[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <95fa9ce2-51f1-4fa8-ad57-990479580b7a@amd.com>
Date: Tue, 3 Feb 2026 11:54:36 +0530
From: Vasant Hegde <vasant.hegde@....com>
To: Ankit Soni <Ankit.Soni@....com>, iommu@...ts.linux.dev
Cc: suravee.suthikulpanit@....com, joro@...tes.org, will@...nel.org,
robin.murphy@....com, linux-kernel@...r.kernel.org, Srikanth.Aithal@....com,
jacob.pan@...ux.microsoft.com
Subject: Re: [PATCH v2] iommu/amd: serialize sequence allocation under
concurrent TLB invalidations
On 1/22/2026 9:00 PM, Ankit Soni wrote:
> With concurrent TLB invalidations, completion wait randomly gets timed out
> because cmd_sem_val was incremented outside the IOMMU spinlock, allowing
> CMD_COMPL_WAIT commands to be queued out of sequence and breaking the
> ordering assumption in wait_on_sem().
> Move the cmd_sem_val increment under iommu->lock so completion sequence
> allocation is serialized with command queuing.
> And remove the unnecessary return.
>
> Fixes: d2a0cac10597 ("iommu/amd: move wait_on_sem() out of spinlock")
>
> Tested-by: Srikanth Aithal <sraithal@....com>
> Reported-by: Srikanth Aithal <sraithal@....com>
> Signed-off-by: Ankit Soni <Ankit.Soni@....com>
Reviewed-by: Vasant Hegde <vasant.hegde@....com>
@Joerg, Can you please pick this one? It fixes regression in iommu/next branch.
-Vasant
Powered by blists - more mailing lists