When only GSO skb was partially ACKed, no hints are reset,
therefore fastpath_cnt_hint must be tweaked too or else it can
corrupt fackets_out. The corruption to occur, one must have
non-trivial ACK/SACK sequence, so this bug is not very often
that harmful. There's a fackets_out state reset in TCP because
fackets_out is known to be inaccurate and that fixes the issue
eventually anyway.
In case there was also at least one skb that got fully ACKed,
the fastpath_skb_hint is set to NULL which causes a recount for
fastpath_cnt_hint (the old value won't be accessed anymore),
thus it can safely be decremented without additional checking.
Reported by Cedric Le Goater <[EMAIL PROTECTED]>
Signed-off-by: Ilpo Järvinen <[EMAIL PROTECTED]>
---
Dave, here's the same fix to upcoming 2.6.23, hopefully it can
make it in time. Probably minor enough to skip stable-2.6.22,
especially since 2.6.23 is soooo close :-), but I leave that
up to you.
net/ipv4/tcp_input.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index bbad2cd..f893e90 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2420,6 +2420,9 @@ static int tcp_tso_acked(struct sock *sk, struct sk_buff
*skb,
__u32 dval = min(tp->fackets_out, packets_acked);
tp->fackets_out -= dval;
}
+ /* hint's skb might be NULL but we don't need to care */
+ tp->fastpath_cnt_hint -= min_t(u32, packets_acked,
+ tp->fastpath_cnt_hint);
tp->packets_out -= packets_acked;
BUG_ON(tcp_skb_pcount(skb) == 0);
--
1.5.0.6
