Messages in this thread | | | Date | Fri, 4 Jun 1999 07:56:10 -0400 (EDT) | From | "Byron R. Stanoszek" <> | Subject | Re: TCP/PPP bug 2.3.5? |
| |
On Fri, 4 Jun 1999, Rodel T. Viado wrote:
> > also experiencing this on 2.2.9 > > > On Fri, 4 Jun 1999, Rik van Riel wrote: > > > On Fri, 4 Jun 1999, Rik van Riel wrote: > > > > > it looks like there's some bug in the TCP stack of > > > at least 2.3.5. > > > > After looking closely, I have seen what seems to be > > a LIFO problem, so it might have something to do with > > bottom halves not getting run (or another LIFO order > > which I'm unaware of). > > > > When the symptoms occur, network packets are in the > > 'out' queue for several _minutes_, until the kernel > > comes around to sending them -- even on an almost > > empty line this problem occurs! > > > > Active Internet connections (w/o servers) > > 0oto Recv-Q Send-Q Local Address Foreign Address State > > tcp 0 64 agratax.demon.nl:1287 k052r.balpol.tudelf:cvs ESTABLISHED > > tcp 0 9 agratax.demon.nl:1286 humbolt.nl.linux.o:pop3 ESTABLISHED > > tcp 0 0 agratax.demon.nl:1284 k052r.balpol.tudelf:cvs TIME_WAIT > > tcp 0 6170 agratax.demon.nl:1274 terra.geo.uu.nl:smtp ESTABLISHED > > tcp 0 0 agratax.demon.nl:1231 pluto.news.demon.n:nntp ESTABLISHED > > tcp 0 0 localhost:7100 localhost:1028 ESTABLISHED > > tcp 0 0 localhost:1028 localhost:7100 ESTABLISHED > > > > The '6170' above has been there for almost 10 minutes (!), > > while newer traffic seems to run without too much problems. > > The newer '64' seems to have problems as well though :( > > > > This simply kills TCP connections. Somebody still awake > > should probably code an emergency patch so that a fix is > > available before Europe wakes up again :) > >
I experience this exact problem, but in a slightly different manner... Actually it's sometimes much worse.. From my box at home, any time I send bursts of data through PPP to a site connected elsewhere on the internet, TCP queues the data up for that one connection. Before that data is fully sent, viewing /proc/net/tcp shows that it retransmits all those packets. After the first retransmit, there is so much data going through that nothing ever returns to its original state and TCP continues to retransmit data, therefore locking up the entire ppp0 port. After about two retransmits, for some reason the kernel seems to lock up from ever sending anything through the modem again, while I know the modem is fully functioning. Thus a ping doesn't even get across. I've had this problem since I started using 2.2 kernels (2.2.5+) and under same hardware configuration had never experienced this the last time I used PPP on 2.1.50+. 2.3.5 currently has the same problem. This is very annoying because I can never send more than 64k of data through the network from my computer without it locking up, sometimes intermittently between 64k and 400k.
--- Byron Stanoszek <byron@math.uakron.edu> System Administrator - University of Akron Applied Mathematics Dept.
- 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/
| |