diff mbox series

[v7] virtio-net: Fix network stall at the host side waiting for kick

Message ID 20240701140017.49223-1-east.moutain.yang@gmail.com (mailing list archive)
State New, archived
Headers show
Series [v7] virtio-net: Fix network stall at the host side waiting for kick | expand

Commit Message

Wencheng Yang July 1, 2024, 2 p.m. UTC
From: thomas <east.moutain.yang@gmail.com>

Patch 06b12970174 ("virtio-net: fix network stall under load")
added double-check to test whether the available buffer size
can satisfy the request or not, in case the guest has added
some buffers to the avail ring simultaneously after the first
check. It will be lucky if the available buffer size becomes
okay after the double-check, then the host can send the packet
to the guest. If the buffer size still can't satisfy the request,
even if the guest has added some buffers, viritio-net would
stall at the host side forever.

The patch checks whether the guest has added some buffers
after last check of avail idx when the available buffers are
not sufficient, if so then recheck the available buffers
in the loop.

The patch also reverts patch "06b12970174".

The case below can reproduce the stall.

                                       Guest 0
                                     +--------+
                                     | iperf  |
                    ---------------> | server |
         Host       |                +--------+
       +--------+   |                    ...
       | iperf  |----
       | client |----                  Guest n
       +--------+   |                +--------+
                    |                | iperf  |
                    ---------------> | server |
                                     +--------+

Boot many guests from qemu with virtio network:
 qemu ... -netdev tap,id=net_x \
    -device virtio-net-pci-non-transitional,\
    iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x

Each guest acts as iperf server with commands below:
 iperf3 -s -D -i 10 -p 8001
 iperf3 -s -D -i 10 -p 8002

The host as iperf client:
 iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
 iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000

After some time, the host loses connection to the guest,
the guest can send packet to the host, but can't receive
packet from host.

It's more likely to happen if SWIOTLB is enabled in the guest,
allocating and freeing bounce buffer takes some CPU ticks,
copying from/to bounce buffer takes more CPU ticks, compared
with that there is no bounce buffer in the guest.
Once the rate of producing packets from the host approximates
the rate of receiveing packets in the guest, the guest would
loop in NAPI.

         receive packets    ---
               |             |
               v             |
           free buf      virtnet_poll
               |             |
               v             |
     add buf to avail ring  ---
               |
               |  need kick the host?
               |  NAPI continues
               v
         receive packets    ---
               |             |
               v             |
           free buf      virtnet_poll
               |             |
               v             |
     add buf to avail ring  ---
               |
               v
              ...           ...

On the other hand, the host fetches free buf from avail
ring, if the buf in the avail ring is not enough, the
host notifies the guest the event by writing the avail
idx read from avail ring to the event idx of used ring,
then the host goes to sleep, waiting for the kick signal
from the guest.

Once the guest finds the host is waiting for kick singal
(in virtqueue_kick_prepare_split()), it kicks the host.

The host may stall forever at the sequences below:

         Host                        Guest
     ------------                 -----------
 fetch buf, send packet           receive packet ---
         ...                          ...         |
 fetch buf, send packet             add buf       |
         ...                        add buf   virtnet_poll
    buf not enough      avail idx-> add buf       |
    read avail idx                  add buf       |
                                    add buf      ---
                                  receive packet ---
    write event idx                   ...         |
    wait for kick                   add buf   virtnet_poll
                                      ...         |
                                                 ---
                                 no more packet, exit NAPI

In the first loop of NAPI above, indicated in the range of
virtnet_poll above, the host is sending packets while the
guest is receiving packets and adding buffers.
 step 1: The buf is not enough, for example, a big packet
         needs 5 buf, but the available buf count is 3.
         The host read current avail idx.
 step 2: The guest adds some buf, then checks whether the
         host is waiting for kick signal, not at this time.
         The used ring is not empty, the guest continues
         the second loop of NAPI.
 step 3: The host writes the avail idx read from avail
         ring to used ring as event idx via
         virtio_queue_set_notification(q->rx_vq, 1).
 step 4: At the end of the second loop of NAPI, recheck
         whether kick is needed, as the event idx in the
         used ring written by the host is beyound the
         range of kick condition, the guest will not
         send kick signal to the host.

Fixes: 06b12970174 ("virtio-net: fix network stall under load")
Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
---

Changelog:
v7:
- Add function virtio_queue_set_notification_and_check()
- Restore the function sequence introduce in v6

v6:
- Take packed packed queue into cosideration
- Adjust function sequence to fix compilation issue

v5:
- Modify return type of virtio_queue_set_notification() to
  bool to indicate whether the guest has added some buffers
  after last check of avail idx
- Loop in virtio_net_has_buffers() if the available buffers
  are not sufficient and the guest has added some buffers.
- Revert patch "06b12970174"
- Update the subject

v4:
- Correct spelling mistake in the subject
- Describe the issue that virtio-net is blocked at host side

v3:
- Add virtio-net tag in the subject
- Refine commit log

v2:
- Add SOB tag at the end of the commit message
- Place Fixes tag at the end of the commit message

v1:
- Initial patch

 hw/net/virtio-net.c        | 19 ++++++++-----------
 hw/virtio/virtio.c         | 39 ++++++++++++++++++++++++++++++++++++++
 include/hw/virtio/virtio.h |  1 +
 3 files changed, 48 insertions(+), 11 deletions(-)

Comments

Michael S. Tsirkin July 1, 2024, 3:05 p.m. UTC | #1
On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> From: thomas <east.moutain.yang@gmail.com>
> 
> Patch 06b12970174 ("virtio-net: fix network stall under load")
> added double-check to test whether the available buffer size
> can satisfy the request or not, in case the guest has added
> some buffers to the avail ring simultaneously after the first
> check. It will be lucky if the available buffer size becomes
> okay after the double-check, then the host can send the packet
> to the guest. If the buffer size still can't satisfy the request,
> even if the guest has added some buffers, viritio-net would
> stall at the host side forever.
> 
> The patch checks whether the guest has added some buffers
> after last check of avail idx when the available buffers are
> not sufficient, if so then recheck the available buffers
> in the loop.
> 
> The patch also reverts patch "06b12970174".
> 
> The case below can reproduce the stall.
> 
>                                        Guest 0
>                                      +--------+
>                                      | iperf  |
>                     ---------------> | server |
>          Host       |                +--------+
>        +--------+   |                    ...
>        | iperf  |----
>        | client |----                  Guest n
>        +--------+   |                +--------+
>                     |                | iperf  |
>                     ---------------> | server |
>                                      +--------+
> 
> Boot many guests from qemu with virtio network:
>  qemu ... -netdev tap,id=net_x \
>     -device virtio-net-pci-non-transitional,\
>     iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> 
> Each guest acts as iperf server with commands below:
>  iperf3 -s -D -i 10 -p 8001
>  iperf3 -s -D -i 10 -p 8002
> 
> The host as iperf client:
>  iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
>  iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> 
> After some time, the host loses connection to the guest,
> the guest can send packet to the host, but can't receive
> packet from host.
> 
> It's more likely to happen if SWIOTLB is enabled in the guest,
> allocating and freeing bounce buffer takes some CPU ticks,
> copying from/to bounce buffer takes more CPU ticks, compared
> with that there is no bounce buffer in the guest.
> Once the rate of producing packets from the host approximates
> the rate of receiveing packets in the guest, the guest would
> loop in NAPI.
> 
>          receive packets    ---
>                |             |
>                v             |
>            free buf      virtnet_poll
>                |             |
>                v             |
>      add buf to avail ring  ---
>                |
>                |  need kick the host?
>                |  NAPI continues
>                v
>          receive packets    ---
>                |             |
>                v             |
>            free buf      virtnet_poll
>                |             |
>                v             |
>      add buf to avail ring  ---
>                |
>                v
>               ...           ...
> 
> On the other hand, the host fetches free buf from avail
> ring, if the buf in the avail ring is not enough, the
> host notifies the guest the event by writing the avail
> idx read from avail ring to the event idx of used ring,
> then the host goes to sleep, waiting for the kick signal
> from the guest.
> 
> Once the guest finds the host is waiting for kick singal
> (in virtqueue_kick_prepare_split()), it kicks the host.
> 
> The host may stall forever at the sequences below:
> 
>          Host                        Guest
>      ------------                 -----------
>  fetch buf, send packet           receive packet ---
>          ...                          ...         |
>  fetch buf, send packet             add buf       |
>          ...                        add buf   virtnet_poll
>     buf not enough      avail idx-> add buf       |
>     read avail idx                  add buf       |
>                                     add buf      ---
>                                   receive packet ---
>     write event idx                   ...         |
>     wait for kick                   add buf   virtnet_poll
>                                       ...         |
>                                                  ---
>                                  no more packet, exit NAPI
> 
> In the first loop of NAPI above, indicated in the range of
> virtnet_poll above, the host is sending packets while the
> guest is receiving packets and adding buffers.
>  step 1: The buf is not enough, for example, a big packet
>          needs 5 buf, but the available buf count is 3.
>          The host read current avail idx.
>  step 2: The guest adds some buf, then checks whether the
>          host is waiting for kick signal, not at this time.
>          The used ring is not empty, the guest continues
>          the second loop of NAPI.
>  step 3: The host writes the avail idx read from avail
>          ring to used ring as event idx via
>          virtio_queue_set_notification(q->rx_vq, 1).
>  step 4: At the end of the second loop of NAPI, recheck
>          whether kick is needed, as the event idx in the
>          used ring written by the host is beyound the
>          range of kick condition, the guest will not
>          send kick signal to the host.
> 
> Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
> ---
> 
> Changelog:
> v7:
> - Add function virtio_queue_set_notification_and_check()
> - Restore the function sequence introduce in v6
> 
> v6:
> - Take packed packed queue into cosideration
> - Adjust function sequence to fix compilation issue
> 
> v5:
> - Modify return type of virtio_queue_set_notification() to
>   bool to indicate whether the guest has added some buffers
>   after last check of avail idx
> - Loop in virtio_net_has_buffers() if the available buffers
>   are not sufficient and the guest has added some buffers.
> - Revert patch "06b12970174"
> - Update the subject
> 
> v4:
> - Correct spelling mistake in the subject
> - Describe the issue that virtio-net is blocked at host side
> 
> v3:
> - Add virtio-net tag in the subject
> - Refine commit log
> 
> v2:
> - Add SOB tag at the end of the commit message
> - Place Fixes tag at the end of the commit message
> 
> v1:
> - Initial patch
> 
>  hw/net/virtio-net.c        | 19 ++++++++-----------
>  hw/virtio/virtio.c         | 39 ++++++++++++++++++++++++++++++++++++++
>  include/hw/virtio/virtio.h |  1 +
>  3 files changed, 48 insertions(+), 11 deletions(-)
> 
> diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> index 9c7e85caea..357c651d9b 100644
> --- a/hw/net/virtio-net.c
> +++ b/hw/net/virtio-net.c
> @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive(NetClientState *nc)
>  static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
>  {
>      VirtIONet *n = q->n;
> -    if (virtio_queue_empty(q->rx_vq) ||
> -        (n->mergeable_rx_bufs &&
> -         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> -        virtio_queue_set_notification(q->rx_vq, 1);
> -
> -        /* To avoid a race condition where the guest has made some buffers
> -         * available after the above check but before notification was
> -         * enabled, check for available buffers again.
> -         */
> -        if (virtio_queue_empty(q->rx_vq) ||
> +
> +    while (virtio_queue_empty(q->rx_vq) ||
>              (n->mergeable_rx_bufs &&
> -             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> +            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> +        /* guest may have made some buf, try again */
> +        if (virtio_queue_set_notification_and_check(q->rx_vq, 1)) {
> +            virtio_queue_set_notification(q->rx_vq, 0);
> +            continue;
> +        } else {
>              return 0;
>          }
>      }
> diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> index 893a072c9d..b165764017 100644
> --- a/hw/virtio/virtio.c
> +++ b/hw/virtio/virtio.c
> @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t flags, bool wrap_counter)
>      return (avail != used) && (avail == wrap_counter);
>  }
>  
> +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> +{
> +    uint16_t shadow_idx;
> +    VRingPackedDesc desc;
> +    VRingMemoryRegionCaches *caches;
> +
> +    vq->notification = enable;
> +
> +    if (!vq->vring.desc) {
> +        return false;
> +    }
> +
> +    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> +        virtio_queue_packed_set_notification(vq, enable);
> +
> +        if (enable) {
> +            caches = vring_get_region_caches(vq);
> +            if (!caches) {
> +                return false;
> +            }
> +
> +            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> +                                   vq->shadow_avail_idx, true);
> +            if (is_desc_avail(desc.flags, vq->shadow_avail_wrap_counter)) {
> +                return true;
> +            }
> +        }
> +    } else {
> +        shadow_idx = vq->shadow_avail_idx;
> +        virtio_queue_split_set_notification(vq, enable);
> +
> +        if (enable) {
> +            return shadow_idx != vring_avail_idx(vq);
> +        }
> +    }
> +
> +    return false;
> +}

Please document what this does.
So this will return false if ring has any availabe buffers?
Equivalent to:

bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
{
        virtio_queue_packed_set_notification(vq, enable);
	return virtio_queue_empty(vq);
}


or am I missing something?

> +
>  /* Fetch avail_idx from VQ memory only when we really need to know if
>   * guest has added some buffers.
>   * Called within rcu_read_lock().  */
> diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
> index 7d5ffdc145..ed85023b87 100644
> --- a/include/hw/virtio/virtio.h
> +++ b/include/hw/virtio/virtio.h
> @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
>  
>  bool virtio_queue_get_notification(VirtQueue *vq);
>  void virtio_queue_set_notification(VirtQueue *vq, int enable);
> +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable);
>  
>  int virtio_queue_ready(VirtQueue *vq);
>  
> -- 
> 2.39.0
Wencheng Yang July 1, 2024, 5:18 p.m. UTC | #2
> Please document what this does.
okay, i will.

> So this will return false if ring has any available buffers?
> Equivalent to:
>
> bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> {
>        virtio_queue_packed_set_notification(vq, enable);
>         return virtio_queue_empty(vq);
> }

No, only when the *shadow_avail_idx *is changed shall the function return
true,
compared with the value seen by the host last time, else return false even
if
there are some buffers available in the queue, as the total size of the
available
buffers in the queue can't satisfy the request.

It maybe better to pass only one arg to the function like this:
bool virtio_queue_set_notification_and_check(VirtQueue *vq)
{
       virtio_queue_packed_set_notification(vq, true);

       return shadow_avail_idx_changed()? true: false;
}

Thanks Michael a lot!

On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com> wrote:

> On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > From: thomas <east.moutain.yang@gmail.com>
> >
> > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > added double-check to test whether the available buffer size
> > can satisfy the request or not, in case the guest has added
> > some buffers to the avail ring simultaneously after the first
> > check. It will be lucky if the available buffer size becomes
> > okay after the double-check, then the host can send the packet
> > to the guest. If the buffer size still can't satisfy the request,
> > even if the guest has added some buffers, viritio-net would
> > stall at the host side forever.
> >
> > The patch checks whether the guest has added some buffers
> > after last check of avail idx when the available buffers are
> > not sufficient, if so then recheck the available buffers
> > in the loop.
> >
> > The patch also reverts patch "06b12970174".
> >
> > The case below can reproduce the stall.
> >
> >                                        Guest 0
> >                                      +--------+
> >                                      | iperf  |
> >                     ---------------> | server |
> >          Host       |                +--------+
> >        +--------+   |                    ...
> >        | iperf  |----
> >        | client |----                  Guest n
> >        +--------+   |                +--------+
> >                     |                | iperf  |
> >                     ---------------> | server |
> >                                      +--------+
> >
> > Boot many guests from qemu with virtio network:
> >  qemu ... -netdev tap,id=net_x \
> >     -device virtio-net-pci-non-transitional,\
> >     iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> >
> > Each guest acts as iperf server with commands below:
> >  iperf3 -s -D -i 10 -p 8001
> >  iperf3 -s -D -i 10 -p 8002
> >
> > The host as iperf client:
> >  iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> >  iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> >
> > After some time, the host loses connection to the guest,
> > the guest can send packet to the host, but can't receive
> > packet from host.
> >
> > It's more likely to happen if SWIOTLB is enabled in the guest,
> > allocating and freeing bounce buffer takes some CPU ticks,
> > copying from/to bounce buffer takes more CPU ticks, compared
> > with that there is no bounce buffer in the guest.
> > Once the rate of producing packets from the host approximates
> > the rate of receiveing packets in the guest, the guest would
> > loop in NAPI.
> >
> >          receive packets    ---
> >                |             |
> >                v             |
> >            free buf      virtnet_poll
> >                |             |
> >                v             |
> >      add buf to avail ring  ---
> >                |
> >                |  need kick the host?
> >                |  NAPI continues
> >                v
> >          receive packets    ---
> >                |             |
> >                v             |
> >            free buf      virtnet_poll
> >                |             |
> >                v             |
> >      add buf to avail ring  ---
> >                |
> >                v
> >               ...           ...
> >
> > On the other hand, the host fetches free buf from avail
> > ring, if the buf in the avail ring is not enough, the
> > host notifies the guest the event by writing the avail
> > idx read from avail ring to the event idx of used ring,
> > then the host goes to sleep, waiting for the kick signal
> > from the guest.
> >
> > Once the guest finds the host is waiting for kick singal
> > (in virtqueue_kick_prepare_split()), it kicks the host.
> >
> > The host may stall forever at the sequences below:
> >
> >          Host                        Guest
> >      ------------                 -----------
> >  fetch buf, send packet           receive packet ---
> >          ...                          ...         |
> >  fetch buf, send packet             add buf       |
> >          ...                        add buf   virtnet_poll
> >     buf not enough      avail idx-> add buf       |
> >     read avail idx                  add buf       |
> >                                     add buf      ---
> >                                   receive packet ---
> >     write event idx                   ...         |
> >     wait for kick                   add buf   virtnet_poll
> >                                       ...         |
> >                                                  ---
> >                                  no more packet, exit NAPI
> >
> > In the first loop of NAPI above, indicated in the range of
> > virtnet_poll above, the host is sending packets while the
> > guest is receiving packets and adding buffers.
> >  step 1: The buf is not enough, for example, a big packet
> >          needs 5 buf, but the available buf count is 3.
> >          The host read current avail idx.
> >  step 2: The guest adds some buf, then checks whether the
> >          host is waiting for kick signal, not at this time.
> >          The used ring is not empty, the guest continues
> >          the second loop of NAPI.
> >  step 3: The host writes the avail idx read from avail
> >          ring to used ring as event idx via
> >          virtio_queue_set_notification(q->rx_vq, 1).
> >  step 4: At the end of the second loop of NAPI, recheck
> >          whether kick is needed, as the event idx in the
> >          used ring written by the host is beyound the
> >          range of kick condition, the guest will not
> >          send kick signal to the host.
> >
> > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
> > ---
> >
> > Changelog:
> > v7:
> > - Add function virtio_queue_set_notification_and_check()
> > - Restore the function sequence introduce in v6
> >
> > v6:
> > - Take packed packed queue into cosideration
> > - Adjust function sequence to fix compilation issue
> >
> > v5:
> > - Modify return type of virtio_queue_set_notification() to
> >   bool to indicate whether the guest has added some buffers
> >   after last check of avail idx
> > - Loop in virtio_net_has_buffers() if the available buffers
> >   are not sufficient and the guest has added some buffers.
> > - Revert patch "06b12970174"
> > - Update the subject
> >
> > v4:
> > - Correct spelling mistake in the subject
> > - Describe the issue that virtio-net is blocked at host side
> >
> > v3:
> > - Add virtio-net tag in the subject
> > - Refine commit log
> >
> > v2:
> > - Add SOB tag at the end of the commit message
> > - Place Fixes tag at the end of the commit message
> >
> > v1:
> > - Initial patch
> >
> >  hw/net/virtio-net.c        | 19 ++++++++-----------
> >  hw/virtio/virtio.c         | 39 ++++++++++++++++++++++++++++++++++++++
> >  include/hw/virtio/virtio.h |  1 +
> >  3 files changed, 48 insertions(+), 11 deletions(-)
> >
> > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > index 9c7e85caea..357c651d9b 100644
> > --- a/hw/net/virtio-net.c
> > +++ b/hw/net/virtio-net.c
> > @@ -1642,18 +1642,15 @@ static bool
> virtio_net_can_receive(NetClientState *nc)
> >  static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> >  {
> >      VirtIONet *n = q->n;
> > -    if (virtio_queue_empty(q->rx_vq) ||
> > -        (n->mergeable_rx_bufs &&
> > -         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > -        virtio_queue_set_notification(q->rx_vq, 1);
> > -
> > -        /* To avoid a race condition where the guest has made some
> buffers
> > -         * available after the above check but before notification was
> > -         * enabled, check for available buffers again.
> > -         */
> > -        if (virtio_queue_empty(q->rx_vq) ||
> > +
> > +    while (virtio_queue_empty(q->rx_vq) ||
> >              (n->mergeable_rx_bufs &&
> > -             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > +            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > +        /* guest may have made some buf, try again */
> > +        if (virtio_queue_set_notification_and_check(q->rx_vq, 1)) {
> > +            virtio_queue_set_notification(q->rx_vq, 0);
> > +            continue;
> > +        } else {
> >              return 0;
> >          }
> >      }
> > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > index 893a072c9d..b165764017 100644
> > --- a/hw/virtio/virtio.c
> > +++ b/hw/virtio/virtio.c
> > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t flags,
> bool wrap_counter)
> >      return (avail != used) && (avail == wrap_counter);
> >  }
> >
> > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > +{
> > +    uint16_t shadow_idx;
> > +    VRingPackedDesc desc;
> > +    VRingMemoryRegionCaches *caches;
> > +
> > +    vq->notification = enable;
> > +
> > +    if (!vq->vring.desc) {
> > +        return false;
> > +    }
> > +
> > +    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > +        virtio_queue_packed_set_notification(vq, enable);
> > +
> > +        if (enable) {
> > +            caches = vring_get_region_caches(vq);
> > +            if (!caches) {
> > +                return false;
> > +            }
> > +
> > +            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > +                                   vq->shadow_avail_idx, true);
> > +            if (is_desc_avail(desc.flags,
> vq->shadow_avail_wrap_counter)) {
> > +                return true;
> > +            }
> > +        }
> > +    } else {
> > +        shadow_idx = vq->shadow_avail_idx;
> > +        virtio_queue_split_set_notification(vq, enable);
> > +
> > +        if (enable) {
> > +            return shadow_idx != vring_avail_idx(vq);
> > +        }
> > +    }
> > +
> > +    return false;
> > +}
>
> Please document what this does.
> So this will return false if ring has any availabe buffers?
> Equivalent to:
>
> bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> {
>         virtio_queue_packed_set_notification(vq, enable);
>         return virtio_queue_empty(vq);
> }
>
>
> or am I missing something?
>
> > +
> >  /* Fetch avail_idx from VQ memory only when we really need to know if
> >   * guest has added some buffers.
> >   * Called within rcu_read_lock().  */
> > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
> > index 7d5ffdc145..ed85023b87 100644
> > --- a/include/hw/virtio/virtio.h
> > +++ b/include/hw/virtio/virtio.h
> > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> >
> >  bool virtio_queue_get_notification(VirtQueue *vq);
> >  void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable);
> >
> >  int virtio_queue_ready(VirtQueue *vq);
> >
> > --
> > 2.39.0
>
>
Michael S. Tsirkin July 1, 2024, 6:46 p.m. UTC | #3
On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > Please document what this does.
> okay, i will.
> 
> > So this will return false if ring has any available buffers?
> > Equivalent to:
> > 
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > {
> >        virtio_queue_packed_set_notification(vq, enable);
> >         return virtio_queue_empty(vq);
> > }
> 
> No, only when the shadow_avail_idx is changed shall the function return true,


what does "changed" mean here? changed compared to what?

> compared with the value seen by the host last time, else return false even if
> there are some buffers available in the queue, as the total size of the
> available
> buffers in the queue can't satisfy the request.
> 
> It maybe better to pass only one arg to the function like this:
> bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> {
>        virtio_queue_packed_set_notification(vq, true);
>        
>        return shadow_avail_idx_changed()? true: false;
> }
> 
> Thanks Michael a lot!
> 
> On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com> wrote:
> 
>     On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
>     > From: thomas <east.moutain.yang@gmail.com>
>     >
>     > Patch 06b12970174 ("virtio-net: fix network stall under load")
>     > added double-check to test whether the available buffer size
>     > can satisfy the request or not, in case the guest has added
>     > some buffers to the avail ring simultaneously after the first
>     > check. It will be lucky if the available buffer size becomes
>     > okay after the double-check, then the host can send the packet
>     > to the guest. If the buffer size still can't satisfy the request,
>     > even if the guest has added some buffers, viritio-net would
>     > stall at the host side forever.
>     >
>     > The patch checks whether the guest has added some buffers
>     > after last check of avail idx when the available buffers are
>     > not sufficient, if so then recheck the available buffers
>     > in the loop.
>     >
>     > The patch also reverts patch "06b12970174".
>     >
>     > The case below can reproduce the stall.
>     >
>     >                                        Guest 0
>     >                                      +--------+
>     >                                      | iperf  |
>     >                     ---------------> | server |
>     >          Host       |                +--------+
>     >        +--------+   |                    ...
>     >        | iperf  |----
>     >        | client |----                  Guest n
>     >        +--------+   |                +--------+
>     >                     |                | iperf  |
>     >                     ---------------> | server |
>     >                                      +--------+
>     >
>     > Boot many guests from qemu with virtio network:
>     >  qemu ... -netdev tap,id=net_x \
>     >     -device virtio-net-pci-non-transitional,\
>     >     iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
>     >
>     > Each guest acts as iperf server with commands below:
>     >  iperf3 -s -D -i 10 -p 8001
>     >  iperf3 -s -D -i 10 -p 8002
>     >
>     > The host as iperf client:
>     >  iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
>     >  iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
>     >
>     > After some time, the host loses connection to the guest,
>     > the guest can send packet to the host, but can't receive
>     > packet from host.
>     >
>     > It's more likely to happen if SWIOTLB is enabled in the guest,
>     > allocating and freeing bounce buffer takes some CPU ticks,
>     > copying from/to bounce buffer takes more CPU ticks, compared
>     > with that there is no bounce buffer in the guest.
>     > Once the rate of producing packets from the host approximates
>     > the rate of receiveing packets in the guest, the guest would
>     > loop in NAPI.
>     >
>     >          receive packets    ---
>     >                |             |
>     >                v             |
>     >            free buf      virtnet_poll
>     >                |             |
>     >                v             |
>     >      add buf to avail ring  ---
>     >                |
>     >                |  need kick the host?
>     >                |  NAPI continues
>     >                v
>     >          receive packets    ---
>     >                |             |
>     >                v             |
>     >            free buf      virtnet_poll
>     >                |             |
>     >                v             |
>     >      add buf to avail ring  ---
>     >                |
>     >                v
>     >               ...           ...
>     >
>     > On the other hand, the host fetches free buf from avail
>     > ring, if the buf in the avail ring is not enough, the
>     > host notifies the guest the event by writing the avail
>     > idx read from avail ring to the event idx of used ring,
>     > then the host goes to sleep, waiting for the kick signal
>     > from the guest.
>     >
>     > Once the guest finds the host is waiting for kick singal
>     > (in virtqueue_kick_prepare_split()), it kicks the host.
>     >
>     > The host may stall forever at the sequences below:
>     >
>     >          Host                        Guest
>     >      ------------                 -----------
>     >  fetch buf, send packet           receive packet ---
>     >          ...                          ...         |
>     >  fetch buf, send packet             add buf       |
>     >          ...                        add buf   virtnet_poll
>     >     buf not enough      avail idx-> add buf       |
>     >     read avail idx                  add buf       |
>     >                                     add buf      ---
>     >                                   receive packet ---
>     >     write event idx                   ...         |
>     >     wait for kick                   add buf   virtnet_poll
>     >                                       ...         |
>     >                                                  ---
>     >                                  no more packet, exit NAPI
>     >
>     > In the first loop of NAPI above, indicated in the range of
>     > virtnet_poll above, the host is sending packets while the
>     > guest is receiving packets and adding buffers.
>     >  step 1: The buf is not enough, for example, a big packet
>     >          needs 5 buf, but the available buf count is 3.
>     >          The host read current avail idx.
>     >  step 2: The guest adds some buf, then checks whether the
>     >          host is waiting for kick signal, not at this time.
>     >          The used ring is not empty, the guest continues
>     >          the second loop of NAPI.
>     >  step 3: The host writes the avail idx read from avail
>     >          ring to used ring as event idx via
>     >          virtio_queue_set_notification(q->rx_vq, 1).
>     >  step 4: At the end of the second loop of NAPI, recheck
>     >          whether kick is needed, as the event idx in the
>     >          used ring written by the host is beyound the
>     >          range of kick condition, the guest will not
>     >          send kick signal to the host.
>     >
>     > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
>     > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
>     > ---
>     >
>     > Changelog:
>     > v7:
>     > - Add function virtio_queue_set_notification_and_check()
>     > - Restore the function sequence introduce in v6
>     >
>     > v6:
>     > - Take packed packed queue into cosideration
>     > - Adjust function sequence to fix compilation issue
>     >
>     > v5:
>     > - Modify return type of virtio_queue_set_notification() to
>     >   bool to indicate whether the guest has added some buffers
>     >   after last check of avail idx
>     > - Loop in virtio_net_has_buffers() if the available buffers
>     >   are not sufficient and the guest has added some buffers.
>     > - Revert patch "06b12970174"
>     > - Update the subject
>     >
>     > v4:
>     > - Correct spelling mistake in the subject
>     > - Describe the issue that virtio-net is blocked at host side
>     >
>     > v3:
>     > - Add virtio-net tag in the subject
>     > - Refine commit log
>     >
>     > v2:
>     > - Add SOB tag at the end of the commit message
>     > - Place Fixes tag at the end of the commit message
>     >
>     > v1:
>     > - Initial patch
>     >
>     >  hw/net/virtio-net.c        | 19 ++++++++-----------
>     >  hw/virtio/virtio.c         | 39 ++++++++++++++++++++++++++++++++++++++
>     >  include/hw/virtio/virtio.h |  1 +
>     >  3 files changed, 48 insertions(+), 11 deletions(-)
>     >
>     > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
>     > index 9c7e85caea..357c651d9b 100644
>     > --- a/hw/net/virtio-net.c
>     > +++ b/hw/net/virtio-net.c
>     > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive(NetClientState
>     *nc)
>     >  static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
>     >  {
>     >      VirtIONet *n = q->n;
>     > -    if (virtio_queue_empty(q->rx_vq) ||
>     > -        (n->mergeable_rx_bufs &&
>     > -         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     > -        virtio_queue_set_notification(q->rx_vq, 1);
>     > -
>     > -        /* To avoid a race condition where the guest has made some
>     buffers
>     > -         * available after the above check but before notification was
>     > -         * enabled, check for available buffers again.
>     > -         */
>     > -        if (virtio_queue_empty(q->rx_vq) ||
>     > +
>     > +    while (virtio_queue_empty(q->rx_vq) ||
>     >              (n->mergeable_rx_bufs &&
>     > -             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     > +            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     > +        /* guest may have made some buf, try again */
>     > +        if (virtio_queue_set_notification_and_check(q->rx_vq, 1)) {
>     > +            virtio_queue_set_notification(q->rx_vq, 0);
>     > +            continue;
>     > +        } else {
>     >              return 0;
>     >          }
>     >      }
>     > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
>     > index 893a072c9d..b165764017 100644
>     > --- a/hw/virtio/virtio.c
>     > +++ b/hw/virtio/virtio.c
>     > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t flags,
>     bool wrap_counter)
>     >      return (avail != used) && (avail == wrap_counter);
>     >  }
>     > 
>     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
>     > +{
>     > +    uint16_t shadow_idx;
>     > +    VRingPackedDesc desc;
>     > +    VRingMemoryRegionCaches *caches;
>     > +
>     > +    vq->notification = enable;
>     > +
>     > +    if (!vq->vring.desc) {
>     > +        return false;
>     > +    }
>     > +
>     > +    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
>     > +        virtio_queue_packed_set_notification(vq, enable);
>     > +
>     > +        if (enable) {
>     > +            caches = vring_get_region_caches(vq);
>     > +            if (!caches) {
>     > +                return false;
>     > +            }
>     > +
>     > +            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
>     > +                                   vq->shadow_avail_idx, true);
>     > +            if (is_desc_avail(desc.flags, vq->
>     shadow_avail_wrap_counter)) {
>     > +                return true;
>     > +            }
>     > +        }
>     > +    } else {
>     > +        shadow_idx = vq->shadow_avail_idx;
>     > +        virtio_queue_split_set_notification(vq, enable);
>     > +
>     > +        if (enable) {
>     > +            return shadow_idx != vring_avail_idx(vq);
>     > +        }
>     > +    }
>     > +
>     > +    return false;
>     > +}
> 
>     Please document what this does.
>     So this will return false if ring has any availabe buffers?
>     Equivalent to:
> 
>     bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
>     {
>             virtio_queue_packed_set_notification(vq, enable);
>             return virtio_queue_empty(vq);
>     }
> 
> 
>     or am I missing something?
> 
>     > +
>     >  /* Fetch avail_idx from VQ memory only when we really need to know if
>     >   * guest has added some buffers.
>     >   * Called within rcu_read_lock().  */
>     > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
>     > index 7d5ffdc145..ed85023b87 100644
>     > --- a/include/hw/virtio/virtio.h
>     > +++ b/include/hw/virtio/virtio.h
>     > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
>     > 
>     >  bool virtio_queue_get_notification(VirtQueue *vq);
>     >  void virtio_queue_set_notification(VirtQueue *vq, int enable);
>     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable);
>     > 
>     >  int virtio_queue_ready(VirtQueue *vq);
>     > 
>     > --
>     > 2.39.0
> 
>
Wencheng Yang July 1, 2024, 11:45 p.m. UTC | #4
> what does "changed" mean here? changed compared to what?
For a split queue, if the shadow_avail_idx synced from avail ring idx
by *vring_avail_idx(vq) last time doesn't equal the current value of avail
ring idx.*

*vq->shadow_avail_idx **!= vring_avail_idx(vq);*

For packed queue, the logic is similar, if vq->shadow_avail_idx

becomes available, it means the guest has added buf at the slot.

*vring_packed_desc_read(vq->vdev, &desc, &caches->desc,*

                        *vq->shadow_avail_idx, true);*

*if (is_desc_avail(desc.flags, vq-> **shadow_avail_wrap_counter))*

*    return true;*



On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com> wrote:

> On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > Please document what this does.
> > okay, i will.
> >
> > > So this will return false if ring has any available buffers?
> > > Equivalent to:
> > >
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > {
> > >        virtio_queue_packed_set_notification(vq, enable);
> > >         return virtio_queue_empty(vq);
> > > }
> >
> > No, only when the shadow_avail_idx is changed shall the function return
> true,
>
>
> what does "changed" mean here? changed compared to what?
>
> > compared with the value seen by the host last time, else return false
> even if
> > there are some buffers available in the queue, as the total size of the
> > available
> > buffers in the queue can't satisfy the request.
> >
> > It maybe better to pass only one arg to the function like this:
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > {
> >        virtio_queue_packed_set_notification(vq, true);
> >
> >        return shadow_avail_idx_changed()? true: false;
> > }
> >
> > Thanks Michael a lot!
> >
> > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com>
> wrote:
> >
> >     On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> >     > From: thomas <east.moutain.yang@gmail.com>
> >     >
> >     > Patch 06b12970174 ("virtio-net: fix network stall under load")
> >     > added double-check to test whether the available buffer size
> >     > can satisfy the request or not, in case the guest has added
> >     > some buffers to the avail ring simultaneously after the first
> >     > check. It will be lucky if the available buffer size becomes
> >     > okay after the double-check, then the host can send the packet
> >     > to the guest. If the buffer size still can't satisfy the request,
> >     > even if the guest has added some buffers, viritio-net would
> >     > stall at the host side forever.
> >     >
> >     > The patch checks whether the guest has added some buffers
> >     > after last check of avail idx when the available buffers are
> >     > not sufficient, if so then recheck the available buffers
> >     > in the loop.
> >     >
> >     > The patch also reverts patch "06b12970174".
> >     >
> >     > The case below can reproduce the stall.
> >     >
> >     >                                        Guest 0
> >     >                                      +--------+
> >     >                                      | iperf  |
> >     >                     ---------------> | server |
> >     >          Host       |                +--------+
> >     >        +--------+   |                    ...
> >     >        | iperf  |----
> >     >        | client |----                  Guest n
> >     >        +--------+   |                +--------+
> >     >                     |                | iperf  |
> >     >                     ---------------> | server |
> >     >                                      +--------+
> >     >
> >     > Boot many guests from qemu with virtio network:
> >     >  qemu ... -netdev tap,id=net_x \
> >     >     -device virtio-net-pci-non-transitional,\
> >     >     iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> >     >
> >     > Each guest acts as iperf server with commands below:
> >     >  iperf3 -s -D -i 10 -p 8001
> >     >  iperf3 -s -D -i 10 -p 8002
> >     >
> >     > The host as iperf client:
> >     >  iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> >     >  iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> >     >
> >     > After some time, the host loses connection to the guest,
> >     > the guest can send packet to the host, but can't receive
> >     > packet from host.
> >     >
> >     > It's more likely to happen if SWIOTLB is enabled in the guest,
> >     > allocating and freeing bounce buffer takes some CPU ticks,
> >     > copying from/to bounce buffer takes more CPU ticks, compared
> >     > with that there is no bounce buffer in the guest.
> >     > Once the rate of producing packets from the host approximates
> >     > the rate of receiveing packets in the guest, the guest would
> >     > loop in NAPI.
> >     >
> >     >          receive packets    ---
> >     >                |             |
> >     >                v             |
> >     >            free buf      virtnet_poll
> >     >                |             |
> >     >                v             |
> >     >      add buf to avail ring  ---
> >     >                |
> >     >                |  need kick the host?
> >     >                |  NAPI continues
> >     >                v
> >     >          receive packets    ---
> >     >                |             |
> >     >                v             |
> >     >            free buf      virtnet_poll
> >     >                |             |
> >     >                v             |
> >     >      add buf to avail ring  ---
> >     >                |
> >     >                v
> >     >               ...           ...
> >     >
> >     > On the other hand, the host fetches free buf from avail
> >     > ring, if the buf in the avail ring is not enough, the
> >     > host notifies the guest the event by writing the avail
> >     > idx read from avail ring to the event idx of used ring,
> >     > then the host goes to sleep, waiting for the kick signal
> >     > from the guest.
> >     >
> >     > Once the guest finds the host is waiting for kick singal
> >     > (in virtqueue_kick_prepare_split()), it kicks the host.
> >     >
> >     > The host may stall forever at the sequences below:
> >     >
> >     >          Host                        Guest
> >     >      ------------                 -----------
> >     >  fetch buf, send packet           receive packet ---
> >     >          ...                          ...         |
> >     >  fetch buf, send packet             add buf       |
> >     >          ...                        add buf   virtnet_poll
> >     >     buf not enough      avail idx-> add buf       |
> >     >     read avail idx                  add buf       |
> >     >                                     add buf      ---
> >     >                                   receive packet ---
> >     >     write event idx                   ...         |
> >     >     wait for kick                   add buf   virtnet_poll
> >     >                                       ...         |
> >     >                                                  ---
> >     >                                  no more packet, exit NAPI
> >     >
> >     > In the first loop of NAPI above, indicated in the range of
> >     > virtnet_poll above, the host is sending packets while the
> >     > guest is receiving packets and adding buffers.
> >     >  step 1: The buf is not enough, for example, a big packet
> >     >          needs 5 buf, but the available buf count is 3.
> >     >          The host read current avail idx.
> >     >  step 2: The guest adds some buf, then checks whether the
> >     >          host is waiting for kick signal, not at this time.
> >     >          The used ring is not empty, the guest continues
> >     >          the second loop of NAPI.
> >     >  step 3: The host writes the avail idx read from avail
> >     >          ring to used ring as event idx via
> >     >          virtio_queue_set_notification(q->rx_vq, 1).
> >     >  step 4: At the end of the second loop of NAPI, recheck
> >     >          whether kick is needed, as the event idx in the
> >     >          used ring written by the host is beyound the
> >     >          range of kick condition, the guest will not
> >     >          send kick signal to the host.
> >     >
> >     > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> >     > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
> >     > ---
> >     >
> >     > Changelog:
> >     > v7:
> >     > - Add function virtio_queue_set_notification_and_check()
> >     > - Restore the function sequence introduce in v6
> >     >
> >     > v6:
> >     > - Take packed packed queue into cosideration
> >     > - Adjust function sequence to fix compilation issue
> >     >
> >     > v5:
> >     > - Modify return type of virtio_queue_set_notification() to
> >     >   bool to indicate whether the guest has added some buffers
> >     >   after last check of avail idx
> >     > - Loop in virtio_net_has_buffers() if the available buffers
> >     >   are not sufficient and the guest has added some buffers.
> >     > - Revert patch "06b12970174"
> >     > - Update the subject
> >     >
> >     > v4:
> >     > - Correct spelling mistake in the subject
> >     > - Describe the issue that virtio-net is blocked at host side
> >     >
> >     > v3:
> >     > - Add virtio-net tag in the subject
> >     > - Refine commit log
> >     >
> >     > v2:
> >     > - Add SOB tag at the end of the commit message
> >     > - Place Fixes tag at the end of the commit message
> >     >
> >     > v1:
> >     > - Initial patch
> >     >
> >     >  hw/net/virtio-net.c        | 19 ++++++++-----------
> >     >  hw/virtio/virtio.c         | 39
> ++++++++++++++++++++++++++++++++++++++
> >     >  include/hw/virtio/virtio.h |  1 +
> >     >  3 files changed, 48 insertions(+), 11 deletions(-)
> >     >
> >     > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> >     > index 9c7e85caea..357c651d9b 100644
> >     > --- a/hw/net/virtio-net.c
> >     > +++ b/hw/net/virtio-net.c
> >     > @@ -1642,18 +1642,15 @@ static bool
> virtio_net_can_receive(NetClientState
> >     *nc)
> >     >  static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> >     >  {
> >     >      VirtIONet *n = q->n;
> >     > -    if (virtio_queue_empty(q->rx_vq) ||
> >     > -        (n->mergeable_rx_bufs &&
> >     > -         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> >     > -        virtio_queue_set_notification(q->rx_vq, 1);
> >     > -
> >     > -        /* To avoid a race condition where the guest has made some
> >     buffers
> >     > -         * available after the above check but before
> notification was
> >     > -         * enabled, check for available buffers again.
> >     > -         */
> >     > -        if (virtio_queue_empty(q->rx_vq) ||
> >     > +
> >     > +    while (virtio_queue_empty(q->rx_vq) ||
> >     >              (n->mergeable_rx_bufs &&
> >     > -             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> >     > +            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> >     > +        /* guest may have made some buf, try again */
> >     > +        if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> {
> >     > +            virtio_queue_set_notification(q->rx_vq, 0);
> >     > +            continue;
> >     > +        } else {
> >     >              return 0;
> >     >          }
> >     >      }
> >     > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> >     > index 893a072c9d..b165764017 100644
> >     > --- a/hw/virtio/virtio.c
> >     > +++ b/hw/virtio/virtio.c
> >     > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> flags,
> >     bool wrap_counter)
> >     >      return (avail != used) && (avail == wrap_counter);
> >     >  }
> >     >
> >     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> >     > +{
> >     > +    uint16_t shadow_idx;
> >     > +    VRingPackedDesc desc;
> >     > +    VRingMemoryRegionCaches *caches;
> >     > +
> >     > +    vq->notification = enable;
> >     > +
> >     > +    if (!vq->vring.desc) {
> >     > +        return false;
> >     > +    }
> >     > +
> >     > +    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> >     > +        virtio_queue_packed_set_notification(vq, enable);
> >     > +
> >     > +        if (enable) {
> >     > +            caches = vring_get_region_caches(vq);
> >     > +            if (!caches) {
> >     > +                return false;
> >     > +            }
> >     > +
> >     > +            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> >     > +                                   vq->shadow_avail_idx, true);
> >     > +            if (is_desc_avail(desc.flags, vq->
> >     shadow_avail_wrap_counter)) {
> >     > +                return true;
> >     > +            }
> >     > +        }
> >     > +    } else {
> >     > +        shadow_idx = vq->shadow_avail_idx;
> >     > +        virtio_queue_split_set_notification(vq, enable);
> >     > +
> >     > +        if (enable) {
> >     > +            return shadow_idx != vring_avail_idx(vq);
> >     > +        }
> >     > +    }
> >     > +
> >     > +    return false;
> >     > +}
> >
> >     Please document what this does.
> >     So this will return false if ring has any availabe buffers?
> >     Equivalent to:
> >
> >     bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> >     {
> >             virtio_queue_packed_set_notification(vq, enable);
> >             return virtio_queue_empty(vq);
> >     }
> >
> >
> >     or am I missing something?
> >
> >     > +
> >     >  /* Fetch avail_idx from VQ memory only when we really need to
> know if
> >     >   * guest has added some buffers.
> >     >   * Called within rcu_read_lock().  */
> >     > diff --git a/include/hw/virtio/virtio.h
> b/include/hw/virtio/virtio.h
> >     > index 7d5ffdc145..ed85023b87 100644
> >     > --- a/include/hw/virtio/virtio.h
> >     > +++ b/include/hw/virtio/virtio.h
> >     > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> >     >
> >     >  bool virtio_queue_get_notification(VirtQueue *vq);
> >     >  void virtio_queue_set_notification(VirtQueue *vq, int enable);
> >     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable);
> >     >
> >     >  int virtio_queue_ready(VirtQueue *vq);
> >     >
> >     > --
> >     > 2.39.0
> >
> >
>
>
Michael S. Tsirkin July 2, 2024, 11:27 a.m. UTC | #5
On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > what does "changed" mean here? changed compared to what?
> For a split queue, if the shadow_avail_idx synced from avail ring idx
> by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> idx.
> 
> vq->shadow_avail_idx != vring_avail_idx(vq);
> 
> For packed queue, the logic is similar, if vq->shadow_avail_idx
> 
> becomes available, it means the guest has added buf at the slot.
> 
> vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> 
>                         vq->shadow_avail_idx, true);
> 
> if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> 
>     return true;
> 
> 

This answer does not make sense from API POV.

My suggestion:

	change virtqueue_get_avail_bytes to return the shadow
	in an opaque unsigned value.

	add virtqueue_poll that gets this opaque and tells us whether any new
	buffers became available in the queue since that value
	was returned.

	accordingly, virtio_queue_set_notification_and_check
	will accept this opaque value and check avail buffers
	against it.



> On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com> wrote:
> 
>     On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
>     > > Please document what this does.
>     > okay, i will.
>     >
>     > > So this will return false if ring has any available buffers?
>     > > Equivalent to:
>     > > 
>     > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
>     > > {
>     > >        virtio_queue_packed_set_notification(vq, enable);
>     > >         return virtio_queue_empty(vq);
>     > > }
>     >
>     > No, only when the shadow_avail_idx is changed shall the function return
>     true,
> 
> 
>     what does "changed" mean here? changed compared to what?
> 
>     > compared with the value seen by the host last time, else return false
>     even if
>     > there are some buffers available in the queue, as the total size of the
>     > available
>     > buffers in the queue can't satisfy the request.
>     >
>     > It maybe better to pass only one arg to the function like this:
>     > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
>     > {
>     >        virtio_queue_packed_set_notification(vq, true);
>     >        
>     >        return shadow_avail_idx_changed()? true: false;
>     > }
>     >
>     > Thanks Michael a lot!
>     >
>     > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com>
>     wrote:
>     >
>     >     On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
>     >     > From: thomas <east.moutain.yang@gmail.com>
>     >     >
>     >     > Patch 06b12970174 ("virtio-net: fix network stall under load")
>     >     > added double-check to test whether the available buffer size
>     >     > can satisfy the request or not, in case the guest has added
>     >     > some buffers to the avail ring simultaneously after the first
>     >     > check. It will be lucky if the available buffer size becomes
>     >     > okay after the double-check, then the host can send the packet
>     >     > to the guest. If the buffer size still can't satisfy the request,
>     >     > even if the guest has added some buffers, viritio-net would
>     >     > stall at the host side forever.
>     >     >
>     >     > The patch checks whether the guest has added some buffers
>     >     > after last check of avail idx when the available buffers are
>     >     > not sufficient, if so then recheck the available buffers
>     >     > in the loop.
>     >     >
>     >     > The patch also reverts patch "06b12970174".
>     >     >
>     >     > The case below can reproduce the stall.
>     >     >
>     >     >                                        Guest 0
>     >     >                                      +--------+
>     >     >                                      | iperf  |
>     >     >                     ---------------> | server |
>     >     >          Host       |                +--------+
>     >     >        +--------+   |                    ...
>     >     >        | iperf  |----
>     >     >        | client |----                  Guest n
>     >     >        +--------+   |                +--------+
>     >     >                     |                | iperf  |
>     >     >                     ---------------> | server |
>     >     >                                      +--------+
>     >     >
>     >     > Boot many guests from qemu with virtio network:
>     >     >  qemu ... -netdev tap,id=net_x \
>     >     >     -device virtio-net-pci-non-transitional,\
>     >     >     iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
>     >     >
>     >     > Each guest acts as iperf server with commands below:
>     >     >  iperf3 -s -D -i 10 -p 8001
>     >     >  iperf3 -s -D -i 10 -p 8002
>     >     >
>     >     > The host as iperf client:
>     >     >  iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
>     >     >  iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
>     >     >
>     >     > After some time, the host loses connection to the guest,
>     >     > the guest can send packet to the host, but can't receive
>     >     > packet from host.
>     >     >
>     >     > It's more likely to happen if SWIOTLB is enabled in the guest,
>     >     > allocating and freeing bounce buffer takes some CPU ticks,
>     >     > copying from/to bounce buffer takes more CPU ticks, compared
>     >     > with that there is no bounce buffer in the guest.
>     >     > Once the rate of producing packets from the host approximates
>     >     > the rate of receiveing packets in the guest, the guest would
>     >     > loop in NAPI.
>     >     >
>     >     >          receive packets    ---
>     >     >                |             |
>     >     >                v             |
>     >     >            free buf      virtnet_poll
>     >     >                |             |
>     >     >                v             |
>     >     >      add buf to avail ring  ---
>     >     >                |
>     >     >                |  need kick the host?
>     >     >                |  NAPI continues
>     >     >                v
>     >     >          receive packets    ---
>     >     >                |             |
>     >     >                v             |
>     >     >            free buf      virtnet_poll
>     >     >                |             |
>     >     >                v             |
>     >     >      add buf to avail ring  ---
>     >     >                |
>     >     >                v
>     >     >               ...           ...
>     >     >
>     >     > On the other hand, the host fetches free buf from avail
>     >     > ring, if the buf in the avail ring is not enough, the
>     >     > host notifies the guest the event by writing the avail
>     >     > idx read from avail ring to the event idx of used ring,
>     >     > then the host goes to sleep, waiting for the kick signal
>     >     > from the guest.
>     >     >
>     >     > Once the guest finds the host is waiting for kick singal
>     >     > (in virtqueue_kick_prepare_split()), it kicks the host.
>     >     >
>     >     > The host may stall forever at the sequences below:
>     >     >
>     >     >          Host                        Guest
>     >     >      ------------                 -----------
>     >     >  fetch buf, send packet           receive packet ---
>     >     >          ...                          ...         |
>     >     >  fetch buf, send packet             add buf       |
>     >     >          ...                        add buf   virtnet_poll
>     >     >     buf not enough      avail idx-> add buf       |
>     >     >     read avail idx                  add buf       |
>     >     >                                     add buf      ---
>     >     >                                   receive packet ---
>     >     >     write event idx                   ...         |
>     >     >     wait for kick                   add buf   virtnet_poll
>     >     >                                       ...         |
>     >     >                                                  ---
>     >     >                                  no more packet, exit NAPI
>     >     >
>     >     > In the first loop of NAPI above, indicated in the range of
>     >     > virtnet_poll above, the host is sending packets while the
>     >     > guest is receiving packets and adding buffers.
>     >     >  step 1: The buf is not enough, for example, a big packet
>     >     >          needs 5 buf, but the available buf count is 3.
>     >     >          The host read current avail idx.
>     >     >  step 2: The guest adds some buf, then checks whether the
>     >     >          host is waiting for kick signal, not at this time.
>     >     >          The used ring is not empty, the guest continues
>     >     >          the second loop of NAPI.
>     >     >  step 3: The host writes the avail idx read from avail
>     >     >          ring to used ring as event idx via
>     >     >          virtio_queue_set_notification(q->rx_vq, 1).
>     >     >  step 4: At the end of the second loop of NAPI, recheck
>     >     >          whether kick is needed, as the event idx in the
>     >     >          used ring written by the host is beyound the
>     >     >          range of kick condition, the guest will not
>     >     >          send kick signal to the host.
>     >     >
>     >     > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
>     >     > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com>
>     >     > ---
>     >     >
>     >     > Changelog:
>     >     > v7:
>     >     > - Add function virtio_queue_set_notification_and_check()
>     >     > - Restore the function sequence introduce in v6
>     >     >
>     >     > v6:
>     >     > - Take packed packed queue into cosideration
>     >     > - Adjust function sequence to fix compilation issue
>     >     >
>     >     > v5:
>     >     > - Modify return type of virtio_queue_set_notification() to
>     >     >   bool to indicate whether the guest has added some buffers
>     >     >   after last check of avail idx
>     >     > - Loop in virtio_net_has_buffers() if the available buffers
>     >     >   are not sufficient and the guest has added some buffers.
>     >     > - Revert patch "06b12970174"
>     >     > - Update the subject
>     >     >
>     >     > v4:
>     >     > - Correct spelling mistake in the subject
>     >     > - Describe the issue that virtio-net is blocked at host side
>     >     >
>     >     > v3:
>     >     > - Add virtio-net tag in the subject
>     >     > - Refine commit log
>     >     >
>     >     > v2:
>     >     > - Add SOB tag at the end of the commit message
>     >     > - Place Fixes tag at the end of the commit message
>     >     >
>     >     > v1:
>     >     > - Initial patch
>     >     >
>     >     >  hw/net/virtio-net.c        | 19 ++++++++-----------
>     >     >  hw/virtio/virtio.c         | 39
>     ++++++++++++++++++++++++++++++++++++++
>     >     >  include/hw/virtio/virtio.h |  1 +
>     >     >  3 files changed, 48 insertions(+), 11 deletions(-)
>     >     >
>     >     > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
>     >     > index 9c7e85caea..357c651d9b 100644
>     >     > --- a/hw/net/virtio-net.c
>     >     > +++ b/hw/net/virtio-net.c
>     >     > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
>     (NetClientState
>     >     *nc)
>     >     >  static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
>     >     >  {
>     >     >      VirtIONet *n = q->n;
>     >     > -    if (virtio_queue_empty(q->rx_vq) ||
>     >     > -        (n->mergeable_rx_bufs &&
>     >     > -         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     >     > -        virtio_queue_set_notification(q->rx_vq, 1);
>     >     > -
>     >     > -        /* To avoid a race condition where the guest has made some
>     >     buffers
>     >     > -         * available after the above check but before notification
>     was
>     >     > -         * enabled, check for available buffers again.
>     >     > -         */
>     >     > -        if (virtio_queue_empty(q->rx_vq) ||
>     >     > +
>     >     > +    while (virtio_queue_empty(q->rx_vq) ||
>     >     >              (n->mergeable_rx_bufs &&
>     >     > -             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     >     > +            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
>     >     > +        /* guest may have made some buf, try again */
>     >     > +        if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
>     {
>     >     > +            virtio_queue_set_notification(q->rx_vq, 0);
>     >     > +            continue;
>     >     > +        } else {
>     >     >              return 0;
>     >     >          }
>     >     >      }
>     >     > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
>     >     > index 893a072c9d..b165764017 100644
>     >     > --- a/hw/virtio/virtio.c
>     >     > +++ b/hw/virtio/virtio.c
>     >     > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
>     flags,
>     >     bool wrap_counter)
>     >     >      return (avail != used) && (avail == wrap_counter);
>     >     >  }
>     >     > 
>     >     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
>     enable)
>     >     > +{
>     >     > +    uint16_t shadow_idx;
>     >     > +    VRingPackedDesc desc;
>     >     > +    VRingMemoryRegionCaches *caches;
>     >     > +
>     >     > +    vq->notification = enable;
>     >     > +
>     >     > +    if (!vq->vring.desc) {
>     >     > +        return false;
>     >     > +    }
>     >     > +
>     >     > +    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
>     >     > +        virtio_queue_packed_set_notification(vq, enable);
>     >     > +
>     >     > +        if (enable) {
>     >     > +            caches = vring_get_region_caches(vq);
>     >     > +            if (!caches) {
>     >     > +                return false;
>     >     > +            }
>     >     > +
>     >     > +            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
>     >     > +                                   vq->shadow_avail_idx, true);
>     >     > +            if (is_desc_avail(desc.flags, vq->
>     >     shadow_avail_wrap_counter)) {
>     >     > +                return true;
>     >     > +            }
>     >     > +        }
>     >     > +    } else {
>     >     > +        shadow_idx = vq->shadow_avail_idx;
>     >     > +        virtio_queue_split_set_notification(vq, enable);
>     >     > +
>     >     > +        if (enable) {
>     >     > +            return shadow_idx != vring_avail_idx(vq);
>     >     > +        }
>     >     > +    }
>     >     > +
>     >     > +    return false;
>     >     > +}
>     >
>     >     Please document what this does.
>     >     So this will return false if ring has any availabe buffers?
>     >     Equivalent to:
>     >
>     >     bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
>     enable)
>     >     {
>     >             virtio_queue_packed_set_notification(vq, enable);
>     >             return virtio_queue_empty(vq);
>     >     }
>     >
>     >
>     >     or am I missing something?
>     >
>     >     > +
>     >     >  /* Fetch avail_idx from VQ memory only when we really need to know
>     if
>     >     >   * guest has added some buffers.
>     >     >   * Called within rcu_read_lock().  */
>     >     > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
>     virtio.h
>     >     > index 7d5ffdc145..ed85023b87 100644
>     >     > --- a/include/hw/virtio/virtio.h
>     >     > +++ b/include/hw/virtio/virtio.h
>     >     > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
>     >     > 
>     >     >  bool virtio_queue_get_notification(VirtQueue *vq);
>     >     >  void virtio_queue_set_notification(VirtQueue *vq, int enable);
>     >     > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
>     enable);
>     >     > 
>     >     >  int virtio_queue_ready(VirtQueue *vq);
>     >     > 
>     >     > --
>     >     > 2.39.0
>     >
>     >
> 
>
Wencheng Yang July 3, 2024, 5:30 a.m. UTC | #6
> This answer does not make sense from API POV.
The core logic of function virtio_queue_set_notification_and_check() 
actually is a copy of function vhost_enable_notify() from
 vhost.c in kernel.

> > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > +{
> > > + uint16_t shadow_idx;
> > > + VRingPackedDesc desc;
> > > + VRingMemoryRegionCaches *caches;
> > > +
> > > + vq->notification = enable;
> > > +
> > > + if (!vq->vring.desc) {
> > > + return false;
> > > + }
> > > +
> > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > + virtio_queue_packed_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + caches = vring_get_region_caches(vq);
> > > + if (!caches) {
> > > + return false;
> > > + }
> > > +
> > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > + vq->shadow_avail_idx, true);
> > > + if (is_desc_avail(desc.flags, vq->
> > shadow_avail_wrap_counter)) {
> > > + return true;
> > > + }
> > > + }
> > > + } else {
> > > + shadow_idx = vq->shadow_avail_idx;
> > > + virtio_queue_split_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + return shadow_idx != vring_avail_idx(vq);
> > > + }
> > > + }
> > > +
> > > + return false;
> > > +}

// vhost.c
bool vhost_enable_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
{
    ...
    r = vhost_update_avail_event(vq, vq->avail_idx);
    ...
    r = vhost_get_avail_idx(vq, &avail_idx);

    return vhost16_to_cpu(vq, avail_idx) != vq->avail_idx;
}

> My suggestion:

> change virtqueue_get_avail_bytes to return the shadow
> in an opaque unsigned value.

virtio_serial_guest_ready() is referenced by:
virtio_serial_guest_ready()  // virtio-serial-bus.c
get_request_size() // virtio-rng.c
virtqueue_avail_bytes() // virtio.c

Change the return type of virtqueue_get_avail_bytes() from void to unsigned,
it seems okay.

> add virtqueue_poll that gets this opaque and tells us whether any new
> buffers became available in the queue since that value
>was returned.
I'm not clear about the purpose of virtqueue_poll(), when will the function
return?
Do you mean that loop in virtqueue_poll() till new buffers become available?

> accordingly, virtio_queue_set_notification_and_check
> will accept this opaque value and check avail buffers
> against it.

Can you show me the pseudo-code?

Thanks Michael

On 2024/7/2, 19:27, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com>> wrote:


On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > what does "changed" mean here? changed compared to what?
> For a split queue, if the shadow_avail_idx synced from avail ring idx
> by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> idx.
> 
> vq->shadow_avail_idx != vring_avail_idx(vq);
> 
> For packed queue, the logic is similar, if vq->shadow_avail_idx
> 
> becomes available, it means the guest has added buf at the slot.
> 
> vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> 
> vq->shadow_avail_idx, true);
> 
> if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> 
> return true;
> 
> 


My suggestion:


change virtqueue_get_avail_bytes to return the shadow
in an opaque unsigned value.


add virtqueue_poll that gets this opaque and tells us whether any new
buffers became available in the queue since that value
was returned.


accordingly, virtio_queue_set_notification_and_check
will accept this opaque value and check avail buffers
against it.






> On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>> wrote:
> 
> On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > Please document what this does.
> > okay, i will.
> >
> > > So this will return false if ring has any available buffers?
> > > Equivalent to:
> > > 
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > {
> > > virtio_queue_packed_set_notification(vq, enable);
> > > return virtio_queue_empty(vq);
> > > }
> >
> > No, only when the shadow_avail_idx is changed shall the function return
> true,
> 
> 
> what does "changed" mean here? changed compared to what?
> 
> > compared with the value seen by the host last time, else return false
> even if
> > there are some buffers available in the queue, as the total size of the
> > available
> > buffers in the queue can't satisfy the request.
> >
> > It maybe better to pass only one arg to the function like this:
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > {
> > virtio_queue_packed_set_notification(vq, true);
> > 
> > return shadow_avail_idx_changed()? true: false;
> > }
> >
> > Thanks Michael a lot!
> >
> > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>>
> wrote:
> >
> > On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > > From: thomas <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > >
> > > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > > added double-check to test whether the available buffer size
> > > can satisfy the request or not, in case the guest has added
> > > some buffers to the avail ring simultaneously after the first
> > > check. It will be lucky if the available buffer size becomes
> > > okay after the double-check, then the host can send the packet
> > > to the guest. If the buffer size still can't satisfy the request,
> > > even if the guest has added some buffers, viritio-net would
> > > stall at the host side forever.
> > >
> > > The patch checks whether the guest has added some buffers
> > > after last check of avail idx when the available buffers are
> > > not sufficient, if so then recheck the available buffers
> > > in the loop.
> > >
> > > The patch also reverts patch "06b12970174".
> > >
> > > The case below can reproduce the stall.
> > >
> > > Guest 0
> > > +--------+
> > > | iperf |
> > > ---------------> | server |
> > > Host | +--------+
> > > +--------+ | ...
> > > | iperf |----
> > > | client |---- Guest n
> > > +--------+ | +--------+
> > > | | iperf |
> > > ---------------> | server |
> > > +--------+
> > >
> > > Boot many guests from qemu with virtio network:
> > > qemu ... -netdev tap,id=net_x \
> > > -device virtio-net-pci-non-transitional,\
> > > iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> > >
> > > Each guest acts as iperf server with commands below:
> > > iperf3 -s -D -i 10 -p 8001
> > > iperf3 -s -D -i 10 -p 8002
> > >
> > > The host as iperf client:
> > > iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> > > iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> > >
> > > After some time, the host loses connection to the guest,
> > > the guest can send packet to the host, but can't receive
> > > packet from host.
> > >
> > > It's more likely to happen if SWIOTLB is enabled in the guest,
> > > allocating and freeing bounce buffer takes some CPU ticks,
> > > copying from/to bounce buffer takes more CPU ticks, compared
> > > with that there is no bounce buffer in the guest.
> > > Once the rate of producing packets from the host approximates
> > > the rate of receiveing packets in the guest, the guest would
> > > loop in NAPI.
> > >
> > > receive packets ---
> > > | |
> > > v |
> > > free buf virtnet_poll
> > > | |
> > > v |
> > > add buf to avail ring ---
> > > |
> > > | need kick the host?
> > > | NAPI continues
> > > v
> > > receive packets ---
> > > | |
> > > v |
> > > free buf virtnet_poll
> > > | |
> > > v |
> > > add buf to avail ring ---
> > > |
> > > v
> > > ... ...
> > >
> > > On the other hand, the host fetches free buf from avail
> > > ring, if the buf in the avail ring is not enough, the
> > > host notifies the guest the event by writing the avail
> > > idx read from avail ring to the event idx of used ring,
> > > then the host goes to sleep, waiting for the kick signal
> > > from the guest.
> > >
> > > Once the guest finds the host is waiting for kick singal
> > > (in virtqueue_kick_prepare_split()), it kicks the host.
> > >
> > > The host may stall forever at the sequences below:
> > >
> > > Host Guest
> > > ------------ -----------
> > > fetch buf, send packet receive packet ---
> > > ... ... |
> > > fetch buf, send packet add buf |
> > > ... add buf virtnet_poll
> > > buf not enough avail idx-> add buf |
> > > read avail idx add buf |
> > > add buf ---
> > > receive packet ---
> > > write event idx ... |
> > > wait for kick add buf virtnet_poll
> > > ... |
> > > ---
> > > no more packet, exit NAPI
> > >
> > > In the first loop of NAPI above, indicated in the range of
> > > virtnet_poll above, the host is sending packets while the
> > > guest is receiving packets and adding buffers.
> > > step 1: The buf is not enough, for example, a big packet
> > > needs 5 buf, but the available buf count is 3.
> > > The host read current avail idx.
> > > step 2: The guest adds some buf, then checks whether the
> > > host is waiting for kick signal, not at this time.
> > > The used ring is not empty, the guest continues
> > > the second loop of NAPI.
> > > step 3: The host writes the avail idx read from avail
> > > ring to used ring as event idx via
> > > virtio_queue_set_notification(q->rx_vq, 1).
> > > step 4: At the end of the second loop of NAPI, recheck
> > > whether kick is needed, as the event idx in the
> > > used ring written by the host is beyound the
> > > range of kick condition, the guest will not
> > > send kick signal to the host.
> > >
> > > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > > ---
> > >
> > > Changelog:
> > > v7:
> > > - Add function virtio_queue_set_notification_and_check()
> > > - Restore the function sequence introduce in v6
> > >
> > > v6:
> > > - Take packed packed queue into cosideration
> > > - Adjust function sequence to fix compilation issue
> > >
> > > v5:
> > > - Modify return type of virtio_queue_set_notification() to
> > > bool to indicate whether the guest has added some buffers
> > > after last check of avail idx
> > > - Loop in virtio_net_has_buffers() if the available buffers
> > > are not sufficient and the guest has added some buffers.
> > > - Revert patch "06b12970174"
> > > - Update the subject
> > >
> > > v4:
> > > - Correct spelling mistake in the subject
> > > - Describe the issue that virtio-net is blocked at host side
> > >
> > > v3:
> > > - Add virtio-net tag in the subject
> > > - Refine commit log
> > >
> > > v2:
> > > - Add SOB tag at the end of the commit message
> > > - Place Fixes tag at the end of the commit message
> > >
> > > v1:
> > > - Initial patch
> > >
> > > hw/net/virtio-net.c | 19 ++++++++-----------
> > > hw/virtio/virtio.c | 39
> ++++++++++++++++++++++++++++++++++++++
> > > include/hw/virtio/virtio.h | 1 +
> > > 3 files changed, 48 insertions(+), 11 deletions(-)
> > >
> > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > index 9c7e85caea..357c651d9b 100644
> > > --- a/hw/net/virtio-net.c
> > > +++ b/hw/net/virtio-net.c
> > > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
> (NetClientState
> > *nc)
> > > static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> > > {
> > > VirtIONet *n = q->n;
> > > - if (virtio_queue_empty(q->rx_vq) ||
> > > - (n->mergeable_rx_bufs &&
> > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > - virtio_queue_set_notification(q->rx_vq, 1);
> > > -
> > > - /* To avoid a race condition where the guest has made some
> > buffers
> > > - * available after the above check but before notification
> was
> > > - * enabled, check for available buffers again.
> > > - */
> > > - if (virtio_queue_empty(q->rx_vq) ||
> > > +
> > > + while (virtio_queue_empty(q->rx_vq) ||
> > > (n->mergeable_rx_bufs &&
> > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > + !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > + /* guest may have made some buf, try again */
> > > + if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> {
> > > + virtio_queue_set_notification(q->rx_vq, 0);
> > > + continue;
> > > + } else {
> > > return 0;
> > > }
> > > }
> > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > index 893a072c9d..b165764017 100644
> > > --- a/hw/virtio/virtio.c
> > > +++ b/hw/virtio/virtio.c
> > > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> flags,
> > bool wrap_counter)
> > > return (avail != used) && (avail == wrap_counter);
> > > }
> > > 
> > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> > > +{
> > > + uint16_t shadow_idx;
> > > + VRingPackedDesc desc;
> > > + VRingMemoryRegionCaches *caches;
> > > +
> > > + vq->notification = enable;
> > > +
> > > + if (!vq->vring.desc) {
> > > + return false;
> > > + }
> > > +
> > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > + virtio_queue_packed_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + caches = vring_get_region_caches(vq);
> > > + if (!caches) {
> > > + return false;
> > > + }
> > > +
> > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > + vq->shadow_avail_idx, true);
> > > + if (is_desc_avail(desc.flags, vq->
> > shadow_avail_wrap_counter)) {
> > > + return true;
> > > + }
> > > + }
> > > + } else {
> > > + shadow_idx = vq->shadow_avail_idx;
> > > + virtio_queue_split_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + return shadow_idx != vring_avail_idx(vq);
> > > + }
> > > + }
> > > +
> > > + return false;
> > > +}
> >
> > Please document what this does.
> > So this will return false if ring has any availabe buffers?
> > Equivalent to:
> >
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> > {
> > virtio_queue_packed_set_notification(vq, enable);
> > return virtio_queue_empty(vq);
> > }
> >
> >
> > or am I missing something?
> >
> > > +
> > > /* Fetch avail_idx from VQ memory only when we really need to know
> if
> > > * guest has added some buffers.
> > > * Called within rcu_read_lock(). */
> > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
> virtio.h
> > > index 7d5ffdc145..ed85023b87 100644
> > > --- a/include/hw/virtio/virtio.h
> > > +++ b/include/hw/virtio/virtio.h
> > > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> > > 
> > > bool virtio_queue_get_notification(VirtQueue *vq);
> > > void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable);
> > > 
> > > int virtio_queue_ready(VirtQueue *vq);
> > > 
> > > --
> > > 2.39.0
> >
> >
> 
>
Wencheng Yang July 4, 2024, 10:20 a.m. UTC | #7
Hi, Michael

> My suggestion:
> 
> 
> change virtqueue_get_avail_bytes to return the shadow
> in an opaque unsigned value.
> 
> 
> add virtqueue_poll that gets this opaque and tells us whether any new
> buffers became available in the queue since that value> 
> was returned.


> accordingly, virtio_queue_set_notification_and_check
> will accept this opaque value and check avail buffers
> against it.

According to your suggestion, it's able to handle the case where the
queue is not empty, when the queue is empty, should I add an API to
get the shadow idx as virtio_queue_set_notification_and_check()
needs the opaque arg.

What value should return from virtqueue_get_avail_bytes() in case of
error branch in the function? 

On 2024/7/2, 19:27, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com>> wrote:


On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > what does "changed" mean here? changed compared to what?
> For a split queue, if the shadow_avail_idx synced from avail ring idx
> by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> idx.
> 
> vq->shadow_avail_idx != vring_avail_idx(vq);
> 
> For packed queue, the logic is similar, if vq->shadow_avail_idx
> 
> becomes available, it means the guest has added buf at the slot.
> 
> vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> 
> vq->shadow_avail_idx, true);
> 
> if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> 
> return true;
> 
> 


This answer does not make sense from API POV.


My suggestion:


change virtqueue_get_avail_bytes to return the shadow
in an opaque unsigned value.


add virtqueue_poll that gets this opaque and tells us whether any new
buffers became available in the queue since that value
was returned.


accordingly, virtio_queue_set_notification_and_check
will accept this opaque value and check avail buffers
against it.






> On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>> wrote:
> 
> On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > Please document what this does.
> > okay, i will.
> >
> > > So this will return false if ring has any available buffers?
> > > Equivalent to:
> > > 
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > {
> > > virtio_queue_packed_set_notification(vq, enable);
> > > return virtio_queue_empty(vq);
> > > }
> >
> > No, only when the shadow_avail_idx is changed shall the function return
> true,
> 
> 
> what does "changed" mean here? changed compared to what?
> 
> > compared with the value seen by the host last time, else return false
> even if
> > there are some buffers available in the queue, as the total size of the
> > available
> > buffers in the queue can't satisfy the request.
> >
> > It maybe better to pass only one arg to the function like this:
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > {
> > virtio_queue_packed_set_notification(vq, true);
> > 
> > return shadow_avail_idx_changed()? true: false;
> > }
> >
> > Thanks Michael a lot!
> >
> > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>>
> wrote:
> >
> > On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > > From: thomas <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > >
> > > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > > added double-check to test whether the available buffer size
> > > can satisfy the request or not, in case the guest has added
> > > some buffers to the avail ring simultaneously after the first
> > > check. It will be lucky if the available buffer size becomes
> > > okay after the double-check, then the host can send the packet
> > > to the guest. If the buffer size still can't satisfy the request,
> > > even if the guest has added some buffers, viritio-net would
> > > stall at the host side forever.
> > >
> > > The patch checks whether the guest has added some buffers
> > > after last check of avail idx when the available buffers are
> > > not sufficient, if so then recheck the available buffers
> > > in the loop.
> > >
> > > The patch also reverts patch "06b12970174".
> > >
> > > The case below can reproduce the stall.
> > >
> > > Guest 0
> > > +--------+
> > > | iperf |
> > > ---------------> | server |
> > > Host | +--------+
> > > +--------+ | ...
> > > | iperf |----
> > > | client |---- Guest n
> > > +--------+ | +--------+
> > > | | iperf |
> > > ---------------> | server |
> > > +--------+
> > >
> > > Boot many guests from qemu with virtio network:
> > > qemu ... -netdev tap,id=net_x \
> > > -device virtio-net-pci-non-transitional,\
> > > iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> > >
> > > Each guest acts as iperf server with commands below:
> > > iperf3 -s -D -i 10 -p 8001
> > > iperf3 -s -D -i 10 -p 8002
> > >
> > > The host as iperf client:
> > > iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> > > iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> > >
> > > After some time, the host loses connection to the guest,
> > > the guest can send packet to the host, but can't receive
> > > packet from host.
> > >
> > > It's more likely to happen if SWIOTLB is enabled in the guest,
> > > allocating and freeing bounce buffer takes some CPU ticks,
> > > copying from/to bounce buffer takes more CPU ticks, compared
> > > with that there is no bounce buffer in the guest.
> > > Once the rate of producing packets from the host approximates
> > > the rate of receiveing packets in the guest, the guest would
> > > loop in NAPI.
> > >
> > > receive packets ---
> > > | |
> > > v |
> > > free buf virtnet_poll
> > > | |
> > > v |
> > > add buf to avail ring ---
> > > |
> > > | need kick the host?
> > > | NAPI continues
> > > v
> > > receive packets ---
> > > | |
> > > v |
> > > free buf virtnet_poll
> > > | |
> > > v |
> > > add buf to avail ring ---
> > > |
> > > v
> > > ... ...
> > >
> > > On the other hand, the host fetches free buf from avail
> > > ring, if the buf in the avail ring is not enough, the
> > > host notifies the guest the event by writing the avail
> > > idx read from avail ring to the event idx of used ring,
> > > then the host goes to sleep, waiting for the kick signal
> > > from the guest.
> > >
> > > Once the guest finds the host is waiting for kick singal
> > > (in virtqueue_kick_prepare_split()), it kicks the host.
> > >
> > > The host may stall forever at the sequences below:
> > >
> > > Host Guest
> > > ------------ -----------
> > > fetch buf, send packet receive packet ---
> > > ... ... |
> > > fetch buf, send packet add buf |
> > > ... add buf virtnet_poll
> > > buf not enough avail idx-> add buf |
> > > read avail idx add buf |
> > > add buf ---
> > > receive packet ---
> > > write event idx ... |
> > > wait for kick add buf virtnet_poll
> > > ... |
> > > ---
> > > no more packet, exit NAPI
> > >
> > > In the first loop of NAPI above, indicated in the range of
> > > virtnet_poll above, the host is sending packets while the
> > > guest is receiving packets and adding buffers.
> > > step 1: The buf is not enough, for example, a big packet
> > > needs 5 buf, but the available buf count is 3.
> > > The host read current avail idx.
> > > step 2: The guest adds some buf, then checks whether the
> > > host is waiting for kick signal, not at this time.
> > > The used ring is not empty, the guest continues
> > > the second loop of NAPI.
> > > step 3: The host writes the avail idx read from avail
> > > ring to used ring as event idx via
> > > virtio_queue_set_notification(q->rx_vq, 1).
> > > step 4: At the end of the second loop of NAPI, recheck
> > > whether kick is needed, as the event idx in the
> > > used ring written by the host is beyound the
> > > range of kick condition, the guest will not
> > > send kick signal to the host.
> > >
> > > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > > ---
> > >
> > > Changelog:
> > > v7:
> > > - Add function virtio_queue_set_notification_and_check()
> > > - Restore the function sequence introduce in v6
> > >
> > > v6:
> > > - Take packed packed queue into cosideration
> > > - Adjust function sequence to fix compilation issue
> > >
> > > v5:
> > > - Modify return type of virtio_queue_set_notification() to
> > > bool to indicate whether the guest has added some buffers
> > > after last check of avail idx
> > > - Loop in virtio_net_has_buffers() if the available buffers
> > > are not sufficient and the guest has added some buffers.
> > > - Revert patch "06b12970174"
> > > - Update the subject
> > >
> > > v4:
> > > - Correct spelling mistake in the subject
> > > - Describe the issue that virtio-net is blocked at host side
> > >
> > > v3:
> > > - Add virtio-net tag in the subject
> > > - Refine commit log
> > >
> > > v2:
> > > - Add SOB tag at the end of the commit message
> > > - Place Fixes tag at the end of the commit message
> > >
> > > v1:
> > > - Initial patch
> > >
> > > hw/net/virtio-net.c | 19 ++++++++-----------
> > > hw/virtio/virtio.c | 39
> ++++++++++++++++++++++++++++++++++++++
> > > include/hw/virtio/virtio.h | 1 +
> > > 3 files changed, 48 insertions(+), 11 deletions(-)
> > >
> > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > index 9c7e85caea..357c651d9b 100644
> > > --- a/hw/net/virtio-net.c
> > > +++ b/hw/net/virtio-net.c
> > > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
> (NetClientState
> > *nc)
> > > static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> > > {
> > > VirtIONet *n = q->n;
> > > - if (virtio_queue_empty(q->rx_vq) ||
> > > - (n->mergeable_rx_bufs &&
> > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > - virtio_queue_set_notification(q->rx_vq, 1);
> > > -
> > > - /* To avoid a race condition where the guest has made some
> > buffers
> > > - * available after the above check but before notification
> was
> > > - * enabled, check for available buffers again.
> > > - */
> > > - if (virtio_queue_empty(q->rx_vq) ||
> > > +
> > > + while (virtio_queue_empty(q->rx_vq) ||
> > > (n->mergeable_rx_bufs &&
> > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > + !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > + /* guest may have made some buf, try again */
> > > + if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> {
> > > + virtio_queue_set_notification(q->rx_vq, 0);
> > > + continue;
> > > + } else {
> > > return 0;
> > > }
> > > }
> > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > index 893a072c9d..b165764017 100644
> > > --- a/hw/virtio/virtio.c
> > > +++ b/hw/virtio/virtio.c
> > > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> flags,
> > bool wrap_counter)
> > > return (avail != used) && (avail == wrap_counter);
> > > }
> > > 
> > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> > > +{
> > > + uint16_t shadow_idx;
> > > + VRingPackedDesc desc;
> > > + VRingMemoryRegionCaches *caches;
> > > +
> > > + vq->notification = enable;
> > > +
> > > + if (!vq->vring.desc) {
> > > + return false;
> > > + }
> > > +
> > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > + virtio_queue_packed_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + caches = vring_get_region_caches(vq);
> > > + if (!caches) {
> > > + return false;
> > > + }
> > > +
> > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > + vq->shadow_avail_idx, true);
> > > + if (is_desc_avail(desc.flags, vq->
> > shadow_avail_wrap_counter)) {
> > > + return true;
> > > + }
> > > + }
> > > + } else {
> > > + shadow_idx = vq->shadow_avail_idx;
> > > + virtio_queue_split_set_notification(vq, enable);
> > > +
> > > + if (enable) {
> > > + return shadow_idx != vring_avail_idx(vq);
> > > + }
> > > + }
> > > +
> > > + return false;
> > > +}
> >
> > Please document what this does.
> > So this will return false if ring has any availabe buffers?
> > Equivalent to:
> >
> > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable)
> > {
> > virtio_queue_packed_set_notification(vq, enable);
> > return virtio_queue_empty(vq);
> > }
> >
> >
> > or am I missing something?
> >
> > > +
> > > /* Fetch avail_idx from VQ memory only when we really need to know
> if
> > > * guest has added some buffers.
> > > * Called within rcu_read_lock(). */
> > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
> virtio.h
> > > index 7d5ffdc145..ed85023b87 100644
> > > --- a/include/hw/virtio/virtio.h
> > > +++ b/include/hw/virtio/virtio.h
> > > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> > > 
> > > bool virtio_queue_get_notification(VirtQueue *vq);
> > > void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> enable);
> > > 
> > > int virtio_queue_ready(VirtQueue *vq);
> > > 
> > > --
> > > 2.39.0
> >
> >
> 
>
Michael S. Tsirkin July 4, 2024, 11:23 a.m. UTC | #8
On Thu, Jul 04, 2024 at 10:20:15AM +0000, Yang Dongshan wrote:
> Hi, Michael
> 
> > My suggestion:
> > 
> > 
> > change virtqueue_get_avail_bytes to return the shadow
> > in an opaque unsigned value.
> > 
> > 
> > add virtqueue_poll that gets this opaque and tells us whether any new
> > buffers became available in the queue since that value> 
> > was returned.
> 
> 
> > accordingly, virtio_queue_set_notification_and_check
> > will accept this opaque value and check avail buffers
> > against it.
> 
> According to your suggestion, it's able to handle the case where the
> queue is not empty, when the queue is empty, should I add an API to
> get the shadow idx as virtio_queue_set_notification_and_check()
> needs the opaque arg.

virtqueue_get_avail_bytes would always return the opaque.

> What value should return from virtqueue_get_avail_bytes() in case of
> error branch in the function? 

One way would be to make opaque int, return a negative value on error,
positive on success.


> On 2024/7/2, 19:27, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com>> wrote:
> 
> 
> On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > > what does "changed" mean here? changed compared to what?
> > For a split queue, if the shadow_avail_idx synced from avail ring idx
> > by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> > idx.
> > 
> > vq->shadow_avail_idx != vring_avail_idx(vq);
> > 
> > For packed queue, the logic is similar, if vq->shadow_avail_idx
> > 
> > becomes available, it means the guest has added buf at the slot.
> > 
> > vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > 
> > vq->shadow_avail_idx, true);
> > 
> > if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> > 
> > return true;
> > 
> > 
> 
> 
> This answer does not make sense from API POV.
> 
> 
> My suggestion:
> 
> 
> change virtqueue_get_avail_bytes to return the shadow
> in an opaque unsigned value.
> 
> 
> add virtqueue_poll that gets this opaque and tells us whether any new
> buffers became available in the queue since that value
> was returned.
> 
> 
> accordingly, virtio_queue_set_notification_and_check
> will accept this opaque value and check avail buffers
> against it.
> 
> 
> 
> 
> 
> 
> > On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>> wrote:
> > 
> > On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > > Please document what this does.
> > > okay, i will.
> > >
> > > > So this will return false if ring has any available buffers?
> > > > Equivalent to:
> > > > 
> > > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > > {
> > > > virtio_queue_packed_set_notification(vq, enable);
> > > > return virtio_queue_empty(vq);
> > > > }
> > >
> > > No, only when the shadow_avail_idx is changed shall the function return
> > true,
> > 
> > 
> > what does "changed" mean here? changed compared to what?
> > 
> > > compared with the value seen by the host last time, else return false
> > even if
> > > there are some buffers available in the queue, as the total size of the
> > > available
> > > buffers in the queue can't satisfy the request.
> > >
> > > It maybe better to pass only one arg to the function like this:
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > > {
> > > virtio_queue_packed_set_notification(vq, true);
> > > 
> > > return shadow_avail_idx_changed()? true: false;
> > > }
> > >
> > > Thanks Michael a lot!
> > >
> > > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com>>
> > wrote:
> > >
> > > On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > > > From: thomas <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > > >
> > > > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > > > added double-check to test whether the available buffer size
> > > > can satisfy the request or not, in case the guest has added
> > > > some buffers to the avail ring simultaneously after the first
> > > > check. It will be lucky if the available buffer size becomes
> > > > okay after the double-check, then the host can send the packet
> > > > to the guest. If the buffer size still can't satisfy the request,
> > > > even if the guest has added some buffers, viritio-net would
> > > > stall at the host side forever.
> > > >
> > > > The patch checks whether the guest has added some buffers
> > > > after last check of avail idx when the available buffers are
> > > > not sufficient, if so then recheck the available buffers
> > > > in the loop.
> > > >
> > > > The patch also reverts patch "06b12970174".
> > > >
> > > > The case below can reproduce the stall.
> > > >
> > > > Guest 0
> > > > +--------+
> > > > | iperf |
> > > > ---------------> | server |
> > > > Host | +--------+
> > > > +--------+ | ...
> > > > | iperf |----
> > > > | client |---- Guest n
> > > > +--------+ | +--------+
> > > > | | iperf |
> > > > ---------------> | server |
> > > > +--------+
> > > >
> > > > Boot many guests from qemu with virtio network:
> > > > qemu ... -netdev tap,id=net_x \
> > > > -device virtio-net-pci-non-transitional,\
> > > > iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> > > >
> > > > Each guest acts as iperf server with commands below:
> > > > iperf3 -s -D -i 10 -p 8001
> > > > iperf3 -s -D -i 10 -p 8002
> > > >
> > > > The host as iperf client:
> > > > iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> > > > iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> > > >
> > > > After some time, the host loses connection to the guest,
> > > > the guest can send packet to the host, but can't receive
> > > > packet from host.
> > > >
> > > > It's more likely to happen if SWIOTLB is enabled in the guest,
> > > > allocating and freeing bounce buffer takes some CPU ticks,
> > > > copying from/to bounce buffer takes more CPU ticks, compared
> > > > with that there is no bounce buffer in the guest.
> > > > Once the rate of producing packets from the host approximates
> > > > the rate of receiveing packets in the guest, the guest would
> > > > loop in NAPI.
> > > >
> > > > receive packets ---
> > > > | |
> > > > v |
> > > > free buf virtnet_poll
> > > > | |
> > > > v |
> > > > add buf to avail ring ---
> > > > |
> > > > | need kick the host?
> > > > | NAPI continues
> > > > v
> > > > receive packets ---
> > > > | |
> > > > v |
> > > > free buf virtnet_poll
> > > > | |
> > > > v |
> > > > add buf to avail ring ---
> > > > |
> > > > v
> > > > ... ...
> > > >
> > > > On the other hand, the host fetches free buf from avail
> > > > ring, if the buf in the avail ring is not enough, the
> > > > host notifies the guest the event by writing the avail
> > > > idx read from avail ring to the event idx of used ring,
> > > > then the host goes to sleep, waiting for the kick signal
> > > > from the guest.
> > > >
> > > > Once the guest finds the host is waiting for kick singal
> > > > (in virtqueue_kick_prepare_split()), it kicks the host.
> > > >
> > > > The host may stall forever at the sequences below:
> > > >
> > > > Host Guest
> > > > ------------ -----------
> > > > fetch buf, send packet receive packet ---
> > > > ... ... |
> > > > fetch buf, send packet add buf |
> > > > ... add buf virtnet_poll
> > > > buf not enough avail idx-> add buf |
> > > > read avail idx add buf |
> > > > add buf ---
> > > > receive packet ---
> > > > write event idx ... |
> > > > wait for kick add buf virtnet_poll
> > > > ... |
> > > > ---
> > > > no more packet, exit NAPI
> > > >
> > > > In the first loop of NAPI above, indicated in the range of
> > > > virtnet_poll above, the host is sending packets while the
> > > > guest is receiving packets and adding buffers.
> > > > step 1: The buf is not enough, for example, a big packet
> > > > needs 5 buf, but the available buf count is 3.
> > > > The host read current avail idx.
> > > > step 2: The guest adds some buf, then checks whether the
> > > > host is waiting for kick signal, not at this time.
> > > > The used ring is not empty, the guest continues
> > > > the second loop of NAPI.
> > > > step 3: The host writes the avail idx read from avail
> > > > ring to used ring as event idx via
> > > > virtio_queue_set_notification(q->rx_vq, 1).
> > > > step 4: At the end of the second loop of NAPI, recheck
> > > > whether kick is needed, as the event idx in the
> > > > used ring written by the host is beyound the
> > > > range of kick condition, the guest will not
> > > > send kick signal to the host.
> > > >
> > > > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > > > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>
> > > > ---
> > > >
> > > > Changelog:
> > > > v7:
> > > > - Add function virtio_queue_set_notification_and_check()
> > > > - Restore the function sequence introduce in v6
> > > >
> > > > v6:
> > > > - Take packed packed queue into cosideration
> > > > - Adjust function sequence to fix compilation issue
> > > >
> > > > v5:
> > > > - Modify return type of virtio_queue_set_notification() to
> > > > bool to indicate whether the guest has added some buffers
> > > > after last check of avail idx
> > > > - Loop in virtio_net_has_buffers() if the available buffers
> > > > are not sufficient and the guest has added some buffers.
> > > > - Revert patch "06b12970174"
> > > > - Update the subject
> > > >
> > > > v4:
> > > > - Correct spelling mistake in the subject
> > > > - Describe the issue that virtio-net is blocked at host side
> > > >
> > > > v3:
> > > > - Add virtio-net tag in the subject
> > > > - Refine commit log
> > > >
> > > > v2:
> > > > - Add SOB tag at the end of the commit message
> > > > - Place Fixes tag at the end of the commit message
> > > >
> > > > v1:
> > > > - Initial patch
> > > >
> > > > hw/net/virtio-net.c | 19 ++++++++-----------
> > > > hw/virtio/virtio.c | 39
> > ++++++++++++++++++++++++++++++++++++++
> > > > include/hw/virtio/virtio.h | 1 +
> > > > 3 files changed, 48 insertions(+), 11 deletions(-)
> > > >
> > > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > > index 9c7e85caea..357c651d9b 100644
> > > > --- a/hw/net/virtio-net.c
> > > > +++ b/hw/net/virtio-net.c
> > > > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
> > (NetClientState
> > > *nc)
> > > > static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> > > > {
> > > > VirtIONet *n = q->n;
> > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > - (n->mergeable_rx_bufs &&
> > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > - virtio_queue_set_notification(q->rx_vq, 1);
> > > > -
> > > > - /* To avoid a race condition where the guest has made some
> > > buffers
> > > > - * available after the above check but before notification
> > was
> > > > - * enabled, check for available buffers again.
> > > > - */
> > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > +
> > > > + while (virtio_queue_empty(q->rx_vq) ||
> > > > (n->mergeable_rx_bufs &&
> > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > + !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > + /* guest may have made some buf, try again */
> > > > + if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> > {
> > > > + virtio_queue_set_notification(q->rx_vq, 0);
> > > > + continue;
> > > > + } else {
> > > > return 0;
> > > > }
> > > > }
> > > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > > index 893a072c9d..b165764017 100644
> > > > --- a/hw/virtio/virtio.c
> > > > +++ b/hw/virtio/virtio.c
> > > > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> > flags,
> > > bool wrap_counter)
> > > > return (avail != used) && (avail == wrap_counter);
> > > > }
> > > > 
> > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable)
> > > > +{
> > > > + uint16_t shadow_idx;
> > > > + VRingPackedDesc desc;
> > > > + VRingMemoryRegionCaches *caches;
> > > > +
> > > > + vq->notification = enable;
> > > > +
> > > > + if (!vq->vring.desc) {
> > > > + return false;
> > > > + }
> > > > +
> > > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > > + virtio_queue_packed_set_notification(vq, enable);
> > > > +
> > > > + if (enable) {
> > > > + caches = vring_get_region_caches(vq);
> > > > + if (!caches) {
> > > > + return false;
> > > > + }
> > > > +
> > > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > > + vq->shadow_avail_idx, true);
> > > > + if (is_desc_avail(desc.flags, vq->
> > > shadow_avail_wrap_counter)) {
> > > > + return true;
> > > > + }
> > > > + }
> > > > + } else {
> > > > + shadow_idx = vq->shadow_avail_idx;
> > > > + virtio_queue_split_set_notification(vq, enable);
> > > > +
> > > > + if (enable) {
> > > > + return shadow_idx != vring_avail_idx(vq);
> > > > + }
> > > > + }
> > > > +
> > > > + return false;
> > > > +}
> > >
> > > Please document what this does.
> > > So this will return false if ring has any availabe buffers?
> > > Equivalent to:
> > >
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable)
> > > {
> > > virtio_queue_packed_set_notification(vq, enable);
> > > return virtio_queue_empty(vq);
> > > }
> > >
> > >
> > > or am I missing something?
> > >
> > > > +
> > > > /* Fetch avail_idx from VQ memory only when we really need to know
> > if
> > > > * guest has added some buffers.
> > > > * Called within rcu_read_lock(). */
> > > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
> > virtio.h
> > > > index 7d5ffdc145..ed85023b87 100644
> > > > --- a/include/hw/virtio/virtio.h
> > > > +++ b/include/hw/virtio/virtio.h
> > > > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> > > > 
> > > > bool virtio_queue_get_notification(VirtQueue *vq);
> > > > void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable);
> > > > 
> > > > int virtio_queue_ready(VirtQueue *vq);
> > > > 
> > > > --
> > > > 2.39.0
> > >
> > >
> > 
> > 
> 
> 
> 
>
Wencheng Yang July 5, 2024, 2:21 a.m. UTC | #9
> virtqueue_get_avail_bytes would always return the opaque.
Then the condition will change from:

static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
{
      if (virtio_queue_empty(q->rx_vq) ||
          (n->mergeable_rx_bufs &&
           !virtqueue_avail_bytes(q->rx_vq, bufsize, 0)))
    ...
}

to:

static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
{
      if (virtio_queue_empty(q->rx_vq) || n->mergeable_rx_bufs) { 
           shadow_Idx = virtqueue_get_avail_bytes (q->rx_vq, &in_total, NULL, bufsize, 0));
           if (bufsize <= in_totail) {
                // bufsize is okay
            } else {
                 // do notification and recheck available buffers.
                If (virtio_queue_set_notification_and_check(q->rx_vq, shadow_idx)) {
                     ...
                } else {
                    ...
               }
           }
    ...
}

When the queue is empty, always call virtqueue_get_avail_bytes() just to
get the opaque, it seems not friendly to the performance.
Why not get the shadow idx within virtio_queue_set_notification_and_check()
directly and change the function name to
virtio_queue_set_notification_and_check_shadow()? 

On 2024/7/4, 19:24, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com>> wrote:


On Thu, Jul 04, 2024 at 10:20:15AM +0000, Yang Dongshan wrote:
> Hi, Michael
> 
> > My suggestion:
> > 
> > 
> > change virtqueue_get_avail_bytes to return the shadow
> > in an opaque unsigned value.
> > 
> > 
> > add virtqueue_poll that gets this opaque and tells us whether any new
> > buffers became available in the queue since that value> 
> > was returned.
> 
> 
> > accordingly, virtio_queue_set_notification_and_check
> > will accept this opaque value and check avail buffers
> > against it.
> 
> According to your suggestion, it's able to handle the case where the
> queue is not empty, when the queue is empty, should I add an API to
> get the shadow idx as virtio_queue_set_notification_and_check()
> needs the opaque arg.


virtqueue_get_avail_bytes would always return the opaque.


> What value should return from virtqueue_get_avail_bytes() in case of
> error branch in the function? 


One way would be to make opaque int, return a negative value on error,
positive on success.




> On 2024/7/2, 19:27, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>> wrote:
> 
> 
> On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > > what does "changed" mean here? changed compared to what?
> > For a split queue, if the shadow_avail_idx synced from avail ring idx
> > by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> > idx.
> > 
> > vq->shadow_avail_idx != vring_avail_idx(vq);
> > 
> > For packed queue, the logic is similar, if vq->shadow_avail_idx
> > 
> > becomes available, it means the guest has added buf at the slot.
> > 
> > vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > 
> > vq->shadow_avail_idx, true);
> > 
> > if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> > 
> > return true;
> > 
> > 
> 
> 
> This answer does not make sense from API POV.
> 
> 
> My suggestion:
> 
> 
> change virtqueue_get_avail_bytes to return the shadow
> in an opaque unsigned value.
> 
> 
> add virtqueue_poll that gets this opaque and tells us whether any new
> buffers became available in the queue since that value
> was returned.
> 
> 
> accordingly, virtio_queue_set_notification_and_check
> will accept this opaque value and check avail buffers
> against it.
> 
> 
> 
> 
> 
> 
> > On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>> wrote:
> > 
> > On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > > Please document what this does.
> > > okay, i will.
> > >
> > > > So this will return false if ring has any available buffers?
> > > > Equivalent to:
> > > > 
> > > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > > {
> > > > virtio_queue_packed_set_notification(vq, enable);
> > > > return virtio_queue_empty(vq);
> > > > }
> > >
> > > No, only when the shadow_avail_idx is changed shall the function return
> > true,
> > 
> > 
> > what does "changed" mean here? changed compared to what?
> > 
> > > compared with the value seen by the host last time, else return false
> > even if
> > > there are some buffers available in the queue, as the total size of the
> > > available
> > > buffers in the queue can't satisfy the request.
> > >
> > > It maybe better to pass only one arg to the function like this:
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > > {
> > > virtio_queue_packed_set_notification(vq, true);
> > > 
> > > return shadow_avail_idx_changed()? true: false;
> > > }
> > >
> > > Thanks Michael a lot!
> > >
> > > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>>
> > wrote:
> > >
> > > On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > > > From: thomas <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com> <mailto:east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>>
> > > >
> > > > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > > > added double-check to test whether the available buffer size
> > > > can satisfy the request or not, in case the guest has added
> > > > some buffers to the avail ring simultaneously after the first
> > > > check. It will be lucky if the available buffer size becomes
> > > > okay after the double-check, then the host can send the packet
> > > > to the guest. If the buffer size still can't satisfy the request,
> > > > even if the guest has added some buffers, viritio-net would
> > > > stall at the host side forever.
> > > >
> > > > The patch checks whether the guest has added some buffers
> > > > after last check of avail idx when the available buffers are
> > > > not sufficient, if so then recheck the available buffers
> > > > in the loop.
> > > >
> > > > The patch also reverts patch "06b12970174".
> > > >
> > > > The case below can reproduce the stall.
> > > >
> > > > Guest 0
> > > > +--------+
> > > > | iperf |
> > > > ---------------> | server |
> > > > Host | +--------+
> > > > +--------+ | ...
> > > > | iperf |----
> > > > | client |---- Guest n
> > > > +--------+ | +--------+
> > > > | | iperf |
> > > > ---------------> | server |
> > > > +--------+
> > > >
> > > > Boot many guests from qemu with virtio network:
> > > > qemu ... -netdev tap,id=net_x \
> > > > -device virtio-net-pci-non-transitional,\
> > > > iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> > > >
> > > > Each guest acts as iperf server with commands below:
> > > > iperf3 -s -D -i 10 -p 8001
> > > > iperf3 -s -D -i 10 -p 8002
> > > >
> > > > The host as iperf client:
> > > > iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> > > > iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> > > >
> > > > After some time, the host loses connection to the guest,
> > > > the guest can send packet to the host, but can't receive
> > > > packet from host.
> > > >
> > > > It's more likely to happen if SWIOTLB is enabled in the guest,
> > > > allocating and freeing bounce buffer takes some CPU ticks,
> > > > copying from/to bounce buffer takes more CPU ticks, compared
> > > > with that there is no bounce buffer in the guest.
> > > > Once the rate of producing packets from the host approximates
> > > > the rate of receiveing packets in the guest, the guest would
> > > > loop in NAPI.
> > > >
> > > > receive packets ---
> > > > | |
> > > > v |
> > > > free buf virtnet_poll
> > > > | |
> > > > v |
> > > > add buf to avail ring ---
> > > > |
> > > > | need kick the host?
> > > > | NAPI continues
> > > > v
> > > > receive packets ---
> > > > | |
> > > > v |
> > > > free buf virtnet_poll
> > > > | |
> > > > v |
> > > > add buf to avail ring ---
> > > > |
> > > > v
> > > > ... ...
> > > >
> > > > On the other hand, the host fetches free buf from avail
> > > > ring, if the buf in the avail ring is not enough, the
> > > > host notifies the guest the event by writing the avail
> > > > idx read from avail ring to the event idx of used ring,
> > > > then the host goes to sleep, waiting for the kick signal
> > > > from the guest.
> > > >
> > > > Once the guest finds the host is waiting for kick singal
> > > > (in virtqueue_kick_prepare_split()), it kicks the host.
> > > >
> > > > The host may stall forever at the sequences below:
> > > >
> > > > Host Guest
> > > > ------------ -----------
> > > > fetch buf, send packet receive packet ---
> > > > ... ... |
> > > > fetch buf, send packet add buf |
> > > > ... add buf virtnet_poll
> > > > buf not enough avail idx-> add buf |
> > > > read avail idx add buf |
> > > > add buf ---
> > > > receive packet ---
> > > > write event idx ... |
> > > > wait for kick add buf virtnet_poll
> > > > ... |
> > > > ---
> > > > no more packet, exit NAPI
> > > >
> > > > In the first loop of NAPI above, indicated in the range of
> > > > virtnet_poll above, the host is sending packets while the
> > > > guest is receiving packets and adding buffers.
> > > > step 1: The buf is not enough, for example, a big packet
> > > > needs 5 buf, but the available buf count is 3.
> > > > The host read current avail idx.
> > > > step 2: The guest adds some buf, then checks whether the
> > > > host is waiting for kick signal, not at this time.
> > > > The used ring is not empty, the guest continues
> > > > the second loop of NAPI.
> > > > step 3: The host writes the avail idx read from avail
> > > > ring to used ring as event idx via
> > > > virtio_queue_set_notification(q->rx_vq, 1).
> > > > step 4: At the end of the second loop of NAPI, recheck
> > > > whether kick is needed, as the event idx in the
> > > > used ring written by the host is beyound the
> > > > range of kick condition, the guest will not
> > > > send kick signal to the host.
> > > >
> > > > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > > > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com> <mailto:east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>>
> > > > ---
> > > >
> > > > Changelog:
> > > > v7:
> > > > - Add function virtio_queue_set_notification_and_check()
> > > > - Restore the function sequence introduce in v6
> > > >
> > > > v6:
> > > > - Take packed packed queue into cosideration
> > > > - Adjust function sequence to fix compilation issue
> > > >
> > > > v5:
> > > > - Modify return type of virtio_queue_set_notification() to
> > > > bool to indicate whether the guest has added some buffers
> > > > after last check of avail idx
> > > > - Loop in virtio_net_has_buffers() if the available buffers
> > > > are not sufficient and the guest has added some buffers.
> > > > - Revert patch "06b12970174"
> > > > - Update the subject
> > > >
> > > > v4:
> > > > - Correct spelling mistake in the subject
> > > > - Describe the issue that virtio-net is blocked at host side
> > > >
> > > > v3:
> > > > - Add virtio-net tag in the subject
> > > > - Refine commit log
> > > >
> > > > v2:
> > > > - Add SOB tag at the end of the commit message
> > > > - Place Fixes tag at the end of the commit message
> > > >
> > > > v1:
> > > > - Initial patch
> > > >
> > > > hw/net/virtio-net.c | 19 ++++++++-----------
> > > > hw/virtio/virtio.c | 39
> > ++++++++++++++++++++++++++++++++++++++
> > > > include/hw/virtio/virtio.h | 1 +
> > > > 3 files changed, 48 insertions(+), 11 deletions(-)
> > > >
> > > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > > index 9c7e85caea..357c651d9b 100644
> > > > --- a/hw/net/virtio-net.c
> > > > +++ b/hw/net/virtio-net.c
> > > > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
> > (NetClientState
> > > *nc)
> > > > static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> > > > {
> > > > VirtIONet *n = q->n;
> > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > - (n->mergeable_rx_bufs &&
> > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > - virtio_queue_set_notification(q->rx_vq, 1);
> > > > -
> > > > - /* To avoid a race condition where the guest has made some
> > > buffers
> > > > - * available after the above check but before notification
> > was
> > > > - * enabled, check for available buffers again.
> > > > - */
> > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > +
> > > > + while (virtio_queue_empty(q->rx_vq) ||
> > > > (n->mergeable_rx_bufs &&
> > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > + !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > + /* guest may have made some buf, try again */
> > > > + if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> > {
> > > > + virtio_queue_set_notification(q->rx_vq, 0);
> > > > + continue;
> > > > + } else {
> > > > return 0;
> > > > }
> > > > }
> > > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > > index 893a072c9d..b165764017 100644
> > > > --- a/hw/virtio/virtio.c
> > > > +++ b/hw/virtio/virtio.c
> > > > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> > flags,
> > > bool wrap_counter)
> > > > return (avail != used) && (avail == wrap_counter);
> > > > }
> > > > 
> > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable)
> > > > +{
> > > > + uint16_t shadow_idx;
> > > > + VRingPackedDesc desc;
> > > > + VRingMemoryRegionCaches *caches;
> > > > +
> > > > + vq->notification = enable;
> > > > +
> > > > + if (!vq->vring.desc) {
> > > > + return false;
> > > > + }
> > > > +
> > > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > > + virtio_queue_packed_set_notification(vq, enable);
> > > > +
> > > > + if (enable) {
> > > > + caches = vring_get_region_caches(vq);
> > > > + if (!caches) {
> > > > + return false;
> > > > + }
> > > > +
> > > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > > + vq->shadow_avail_idx, true);
> > > > + if (is_desc_avail(desc.flags, vq->
> > > shadow_avail_wrap_counter)) {
> > > > + return true;
> > > > + }
> > > > + }
> > > > + } else {
> > > > + shadow_idx = vq->shadow_avail_idx;
> > > > + virtio_queue_split_set_notification(vq, enable);
> > > > +
> > > > + if (enable) {
> > > > + return shadow_idx != vring_avail_idx(vq);
> > > > + }
> > > > + }
> > > > +
> > > > + return false;
> > > > +}
> > >
> > > Please document what this does.
> > > So this will return false if ring has any availabe buffers?
> > > Equivalent to:
> > >
> > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable)
> > > {
> > > virtio_queue_packed_set_notification(vq, enable);
> > > return virtio_queue_empty(vq);
> > > }
> > >
> > >
> > > or am I missing something?
> > >
> > > > +
> > > > /* Fetch avail_idx from VQ memory only when we really need to know
> > if
> > > > * guest has added some buffers.
> > > > * Called within rcu_read_lock(). */
> > > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
> > virtio.h
> > > > index 7d5ffdc145..ed85023b87 100644
> > > > --- a/include/hw/virtio/virtio.h
> > > > +++ b/include/hw/virtio/virtio.h
> > > > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> > > > 
> > > > bool virtio_queue_get_notification(VirtQueue *vq);
> > > > void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > enable);
> > > > 
> > > > int virtio_queue_ready(VirtQueue *vq);
> > > > 
> > > > --
> > > > 2.39.0
> > >
> > >
> > 
> > 
> 
> 
> 
>
Michael S. Tsirkin July 5, 2024, 5:02 a.m. UTC | #10
On Fri, Jul 05, 2024 at 02:21:35AM +0000, Yang Dongshan wrote:
> > virtqueue_get_avail_bytes would always return the opaque.
> Then the condition will change from:
> 
> static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> {
>       if (virtio_queue_empty(q->rx_vq) ||
>           (n->mergeable_rx_bufs &&
>            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0)))
>     ...
> }
> 
> to:
> 
> static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> {
>       if (virtio_queue_empty(q->rx_vq) || n->mergeable_rx_bufs) { 
>            shadow_Idx = virtqueue_get_avail_bytes (q->rx_vq, &in_total, NULL, bufsize, 0));
>            if (bufsize <= in_totail) {
>                 // bufsize is okay
>             } else {
>                  // do notification and recheck available buffers.
>                 If (virtio_queue_set_notification_and_check(q->rx_vq, shadow_idx)) {
>                      ...
>                 } else {
>                     ...
>                }
>            }
>     ...
> }
> 
> When the queue is empty, always call virtqueue_get_avail_bytes() just to
> get the opaque, it seems not friendly to the performance.
> Why not get the shadow idx within virtio_queue_set_notification_and_check()
> directly and change the function name to
> virtio_queue_set_notification_and_check_shadow()? 

Keep shadow an internal implementation detail otherwise
it's unmaintainable.

We already call virtqueue_get_avail_bytes.
The situation with the stall is super unlikely, it is not
worth worrying about the performance in that case.




> On 2024/7/4, 19:24, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com>> wrote:
> 
> 
> On Thu, Jul 04, 2024 at 10:20:15AM +0000, Yang Dongshan wrote:
> > Hi, Michael
> > 
> > > My suggestion:
> > > 
> > > 
> > > change virtqueue_get_avail_bytes to return the shadow
> > > in an opaque unsigned value.
> > > 
> > > 
> > > add virtqueue_poll that gets this opaque and tells us whether any new
> > > buffers became available in the queue since that value> 
> > > was returned.
> > 
> > 
> > > accordingly, virtio_queue_set_notification_and_check
> > > will accept this opaque value and check avail buffers
> > > against it.
> > 
> > According to your suggestion, it's able to handle the case where the
> > queue is not empty, when the queue is empty, should I add an API to
> > get the shadow idx as virtio_queue_set_notification_and_check()
> > needs the opaque arg.
> 
> 
> virtqueue_get_avail_bytes would always return the opaque.
> 
> 
> > What value should return from virtqueue_get_avail_bytes() in case of
> > error branch in the function? 
> 
> 
> One way would be to make opaque int, return a negative value on error,
> positive on success.
> 
> 
> 
> 
> > On 2024/7/2, 19:27, "Michael S. Tsirkin" <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>> wrote:
> > 
> > 
> > On Tue, Jul 02, 2024 at 07:45:31AM +0800, Yang Dongshan wrote:
> > > > what does "changed" mean here? changed compared to what?
> > > For a split queue, if the shadow_avail_idx synced from avail ring idx
> > > by vring_avail_idx(vq) last time doesn't equal the current value of avail ring
> > > idx.
> > > 
> > > vq->shadow_avail_idx != vring_avail_idx(vq);
> > > 
> > > For packed queue, the logic is similar, if vq->shadow_avail_idx
> > > 
> > > becomes available, it means the guest has added buf at the slot.
> > > 
> > > vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > 
> > > vq->shadow_avail_idx, true);
> > > 
> > > if (is_desc_avail(desc.flags, vq-> shadow_avail_wrap_counter))
> > > 
> > > return true;
> > > 
> > > 
> > 
> > 
> > This answer does not make sense from API POV.
> > 
> > 
> > My suggestion:
> > 
> > 
> > change virtqueue_get_avail_bytes to return the shadow
> > in an opaque unsigned value.
> > 
> > 
> > add virtqueue_poll that gets this opaque and tells us whether any new
> > buffers became available in the queue since that value
> > was returned.
> > 
> > 
> > accordingly, virtio_queue_set_notification_and_check
> > will accept this opaque value and check avail buffers
> > against it.
> > 
> > 
> > 
> > 
> > 
> > 
> > > On Tue, Jul 2, 2024 at 2:46 AM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>> wrote:
> > > 
> > > On Tue, Jul 02, 2024 at 01:18:15AM +0800, Yang Dongshan wrote:
> > > > > Please document what this does.
> > > > okay, i will.
> > > >
> > > > > So this will return false if ring has any available buffers?
> > > > > Equivalent to:
> > > > > 
> > > > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
> > > > > {
> > > > > virtio_queue_packed_set_notification(vq, enable);
> > > > > return virtio_queue_empty(vq);
> > > > > }
> > > >
> > > > No, only when the shadow_avail_idx is changed shall the function return
> > > true,
> > > 
> > > 
> > > what does "changed" mean here? changed compared to what?
> > > 
> > > > compared with the value seen by the host last time, else return false
> > > even if
> > > > there are some buffers available in the queue, as the total size of the
> > > > available
> > > > buffers in the queue can't satisfy the request.
> > > >
> > > > It maybe better to pass only one arg to the function like this:
> > > > bool virtio_queue_set_notification_and_check(VirtQueue *vq)
> > > > {
> > > > virtio_queue_packed_set_notification(vq, true);
> > > > 
> > > > return shadow_avail_idx_changed()? true: false;
> > > > }
> > > >
> > > > Thanks Michael a lot!
> > > >
> > > > On Mon, Jul 1, 2024 at 11:05 PM Michael S. Tsirkin <mst@redhat.com <mailto:mst@redhat.com> <mailto:mst@redhat.com <mailto:mst@redhat.com>>>
> > > wrote:
> > > >
> > > > On Mon, Jul 01, 2024 at 10:00:17PM +0800, Wencheng Yang wrote:
> > > > > From: thomas <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com> <mailto:east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>>
> > > > >
> > > > > Patch 06b12970174 ("virtio-net: fix network stall under load")
> > > > > added double-check to test whether the available buffer size
> > > > > can satisfy the request or not, in case the guest has added
> > > > > some buffers to the avail ring simultaneously after the first
> > > > > check. It will be lucky if the available buffer size becomes
> > > > > okay after the double-check, then the host can send the packet
> > > > > to the guest. If the buffer size still can't satisfy the request,
> > > > > even if the guest has added some buffers, viritio-net would
> > > > > stall at the host side forever.
> > > > >
> > > > > The patch checks whether the guest has added some buffers
> > > > > after last check of avail idx when the available buffers are
> > > > > not sufficient, if so then recheck the available buffers
> > > > > in the loop.
> > > > >
> > > > > The patch also reverts patch "06b12970174".
> > > > >
> > > > > The case below can reproduce the stall.
> > > > >
> > > > > Guest 0
> > > > > +--------+
> > > > > | iperf |
> > > > > ---------------> | server |
> > > > > Host | +--------+
> > > > > +--------+ | ...
> > > > > | iperf |----
> > > > > | client |---- Guest n
> > > > > +--------+ | +--------+
> > > > > | | iperf |
> > > > > ---------------> | server |
> > > > > +--------+
> > > > >
> > > > > Boot many guests from qemu with virtio network:
> > > > > qemu ... -netdev tap,id=net_x \
> > > > > -device virtio-net-pci-non-transitional,\
> > > > > iommu_platform=on,mac=xx:xx:xx:xx:xx:xx,netdev=net_x
> > > > >
> > > > > Each guest acts as iperf server with commands below:
> > > > > iperf3 -s -D -i 10 -p 8001
> > > > > iperf3 -s -D -i 10 -p 8002
> > > > >
> > > > > The host as iperf client:
> > > > > iperf3 -c guest_IP -p 8001 -i 30 -w 256k -P 20 -t 40000
> > > > > iperf3 -c guest_IP -p 8002 -i 30 -w 256k -P 20 -t 40000
> > > > >
> > > > > After some time, the host loses connection to the guest,
> > > > > the guest can send packet to the host, but can't receive
> > > > > packet from host.
> > > > >
> > > > > It's more likely to happen if SWIOTLB is enabled in the guest,
> > > > > allocating and freeing bounce buffer takes some CPU ticks,
> > > > > copying from/to bounce buffer takes more CPU ticks, compared
> > > > > with that there is no bounce buffer in the guest.
> > > > > Once the rate of producing packets from the host approximates
> > > > > the rate of receiveing packets in the guest, the guest would
> > > > > loop in NAPI.
> > > > >
> > > > > receive packets ---
> > > > > | |
> > > > > v |
> > > > > free buf virtnet_poll
> > > > > | |
> > > > > v |
> > > > > add buf to avail ring ---
> > > > > |
> > > > > | need kick the host?
> > > > > | NAPI continues
> > > > > v
> > > > > receive packets ---
> > > > > | |
> > > > > v |
> > > > > free buf virtnet_poll
> > > > > | |
> > > > > v |
> > > > > add buf to avail ring ---
> > > > > |
> > > > > v
> > > > > ... ...
> > > > >
> > > > > On the other hand, the host fetches free buf from avail
> > > > > ring, if the buf in the avail ring is not enough, the
> > > > > host notifies the guest the event by writing the avail
> > > > > idx read from avail ring to the event idx of used ring,
> > > > > then the host goes to sleep, waiting for the kick signal
> > > > > from the guest.
> > > > >
> > > > > Once the guest finds the host is waiting for kick singal
> > > > > (in virtqueue_kick_prepare_split()), it kicks the host.
> > > > >
> > > > > The host may stall forever at the sequences below:
> > > > >
> > > > > Host Guest
> > > > > ------------ -----------
> > > > > fetch buf, send packet receive packet ---
> > > > > ... ... |
> > > > > fetch buf, send packet add buf |
> > > > > ... add buf virtnet_poll
> > > > > buf not enough avail idx-> add buf |
> > > > > read avail idx add buf |
> > > > > add buf ---
> > > > > receive packet ---
> > > > > write event idx ... |
> > > > > wait for kick add buf virtnet_poll
> > > > > ... |
> > > > > ---
> > > > > no more packet, exit NAPI
> > > > >
> > > > > In the first loop of NAPI above, indicated in the range of
> > > > > virtnet_poll above, the host is sending packets while the
> > > > > guest is receiving packets and adding buffers.
> > > > > step 1: The buf is not enough, for example, a big packet
> > > > > needs 5 buf, but the available buf count is 3.
> > > > > The host read current avail idx.
> > > > > step 2: The guest adds some buf, then checks whether the
> > > > > host is waiting for kick signal, not at this time.
> > > > > The used ring is not empty, the guest continues
> > > > > the second loop of NAPI.
> > > > > step 3: The host writes the avail idx read from avail
> > > > > ring to used ring as event idx via
> > > > > virtio_queue_set_notification(q->rx_vq, 1).
> > > > > step 4: At the end of the second loop of NAPI, recheck
> > > > > whether kick is needed, as the event idx in the
> > > > > used ring written by the host is beyound the
> > > > > range of kick condition, the guest will not
> > > > > send kick signal to the host.
> > > > >
> > > > > Fixes: 06b12970174 ("virtio-net: fix network stall under load")
> > > > > Signed-off-by: Wencheng Yang <east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com> <mailto:east.moutain.yang@gmail.com <mailto:east.moutain.yang@gmail.com>>>
> > > > > ---
> > > > >
> > > > > Changelog:
> > > > > v7:
> > > > > - Add function virtio_queue_set_notification_and_check()
> > > > > - Restore the function sequence introduce in v6
> > > > >
> > > > > v6:
> > > > > - Take packed packed queue into cosideration
> > > > > - Adjust function sequence to fix compilation issue
> > > > >
> > > > > v5:
> > > > > - Modify return type of virtio_queue_set_notification() to
> > > > > bool to indicate whether the guest has added some buffers
> > > > > after last check of avail idx
> > > > > - Loop in virtio_net_has_buffers() if the available buffers
> > > > > are not sufficient and the guest has added some buffers.
> > > > > - Revert patch "06b12970174"
> > > > > - Update the subject
> > > > >
> > > > > v4:
> > > > > - Correct spelling mistake in the subject
> > > > > - Describe the issue that virtio-net is blocked at host side
> > > > >
> > > > > v3:
> > > > > - Add virtio-net tag in the subject
> > > > > - Refine commit log
> > > > >
> > > > > v2:
> > > > > - Add SOB tag at the end of the commit message
> > > > > - Place Fixes tag at the end of the commit message
> > > > >
> > > > > v1:
> > > > > - Initial patch
> > > > >
> > > > > hw/net/virtio-net.c | 19 ++++++++-----------
> > > > > hw/virtio/virtio.c | 39
> > > ++++++++++++++++++++++++++++++++++++++
> > > > > include/hw/virtio/virtio.h | 1 +
> > > > > 3 files changed, 48 insertions(+), 11 deletions(-)
> > > > >
> > > > > diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
> > > > > index 9c7e85caea..357c651d9b 100644
> > > > > --- a/hw/net/virtio-net.c
> > > > > +++ b/hw/net/virtio-net.c
> > > > > @@ -1642,18 +1642,15 @@ static bool virtio_net_can_receive
> > > (NetClientState
> > > > *nc)
> > > > > static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
> > > > > {
> > > > > VirtIONet *n = q->n;
> > > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > > - (n->mergeable_rx_bufs &&
> > > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > > - virtio_queue_set_notification(q->rx_vq, 1);
> > > > > -
> > > > > - /* To avoid a race condition where the guest has made some
> > > > buffers
> > > > > - * available after the above check but before notification
> > > was
> > > > > - * enabled, check for available buffers again.
> > > > > - */
> > > > > - if (virtio_queue_empty(q->rx_vq) ||
> > > > > +
> > > > > + while (virtio_queue_empty(q->rx_vq) ||
> > > > > (n->mergeable_rx_bufs &&
> > > > > - !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > > + !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
> > > > > + /* guest may have made some buf, try again */
> > > > > + if (virtio_queue_set_notification_and_check(q->rx_vq, 1))
> > > {
> > > > > + virtio_queue_set_notification(q->rx_vq, 0);
> > > > > + continue;
> > > > > + } else {
> > > > > return 0;
> > > > > }
> > > > > }
> > > > > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > > > > index 893a072c9d..b165764017 100644
> > > > > --- a/hw/virtio/virtio.c
> > > > > +++ b/hw/virtio/virtio.c
> > > > > @@ -668,6 +668,45 @@ static inline bool is_desc_avail(uint16_t
> > > flags,
> > > > bool wrap_counter)
> > > > > return (avail != used) && (avail == wrap_counter);
> > > > > }
> > > > > 
> > > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > > enable)
> > > > > +{
> > > > > + uint16_t shadow_idx;
> > > > > + VRingPackedDesc desc;
> > > > > + VRingMemoryRegionCaches *caches;
> > > > > +
> > > > > + vq->notification = enable;
> > > > > +
> > > > > + if (!vq->vring.desc) {
> > > > > + return false;
> > > > > + }
> > > > > +
> > > > > + if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
> > > > > + virtio_queue_packed_set_notification(vq, enable);
> > > > > +
> > > > > + if (enable) {
> > > > > + caches = vring_get_region_caches(vq);
> > > > > + if (!caches) {
> > > > > + return false;
> > > > > + }
> > > > > +
> > > > > + vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
> > > > > + vq->shadow_avail_idx, true);
> > > > > + if (is_desc_avail(desc.flags, vq->
> > > > shadow_avail_wrap_counter)) {
> > > > > + return true;
> > > > > + }
> > > > > + }
> > > > > + } else {
> > > > > + shadow_idx = vq->shadow_avail_idx;
> > > > > + virtio_queue_split_set_notification(vq, enable);
> > > > > +
> > > > > + if (enable) {
> > > > > + return shadow_idx != vring_avail_idx(vq);
> > > > > + }
> > > > > + }
> > > > > +
> > > > > + return false;
> > > > > +}
> > > >
> > > > Please document what this does.
> > > > So this will return false if ring has any availabe buffers?
> > > > Equivalent to:
> > > >
> > > > bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > > enable)
> > > > {
> > > > virtio_queue_packed_set_notification(vq, enable);
> > > > return virtio_queue_empty(vq);
> > > > }
> > > >
> > > >
> > > > or am I missing something?
> > > >
> > > > > +
> > > > > /* Fetch avail_idx from VQ memory only when we really need to know
> > > if
> > > > > * guest has added some buffers.
> > > > > * Called within rcu_read_lock(). */
> > > > > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/
> > > virtio.h
> > > > > index 7d5ffdc145..ed85023b87 100644
> > > > > --- a/include/hw/virtio/virtio.h
> > > > > +++ b/include/hw/virtio/virtio.h
> > > > > @@ -302,6 +302,7 @@ void virtio_notify_config(VirtIODevice *vdev);
> > > > > 
> > > > > bool virtio_queue_get_notification(VirtQueue *vq);
> > > > > void virtio_queue_set_notification(VirtQueue *vq, int enable);
> > > > > +bool virtio_queue_set_notification_and_check(VirtQueue *vq, int
> > > enable);
> > > > > 
> > > > > int virtio_queue_ready(VirtQueue *vq);
> > > > > 
> > > > > --
> > > > > 2.39.0
> > > >
> > > >
> > > 
> > > 
> > 
> > 
> > 
> > 
> 
> 
> 
>
diff mbox series

Patch

diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
index 9c7e85caea..357c651d9b 100644
--- a/hw/net/virtio-net.c
+++ b/hw/net/virtio-net.c
@@ -1642,18 +1642,15 @@  static bool virtio_net_can_receive(NetClientState *nc)
 static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize)
 {
     VirtIONet *n = q->n;
-    if (virtio_queue_empty(q->rx_vq) ||
-        (n->mergeable_rx_bufs &&
-         !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
-        virtio_queue_set_notification(q->rx_vq, 1);
-
-        /* To avoid a race condition where the guest has made some buffers
-         * available after the above check but before notification was
-         * enabled, check for available buffers again.
-         */
-        if (virtio_queue_empty(q->rx_vq) ||
+
+    while (virtio_queue_empty(q->rx_vq) ||
             (n->mergeable_rx_bufs &&
-             !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
+            !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) {
+        /* guest may have made some buf, try again */
+        if (virtio_queue_set_notification_and_check(q->rx_vq, 1)) {
+            virtio_queue_set_notification(q->rx_vq, 0);
+            continue;
+        } else {
             return 0;
         }
     }
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 893a072c9d..b165764017 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -668,6 +668,45 @@  static inline bool is_desc_avail(uint16_t flags, bool wrap_counter)
     return (avail != used) && (avail == wrap_counter);
 }
 
+bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable)
+{
+    uint16_t shadow_idx;
+    VRingPackedDesc desc;
+    VRingMemoryRegionCaches *caches;
+
+    vq->notification = enable;
+
+    if (!vq->vring.desc) {
+        return false;
+    }
+
+    if (virtio_vdev_has_feature(vq->vdev, VIRTIO_F_RING_PACKED)) {
+        virtio_queue_packed_set_notification(vq, enable);
+
+        if (enable) {
+            caches = vring_get_region_caches(vq);
+            if (!caches) {
+                return false;
+            }
+
+            vring_packed_desc_read(vq->vdev, &desc, &caches->desc,
+                                   vq->shadow_avail_idx, true);
+            if (is_desc_avail(desc.flags, vq->shadow_avail_wrap_counter)) {
+                return true;
+            }
+        }
+    } else {
+        shadow_idx = vq->shadow_avail_idx;
+        virtio_queue_split_set_notification(vq, enable);
+
+        if (enable) {
+            return shadow_idx != vring_avail_idx(vq);
+        }
+    }
+
+    return false;
+}
+
 /* Fetch avail_idx from VQ memory only when we really need to know if
  * guest has added some buffers.
  * Called within rcu_read_lock().  */
diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
index 7d5ffdc145..ed85023b87 100644
--- a/include/hw/virtio/virtio.h
+++ b/include/hw/virtio/virtio.h
@@ -302,6 +302,7 @@  void virtio_notify_config(VirtIODevice *vdev);
 
 bool virtio_queue_get_notification(VirtQueue *vq);
 void virtio_queue_set_notification(VirtQueue *vq, int enable);
+bool virtio_queue_set_notification_and_check(VirtQueue *vq, int enable);
 
 int virtio_queue_ready(VirtQueue *vq);