home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sgi
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!mips!odin!sgihub!zola!zuni!anchor!olson
- From: olson@anchor.esd.sgi.com (Dave Olson)
- Subject: Re: ec0: too many oflows, changing modes?
- Message-ID: <or2pgqs@zuni.esd.sgi.com>
- Sender: news@zuni.esd.sgi.com (Net News)
- Organization: Silicon Graphics, Inc. Mountain View, CA
- References: <1992Aug20.091238.17149@news.uni-stuttgart.de>
- Date: Fri, 21 Aug 92 02:07:29 GMT
- Lines: 51
-
- In <1992Aug20.091238.17149@news.uni-stuttgart.de> andreas@awstar.rus.uni-stuttgart.de (Andreas Wierse) writes:
-
- | Hi,
- |
- | when I did "ifconfig debug" the following entries appeared in the SYSLOG:
- |
- | Aug 20 02:43:48 awstar unix: ec0: too many oflows, changing modes
- | Aug 20 03:48:13 awstar unix: ec0: dma stopped, rstat = 0x11, crbdp = 0x8015FAE0
- | Aug 20 04:49:18 awstar last message repeated 2 times
- | Aug 20 04:49:18 awstar unix: ec0: too many oflows, changing modes
- | Aug 20 09:21:42 awstar unix: ec0: dma stopped, rstat = 0x11, crbdp = 0x8015FB40
- |
- | Can anyone explain what these messages mean? I didn't find anything in
- | the manual pages that could help me.
-
- They are debug messages, not surprisingly. The interesting one is the
- 'too many oflows' message. Basicly, we are working around a chip bug
- by putting the ethernet chip into promiscuous mode, and leaving in that
- mode for a while. Some nets run into this a lot, others don't. When
- the mode switches, and how long it stays that way can be controlled
- by the variables in /usr/sysgen/master.d/if_ec2.
-
- Later versions of the ethernet chip have some changes to remove this
- problem. There are no real problems associated with running the
- chip in promiscuous mode, aside from more interrupts, which have
- negligible impact on performance (the chip is run in promiscuous mode
- anyway, if you use the network monitoring prodcut).
-
- | Additionally we have a problem with NFS in that it transfers data
- | (write only) in a kind of burst mode. This is like: two or three
- | hundred KB are transferred, then there are a few seconds delay, then
- | the next some hundred KB are transfered and so on. Watching this with
- | nfsstat shows that everytime when it stops for a while afterwards
- | timeout and retrans are incremented by one. Maybe the reason for the
- | timeout has something to do with the ifconfig error messages. (they
- | don't coincide in time).
-
- It is possible that they are related, but more likely you are having
- physical network problems resulting in dropped packets. It is surprising
- how significant an impact a few dropped packets will have on NFS performance.
- One of the symptoms of the problem that the promiscuous mode workaround
- fixes was in fact dropped packets, but it doesn't sound like they match
- up in your case. You could reconfigure your kernel to remain in
- promiscuous mode for longer periods once entered, and see if that has
- an effect on your performance; you could also look around for other
- network problems.
-
- --
- Let no one tell me that silence gives consent, | Dave Olson
- because whoever is silent dissents. | Silicon Graphics, Inc.
- Maria Isabel Barreno | olson@sgi.com
-