[PATCH for 2.6.23] [0/3] More last minute bug fixes for x86

[PATCH for 2.6.23] [0/3] More last minute bug fixes for x86

[Andi Kleen]

Linus,

Some more x86 bug fixes for .23rc6

- Fix the infamous "5 minutes" vdso problem on 32bit HPET with new glibc
- Error handling fix in cpuid4 handling (Jan Beulich) 
- Fix leak of ../kernel out of build dir (Ben LaHaise & Sam Ravnborg) 

Please merge.

-Andi

[PATCH for 2.6.23] [1/3] x86_64: Add missing mask operation to vdso

[Andi Kleen]

vdso vgetns() didn't mask the time source offset calculation, which could
lead to time problems with 32bit HPET. Add the masking.

Thanks to Chuck Ebbert for tracking down.

Cc: cebbert@redhat.com

Signed-off-by: Andi Kleen <ak@suse.de>

Index: linux/arch/x86_64/vdso/vclock_gettime.c
===================================================================
--- linux.orig/arch/x86_64/vdso/vclock_gettime.c
+++ linux/arch/x86_64/vdso/vclock_gettime.c
@@ -34,10 +34,11 @@ static long vdso_fallback_gettime(long c
 
 static inline long vgetns(void)
 {
+	long v;
 	cycles_t (*vread)(void);
 	vread = gtod->clock.vread;
-	return ((vread() - gtod->clock.cycle_last) * gtod->clock.mult) >>
-		gtod->clock.shift;
+	v = (vread() - gtod->clock.cycle_last) & gtod->clock.mask;
+	return (v * gtod->clock.mult) >> gtod->clock.shift;
 }
 
 static noinline int do_realtime(struct timespec *ts)

[PATCH for 2.6.23] [2/3] x86_64: Prevent doing anything from cache_remove_dev() when info setup failed.

[Andi Kleen]

AK: Removed the unlikelies because gcc heuristics default to unlikely for test
AK: == NULL and for negative returns.

Cc: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>

Index: linux/arch/i386/kernel/cpu/intel_cacheinfo.c
===================================================================
--- linux.orig/arch/i386/kernel/cpu/intel_cacheinfo.c
+++ linux/arch/i386/kernel/cpu/intel_cacheinfo.c
@@ -515,7 +515,7 @@ static int __cpuinit detect_cache_attrib
 
 	cpuid4_info[cpu] = kzalloc(
 	    sizeof(struct _cpuid4_info) * num_cache_leaves, GFP_KERNEL);
-	if (unlikely(cpuid4_info[cpu] == NULL))
+	if (cpuid4_info[cpu] == NULL)
 		return -ENOMEM;
 
 	oldmask = current->cpus_allowed;
@@ -748,6 +748,8 @@ static void __cpuinit cache_remove_dev(s
 	unsigned int cpu = sys_dev->id;
 	unsigned long i;
 
+	if (cpuid4_info[cpu] == NULL)
+		return;
 	for (i = 0; i < num_cache_leaves; i++) {
 		cache_remove_shared_cpu_map(cpu, i);
 		kobject_unregister(&(INDEX_KOBJECT_PTR(cpu,i)->kobj));

[PATCH for 2.6.23] [3/3] i386: Fix leak of ../kernel from top level

[Andi Kleen]

Fix a compile error when the directory above the kernel source contains
a file named "kernel". Originally from Ben La Haise, modified based
on feedback from Sam Ravnborg

Cc: sam@ravnborg.org
Cc: bcrl@kvack.org

Signed-off-by: Andi Kleen <ak@suse.de>

Index: linux/arch/i386/mach-generic/Makefile
===================================================================
--- linux.orig/arch/i386/mach-generic/Makefile
+++ linux/arch/i386/mach-generic/Makefile
@@ -2,6 +2,6 @@
 # Makefile for the generic architecture
 #
 
-EXTRA_CFLAGS	+= -I../kernel
+EXTRA_CFLAGS	:= -Iarch/i386/kernel
 
 obj-y				:= probe.o summit.o bigsmp.o es7000.o default.o ../mach-es7000/
Index: linux/arch/i386/mach-voyager/Makefile
===================================================================
--- linux.orig/arch/i386/mach-voyager/Makefile
+++ linux/arch/i386/mach-voyager/Makefile
@@ -2,7 +2,7 @@
 # Makefile for the linux kernel.
 #
 
-EXTRA_CFLAGS	+= -I../kernel
+EXTRA_CFLAGS	:= -Iarch/i386/kernel
 obj-y			:= setup.o voyager_basic.o voyager_thread.o
 
 obj-$(CONFIG_SMP)	+= voyager_smp.o voyager_cat.o

Re: [PATCH for 2.6.23] [2/3] x86_64: Prevent doing anything from cache_remove_dev() when info setup failed.

[Andi Kleen]

Andi Kleen <ak@suse.de> writes:

> AK: Removed the unlikelies because gcc heuristics default to unlikely for tes
> AK: == NULL and for negative returns.

This patch was actually from Jan Beulich, but somehow the attribution
got dropped. Sorry for that.

-Andi