From owner-freebsd-stable Thu Jan 25 11:53:25 2001 Delivered-To: freebsd-stable@freebsd.org Received: from mail2.wi.rr.com (fe2.rdc-kc.rr.com [24.94.163.49]) by hub.freebsd.org (Postfix) with ESMTP id 22BFA37B400 for ; Thu, 25 Jan 2001 11:53:02 -0800 (PST) Received: from woodstock.monkey.net ([24.167.246.123]) by mail2.wi.rr.com with Microsoft SMTPSVC(5.5.1877.357.35); Thu, 25 Jan 2001 13:50:35 -0600 Received: from pobox.com (localhost [127.0.0.1]) by woodstock.monkey.net (Postfix) with ESMTP id 67841201; Thu, 25 Jan 2001 13:52:56 -0600 (CST) X-Mailer: exmh version 2.3.1 01/23/2001 with version: MH 6.8.4 #1[UCI] To: Stephen Montgomery-Smith Cc: Luc Morin , freebsd-stable@FreeBSD.ORG Subject: Re: Network stops working In-reply-to: Your message of "Thu, 25 Jan 2001 12:02:28 CST." <3A706A34.E5F1C502@math.missouri.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Thu, 25 Jan 2001 13:52:56 -0600 From: Jon Hamilton Message-Id: <20010125195256.67841201@woodstock.monkey.net> Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG In message <3A706A34.E5F1C502@math.missouri.edu>, Stephen Montgomery-Smith wrot e: [ responding to a message from Luc Morin, who mentioned that in a recent-ish -STABLE, his network would stop working after a while ] } That's interesting - I have a similar problem that also } just started recently (coincided with changing my power } supply but I'm sure that it is unrelated). I find that } one of three things happens: } } 1. dhcp simply does not successfully negotiate } } 2. dhcp negotiates, gets an IP number and a router, but } then I get 100% packet loss. (Never tried calling them to } see if they got through to me.) } } 3. It actually works. } } Option 2 is what usually happens, but if I restarting } dhcp enough times, I get option 3. Since I keep the } connection up all the time, I only worry about the } problem sporadically. So I have never tried to investigate } its cause. (The ISP is att@home if that is useful. My } NIC is a RealTek 8139.) I've recently had problems too, that don't exactly match any of what I've seen in this thread, but which "feel" similar, and may be related. The background: I have a Compaq 1200 laptop, which travels between two or three netoworks. At home, it's behind a NAT/firewall connection, and at work it's on one of two LANs. The laptop uses DHCP to get its IP address in all situations. With -STABLE from 12/31/00, everything is fine. When I went to -STABLE from 1/24/01, I saw the following behavior: At home: gets IP address from DHCP server fine. Can connect to various other machines on the subnet ok. Can ssh to various machines both on the home LAN and outside the NATted subnet, but cannot connect from other machines to the laptop. Packets come in, but for some reason (according to both the ipfw logs and tcpdump) the kernel on the laptop doesn't seem to respond with outgoing packets at all. The firewall is not blocking the packets; they don't even get that far (tested by logging all traffic to/from all interfaces on the laptop). This bizarreness holds for TCP and ICMP traffic. The odd thing is, for "outgoing" connections (ssh from laptop to server) everything is fine; traffic moves just fine both directions. At work (two different subnets, but similar configuration between them): can get IP address via DHCP (it's not falling back to its last lease, but is actually getting a response from the DHCP server), but dhclient can't ping the gateway. Configuring everything by hand, I still can't get traffic to go out. I also get lots of complaints about corrupt NIC memory, with odd packet sizes ranging from 1 to about the MTU. I've tested this with two different Netgear FA-410tx cards (which I've been using for about 6 months, though these _are_ the cards that require fa_select.c to set them up). I did try without fa_select hoping that there were some changes to the driver that made it unnecessary, but that wasn't the case. This card uses the ed driver; unfortunately I don't have any PCMCIA cards I can use to test other drivers. I've backed down to a src tree from 12/31/00 and everything is once again working normally. I wasn't able to keep around the bogus tree, though I can check it out again and poke at it on an unused partition if it'll help. I did blow away /usr/src and /usr/obj when the problems first started and rebuilt from scratch, with the same results. Even GENERIC had the same symptoms. I did run mergemaster after updating. My NIC is configured for: ed0 at port 0x240-0x25f irq 3 slot 0 on pccard0 One last bit of info: the loopback driver does respond normally, for whatever that's worth. I can ping localhost and tcpdump -i lo0 will show both the incoming and outgoing packets. -- Jon Hamilton hamilton@pobox.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message