[PATCH] virtio_net: Fix duplicated return values in virtnet_get_hw_stats

[PATCH] virtio_net: Fix duplicated return values in virtnet_get_hw_stats

[Tang Longjun]

From: Tang Longjun <tanglongjun@kylinos.cn>

virtnet_get_hw_stats will return same value of 0 in the following three cases:
1. VIRTIO_NET_F_DEVICE_STATS feature is not configured on the device side.
2. success to obtain device stats.
3. fail to obtain device stats.
This would lead to case 1 and case 3 being unable to be recognized.

So, it's fixed by returning different values for the three cases.

Fixes: 941168f8b40e ("virtio_net: support device stats")
Signed-off-by: Tang Longjun <tanglongjun@kylinos.cn>
---
 drivers/net/virtio_net.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index e53ba600605a..c9a86f325619 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -4897,7 +4897,7 @@ static int __virtnet_get_hw_stats(struct virtnet_info *vi,
 					&sgs_out, &sgs_in);
 
 	if (!ok)
-		return ok;
+		return 1;
 
 	for (p = reply; p - reply < res_size; p += le16_to_cpu(hdr->size)) {
 		hdr = p;
@@ -4937,7 +4937,7 @@ static int virtnet_get_hw_stats(struct virtnet_info *vi,
 	int ok;
 
 	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_DEVICE_STATS))
-		return 0;
+		return -EOPNOTSUPP;
 
 	if (qid == -1) {
 		last_vq = vi->curr_queue_pairs * 2 - 1;
@@ -5038,10 +5038,12 @@ static void virtnet_get_ethtool_stats(struct net_device *dev,
 	struct virtnet_stats_ctx ctx = {0};
 	unsigned int start, i;
 	const u8 *stats_base;
+	int ret;
 
 	virtnet_stats_ctx_init(vi, &ctx, data, false);
-	if (virtnet_get_hw_stats(vi, &ctx, -1))
-		dev_warn(&vi->dev->dev, "Failed to get hw stats.\n");
+	ret = virtnet_get_hw_stats(vi, &ctx, -1);
+	if (ret)
+		dev_warn(&vi->dev->dev, "Failed to get hw stats. err code:%d\n", ret);
 
 	for (i = 0; i < vi->curr_queue_pairs; i++) {
 		struct receive_queue *rq = &vi->rq[i];
-- 
2.43.0

Re: [PATCH] virtio_net: Fix duplicated return values in virtnet_get_hw_stats

[Jakub Kicinski]

On Wed, 14 May 2025 13:44:33 +0800 Tang Longjun wrote:
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index e53ba600605a..c9a86f325619 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -4897,7 +4897,7 @@ static int __virtnet_get_hw_stats(struct virtnet_info *vi,
>  					&sgs_out, &sgs_in);
>  
>  	if (!ok)
> -		return ok;
> +		return 1;

This makes sense, looks like a typo in the original code.
But we are now returning the reverse polarity of "ok", so we should
probably rename the variable in virtnet_get_hw_stats() ok -> failed
Or invert the polarity here and in virtnet_get_hw_stats()

>  	for (p = reply; p - reply < res_size; p += le16_to_cpu(hdr->size)) {
>  		hdr = p;
> @@ -4937,7 +4937,7 @@ static int virtnet_get_hw_stats(struct virtnet_info *vi,
>  	int ok;
>  
>  	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_DEVICE_STATS))
> -		return 0;
> +		return -EOPNOTSUPP;

IDK about this part. We should not spam the logs if the device does not
support a feature. We should instead skip reporting the relevant stats.
User can tell that those stats are not supported from the fact they are
not present.
-- 
pw-bot: cr

Re:Re: [PATCH] virtio_net: Fix duplicated return values in virtnet_get_hw_stats

[Lange Tang]

At 2025-05-15 22:10:42, "Jakub Kicinski" <kuba@kernel.org> wrote:
>On Wed, 14 May 2025 13:44:33 +0800 Tang Longjun wrote:
>> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
>> index e53ba600605a..c9a86f325619 100644
>> --- a/drivers/net/virtio_net.c
>> +++ b/drivers/net/virtio_net.c
>> @@ -4897,7 +4897,7 @@ static int __virtnet_get_hw_stats(struct virtnet_info *vi,
>>  					&sgs_out, &sgs_in);
>>  
>>  	if (!ok)
>> -		return ok;
>> +		return 1;
>
>This makes sense, looks like a typo in the original code.
>But we are now returning the reverse polarity of "ok", so we should
>probably rename the variable in virtnet_get_hw_stats() ok -> failed
>Or invert the polarity here and in virtnet_get_hw_stats()
>
>>  	for (p = reply; p - reply < res_size; p += le16_to_cpu(hdr->size)) {
>>  		hdr = p;
>> @@ -4937,7 +4937,7 @@ static int virtnet_get_hw_stats(struct virtnet_info *vi,
>>  	int ok;
>>  
>>  	if (!virtio_has_feature(vi->vdev, VIRTIO_NET_F_DEVICE_STATS))
>> -		return 0;
>> +		return -EOPNOTSUPP;
>
>IDK about this part. We should not spam the logs if the device does not
>support a feature. We should instead skip reporting the relevant stats.
>User can tell that those stats are not supported from the fact they are
>not present.
>-- 

>pw-bot: cr


I found that Dan Carpenter had discovered the same issue before, and he had already submitted a patch, so I'm backing out.
https://lore.kernel.org/netdev/20240515110626-mutt-send-email-mst@kernel.org/ 



regards,
Tang Longjun