summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorPaul Mackerras <paulus@samba.org>2010-04-13 20:46:04 +0000
committerGreg Kroah-Hartman <gregkh@suse.de>2010-05-26 14:29:16 -0700
commit73ef533de7533fcf514403125a0d4747d3390317 (patch)
tree83865693cb3792ac3b358733cf8203b6a820eeee
parentfa157eec79e78a875790df6faf53b3501c33a462 (diff)
downloadlwn-73ef533de7533fcf514403125a0d4747d3390317.tar.gz
lwn-73ef533de7533fcf514403125a0d4747d3390317.zip
powerpc/perf_event: Fix oops due to perf_event_do_pending call
commit 0fe1ac48bef018bed896307cd12f6ca9b5e704ab upstream. Anton Blanchard found that large POWER systems would occasionally crash in the exception exit path when profiling with perf_events. The symptom was that an interrupt would occur late in the exit path when the MSR[RI] (recoverable interrupt) bit was clear. Interrupts should be hard-disabled at this point but they were enabled. Because the interrupt was not recoverable the system panicked. The reason is that the exception exit path was calling perf_event_do_pending after hard-disabling interrupts, and perf_event_do_pending will re-enable interrupts. The simplest and cleanest fix for this is to use the same mechanism that 32-bit powerpc does, namely to cause a self-IPI by setting the decrementer to 1. This means we can remove the tests in the exception exit path and raw_local_irq_restore. This also makes sure that the call to perf_event_do_pending from timer_interrupt() happens within irq_enter/irq_exit. (Note that calling perf_event_do_pending from timer_interrupt does not mean that there is a possible 1/HZ latency; setting the decrementer to 1 ensures that the timer interrupt will happen immediately, i.e. within one timebase tick, which is a few nanoseconds or 10s of nanoseconds.) Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
-rw-r--r--arch/powerpc/include/asm/hw_irq.h38
-rw-r--r--arch/powerpc/kernel/asm-offsets.c1
-rw-r--r--arch/powerpc/kernel/entry_64.S9
-rw-r--r--arch/powerpc/kernel/irq.c6
-rw-r--r--arch/powerpc/kernel/time.c60
5 files changed, 48 insertions, 66 deletions
diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h
index abbc2aaaced5..c1de3c9a1d65 100644
--- a/arch/powerpc/include/asm/hw_irq.h
+++ b/arch/powerpc/include/asm/hw_irq.h
@@ -135,43 +135,5 @@ static inline int irqs_disabled_flags(unsigned long flags)
*/
struct irq_chip;
-#ifdef CONFIG_PERF_EVENTS
-
-#ifdef CONFIG_PPC64
-static inline unsigned long test_perf_event_pending(void)
-{
- unsigned long x;
-
- asm volatile("lbz %0,%1(13)"
- : "=r" (x)
- : "i" (offsetof(struct paca_struct, perf_event_pending)));
- return x;
-}
-
-static inline void set_perf_event_pending(void)
-{
- asm volatile("stb %0,%1(13)" : :
- "r" (1),
- "i" (offsetof(struct paca_struct, perf_event_pending)));
-}
-
-static inline void clear_perf_event_pending(void)
-{
- asm volatile("stb %0,%1(13)" : :
- "r" (0),
- "i" (offsetof(struct paca_struct, perf_event_pending)));
-}
-#endif /* CONFIG_PPC64 */
-
-#else /* CONFIG_PERF_EVENTS */
-
-static inline unsigned long test_perf_event_pending(void)
-{
- return 0;
-}
-
-static inline void clear_perf_event_pending(void) {}
-#endif /* CONFIG_PERF_EVENTS */
-
#endif /* __KERNEL__ */
#endif /* _ASM_POWERPC_HW_IRQ_H */
diff --git a/arch/powerpc/kernel/asm-offsets.c b/arch/powerpc/kernel/asm-offsets.c
index 0812b0f414bb..692c056566c7 100644
--- a/arch/powerpc/kernel/asm-offsets.c
+++ b/arch/powerpc/kernel/asm-offsets.c
@@ -133,7 +133,6 @@ int main(void)
DEFINE(PACAKMSR, offsetof(struct paca_struct, kernel_msr));
DEFINE(PACASOFTIRQEN, offsetof(struct paca_struct, soft_enabled));
DEFINE(PACAHARDIRQEN, offsetof(struct paca_struct, hard_enabled));
- DEFINE(PACAPERFPEND, offsetof(struct paca_struct, perf_event_pending));
DEFINE(PACACONTEXTID, offsetof(struct paca_struct, context.id));
#ifdef CONFIG_PPC_MM_SLICES
DEFINE(PACALOWSLICESPSIZE, offsetof(struct paca_struct,
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 9763267e38b4..917cebc2f262 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -556,15 +556,6 @@ ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
2:
TRACE_AND_RESTORE_IRQ(r5);
-#ifdef CONFIG_PERF_EVENTS
- /* check paca->perf_event_pending if we're enabling ints */
- lbz r3,PACAPERFPEND(r13)
- and. r3,r3,r5
- beq 27f
- bl .perf_event_do_pending
-27:
-#endif /* CONFIG_PERF_EVENTS */
-
/* extract EE bit and use it to restore paca->hard_enabled */
ld r3,_MSR(r1)
rldicl r4,r3,49,63 /* r0 = (r3 >> 15) & 1 */
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index e5d121177984..8564a412e7a6 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -53,7 +53,6 @@
#include <linux/bootmem.h>
#include <linux/pci.h>
#include <linux/debugfs.h>
-#include <linux/perf_event.h>
#include <asm/uaccess.h>
#include <asm/system.h>
@@ -138,11 +137,6 @@ notrace void raw_local_irq_restore(unsigned long en)
}
#endif /* CONFIG_PPC_STD_MMU_64 */
- if (test_perf_event_pending()) {
- clear_perf_event_pending();
- perf_event_do_pending();
- }
-
/*
* if (get_paca()->hard_enabled) return;
* But again we need to take care that gcc gets hard_enabled directly
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index a136a11c490d..02aab7f90295 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -530,25 +530,60 @@ void __init iSeries_time_init_early(void)
}
#endif /* CONFIG_PPC_ISERIES */
-#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_PPC32)
-DEFINE_PER_CPU(u8, perf_event_pending);
+#ifdef CONFIG_PERF_EVENTS
-void set_perf_event_pending(void)
+/*
+ * 64-bit uses a byte in the PACA, 32-bit uses a per-cpu variable...
+ */
+#ifdef CONFIG_PPC64
+static inline unsigned long test_perf_event_pending(void)
{
- get_cpu_var(perf_event_pending) = 1;
- set_dec(1);
- put_cpu_var(perf_event_pending);
+ unsigned long x;
+
+ asm volatile("lbz %0,%1(13)"
+ : "=r" (x)
+ : "i" (offsetof(struct paca_struct, perf_event_pending)));
+ return x;
}
+static inline void set_perf_event_pending_flag(void)
+{
+ asm volatile("stb %0,%1(13)" : :
+ "r" (1),
+ "i" (offsetof(struct paca_struct, perf_event_pending)));
+}
+
+static inline void clear_perf_event_pending(void)
+{
+ asm volatile("stb %0,%1(13)" : :
+ "r" (0),
+ "i" (offsetof(struct paca_struct, perf_event_pending)));
+}
+
+#else /* 32-bit */
+
+DEFINE_PER_CPU(u8, perf_event_pending);
+
+#define set_perf_event_pending_flag() __get_cpu_var(perf_event_pending) = 1
#define test_perf_event_pending() __get_cpu_var(perf_event_pending)
#define clear_perf_event_pending() __get_cpu_var(perf_event_pending) = 0
-#else /* CONFIG_PERF_EVENTS && CONFIG_PPC32 */
+#endif /* 32 vs 64 bit */
+
+void set_perf_event_pending(void)
+{
+ preempt_disable();
+ set_perf_event_pending_flag();
+ set_dec(1);
+ preempt_enable();
+}
+
+#else /* CONFIG_PERF_EVENTS */
#define test_perf_event_pending() 0
#define clear_perf_event_pending()
-#endif /* CONFIG_PERF_EVENTS && CONFIG_PPC32 */
+#endif /* CONFIG_PERF_EVENTS */
/*
* For iSeries shared processors, we have to let the hypervisor
@@ -576,10 +611,6 @@ void timer_interrupt(struct pt_regs * regs)
set_dec(DECREMENTER_MAX);
#ifdef CONFIG_PPC32
- if (test_perf_event_pending()) {
- clear_perf_event_pending();
- perf_event_do_pending();
- }
if (atomic_read(&ppc_n_lost_interrupts) != 0)
do_IRQ(regs);
#endif
@@ -597,6 +628,11 @@ void timer_interrupt(struct pt_regs * regs)
calculate_steal_time();
+ if (test_perf_event_pending()) {
+ clear_perf_event_pending();
+ perf_event_do_pending();
+ }
+
#ifdef CONFIG_PPC_ISERIES
if (firmware_has_feature(FW_FEATURE_ISERIES))
get_lppaca()->int_dword.fields.decr_int = 0;