[RFC] patch to AHCI device detection code

Søren Schmidt sos at FreeBSD.ORG
Wed Jun 11 08:42:53 UTC 2008


You are right on that one of course, from my quick eyeballing the  
patch I thought you enabled the reset sequence too, but on a closer  
look its only the extra timeout, my bad.

However, the most usual problem is that the devices does not leave a  
signature behind due to the missing reset sequence. That problem is  
50/50 as some controllers apparently doesn't like the resetting and  
will leave busy set and a hanging channel.

I'll be getting some new HW here in the lab tomorrow if all goes well,  
that will let me play with this on one of the failing devices again.  
Maybe I'll find a generic solution this time around.

I'll commit this patch for now, as you said, it doesn't hurt..


On 11Jun, 2008, at 10:00 , Andrey V. Elsukov wrote:

> Søren Schmidt wrote:
>> Well, as you can see I tried this when I added the PM code to AHCI,  
>> however one of the reasons I left it out again (for now) is that it  
>> doesn't work on more than about 50% of the AHCI chipsets out there..
>> Modern HW in a nutshell, you win some you loose some ;)
> Hm.. I think it can't made regression. I don't see nothing dangerous
> in addition 1s waiting (it doesn't enable softreset code, only PxTFD
> reading loop)..
> -- 
> WBR, Andrey V. Elsukov

More information about the freebsd-hackers mailing list