-stable review patch. If anyone has any objections, please let us know.
------------------
From: Ravikiran G Thirumalai <[email protected]>
2.6.19 stopped booting (or booted based on build/config) on our x86_64
systems due to a bug introduced in 2.6.19. check_nmi_watchdog schedules an
IPI on all cpus to busy wait on a flag, but fails to set the busywait flag
if NMI functionality is disabled.
This causes the secondary cpus to spin in an endless loop, causing the
kernel bootup to hang.
Depending upon the build, the busywait flag got overwritten (stack
variable) and caused the kernel to bootup on certain builds. Following
patch fixes the bug by setting the busywait flag before returning from
check_nmi_watchdog.
I guess using a stack variable is not good here as the calling function
could potentially return while the busy wait loop is still spinning on the
flag. I would think this is a good candidate for 2.6.19 stable as well.
[[email protected]: cleanups]
Signed-off-by: Ravikiran Thirumalai <[email protected]>
Signed-off-by: Shai Fultheim <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Chris Wright <[email protected]>
---
arch/x86_64/kernel/nmi.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--- linux-2.6.19.orig/arch/x86_64/kernel/nmi.c
+++ linux-2.6.19/arch/x86_64/kernel/nmi.c
@@ -212,7 +212,7 @@ static __init void nmi_cpu_busy(void *da
int __init check_nmi_watchdog (void)
{
- volatile int endflag = 0;
+ static int __initdata endflag;
int *counts;
int cpu;
@@ -253,6 +253,7 @@ int __init check_nmi_watchdog (void)
if (!atomic_read(&nmi_active)) {
kfree(counts);
atomic_set(&nmi_active, -1);
+ endflag = 1;
return -1;
}
endflag = 1;
--
* Chris Wright ([email protected]) wrote:
> -stable review patch. If anyone has any objections, please let us know.
> ------------------
replaced with upstream version, which is slightly changed by Andi.
--
From: Ravikiran G Thirumalai <[email protected]>
2.6.19 stopped booting (or booted based on build/config) on our x86_64
systems due to a bug introduced in 2.6.19. check_nmi_watchdog schedules an
IPI on all cpus to busy wait on a flag, but fails to set the busywait
flag if NMI functionality is disabled. This causes the secondary cpus
to spin in an endless loop, causing the kernel bootup to hang.
Depending upon the build, the busywait flag got overwritten (stack variable)
and caused the kernel to bootup on certain builds. Following patch fixes
the bug by setting the busywait flag before returning from check_nmi_watchdog.
I guess using a stack variable is not good here as the calling function could
potentially return while the busy wait loop is still spinning on the flag.
AK: I redid the patch significantly to be cleaner
Signed-off-by: Ravikiran Thirumalai <[email protected]>
Signed-off-by: Shai Fultheim <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Signed-off-by: Chris Wright <[email protected]>
---
arch/i386/kernel/nmi.c | 8 ++++----
arch/x86_64/kernel/nmi.c | 9 +++++----
2 files changed, 9 insertions(+), 8 deletions(-)
--- linux-2.6.19.orig/arch/i386/kernel/nmi.c
+++ linux-2.6.19/arch/i386/kernel/nmi.c
@@ -192,6 +192,8 @@ static __cpuinit inline int nmi_known_cp
return 0;
}
+static int endflag __initdata = 0;
+
#ifdef CONFIG_SMP
/* The performance counters used by NMI_LOCAL_APIC don't trigger when
* the CPU is idle. To make sure the NMI watchdog really ticks on all
@@ -199,7 +201,6 @@ static __cpuinit inline int nmi_known_cp
*/
static __init void nmi_cpu_busy(void *data)
{
- volatile int *endflag = data;
local_irq_enable_in_hardirq();
/* Intentionally don't use cpu_relax here. This is
to make sure that the performance counter really ticks,
@@ -207,14 +208,13 @@ static __init void nmi_cpu_busy(void *da
pause instruction. On a real HT machine this is fine because
all other CPUs are busy with "useless" delay loops and don't
care if they get somewhat less cycles. */
- while (*endflag == 0)
- barrier();
+ while (endflag == 0)
+ mb();
}
#endif
static int __init check_nmi_watchdog(void)
{
- volatile int endflag = 0;
unsigned int *prev_nmi_count;
int cpu;
--- linux-2.6.19.orig/arch/x86_64/kernel/nmi.c
+++ linux-2.6.19/arch/x86_64/kernel/nmi.c
@@ -190,6 +190,8 @@ void nmi_watchdog_default(void)
nmi_watchdog = NMI_IO_APIC;
}
+static int endflag __initdata = 0;
+
#ifdef CONFIG_SMP
/* The performance counters used by NMI_LOCAL_APIC don't trigger when
* the CPU is idle. To make sure the NMI watchdog really ticks on all
@@ -197,7 +199,6 @@ void nmi_watchdog_default(void)
*/
static __init void nmi_cpu_busy(void *data)
{
- volatile int *endflag = data;
local_irq_enable_in_hardirq();
/* Intentionally don't use cpu_relax here. This is
to make sure that the performance counter really ticks,
@@ -205,14 +206,13 @@ static __init void nmi_cpu_busy(void *da
pause instruction. On a real HT machine this is fine because
all other CPUs are busy with "useless" delay loops and don't
care if they get somewhat less cycles. */
- while (*endflag == 0)
- barrier();
+ while (endflag == 0)
+ mb();
}
#endif
int __init check_nmi_watchdog (void)
{
- volatile int endflag = 0;
int *counts;
int cpu;
@@ -253,6 +253,7 @@ int __init check_nmi_watchdog (void)
if (!atomic_read(&nmi_active)) {
kfree(counts);
atomic_set(&nmi_active, -1);
+ endflag = 1;
return -1;
}
endflag = 1;