Messages in this thread | | | From | Whit Blauvelt <> | Subject | Re: tcp_keepalive crash - not egcs caused | Date | Wed, 1 Mar 2000 21:09:50 -0500 |
| |
On Wed, 01 Mar 2000, Alan Cox wrote: > > reverted to gcc 2.7.2.3 from the current Red Hat egcs. Do you suspect that > > the bug came about through egcs compilation (some of your changes, Dave, > > It would be very interesting if it did, since we believe egcs 1.1.x and > gcc 2.7.2 are both totally solid compilers
Alan, quite right, at least that it makes no difference. Recompiled 2.2.15.pre10 with the old gcc late this afternoon, set the tcp_keepalive_time to 60 (seconds) in hopes of triggering a crash while I was around, and by midevening, voila, the same crash. At the time there were two tcp sockets open, one to the machine on which I was watching "netstat -tc" for the server, and one for sendmail which was attempting to pass on a message to a nonresponsive system, as this shows:
tcp 0 1 china.patternbook.:1264 b.nonamecorp.com:smtp SYN_SENT tcp 0 0 china.patternboo:telnet whit.transpect.com:1136 ESTABLISHED
(those two lines repeated up off the screen in the buffer)
and the maillog on that one looked like this:
Mar 1 19:56:02 china sendmail[2028]: QAA01576: to=Tom Campbell <Tom.campbell@sid.net>, delay=03:21:16, xdelay=00:13:10, mailer=esmtp, relay=puffy.noname.com. [216.112.75.129], stat=Deferred: Connection timed out with puffy.noname.com.
So I'll set tcp_keepalive_time back to insanity, and revel in stability.
Whit
- 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/
| |