Re: [PATCH net-next] net: fix __skb_try_recv_from_queue to return the old behavior

[Paolo Abeni]

On Tue, 2017-05-16 at 21:47 -0700, Andrei Vagin wrote:
> This function has to return NULL on a error case, because there is a
> separate error variable.
> 
> The offset has to be changed only if skb is returned
> 
> Cc: Paolo Abeni 
> Cc: Eric Dumazet 
> Cc: David S. Miller 
> Fixes: 65101aeca522 ("net/sock: factor out dequeue/peek with offset cod")
> Signed-off-by: Andrei Vagin 
> ---
>  net/core/datagram.c | 14 --
>  1 file changed, 8 insertions(+), 6 deletions(-)
> 
> diff --git a/net/core/datagram.c b/net/core/datagram.c
> index a4592b4..bc46118 100644
> --- a/net/core/datagram.c
> +++ b/net/core/datagram.c
> @@ -170,20 +170,21 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock 
> *sk,
> struct sk_buff **last)
>  {
>   struct sk_buff *skb;
> + int _off = *off;
>  
>   *last = queue->prev;
>   skb_queue_walk(queue, skb) {
>   if (flags & MSG_PEEK) {
> - if (*off >= skb->len && (skb->len || *off ||
> + if (_off >= skb->len && (skb->len || _off ||
>skb->peeked)) {
> - *off -= skb->len;
> + _off -= skb->len;
>   continue;
>   }
>   if (!skb->len) {
>   skb = skb_set_peeked(skb);
>   if (unlikely(IS_ERR(skb))) {
>   *err = PTR_ERR(skb);
> - return skb;
> + return NULL;
>   }
>   }
>   *peeked = 1;
> @@ -193,6 +194,7 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
>   if (destructor)
>   destructor(sk, skb);
>   
>   }
> + *off = _off;
>   return skb;
>   }
>   return NULL;
> @@ -253,8 +255,6 @@ struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 
> unsigned int flags,
>  
>   *peeked = 0;
>   do {
> - int _off = *off;
> -
>   /* Again only user level code calls this function, so nothing
>* interrupt level will suddenly eat the receive_queue.
>*
> @@ -263,8 +263,10 @@ struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 
> unsigned int flags,
>*/
>   spin_lock_irqsave(>lock, cpu_flags);
>   skb = __skb_try_recv_from_queue(sk, queue, flags, destructor,
> - peeked, &_off, err, last);
> + peeked, off, , last);
>   spin_unlock_irqrestore(>lock, cpu_flags);
> + if (error)
> + goto no_packet;
>   if (skb)
>   return skb;
>  

Thank you for catching this so early! 

If __skb_try_recv_from_queue() updates the offset if/only if the skb is
returned, than we can also add something like the following ?
(only compile tested, should not bring any functional changes, only
code cleanup)

BTW can you please share some entry pointer/walk-through for the CRIU
tests ? I'd like to try to integrate them in my workflow, thank you!

Paolo
---
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 8ad5862..e65c7ed 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -1555,16 +1555,13 @@ struct sk_buff *__skb_recv_udp(struct sock *sk, 
unsigned int flags,
error = -EAGAIN;
*peeked = 0;
do {
-   int _off = *off;
-
spin_lock_bh(>lock);
skb = __skb_try_recv_from_queue(sk, queue, flags,
udp_skb_destructor,
-   peeked, &_off, err,
+   peeked, off, err,
);
if (skb) {
spin_unlock_bh(>lock);
-   *off = _off;
return skb;
}
 
@@ -1578,20 +1575,17 @@ struct sk_buff *__skb_recv_udp(struct sock *sk, 
unsigned int flags,
 * the sk_receive_queue lock if fwd memory scheduling
 * is needed.
 */
-   _off = *off;
spin_lock(_queue->lock);
skb_queue_splice_tail_init(sk_queue, queue);
 
skb = __skb_try_recv_from_queue(sk, queue, flags,
udp_skb_dtor_locked,
-   

[PATCH net-next] net: fix __skb_try_recv_from_queue to return the old behavior

[Andrei Vagin]

This function has to return NULL on a error case, because there is a
separate error variable.

The offset has to be changed only if skb is returned

Cc: Paolo Abeni 
Cc: Eric Dumazet 
Cc: David S. Miller 
Fixes: 65101aeca522 ("net/sock: factor out dequeue/peek with offset cod")
Signed-off-by: Andrei Vagin 
---
 net/core/datagram.c | 14 --
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/net/core/datagram.c b/net/core/datagram.c
index a4592b4..bc46118 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -170,20 +170,21 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  struct sk_buff **last)
 {
struct sk_buff *skb;
+   int _off = *off;
 
*last = queue->prev;
skb_queue_walk(queue, skb) {
if (flags & MSG_PEEK) {
-   if (*off >= skb->len && (skb->len || *off ||
+   if (_off >= skb->len && (skb->len || _off ||
 skb->peeked)) {
-   *off -= skb->len;
+   _off -= skb->len;
continue;
}
if (!skb->len) {
skb = skb_set_peeked(skb);
if (unlikely(IS_ERR(skb))) {
*err = PTR_ERR(skb);
-   return skb;
+   return NULL;
}
}
*peeked = 1;
@@ -193,6 +194,7 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
if (destructor)
destructor(sk, skb);
}
+   *off = _off;
return skb;
}
return NULL;
@@ -253,8 +255,6 @@ struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 
unsigned int flags,
 
*peeked = 0;
do {
-   int _off = *off;
-
/* Again only user level code calls this function, so nothing
 * interrupt level will suddenly eat the receive_queue.
 *
@@ -263,8 +263,10 @@ struct sk_buff *__skb_try_recv_datagram(struct sock *sk, 
unsigned int flags,
 */
spin_lock_irqsave(>lock, cpu_flags);
skb = __skb_try_recv_from_queue(sk, queue, flags, destructor,
-   peeked, &_off, err, last);
+   peeked, off, , last);
spin_unlock_irqrestore(>lock, cpu_flags);
+   if (error)
+   goto no_packet;
if (skb)
return skb;
 
-- 
2.9.3