Trouble getting Compact Flash IDE Interface to PPC440GP working

[Gregg Nemas]

I have a PPC440GP-based CPU board that has a compact flash connected
to the peripheral bus in so-called "True IDE" mode. I am able to read
and write to the device using u-boot, and now I am trying to get it
working with linux 2.6.

The connection is like this:

CF A0..A2 => PPC440 A30..A28 (ppc A31 not used for 16-bit bus)
CF D0..D15 => PPC440 D15..D0
CF INTRQ => PPC440 IRQ0 (GPIO0)

The peripheral bus is configured for 16-bit wide accesses.

I set the offsets passed to ide_setup_ports as follows:

#define CF_HD_DATA       0x00
#define CF_HD_ERROR      0x03            /* see err-bits */
#define CF_HD_NSECTOR    0x05            /* nr of sectors to read/write */
#define CF_HD_SECTOR     0x07            /* starting sector */
#define CF_HD_LCYL       0x09            /* starting cylinder */
#define CF_HD_HCYL       0x0b            /* high byte of starting cyl */
#define CF_HD_SELECT     0x0d            /* 101dhhhh, d=drive, hhhh=head */
#define CF_HD_STATUS     0x0f            /* see status-bits */
#define CF_HD_CONTROL    0x0010000d      /* control/altstatus */

In order to get this working with u-boot, I had to endian swap all
16-bit data register accesses. I found that I had to do the same thing
in the ata_input_data and ata_output_data functions in ide-iops.c file
in linux before it would correctly identify the drive. I don't
understand why I have to do this in either place, given the
connections I described above, but it seems to be required.

After doing this, I get the following at bootup:

>Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
>ide: Assuming 50MHz system bus speed for PIO modes; override with
idebus=xx
>ide0: CF IDE interface
>Probing IDE interface ide0...
>hda: SanDisk SDCFJ-128, CFA DISK drive
>ide0 at 0xd1080000-0xd1080007,0xd118000d on irq 23
>hda: max request size: 128KiB
>hda: 250880 sectors (128 MB) w/1KiB Cache, CHS=980/8/32
>hda: cache flushes not supported
> hda: unknown partition table

I partitioned the drive using debian linux workstation with a pair of
type 83 partitions, and added a root file system to one of them. I can
see all this from u-boot, and can even load the kernel from it. I am
lost as to where to proceed to figure out why the kernel is having
trouble.

The other problem I am having is that after booting up, any attempt to
access /dev/hda using fdisk results in fdisk hanging, and "lost
interrupt" messages appearing on the console. If I look at
/proc/interrupts, it shows that 11 interrupts have occurred on
interrupt number 23. I have this set to level senstitive, positive
polarity.

Can someone offer some guidance?

Thanks.

Gregg