From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from lists1p.gnu.org (lists1p.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id C35D3CD5BAB for ; Sun, 24 May 2026 16:42:54 +0000 (UTC) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists1p.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1wRBu1-0005wq-Cs; Sun, 24 May 2026 12:42:13 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists1p.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1wRBtz-0005wO-13 for qemu-devel@nongnu.org; Sun, 24 May 2026 12:42:11 -0400 Received: from us-smtp-delivery-124.mimecast.com ([170.10.129.124]) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1wRBtv-0007eX-GQ for qemu-devel@nongnu.org; Sun, 24 May 2026 12:42:10 -0400 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1779640925; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: in-reply-to:in-reply-to:references:references; bh=gj/pxQhMND4bnnv1xI6JUam1/dKyuuPjSu/KkdyeQY4=; b=B1HsTtcL9MNxiyB4z3Ny8XRwm4t4iUqCl3/Lrt9ncwR+XVHQlhHh4TziPW12U2lWYKI+wE 2PbRNzfpOqiafaBapPEonqJSmr2q1us8p5FAeowS3mPe7tmKAQRZL61Q8kO8X11OqUcqFH OVDyIPAdQeI4ZEorITDM/jC7brih+Kw= Received: from mail-wm1-f69.google.com (mail-wm1-f69.google.com [209.85.128.69]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.3, cipher=TLS_AES_256_GCM_SHA384) id us-mta-145-ScHy7X5VPe21WBc2hIxHHw-1; Sun, 24 May 2026 12:42:03 -0400 X-MC-Unique: ScHy7X5VPe21WBc2hIxHHw-1 X-Mimecast-MFC-AGG-ID: ScHy7X5VPe21WBc2hIxHHw_1779640922 Received: by mail-wm1-f69.google.com with SMTP id 5b1f17b1804b1-490402ae2c1so25205555e9.0 for ; Sun, 24 May 2026 09:42:03 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=google; t=1779640922; x=1780245722; darn=nongnu.org; h=in-reply-to:content-disposition:mime-version:references:message-id :subject:cc:to:from:date:from:to:cc:subject:date:message-id:reply-to; bh=gj/pxQhMND4bnnv1xI6JUam1/dKyuuPjSu/KkdyeQY4=; b=s9cSYGN7SJN9PBP5zxVGufpRVAlkDXPVJDhhHaYIJT98EI2NsVF4Spou0WnIcoQ4CK LKTmzwU7OzxftziY05GRs9zCOIwbx3p/Yx0eRhlzq5/ofHRRA00IXGfCxjGeA0IUKpCh KdiTDtalWgNn7EEcL1tyMGTwxi2XRVmIPu8zeCOEK2cBASrw18aMFHOdptk9Uu8tWL++ WV7Er/75V2o+pG7cehNuv5zspzwIOcA99ddxD2nTXE2b8TwI66g1Z5pUbcGSnzkaviqQ 7PuKH5bwNaED0s07iMiI2NtH1YRbk+H3VF+HlvkgDx10bedJckXLSVBpy2M2Gj2wFfLW o3/w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1779640922; x=1780245722; h=in-reply-to:content-disposition:mime-version:references:message-id :subject:cc:to:from:date:x-gm-gg:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=gj/pxQhMND4bnnv1xI6JUam1/dKyuuPjSu/KkdyeQY4=; b=hP5TgfmkVuVdF9Pq2Kgkaa0HvwhGrVE6aGX/nnQGg2fcWLmWbvosCVgBmKkv8MJeT3 jDl6Au0juq+H04nBtSppig2x0FtOt5c3rmD80GwTsS7NxosudsRS8Um9613iWPHflBsl ns0OCTg4PJbB7BAOhH9SdNlkvJkfICLTma1y+50/QJF2R9gneXVMnFYSaBItg4O7BaHx tK7JeZOpYNl4cF/xBVlXPQKJl4NymNOoUbHI3RXL4RS8lOI7Ft9+8owM+cXsI65TJaB+ 06A9JRhwW7WAPKjn+u+AYHImvN9SGfaBKwsEAwGclWPDi79tH8BMVMwvoNU01pQeQyY4 ocqw== X-Gm-Message-State: AOJu0YwK4ieG6OZp5tvJmEuRc0Hd3TM11L7JmN0yQmkU3ZjeKiCJKyOu CBpFzqU1lXztp9ayonqMOyd5DSz7lWLTIphLsaA5xLjbP+kjMbplesmzMGyFSWhliJUznCWNDMG KZwFBgg0YvAIUK/4QtIHvQNmBPHPQ01iJjDlgvIgxCh63tpK6anRGZP7W X-Gm-Gg: Acq92OG3BJiQNAFJTIhUbwNbkDWUEDW8W6xSgSAsm2SwqIRLhgQ2uP6L/ZSXAsii5Jw jQLRTRgx4qCNssDq79uvA3Zlv6UWLJbzg7yktDIlj93+LsIkx7+8+h4TAxSy2qqcxokAr5wLkR5 NqrNq40Fti0rFb2b5wG39Y7rp2UWUi5m6vw5VFTs5oY4IO+7z71LoZ5GTyPeADhKrg6zbj3Ede8 AZPsuIA+gnfo26jJIyCZv5PIwTpeTedm0wRWl5XwhiT8hCRIdzXmPlLKortfprVp1YaKuUw/wWT aq786/WfOCLTR9D3kGLZBVZ2yrRdO0mzMUhlQU+isWy2MBg/t0OrtQkWhnOer18QCyK0jzfdVXk /Z14e8NOzkeY5sgUVGcG9W+EosfsJsNSBZNIBubEV/dA= X-Received: by 2002:a05:600c:1992:b0:48a:79d8:a8d6 with SMTP id 5b1f17b1804b1-4904245f54cmr147790585e9.7.1779640922291; Sun, 24 May 2026 09:42:02 -0700 (PDT) X-Received: by 2002:a05:600c:1992:b0:48a:79d8:a8d6 with SMTP id 5b1f17b1804b1-4904245f54cmr147790085e9.7.1779640921665; Sun, 24 May 2026 09:42:01 -0700 (PDT) Received: from redhat.com (IGLD-80-230-25-45.inter.net.il. [80.230.25.45]) by smtp.gmail.com with ESMTPSA id 5b1f17b1804b1-490418b27b5sm66685745e9.11.2026.05.24.09.41.59 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sun, 24 May 2026 09:42:00 -0700 (PDT) Date: Sun, 24 May 2026 12:41:57 -0400 From: "Michael S. Tsirkin" To: Koushik Dutta Cc: qemu-devel@nongnu.org, Jason Wang , Eugenio =?iso-8859-1?Q?P=E9rez?= , Stefano Garzarella Subject: Re: [PATCH v7] hw/net/virtio-net: add support for notification coalescing Message-ID: <20260524123115-mutt-send-email-mst@kernel.org> References: <20260515181620.1774378-1-kdutta@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20260515181620.1774378-1-kdutta@redhat.com> Received-SPF: pass client-ip=170.10.129.124; envelope-from=mst@redhat.com; helo=us-smtp-delivery-124.mimecast.com X-Spam_score_int: -24 X-Spam_score: -2.5 X-Spam_bar: -- X-Spam_report: (-2.5 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.445, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_NONE=-0.0001, RCVD_IN_MSPIKE_H4=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: qemu development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org On Fri, May 15, 2026 at 11:46:20PM +0530, Koushik Dutta wrote: > Implement VirtIO Network Notification Coalescing (Bit 53). > This allows the guest to manage interrupt frequency using ethtool > -C for both RX and TX paths. > > - Added VIRTIO_NET_F_NOTF_COAL to host features. > - Implemented VIRTIO_NET_CTRL_NOTF_COAL class handling in > virtio_net_handle_ctrl_iov. > - Added logic to store and apply rx/tx usecs and max_packets. > - Added packet counters and threshold logic for both RX and TX data paths. > - Dynamic Dispatcher: Implemented a dispatcher mechanism that > dynamically switches/activates the notification callback logic > only after the guest enables TX coalescing via ethtool. > - After VM LM coalescing parameters persist in the destination VM. > > This reduces interrupt overhead by batching notifications based on > either a packet count or a time-based threshold. > > Signed-off-by: Koushik Dutta > --- > v7 changes: > - Fixed time unit consistency: TX path now uses microseconds matching RX > - Changed timer_new_ns to timer_new_us for TX timers > - Changed qemu_clock_get_ns to qemu_clock_get_us in TX path > - Convert txtimer from nanoseconds to microseconds (txtimer / 1000) > - Removed unused rx_index_timer variable from struct and initialization > - Fixed feature flag handling: removed conditional check in ctrl handler Thanks! Yet something to improve: > --- > hw/net/virtio-net.c | 193 +++++++++++++++++++++++++++++---- > include/hw/virtio/virtio-net.h | 9 +- > net/passt.c | 1 + > net/tap.c | 1 + > net/vhost-user.c | 1 + > net/vhost-vdpa.c | 1 + > 6 files changed, 183 insertions(+), 23 deletions(-) > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c > index 2a5d642a64..5a98b52ded 100644 > --- a/hw/net/virtio-net.c > +++ b/hw/net/virtio-net.c > @@ -157,6 +157,16 @@ static void flush_or_purge_queued_packets(NetClientState *nc) > * - we could suppress RX interrupt if we were so inclined. > */ > > +static void virtio_net_rx_notify(void *opaque) > +{ > + VirtIONetQueue *q = opaque; > + VirtIONet *n = q->n; > + VirtIODevice *vdev = VIRTIO_DEVICE(n); > + > + n->rx_pkt_cnt = 0; > + virtio_notify(vdev, q->rx_vq); > +} > + > static void virtio_net_get_config(VirtIODevice *vdev, uint8_t *config) > { > VirtIONet *n = VIRTIO_NET(vdev); > @@ -435,7 +445,7 @@ static int virtio_net_set_status(struct VirtIODevice *vdev, uint8_t status) > if (queue_started) { > if (q->tx_timer) { > timer_mod(q->tx_timer, > - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > } else { > replay_bh_schedule_event(q->tx_bh); > } > @@ -1080,6 +1090,52 @@ static int virtio_net_handle_offloads(VirtIONet *n, uint8_t cmd, > } > } > > +static void virtio_net_tx_timer(void *opaque); > + I do not like forward declarations like this - pls just order the code sensibly. If you need to move some code, it can be a separate preparatory patch. > +static int virtio_net_handle_coal(VirtIONet *n, uint8_t cmd, > + struct iovec *iov, unsigned int iov_cnt) > +{ > + struct virtio_net_ctrl_coal coal; > + VirtIONetQueue *q; > + size_t s; > + int i; > + > + s = iov_to_buf(iov, iov_cnt, 0, &coal, sizeof(coal)); > + if (s != sizeof(coal)) { > + return VIRTIO_NET_ERR; > + } > + > + if (cmd == VIRTIO_NET_CTRL_NOTF_COAL_RX_SET) { > + n->rx_coal_usecs = le32_to_cpu(coal.max_usecs); > + n->rx_coal_packets = le32_to_cpu(coal.max_packets); > + if (n->rx_coal_usecs > 0) { > + for (i = 0; i < n->max_queue_pairs; i++) { > + q = &n->vqs[i]; > + if (!q->rx_timer) { > + q->rx_timer = timer_new_us(QEMU_CLOCK_VIRTUAL, > + virtio_net_rx_notify, > + q); > + } > + } > + } > + } else if (cmd == VIRTIO_NET_CTRL_NOTF_COAL_TX_SET) { > + n->tx_coal_usecs = le32_to_cpu(coal.max_usecs); > + n->tx_coal_packets = le32_to_cpu(coal.max_packets); > + if (n->tx_coal_usecs > 0) { > + for (i = 0; i < n->max_queue_pairs; i++) { > + q = &n->vqs[i]; > + if (!q->tx_timer && n->tx_coal_usecs > 0) { > + q->tx_timer = timer_new_us(QEMU_CLOCK_VIRTUAL, > + virtio_net_tx_timer, > + q); > + } > + } > + } > + } So, the value is never propagated to any of vhost/vhost-user/vdpa. Thus if they decide to implement and advertise it, They will not get the value and will not coalesce appropriately. Makes this of a rather limited use - most deployments use some kind of offload. > + > + return VIRTIO_NET_OK; > +} > + > static int virtio_net_handle_mac(VirtIONet *n, uint8_t cmd, > struct iovec *iov, unsigned int iov_cnt) > { > @@ -1581,6 +1637,8 @@ size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev, > status = virtio_net_handle_mq(n, ctrl.cmd, iov, out_num); > } else if (ctrl.class == VIRTIO_NET_CTRL_GUEST_OFFLOADS) { > status = virtio_net_handle_offloads(n, ctrl.cmd, iov, out_num); > + } else if (ctrl.class == VIRTIO_NET_CTRL_NOTF_COAL) { > + status = virtio_net_handle_coal(n, ctrl.cmd, iov, out_num); > } > > s = iov_from_buf(in_sg, in_num, 0, &status, sizeof(status)); > @@ -2040,7 +2098,22 @@ static ssize_t virtio_net_receive_rcu(NetClientState *nc, const uint8_t *buf, > } > > virtqueue_flush(q->rx_vq, i); > - virtio_notify(vdev, q->rx_vq); > + > + /* rx coalescing */ > + n->rx_pkt_cnt += i; > + if (n->rx_coal_usecs == 0 || n->rx_pkt_cnt >= n->rx_coal_packets) { > + if (q->rx_timer) { > + timer_del(q->rx_timer); > + } > + virtio_net_rx_notify(q); > + } else { > + if (q->rx_timer) { > + if (!timer_pending(q->rx_timer)) { > + timer_mod(q->rx_timer, > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->rx_coal_usecs); > + } > + } > + } > > return size; > > @@ -2708,7 +2781,7 @@ static void virtio_net_tx_complete(NetClientState *nc, ssize_t len) > replay_bh_schedule_event(q->tx_bh); > } else { > timer_mod(q->tx_timer, > - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > } > q->tx_waiting = 1; > } > @@ -2817,7 +2890,6 @@ detach: > return -EINVAL; > } > > -static void virtio_net_tx_timer(void *opaque); > > static void virtio_net_handle_tx_timer(VirtIODevice *vdev, VirtQueue *vq) > { > @@ -2842,7 +2914,7 @@ static void virtio_net_handle_tx_timer(VirtIODevice *vdev, VirtQueue *vq) > } else { > /* re-arm timer to flush it (and more) on next tick */ > timer_mod(q->tx_timer, > - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > q->tx_waiting = 1; > virtio_queue_set_notification(vq, 0); > } > @@ -2899,6 +2971,12 @@ static void virtio_net_tx_timer(void *opaque) > if (ret == -EBUSY || ret == -EINVAL) { > return; > } > + if (n->tx_pkt_cnt < ret) { > + n->tx_pkt_cnt = 0; > + } else { > + n->tx_pkt_cnt -= ret; > + } > + > /* > * If we flush a full burst of packets, assume there are > * more coming and immediately rearm > @@ -2906,7 +2984,7 @@ static void virtio_net_tx_timer(void *opaque) > if (ret >= n->tx_burst) { > q->tx_waiting = 1; > timer_mod(q->tx_timer, > - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > return; > } > /* > @@ -2918,9 +2996,10 @@ static void virtio_net_tx_timer(void *opaque) > ret = virtio_net_flush_tx(q); > if (ret > 0) { > virtio_queue_set_notification(q->tx_vq, 0); > + n->tx_pkt_cnt -= ret; > q->tx_waiting = 1; > timer_mod(q->tx_timer, > - qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > } > } > > @@ -2973,6 +3052,32 @@ static void virtio_net_tx_bh(void *opaque) > } > } > > +static void virtio_net_handle_tx_dispatch(VirtIODevice *vdev, VirtQueue *vq) > +{ > + VirtIONet *n = VIRTIO_NET(vdev); > + VirtIONetQueue *q = &n->vqs[vq2q(virtio_get_queue_index(vq))]; > + bool use_timer = n->tx_timer_activate || n->tx_coal_usecs > 0 || > + n->tx_coal_packets > 0; > + bool pkt_limit = (n->tx_coal_packets > 0); > + > + if (use_timer) { > + n->tx_pkt_cnt++; > + if (!pkt_limit || n->tx_pkt_cnt < n->tx_coal_packets) { It seems coalescing is exclusive with bh use? > + if (q->tx_timer) { > + virtio_net_handle_tx_timer(vdev, vq); > + return; > + } > + } > + n->tx_pkt_cnt = 0; > + if (q->tx_timer) { > + timer_del(q->tx_timer); > + } > + virtio_net_handle_tx_bh(vdev, vq); > + } else { > + virtio_net_handle_tx_bh(vdev, vq); > + } > +} > + So, you unified tx handling in one place? this is better done in a preparatory patch. > static void virtio_net_add_queue(VirtIONet *n, int index) > { > VirtIODevice *vdev = VIRTIO_DEVICE(n); > @@ -2980,20 +3085,15 @@ static void virtio_net_add_queue(VirtIONet *n, int index) > n->vqs[index].rx_vq = virtio_add_queue(vdev, n->net_conf.rx_queue_size, > virtio_net_handle_rx); > > - if (n->net_conf.tx && !strcmp(n->net_conf.tx, "timer")) { > - n->vqs[index].tx_vq = > - virtio_add_queue(vdev, n->net_conf.tx_queue_size, > - virtio_net_handle_tx_timer); > - n->vqs[index].tx_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, > - virtio_net_tx_timer, > - &n->vqs[index]); > - } else { > - n->vqs[index].tx_vq = > - virtio_add_queue(vdev, n->net_conf.tx_queue_size, > - virtio_net_handle_tx_bh); > - n->vqs[index].tx_bh = qemu_bh_new_guarded(virtio_net_tx_bh, &n->vqs[index], > - &DEVICE(vdev)->mem_reentrancy_guard); > - } > + n->vqs[index].tx_vq = > + virtio_add_queue(vdev, > + n->net_conf.tx_queue_size, > + virtio_net_handle_tx_dispatch); > + > + n->vqs[index].tx_bh = > + qemu_bh_new_guarded(virtio_net_tx_bh, > + &n->vqs[index], > + &DEVICE(vdev)->mem_reentrancy_guard); > > n->vqs[index].tx_waiting = 0; > n->vqs[index].n = n; > @@ -3088,6 +3188,9 @@ static void virtio_net_get_features(VirtIODevice *vdev, uint64_t *features, > virtio_features_or(features, features, n->host_features_ex); > > virtio_add_feature_ex(features, VIRTIO_NET_F_MAC); > + if (n->tx_timer_activate) { > + virtio_clear_feature_ex(features, VIRTIO_NET_F_NOTF_COAL); > + } > > if (!peer_has_vnet_hdr(n)) { > virtio_clear_feature_ex(features, VIRTIO_NET_F_CSUM); > @@ -3242,6 +3345,35 @@ static int virtio_net_post_load_device(void *opaque, int version_id) > } > > virtio_net_commit_rss_config(n); > + if (n->tx_coal_usecs > 0 || n->rx_coal_usecs > 0) { > + > + for (i = 0; i < n->max_queue_pairs; i++) { > + VirtIONetQueue *q = &n->vqs[i]; > + if (!q->rx_timer && n->rx_coal_usecs > 0) { > + q->rx_timer = timer_new_us(QEMU_CLOCK_VIRTUAL, > + virtio_net_rx_notify, > + q); > + } > + > + if (!q->tx_timer && n->tx_coal_usecs > 0) { > + q->tx_timer = timer_new_us(QEMU_CLOCK_VIRTUAL, > + virtio_net_tx_timer, > + q); > + } > + > + if (n->tx_coal_usecs > 0 && q->tx_timer) { > + n->tx_pkt_cnt = 0; > + timer_mod(q->tx_timer, > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->tx_coal_usecs); > + } > + > + if (n->rx_coal_usecs > 0 && q->rx_timer) { > + timer_mod(q->rx_timer, > + qemu_clock_get_us(QEMU_CLOCK_VIRTUAL) + n->rx_coal_usecs); > + } > + } > + } > + > return 0; > } > > @@ -3617,6 +3749,10 @@ static const VMStateDescription vmstate_virtio_net_device = { > vmstate_virtio_net_tx_waiting), > VMSTATE_UINT64_TEST(curr_guest_offloads, VirtIONet, > has_ctrl_guest_offloads), > + VMSTATE_UINT32(rx_coal_usecs, VirtIONet), > + VMSTATE_UINT32(tx_coal_usecs, VirtIONet), > + VMSTATE_UINT32(rx_coal_packets, VirtIONet), > + VMSTATE_UINT32(tx_coal_packets, VirtIONet), > VMSTATE_END_OF_LIST() > }, > .subsections = (const VMStateDescription * const []) { > @@ -3960,7 +4096,6 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp) > } > n->vqs = g_new0(VirtIONetQueue, n->max_queue_pairs); > n->curr_queue_pairs = 1; > - n->tx_timeout = n->net_conf.txtimer; > > if (n->net_conf.tx && strcmp(n->net_conf.tx, "timer") > && strcmp(n->net_conf.tx, "bh")) { > @@ -3970,6 +4105,13 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp) > error_printf("Defaulting to \"bh\""); > } > > + if (n->net_conf.tx && strcmp(n->net_conf.tx, "timer") == 0) { > + n->tx_coal_usecs = n->net_conf.txtimer / 1000; add a code comment explaining what is going on. why is losing precision not a concern? maybe we should do the reverse and multiply? > + n->tx_timer_activate = true; > + } else { > + n->tx_coal_usecs = 0; > + } > + > n->net_conf.tx_queue_size = MIN(virtio_net_max_tx_queue_size(n), > n->net_conf.tx_queue_size); > > @@ -4046,6 +4188,11 @@ static void virtio_net_device_realize(DeviceState *dev, Error **errp) > n->rss_data.specified_hash_types.on_bits | > n->rss_data.specified_hash_types.auto_bits; > } > + n->rx_pkt_cnt = 0; > + n->tx_pkt_cnt = 0; > + n->rx_coal_usecs = 0; > + n->rx_coal_packets = 0; > + n->tx_coal_packets = 0; > } > > static void virtio_net_device_unrealize(DeviceState *dev) > @@ -4258,6 +4405,8 @@ static const Property virtio_net_properties[] = { > VIRTIO_NET_F_GUEST_USO6, true), > DEFINE_PROP_BIT64("host_uso", VirtIONet, host_features, > VIRTIO_NET_F_HOST_USO, true), > + DEFINE_PROP_BIT64("vq_notf_coal", VirtIONet, host_features, > + VIRTIO_NET_F_NOTF_COAL, true), We can't change host features like this without compat machinery. > DEFINE_PROP_ON_OFF_AUTO_BIT64("hash-ipv4", VirtIONet, > rss_data.specified_hash_types, > VIRTIO_NET_HASH_REPORT_IPv4 - 1, > diff --git a/include/hw/virtio/virtio-net.h b/include/hw/virtio/virtio-net.h > index 371e376428..b3a7df5ad8 100644 > --- a/include/hw/virtio/virtio-net.h > +++ b/include/hw/virtio/virtio-net.h > @@ -158,6 +158,7 @@ typedef struct VirtioNetRssData { > typedef struct VirtIONetQueue { > VirtQueue *rx_vq; > VirtQueue *tx_vq; > + QEMUTimer *rx_timer; > QEMUTimer *tx_timer; > QEMUBH *tx_bh; > uint32_t tx_waiting; > @@ -177,7 +178,6 @@ struct VirtIONet { > /* RSC Chains - temporary storage of coalesced data, > all these data are lost in case of migration */ > QTAILQ_HEAD(, VirtioNetRscChain) rsc_chains; > - uint32_t tx_timeout; > int32_t tx_burst; > uint32_t has_vnet_hdr; > size_t host_hdr_len; > @@ -230,6 +230,13 @@ struct VirtIONet { > struct EBPFRSSContext ebpf_rss; > uint32_t nr_ebpf_rss_fds; > char **ebpf_rss_fds; > + uint32_t rx_coal_usecs; > + uint32_t rx_coal_packets; > + uint32_t rx_pkt_cnt; > + uint32_t tx_coal_usecs; > + uint32_t tx_coal_packets; > + uint32_t tx_pkt_cnt; > + bool tx_timer_activate; > }; Can we get some documentation on what each of these is? > > size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev, > diff --git a/net/passt.c b/net/passt.c > index 4ff94ee509..0b0d9e222a 100644 > --- a/net/passt.c > +++ b/net/passt.c > @@ -52,6 +52,7 @@ static const int user_feature_bits[] = { > VIRTIO_NET_F_GUEST_USO4, > VIRTIO_NET_F_GUEST_USO6, > VIRTIO_NET_F_HOST_USO, > + VIRTIO_NET_F_NOTF_COAL, > > /* This bit implies RARP isn't sent by QEMU out of band */ > VIRTIO_NET_F_GUEST_ANNOUNCE, > diff --git a/net/tap.c b/net/tap.c > index 8d7ab6ba6f..ea5987a3dc 100644 > --- a/net/tap.c > +++ b/net/tap.c > @@ -62,6 +62,7 @@ static const int kernel_feature_bits[] = { > VIRTIO_F_NOTIFICATION_DATA, > VIRTIO_NET_F_RSC_EXT, > VIRTIO_NET_F_HASH_REPORT, > + VIRTIO_NET_F_NOTF_COAL, > VIRTIO_NET_F_GUEST_UDP_TUNNEL_GSO, > VIRTIO_NET_F_HOST_UDP_TUNNEL_GSO, > VHOST_INVALID_FEATURE_BIT > diff --git a/net/vhost-user.c b/net/vhost-user.c > index a4bb49bbcf..f0b3752d7c 100644 > --- a/net/vhost-user.c > +++ b/net/vhost-user.c > @@ -54,6 +54,7 @@ static const int user_feature_bits[] = { > VIRTIO_NET_F_GUEST_USO4, > VIRTIO_NET_F_GUEST_USO6, > VIRTIO_NET_F_HOST_USO, > + VIRTIO_NET_F_NOTF_COAL, > > /* This bit implies RARP isn't sent by QEMU out of band */ > VIRTIO_NET_F_GUEST_ANNOUNCE, > diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c > index 3df6091274..4ab8f26ceb 100644 > --- a/net/vhost-vdpa.c > +++ b/net/vhost-vdpa.c > @@ -70,6 +70,7 @@ static const int vdpa_feature_bits[] = { > VIRTIO_NET_F_CTRL_RX, > VIRTIO_NET_F_CTRL_RX_EXTRA, > VIRTIO_NET_F_CTRL_VLAN, > + VIRTIO_NET_F_NOTF_COAL, > VIRTIO_NET_F_CTRL_VQ, > VIRTIO_NET_F_GSO, > VIRTIO_NET_F_GUEST_CSUM, > -- > 2.53.0