Re: [PATCH 1/2] iommu/arm-smmu-qcom: Fix missing adreno_smmu's

[Dmitry Baryshkov]

On 04/05/2023 21:08, Rob Clark wrote:
> On Thu, May 4, 2023 at 10:41 AM Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
> > On 2.05.2023 18:09, Rob Clark wrote:
> > > From: Rob Clark <robdclark@xxxxxxxxxxxx>
> > >
> > > When the special handling of qcom,adreno-smmu was moved into
> > > qcom_smmu_create(), it was overlooked that we didn't have all the
> > > required entries in qcom_smmu_impl_of_match.  So we stopped getting
> > > adreno_smmu_priv on sc7180, breaking per-process pgtables.
> > >
> > > Fixes: 30b912a03d91 ("iommu/arm-smmu-qcom: Move the qcom,adreno-smmu check into qcom_smmu_create")
> > > Signed-off-by: Rob Clark <robdclark@xxxxxxxxxxxx>
> > > ---
> > I believe the issue here is the lack of qcom,sc7180-smmu-v2 instead.
> >
> > qcom,adreno-smmu does not have to imply the "qcom smmu v2" impl
>
> Yes, but the ordering after "qcom,smmu-500" does.  Currently we just
> need the one missing "qcom,sc7180-smmu-v2" but that seemed kind of
> fragile to me, which is why I went with "qcom,adreno-smmu" as a
> catch-all

I think, the order is not relevant when comparing compatible entries, 
see __of_match_node(). It uses scoring and the more specific compatible 
wins. This way, "qcom,adreno-smmu", "qcom,smmu-500" will always end up 
with the v2 ops instead smmu500. It was not included into the list 
intentionally, rather than being by omission.

Thus I'd also suggest adding "qcom,sc7180-smmu-v2" to the list, as we 
currently do for all SMMU_v2 devices.

> BR,
> -R
>
> > Konrad
> > >   drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c | 5 +++++
> > >   1 file changed, 5 insertions(+)
> > >
> > > diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> > > index d1b296b95c86..88c89424485b 100644
> > > --- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> > > +++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
> > > @@ -512,20 +512,25 @@ static const struct of_device_id __maybe_unused qcom_smmu_impl_of_match[] = {
> > >        { .compatible = "qcom,sm6115-smmu-500", .data = &qcom_smmu_500_impl0_data},
> > >        { .compatible = "qcom,sm6125-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm6350-smmu-v2", .data = &qcom_smmu_v2_data },
> > >        { .compatible = "qcom,sm6350-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm6375-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm8150-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm8250-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm8350-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,sm8450-smmu-500", .data = &qcom_smmu_500_impl0_data },
> > >        { .compatible = "qcom,smmu-500", .data = &qcom_smmu_500_impl0_data },
> > > +     /*
> > > +      * Should come after the qcom,smmu-500 fallback so smmu-500 variants of
> > > +      * adreno-smmu get qcom_adreno_smmu_500_impl:
> > > +      */
> > > +     { .compatible = "qcom,adreno-smmu", .data = &qcom_smmu_v2_data },
> > >        { }
> > >   };
> > >
> > >   #ifdef CONFIG_ACPI
> > >   static struct acpi_platform_list qcom_acpi_platlist[] = {
> > >        { "LENOVO", "CB-01   ", 0x8180, ACPI_SIG_IORT, equal, "QCOM SMMU" },
> > >        { "QCOM  ", "QCOMEDK2", 0x8180, ACPI_SIG_IORT, equal, "QCOM SMMU" },
> > >        { }
> > >   };
> > >   #endif

--
With best wishes
Dmitry