The loop may exist if vq->broken is true,
virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
will return NULL, so virtnet_poll will reschedule napi to
receive packet, it will lead cpu usage(si) to 100%.
call trace as below:
virtnet_poll
virtnet_receive
virtqueue_get_buf_ctx
virtqueue_get_buf_ctx_packed
virtqueue_get_buf_ctx_split
virtqueue_napi_complete
virtqueue_poll //return true
virtqueue_napi_schedule //it will reschedule napi
To fix this, return false if vq is broken in virtqueue_poll.
Signed-off-by: Mao Wenan <[email protected]>
---
v1->v2: fix it in virtqueue_poll suggested by Michael S. Tsirkin <[email protected]>
drivers/virtio/virtio_ring.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 58b96ba..4f7c73e 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -1960,6 +1960,9 @@ bool virtqueue_poll(struct virtqueue *_vq, unsigned last_used_idx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
+ if (unlikely(vq->broken))
+ return false;
+
virtio_mb(vq->weak_barriers);
return vq->packed_ring ? virtqueue_poll_packed(_vq, last_used_idx) :
virtqueue_poll_split(_vq, last_used_idx);
--
1.8.3.1
On Sun, Aug 02, 2020 at 01:56:33PM +0800, Mao Wenan wrote:
> The loop may exist if vq->broken is true,
> virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
> will return NULL, so virtnet_poll will reschedule napi to
> receive packet, it will lead cpu usage(si) to 100%.
>
> call trace as below:
> virtnet_poll
> virtnet_receive
> virtqueue_get_buf_ctx
> virtqueue_get_buf_ctx_packed
> virtqueue_get_buf_ctx_split
> virtqueue_napi_complete
> virtqueue_poll //return true
> virtqueue_napi_schedule //it will reschedule napi
>
> To fix this, return false if vq is broken in virtqueue_poll.
>
> Signed-off-by: Mao Wenan <[email protected]>
Looks good:
Acked-by: Michael S. Tsirkin <[email protected]>
> ---
> v1->v2: fix it in virtqueue_poll suggested by Michael S. Tsirkin <[email protected]>
> drivers/virtio/virtio_ring.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 58b96ba..4f7c73e 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -1960,6 +1960,9 @@ bool virtqueue_poll(struct virtqueue *_vq, unsigned last_used_idx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
>
> + if (unlikely(vq->broken))
> + return false;
> +
> virtio_mb(vq->weak_barriers);
> return vq->packed_ring ? virtqueue_poll_packed(_vq, last_used_idx) :
> virtqueue_poll_split(_vq, last_used_idx);
> --
> 1.8.3.1
Just noticed the subject is wrong: this is no longer
a virtio_net patch.
On Sun, Aug 02, 2020 at 01:56:33PM +0800, Mao Wenan wrote:
> The loop may exist if vq->broken is true,
> virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
> will return NULL, so virtnet_poll will reschedule napi to
> receive packet, it will lead cpu usage(si) to 100%.
>
> call trace as below:
> virtnet_poll
> virtnet_receive
> virtqueue_get_buf_ctx
> virtqueue_get_buf_ctx_packed
> virtqueue_get_buf_ctx_split
> virtqueue_napi_complete
> virtqueue_poll //return true
> virtqueue_napi_schedule //it will reschedule napi
>
> To fix this, return false if vq is broken in virtqueue_poll.
>
> Signed-off-by: Mao Wenan <[email protected]>
> ---
> v1->v2: fix it in virtqueue_poll suggested by Michael S. Tsirkin <[email protected]>
> drivers/virtio/virtio_ring.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 58b96ba..4f7c73e 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -1960,6 +1960,9 @@ bool virtqueue_poll(struct virtqueue *_vq, unsigned last_used_idx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
>
> + if (unlikely(vq->broken))
> + return false;
> +
> virtio_mb(vq->weak_barriers);
> return vq->packed_ring ? virtqueue_poll_packed(_vq, last_used_idx) :
> virtqueue_poll_split(_vq, last_used_idx);
> --
> 1.8.3.1
On 8/2/20 2:26 PM, Michael S. Tsirkin wrote:
>
> Just noticed the subject is wrong: this is no longer
> a virtio_net patch.
thanks, I will change the subject and send v3.
The loop may exist if vq->broken is true,
virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
will return NULL, so virtnet_poll will reschedule napi to
receive packet, it will lead cpu usage(si) to 100%.
call trace as below:
virtnet_poll
virtnet_receive
virtqueue_get_buf_ctx
virtqueue_get_buf_ctx_packed
virtqueue_get_buf_ctx_split
virtqueue_napi_complete
virtqueue_poll //return true
virtqueue_napi_schedule //it will reschedule napi
to fix this, return false if vq is broken in virtqueue_poll.
Signed-off-by: Mao Wenan <[email protected]>
Acked-by: Michael S. Tsirkin <[email protected]>
---
v2->v3: change subject, original is : "virtio_net: Avoid loop in virtnet_poll"
v1->v2: fix it in virtqueue_poll suggested by Michael S. Tsirkin <[email protected]>
drivers/virtio/virtio_ring.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 58b96ba..4f7c73e 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -1960,6 +1960,9 @@ bool virtqueue_poll(struct virtqueue *_vq, unsigned last_used_idx)
{
struct vring_virtqueue *vq = to_vvq(_vq);
+ if (unlikely(vq->broken))
+ return false;
+
virtio_mb(vq->weak_barriers);
return vq->packed_ring ? virtqueue_poll_packed(_vq, last_used_idx) :
virtqueue_poll_split(_vq, last_used_idx);
--
1.8.3.1
On 2020/8/2 下午3:44, Mao Wenan wrote:
> The loop may exist if vq->broken is true,
> virtqueue_get_buf_ctx_packed or virtqueue_get_buf_ctx_split
> will return NULL, so virtnet_poll will reschedule napi to
> receive packet, it will lead cpu usage(si) to 100%.
>
> call trace as below:
> virtnet_poll
> virtnet_receive
> virtqueue_get_buf_ctx
> virtqueue_get_buf_ctx_packed
> virtqueue_get_buf_ctx_split
> virtqueue_napi_complete
> virtqueue_poll //return true
> virtqueue_napi_schedule //it will reschedule napi
>
> to fix this, return false if vq is broken in virtqueue_poll.
>
> Signed-off-by: Mao Wenan <[email protected]>
> Acked-by: Michael S. Tsirkin <[email protected]>
> ---
> v2->v3: change subject, original is : "virtio_net: Avoid loop in virtnet_poll"
> v1->v2: fix it in virtqueue_poll suggested by Michael S. Tsirkin <[email protected]>
> drivers/virtio/virtio_ring.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 58b96ba..4f7c73e 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -1960,6 +1960,9 @@ bool virtqueue_poll(struct virtqueue *_vq, unsigned last_used_idx)
> {
> struct vring_virtqueue *vq = to_vvq(_vq);
>
> + if (unlikely(vq->broken))
> + return false;
> +
> virtio_mb(vq->weak_barriers);
> return vq->packed_ring ? virtqueue_poll_packed(_vq, last_used_idx) :
> virtqueue_poll_split(_vq, last_used_idx);
Acked-by: Jason Wang <[email protected]>