block/nvme: poll queues without q->lock

A lot of CPU time is spent simply locking/unlocking q->lock during
polling. Check for completion outside the lock to make q->lock disappear
from the profile.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200617132201.1832152-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

1 files changed, 12 insertions, 0 deletions

diff --git a/block/nvme.c b/block/nvme.c
index eb2f54dd9d..e4375ec245 100644
--- a/block/nvme.c
+++ b/block/nvme.c
@@ -512,6 +512,18 @@ static bool nvme_poll_queues(BDRVNVMeState *s)
 
     for (i = 0; i < s->nr_queues; i++) {
         NVMeQueuePair *q = s->queues[i];
+        const size_t cqe_offset = q->cq.head * NVME_CQ_ENTRY_BYTES;
+        NvmeCqe *cqe = (NvmeCqe *)&q->cq.queue[cqe_offset];
+
+        /*
+         * Do an early check for completions. q->lock isn't needed because
+         * nvme_process_completion() only runs in the event loop thread and
+         * cannot race with itself.
+         */
+        if ((le16_to_cpu(cqe->status) & 0x1) == q->cq_phase) {
+            continue;
+        }
+
         qemu_mutex_lock(&q->lock);
         while (nvme_process_completion(s, q)) {
             /* Keep polling */