Skip to contentSkip to conversation

vhost: make vhost_zerocopy_callback more efficient by poll_queue base on vhost status

PATCH In review Qin Chuanyu 2 participants 2 messages Feb 26, 2014

Subsystems (2)
KVM virtio host (vhost)
lore
Share & tools
lore

Qin Chuanyu

author
#0 lore raw

PATCH 1/1 vhost: make vhost_zerocopy_callback more efficient by poll_queue base on vhost status

guest kick vhost base on vring flag status and get perfermance improved, vhost_zerocopy_callback could do this in the same way, as virtqueue_enable_cb need one more check after change the status of avail_ring flags, vhost also do the same thing after vhost_enable_notify

test result list as below:
guest and host: suse11sp3, netperf, intel 2.4GHz

+------+----------+--------+----------+--------+--------+---------+
|      |             old              |            new            |
+------+----------+--------+----------+--------+--------+---------+
| UDP  |  Gbit/s  |  PPS   |CPU idle% | Gbit/s |   PPS  |CPU idle%|
| 256  | 0.74805  | 321309 |  87.16   | 0.77933| 334743 |  90.71  |
| 512  |   1.42   | 328475 |  87.03   |  1.44  | 333550 |  90.43  |
| 1024 |   2.79   | 334426 |  89.09   |  2.81  | 336986 |  89.55  |
| 1460 |   3.71   | 316215 |  87.53   |  4.02  | 342325 |  89.58  |
+------+----------+--------+----------+--------+--------+---------+
Signed-off-by Chuanyu Qin <[email protected]>
---
 drivers/vhost/net.c |   12 ++++++++++--
 1 files changed, 10 insertions(+), 2 deletions(-)
drivers/vhost/net.c +10 -2
index a0fa5de..3c76b89 100644
@@ -314,7 +314,8 @@ static void vhost_zerocopy_callback(struct ubuf_info *ubuf, bool success)
 	vq->heads[ubuf->desc].len = success ?
 		VHOST_DMA_DONE_LEN : VHOST_DMA_FAILED_LEN;
 	cnt = vhost_net_ubuf_put(ubufs);
-
+	/* make sure len has been updated because handle_tx would use it */
+	smp_wmb();
 	/*
 	 * Trigger polling thread if guest stopped submitting new buffers:
 	 * in this case, the refcount after decrement will eventually reach 1.

@@ -322,7 +323,8 @@ static void vhost_zerocopy_callback(struct ubuf_info *ubuf, bool success)
 	 * (the value 16 here is more or less arbitrary, it's tuned to trigger
 	 * less than 10% of times).
 	 */
-	if (cnt <= 1 || !(cnt % 16))
+	if ((!(vq->used_flags & VRING_USED_F_NO_NOTIFY))
+			&& (cnt <= 1 || !(cnt % 16)))
 		vhost_poll_queue(&vq->poll);
 
 	rcu_read_unlock_bh();

@@ -386,6 +388,12 @@ static void handle_tx(struct vhost_net *net)
 				vhost_disable_notify(&net->dev, vq);
 				continue;
 			}
+			/* there might skb been freed between last
+			* vhost_zerocopy_signal_used and vhost_enable_notify,
+			* so one more check is needed.
+			*/
+			if (zcopy)
+				vhost_zerocopy_signal_used(net, vq);
 			break;
 		}
 		if (in) {
Signature · 1 line
1.7.3.1.msysgit.0
#1 lore raw

----- Original Message -----

Quoted 27 lines

guest kick vhost base on vring flag status and get perfermance improved, vhost_zerocopy_callback could do this in the same way, as virtqueue_enable_cb need one more check after change the status of avail_ring flags, vhost also do the same thing after vhost_enable_notify

test result list as below:
guest and host: suse11sp3, netperf, intel 2.4GHz

+------+----------+--------+----------+--------+--------+---------+
|      |             old              |            new            |
+------+----------+--------+----------+--------+--------+---------+
| UDP  |  Gbit/s  |  PPS   |CPU idle% | Gbit/s |   PPS  |CPU idle%|
| 256  | 0.74805  | 321309 |  87.16   | 0.77933| 334743 |  90.71  |
| 512  |   1.42   | 328475 |  87.03   |  1.44  | 333550 |  90.43  |
| 1024 |   2.79   | 334426 |  89.09   |  2.81  | 336986 |  89.55  |
| 1460 |   3.71   | 316215 |  87.53   |  4.02  | 342325 |  89.58  |
+------+----------+--------+----------+--------+--------+---------+
Signed-off-by Chuanyu Qin <[email protected]>
---
  drivers/vhost/net.c |   12 ++++++++++--
  1 files changed, 10 insertions(+), 2 deletions(-)
drivers/vhost/net.c +0 -0
index a0fa5de..3c76b89 100644

drivers/vhost/net.c static void vhost_zerocopy_callback(struct ubuf_info · line 319

Quoted 7 lines
drivers/vhost/net.c +2 -1
@@ -314,7 +314,8 @@ static void vhost_zerocopy_callback(struct ubuf_info
 *ubuf, bool success)
  	vq->heads[ubuf->desc].len = success ?
  		VHOST_DMA_DONE_LEN : VHOST_DMA_FAILED_LEN;
  	cnt = vhost_net_ubuf_put(ubufs);
-
+	/* make sure len has been updated because handle_tx would use it */
+	smp_wmb();

Hi Chuanyu:

Apologize first for my typo in the last comment, should be a smp_mb() here to prevent the used_flags to be checked(read) before updating the len (writing).

smp_wmb() can only prevent re-ordering between writes.

btw. If you want you patch to be applied for David tree, you need also cc or to netdev.

Quoted 35 lines
drivers/vhost/net.c +0 -0
  	/*
  	 * Trigger polling thread if guest stopped submitting new buffers:

* in this case, the refcount after decrement will eventually reach 1.

drivers/vhost/net.c +2 -1
@@ -322,7 +323,8 @@ static void vhost_zerocopy_callback(struct ubuf_info
 *ubuf, bool success)
  	 * (the value 16 here is more or less arbitrary, it's tuned to trigger
  	 * less than 10% of times).
  	 */
-	if (cnt <= 1 || !(cnt % 16))
+	if ((!(vq->used_flags & VRING_USED_F_NO_NOTIFY))
+			&& (cnt <= 1 || !(cnt % 16)))
  		vhost_poll_queue(&vq->poll);
 

  	rcu_read_unlock_bh();
drivers/vhost/net.c +6 -1
@@ -386,6 +388,12 @@ static void handle_tx(struct vhost_net *net)
  				vhost_disable_notify(&net->dev, vq);
  				continue;
  			}
+			/* there might skb been freed between last
+			* vhost_zerocopy_signal_used and vhost_enable_notify,
+			* so one more check is needed.
+			*/
+			if (zcopy)
+				vhost_zerocopy_signal_used(net, vq);
  			break;
  		}
  		if (in) {
--

1.7.3.1.msysgit.0

--
To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html