From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga03.intel.com (mga03.intel.com [134.134.136.65]) by dpdk.org (Postfix) with ESMTP id 3DA431B847 for ; Tue, 10 Apr 2018 11:39:06 +0200 (CEST) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga006.jf.intel.com ([10.7.209.51]) by orsmga103.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 10 Apr 2018 02:39:06 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.48,431,1517904000"; d="scan'208";a="33105527" Received: from tanjianf-mobl.ccr.corp.intel.com (HELO [10.67.64.80]) ([10.67.64.80]) by orsmga006.jf.intel.com with ESMTP; 10 Apr 2018 02:39:04 -0700 To: Junjie Chen , maxime.coquelin@redhat.com, mtetsuyah@gmail.com References: <1523369889-73457-1-git-send-email-junjie.j.chen@intel.com> Cc: dev@dpdk.org From: "Tan, Jianfeng" Message-ID: <089d4be2-811d-8bf3-1ae4-58b73b10a07b@intel.com> Date: Tue, 10 Apr 2018 17:39:04 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 MIME-Version: 1.0 In-Reply-To: <1523369889-73457-1-git-send-email-junjie.j.chen@intel.com> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [dpdk-dev] [PATCH] net/vhost: fix vhost invalid state 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: , X-List-Received-Date: Tue, 10 Apr 2018 09:39:07 -0000 Hi Junjie, I think the code is still buggy. As vhost thread and master thread are separately invoking new_device() and dev_start(). On 4/10/2018 10:18 PM, Junjie Chen wrote: > dev_start sets *dev_attached* after setup queues, this sets device to > invalid state since no frontend is attached. Also destroy_device set > *started* to zero which makes *allow_queuing* always zero until dev_start > get called again. Actually, we should not determine queues existence by > *dev_attached* but by queues pointers or other separated variable(s). > > Fixes: 30a701a53737 ("net/vhost: fix crash when creating vdev > dynamically") > > Signed-off-by: Junjie Chen > --- > drivers/net/vhost/rte_eth_vhost.c | 64 +++++++++++++++++++++++---------------- > 1 file changed, 38 insertions(+), 26 deletions(-) > > diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c > index 11b6076..6a2ff76 100644 > --- a/drivers/net/vhost/rte_eth_vhost.c > +++ b/drivers/net/vhost/rte_eth_vhost.c > @@ -118,6 +118,7 @@ struct pmd_internal { > char *iface_name; > uint16_t max_queues; > uint16_t vid; > + uint16_t queue_ready; If we can reuse below *started* field, instead of introducing a new variable? > rte_atomic32_t started; > }; > > @@ -528,10 +529,13 @@ update_queuing_status(struct rte_eth_dev *dev) > unsigned int i; > int allow_queuing = 1; > > - if (rte_atomic32_read(&internal->dev_attached) == 0) > + if (!dev->data->rx_queues || !dev->data->tx_queues) { > + RTE_LOG(ERR, PMD, "RX/TX queues not setup yet\n"); > return; > + } > > - if (rte_atomic32_read(&internal->started) == 0) > + if (rte_atomic32_read(&internal->started) == 0 || > + rte_atomic32_read(&internal->dev_attached) == 0) > allow_queuing = 0; > > /* Wait until rx/tx_pkt_burst stops accessing vhost device */ > @@ -576,6 +580,8 @@ queue_setup(struct rte_eth_dev *eth_dev, struct pmd_internal *internal) > vq->internal = internal; > vq->port = eth_dev->data->port_id; > } > + > + internal->queue_ready = 1; > } > > static int > @@ -607,13 +613,10 @@ new_device(int vid) > #endif > > internal->vid = vid; > - if (eth_dev->data->rx_queues && eth_dev->data->tx_queues) { > + if (eth_dev->data->rx_queues && eth_dev->data->tx_queues) > queue_setup(eth_dev, internal); > - rte_atomic32_set(&internal->dev_attached, 1); > - } else { > - RTE_LOG(INFO, PMD, "RX/TX queues have not setup yet\n"); > - rte_atomic32_set(&internal->dev_attached, 0); > - } > + else vhost thread (t1): goes here and before setting dev_attached ... > + RTE_LOG(INFO, PMD, "RX/TX queues not setup yet\n"); > > for (i = 0; i < rte_vhost_get_vring_num(vid); i++) > rte_vhost_enable_guest_notification(vid, i, 0); > @@ -622,6 +625,7 @@ new_device(int vid) > > eth_dev->data->dev_link.link_status = ETH_LINK_UP; > > + rte_atomic32_set(&internal->dev_attached, 1); > update_queuing_status(eth_dev); > > RTE_LOG(INFO, PMD, "Vhost device %d created\n", vid); > @@ -657,17 +661,19 @@ destroy_device(int vid) > > eth_dev->data->dev_link.link_status = ETH_LINK_DOWN; > > - for (i = 0; i < eth_dev->data->nb_rx_queues; i++) { > - vq = eth_dev->data->rx_queues[i]; > - if (vq == NULL) > - continue; > - vq->vid = -1; > - } > - for (i = 0; i < eth_dev->data->nb_tx_queues; i++) { > - vq = eth_dev->data->tx_queues[i]; > - if (vq == NULL) > - continue; > - vq->vid = -1; > + if (eth_dev->data->rx_queues && eth_dev->data->tx_queues) { > + for (i = 0; i < eth_dev->data->nb_rx_queues; i++) { > + vq = eth_dev->data->rx_queues[i]; > + if (!vq) > + continue; > + vq->vid = -1; > + } > + for (i = 0; i < eth_dev->data->nb_tx_queues; i++) { > + vq = eth_dev->data->tx_queues[i]; > + if (!vq) > + continue; > + vq->vid = -1; > + } > } > > state = vring_states[eth_dev->data->port_id]; > @@ -792,11 +798,14 @@ eth_dev_start(struct rte_eth_dev *eth_dev) > { > struct pmd_internal *internal = eth_dev->data->dev_private; > > - if (unlikely(rte_atomic32_read(&internal->dev_attached) == 0)) { > - queue_setup(eth_dev, internal); > - rte_atomic32_set(&internal->dev_attached, 1); > + if (!eth_dev->data->rx_queues || !eth_dev->data->tx_queues) { > + RTE_LOG(ERR, PMD, "RX/TX queues not exist yet\n"); > + return -1; > } I think in DPDK, we can make the assumption that dev_start() comes later than queue setup. So we don't need this check. > > + if (!internal->queue_ready) > + queue_setup(eth_dev, internal); Master thread (t2): setting invalid values. > + > rte_atomic32_set(&internal->started, 1); > update_queuing_status(eth_dev); > > @@ -836,10 +845,13 @@ eth_dev_close(struct rte_eth_dev *dev) > pthread_mutex_unlock(&internal_list_lock); > rte_free(list); > > - for (i = 0; i < dev->data->nb_rx_queues; i++) > - rte_free(dev->data->rx_queues[i]); > - for (i = 0; i < dev->data->nb_tx_queues; i++) > - rte_free(dev->data->tx_queues[i]); > + if (dev->data->rx_queues) > + for (i = 0; i < dev->data->nb_rx_queues; i++) > + rte_free(dev->data->rx_queues[i]); > + > + if (dev->data->tx_queues) > + for (i = 0; i < dev->data->nb_tx_queues; i++) > + rte_free(dev->data->tx_queues[i]); > > rte_free(dev->data->mac_addrs); > free(internal->dev_name);