Re: libata dev_config call order wrong.

[Jeff Garzik <jgarzik@pobox.com>]

Brad Campbell wrote:
> Jeff Garzik wrote:
> 
>>
>> Take a look at what ATA_DFLAG_LOCK_SECTORS does ;-)  Here's hoping 
>> that's the last piece of the puzzle...
> 
> 
> HAHA! He says. That was not in 2.6.8.1! That solves that one.
> 
> Frankly, now I'm running a WD2500BB on one of those converters on a 
> RAID-5 I'm just going to leave the 200 sector hack in my kernel until we 
> figure out a way to detect the bridges. I can't see any other clean way 
> of doing it.

me either, unless I can figure out a way to detect that a disk is PATA 
not SATA.  Nothing is obvious from the specs on www.t13.org, but who knows.

Honestly I would prefer to apply some "if it's PATA-over-SATA bridge" 
rules, namely,
* limit to udma5 (udma/100)
* limit to 256 sectors [actually 200 in libata]


>> BTW I just sent email to SiI to see if there is a way to detect a 
>> PATA->SATA bridge.
> 
> 
> Man I hope so.. I don't want to resort to having to throw these board 
> away or constantly patch my kernel! I could just go and buy some more 
> SATA drives I guess, but where would the fun be in that!

hehe

Another option is to have a "no large transfers" module option that 
users could specify; that way you wouldn't have to patch the kernel, and 
it would work for all drivers without having to maintain a blacklist.

	Jeff