Re: [PATCH virtio v2 12/13] virtio_pci_modern: use completion instead of busy loop to wait on admin cmd result

["Michael S. Tsirkin" <mst@redhat.com>]

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 <jiri@nvidia.com>
> >> 
> >> 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 <jiri@nvidia.com>
> >> ---
> >> 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.


> 
> >You need to add a callback and signal the completion.
> 
> Callback to what?

for example, remove callback can signal completion.


> 
> 
> 
> >
> >-- 
> >MST
> >