[PATCH net] net: bcmgenet: fix accounting of packet drops vs errors

[PATCH net] net: bcmgenet: fix accounting of packet drops vs errors

[Petri Gynther]

bcmgenet driver needs to separate packet drops from packet errors.

When the driver has to drop a *good* packet, due to lack of buffers or
replacement skbs, increment only dev->stats.[rx|tx]_dropped.

When the driver encounters a bad Rx packet or Tx error, increment only
dev->stats.[rx|tx]_errors + relevant detailed error counter.

Signed-off-by: Petri Gynther <pgynther@google.com>
---
  drivers/net/ethernet/broadcom/genet/bcmgenet.c | 4 ----
  1 file changed, 4 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c  
b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
index b43b2cb..64c1e9d 100644
--- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
+++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
@@ -1230,7 +1230,6 @@ static struct sk_buff *bcmgenet_put_tx_csum(struct  
net_device *dev,
  		new_skb = skb_realloc_headroom(skb, sizeof(*status));
  		dev_kfree_skb(skb);
  		if (!new_skb) {
-			dev->stats.tx_errors++;
  			dev->stats.tx_dropped++;
  			return NULL;
  		}
@@ -1465,7 +1464,6 @@ static unsigned int bcmgenet_desc_rx(struct  
bcmgenet_rx_ring *ring,

  		if (unlikely(!skb)) {
  			dev->stats.rx_dropped++;
-			dev->stats.rx_errors++;
  			goto next;
  		}

@@ -1493,7 +1491,6 @@ static unsigned int bcmgenet_desc_rx(struct  
bcmgenet_rx_ring *ring,
  		if (unlikely(!(dma_flag & DMA_EOP) || !(dma_flag & DMA_SOP))) {
  			netif_err(priv, rx_status, dev,
  				  "dropping fragmented packet!\n");
-			dev->stats.rx_dropped++;
  			dev->stats.rx_errors++;
  			dev_kfree_skb_any(skb);
  			goto next;
@@ -1515,7 +1512,6 @@ static unsigned int bcmgenet_desc_rx(struct  
bcmgenet_rx_ring *ring,
  				dev->stats.rx_frame_errors++;
  			if (dma_flag & DMA_RX_LG)
  				dev->stats.rx_length_errors++;
-			dev->stats.rx_dropped++;
  			dev->stats.rx_errors++;
  			dev_kfree_skb_any(skb);
  			goto next;
-- 
2.4.3.573.g4eafbef

Re: [PATCH net] net: bcmgenet: fix accounting of packet drops vs errors

[Florian Fainelli]

Le 07/08/15 18:55, Petri Gynther a écrit :
> bcmgenet driver needs to separate packet drops from packet errors.
> 
> When the driver has to drop a *good* packet, due to lack of buffers or
> replacement skbs, increment only dev->stats.[rx|tx]_dropped.
> 
> When the driver encounters a bad Rx packet or Tx error, increment only
> dev->stats.[rx|tx]_errors + relevant detailed error counter.
> 
> Signed-off-by: Petri Gynther <pgynther@google.com>

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>

> ---
>  drivers/net/ethernet/broadcom/genet/bcmgenet.c | 4 ----
>  1 file changed, 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> index b43b2cb..64c1e9d 100644
> --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> @@ -1230,7 +1230,6 @@ static struct sk_buff *bcmgenet_put_tx_csum(struct
> net_device *dev,
>          new_skb = skb_realloc_headroom(skb, sizeof(*status));
>          dev_kfree_skb(skb);
>          if (!new_skb) {
> -            dev->stats.tx_errors++;
>              dev->stats.tx_dropped++;
>              return NULL;
>          }
> @@ -1465,7 +1464,6 @@ static unsigned int bcmgenet_desc_rx(struct
> bcmgenet_rx_ring *ring,
> 
>          if (unlikely(!skb)) {
>              dev->stats.rx_dropped++;
> -            dev->stats.rx_errors++;
>              goto next;
>          }
> 
> @@ -1493,7 +1491,6 @@ static unsigned int bcmgenet_desc_rx(struct
> bcmgenet_rx_ring *ring,
>          if (unlikely(!(dma_flag & DMA_EOP) || !(dma_flag & DMA_SOP))) {
>              netif_err(priv, rx_status, dev,
>                    "dropping fragmented packet!\n");
> -            dev->stats.rx_dropped++;
>              dev->stats.rx_errors++;
>              dev_kfree_skb_any(skb);
>              goto next;
> @@ -1515,7 +1512,6 @@ static unsigned int bcmgenet_desc_rx(struct
> bcmgenet_rx_ring *ring,
>                  dev->stats.rx_frame_errors++;
>              if (dma_flag & DMA_RX_LG)
>                  dev->stats.rx_length_errors++;
> -            dev->stats.rx_dropped++;
>              dev->stats.rx_errors++;
>              dev_kfree_skb_any(skb);
>              goto next;


-- 
Florian

Re: [PATCH net] net: bcmgenet: fix accounting of packet drops vs errors

[David Miller]

From: Petri Gynther <pgynther@google.com>
Date: Thu, 09 Jul 2015 01:55:02 +0000

> bcmgenet driver needs to separate packet drops from packet errors.
> 
> When the driver has to drop a *good* packet, due to lack of buffers or
> replacement skbs, increment only dev->stats.[rx|tx]_dropped.
> 
> When the driver encounters a bad Rx packet or Tx error, increment only
> dev->stats.[rx|tx]_errors + relevant detailed error counter.
> 
> Signed-off-by: Petri Gynther <pgynther@google.com>
> ---
>  drivers/net/ethernet/broadcom/genet/bcmgenet.c | 4 ----
>  1 file changed, 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> index b43b2cb..64c1e9d 100644
> --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> @@ -1230,7 +1230,6 @@ static struct sk_buff
> *bcmgenet_put_tx_csum(struct net_device *dev,

This is severely corrupted by your email client.

Please fix this, email a test patch to yourself, and only resubmit
this patch to the mailing list if you can successfully apply the
patch you receive in that test email.

Thanks.