[dpdk-dev] vhost: vhost injects interrupts to guest for each packet

Message ID 1442823374-5233-1-git-send-email-huawei.xie@intel.com (mailing list archive)
State Accepted, archived
Headers

Commit Message

Huawei Xie Sept. 21, 2015, 8:16 a.m. UTC
  In merge-able RX path, vhost injects interrupts to guest for each packet.
This should degrade performance a lot.
This patch fixes this issue by injecting one interrupt for a batch of packets. 

Signed-off-by: Huawei Xie <huawei.xie@intel.com>
---
 lib/librte_vhost/vhost_rxtx.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
  

Comments

Jianfeng Tan Sept. 21, 2015, 9:40 a.m. UTC | #1
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Huawei Xie
> Sent: Monday, September 21, 2015 4:16 PM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH] vhost: vhost injects interrupts to guest for each
> packet
> 
> In merge-able RX path, vhost injects interrupts to guest for each packet.
> This should degrade performance a lot.
> This patch fixes this issue by injecting one interrupt for a batch of packets.
> 
> Signed-off-by: Huawei Xie <huawei.xie@intel.com>
> ---
>  lib/librte_vhost/vhost_rxtx.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
> index 0d07338..9d3165a 100644
> --- a/lib/librte_vhost/vhost_rxtx.c
> +++ b/lib/librte_vhost/vhost_rxtx.c
> @@ -496,7 +496,7 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t
> queue_id,
>  						"to get enough desc from "
>  						"vring\n",
>  						dev->device_fh);
> -					return pkt_idx;
> +					goto merge_rx_exit;
>  				} else {
>  					update_secure_len(vq, res_cur_idx,
> &secure_len, &vec_idx);
>  					res_cur_idx++;
> @@ -523,7 +523,10 @@ virtio_dev_merge_rx(struct virtio_net *dev,
> uint16_t queue_id,
> 
>  		*(volatile uint16_t *)&vq->used->idx += entry_success;
>  		vq->last_used_idx = res_cur_idx;
> +	}
> 
> +merge_rx_exit:
> +	if (likely(pkt_idx)) {
>  		/* flush used->idx update before we read avail->flags. */
>  		rte_mb();
> 
> @@ -532,7 +535,7 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t
> queue_id,
>  			eventfd_write((int)vq->callfd, 1);
>  	}
> 
> -	return count;
> +	return pkt_idx;
>  }
> 
>  uint16_t
> --

Acked-by: Jianfeng Tan <Jianfeng.tan@intel.com>

> 1.9.3
  
Thomas Monjalon Sept. 24, 2015, 7:58 p.m. UTC | #2
Please reword the title (no need to re-send the patch),
so that you describe what the patch do (batch interrupt).

2015-09-21 16:16, Huawei Xie:
> In merge-able RX path, vhost injects interrupts to guest for each packet.
> This should degrade performance a lot.
> This patch fixes this issue by injecting one interrupt for a batch of packets.
  
Huawei Xie Sept. 29, 2015, 3:31 p.m. UTC | #3
On 9/25/2015 4:00 AM, Thomas Monjalon wrote:
> Please reword the title (no need to re-send the patch),
> so that you describe what the patch do (batch interrupt).
>
> 2015-09-21 16:16, Huawei Xie:
>> In merge-able RX path, vhost injects interrupts to guest for each packet.
>> This should degrade performance a lot.
>> This patch fixes this issue by injecting one interrupt for a batch of packets. 
>
Hi Thomas:
title:
    vhost: inject one interrupt only for a batch of packets
Do you mean this?
  
Thomas Monjalon Sept. 29, 2015, 7:21 p.m. UTC | #4
2015-09-29 15:31, Xie, Huawei:
> On 9/25/2015 4:00 AM, Thomas Monjalon wrote:
> > Please reword the title (no need to re-send the patch),
> > so that you describe what the patch do (batch interrupt).
> >
> > 2015-09-21 16:16, Huawei Xie:
> >> In merge-able RX path, vhost injects interrupts to guest for each packet.
> >> This should degrade performance a lot.
> >> This patch fixes this issue by injecting one interrupt for a batch of packets. 
> >
> Hi Thomas:
> title:
>     vhost: inject one interrupt only for a batch of packets
> Do you mean this?

Yes, or
"vhost: inject only one interrupt for a batch of packets"

Applied, thanks
  

Patch

diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
index 0d07338..9d3165a 100644
--- a/lib/librte_vhost/vhost_rxtx.c
+++ b/lib/librte_vhost/vhost_rxtx.c
@@ -496,7 +496,7 @@  virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 						"to get enough desc from "
 						"vring\n",
 						dev->device_fh);
-					return pkt_idx;
+					goto merge_rx_exit;
 				} else {
 					update_secure_len(vq, res_cur_idx, &secure_len, &vec_idx);
 					res_cur_idx++;
@@ -523,7 +523,10 @@  virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 
 		*(volatile uint16_t *)&vq->used->idx += entry_success;
 		vq->last_used_idx = res_cur_idx;
+	}
 
+merge_rx_exit:
+	if (likely(pkt_idx)) {
 		/* flush used->idx update before we read avail->flags. */
 		rte_mb();
 
@@ -532,7 +535,7 @@  virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 			eventfd_write((int)vq->callfd, 1);
 	}
 
-	return count;
+	return pkt_idx;
 }
 
 uint16_t