Re: Complete support PC-9800 for 2.5.64-ac4 (11/11) SCSI

[James Bottomley <James.Bottomley@steeleye.com>]

On Sun, 2003-03-16 at 14:00, Geert Uytterhoeven wrote:
> On 16 Mar 2003, James Bottomley wrote:
> > On Sun, 2003-03-16 at 12:36, Geert Uytterhoeven wrote:
> > > Actually, it was my suggestion to remove the dereference for PIO accesses. In
> > > that case SASR contains the I/O port register.
> > 
> > There's still something wrong with the implementation in this patch. 
> > For non PIO SASR is defined as volatile unsigned char *SASR.  Its access
> > has gone from being outb(n, *regs.SASR) to outb(n, regs.SASR).  What
> > expansion can outb have on m68k and MIPS that makes this change
> > idempotent?
> 
> outb() and friends are only used if CONFIG_WD33C93_PIO is set. In all other
> cases, it uses the old implementation, e.g. `*regs.SASR = reg_num'.

Ah, OK, I see what it's doing.  Instead of having a single redefine of
wd33c93_outb to be either outb or readb etc, it has a whole chunk of
code in wd33c93.c that #ifdef's for this.

Perhaps, while you're cleaning this up, you'd like to move this into the
header?  Shouldn't it also be using readb/writeb instead of just direct
memory accesses?

James