Date: Mon, 7 Sep 1998 11:47:15 -0400 (EDT) From: "Viren R. Shah" <viren@rstcorp.com> To: Bill Paul <wpaul@skynet.ctr.columbia.edu> Cc: freebsd-stable@FreeBSD.ORG Subject: Re: auto-negotiation in xl driver takes too long? Message-ID: <199809071547.LAA01012@jabberwock.rstcorp.com> In-Reply-To: <199809071509.LAA19002@skynet.ctr.columbia.edu> References: <199809071353.JAA00940@jabberwock.rstcorp.com> <199809071509.LAA19002@skynet.ctr.columbia.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> "Bill" == Bill Paul <wpaul@skynet.ctr.columbia.edu> writes:
Bill> Of all the gin joints in all the towns in all the world, Viren R. Shah
Bill> had to walk into mine and say:
Bill> It should not be from memory: it should be written down on paper and
Bill> transcribed exactly. I have no idea what this error message means or
Bill> where it comes from. It's possible you remembered it wrong and
Bill> it actually said something different which might have been
Bill> useful, but now we'll never know.
Here's the exact boot message. It's easily reproducible, by removing
the "sleep 5" from /etc/rc
*Network pass 1 done.
*Starting the mount of all NFS filesystems.
xl0: autoneg not complete, no carrier
xl1: autoneg complete, link status good (half-duplex, 10Mbps)
[hangs for a long time]
nfs: can't get net id for host
*Finished mounting all NFS filesystems.
recording kernel -c changes
additional daemons: syslogd.
The statements prefixed with an * are the ones I introduced into the
rc files.
When it does continue after being hung during the NFS mount, all of my
NFS filesystems are not mounted (although most of them are). In this
state when I do a mount -a -t nfs again in order to mount the ones
that haven't been mounted, all the NFS mounts that were already
mounted get mounted again (as in they are listed twice in the df
output).
Bill> The autonegotiation doesn't happen sequentially. That is, it takes 5
Aah.
Bill> I'm not quite certain how to deal with this. If I reduce the timeout,
Bill> it will break autonegotiation with some of the 3c905 boards (which is
Bill> what I increased the timeout to fix in the first place). NFS mounts are
Bill> done via RPC, and the RPC library is supposed to retry a couple times
Bill> before giving up. I would have expected it to succeed eventually.
It does seem to retry and succeed for *some* of the NFS mounts.
[options]
1. add "bg" to fstab
2. force the media type
3. change the code
Like I mentioned the first time, I'm perfectly happy with putting a
sleep statement in the rc file. If this is going to affect only a few
people (namely me), then that is fine. However, if you want me to test
the above 3 options to see whether they work, and give you feedback,
let me know, and I'll do so. This seems to be more the fault of the
NFS/RPC subsystem than of the NIC driver.
Bill> -Bill
Viren
--
Viren R. Shah, {viren @ rstcorp . com}
"SCSI *is* God's bus and it's hardly an old-fashioned (or academic)
attitude to think so, it's simply an informed attitude."
-- Jordan Hubbard
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199809071547.LAA01012>