Merge branch 'p2pdma-mmio-6.19.v5' into for-6.19/block

Merge MMIO P2P DMA series from Leon:

"This patch series improves block layer and NVMe driver support for MMIO
 memory regions, particularly for peer-to-peer (P2P) DMA transfers that
 go through the host bridge.

 The series addresses a critical gap where P2P transfers through the
 host bridge (PCI_P2PDMA_MAP_THRU_HOST_BRIDGE) were not properly marked
 as MMIO memory, leading to potential issues with:

 - Inappropriate CPU cache synchronization operations on MMIO regions
 - Incorrect DMA mapping/unmapping that doesn't respect MMIO semantics
 - Missing IOMMU configuration for MMIO memory handling

 This work is extracted from the larger DMA physical API improvement
 series [1] and focuses specifically on block layer and NVMe
 requirements for MMIO memory support.

 [1] https://lore.kernel.org/all/cover.1757423202.git.leonro@nvidia.com/"

Link: https://lore.kernel.org/linux-block/20251114-block-with-mmio-v5-0-69d00f73d766@nvidia.com/
Signed-off-by: Jens Axboe <axboe@kernel.dk>

* p2pdma-mmio-6.19.v5:
  block-dma: properly take MMIO path
  nvme-pci: migrate to dma_map_phys instead of map_page

block/blk-mq-dma.c

@@ -92,8 +92,13 @@ static bool blk_dma_map_bus(struct blk_dma_iter *iter, struct phys_vec *vec)
 static bool blk_dma_map_direct(struct request *req, struct device *dma_dev,
 		struct blk_dma_iter *iter, struct phys_vec *vec)
 {
-	iter->addr = dma_map_page(dma_dev, phys_to_page(vec->paddr),
-			offset_in_page(vec->paddr), vec->len, rq_dma_dir(req));
+	unsigned int attrs = 0;
+
+	if (iter->p2pdma.map == PCI_P2PDMA_MAP_THRU_HOST_BRIDGE)
+		attrs |= DMA_ATTR_MMIO;
+
+	iter->addr = dma_map_phys(dma_dev, vec->paddr, vec->len,
+			rq_dma_dir(req), attrs);
 	if (dma_mapping_error(dma_dev, iter->addr)) {
 		iter->status = BLK_STS_RESOURCE;
 		return false;
@@ -108,14 +113,18 @@ static bool blk_rq_dma_map_iova(struct request *req, struct device *dma_dev,
 {
 	enum dma_data_direction dir = rq_dma_dir(req);
 	unsigned int mapped = 0;
+	unsigned int attrs = 0;
 	int error;
 
 	iter->addr = state->addr;
 	iter->len = dma_iova_size(state);
 
+	if (iter->p2pdma.map == PCI_P2PDMA_MAP_THRU_HOST_BRIDGE)
+		attrs |= DMA_ATTR_MMIO;
+
 	do {
 		error = dma_iova_link(dma_dev, state, vec->paddr, mapped,
-				vec->len, dir, 0);
+				vec->len, dir, attrs);
 		if (error)
 			break;
 		mapped += vec->len;
@@ -162,6 +171,7 @@ static bool blk_dma_map_iter_start(struct request *req, struct device *dma_dev,
 
 	memset(&iter->p2pdma, 0, sizeof(iter->p2pdma));
 	iter->status = BLK_STS_OK;
+	iter->p2pdma.map = PCI_P2PDMA_MAP_NONE;
 
 	/*
 	 * Grab the first segment ASAP because we'll need it to check for P2P
@@ -173,10 +183,6 @@ static bool blk_dma_map_iter_start(struct request *req, struct device *dma_dev,
 	switch (pci_p2pdma_state(&iter->p2pdma, dma_dev,
 				 phys_to_page(vec.paddr))) {
 	case PCI_P2PDMA_MAP_BUS_ADDR:
-		if (iter->iter.is_integrity)
-			bio_integrity(req->bio)->bip_flags |= BIP_P2P_DMA;
-		else
-			req->cmd_flags |= REQ_P2PDMA;
 		return blk_dma_map_bus(iter, &vec);
 	case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
 		/*

drivers/nvme/host/pci.c

@@ -260,8 +260,20 @@ enum nvme_iod_flags {
 	/* single segment dma mapping */
 	IOD_SINGLE_SEGMENT	= 1U << 2,
 
+	/* Data payload contains p2p memory */
+	IOD_DATA_P2P		= 1U << 3,
+
+	/* Metadata contains p2p memory */
+	IOD_META_P2P		= 1U << 4,
+
+	/* Data payload contains MMIO memory */
+	IOD_DATA_MMIO		= 1U << 5,
+
+	/* Metadata contains MMIO memory */
+	IOD_META_MMIO		= 1U << 6,
+
 	/* Metadata using non-coalesced MPTR */
-	IOD_SINGLE_META_SEGMENT	= 1U << 5,
+	IOD_SINGLE_META_SEGMENT	= 1U << 7,
 };
 
 struct nvme_dma_vec {
@@ -698,20 +710,20 @@ static void nvme_free_descriptors(struct request *req)
 	}
 }
 
-static void nvme_free_prps(struct request *req)
+static void nvme_free_prps(struct request *req, unsigned int attrs)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 	unsigned int i;
 
 	for (i = 0; i < iod->nr_dma_vecs; i++)
-		dma_unmap_page(nvmeq->dev->dev, iod->dma_vecs[i].addr,
-				iod->dma_vecs[i].len, rq_dma_dir(req));
+		dma_unmap_phys(nvmeq->dev->dev, iod->dma_vecs[i].addr,
+			       iod->dma_vecs[i].len, rq_dma_dir(req), attrs);
 	mempool_free(iod->dma_vecs, nvmeq->dev->dmavec_mempool);
 }
 
 static void nvme_free_sgls(struct request *req, struct nvme_sgl_desc *sge,
-		struct nvme_sgl_desc *sg_list)
+		struct nvme_sgl_desc *sg_list, unsigned int attrs)
 {
 	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 	enum dma_data_direction dir = rq_dma_dir(req);
@@ -720,22 +732,25 @@ static void nvme_free_sgls(struct request *req, struct nvme_sgl_desc *sge,
 	unsigned int i;
 
 	if (sge->type == (NVME_SGL_FMT_DATA_DESC << 4)) {
-		dma_unmap_page(dma_dev, le64_to_cpu(sge->addr), len, dir);
+		dma_unmap_phys(dma_dev, le64_to_cpu(sge->addr), len, dir,
+			       attrs);
 		return;
 	}
 
 	for (i = 0; i < len / sizeof(*sg_list); i++)
-		dma_unmap_page(dma_dev, le64_to_cpu(sg_list[i].addr),
-			le32_to_cpu(sg_list[i].length), dir);
+		dma_unmap_phys(dma_dev, le64_to_cpu(sg_list[i].addr),
+			le32_to_cpu(sg_list[i].length), dir, attrs);
 }
 
 static void nvme_unmap_metadata(struct request *req)
 {
 	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
+	enum pci_p2pdma_map_type map = PCI_P2PDMA_MAP_NONE;
 	enum dma_data_direction dir = rq_dma_dir(req);
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct device *dma_dev = nvmeq->dev->dev;
 	struct nvme_sgl_desc *sge = iod->meta_descriptor;
+	unsigned int attrs = 0;
 
 	if (iod->flags & IOD_SINGLE_META_SEGMENT) {
 		dma_unmap_page(dma_dev, iod->meta_dma,
@@ -744,13 +759,20 @@ static void nvme_unmap_metadata(struct request *req)
 		return;
 	}
 
-	if (!blk_rq_integrity_dma_unmap(req, dma_dev, &iod->meta_dma_state,
-					iod->meta_total_len)) {
+	if (iod->flags & IOD_META_P2P)
+		map = PCI_P2PDMA_MAP_BUS_ADDR;
+	else if (iod->flags & IOD_META_MMIO) {
+		map = PCI_P2PDMA_MAP_THRU_HOST_BRIDGE;
+		attrs |= DMA_ATTR_MMIO;
+	}
+
+	if (!blk_rq_dma_unmap(req, dma_dev, &iod->meta_dma_state,
+			      iod->meta_total_len, map)) {
 		if (nvme_pci_cmd_use_meta_sgl(&iod->cmd))
-			nvme_free_sgls(req, sge, &sge[1]);
+			nvme_free_sgls(req, sge, &sge[1], attrs);
 		else
-			dma_unmap_page(dma_dev, iod->meta_dma,
-				       iod->meta_total_len, dir);
+			dma_unmap_phys(dma_dev, iod->meta_dma,
+				       iod->meta_total_len, dir, attrs);
 	}
 
 	if (iod->meta_descriptor)
@@ -760,9 +782,11 @@ static void nvme_unmap_metadata(struct request *req)
 
 static void nvme_unmap_data(struct request *req)
 {
+	enum pci_p2pdma_map_type map = PCI_P2PDMA_MAP_NONE;
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 	struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
 	struct device *dma_dev = nvmeq->dev->dev;
+	unsigned int attrs = 0;
 
 	if (iod->flags & IOD_SINGLE_SEGMENT) {
 		static_assert(offsetof(union nvme_data_ptr, prp1) ==
@@ -772,12 +796,20 @@ static void nvme_unmap_data(struct request *req)
 		return;
 	}
 
-	if (!blk_rq_dma_unmap(req, dma_dev, &iod->dma_state, iod->total_len)) {
+	if (iod->flags & IOD_DATA_P2P)
+		map = PCI_P2PDMA_MAP_BUS_ADDR;
+	else if (iod->flags & IOD_DATA_MMIO) {
+		map = PCI_P2PDMA_MAP_THRU_HOST_BRIDGE;
+		attrs |= DMA_ATTR_MMIO;
+	}
+
+	if (!blk_rq_dma_unmap(req, dma_dev, &iod->dma_state, iod->total_len,
+			      map)) {
 		if (nvme_pci_cmd_use_sgl(&iod->cmd))
 			nvme_free_sgls(req, iod->descriptors[0],
-				       &iod->cmd.common.dptr.sgl);
+				       &iod->cmd.common.dptr.sgl, attrs);
 		else
-			nvme_free_prps(req);
+			nvme_free_prps(req, attrs);
 	}
 
 	if (iod->nr_descriptors)
@@ -1048,6 +1080,19 @@ static blk_status_t nvme_map_data(struct request *req)
 	if (!blk_rq_dma_map_iter_start(req, dev->dev, &iod->dma_state, &iter))
 		return iter.status;
 
+	switch (iter.p2pdma.map) {
+	case PCI_P2PDMA_MAP_BUS_ADDR:
+		iod->flags |= IOD_DATA_P2P;
+		break;
+	case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
+		iod->flags |= IOD_DATA_MMIO;
+		break;
+	case PCI_P2PDMA_MAP_NONE:
+		break;
+	default:
+		return BLK_STS_RESOURCE;
+	}
+
 	if (use_sgl == SGL_FORCED ||
 	    (use_sgl == SGL_SUPPORTED &&
 	     (sgl_threshold && nvme_pci_avg_seg_size(req) >= sgl_threshold)))
@@ -1070,6 +1115,19 @@ static blk_status_t nvme_pci_setup_meta_sgls(struct request *req)
 						&iod->meta_dma_state, &iter))
 		return iter.status;
 
+	switch (iter.p2pdma.map) {
+	case PCI_P2PDMA_MAP_BUS_ADDR:
+		iod->flags |= IOD_META_P2P;
+		break;
+	case PCI_P2PDMA_MAP_THRU_HOST_BRIDGE:
+		iod->flags |= IOD_META_MMIO;
+		break;
+	case PCI_P2PDMA_MAP_NONE:
+		break;
+	default:
+		return BLK_STS_RESOURCE;
+	}
+
 	if (blk_rq_dma_map_coalesce(&iod->meta_dma_state))
 		entries = 1;
 

include/linux/bio-integrity.h

@@ -13,7 +13,6 @@ enum bip_flags {
 	BIP_CHECK_GUARD		= 1 << 5, /* guard check */
 	BIP_CHECK_REFTAG	= 1 << 6, /* reftag check */
 	BIP_CHECK_APPTAG	= 1 << 7, /* apptag check */
-	BIP_P2P_DMA		= 1 << 8, /* using P2P address */
 
 	BIP_MEMPOOL		= 1 << 15, /* buffer backed by mempool */
 };

include/linux/blk-integrity.h

@@ -33,14 +33,6 @@ static inline bool queue_limits_stack_integrity_bdev(struct queue_limits *t,
 #ifdef CONFIG_BLK_DEV_INTEGRITY
 int blk_rq_map_integrity_sg(struct request *, struct scatterlist *);
 
-static inline bool blk_rq_integrity_dma_unmap(struct request *req,
-		struct device *dma_dev, struct dma_iova_state *state,
-		size_t mapped_len)
-{
-	return blk_dma_unmap(req, dma_dev, state, mapped_len,
-			bio_integrity(req->bio)->bip_flags & BIP_P2P_DMA);
-}
-
 int blk_rq_count_integrity_sg(struct request_queue *, struct bio *);
 int blk_rq_integrity_map_user(struct request *rq, void __user *ubuf,
 			      ssize_t bytes);
@@ -129,12 +121,6 @@ static inline int blk_rq_map_integrity_sg(struct request *q,
 {
 	return 0;
 }
-static inline bool blk_rq_integrity_dma_unmap(struct request *req,
-		struct device *dma_dev, struct dma_iova_state *state,
-		size_t mapped_len)
-{
-	return false;
-}
 static inline int blk_rq_integrity_map_user(struct request *rq,
 					    void __user *ubuf,
 					    ssize_t bytes)

include/linux/blk-mq-dma.h

@@ -16,13 +16,13 @@ struct blk_dma_iter {
 	/* Output address range for this iteration */
 	dma_addr_t			addr;
 	u32				len;
+	struct pci_p2pdma_map_state	p2pdma;
 
 	/* Status code. Only valid when blk_rq_dma_map_iter_* returned false */
 	blk_status_t			status;
 
 	/* Internal to blk_rq_dma_map_iter_* */
 	struct blk_map_iter		iter;
-	struct pci_p2pdma_map_state	p2pdma;
 };
 
 bool blk_rq_dma_map_iter_start(struct request *req, struct device *dma_dev,
@@ -43,36 +43,34 @@ static inline bool blk_rq_dma_map_coalesce(struct dma_iova_state *state)
 }
 
 /**
- * blk_dma_unmap - try to DMA unmap a request
+ * blk_rq_dma_unmap - try to DMA unmap a request
  * @req:	request to unmap
  * @dma_dev:	device to unmap from
  * @state:	DMA IOVA state
  * @mapped_len: number of bytes to unmap
- * @is_p2p:	true if mapped with PCI_P2PDMA_MAP_BUS_ADDR
+ * @map:	peer-to-peer mapping type
  *
  * Returns %false if the callers need to manually unmap every DMA segment
  * mapped using @iter or %true if no work is left to be done.
  */
-static inline bool blk_dma_unmap(struct request *req, struct device *dma_dev,
-		struct dma_iova_state *state, size_t mapped_len, bool is_p2p)
+static inline bool blk_rq_dma_unmap(struct request *req, struct device *dma_dev,
+		struct dma_iova_state *state, size_t mapped_len,
+		enum pci_p2pdma_map_type map)
 {
-	if (is_p2p)
+	if (map == PCI_P2PDMA_MAP_BUS_ADDR)
 		return true;
 
 	if (dma_use_iova(state)) {
+		unsigned int attrs = 0;
+
+		if (map == PCI_P2PDMA_MAP_THRU_HOST_BRIDGE)
+			attrs |= DMA_ATTR_MMIO;
+
 		dma_iova_destroy(dma_dev, state, mapped_len, rq_dma_dir(req),
-				 0);
+				 attrs);
 		return true;
 	}
 
 	return !dma_need_unmap(dma_dev);
 }
-
-static inline bool blk_rq_dma_unmap(struct request *req, struct device *dma_dev,
-		struct dma_iova_state *state, size_t mapped_len)
-{
-	return blk_dma_unmap(req, dma_dev, state, mapped_len,
-				req->cmd_flags & REQ_P2PDMA);
-}
-
 #endif /* BLK_MQ_DMA_H */

include/linux/blk_types.h

@@ -393,7 +393,6 @@ enum req_flag_bits {
 	__REQ_DRV,		/* for driver use */
 	__REQ_FS_PRIVATE,	/* for file system (submitter) use */
 	__REQ_ATOMIC,		/* for atomic write operations */
-	__REQ_P2PDMA,		/* contains P2P DMA pages */
 	/*
 	 * Command specific flags, keep last:
 	 */
@@ -426,7 +425,6 @@ enum req_flag_bits {
 #define REQ_DRV		(__force blk_opf_t)(1ULL << __REQ_DRV)
 #define REQ_FS_PRIVATE	(__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE)
 #define REQ_ATOMIC	(__force blk_opf_t)(1ULL << __REQ_ATOMIC)
-#define REQ_P2PDMA	(__force blk_opf_t)(1ULL << __REQ_P2PDMA)
 
 #define REQ_NOUNMAP	(__force blk_opf_t)(1ULL << __REQ_NOUNMAP)