[PATCH -v4 0/4] x86: MCE: testing enhancement

[PATCH -v4 0/4] x86: MCE: testing enhancement

[Huang Ying]

Hi, All,

This patchset is necessary for MCE regression testing suite mce-test 
(http://www.kernel.org/pub/scm/utils/cpu/mce/mce-test.git) to work 
properly.

Best Regards,
Huang Ying

[PATCH -v4 1/4] x86: MCE: Support specifying context for software mce injection

[Huang Ying]

The cpu context is specified via the new mce.inject_flags fields.
This allows more realistic machine check testing in different
situations. "RANDOM" context is implemented via NMI broadcasting to
add randomization to testing.

AK: Fix NMI broadcasting check. Fix 32-bit building. Some race
fixes. Move to module. Various changes

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>


ChangeLog:

v3:

- Re-based on latest x86-tip.git/mce4

- Fix 32-bit building

v2:

- Re-base on latest x86-tip.git/mce3
---
 arch/x86/include/asm/mce.h              |   11 ++-
 arch/x86/kernel/cpu/mcheck/mce-inject.c |  156 +++++++++++++++++++++++++------
 2 files changed, 135 insertions(+), 32 deletions(-)

diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 5cdd8d1..7196cdc 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -38,6 +38,13 @@
 #define MCM_ADDR_MEM	 3	/* memory address */
 #define MCM_ADDR_GENERIC 7	/* generic */
 
+#define MCJ_CTX_MASK		3
+#define MCJ_CTX(flags)		((flags) & MCJ_CTX_MASK)
+#define MCJ_CTX_RANDOM		0    /* inject context: random */
+#define MCJ_CTX_PROCESS		1    /* inject context: process */
+#define MCJ_CTX_IRQ		2    /* inject context: IRQ */
+#define MCJ_NMI_BROADCAST	4    /* do NMI broadcasting */
+
 /* Fields are zero when not available */
 struct mce {
 	__u64 status;
@@ -48,8 +55,8 @@ struct mce {
 	__u64 tsc;	/* cpu time stamp counter */
 	__u64 time;	/* wall time_t when error was detected */
 	__u8  cpuvendor;	/* cpu vendor as encoded in system.h */
-	__u8  pad1;
-	__u16 pad2;
+	__u8  inject_flags;	/* software inject flags */
+	__u16  pad;
 	__u32 cpuid;	/* CPUID 1 EAX */
 	__u8  cs;		/* code segment */
 	__u8  bank;	/* machine check bank */
diff --git a/arch/x86/kernel/cpu/mcheck/mce-inject.c b/arch/x86/kernel/cpu/mcheck/mce-inject.c
index a3a235a..ad5d927 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-inject.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-inject.c
@@ -18,7 +18,12 @@
 #include <linux/string.h>
 #include <linux/fs.h>
 #include <linux/smp.h>
+#include <linux/notifier.h>
+#include <linux/kdebug.h>
+#include <linux/cpu.h>
+#include <linux/sched.h>
 #include <asm/mce.h>
+#include <asm/apic.h>
 
 /* Update fake mce registers on current CPU. */
 static void inject_mce(struct mce *m)
@@ -39,44 +44,141 @@ static void inject_mce(struct mce *m)
 	i->finished = 1;
 }
 
-struct delayed_mce {
-	struct timer_list timer;
-	struct mce m;
+static void raise_corrected(struct mce *m)
+{
+	unsigned long flags;
+	mce_banks_t b;
+
+	memset(&b, 0xff, sizeof(mce_banks_t));
+	local_irq_save(flags);
+	machine_check_poll(0, &b);
+	local_irq_restore(flags);
+	m->finished = 0;
+}
+
+static void raise_uncorrected(struct mce *m, struct pt_regs *pregs)
+{
+	struct pt_regs regs;
+	unsigned long flags;
+
+	if (!pregs) {
+		memset(&regs, 0, sizeof(struct pt_regs));
+		regs.ip = m->ip;
+		regs.cs = m->cs;
+		pregs = &regs;
+	}
+	/* in mcheck exeception handler, irq will be disabled */
+	local_irq_save(flags);
+	do_machine_check(pregs, 0);
+	local_irq_restore(flags);
+	m->finished = 0;
+}
+
+static cpumask_t mce_inject_cpumask;
+
+static int mce_raise_notify(struct notifier_block *self,
+			    unsigned long val, void *data)
+{
+	struct die_args *args = (struct die_args *)data;
+	int cpu = smp_processor_id();
+	struct mce *m = &__get_cpu_var(injectm);
+	if (val != DIE_NMI_IPI || !cpu_isset(cpu, mce_inject_cpumask))
+		return NOTIFY_DONE;
+	cpu_clear(cpu, mce_inject_cpumask);
+	if (m->status & MCI_STATUS_UC)
+		raise_uncorrected(m, args->regs);
+	else if (m->status)
+		raise_corrected(m);
+	return NOTIFY_STOP;
+}
+
+static struct notifier_block mce_raise_nb = {
+	.notifier_call = mce_raise_notify,
+	.priority = 1000,
 };
 
 /* Inject mce on current CPU */
-static void raise_mce(unsigned long data)
+static int raise_local(struct mce *m)
 {
-	struct delayed_mce *dm = (struct delayed_mce *)data;
-	struct mce *m = &dm->m;
+	int context = MCJ_CTX(m->inject_flags);
+	int ret = 0;
 	int cpu = m->extcpu;
 
-	inject_mce(m);
 	if (m->status & MCI_STATUS_UC) {
-		struct pt_regs regs;
-		memset(&regs, 0, sizeof(struct pt_regs));
-		regs.ip = m->ip;
-		regs.cs = m->cs;
 		printk(KERN_INFO "Triggering MCE exception on CPU %d\n", cpu);
-		do_machine_check(&regs, 0);
+		switch (context) {
+		case MCJ_CTX_IRQ:
+			/*
+			 * Could do more to fake interrupts like
+			 * calling irq_enter, but the necessary
+			 * machinery isn't exported currently.
+			 */
+			/*FALL THROUGH*/
+		case MCJ_CTX_PROCESS:
+			raise_uncorrected(m, NULL);
+			break;
+		default:
+			printk(KERN_INFO "Invalid MCE context\n");
+			ret = -EINVAL;
+		}
 		printk(KERN_INFO "MCE exception done on CPU %d\n", cpu);
-	} else {
-		mce_banks_t b;
-		memset(&b, 0xff, sizeof(mce_banks_t));
+	} else if (m->status) {
 		printk(KERN_INFO "Starting machine check poll CPU %d\n", cpu);
-		machine_check_poll(0, &b);
+		raise_corrected(m);
 		mce_notify_irq();
-		printk(KERN_INFO "Finished machine check poll on CPU %d\n",
-		       cpu);
-	}
-	kfree(dm);
+		printk(KERN_INFO "Machine check poll done on CPU %d\n", cpu);
+	} else
+		m->finished = 0;
+
+	return ret;
+}
+
+static void raise_mce(struct mce *m)
+{
+	int context = MCJ_CTX(m->inject_flags);
+
+	inject_mce(m);
+
+	if (context == MCJ_CTX_RANDOM)
+		return;
+
+#ifdef CONFIG_X86_LOCAL_APIC
+	if (m->inject_flags & MCJ_NMI_BROADCAST) {
+		unsigned long start;
+		int cpu;
+		get_online_cpus();
+		mce_inject_cpumask = cpu_online_map;
+		cpu_clear(get_cpu(), mce_inject_cpumask);
+		for_each_online_cpu(cpu) {
+			struct mce *mcpu = &per_cpu(injectm, cpu);
+			if (!mcpu->finished ||
+			    MCJ_CTX(mcpu->inject_flags) != MCJ_CTX_RANDOM)
+				cpu_clear(cpu, mce_inject_cpumask);
+		}
+		if (!cpus_empty(mce_inject_cpumask))
+			apic->send_IPI_mask(&mce_inject_cpumask, NMI_VECTOR);
+		start = jiffies;
+		while (!cpus_empty(mce_inject_cpumask)) {
+			if (!time_before(jiffies, start + 2*HZ)) {
+				printk(KERN_ERR
+				"Timeout waiting for mce inject NMI %lx\n",
+					*cpus_addr(mce_inject_cpumask));
+				break;
+			}
+			cpu_relax();
+		}
+		raise_local(m);
+		put_cpu();
+		put_online_cpus();
+	} else
+#endif
+		raise_local(m);
 }
 
 /* Error injection interface */
 static ssize_t mce_write(struct file *filp, const char __user *ubuf,
 			 size_t usize, loff_t *off)
 {
-	struct delayed_mce *dm;
 	struct mce m;
 
 	if (!capable(CAP_SYS_ADMIN))
@@ -96,19 +198,12 @@ static ssize_t mce_write(struct file *filp, const char __user *ubuf,
 	if (m.extcpu >= num_possible_cpus() || !cpu_online(m.extcpu))
 		return -EINVAL;
 
-	dm = kmalloc(sizeof(struct delayed_mce), GFP_KERNEL);
-	if (!dm)
-		return -ENOMEM;
-
 	/*
 	 * Need to give user space some time to set everything up,
 	 * so do it a jiffie or two later everywhere.
-	 * Should we use a hrtimer here for better synchronization?
 	 */
-	memcpy(&dm->m, &m, sizeof(struct mce));
-	setup_timer(&dm->timer, raise_mce, (unsigned long)dm);
-	dm->timer.expires = jiffies + 2;
-	add_timer_on(&dm->timer, m.extcpu);
+	schedule_timeout(2);
+	raise_mce(&m);
 	return usize;
 }
 
@@ -116,6 +211,7 @@ static int inject_init(void)
 {
 	printk(KERN_INFO "Machine check injector initialized\n");
 	mce_chrdev_ops.write = mce_write;
+	register_die_notifier(&mce_raise_nb);
 	return 0;
 }
 
-- 
1.6.3.3

[PATCH -v4 2/4] x86: MCE: Support specifying raise mode for software MCE injection

[Huang Ying]

Raise mode include raising as exception or raising as poll, it is
specified via the mce.inject_flags field.

This can be used to specify raise mode of UCNA, which is UC error but
raised not as exception. And this can be used to test the filter code
of poll handler or exception handler too. For example, enforce a poll
raise mode for a fatal MCE.

Signed-off-by: Huang Ying <ying.huang@intel.com>


ChangeLog:

v2:

- Re-base on latest x86-tip.git/mce3
---
 arch/x86/include/asm/mce.h              |    1 +
 arch/x86/kernel/cpu/mcheck/mce-inject.c |   16 ++++++++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index 7196cdc..871f9a9 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -44,6 +44,7 @@
 #define MCJ_CTX_PROCESS		1    /* inject context: process */
 #define MCJ_CTX_IRQ		2    /* inject context: IRQ */
 #define MCJ_NMI_BROADCAST	4    /* do NMI broadcasting */
+#define MCJ_EXCEPTION		8    /* raise as exception */
 
 /* Fields are zero when not available */
 struct mce {
diff --git a/arch/x86/kernel/cpu/mcheck/mce-inject.c b/arch/x86/kernel/cpu/mcheck/mce-inject.c
index ad5d927..7029f0e 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-inject.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-inject.c
@@ -44,7 +44,7 @@ static void inject_mce(struct mce *m)
 	i->finished = 1;
 }
 
-static void raise_corrected(struct mce *m)
+static void raise_poll(struct mce *m)
 {
 	unsigned long flags;
 	mce_banks_t b;
@@ -56,7 +56,7 @@ static void raise_corrected(struct mce *m)
 	m->finished = 0;
 }
 
-static void raise_uncorrected(struct mce *m, struct pt_regs *pregs)
+static void raise_exception(struct mce *m, struct pt_regs *pregs)
 {
 	struct pt_regs regs;
 	unsigned long flags;
@@ -85,10 +85,10 @@ static int mce_raise_notify(struct notifier_block *self,
 	if (val != DIE_NMI_IPI || !cpu_isset(cpu, mce_inject_cpumask))
 		return NOTIFY_DONE;
 	cpu_clear(cpu, mce_inject_cpumask);
-	if (m->status & MCI_STATUS_UC)
-		raise_uncorrected(m, args->regs);
+	if (m->inject_flags & MCJ_EXCEPTION)
+		raise_exception(m, args->regs);
 	else if (m->status)
-		raise_corrected(m);
+		raise_poll(m);
 	return NOTIFY_STOP;
 }
 
@@ -104,7 +104,7 @@ static int raise_local(struct mce *m)
 	int ret = 0;
 	int cpu = m->extcpu;
 
-	if (m->status & MCI_STATUS_UC) {
+	if (m->inject_flags & MCJ_EXCEPTION) {
 		printk(KERN_INFO "Triggering MCE exception on CPU %d\n", cpu);
 		switch (context) {
 		case MCJ_CTX_IRQ:
@@ -115,7 +115,7 @@ static int raise_local(struct mce *m)
 			 */
 			/*FALL THROUGH*/
 		case MCJ_CTX_PROCESS:
-			raise_uncorrected(m, NULL);
+			raise_exception(m, NULL);
 			break;
 		default:
 			printk(KERN_INFO "Invalid MCE context\n");
@@ -124,7 +124,7 @@ static int raise_local(struct mce *m)
 		printk(KERN_INFO "MCE exception done on CPU %d\n", cpu);
 	} else if (m->status) {
 		printk(KERN_INFO "Starting machine check poll CPU %d\n", cpu);
-		raise_corrected(m);
+		raise_poll(m);
 		mce_notify_irq();
 		printk(KERN_INFO "Machine check poll done on CPU %d\n", cpu);
 	} else
-- 
1.6.3.3

[PATCH -v4 3/4] x86: MCE: Move debugfs mce dir creating to mce.c

[Huang Ying]

Because more debugfs files under mce dir will be create in mce.c.

Signed-off-by: Huang Ying <ying.huang@intel.com>
---
 arch/x86/kernel/cpu/mcheck/mce-internal.h |    1 +
 arch/x86/kernel/cpu/mcheck/mce-severity.c |    6 +++---
 arch/x86/kernel/cpu/mcheck/mce.c          |   13 +++++++++++++
 3 files changed, 17 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/cpu/mcheck/mce-internal.h b/arch/x86/kernel/cpu/mcheck/mce-internal.h
index 54dcb8f..2179f87 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-internal.h
+++ b/arch/x86/kernel/cpu/mcheck/mce-internal.h
@@ -11,5 +11,6 @@ enum severity_level {
 };
 
 int mce_severity(struct mce *a, int tolerant, char **msg);
+struct dentry *mce_get_debugfs_dir(void);
 
 extern int mce_ser;
diff --git a/arch/x86/kernel/cpu/mcheck/mce-severity.c b/arch/x86/kernel/cpu/mcheck/mce-severity.c
index ff0807f..1392c96 100644
--- a/arch/x86/kernel/cpu/mcheck/mce-severity.c
+++ b/arch/x86/kernel/cpu/mcheck/mce-severity.c
@@ -139,6 +139,7 @@ int mce_severity(struct mce *a, int tolerant, char **msg)
 	}
 }
 
+#ifdef CONFIG_DEBUG_FS
 static void *s_start(struct seq_file *f, loff_t *pos)
 {
 	if (*pos >= ARRAY_SIZE(severities))
@@ -197,7 +198,7 @@ static int __init severities_debugfs_init(void)
 {
 	struct dentry *dmce = NULL, *fseverities_coverage = NULL;
 
-	dmce = debugfs_create_dir("mce", NULL);
+	dmce = mce_get_debugfs_dir();
 	if (dmce == NULL)
 		goto err_out;
 	fseverities_coverage = debugfs_create_file("severities-coverage",
@@ -211,8 +212,7 @@ static int __init severities_debugfs_init(void)
 err_out:
 	if (fseverities_coverage)
 		debugfs_remove(fseverities_coverage);
-	if (dmce)
-		debugfs_remove(dmce);
 	return -ENOMEM;
 }
 late_initcall(severities_debugfs_init);
+#endif
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 484c1e5..b57b6aa 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -34,6 +34,7 @@
 #include <linux/smp.h>
 #include <linux/fs.h>
 #include <linux/mm.h>
+#include <linux/debugfs.h>
 
 #include <asm/processor.h>
 #include <asm/hw_irq.h>
@@ -2047,3 +2048,15 @@ static int __init mcheck_disable(char *str)
 	return 1;
 }
 __setup("nomce", mcheck_disable);
+
+#ifdef CONFIG_DEBUG_FS
+struct dentry *mce_get_debugfs_dir(void)
+{
+	static struct dentry *dmce;
+
+	if (!dmce)
+		dmce = debugfs_create_dir("mce", NULL);
+
+	return dmce;
+}
+#endif
-- 
1.6.3.3

[PATCH -v4 4/4] x86: MCE: Fake panic support for MCE testing

[Huang Ying]

If "fake panic" mode is turned on, just log panic message instead of
go real panic. This is used for testing only, so that the test suite
can check for the correct panic message and do regression testing for
MCE would go panic.

This patch is based on x86-tip.git/mce3.

Signed-off-by: Huang Ying <ying.huang@intel.com>


ChangeLog:

v4:

- Move config file from sysfs to debugfs
---
 arch/x86/kernel/cpu/mcheck/mce.c |   75 ++++++++++++++++++++++++++++++++------
 1 files changed, 64 insertions(+), 11 deletions(-)

diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index b57b6aa..c3a9f07 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -223,6 +223,9 @@ static void print_mce_tail(void)
 
 static atomic_t mce_paniced;
 
+static int fake_panic;
+static atomic_t mce_fake_paniced;
+
 /* Panic in progress. Enable interrupts and wait for final IPI */
 static void wait_for_panic(void)
 {
@@ -240,15 +243,21 @@ static void mce_panic(char *msg, struct mce *final, char *exp)
 {
 	int i;
 
-	/*
-	 * Make sure only one CPU runs in machine check panic
-	 */
-	if (atomic_add_return(1, &mce_paniced) > 1)
-		wait_for_panic();
-	barrier();
+	if (!fake_panic) {
+		/*
+		 * Make sure only one CPU runs in machine check panic
+		 */
+		if (atomic_add_return(1, &mce_paniced) > 1)
+			wait_for_panic();
+		barrier();
 
-	bust_spinlocks(1);
-	console_verbose();
+		bust_spinlocks(1);
+		console_verbose();
+	} else {
+		/* Don't log too much for fake panic */
+		if (atomic_add_return(1, &mce_fake_paniced) > 1)
+			return;
+	}
 	print_mce_head();
 	/* First print corrected ones that are still unlogged */
 	for (i = 0; i < MCE_LOG_LEN; i++) {
@@ -275,9 +284,12 @@ static void mce_panic(char *msg, struct mce *final, char *exp)
 	print_mce_tail();
 	if (exp)
 		printk(KERN_EMERG "Machine check: %s\n", exp);
-	if (panic_timeout == 0)
-		panic_timeout = mce_panic_timeout;
-	panic(msg);
+	if (!fake_panic) {
+		if (panic_timeout == 0)
+			panic_timeout = mce_panic_timeout;
+		panic(msg);
+	} else
+		printk(KERN_EMERG "Fake kernel panic: %s\n", msg);
 }
 
 /* Support code for software error injection */
@@ -2059,4 +2071,45 @@ struct dentry *mce_get_debugfs_dir(void)
 
 	return dmce;
 }
+
+static void mce_reset(void)
+{
+	cpu_missing = 0;
+	atomic_set(&mce_fake_paniced, 0);
+	atomic_set(&mce_executing, 0);
+	atomic_set(&mce_callin, 0);
+	atomic_set(&global_nwo, 0);
+}
+
+static int fake_panic_get(void *data, u64 *val)
+{
+	*val = fake_panic;
+	return 0;
+}
+
+static int fake_panic_set(void *data, u64 val)
+{
+	mce_reset();
+	fake_panic = val;
+	return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(fake_panic_fops, fake_panic_get,
+			fake_panic_set, "%llu\n");
+
+static int __init mce_debugfs_init(void)
+{
+	struct dentry *dmce, *ffake_panic;
+
+	dmce = mce_get_debugfs_dir();
+	if (!dmce)
+		return -ENOMEM;
+	ffake_panic = debugfs_create_file("fake_panic", 0444, dmce, NULL,
+					  &fake_panic_fops);
+	if (!ffake_panic)
+		return -ENOMEM;
+
+	return 0;
+}
+late_initcall(mce_debugfs_init);
 #endif
-- 
1.6.3.3