On Mon, 21 Jul 2008, Michael Schmitz wrote:
Here goes ...
On Wed, Jul 16, 2008 at 11:59:53PM +0200, Michael Schmitz wrote:
We should close it with the next kernel upload... with any luck, I'll
have the SMC91C111 driver working by then (it did finally own up to
detecting the proper hardware, instead of stuffing up the ROM-port card
or just hanging the kernel).
Does Geert have this in his queue yet?
No. Where's the patch?
This patch fixes errors I introduced in preparing the io.h code for ROM port
ISA access (for the EtherNEC driver, originally).
My assumption was that insb/outsb and friends should be mapped to the ROM port
accessors unconditionally, and the read*/write* functions should behave like
regular ISA accesses. As it turned out, it's just the other way 'round.
"Regular" ISA readb would apply the ROM port address mapping without actually
using the correct access function.
ins*/outs* are conditionalized now (port < 1024 goes to the ROM port, all else
via MMIO), read*/write* use straight MMIO no questions asked.
This fixes register access problems for the EtherNAT (SMC91C111) when used
together with the EtherNEC (ROM port NE2k clone). Tested on my Falcon.
Applies after Geert's linux-m68k-patches-2.6/atari-rom-isa.diff .... should
integrate accordingly in the Debian kernel source.
Sorry, it took a while...
The patch idea looks fine to me.
--- a/include/asm-m68k/io.h
+++ b/include/asm-m68k/io.h
@@ -429,8 +429,8 @@
/*
* kernel with both ROM port ISA and IDE compiled in, those have
* conflicting defs for in/out. Simply consider port < 1024
- * ROM port ISA and everything else regular ISA for IDE. read,write not
defined
But it's seriously whitespace-challenged. Time to tame alpine?
+ * Originally, insb/outsb and insw/outsw were set to the ROM variants in
2.6.19
Do we need this explanation in the comments? It's already in the patch
comment.
+ * insl/outsl not used. isa_ins* and isa_outs* do the Right Thing so this is
OK.
+ */
+#define insb(port, buf, nr) ((port) < 1024 ? isa_rom_insb((port), (buf),
(nr)) : isa_insb((port), (buf), (nr)))
+#define insw(port, buf, nr) ((port) < 1024 ? isa_rom_insw((port), (buf),
(nr)) : isa_insw((port), (buf), (nr)))
+#define insl(port, buf, nr) ((port) < 1024 ? isa_rom_insl((port), (buf),
(nr)) : isa_insl((port), (buf), (nr)))
+#define outsb(port, buf, nr) ((port) < 1024 ? isa_rom_outsb((port), (buf),
(nr)) : isa_outsb((port), (buf), (nr)))
+#define outsw(port, buf, nr) ((port) < 1024 ? isa_rom_outsw((port), (buf),
(nr)) : isa_outsw((port), (buf), (nr)))
+#define outsl(port, buf, nr) ((port) < 1024 ? isa_rom_outsl((port), (buf),
(nr)) : isa_outsl((port), (buf), (nr)))
+
+/*
+ * Previously, read* / write* used the isa_ variants here.
Same here.
+ * This is absolutely deadly: with CONFIG_ATARI_ROM_ISA defined, isa_readb +
* will use the ROM port space mtb translation, but at the same time, uses + *
plain old in_8() instead of the ROM port access code. So that'll bang + * on
the ROM port using invalid addresses, actually. Mayhem ensues.
+ *
+ * Using in_* / out_*, both EtherNEC and EtherNAT work. If things get more +
* complex than that, I'll have to make the ENEC_ISA_* mappings conditional + *
on the 'port' addresses, perhaps. + * First try to make read[bw]/write[bw]
conditional, though. + */
+#define readb(addr) in_8(addr)
+#define writeb(val,addr) out_8((addr),(val))
+#define readw(addr) in_le16(addr)
+#define writew(val,addr) out_le16((addr),(val))
+
+#define readsw raw_insw
+#define writesw raw_outsw
#define readsl raw_insl
#define writesl raw_outsl
#endif
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
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
--
To unsubscribe from this list: send the line "unsubscribe linux-m68k" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
