From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 05DDCA0032; Mon, 13 Dec 2021 09:02:40 +0100 (CET) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id C71A940140; Mon, 13 Dec 2021 09:02:39 +0100 (CET) Received: from mga06.intel.com (mga06.intel.com [134.134.136.31]) by mails.dpdk.org (Postfix) with ESMTP id 638F640042 for ; Mon, 13 Dec 2021 09:02:38 +0100 (CET) DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1639382558; x=1670918558; h=from:to:cc:subject:date:message-id; bh=A9RDeI1day9X5KlQzM2Vs2a1EVdrVbvtYKeSlVTOZHA=; b=lviHpBZxWAcSp36MCwKfbkWPZF1zNfq5pZycxtmjMzyIDnwW7RmHcsEI MEidc1jn9TcOODto6KEwNvhJ5g0lE3mmgTQGb1qFmHEFEtjcM7WEFdwfS HLvLJU3QfnjZFvhKvCNN2jAC0+rwrWoTgMhf8g+OHXOgg41wmC5Hvp6kh nP59ef7RB+4B9EswvJC9TPbU3JuQfCGNZcOuOIHUquG0JTchVIBSL2IiK QNCWu2XCe3YSkvnacTApij2wshdQVyIMkoZ9PM29l2kBuU/OJzWH634+l dFD6CU0IKCTMTSGJz4oWS8+om9P9gvYcmnlziZjC2UnWSwx4VlEehmSB4 Q==; X-IronPort-AV: E=McAfee;i="6200,9189,10196"; a="299465290" X-IronPort-AV: E=Sophos;i="5.88,202,1635231600"; d="scan'208";a="299465290" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by orsmga104.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 13 Dec 2021 00:02:37 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.88,202,1635231600"; d="scan'208";a="517650706" Received: from npg-dpdk-xuan-cbdma.sh.intel.com ([10.67.110.228]) by orsmga008.jf.intel.com with ESMTP; 13 Dec 2021 00:02:33 -0800 From: xuan.ding@intel.com To: maxime.coquelin@redhat.com, chenbo.xia@intel.com Cc: dev@dpdk.org, yuanx.wang@intel.com, jiayu.hu@intel.com, yinan.wang@intel.com, xingguang.he@intel.com, Xuan Ding 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 List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org From: Xuan Ding 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 Signed-off-by: Yuan Wang --- 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 7085e0885c..d246538ca5 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_phys_addr; + uint64_t host_user_addr; uint64_t size; }; diff --git a/lib/vhost/vhost_user.c b/lib/vhost/vhost_user.c index a781346c4d..6d888766b0 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_phys_addr, + 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_phys_addr, + 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_phys_addr, uint64_t size) + uint64_t host_phys_addr, 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_phys_addr == last_page->host_phys_addr + - last_page->size) { + if (host_phys_addr == last_page->host_phys_addr + 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_phys_addr = host_phys_addr; + 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_phys_addr, size) < 0) + if (add_one_guest_page(dev, guest_phys_addr, host_phys_addr, + 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_phys_addr = rte_mem_virt2iova((void *)(uintptr_t) host_user_addr); if (add_one_guest_page(dev, guest_phys_addr, host_phys_addr, - 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; -- 2.25.1