It could happen to you (was Re: network cards and POST failures)
at last, the strange problems I was having with my Linux box at reboot time
is solved. I followed Donald Becker's recommendation about putting
"/sbin/ifconfig eth0 down" at the very end of the reboot script and this
fixes everything. Now everything is crystal-clear to me.
I agree with Bruce that the Debian distribution should take care of shutting
down Ethernet devices before rebooting.
As an example, I was getting this error in my desktop machine everytime I
rebooted the Linux box: "eth0: Bus master arbitration failure, status 88f2"
Since my desktop machine is not a production machine (I am the only one that
uses it) and that I first saw this error after a kernel upgrade, I did not
waste much time finding the cause. Each time I got that error I just cycled
power to the computer and everything worked fine after that. Well, the
Ethernet card here is also a NE2100 compatible (it's a BocaLAN Card, using
the lance.c driver) and it does DMA. After I put "/sbin/ifconfig eth0 down"
at the bottom of /etc/init.d/reboot the problem went away.
Beware!!! This problem could happen to you and you'll waste hours trying to
find what's wrong.
P.S. Somebody may want to post this to firstname.lastname@example.org because
I have no posting priviledges there.
At 12:46 PM 12/2/96 PST, Bruce Perens wrote:
>Here's what Donald Becker had to say about Eloy Paris' problem in which
>a network card caused RAM POST failures. This is perhaps something that
>we should take care of in the shutdown scripts after NFS partitions are
>unmounted and daemons are killed. We can use the output of "netstat -i"
>to figure out what interfaces to shut down.
>From: Donald Becker <email@example.com>
>On Sat, 30 Nov 1996, Bruce Perens wrote:
>> Eloy Paris <Eloy.Paris@ven.ra.rockwell.edu> has a DEC PC that fails
>> the RAM POST only when the Boca PCnet32 card is present and a recent
>> Linux kernel has just shut down. My uninformed speculation is that the
>> card could be doing wild DMA, and I'm wondering if there's any special
>> step that should be taken to shut down this card before rebooting.
>That's a common BIOS bug.
>The warm boot code should disable the bus-master capability on all cards.
>What is happening is that the card is continuing to receive and store
>packets into memory. The work-around, which we must use here as well, is to
> /sbin/ifconfig eth0 down
> /sbin/ifconfig eth1 down
> /sbin/ifconfig eth2 down
> /sbin/ifconfig eth3 down
>in you shutdown script. (Usually /etc/rc.d/rc.6, but Redhat is different.)
>This will disable the cards.
Eloy A. Paris
Rockwell Automation de Venezuela
Telephone: +58-2-9432311 Fax: +58-2-9430323
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-REQUEST@lists.debian.org . Trouble? e-mail to Bruce@Pixar.com