Re: [PATCH v11 03/16] iommu/arm-smmu-v3: Add arm_smmu_drain_queue() helper

[Pranjal Shrivastava <praan@google.com>]

On Wed, Sep 30, 2026 at 11:24:21AM -0700, Nicolin Chen wrote:
> On Tue, Sep 29, 2026 at 03:44:57AM +0000, Pranjal Shrivastava wrote:
> > +/**
> > + * arm_smmu_drain_queue - Drain an SMMU queue
> > + * @smmu: the SMMU device
> > + * @q: the queue to drain
> > + * @until_empty: target selection
> > + *
> > + * With @until_empty == true (for CMDQ), exit once the queue is observed empty:
> > + *
> > + *   cons0                cons                                prod
> > + *     |                   |                                   |
> > + *  ---+###################+=====================+=============+--->
> > + *                         |<--------- undrained==0? --------->|
> 
> Jason gave some remarks against this API:
> https://lore.kernel.org/linux-iommu/179018862538.3334538.17643821143626392419.b4-review@b4/
> 

I see!

> So, my PRI series would drop this patch. I suspect that it still
> has some value for the RPM/CMDQ case, compared to a busy-waiting
> alternative. Yet, it probably should be simplified for CMDQ case
> only.
> 
> In RPM case, the "prod" would be fixed/stable right? IOW, there
> should be no new commands added to the cmdq, so it only needs to
> wait for cons == prod?
> 

Yes, that's correct. By the time we drain, the STOP_FLAG is set and
we've waited for owner_prod to reach the target & cmdq->lock to drop..
so PROD is stable and only the HW moves CONS (same for the lvcmdqs).

I'll simplify this to just serve the CMDQ in v12, maybe just a sleeping
readl_relaxed_poll_timeout() on CONS until it matches PROD..? 
I'll think try to see if I could repurpose this too.

Thanks,
Praan