Re: [PATCH v2 2/5] iommu/arm-smmu: Emulate bypass by using context banks

From: Will Deacon
Date: Mon Jul 20 2020 - 04:58:50 EST


On Thu, Jul 16, 2020 at 05:16:16PM -0700, Bjorn Andersson wrote:
> Some firmware found on various Qualcomm platforms traps writes to S2CR
> of type BYPASS and writes FAULT into the register. This prevents us from
> marking the streams for the display controller as BYPASS to allow
> continued scanout of the screen through the initialization of the ARM
> SMMU.
>
> This adds a Qualcomm specific cfg_probe function, which probes the
> behavior of the S2CR registers and if found faulty enables the related
> quirk. Based on this quirk context banks are allocated for IDENTITY
> domains as well, but with ARM_SMMU_SCTLR_M omitted.
>
> The result is valid stream mappings, without translation.
>
> Tested-by: John Stultz <john.stultz@xxxxxxxxxx>
> Tested-by: Vinod Koul <vkoul@xxxxxxxxxx>
> Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
> ---
>
> Changes since v1:
> - Picked up tested-by
>
> drivers/iommu/arm-smmu-qcom.c | 21 +++++++++++++++++++++
> drivers/iommu/arm-smmu.c | 14 ++++++++++++--
> drivers/iommu/arm-smmu.h | 3 +++
> 3 files changed, 36 insertions(+), 2 deletions(-)

[...]

> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> index fb85e716ae9a..5d5fe6741ed4 100644
> --- a/drivers/iommu/arm-smmu.c
> +++ b/drivers/iommu/arm-smmu.c
> @@ -654,7 +654,9 @@ static void arm_smmu_write_context_bank(struct arm_smmu_device *smmu, int idx)
>
> /* SCTLR */
> reg = ARM_SMMU_SCTLR_CFIE | ARM_SMMU_SCTLR_CFRE | ARM_SMMU_SCTLR_AFE |
> - ARM_SMMU_SCTLR_TRE | ARM_SMMU_SCTLR_M;
> + ARM_SMMU_SCTLR_TRE;
> + if (cfg->m)
> + reg |= ARM_SMMU_SCTLR_M;
> if (stage1)
> reg |= ARM_SMMU_SCTLR_S1_ASIDPNE;
> if (IS_ENABLED(CONFIG_CPU_BIG_ENDIAN))
> @@ -678,7 +680,11 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
> if (smmu_domain->smmu)
> goto out_unlock;
>
> - if (domain->type == IOMMU_DOMAIN_IDENTITY) {
> + /*
> + * Nothing to do for IDENTITY domains,unless disabled context banks are
> + * used to emulate bypass mappings on Qualcomm platforms.
> + */
> + if (domain->type == IOMMU_DOMAIN_IDENTITY && !smmu->qcom_bypass_quirk) {

Given that the other thread [1] with Jordan (why haven't you cc'd him?! --
adding him now) has identified the need for a callback to allocate the
context bank, why don't we use the same sort of idea here? If the impl
provides a CB allocator function, call it irrespective of the domain type.
If it allocates a domain even for an identity domain, then we can install
if with SCTLR.M clear.

Will

[1] https://lore.kernel.org/r/20200716151625.GA14526@xxxxxxxxxxxxxxxxxxxxxxxxx