aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--arch/x86/include/asm/mce.h1
-rw-r--r--arch/x86/kernel/cpu/mcheck/mce_64.c33
2 files changed, 18 insertions, 16 deletions
diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 563933e06a35..4f8c199584e7 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -137,6 +137,7 @@ DECLARE_PER_CPU(mce_banks_t, mce_poll_banks);
137enum mcp_flags { 137enum mcp_flags {
138 MCP_TIMESTAMP = (1 << 0), /* log time stamp */ 138 MCP_TIMESTAMP = (1 << 0), /* log time stamp */
139 MCP_UC = (1 << 1), /* log uncorrected errors */ 139 MCP_UC = (1 << 1), /* log uncorrected errors */
140 MCP_DONTLOG = (1 << 2), /* only clear, don't log */
140}; 141};
141extern void machine_check_poll(enum mcp_flags flags, mce_banks_t *b); 142extern void machine_check_poll(enum mcp_flags flags, mce_banks_t *b);
142 143
diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c
index 863f89568b1a..6fb0b359d2a5 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -239,9 +239,10 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b)
239 * Don't get the IP here because it's unlikely to 239 * Don't get the IP here because it's unlikely to
240 * have anything to do with the actual error location. 240 * have anything to do with the actual error location.
241 */ 241 */
242 242 if (!(flags & MCP_DONTLOG)) {
243 mce_log(&m); 243 mce_log(&m);
244 add_taint(TAINT_MACHINE_CHECK); 244 add_taint(TAINT_MACHINE_CHECK);
245 }
245 246
246 /* 247 /*
247 * Clear state for this bank. 248 * Clear state for this bank.
@@ -452,13 +453,14 @@ void mce_log_therm_throt_event(__u64 status)
452 */ 453 */
453 454
454static int check_interval = 5 * 60; /* 5 minutes */ 455static int check_interval = 5 * 60; /* 5 minutes */
455static int next_interval; /* in jiffies */ 456static DEFINE_PER_CPU(int, next_interval); /* in jiffies */
456static void mcheck_timer(unsigned long); 457static void mcheck_timer(unsigned long);
457static DEFINE_PER_CPU(struct timer_list, mce_timer); 458static DEFINE_PER_CPU(struct timer_list, mce_timer);
458 459
459static void mcheck_timer(unsigned long data) 460static void mcheck_timer(unsigned long data)
460{ 461{
461 struct timer_list *t = &per_cpu(mce_timer, data); 462 struct timer_list *t = &per_cpu(mce_timer, data);
463 int *n;
462 464
463 WARN_ON(smp_processor_id() != data); 465 WARN_ON(smp_processor_id() != data);
464 466
@@ -470,14 +472,14 @@ static void mcheck_timer(unsigned long data)
470 * Alert userspace if needed. If we logged an MCE, reduce the 472 * Alert userspace if needed. If we logged an MCE, reduce the
471 * polling interval, otherwise increase the polling interval. 473 * polling interval, otherwise increase the polling interval.
472 */ 474 */
475 n = &__get_cpu_var(next_interval);
473 if (mce_notify_user()) { 476 if (mce_notify_user()) {
474 next_interval = max(next_interval/2, HZ/100); 477 *n = max(*n/2, HZ/100);
475 } else { 478 } else {
476 next_interval = min(next_interval * 2, 479 *n = min(*n*2, (int)round_jiffies_relative(check_interval*HZ));
477 (int)round_jiffies_relative(check_interval*HZ));
478 } 480 }
479 481
480 t->expires = jiffies + next_interval; 482 t->expires = jiffies + *n;
481 add_timer(t); 483 add_timer(t);
482} 484}
483 485
@@ -584,7 +586,7 @@ static void mce_init(void *dummy)
584 * Log the machine checks left over from the previous reset. 586 * Log the machine checks left over from the previous reset.
585 */ 587 */
586 bitmap_fill(all_banks, MAX_NR_BANKS); 588 bitmap_fill(all_banks, MAX_NR_BANKS);
587 machine_check_poll(MCP_UC, &all_banks); 589 machine_check_poll(MCP_UC|(!mce_bootlog ? MCP_DONTLOG : 0), &all_banks);
588 590
589 set_in_cr4(X86_CR4_MCE); 591 set_in_cr4(X86_CR4_MCE);
590 592
@@ -632,14 +634,13 @@ static void mce_cpu_features(struct cpuinfo_x86 *c)
632static void mce_init_timer(void) 634static void mce_init_timer(void)
633{ 635{
634 struct timer_list *t = &__get_cpu_var(mce_timer); 636 struct timer_list *t = &__get_cpu_var(mce_timer);
637 int *n = &__get_cpu_var(next_interval);
635 638
636 /* data race harmless because everyone sets to the same value */ 639 *n = check_interval * HZ;
637 if (!next_interval) 640 if (!*n)
638 next_interval = check_interval * HZ;
639 if (!next_interval)
640 return; 641 return;
641 setup_timer(t, mcheck_timer, smp_processor_id()); 642 setup_timer(t, mcheck_timer, smp_processor_id());
642 t->expires = round_jiffies(jiffies + next_interval); 643 t->expires = round_jiffies(jiffies + *n);
643 add_timer(t); 644 add_timer(t);
644} 645}
645 646
@@ -907,7 +908,6 @@ static void mce_cpu_restart(void *data)
907/* Reinit MCEs after user configuration changes */ 908/* Reinit MCEs after user configuration changes */
908static void mce_restart(void) 909static void mce_restart(void)
909{ 910{
910 next_interval = check_interval * HZ;
911 on_each_cpu(mce_cpu_restart, NULL, 1); 911 on_each_cpu(mce_cpu_restart, NULL, 1);
912} 912}
913 913
@@ -1110,7 +1110,8 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb,
1110 break; 1110 break;
1111 case CPU_DOWN_FAILED: 1111 case CPU_DOWN_FAILED:
1112 case CPU_DOWN_FAILED_FROZEN: 1112 case CPU_DOWN_FAILED_FROZEN:
1113 t->expires = round_jiffies(jiffies + next_interval); 1113 t->expires = round_jiffies(jiffies +
1114 __get_cpu_var(next_interval));
1114 add_timer_on(t, cpu); 1115 add_timer_on(t, cpu);
1115 smp_call_function_single(cpu, mce_reenable_cpu, &action, 1); 1116 smp_call_function_single(cpu, mce_reenable_cpu, &action, 1);
1116 break; 1117 break;