Re: Marvell SATA driver

[KIYOHARA Takashi <kiyohara%kk.iij4u.or.jp@localhost>]

Hi! Tsutsui-san,


From: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
Date: Sun, 19 Jul 2009 23:36:01 +0900

> > From: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
> > > > > But isn't it better to use cf_flags like pciide(4) rather than
> > > > > options in config(5) and #ifdefs?
> > > > 
> > > > Please specify flags of wd(4) if you do not use DMA for a HDD in which
> > > > you are.  see man wd(4).
> > > 
> > > I'm not asking howto, but design.
> > > 
> > > My question is:
> > > "Is there any particular reason why do you choose a different method
> > >  (i.e. #ifdef with options) from other ATA drivers to disable DMA?"
> > 
> > The DMA of mvsata is different from pciide and wdc.  However, it becomes
> > compatible wdc when operating only by PIO.  I think that the size of the
> > processing of DMA is obstructive in this case.
> 
> Are you saying you choose different method (#ifdef or cf_flags)
> because implementation is different? I don't see the particular point.
> 
> Internal implementation is invisible by user
> but configuration is user visible, so I think
> unified config(5) format among the similar devices
> is useful for users. But if you don't think that's
> a matter, I'll shut up.

I think, recent SATA controllers(siisata, ahcisata) supports DMA.
This function is not offered for the end-users.  This is offered to
developers.
And, if the end-users is confused, all '#ifdef - #endif' will be deleted
to begin with.  DMA is always enabled.

Thanks,
--
kiyohara