[PATCH 4/5] net: wwan: iosm: fix netdev tx stats

[PATCH 4/5] net: wwan: iosm: fix netdev tx stats

[M Chetan Kumar]

Update tx stats on successful packet consume, drop.

Signed-off-by: M Chetan Kumar <m.chetan.kumar@linux.intel.com>
---
 drivers/net/wwan/iosm/iosm_ipc_wwan.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/net/wwan/iosm/iosm_ipc_wwan.c b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
index 84e37c4b0f74..561944a33725 100644
--- a/drivers/net/wwan/iosm/iosm_ipc_wwan.c
+++ b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
@@ -123,6 +123,8 @@ static int ipc_wwan_link_transmit(struct sk_buff *skb,
 
 	/* Return code of zero is success */
 	if (ret == 0) {
+		netdev->stats.tx_packets++;
+		netdev->stats.tx_bytes += skb->len;
 		ret = NETDEV_TX_OK;
 	} else if (ret == -EBUSY) {
 		ret = NETDEV_TX_BUSY;
@@ -140,7 +142,8 @@ static int ipc_wwan_link_transmit(struct sk_buff *skb,
 			ret);
 
 	dev_kfree_skb_any(skb);
-	return ret;
+	netdev->stats.tx_dropped++;
+	return NETDEV_TX_OK;
 }
 
 /* Ops structure for wwan net link */
-- 
2.25.1

Re: [PATCH 4/5] net: wwan: iosm: fix netdev tx stats

[Eric Dumazet]

On 6/30/21 8:27 PM, M Chetan Kumar wrote:
> Update tx stats on successful packet consume, drop.
> 
> Signed-off-by: M Chetan Kumar <m.chetan.kumar@linux.intel.com>
> ---
>  drivers/net/wwan/iosm/iosm_ipc_wwan.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/wwan/iosm/iosm_ipc_wwan.c b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
> index 84e37c4b0f74..561944a33725 100644
> --- a/drivers/net/wwan/iosm/iosm_ipc_wwan.c
> +++ b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
> @@ -123,6 +123,8 @@ static int ipc_wwan_link_transmit(struct sk_buff *skb,
>  
>  	/* Return code of zero is success */
>  	if (ret == 0) {
> +		netdev->stats.tx_packets++;
> +		netdev->stats.tx_bytes += skb->len;

What makes you think skb has not been consumed already ?
It seems clear it has been given, this thread can not expect skb has not been mangled/freed.
skb->len might now contain garbage, or even crash the kernel under appropriate debug features.

>  		ret = NETDEV_TX_OK;
>  	} else if (ret == -EBUSY) {
>  		ret = NETDEV_TX_BUSY;
> @@ -140,7 +142,8 @@ static int ipc_wwan_link_transmit(struct sk_buff *skb,
>  			ret);
>  
>  	dev_kfree_skb_any(skb);
> -	return ret;
> +	netdev->stats.tx_dropped++;
> +	return NETDEV_TX_OK;
>  }
>  
>  /* Ops structure for wwan net link */
> 

Re: [PATCH 4/5] net: wwan: iosm: fix netdev tx stats

[Kumar, M Chetan]

On 7/1/2021 2:06 AM, Eric Dumazet wrote:
> 
> 
> On 6/30/21 8:27 PM, M Chetan Kumar wrote:
>> Update tx stats on successful packet consume, drop.
>>
>> Signed-off-by: M Chetan Kumar <m.chetan.kumar@linux.intel.com>
>> ---
>>  drivers/net/wwan/iosm/iosm_ipc_wwan.c | 5 ++++-
>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/wwan/iosm/iosm_ipc_wwan.c b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
>> index 84e37c4b0f74..561944a33725 100644
>> --- a/drivers/net/wwan/iosm/iosm_ipc_wwan.c
>> +++ b/drivers/net/wwan/iosm/iosm_ipc_wwan.c
>> @@ -123,6 +123,8 @@ static int ipc_wwan_link_transmit(struct sk_buff *skb,
>>  
>>  	/* Return code of zero is success */
>>  	if (ret == 0) {
>> +		netdev->stats.tx_packets++;
>> +		netdev->stats.tx_bytes += skb->len;
> 
> What makes you think skb has not been consumed already ?
> It seems clear it has been given, this thread can not expect skb has not been mangled/freed.
> skb->len might now contain garbage, or even crash the kernel under appropriate debug features.

Ya. there could be a possibility skb might have been dequeued from ul_list and
UL task is already processing it. We can't rule out.

Will backup skb->len to local var and use it for tx_bytes update.

Regards,
Chetan