Re: Unaligned Access in hp6xx

[Paul M. <le...@li...>]

On Fri, Mar 09, 2007 at 09:39:54AM +0000, Kristoffer Ericson wrote:
> Still getting those unaligned access issues. I've setup the setup.c to have 
> the proper cf_ide IO values but getting same error.
> 
Current git should work for this case now that the PG_dcache_dirty bits
have been reverted. It is quite interesting that you hit this on SH-3
though, this suggests that it's not actually a fault with the lazy
write-back itself, but simply a missing write-back somewhere else that
PG_mapped happens to paper over.

> As soon as I get it working, I plan on start removing the PIO mapping so we 
> can get rid of io.c as Paul suggested.
> 
You should be able to do that from the beginning. As long as
pata_platform is capable of detecting the device and you're able to mount
it, you're in reasonable shape as far as the I/O mappings are concerned.

The only potential issue would be the access width, we've seen the
ioread8() cause bus locks on early R7780RP-1 boards where it really does
need to be a 16-bit access with the 8-bits masked. Looking at the hd64461
code, this should not be a problem.