Date: Mon, 1 Mar 1999 11:55:08 +1030 From: Greg Lehey <grog@lemis.com> To: Edward Ing <inge@home.com> Cc: freebsd-questions@freebsd.org Subject: Re: TCP/IP on LAN drops after a few minutes. Message-ID: <19990301115508.W7279@lemis.com> In-Reply-To: <001801be6f3b$6cc86aa0$4f4b7018@mimico.firstmaple.ca>; from Edward Ing on Mon, Mar 15, 1999 at 06:27:42PM -0500 References: <001801be6f3b$6cc86aa0$4f4b7018@mimico.firstmaple.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
[Format recovered at freebie.lemis.com] On Monday, 15 March 1999 at 18:27:42 -0500, Edward Ing wrote: > On Sunday, February 28, 1999 5:44 PM, Greg Lehey <grog@lemis.com> wrote: > >> On Monday, 15 March 1999 at 9:17:18 -0500, Edward Ing wrote: >>> I have a ethernet lan network with a FreeBSD machine and a dual boot >>> window98/windowsNT workstation machine. The FreeBSD machine has two NICs, >>> one on the LAN and one on cable modem on the internet. >>> >>> Whenever I first boot the FreeBSD machine the TCP/IP is working >>> fine. But after several miniutes the TCP/IP drops. If I reboot >>> FreeBSD, everthing work again for five minutes. >> >> What do you mean by "drops"? What are the symptoms? What version of >> FreeBSD? Please describe what you observe. > > Version: FreeBSD 2.2.8. Using a DEC203 (Etherworks3) card. (le0 device). > Running on a VESA bus 486 with 32M of ram and scsi disks. The network card > is ISA. > I configured FreeBSD to have 2 NICs. One to the Internet and one to my > private (non-registered ip) with ip forwarding turned off. I also ensure > that only TCP/IP was running on the Windows Machines.(i.e. no windows > networking). > > The TCP/IP communications stops working at about 5 minutes every time after > the FreeBSD machine is booted and rebooted. If I try to ftp or ping the > FreeBSD box from WindowsNT or Window98 ping fails (times out) and ftp fails. > From the FreeBSD side if I start pinging the Windows machine and let it > continue right after I reboot, ping is okay at first but within 5 minutes > the pinging will fail and a message will say something to the effect that > "There is no buffer space available" everytime it tries to ping. > > If I start pinging after a period after reboot when I know the TCP/IP is > down ping will tell me that there is a timeout. > > I am pretty sure the problem is on FreeBSD for the following reasons, if I > try WindowsNT or Windows95, I have the problem. Using the same network card > with Linux as the server rather that FreeBSD, I have no problem. If I swap > the DEC 203 with another DEC 203 on FreeBSD, I still get the problem. But > now I have swapped out the DEC and replaced it with a SMC adapter (ed0) and > I have no problems. OK, this is rather pointing towards the DEC board, or the FreeBSD le0 driver. The next thing would be to query the PR database. There I see: > Number: 4292 > Category: kern > Synopsis: le0 (DE203) goes OACTIVE after some time > Confidential: no > Severity: serious > Priority: medium > Responsible: freebsd-bugs > State: closed > Class: sw-bug > Quarter: > Keywords: > Date-Required: > Submitter-Id: current-users > Arrival-Date: Wed Aug 13 03:10:01 PDT 1997 > Closed-Date: > Last-Modified: Sun Oct 12 06:57:35 PDT 1997 > Originator: Olaf Erb > Release: FreeBSD 2.2-STABLE i386 Looking in more detail, >> Description: > > le0 goes OACTIVE after some time, errors like "no buffer space > available" occur. Only downing/deleting and re-configuring the interface > brings it back, for some packets though, then it hangs again. The problem > mainly appears while acting as bootp/tftp server, after some < 100kB of data > transfered. > Have a look at > f [1995/07/04] kern/587 if_le hangs on OACTIVE with 2k buffer > too. > > >> How-To-Repeat: > > use a DE203 with msize 2048 and serve bootp/tftp/nfs over it, or just > large transfers over some time. > >> Fix: > > switched the card with nicsetup.exe to 32768 bytes msize. This isn't > recognised by the driver, because msize is hardcoded there to 2048. > > A quick change in if_le.c to 32768 (lemac_probe()) solved this problem, > but this is no real fix. > Hardcoded value should be replaced by using the probed msize, if it's > possible (this info should be in eeprom, though I don't know how to figure > it out). > > This doesn't solve the problem. It just avoids it. I wish I know > enough about FreeBSD to poke around an figure out what is going on. Yup, it looks as if this problem hasn't really been solved. I'd suggest you put in a new PR, referring to this one, and maybe somebody will fix it this time. Greg -- See complete headers for address, home page and phone numbers finger grog@lemis.com for PGP public key To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990301115508.W7279>