From owner-freebsd-questions Fri Jul 30 14: 3: 2 1999 Delivered-To: freebsd-questions@freebsd.org Received: from gloworm.Stanford.EDU (gloworm.Stanford.EDU [171.64.96.52]) by hub.freebsd.org (Postfix) with ESMTP id 28A6914E6F for ; Fri, 30 Jul 1999 14:02:59 -0700 (PDT) (envelope-from yergeau@gloworm.Stanford.EDU) Received: from localhost (yergeau@localhost) by gloworm.Stanford.EDU (8.8.8/8.8.8) with ESMTP id OAA27981; Fri, 30 Jul 1999 14:01:05 -0700 (PDT) Message-Id: <199907302101.OAA27981@gloworm.Stanford.EDU> To: freebsd-questions@freebsd.org Cc: yergeau@gloworm.Stanford.EDU Subject: problems with Netgear FA310-TX (rev D2) Reply-To: yergeau@gloworm.Stanford.EDU Date: Fri, 30 Jul 1999 14:01:04 -0700 From: Dan Yergeau Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I just dropped a Netgear FA310-TX (rev D2 w/the 82c169 PNIC chip) into my machine. It is recognized by the pn driver and negotiates 100Mbps-FD fine, but performance is *extremely* poor under FreeBSD (tcp seems to stall due to lost[?] packets). The interface functions well under NT with the Netgear driver (i.e. I'm definitely seeing 100Mbps-level performance; same machine, network port, etc.), so it doesn't look like a problem with the card, the switch, or the combo of the card and switch. If it matters, according to our LNA, the other end of the cable is plugged into a 3COM 3300. "netstat -I pn0 -w 1" showed that some errs were being encountered when trying to transfer data in/out, so I modified the pn driver to print out the type of the error (instead of just incrementing if_{i,o}errors). The errors encountered are: PN_RXSTAT_RXERR PN_ISR_TX_UNDERRUN PN_TXSTAT_ERRSUM The PN_ISR_TX_UNDERRUN and PN_TXSTAT_ERRSUM always occur in tandem (and ERRSUM is not due to late or excessive collisions; it is probably due to the underrun). I've tried the various media and mediaopts supported by the pn driver: 100-FD - unusable (any significant size TCP transfer in either direction stalls) 100-HD - unusable (any significant size TCP transfer in either direction stalls). I don't see any PN_RXSTAT_RXERR's. I also don't see any collisions (I'm assuming that "pacing" collisions for HD would be reported by netstat). 10-FD - TCP data into(*) the NIC peaks around 500KB/s with frequent PN_RXSTAT_RXERR's. Data out of the NIC peaks around 15KB/s, but I don't see any TX or RX diagnostics, and data is definitely not getting out very quickly. 10-HD - seems to work perfectly (1MB/s in either direction). No errors. (*) "into" means from a remote machine into the NIC on the FreeBSD host Some of the above seems to indicate that FD flow control may not be working on that port, but with a Sun Ultra 10 (hme) plugged into the same port on the 3COM 3300 with the same cable, FD works just fine. I don't know whether NT is using FD or HD. Boot and if_pn.c version information are below. CPU: Pentium Pro (199.43-MHz 686-class CPU) Origin = "GenuineIntel" Id = 0x616 Stepping=6 Features=0xf9ff real memory = 33554432 (32768K bytes) avail memory = 30150656 (29444K bytes) Probing for devices on PCI bus 0: chip0: rev 0x02 on pci0.2.0 ahc0: rev 0x00 int a irq 3 on pci0.6.0 ahc0: aic7880 Single Channel A, SCSI Id=7, 16/255 SCBs pn0: <82c169 PNIC 10/100BaseTX> rev 0x20 int a irq 9 on pci0.12.0 pn0: Ethernet address: 00:a0:cc:41:8a:95 pn0: autoneg complete, link status good (full-duplex, 100Mbps) vga0: rev 0x01 int a irq 11 on pci0.14.0 chip1: rev 0x04 on pci0.20.0 chip2: rev 0x04 on pci0.25.0 ... [full "-v" boot at http://www-tcad.stanford.edu/~yergeau/fbsd/boot-v.txt] if_pn.c is from 3.2-RELEASE * $Id: if_pn.c,v 1.6.2.10 1999/04/14 19:44:53 wpaul Exp $ Any ideas? Any successes or other horror stories with revision D2 of this NIC? Cheers, Dan To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message