[net/9p] unconditional wake_up to proc waiting for space on VirtIO ring
authorVenkateswararao Jujjuri (JV) <jvrao@linux.vnet.ibm.com>
Tue, 8 Mar 2011 23:34:20 +0000 (15:34 -0800)
committerEric Van Hensbergen <ericvh@gmail.com>
Tue, 22 Mar 2011 21:32:19 +0000 (16:32 -0500)
Process may wait to get space on VirtIO ring to send a transaction to
VirtFS server. Current code just does a conditional wake_up() which
means only one process will be woken up even if multiple processes
are waiting.

This fix makes the wake_up unconditional. Hence we won't have any
processes waiting for-ever.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
net/9p/trans_virtio.c

index 9b550ed..961e025 100644 (file)
@@ -146,11 +146,10 @@ static void req_done(struct virtqueue *vq)
                rc = virtqueue_get_buf(chan->vq, &len);
 
                if (rc != NULL) {
-                       if (!chan->ring_bufs_avail) {
-                               chan->ring_bufs_avail = 1;
-                               wake_up(chan->vc_wq);
-                       }
+                       chan->ring_bufs_avail = 1;
                        spin_unlock_irqrestore(&chan->lock, flags);
+                       /* Wakeup if anyone waiting for VirtIO ring space. */
+                       wake_up(chan->vc_wq);
                        P9_DPRINTK(P9_DEBUG_TRANS, ": rc %p\n", rc);
                        P9_DPRINTK(P9_DEBUG_TRANS, ": lookup tag %d\n",
                                        rc->tag);