cm@jet.uucp (colin manning) (04/30/91)
I've just got a Intel 403e (33Mhz 486) with 16Mb RAM, 1740 EISA SCSI and a Fujitsu 1.2Gb SCSI drive (model 2266sa I believe). Q1: Machine seems to work fine, but for the following *very* serious problem. If I do something that is very disk intensive, I can get the SCSI bus/drive/controller hung, and all I can do is reboot. The disk access light stays on permanently (suggesting system is midway through a SCSI command ?), and the all Unix activity ceases. The kernel does not appear to have crashed, since it will still echo keyboard input to the console. But thats all it will do. You cant interrupt the cp or whatever caused the problem. I can make this happen easily by copying or uncompressing a 10Mb file. The 1740 is configured in its standard 1540 compatibility mode, with synchronous SCSI enabled, parity enabled, and disconnect reconnect enabled. I have configured the kernel's 'high performance disk driver' according to the documentation, using all the default values. System also has a Wangtek 150Mb SCSI tape streamer, which is not in use when the problem occurs. Any ideas ? Q2: The Fujitsu 2266 has 1029 cylinders. Although the fdisk table only caters for disks with up to 1024 cylinders, ISC's documentation says that it will use the additional cylinders, if you try to allocate them when partitioning the disk during installation. I had a mail msg some time back from someone who said that this does not work, and furthermore that by attempting to use cylinders >1024, the system will wrap around to cylinder 0 and clobber the start of your disk (which would be *serious*). Anyone else experienced such problems ? It would seem unlikely to me that ISC would state that such a facility works if there is a bug of this order of magnitude in it. Many thanks in advance, -- - Colin Manning, cm%jet.uucp@ukc.ac.uk (world) OR cm@jet.uucp (UK only) - Disclaimer: Please note that the above is a personal view and should not be construed as an official comment from the JET project.
cpcahil@virtech.uucp (Conor P. Cahill) (05/01/91)
cm@jet.uucp (colin manning) writes: >The Fujitsu 2266 has 1029 cylinders. Although the fdisk table only >caters for disks with up to 1024 cylinders, ISC's documentation says >that it will use the additional cylinders, if you try to allocate >them when partitioning the disk during installation. I had a mail >msg some time back from someone who said that this does not work, and >furthermore that by attempting to use cylinders >1024, the system will wrap >around to cylinder 0 and clobber the start of your disk (which would >be *serious*). The wrap-around problem is not related to 1024 cylinders. It is some function of disk size and as far as I know only happens on 1.2GB or higher drives. Last I heard ISC knew about the problem and was working on it, but that was months ago. Perhaps it has been solved by now. >Anyone else experienced such problems ? It would seem unlikely to me >that ISC would state that such a facility works if there is a bug of >this order of magnitude in it. The facility for using more than 1024 cyls does work. We have used it several times with no problems. Our 680MB drives have 1543 cyls. -- Conor P. Cahill (703)430-9247 Virtual Technologies, Inc. uunet!virtech!cpcahil 46030 Manekin Plaza, Suite 160 Sterling, VA 22170
pim@cti-software.nl (Pim Zandbergen) (05/02/91)
cm@jet.uucp (colin manning) writes: >I've just got a Intel 403e (33Mhz 486) with 16Mb RAM, 1740 EISA SCSI >and a Fujitsu 1.2Gb SCSI drive (model 2266sa I believe). I've just got a Intel 403e (33Mhz 486) with 8Mb RAM, 1740 EISA SCSI and a HP 680Mb SCSI drive (model 97548su I believe). >Q1: >Machine seems to work fine, but for the following *very* serious problem. >If I do something that is very disk intensive, I can get the SCSI >bus/drive/controller hung, and all I can do is reboot. I do not have this problem. >System also has a Wangtek 150Mb SCSI tape streamer, which is >not in use when the problem occurs. I do have a problem that Colin obviously has not. If I have a SCSI tape unit on the SCSI bus AND the SCSI tape driver is configured, the kernel will panic while booting. I have tried Archive VIPER 2150s and 2060s and WangDAT drives, without any difference. These tape units are recognized by and can be erased by Roy Neese's utilities. According to our Intel representative, this is a known problem and ISC should have a fix. Nor Intel, nor our ISC dealer could acknowledge whether this is the SSU.7 fix that I am awaiting. If it is, I'll let the net know. -- Pim Zandbergen domain : pim@cti-software.nl CTI Software BV uucp : uunet!mcsun!hp4nl!ctisbv!pim Laan Copes van Cattenburch 70 phone : +31 70 3542302 2585 GD The Hague, The Netherlands fax : +31 70 3512837
tmh@prosun.first.gmd.de (Thomas Hoberg) (05/08/91)
In article <1991May01.114725.14859@virtech.uucp>, cpcahil@virtech.uucp (Conor P. Cahill) writes: |> cm@jet.uucp (colin manning) writes: |> >The Fujitsu 2266 has 1029 cylinders. Although the fdisk table only |> >caters for disks with up to 1024 cylinders, ISC's documentation says |> >that it will use the additional cylinders, if you try to allocate |> >them when partitioning the disk during installation. I had a mail |> >msg some time back from someone who said that this does not work, and |> >furthermore that by attempting to use cylinders >1024, the system will wrap |> >around to cylinder 0 and clobber the start of your disk (which would |> >be *serious*). |> |> The wrap-around problem is not related to 1024 cylinders. It is some |> function of disk size and as far as I know only happens on 1.2GB or |> higher drives. Last I heard ISC knew about the problem and was working |> on it, but that was months ago. Perhaps it has been solved by now. [...] When using the Adaptec UNIX and any other OS care rather little about the physical layout of the disk. The Adaptec will remap any SCSI drive to 32 sectors, 64 heads and as many cylinders as there is megabytes on the disk. This comes rather handy when you are asked by 'fdisk'just how many cylinders you want to allocate for a partition. This induces some latency due to the Adaptecs 8085 having to do some calculations, but with SCSI's defect management a sector might not end up where you believe it should, anyway. With my 1.2GB I am still 'lucky', because it only has about 970MB after formatting, meaning 970 cylinders. Once you got a true 1.2GB drive you are in for trouble, because soft cylinder numbers will actually exceed 1024. On the other hand huge ESDI drive with >1024 cylinders can be made to work with ISC, so it's probably not a big deal to fix it. Until then you will just have to forego some of your disks capacity. -- tom ---- Thomas M. Hoberg | UUCP: tmh@bigfoot.first.gmd.de or tmh%gmdtub@tub.UUCP c/o GMD Berlin | ...!unido!tub!gmdtub!tmh (Europe) or D-1000 Berlin 12 | ...!unido!tub!tmh Hardenbergplatz 2 | ...!pyramid!tub!tmh (World) Germany | BITNET: tmh%DB0TUI6.BITNET@DB0TUI11 or +49-30-254 99 160 | tmh@tub.BITNET