Patch Detail
get:
Show a patch.
patch:
Update a patch.
put:
Update a patch.
GET /api/patches/105082/?format=api
{ "id": 105082, "url": "http://patches.dpdk.org/api/patches/105082/?format=api", "web_url": "http://patches.dpdk.org/project/dpdk/patch/20211213164111.60015-1-xuan.ding@intel.com/", "project": { "id": 1, "url": "http://patches.dpdk.org/api/projects/1/?format=api", "name": "DPDK", "link_name": "dpdk", "list_id": "dev.dpdk.org", "list_email": "dev@dpdk.org", "web_url": "http://core.dpdk.org", "scm_url": "git://dpdk.org/dpdk", "webscm_url": "http://git.dpdk.org/dpdk", "list_archive_url": "https://inbox.dpdk.org/dev", "list_archive_url_format": "https://inbox.dpdk.org/dev/{}", "commit_url_format": "" }, "msgid": "<20211213164111.60015-1-xuan.ding@intel.com>", "list_archive_url": "https://inbox.dpdk.org/dev/20211213164111.60015-1-xuan.ding@intel.com", "date": "2021-12-13T16:41:11", "name": "vhost: fix physical address mapping", "commit_ref": null, "pull_url": null, "state": "superseded", "archived": true, "hash": "d25cf264f4bf8521910c291613c5bf29b1e1b8ec", "submitter": { "id": 1401, "url": "http://patches.dpdk.org/api/people/1401/?format=api", "name": "Ding, Xuan", "email": "xuan.ding@intel.com" }, "delegate": { "id": 2642, "url": "http://patches.dpdk.org/api/users/2642/?format=api", "username": "mcoquelin", "first_name": "Maxime", "last_name": "Coquelin", "email": "maxime.coquelin@redhat.com" }, "mbox": "http://patches.dpdk.org/project/dpdk/patch/20211213164111.60015-1-xuan.ding@intel.com/mbox/", "series": [ { "id": 20920, "url": "http://patches.dpdk.org/api/series/20920/?format=api", "web_url": "http://patches.dpdk.org/project/dpdk/list/?series=20920", "date": "2021-12-13T16:41:11", "name": "vhost: fix physical address mapping", "version": 1, "mbox": "http://patches.dpdk.org/series/20920/mbox/" } ], "comments": "http://patches.dpdk.org/api/patches/105082/comments/", "check": "success", "checks": "http://patches.dpdk.org/api/patches/105082/checks/", "tags": {}, "related": [], "headers": { "Return-Path": "<dev-bounces@dpdk.org>", "X-Original-To": "patchwork@inbox.dpdk.org", "Delivered-To": "patchwork@inbox.dpdk.org", "Received": [ "from mails.dpdk.org (mails.dpdk.org [217.70.189.124])\n\tby inbox.dpdk.org (Postfix) with ESMTP id 05DDCA0032;\n\tMon, 13 Dec 2021 09:02:40 +0100 (CET)", "from [217.70.189.124] (localhost [127.0.0.1])\n\tby mails.dpdk.org (Postfix) with ESMTP id C71A940140;\n\tMon, 13 Dec 2021 09:02:39 +0100 (CET)", "from mga06.intel.com (mga06.intel.com [134.134.136.31])\n by mails.dpdk.org (Postfix) with ESMTP id 638F640042\n for <dev@dpdk.org>; Mon, 13 Dec 2021 09:02:38 +0100 (CET)", "from orsmga008.jf.intel.com ([10.7.209.65])\n by orsmga104.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384;\n 13 Dec 2021 00:02:37 -0800", "from npg-dpdk-xuan-cbdma.sh.intel.com ([10.67.110.228])\n by orsmga008.jf.intel.com with ESMTP; 13 Dec 2021 00:02:33 -0800" ], "DKIM-Signature": "v=1; a=rsa-sha256; c=simple/simple;\n d=intel.com; i=@intel.com; q=dns/txt; s=Intel;\n t=1639382558; x=1670918558;\n h=from:to:cc:subject:date:message-id;\n bh=A9RDeI1day9X5KlQzM2Vs2a1EVdrVbvtYKeSlVTOZHA=;\n b=lviHpBZxWAcSp36MCwKfbkWPZF1zNfq5pZycxtmjMzyIDnwW7RmHcsEI\n MEidc1jn9TcOODto6KEwNvhJ5g0lE3mmgTQGb1qFmHEFEtjcM7WEFdwfS\n HLvLJU3QfnjZFvhKvCNN2jAC0+rwrWoTgMhf8g+OHXOgg41wmC5Hvp6kh\n nP59ef7RB+4B9EswvJC9TPbU3JuQfCGNZcOuOIHUquG0JTchVIBSL2IiK\n QNCWu2XCe3YSkvnacTApij2wshdQVyIMkoZ9PM29l2kBuU/OJzWH634+l\n dFD6CU0IKCTMTSGJz4oWS8+om9P9gvYcmnlziZjC2UnWSwx4VlEehmSB4 Q==;", "X-IronPort-AV": [ "E=McAfee;i=\"6200,9189,10196\"; a=\"299465290\"", "E=Sophos;i=\"5.88,202,1635231600\"; d=\"scan'208\";a=\"299465290\"", "E=Sophos;i=\"5.88,202,1635231600\"; d=\"scan'208\";a=\"517650706\"" ], "X-ExtLoop1": "1", "From": "xuan.ding@intel.com", "To": "maxime.coquelin@redhat.com,\n\tchenbo.xia@intel.com", "Cc": "dev@dpdk.org, yuanx.wang@intel.com, jiayu.hu@intel.com,\n yinan.wang@intel.com, xingguang.he@intel.com,\n Xuan Ding <xuan.ding@intel.com>", "Subject": "[PATCH] vhost: fix physical address mapping", "Date": "Mon, 13 Dec 2021 16:41:11 +0000", "Message-Id": "<20211213164111.60015-1-xuan.ding@intel.com>", "X-Mailer": "git-send-email 2.17.1", "X-BeenThere": "dev@dpdk.org", "X-Mailman-Version": "2.1.29", "Precedence": "list", "List-Id": "DPDK patches and discussions <dev.dpdk.org>", "List-Unsubscribe": "<https://mails.dpdk.org/options/dev>,\n <mailto:dev-request@dpdk.org?subject=unsubscribe>", "List-Archive": "<http://mails.dpdk.org/archives/dev/>", "List-Post": "<mailto:dev@dpdk.org>", "List-Help": "<mailto:dev-request@dpdk.org?subject=help>", "List-Subscribe": "<https://mails.dpdk.org/listinfo/dev>,\n <mailto:dev-request@dpdk.org?subject=subscribe>", "Errors-To": "dev-bounces@dpdk.org" }, "content": "From: Xuan Ding <xuan.ding@intel.com>\n\nWhen choosing IOVA as PA mode, IOVA is likely to be discontinuous,\nwhich requires page by page mapping for DMA devices. To be consistent,\nthis patch implements page by page mapping instead of mapping at the\nregion granularity for both IOVA as VA and PA mode.\n\nFixes: 7c61fa08b716 (\"vhost: enable IOMMU for async vhost\")\n\nSigned-off-by: Xuan Ding <xuan.ding@intel.com>\nSigned-off-by: Yuan Wang <yuanx.wang@intel.com>\n---\n lib/vhost/vhost.h | 1 +\n lib/vhost/vhost_user.c | 116 ++++++++++++++++++++---------------------\n 2 files changed, 57 insertions(+), 60 deletions(-)", "diff": "diff --git a/lib/vhost/vhost.h b/lib/vhost/vhost.h\nindex 7085e0885c..d246538ca5 100644\n--- a/lib/vhost/vhost.h\n+++ b/lib/vhost/vhost.h\n@@ -355,6 +355,7 @@ struct vring_packed_desc_event {\n struct guest_page {\n \tuint64_t guest_phys_addr;\n \tuint64_t host_phys_addr;\n+\tuint64_t host_user_addr;\n \tuint64_t size;\n };\n \ndiff --git a/lib/vhost/vhost_user.c b/lib/vhost/vhost_user.c\nindex a781346c4d..6d888766b0 100644\n--- a/lib/vhost/vhost_user.c\n+++ b/lib/vhost/vhost_user.c\n@@ -143,57 +143,56 @@ get_blk_size(int fd)\n \treturn ret == -1 ? (uint64_t)-1 : (uint64_t)stat.st_blksize;\n }\n \n-static int\n-async_dma_map(struct rte_vhost_mem_region *region, bool do_map)\n+static void\n+async_dma_map(struct virtio_net *dev, bool do_map)\n {\n-\tuint64_t host_iova;\n \tint ret = 0;\n-\n-\thost_iova = rte_mem_virt2iova((void *)(uintptr_t)region->host_user_addr);\n+\tuint32_t i;\n+\tstruct guest_page *page;\n \tif (do_map) {\n-\t\t/* Add mapped region into the default container of DPDK. */\n-\t\tret = rte_vfio_container_dma_map(RTE_VFIO_DEFAULT_CONTAINER_FD,\n-\t\t\t\t\t\t region->host_user_addr,\n-\t\t\t\t\t\t host_iova,\n-\t\t\t\t\t\t region->size);\n-\t\tif (ret) {\n-\t\t\t/*\n-\t\t\t * DMA device may bind with kernel driver, in this case,\n-\t\t\t * we don't need to program IOMMU manually. However, if no\n-\t\t\t * device is bound with vfio/uio in DPDK, and vfio kernel\n-\t\t\t * module is loaded, the API will still be called and return\n-\t\t\t * with ENODEV/ENOSUP.\n-\t\t\t *\n-\t\t\t * DPDK vfio only returns ENODEV/ENOSUP in very similar\n-\t\t\t * situations(vfio either unsupported, or supported\n-\t\t\t * but no devices found). Either way, no mappings could be\n-\t\t\t * performed. We treat it as normal case in async path.\n-\t\t\t */\n-\t\t\tif (rte_errno == ENODEV || rte_errno == ENOTSUP)\n-\t\t\t\treturn 0;\n-\n-\t\t\tVHOST_LOG_CONFIG(ERR, \"DMA engine map failed\\n\");\n-\t\t\t/* DMA mapping errors won't stop VHST_USER_SET_MEM_TABLE. */\n-\t\t\treturn 0;\n+\t\tfor (i = 0; i < dev->nr_guest_pages; i++) {\n+\t\t\tpage = &dev->guest_pages[i];\n+\t\t\tret = rte_vfio_container_dma_map(RTE_VFIO_DEFAULT_CONTAINER_FD,\n+\t\t\t\t\t\t\t page->host_user_addr,\n+\t\t\t\t\t\t\t page->host_phys_addr,\n+\t\t\t\t\t\t\t page->size);\n+\t\t\tif (ret) {\n+\t\t\t\t/*\n+\t\t\t\t * DMA device may bind with kernel driver, in this case,\n+\t\t\t\t * we don't need to program IOMMU manually. However, if no\n+\t\t\t\t * device is bound with vfio/uio in DPDK, and vfio kernel\n+\t\t\t\t * module is loaded, the API will still be called and return\n+\t\t\t\t * with ENODEV.\n+\t\t\t\t *\n+\t\t\t\t * DPDK vfio only returns ENODEV in very similar situations\n+\t\t\t\t * (vfio either unsupported, or supported but no devices found).\n+\t\t\t\t * Either way, no mappings could be performed. We treat it as\n+\t\t\t\t * normal case in async path. This is a workaround.\n+\t\t\t\t */\n+\t\t\t\tif (rte_errno == ENODEV)\n+\t\t\t\t\treturn;\n+\n+\t\t\t\t/* DMA mapping errors won't stop VHOST_USER_SET_MEM_TABLE. */\n+\t\t\t\tVHOST_LOG_CONFIG(ERR, \"DMA engine map failed\\n\");\n+\t\t\t}\n \t\t}\n \n \t} else {\n-\t\t/* Remove mapped region from the default container of DPDK. */\n-\t\tret = rte_vfio_container_dma_unmap(RTE_VFIO_DEFAULT_CONTAINER_FD,\n-\t\t\t\t\t\t region->host_user_addr,\n-\t\t\t\t\t\t host_iova,\n-\t\t\t\t\t\t region->size);\n-\t\tif (ret) {\n-\t\t\t/* like DMA map, ignore the kernel driver case when unmap. */\n-\t\t\tif (rte_errno == EINVAL)\n-\t\t\t\treturn 0;\n-\n-\t\t\tVHOST_LOG_CONFIG(ERR, \"DMA engine unmap failed\\n\");\n-\t\t\treturn ret;\n+\t\tfor (i = 0; i < dev->nr_guest_pages; i++) {\n+\t\t\tpage = &dev->guest_pages[i];\n+\t\t\tret = rte_vfio_container_dma_unmap(RTE_VFIO_DEFAULT_CONTAINER_FD,\n+\t\t\t\t\t\t\t page->host_user_addr,\n+\t\t\t\t\t\t\t page->host_phys_addr,\n+\t\t\t\t\t\t\t page->size);\n+\t\t\tif (ret) {\n+\t\t\t\t/* like DMA map, ignore the kernel driver case when unmap. */\n+\t\t\t\tif (rte_errno == EINVAL)\n+\t\t\t\t\treturn;\n+\n+\t\t\t\tVHOST_LOG_CONFIG(ERR, \"DMA engine unmap failed\\n\");\n+\t\t\t}\n \t\t}\n \t}\n-\n-\treturn ret;\n }\n \n static void\n@@ -205,12 +204,12 @@ free_mem_region(struct virtio_net *dev)\n \tif (!dev || !dev->mem)\n \t\treturn;\n \n+\tif (dev->async_copy && rte_vfio_is_enabled(\"vfio\"))\n+\t\tasync_dma_map(dev, false);\n+\n \tfor (i = 0; i < dev->mem->nregions; i++) {\n \t\treg = &dev->mem->regions[i];\n \t\tif (reg->host_user_addr) {\n-\t\t\tif (dev->async_copy && rte_vfio_is_enabled(\"vfio\"))\n-\t\t\t\tasync_dma_map(reg, false);\n-\n \t\t\tmunmap(reg->mmap_addr, reg->mmap_size);\n \t\t\tclose(reg->fd);\n \t\t}\n@@ -978,7 +977,7 @@ vhost_user_set_vring_base(struct virtio_net **pdev,\n \n static int\n add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,\n-\t\t uint64_t host_phys_addr, uint64_t size)\n+\t\t uint64_t host_phys_addr, uint64_t host_user_addr, uint64_t size)\n {\n \tstruct guest_page *page, *last_page;\n \tstruct guest_page *old_pages;\n@@ -999,8 +998,9 @@ add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,\n \tif (dev->nr_guest_pages > 0) {\n \t\tlast_page = &dev->guest_pages[dev->nr_guest_pages - 1];\n \t\t/* merge if the two pages are continuous */\n-\t\tif (host_phys_addr == last_page->host_phys_addr +\n-\t\t\t\t last_page->size) {\n+\t\tif (host_phys_addr == last_page->host_phys_addr + last_page->size\n+\t\t && guest_phys_addr == last_page->guest_phys_addr + last_page->size\n+\t\t && host_user_addr == last_page->host_user_addr + last_page->size) {\n \t\t\tlast_page->size += size;\n \t\t\treturn 0;\n \t\t}\n@@ -1009,6 +1009,7 @@ add_one_guest_page(struct virtio_net *dev, uint64_t guest_phys_addr,\n \tpage = &dev->guest_pages[dev->nr_guest_pages++];\n \tpage->guest_phys_addr = guest_phys_addr;\n \tpage->host_phys_addr = host_phys_addr;\n+\tpage->host_user_addr = host_user_addr;\n \tpage->size = size;\n \n \treturn 0;\n@@ -1028,7 +1029,8 @@ add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,\n \tsize = page_size - (guest_phys_addr & (page_size - 1));\n \tsize = RTE_MIN(size, reg_size);\n \n-\tif (add_one_guest_page(dev, guest_phys_addr, host_phys_addr, size) < 0)\n+\tif (add_one_guest_page(dev, guest_phys_addr, host_phys_addr,\n+\t\t\t host_user_addr, size) < 0)\n \t\treturn -1;\n \n \thost_user_addr += size;\n@@ -1040,7 +1042,7 @@ add_guest_pages(struct virtio_net *dev, struct rte_vhost_mem_region *reg,\n \t\thost_phys_addr = rte_mem_virt2iova((void *)(uintptr_t)\n \t\t\t\t\t\t host_user_addr);\n \t\tif (add_one_guest_page(dev, guest_phys_addr, host_phys_addr,\n-\t\t\t\tsize) < 0)\n+\t\t\t\t host_user_addr, size) < 0)\n \t\t\treturn -1;\n \n \t\thost_user_addr += size;\n@@ -1215,7 +1217,6 @@ vhost_user_mmap_region(struct virtio_net *dev,\n \tuint64_t mmap_size;\n \tuint64_t alignment;\n \tint populate;\n-\tint ret;\n \n \t/* Check for memory_size + mmap_offset overflow */\n \tif (mmap_offset >= -region->size) {\n@@ -1274,14 +1275,6 @@ vhost_user_mmap_region(struct virtio_net *dev,\n \t\t\tVHOST_LOG_CONFIG(ERR, \"adding guest pages to region failed.\\n\");\n \t\t\treturn -1;\n \t\t}\n-\n-\t\tif (rte_vfio_is_enabled(\"vfio\")) {\n-\t\t\tret = async_dma_map(region, true);\n-\t\t\tif (ret) {\n-\t\t\t\tVHOST_LOG_CONFIG(ERR, \"Configure IOMMU for DMA engine failed\\n\");\n-\t\t\t\treturn -1;\n-\t\t\t}\n-\t\t}\n \t}\n \n \tVHOST_LOG_CONFIG(INFO,\n@@ -1420,6 +1413,9 @@ vhost_user_set_mem_table(struct virtio_net **pdev, struct VhostUserMsg *msg,\n \t\tdev->mem->nregions++;\n \t}\n \n+\tif (dev->async_copy && rte_vfio_is_enabled(\"vfio\"))\n+\t\tasync_dma_map(dev, true);\n+\n \tif (vhost_user_postcopy_register(dev, main_fd, msg) < 0)\n \t\tgoto free_mem_table;\n \n", "prefixes": [] }