Erik Benada wrote:

Jeff,

included is patch to add support for PATA port on Promise PDC2037x controllers.
Patch is against 2.6.8.1 vanilla kernel.
I tested this patch on my PDC20378 on-board controller with 1 PATA drive attached.
I didn't do any large testing but my drive was detected properly and I was able to 
mount and use filesystem on it.

I tried to minimize changes to libata code. I just added flags for each port to ata_probe_ent
structure and modified ata_host_init() function. 

Promise SATA driver was changed to use new ata_probe_ent->port_flags, check for presence of
PATA port and pdc_phy_reset will use different reset code for PATA and SATA ports.

Any comments, suggestions?

Very nice!

I don't see any bugs in the code...  my main concern is that I 
definitely want to see host controllers (struct ata_host_set) support 
multiple ata_port_operations structures.  That would eliminate the need 
to do

+	if (ap->flags & ATA_FLAG_SATA)
+		sata_phy_reset(ap);
+	else
+		pdc_pata_phy_reset(ap);

and also eliminate a few other roadblocks.

However, implementing such is not a simple as your solution.

I will definitely merge this patch into my libata-dev queue, so that 
others can download and test it (and/or simply get their PATA port 
working).  We'll see where it goes from there.   If nobody (including 
me) implements the ata_port_operations solution I described, then I will 
probably push this into upstream 2.6.x as an incremental step towards 
the proper solution.

Thanks,

	Jeff