Re: [PATCH 1/1] Revert "iommu/arm-smmu-v3: Set TTL invalidation hint better"

[Will Deacon]

On Mon, Jul 31, 2023 at 02:21:22PM +0800, wangwudi wrote:
> From: Rui Zhu <zhurui3@xxxxxxxxxx>
> 
> This reverts commit 6833b8f2e19945a41e4d5efd8c6d9f4cae9a5b7d.
> 
> This constraint violates the protocol. When tg is not 0 but ttl, scale,
> and num are 0, the hardware reports the CERROR_IL gerror. In the
> protocol, leaf is not a prerequisite for TTL.
> 
> Cc: Will Deacon <will@xxxxxxxxxx>
> Cc: Robin Murphy <robin.murphy@xxxxxxx>
> Cc: Joerg Roedel <joro@xxxxxxxxxx>
> Cc: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> Cc: Jason Gunthorpe <jgg@xxxxxxxx>
> Cc: Yicong Yang <yangyicong@xxxxxxxxxxxxx>
> Cc: Tomas Krcka <krckatom@xxxxxxxxx>
> Cc: Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx>
> Cc: Nicolin Chen <nicolinc@xxxxxxxxxx>
> Cc: Rui Zhu <zhurui3@xxxxxxxxxx>
> 
> Signed-off-by: Rui Zhu <zhurui3@xxxxxxxxxx>
> ---
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 9 ++-------
>  1 file changed, 2 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> index 9b0dc3505601..098e84cfa82f 100644
> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> @@ -1898,13 +1898,8 @@ static void __arm_smmu_tlb_inv_range(struct arm_smmu_cmdq_ent *cmd,
>  		/* Convert page size of 12,14,16 (log2) to 1,2,3 */
>  		cmd->tlbi.tg = (tg - 10) / 2;
>  
> -		/*
> -		 * Determine what level the granule is at. For non-leaf, io-pgtable
> -		 * assumes .tlb_flush_walk can invalidate multiple levels at once,
> -		 * so ignore the nominal last-level granule and leave TTL=0.
> -		 */
> -		if (cmd->tlbi.leaf)
> -			cmd->tlbi.ttl = 4 - ((ilog2(granule) - 3) / (tg - 3));
> +		/* Determine what level the granule is at */
> +		cmd->tlbi.ttl = 4 - ((ilog2(granule) - 3) / (tg - 3));

Doesn't this reintroduce the bug that 6833b8f2e199 tried to fix?

afaict, we should only hit the problematic case of tg != 0 but ttl, scale
and num all 0 if we're invalidating a single page, so shouldn't we just
zap tg in that case, since it's not doing anything useful?

I hesitate to say we should avoid range invalidation altogether for
single-page invalidations because I think some errata workarounds might
need that to work.

Will