[PATCH] arc_emac: fix

[PATCH] arc_emac: fix

[Eric Dumazet]

Signed-off-by: Eric Dumazet <edumazet@google.com>

skb_tx_timestamp(skb) should be called _before_ TX completion
has a chance to trigger, otherwise it is too late and we access
freed memory.

Fixes: e4f2379db6c6 ("ethernet/arc/arc_emac - Add new driver")
From: Eric Dumazet <edumazet@google.com>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Richard Cochran <richardcochran@gmail.com>
---
 drivers/net/ethernet/arc/emac_main.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/arc/emac_main.c b/drivers/net/ethernet/arc/emac_main.c
index b2ffad1304d2..248baf6273fb 100644
--- a/drivers/net/ethernet/arc/emac_main.c
+++ b/drivers/net/ethernet/arc/emac_main.c
@@ -565,6 +565,8 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
 	/* Make sure pointer to data buffer is set */
 	wmb();
 
+	skb_tx_timestamp(skb);
+
 	*info = cpu_to_le32(FOR_EMAC | FIRST_OR_LAST_MASK | len);
 
 	/* Increment index to point to the next BD */
@@ -579,8 +581,6 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
 
 	arc_reg_set(priv, R_STATUS, TXPL_MASK);
 
-	skb_tx_timestamp(skb);
-
 	return NETDEV_TX_OK;
 }
 

Re: [PATCH] arc_emac: fix

[Eric Dumazet]

On Thu, 2013-12-19 at 17:58 -0800, Eric Dumazet wrote:
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> 
> skb_tx_timestamp(skb) should be called _before_ TX completion
> has a chance to trigger, otherwise it is too late and we access
> freed memory.
> 
> Fixes: e4f2379db6c6 ("ethernet/arc/arc_emac - Add new driver")
> From: Eric Dumazet <edumazet@google.com>
> Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
> Cc: Richard Cochran <richardcochran@gmail.com>
> ---
>  drivers/net/ethernet/arc/emac_main.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/arc/emac_main.c b/drivers/net/ethernet/arc/emac_main.c
> index b2ffad1304d2..248baf6273fb 100644
> --- a/drivers/net/ethernet/arc/emac_main.c
> +++ b/drivers/net/ethernet/arc/emac_main.c
> @@ -565,6 +565,8 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
>  	/* Make sure pointer to data buffer is set */
>  	wmb();
>  
> +	skb_tx_timestamp(skb);
> +
>  	*info = cpu_to_le32(FOR_EMAC | FIRST_OR_LAST_MASK | len);
>  
>  	/* Increment index to point to the next BD */
> @@ -579,8 +581,6 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
>  
>  	arc_reg_set(priv, R_STATUS, TXPL_MASK);
>  
> -	skb_tx_timestamp(skb);
> -
>  	return NETDEV_TX_OK;
>  }
>  

Sorry I messed patch title, will resend

[PATCH v2] arc_emac: fix potential use after free

[Eric Dumazet]

Signed-off-by: Eric Dumazet <edumazet@google.com>

skb_tx_timestamp(skb) should be called _before_ TX completion
has a chance to trigger, otherwise it is too late and we access
freed memory.

Fixes: e4f2379db6c6 ("ethernet/arc/arc_emac - Add new driver")
From: Eric Dumazet <edumazet@google.com>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Richard Cochran <richardcochran@gmail.com>
---
 drivers/net/ethernet/arc/emac_main.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/arc/emac_main.c b/drivers/net/ethernet/arc/emac_main.c
index b2ffad1304d2..248baf6273fb 100644
--- a/drivers/net/ethernet/arc/emac_main.c
+++ b/drivers/net/ethernet/arc/emac_main.c
@@ -565,6 +565,8 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
 	/* Make sure pointer to data buffer is set */
 	wmb();
 
+	skb_tx_timestamp(skb);
+
 	*info = cpu_to_le32(FOR_EMAC | FIRST_OR_LAST_MASK | len);
 
 	/* Increment index to point to the next BD */
@@ -579,8 +581,6 @@ static int arc_emac_tx(struct sk_buff *skb, struct net_device *ndev)
 
 	arc_reg_set(priv, R_STATUS, TXPL_MASK);
 
-	skb_tx_timestamp(skb);
-
 	return NETDEV_TX_OK;
 }
 

Re: [PATCH v2] arc_emac: fix potential use after free

[Alexey Brodkin]

On Thu, 2013-12-19 at 18:10 -0800, Eric Dumazet wrote:
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> 
> skb_tx_timestamp(skb) should be called _before_ TX completion
> has a chance to trigger, otherwise it is too late and we access
> freed memory.
> 
> Fixes: e4f2379db6c6 ("ethernet/arc/arc_emac - Add new driver")

Hi Eric,

thanks for this fix.

Indeed I missed this part and description of "skb_tx_timestamp" which
clearly directs to use it earlier:
=======
 * skb_tx_timestamp() - Driver hook for transmit timestamping
 *
 * Ethernet MAC Drivers should call this function in their hard_xmit()
 * function immediately before giving the sk_buff to the MAC hardware.
=======

-Alexey

Re: [PATCH v2] arc_emac: fix potential use after free

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 19 Dec 2013 18:10:40 -0800

> Signed-off-by: Eric Dumazet <edumazet@google.com>
> 
> skb_tx_timestamp(skb) should be called _before_ TX completion
> has a chance to trigger, otherwise it is too late and we access
> freed memory.
> 
> Fixes: e4f2379db6c6 ("ethernet/arc/arc_emac - Add new driver")
> From: Eric Dumazet <edumazet@google.com>
> Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
> Cc: Richard Cochran <richardcochran@gmail.com>

Applied, thanks.

I've seen this bug enough times that I've expanded the comment
a bit more.

====================
[PATCH] net: Add some clarification to skb_tx_timestamp() comment.

We've seen so many instances of people invoking skb_tx_timestamp()
after the device already has been given the packet, that it's worth
being a little bit more verbose and explicit in this comment.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 include/linux/skbuff.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 6aae838..6f69b3f 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -2531,6 +2531,10 @@ static inline void sw_tx_timestamp(struct sk_buff *skb)
  * Ethernet MAC Drivers should call this function in their hard_xmit()
  * function immediately before giving the sk_buff to the MAC hardware.
  *
+ * Specifically, one should make absolutely sure that this function is
+ * called before TX completion of this packet can trigger.  Otherwise
+ * the packet could potentially already be freed.
+ *
  * @skb: A socket buffer.
  */
 static inline void skb_tx_timestamp(struct sk_buff *skb)
-- 
1.7.11.7