Re: [PATCH v3 2/5] iommu/arm-smmu: Add support for split pagetables

[Robin Murphy <robin.murphy@xxxxxxx>]

Oh, one more thing...

On 16/12/2019 4:37 pm, Jordan Crouse wrote:
[...]
> @@ -651,6 +659,7 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
>   	enum io_pgtable_fmt fmt;
>   	struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
>   	struct arm_smmu_cfg *cfg = &smmu_domain->cfg;
> +	u32 quirks = 0;
>   
>   	mutex_lock(&smmu_domain->init_mutex);
>   	if (smmu_domain->smmu)
> @@ -719,6 +728,8 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain,
>   		oas = smmu->ipa_size;
>   		if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH64) {
>   			fmt = ARM_64_LPAE_S1;
> +			if (smmu_domain->split_pagetables)
> +				quirks |= IO_PGTABLE_QUIRK_ARM_TTBR1;

To avoid me forgetting and questioning it again in future, I'd recommend 
sticking a comment somewhere near here that we don't reduce cfg->ias in 
this case because we're currently assuming SEP_UPSTREAM.

Robin.

>   		} else if (cfg->fmt == ARM_SMMU_CTX_FMT_AARCH32_L) {
>   			fmt = ARM_32_LPAE_S1;
>   			ias = min(ias, 32UL);