[TRIVIAL] fix location of zap_low_mappings

[TRIVIAL] fix location of zap_low_mappings

[Jasper Spaans]

Hello list,

When compiling current BK 2.5, I get a warning about zap_low_mappings not
being declared. Moving it from smp.h to pgtable.h fixes this (and doesn't
break my setup). 

Does anyone object to this fix?

[not Cc:-ed to the trivial patch monkey, as I'm not sure whether pgtable.h
 is the right place to put this]

Bye,
Jasper

Index: linux-2.5/include/asm-i386/pgtable.h
===================================================================
RCS file: /home/cvs/linux-2.5/include/asm-i386/pgtable.h,v
retrieving revision 1.33
diff -u -r1.33 pgtable.h
--- l/include/asm-i386/pgtable.h	4 May 2003 01:50:19 -0000	1.33
+++ l/include/asm-i386/pgtable.h	6 Jun 2003 08:40:18 -0000
@@ -23,6 +23,7 @@
 
 extern pgd_t swapper_pg_dir[1024];
 extern void paging_init(void);
+extern void zap_low_mappings (void);
 
 /*
  * ZERO_PAGE is a global shared page that is always zero: used
Index: l/include/asm-i386//smp.h
===================================================================
RCS file: /home/cvs/linux-2.5/include/asm-i386/smp.h,v
retrieving revision 1.24
diff -u -r1.24 smp.h
--- l/include/asm-i386/smp.h	4 Jun 2003 00:16:47 -0000	1.24
+++ l/include/asm-i386/smp.h	6 Jun 2003 08:40:18 -0000
@@ -43,7 +43,6 @@
 extern void smp_send_reschedule(int cpu);
 extern void smp_invalidate_rcv(void);		/* Process an NMI */
 extern void (*mtrr_hook) (void);
-extern void zap_low_mappings (void);
 
 #define MAX_APICID 256
 

-- 
Jasper Spaans
http://jsp.vs19.net/contact/

``Got no clue? Too bad for you.''

Re: [TRIVIAL] fix location of zap_low_mappings

[William Lee Irwin III]

On Fri, Jun 06, 2003 at 11:57:49AM +0200, Jasper Spaans wrote:
> When compiling current BK 2.5, I get a warning about zap_low_mappings not
> being declared. Moving it from smp.h to pgtable.h fixes this (and doesn't
> break my setup). 
> Does anyone object to this fix?
> [not Cc:-ed to the trivial patch monkey, as I'm not sure whether pgtable.h
>  is the right place to put this]

It's basically not supposed to be visible on UP. Perhaps a better
approach would be declare it in pgtable.h as you did, stub out the UP
case with an empty function, and un-#ifdef it from mem_init().


-- wli

Re: [TRIVIAL] fix location of zap_low_mappings

[Jasper Spaans]

On Fri, Jun 06, 2003 at 04:34:43AM -0700, William Lee Irwin III wrote:
> On Fri, Jun 06, 2003 at 11:57:49AM +0200, Jasper Spaans wrote:
> > When compiling current BK 2.5, I get a warning about zap_low_mappings not
> > being declared. Moving it from smp.h to pgtable.h fixes this (and doesn't
> > break my setup). 
> > Does anyone object to this fix?

> It's basically not supposed to be visible on UP. Perhaps a better
> approach would be declare it in pgtable.h as you did, stub out the UP
> case with an empty function, and un-#ifdef it from mem_init().

That wouldn't seem right to me:

* in the UP-case, it is explicitly called in mem_init() [mm/init.c]:

void __init mem_init(void)
{
[...]
#ifndef CONFIG_SMP
        zap_low_mappings();
#endif
}

* in the SMP-case, this call is delayed until 
  smp_cpus_done() [kernel/smpboot.c]

These two cases are fine however, as the UP-case defines it in mm/init.c,
and the SMP-case has CONFIG_SMP enabled and includes <smp.h>; the warning
comes from this function being called from acpi_restore_state_mem()
[kernel/acpi/sleep.c]) in the UP-case, in which case it isn't declared in
<smp.h>.

Bye,

Jasper
-- 
Jasper Spaans
http://jsp.vs19.net/contact/

``Got no clue? Too bad for you.''

Re: [TRIVIAL] fix location of zap_low_mappings

[William Lee Irwin III]

On Fri, Jun 06, 2003 at 04:34:43AM -0700, William Lee Irwin III wrote:
>> It's basically not supposed to be visible on UP. Perhaps a better
>> approach would be declare it in pgtable.h as you did, stub out the UP
>> case with an empty function, and un-#ifdef it from mem_init().

On Fri, Jun 06, 2003 at 04:33:35PM +0200, Jasper Spaans wrote:
> That wouldn't seem right to me:

Yeah. That suggestion was totally bogus. Fine as-is then.

-- wli