On Fri, 27 Aug 2004 12:52:25 +0200, Pablo Neira wrote:
[ added some additional context ]
> if (!timeo) {
> if (!nlk->pid)
> netlink_overrun(sk);
> sock_put(sk);
> kfree_skb(skb);
> return -EAGAIN;
> }
>
> + if (nlk->pid) {
> + /* Kernel is sending information to user space
> + * and socket buffer is full: Wake up user
> + * process */
> + client = find_task_by_pid(nlk->pid);
> + if (!client) {
> + sock_put(sk);
> + kfree_skb(skb);
> + return -EAGAIN;
> + }
> + wake_up_process(client);
> + }
> +
> __set_current_state(TASK_INTERRUPTIBLE);
> add_wait_queue(&nlk->wait, &wait);
What test case did you use for this patch? AFAICS any netlink_unicast
caller passing MSG_DONTWAIT (i.e. all of them) will still have the skb
freed and return with -EAGAIN. However, the caller itself cannot block
no matter if MSG_DONTWAIT is passed or not. This suggests an inspection
of existing callers, to check whether they are still optimal with the
new semantics. No?
Roger
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
