On 26-01-2007 04:37, Masayuki Nakagawa wrote:

This patch simply replaces __kfree_skb() in exit path with kfree_skb().
In tcp_rcv_state_process(), generally skbs should be destroyed only when
the ref count is zero.
That is the way things are supposed to be done in the kernel.

This change might reveals a memory leak of skb.
If it happens, it would be because someone doesn't deal with the skb properly.

So maybe for some time there should be added skb->users
test with a warning to reveal this?

Regards,
Jarek P.