Date: Fri, 26 Sep 1997 13:36:42 -0700 From: pius@ienet.com To: dg@root.com Cc: hackers@FreeBSD.ORG, terryl@ienet.com Subject: fxp driver/Intel EtherExpress Pro 10/100B problem (was Re: your mail) Message-ID: <199709262036.NAA09474@iago.ienet.com>
next in thread | raw e-mail | index | archive | help
On Fri, 26 Sep 1997, dg@root.com wrote: >>I'm using 3 EtherExpress cards: >> >>fxp0 <Intel EtherExpress Pro 10/100B Ethernet> rev 2 int a irq 10 on pci0:10 >>fxp0: Ethernet address 00:a0:c9:55:13:22 >>fxp1 <Intel EtherExpress Pro 10/100B Ethernet> rev 1 int a irq 9 on pci0:12 >>fxp1: Ethernet address 00:a0:c9:10:68:a0, 10Mbps >>fxp2 <Intel EtherExpress Pro 10/100B Ethernet> rev 1 int a irq 9 on pci0:14 >>fxp2: Ethernet address 00:a0:c9:5a:51:f9, 10Mbps >> >>Sometimes, ~ 1/2 per week, my fxp0, which is revision 2 card, silently >>freezes. >>When I `ifconfig down delete fxp0' and then ifconfig it up, the situation >>resolves. > > If this is true, then Intel has apparantly lied to me about fixing the >problem in the "next stepping" of the chip. The lock up should only occur >when garbage bits occur in the preamble of a packet. It's not supposed to >happen in 100Mbps mode (only 10Mbps), but since I've personnally been able >to get it to happen in 100Mbps, I don't believe it. I can reproduce the >problem only under special circumstance like power-cycling the hub. > You might try using the "link0 link1 -link2" flags on the 100Mbps link >to force it to stay in 100/half...this might prevent the lock up from >occuring (but I've not tested this myself). For what it's worth, we also saw problems with the Intel EtherExpress Pro 10/100B card (rev 2) under not so good network conditions. We're using 4 such cards on a router: fxp0 <Intel EtherExpress Pro 10/100B Ethernet> rev 2 int a irq 9 on pci0:9 fxp0: Ethernet address 00:a0:c9:5f:f8:f9 fxp1 <Intel EtherExpress Pro 10/100B Ethernet> rev 2 int a irq 12 on pci0:10 fxp1: Ethernet address 00:a0:c9:5f:f9:0a fxp2 <Intel EtherExpress Pro 10/100B Ethernet> rev 2 int a irq 10 on pci0:11 fxp2: Ethernet address 00:a0:c9:6c:78:6d fxp3 <Intel EtherExpress Pro 10/100B Ethernet> rev 2 int a irq 11 on pci0:12 fxp3: Ethernet address 00:a0:c9:6c:a6:e7 fxp3 (running in 10Mbps) was connected on the other end to a cheap hub in a room 11 floors below our machine. So, it's a long haul and the wiring at certain places may be a little sketchy. The connection would freeze-up at least every two days (sometimes more frequently) and we'd get messages like the following in /var/log/messages: Sep 8 11:41:04 core5 routed[83]: interface fxp3 to 206.253.23.1 bad: in=0 ierr=511 out=0 oerr=0 Sep 8 11:41:39 core5 routed[83]: interface fxp3 to 206.253.23.1 restored These seem to indicate garbage on the connection. We don't have control over the hub - so I guess it's possible that the hub was power-cycled around the times that the connection froze. We then added a 5th card to the machine (running the FreeBSD 2.2-970618 snapshot, by the way) - an SMC card (ed0 driver) - and switched over to using that card for the connection that was freezing up. We haven't had any problems since we switched to the SMC card. Best regards, Pius
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199709262036.NAA09474>