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 A43F0A04FD; Wed, 15 Jan 2020 07:14:47 +0100 (CET) Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id DCE891C044; Wed, 15 Jan 2020 07:14:46 +0100 (CET) Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by dpdk.org (Postfix) with ESMTP id DC9FB1C043; Wed, 15 Jan 2020 07:14:44 +0100 (CET) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by orsmga102.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 14 Jan 2020 22:14:43 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.70,321,1574150400"; d="scan'208";a="256628411" Received: from dpdk-xuanding-dev2.sh.intel.com ([10.67.119.57]) by fmsmga002.fm.intel.com with ESMTP; 14 Jan 2020 22:14:42 -0800 From: Xuan Ding To: maxime.coquelin@redhat.com, tiwei.bie@intel.com, zhihong.wang@intel.com, yong.liu@intel.com, xiaolong.ye@intel.com Cc: dev@dpdk.org, Xuan Ding , stable@dpdk.org Date: Wed, 15 Jan 2020 06:13:58 +0000 Message-Id: <20200115061358.5666-1-xuan.ding@intel.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20191223072519.18899-1-xuan.ding@intel.com> References: <20191223072519.18899-1-xuan.ding@intel.com> Subject: [dpdk-dev] [PATCH v5] net/virtio-user: fix packed ring server mode 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" This patch fixes the situation where data path does not work properly when vhost reconnects to virtio in server mode with packed ring. Currently, virtio and vhost share memory of vring. For split ring, vhost can read the status of descriptors directly from the available ring and the used ring during reconnection. Therefore, the data path can continue. But for packed ring, when reconnecting to virtio, vhost cannot get the status of descriptors via the descriptor ring. By resetting Tx and Rx queues, the data path can restart from the beginning. Fixes: 4c3f5822eb214 ("net/virtio: add packed virtqueue defines") Cc: stable@dpdk.org Signed-off-by: Xuan Ding --- v5: * Fixed two spelling mistakes in the commit log. * Added notice message when resetting vring. v4: * Moved change log below '---' marker. v3: * Removed an extra asterisk from a comment. * Renamed device reset function and moved it to virtio_user_ethdev.c. v2: * Renamed queue reset functions and moved them to virtqueue.c. --- drivers/net/virtio/virtio_ethdev.c | 4 +- drivers/net/virtio/virtio_user_ethdev.c | 42 +++++++++++++++ drivers/net/virtio/virtqueue.c | 71 +++++++++++++++++++++++++ drivers/net/virtio/virtqueue.h | 4 ++ 4 files changed, 119 insertions(+), 2 deletions(-) diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c index 044eb10a7..f9d0ea70d 100644 --- a/drivers/net/virtio/virtio_ethdev.c +++ b/drivers/net/virtio/virtio_ethdev.c @@ -1913,6 +1913,8 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev) goto err_vtpci_init; } + rte_spinlock_init(&hw->state_lock); + /* reset device and negotiate default features */ ret = virtio_init_device(eth_dev, VIRTIO_PMD_DEFAULT_GUEST_FEATURES); if (ret < 0) @@ -2155,8 +2157,6 @@ virtio_dev_configure(struct rte_eth_dev *dev) return -EBUSY; } - rte_spinlock_init(&hw->state_lock); - hw->use_simple_rx = 1; if (vtpci_with_feature(hw, VIRTIO_F_IN_ORDER)) { diff --git a/drivers/net/virtio/virtio_user_ethdev.c b/drivers/net/virtio/virtio_user_ethdev.c index 3fc172573..9c9d3407f 100644 --- a/drivers/net/virtio/virtio_user_ethdev.c +++ b/drivers/net/virtio/virtio_user_ethdev.c @@ -25,12 +25,48 @@ #define virtio_user_get_dev(hw) \ ((struct virtio_user_dev *)(hw)->virtio_user_dev) +static void +virtio_user_reset_queues_packed(struct rte_eth_dev *dev) +{ + struct virtio_hw *hw = dev->data->dev_private; + struct virtnet_rx *rxvq; + struct virtnet_tx *txvq; + uint16_t i; + + /* Add lock to avoid queue contention. */ + rte_spinlock_lock(&hw->state_lock); + hw->started = 0; + + /* + * Waitting for datapath to complete before resetting queues. + * 1 ms should be enough for the ongoing Tx/Rx function to finish. + */ + rte_delay_ms(1); + + /* Vring reset for each Tx queue and Rx queue. */ + for (i = 0; i < dev->data->nb_rx_queues; i++) { + rxvq = dev->data->rx_queues[i]; + virtqueue_rxvq_reset_packed(rxvq->vq); + virtio_dev_rx_queue_setup_finish(dev, i); + } + + for (i = 0; i < dev->data->nb_tx_queues; i++) { + txvq = dev->data->tx_queues[i]; + virtqueue_txvq_reset_packed(txvq->vq); + } + + hw->started = 1; + rte_spinlock_unlock(&hw->state_lock); +} + + static int virtio_user_server_reconnect(struct virtio_user_dev *dev) { int ret; int connectfd; struct rte_eth_dev *eth_dev = &rte_eth_devices[dev->port_id]; + struct virtio_hw *hw = eth_dev->data->dev_private; connectfd = accept(dev->listenfd, NULL, NULL); if (connectfd < 0) @@ -51,6 +87,12 @@ virtio_user_server_reconnect(struct virtio_user_dev *dev) dev->features &= dev->device_features; + /* For packed ring, resetting queues is required in reconnection. */ + if (vtpci_packed_queue(hw)) + PMD_INIT_LOG(NOTICE, "Packets on the fly will be dropped" + " when packed ring reconnecting."); + virtio_user_reset_queues_packed(eth_dev); + ret = virtio_user_start_device(dev); if (ret < 0) return -1; diff --git a/drivers/net/virtio/virtqueue.c b/drivers/net/virtio/virtqueue.c index 5ff1e3587..0b4e3bf3e 100644 --- a/drivers/net/virtio/virtqueue.c +++ b/drivers/net/virtio/virtqueue.c @@ -141,3 +141,74 @@ virtqueue_rxvq_flush(struct virtqueue *vq) else virtqueue_rxvq_flush_split(vq); } + +int +virtqueue_rxvq_reset_packed(struct virtqueue *vq) +{ + int size = vq->vq_nentries; + struct vq_desc_extra *dxp; + struct virtnet_rx *rxvq; + uint16_t desc_idx; + + vq->vq_used_cons_idx = 0; + vq->vq_desc_head_idx = 0; + vq->vq_avail_idx = 0; + vq->vq_desc_tail_idx = (uint16_t)(vq->vq_nentries - 1); + vq->vq_free_cnt = vq->vq_nentries; + + vq->vq_packed.used_wrap_counter = 1; + vq->vq_packed.cached_flags = VRING_PACKED_DESC_F_AVAIL; + vq->vq_packed.event_flags_shadow = 0; + vq->vq_packed.cached_flags |= VRING_DESC_F_WRITE; + + rxvq = &vq->rxq; + memset(rxvq->mz->addr, 0, rxvq->mz->len); + + for (desc_idx = 0; desc_idx < vq->vq_nentries; desc_idx++) { + dxp = &vq->vq_descx[desc_idx]; + if (dxp->cookie != NULL) { + rte_pktmbuf_free(dxp->cookie); + dxp->cookie = NULL; + } + } + + vring_desc_init_packed(vq, size); + + return 0; +} + +int +virtqueue_txvq_reset_packed(struct virtqueue *vq) +{ + int size = vq->vq_nentries; + struct vq_desc_extra *dxp; + struct virtnet_tx *txvq; + uint16_t desc_idx; + + vq->vq_used_cons_idx = 0; + vq->vq_desc_head_idx = 0; + vq->vq_avail_idx = 0; + vq->vq_desc_tail_idx = (uint16_t)(vq->vq_nentries - 1); + vq->vq_free_cnt = vq->vq_nentries; + + vq->vq_packed.used_wrap_counter = 1; + vq->vq_packed.cached_flags = VRING_PACKED_DESC_F_AVAIL; + vq->vq_packed.event_flags_shadow = 0; + + txvq = &vq->txq; + memset(txvq->mz->addr, 0, txvq->mz->len); + memset(txvq->virtio_net_hdr_mz->addr, 0, + txvq->virtio_net_hdr_mz->len); + + for (desc_idx = 0; desc_idx < vq->vq_nentries; desc_idx++) { + dxp = &vq->vq_descx[desc_idx]; + if (dxp->cookie != NULL) { + rte_pktmbuf_free(dxp->cookie); + dxp->cookie = NULL; + } + } + + vring_desc_init_packed(vq, size); + + return 0; +} diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h index 8d7f197b1..58ad7309a 100644 --- a/drivers/net/virtio/virtqueue.h +++ b/drivers/net/virtio/virtqueue.h @@ -443,6 +443,10 @@ struct rte_mbuf *virtqueue_detach_unused(struct virtqueue *vq); /* Flush the elements in the used ring. */ void virtqueue_rxvq_flush(struct virtqueue *vq); +int virtqueue_rxvq_reset_packed(struct virtqueue *vq); + +int virtqueue_txvq_reset_packed(struct virtqueue *vq); + static inline int virtqueue_full(const struct virtqueue *vq) { -- 2.17.1