From: Jason Wang <jasowang@redhat.com> To: mst@redhat.com, jasowang@redhat.com Cc: kvm@vger.kernel.org, virtualization@lists.linux-foundation.org, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, rob.miller@broadcom.com, lingshan.zhu@intel.com, eperezma@redhat.com, lulu@redhat.com, shahafs@mellanox.com, hanand@xilinx.com, mhabets@solarflare.com, gdawar@xilinx.com, saugatm@xilinx.com, vmireyno@marvell.com, zhangweining@ruijie.com.cn, eli@mellanox.com Subject: [PATCH 1/6] vhost: allow device that does not depend on vhost worker Date: Fri, 29 May 2020 16:02:58 +0800 [thread overview] Message-ID: <20200529080303.15449-2-jasowang@redhat.com> (raw) In-Reply-To: <20200529080303.15449-1-jasowang@redhat.com> vDPA device currently relays the eventfd via vhost worker. This is inefficient due the latency of wakeup and scheduling, so this patch tries to introduce a use_worker attribute for the vhost device. When use_worker is not set with vhost_dev_init(), vhost won't try to allocate a worker thread and the vhost_poll will be processed directly in the wakeup function. This help for vDPA since it reduces the latency caused by vhost worker. In my testing, it saves 0.2 ms in pings between VMs on a mutual host. Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- drivers/vhost/net.c | 2 +- drivers/vhost/scsi.c | 2 +- drivers/vhost/vdpa.c | 2 +- drivers/vhost/vhost.c | 38 +++++++++++++++++++++++++------------- drivers/vhost/vhost.h | 2 ++ drivers/vhost/vsock.c | 2 +- 6 files changed, 31 insertions(+), 17 deletions(-) diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index 2927f02cc7e1..bf5e1d81ae25 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -1326,7 +1326,7 @@ static int vhost_net_open(struct inode *inode, struct file *f) } vhost_dev_init(dev, vqs, VHOST_NET_VQ_MAX, UIO_MAXIOV + VHOST_NET_BATCH, - VHOST_NET_PKT_WEIGHT, VHOST_NET_WEIGHT, + VHOST_NET_PKT_WEIGHT, VHOST_NET_WEIGHT, true, NULL); vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, EPOLLOUT, dev); diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c index c39952243fd3..0cbaa0b3893d 100644 --- a/drivers/vhost/scsi.c +++ b/drivers/vhost/scsi.c @@ -1628,7 +1628,7 @@ static int vhost_scsi_open(struct inode *inode, struct file *f) vs->vqs[i].vq.handle_kick = vhost_scsi_handle_kick; } vhost_dev_init(&vs->dev, vqs, VHOST_SCSI_MAX_VQ, UIO_MAXIOV, - VHOST_SCSI_WEIGHT, 0, NULL); + VHOST_SCSI_WEIGHT, 0, true, NULL); vhost_scsi_init_inflight(vs, NULL); diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c index 20476b505d99..6ff72289f488 100644 --- a/drivers/vhost/vdpa.c +++ b/drivers/vhost/vdpa.c @@ -696,7 +696,7 @@ static int vhost_vdpa_open(struct inode *inode, struct file *filep) vqs[i] = &v->vqs[i]; vqs[i]->handle_kick = handle_vq_kick; } - vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, + vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, false, vhost_vdpa_process_iotlb_msg); dev->iotlb = vhost_iotlb_alloc(0, 0); diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index d450e16c5c25..70105e045768 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -166,11 +166,16 @@ static int vhost_poll_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync, void *key) { struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait); + struct vhost_work *work = &poll->work; if (!(key_to_poll(key) & poll->mask)) return 0; - vhost_poll_queue(poll); + if (!poll->dev->use_worker) + work->fn(work); + else + vhost_poll_queue(poll); + return 0; } @@ -454,6 +459,7 @@ static size_t vhost_get_desc_size(struct vhost_virtqueue *vq, void vhost_dev_init(struct vhost_dev *dev, struct vhost_virtqueue **vqs, int nvqs, int iov_limit, int weight, int byte_weight, + bool use_worker, int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg)) { @@ -471,6 +477,7 @@ void vhost_dev_init(struct vhost_dev *dev, dev->iov_limit = iov_limit; dev->weight = weight; dev->byte_weight = byte_weight; + dev->use_worker = use_worker; dev->msg_handler = msg_handler; init_llist_head(&dev->work_list); init_waitqueue_head(&dev->wait); @@ -549,18 +556,21 @@ long vhost_dev_set_owner(struct vhost_dev *dev) /* No owner, become one */ dev->mm = get_task_mm(current); dev->kcov_handle = kcov_common_handle(); - worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid); - if (IS_ERR(worker)) { - err = PTR_ERR(worker); - goto err_worker; - } + if (dev->use_worker) { + worker = kthread_create(vhost_worker, dev, + "vhost-%d", current->pid); + if (IS_ERR(worker)) { + err = PTR_ERR(worker); + goto err_worker; + } - dev->worker = worker; - wake_up_process(worker); /* avoid contributing to loadavg */ + dev->worker = worker; + wake_up_process(worker); /* avoid contributing to loadavg */ - err = vhost_attach_cgroups(dev); - if (err) - goto err_cgroup; + err = vhost_attach_cgroups(dev); + if (err) + goto err_cgroup; + } err = vhost_dev_alloc_iovecs(dev); if (err) @@ -568,8 +578,10 @@ long vhost_dev_set_owner(struct vhost_dev *dev) return 0; err_cgroup: - kthread_stop(worker); - dev->worker = NULL; + if (dev->worker) { + kthread_stop(dev->worker); + dev->worker = NULL; + } err_worker: if (dev->mm) mmput(dev->mm); diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h index f8403bd46b85..0feb6701e273 100644 --- a/drivers/vhost/vhost.h +++ b/drivers/vhost/vhost.h @@ -154,6 +154,7 @@ struct vhost_dev { int weight; int byte_weight; u64 kcov_handle; + bool use_worker; int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg); }; @@ -161,6 +162,7 @@ struct vhost_dev { bool vhost_exceeds_weight(struct vhost_virtqueue *vq, int pkts, int total_len); void vhost_dev_init(struct vhost_dev *, struct vhost_virtqueue **vqs, int nvqs, int iov_limit, int weight, int byte_weight, + bool use_worker, int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg)); long vhost_dev_set_owner(struct vhost_dev *dev); diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c index e36aaf9ba7bd..2eb85c42bac4 100644 --- a/drivers/vhost/vsock.c +++ b/drivers/vhost/vsock.c @@ -621,7 +621,7 @@ static int vhost_vsock_dev_open(struct inode *inode, struct file *file) vhost_dev_init(&vsock->dev, vqs, ARRAY_SIZE(vsock->vqs), UIO_MAXIOV, VHOST_VSOCK_PKT_WEIGHT, - VHOST_VSOCK_WEIGHT, NULL); + VHOST_VSOCK_WEIGHT, true, NULL); file->private_data = vsock; spin_lock_init(&vsock->send_pkt_list_lock); -- 2.20.1
WARNING: multiple messages have this Message-ID (diff)
From: Jason Wang <jasowang@redhat.com> To: mst@redhat.com, jasowang@redhat.com Cc: shahafs@mellanox.com, lulu@redhat.com, kvm@vger.kernel.org, saugatm@xilinx.com, netdev@vger.kernel.org, mhabets@solarflare.com, vmireyno@marvell.com, linux-kernel@vger.kernel.org, gdawar@xilinx.com, virtualization@lists.linux-foundation.org, eperezma@redhat.com, hanand@xilinx.com, zhangweining@ruijie.com.cn, eli@mellanox.com, lingshan.zhu@intel.com, rob.miller@broadcom.com Subject: [PATCH 1/6] vhost: allow device that does not depend on vhost worker Date: Fri, 29 May 2020 16:02:58 +0800 [thread overview] Message-ID: <20200529080303.15449-2-jasowang@redhat.com> (raw) In-Reply-To: <20200529080303.15449-1-jasowang@redhat.com> vDPA device currently relays the eventfd via vhost worker. This is inefficient due the latency of wakeup and scheduling, so this patch tries to introduce a use_worker attribute for the vhost device. When use_worker is not set with vhost_dev_init(), vhost won't try to allocate a worker thread and the vhost_poll will be processed directly in the wakeup function. This help for vDPA since it reduces the latency caused by vhost worker. In my testing, it saves 0.2 ms in pings between VMs on a mutual host. Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- drivers/vhost/net.c | 2 +- drivers/vhost/scsi.c | 2 +- drivers/vhost/vdpa.c | 2 +- drivers/vhost/vhost.c | 38 +++++++++++++++++++++++++------------- drivers/vhost/vhost.h | 2 ++ drivers/vhost/vsock.c | 2 +- 6 files changed, 31 insertions(+), 17 deletions(-) diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c index 2927f02cc7e1..bf5e1d81ae25 100644 --- a/drivers/vhost/net.c +++ b/drivers/vhost/net.c @@ -1326,7 +1326,7 @@ static int vhost_net_open(struct inode *inode, struct file *f) } vhost_dev_init(dev, vqs, VHOST_NET_VQ_MAX, UIO_MAXIOV + VHOST_NET_BATCH, - VHOST_NET_PKT_WEIGHT, VHOST_NET_WEIGHT, + VHOST_NET_PKT_WEIGHT, VHOST_NET_WEIGHT, true, NULL); vhost_poll_init(n->poll + VHOST_NET_VQ_TX, handle_tx_net, EPOLLOUT, dev); diff --git a/drivers/vhost/scsi.c b/drivers/vhost/scsi.c index c39952243fd3..0cbaa0b3893d 100644 --- a/drivers/vhost/scsi.c +++ b/drivers/vhost/scsi.c @@ -1628,7 +1628,7 @@ static int vhost_scsi_open(struct inode *inode, struct file *f) vs->vqs[i].vq.handle_kick = vhost_scsi_handle_kick; } vhost_dev_init(&vs->dev, vqs, VHOST_SCSI_MAX_VQ, UIO_MAXIOV, - VHOST_SCSI_WEIGHT, 0, NULL); + VHOST_SCSI_WEIGHT, 0, true, NULL); vhost_scsi_init_inflight(vs, NULL); diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c index 20476b505d99..6ff72289f488 100644 --- a/drivers/vhost/vdpa.c +++ b/drivers/vhost/vdpa.c @@ -696,7 +696,7 @@ static int vhost_vdpa_open(struct inode *inode, struct file *filep) vqs[i] = &v->vqs[i]; vqs[i]->handle_kick = handle_vq_kick; } - vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, + vhost_dev_init(dev, vqs, nvqs, 0, 0, 0, false, vhost_vdpa_process_iotlb_msg); dev->iotlb = vhost_iotlb_alloc(0, 0); diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c index d450e16c5c25..70105e045768 100644 --- a/drivers/vhost/vhost.c +++ b/drivers/vhost/vhost.c @@ -166,11 +166,16 @@ static int vhost_poll_wakeup(wait_queue_entry_t *wait, unsigned mode, int sync, void *key) { struct vhost_poll *poll = container_of(wait, struct vhost_poll, wait); + struct vhost_work *work = &poll->work; if (!(key_to_poll(key) & poll->mask)) return 0; - vhost_poll_queue(poll); + if (!poll->dev->use_worker) + work->fn(work); + else + vhost_poll_queue(poll); + return 0; } @@ -454,6 +459,7 @@ static size_t vhost_get_desc_size(struct vhost_virtqueue *vq, void vhost_dev_init(struct vhost_dev *dev, struct vhost_virtqueue **vqs, int nvqs, int iov_limit, int weight, int byte_weight, + bool use_worker, int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg)) { @@ -471,6 +477,7 @@ void vhost_dev_init(struct vhost_dev *dev, dev->iov_limit = iov_limit; dev->weight = weight; dev->byte_weight = byte_weight; + dev->use_worker = use_worker; dev->msg_handler = msg_handler; init_llist_head(&dev->work_list); init_waitqueue_head(&dev->wait); @@ -549,18 +556,21 @@ long vhost_dev_set_owner(struct vhost_dev *dev) /* No owner, become one */ dev->mm = get_task_mm(current); dev->kcov_handle = kcov_common_handle(); - worker = kthread_create(vhost_worker, dev, "vhost-%d", current->pid); - if (IS_ERR(worker)) { - err = PTR_ERR(worker); - goto err_worker; - } + if (dev->use_worker) { + worker = kthread_create(vhost_worker, dev, + "vhost-%d", current->pid); + if (IS_ERR(worker)) { + err = PTR_ERR(worker); + goto err_worker; + } - dev->worker = worker; - wake_up_process(worker); /* avoid contributing to loadavg */ + dev->worker = worker; + wake_up_process(worker); /* avoid contributing to loadavg */ - err = vhost_attach_cgroups(dev); - if (err) - goto err_cgroup; + err = vhost_attach_cgroups(dev); + if (err) + goto err_cgroup; + } err = vhost_dev_alloc_iovecs(dev); if (err) @@ -568,8 +578,10 @@ long vhost_dev_set_owner(struct vhost_dev *dev) return 0; err_cgroup: - kthread_stop(worker); - dev->worker = NULL; + if (dev->worker) { + kthread_stop(dev->worker); + dev->worker = NULL; + } err_worker: if (dev->mm) mmput(dev->mm); diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h index f8403bd46b85..0feb6701e273 100644 --- a/drivers/vhost/vhost.h +++ b/drivers/vhost/vhost.h @@ -154,6 +154,7 @@ struct vhost_dev { int weight; int byte_weight; u64 kcov_handle; + bool use_worker; int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg); }; @@ -161,6 +162,7 @@ struct vhost_dev { bool vhost_exceeds_weight(struct vhost_virtqueue *vq, int pkts, int total_len); void vhost_dev_init(struct vhost_dev *, struct vhost_virtqueue **vqs, int nvqs, int iov_limit, int weight, int byte_weight, + bool use_worker, int (*msg_handler)(struct vhost_dev *dev, struct vhost_iotlb_msg *msg)); long vhost_dev_set_owner(struct vhost_dev *dev); diff --git a/drivers/vhost/vsock.c b/drivers/vhost/vsock.c index e36aaf9ba7bd..2eb85c42bac4 100644 --- a/drivers/vhost/vsock.c +++ b/drivers/vhost/vsock.c @@ -621,7 +621,7 @@ static int vhost_vsock_dev_open(struct inode *inode, struct file *file) vhost_dev_init(&vsock->dev, vqs, ARRAY_SIZE(vsock->vqs), UIO_MAXIOV, VHOST_VSOCK_PKT_WEIGHT, - VHOST_VSOCK_WEIGHT, NULL); + VHOST_VSOCK_WEIGHT, true, NULL); file->private_data = vsock; spin_lock_init(&vsock->send_pkt_list_lock); -- 2.20.1
next prev parent reply other threads:[~2020-05-29 8:03 UTC|newest] Thread overview: 60+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-05-29 8:02 [PATCH 0/6] vDPA: doorbell mapping Jason Wang 2020-05-29 8:02 ` Jason Wang 2020-05-29 8:02 ` Jason Wang [this message] 2020-05-29 8:02 ` [PATCH 1/6] vhost: allow device that does not depend on vhost worker Jason Wang 2020-06-02 5:01 ` Michael S. Tsirkin 2020-06-02 7:04 ` Jason Wang 2020-06-02 7:04 ` Jason Wang 2020-05-29 8:02 ` [PATCH 2/6] vhost: use mmgrab() instead of mmget() for non worker device Jason Wang 2020-05-29 8:03 ` [PATCH 3/6] vdpa: introduce get_vq_notification method Jason Wang 2020-05-29 8:03 ` Jason Wang 2020-05-29 8:03 ` [PATCH 4/6] vhost_vdpa: support doorbell mapping via mmap Jason Wang 2020-05-29 9:16 ` Mika Penttilä 2020-05-29 9:24 ` Jason Wang 2020-05-29 18:30 ` Rob Miller 2020-05-29 18:30 ` [virtio-dev] " Rob Miller 2020-06-02 2:04 ` Jason Wang 2020-06-02 2:04 ` Jason Wang 2020-06-02 2:04 ` Jason Wang 2020-06-01 19:22 ` kbuild test robot 2020-06-01 19:22 ` kbuild test robot 2020-06-01 19:22 ` kbuild test robot 2020-06-02 4:56 ` Michael S. Tsirkin 2020-06-02 4:56 ` Michael S. Tsirkin 2020-06-02 6:49 ` Jason Wang 2020-06-02 6:49 ` Jason Wang 2020-06-02 13:31 ` Michael S. Tsirkin 2020-06-02 13:31 ` Michael S. Tsirkin 2020-06-03 4:18 ` Jason Wang 2020-06-03 4:18 ` Jason Wang 2020-06-03 6:34 ` Michael S. Tsirkin 2020-06-03 6:34 ` Michael S. Tsirkin 2020-06-03 6:37 ` Jason Wang 2020-06-03 6:37 ` Jason Wang 2020-05-29 8:03 ` [PATCH 5/6] vdpa: introduce virtio pci driver Jason Wang 2020-05-29 8:03 ` Jason Wang 2020-06-02 5:08 ` Michael S. Tsirkin 2020-06-02 7:08 ` Jason Wang 2020-06-05 8:54 ` Jason Wang 2020-06-05 8:54 ` Jason Wang 2020-06-07 13:51 ` Michael S. Tsirkin 2020-06-07 13:51 ` Michael S. Tsirkin 2020-06-08 3:32 ` Jason Wang 2020-06-08 3:32 ` Jason Wang 2020-06-08 6:32 ` Michael S. Tsirkin 2020-06-08 9:18 ` Jason Wang 2020-06-08 9:31 ` Michael S. Tsirkin 2020-06-08 9:43 ` Jason Wang 2020-06-08 9:45 ` Michael S. Tsirkin 2020-06-08 9:45 ` Michael S. Tsirkin 2020-06-08 9:46 ` Jason Wang 2020-06-08 9:54 ` Michael S. Tsirkin 2020-06-08 10:07 ` Jason Wang 2020-06-08 13:29 ` Michael S. Tsirkin 2020-06-08 13:29 ` Michael S. Tsirkin 2020-06-09 5:55 ` Jason Wang 2020-06-02 5:09 ` Michael S. Tsirkin 2020-06-02 7:12 ` Jason Wang 2020-06-04 18:50 ` Michael S. Tsirkin 2020-05-29 8:03 ` [PATCH 6/6] vdpa: vp_vdpa: report doorbell location Jason Wang 2020-05-29 8:03 ` Jason Wang
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20200529080303.15449-2-jasowang@redhat.com \ --to=jasowang@redhat.com \ --cc=eli@mellanox.com \ --cc=eperezma@redhat.com \ --cc=gdawar@xilinx.com \ --cc=hanand@xilinx.com \ --cc=kvm@vger.kernel.org \ --cc=lingshan.zhu@intel.com \ --cc=linux-kernel@vger.kernel.org \ --cc=lulu@redhat.com \ --cc=mhabets@solarflare.com \ --cc=mst@redhat.com \ --cc=netdev@vger.kernel.org \ --cc=rob.miller@broadcom.com \ --cc=saugatm@xilinx.com \ --cc=shahafs@mellanox.com \ --cc=virtualization@lists.linux-foundation.org \ --cc=vmireyno@marvell.com \ --cc=zhangweining@ruijie.com.cn \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.