http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=
f354942cb301fed273f423fb5c4f57bde3efc5b2
converted the check_region() calls in drivers/pcmcia/i82365.c into
request_regions. Unfortunately this seems to have broken things.
isa_probe() used to call check_region() and then call add_pcic() which would
request_region().
Now isa_probe() calls request_region() and then calls add_pcic() which calls
request_region() again, this fails and add_pcic() returns immediately without
doing all the setup etc.
On the face of it the patch below fixes the problem, by not doing the second
request region in add_pcic(). I think this is preferable to remove the call
in isa_probe() since identify() touches the I/O regions and is called before
add_pcic().
However I haven't fully grokked the meaning of the code which follows the
request_region() in isa_probe(), so I'm not sure that the handling WRT
multiple sockets and multiple bridge chips is correct. In particular I'm not
convinced that the regions for subsequent sockets and/or bridges will be
requested at all. I suspect a more thorough reworking by someone who
understands what is going on there might be in order.
I should mention that I'm actually messing about with this on an ARM platform
with wacky memory and i/o mapping offsets etc, it doesn't quite work yet for
other reasons which preclude full testing etc, but I think the problem above
is still present for more normal x86 stuff.
Signed-off-by: Ian Campbell <icampbell@arcom.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
struct i82365_socket *t = &socket[sockets-ns];
base = sockets-ns;
- if (t->ioaddr > 0) {
- if (!request_region(t->ioaddr, 2, "i82365")) {
- printk(KERN_ERR "i82365: IO region conflict at %#lx, not available\n",
- t->ioaddr);
- return;
- }
- }
-
if (base == 0) printk("\n");
printk(KERN_INFO " %s", pcic[type].name);
printk(" ISA-to-PCMCIA at port %#lx ofs 0x%02x",