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

Benjamin Gaignard benjamin.gaignard at collabora.com
Thu Nov 12 06:13:54 PST 2026


Le 12/11/2026 à 11:19, Philipp Zabel a écrit :
> On Do, 2026-11-12 at 08:30 +0000, Arun Iyengar via B4 Relay wrote:
>> From: Arun Iyengar <aiyengar1990 at gmail.com>
>>
>> On the RK3588, the resets of the AV1 decoder include the AXI and APB
>> bus interface resets (SRST_A_AV1_BIU, SRST_P_AV1_BIU), which also cover
>> the Verisilicon IOMMU of the decoder. When the driver is unbound, the
>> driver core runtime resumes the device and, through the runtime PM
>> device link, its IOMMU. hantro_remove() then asserts the resets, and
>> when the IOMMU autosuspends 100 ms later, vsi_iommu_suspend() writes to
>> its registers while they are held in reset, which raises an SError and
>> panics the system:
> 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.

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.

Regards,
Benjamin

>
> regards
> Philipp



More information about the Linux-rockchip mailing list