Re: [dpdk-dev] [PATCH 1/2] vhost: support to clear in-flight packets for async dequeue
From: Xia, Chenbo <hidden>
Date: 2021-09-15 07:02:26
Hi Yuan,
quoted hunk ↗ jump to hunk
-----Original Message----- From: Wang, YuanX <redacted> Sent: Thursday, September 9, 2021 2:58 PM To: dev@dpdk.org Cc: maxime.coquelin@redhat.com; Xia, Chenbo <redacted>; Pai G, Sunil [off-list ref]; Hu, Jiayu [off-list ref]; Ding, Xuan [off-list ref]; Jiang, Cheng1 [off-list ref]; Ma, WenwuX [off-list ref]; Yang, YvonneX [off-list ref]; Wang, YuanX [off-list ref] Subject: [PATCH 1/2] vhost: support to clear in-flight packets for async dequeue rte_vhost_clear_queue_thread_unsafe() supports to clear in-flight packets for async enqueue only. But after supporting async dequeue, this API should support async dequeue too. Signed-off-by: Yuan Wang <redacted> --- lib/vhost/virtio_net.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-)diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c index e0159b53e3..7f6183a929 100644 --- a/lib/vhost/virtio_net.c +++ b/lib/vhost/virtio_net.c@@ -27,6 +27,11 @@ #define VHOST_ASYNC_BATCH_THRESHOLD 32 +static __rte_always_inline uint16_t +async_poll_dequeue_completed_split(struct virtio_net *dev, + struct vhost_virtqueue *vq, uint16_t queue_id, + struct rte_mbuf **pkts, uint16_t count, bool legacy_ol_flags); + static __rte_always_inline bool rxvq_is_mergeable(struct virtio_net *dev) {@@ -2119,11 +2124,6 @@ rte_vhost_clear_queue_thread_unsafe(int vid, uint16_tqueue_id, return 0; VHOST_LOG_DATA(DEBUG, "(%d) %s\n", dev->vid, __func__); - if (unlikely(!is_valid_virt_queue_idx(queue_id, 0, dev->nr_vring))) { - VHOST_LOG_DATA(ERR, "(%d) %s: invalid virtqueue idx %d.\n", - dev->vid, __func__, queue_id); - return 0; - } vq = dev->virtqueue[queue_id];@@ -2133,7 +2133,11 @@ rte_vhost_clear_queue_thread_unsafe(int vid, uint16_tqueue_id, return 0; } - n_pkts_cpl = vhost_poll_enqueue_completed(dev, queue_id, pkts, count); + if ((queue_id % 2) == 0)
You can remove the internal '()'.
+ n_pkts_cpl = vhost_poll_enqueue_completed(dev, queue_id, pkts, count); + else + n_pkts_cpl = async_poll_dequeue_completed_split(dev, vq, queue_id, pkts, count,
You should check we are using split queue before entering this split queue function. Thanks, Chenbo
+ dev->flags & VIRTIO_DEV_LEGACY_OL_FLAGS); return n_pkts_cpl; } -- 2.25.1