> Networking is a good one for showing up very specific time related
> problems. Does your stall remain if you tcpdump the session, and does
> disabling sack make any change (its in /proc/sys/net/...).
Yes, the stall remains if I tcpdump the session. Disabling sack does
not make any change.
The stall only appears if the client reads small amounts of data
(10-200 bytes) per read() while the server sends lots of data. So
many real-world-applications are not affected.
Here is a fresh tcpdump, the client stalls after reading 30kB (of 5Mb)
in a select() call, waiting for the socket:
15:57:16.361625 localhost.1024 > localhost.3333: S 891136714:891136714(0) win 31072 <mss 3884,sackOK,timestamp 7039 0,nop,wscale 0> (DF)
15:57:16.361625 localhost.1024 > localhost.3333: S 891136714:891136714(0) win 31072 <mss 3884,sackOK,timestamp 7039 0,nop,wscale 0> (DF)
15:57:16.361685 localhost.3333 > localhost.1024: S 882442719:882442719(0) ack 891136715 win 31072 <mss 3884,sackOK,timestamp 7039 7039,nop,wscale 0> (DF)
15:57:16.361685 localhost.3333 > localhost.1024: S 882442719:882442719(0) ack 891136715 win 31072 <mss 3884,sackOK,timestamp 7039 7039,nop,wscale 0> (DF)
15:57:16.361711 localhost.1024 > localhost.3333: . ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.361711 localhost.1024 > localhost.3333: . ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362897 localhost.3333 > localhost.1024: P 1:1586(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362897 localhost.3333 > localhost.1024: P 1:1586(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362929 localhost.1024 > localhost.3333: . ack 1586 win 29487 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362929 localhost.1024 > localhost.3333: . ack 1586 win 29487 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362975 localhost.3333 > localhost.1024: P 1586:3171(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362975 localhost.3333 > localhost.1024: P 1586:3171(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362991 localhost.1024 > localhost.3333: . ack 3171 win 27902 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.362991 localhost.1024 > localhost.3333: . ack 3171 win 27902 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363030 localhost.3333 > localhost.1024: P 3171:4756(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363030 localhost.3333 > localhost.1024: P 3171:4756(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363043 localhost.1024 > localhost.3333: . ack 4756 win 26317 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363043 localhost.1024 > localhost.3333: . ack 4756 win 26317 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363083 localhost.3333 > localhost.1024: P 4756:6341(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363083 localhost.3333 > localhost.1024: P 4756:6341(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363097 localhost.1024 > localhost.3333: . ack 6341 win 24732 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363097 localhost.1024 > localhost.3333: . ack 6341 win 24732 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363135 localhost.3333 > localhost.1024: P 6341:7926(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363135 localhost.3333 > localhost.1024: P 6341:7926(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363153 localhost.1024 > localhost.3333: . ack 7926 win 23147 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363153 localhost.1024 > localhost.3333: . ack 7926 win 23147 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363194 localhost.3333 > localhost.1024: P 7926:9511(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363194 localhost.3333 > localhost.1024: P 7926:9511(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363207 localhost.1024 > localhost.3333: . ack 9511 win 21562 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363207 localhost.1024 > localhost.3333: . ack 9511 win 21562 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363244 localhost.3333 > localhost.1024: P 9511:11096(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363244 localhost.3333 > localhost.1024: P 9511:11096(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363257 localhost.1024 > localhost.3333: . ack 11096 win 19977 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363257 localhost.1024 > localhost.3333: . ack 11096 win 19977 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363295 localhost.3333 > localhost.1024: P 11096:12681(1585) ack 1 win 31072 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363309 localhost.1024 > localhost.3333: . ack 12681 win 18392 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363309 localhost.1024 > localhost.3333: . ack 12681 win 18392 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363358 localhost.1024 > localhost.3333: . ack 14266 win 16807 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363358 localhost.1024 > localhost.3333: . ack 14266 win 16807 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363412 localhost.1024 > localhost.3333: . ack 15851 win 15222 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363412 localhost.1024 > localhost.3333: . ack 15851 win 15222 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363462 localhost.1024 > localhost.3333: . ack 17436 win 13637 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363462 localhost.1024 > localhost.3333: . ack 17436 win 13637 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363515 localhost.1024 > localhost.3333: . ack 19021 win 12052 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363515 localhost.1024 > localhost.3333: . ack 19021 win 12052 <nop,nop,timestamp 7039 7039> (DF)
15:57:16.363564 localhost.1024 > localhost.3333: . ack 20606 win 10467 <nop,nop,timestamp 7039 7039> (DF)
netstat:
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 39625 localhost:3333 localhost:1040 ESTABLISHED
tcp 0 0 localhost:1040 localhost:3333 ESTABLISHED
I think the dump looks okay? Could I do some more debugging?
Matthias
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/