Thread (5 messages) flat view 5 messages, 4 authors, 2000-03-14

Re: Byte swapped inw/outw again....

From: Mark S. Mathews <hidden>
Date: 2000-03-14 14:39:38

Just to finish off this thread...

My thanks for Geert's and Gabriel's responses.  While _I_ think it's
wierd, I'll just smile and change my code per Geert's suggestions.  I get
the feeling this is a subject that's been debated heatedly in the past and
don't want to refan the flames.

Why do I think it's wierd?  It just bothers me that the following two
pieces of code don't give you the same thing:

   uint16  a;
   uint8   *p = (uint8*)&a;
   *p = inb(portaddr);
   *(p+1) = inb(portaddr+1);
   printk("%d\n", a);

and

   a = inw(portaddr);
   printk("%d\n", a);

It just seems more logical (and readable) to me to have in[wl]/out[wl]
variants that explicitly state (as part of their name) that they are going
to do the swap.  This whole thing smells like 'magic' to me.  ;-)

Oh well, I never did mind the little things....

Thanks Again,
-Mark

On Tue, 14 Mar 2000, Geert Uytterhoeven wrote:
On Mon, 13 Mar 2000, Mark S. Mathews wrote:
quoted
Another driver port with inw/outw issues again.  I've looked through the
linuxppc-dev archive and I'm not quite seeing the answer to my question
(below).

I'm porting a driver for a WLAN card whose interface is all 16-bit
registers.  Sometimes the reads/writes are strictly a 16-bit integer
quantity where the byte swapping in inw/outw wouldn't be too much of a
problem.  Other times we're reading/writing buffer contents from/to the
card 2-bytes at a time (swapping in this case is exceptionally bad).

When writing the driver originally on the x86 I put in xxx_to_cpu and
cpu_to_xxx calls everywhere assuming that the inw/outw words would be
coming up byteswapped on the PPC and other platforms.  Now, after I
realize that inw/outw are trying to do the byte swapping....my code is
swapping them back and we're _still_ backwards.  ;-)
Hence, you should remove the le*_to_cpu() calls. Inw()/outw() are defined to
operate on PCI/ISA I/O space, which is defined to be little endian.
quoted
What I'd like to do is set/unset or undef/define an item such that the
inw/outw I'm using _will not_ do the byte swapping.  On reading the code
and looking through the archives, I get the feeling there is a RIGHT way
to do this.  I just haven't been able to figure it out precisely.
There is indeed a way to do this, but since it's not `The Right Way' for this
case I won't tell you :-) Doing this would still make it break on other big
endian platforms, which is not what you want, I assume.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Mark S. Mathews

AbsoluteValue Software     Web:    http://www.absoval.com
P.O. Box 941149            e-mail: mark@absoval.com
Maitland, FL 32794-1149    Phone:  407.644.8582
USA                        Fax:    407.539.1294

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help