| To: | kuznet@xxxxxxxxxxxxx |
|---|---|
| Subject: | Re: TCP hang with 2.2.14 <-> 2.2.15pre5 |
| From: | "David S. Miller" <davem@xxxxxxxxxx> |
| Date: | Sun, 26 Mar 2000 03:27:41 -0800 |
| Cc: | sfr@xxxxxxxxxxxxxxxx, netdev@xxxxxxxxxxx |
| In-reply-to: | <E12ZBGU-0000ST-00@xxxxxxxxxxxxx> (kuznet@xxxxxxxxxxxxx) |
| References: | <E12ZBGU-0000ST-00@xxxxxxxxxxxxx> |
| Sender: | owner-netdev@xxxxxxxxxxx |
Date: Sun, 26 Mar 2000 19:31:30 +0800 (SGT)
From: kuznet@xxxxxxxxxxxxx (Alexey Kuznetosv)
> I would be happy to be told that this is a bug in 2.2.14 that is
> already fixed.
I do not remember. Let's search.
But! Despite of all this, rsync behaves strangely. Window became
zero not because of bug in tcp, but because of bug in rsync. If it
forgets to read data, it is deemed to dead-lock soon or later.
It is bug both in rsync and kernel :-)
Kernel fix is in 2.2.15-pre* for some time now, look at this
change to tcp_timer.c:tcp_write_err()
@@ -131,6 +131,7 @@
} else {
/* Clean up time. */
tcp_set_state(sk, TCP_CLOSE);
+ sk->shutdown |= SHUTDOWN_MASK;
return 0;
}
return 1;
Later,
David S. Miller
davem@xxxxxxxxxx
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: TCP hang with 2.2.14 <-> 2.2.15pre5, Alexey Kuznetosv |
|---|---|
| Next by Date: | Re: TCP hang with 2.2.14 <-> 2.2.15pre5, Alexey Kuznetosv |
| Previous by Thread: | Re: TCP hang with 2.2.14 <-> 2.2.15pre5, Alexey Kuznetosv |
| Next by Thread: | Re: TCP hang with 2.2.14 <-> 2.2.15pre5, Alexey Kuznetosv |
| Indexes: | [Date] [Thread] [Top] [All Lists] |