Re: [PATCH] make ata_exec_internal_sg honor DMADIR

[Vincent Pelletier <plr.vincent@gmail.com>]

Le lundi 20 mai 2013 01:38:18, Tejun Heo a écrit :
> I don't really wanna go that way.  Those bridges always have been
> something fringe and broken in ways which aren't fundamentally
> fixable.  Fixing one would break another without anyway to properly
> detect them.
> 
> So, I'm okay with having a knob for cases where the user knows what to
> do but I don't think even that is something of much importance, and
> I'm definitely not gonna do anything which may affect !bridge case
> adversely.

I understand.

> Those bridges have always been a second-class citizen and
> their importance has waned a lot.

Just a bit of background on why I got interested in Csaba's patch:

I never used my bridges when I originally got them (serillel was bundled with 
some motherboards I bought circa 2003), because a PATA controller was 
available anyway.
Since then, I changed all my hard drives progressively to SATA ones because of 
the capacity increase. But I still have many DVD drives from that time, 
because there was just no motivation to change them, and I don't have any 
optical SATA drive.
Then I bought a new motherboards a few weeks ago, without paying enough 
attention: there is no PATA controller on it. So I remembered those bridges 
and finally tried to use them.

Now, given that those bridges are old and were originally IMHO very close to 
useless, I'm indeed probably part of a fringe who didn't throw them away and 
remember them. SATA optical drives are quite cheap, maybe there is not even a 
financial motivation to look for such bridge.

> EH stands for exception handling and discovery / init definitely are
> part of exception handling in libata speak.

Thanks for the clarification.

> So, nope, I really don't want this.

Err, the body of this patch didn't change from my original submission, only 
the commit message has changed.

> > +atapi_dmadir
> > +
> > +	Bitmask enabling dmadir for corresponding device if ATAPI.
> > +	1:	Enable dmadir for port's device 0
> > +	2:	Enable dmadir for port's device 1
> > +	(etc)
> > +	See also libata's atapi_dmadir module parameter.
> 
> Shouldn't this be a device property?

Unplugging the drive would, in my understanding, loose the setting if stored 
at the device level. Is there another way to trigger a new initialisation 
attempt after changing the setting ?
Should I add a "rescan" device attribute ?

Regards,
-- 
Vincent Pelletier