"Benson, Chris M" wrote:
>
> In my quest to set up a simple embedded software base that I could flash to
> the DOC (which I am having problems with, but plugging along with the image
> with expectation that this will get solved) I have run into an interesting
> item working with a 16 MB (SanDisk) Compact Flash interfaced through the IDE
> chain. Under a normal BIOS the CF comes up and I am able to fdisk, format
> and access the disk. If I then switch over to the Linux BIOS expecting it to
> load from /dev/hda1 which is now the compact flash I get a kernel panic. I
> move the CF to /dev/hdc1 and re-install HD as hda1 and the system comes up,
> but the partition type of the CF is unknown. If I use a blank CF and try to
> fdisk it under the LinuxBIOS when I try to write the partition table to the
> part the system locks up while trying to call the ioctl to re-synch the
> drive. (Note - under a normal BIOS the CF loads just fine with the bootdisk
> and CF as hda1.)
>
Are you using SiS 630 ?? There is something (bugs??) in the compability
of
CF and SiS 630. You have disable PCI prefetch for the IDE controller or
the CF will never been detected. I think it is because CF is SLOW and
support only PIO modes.
> Is there something special done by the normal PC BIOS that is not being done
> at start-up of the LinuxBIOS? I can only think that some part is not being
> initialized or something of the like, because the part is valid and the same
> kernel appears to work fine with the part when loading from a floppy.
>
I think normal BIOS tries every DMA/PIO etc combination on the POST.
So it can detect the CF.
Ollie
