On Wed, Jul 17, 2019 at 4:51 PM Michael S. Tsirkin [off-list ref] wrote:

On Wed, Jul 17, 2019 at 01:30:28PM +0200, Stefano Garzarella wrote:

quoted

fwd_cnt and last_fwd_cnt are protected by rx_lock, so we should use
the same spinlock also if we are in the TX path.

Move also buf_alloc under the same lock.

Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>

Wait a second is this a bugfix?
If it's used under the wrong lock won't values get corrupted?
Won't traffic then stall or more data get to sent than
credits?

Before this series, we only read vvs->fwd_cnt and vvs->buf_alloc in this
function, but using a different lock than the one used to write them.
I'm not sure if a corruption can happen, but if we want to avoid the
lock, we should use an atomic operation or memory barriers.

Since now we also need to update vvs->last_fwd_cnt, in order to limit the
credit message, I decided to take the same lock used to protect vvs->fwd_cnt
and vvs->last_fwd_cnt.


Thanks,
Stefano

quoted

---
 include/linux/virtio_vsock.h            | 2 +-
 net/vmw_vsock/virtio_transport_common.c | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/include/linux/virtio_vsock.h b/include/linux/virtio_vsock.h
index 49fc9d20bc43..4c7781f4b29b 100644
--- a/include/linux/virtio_vsock.h
+++ b/include/linux/virtio_vsock.h

@@ -35,7 +35,6 @@ struct virtio_vsock_sock {

      /* Protected by tx_lock */
      u32 tx_cnt;
-     u32 buf_alloc;
      u32 peer_fwd_cnt;
      u32 peer_buf_alloc;

@@ -43,6 +42,7 @@ struct virtio_vsock_sock {
      u32 fwd_cnt;
      u32 last_fwd_cnt;
      u32 rx_bytes;
+     u32 buf_alloc;
      struct list_head rx_queue;
 };

diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
index a85559d4d974..34a2b42313b7 100644
--- a/net/vmw_vsock/virtio_transport_common.c
+++ b/net/vmw_vsock/virtio_transport_common.c

@@ -210,11 +210,11 @@ static void virtio_transport_dec_rx_pkt(struct virtio_vsock_sock *vvs,

 void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs, struct virtio_vsock_pkt *pkt)
 {
-     spin_lock_bh(&vvs->tx_lock);
+     spin_lock_bh(&vvs->rx_lock);
      vvs->last_fwd_cnt = vvs->fwd_cnt;
      pkt->hdr.fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
      pkt->hdr.buf_alloc = cpu_to_le32(vvs->buf_alloc);
-     spin_unlock_bh(&vvs->tx_lock);
+     spin_unlock_bh(&vvs->rx_lock);
 }
 EXPORT_SYMBOL_GPL(virtio_transport_inc_tx_pkt);

--

2.20.1