Hi,
The 2.6.24-git18 kernel build fails on the power machine with following message
drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
make[3]: *** [drivers/input/mouse/psmouse-base.o] Error 1
make[2]: *** [drivers/input/mouse] Error 2
make[1]: *** [drivers/input] Error 2
make: *** [drivers] Error 2
this failure was reported earlier http://lkml.org/lkml/2007/9/22/64
--
Thanks & Regards,
Kamalesh Babulal,
Linux Technology Center,
IBM, ISTL.
On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
> Hi,
>
> The 2.6.24-git18 kernel build fails on the power machine with following message
>
> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
> make[3]: *** [drivers/input/mouse/psmouse-base.o] Error 1
> make[2]: *** [drivers/input/mouse] Error 2
> make[1]: *** [drivers/input] Error 2
> make: *** [drivers] Error 2
>
> this failure was reported earlier http://lkml.org/lkml/2007/9/22/64
You need the patch from [1] manually applied (due to changed context)
and with " || defined(__powerpc__)" added for getting it compiling.
@Rusty:
Can you look at this issue? It's a modules problem on 3 architectures.
cu
Adrian
[1] http://lkml.org/lkml/2007/12/27/29
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
Adrian Bunk wrote:
> On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
>> Hi,
>>
>> The 2.6.24-git18 kernel build fails on the power machine with following message
>>
>> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
>> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
>> make[3]: *** [drivers/input/mouse/psmouse-base.o] Error 1
>> make[2]: *** [drivers/input/mouse] Error 2
>> make[1]: *** [drivers/input] Error 2
>> make: *** [drivers] Error 2
>>
>> this failure was reported earlier http://lkml.org/lkml/2007/9/22/64
>
> You need the patch from [1] manually applied (due to changed context)
> and with " || defined(__powerpc__)" added for getting it compiling.
>
> @Rusty:
> Can you look at this issue? It's a modules problem on 3 architectures.
>
> cu
> Adrian
>
> [1] http://lkml.org/lkml/2007/12/27/29
>
I tried manually applying the patch and added ||defined(__powerpc__), the above
build failure was resolved but it started failing at another place,
CC kernel/printk.o
kernel/printk.c:568: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘static’
make[1]: *** [kernel/printk.o] Error 1
make: *** [kernel] Error 2
--
Thanks & Regards,
Kamalesh Babulal,
Linux Technology Center,
IBM, ISTL.
On Fri, Feb 08, 2008 at 06:20:52PM +0530, Kamalesh Babulal wrote:
> Adrian Bunk wrote:
> > On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
> >> Hi,
> >>
> >> The 2.6.24-git18 kernel build fails on the power machine with following message
> >>
> >> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
> >> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
> >> make[3]: *** [drivers/input/mouse/psmouse-base.o] Error 1
> >> make[2]: *** [drivers/input/mouse] Error 2
> >> make[1]: *** [drivers/input] Error 2
> >> make: *** [drivers] Error 2
> >>
> >> this failure was reported earlier http://lkml.org/lkml/2007/9/22/64
> >
> > You need the patch from [1] manually applied (due to changed context)
> > and with " || defined(__powerpc__)" added for getting it compiling.
> >
> > @Rusty:
> > Can you look at this issue? It's a modules problem on 3 architectures.
> >
> > cu
> > Adrian
> >
> > [1] http://lkml.org/lkml/2007/12/27/29
> >
> I tried manually applying the patch and added ||defined(__powerpc__), the above
> build failure was resolved but it started failing at another place,
>
> CC kernel/printk.o
> kernel/printk.c:568: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘static’
> make[1]: *** [kernel/printk.o] Error 1
> make: *** [kernel] Error 2
Below is what worked for me (based on Ivan's patch).
The comment is wrong now, the #if's should refer to kconfig variables,
and I don't know whether this patch is really the best solution.
cu
Adrian
1fffe660ace75bd875fc23f7b2121d7341f08464 diff --git a/include/linux/moduleparam.h b/include/linux/moduleparam.h
index 8126e55..d17ce12 100644
--- a/include/linux/moduleparam.h
+++ b/include/linux/moduleparam.h
@@ -62,6 +62,15 @@ struct kparam_array
void *elem;
};
+/* On alpha and ia64 relocations to global data cannot go into read-only
+ sections, so 'const' makes no sense and even causes compile failures
+ with some compilers. */
+#if defined(__alpha__) || defined(__ia64__) || defined(__powerpc__)
+#define __moduleparam_const
+#else
+#define __moduleparam_const const
+#endif
+
/* This is the fundamental function for registering boot/module
parameters. perm sets the visibility in sysfs: 000 means it's
not there, read bits mean it's readable, write bits mean it's
@@ -71,7 +80,7 @@ struct kparam_array
static int __param_perm_check_##name __attribute__((unused)) = \
BUILD_BUG_ON_ZERO((perm) < 0 || (perm) > 0777 || ((perm) & 2)); \
static const char __param_str_##name[] = prefix #name; \
- static struct kernel_param const __param_##name \
+ static struct kernel_param __moduleparam_const __param_##name \
__used \
__attribute__ ((unused,__section__ ("__param"),aligned(sizeof(void *)))) \
= { __param_str_##name, perm, set, get, { arg } }
Adrian Bunk wrote:
> On Fri, Feb 08, 2008 at 06:20:52PM +0530, Kamalesh Babulal wrote:
>> Adrian Bunk wrote:
>>> On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
>>>> Hi,
>>>>
>>>> The 2.6.24-git18 kernel build fails on the power machine with following message
>>>>
>>>> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
>>>> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
>>>> make[3]: *** [drivers/input/mouse/psmouse-base.o] Error 1
>>>> make[2]: *** [drivers/input/mouse] Error 2
>>>> make[1]: *** [drivers/input] Error 2
>>>> make: *** [drivers] Error 2
>>>>
>>>> this failure was reported earlier http://lkml.org/lkml/2007/9/22/64
>>> You need the patch from [1] manually applied (due to changed context)
>>> and with " || defined(__powerpc__)" added for getting it compiling.
>>>
>>> @Rusty:
>>> Can you look at this issue? It's a modules problem on 3 architectures.
>>>
>>> cu
>>> Adrian
>>>
>>> [1] http://lkml.org/lkml/2007/12/27/29
>>>
>> I tried manually applying the patch and added ||defined(__powerpc__), the above
>> build failure was resolved but it started failing at another place,
>>
>> CC kernel/printk.o
>> kernel/printk.c:568: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘static’
>> make[1]: *** [kernel/printk.o] Error 1
>> make: *** [kernel] Error 2
>
> Below is what worked for me (based on Ivan's patch).
>
> The comment is wrong now, the #if's should refer to kconfig variables,
> and I don't know whether this patch is really the best solution.
>
> cu
> Adrian
>
>
> 1fffe660ace75bd875fc23f7b2121d7341f08464 diff --git a/include/linux/moduleparam.h b/include/linux/moduleparam.h
> index 8126e55..d17ce12 100644
> --- a/include/linux/moduleparam.h
> +++ b/include/linux/moduleparam.h
> @@ -62,6 +62,15 @@ struct kparam_array
> void *elem;
> };
>
> +/* On alpha and ia64 relocations to global data cannot go into read-only
> + sections, so 'const' makes no sense and even causes compile failures
> + with some compilers. */
> +#if defined(__alpha__) || defined(__ia64__) || defined(__powerpc__)
> +#define __moduleparam_const
> +#else
> +#define __moduleparam_const const
> +#endif
> +
> /* This is the fundamental function for registering boot/module
> parameters. perm sets the visibility in sysfs: 000 means it's
> not there, read bits mean it's readable, write bits mean it's
> @@ -71,7 +80,7 @@ struct kparam_array
> static int __param_perm_check_##name __attribute__((unused)) = \
> BUILD_BUG_ON_ZERO((perm) < 0 || (perm) > 0777 || ((perm) & 2)); \
> static const char __param_str_##name[] = prefix #name; \
> - static struct kernel_param const __param_##name \
> + static struct kernel_param __moduleparam_const __param_##name \
> __used \
> __attribute__ ((unused,__section__ ("__param"),aligned(sizeof(void *)))) \
> = { __param_str_##name, perm, set, get, { arg } }
> --
Hi Adrian,
Thanks the patch resolves the problem, I had a typo when applied the manually :-(.
--
Thanks & Regards,
Kamalesh Babulal,
Linux Technology Center,
IBM, ISTL.
On Fri, Feb 08, 2008 at 03:30:22PM +0200, Adrian Bunk wrote:
> > > On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
> > >> The 2.6.24-git18 kernel build fails on the power machine with following message
> > >> drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a section type conflict
My fault - I somehow overlooked that ppc64 is the club member...
> The comment is wrong now, the #if's should refer to kconfig variables,
> and I don't know whether this patch is really the best solution.
I think it's the best, because it at least points at some of
the problems which arise when you blindly rely on some undocumented
behaviour of __attribute__(section()) on certain platform, and
assume it's the same for the rest of the world...
Alternatives are:
- drop the 'const' qualifier from that macro unconditionally, so that
__param section gets read/write on all arches. I'd be fine with it,
but then we'll probably have the same problem again and again...
- make the offending module_param functions global (psmouse and 3-4
other places). But how do you explain that to driver writers?
These functions are logically private to this particular module,
so it's obviously not the same as "we don't export static functions",
though underlying problem is precisely the same on our three
platforms...
- hack up gcc. But I don't even have a proposal, because it's not a bug
from gcc point of view. Ideally, I'd like to pass some extra parameters
to __attribute__(section()), like r/w control and some equivalent of
-fno-jump-tables...
Oh, well. Updated patch in a minute.
Ivan.
On Friday 08 February 2008 21:48:51 Adrian Bunk wrote:
> On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
> > drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a
> > section type conflict drivers/input/mouse/psmouse-base.c:44: error:
> > __param_proto causes a section type conflict
> @Rusty:
> Can you look at this issue? It's a modules problem on 3 architectures.
>
> cu
> Adrian
>
> [1] http://lkml.org/lkml/2007/12/27/29
Erk... This is horrible, and arguably something gcc should try to deal with.
How about we just drop the const and add a comment, rather than breaking one
arch at a time?
Thanks,
Rusty.
On Mon, Mar 03, 2008 at 03:35:16PM +1100, Rusty Russell wrote:
> On Friday 08 February 2008 21:48:51 Adrian Bunk wrote:
> > On Fri, Feb 08, 2008 at 04:02:59PM +0530, Kamalesh Babulal wrote:
> > > drivers/input/mouse/psmouse-base.c:44: error: __param_proto causes a
> > > section type conflict drivers/input/mouse/psmouse-base.c:44: error:
> > > __param_proto causes a section type conflict
> > @Rusty:
> > Can you look at this issue? It's a modules problem on 3 architectures.
> >
> > cu
> > Adrian
> >
> > [1] http://lkml.org/lkml/2007/12/27/29
>
> Erk... This is horrible, and arguably something gcc should try to deal with.
> How about we just drop the const and add a comment, rather than breaking one
> arch at a time?
We should have caught all architectures having problems, and a slightly
improved patch is now in Linus' tree.
> Thanks,
> Rusty.
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
On Monday 03 March 2008 19:13:45 Adrian Bunk wrote:
> On Mon, Mar 03, 2008 at 03:35:16PM +1100, Rusty Russell wrote:
> > Erk... This is horrible, and arguably something gcc should try to deal
> > with. How about we just drop the const and add a comment, rather than
> > breaking one arch at a time?
>
> We should have caught all architectures having problems, and a slightly
> improved patch is now in Linus' tree.
Yes, thanks, I saw it later on in the thread. 'Tis still horrible tho :)
Rusty.