[TRIVIAL PATCH] 2.4.25pre7 warning fix

[TRIVIAL PATCH] 2.4.25pre7 warning fix

[Krzysztof Halasa]

[-- Attachment #1: Type: text/plain, Size: 476 bytes --]

Hi,

The attached patch fixes the following warning msg:

time.c:435: warning: `do_gettimeoffset_cyclone' defined but not used

There is no need to define functions which do just { return 0; } and
which aren't called by anything.

(In case CONFIG_X86_SUMMIT is defined, there is another (real)
do_gettimeoffset_cyclone() function, and it is referenced - but
it's simply not related to this empty function).

Please apply to 2.4 kernel tree. Thanks.
-- 
Krzysztof Halasa, B*FH

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: time-2.4.23pre9.patch --]
[-- Type: text/x-patch, Size: 433 bytes --]

--- linux-2.4.orig/arch/i386/kernel/time.c	2003-10-24 22:48:15.000000000 +0200
+++ linux-2.4/arch/i386/kernel/time.c	2003-11-09 17:38:18.000000000 +0100
@@ -430,7 +430,6 @@
 
 const int use_cyclone = 0;
 static void mark_timeoffset_cyclone(void) {}
-static unsigned long do_gettimeoffset_cyclone(void) {return 0;}
 static void init_cyclone_clock(void) {}
 void __cyclone_delay(unsigned long loops) {}
 #endif /* CONFIG_X86_SUMMIT */

Re: [TRIVIAL PATCH] 2.4.25pre7 warning fix

[Marcelo Tosatti]

On Tue, 27 Jan 2004, Krzysztof Halasa wrote:

> Hi,
>
> The attached patch fixes the following warning msg:
>
> time.c:435: warning: `do_gettimeoffset_cyclone' defined but not used
>
> There is no need to define functions which do just { return 0; } and
> which aren't called by anything.
>
> (In case CONFIG_X86_SUMMIT is defined, there is another (real)
> do_gettimeoffset_cyclone() function, and it is referenced - but
> it's simply not related to this empty function).

Applied, thanks.

Btw, why do we need cyclone_setup() for !CONFIG_X86_SUMMIT ?

/* No-cyclone stubs */
#ifndef CONFIG_X86_SUMMIT
int __init cyclone_setup(char *str)
{
        printk(KERN_ERR "cyclone: Kernel not compiled with
CONFIG_X86_SUMMIT, cannot use the cyclone-timer.\n");
        return 1;
}

Re: [TRIVIAL PATCH] 2.4.25pre7 warning fix

[Krzysztof Halasa]

Marcelo Tosatti <marcelo.tosatti@cyclades.com> writes:

> Btw, why do we need cyclone_setup() for !CONFIG_X86_SUMMIT ?
>
> /* No-cyclone stubs */
> #ifndef CONFIG_X86_SUMMIT
> int __init cyclone_setup(char *str)
> {
>         printk(KERN_ERR "cyclone: Kernel not compiled with
> CONFIG_X86_SUMMIT, cannot use the cyclone-timer.\n");
>         return 1;
> }

No idea. All the stubs seem strange to me. I will make another patch
to address that.
-- 
Krzysztof Halasa, B*FH

Re: [TRIVIAL PATCH] 2.4.25pre7 warning fix

[Krzysztof Halasa]

Marcelo Tosatti <marcelo.tosatti@cyclades.com> writes:

> Btw, why do we need cyclone_setup() for !CONFIG_X86_SUMMIT ?
>
> /* No-cyclone stubs */
> #ifndef CONFIG_X86_SUMMIT
> int __init cyclone_setup(char *str)
> {
>         printk(KERN_ERR "cyclone: Kernel not compiled with
> CONFIG_X86_SUMMIT, cannot use the cyclone-timer.\n");
>         return 1;
> }

After having a closer look at it I think we should:

1. if CONFIG_X86_TSC is set:
   - make calibrate_tsc() failure a fatal error
   - assume use_tsc = 1 and x86_udelay_tsc = 1 and optimize them out
     with preprocessor

2. if CONFIG_X86_SUMMIT is _not_ set:
   - assume use_cyclone = 0 and optimize it out as well.
   - cyclone_setup() etc should go out.

3. I would rename CONFIG_X86_TSC to something like CONFIG_X86_TSC_FORCE
   - the current name is misleading. It wouldn't affect .config.

This is all 2.4-only, as 2.6 is a little different here.

Comments?
-- 
Krzysztof Halasa, B*FH

Re: [TRIVIAL PATCH] 2.4.25pre7 warning fix

[john stultz]

On Thu, 2004-01-29 at 08:38, Krzysztof Halasa wrote:
> Marcelo Tosatti <marcelo.tosatti@cyclades.com> writes:
> 
> > Btw, why do we need cyclone_setup() for !CONFIG_X86_SUMMIT ?
> >
> > /* No-cyclone stubs */
> > #ifndef CONFIG_X86_SUMMIT
> > int __init cyclone_setup(char *str)
> > {
> >         printk(KERN_ERR "cyclone: Kernel not compiled with
> > CONFIG_X86_SUMMIT, cannot use the cyclone-timer.\n");
> >         return 1;
> > }

This is needed because cyclone_setup() is called by
detect_clustered_apic(), which may or may not be done on a kernel w/
CONFIG_X86_SUMMIT enabled. 

> 
> After having a closer look at it I think we should:
> 
> 1. if CONFIG_X86_TSC is set:
>    - make calibrate_tsc() failure a fatal error
>    - assume use_tsc = 1 and x86_udelay_tsc = 1 and optimize them out
>      with preprocessor

Sounds fair. 


> 
> 2. if CONFIG_X86_SUMMIT is _not_ set:
>    - assume use_cyclone = 0 and optimize it out as well.

We already do this. 

>    - cyclone_setup() etc should go out.

cyclone_setup() is still needed. 

> 3. I would rename CONFIG_X86_TSC to something like CONFIG_X86_TSC_FORCE
>    - the current name is misleading. It wouldn't affect .config.

Agreed. 


thanks
-john