[PATCH v2 0/4] iommu/amd: Convert the AMD iommu driver to the dma-iommu api

[PATCH v2 0/4] iommu/amd: Convert the AMD iommu driver to the dma-iommu api

[Tom Murphy]

Convert the AMD iommu driver to the dma-iommu api. Remove the iova
handling and reserve region code from the AMD iommu driver.

Change-log:
v2:
-Rebase on top of this series:
 http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/dma-iommu-ops.3
-Add a gfp_t parameter to the iommu_ops::map function.
-Made use of the reserve region code inside the dma-iommu api

Tom Murphy (4):
  iommu: Add gfp parameter to iommu_ops::map
  iommu/dma-iommu: Handle deferred devices
  iommu/dma-iommu: Use the dev->coherent_dma_mask
  iommu/amd: Convert the AMD iommu driver to the dma-iommu api

 drivers/iommu/Kconfig          |   1 +
 drivers/iommu/amd_iommu.c      | 694 ++++-----------------------------
 drivers/iommu/arm-smmu-v3.c    |   2 +-
 drivers/iommu/arm-smmu.c       |   2 +-
 drivers/iommu/dma-iommu.c      |  52 ++-
 drivers/iommu/exynos-iommu.c   |   2 +-
 drivers/iommu/intel-iommu.c    |   2 +-
 drivers/iommu/iommu.c          |  43 +-
 drivers/iommu/ipmmu-vmsa.c     |   2 +-
 drivers/iommu/msm_iommu.c      |   2 +-
 drivers/iommu/mtk_iommu.c      |   2 +-
 drivers/iommu/mtk_iommu_v1.c   |   2 +-
 drivers/iommu/omap-iommu.c     |   2 +-
 drivers/iommu/qcom_iommu.c     |   2 +-
 drivers/iommu/rockchip-iommu.c |   2 +-
 drivers/iommu/s390-iommu.c     |   2 +-
 drivers/iommu/tegra-gart.c     |   2 +-
 drivers/iommu/tegra-smmu.c     |   2 +-
 include/linux/iommu.h          |  21 +-
 19 files changed, 187 insertions(+), 652 deletions(-)

-- 
2.17.1

[PATCH v2 1/4] iommu: Add gfp parameter to iommu_ops::map

[Tom Murphy]

Add a gfp_t parameter to the iommu_ops::map function.
Remove the needless locking in the AMD iommu driver.

The iommu_ops::map function (or the iommu_map function which calls it)
was always supposed to be sleepable (according to Joerg's comment in
this thread: https://lore.kernel.org/patchwork/patch/977520/ ) and so
should probably have had a "might_sleep()" since it was written. However
currently the dma-iommu api can call iommu_map in an atomic context,
which it shouldn't do. This doesn't cause any problems because any iommu
driver which uses the dma-iommu api uses gfp_atomic in it's
iommu_ops::map function. But doing this wastes the memory allocators
atomic pools.

We can remove the mutex lock from amd_iommu_map and amd_iommu_unmap.
iommu_map doesn’t lock while mapping and so no two calls should touch
the same iova range. The AMD driver already handles the page table page
allocations without locks so we can safely remove the locks.

Signed-off-by: Tom Murphy <tmurphy@arista.com>
---
 drivers/iommu/amd_iommu.c      | 14 ++++-------
 drivers/iommu/arm-smmu-v3.c    |  2 +-
 drivers/iommu/arm-smmu.c       |  2 +-
 drivers/iommu/dma-iommu.c      |  6 ++---
 drivers/iommu/exynos-iommu.c   |  2 +-
 drivers/iommu/intel-iommu.c    |  2 +-
 drivers/iommu/iommu.c          | 43 +++++++++++++++++++++++++++++-----
 drivers/iommu/ipmmu-vmsa.c     |  2 +-
 drivers/iommu/msm_iommu.c      |  2 +-
 drivers/iommu/mtk_iommu.c      |  2 +-
 drivers/iommu/mtk_iommu_v1.c   |  2 +-
 drivers/iommu/omap-iommu.c     |  2 +-
 drivers/iommu/qcom_iommu.c     |  2 +-
 drivers/iommu/rockchip-iommu.c |  2 +-
 drivers/iommu/s390-iommu.c     |  2 +-
 drivers/iommu/tegra-gart.c     |  2 +-
 drivers/iommu/tegra-smmu.c     |  2 +-
 include/linux/iommu.h          | 21 ++++++++++++++++-
 18 files changed, 78 insertions(+), 34 deletions(-)

diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
index ebd062522cf5..ea3a5dc61bb0 100644
--- a/drivers/iommu/amd_iommu.c
+++ b/drivers/iommu/amd_iommu.c
@@ -3092,7 +3092,8 @@ static int amd_iommu_attach_device(struct iommu_domain *dom,
 }
 
 static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
-			 phys_addr_t paddr, size_t page_size, int iommu_prot)
+			 phys_addr_t paddr, size_t page_size, int iommu_prot,
+			 gfp_t gfp)
 {
 	struct protection_domain *domain = to_pdomain(dom);
 	int prot = 0;
@@ -3106,9 +3107,7 @@ static int amd_iommu_map(struct iommu_domain *dom, unsigned long iova,
 	if (iommu_prot & IOMMU_WRITE)
 		prot |= IOMMU_PROT_IW;
 
-	mutex_lock(&domain->api_lock);
-	ret = iommu_map_page(domain, iova, paddr, page_size, prot, GFP_KERNEL);
-	mutex_unlock(&domain->api_lock);
+	ret = iommu_map_page(domain, iova, paddr, page_size, prot, gfp);
 
 	domain_flush_np_cache(domain, iova, page_size);
 
@@ -3119,16 +3118,11 @@ static size_t amd_iommu_unmap(struct iommu_domain *dom, unsigned long iova,
 			   size_t page_size)
 {
 	struct protection_domain *domain = to_pdomain(dom);
-	size_t unmap_size;
 
 	if (domain->mode == PAGE_MODE_NONE)
 		return 0;
 
-	mutex_lock(&domain->api_lock);
-	unmap_size = iommu_unmap_page(domain, iova, page_size);
-	mutex_unlock(&domain->api_lock);
-
-	return unmap_size;
+	return iommu_unmap_page(domain, iova, page_size);
 }
 
 static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c
index d3880010c6cf..e5c48089b49f 100644
--- a/drivers/iommu/arm-smmu-v3.c
+++ b/drivers/iommu/arm-smmu-v3.c
@@ -1777,7 +1777,7 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
 }
 
 static int arm_smmu_map(struct iommu_domain *domain, unsigned long iova,
-			phys_addr_t paddr, size_t size, int prot)
+			phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct io_pgtable_ops *ops = to_smmu_domain(domain)->pgtbl_ops;
 
diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
index 045d93884164..2d50db55b788 100644
--- a/drivers/iommu/arm-smmu.c
+++ b/drivers/iommu/arm-smmu.c
@@ -1286,7 +1286,7 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
 }
 
 static int arm_smmu_map(struct iommu_domain *domain, unsigned long iova,
-			phys_addr_t paddr, size_t size, int prot)
+			phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct io_pgtable_ops *ops = to_smmu_domain(domain)->pgtbl_ops;
 	struct arm_smmu_device *smmu = to_smmu_domain(domain)->smmu;
diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index fa5713a4f6f8..7a96c2c8f56b 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -440,7 +440,7 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
 	if (!iova)
 		return DMA_MAPPING_ERROR;
 
-	if (iommu_map(domain, iova, phys - iova_off, size, prot)) {
+	if (iommu_map_atomic(domain, iova, phys - iova_off, size, prot)) {
 		iommu_dma_free_iova(cookie, iova, size);
 		return DMA_MAPPING_ERROR;
 	}
@@ -641,7 +641,7 @@ static void *iommu_dma_alloc_remap(struct device *dev, size_t size,
 			arch_dma_prep_coherent(sg_page(sg), sg->length);
 	}
 
-	if (iommu_map_sg(domain, iova, sgt.sgl, sgt.orig_nents, ioprot)
+	if (iommu_map_sg_atomic(domain, iova, sgt.sgl, sgt.orig_nents, ioprot)
 			< size)
 		goto out_free_sg;
 
@@ -1003,7 +1003,7 @@ static int iommu_dma_map_sg(struct device *dev, struct scatterlist *sg,
 	 * We'll leave any physical concatenation to the IOMMU driver's
 	 * implementation - it knows better than we do.
 	 */
-	if (iommu_map_sg(domain, iova, sg, nents, prot) < iova_len)
+	if (iommu_map_sg_atomic(domain, iova, sg, nents, prot) < iova_len)
 		goto out_free_iova;
 
 	return __finalise_sg(dev, sg, nents, iova);
diff --git a/drivers/iommu/exynos-iommu.c b/drivers/iommu/exynos-iommu.c
index 05c6bc099d62..46414234c179 100644
--- a/drivers/iommu/exynos-iommu.c
+++ b/drivers/iommu/exynos-iommu.c
@@ -1078,7 +1078,7 @@ static int lv2set_page(sysmmu_pte_t *pent, phys_addr_t paddr, size_t size,
  */
 static int exynos_iommu_map(struct iommu_domain *iommu_domain,
 			    unsigned long l_iova, phys_addr_t paddr, size_t size,
-			    int prot)
+			    int prot, gfp_t gfp)
 {
 	struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
 	sysmmu_pte_t *entry;
diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 28cb713d728c..4f0ff28f7cb9 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -5137,7 +5137,7 @@ static void intel_iommu_detach_device(struct iommu_domain *domain,
 
 static int intel_iommu_map(struct iommu_domain *domain,
 			   unsigned long iova, phys_addr_t hpa,
-			   size_t size, int iommu_prot)
+			   size_t size, int iommu_prot, gfp_t gfp)
 {
 	struct dmar_domain *dmar_domain = to_dmar_domain(domain);
 	u64 max_addr;
diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index 109de67d5d72..1b49841c177e 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -1584,8 +1584,8 @@ static size_t iommu_pgsize(struct iommu_domain *domain,
 	return pgsize;
 }
 
-int iommu_map(struct iommu_domain *domain, unsigned long iova,
-	      phys_addr_t paddr, size_t size, int prot)
+int __iommu_map(struct iommu_domain *domain, unsigned long iova,
+	      phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	const struct iommu_ops *ops = domain->ops;
 	unsigned long orig_iova = iova;
@@ -1622,8 +1622,8 @@ int iommu_map(struct iommu_domain *domain, unsigned long iova,
 
 		pr_debug("mapping: iova 0x%lx pa %pa pgsize 0x%zx\n",
 			 iova, &paddr, pgsize);
+		ret = ops->map(domain, iova, paddr, pgsize, prot, gfp);
 
-		ret = ops->map(domain, iova, paddr, pgsize, prot);
 		if (ret)
 			break;
 
@@ -1643,8 +1643,22 @@ int iommu_map(struct iommu_domain *domain, unsigned long iova,
 
 	return ret;
 }
+
+int iommu_map(struct iommu_domain *domain, unsigned long iova,
+	      phys_addr_t paddr, size_t size, int prot)
+{
+	might_sleep();
+	return __iommu_map(domain, iova, paddr, size, prot, GFP_KERNEL);
+}
 EXPORT_SYMBOL_GPL(iommu_map);
 
+int iommu_map_atomic(struct iommu_domain *domain, unsigned long iova,
+	      phys_addr_t paddr, size_t size, int prot)
+{
+	return __iommu_map(domain, iova, paddr, size, prot, GFP_ATOMIC);
+}
+EXPORT_SYMBOL_GPL(iommu_map_atomic);
+
 static size_t __iommu_unmap(struct iommu_domain *domain,
 			    unsigned long iova, size_t size,
 			    bool sync)
@@ -1719,8 +1733,9 @@ size_t iommu_unmap_fast(struct iommu_domain *domain,
 }
 EXPORT_SYMBOL_GPL(iommu_unmap_fast);
 
-size_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
-		    struct scatterlist *sg, unsigned int nents, int prot)
+size_t __iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
+		    struct scatterlist *sg, unsigned int nents, int prot,
+		    gfp_t gfp)
 {
 	size_t len = 0, mapped = 0;
 	phys_addr_t start;
@@ -1731,7 +1746,9 @@ size_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
 		phys_addr_t s_phys = sg_phys(sg);
 
 		if (len && s_phys != start + len) {
-			ret = iommu_map(domain, iova + mapped, start, len, prot);
+			ret = __iommu_map(domain, iova + mapped, start,
+					len, prot, gfp);
+
 			if (ret)
 				goto out_err;
 
@@ -1759,8 +1776,22 @@ size_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
 	return 0;
 
 }
+
+size_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
+		    struct scatterlist *sg, unsigned int nents, int prot)
+{
+	might_sleep();
+	return __iommu_map_sg(domain, iova, sg, nents, prot, GFP_KERNEL);
+}
 EXPORT_SYMBOL_GPL(iommu_map_sg);
 
+size_t iommu_map_sg_atomic(struct iommu_domain *domain, unsigned long iova,
+		    struct scatterlist *sg, unsigned int nents, int prot)
+{
+	return __iommu_map_sg(domain, iova, sg, nents, prot, GFP_ATOMIC);
+}
+EXPORT_SYMBOL_GPL(iommu_map_sg_atomic);
+
 int iommu_domain_window_enable(struct iommu_domain *domain, u32 wnd_nr,
 			       phys_addr_t paddr, u64 size, int prot)
 {
diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c
index 9a380c10655e..e005c83d49d8 100644
--- a/drivers/iommu/ipmmu-vmsa.c
+++ b/drivers/iommu/ipmmu-vmsa.c
@@ -707,7 +707,7 @@ static void ipmmu_detach_device(struct iommu_domain *io_domain,
 }
 
 static int ipmmu_map(struct iommu_domain *io_domain, unsigned long iova,
-		     phys_addr_t paddr, size_t size, int prot)
+		     phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct ipmmu_vmsa_domain *domain = to_vmsa_domain(io_domain);
 
diff --git a/drivers/iommu/msm_iommu.c b/drivers/iommu/msm_iommu.c
index 9fb0eb7a4d02..3f6bf3653aa2 100644
--- a/drivers/iommu/msm_iommu.c
+++ b/drivers/iommu/msm_iommu.c
@@ -508,7 +508,7 @@ static void msm_iommu_detach_dev(struct iommu_domain *domain,
 }
 
 static int msm_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			 phys_addr_t pa, size_t len, int prot)
+			 phys_addr_t pa, size_t len, int prot, gfp_t gfp)
 {
 	struct msm_priv *priv = to_msm_priv(domain);
 	unsigned long flags;
diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
index de3e02277b70..3176b9b54d4d 100644
--- a/drivers/iommu/mtk_iommu.c
+++ b/drivers/iommu/mtk_iommu.c
@@ -364,7 +364,7 @@ static void mtk_iommu_detach_device(struct iommu_domain *domain,
 }
 
 static int mtk_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			 phys_addr_t paddr, size_t size, int prot)
+			 phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct mtk_iommu_domain *dom = to_mtk_domain(domain);
 	unsigned long flags;
diff --git a/drivers/iommu/mtk_iommu_v1.c b/drivers/iommu/mtk_iommu_v1.c
index 52b01e3a49df..e7b1907faec1 100644
--- a/drivers/iommu/mtk_iommu_v1.c
+++ b/drivers/iommu/mtk_iommu_v1.c
@@ -303,7 +303,7 @@ static void mtk_iommu_detach_device(struct iommu_domain *domain,
 }
 
 static int mtk_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			 phys_addr_t paddr, size_t size, int prot)
+			 phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct mtk_iommu_domain *dom = to_mtk_domain(domain);
 	unsigned int page_num = size >> MT2701_IOMMU_PAGE_SHIFT;
diff --git a/drivers/iommu/omap-iommu.c b/drivers/iommu/omap-iommu.c
index d2fb347aa4ff..c1d5a71285dc 100644
--- a/drivers/iommu/omap-iommu.c
+++ b/drivers/iommu/omap-iommu.c
@@ -1109,7 +1109,7 @@ static u32 iotlb_init_entry(struct iotlb_entry *e, u32 da, u32 pa, int pgsz)
 }
 
 static int omap_iommu_map(struct iommu_domain *domain, unsigned long da,
-			  phys_addr_t pa, size_t bytes, int prot)
+			  phys_addr_t pa, size_t bytes, int prot, gfp_t gfp)
 {
 	struct omap_iommu_domain *omap_domain = to_omap_domain(domain);
 	struct device *dev = omap_domain->dev;
diff --git a/drivers/iommu/qcom_iommu.c b/drivers/iommu/qcom_iommu.c
index 8cdd3f059513..a01e07a4e76f 100644
--- a/drivers/iommu/qcom_iommu.c
+++ b/drivers/iommu/qcom_iommu.c
@@ -411,7 +411,7 @@ static void qcom_iommu_detach_dev(struct iommu_domain *domain, struct device *de
 }
 
 static int qcom_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			  phys_addr_t paddr, size_t size, int prot)
+			  phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	int ret;
 	unsigned long flags;
diff --git a/drivers/iommu/rockchip-iommu.c b/drivers/iommu/rockchip-iommu.c
index 77d4bd93fe4b..aa3507f35107 100644
--- a/drivers/iommu/rockchip-iommu.c
+++ b/drivers/iommu/rockchip-iommu.c
@@ -760,7 +760,7 @@ static int rk_iommu_map_iova(struct rk_iommu_domain *rk_domain, u32 *pte_addr,
 }
 
 static int rk_iommu_map(struct iommu_domain *domain, unsigned long _iova,
-			phys_addr_t paddr, size_t size, int prot)
+			phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct rk_iommu_domain *rk_domain = to_rk_domain(domain);
 	unsigned long flags;
diff --git a/drivers/iommu/s390-iommu.c b/drivers/iommu/s390-iommu.c
index 22d4db302c1c..efa6aa68521d 100644
--- a/drivers/iommu/s390-iommu.c
+++ b/drivers/iommu/s390-iommu.c
@@ -265,7 +265,7 @@ static int s390_iommu_update_trans(struct s390_domain *s390_domain,
 }
 
 static int s390_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			  phys_addr_t paddr, size_t size, int prot)
+			  phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct s390_domain *s390_domain = to_s390_domain(domain);
 	int flags = ZPCI_PTE_VALID, rc = 0;
diff --git a/drivers/iommu/tegra-gart.c b/drivers/iommu/tegra-gart.c
index 4d8057916552..f300099852b1 100644
--- a/drivers/iommu/tegra-gart.c
+++ b/drivers/iommu/tegra-gart.c
@@ -190,7 +190,7 @@ static inline int __gart_iommu_map(struct gart_device *gart, unsigned long iova,
 }
 
 static int gart_iommu_map(struct iommu_domain *domain, unsigned long iova,
-			  phys_addr_t pa, size_t bytes, int prot)
+			  phys_addr_t pa, size_t bytes, int prot, gfp_t gfp)
 {
 	struct gart_device *gart = gart_handle;
 	int ret;
diff --git a/drivers/iommu/tegra-smmu.c b/drivers/iommu/tegra-smmu.c
index 5182c7d6171e..e1bf867e0607 100644
--- a/drivers/iommu/tegra-smmu.c
+++ b/drivers/iommu/tegra-smmu.c
@@ -641,7 +641,7 @@ static void tegra_smmu_set_pte(struct tegra_smmu_as *as, unsigned long iova,
 }
 
 static int tegra_smmu_map(struct iommu_domain *domain, unsigned long iova,
-			  phys_addr_t paddr, size_t size, int prot)
+			  phys_addr_t paddr, size_t size, int prot, gfp_t gfp)
 {
 	struct tegra_smmu_as *as = to_smmu_as(domain);
 	dma_addr_t pte_dma;
diff --git a/include/linux/iommu.h b/include/linux/iommu.h
index ffbbc7e39cee..76b8e7fe3ed0 100644
--- a/include/linux/iommu.h
+++ b/include/linux/iommu.h
@@ -198,7 +198,7 @@ struct iommu_ops {
 	int (*attach_dev)(struct iommu_domain *domain, struct device *dev);
 	void (*detach_dev)(struct iommu_domain *domain, struct device *dev);
 	int (*map)(struct iommu_domain *domain, unsigned long iova,
-		   phys_addr_t paddr, size_t size, int prot);
+		   phys_addr_t paddr, size_t size, int prot, gfp_t gfp);
 	size_t (*unmap)(struct iommu_domain *domain, unsigned long iova,
 		     size_t size);
 	void (*flush_iotlb_all)(struct iommu_domain *domain);
@@ -295,12 +295,17 @@ extern struct iommu_domain *iommu_get_domain_for_dev(struct device *dev);
 extern struct iommu_domain *iommu_get_dma_domain(struct device *dev);
 extern int iommu_map(struct iommu_domain *domain, unsigned long iova,
 		     phys_addr_t paddr, size_t size, int prot);
+extern int iommu_map_atomic(struct iommu_domain *domain, unsigned long iova,
+		     phys_addr_t paddr, size_t size, int prot);
 extern size_t iommu_unmap(struct iommu_domain *domain, unsigned long iova,
 			  size_t size);
 extern size_t iommu_unmap_fast(struct iommu_domain *domain,
 			       unsigned long iova, size_t size);
 extern size_t iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
 			   struct scatterlist *sg,unsigned int nents, int prot);
+extern size_t iommu_map_sg_atomic(struct iommu_domain *domain,
+				  unsigned long iova, struct scatterlist *sg,
+				  unsigned int nents, int prot);
 extern phys_addr_t iommu_iova_to_phys(struct iommu_domain *domain, dma_addr_t iova);
 extern void iommu_set_fault_handler(struct iommu_domain *domain,
 			iommu_fault_handler_t handler, void *token);
@@ -469,6 +474,13 @@ static inline int iommu_map(struct iommu_domain *domain, unsigned long iova,
 	return -ENODEV;
 }
 
+static inline int iommu_map_atomic(struct iommu_domain *domain,
+				   unsigned long iova, phys_addr_t paddr,
+				   size_t size, int prot)
+{
+	return -ENODEV;
+}
+
 static inline size_t iommu_unmap(struct iommu_domain *domain,
 				 unsigned long iova, size_t size)
 {
@@ -488,6 +500,13 @@ static inline size_t iommu_map_sg(struct iommu_domain *domain,
 	return 0;
 }
 
+static inline size_t iommu_map_sg_atomic(struct iommu_domain *domain,
+				  unsigned long iova, struct scatterlist *sg,
+				  unsigned int nents, int prot)
+{
+	return 0;
+}
+
 static inline void iommu_flush_tlb_all(struct iommu_domain *domain)
 {
 }
-- 
2.17.1

[PATCH v2 2/4] iommu/dma-iommu: Handle deferred devices

[Tom Murphy]

Handle devices which defer their attach to the iommu in the dma-iommu api

Signed-off-by: Tom Murphy <tmurphy@arista.com>
---
 drivers/iommu/dma-iommu.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 7a96c2c8f56b..c18f74ad1e8b 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -322,6 +322,17 @@ static int iommu_dma_init_domain(struct iommu_domain *domain, dma_addr_t base,
 	return iova_reserve_iommu_regions(dev, domain);
 }
 
+static int handle_deferred_device(struct device *dev)
+{
+	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);
+	const struct iommu_ops *ops = domain->ops;
+
+	if (ops->is_attach_deferred && ops->is_attach_deferred(domain, dev))
+		return iommu_attach_device(domain, dev);
+
+	return 0;
+}
+
 /**
  * dma_info_to_prot - Translate DMA API directions and attributes to IOMMU API
  *                    page flags.
@@ -835,6 +846,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
 	bool coherent = dev_is_dma_coherent(dev);
 	dma_addr_t dma_handle;
 
+	handle_deferred_device(dev);
+
 	dma_handle =__iommu_dma_map(dev, phys, size,
 			dma_info_to_prot(dir, coherent, attrs),
 			iommu_get_dma_domain(dev));
@@ -849,6 +862,8 @@ static void iommu_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
 {
 	struct iommu_domain *domain = iommu_get_dma_domain(dev);
 
+	handle_deferred_device(dev);
+
 	if (!dev_is_dma_coherent(dev) && !(attrs & DMA_ATTR_SKIP_CPU_SYNC)) {
 		phys_addr_t phys = iommu_iova_to_phys(domain, dma_handle);
 
@@ -873,6 +888,8 @@ static int __finalise_sg(struct device *dev, struct scatterlist *sg, int nents,
 	unsigned int cur_len = 0, max_len = dma_get_max_seg_size(dev);
 	int i, count = 0;
 
+	handle_deferred_device(dev);
+
 	for_each_sg(sg, s, nents, i) {
 		/* Restore this segment's original unaligned fields first */
 		unsigned int s_iova_off = sg_dma_address(s);
@@ -1022,6 +1039,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
 	struct scatterlist *tmp;
 	int i;
 
+	handle_deferred_device(dev);
+
 	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC) == 0)
 		iommu_dma_sync_sg_for_cpu(dev, sg, nents, dir);
 
@@ -1042,6 +1061,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
 static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
 		size_t size, enum dma_data_direction dir, unsigned long attrs)
 {
+	handle_deferred_device(dev);
+
 	return __iommu_dma_map(dev, phys, size,
 			dma_info_to_prot(dir, false, attrs) | IOMMU_MMIO,
 			iommu_get_dma_domain(dev));
@@ -1050,12 +1071,15 @@ static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
 static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
 		size_t size, enum dma_data_direction dir, unsigned long attrs)
 {
+	handle_deferred_device(dev);
+
 	__iommu_dma_unmap(iommu_get_dma_domain(dev), handle, size);
 }
 
 static void *iommu_dma_alloc(struct device *dev, size_t size,
 		dma_addr_t *dma_handle, gfp_t gfp, unsigned long attrs)
 {
+	handle_deferred_device(dev);
 	gfp |= __GFP_ZERO;
 
 #ifdef CONFIG_DMA_DIRECT_REMAP
@@ -1076,6 +1100,8 @@ static void iommu_dma_free(struct device *dev, size_t size, void *cpu_addr,
 {
 	struct page *page;
 
+	handle_deferred_device(dev);
+
 	/*
 	 * cpu_addr can be one of 4 things depending on how it was allocated:
 	 *
@@ -1115,6 +1141,8 @@ static int iommu_dma_mmap(struct device *dev, struct vm_area_struct *vma,
 	unsigned long pfn;
 	int ret;
 
+	handle_deferred_device(dev);
+
 	vma->vm_page_prot = arch_dma_mmap_pgprot(dev, vma->vm_page_prot, attrs);
 
 	if (dma_mmap_from_dev_coherent(dev, vma, cpu_addr, size, &ret))
@@ -1143,6 +1171,8 @@ static int iommu_dma_get_sgtable(struct device *dev, struct sg_table *sgt,
 	struct page *page;
 	int ret;
 
+	handle_deferred_device(dev);
+
 #ifdef CONFIG_DMA_DIRECT_REMAP
 	if (is_vmalloc_addr(cpu_addr)) {
 		if (!(attrs & DMA_ATTR_FORCE_CONTIGUOUS))
-- 
2.17.1

Re: [PATCH v2 2/4] iommu/dma-iommu: Handle deferred devices

[Robin Murphy]

On 30/04/2019 01:29, Tom Murphy wrote:
> Handle devices which defer their attach to the iommu in the dma-iommu api

I've just spent a while trying to understand what this is about...

AFAICS it's a kdump thing where the regular default domain attachment 
may lead to ongoing DMA traffic from the crashed kernel raising a fault 
storm, so we put off the "real" attach of a given device until we know 
it's been reset and brought into a sane state, but the only way to 
reliably detect that is to wait until the kdump kernel driver starts 
making new DMA mappings. Is that about right?

(I note that for SMMUv3 we now handle that situation with the slightly 
more heavy-handed approach of just turning off reporting and letting the 
'rogue' devices fault silently, but I appreciate that not all IOMMUs may 
have that option)

> Signed-off-by: Tom Murphy <tmurphy@arista.com>
> ---
>   drivers/iommu/dma-iommu.c | 30 ++++++++++++++++++++++++++++++
>   1 file changed, 30 insertions(+)
> 
> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> index 7a96c2c8f56b..c18f74ad1e8b 100644
> --- a/drivers/iommu/dma-iommu.c
> +++ b/drivers/iommu/dma-iommu.c
> @@ -322,6 +322,17 @@ static int iommu_dma_init_domain(struct iommu_domain *domain, dma_addr_t base,
>   	return iova_reserve_iommu_regions(dev, domain);
>   }
>   
> +static int handle_deferred_device(struct device *dev)
> +{
> +	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);

We don't want iommu_get_domain_for_dev() in fast-paths, as the 
contention on the group refcount has proven to have a surprisingly high 
overhead on some large systems. That's what iommu_get_dma_domain() 
solves, but ideally, can this be wrapped in is_kdump_kernel() such as to 
have no impact at all on the general case?

> +	const struct iommu_ops *ops = domain->ops;
> +
> +	if (ops->is_attach_deferred && ops->is_attach_deferred(domain, dev))
> +		return iommu_attach_device(domain, dev);
> +
> +	return 0;
> +}
> +
>   /**
>    * dma_info_to_prot - Translate DMA API directions and attributes to IOMMU API
>    *                    page flags.
> @@ -835,6 +846,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
>   	bool coherent = dev_is_dma_coherent(dev);
>   	dma_addr_t dma_handle;
>   
> +	handle_deferred_device(dev);
> +
>   	dma_handle =__iommu_dma_map(dev, phys, size,
>   			dma_info_to_prot(dir, coherent, attrs),
>   			iommu_get_dma_domain(dev));
> @@ -849,6 +862,8 @@ static void iommu_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
>   {
>   	struct iommu_domain *domain = iommu_get_dma_domain(dev);
>   
> +	handle_deferred_device(dev);

You don't need this - it's completely bogus to make an unmap call 
without having already called the corresponding map function, so we can 
assume it's already been dealt with.

> +
>   	if (!dev_is_dma_coherent(dev) && !(attrs & DMA_ATTR_SKIP_CPU_SYNC)) {
>   		phys_addr_t phys = iommu_iova_to_phys(domain, dma_handle);
>   
> @@ -873,6 +888,8 @@ static int __finalise_sg(struct device *dev, struct scatterlist *sg, int nents,
>   	unsigned int cur_len = 0, max_len = dma_get_max_seg_size(dev);
>   	int i, count = 0;
>   
> +	handle_deferred_device(dev);

Hmm, this should be in iommu_dma_map_sg() - that's the guy that needs a 
valid domain, and it's impossible to get to __finalise_sg() without 
having been through there anyway.

> +
>   	for_each_sg(sg, s, nents, i) {
>   		/* Restore this segment's original unaligned fields first */
>   		unsigned int s_iova_off = sg_dma_address(s);
> @@ -1022,6 +1039,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
>   	struct scatterlist *tmp;
>   	int i;
>   
> +	handle_deferred_device(dev);

Again, not necessary.

> +
>   	if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC) == 0)
>   		iommu_dma_sync_sg_for_cpu(dev, sg, nents, dir);
>   
> @@ -1042,6 +1061,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
>   static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
>   		size_t size, enum dma_data_direction dir, unsigned long attrs)
>   {
> +	handle_deferred_device(dev);

Ditto.

> +
>   	return __iommu_dma_map(dev, phys, size,
>   			dma_info_to_prot(dir, false, attrs) | IOMMU_MMIO,
>   			iommu_get_dma_domain(dev));
> @@ -1050,12 +1071,15 @@ static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
>   static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
>   		size_t size, enum dma_data_direction dir, unsigned long attrs)
>   {
> +	handle_deferred_device(dev);

Ditto.

> +
>   	__iommu_dma_unmap(iommu_get_dma_domain(dev), handle, size);
>   }
>   
>   static void *iommu_dma_alloc(struct device *dev, size_t size,
>   		dma_addr_t *dma_handle, gfp_t gfp, unsigned long attrs)
>   {
> +	handle_deferred_device(dev);
>   	gfp |= __GFP_ZERO;
>   
>   #ifdef CONFIG_DMA_DIRECT_REMAP
> @@ -1076,6 +1100,8 @@ static void iommu_dma_free(struct device *dev, size_t size, void *cpu_addr,
>   {
>   	struct page *page;
>   
> +	handle_deferred_device(dev);

Similarly, you can't free anything that hasn't already come from a 
successful call to iommu_dma_alloc()...

> +
>   	/*
>   	 * cpu_addr can be one of 4 things depending on how it was allocated:
>   	 *
> @@ -1115,6 +1141,8 @@ static int iommu_dma_mmap(struct device *dev, struct vm_area_struct *vma,
>   	unsigned long pfn;
>   	int ret;
>   
> +	handle_deferred_device(dev);

...nor can you mmap() it...

> +
>   	vma->vm_page_prot = arch_dma_mmap_pgprot(dev, vma->vm_page_prot, attrs);
>   
>   	if (dma_mmap_from_dev_coherent(dev, vma, cpu_addr, size, &ret))
> @@ -1143,6 +1171,8 @@ static int iommu_dma_get_sgtable(struct device *dev, struct sg_table *sgt,
>   	struct page *page;
>   	int ret;
>   
> +	handle_deferred_device(dev);

...nor attempt to export it.

Robin.

> +
>   #ifdef CONFIG_DMA_DIRECT_REMAP
>   	if (is_vmalloc_addr(cpu_addr)) {
>   		if (!(attrs & DMA_ATTR_FORCE_CONTIGUOUS))
> 

Re: [PATCH v2 2/4] iommu/dma-iommu: Handle deferred devices

[Tom Murphy]

On Tue, Apr 30, 2019 at 2:42 PM Robin Murphy <robin.murphy@arm.com> wrote:
>
> On 30/04/2019 01:29, Tom Murphy wrote:
> > Handle devices which defer their attach to the iommu in the dma-iommu api
>
> I've just spent a while trying to understand what this is about...
>
> AFAICS it's a kdump thing where the regular default domain attachment
> may lead to ongoing DMA traffic from the crashed kernel raising a fault
> storm, so we put off the "real" attach of a given device until we know
> it's been reset and brought into a sane state, but the only way to
> reliably detect that is to wait until the kdump kernel driver starts
> making new DMA mappings. Is that about right?

That's the impression I got too. The many iterations of the patch
series which contributed this code makes it hard to figure out exactly
why it's doing what it's doing but AFAIK it works how you described
it.

>
> (I note that for SMMUv3 we now handle that situation with the slightly
> more heavy-handed approach of just turning off reporting and letting the
> 'rogue' devices fault silently, but I appreciate that not all IOMMUs may
> have that option)
>
> > Signed-off-by: Tom Murphy <tmurphy@arista.com>
> > ---
> >   drivers/iommu/dma-iommu.c | 30 ++++++++++++++++++++++++++++++
> >   1 file changed, 30 insertions(+)
> >
> > diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> > index 7a96c2c8f56b..c18f74ad1e8b 100644
> > --- a/drivers/iommu/dma-iommu.c
> > +++ b/drivers/iommu/dma-iommu.c
> > @@ -322,6 +322,17 @@ static int iommu_dma_init_domain(struct iommu_domain *domain, dma_addr_t base,
> >       return iova_reserve_iommu_regions(dev, domain);
> >   }
> >
> > +static int handle_deferred_device(struct device *dev)
> > +{
> > +     struct iommu_domain *domain = iommu_get_domain_for_dev(dev);
>
> We don't want iommu_get_domain_for_dev() in fast-paths, as the
> contention on the group refcount has proven to have a surprisingly high
> overhead on some large systems. That's what iommu_get_dma_domain()
> solves, but ideally, can this be wrapped in is_kdump_kernel() such as to
> have no impact at all on the general case?

will do.

>
> > +     const struct iommu_ops *ops = domain->ops;
> > +
> > +     if (ops->is_attach_deferred && ops->is_attach_deferred(domain, dev))
> > +             return iommu_attach_device(domain, dev);
> > +
> > +     return 0;
> > +}
> > +
> >   /**
> >    * dma_info_to_prot - Translate DMA API directions and attributes to IOMMU API
> >    *                    page flags.
> > @@ -835,6 +846,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> >       bool coherent = dev_is_dma_coherent(dev);
> >       dma_addr_t dma_handle;
> >
> > +     handle_deferred_device(dev);
> > +
> >       dma_handle =__iommu_dma_map(dev, phys, size,
> >                       dma_info_to_prot(dir, coherent, attrs),
> >                       iommu_get_dma_domain(dev));
> > @@ -849,6 +862,8 @@ static void iommu_dma_unmap_page(struct device *dev, dma_addr_t dma_handle,
> >   {
> >       struct iommu_domain *domain = iommu_get_dma_domain(dev);
> >
> > +     handle_deferred_device(dev);
>
> You don't need this - it's completely bogus to make an unmap call
> without having already called the corresponding map function, so we can
> assume it's already been dealt with.
>
> > +
> >       if (!dev_is_dma_coherent(dev) && !(attrs & DMA_ATTR_SKIP_CPU_SYNC)) {
> >               phys_addr_t phys = iommu_iova_to_phys(domain, dma_handle);
> >
> > @@ -873,6 +888,8 @@ static int __finalise_sg(struct device *dev, struct scatterlist *sg, int nents,
> >       unsigned int cur_len = 0, max_len = dma_get_max_seg_size(dev);
> >       int i, count = 0;
> >
> > +     handle_deferred_device(dev);
>
> Hmm, this should be in iommu_dma_map_sg() - that's the guy that needs a
> valid domain, and it's impossible to get to __finalise_sg() without
> having been through there anyway.
>
> > +
> >       for_each_sg(sg, s, nents, i) {
> >               /* Restore this segment's original unaligned fields first */
> >               unsigned int s_iova_off = sg_dma_address(s);
> > @@ -1022,6 +1039,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
> >       struct scatterlist *tmp;
> >       int i;
> >
> > +     handle_deferred_device(dev);
>
> Again, not necessary.
>
> > +
> >       if (!(attrs & DMA_ATTR_SKIP_CPU_SYNC) == 0)
> >               iommu_dma_sync_sg_for_cpu(dev, sg, nents, dir);
> >
> > @@ -1042,6 +1061,8 @@ static void iommu_dma_unmap_sg(struct device *dev, struct scatterlist *sg,
> >   static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
> >               size_t size, enum dma_data_direction dir, unsigned long attrs)
> >   {
> > +     handle_deferred_device(dev);
>
> Ditto.
>
> > +
> >       return __iommu_dma_map(dev, phys, size,
> >                       dma_info_to_prot(dir, false, attrs) | IOMMU_MMIO,
> >                       iommu_get_dma_domain(dev));
> > @@ -1050,12 +1071,15 @@ static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
> >   static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
> >               size_t size, enum dma_data_direction dir, unsigned long attrs)
> >   {
> > +     handle_deferred_device(dev);
>
> Ditto.
>
> > +
> >       __iommu_dma_unmap(iommu_get_dma_domain(dev), handle, size);
> >   }
> >
> >   static void *iommu_dma_alloc(struct device *dev, size_t size,
> >               dma_addr_t *dma_handle, gfp_t gfp, unsigned long attrs)
> >   {
> > +     handle_deferred_device(dev);
> >       gfp |= __GFP_ZERO;
> >
> >   #ifdef CONFIG_DMA_DIRECT_REMAP
> > @@ -1076,6 +1100,8 @@ static void iommu_dma_free(struct device *dev, size_t size, void *cpu_addr,
> >   {
> >       struct page *page;
> >
> > +     handle_deferred_device(dev);
>
> Similarly, you can't free anything that hasn't already come from a
> successful call to iommu_dma_alloc()...
>
> > +
> >       /*
> >        * cpu_addr can be one of 4 things depending on how it was allocated:
> >        *
> > @@ -1115,6 +1141,8 @@ static int iommu_dma_mmap(struct device *dev, struct vm_area_struct *vma,
> >       unsigned long pfn;
> >       int ret;
> >
> > +     handle_deferred_device(dev);
>
> ...nor can you mmap() it...
>
> > +
> >       vma->vm_page_prot = arch_dma_mmap_pgprot(dev, vma->vm_page_prot, attrs);
> >
> >       if (dma_mmap_from_dev_coherent(dev, vma, cpu_addr, size, &ret))
> > @@ -1143,6 +1171,8 @@ static int iommu_dma_get_sgtable(struct device *dev, struct sg_table *sgt,
> >       struct page *page;
> >       int ret;
> >
> > +     handle_deferred_device(dev);
>
> ...nor attempt to export it.

That all makes sense. Will fix.

>
> Robin.
>
> > +
> >   #ifdef CONFIG_DMA_DIRECT_REMAP
> >       if (is_vmalloc_addr(cpu_addr)) {
> >               if (!(attrs & DMA_ATTR_FORCE_CONTIGUOUS))
> >

[PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Tom Murphy]

Use the dev->coherent_dma_mask when allocating in the dma-iommu ops api.

Signed-off-by: Tom Murphy <tmurphy@arista.com>
---
 drivers/iommu/dma-iommu.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index c18f74ad1e8b..df03104978d7 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -436,7 +436,8 @@ static void __iommu_dma_unmap(struct iommu_domain *domain, dma_addr_t dma_addr,
 }
 
 static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
-		size_t size, int prot, struct iommu_domain *domain)
+		size_t size, int prot, struct iommu_domain *domain,
+		dma_addr_t dma_limit)
 {
 	struct iommu_dma_cookie *cookie = domain->iova_cookie;
 	size_t iova_off = 0;
@@ -447,7 +448,7 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
 		size = iova_align(&cookie->iovad, size + iova_off);
 	}
 
-	iova = iommu_dma_alloc_iova(domain, size, dma_get_mask(dev), dev);
+	iova = iommu_dma_alloc_iova(domain, size, dma_limit, dev);
 	if (!iova)
 		return DMA_MAPPING_ERROR;
 
@@ -490,7 +491,7 @@ static void *iommu_dma_alloc_contiguous(struct device *dev, size_t size,
 		return NULL;
 
 	*dma_handle = __iommu_dma_map(dev, page_to_phys(page), size, ioprot,
-			iommu_get_dma_domain(dev));
+			iommu_get_dma_domain(dev), dev->coherent_dma_mask);
 	if (*dma_handle == DMA_MAPPING_ERROR) {
 		if (!dma_release_from_contiguous(dev, page, count))
 			__free_pages(page, page_order);
@@ -760,7 +761,7 @@ static void *iommu_dma_alloc_pool(struct device *dev, size_t size,
 
 	*dma_handle = __iommu_dma_map(dev, page_to_phys(page), size,
 			dma_info_to_prot(DMA_BIDIRECTIONAL, coherent, attrs),
-			iommu_get_domain_for_dev(dev));
+			iommu_get_domain_for_dev(dev), dev->coherent_dma_mask);
 	if (*dma_handle == DMA_MAPPING_ERROR) {
 		dma_free_from_pool(vaddr, PAGE_ALIGN(size));
 		return NULL;
@@ -850,7 +851,7 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
 
 	dma_handle =__iommu_dma_map(dev, phys, size,
 			dma_info_to_prot(dir, coherent, attrs),
-			iommu_get_dma_domain(dev));
+			iommu_get_dma_domain(dev), dma_get_mask(dev));
 	if (!coherent && !(attrs & DMA_ATTR_SKIP_CPU_SYNC) &&
 	    dma_handle != DMA_MAPPING_ERROR)
 		arch_sync_dma_for_device(dev, phys, size, dir);
@@ -1065,7 +1066,7 @@ static dma_addr_t iommu_dma_map_resource(struct device *dev, phys_addr_t phys,
 
 	return __iommu_dma_map(dev, phys, size,
 			dma_info_to_prot(dir, false, attrs) | IOMMU_MMIO,
-			iommu_get_dma_domain(dev));
+			iommu_get_dma_domain(dev), dma_get_mask(dev));
 }
 
 static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
@@ -1250,7 +1251,8 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
 	if (!msi_page)
 		return NULL;
 
-	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain);
+	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain,
+			dma_get_mask(dev));
 	if (iova == DMA_MAPPING_ERROR)
 		goto out_free_page;
 
-- 
2.17.1

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Christoph Hellwig]

>  static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
> -		size_t size, int prot, struct iommu_domain *domain)
> +		size_t size, int prot, struct iommu_domain *domain,
> +		dma_addr_t dma_limit)

Can we just call this dma_mask?

>  static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
> @@ -1250,7 +1251,8 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
>  	if (!msi_page)
>  		return NULL;
>  
> -	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain);
> +	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain,
> +			dma_get_mask(dev));

Hmm, I don't think we need the DMA mask for the MSI mapping, this
should probably always use a 64-bit mask.  Or we could just untangle
it from the DMA mapping fast path entire, something like:

---
From 0debafc85174ca830f2e371ff8e8f7465bde3ad8 Mon Sep 17 00:00:00 2001
From: Christoph Hellwig <hch@lst.de>
Date: Tue, 30 Apr 2019 07:06:23 -0400
Subject: iommu/dma: opencode __iommu_dma_map in iommu_dma_get_msi_page

The MSI page mapping really is a little different from the normal DMA
mappings and doesn't need to look at the DMA mask.  Just open code
it instead of trying to squeeze the behavior into the DMA path helpers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/iommu/dma-iommu.c | 27 +++++++--------------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 58c35bab7626..2ac0df0879d7 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -358,11 +358,6 @@ static dma_addr_t iommu_dma_alloc_iova(struct iommu_domain *domain,
 	struct iova_domain *iovad = &cookie->iovad;
 	unsigned long shift, iova_len, iova = 0;
 
-	if (cookie->type == IOMMU_DMA_MSI_COOKIE) {
-		cookie->msi_iova += size;
-		return cookie->msi_iova - size;
-	}
-
 	shift = iova_shift(iovad);
 	iova_len = size >> shift;
 	/*
@@ -397,10 +392,7 @@ static void iommu_dma_free_iova(struct iommu_dma_cookie *cookie,
 {
 	struct iova_domain *iovad = &cookie->iovad;
 
-	/* The MSI case is only ever cleaning up its most recent allocation */
-	if (cookie->type == IOMMU_DMA_MSI_COOKIE)
-		cookie->msi_iova -= size;
-	else if (cookie->fq_domain)	/* non-strict mode */
+	if (cookie->fq_domain)	/* non-strict mode */
 		queue_iova(iovad, iova_pfn(iovad, iova),
 				size >> iova_shift(iovad), 0);
 	else
@@ -430,14 +422,10 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
 {
 	struct iommu_domain *domain = iommu_get_dma_domain(dev);
 	struct iommu_dma_cookie *cookie = domain->iova_cookie;
-	size_t iova_off = 0;
+	size_t iova_off = iova_offset(&cookie->iovad, phys);
 	dma_addr_t iova;
 
-	if (cookie->type == IOMMU_DMA_IOVA_COOKIE) {
-		iova_off = iova_offset(&cookie->iovad, phys);
-		size = iova_align(&cookie->iovad, size + iova_off);
-	}
-
+	size = iova_align(&cookie->iovad, size + iova_off);
 	iova = iommu_dma_alloc_iova(domain, size, dma_get_mask(dev), dev);
 	if (!iova)
 		return DMA_MAPPING_ERROR;
@@ -1121,7 +1109,6 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
 {
 	struct iommu_dma_cookie *cookie = domain->iova_cookie;
 	struct iommu_dma_msi_page *msi_page;
-	dma_addr_t iova;
 	int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
 	size_t size = cookie_msi_granule(cookie);
 
@@ -1134,16 +1121,16 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
 	if (!msi_page)
 		return NULL;
 
-	iova = __iommu_dma_map(dev, msi_addr, size, prot);
-	if (iova == DMA_MAPPING_ERROR)
+	if (iommu_map(domain, cookie->msi_iova, msi_addr, size, prot))
 		goto out_free_page;
 
 	INIT_LIST_HEAD(&msi_page->list);
 	msi_page->phys = msi_addr;
-	msi_page->iova = iova;
+	msi_page->iova = cookie->msi_iova;
 	list_add(&msi_page->list, &cookie->msi_page_list);
-	return msi_page;
 
+	cookie->msi_iova += size;
+	return msi_page;
 out_free_page:
 	kfree(msi_page);
 	return NULL;
-- 
2.20.1

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Robin Murphy]

On 30/04/2019 12:12, Christoph Hellwig wrote:
>>   static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
>> -		size_t size, int prot, struct iommu_domain *domain)
>> +		size_t size, int prot, struct iommu_domain *domain,
>> +		dma_addr_t dma_limit)
> 
> Can we just call this dma_mask?
> 
>>   static void iommu_dma_unmap_resource(struct device *dev, dma_addr_t handle,
>> @@ -1250,7 +1251,8 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
>>   	if (!msi_page)
>>   		return NULL;
>>   
>> -	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain);
>> +	iova = __iommu_dma_map(dev, msi_addr, size, prot, domain,
>> +			dma_get_mask(dev));
> 
> Hmm, I don't think we need the DMA mask for the MSI mapping, this
> should probably always use a 64-bit mask.

If that were true then we wouldn't need DMA masks for regular mappings 
either. If we have to map the MSI doorbell at all, then we certainly 
have to place it at an IOVA that the relevant device is actually capable 
of addressing.

Robin.

>  Or we could just untangle
> it from the DMA mapping fast path entire, something like:
> 
> ---
>  From 0debafc85174ca830f2e371ff8e8f7465bde3ad8 Mon Sep 17 00:00:00 2001
> From: Christoph Hellwig <hch@lst.de>
> Date: Tue, 30 Apr 2019 07:06:23 -0400
> Subject: iommu/dma: opencode __iommu_dma_map in iommu_dma_get_msi_page
> 
> The MSI page mapping really is a little different from the normal DMA
> mappings and doesn't need to look at the DMA mask.  Just open code
> it instead of trying to squeeze the behavior into the DMA path helpers.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   drivers/iommu/dma-iommu.c | 27 +++++++--------------------
>   1 file changed, 7 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> index 58c35bab7626..2ac0df0879d7 100644
> --- a/drivers/iommu/dma-iommu.c
> +++ b/drivers/iommu/dma-iommu.c
> @@ -358,11 +358,6 @@ static dma_addr_t iommu_dma_alloc_iova(struct iommu_domain *domain,
>   	struct iova_domain *iovad = &cookie->iovad;
>   	unsigned long shift, iova_len, iova = 0;
>   
> -	if (cookie->type == IOMMU_DMA_MSI_COOKIE) {
> -		cookie->msi_iova += size;
> -		return cookie->msi_iova - size;
> -	}
> -
>   	shift = iova_shift(iovad);
>   	iova_len = size >> shift;
>   	/*
> @@ -397,10 +392,7 @@ static void iommu_dma_free_iova(struct iommu_dma_cookie *cookie,
>   {
>   	struct iova_domain *iovad = &cookie->iovad;
>   
> -	/* The MSI case is only ever cleaning up its most recent allocation */
> -	if (cookie->type == IOMMU_DMA_MSI_COOKIE)
> -		cookie->msi_iova -= size;
> -	else if (cookie->fq_domain)	/* non-strict mode */
> +	if (cookie->fq_domain)	/* non-strict mode */
>   		queue_iova(iovad, iova_pfn(iovad, iova),
>   				size >> iova_shift(iovad), 0);
>   	else
> @@ -430,14 +422,10 @@ static dma_addr_t __iommu_dma_map(struct device *dev, phys_addr_t phys,
>   {
>   	struct iommu_domain *domain = iommu_get_dma_domain(dev);
>   	struct iommu_dma_cookie *cookie = domain->iova_cookie;
> -	size_t iova_off = 0;
> +	size_t iova_off = iova_offset(&cookie->iovad, phys);
>   	dma_addr_t iova;
>   
> -	if (cookie->type == IOMMU_DMA_IOVA_COOKIE) {
> -		iova_off = iova_offset(&cookie->iovad, phys);
> -		size = iova_align(&cookie->iovad, size + iova_off);
> -	}
> -
> +	size = iova_align(&cookie->iovad, size + iova_off);
>   	iova = iommu_dma_alloc_iova(domain, size, dma_get_mask(dev), dev);
>   	if (!iova)
>   		return DMA_MAPPING_ERROR;
> @@ -1121,7 +1109,6 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
>   {
>   	struct iommu_dma_cookie *cookie = domain->iova_cookie;
>   	struct iommu_dma_msi_page *msi_page;
> -	dma_addr_t iova;
>   	int prot = IOMMU_WRITE | IOMMU_NOEXEC | IOMMU_MMIO;
>   	size_t size = cookie_msi_granule(cookie);
>   
> @@ -1134,16 +1121,16 @@ static struct iommu_dma_msi_page *iommu_dma_get_msi_page(struct device *dev,
>   	if (!msi_page)
>   		return NULL;
>   
> -	iova = __iommu_dma_map(dev, msi_addr, size, prot);
> -	if (iova == DMA_MAPPING_ERROR)
> +	if (iommu_map(domain, cookie->msi_iova, msi_addr, size, prot))
>   		goto out_free_page;
>   
>   	INIT_LIST_HEAD(&msi_page->list);
>   	msi_page->phys = msi_addr;
> -	msi_page->iova = iova;
> +	msi_page->iova = cookie->msi_iova;
>   	list_add(&msi_page->list, &cookie->msi_page_list);
> -	return msi_page;
>   
> +	cookie->msi_iova += size;
> +	return msi_page;
>   out_free_page:
>   	kfree(msi_page);
>   	return NULL;
> 

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Christoph Hellwig]

On Tue, Apr 30, 2019 at 12:27:02PM +0100, Robin Murphy wrote:
> > Hmm, I don't think we need the DMA mask for the MSI mapping, this
> > should probably always use a 64-bit mask.
> 
> If that were true then we wouldn't need DMA masks for regular mappings
> either. If we have to map the MSI doorbell at all, then we certainly have to
> place it at an IOVA that the relevant device is actually capable of
> addressing.

Well, as shown by the patch below we don't even look at the DMA mask
for the MSI page - we just allocate from bottom to top.

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Robin Murphy]

On 30/04/2019 12:32, Christoph Hellwig wrote:
> On Tue, Apr 30, 2019 at 12:27:02PM +0100, Robin Murphy wrote:
>>> Hmm, I don't think we need the DMA mask for the MSI mapping, this
>>> should probably always use a 64-bit mask.
>>
>> If that were true then we wouldn't need DMA masks for regular mappings
>> either. If we have to map the MSI doorbell at all, then we certainly have to
>> place it at an IOVA that the relevant device is actually capable of
>> addressing.
> 
> Well, as shown by the patch below we don't even look at the DMA mask
> for the MSI page - we just allocate from bottom to top.

In the trivial cookie for unmanaged domains, yes, but in that case the 
responsibility is on VFIO to provide a suitable (i.e. sub-32-bit) 
address range for that cookie in the first place. In the managed case, 
allocation uses the streaming mask via iommu_dma_get_msi_page() calling 
__iommu_dma_map(). Admittedly the mask can then get overlooked when 
reusing an existing mapping, which strictly could pose a problem if you 
have multiple devices with incompatible masks in the same group (and 
such that the PCI stuff doesn't already mitigate it), but that's such an 
obscure corner case that I'm reticent to introduce the complication to 
handle it until it's actually proven necessary.

Robin.

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Tom Murphy]

Just to make this clear, I won't apply Christoph's patch (the one in
this email thread) and instead the only change I will make is to
rename dma_limit to dma_mask.

On Tue, Apr 30, 2019 at 1:05 PM Robin Murphy <robin.murphy@arm.com> wrote:
>
> On 30/04/2019 12:32, Christoph Hellwig wrote:
> > On Tue, Apr 30, 2019 at 12:27:02PM +0100, Robin Murphy wrote:
> >>> Hmm, I don't think we need the DMA mask for the MSI mapping, this
> >>> should probably always use a 64-bit mask.
> >>
> >> If that were true then we wouldn't need DMA masks for regular mappings
> >> either. If we have to map the MSI doorbell at all, then we certainly have to
> >> place it at an IOVA that the relevant device is actually capable of
> >> addressing.
> >
> > Well, as shown by the patch below we don't even look at the DMA mask
> > for the MSI page - we just allocate from bottom to top.
>
> In the trivial cookie for unmanaged domains, yes, but in that case the
> responsibility is on VFIO to provide a suitable (i.e. sub-32-bit)
> address range for that cookie in the first place. In the managed case,
> allocation uses the streaming mask via iommu_dma_get_msi_page() calling
> __iommu_dma_map(). Admittedly the mask can then get overlooked when
> reusing an existing mapping, which strictly could pose a problem if you
> have multiple devices with incompatible masks in the same group (and
> such that the PCI stuff doesn't already mitigate it), but that's such an
> obscure corner case that I'm reticent to introduce the complication to
> handle it until it's actually proven necessary.
>
> Robin.

Re: [PATCH v2 3/4] iommu/dma-iommu: Use the dev->coherent_dma_mask

[Christoph Hellwig]

On Mon, May 06, 2019 at 06:56:13PM +0100, Tom Murphy wrote:
> Just to make this clear, I won't apply Christoph's patch (the one in
> this email thread) and instead the only change I will make is to
> rename dma_limit to dma_mask.

Sounds good for now.

[PATCH v2 4/4] iommu/amd: Convert the AMD iommu driver to the dma-iommu api

[Tom Murphy]

Convert the AMD iommu driver to the dma-iommu api. Remove the iova
handling and reserve region code from the AMD iommu driver.

Signed-off-by: Tom Murphy <tmurphy@arista.com>
---
 drivers/iommu/Kconfig     |   1 +
 drivers/iommu/amd_iommu.c | 680 ++++----------------------------------
 2 files changed, 70 insertions(+), 611 deletions(-)

diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig
index 6f07f3b21816..5914ba85180b 100644
--- a/drivers/iommu/Kconfig
+++ b/drivers/iommu/Kconfig
@@ -136,6 +136,7 @@ config AMD_IOMMU
 	select PCI_PASID
 	select IOMMU_API
 	select IOMMU_IOVA
+	select IOMMU_DMA
 	depends on X86_64 && PCI && ACPI
 	---help---
 	  With this option you can enable support for AMD IOMMU hardware in
diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
index ea3a5dc61bb0..366af2b27e7f 100644
--- a/drivers/iommu/amd_iommu.c
+++ b/drivers/iommu/amd_iommu.c
@@ -32,6 +32,7 @@
 #include <linux/scatterlist.h>
 #include <linux/dma-mapping.h>
 #include <linux/dma-direct.h>
+#include <linux/dma-iommu.h>
 #include <linux/iommu-helper.h>
 #include <linux/iommu.h>
 #include <linux/delay.h>
@@ -101,8 +102,6 @@ const struct iommu_ops amd_iommu_ops;
 static ATOMIC_NOTIFIER_HEAD(ppr_notifier);
 int amd_iommu_max_glx_val = -1;
 
-static const struct dma_map_ops amd_iommu_dma_ops;
-
 /*
  * general struct to manage commands send to an IOMMU
  */
@@ -115,21 +114,6 @@ struct kmem_cache *amd_iommu_irq_cache;
 static void update_domain(struct protection_domain *domain);
 static int protection_domain_init(struct protection_domain *domain);
 static void detach_device(struct device *dev);
-static void iova_domain_flush_tlb(struct iova_domain *iovad);
-
-/*
- * Data container for a dma_ops specific protection domain
- */
-struct dma_ops_domain {
-	/* generic protection domain information */
-	struct protection_domain domain;
-
-	/* IOVA RB-Tree */
-	struct iova_domain iovad;
-};
-
-static struct iova_domain reserved_iova_ranges;
-static struct lock_class_key reserved_rbtree_key;
 
 /****************************************************************************
  *
@@ -200,12 +184,6 @@ static struct protection_domain *to_pdomain(struct iommu_domain *dom)
 	return container_of(dom, struct protection_domain, domain);
 }
 
-static struct dma_ops_domain* to_dma_ops_domain(struct protection_domain *domain)
-{
-	BUG_ON(domain->flags != PD_DMA_OPS_MASK);
-	return container_of(domain, struct dma_ops_domain, domain);
-}
-
 static struct iommu_dev_data *alloc_dev_data(u16 devid)
 {
 	struct iommu_dev_data *dev_data;
@@ -1279,12 +1257,6 @@ static void domain_flush_pages(struct protection_domain *domain,
 	__domain_flush_pages(domain, address, size, 0);
 }
 
-/* Flush the whole IO/TLB for a given protection domain */
-static void domain_flush_tlb(struct protection_domain *domain)
-{
-	__domain_flush_pages(domain, 0, CMD_INV_IOMMU_ALL_PAGES_ADDRESS, 0);
-}
-
 /* Flush the whole IO/TLB for a given protection domain - including PDE */
 static void domain_flush_tlb_pde(struct protection_domain *domain)
 {
@@ -1686,43 +1658,6 @@ static unsigned long iommu_unmap_page(struct protection_domain *dom,
 	return unmapped;
 }
 
-/****************************************************************************
- *
- * The next functions belong to the address allocator for the dma_ops
- * interface functions.
- *
- ****************************************************************************/
-
-
-static unsigned long dma_ops_alloc_iova(struct device *dev,
-					struct dma_ops_domain *dma_dom,
-					unsigned int pages, u64 dma_mask)
-{
-	unsigned long pfn = 0;
-
-	pages = __roundup_pow_of_two(pages);
-
-	if (dma_mask > DMA_BIT_MASK(32))
-		pfn = alloc_iova_fast(&dma_dom->iovad, pages,
-				      IOVA_PFN(DMA_BIT_MASK(32)), false);
-
-	if (!pfn)
-		pfn = alloc_iova_fast(&dma_dom->iovad, pages,
-				      IOVA_PFN(dma_mask), true);
-
-	return (pfn << PAGE_SHIFT);
-}
-
-static void dma_ops_free_iova(struct dma_ops_domain *dma_dom,
-			      unsigned long address,
-			      unsigned int pages)
-{
-	pages = __roundup_pow_of_two(pages);
-	address >>= PAGE_SHIFT;
-
-	free_iova_fast(&dma_dom->iovad, address, pages);
-}
-
 /****************************************************************************
  *
  * The next functions belong to the domain allocation. A domain is
@@ -1824,40 +1759,25 @@ static void free_gcr3_table(struct protection_domain *domain)
 	free_page((unsigned long)domain->gcr3_tbl);
 }
 
-static void dma_ops_domain_flush_tlb(struct dma_ops_domain *dom)
-{
-	domain_flush_tlb(&dom->domain);
-	domain_flush_complete(&dom->domain);
-}
-
-static void iova_domain_flush_tlb(struct iova_domain *iovad)
-{
-	struct dma_ops_domain *dom;
-
-	dom = container_of(iovad, struct dma_ops_domain, iovad);
-
-	dma_ops_domain_flush_tlb(dom);
-}
-
 /*
  * Free a domain, only used if something went wrong in the
  * allocation path and we need to free an already allocated page table
  */
-static void dma_ops_domain_free(struct dma_ops_domain *dom)
+static void dma_ops_domain_free(struct protection_domain *domain)
 {
-	if (!dom)
+	if (!domain)
 		return;
 
-	del_domain_from_list(&dom->domain);
+	del_domain_from_list(domain);
 
-	put_iova_domain(&dom->iovad);
+	iommu_put_dma_cookie(&domain->domain);
 
-	free_pagetable(&dom->domain);
+	free_pagetable(domain);
 
-	if (dom->domain.id)
-		domain_id_free(dom->domain.id);
+	if (domain->id)
+		domain_id_free(domain->id);
 
-	kfree(dom);
+	kfree(domain);
 }
 
 /*
@@ -1865,37 +1785,32 @@ static void dma_ops_domain_free(struct dma_ops_domain *dom)
  * It also initializes the page table and the address allocator data
  * structures required for the dma_ops interface
  */
-static struct dma_ops_domain *dma_ops_domain_alloc(void)
+static struct protection_domain *dma_ops_domain_alloc(void)
 {
-	struct dma_ops_domain *dma_dom;
+	struct protection_domain *domain;
 
-	dma_dom = kzalloc(sizeof(struct dma_ops_domain), GFP_KERNEL);
-	if (!dma_dom)
+	domain = kzalloc(sizeof(struct protection_domain), GFP_KERNEL);
+	if (!domain)
 		return NULL;
 
-	if (protection_domain_init(&dma_dom->domain))
-		goto free_dma_dom;
-
-	dma_dom->domain.mode = PAGE_MODE_3_LEVEL;
-	dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL);
-	dma_dom->domain.flags = PD_DMA_OPS_MASK;
-	if (!dma_dom->domain.pt_root)
-		goto free_dma_dom;
-
-	init_iova_domain(&dma_dom->iovad, PAGE_SIZE, IOVA_START_PFN);
+	if (protection_domain_init(domain))
+		goto free_domain;
 
-	if (init_iova_flush_queue(&dma_dom->iovad, iova_domain_flush_tlb, NULL))
-		goto free_dma_dom;
+	domain->mode = PAGE_MODE_3_LEVEL;
+	domain->pt_root = (void *)get_zeroed_page(GFP_KERNEL);
+	domain->flags = PD_DMA_OPS_MASK;
+	if (!domain->pt_root)
+		goto free_domain;
 
-	/* Initialize reserved ranges */
-	copy_reserved_iova(&reserved_iova_ranges, &dma_dom->iovad);
+	if (iommu_get_dma_cookie(&domain->domain) == -ENOMEM)
+		goto free_domain;
 
-	add_domain_to_list(&dma_dom->domain);
+	add_domain_to_list(domain);
 
-	return dma_dom;
+	return domain;
 
-free_dma_dom:
-	dma_ops_domain_free(dma_dom);
+free_domain:
+	dma_ops_domain_free(domain);
 
 	return NULL;
 }
@@ -2291,8 +2206,8 @@ static int amd_iommu_add_device(struct device *dev)
 	domain = iommu_get_domain_for_dev(dev);
 	if (domain->type == IOMMU_DOMAIN_IDENTITY)
 		dev_data->passthrough = true;
-	else
-		dev->dma_ops = &amd_iommu_dma_ops;
+	else if (domain->type == IOMMU_DOMAIN_DMA)
+		iommu_setup_dma_ops(dev, IOVA_START_PFN << PAGE_SHIFT, 0);
 
 out:
 	iommu_completion_wait(iommu);
@@ -2326,43 +2241,32 @@ static struct iommu_group *amd_iommu_device_group(struct device *dev)
 	return acpihid_device_group(dev);
 }
 
+static int amd_iommu_domain_get_attr(struct iommu_domain *domain,
+		enum iommu_attr attr, void *data)
+{
+	switch (domain->type) {
+	case IOMMU_DOMAIN_UNMANAGED:
+		return -ENODEV;
+	case IOMMU_DOMAIN_DMA:
+		switch (attr) {
+		case DOMAIN_ATTR_DMA_USE_FLUSH_QUEUE:
+			*(int *)data = !amd_iommu_unmap_flush;
+			return 0;
+		default:
+			return -ENODEV;
+		}
+		break;
+	default:
+		return -EINVAL;
+	}
+}
+
 /*****************************************************************************
  *
  * The next functions belong to the dma_ops mapping/unmapping code.
  *
  *****************************************************************************/
 
-/*
- * In the dma_ops path we only have the struct device. This function
- * finds the corresponding IOMMU, the protection domain and the
- * requestor id for a given device.
- * If the device is not yet associated with a domain this is also done
- * in this function.
- */
-static struct protection_domain *get_domain(struct device *dev)
-{
-	struct protection_domain *domain;
-	struct iommu_domain *io_domain;
-
-	if (!check_device(dev))
-		return ERR_PTR(-EINVAL);
-
-	domain = get_dev_data(dev)->domain;
-	if (domain == NULL && get_dev_data(dev)->defer_attach) {
-		get_dev_data(dev)->defer_attach = false;
-		io_domain = iommu_get_domain_for_dev(dev);
-		domain = to_pdomain(io_domain);
-		attach_device(dev, domain);
-	}
-	if (domain == NULL)
-		return ERR_PTR(-EBUSY);
-
-	if (!dma_ops_domain(domain))
-		return ERR_PTR(-EBUSY);
-
-	return domain;
-}
-
 static void update_device_table(struct protection_domain *domain)
 {
 	struct iommu_dev_data *dev_data;
@@ -2393,446 +2297,6 @@ static void update_domain(struct protection_domain *domain)
 	domain->updated = false;
 }
 
-static int dir2prot(enum dma_data_direction direction)
-{
-	if (direction == DMA_TO_DEVICE)
-		return IOMMU_PROT_IR;
-	else if (direction == DMA_FROM_DEVICE)
-		return IOMMU_PROT_IW;
-	else if (direction == DMA_BIDIRECTIONAL)
-		return IOMMU_PROT_IW | IOMMU_PROT_IR;
-	else
-		return 0;
-}
-
-/*
- * This function contains common code for mapping of a physically
- * contiguous memory region into DMA address space. It is used by all
- * mapping functions provided with this IOMMU driver.
- * Must be called with the domain lock held.
- */
-static dma_addr_t __map_single(struct device *dev,
-			       struct dma_ops_domain *dma_dom,
-			       phys_addr_t paddr,
-			       size_t size,
-			       enum dma_data_direction direction,
-			       u64 dma_mask)
-{
-	dma_addr_t offset = paddr & ~PAGE_MASK;
-	dma_addr_t address, start, ret;
-	unsigned int pages;
-	int prot = 0;
-	int i;
-
-	pages = iommu_num_pages(paddr, size, PAGE_SIZE);
-	paddr &= PAGE_MASK;
-
-	address = dma_ops_alloc_iova(dev, dma_dom, pages, dma_mask);
-	if (!address)
-		goto out;
-
-	prot = dir2prot(direction);
-
-	start = address;
-	for (i = 0; i < pages; ++i) {
-		ret = iommu_map_page(&dma_dom->domain, start, paddr,
-				     PAGE_SIZE, prot, GFP_ATOMIC);
-		if (ret)
-			goto out_unmap;
-
-		paddr += PAGE_SIZE;
-		start += PAGE_SIZE;
-	}
-	address += offset;
-
-	domain_flush_np_cache(&dma_dom->domain, address, size);
-
-out:
-	return address;
-
-out_unmap:
-
-	for (--i; i >= 0; --i) {
-		start -= PAGE_SIZE;
-		iommu_unmap_page(&dma_dom->domain, start, PAGE_SIZE);
-	}
-
-	domain_flush_tlb(&dma_dom->domain);
-	domain_flush_complete(&dma_dom->domain);
-
-	dma_ops_free_iova(dma_dom, address, pages);
-
-	return DMA_MAPPING_ERROR;
-}
-
-/*
- * Does the reverse of the __map_single function. Must be called with
- * the domain lock held too
- */
-static void __unmap_single(struct dma_ops_domain *dma_dom,
-			   dma_addr_t dma_addr,
-			   size_t size,
-			   int dir)
-{
-	dma_addr_t i, start;
-	unsigned int pages;
-
-	pages = iommu_num_pages(dma_addr, size, PAGE_SIZE);
-	dma_addr &= PAGE_MASK;
-	start = dma_addr;
-
-	for (i = 0; i < pages; ++i) {
-		iommu_unmap_page(&dma_dom->domain, start, PAGE_SIZE);
-		start += PAGE_SIZE;
-	}
-
-	if (amd_iommu_unmap_flush) {
-		domain_flush_tlb(&dma_dom->domain);
-		domain_flush_complete(&dma_dom->domain);
-		dma_ops_free_iova(dma_dom, dma_addr, pages);
-	} else {
-		pages = __roundup_pow_of_two(pages);
-		queue_iova(&dma_dom->iovad, dma_addr >> PAGE_SHIFT, pages, 0);
-	}
-}
-
-/*
- * The exported map_single function for dma_ops.
- */
-static dma_addr_t map_page(struct device *dev, struct page *page,
-			   unsigned long offset, size_t size,
-			   enum dma_data_direction dir,
-			   unsigned long attrs)
-{
-	phys_addr_t paddr = page_to_phys(page) + offset;
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-	u64 dma_mask;
-
-	domain = get_domain(dev);
-	if (PTR_ERR(domain) == -EINVAL)
-		return (dma_addr_t)paddr;
-	else if (IS_ERR(domain))
-		return DMA_MAPPING_ERROR;
-
-	dma_mask = *dev->dma_mask;
-	dma_dom = to_dma_ops_domain(domain);
-
-	return __map_single(dev, dma_dom, paddr, size, dir, dma_mask);
-}
-
-/*
- * The exported unmap_single function for dma_ops.
- */
-static void unmap_page(struct device *dev, dma_addr_t dma_addr, size_t size,
-		       enum dma_data_direction dir, unsigned long attrs)
-{
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-
-	domain = get_domain(dev);
-	if (IS_ERR(domain))
-		return;
-
-	dma_dom = to_dma_ops_domain(domain);
-
-	__unmap_single(dma_dom, dma_addr, size, dir);
-}
-
-static int sg_num_pages(struct device *dev,
-			struct scatterlist *sglist,
-			int nelems)
-{
-	unsigned long mask, boundary_size;
-	struct scatterlist *s;
-	int i, npages = 0;
-
-	mask          = dma_get_seg_boundary(dev);
-	boundary_size = mask + 1 ? ALIGN(mask + 1, PAGE_SIZE) >> PAGE_SHIFT :
-				   1UL << (BITS_PER_LONG - PAGE_SHIFT);
-
-	for_each_sg(sglist, s, nelems, i) {
-		int p, n;
-
-		s->dma_address = npages << PAGE_SHIFT;
-		p = npages % boundary_size;
-		n = iommu_num_pages(sg_phys(s), s->length, PAGE_SIZE);
-		if (p + n > boundary_size)
-			npages += boundary_size - p;
-		npages += n;
-	}
-
-	return npages;
-}
-
-/*
- * The exported map_sg function for dma_ops (handles scatter-gather
- * lists).
- */
-static int map_sg(struct device *dev, struct scatterlist *sglist,
-		  int nelems, enum dma_data_direction direction,
-		  unsigned long attrs)
-{
-	int mapped_pages = 0, npages = 0, prot = 0, i;
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-	struct scatterlist *s;
-	unsigned long address;
-	u64 dma_mask;
-	int ret;
-
-	domain = get_domain(dev);
-	if (IS_ERR(domain))
-		return 0;
-
-	dma_dom  = to_dma_ops_domain(domain);
-	dma_mask = *dev->dma_mask;
-
-	npages = sg_num_pages(dev, sglist, nelems);
-
-	address = dma_ops_alloc_iova(dev, dma_dom, npages, dma_mask);
-	if (address == DMA_MAPPING_ERROR)
-		goto out_err;
-
-	prot = dir2prot(direction);
-
-	/* Map all sg entries */
-	for_each_sg(sglist, s, nelems, i) {
-		int j, pages = iommu_num_pages(sg_phys(s), s->length, PAGE_SIZE);
-
-		for (j = 0; j < pages; ++j) {
-			unsigned long bus_addr, phys_addr;
-
-			bus_addr  = address + s->dma_address + (j << PAGE_SHIFT);
-			phys_addr = (sg_phys(s) & PAGE_MASK) + (j << PAGE_SHIFT);
-			ret = iommu_map_page(domain, bus_addr, phys_addr, PAGE_SIZE, prot, GFP_ATOMIC);
-			if (ret)
-				goto out_unmap;
-
-			mapped_pages += 1;
-		}
-	}
-
-	/* Everything is mapped - write the right values into s->dma_address */
-	for_each_sg(sglist, s, nelems, i) {
-		/*
-		 * Add in the remaining piece of the scatter-gather offset that
-		 * was masked out when we were determining the physical address
-		 * via (sg_phys(s) & PAGE_MASK) earlier.
-		 */
-		s->dma_address += address + (s->offset & ~PAGE_MASK);
-		s->dma_length   = s->length;
-	}
-
-	domain_flush_np_cache(domain, s->dma_address, s->dma_length);
-
-	return nelems;
-
-out_unmap:
-	dev_err(dev, "IOMMU mapping error in map_sg (io-pages: %d reason: %d)\n",
-		npages, ret);
-
-	for_each_sg(sglist, s, nelems, i) {
-		int j, pages = iommu_num_pages(sg_phys(s), s->length, PAGE_SIZE);
-
-		for (j = 0; j < pages; ++j) {
-			unsigned long bus_addr;
-
-			bus_addr  = address + s->dma_address + (j << PAGE_SHIFT);
-			iommu_unmap_page(domain, bus_addr, PAGE_SIZE);
-
-			if (--mapped_pages == 0)
-				goto out_free_iova;
-		}
-	}
-
-out_free_iova:
-	free_iova_fast(&dma_dom->iovad, address >> PAGE_SHIFT, npages);
-
-out_err:
-	return 0;
-}
-
-/*
- * The exported map_sg function for dma_ops (handles scatter-gather
- * lists).
- */
-static void unmap_sg(struct device *dev, struct scatterlist *sglist,
-		     int nelems, enum dma_data_direction dir,
-		     unsigned long attrs)
-{
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-	unsigned long startaddr;
-	int npages = 2;
-
-	domain = get_domain(dev);
-	if (IS_ERR(domain))
-		return;
-
-	startaddr = sg_dma_address(sglist) & PAGE_MASK;
-	dma_dom   = to_dma_ops_domain(domain);
-	npages    = sg_num_pages(dev, sglist, nelems);
-
-	__unmap_single(dma_dom, startaddr, npages << PAGE_SHIFT, dir);
-}
-
-/*
- * The exported alloc_coherent function for dma_ops.
- */
-static void *alloc_coherent(struct device *dev, size_t size,
-			    dma_addr_t *dma_addr, gfp_t flag,
-			    unsigned long attrs)
-{
-	u64 dma_mask = dev->coherent_dma_mask;
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-	struct page *page;
-
-	domain = get_domain(dev);
-	if (PTR_ERR(domain) == -EINVAL) {
-		page = alloc_pages(flag, get_order(size));
-		*dma_addr = page_to_phys(page);
-		return page_address(page);
-	} else if (IS_ERR(domain))
-		return NULL;
-
-	dma_dom   = to_dma_ops_domain(domain);
-	size	  = PAGE_ALIGN(size);
-	dma_mask  = dev->coherent_dma_mask;
-	flag     &= ~(__GFP_DMA | __GFP_HIGHMEM | __GFP_DMA32);
-	flag     |= __GFP_ZERO;
-
-	page = alloc_pages(flag | __GFP_NOWARN,  get_order(size));
-	if (!page) {
-		if (!gfpflags_allow_blocking(flag))
-			return NULL;
-
-		page = dma_alloc_from_contiguous(dev, size >> PAGE_SHIFT,
-					get_order(size), flag & __GFP_NOWARN);
-		if (!page)
-			return NULL;
-	}
-
-	if (!dma_mask)
-		dma_mask = *dev->dma_mask;
-
-	*dma_addr = __map_single(dev, dma_dom, page_to_phys(page),
-				 size, DMA_BIDIRECTIONAL, dma_mask);
-
-	if (*dma_addr == DMA_MAPPING_ERROR)
-		goto out_free;
-
-	return page_address(page);
-
-out_free:
-
-	if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT))
-		__free_pages(page, get_order(size));
-
-	return NULL;
-}
-
-/*
- * The exported free_coherent function for dma_ops.
- */
-static void free_coherent(struct device *dev, size_t size,
-			  void *virt_addr, dma_addr_t dma_addr,
-			  unsigned long attrs)
-{
-	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
-	struct page *page;
-
-	page = virt_to_page(virt_addr);
-	size = PAGE_ALIGN(size);
-
-	domain = get_domain(dev);
-	if (IS_ERR(domain))
-		goto free_mem;
-
-	dma_dom = to_dma_ops_domain(domain);
-
-	__unmap_single(dma_dom, dma_addr, size, DMA_BIDIRECTIONAL);
-
-free_mem:
-	if (!dma_release_from_contiguous(dev, page, size >> PAGE_SHIFT))
-		__free_pages(page, get_order(size));
-}
-
-/*
- * This function is called by the DMA layer to find out if we can handle a
- * particular device. It is part of the dma_ops.
- */
-static int amd_iommu_dma_supported(struct device *dev, u64 mask)
-{
-	if (!dma_direct_supported(dev, mask))
-		return 0;
-	return check_device(dev);
-}
-
-static const struct dma_map_ops amd_iommu_dma_ops = {
-	.alloc		= alloc_coherent,
-	.free		= free_coherent,
-	.map_page	= map_page,
-	.unmap_page	= unmap_page,
-	.map_sg		= map_sg,
-	.unmap_sg	= unmap_sg,
-	.dma_supported	= amd_iommu_dma_supported,
-};
-
-static int init_reserved_iova_ranges(void)
-{
-	struct pci_dev *pdev = NULL;
-	struct iova *val;
-
-	init_iova_domain(&reserved_iova_ranges, PAGE_SIZE, IOVA_START_PFN);
-
-	lockdep_set_class(&reserved_iova_ranges.iova_rbtree_lock,
-			  &reserved_rbtree_key);
-
-	/* MSI memory range */
-	val = reserve_iova(&reserved_iova_ranges,
-			   IOVA_PFN(MSI_RANGE_START), IOVA_PFN(MSI_RANGE_END));
-	if (!val) {
-		pr_err("Reserving MSI range failed\n");
-		return -ENOMEM;
-	}
-
-	/* HT memory range */
-	val = reserve_iova(&reserved_iova_ranges,
-			   IOVA_PFN(HT_RANGE_START), IOVA_PFN(HT_RANGE_END));
-	if (!val) {
-		pr_err("Reserving HT range failed\n");
-		return -ENOMEM;
-	}
-
-	/*
-	 * Memory used for PCI resources
-	 * FIXME: Check whether we can reserve the PCI-hole completly
-	 */
-	for_each_pci_dev(pdev) {
-		int i;
-
-		for (i = 0; i < PCI_NUM_RESOURCES; ++i) {
-			struct resource *r = &pdev->resource[i];
-
-			if (!(r->flags & IORESOURCE_MEM))
-				continue;
-
-			val = reserve_iova(&reserved_iova_ranges,
-					   IOVA_PFN(r->start),
-					   IOVA_PFN(r->end));
-			if (!val) {
-				pci_err(pdev, "Reserve pci-resource range %pR failed\n", r);
-				return -ENOMEM;
-			}
-		}
-	}
-
-	return 0;
-}
-
 int __init amd_iommu_init_api(void)
 {
 	int ret, err = 0;
@@ -2841,10 +2305,6 @@ int __init amd_iommu_init_api(void)
 	if (ret)
 		return ret;
 
-	ret = init_reserved_iova_ranges();
-	if (ret)
-		return ret;
-
 	err = bus_set_iommu(&pci_bus_type, &amd_iommu_ops);
 	if (err)
 		return err;
@@ -2950,7 +2410,6 @@ static struct protection_domain *protection_domain_alloc(void)
 static struct iommu_domain *amd_iommu_domain_alloc(unsigned type)
 {
 	struct protection_domain *pdomain;
-	struct dma_ops_domain *dma_domain;
 
 	switch (type) {
 	case IOMMU_DOMAIN_UNMANAGED:
@@ -2971,12 +2430,11 @@ static struct iommu_domain *amd_iommu_domain_alloc(unsigned type)
 
 		break;
 	case IOMMU_DOMAIN_DMA:
-		dma_domain = dma_ops_domain_alloc();
-		if (!dma_domain) {
+		pdomain = dma_ops_domain_alloc();
+		if (!pdomain) {
 			pr_err("Failed to allocate\n");
 			return NULL;
 		}
-		pdomain = &dma_domain->domain;
 		break;
 	case IOMMU_DOMAIN_IDENTITY:
 		pdomain = protection_domain_alloc();
@@ -2995,7 +2453,6 @@ static struct iommu_domain *amd_iommu_domain_alloc(unsigned type)
 static void amd_iommu_domain_free(struct iommu_domain *dom)
 {
 	struct protection_domain *domain;
-	struct dma_ops_domain *dma_dom;
 
 	domain = to_pdomain(dom);
 
@@ -3010,8 +2467,7 @@ static void amd_iommu_domain_free(struct iommu_domain *dom)
 	switch (dom->type) {
 	case IOMMU_DOMAIN_DMA:
 		/* Now release the domain */
-		dma_dom = to_dma_ops_domain(domain);
-		dma_ops_domain_free(dma_dom);
+		dma_ops_domain_free(domain);
 		break;
 	default:
 		if (domain->mode != PAGE_MODE_NONE)
@@ -3067,6 +2523,7 @@ static int amd_iommu_attach_device(struct iommu_domain *dom,
 		return -EINVAL;
 
 	dev_data = dev->archdata.iommu;
+	dev_data->defer_attach = false;
 
 	iommu = amd_iommu_rlookup_table[dev_data->devid];
 	if (!iommu)
@@ -3223,19 +2680,6 @@ static void amd_iommu_put_resv_regions(struct device *dev,
 		kfree(entry);
 }
 
-static void amd_iommu_apply_resv_region(struct device *dev,
-				      struct iommu_domain *domain,
-				      struct iommu_resv_region *region)
-{
-	struct dma_ops_domain *dma_dom = to_dma_ops_domain(to_pdomain(domain));
-	unsigned long start, end;
-
-	start = IOVA_PFN(region->start);
-	end   = IOVA_PFN(region->start + region->length - 1);
-
-	WARN_ON_ONCE(reserve_iova(&dma_dom->iovad, start, end) == NULL);
-}
-
 static bool amd_iommu_is_attach_deferred(struct iommu_domain *domain,
 					 struct device *dev)
 {
@@ -3268,9 +2712,9 @@ const struct iommu_ops amd_iommu_ops = {
 	.add_device = amd_iommu_add_device,
 	.remove_device = amd_iommu_remove_device,
 	.device_group = amd_iommu_device_group,
+	.domain_get_attr = amd_iommu_domain_get_attr,
 	.get_resv_regions = amd_iommu_get_resv_regions,
 	.put_resv_regions = amd_iommu_put_resv_regions,
-	.apply_resv_region = amd_iommu_apply_resv_region,
 	.is_attach_deferred = amd_iommu_is_attach_deferred,
 	.pgsize_bitmap	= AMD_IOMMU_PGSIZES,
 	.flush_iotlb_all = amd_iommu_flush_iotlb_all,
@@ -3585,9 +3029,23 @@ EXPORT_SYMBOL(amd_iommu_complete_ppr);
 struct iommu_domain *amd_iommu_get_v2_domain(struct pci_dev *pdev)
 {
 	struct protection_domain *pdomain;
+	struct iommu_domain *io_domain;
+	struct device *dev = &pdev->dev;
+
+	if (!check_device(dev))
+		return NULL;
+
+	pdomain = get_dev_data(dev)->domain;
+	if (pdomain == NULL && get_dev_data(dev)->defer_attach) {
+		get_dev_data(dev)->defer_attach = false;
+		io_domain = iommu_get_domain_for_dev(dev);
+		pdomain = to_pdomain(io_domain);
+		attach_device(dev, pdomain);
+	}
+	if (pdomain == NULL)
+		return NULL;
 
-	pdomain = get_domain(&pdev->dev);
-	if (IS_ERR(pdomain))
+	if (!dma_ops_domain(pdomain))
 		return NULL;
 
 	/* Only return IOMMUv2 domains */
-- 
2.17.1