linux-kernel - Re: amdxdna breaks suspend (was: Re: [PATCH v5 2/8] iommu: Remove IOMMU_DEV_FEAT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <0a7f0ae1-3ce6-c701-6217-422d2728027c@amd.com>
Date: Fri, 16 Jan 2026 09:01:27 -0800
From: Lizhi Hou <lizhi.hou@....com>
To: Linus Heckemann <linus@...reibt.jetzt>, Alex Deucher
	<alexdeucher@...il.com>, Jason Gunthorpe <jgg@...dia.com>
CC: Min Ma <min.ma@....com>, <dri-devel@...ts.freedesktop.org>,
	<iommu@...ts.linux.dev>, <linux-kernel@...r.kernel.org>
Subject: Re: amdxdna breaks suspend (was: Re: [PATCH v5 2/8] iommu: Remove
 IOMMU_DEV_FEAT_SVA)


On 1/16/26 08:02, Jason Gunthorpe wrote:
> On Fri, Jan 16, 2026 at 10:27:56AM -0500, Alex Deucher wrote:
>> On Fri, Jan 16, 2026 at 10:12 AM Linus Heckemann <linus@...reibt.jetzt> wrote:
>>> Jason Gunthorpe <jgg@...dia.com> writes:
>>>
>>>> On Thu, Dec 25, 2025 at 10:05:59PM +0100, Linus Heckemann wrote:
>>>>> It appears the code removed here was not in fact entirely dead; my 2024
>>>>> gpd win mini ("G1617-01", with a Ryzen 8840U) fails to suspend
>>>>> correctly, and I bisected the issue to this commit.
>>>> The only behavior change this patch had that could be relavent to a
>>>> Ryzen was in drivers/accel/amdxdna/aie2_pci.c - are you using this
>>>> driver?
>>>>
>>>> Prior to this patch amdxdna would have failed to load in configs
>>>> without an iommu as iommu_dev_enable_feature() would have
>>>> failed. After this patch it will load successfully.

I am wondering if the amdxdna loads firmware successfully or the loading 
actually failed?

When running under a hypervisor, we have observed that amdxdna firmware 
loading failures can cause issues. And a quick fix was recently 
upstreamed. Please see:

https://github.com/torvalds/linux/commit/7bbf6d15e935abbb3d604c1fa157350e84a26f98

Are you running under a hypervisor, like xen dom0?


Thanks,

Lizhi

>>>>
>>>> If so then that driver presmuably doesn't have working power
>>>> management in your system.
>>> You're right, blacklisting the amdxdna driver fixes suspend, thanks for
>>> the pointer!
>>>
>>> @Min Ma, Lizhi Hou, dri-devel: do you have any insight into why this
>>> might be happening?
>> The xdna driver requires SVA and the IOMMU.
> It should probably call device_iommu_mapped() during probe then
>
> Jason