Re: [PATCH v2 2/6] media: verisilicon: Don't hold the VPU in reset after remove

From: Arun Iyengar

Date: Fri Nov 13 2026 - 06:40:10 EST


> Le 12/11/2026 à 11:19, Philipp Zabel a écrit :
>> If vsi-iommu accesses require these bus interface resets to be
>> deasserted at certain points, why doesn't it request its own reset
>> controls?
>>
>> Depending on implicit assumptions like another driver not asserting its
>> exclusively controlled resets seems fragile.

The AV1 node lists these resets for hantro only. vsi-iommu would need
them in its own node and binding, and both drivers would have to move
to shared reset controls. Device trees without them would still hit
the SError. With this patch hantro never asserts the AV1 resets, so
vsi-iommu no longer depends on what hantro does with them.

> There is a side note on IOMMU doc saying that once the IOMMU has been
> enabled it can't be disabled for security reasons.
> That means that writing 0 on VSI_MMU_AHB_CONTROL_BASE register in vsi_iommu_disable()
> could raise an SError so I would suggest to remove this line from IOMMU driver.

I only see the SError after hantro_remove() has asserted the BIU
resets. vsi_iommu_suspend() does the same write on every autosuspend in
normal use, and that has not faulted here.

I don't think dropping the write is enough on its own. With the resets
still asserted, the AV1 domain is powered off when the IOMMU suspends,
which is the fault 8e26d4c20ed2 fixed for hantro's own suspend. The
next bind then resumes the IOMMU, and vsi_iommu_enable() writes its
registers, before hantro_probe() deasserts the resets. I have not
tested that.

If you would like the write gone as well, I can send
that as a separate vsi-iommu patch.

Arun