Re: Regression 3.2 -> 3.3-rc1 10 sec hang at boot and resume, COMRESET failed

On 02/15/2012 06:39 PM, Brian Norris wrote:

On Wed, Feb 15, 2012 at 11:18 AM, Jeff Garzik[off-list ref]  wrote:

quoted

On 02/15/2012 01:31 PM, Tejun Heo wrote:

quoted

The problem is that there's no way to identify the controller in
question, so we can't do this automatically, so might just as well do
it in the simplest way for now.  :(

See ahci_devtype[] and ahci_port_info[] in ahci_platform.c for how to do
this.  Brian would not have to tweak the driver out of tree as claimed; we
put all changes in tree, and the platform calls itself "spec-strict-ahci" or
whatever string you prefer.

OK, so can it seems I would add a flag in include/linux/libata.h that
can be added to a new entry in ahci_port_info[]? It seems a little
awkward to put in "libata.h", plus I'm not sure if all the flags have
direct meaning to ATA or if they can be AHCI-specific.

Nothing needs to go into libata.h.  grep for IMX53_AHCI, and see how 
that differentiates ata_port_info information.  There you may vary 
AHCI-specific flags (defined in ahci.h), port operations, or anything 
else specific to the controller.

Create a flag, add it to ahci.h, guard the spec-strict behavior creating 
problems with said flag, and apply said flag to your controller in 
ahci_platform.

	Jeff