non-decreasing NextLevel in fetch_pte() (2026-03-30 17:23:54 +0000)

----------------------------------------------------------------
amd_iommu fixes for 11.0-rc2

- Fix hugepage mappings being reported as 4K pages during page walks
- Reject non-decreasing NextLevel values to avoid invalid page-walk loops
- Both patches should be picked when porting to stable release

----------------------------------------------------------------

Alejandro Jimenez (2):
  amd_iommu: Follow root pointer before page walk and use 1-based levels
  amd_iommu: Reject non-decreasing NextLevel in fetch_pte()

 hw/i386/amd_iommu.c | 136 +++++++++++++++++++++++++++++++-------------
 hw/i386/amd_iommu.h |  11 ++--
 2 files changed, 101 insertions(+), 46 deletions(-)


base-commit: dd3ee9c63d96b223c8d41bf5d65392eb898be626
-- 
2.47.3

On Mon, 30 Mar 2026 at 22:29, Alejandro Jimenez
<alejandro.j.jimenez@oracle.com> wrote:
>
> Hi,
>
> Please consider these AMD vIOMMU fixes for 11.0-rc2.
>
> This series fixes a bug in the current page walker implementation.
> When triggered, the issue leads to extremely slow boot or hangs and missing
> devices in the guest, as reported in:
>
> https://lore.kernel.org/qemu-devel/20260225145831.28275-3-qemu@demindiro.com/
>
> Thank you,
> Alejandro
>
> ---
> Also available on github:
>
> The following changes since commit dd3ee9c63d96b223c8d41bf5d65392eb898be626:
>
>   Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2026-03-30 13:54:56 +0100)
>
> are available in the Git repository at:

Hi -- please don't send pull request emails to the mailing list for
normal patch series. It's fine to put the URL of "here's a git tree
you can find these in" in the cover letter, but an actual pull
request email triggers tooling that puts the patches on the list
to consider applying directly.

thanks
-- PMM

On 3/31/26 4:39 AM, Peter Maydell wrote:
> On Mon, 30 Mar 2026 at 22:29, Alejandro Jimenez
> <alejandro.j.jimenez@oracle.com> wrote:
>>
>> Hi,
>>
>> Please consider these AMD vIOMMU fixes for 11.0-rc2.
>>
>> This series fixes a bug in the current page walker implementation.
>> When triggered, the issue leads to extremely slow boot or hangs and missing
>> devices in the guest, as reported in:
>>
>> https://lore.kernel.org/qemu-devel/20260225145831.28275-3-qemu@demindiro.com/
>>
>> Thank you,
>> Alejandro
>>
>> ---
>> Also available on github:
>>
>> The following changes since commit dd3ee9c63d96b223c8d41bf5d65392eb898be626:
>>
>>   Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging (2026-03-30 13:54:56 +0100)
>>
>> are available in the Git repository at:
> 
> Hi -- please don't send pull request emails to the mailing list for
> normal patch series. It's fine to put the URL of "here's a git tree
> you can find these in" in the cover letter, but an actual pull
> request email triggers tooling that puts the patches on the list
> to consider applying directly.
> 

Understood, thank you.

From now on I will remove the git request-pull boilerplate from the cover
letter, and include just the URL, tag and summary changelog when sending
the AMD vIOMMU subsystem changes to MST.

Sorry for the noise,
Alejandro

> thanks
> -- PMM