On Wed, 2007-07-04 at 08:33 +0200, Bernd Eckenfels wrote:
> In article <E1I5pNO-0006Xb-00@xxxxxxxxxxxxxxxxxxxxx> you wrote:
> > Search for UDP_MIB_INERRORS in linux/net/**.c
>
> ... but most likely if you see nothing in dmesg (checksum error) it was
> dropped due to a filled up socket buffer.
>
I checked this and I see that indeed the socket buffer is exhausted; I
fail here:
int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
{
int err = 0;
int skb_len;
/* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
number of warnings when compiling with -W --ANK
*/
*here* if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
(unsigned)sk->sk_rcvbuf) {
err = -ENOMEM;
goto out;
}
My receive ring allocates 2KB SKBs - when the sender sends small
messages, the packet rate is high and I fail the condition above. To
check my assumption and possibly use as a solution, I would like to
check for all incoming SKBs if they are small to a certain extent, I
will create a new SKB of smaller size and copy the data from the
original SKB. This way I expect to not fail as before. The problem is
that I can't find a proper function from the kernel to do this.
copy_skb() will allocate an SKB with the original size so I don't earn
anything. Can you suggest a way to do this?
Thanks,
Eli
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
