* [PATCH] iommu/amd: serialize sequence allocation under concurrent TLB invalidations
@ 2026-01-20 9:05 Ankit Soni
2026-01-20 17:24 ` Jacob Pan
0 siblings, 1 reply; 3+ messages in thread
From: Ankit Soni @ 2026-01-20 9:05 UTC (permalink / raw)
To: iommu
Cc: vasant.hegde, suravee.suthikulpanit, joro, will, robin.murphy,
linux-kernel, Srikanth.Aithal
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@amd.com>
Reported-by: Srikanth Aithal <sraithal@amd.com>
Signed-off-by: Ankit Soni <Ankit.Soni@amd.com>
---
drivers/iommu/amd/iommu.c | 18 ++++++++++++------
1 file changed, 12 insertions(+), 6 deletions(-)
diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
index d7f457338de7..593fb879b7b0 100644
--- a/drivers/iommu/amd/iommu.c
+++ b/drivers/iommu/amd/iommu.c
@@ -1422,6 +1422,12 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
return iommu_queue_command_sync(iommu, cmd, true);
}
+static u64 get_cmdsem_val(struct amd_iommu *iommu)
+{
+ lockdep_assert_held(&iommu->lock);
+ return atomic64_inc_return(&iommu->cmd_sem_val);
+}
+
/*
* This function queues a completion wait command into the command
* buffer of an IOMMU
@@ -1436,11 +1442,11 @@ static int iommu_completion_wait(struct amd_iommu *iommu)
if (!iommu->need_sync)
return 0;
- data = atomic64_inc_return(&iommu->cmd_sem_val);
- build_completion_wait(&cmd, iommu, data);
-
raw_spin_lock_irqsave(&iommu->lock, flags);
+ data = get_cmdsem_val(iommu);
+ build_completion_wait(&cmd, iommu, data);
+
ret = __iommu_queue_command_sync(iommu, &cmd, false);
raw_spin_unlock_irqrestore(&iommu->lock, flags);
@@ -3119,10 +3125,11 @@ static void iommu_flush_irt_and_complete(struct amd_iommu *iommu, u16 devid)
return;
build_inv_irt(&cmd, devid);
- data = atomic64_inc_return(&iommu->cmd_sem_val);
- build_completion_wait(&cmd2, iommu, data);
raw_spin_lock_irqsave(&iommu->lock, flags);
+ data = get_cmdsem_val(iommu);
+ build_completion_wait(&cmd2, iommu, data);
+
ret = __iommu_queue_command_sync(iommu, &cmd, true);
if (ret)
goto out_err;
@@ -3136,7 +3143,6 @@ static void iommu_flush_irt_and_complete(struct amd_iommu *iommu, u16 devid)
out_err:
raw_spin_unlock_irqrestore(&iommu->lock, flags);
- return;
}
static inline u8 iommu_get_int_tablen(struct iommu_dev_data *dev_data)
--
2.43.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH] iommu/amd: serialize sequence allocation under concurrent TLB invalidations
2026-01-20 9:05 [PATCH] iommu/amd: serialize sequence allocation under concurrent TLB invalidations Ankit Soni
@ 2026-01-20 17:24 ` Jacob Pan
2026-01-21 6:49 ` Ankit Soni
0 siblings, 1 reply; 3+ messages in thread
From: Jacob Pan @ 2026-01-20 17:24 UTC (permalink / raw)
To: Ankit Soni
Cc: iommu, vasant.hegde, suravee.suthikulpanit, joro, will,
robin.murphy, linux-kernel, Srikanth.Aithal
Hi Ankit,
On Tue, 20 Jan 2026 09:05:07 +0000
Ankit Soni <Ankit.Soni@amd.com> 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@amd.com>
> Reported-by: Srikanth Aithal <sraithal@amd.com>
> Signed-off-by: Ankit Soni <Ankit.Soni@amd.com>
> ---
> drivers/iommu/amd/iommu.c | 18 ++++++++++++------
> 1 file changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
> index d7f457338de7..593fb879b7b0 100644
> --- a/drivers/iommu/amd/iommu.c
> +++ b/drivers/iommu/amd/iommu.c
> @@ -1422,6 +1422,12 @@ static int iommu_queue_command(struct
> amd_iommu *iommu, struct iommu_cmd *cmd) return
> iommu_queue_command_sync(iommu, cmd, true); }
>
> +static u64 get_cmdsem_val(struct amd_iommu *iommu)
> +{
> + lockdep_assert_held(&iommu->lock);
> + return atomic64_inc_return(&iommu->cmd_sem_val);
Do we still need this to be atomic now that it’s protected by a
spinlock?
> +}
> +
> /*
> * This function queues a completion wait command into the command
> * buffer of an IOMMU
> @@ -1436,11 +1442,11 @@ static int iommu_completion_wait(struct
> amd_iommu *iommu) if (!iommu->need_sync)
> return 0;
>
> - data = atomic64_inc_return(&iommu->cmd_sem_val);
> - build_completion_wait(&cmd, iommu, data);
> -
> raw_spin_lock_irqsave(&iommu->lock, flags);
>
> + data = get_cmdsem_val(iommu);
> + build_completion_wait(&cmd, iommu, data);
> +
> ret = __iommu_queue_command_sync(iommu, &cmd, false);
> raw_spin_unlock_irqrestore(&iommu->lock, flags);
>
> @@ -3119,10 +3125,11 @@ static void
> iommu_flush_irt_and_complete(struct amd_iommu *iommu, u16 devid)
> return;
> build_inv_irt(&cmd, devid);
> - data = atomic64_inc_return(&iommu->cmd_sem_val);
> - build_completion_wait(&cmd2, iommu, data);
>
> raw_spin_lock_irqsave(&iommu->lock, flags);
> + data = get_cmdsem_val(iommu);
> + build_completion_wait(&cmd2, iommu, data);
> +
> ret = __iommu_queue_command_sync(iommu, &cmd, true);
> if (ret)
> goto out_err;
> @@ -3136,7 +3143,6 @@ static void iommu_flush_irt_and_complete(struct
> amd_iommu *iommu, u16 devid)
> out_err:
> raw_spin_unlock_irqrestore(&iommu->lock, flags);
> - return;
> }
>
> static inline u8 iommu_get_int_tablen(struct iommu_dev_data
> *dev_data)
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] iommu/amd: serialize sequence allocation under concurrent TLB invalidations
2026-01-20 17:24 ` Jacob Pan
@ 2026-01-21 6:49 ` Ankit Soni
0 siblings, 0 replies; 3+ messages in thread
From: Ankit Soni @ 2026-01-21 6:49 UTC (permalink / raw)
To: Jacob Pan
Cc: iommu, vasant.hegde, suravee.suthikulpanit, joro, will,
robin.murphy, linux-kernel, Srikanth.Aithal
On Tue, Jan 20, 2026 at 09:24:27AM -0800, Jacob Pan wrote:
> Hi Ankit,
>
> On Tue, 20 Jan 2026 09:05:07 +0000
> Ankit Soni <Ankit.Soni@amd.com> 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@amd.com>
> > Reported-by: Srikanth Aithal <sraithal@amd.com>
> > Signed-off-by: Ankit Soni <Ankit.Soni@amd.com>
> > ---
> > drivers/iommu/amd/iommu.c | 18 ++++++++++++------
> > 1 file changed, 12 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
> > index d7f457338de7..593fb879b7b0 100644
> > --- a/drivers/iommu/amd/iommu.c
> > +++ b/drivers/iommu/amd/iommu.c
> > @@ -1422,6 +1422,12 @@ static int iommu_queue_command(struct
> > amd_iommu *iommu, struct iommu_cmd *cmd) return
> > iommu_queue_command_sync(iommu, cmd, true); }
> >
> > +static u64 get_cmdsem_val(struct amd_iommu *iommu)
> > +{
> > + lockdep_assert_held(&iommu->lock);
> > + return atomic64_inc_return(&iommu->cmd_sem_val);
> Do we still need this to be atomic now that it’s protected by a
> spinlock?
>
Hi Jacob,
Thanks for pointing this out, we can remove atomic operation here.
I will change and post v2.
-Ankit
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-01-21 6:49 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-01-20 9:05 [PATCH] iommu/amd: serialize sequence allocation under concurrent TLB invalidations Ankit Soni
2026-01-20 17:24 ` Jacob Pan
2026-01-21 6:49 ` Ankit Soni
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox