| Message ID | a472d061559a6afcc4631e881e9420944471f461.1505747866.git.robin.murphy@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Sep 18, 2017 at 04:21:53PM +0100, Robin Murphy wrote: > Now that the core API issues its own post-unmap TLB sync call, push that > operation out from the io-pgtable-arm internals into the users. For now, > we leave the invalidation implicit in the unmap operation, since none of > the current users would benefit much from any change to that. > > CC: Magnus Damm <damm+renesas@opensource.se> > CC: Laurent Pinchart <laurent.pinchart@ideasonboard.com> > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > --- > drivers/iommu/arm-smmu-v3.c | 6 ++++++ > drivers/iommu/arm-smmu.c | 18 +++++++++++++----- > drivers/iommu/io-pgtable-arm.c | 7 +------ > drivers/iommu/ipmmu-vmsa.c | 6 ++++++ > 4 files changed, 26 insertions(+), 11 deletions(-) > > diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c > index e67ba6c40faf..6a866e029173 100644 > --- a/drivers/iommu/arm-smmu-v3.c > +++ b/drivers/iommu/arm-smmu-v3.c > @@ -1743,6 +1743,11 @@ arm_smmu_unmap(struct iommu_domain *domain, unsigned long iova, size_t size) > return ops->unmap(ops, iova, size); > } > > +static void arm_smmu_iotlb_sync(struct iommu_domain *domain) > +{ > + __arm_smmu_tlb_sync(to_smmu_domain(domain)->smmu); > +} > + > static phys_addr_t > arm_smmu_iova_to_phys(struct iommu_domain *domain, dma_addr_t iova) > { > @@ -1963,6 +1968,7 @@ static struct iommu_ops arm_smmu_ops = { > .map = arm_smmu_map, > .unmap = arm_smmu_unmap, > .map_sg = default_iommu_map_sg, > + .iotlb_sync = arm_smmu_iotlb_sync, > .iova_to_phys = arm_smmu_iova_to_phys, > .add_device = arm_smmu_add_device, > .remove_device = arm_smmu_remove_device, > diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c > index 3bdb799d3b4b..c4b7efe08ce4 100644 > --- a/drivers/iommu/arm-smmu.c > +++ b/drivers/iommu/arm-smmu.c > @@ -250,6 +250,7 @@ enum arm_smmu_domain_stage { > struct arm_smmu_domain { > struct arm_smmu_device *smmu; > struct io_pgtable_ops *pgtbl_ops; > + const struct iommu_gather_ops *tlb_ops; Hmmm.... I *suppose* that means there's a way to the TLB invalidation functions from the iommu_device field of the io_pgtable_cfg via something like iommu_ops_from_fwnode, but it feels horribly ugly doing that so I'm happy to keep the extra gather_ops pointer in the io_pgtable_cfg for now. This patch and the subsequent one look fine to me, but I'd like to go a step further and hook-up the add_flush callback too so that it can be (mostly) removed from the io_pgtable code and make the *_fast operations even cheaper for VFIO when it starts using them. So, Acked-by: Will Deacon <will.deacon@arm.com> Joerg, do you reckon it's worth merging this as-is, or should we also hook up add_flush before implementing this? Will
Hi Will, Robin, On Fri, Sep 22, 2017 at 04:43:22PM +0100, Will Deacon wrote: > Joerg, do you reckon it's worth merging this as-is, or should we also > hook up add_flush before implementing this? The patches implement .iotlb_sync() so that it is okay to not have a .iotlb_range_add() call-back for now. But the patches lack a .flush_iotlb_range() call-back, which is strictly necessary for correctness because it is defined to be synchronous and doesn't require a sync-call afterwards. Regards, Joerg
On 27/09/17 13:27, Joerg Roedel wrote: > Hi Will, Robin, > > On Fri, Sep 22, 2017 at 04:43:22PM +0100, Will Deacon wrote: >> Joerg, do you reckon it's worth merging this as-is, or should we also >> hook up add_flush before implementing this? > > The patches implement .iotlb_sync() so that it is okay to not have a > .iotlb_range_add() call-back for now. But the patches lack a > .flush_iotlb_range() call-back, which is strictly necessary for > correctness because it is defined to be synchronous and doesn't require > a sync-call afterwards. Good point - even with the flush still implicit in the unmap, an iommu_unmap_fast()/iommu_flush_iotlb_all() sequence would indeed be missing a sync at the end. I'll fix that up and post a new version shortly. Thanks, Robin.
diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c index e67ba6c40faf..6a866e029173 100644 --- a/drivers/iommu/arm-smmu-v3.c +++ b/drivers/iommu/arm-smmu-v3.c @@ -1743,6 +1743,11 @@ arm_smmu_unmap(struct iommu_domain *domain, unsigned long iova, size_t size) return ops->unmap(ops, iova, size); } +static void arm_smmu_iotlb_sync(struct iommu_domain *domain) +{ + __arm_smmu_tlb_sync(to_smmu_domain(domain)->smmu); +} + static phys_addr_t arm_smmu_iova_to_phys(struct iommu_domain *domain, dma_addr_t iova) { @@ -1963,6 +1968,7 @@ static struct iommu_ops arm_smmu_ops = { .map = arm_smmu_map, .unmap = arm_smmu_unmap, .map_sg = default_iommu_map_sg, + .iotlb_sync = arm_smmu_iotlb_sync, .iova_to_phys = arm_smmu_iova_to_phys, .add_device = arm_smmu_add_device, .remove_device = arm_smmu_remove_device, diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c index 3bdb799d3b4b..c4b7efe08ce4 100644 --- a/drivers/iommu/arm-smmu.c +++ b/drivers/iommu/arm-smmu.c @@ -250,6 +250,7 @@ enum arm_smmu_domain_stage { struct arm_smmu_domain { struct arm_smmu_device *smmu; struct io_pgtable_ops *pgtbl_ops; + const struct iommu_gather_ops *tlb_ops; struct arm_smmu_cfg cfg; enum arm_smmu_domain_stage stage; struct mutex init_mutex; /* Protects smmu pointer */ @@ -735,7 +736,6 @@ 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; - const struct iommu_gather_ops *tlb_ops; mutex_lock(&smmu_domain->init_mutex); if (smmu_domain->smmu) @@ -813,7 +813,7 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain, ias = min(ias, 32UL); oas = min(oas, 32UL); } - tlb_ops = &arm_smmu_s1_tlb_ops; + smmu_domain->tlb_ops = &arm_smmu_s1_tlb_ops; break; case ARM_SMMU_DOMAIN_NESTED: /* @@ -833,9 +833,9 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain, oas = min(oas, 40UL); } if (smmu->version == ARM_SMMU_V2) - tlb_ops = &arm_smmu_s2_tlb_ops_v2; + smmu_domain->tlb_ops = &arm_smmu_s2_tlb_ops_v2; else - tlb_ops = &arm_smmu_s2_tlb_ops_v1; + smmu_domain->tlb_ops = &arm_smmu_s2_tlb_ops_v1; break; default: ret = -EINVAL; @@ -863,7 +863,7 @@ static int arm_smmu_init_domain_context(struct iommu_domain *domain, .pgsize_bitmap = smmu->pgsize_bitmap, .ias = ias, .oas = oas, - .tlb = tlb_ops, + .tlb = smmu_domain->tlb_ops, .iommu_dev = smmu->dev, }; @@ -1259,6 +1259,13 @@ static size_t arm_smmu_unmap(struct iommu_domain *domain, unsigned long iova, return ops->unmap(ops, iova, size); } +static void arm_smmu_iotlb_sync(struct iommu_domain *domain) +{ + struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain); + + smmu_domain->tlb_ops->tlb_sync(smmu_domain); +} + static phys_addr_t arm_smmu_iova_to_phys_hard(struct iommu_domain *domain, dma_addr_t iova) { @@ -1562,6 +1569,7 @@ static struct iommu_ops arm_smmu_ops = { .map = arm_smmu_map, .unmap = arm_smmu_unmap, .map_sg = default_iommu_map_sg, + .iotlb_sync = arm_smmu_iotlb_sync, .iova_to_phys = arm_smmu_iova_to_phys, .add_device = arm_smmu_add_device, .remove_device = arm_smmu_remove_device, diff --git a/drivers/iommu/io-pgtable-arm.c b/drivers/iommu/io-pgtable-arm.c index e8018a308868..51e5c43caed1 100644 --- a/drivers/iommu/io-pgtable-arm.c +++ b/drivers/iommu/io-pgtable-arm.c @@ -609,7 +609,6 @@ static int __arm_lpae_unmap(struct arm_lpae_io_pgtable *data, static int arm_lpae_unmap(struct io_pgtable_ops *ops, unsigned long iova, size_t size) { - size_t unmapped; struct arm_lpae_io_pgtable *data = io_pgtable_ops_to_data(ops); arm_lpae_iopte *ptep = data->pgd; int lvl = ARM_LPAE_START_LVL(data); @@ -617,11 +616,7 @@ static int arm_lpae_unmap(struct io_pgtable_ops *ops, unsigned long iova, if (WARN_ON(iova >= (1ULL << data->iop.cfg.ias))) return 0; - unmapped = __arm_lpae_unmap(data, iova, size, lvl, ptep); - if (unmapped) - io_pgtable_tlb_sync(&data->iop); - - return unmapped; + return __arm_lpae_unmap(data, iova, size, lvl, ptep); } static phys_addr_t arm_lpae_iova_to_phys(struct io_pgtable_ops *ops, diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c index 195d6e93ac71..2c045b8d25f8 100644 --- a/drivers/iommu/ipmmu-vmsa.c +++ b/drivers/iommu/ipmmu-vmsa.c @@ -619,6 +619,11 @@ static size_t ipmmu_unmap(struct iommu_domain *io_domain, unsigned long iova, return domain->iop->unmap(domain->iop, iova, size); } +static void ipmmu_iotlb_sync(struct iommu_domain *io_domain) +{ + ipmmu_tlb_flush_all(to_vmsa_domain(io_domain)); +} + static phys_addr_t ipmmu_iova_to_phys(struct iommu_domain *io_domain, dma_addr_t iova) { @@ -876,6 +881,7 @@ static const struct iommu_ops ipmmu_ops = { .detach_dev = ipmmu_detach_device, .map = ipmmu_map, .unmap = ipmmu_unmap, + .iotlb_sync = ipmmu_iotlb_sync, .map_sg = default_iommu_map_sg, .iova_to_phys = ipmmu_iova_to_phys, .add_device = ipmmu_add_device_dma,
Now that the core API issues its own post-unmap TLB sync call, push that operation out from the io-pgtable-arm internals into the users. For now, we leave the invalidation implicit in the unmap operation, since none of the current users would benefit much from any change to that. CC: Magnus Damm <damm+renesas@opensource.se> CC: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Robin Murphy <robin.murphy@arm.com> --- drivers/iommu/arm-smmu-v3.c | 6 ++++++ drivers/iommu/arm-smmu.c | 18 +++++++++++++----- drivers/iommu/io-pgtable-arm.c | 7 +------ drivers/iommu/ipmmu-vmsa.c | 6 ++++++ 4 files changed, 26 insertions(+), 11 deletions(-)