Promise Ultra100 IDE card(s) and Raid-5
Jeff Folts
I have not seen any specific problems with the two Promise Ultra100 controllers working together, properly configured with one drive per channel. I can access the drives with fdisk and hdparm with no apparent problems. I did not try to create a file system on a non raid partition, though.
With test12, I can create raid 1 or raid 5 arrays and the resync processes do their thing in the background. As soon as I try to create a file system on an array, I get a kernel BUG() call in buffer.c at line 765 I think. The function is called something like end_buffer_io_bad().
With test11, I can create raid 5, create a file system, but then as soon as I try to copy any files to file system, the kernel crashes somewhere other than a BUG() call. I saw other problems with test11 and raid that I don't recall specifically.
Hardware Configuration:
SuperMicro 370DLE motherboard
2 Promise Ultra100 controllers
4 IBM DTLA 45G Ultra100 drives
Jim Deas wrote:
> I am testing the new 2.4.0-test12 kernel as a I686 build. When installing
> more that one Ultra100 IDE controller card (I am building a raid-5), the
> kernel halts as it tries to read the second two HD's (hdi1 and hdk1). The
> four drives are all setup as masters on there own cables.
> The bios shows the drives correctly and even linux sees the cards on the
> first check.
> I have been able to see all the drives if I set them up on the first
> Ultra100 card as master/slave combinations but I am sure this would cause
> the raid to run very slow.
>
> On a second front. When trying to configure a Raid-5 on these drives (as
> master/slave x2), I get a system that seems to be created fine. Even
> /proc/mdstat shows the raid as existing and rebuilding.
> When I try to run 'mkfs' on the raid I get a kernel crash that siezes the
> system with no way out but a hardware reset.
>
> Any help? suggestions? newbee bashing?
> Regards,
> J. Deas