From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from dpdk.org (dpdk.org [92.243.14.124]) by inbox.dpdk.org (Postfix) with ESMTP id 252F9A2E1B for ; Tue, 3 Sep 2019 09:40:34 +0200 (CEST) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id 9A1ED1E870; Tue, 3 Sep 2019 09:40:33 +0200 (CEST) Received: from mx1.redhat.com (mx1.redhat.com [209.132.183.28]) by dpdk.org (Postfix) with ESMTP id 304A01BF41; Tue, 3 Sep 2019 09:40:32 +0200 (CEST) Received: from smtp.corp.redhat.com (int-mx04.intmail.prod.int.phx2.redhat.com [10.5.11.14]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 666A57F748; Tue, 3 Sep 2019 07:40:31 +0000 (UTC) Received: from [10.36.112.25] (ovpn-112-25.ams2.redhat.com [10.36.112.25]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 54CFE5D9DC; Tue, 3 Sep 2019 07:40:27 +0000 (UTC) To: Tiwei Bie Cc: zhihong.wang@intel.com, amorenoz@redhat.com, xiao.w.wang@intel.com, dev@dpdk.org, jfreimann@redhat.com, stable@dpdk.org References: <20190829080000.20806-1-maxime.coquelin@redhat.com> <20190829080000.20806-11-maxime.coquelin@redhat.com> <20190903053006.GA32701@___> From: Maxime Coquelin Openpgp: preference=signencrypt Autocrypt: addr=maxime.coquelin@redhat.com; keydata= mQINBFOEQQIBEADjNLYZZqghYuWv1nlLisptPJp+TSxE/KuP7x47e1Gr5/oMDJ1OKNG8rlNg kLgBQUki3voWhUbMb69ybqdMUHOl21DGCj0BTU3lXwapYXOAnsh8q6RRM+deUpasyT+Jvf3a gU35dgZcomRh5HPmKMU4KfeA38cVUebsFec1HuJAWzOb/UdtQkYyZR4rbzw8SbsOemtMtwOx YdXodneQD7KuRU9IhJKiEfipwqk2pufm2VSGl570l5ANyWMA/XADNhcEXhpkZ1Iwj3TWO7XR uH4xfvPl8nBsLo/EbEI7fbuUULcAnHfowQslPUm6/yaGv6cT5160SPXT1t8U9QDO6aTSo59N jH519JS8oeKZB1n1eLDslCfBpIpWkW8ZElGkOGWAN0vmpLfdyiqBNNyS3eGAfMkJ6b1A24un /TKc6j2QxM0QK4yZGfAxDxtvDv9LFXec8ENJYsbiR6WHRHq7wXl/n8guyh5AuBNQ3LIK44x0 KjGXP1FJkUhUuruGyZsMrDLBRHYi+hhDAgRjqHgoXi5XGETA1PAiNBNnQwMf5aubt+mE2Q5r qLNTgwSo2dpTU3+mJ3y3KlsIfoaxYI7XNsPRXGnZi4hbxmeb2NSXgdCXhX3nELUNYm4ArKBP LugOIT/zRwk0H0+RVwL2zHdMO1Tht1UOFGfOZpvuBF60jhMzbQARAQABtCxNYXhpbWUgQ29x dWVsaW4gPG1heGltZS5jb3F1ZWxpbkByZWRoYXQuY29tPokCOAQTAQIAIgUCV3u/5QIbAwYL CQgHAwIGFQgCCQoLBBYCAwECHgECF4AACgkQyjiNKEaHD4ma2g/+P+Hg9WkONPaY1J4AR7Uf kBneosS4NO3CRy0x4WYmUSLYMLx1I3VH6SVjqZ6uBoYy6Fs6TbF6SHNc7QbB6Qjo3neqnQR1 71Ua1MFvIob8vUEl3jAR/+oaE1UJKrxjWztpppQTukIk4oJOmXbL0nj3d8dA2QgHdTyttZ1H xzZJWWz6vqxCrUqHU7RSH9iWg9R2iuTzii4/vk1oi4Qz7y/q8ONOq6ffOy/t5xSZOMtZCspu Mll2Szzpc/trFO0pLH4LZZfz/nXh2uuUbk8qRIJBIjZH3ZQfACffgfNefLe2PxMqJZ8mFJXc RQO0ONZvwoOoHL6CcnFZp2i0P5ddduzwPdGsPq1bnIXnZqJSl3dUfh3xG5ArkliZ/++zGF1O wvpGvpIuOgLqjyCNNRoR7cP7y8F24gWE/HqJBXs1qzdj/5Hr68NVPV1Tu/l2D1KMOcL5sOrz 2jLXauqDWn1Okk9hkXAP7+0Cmi6QwAPuBT3i6t2e8UdtMtCE4sLesWS/XohnSFFscZR6Vaf3 gKdWiJ/fW64L6b9gjkWtHd4jAJBAIAx1JM6xcA1xMbAFsD8gA2oDBWogHGYcScY/4riDNKXi lw92d6IEHnSf6y7KJCKq8F+Jrj2BwRJiFKTJ6ChbOpyyR6nGTckzsLgday2KxBIyuh4w+hMq TGDSp2rmWGJjASq5Ag0EVPSbkwEQAMkaNc084Qvql+XW+wcUIY+Dn9A2D1gMr2BVwdSfVDN7 0ZYxo9PvSkzh6eQmnZNQtl8WSHl3VG3IEDQzsMQ2ftZn2sxjcCadexrQQv3Lu60Tgj7YVYRM H+fLYt9W5YuWduJ+FPLbjIKynBf6JCRMWr75QAOhhhaI0tsie3eDsKQBA0w7WCuPiZiheJaL 4MDe9hcH4rM3ybnRW7K2dLszWNhHVoYSFlZGYh+MGpuODeQKDS035+4H2rEWgg+iaOwqD7bg CQXwTZ1kSrm8NxIRVD3MBtzp9SZdUHLfmBl/tLVwDSZvHZhhvJHC6Lj6VL4jPXF5K2+Nn/Su CQmEBisOmwnXZhhu8ulAZ7S2tcl94DCo60ReheDoPBU8PR2TLg8rS5f9w6mLYarvQWL7cDtT d2eX3Z6TggfNINr/RTFrrAd7NHl5h3OnlXj7PQ1f0kfufduOeCQddJN4gsQfxo/qvWVB7PaE 1WTIggPmWS+Xxijk7xG6x9McTdmGhYaPZBpAxewK8ypl5+yubVsE9yOOhKMVo9DoVCjh5To5 aph7CQWfQsV7cd9PfSJjI2lXI0dhEXhQ7lRCFpf3V3mD6CyrhpcJpV6XVGjxJvGUale7+IOp sQIbPKUHpB2F+ZUPWds9yyVxGwDxD8WLqKKy0WLIjkkSsOb9UBNzgRyzrEC9lgQ/ABEBAAGJ Ah8EGAECAAkFAlT0m5MCGwwACgkQyjiNKEaHD4nU8hAAtt0xFJAy0sOWqSmyxTc7FUcX+pbD KVyPlpl6urKKMk1XtVMUPuae/+UwvIt0urk1mXi6DnrAN50TmQqvdjcPTQ6uoZ8zjgGeASZg jj0/bJGhgUr9U7oG7Hh2F8vzpOqZrdd65MRkxmc7bWj1k81tOU2woR/Gy8xLzi0k0KUa8ueB iYOcZcIGTcs9CssVwQjYaXRoeT65LJnTxYZif2pfNxfINFzCGw42s3EtZFteczClKcVSJ1+L +QUY/J24x0/ocQX/M1PwtZbB4c/2Pg/t5FS+s6UB1Ce08xsJDcwyOPIH6O3tccZuriHgvqKP yKz/Ble76+NFlTK1mpUlfM7PVhD5XzrDUEHWRTeTJSvJ8TIPL4uyfzhjHhlkCU0mw7Pscyxn DE8G0UYMEaNgaZap8dcGMYH/96EfE5s/nTX0M6MXV0yots7U2BDb4soLCxLOJz4tAFDtNFtA wLBhXRSvWhdBJZiig/9CG3dXmKfi2H+wdUCSvEFHRpgo7GK8/Kh3vGhgKmnnxhl8ACBaGy9n fxjSxjSO6rj4/MeenmlJw1yebzkX8ZmaSi8BHe+n6jTGEFNrbiOdWpJgc5yHIZZnwXaW54QT UhhSjDL1rV2B4F28w30jYmlRmm2RdN7iCZfbyP3dvFQTzQ4ySquuPkIGcOOHrvZzxbRjzMx1 Mwqu3GQ= Message-ID: Date: Tue, 3 Sep 2019 09:40:25 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.8.0 MIME-Version: 1.0 In-Reply-To: <20190903053006.GA32701@___> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit X-Scanned-By: MIMEDefang 2.79 on 10.5.11.14 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.6.2 (mx1.redhat.com [10.5.110.71]); Tue, 03 Sep 2019 07:40:31 +0000 (UTC) Subject: Re: [dpdk-dev] [PATCH 10/15] net/virtio: add vDPA op to configure and start the device X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 9/3/19 7:30 AM, Tiwei Bie wrote: > On Thu, Aug 29, 2019 at 09:59:55AM +0200, Maxime Coquelin wrote: >> In order to support multi-queue, we need to implement the control >> path. The problem is that both the Vhost-user master and slave use >> VAs in their processes address spaces as IOVAs, which creates >> collusions between the data rings IOVAs managed the master, and >> the Control ring IOVAs. The trick here is to remmap the Control >> ring memory to another range, after the slave is aware of master's >> ranges. >> >> Signed-off-by: Maxime Coquelin >> --- >> drivers/net/virtio/virtio_vdpa.c | 255 +++++++++++++++++++++++++++++++ >> 1 file changed, 255 insertions(+) >> >> diff --git a/drivers/net/virtio/virtio_vdpa.c b/drivers/net/virtio/virtio_vdpa.c >> index fc52a8e92..13b4dd07d 100644 >> --- a/drivers/net/virtio/virtio_vdpa.c >> +++ b/drivers/net/virtio/virtio_vdpa.c >> @@ -106,6 +106,127 @@ find_internal_resource_by_dev(struct rte_pci_device *pdev) >> return list; >> } >> >> +static int >> +virtio_vdpa_dma_map_ctrl_queue(struct virtio_vdpa_device *dev, int do_map, >> + uint64_t iova) >> +{ >> + const struct rte_memzone *mz; >> + int ret; >> + >> + /* >> + * IOVAs are processes VAs. We cannot use them as the Data and Control >> + * paths are run in different processes, which may (does) lead to >> + * collusions. The trick here is to fixup Ctrl path IOVAs so that they >> + * start after the Data path ranges. >> + */ >> + if (do_map) { >> + mz = dev->cvq->cq.mz; >> + ret = rte_vfio_container_dma_map(dev->vfio_container_fd, >> + (uint64_t)(uintptr_t)mz->addr, >> + iova, mz->len); >> + if (ret < 0) { >> + DRV_LOG(ERR, "Failed to map ctrl ring (%d)", ret); >> + return ret; >> + } >> + >> + dev->cvq->vq_ring_mem = iova; >> + iova += mz->len; >> + >> + mz = dev->cvq->cq.virtio_net_hdr_mz; >> + ret = rte_vfio_container_dma_map(dev->vfio_container_fd, >> + (uint64_t)(uintptr_t)mz->addr, >> + iova, mz->len); >> + if (ret < 0) { >> + DRV_LOG(ERR, "Failed to map ctrl headers (%d)", ret); >> + return ret; >> + } > > This will allow guest to access the cq.mz and cq.virtio_net_hdr_mz > via the device which may have potential risks. I get what you mean, but I'm not sure to see how we could avoid that. AFAIU, we need to map the control queue in the device IOMMU, otherwise how could the host (in case of virtual device) or the NIC (in case of Virtio offload), could access the ring? Any thoughts? Thanks, Maxime > Regards, > Tiwei > >> + >> + dev->cvq->cq.virtio_net_hdr_mem = iova; >> + } else { >> + mz = dev->cvq->cq.mz; >> + ret = rte_vfio_container_dma_unmap(dev->vfio_container_fd, >> + (uint64_t)(uintptr_t)mz->addr, >> + iova, mz->len); >> + if (ret < 0) { >> + DRV_LOG(ERR, "Failed to unmap ctrl ring (%d)", ret); >> + return ret; >> + } >> + >> + dev->cvq->vq_ring_mem = 0; >> + iova += mz->len; >> + >> + mz = dev->cvq->cq.virtio_net_hdr_mz; >> + ret = rte_vfio_container_dma_unmap(dev->vfio_container_fd, >> + (uint64_t)(uintptr_t)mz->addr, >> + iova, mz->len); >> + if (ret < 0) { >> + DRV_LOG(ERR, "Failed to unmap ctrl headers (%d)", ret); >> + return ret; >> + } >> + >> + dev->cvq->cq.virtio_net_hdr_mem = 0; >> + } >> + >> + return 0; >> +} >> + >> +static int >> +virtio_vdpa_dma_map(struct virtio_vdpa_device *dev, int do_map) >> +{ >> + uint32_t i; >> + int ret; >> + struct rte_vhost_memory *mem = NULL; >> + int vfio_container_fd; >> + uint64_t avail_iova = 0; >> + >> + ret = rte_vhost_get_mem_table(dev->vid, &mem); >> + if (ret < 0 || !mem) { >> + DRV_LOG(ERR, "failed to get VM memory layout."); >> + return ret; >> + } >> + >> + vfio_container_fd = dev->vfio_container_fd; >> + >> + for (i = 0; i < mem->nregions; i++) { >> + struct rte_vhost_mem_region *reg; >> + >> + reg = &mem->regions[i]; >> + DRV_LOG(INFO, "%s, region %u: HVA 0x%" PRIx64 ", " >> + "GPA 0x%" PRIx64 ", size 0x%" PRIx64 ".", >> + do_map ? "DMA map" : "DMA unmap", i, >> + reg->host_user_addr, reg->guest_phys_addr, reg->size); >> + >> + if (reg->guest_phys_addr + reg->size > avail_iova) >> + avail_iova = reg->guest_phys_addr + reg->size; >> + >> + if (do_map) { >> + ret = rte_vfio_container_dma_map(vfio_container_fd, >> + reg->host_user_addr, reg->guest_phys_addr, >> + reg->size); >> + if (ret < 0) { >> + DRV_LOG(ERR, "DMA map failed."); >> + goto exit; >> + } >> + } else { >> + ret = rte_vfio_container_dma_unmap(vfio_container_fd, >> + reg->host_user_addr, reg->guest_phys_addr, >> + reg->size); >> + if (ret < 0) { >> + DRV_LOG(ERR, "DMA unmap failed."); >> + goto exit; >> + } >> + } >> + } >> + >> + if (dev->cvq) >> + ret = virtio_vdpa_dma_map_ctrl_queue(dev, do_map, avail_iova); >> + >> +exit: >> + free(mem); >> + >> + return ret; >> +} >> + >> static int >> virtio_vdpa_vfio_setup(struct virtio_vdpa_device *dev) >> { >> @@ -216,10 +337,144 @@ virtio_vdpa_get_protocol_features(int did __rte_unused, uint64_t *features) >> return 0; >> } >> >> +static uint64_t >> +hva_to_gpa(int vid, uint64_t hva) >> +{ >> + struct rte_vhost_memory *mem = NULL; >> + struct rte_vhost_mem_region *reg; >> + uint32_t i; >> + uint64_t gpa = 0; >> + >> + if (rte_vhost_get_mem_table(vid, &mem) < 0) >> + goto exit; >> + >> + for (i = 0; i < mem->nregions; i++) { >> + reg = &mem->regions[i]; >> + >> + if (hva >= reg->host_user_addr && >> + hva < reg->host_user_addr + reg->size) { >> + gpa = hva - reg->host_user_addr + reg->guest_phys_addr; >> + break; >> + } >> + } >> + >> +exit: >> + if (mem) >> + free(mem); >> + return gpa; >> +} >> + >> +static int >> +virtio_vdpa_start(struct virtio_vdpa_device *dev) >> +{ >> + struct virtio_hw *hw = &dev->hw; >> + int i, vid, nr_vring, ret; >> + struct rte_vhost_vring vr; >> + struct virtio_pmd_ctrl ctrl; >> + int dlen[1]; >> + >> + vid = dev->vid; >> + nr_vring = rte_vhost_get_vring_num(vid); >> + >> + if (dev->vqs) >> + rte_free(dev->vqs); >> + >> + dev->vqs = rte_zmalloc("virtio_vdpa", sizeof(*dev->vqs) * nr_vring, 0); >> + >> + for (i = 0; i < nr_vring; i++) { >> + struct virtqueue *vq = &dev->vqs[i]; >> + >> + rte_vhost_get_vhost_vring(vid, i, &vr); >> + >> + vq->vq_queue_index = i; >> + vq->vq_nentries = vr.size; >> + vq->vq_ring_mem = hva_to_gpa(vid, (uint64_t)(uintptr_t)vr.desc); >> + if (vq->vq_ring_mem == 0) { >> + DRV_LOG(ERR, "Fail to get GPA for descriptor ring."); >> + ret = -1; >> + goto out_free_vqs; >> + } >> + >> + ret = VTPCI_OPS(hw)->setup_queue(hw, vq); >> + if (ret) { >> + DRV_LOG(ERR, "Fail to setup queue."); >> + goto out_free_vqs; >> + } >> + } >> + >> + if (dev->cvq) { >> + ret = VTPCI_OPS(hw)->setup_queue(hw, dev->cvq); >> + if (ret) { >> + DRV_LOG(ERR, "Fail to setup ctrl queue."); >> + goto out_free_vqs; >> + } >> + } >> + >> + vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_DRIVER_OK); >> + >> + if (!dev->cvq) >> + return 0; >> + >> + ctrl.hdr.class = VIRTIO_NET_CTRL_MQ; >> + ctrl.hdr.cmd = VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET; >> + memcpy(ctrl.data, &dev->max_queue_pairs, sizeof(uint16_t)); >> + >> + dlen[0] = sizeof(uint16_t); >> + >> + ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1); >> + if (ret) { >> + DRV_LOG(ERR, "Multiqueue configured but send command " >> + "failed, this is too late now..."); >> + ret = -EINVAL; >> + goto out_free_vqs; >> + } >> + >> + return 0; >> +out_free_vqs: >> + rte_free(dev->vqs); >> + >> + return ret; >> +} >> + >> +static int >> +virtio_vdpa_dev_config(int vid) >> +{ >> + int did, ret; >> + struct internal_list *list; >> + struct virtio_vdpa_device *dev; >> + >> + did = rte_vhost_get_vdpa_device_id(vid); >> + list = find_internal_resource_by_did(did); >> + if (list == NULL) { >> + DRV_LOG(ERR, "Invalid device id: %d", did); >> + return -1; >> + } >> + >> + dev = list->dev; >> + dev->vid = vid; >> + >> + rte_spinlock_lock(&dev->lock); >> + >> + ret = virtio_vdpa_dma_map(dev, 1); >> + if (ret) >> + goto out_unlock; >> + >> + ret = virtio_vdpa_start(dev); >> + >> + if (rte_vhost_host_notifier_ctrl(vid, true) != 0) >> + DRV_LOG(NOTICE, "vDPA (%d): software relay is used.", did); >> + >> +out_unlock: >> + rte_spinlock_unlock(&dev->lock); >> + >> + return ret; >> +} >> + >> static struct rte_vdpa_dev_ops virtio_vdpa_ops = { >> .get_queue_num = virtio_vdpa_get_queue_num, >> .get_features = virtio_vdpa_get_features, >> .get_protocol_features = virtio_vdpa_get_protocol_features, >> + .dev_conf = virtio_vdpa_dev_config, >> }; >> >> static inline int >> -- >> 2.21.0 >>