On 12.12.2023 19:11, Michael S. Tsirkin wrote: > On Tue, Dec 12, 2023 at 06:50:39PM +0300, Arseniy Krasnov wrote: >> >> >> On 12.12.2023 18:54, Michael S. Tsirkin wrote: >>> On Tue, Dec 12, 2023 at 12:16:57AM +0300, Arseniy Krasnov wrote: >>>> Add one more condition for sending credit update during dequeue from >>>> stream socket: when number of bytes in the rx queue is smaller than >>>> SO_RCVLOWAT value of the socket. This is actual for non-default value >>>> of SO_RCVLOWAT (e.g. not 1) - idea is to "kick" peer to continue data >>>> transmission, because we need at least SO_RCVLOWAT bytes in our rx >>>> queue to wake up user for reading data (in corner case it is also >>>> possible to stuck both tx and rx sides, this is why 'Fixes' is used). >>> >>> I don't get what does "to stuck both tx and rx sides" mean. >> >> I meant situation when tx waits for the free space, while rx doesn't send >> credit update, just waiting for more data. Sorry for my English :) >> >>> Besides being agrammatical, is there a way to do this without >>> playing with SO_RCVLOWAT? >> >> No, this may happen only with non-default SO_RCVLOWAT values (e.g. != 1) >> >> Thanks, Arseniy > > I am split on whether we need the Fixes tag. I guess if the other side > is vhost with SO_RCVLOWAT then it might be stuck and it might apply > without SO_RCVLOWAT on the local kernel? IIUC your question, then this problem is actual for any transports: g2h, h2g and loopback. Thanks, Arseniy > > >>> >>>> >>>> Fixes: b89d882dc9fc ("vsock/virtio: reduce credit update messages") >>>> Signed-off-by: Arseniy Krasnov <avkrasnov@xxxxxxxxxxxxxxxxx> >>>> --- >>>> Changelog: >>>> v6 -> v7: >>>> * Handle wrap of 'fwd_cnt'. >>>> * Do to send credit update when 'fwd_cnt' == 'last_fwd_cnt'. >>>> v7 -> v8: >>>> * Remove unneeded/wrong handling of wrap for 'fwd_cnt'. >>>> >>>> net/vmw_vsock/virtio_transport_common.c | 13 ++++++++++--- >>>> 1 file changed, 10 insertions(+), 3 deletions(-) >>>> >>>> diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c >>>> index e137d740804e..8572f94bba88 100644 >>>> --- a/net/vmw_vsock/virtio_transport_common.c >>>> +++ b/net/vmw_vsock/virtio_transport_common.c >>>> @@ -558,6 +558,8 @@ virtio_transport_stream_do_dequeue(struct vsock_sock *vsk, >>>> struct virtio_vsock_sock *vvs = vsk->trans; >>>> size_t bytes, total = 0; >>>> struct sk_buff *skb; >>>> + u32 fwd_cnt_delta; >>>> + bool low_rx_bytes; >>>> int err = -EFAULT; >>>> u32 free_space; >>>> >>>> @@ -601,7 +603,10 @@ virtio_transport_stream_do_dequeue(struct vsock_sock *vsk, >>>> } >>>> } >>>> >>>> - free_space = vvs->buf_alloc - (vvs->fwd_cnt - vvs->last_fwd_cnt); >>>> + fwd_cnt_delta = vvs->fwd_cnt - vvs->last_fwd_cnt; >>>> + free_space = vvs->buf_alloc - fwd_cnt_delta; >>>> + low_rx_bytes = (vvs->rx_bytes < >>>> + sock_rcvlowat(sk_vsock(vsk), 0, INT_MAX)); >>>> >>>> spin_unlock_bh(&vvs->rx_lock); >>>> >>>> @@ -611,9 +616,11 @@ virtio_transport_stream_do_dequeue(struct vsock_sock *vsk, >>>> * too high causes extra messages. Too low causes transmitter >>>> * stalls. As stalls are in theory more expensive than extra >>>> * messages, we set the limit to a high value. TODO: experiment >>>> - * with different values. >>>> + * with different values. Also send credit update message when >>>> + * number of bytes in rx queue is not enough to wake up reader. >>>> */ >>>> - if (free_space < VIRTIO_VSOCK_MAX_PKT_BUF_SIZE) >>>> + if (fwd_cnt_delta && >>>> + (free_space < VIRTIO_VSOCK_MAX_PKT_BUF_SIZE || low_rx_bytes)) >>>> virtio_transport_send_credit_update(vsk); >>>> >>>> return total; >>>> -- >>>> 2.25.1 >>> >