DPDK patches and discussions
 help / color / mirror / Atom feed
From: xuan.ding@intel.com
To: maxime.coquelin@redhat.com, chenbo.xia@intel.com
Cc: dev@dpdk.org, jiayu.hu@intel.com, yuanx.wang@intel.com,
	Xuan Ding <xuan.ding@intel.com>
Subject: [PATCH v2 2/2] vhost: fix physical address mapping
Date: Wed, 19 Jan 2022 15:10:16 +0000	[thread overview]
Message-ID: <20220119151016.9970-3-xuan.ding@intel.com> (raw)
In-Reply-To: <20220119151016.9970-1-xuan.ding@intel.com>

From: Xuan Ding <xuan.ding@intel.com>

When choosing IOVA as PA mode, IOVA is likely to be discontinuous,
which requires page by page mapping for DMA devices. To be consistent,
this patch implements page by page mapping instead of mapping at the
region granularity for both IOVA as VA and PA mode.

Fixes: 7c61fa08b716 ("vhost: enable IOMMU for async vhost")

Signed-off-by: Xuan Ding <xuan.ding@intel.com>
Signed-off-by: Yuan Wang <yuanx.wang@intel.com>
---
 lib/vhost/vhost.h      |   1 +
 lib/vhost/vhost_user.c | 116 ++++++++++++++++++++---------------------
 2 files changed, 57 insertions(+), 60 deletions(-)

diff --git a/lib/vhost/vhost.h b/lib/vhost/vhost.h
index ca7f58039d..9521ae56da 100644
--- a/lib/vhost/vhost.h
+++ b/lib/vhost/vhost.h
@@ -355,6 +355,7 @@ struct vring_packed_desc_event {
 struct guest_page {
 	uint64_t guest_phys_addr;
 	uint64_t host_iova;
+	uint64_t host_user_addr;
 	uint64_t size;
 };
 
diff --git a/lib/vhost/vhost_user.c b/lib/vhost/vhost_user.c
index 95c9df697e..48c08716ba 100644
--- a/lib/vhost/vhost_user.c
+++ b/lib/vhost/vhost_user.c
@@ -143,57 +143,56 @@ get_blk_size(int fd)
 	return ret == -1 ? (uint64_t)-1 : (uint64_t)stat.st_blksize;
 }
 
-static int
-async_dma_map(struct rte_vhost_mem_region *region, bool do_map)
+static void
+async_dma_map(struct virtio_net *dev, bool do_map)
 {
-	uint64_t host_iova;
 	int ret = 0;
-
-	host_iova = rte_mem_virt2iova((void *)(uintptr_t)region->host_user_addr);
+	uint32_t i;
+	struct guest_page *page;
 	if (do_map) {
-		/* Add mapped region into the default container of DPDK. */
-		ret = rte_vfio_container_dma_map(RTE_VFIO_DEFAULT_CONTAINER_FD,
-						 region->host_user_addr,
-						 host_iova,
-						 region->size);
-		if (ret) {
-			/*
-			 * DMA device may bind with kernel driver, in this case,
-			 * we don't need to program IOMMU manually. However, if no
-			 * device is bound with vfio/uio in DPDK, and vfio kernel
-			 * module is loaded, the API will still be called and return
-			 * with ENODEV/ENOSUP.
-			 *
-			 * DPDK vfio only returns ENODEV/ENOSUP in very similar
-			 * situations(vfio either unsupported, or supported
-			 * but no devices found). Either way, no mappings could be
-			 * performed. We treat it as normal case in async path.
-			 */
-			if (rte_errno == ENODEV || rte_errno == ENOTSUP)
-				return 0;
-
-			VHOST_LOG_CONFIG(ERR, "DMA engine map failed\n");
-			/* DMA mapping errors won't stop VHST_USER_SET_MEM_TABLE. */
-			return 0;
+		for (i = 0; i < dev->nr_guest_pages; i++) {
+			page = &dev->guest_pages[i];
+			ret = rte_vfio_container_dma_map(RTE_VFIO_DEFAULT_CONTAINER_FD,
+							 page->host_user_addr,
+							 page->host_iova,
+							 page->size);
+			if (ret) {
+				/*
+				 * DMA device may bind with kernel driver, in this case,
+				 * we don't need to program IOMMU manually. However, if no
+				 * device is bound with vfio/uio in DPDK, and vfio kernel
+				 * module is loaded, the API will still be called and return
+				 * with ENODEV.
+				 *
+				 * DPDK vfio only returns ENODEV in very similar situations
+				 * (vfio either unsupported, or supported but no devices found).
+				 * Either way, no mappings could be performed. We treat it as
+				 * normal case in async path. This is a workaround.
+				 */
+				if (rte_errno == ENODEV)
+					return;
+
+				/* DMA mapping errors won't stop VHOST_USER_SET_MEM_TABLE. */
+				VHOST_LOG_CONFIG(ERR, "DMA engine map failed\n");
+			}
 		}
 
 	} else {
-		/* Remove mapped region from the default container of DPDK. */
-		ret = rte_vfio_container_dma_unmap(RTE_VFIO_DEFAULT_CONTAINER_FD,
-						   region->host_user_addr,
-						   host_iova,
-						   region->size);
-		if (ret) {
-			/* like DMA map, ignore the kernel driver case when unmap. */
-			if (rte_errno == EINVAL)
-				return 0;
-
-			VHOST_LOG_CONFIG(ERR, "DMA engine unmap failed\n");
-			return ret;
+		for (i = 0; i < dev->nr_guest_pages; i++) {
+			page = &dev->guest_pages[i];
+			ret = rte_vfio_container_dma_unmap(RTE_VFIO_DEFAULT_CONTAINER_FD,
+							   page->host_user_addr,
+							   page->host_iova,
+							   page->size);
+			if (ret) {
+				/* like DMA map, ignore the kernel driver case when unmap. */
+				if (rte_errno == EINVAL)
+					return;
+
+				VHOST_LOG_CONFIG(ERR, "DMA engine unmap failed\n");
+			}
 		}
 	}
-
-	return ret;
 }
 
 static void
@@ -205,12 +204,12 @@ free_mem_region(struct virtio_net *dev)
 	if (!dev || !dev->mem)
 		return;
 
+	if (dev->async_copy && rte_vfio_is_enabled("vfio"))
+		async_dma_map(dev, false);
+
 	for (i = 0; i < dev->mem->nregions; i++) {
 		reg = &dev->mem->regions[i];
 		if (reg->host_user_addr) {
-			if (dev->async_copy && rte_vfio_is_enabled("vfio"))
-				async_dma_map(reg, false);
-
 			munmap(reg->mmap_addr, reg->mmap_size);
 			close(reg->fd);
 		}
@@ -978,7 +977,7 @@ vhost_user_set_vring_base(struct virtio_net **pdev,
 
 static int
 add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
-		   uint64_t host_iova, uint64_t size)
+		   uint64_t host_iova, uint64_t host_user_addr, uint64_t size)
 {
 	struct guest_page *page, *last_page;
 	struct guest_page *old_pages;
@@ -999,8 +998,9 @@ add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
 	if (dev->nr_guest_pages > 0) {
 		last_page = &dev->guest_pages[dev->nr_guest_pages - 1];
 		/* merge if the two pages are continuous */
-		if (host_iova == last_page->host_iova +
-				      last_page->size) {
+		if (host_iova == last_page->host_iova + last_page->size
+		&& guest_phys_addr == last_page->guest_phys_addr + last_page->size
+		&& host_user_addr == last_page->host_user_addr + last_page->size) {
 			last_page->size += size;
 			return 0;
 		}
@@ -1009,6 +1009,7 @@ add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,
 	page = &dev->guest_pages[dev->nr_guest_pages++];
 	page->guest_phys_addr = guest_phys_addr;
 	page->host_iova  = host_iova;
+	page->host_user_addr = host_user_addr;
 	page->size = size;
 
 	return 0;
@@ -1028,7 +1029,8 @@ add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,
 	size = page_size - (guest_phys_addr & (page_size - 1));
 	size = RTE_MIN(size, reg_size);
 
-	if (add_one_guest_page(dev, guest_phys_addr, host_iova, size) < 0)
+	if (add_one_guest_page(dev, guest_phys_addr, host_iova,
+			       host_user_addr, size) < 0)
 		return -1;
 
 	host_user_addr  += size;
@@ -1040,7 +1042,7 @@ add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,
 		host_iova = rte_mem_virt2iova((void *)(uintptr_t)
 						  host_user_addr);
 		if (add_one_guest_page(dev, guest_phys_addr, host_iova,
-				size) < 0)
+				       host_user_addr, size) < 0)
 			return -1;
 
 		host_user_addr  += size;
@@ -1215,7 +1217,6 @@ vhost_user_mmap_region(struct virtio_net *dev,
 	uint64_t mmap_size;
 	uint64_t alignment;
 	int populate;
-	int ret;
 
 	/* Check for memory_size + mmap_offset overflow */
 	if (mmap_offset >= -region->size) {
@@ -1274,14 +1275,6 @@ vhost_user_mmap_region(struct virtio_net *dev,
 			VHOST_LOG_CONFIG(ERR, "adding guest pages to region failed.\n");
 			return -1;
 		}
-
-		if (rte_vfio_is_enabled("vfio")) {
-			ret = async_dma_map(region, true);
-			if (ret) {
-				VHOST_LOG_CONFIG(ERR, "Configure IOMMU for DMA engine failed\n");
-				return -1;
-			}
-		}
 	}
 
 	VHOST_LOG_CONFIG(INFO,
@@ -1420,6 +1413,9 @@ vhost_user_set_mem_table(struct virtio_net **pdev, struct VhostUserMsg *msg,
 		dev->mem->nregions++;
 	}
 
+	if (dev->async_copy && rte_vfio_is_enabled("vfio"))
+		async_dma_map(dev, true);
+
 	if (vhost_user_postcopy_register(dev, main_fd, msg) < 0)
 		goto free_mem_table;
 
-- 
2.17.1


  parent reply	other threads:[~2022-01-19  6:36 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-01-19 15:10 [PATCH v2 0/2] vhost: fix async " xuan.ding
2022-01-19 15:10 ` [PATCH v2 1/2] vhost: rename field in guest page struct xuan.ding
2022-02-01  8:47   ` Maxime Coquelin
2022-01-19 15:10 ` xuan.ding [this message]
2022-02-01  8:51   ` [PATCH v2 2/2] vhost: fix physical address mapping Maxime Coquelin
2022-02-04 10:43   ` Maxime Coquelin
2022-02-01  8:28 ` [PATCH v2 0/2] vhost: fix async " Maxime Coquelin
2022-02-01 11:29   ` Kevin Traynor
2022-02-04 10:43     ` Maxime Coquelin
2022-02-04 10:56 ` Maxime Coquelin
2022-02-11  2:48   ` Ding, Xuan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220119151016.9970-3-xuan.ding@intel.com \
    --to=xuan.ding@intel.com \
    --cc=chenbo.xia@intel.com \
    --cc=dev@dpdk.org \
    --cc=jiayu.hu@intel.com \
    --cc=maxime.coquelin@redhat.com \
    --cc=yuanx.wang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).