Merge branch 'x86/mce2' into x86/core

8 files changed, 666 insertions, 169 deletions

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 6907b8e85d5..4c80f155743 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -414,9 +414,17 @@ void __init alternative_instructions(void)
 	   that might execute the to be patched code.
 	   Other CPUs are not running. */
 	stop_nmi();
-#ifdef CONFIG_X86_MCE
-	stop_mce();
-#endif
+
+	/*
+	 * Don't stop machine check exceptions while patching.
+	 * MCEs only happen when something got corrupted and in this
+	 * case we must do something about the corruption.
+	 * Ignoring it is worse than a unlikely patching race.
+	 * Also machine checks tend to be broadcast and if one CPU
+	 * goes into machine check the others follow quickly, so we don't
+	 * expect a machine check to cause undue problems during to code
+	 * patching.
+	 */
 
 	apply_alternatives(__alt_instructions, __alt_instructions_end);
 
@@ -456,9 +464,6 @@ void __init alternative_instructions(void)
 				(unsigned long)__smp_locks_end);
 
 	restart_nmi();
-#ifdef CONFIG_X86_MCE
-	restart_mce();
-#endif
 }
 
 /**
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index f9cecdfd05c..30909a258d0 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -46,6 +46,7 @@
 #include <asm/idle.h>
 #include <asm/mtrr.h>
 #include <asm/smp.h>
+#include <asm/mce.h>
 
 unsigned int num_processors;
 
@@ -842,6 +843,14 @@ void clear_local_APIC(void)
 		apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
 	}
 #endif
+#ifdef CONFIG_X86_MCE_INTEL
+	if (maxlvt >= 6) {
+		v = apic_read(APIC_LVTCMCI);
+		if (!(v & APIC_LVT_MASKED))
+			apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
+	}
+#endif
+
 	/*
 	 * Clean APIC state for other OSs:
 	 */
@@ -1241,6 +1250,12 @@ void __cpuinit setup_local_APIC(void)
 	apic_write(APIC_LVT1, value);
 
 	preempt_enable();
+
+#ifdef CONFIG_X86_MCE_INTEL
+	/* Recheck CMCI information after local APIC is up on CPU #0 */
+	if (smp_processor_id() == 0)
+		cmci_recheck();
+#endif
 }
 
 void __cpuinit end_local_APIC_setup(void)
diff --git a/arch/x86/kernel/cpu/mcheck/Makefile b/arch/x86/kernel/cpu/mcheck/Makefile
index d7d2323bbb6..b2f89829bbe 100644
--- a/arch/x86/kernel/cpu/mcheck/Makefile
+++ b/arch/x86/kernel/cpu/mcheck/Makefile
@@ -4,3 +4,4 @@ obj-$(CONFIG_X86_32)		+= k7.o p4.o p5.o p6.o winchip.o
 obj-$(CONFIG_X86_MCE_INTEL)	+= mce_intel_64.o
 obj-$(CONFIG_X86_MCE_AMD)	+= mce_amd_64.o
 obj-$(CONFIG_X86_MCE_NONFATAL)	+= non-fatal.o
+obj-$(CONFIG_X86_MCE_THRESHOLD) += threshold.o
diff --git a/arch/x86/kernel/cpu/mcheck/mce_32.c b/arch/x86/kernel/cpu/mcheck/mce_32.c
index dfaebce3633..3552119b091 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_32.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_32.c
@@ -60,20 +60,6 @@ void mcheck_init(struct cpuinfo_x86 *c)
 	}
 }
 
-static unsigned long old_cr4 __initdata;
-
-void __init stop_mce(void)
-{
-	old_cr4 = read_cr4();
-	clear_in_cr4(X86_CR4_MCE);
-}
-
-void __init restart_mce(void)
-{
-	if (old_cr4 & X86_CR4_MCE)
-		set_in_cr4(X86_CR4_MCE);
-}
-
 static int __init mcheck_disable(char *str)
 {
 	mce_disabled = 1;
diff --git a/arch/x86/kernel/cpu/mcheck/mce_64.c b/arch/x86/kernel/cpu/mcheck/mce_64.c
index fe79985ce0f..bfbd5323a63 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_64.c
@@ -3,6 +3,8 @@
  * K8 parts Copyright 2002,2003 Andi Kleen, SuSE Labs.
  * Rest from unknown author(s).
  * 2004 Andi Kleen. Rewrote most of it.
+ * Copyright 2008 Intel Corporation
+ * Author: Andi Kleen
  */
 
 #include <linux/init.h>
@@ -24,6 +26,9 @@
 #include <linux/ctype.h>
 #include <linux/kmod.h>
 #include <linux/kdebug.h>
+#include <linux/kobject.h>
+#include <linux/sysfs.h>
+#include <linux/ratelimit.h>
 #include <asm/processor.h>
 #include <asm/msr.h>
 #include <asm/mce.h>
@@ -32,7 +37,6 @@
 #include <asm/idle.h>
 
 #define MISC_MCELOG_MINOR 227
-#define NR_SYSFS_BANKS 6
 
 atomic_t mce_entry;
 
@@ -47,7 +51,7 @@ static int mce_dont_init;
  */
 static int tolerant = 1;
 static int banks;
-static unsigned long bank[NR_SYSFS_BANKS] = { [0 ... NR_SYSFS_BANKS-1] = ~0UL };
+static u64 *bank;
 static unsigned long notify_user;
 static int rip_msr;
 static int mce_bootlog = -1;
@@ -58,6 +62,19 @@ static char *trigger_argv[2] = { trigger, NULL };
 
 static DECLARE_WAIT_QUEUE_HEAD(mce_wait);
 
+/* MCA banks polled by the period polling timer for corrected events */
+DEFINE_PER_CPU(mce_banks_t, mce_poll_banks) = {
+	[0 ... BITS_TO_LONGS(MAX_NR_BANKS)-1] = ~0UL
+};
+
+/* Do initial initialization of a struct mce */
+void mce_setup(struct mce *m)
+{
+	memset(m, 0, sizeof(struct mce));
+	m->cpu = smp_processor_id();
+	rdtscll(m->tsc);
+}
+
 /*
  * Lockless MCE logging infrastructure.
  * This avoids deadlocks on printk locks without having to break locks. Also
@@ -119,11 +136,11 @@ static void print_mce(struct mce *m)
 			print_symbol("{%s}", m->ip);
 		printk("\n");
 	}
-	printk(KERN_EMERG "TSC %Lx ", m->tsc);
+	printk(KERN_EMERG "TSC %llx ", m->tsc);
 	if (m->addr)
-		printk("ADDR %Lx ", m->addr);
+		printk("ADDR %llx ", m->addr);
 	if (m->misc)
-		printk("MISC %Lx ", m->misc);
+		printk("MISC %llx ", m->misc);
 	printk("\n");
 	printk(KERN_EMERG "This is not a software problem!\n");
 	printk(KERN_EMERG "Run through mcelog --ascii to decode "
@@ -149,8 +166,10 @@ static void mce_panic(char *msg, struct mce *backup, unsigned long start)
 	panic(msg);
 }
 
-static int mce_available(struct cpuinfo_x86 *c)
+int mce_available(struct cpuinfo_x86 *c)
 {
+	if (mce_dont_init)
+		return 0;
 	return cpu_has(c, X86_FEATURE_MCE) && cpu_has(c, X86_FEATURE_MCA);
 }
 
@@ -172,7 +191,77 @@ static inline void mce_get_rip(struct mce *m, struct pt_regs *regs)
 }
 
 /*
- * The actual machine check handler
+ * Poll for corrected events or events that happened before reset.
+ * Those are just logged through /dev/mcelog.
+ *
+ * This is executed in standard interrupt context.
+ */
+void machine_check_poll(enum mcp_flags flags, mce_banks_t *b)
+{
+	struct mce m;
+	int i;
+
+	mce_setup(&m);
+
+	rdmsrl(MSR_IA32_MCG_STATUS, m.mcgstatus);
+	for (i = 0; i < banks; i++) {
+		if (!bank[i] || !test_bit(i, *b))
+			continue;
+
+		m.misc = 0;
+		m.addr = 0;
+		m.bank = i;
+		m.tsc = 0;
+
+		barrier();
+		rdmsrl(MSR_IA32_MC0_STATUS + i*4, m.status);
+		if (!(m.status & MCI_STATUS_VAL))
+			continue;
+
+		/*
+		 * Uncorrected events are handled by the exception handler
+		 * when it is enabled. But when the exception is disabled log
+		 * everything.
+		 *
+		 * TBD do the same check for MCI_STATUS_EN here?
+		 */
+		if ((m.status & MCI_STATUS_UC) && !(flags & MCP_UC))
+			continue;
+
+		if (m.status & MCI_STATUS_MISCV)
+			rdmsrl(MSR_IA32_MC0_MISC + i*4, m.misc);
+		if (m.status & MCI_STATUS_ADDRV)
+			rdmsrl(MSR_IA32_MC0_ADDR + i*4, m.addr);
+
+		if (!(flags & MCP_TIMESTAMP))
+			m.tsc = 0;
+		/*
+		 * Don't get the IP here because it's unlikely to
+		 * have anything to do with the actual error location.
+		 */
+
+		mce_log(&m);
+		add_taint(TAINT_MACHINE_CHECK);
+
+		/*
+		 * Clear state for this bank.
+		 */
+		wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0);
+	}
+
+	/*
+	 * Don't clear MCG_STATUS here because it's only defined for
+	 * exceptions.
+	 */
+}
+
+/*
+ * The actual machine check handler. This only handles real
+ * exceptions when something got corrupted coming in through int 18.
+ *
+ * This is executed in NMI context not subject to normal locking rules. This
+ * implies that most kernel services cannot be safely used. Don't even
+ * think about putting a printk in there!
  */
 void do_machine_check(struct pt_regs * regs, long error_code)
 {
@@ -190,17 +279,18 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 	 * error.
 	 */
 	int kill_it = 0;
+	DECLARE_BITMAP(toclear, MAX_NR_BANKS);
 
 	atomic_inc(&mce_entry);
 
-	if ((regs
-	     && notify_die(DIE_NMI, "machine check", regs, error_code,
+	if (notify_die(DIE_NMI, "machine check", regs, error_code,
 			   18, SIGKILL) == NOTIFY_STOP)
-	    || !banks)
+		goto out2;
+	if (!banks)
 		goto out2;
 
-	memset(&m, 0, sizeof(struct mce));
-	m.cpu = smp_processor_id();
+	mce_setup(&m);
+
 	rdmsrl(MSR_IA32_MCG_STATUS, m.mcgstatus);
 	/* if the restart IP is not valid, we're done for */
 	if (!(m.mcgstatus & MCG_STATUS_RIPV))
@@ -210,18 +300,32 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 	barrier();
 
 	for (i = 0; i < banks; i++) {
-		if (i < NR_SYSFS_BANKS && !bank[i])
+		__clear_bit(i, toclear);
+		if (!bank[i])
 			continue;
 
 		m.misc = 0;
 		m.addr = 0;
 		m.bank = i;
-		m.tsc = 0;
 
 		rdmsrl(MSR_IA32_MC0_STATUS + i*4, m.status);
 		if ((m.status & MCI_STATUS_VAL) == 0)
 			continue;
 
+		/*
+		 * Non uncorrected errors are handled by machine_check_poll
+		 * Leave them alone.
+		 */
+		if ((m.status & MCI_STATUS_UC) == 0)
+			continue;
+
+		/*
+		 * Set taint even when machine check was not enabled.
+		 */
+		add_taint(TAINT_MACHINE_CHECK);
+
+		__set_bit(i, toclear);
+
 		if (m.status & MCI_STATUS_EN) {
 			/* if PCC was set, there's no way out */
 			no_way_out |= !!(m.status & MCI_STATUS_PCC);
@@ -235,6 +339,12 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 					no_way_out = 1;
 				kill_it = 1;
 			}
+		} else {
+			/*
+			 * Machine check event was not enabled. Clear, but
+			 * ignore.
+			 */
+			continue;
 		}
 
 		if (m.status & MCI_STATUS_MISCV)
@@ -243,10 +353,7 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 			rdmsrl(MSR_IA32_MC0_ADDR + i*4, m.addr);
 
 		mce_get_rip(&m, regs);
-		if (error_code >= 0)
-			rdtscll(m.tsc);
-		if (error_code != -2)
-			mce_log(&m);
+		mce_log(&m);
 
 		/* Did this bank cause the exception? */
 		/* Assume that the bank with uncorrectable errors did it,
@@ -255,14 +362,8 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 			panicm = m;
 			panicm_found = 1;
 		}
-
-		add_taint(TAINT_MACHINE_CHECK);
 	}
 
-	/* Never do anything final in the polling timer */
-	if (!regs)
-		goto out;
-
 	/* If we didn't find an uncorrectable error, pick
 	   the last one (shouldn't happen, just being safe). */
 	if (!panicm_found)
@@ -309,10 +410,11 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 	/* notify userspace ASAP */
 	set_thread_flag(TIF_MCE_NOTIFY);
 
- out:
 	/* the last thing we do is clear state */
-	for (i = 0; i < banks; i++)
-		wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0);
+	for (i = 0; i < banks; i++) {
+		if (test_bit(i, toclear))
+			wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0);
+	}
 	wrmsrl(MSR_IA32_MCG_STATUS, 0);
  out2:
 	atomic_dec(&mce_entry);
@@ -332,15 +434,13 @@ void do_machine_check(struct pt_regs * regs, long error_code)
  * and historically has been the register value of the
  * MSR_IA32_THERMAL_STATUS (Intel) msr.
  */
-void mce_log_therm_throt_event(unsigned int cpu, __u64 status)
+void mce_log_therm_throt_event(__u64 status)
 {
 	struct mce m;
 
-	memset(&m, 0, sizeof(m));
-	m.cpu = cpu;
+	mce_setup(&m);
 	m.bank = MCE_THERMAL_BANK;
 	m.status = status;
-	rdtscll(m.tsc);
 	mce_log(&m);
 }
 #endif /* CONFIG_X86_MCE_INTEL */
@@ -353,18 +453,18 @@ void mce_log_therm_throt_event(unsigned int cpu, __u64 status)
 
 static int check_interval = 5 * 60; /* 5 minutes */
 static int next_interval; /* in jiffies */
-static void mcheck_timer(struct work_struct *work);
-static DECLARE_DELAYED_WORK(mcheck_work, mcheck_timer);
+static void mcheck_timer(unsigned long);
+static DEFINE_PER_CPU(struct timer_list, mce_timer);
 
-static void mcheck_check_cpu(void *info)
+static void mcheck_timer(unsigned long data)
 {
-	if (mce_available(&current_cpu_data))
-		do_machine_check(NULL, 0);
-}
+	struct timer_list *t = &per_cpu(mce_timer, data);
 
-static void mcheck_timer(struct work_struct *work)
-{
-	on_each_cpu(mcheck_check_cpu, NULL, 1);
+	WARN_ON(smp_processor_id() != data);
+
+	if (mce_available(&current_cpu_data))
+		machine_check_poll(MCP_TIMESTAMP,
+				&__get_cpu_var(mce_poll_banks));
 
 	/*
 	 * Alert userspace if needed.  If we logged an MCE, reduce the
@@ -377,31 +477,41 @@ static void mcheck_timer(struct work_struct *work)
 				(int)round_jiffies_relative(check_interval*HZ));
 	}
 
-	schedule_delayed_work(&mcheck_work, next_interval);
+	t->expires = jiffies + next_interval;
+	add_timer(t);
+}
+
+static void mce_do_trigger(struct work_struct *work)
+{
+	call_usermodehelper(trigger, trigger_argv, NULL, UMH_NO_WAIT);
 }
 
+static DECLARE_WORK(mce_trigger_work, mce_do_trigger);
+
 /*
- * This is only called from process context.  This is where we do
- * anything we need to alert userspace about new MCEs.  This is called
- * directly from the poller and also from entry.S and idle, thanks to
- * TIF_MCE_NOTIFY.
+ * Notify the user(s) about new machine check events.
+ * Can be called from interrupt context, but not from machine check/NMI
+ * context.
  */
 int mce_notify_user(void)
 {
+	/* Not more than two messages every minute */
+	static DEFINE_RATELIMIT_STATE(ratelimit, 60*HZ, 2);
+
 	clear_thread_flag(TIF_MCE_NOTIFY);
 	if (test_and_clear_bit(0, &notify_user)) {
-		static unsigned long last_print;
-		unsigned long now = jiffies;
-
 		wake_up_interruptible(&mce_wait);
-		if (trigger[0])
-			call_usermodehelper(trigger, trigger_argv, NULL,
-						UMH_NO_WAIT);
 
-		if (time_after_eq(now, last_print + (check_interval*HZ))) {
-			last_print = now;
+		/*
+		 * There is no risk of missing notifications because
+		 * work_pending is always cleared before the function is
+		 * executed.
+		 */
+		if (trigger[0] && !work_pending(&mce_trigger_work))
+			schedule_work(&mce_trigger_work);
+
+		if (__ratelimit(&ratelimit))
 			printk(KERN_INFO "Machine check events logged\n");
-		}
 
 		return 1;
 	}
@@ -425,63 +535,78 @@ static struct notifier_block mce_idle_notifier = {
 
 static __init int periodic_mcheck_init(void)
 {
-	next_interval = check_interval * HZ;
-	if (next_interval)
-		schedule_delayed_work(&mcheck_work,
-				      round_jiffies_relative(next_interval));
-	idle_notifier_register(&mce_idle_notifier);
-	return 0;
+       idle_notifier_register(&mce_idle_notifier);
+       return 0;
 }
 __initcall(periodic_mcheck_init);
 
-
 /*
  * Initialize Machine Checks for a CPU.
  */
-static void mce_init(void *dummy)
+static int mce_cap_init(void)
 {
 	u64 cap;
-	int i;
+	unsigned b;
 
 	rdmsrl(MSR_IA32_MCG_CAP, cap);
-	banks = cap & 0xff;
-	if (banks > MCE_EXTENDED_BANK) {
-		banks = MCE_EXTENDED_BANK;
-		printk(KERN_INFO "MCE: warning: using only %d banks\n",
-		       MCE_EXTENDED_BANK);
+	b = cap & 0xff;
+	if (b > MAX_NR_BANKS) {
+		printk(KERN_WARNING
+		       "MCE: Using only %u machine check banks out of %u\n",
+			MAX_NR_BANKS, b);
+		b = MAX_NR_BANKS;
 	}
+
+	/* Don't support asymmetric configurations today */
+	WARN_ON(banks != 0 && b != banks);
+	banks = b;
+	if (!bank) {
+		bank = kmalloc(banks * sizeof(u64), GFP_KERNEL);
+		if (!bank)
+			return -ENOMEM;
+		memset(bank, 0xff, banks * sizeof(u64));
+	}
+
 	/* Use accurate RIP reporting if available. */
 	if ((cap & (1<<9)) && ((cap >> 16) & 0xff) >= 9)
 		rip_msr = MSR_IA32_MCG_EIP;
 
-	/* Log the machine checks left over from the previous reset.
-	   This also clears all registers */
-	do_machine_check(NULL, mce_bootlog ? -1 : -2);
+	return 0;
+}
+
+static void mce_init(void *dummy)
+{
+	u64 cap;
+	int i;
+	mce_banks_t all_banks;
+
+	/*
+	 * Log the machine checks left over from the previous reset.
+	 */
+	bitmap_fill(all_banks, MAX_NR_BANKS);
+	machine_check_poll(MCP_UC, &all_banks);
 
 	set_in_cr4(X86_CR4_MCE);
 
+	rdmsrl(MSR_IA32_MCG_CAP, cap);
 	if (cap & MCG_CTL_P)
 		wrmsr(MSR_IA32_MCG_CTL, 0xffffffff, 0xffffffff);
 
 	for (i = 0; i < banks; i++) {
-		if (i < NR_SYSFS_BANKS)
-			wrmsrl(MSR_IA32_MC0_CTL+4*i, bank[i]);
-		else
-			wrmsrl(MSR_IA32_MC0_CTL+4*i, ~0UL);
-
+		wrmsrl(MSR_IA32_MC0_CTL+4*i, bank[i]);
 		wrmsrl(MSR_IA32_MC0_STATUS+4*i, 0);
 	}
 }
 
 /* Add per CPU specific workarounds here */
-static void __cpuinit mce_cpu_quirks(struct cpuinfo_x86 *c)
+static void mce_cpu_quirks(struct cpuinfo_x86 *c)
 {
 	/* This should be disabled by the BIOS, but isn't always */
 	if (c->x86_vendor == X86_VENDOR_AMD) {
-		if(c->x86 == 15)
+		if (c->x86 == 15 && banks > 4)
 			/* disable GART TBL walk error reporting, which trips off
 			   incorrectly with the IOMMU & 3ware & Cerberus. */
-			clear_bit(10, &bank[4]);
+			clear_bit(10, (unsigned long *)&bank[4]);
 		if(c->x86 <= 17 && mce_bootlog < 0)
 			/* Lots of broken BIOS around that don't clear them
 			   by default and leave crap in there. Don't log. */
@@ -504,20 +629,38 @@ static void mce_cpu_features(struct cpuinfo_x86 *c)
 	}
 }
 
+static void mce_init_timer(void)
+{
+	struct timer_list *t = &__get_cpu_var(mce_timer);
+
+	/* data race harmless because everyone sets to the same value */
+	if (!next_interval)
+		next_interval = check_interval * HZ;
+	if (!next_interval)
+		return;
+	setup_timer(t, mcheck_timer, smp_processor_id());
+	t->expires = round_jiffies_relative(jiffies + next_interval);
+	add_timer(t);
+}
+
 /*
  * Called for each booted CPU to set up machine checks.
  * Must be called with preempt off.
  */
 void __cpuinit mcheck_init(struct cpuinfo_x86 *c)
 {
-	mce_cpu_quirks(c);
+	if (!mce_available(c))
+		return;
 
-	if (mce_dont_init ||
-	    !mce_available(c))
+	if (mce_cap_init() < 0) {
+		mce_dont_init = 1;
 		return;
+	}
+	mce_cpu_quirks(c);
 
 	mce_init(NULL);
 	mce_cpu_features(c);
+	mce_init_timer();
 }
 
 /*
@@ -573,7 +716,7 @@ static ssize_t mce_read(struct file *filp, char __user *ubuf, size_t usize,
 {
 	unsigned long *cpu_tsc;
 	static DEFINE_MUTEX(mce_read_mutex);
-	unsigned next;
+	unsigned prev, next;
 	char __user *buf = ubuf;
 	int i, err;
 
@@ -592,25 +735,32 @@ static ssize_t mce_read(struct file *filp, char __user *ubuf, size_t usize,
 	}
 
 	err = 0;
-	for (i = 0; i < next; i++) {
-		unsigned long start = jiffies;
-
-		while (!mcelog.entry[i].finished) {
-			if (time_after_eq(jiffies, start + 2)) {
-				memset(mcelog.entry + i,0, sizeof(struct mce));
-				goto timeout;
+	prev = 0;
+	do {
+		for (i = prev; i < next; i++) {
+			unsigned long start = jiffies;
+
+			while (!mcelog.entry[i].finished) {
+				if (time_after_eq(jiffies, start + 2)) {
+					memset(mcelog.entry + i, 0,
+					       sizeof(struct mce));
+					goto timeout;
+				}
+				cpu_relax();
 			}
-			cpu_relax();
+			smp_rmb();
+			err |= copy_to_user(buf, mcelog.entry + i,
+					    sizeof(struct mce));
+			buf += sizeof(struct mce);
+timeout:
+			;
 		}
-		smp_rmb();
-		err |= copy_to_user(buf, mcelog.entry + i, sizeof(struct mce));
-		buf += sizeof(struct mce);
- timeout:
-		;
-	}
 
-	memset(mcelog.entry, 0, next * sizeof(struct mce));
-	mcelog.next = 0;
+		memset(mcelog.entry + prev, 0,
+		       (next - prev) * sizeof(struct mce));
+		prev = next;
+		next = cmpxchg(&mcelog.next, prev, 0);
+	} while (next != prev);
 
 	synchronize_sched();
 
@@ -680,20 +830,6 @@ static struct miscdevice mce_log_device = {
 	&mce_chrdev_ops,
 };
 
-static unsigned long old_cr4 __initdata;
-
-void __init stop_mce(void)
-{
-	old_cr4 = read_cr4();
-	clear_in_cr4(X86_CR4_MCE);
-}
-
-void __init restart_mce(void)
-{
-	if (old_cr4 & X86_CR4_MCE)
-		set_in_cr4(X86_CR4_MCE);
-}
-
 /*
  * Old style boot options parsing. Only for compatibility.
  */
@@ -703,8 +839,7 @@ static int __init mcheck_disable(char *str)
 	return 1;
 }
 
-/* mce=off disables machine check. Note you can re-enable it later
-   using sysfs.
+/* mce=off disables machine check.
    mce=TOLERANCELEVEL (number, see above)
    mce=bootlog Log MCEs from before booting. Disabled by default on AMD.
    mce=nobootlog Don't log MCEs from before booting. */
@@ -728,6 +863,29 @@ __setup("mce=", mcheck_enable);
  * Sysfs support
  */
 
+/*
+ * Disable machine checks on suspend and shutdown. We can't really handle
+ * them later.
+ */
+static int mce_disable(void)
+{
+	int i;
+
+	for (i = 0; i < banks; i++)
+		wrmsrl(MSR_IA32_MC0_CTL + i*4, 0);
+	return 0;
+}
+
+static int mce_suspend(struct sys_device *dev, pm_message_t state)
+{
+	return mce_disable();
+}
+
+static int mce_shutdown(struct sys_device *dev)
+{
+	return mce_disable();
+}
+
 /* On resume clear all MCE state. Don't want to see leftovers from the BIOS.
    Only one CPU is active at this time, the others get readded later using
    CPU hotplug. */
@@ -738,20 +896,24 @@ static int mce_resume(struct sys_device *dev)
 	return 0;
 }
 
+static void mce_cpu_restart(void *data)
+{
+	del_timer_sync(&__get_cpu_var(mce_timer));
+	if (mce_available(&current_cpu_data))
+		mce_init(NULL);
+	mce_init_timer();
+}
+
 /* Reinit MCEs after user configuration changes */
 static void mce_restart(void)
 {
-	if (next_interval)
-		cancel_delayed_work(&mcheck_work);
-	/* Timer race is harmless here */
-	on_each_cpu(mce_init, NULL, 1);
 	next_interval = check_interval * HZ;
-	if (next_interval)
-		schedule_delayed_work(&mcheck_work,
-				      round_jiffies_relative(next_interval));
+	on_each_cpu(mce_cpu_restart, NULL, 1);
 }
 
 static struct sysdev_class mce_sysclass = {
+	.suspend = mce_suspend,
+	.shutdown = mce_shutdown,
 	.resume = mce_resume,
 	.name = "machinecheck",
 };
@@ -778,16 +940,26 @@ void (*threshold_cpu_callback)(unsigned long action, unsigned int cpu) __cpuinit
 	}								\
 	static SYSDEV_ATTR(name, 0644, show_ ## name, set_ ## name);
 
-/*
- * TBD should generate these dynamically based on number of available banks.
- * Have only 6 contol banks in /sysfs until then.
- */
-ACCESSOR(bank0ctl,bank[0],mce_restart())
-ACCESSOR(bank1ctl,bank[1],mce_restart())
-ACCESSOR(bank2ctl,bank[2],mce_restart())
-ACCESSOR(bank3ctl,bank[3],mce_restart())
-ACCESSOR(bank4ctl,bank[4],mce_restart())
-ACCESSOR(bank5ctl,bank[5],mce_restart())
+static struct sysdev_attribute *bank_attrs;
+
+static ssize_t show_bank(struct sys_device *s, struct sysdev_attribute *attr,
+			 char *buf)
+{
+	u64 b = bank[attr - bank_attrs];
+	return sprintf(buf, "%llx\n", b);
+}
+
+static ssize_t set_bank(struct sys_device *s, struct sysdev_attribute *attr,
+			const char *buf, size_t siz)
+{
+	char *end;
+	u64 new = simple_strtoull(buf, &end, 0);
+	if (end == buf)
+		return -EINVAL;
+	bank[attr - bank_attrs] = new;
+	mce_restart();
+	return end-buf;
+}
 
 static ssize_t show_trigger(struct sys_device *s, struct sysdev_attribute *attr,
 				char *buf)
@@ -814,8 +986,6 @@ static SYSDEV_ATTR(trigger, 0644, show_trigger, set_trigger);
 static SYSDEV_INT_ATTR(tolerant, 0644, tolerant);
 ACCESSOR(check_interval,check_interval,mce_restart())
 static struct sysdev_attribute *mce_attributes[] = {
-	&attr_bank0ctl, &attr_bank1ctl, &attr_bank2ctl,
-	&attr_bank3ctl, &attr_bank4ctl, &attr_bank5ctl,
 	&attr_tolerant.attr, &attr_check_interval, &attr_trigger,
 	NULL
 };
@@ -845,11 +1015,22 @@ static __cpuinit int mce_create_device(unsigned int cpu)
 		if (err)
 			goto error;
 	}
+	for (i = 0; i < banks; i++) {
+		err = sysdev_create_file(&per_cpu(device_mce, cpu),
+					&bank_attrs[i]);
+		if (err)
+			goto error2;
+	}
 	cpu_set(cpu, mce_device_initialized);
 
 	return 0;
+error2:
+	while (--i >= 0) {
+		sysdev_remove_file(&per_cpu(device_mce, cpu),
+					&bank_attrs[i]);
+	}
 error:
-	while (i--) {
+	while (--i >= 0) {
 		sysdev_remove_file(&per_cpu(device_mce,cpu),
 				   mce_attributes[i]);
 	}
@@ -868,15 +1049,46 @@ static __cpuinit void mce_remove_device(unsigned int cpu)
 	for (i = 0; mce_attributes[i]; i++)
 		sysdev_remove_file(&per_cpu(device_mce,cpu),
 			mce_attributes[i]);
+	for (i = 0; i < banks; i++)
+		sysdev_remove_file(&per_cpu(device_mce, cpu),
+			&bank_attrs[i]);
 	sysdev_unregister(&per_cpu(device_mce,cpu));
 	cpu_clear(cpu, mce_device_initialized);
 }
 
+/* Make sure there are no machine checks on offlined CPUs. */
+static void mce_disable_cpu(void *h)
+{
+	int i;
+	unsigned long action = *(unsigned long *)h;
+
+	if (!mce_available(&current_cpu_data))
+		return;
+	if (!(action & CPU_TASKS_FROZEN))
+		cmci_clear();
+	for (i = 0; i < banks; i++)
+		wrmsrl(MSR_IA32_MC0_CTL + i*4, 0);
+}
+
+static void mce_reenable_cpu(void *h)
+{
+	int i;
+	unsigned long action = *(unsigned long *)h;
+
+	if (!mce_available(&current_cpu_data))
+		return;
+	if (!(action & CPU_TASKS_FROZEN))
+		cmci_reenable();
+	for (i = 0; i < banks; i++)
+		wrmsrl(MSR_IA32_MC0_CTL + i*4, bank[i]);
+}
+
 /* Get notified when a cpu comes on/off. Be hotplug friendly. */
 static int __cpuinit mce_cpu_callback(struct notifier_block *nfb,
 				      unsigned long action, void *hcpu)
 {
 	unsigned int cpu = (unsigned long)hcpu;
+	struct timer_list *t = &per_cpu(mce_timer, cpu);
 
 	switch (action) {
 	case CPU_ONLINE:
@@ -891,6 +1103,21 @@ static int __cpuinit mce_cpu_callback(struct notifier_block *nfb,
 			threshold_cpu_callback(action, cpu);
 		mce_remove_device(cpu);
 		break;
+	case CPU_DOWN_PREPARE:
+	case CPU_DOWN_PREPARE_FROZEN:
+		del_timer_sync(t);
+		smp_call_function_single(cpu, mce_disable_cpu, &action, 1);
+		break;
+	case CPU_DOWN_FAILED:
+	case CPU_DOWN_FAILED_FROZEN:
+		t->expires = round_jiffies_relative(jiffies + next_interval);
+		add_timer_on(t, cpu);
+		smp_call_function_single(cpu, mce_reenable_cpu, &action, 1);
+		break;
+	case CPU_POST_DEAD:
+		/* intentionally ignoring frozen here */
+		cmci_rediscover(cpu);
+		break;
 	}
 	return NOTIFY_OK;
 }
@@ -899,6 +1126,34 @@ static struct notifier_block mce_cpu_notifier __cpuinitdata = {
 	.notifier_call = mce_cpu_callback,
 };
 
+static __init int mce_init_banks(void)
+{
+	int i;
+
+	bank_attrs = kzalloc(sizeof(struct sysdev_attribute) * banks,
+				GFP_KERNEL);
+	if (!bank_attrs)
+		return -ENOMEM;
+
+	for (i = 0; i < banks; i++) {
+		struct sysdev_attribute *a = &bank_attrs[i];
+		a->attr.name = kasprintf(GFP_KERNEL, "bank%d", i);
+		if (!a->attr.name)
+			goto nomem;
+		a->attr.mode = 0644;
+		a->show = show_bank;
+		a->store = set_bank;
+	}
+	return 0;
+
+nomem:
+	while (--i >= 0)
+		kfree(bank_attrs[i].attr.name);
+	kfree(bank_attrs);
+	bank_attrs = NULL;
+	return -ENOMEM;
+}
+
 static __init int mce_init_device(void)
 {
 	int err;
@@ -906,6 +1161,11 @@ static __init int mce_init_device(void)
 
 	if (!mce_available(&boot_cpu_data))
 		return -EIO;
+
+	err = mce_init_banks();
+	if (err)
+		return err;
+
 	err = sysdev_class_register(&mce_sysclass);
 	if (err)
 		return err;
diff --git a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
index 9817506dd46..c5a32f92d07 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_amd_64.c
@@ -79,6 +79,8 @@ static unsigned char shared_bank[NR_BANKS] = {
 
 static DEFINE_PER_CPU(unsigned char, bank_map);	/* see which banks are on */
 
+static void amd_threshold_interrupt(void);
+
 /*
  * CPU Initialization
  */
@@ -174,6 +176,8 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c)
 			tr.reset = 0;
 			tr.old_limit = 0;
 			threshold_restart_bank(&tr);
+
+			mce_threshold_vector = amd_threshold_interrupt;
 		}
 	}
 }
@@ -187,19 +191,13 @@ void mce_amd_feature_init(struct cpuinfo_x86 *c)
  * the interrupt goes off when error_count reaches threshold_limit.
  * the handler will simply log mcelog w/ software defined bank number.
  */
-asmlinkage void mce_threshold_interrupt(void)
+static void amd_threshold_interrupt(void)
 {
 	unsigned int bank, block;
 	struct mce m;
 	u32 low = 0, high = 0, address = 0;
 
-	ack_APIC_irq();
-	exit_idle();
-	irq_enter();
-
-	memset(&m, 0, sizeof(m));
-	rdtscll(m.tsc);
-	m.cpu = smp_processor_id();
+	mce_setup(&m);
 
 	/* assume first bank caused it */
 	for (bank = 0; bank < NR_BANKS; ++bank) {
@@ -233,7 +231,8 @@ asmlinkage void mce_threshold_interrupt(void)
 
 			/* Log the machine check that caused the threshold
 			   event. */
-			do_machine_check(NULL, 0);
+			machine_check_poll(MCP_TIMESTAMP,
+					&__get_cpu_var(mce_poll_banks));
 
 			if (high & MASK_OVERFLOW_HI) {
 				rdmsrl(address, m.misc);
@@ -243,13 +242,10 @@ asmlinkage void mce_threshold_interrupt(void)
 				       + bank * NR_BLOCKS
 				       + block;
 				mce_log(&m);
-				goto out;
+				return;
 			}
 		}
 	}
-out:
-	inc_irq_stat(irq_threshold_count);
-	irq_exit();
 }
 
 /*
diff --git a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
index aa5e287c98e..aaa7d973093 100644
--- a/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
+++ b/arch/x86/kernel/cpu/mcheck/mce_intel_64.c
@@ -1,6 +1,8 @@
 /*
  * Intel specific MCE features.
  * Copyright 2004 Zwane Mwaikambo <zwane@linuxpower.ca>
+ * Copyright (C) 2008, 2009 Intel Corporation
+ * Author: Andi Kleen
  */
 
 #include <linux/init.h>
@@ -13,6 +15,7 @@
 #include <asm/hw_irq.h>
 #include <asm/idle.h>
 #include <asm/therm_throt.h>
+#include <asm/apic.h>
 
 asmlinkage void smp_thermal_interrupt(void)
 {
@@ -25,7 +28,7 @@ asmlinkage void smp_thermal_interrupt(void)
 
 	rdmsrl(MSR_IA32_THERM_STATUS, msr_val);
 	if (therm_throt_process(msr_val & 1))
-		mce_log_therm_throt_event(smp_processor_id(), msr_val);
+		mce_log_therm_throt_event(msr_val);
 
 	inc_irq_stat(irq_thermal_count);
 	irq_exit();
@@ -85,7 +88,209 @@ static void intel_init_thermal(struct cpuinfo_x86 *c)
 	return;
 }
 
+/*
+ * Support for Intel Correct Machine Check Interrupts. This allows
+ * the CPU to raise an interrupt when a corrected machine check happened.
+ * Normally we pick those up using a regular polling timer.
+ * Also supports reliable discovery of shared banks.
+ */
+
+static DEFINE_PER_CPU(mce_banks_t, mce_banks_owned);
+
+/*
+ * cmci_discover_lock protects against parallel discovery attempts
+ * which could race against each other.
+ */
+static DEFINE_SPINLOCK(cmci_discover_lock);
+
+#define CMCI_THRESHOLD 1
+
+static int cmci_supported(int *banks)
+{
+	u64 cap;
+
+	/*
+	 * Vendor check is not strictly needed, but the initial
+	 * initialization is vendor keyed and this
+	 * makes sure none of the backdoors are entered otherwise.
+	 */
+	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
+		return 0;
+	if (!cpu_has_apic || lapic_get_maxlvt() < 6)
+		return 0;
+	rdmsrl(MSR_IA32_MCG_CAP, cap);
+	*banks = min_t(unsigned, MAX_NR_BANKS, cap & 0xff);
+	return !!(cap & MCG_CMCI_P);
+}
+
+/*
+ * The interrupt handler. This is called on every event.
+ * Just call the poller directly to log any events.
+ * This could in theory increase the threshold under high load,
+ * but doesn't for now.
+ */
+static void intel_threshold_interrupt(void)
+{
+	machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
+	mce_notify_user();
+}
+
+static void print_update(char *type, int *hdr, int num)
+{
+	if (*hdr == 0)
+		printk(KERN_INFO "CPU %d MCA banks", smp_processor_id());
+	*hdr = 1;
+	printk(KERN_CONT " %s:%d", type, num);
+}
+
+/*
+ * Enable CMCI (Corrected Machine Check Interrupt) for available MCE banks
+ * on this CPU. Use the algorithm recommended in the SDM to discover shared
+ * banks.
+ */
+static void cmci_discover(int banks, int boot)
+{
+	unsigned long *owned = (void *)&__get_cpu_var(mce_banks_owned);
+	int hdr = 0;
+	int i;
+
+	spin_lock(&cmci_discover_lock);
+	for (i = 0; i < banks; i++) {
+		u64 val;
+
+		if (test_bit(i, owned))
+			continue;
+
+		rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
+
+		/* Already owned by someone else? */
+		if (val & CMCI_EN) {
+			if (test_and_clear_bit(i, owned) || boot)
+				print_update("SHD", &hdr, i);
+			__clear_bit(i, __get_cpu_var(mce_poll_banks));
+			continue;
+		}
+
+		val |= CMCI_EN | CMCI_THRESHOLD;
+		wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
+		rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
+
+		/* Did the enable bit stick? -- the bank supports CMCI */
+		if (val & CMCI_EN) {
+			if (!test_and_set_bit(i, owned) || boot)
+				print_update("CMCI", &hdr, i);
+			__clear_bit(i, __get_cpu_var(mce_poll_banks));
+		} else {
+			WARN_ON(!test_bit(i, __get_cpu_var(mce_poll_banks)));
+		}
+	}
+	spin_unlock(&cmci_discover_lock);
+	if (hdr)
+		printk(KERN_CONT "\n");
+}
+
+/*
+ * Just in case we missed an event during initialization check
+ * all the CMCI owned banks.
+ */
+void cmci_recheck(void)
+{
+	unsigned long flags;
+	int banks;
+
+	if (!mce_available(&current_cpu_data) || !cmci_supported(&banks))
+		return;
+	local_irq_save(flags);
+	machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
+	local_irq_restore(flags);
+}
+
+/*
+ * Disable CMCI on this CPU for all banks it owns when it goes down.
+ * This allows other CPUs to claim the banks on rediscovery.
+ */
+void cmci_clear(void)
+{
+	int i;
+	int banks;
+	u64 val;
+
+	if (!cmci_supported(&banks))
+		return;
+	spin_lock(&cmci_discover_lock);
+	for (i = 0; i < banks; i++) {
+		if (!test_bit(i, __get_cpu_var(mce_banks_owned)))
+			continue;
+		/* Disable CMCI */
+		rdmsrl(MSR_IA32_MC0_CTL2 + i, val);
+		val &= ~(CMCI_EN|CMCI_THRESHOLD_MASK);
+		wrmsrl(MSR_IA32_MC0_CTL2 + i, val);
+		__clear_bit(i, __get_cpu_var(mce_banks_owned));
+	}
+	spin_unlock(&cmci_discover_lock);
+}
+
+/*
+ * After a CPU went down cycle through all the others and rediscover
+ * Must run in process context.
+ */
+void cmci_rediscover(int dying)
+{
+	int banks;
+	int cpu;
+	cpumask_var_t old;
+
+	if (!cmci_supported(&banks))
+		return;
+	if (!alloc_cpumask_var(&old, GFP_KERNEL))
+		return;
+	cpumask_copy(old, &current->cpus_allowed);
+
+	for_each_online_cpu (cpu) {
+		if (cpu == dying)
+			continue;
+		if (set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu)))
+			continue;
+		/* Recheck banks in case CPUs don't all have the same */
+		if (cmci_supported(&banks))
+			cmci_discover(banks, 0);
+	}
+
+	set_cpus_allowed_ptr(current, old);
+	free_cpumask_var(old);
+}
+
+/*
+ * Reenable CMCI on this CPU in case a CPU down failed.
+ */
+void cmci_reenable(void)
+{
+	int banks;
+	if (cmci_supported(&banks))
+		cmci_discover(banks, 0);
+}
+
+static __cpuinit void intel_init_cmci(void)
+{
+	int banks;
+
+	if (!cmci_supported(&banks))
+		return;
+
+	mce_threshold_vector = intel_threshold_interrupt;
+	cmci_discover(banks, 1);
+	/*
+	 * For CPU #0 this runs with still disabled APIC, but that's
+	 * ok because only the vector is set up. We still do another
+	 * check for the banks later for CPU #0 just to make sure
+	 * to not miss any events.
+	 */
+	apic_write(APIC_LVTCMCI, THRESHOLD_APIC_VECTOR|APIC_DM_FIXED);
+	cmci_recheck();
+}
+
 void mce_intel_feature_init(struct cpuinfo_x86 *c)
 {
 	intel_init_thermal(c);
+	intel_init_cmci();
 }
diff --git a/arch/x86/kernel/cpu/mcheck/threshold.c b/arch/x86/kernel/cpu/mcheck/threshold.c
new file mode 100644
index 00000000000..23ee9e730f7
--- /dev/null
+++ b/arch/x86/kernel/cpu/mcheck/threshold.c
@@ -0,0 +1,29 @@
+/*
+ * Common corrected MCE threshold handler code:
+ */
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+
+#include <asm/irq_vectors.h>
+#include <asm/apic.h>
+#include <asm/idle.h>
+#include <asm/mce.h>
+
+static void default_threshold_interrupt(void)
+{
+	printk(KERN_ERR "Unexpected threshold interrupt at vector %x\n",
+			 THRESHOLD_APIC_VECTOR);
+}
+
+void (*mce_threshold_vector)(void) = default_threshold_interrupt;
+
+asmlinkage void mce_threshold_interrupt(void)
+{
+	exit_idle();
+	irq_enter();
+	inc_irq_stat(irq_threshold_count);
+	mce_threshold_vector();
+	irq_exit();
+	/* Ack only at the end to avoid potential reentry */
+	ack_APIC_irq();
+}