Fix hd section references:
make parse_hd_setup() __init
Error: ./drivers/ide/legacy/hd.o .text refers to 00000943 R_386_PC32
.init.text
Signed-off-by: maximilian attems <[email protected]>
--- linux-2.6.12-rc1-bk5/drivers/ide/legacy/hd.c.orig 2005-04-04 18:39:04.000000000 +0200
+++ linux-2.6.12-rc1-bk5/drivers/ide/legacy/hd.c 2005-04-04 19:02:57.908576221 +0200
@@ -851,7 +851,7 @@
goto out;
}
-static int parse_hd_setup (char *line) {
+static int __init parse_hd_setup (char *line) {
int ints[6];
(void) get_options(line, ARRAY_SIZE(ints), ints);
maximilian attems wrote:
> Fix hd section references:
> make parse_hd_setup() __init
>
> Error: ./drivers/ide/legacy/hd.o .text refers to 00000943 R_386_PC32
> .init.text
>
> Signed-off-by: maximilian attems <[email protected]>
>
>
> --- linux-2.6.12-rc1-bk5/drivers/ide/legacy/hd.c.orig 2005-04-04 18:39:04.000000000 +0200
> +++ linux-2.6.12-rc1-bk5/drivers/ide/legacy/hd.c 2005-04-04 19:02:57.908576221 +0200
> @@ -851,7 +851,7 @@
> goto out;
> }
>
> -static int parse_hd_setup (char *line) {
> +static int __init parse_hd_setup (char *line) {
> int ints[6];
>
> (void) get_options(line, ARRAY_SIZE(ints), ints);
This one is fairly interesting and needs some resolution by someone
who knows....
On the surface, the patch is correct.
Rusty, can you explain when __setup functions are called relative
to in-kernel init functions? or put another way, can a __setup
function safely call in __init function?
Here's the function in question:
static int parse_hd_setup (char *line) {
int ints[6];
(void) get_options(line, ARRAY_SIZE(ints), ints);
hd_setup(NULL, ints);
return 1;
}
__setup("hd=", parse_hd_setup);
Should we make parse_hd_setup() __init,
or make hd_setup() non-__init, or something else?
{time passes, he looks]
OK, I looked at include/linux/init.h. From what I can see
there, __setup() causes an .init.setup section to be emitted,
so marking __setup() function as __init would make sense.
I think that this patch is good.
Thanks.
--
~Randy
On Mon, 04 Apr 2005, Randy.Dunlap wrote:
> >-static int parse_hd_setup (char *line) {
> >+static int __init parse_hd_setup (char *line) {
..
> This one is fairly interesting and needs some resolution by someone
> who knows....
thanks a lot for your quick and profund feedback.
> On the surface, the patch is correct.
>
> Rusty, can you explain when __setup functions are called relative
> to in-kernel init functions? or put another way, can a __setup
> function safely call in __init function?
>
> Here's the function in question:
>
> static int parse_hd_setup (char *line) {
> int ints[6];
>
> (void) get_options(line, ARRAY_SIZE(ints), ints);
> hd_setup(NULL, ints);
>
> return 1;
> }
> __setup("hd=", parse_hd_setup);
>
>
>
> Should we make parse_hd_setup() __init,
> or make hd_setup() non-__init, or something else?
>
> {time passes, he looks]
>
> OK, I looked at include/linux/init.h. From what I can see
> there, __setup() causes an .init.setup section to be emitted,
> so marking __setup() function as __init would make sense.
> I think that this patch is good.
i saw that ide_setup() is __init as a bunch of lots others.
yes init.h confirms that. :)
--
maks
kernel janitor http://janitor.kernelnewbies.org/
On Tue, 2005-04-05 at 01:05 +0200, maximilian attems wrote:
> On Mon, 04 Apr 2005, Randy.Dunlap wrote:
> > Rusty, can you explain when __setup functions are called relative
> > to in-kernel init functions? or put another way, can a __setup
> > function safely call in __init function?
Yes. init sections are discarded just before "init" is exec'ed, ie. at
the last moment in the boot process. __setup() functions are called at
the same time as module_param() parsing, which is all done before the
various initcall/module_init calls.
Cheers,
Rusty.
--
A bad analogy is like a leaky screwdriver -- Richard Braakman