Re: [PATCH v5 03/14] media: iris: Fix VM count passed to firmware

[Dmitry Baryshkov]

On Fri, May 15, 2026 at 04:48:44PM +0530, Vishnu Reddy wrote:
> 
> On 5/9/2026 12:50 AM, Dmitry Baryshkov wrote:
> > On Sat, May 09, 2026 at 12:29:52AM +0530, Vishnu Reddy wrote:
> >> On Glymur, firmware interprets the value written to CPU_CS_SCIACMDARG3 as
> >> the number of virtual machines (VMs) and internally adds 1 to it. Writing
> >> 1 causes firmware to treat it as 2 VMs. Since only one VM is required,
> >> remove this write to leave the register at its reset value of 0. This does
> >> not affect other platforms as only Glymur firmware uses this register,
> >> earlier platform firmwares ignore it.
> > The explanation is pretty suspicious. I can see this write in venus
> > sources too and it was added in the initial submission, dating 2017. The
> > driver targeted two platforms, MSM8916 and MSM8996, so this write
> > predates Glymur pretty much.
> 
> Thank you for the historical context! I checked with the firmware team and
> confirmed that this register is not read by any of the platform firmwares
> currently supported in the Iris driver. Regarding MSM8916 and MSM8996, those
> are not supported in the Iris driver.

So, which platforms actually needed that register? The "currently
supported" is not strong enough. We are bringing in Agatti support and
support for SM8150 and SDM845 is in discussion. I would not be surprised
if we end up porting other older platforms too. So, you are changing the
historical behaviour. You need to document the applicability of the
change.

> 
> >> Fixes: abf5bac63f68 ("media: iris: implement the boot sequence of the firmware")
> >> Cc: stable@xxxxxxxxxxxxxxx
> >> Reviewed-by: Vikash Garodia <vikash.garodia@xxxxxxxxxxxxxxxx>
> >> Signed-off-by: Vishnu Reddy <busanna.reddy@xxxxxxxxxxxxxxxx>
> >> ---
> >>  drivers/media/platform/qcom/iris/iris_vpu_common.c | 1 -
> >>  1 file changed, 1 deletion(-)
> >>

-- 
With best wishes
Dmitry