[rescue] OpenBSD on Intel (actually)

[Sheldon T. Hall]

So ... sometimes it boots, sometimes it doesn't.  It boots if I leave it off
for a while, like overnight, but not if it's been on already.

When it hangs, it hangs right after "ne3", but when it boots, there _is_no_
"ne3" if I "find ne" in config.  There is, however, an /etc/hostname.ne3
file ... and when it boots, it runs fine.

ne0 is reported to be on irq 9 if I "find ne" in config
ne1 is reported to be on irq 10 if I "find ne" in config
ne2 is reported to be on irq 9 if I "find ne" in config
ne3 is reported to be on irq 15 when it hangs during the boot process, and
doesn't show up in "find ne" if the machine boots.

There's only one actual NIC in the machine, a "LINKSYS EtherPCI LAN card
II".  When it hangs after the "ne3" line, "ne3" claims to be something like
"LinkSys Winbond" ... and the LinkSys does have a Winbond chip on it.

The irqs actually in use are the regular PC set, plus 5 for the soundcard,
11 for the SCSI controller, and whatever the NIC uses.  An old MS-DOS tool
indicates that something is on irq 9, and I expect that's the NIC.

Question #1:  If I have one NIC, how many "ne" devices should I have in
"find ne"?  Are the names significant?

Question #2:  Does OpenBSD put the stuff on the screen _before_ it probes
for the device, or afterwards?

If it puts it on the screen before, then "ne3" would seem to be the problem.
If, however, it puts it on the screen only after it probes and finds
something, then whatever-is-after-"ne3" would be it, right?

Question #3:  If OpenBSD probes first, then displays after probing, how do I
fingure out what it's looking for after "ne3", i.e. the thing it's actually
hanging on.  Is the order of things at boot time the same as the order of
things in "UKC" or "config -ef"?

FWIW, I've done the "boot -c" trick and disabled "pcibios" ... still hangs.

Clues solicited.

Thanks.

-Shel