[linuxsh-dev] Re: Solution Engin's ATA PC-card enabler

[NIIBE Y. <gn...@ch...>]

kaz Kojima writes:
 > MR-SHPC-01 chip has 1MB window which is constructed as the set of
 > four 256KB spaces. First 256KB space is "real word" space and the
 > second is "real byte" space. So the above change makes byte input
 > from the "real word" space. It seems not good in the i/o window
 > case since the access to this area has 16-bit sizing which will
 > cause problems on some PCMCIA cards. Of course, it might not cause
 > any problem on CF(ATA) cards, but I'd like to be defensive for such
 > chip :-)

Let's clarify things.  

#1:
Is this correct?

 >  /* SH pcmcia io window base, start and end.  */
 > -int sh_pcic_io_wbase = 0xb8400000;
 > -int sh_pcic_io_start;
 > -int sh_pcic_io_stop;
 > +int sh_pcic_io_wbase = 0xb8600000;
 > +int sh_pcic_io_start = 0x3000;
 > +int sh_pcic_io_stop  = 0x3fff;

Specifically, io_wbase has been changed.  Does not cause any problem?

#2:
 > unsigned long inb(unsigned int port)
 > {
 > 	if (sh_pcic_io_start <= port && port <= sh_pcic_io_stop)
 >-		return *(__u8 *) (sh_pcic_io_wbase + 0x40000 + port); 
 >+		return *(__u8 *) port2adr(port); 
 > 	else if (shifted_port(port))

It seems that your comments above means we need this "0x400000" for
byte access of "read" (in).  Is there any reason why we don't need
this "0x40000" for "write" (out)?
--