From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-lj1-f178.google.com (mail-lj1-f178.google.com [209.85.208.178]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 30631158DAE for ; Thu, 11 Jul 2024 08:11:01 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.208.178 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1720685465; cv=none; b=WQdVVpIQvMswxfHnBQnwnJx/PAAo1i7MuLQeNccQdYeiSZf1XvAvs9GmuagczZDmJ3U2LvDp8ffEP+LJXbsth7voP7hwHYfQekANEPE8KdQMqaC77S43U8u8su1UQWBJk1v77ZgAgnIHKIHnYXCw8ZFO3s7H4uBC22yWwEV5s5U= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1720685465; c=relaxed/simple; bh=nAWH+kqovGyLzHOA4tYUSNTDuQGHcfZnaJKc0XwAgn4=; h=Date:From:To:Cc:Subject:Message-ID:References:MIME-Version: Content-Type:Content-Disposition:In-Reply-To; b=naNRiT/DYNoqWfbqWLoDp2oDtgpMqD0IUDNLm+FsxHGXrBm0MEkbwtx/JM4rIVM36jWeV0uG8Y5t/9/AN9Z42xUC+h00l3greoy3m+RXAdCEqNSSWlTdbgeasMpuD3r1NP8FICbOOtWDnPU/oCMMc1I8SFrLJlBkhAv1lVGJfOA= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=none (p=none dis=none) header.from=resnulli.us; spf=none smtp.mailfrom=resnulli.us; dkim=pass (2048-bit key) header.d=resnulli-us.20230601.gappssmtp.com header.i=@resnulli-us.20230601.gappssmtp.com header.b=SVOmnhkP; arc=none smtp.client-ip=209.85.208.178 Authentication-Results: smtp.subspace.kernel.org; dmarc=none (p=none dis=none) header.from=resnulli.us Authentication-Results: smtp.subspace.kernel.org; spf=none smtp.mailfrom=resnulli.us Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=resnulli-us.20230601.gappssmtp.com header.i=@resnulli-us.20230601.gappssmtp.com header.b="SVOmnhkP" Received: by mail-lj1-f178.google.com with SMTP id 38308e7fff4ca-2ebe40673e8so7509631fa.3 for ; Thu, 11 Jul 2024 01:11:01 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=resnulli-us.20230601.gappssmtp.com; s=20230601; t=1720685460; x=1721290260; darn=lists.linux.dev; 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=1ECIVSyBZeD5Sr44m029+DyqMzD5KxWWXnVzF5mUsh8=; b=SVOmnhkPg20yz5eHiByVQUfuhbBLgIcAqvKXWWiwSjSgkkyLxWLvpunpX+JtHgdyuL QlUIpSD5dWhmx9wwAazGG4TCTgUBrsXRML9MGXHtsiA1nVsadkg1IHDwLhp9HYJGN8UM YCu6PRma+Tq6Is/F0G1DyJhoXNIhOHMY9f9PpUw2N4bvE6HAXFqtIhmUPVTEWa7EkGad cR2ibgwPoLr9H/4643tGUXLdFaULKsuUQnUahSZWx30xnon/T+o8UAZZ9te1n9dDVHqm reSazbqVcGJM3lbH8ne/NBRJ8CJAeoYF5MW59dXnYXGQkT9No108pQR4zfzPMKHW7J15 IauA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1720685460; x=1721290260; h=in-reply-to:content-disposition:mime-version:references:message-id :subject:cc:to:from:date:x-gm-message-state:from:to:cc:subject:date :message-id:reply-to; bh=1ECIVSyBZeD5Sr44m029+DyqMzD5KxWWXnVzF5mUsh8=; b=LlcmzyDte3owp1ZUisiCSFpZ6aClHG3GZ42DK9ggMd7mC/aKsKAicL1g7YkcMwZctq 18d/+1yeUGvHamgd9b3fhZDNDA+AI6BDFVYNivLAp06+Q2dQHxJPIZyv4W9pN9J2GBzo JgzhEiL52aOoVhrL8B17YahBAT3UWSeKL/3XsMiPqVHWwGBxfD4rsKDyKEWS6347/CIY h+FzLVMnRWNsx7zIPkzexBp353vXpDF5t9yzytiCcFu+qSVEDG1fGws3zDjtLkGHrJ0A Sr3FPQQufL5q+DGdD+5v8JRC13QjEeW4CDU6GyGJdsPjfbhWV3b8pV89L0sAAdd4DJhB 2Njw== X-Gm-Message-State: AOJu0YwEir2m3J0ulVzWZMRqakIJtURtp8MNraDSINFiMTfQtjKlaxgM 8mChDnGfRpBKeaMpjyNqRot+2vmtvHO9BYkDPsogtYASrYvJiUc0GDTyWSjhvWM= X-Google-Smtp-Source: AGHT+IEzp42ox2embjz5mqBO9hbK+nGxZ/NOOBUgoiQCAL8pcwdZbz03MmWsw/oHpMA/87ymv5OefA== X-Received: by 2002:a19:a414:0:b0:52d:582e:8093 with SMTP id 2adb3069b0e04-52eb9995541mr3725657e87.23.1720685459885; Thu, 11 Jul 2024 01:10:59 -0700 (PDT) Received: from localhost ([37.48.50.18]) by smtp.gmail.com with ESMTPSA id 5b1f17b1804b1-4266f6f1138sm107131405e9.19.2024.07.11.01.10.58 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 11 Jul 2024 01:10:59 -0700 (PDT) Date: Thu, 11 Jul 2024 10:10:57 +0200 From: Jiri Pirko To: "Michael S. Tsirkin" Cc: virtualization@lists.linux.dev, jasowang@redhat.com, xuanzhuo@linux.alibaba.com, eperezma@redhat.com, parav@nvidia.com, feliu@nvidia.com, hengqi@linux.alibaba.com Subject: Re: [PATCH virtio v2 12/13] virtio_pci_modern: use completion instead of busy loop to wait on admin cmd result Message-ID: References: <20240710063601.2000149-1-jiri@resnulli.us> <20240710063601.2000149-13-jiri@resnulli.us> <20240710073925-mutt-send-email-mst@kernel.org> <20240710090418-mutt-send-email-mst@kernel.org> Precedence: bulk X-Mailing-List: virtualization@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20240710090418-mutt-send-email-mst@kernel.org> Wed, Jul 10, 2024 at 03:06:57PM CEST, mst@redhat.com wrote: >On Wed, Jul 10, 2024 at 03:03:33PM +0200, Jiri Pirko wrote: >> Wed, Jul 10, 2024 at 01:47:22PM CEST, mst@redhat.com wrote: >> >On Wed, Jul 10, 2024 at 08:36:00AM +0200, Jiri Pirko wrote: >> >> From: Jiri Pirko >> >> >> >> Currently, the code waits in a busy loop on every admin virtqueue issued >> >> command to get a reply. That prevents callers from issuing multiple >> >> commands in parallel. >> >> >> >> To overcome this limitation, introduce a virtqueue event callback for >> >> admin virtqueue. For every issued command, use completion mechanism >> >> to wait on a reply. In the event callback, trigger the completion >> >> is done for every incoming reply. >> >> >> >> Alongside with that, introduce a spin lock to protect the admin >> >> virtqueue operations. >> >> >> >> Signed-off-by: Jiri Pirko >> >> --- >> >> v1->v2: >> >> - rebased on top of newly added patches >> >> - rebased on top of changes in previous patches (vq info, vqs[]) >> >> - removed WARN_ON_ONCE() when calling virtqueue_kick() >> >> - added virtqueue_is_broken check in virtqueue_exec_admin_cmd() loop >> >> - added vp_modern_avq_cleanup() implementation to handle surprise >> >> removal case >> >> --- >> >> drivers/virtio/virtio_pci_common.c | 13 ++++-- >> >> drivers/virtio/virtio_pci_common.h | 3 ++ >> >> drivers/virtio/virtio_pci_modern.c | 74 +++++++++++++++++++++++++----- >> >> include/linux/virtio.h | 3 ++ >> >> 4 files changed, 77 insertions(+), 16 deletions(-) >> >> >> >> diff --git a/drivers/virtio/virtio_pci_common.c b/drivers/virtio/virtio_pci_common.c >> >> index 267643bb1cd5..c44d8ba00c02 100644 >> >> --- a/drivers/virtio/virtio_pci_common.c >> >> +++ b/drivers/virtio/virtio_pci_common.c >> >> @@ -395,6 +395,8 @@ static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned int nvqs, >> >> if (vqi->name && vqi->callback) >> >> ++nvectors; >> >> } >> >> + if (avq_num && vector_policy == VP_VQ_VECTOR_POLICY_EACH) >> >> + ++nvectors; >> >> } else { >> >> /* Second best: one for change, shared for all vqs. */ >> >> nvectors = 2; >> >> @@ -425,9 +427,9 @@ static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned int nvqs, >> >> if (!avq_num) >> >> return 0; >> >> sprintf(avq->name, "avq.%u", avq->vq_index); >> >> - vq = vp_find_one_vq_msix(vdev, avq->vq_index, NULL, avq->name, false, >> >> - true, &allocated_vectors, vector_policy, >> >> - &vp_dev->admin_vq.info); >> >> + vq = vp_find_one_vq_msix(vdev, avq->vq_index, vp_modern_avq_done, >> >> + avq->name, false, true, &allocated_vectors, >> >> + vector_policy, &vp_dev->admin_vq.info); >> >> if (IS_ERR(vq)) { >> >> err = PTR_ERR(vq); >> >> goto error_find; >> >> @@ -486,8 +488,9 @@ static int vp_find_vqs_intx(struct virtio_device *vdev, unsigned int nvqs, >> >> if (!avq_num) >> >> return 0; >> >> sprintf(avq->name, "avq.%u", avq->vq_index); >> >> - vq = vp_setup_vq(vdev, queue_idx++, NULL, avq->name, false, >> >> - VIRTIO_MSI_NO_VECTOR, &vp_dev->admin_vq.info); >> >> + vq = vp_setup_vq(vdev, queue_idx++, vp_modern_avq_done, avq->name, >> >> + false, VIRTIO_MSI_NO_VECTOR, >> >> + &vp_dev->admin_vq.info); >> >> if (IS_ERR(vq)) { >> >> err = PTR_ERR(vq); >> >> goto out_del_vqs; >> >> diff --git a/drivers/virtio/virtio_pci_common.h b/drivers/virtio/virtio_pci_common.h >> >> index de59bb06ec3c..90df381fbbcf 100644 >> >> --- a/drivers/virtio/virtio_pci_common.h >> >> +++ b/drivers/virtio/virtio_pci_common.h >> >> @@ -47,6 +47,8 @@ struct virtio_pci_admin_vq { >> >> struct virtio_pci_vq_info *info; >> >> /* serializing admin commands execution. */ >> >> struct mutex cmd_lock; >> >> + /* Protects virtqueue access. */ >> >> + spinlock_t lock; >> >> u64 supported_cmds; >> >> /* Name of the admin queue: avq.$vq_index. */ >> >> char name[10]; >> >> @@ -178,6 +180,7 @@ struct virtio_device *virtio_pci_vf_get_pf_dev(struct pci_dev *pdev); >> >> #define VIRTIO_ADMIN_CMD_BITMAP 0 >> >> #endif >> >> >> >> +void vp_modern_avq_done(struct virtqueue *vq); >> >> int vp_modern_admin_cmd_exec(struct virtio_device *vdev, >> >> struct virtio_admin_cmd *cmd); >> >> >> >> diff --git a/drivers/virtio/virtio_pci_modern.c b/drivers/virtio/virtio_pci_modern.c >> >> index 0fd344d1eaf9..608df3263df1 100644 >> >> --- a/drivers/virtio/virtio_pci_modern.c >> >> +++ b/drivers/virtio/virtio_pci_modern.c >> >> @@ -53,6 +53,23 @@ static bool vp_is_avq(struct virtio_device *vdev, unsigned int index) >> >> return index == vp_dev->admin_vq.vq_index; >> >> } >> >> >> >> +void vp_modern_avq_done(struct virtqueue *vq) >> >> +{ >> >> + struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev); >> >> + struct virtio_pci_admin_vq *admin_vq = &vp_dev->admin_vq; >> >> + struct virtio_admin_cmd *cmd; >> >> + unsigned long flags; >> >> + unsigned int len; >> >> + >> >> + spin_lock_irqsave(&admin_vq->lock, flags); >> >> + do { >> >> + virtqueue_disable_cb(vq); >> >> + while ((cmd = virtqueue_get_buf(vq, &len))) >> >> + complete(&cmd->completion); >> >> + } while (!virtqueue_enable_cb(vq)); >> >> + spin_unlock_irqrestore(&admin_vq->lock, flags); >> >> +} >> >> + >> >> static int virtqueue_exec_admin_cmd(struct virtio_pci_admin_vq *admin_vq, >> >> u16 opcode, >> >> struct scatterlist **sgs, >> >> @@ -61,7 +78,8 @@ static int virtqueue_exec_admin_cmd(struct virtio_pci_admin_vq *admin_vq, >> >> struct virtio_admin_cmd *cmd) >> >> { >> >> struct virtqueue *vq; >> >> - int ret, len; >> >> + unsigned long flags; >> >> + int ret; >> >> >> >> vq = admin_vq->info->vq; >> >> if (!vq) >> >> @@ -72,21 +90,33 @@ static int virtqueue_exec_admin_cmd(struct virtio_pci_admin_vq *admin_vq, >> >> !((1ULL << opcode) & admin_vq->supported_cmds)) >> >> return -EOPNOTSUPP; >> >> >> >> - ret = virtqueue_add_sgs(vq, sgs, out_num, in_num, cmd, GFP_KERNEL); >> >> - if (ret < 0) >> >> - return -EIO; >> >> + init_completion(&cmd->completion); >> >> >> >> - if (unlikely(!virtqueue_kick(vq))) >> >> +again: >> >> + if (virtqueue_is_broken(vq)) >> >> return -EIO; >> >> >> >> - while (!virtqueue_get_buf(vq, &len) && >> >> - !virtqueue_is_broken(vq)) >> >> - cpu_relax(); >> >> + spin_lock_irqsave(&admin_vq->lock, flags); >> >> + ret = virtqueue_add_sgs(vq, sgs, out_num, in_num, cmd, GFP_KERNEL); >> >> + if (ret < 0) { >> >> + if (ret == -ENOSPC) { >> >> + spin_unlock_irqrestore(&admin_vq->lock, flags); >> >> + cpu_relax(); >> >> + goto again; >> >> + } >> >> + goto unlock_err; >> >> + } >> >> + if (!virtqueue_kick(vq)) >> >> + goto unlock_err; >> >> + spin_unlock_irqrestore(&admin_vq->lock, flags); >> >> >> >> - if (virtqueue_is_broken(vq)) >> >> - return -EIO; >> >> + wait_for_completion(&cmd->completion); >> >> >> >> - return 0; >> >> + return cmd->ret; >> >> + >> >> +unlock_err: >> >> + spin_unlock_irqrestore(&admin_vq->lock, flags); >> >> + return -EIO; >> >> } >> >> >> >> int vp_modern_admin_cmd_exec(struct virtio_device *vdev, >> >> @@ -209,6 +239,25 @@ static void vp_modern_avq_activate(struct virtio_device *vdev) >> >> virtio_pci_admin_cmd_list_init(vdev); >> >> } >> >> >> >> +static void vp_modern_avq_cleanup(struct virtio_device *vdev) >> >> +{ >> >> + struct virtio_pci_device *vp_dev = to_vp_device(vdev); >> >> + struct virtio_admin_cmd *cmd; >> >> + struct virtqueue *vq; >> >> + >> >> + if (!virtio_has_feature(vdev, VIRTIO_F_ADMIN_VQ)) >> >> + return; >> >> + >> >> + vq = vp_dev->vqs[vp_dev->admin_vq.vq_index]->vq; >> >> + if (!vq) >> >> + return; >> >> + >> >> + while ((cmd = virtqueue_detach_unused_buf(vq))) { >> >> + cmd->ret = -EIO; >> >> + complete(&cmd->completion); >> >> + } >> >> +} >> >> + >> > >> > >> >I think surprise removal is still broken with this. >> >> Why do you think so? I will drain the remaining buffers from the queue >> and complete them all. What's missing? > >You have this: > >> >> + wait_for_completion(&cmd->completion); > >if surprise removal triggers after you submitted command >but before it was used, no callback triggers and so no >completion either. Well. In that case virtio_pci_remove() -> unregister_virtio_device() -> device_unregister() -> virtnet_remove() -> remove_vq_common() -> virtio_reset_device() -> vp_reset() -> vp_modern_avq_cleanup() which goes over these cmds and does complete them all. See vp_modern_avq_cleanup() above. What am I missing? > > >> >> >You need to add a callback and signal the completion. >> >> Callback to what? > >for example, remove callback can signal completion. > > >> >> >> >> > >> >-- >> >MST >> > >