Frank van Maarseveen wrote:
> The 2.6.13.2 data is inconsistent. The bug appears to be present there at
> well after closer examination. So there must be another factor involved
> because I have at least one case logged where 2.6.13.2 did work (the
> "sirkka" log in my previous mail). Applying your patch on 2.6.13.2
> again removes the protocol is buggy messages (when doing a tcpdump)
> but the problem of the 10 missing packets persists.
Which network driver are you using? Does this patch show anything in
the ringbuffer?
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index e8e05ce..2b12280 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -302,6 +302,9 @@ static void netpoll_send_skb(struct netp
netpoll_poll(np);
udelay(50);
} while (npinfo->tries > 0);
+
+ printk("failed to transmit\n");
+ kfree_skb(skb);
}
void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
