Date: Fri, 6 Oct 2000 13:22:49 -0400 (EDT) From: Kenneth W Cochran <kwc@world.std.com> To: John Reynolds~ <jreynold@sedona.ch.intel.com> Cc: freebsd-stable@freebsd.org Subject: Re: breakage with two ed network devices Message-ID: <200010061722.NAA13450@world.std.com> References: <14812.58143.609625.133015@hip186.ch.intel.com> <A0E035400B00D4118F9E0008C70D4D77A88A@ITC1> <200010051637.KAA51557@harmony.village.org> <200010060408.WAA05189@harmony.village.org> <200010061618.MAA07034@world.std.com>
next in thread | previous in thread | raw e-mail | index | archive | help
>From owner-freebsd-stable@freebsd.org Fri Oct 6 12:39:56 2000 >From: John Reynolds~ <jreynold@sedona.ch.intel.com> >Date: Fri, 6 Oct 2000 09:35:51 -0700 (MST) >To: Kenneth W Cochran <kwc@world.std.com> >Cc: freebsd-stable@freebsd.org >Subject: Re: breakage with two ed network devices > >[ On Friday, October 6, Kenneth W Cochran wrote: ] >> Please pardon my "jumping in" and/or ignorance... >> >> ed0 living on an IRQ that is "reserved" (somehow) for one >> of the ATA "channels?" (ie. 14 and/or 15?) > >Hmmmm. Yeah, I've got ed0 living on IRQ15. However, this h/w >combo has been running since 3.3 (including 4.0->4.1-RELEASE). I wonder if its "previous" running/probing might have been a "glitch" (not as severe as a "bug" :) with prior drivers and that now, this "glitch" is being "exposed" by the "new" stuff(?) >> Hmmm... I might "question" an ATA-probe "there..." (?) >> >> Can this card (ed0) go to a different IRQ? Is that IRQ >> disabled/reassigned (from PCI) in the BIOS/setup? > >No, it's hardwired in the card. I read on the archives when >I bought it to disable the PnP stuff on the card because it >wouldn't work without it. So, I assigned it IRQs manually >through their stupid little DOG program. 15 and 9 were the >only ones their setup program could find that weren't >"conflicts" with something else. > >I can certainly try to move ed0 onto a different IRQ. Maybe remove something & change the IRQ just for testing? Is IRQ 15 BIOS-disabled (therefore making it available for things like ISA cards). Might be useful datapoints... >> I might want to "move" this one, too; IRQ 9 is the "shared" >> one & it has always "frightened" me some... :) >> >> (Brain-cobweb-digging) I also notice that that the "iomem" >> is the same; shouldn't those be different segments? > >Probably so. Any suggestions for the second segment's >starting position? LINT says nothing about it. Hmmm... These cards appear to be "shared-memory" cards. 1. IIRC (deep cobwebs now :), shared memory should not (cannot?) be cached, thus the possible need to "enable" the "memory hole" at 15M (in BIOS setup). IIRC (again :) this makes a 1mb non-cacheable "region" starting at 15M. 2. We would need to know the "shared memory segment size" of the cards & how to set their (starting) addresses (any card -doc?); (conceivably) these would be values for "iomem" in your kernel-config. I couldn't find anything about "iomem" either, but that would be my guess. Hopefully Someone Who Knows will answer here (& maybe document that parameter? :) (I see Warner replying... ;) >> Hope I was at least slightly helpful... > >yeah, it was at least enlightening to see that IRQ14/IRQ15 >are "meant" for ATA. That certainly does look like a >smoking gun. However, I bring up the canonical fact that >"it worked before this commit" .... As I mentioned above, its "working before" miiiiight have been a "glitch" (or maybe a "boo-boo" :). But I'd think that would be "cleared" if you "turn off" IRQ 15 in your BIOS. Fwiw, I run 4.1.1-s on a box so configured, & IRQ 15 gets used "elsewhere" (the (PCI) NIC at this time). >Hopefully I'll have time over the weekend to futz with the >IRQs on these cards. Maybe I'll just ditch the damn things >and go get two PCI NICs ... who knows ... I believe PCI NICs would be The Cure... My experience with (especially PnP) ISA cards on modern PCI systems leads me to conclude that (ISA/PnP) cards are Truly Evil(tm)... >Thanks, >-Jr I'm curious as to how it turns out... -kc 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?200010061722.NAA13450>