Re: 2.6.16-rc6-mm2 uninitialized online_policy_cpus.bits[0]

Re: 2.6.16-rc6-mm2 uninitialized online_policy_cpus.bits[0]

[Andrew Morton]

Con Kolivas <kernel@kolivas.org> wrote:
>
> Wonder if this is related to rc6's oops?
>  gcc 4.0.3
> 
>    CC [M]  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.o
>  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c: In function 
>  'centrino_target':
>  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is used 
>  uninitialized in this function
>    CC [M]  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.o
>  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c: In function 
>  'acpi_cpufreq_target':
>  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is used 
>  uninitialized in this function

Well conceivably.  That warning is a consequence of my quick hack to make
the ACPI tree compile on uniprocessor.

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.16-rc6/2.6.16-rc6-mm2/broken-out/git-acpi-up-fix.patch

My patch is, as the compiler points out, wrong.

I've sent that patch two or three times to the APCI maintainers, to the
ACPI mailing list and to the author of the original buggy patch.  The
response thus far has been dead silence.

IOW, despite my efforts, the ACPI tree has been in a non-compiling state on
uniprocessor since February 11.

This is pathetic.  People are trying to get things done here and ACPI is
getting in the way.  But *need* to get the ACPI development tree out for
people to test else we'll never be able to take another ACPI update into
mainline.

Re: 2.6.16-rc6-mm2 uninitialized online_policy_cpus.bits[0]

[Con Kolivas]

On Sunday 19 March 2006 12:35, Andrew Morton wrote:
> Con Kolivas <kernel@kolivas.org> wrote:
> > Wonder if this is related to rc6's oops?
> >  gcc 4.0.3
> >
> >    CC [M]  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.o
> >  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c: In function
> >  'centrino_target':
> >  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is
> > used uninitialized in this function
> >    CC [M]  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.o
> >  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c: In function
> >  'acpi_cpufreq_target':
> >  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is
> > used uninitialized in this function
>
> Well conceivably.  That warning is a consequence of my quick hack to make
> the ACPI tree compile on uniprocessor.
>
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.16-rc6/2.
>6.16-rc6-mm2/broken-out/git-acpi-up-fix.patch
>
> My patch is, as the compiler points out, wrong.
>
> I've sent that patch two or three times to the APCI maintainers, to the
> ACPI mailing list and to the author of the original buggy patch.  The
> response thus far has been dead silence.

Well this will end up being the wrong place to do it but I needed it to work
 now so this patch fixes it for me. Dunno what else it will break. Works on
a couple of configs fine.

Cheers,
Con
---
Hacky workaround for cpu_online_map not being defined

 arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c       |    2 --
 arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c |    2 --
 kernel/sched.c                                    |    4 ++++
 3 files changed, 4 insertions(+), 4 deletions(-)

Index: linux-2.6.16-rc6-mm2/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c
===================================================================
--- linux-2.6.16-rc6-mm2.orig/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c	2006-03-19 11:15:05.000000000 +1100
+++ linux-2.6.16-rc6-mm2/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c	2006-03-19 12:45:56.000000000 +1100
@@ -225,10 +225,8 @@ acpi_cpufreq_target (
 	freqs.old = data->freq_table[cur_state].frequency;
 	freqs.new = data->freq_table[next_state].frequency;
 
-#ifdef CONFIG_SMP
 	/* cpufreq holds the hotplug lock, so we are safe from here on */
 	cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
-#endif
 
 	for_each_cpu_mask(j, online_policy_cpus) {
 		freqs.cpu = j;
Index: linux-2.6.16-rc6-mm2/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
===================================================================
--- linux-2.6.16-rc6-mm2.orig/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c	2006-03-19 11:15:05.000000000 +1100
+++ linux-2.6.16-rc6-mm2/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c	2006-03-19 12:45:42.000000000 +1100
@@ -652,10 +652,8 @@ static int centrino_target (struct cpufr
 		return -EINVAL;
 	}
 
-#ifdef CONFIG_SMP
 	/* cpufreq holds the hotplug lock, so we are safe from here on */
 	cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
-#endif
 
 	saved_mask = current->cpus_allowed;
 	first_cpu = 1;
Index: linux-2.6.16-rc6-mm2/kernel/sched.c
===================================================================
--- linux-2.6.16-rc6-mm2.orig/kernel/sched.c	2006-03-19 13:25:25.000000000 +1100
+++ linux-2.6.16-rc6-mm2/kernel/sched.c	2006-03-19 13:25:36.000000000 +1100
@@ -6366,6 +6366,10 @@ void __init sched_init_smp(void)
 	init_sched_domain_sysctl();
 }
 #else
+/* bitmap of online cpus */
+cpumask_t cpu_online_map __read_mostly;
+EXPORT_SYMBOL(cpu_online_map);
+
 void __init sched_init_smp(void)
 {
 }

Re: 2.6.16-rc6-mm2 uninitialized online_policy_cpus.bits[0]

[Venkatesh Pallipadi]

On Sat, Mar 18, 2006 at 05:35:12PM -0800, Andrew Morton wrote:
> Con Kolivas <kernel@kolivas.org> wrote:
> >
> > Wonder if this is related to rc6's oops?
> >  gcc 4.0.3
> > 
> >    CC [M]  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.o
> >  arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c: In function 
> >  'centrino_target':
> >  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is used 
> >  uninitialized in this function
> >    CC [M]  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.o
> >  arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c: In function 
> >  'acpi_cpufreq_target':
> >  include/linux/bitmap.h:170: warning: 'online_policy_cpus.bits[0]' is used 
> >  uninitialized in this function
> 
> Well conceivably.  That warning is a consequence of my quick hack to make
> the ACPI tree compile on uniprocessor.
> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.16-rc6/2.6.16-rc6-mm2/broken-out/git-acpi-up-fix.patch
> 
> My patch is, as the compiler points out, wrong.
> 
> I've sent that patch two or three times to the APCI maintainers, to the
> ACPI mailing list and to the author of the original buggy patch.  The
> response thus far has been dead silence.
> 
> IOW, despite my efforts, the ACPI tree has been in a non-compiling state on
> uniprocessor since February 11.
> 
> This is pathetic.  People are trying to get things done here and ACPI is
> getting in the way.  


Oops. Sorry. It is me who dropped the ball here. I somehow assumed that
your original patch fixed the issue and didn't look at the actual code.

Here is the patch against mm.

Thanks,
Venki



Fix the UP build breakage in acpi-cpufreq and speedstep-centrino due to
previous p-state software coordination patch.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>

diff -purN linux-2.6.15/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c linux-2.6.15-new/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c
--- linux-2.6.15/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c	2006-03-18 19:41:25.000000000 -0800
+++ linux-2.6.15-new/arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c	2006-03-18 19:48:10.000000000 -0800
@@ -225,9 +225,11 @@ acpi_cpufreq_target (
 	freqs.old = data->freq_table[cur_state].frequency;
 	freqs.new = data->freq_table[next_state].frequency;
 
-#ifdef CONFIG_SMP
+#ifdef CONFIG_HOTPLUG_CPU
 	/* cpufreq holds the hotplug lock, so we are safe from here on */
 	cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
+#else
+	online_policy_cpus = policy->cpus;
 #endif
 
 	for_each_cpu_mask(j, online_policy_cpus) {
diff -purN linux-2.6.15/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c linux-2.6.15-new/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
--- linux-2.6.15/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c	2006-03-18 19:41:25.000000000 -0800
+++ linux-2.6.15-new/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c	2006-03-18 19:47:46.000000000 -0800
@@ -652,9 +652,11 @@ static int centrino_target (struct cpufr
 		return -EINVAL;
 	}
 
-#ifdef CONFIG_SMP
+#ifdef CONFIG_HOTPLUG_CPU
 	/* cpufreq holds the hotplug lock, so we are safe from here on */
 	cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
+#else
+	online_policy_cpus = policy->cpus;
 #endif
 
 	saved_mask = current->cpus_allowed;