[PATCH V2] iommu/arm-smmu-v2: Add support for 16 bit VMID

[Will Deacon]

Hi Tirumalesh,

I still have some questions and comments about this.

On Fri, Feb 19, 2016 at 10:33:33AM -0800, tchalamarla at caviumnetworks.com wrote:
> From: Tirumalesh Chalamarla <tchalamarla at caviumnetworks.com>
> 
> ARM-SMMUv2 supports upto 16 bit VMID. This patch enables
> 16 bit VMID when HW supports.
> 
> changes from V1:
> 	- Remove DT Property and enable 16 bit VMID if ID says.

Does this mean we can remove the comment from arm_smmu_init_context_bank
about CBA2R initialisation, or is that behaviour still required by some
parts? Further, what happens if we set CR0.VMID16EN on those
implementations?

> Signed-off-by: Tirumalesh Chalamarla <tchalamarla at caviumnetworks.com>
> ---
>  drivers/iommu/arm-smmu.c | 24 +++++++++++++++++++++++-
>  1 file changed, 23 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> index 59ee4b8..92caab8 100644
> --- a/drivers/iommu/arm-smmu.c
> +++ b/drivers/iommu/arm-smmu.c
> @@ -93,6 +93,7 @@
>  #define sCR0_VMIDPNE			(1 << 11)
>  #define sCR0_PTM			(1 << 12)
>  #define sCR0_FB				(1 << 13)
> +#define sCR0_VMID16EN			(1 << 31)
>  #define sCR0_BSU_SHIFT			14
>  #define sCR0_BSU_MASK			0x3
>  
> @@ -140,6 +141,7 @@
>  #define ID2_PTFS_4K			(1 << 12)
>  #define ID2_PTFS_16K			(1 << 13)
>  #define ID2_PTFS_64K			(1 << 14)
> +#define ID2_VMID16			(1 << 15)
>  
>  /* Global TLB invalidation */
>  #define ARM_SMMU_GR0_TLBIVMID		0x64
> @@ -189,6 +191,8 @@
>  #define ARM_SMMU_GR1_CBA2R(n)		(0x800 + ((n) << 2))
>  #define CBA2R_RW64_32BIT		(0 << 0)
>  #define CBA2R_RW64_64BIT		(1 << 0)
> +#define CBA2R_VMID_SHIFT		16
> +#define CBA2R_VMID_MASK			0xffff
>  
>  /* Translation context bank */
>  #define ARM_SMMU_CB_BASE(smmu)		((smmu)->base + ((smmu)->size >> 1))
> @@ -297,6 +301,7 @@ struct arm_smmu_device {
>  #define ARM_SMMU_FEAT_TRANS_S2		(1 << 3)
>  #define ARM_SMMU_FEAT_TRANS_NESTED	(1 << 4)
>  #define ARM_SMMU_FEAT_TRANS_OPS		(1 << 5)
> +#define ARM_SMMU_FEAT_VMID16		(1 << 6)
>  	u32				features;
>  
>  #define ARM_SMMU_OPT_SECURE_CFG_ACCESS (1 << 0)
> @@ -736,6 +741,10 @@ static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
>  #else
>  		reg = CBA2R_RW64_32BIT;
>  #endif
> +		/* if 16bit VMID supported set VMID in CBA2R */
> +		if (smmu->features & ARM_SMMU_FEAT_VMID16)
> +			reg |= ARM_SMMU_CB_VMID(cfg) << CBA2R_VMID_SHIFT;
> +
>  		writel_relaxed(reg, gr1_base + ARM_SMMU_GR1_CBA2R(cfg->cbndx));
>  	}
>  
> @@ -751,7 +760,8 @@ static void arm_smmu_init_context_bank(struct arm_smmu_domain *smmu_domain,
>  	if (stage1) {
>  		reg |= (CBAR_S1_BPSHCFG_NSH << CBAR_S1_BPSHCFG_SHIFT) |
>  			(CBAR_S1_MEMATTR_WB << CBAR_S1_MEMATTR_SHIFT);
> -	} else {
> +	} else if (!(smmu->features & ARM_SMMU_FEAT_VMID16)) {
> +		/*16 bit VMID is not supported set 8 bit VMID here */
>  		reg |= ARM_SMMU_CB_VMID(cfg) << CBAR_VMID_SHIFT;
>  	}
>  	writel_relaxed(reg, gr1_base + ARM_SMMU_GR1_CBAR(cfg->cbndx));
> @@ -1508,6 +1518,9 @@ static void arm_smmu_device_reset(struct arm_smmu_device *smmu)
>  	/* Don't upgrade barriers */
>  	reg &= ~(sCR0_BSU_MASK << sCR0_BSU_SHIFT);
>  
> +	if (smmu->features & ARM_SMMU_FEAT_VMID16)
> +		reg |= sCR0_VMID16EN;
> +
>  	/* Push the button */
>  	__arm_smmu_tlb_sync(smmu);
>  	writel(reg, ARM_SMMU_GR0_NS(smmu) + ARM_SMMU_GR0_sCR0);
> @@ -1658,6 +1671,10 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
>  	size = arm_smmu_id_size_to_bits((id >> ID2_OAS_SHIFT) & ID2_OAS_MASK);
>  	smmu->pa_size = size;
>  
> +	/* See if 16bit VMID is supported */
> +	if (id & ID2_VMID16)
> +		smmu->features = ARM_SMMU_FEAT_VMID16;

This should be |=

> +
>  	/*
>  	 * What the page table walker can address actually depends on which
>  	 * descriptor format is in use, but since a) we don't know that yet,
> @@ -1696,6 +1713,11 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
>  		dev_notice(smmu->dev, "\tStage-2: %lu-bit IPA -> %lu-bit PA\n",
>  			   smmu->ipa_size, smmu->pa_size);
>  
> +	if (smmu->features & ARM_SMMU_FEAT_VMID16)
> +		dev_notice(smmu->dev, "\t 16 bit VMID Supported.\n");
> +	else
> +		dev_notice(smmu->dev, "\t Only 8 bit VMID Supported.\n");

Supporting "only" 8-bit VMIDs is usually plenty ;) In fact, your patch
doesn't change the allocator, so we'll still use a zero-extended 8-bit
VMID, so this print is fairly useless.

Will