[PATCH V2] iommu/arm-smmu: add actlr settings for mdss on sa8775p platform

Dmitry Baryshkov dmitry.baryshkov at oss.qualcomm.com
Fri Nov 21 08:40:16 PST 2025


On Thu, Nov 20, 2025 at 01:33:29PM +0100, Konrad Dybcio wrote:
> On 11/19/25 7:06 PM, Charan Teja Kalla wrote:
> > 
> > 
> > On 11/18/2025 11:33 PM, Dmitry Baryshkov wrote:
> >>> Add ACTLR settings for the mdss device on Qualcomm SA8775P platform.
> >>> This is achieved by adding compatibility string for mdss in the actlr
> >>> client of match table.
> >>>
> >>> Signed-off-by: Charan Teja Kalla <charan.kalla at oss.qualcomm.com>
> >>> ---
> >>>
> >>> Changed from V1:
> >>>   1) Added actlr setting only for MDSS and dropped for fastrpc. -- konrad
> >>>   2) ACTLR table is updated per alphanumeric order -- konrad 
> >>>   https://lore.kernel.org/all/20251105075307.1658329-1-
> >>> charan.kalla at oss.qualcomm.com/
> >>>
> >>>  drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c | 2 ++
> >>>  1 file changed, 2 insertions(+)
> >>>
> >>> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> >>> index 62874b18f645..0b400e22cb00 100644
> >>> --- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> >>> +++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> >>> @@ -41,6 +41,8 @@ static const struct of_device_id qcom_smmu_actlr_client_of_match[] = {
> >>>  			.data = (const void *) (PREFETCH_DEEP | CPRE | CMTLB) },
> >>>  	{ .compatible = "qcom,fastrpc",
> >>>  			.data = (const void *) (PREFETCH_DEEP | CPRE | CMTLB) },
> >>> +	{ .compatible = "qcom,sa8775p-mdss",
> >>> +			.data = (const void *) (PREFETCH_DEFAULT | CMTLB) },
> >> What's the difference from SC7280? Why don't we need to set CPRE?
> > 
> > Sorry, I'm really not sure If your question imply to talk about the IP
> > level difference compared to SC7280 for mdss?
> 
> "SC7280 has MDSS, SA8775P also has (a slightly different) MDSS, is there
> a reason these two have different settings"
> 
> > Regarding why don't we need CPRE -- these are QoS settings that does
> > control how well hw behaves i.e., related to performance settings not
> > the functional ones.  I guess, these settings are derived from factors
> > like how many masters are sharing a TBU, how crucial are masters using
> > that TBU(like important ones may be allowed to +7 or +15 and may be
> > others upto only +1 prefetch).
> 
> And this is a good answer
> 
> Perhaps you could reword it a little and put in the commit message to
> make it clear in case anyone has the same question in a year or so

Yes, please. It would be really nice if you can add settings for other
MDSS devices too, not just the very few that management is caring
about.

-- 
With best wishes
Dmitry



More information about the linux-arm-kernel mailing list