Date: Sun, 26 Nov 2000 21:36:28 +0100 (CET) From: =?ISO-8859-1?Q?G=E9rard_Roudier?= <groudier@club-internet.fr> To: John-Mark Gurney <gurney_j@resnet.uoregon.edu> Cc: freebsd-scsi@FreeBSD.ORG Subject: Re: ncr problems (and amd) Message-ID: <Pine.LNX.4.10.10011262115461.669-100000@linux.local> In-Reply-To: <20001126123040.16010@hydrogen.funkthat.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 26 Nov 2000, John-Mark Gurney wrote: > G=E9rard Roudier scribbled this message on Nov 26: > > Some VIA based chipset motherboards can mess-up a lot with your control= ler > > when some feature is enabled on the chipset. > >=20 > > I donnot remember exactly the feature that makes problems (DMA problems= ), > > but it should have something to do with `cache streaming' or `PCI > > streaming'. > >=20 > > Could you look into your mother-board set-up if a feature that resemble= s > > `<something> streaming' is enabled and then disable it. >=20 > there is a burst option I believe.. and I'll try to do that the next May-be it is '<something> bursting' instead, but `streaming' sounds better to me. As I wrote, I forgot the exact feature that breaks PCI DMA with the 53c875 using some VIA chipset based motherboards. > time it crashes... right now the machine doesn't even have a video > card in it.. so, when I put in the shorter cable, I'll do this config > change at the same time... >=20 > but I don't think it's a PCI bus issue, as it looks like the ncr driver If you have time for giving a try with another O/S and/or driver for your chip, this will help better that thinking, in my opinion. > tries to retry, but the ncr is completely out to lunch and needs to be > reset.. is there a way for it to detect excessive timeouts (like no > command completed and all commands have timedout twice), and reinitalize > the card? If you have time for installing 4.X you will be able to use `sym', but this will probably not work better, if as I beleive, the real cause is bad DMA happening between the PCI-SCSI chip and core-logic/memory. G=E9rard. > > On Sun, 26 Nov 2000, John-Mark Gurney wrote: > >=20 > > > G=E9rard Roudier scribbled this message on Nov 26: > > > > I read your 2 messages, and the problems reported by the `ncr' driv= er > > > > seems to me rather due to flaws in the hardware than to the `ncr' i= tself. > > > >=20 > > > > Could you describe the hardware you are using and what you are doin= g with > > > > it. You also may check that it is properly configured in all its pa= rts. I > > > > mean, clockings, cabling, cooling, etc ... > > >=20 > > > ok... well, the hardware is a Tekram DC-390F. Attached hardware is: > > > <FUJITSU M2949Q-512 0155> at scbus0 target 0 lun 0 (pass0,da= 0) > > > <MICROP 4743 S150> at scbus0 target 1 lun 0 (pass1,da= 1) > > > <MICROP 4743 S150> at scbus0 target 2 lun 0 (pass2,da= 4) > > > <SAMSUNG WN32162U 0100> at scbus0 target 4 lun 0 (pass3,da= 2) > > >=20 > > > All are narrow except the Fujitsu. All are internal, using no specia= l > > > addapters. Termination is turned on on both drives at the end of the > > > chain. Motherboard is a FICA VA-503+ VIA based chipset that is *NOT* > > > overclocked. As for cooling, nothing special, but the case is closed > > > and has two exhaust fans and one intake fan. When I upgraded from th= e > > > Advansys Ultra SCSI card, I didn't change internal cabling for the > > > three narrow devices (and I'd been running that for 9+ months). > > >=20 > > > The cable length on the narror part of the system might be a bit long= , > > > so I plan on next crash replacing it with a cable that only has three > > > connectors for drives instead of the current five. Though the Tekram > > > manual lists a length of 3 meters for 4 devices or less, which I know > > > I'm less than, for the 1.5 meters (max of 8 devices) I'm probably pre= tty > > > close to. > > >=20 > > > Well, it also survived daily, but I did reduce the tags all above > > > mentioned drives to 16, which may have made a difference. But only > > > 16 tags is so few. :( > > >=20 > > > Thanks for any help you can provide. >=20 > --=20 > John-Mark Gurney=09=09=09=09Voice: +1 408 975 9651 > Cu Networking > =09"Thank God I'm an atheist, that'd just be confusing." -- cmc To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.10.10011262115461.669-100000>