David Miller <[EMAIL PROTECTED]> wrote on 01/23/2008 01:27:23 PM:
> > iperf with multiple threads almost always gets these 4, *especially*
when I
> > do some batching :).
> >
> > static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int
flag)
> > {
> > ...
> > if (WARN_ON(!tp->sacked_out && tp->fackets_out))
> > tp->fackets_out = 0;
> > ...
> > }
>
> Does this assertion show up first or do you get the other TCP
> ones first? It might be important, in that if you get the
> others ones first that corrupted state might be what leads to
> this one.
Hi Dave,
I looked at my *old* messages file and found this assert (2506) was first
to hit (atleast in
two messages file). It hit 5 times, then I got a different one that I had
not reported earlier:
"KERNEL: assertion (packets <= tp->packets_out) failed at
net/ipv4/tcp_input.c (2139)"
(though this was hidden in my report under the panic for tcp_input.c:2528.
Then another two thousand times of the 2506 asserts.
Today I installed the latest untouched kernel, rebooted system and got the
following errors
in sequence, but no 2506 errors (which I have always got when running
batching in the last
2-3 weeks):
Jan 22 02:07:55 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:56 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:56 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:57 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:58 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:59 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:07:59 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:00 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:01 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:01 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:02 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:03 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:03 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:04 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:05 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:06 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:06 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:07 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:1767
Jan 22 02:08:07 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:08 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:09 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:10 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:10 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:11 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:12 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:12 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:13 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:14 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:15 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:15 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:1767
Jan 22 02:08:16 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:1767
Jan 22 02:08:16 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:1767
Jan 22 02:08:17 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:18 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:18 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
Jan 22 02:08:19 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2528
Jan 22 02:08:19 elm3b39 kernel: Badness at net/ipv4/tcp_input.c:2169
and so on for another 700 counts.
The unique asserts are:
1767: tcp_verify_left_out (from tcp_entry_frto)
2169: tcp_verify_left_out (from tcp_mark_head_lost)
2528: tcp_verify_left_out (from tcp_fastretrans_alert)
3063: tcp_verify_left_out (from tcp_process_frto)
(where 2169 seems to preceed any other asserts)
The other two asserts that I got only with batching are:
2139: BUG_TRAP(packets <= tp->packets_out); (in tcp_mark_head_lost)
2506: WARN_ON(!tp->sacked_out && tp->fackets_out) (in
tcp_fastretrans_alert)
(where 2506 always seems to preceed any other asserts).
thanks,
- KK
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
