1 files changed, 166 insertions, 98 deletions

diff --git a/arch/ia64/kernel/salinfo.c b/arch/ia64/kernel/salinfo.c
index ca68e6e44a7..ee9719eebb1 100644
--- a/arch/ia64/kernel/salinfo.c
+++ b/arch/ia64/kernel/salinfo.c
@@ -3,7 +3,7 @@
  *
  * Creates entries in /proc/sal for various system features.
  *
- * Copyright (c) 2003 Silicon Graphics, Inc.  All rights reserved.
+ * Copyright (c) 2003, 2006 Silicon Graphics, Inc.  All rights reserved.
  * Copyright (c) 2003 Hewlett-Packard Co
  *	Bjorn Helgaas <bjorn.helgaas@hp.com>
  *
@@ -27,17 +27,26 @@
  *   mca.c may not pass a buffer, a NULL buffer just indicates that a new
  *   record is available in SAL.
  *   Replace some NR_CPUS by cpus_online, for hotplug cpu.
+ *
+ * Jan  5 2006        kaos@sgi.com
+ *   Handle hotplug cpus coming online.
+ *   Handle hotplug cpus going offline while they still have outstanding records.
+ *   Use the cpu_* macros consistently.
+ *   Replace the counting semaphore with a mutex and a test if the cpumask is non-empty.
+ *   Modify the locking to make the test for "work to do" an atomic operation.
  */
 
+#include <linux/capability.h>
+#include <linux/cpu.h>
 #include <linux/types.h>
 #include <linux/proc_fs.h>
+#include <linux/seq_file.h>
 #include <linux/module.h>
 #include <linux/smp.h>
-#include <linux/smp_lock.h>
 #include <linux/timer.h>
 #include <linux/vmalloc.h>
+#include <linux/semaphore.h>
 
-#include <asm/semaphore.h>
 #include <asm/sal.h>
 #include <asm/uaccess.h>
 
@@ -45,7 +54,7 @@ MODULE_AUTHOR("Jesse Barnes <jbarnes@sgi.com>");
 MODULE_DESCRIPTION("/proc interface to IA-64 SAL features");
 MODULE_LICENSE("GPL");
 
-static int salinfo_read(char *page, char **start, off_t off, int count, int *eof, void *data);
+static const struct file_operations proc_salinfo_fops;
 
 typedef struct {
 	const char		*name;		/* name of the proc entry */
@@ -57,7 +66,7 @@ typedef struct {
  * List {name,feature} pairs for every entry in /proc/sal/<feature>
  * that this module exports
  */
-static salinfo_entry_t salinfo_entries[]={
+static const salinfo_entry_t salinfo_entries[]={
 	{ "bus_lock",           IA64_SAL_PLATFORM_FEATURE_BUS_LOCK, },
 	{ "irq_redirection",	IA64_SAL_PLATFORM_FEATURE_IRQ_REDIR_HINT, },
 	{ "ipi_redirection",	IA64_SAL_PLATFORM_FEATURE_IPI_REDIR_HINT, },
@@ -131,8 +140,8 @@ enum salinfo_state {
 };
 
 struct salinfo_data {
-	volatile cpumask_t	cpu_event;	/* which cpus have outstanding events */
-	struct semaphore	sem;		/* count of cpus with outstanding events (bits set in cpu_event) */
+	cpumask_t		cpu_event;	/* which cpus have outstanding events */
+	struct semaphore	mutex;
 	u8			*log_buffer;
 	u64			log_size;
 	u8			*oemdata;	/* decoded oem data */
@@ -154,7 +163,7 @@ static DEFINE_SPINLOCK(data_saved_lock);
 /** salinfo_platform_oemdata - optional callback to decode oemdata from an error
  * record.
  * @sect_header: pointer to the start of the section to decode.
- * @oemdata: returns vmalloc area containing the decded output.
+ * @oemdata: returns vmalloc area containing the decoded output.
  * @oemdata_size: returns length of decoded output (strlen).
  *
  * Description: If user space asks for oem data to be decoded by the kernel
@@ -173,6 +182,21 @@ struct salinfo_platform_oemdata_parms {
 	int ret;
 };
 
+/* Kick the mutex that tells user space that there is work to do.  Instead of
+ * trying to track the state of the mutex across multiple cpus, in user
+ * context, interrupt context, non-maskable interrupt context and hotplug cpu,
+ * it is far easier just to grab the mutex if it is free then release it.
+ *
+ * This routine must be called with data_saved_lock held, to make the down/up
+ * operation atomic.
+ */
+static void
+salinfo_work_to_do(struct salinfo_data *data)
+{
+	(void)(down_trylock(&data->mutex) ?: 0);
+	up(&data->mutex);
+}
+
 static void
 salinfo_platform_oemdata_cpu(void *context)
 {
@@ -211,9 +235,9 @@ salinfo_log_wakeup(int type, u8 *buffer, u64 size, int irqsafe)
 
 	BUG_ON(type >= ARRAY_SIZE(salinfo_log_name));
 
+	if (irqsafe)
+		spin_lock_irqsave(&data_saved_lock, flags);
 	if (buffer) {
-		if (irqsafe)
-			spin_lock_irqsave(&data_saved_lock, flags);
 		for (i = 0, data_saved = data->data_saved; i < saved_size; ++i, ++data_saved) {
 			if (!data_saved->buffer)
 				break;
@@ -231,39 +255,36 @@ salinfo_log_wakeup(int type, u8 *buffer, u64 size, int irqsafe)
 			data_saved->size = size;
 			data_saved->buffer = buffer;
 		}
-		if (irqsafe)
-			spin_unlock_irqrestore(&data_saved_lock, flags);
 	}
-
-	if (!test_and_set_bit(smp_processor_id(), &data->cpu_event)) {
-		if (irqsafe)
-			up(&data->sem);
+	cpu_set(smp_processor_id(), data->cpu_event);
+	if (irqsafe) {
+		salinfo_work_to_do(data);
+		spin_unlock_irqrestore(&data_saved_lock, flags);
 	}
 }
 
 /* Check for outstanding MCA/INIT records every minute (arbitrary) */
 #define SALINFO_TIMER_DELAY (60*HZ)
 static struct timer_list salinfo_timer;
+extern void ia64_mlogbuf_dump(void);
 
 static void
 salinfo_timeout_check(struct salinfo_data *data)
 {
-	int i;
+	unsigned long flags;
 	if (!data->open)
 		return;
-	for_each_online_cpu(i) {
-		if (test_bit(i, &data->cpu_event)) {
-			/* double up() is not a problem, user space will see no
-			 * records for the additional "events".
-			 */
-			up(&data->sem);
-		}
+	if (!cpus_empty(data->cpu_event)) {
+		spin_lock_irqsave(&data_saved_lock, flags);
+		salinfo_work_to_do(data);
+		spin_unlock_irqrestore(&data_saved_lock, flags);
 	}
 }
 
-static void 
+static void
 salinfo_timeout (unsigned long arg)
 {
+	ia64_mlogbuf_dump();
 	salinfo_timeout_check(salinfo_data + SAL_INFO_TYPE_MCA);
 	salinfo_timeout_check(salinfo_data + SAL_INFO_TYPE_INIT);
 	salinfo_timer.expires = jiffies + SALINFO_TIMER_DELAY;
@@ -281,40 +302,41 @@ salinfo_event_open(struct inode *inode, struct file *file)
 static ssize_t
 salinfo_event_read(struct file *file, char __user *buffer, size_t count, loff_t *ppos)
 {
-	struct inode *inode = file->f_dentry->d_inode;
-	struct proc_dir_entry *entry = PDE(inode);
-	struct salinfo_data *data = entry->data;
+	struct salinfo_data *data = PDE_DATA(file_inode(file));
 	char cmd[32];
 	size_t size;
 	int i, n, cpu = -1;
 
 retry:
-	if (down_trylock(&data->sem)) {
+	if (cpus_empty(data->cpu_event) && down_trylock(&data->mutex)) {
 		if (file->f_flags & O_NONBLOCK)
 			return -EAGAIN;
-		if (down_interruptible(&data->sem))
-			return -ERESTARTSYS;
+		if (down_interruptible(&data->mutex))
+			return -EINTR;
 	}
 
 	n = data->cpu_check;
-	for (i = 0; i < NR_CPUS; i++) {
-		if (test_bit(n, &data->cpu_event) && cpu_online(n)) {
+	for (i = 0; i < nr_cpu_ids; i++) {
+		if (cpu_isset(n, data->cpu_event)) {
+			if (!cpu_online(n)) {
+				cpu_clear(n, data->cpu_event);
+				continue;
+			}
 			cpu = n;
 			break;
 		}
-		if (++n == NR_CPUS)
+		if (++n == nr_cpu_ids)
 			n = 0;
 	}
 
 	if (cpu == -1)
 		goto retry;
 
-	/* events are sticky until the user says "clear" */
-	up(&data->sem);
+	ia64_mlogbuf_dump();
 
 	/* for next read, start checking at next CPU */
 	data->cpu_check = cpu;
-	if (++data->cpu_check == NR_CPUS)
+	if (++data->cpu_check == nr_cpu_ids)
 		data->cpu_check = 0;
 
 	snprintf(cmd, sizeof(cmd), "read %d\n", cpu);
@@ -328,16 +350,16 @@ retry:
 	return size;
 }
 
-static struct file_operations salinfo_event_fops = {
+static const struct file_operations salinfo_event_fops = {
 	.open  = salinfo_event_open,
 	.read  = salinfo_event_read,
+	.llseek = noop_llseek,
 };
 
 static int
 salinfo_log_open(struct inode *inode, struct file *file)
 {
-	struct proc_dir_entry *entry = PDE(inode);
-	struct salinfo_data *data = entry->data;
+	struct salinfo_data *data = PDE_DATA(inode);
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
@@ -362,8 +384,7 @@ salinfo_log_open(struct inode *inode, struct file *file)
 static int
 salinfo_log_release(struct inode *inode, struct file *file)
 {
-	struct proc_dir_entry *entry = PDE(inode);
-	struct salinfo_data *data = entry->data;
+	struct salinfo_data *data = PDE_DATA(inode);
 
 	if (data->state == STATE_NO_DATA) {
 		vfree(data->log_buffer);
@@ -380,13 +401,10 @@ salinfo_log_release(struct inode *inode, struct file *file)
 static void
 call_on_cpu(int cpu, void (*fn)(void *), void *arg)
 {
-	cpumask_t save_cpus_allowed, new_cpus_allowed;
-	memcpy(&save_cpus_allowed, &current->cpus_allowed, sizeof(save_cpus_allowed));
-	memset(&new_cpus_allowed, 0, sizeof(new_cpus_allowed));
-	set_bit(cpu, &new_cpus_allowed);
-	set_cpus_allowed(current, new_cpus_allowed);
+	cpumask_t save_cpus_allowed = current->cpus_allowed;
+	set_cpus_allowed_ptr(current, cpumask_of(cpu));
 	(*fn)(arg);
-	set_cpus_allowed(current, save_cpus_allowed);
+	set_cpus_allowed_ptr(current, &save_cpus_allowed);
 }
 
 static void
@@ -432,19 +450,17 @@ retry:
 	if (!data->saved_num)
 		call_on_cpu(cpu, salinfo_log_read_cpu, data);
 	if (!data->log_size) {
-	        data->state = STATE_NO_DATA;
-	        clear_bit(cpu, &data->cpu_event);
+		data->state = STATE_NO_DATA;
+		cpu_clear(cpu, data->cpu_event);
 	} else {
-	        data->state = STATE_LOG_RECORD;
+		data->state = STATE_LOG_RECORD;
 	}
 }
 
 static ssize_t
 salinfo_log_read(struct file *file, char __user *buffer, size_t count, loff_t *ppos)
 {
-	struct inode *inode = file->f_dentry->d_inode;
-	struct proc_dir_entry *entry = PDE(inode);
-	struct salinfo_data *data = entry->data;
+	struct salinfo_data *data = PDE_DATA(file_inode(file));
 	u8 *buf;
 	u64 bufsize;
 
@@ -472,36 +488,38 @@ static int
 salinfo_log_clear(struct salinfo_data *data, int cpu)
 {
 	sal_log_record_header_t *rh;
+	unsigned long flags;
+	spin_lock_irqsave(&data_saved_lock, flags);
 	data->state = STATE_NO_DATA;
-	if (!test_bit(cpu, &data->cpu_event))
+	if (!cpu_isset(cpu, data->cpu_event)) {
+		spin_unlock_irqrestore(&data_saved_lock, flags);
 		return 0;
-	down(&data->sem);
-	clear_bit(cpu, &data->cpu_event);
+	}
+	cpu_clear(cpu, data->cpu_event);
 	if (data->saved_num) {
-		unsigned long flags;
-		spin_lock_irqsave(&data_saved_lock, flags);
-		shift1_data_saved(data, data->saved_num - 1 );
+		shift1_data_saved(data, data->saved_num - 1);
 		data->saved_num = 0;
-		spin_unlock_irqrestore(&data_saved_lock, flags);
 	}
+	spin_unlock_irqrestore(&data_saved_lock, flags);
 	rh = (sal_log_record_header_t *)(data->log_buffer);
 	/* Corrected errors have already been cleared from SAL */
 	if (rh->severity != sal_log_severity_corrected)
 		call_on_cpu(cpu, salinfo_log_clear_cpu, data);
 	/* clearing a record may make a new record visible */
 	salinfo_log_new_read(cpu, data);
-	if (data->state == STATE_LOG_RECORD &&
-	    !test_and_set_bit(cpu,  &data->cpu_event))
-		up(&data->sem);
+	if (data->state == STATE_LOG_RECORD) {
+		spin_lock_irqsave(&data_saved_lock, flags);
+		cpu_set(cpu, data->cpu_event);
+		salinfo_work_to_do(data);
+		spin_unlock_irqrestore(&data_saved_lock, flags);
+	}
 	return 0;
 }
 
 static ssize_t
 salinfo_log_write(struct file *file, const char __user *buffer, size_t count, loff_t *ppos)
 {
-	struct inode *inode = file->f_dentry->d_inode;
-	struct proc_dir_entry *entry = PDE(inode);
-	struct salinfo_data *data = entry->data;
+	struct salinfo_data *data = PDE_DATA(file_inode(file));
 	char cmd[32];
 	size_t size;
 	u32 offset;
@@ -542,11 +560,59 @@ salinfo_log_write(struct file *file, const char __user *buffer, size_t count, lo
 	return count;
 }
 
-static struct file_operations salinfo_data_fops = {
+static const struct file_operations salinfo_data_fops = {
 	.open    = salinfo_log_open,
 	.release = salinfo_log_release,
 	.read    = salinfo_log_read,
 	.write   = salinfo_log_write,
+	.llseek  = default_llseek,
+};
+
+static int
+salinfo_cpu_callback(struct notifier_block *nb, unsigned long action, void *hcpu)
+{
+	unsigned int i, cpu = (unsigned long)hcpu;
+	unsigned long flags;
+	struct salinfo_data *data;
+	switch (action) {
+	case CPU_ONLINE:
+	case CPU_ONLINE_FROZEN:
+		spin_lock_irqsave(&data_saved_lock, flags);
+		for (i = 0, data = salinfo_data;
+		     i < ARRAY_SIZE(salinfo_data);
+		     ++i, ++data) {
+			cpu_set(cpu, data->cpu_event);
+			salinfo_work_to_do(data);
+		}
+		spin_unlock_irqrestore(&data_saved_lock, flags);
+		break;
+	case CPU_DEAD:
+	case CPU_DEAD_FROZEN:
+		spin_lock_irqsave(&data_saved_lock, flags);
+		for (i = 0, data = salinfo_data;
+		     i < ARRAY_SIZE(salinfo_data);
+		     ++i, ++data) {
+			struct salinfo_data_saved *data_saved;
+			int j;
+			for (j = ARRAY_SIZE(data->data_saved) - 1, data_saved = data->data_saved + j;
+			     j >= 0;
+			     --j, --data_saved) {
+				if (data_saved->buffer && data_saved->cpu == cpu) {
+					shift1_data_saved(data, j);
+				}
+			}
+			cpu_clear(cpu, data->cpu_event);
+		}
+		spin_unlock_irqrestore(&data_saved_lock, flags);
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block salinfo_cpu_notifier =
+{
+	.notifier_call = salinfo_cpu_callback,
+	.priority = 0,
 };
 
 static int __init
@@ -556,7 +622,7 @@ salinfo_init(void)
 	struct proc_dir_entry **sdir = salinfo_proc_entries; /* keeps track of every entry */
 	struct proc_dir_entry *dir, *entry;
 	struct salinfo_data *data;
-	int i, j, online;
+	int i, j;
 
 	salinfo_dir = proc_mkdir("sal", NULL);
 	if (!salinfo_dir)
@@ -564,39 +630,36 @@ salinfo_init(void)
 
 	for (i=0; i < NR_SALINFO_ENTRIES; i++) {
 		/* pass the feature bit in question as misc data */
-		*sdir++ = create_proc_read_entry (salinfo_entries[i].name, 0, salinfo_dir,
-						  salinfo_read, (void *)salinfo_entries[i].feature);
+		*sdir++ = proc_create_data(salinfo_entries[i].name, 0, salinfo_dir,
+					   &proc_salinfo_fops,
+					   (void *)salinfo_entries[i].feature);
 	}
 
+	cpu_notifier_register_begin();
+
 	for (i = 0; i < ARRAY_SIZE(salinfo_log_name); i++) {
 		data = salinfo_data + i;
 		data->type = i;
-		sema_init(&data->sem, 0);
+		sema_init(&data->mutex, 1);
 		dir = proc_mkdir(salinfo_log_name[i], salinfo_dir);
 		if (!dir)
 			continue;
 
-		entry = create_proc_entry("event", S_IRUSR, dir);
+		entry = proc_create_data("event", S_IRUSR, dir,
+					 &salinfo_event_fops, data);
 		if (!entry)
 			continue;
-		entry->data = data;
-		entry->proc_fops = &salinfo_event_fops;
 		*sdir++ = entry;
 
-		entry = create_proc_entry("data", S_IRUSR | S_IWUSR, dir);
+		entry = proc_create_data("data", S_IRUSR | S_IWUSR, dir,
+					 &salinfo_data_fops, data);
 		if (!entry)
 			continue;
-		entry->data = data;
-		entry->proc_fops = &salinfo_data_fops;
 		*sdir++ = entry;
 
 		/* we missed any events before now */
-		online = 0;
-		for_each_online_cpu(j) {
-			set_bit(j, &data->cpu_event);
-			++online;
-		}
-		sema_init(&data->sem, online);
+		for_each_online_cpu(j)
+			cpu_set(j, data->cpu_event);
 
 		*sdir++ = dir;
 	}
@@ -608,6 +671,10 @@ salinfo_init(void)
 	salinfo_timer.function = &salinfo_timeout;
 	add_timer(&salinfo_timer);
 
+	__register_hotcpu_notifier(&salinfo_cpu_notifier);
+
+	cpu_notifier_register_done();
+
 	return 0;
 }
 
@@ -615,22 +682,23 @@ salinfo_init(void)
  * 'data' contains an integer that corresponds to the feature we're
  * testing
  */
-static int
-salinfo_read(char *page, char **start, off_t off, int count, int *eof, void *data)
+static int proc_salinfo_show(struct seq_file *m, void *v)
 {
-	int len = 0;
-
-	len = sprintf(page, (sal_platform_features & (unsigned long)data) ? "1\n" : "0\n");
-
-	if (len <= off+count) *eof = 1;
-
-	*start = page + off;
-	len   -= off;
-
-	if (len>count) len = count;
-	if (len<0) len = 0;
+	unsigned long data = (unsigned long)v;
+	seq_puts(m, (sal_platform_features & data) ? "1\n" : "0\n");
+	return 0;
+}
 
-	return len;
+static int proc_salinfo_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, proc_salinfo_show, PDE_DATA(inode));
 }
 
+static const struct file_operations proc_salinfo_fops = {
+	.open		= proc_salinfo_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 module_init(salinfo_init);