async: Asynchronous function calls to speed up kernel boot

Right now, most of the kernel boot is strictly synchronous, such that various hardware delays are done sequentially. In order to make the kernel boot faster, this patch introduces infrastructure to allow doing some of the initialization steps asynchronously, which will hide significant portions of the hardware delays in practice. In order to not change device order and other similar observables, this patch does NOT do full parallel initialization. Rather, it operates more in the way an out of order CPU does; the work may be done out of order and asynchronous, but the observable effects (instruction retiring for the CPU) are still done in the original sequence. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>

async: Asynchronous function calls to speed up kernel boot
Right now, most of the kernel boot is strictly synchronous, such that various hardware delays are done sequentially. In order to make the kernel boot faster, this patch introduces infrastructure to allow doing some of the initialization steps asynchronously, which will hide significant portions of the hardware delays in practice. In order to not change device order and other similar observables, this patch does NOT do full parallel initialization. Rather, it operates more in the way an out of order CPU does; the work may be done out of order and asynchronous, but the observable effects (instruction retiring for the CPU) are still done in the original sequence. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Arjan van de Ven
1 parent ede6f5aea0
Showing 7 changed files with 361 additions and 2 deletions Side-by-side Diff
include/linux/async.h
init/do_mounts.c
init/main.c
kernel/Makefile
kernel/async.c
kernel/irq/autoprobe.c
kernel/module.c
+/*
+ * async.h: Asynchronous function calls for boot performance
+ *
+ * (C) Copyright 2009 Intel Corporation
+ * Author: Arjan van de Ven <arjan@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+
+#include <linux/types.h>
+#include <linux/list.h>
+
+typedef u64 async_cookie_t;
+typedef void (async_func_ptr) (void *data, async_cookie_t cookie);
+
+extern async_cookie_t async_schedule(async_func_ptr *ptr, void *data);
+extern async_cookie_t async_schedule_special(async_func_ptr *ptr, void *data, struct list_head *list);
+extern void async_synchronize_full(void);
+extern void async_synchronize_full_special(struct list_head *list);
+extern void async_synchronize_cookie(async_cookie_t cookie);
+extern void async_synchronize_cookie_special(async_cookie_t cookie, struct list_head *list);
@@ -13,6 +13,7 @@
 #include <linux/init.h>
 #include <linux/fs.h>
 #include <linux/initrd.h>
+#include <linux/async.h>
  
 #include <linux/nfs_fs.h>
 #include <linux/nfs_fs_sb.h>
@@ -372,6 +373,7 @@
 	/* wait for the known devices to complete their probing */
 	while (driver_probe_done() != 0)
 		msleep(100);
+	async_synchronize_full();
  
 	md_run_setup();
  
@@ -62,6 +62,7 @@
 #include <linux/signal.h>
 #include <linux/idr.h>
 #include <linux/ftrace.h>
+#include <linux/async.h>
 #include <trace/boot.h>
  
 #include <asm/io.h>
@@ -684,7 +685,7 @@
 	rest_init();
 }
  
-static int initcall_debug;
+int initcall_debug;
 core_param(initcall_debug, initcall_debug, bool, 0644);
  
 int do_one_initcall(initcall_t fn)
@@ -785,6 +786,8 @@
  */
 static noinline int init_post(void)
 {
+	/* need to finish all async __init code before freeing the memory */
+	async_synchronize_full();
 	free_initmem();
 	unlock_kernel();
 	mark_rodata_ro();
@@ -9,7 +9,8 @@
 	    rcupdate.o extable.o params.o posix-timers.o \
 	    kthread.o wait.o kfifo.o sys_ni.o posix-cpu-timers.o mutex.o \
 	    hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \
-	    notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o
+	    notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o \
+	    async.o
  
 ifdef CONFIG_FUNCTION_TRACER
 # Do not trace debug files and internal ftrace files
+/*
+ * async.c: Asynchronous function calls for boot performance
+ *
+ * (C) Copyright 2009 Intel Corporation
+ * Author: Arjan van de Ven <arjan@linux.intel.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2
+ * of the License.
+ */
+
+
+/*
+
+Goals and Theory of Operation
+
+The primary goal of this feature is to reduce the kernel boot time,
+by doing various independent hardware delays and discovery operations
+decoupled and not strictly serialized.
+
+More specifically, the asynchronous function call concept allows
+certain operations (primarily during system boot) to happen
+asynchronously, out of order, while these operations still
+have their externally visible parts happen sequentially and in-order.
+(not unlike how out-of-order CPUs retire their instructions in order)
+
+Key to the asynchronous function call implementation is the concept of
+a "sequence cookie" (which, although it has an abstracted type, can be
+thought of as a monotonically incrementing number).
+
+The async core will assign each scheduled event such a sequence cookie and
+pass this to the called functions.
+
+The asynchronously called function should before doing a globally visible
+operation, such as registering device numbers, call the
+async_synchronize_cookie() function and pass in its own cookie. The
+async_synchronize_cookie() function will make sure that all asynchronous
+operations that were scheduled prior to the operation corresponding with the
+cookie have completed.
+
+Subsystem/driver initialization code that scheduled asynchronous probe
+functions, but which shares global resources with other drivers/subsystems
+that do not use the asynchronous call feature, need to do a full
+synchronization with the async_synchronize_full() function, before returning
+from their init function. This is to maintain strict ordering between the
+asynchronous and synchronous parts of the kernel.
+
+*/
+
+#include <linux/async.h>
+#include <linux/module.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/init.h>
+#include <linux/kthread.h>
+#include <asm/atomic.h>
+
+static async_cookie_t next_cookie = 1;
+
+#define MAX_THREADS	256
+#define MAX_WORK	32768
+
+static LIST_HEAD(async_pending);
+static LIST_HEAD(async_running);
+static DEFINE_SPINLOCK(async_lock);
+
+struct async_entry {
+	struct list_head list;
+	async_cookie_t   cookie;
+	async_func_ptr	 *func;
+	void             *data;
+	struct list_head *running;
+};
+
+static DECLARE_WAIT_QUEUE_HEAD(async_done);
+static DECLARE_WAIT_QUEUE_HEAD(async_new);
+
+static atomic_t entry_count;
+static atomic_t thread_count;
+
+extern int initcall_debug;
+
+
+/*
+ * MUST be called with the lock held!
+ */
+static async_cookie_t  __lowest_in_progress(struct list_head *running)
+{
+	struct async_entry *entry;
+	if (!list_empty(&async_pending)) {
+		entry = list_first_entry(&async_pending,
+			struct async_entry, list);
+		return entry->cookie;
+	} else if (!list_empty(running)) {
+		entry = list_first_entry(running,
+			struct async_entry, list);
+		return entry->cookie;
+	} else {
+		/* nothing in progress... next_cookie is "infinity" */
+		return next_cookie;
+	}
+
+}
+/*
+ * pick the first pending entry and run it
+ */
+static void run_one_entry(void)
+{
+	unsigned long flags;
+	struct async_entry *entry;
+	ktime_t calltime, delta, rettime;
+
+	/* 1) pick one task from the pending queue */
+
+	spin_lock_irqsave(&async_lock, flags);
+	if (list_empty(&async_pending))
+		goto out;
+	entry = list_first_entry(&async_pending, struct async_entry, list);
+
+	/* 2) move it to the running queue */
+	list_del(&entry->list);
+	list_add_tail(&entry->list, &async_running);
+	spin_unlock_irqrestore(&async_lock, flags);
+
+	/* 3) run it (and print duration)*/
+	if (initcall_debug) {
+		printk("calling  %lli_%pF @ %i\n", entry->cookie, entry->func, task_pid_nr(current));
+		calltime = ktime_get();
+	}
+	entry->func(entry->data, entry->cookie);
+	if (initcall_debug) {
+		rettime = ktime_get();
+		delta = ktime_sub(rettime, calltime);
+		printk("initcall %lli_%pF returned 0 after %lld usecs\n", entry->cookie,
+			entry->func, ktime_to_ns(delta) >> 10);
+	}
+
+	/* 4) remove it from the running queue */
+	spin_lock_irqsave(&async_lock, flags);
+	list_del(&entry->list);
+
+	/* 5) free the entry  */
+	kfree(entry);
+	atomic_dec(&entry_count);
+
+	spin_unlock_irqrestore(&async_lock, flags);
+
+	/* 6) wake up any waiters. */
+	wake_up(&async_done);
+	return;
+
+out:
+	spin_unlock_irqrestore(&async_lock, flags);
+}
+
+
+static async_cookie_t __async_schedule(async_func_ptr *ptr, void *data, struct list_head *running)
+{
+	struct async_entry *entry;
+	unsigned long flags;
+	async_cookie_t newcookie;
+	
+
+	/* allow irq-off callers */
+	entry = kzalloc(sizeof(struct async_entry), GFP_ATOMIC);
+
+	/*
+	 * If we're out of memory or if there's too much work
+	 * pending already, we execute synchronously.
+	 */
+	if (!entry || atomic_read(&entry_count) > MAX_WORK) {
+		kfree(entry);
+		spin_lock_irqsave(&async_lock, flags);
+		newcookie = next_cookie++;
+		spin_unlock_irqrestore(&async_lock, flags);
+
+		/* low on memory.. run synchronously */
+		ptr(data, newcookie);
+		return newcookie;
+	}
+	entry->func = ptr;
+	entry->data = data;
+	entry->running = running;
+
+	spin_lock_irqsave(&async_lock, flags);
+	newcookie = entry->cookie = next_cookie++;
+	list_add_tail(&entry->list, &async_pending);
+	atomic_inc(&entry_count);
+	spin_unlock_irqrestore(&async_lock, flags);
+	wake_up(&async_new);
+	return newcookie;
+}
+
+async_cookie_t async_schedule(async_func_ptr *ptr, void *data)
+{
+	return __async_schedule(ptr, data, &async_pending);
+}
+EXPORT_SYMBOL_GPL(async_schedule);
+
+async_cookie_t async_schedule_special(async_func_ptr *ptr, void *data, struct list_head *running)
+{
+	return __async_schedule(ptr, data, running);
+}
+EXPORT_SYMBOL_GPL(async_schedule_special);
+
+void async_synchronize_full(void)
+{
+	async_synchronize_cookie(next_cookie);
+}
+EXPORT_SYMBOL_GPL(async_synchronize_full);
+
+void async_synchronize_full_special(struct list_head *list)
+{
+	async_synchronize_cookie_special(next_cookie, list);
+}
+EXPORT_SYMBOL_GPL(async_synchronize_full_special);
+
+void async_synchronize_cookie_special(async_cookie_t cookie, struct list_head *running)
+{
+	ktime_t starttime, delta, endtime;
+
+	if (initcall_debug) {
+		printk("async_waiting @ %i\n", task_pid_nr(current));
+		starttime = ktime_get();
+	}
+
+	wait_event(async_done, __lowest_in_progress(running) >= cookie);
+
+	if (initcall_debug) {
+		endtime = ktime_get();
+		delta = ktime_sub(endtime, starttime);
+
+		printk("async_continuing @ %i after %lli usec\n",
+			task_pid_nr(current), ktime_to_ns(delta) >> 10);
+	}
+}
+EXPORT_SYMBOL_GPL(async_synchronize_cookie_special);
+
+void async_synchronize_cookie(async_cookie_t cookie)
+{
+	async_synchronize_cookie_special(cookie, &async_running);
+}
+EXPORT_SYMBOL_GPL(async_synchronize_cookie);
+
+
+static int async_thread(void *unused)
+{
+	DECLARE_WAITQUEUE(wq, current);
+	add_wait_queue(&async_new, &wq);
+
+	while (!kthread_should_stop()) {
+		int ret = HZ;
+		set_current_state(TASK_INTERRUPTIBLE);
+		/*
+		 * check the list head without lock.. false positives
+		 * are dealt with inside run_one_entry() while holding
+		 * the lock.
+		 */
+		rmb();
+		if (!list_empty(&async_pending))
+			run_one_entry();
+		else
+			ret = schedule_timeout(HZ);
+
+		if (ret == 0) {
+			/*
+			 * we timed out, this means we as thread are redundant.
+			 * we sign off and die, but we to avoid any races there
+			 * is a last-straw check to see if work snuck in.
+			 */
+			atomic_dec(&thread_count);
+			wmb(); /* manager must see our departure first */
+			if (list_empty(&async_pending))
+				break;
+			/*
+			 * woops work came in between us timing out and us
+			 * signing off; we need to stay alive and keep working.
+			 */
+			atomic_inc(&thread_count);
+		}
+	}
+	remove_wait_queue(&async_new, &wq);
+
+	return 0;
+}
+
+static int async_manager_thread(void *unused)
+{
+	DECLARE_WAITQUEUE(wq, current);
+	add_wait_queue(&async_new, &wq);
+
+	while (!kthread_should_stop()) {
+		int tc, ec;
+
+		set_current_state(TASK_INTERRUPTIBLE);
+
+		tc = atomic_read(&thread_count);
+		rmb();
+		ec = atomic_read(&entry_count);
+
+		while (tc < ec && tc < MAX_THREADS) {
+			kthread_run(async_thread, NULL, "async/%i", tc);
+			atomic_inc(&thread_count);
+			tc++;
+		}
+
+		schedule();
+	}
+	remove_wait_queue(&async_new, &wq);
+
+	return 0;
+}
+
+static int __init async_init(void)
+{
+	kthread_run(async_manager_thread, NULL, "async/mgr");
+	return 0;
+}
+
+core_initcall(async_init);
@@ -10,6 +10,7 @@
 #include <linux/module.h>
 #include <linux/interrupt.h>
 #include <linux/delay.h>
+#include <linux/async.h>
  
 #include "internals.h"
  
@@ -34,6 +35,10 @@
 	unsigned int status;
 	int i;
  
+	/*
+	 * quiesce the kernel, or at least the asynchronous portion
+	 */
+	async_synchronize_full();
 	mutex_lock(&probing_active);
 	/*
 	 * something may have generated an irq long ago and we want to
@@ -50,6 +50,7 @@
 #include <asm/sections.h>
 #include <linux/tracepoint.h>
 #include <linux/ftrace.h>
+#include <linux/async.h>
  
 #if 0
 #define DEBUGP printk
@@ -816,6 +817,7 @@
 		mod->exit();
 	blocking_notifier_call_chain(&module_notify_list,
 				     MODULE_STATE_GOING, mod);
+	async_synchronize_full();
 	mutex_lock(&module_mutex);
 	/* Store the name of the last unloaded module for diagnostic purposes */
 	strlcpy(last_unloaded_module, mod->name, sizeof(last_unloaded_module));
	1	+/*
	2	+ * async.h: Asynchronous function calls for boot performance
	3	+ *
	4	+ * (C) Copyright 2009 Intel Corporation
	5	+ * Author: Arjan van de Ven <arjan@linux.intel.com>
	6	+ *
	7	+ * This program is free software; you can redistribute it and/or
	8	+ * modify it under the terms of the GNU General Public License
	9	+ * as published by the Free Software Foundation; version 2
	10	+ * of the License.
	11	+ */
	12	+
	13	+#include <linux/types.h>
	14	+#include <linux/list.h>
	15	+
	16	+typedef u64 async_cookie_t;
	17	+typedef void (async_func_ptr) (void *data, async_cookie_t cookie);
	18	+
	19	+extern async_cookie_t async_schedule(async_func_ptr ptr, void data);
	20	+extern async_cookie_t async_schedule_special(async_func_ptr ptr, void data, struct list_head *list);
	21	+extern void async_synchronize_full(void);
	22	+extern void async_synchronize_full_special(struct list_head *list);
	23	+extern void async_synchronize_cookie(async_cookie_t cookie);
	24	+extern void async_synchronize_cookie_special(async_cookie_t cookie, struct list_head *list);
...	...	@@ -13,6 +13,7 @@
13	13	#include <linux/init.h>
14	14	#include <linux/fs.h>
15	15	#include <linux/initrd.h>
	16	+#include <linux/async.h>
16	17
17	18	#include <linux/nfs_fs.h>
18	19	#include <linux/nfs_fs_sb.h>
...	...	@@ -372,6 +373,7 @@
372	373	/* wait for the known devices to complete their probing */
373	374	while (driver_probe_done() != 0)
374	375	msleep(100);
	376	+ async_synchronize_full();
375	377
376	378	md_run_setup();
377	379
...	...	@@ -62,6 +62,7 @@
62	62	#include <linux/signal.h>
63	63	#include <linux/idr.h>
64	64	#include <linux/ftrace.h>
	65	+#include <linux/async.h>
65	66	#include <trace/boot.h>
66	67
67	68	#include <asm/io.h>
...	...	@@ -684,7 +685,7 @@
684	685	rest_init();
685	686	}
686	687
687		-static int initcall_debug;
	688	+int initcall_debug;
688	689	core_param(initcall_debug, initcall_debug, bool, 0644);
689	690
690	691	int do_one_initcall(initcall_t fn)
...	...	@@ -785,6 +786,8 @@
785	786	*/
786	787	static noinline int init_post(void)
787	788	{
	789	+ /* need to finish all async __init code before freeing the memory */
	790	+ async_synchronize_full();
788	791	free_initmem();
789	792	unlock_kernel();
790	793	mark_rodata_ro();
...	...	@@ -9,7 +9,8 @@
9	9	rcupdate.o extable.o params.o posix-timers.o \
10	10	kthread.o wait.o kfifo.o sys_ni.o posix-cpu-timers.o mutex.o \
11	11	hrtimer.o rwsem.o nsproxy.o srcu.o semaphore.o \
12		- notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o
	12	+ notifier.o ksysfs.o pm_qos_params.o sched_clock.o cred.o \
	13	+ async.o
13	14
14	15	ifdef CONFIG_FUNCTION_TRACER
15	16	# Do not trace debug files and internal ftrace files
	1	+/*
	2	+ * async.c: Asynchronous function calls for boot performance
	3	+ *
	4	+ * (C) Copyright 2009 Intel Corporation
	5	+ * Author: Arjan van de Ven <arjan@linux.intel.com>
	6	+ *
	7	+ * This program is free software; you can redistribute it and/or
	8	+ * modify it under the terms of the GNU General Public License
	9	+ * as published by the Free Software Foundation; version 2
	10	+ * of the License.
	11	+ */
	12	+
	13	+
	14	+/*
	15	+
	16	+Goals and Theory of Operation
	17	+
	18	+The primary goal of this feature is to reduce the kernel boot time,
	19	+by doing various independent hardware delays and discovery operations
	20	+decoupled and not strictly serialized.
	21	+
	22	+More specifically, the asynchronous function call concept allows
	23	+certain operations (primarily during system boot) to happen
	24	+asynchronously, out of order, while these operations still
	25	+have their externally visible parts happen sequentially and in-order.
	26	+(not unlike how out-of-order CPUs retire their instructions in order)
	27	+
	28	+Key to the asynchronous function call implementation is the concept of
	29	+a "sequence cookie" (which, although it has an abstracted type, can be
	30	+thought of as a monotonically incrementing number).
	31	+
	32	+The async core will assign each scheduled event such a sequence cookie and
	33	+pass this to the called functions.
	34	+
	35	+The asynchronously called function should before doing a globally visible
	36	+operation, such as registering device numbers, call the
	37	+async_synchronize_cookie() function and pass in its own cookie. The
	38	+async_synchronize_cookie() function will make sure that all asynchronous
	39	+operations that were scheduled prior to the operation corresponding with the
	40	+cookie have completed.
	41	+
	42	+Subsystem/driver initialization code that scheduled asynchronous probe
	43	+functions, but which shares global resources with other drivers/subsystems
	44	+that do not use the asynchronous call feature, need to do a full
	45	+synchronization with the async_synchronize_full() function, before returning
	46	+from their init function. This is to maintain strict ordering between the
	47	+asynchronous and synchronous parts of the kernel.
	48	+
	49	+*/
	50	+
	51	+#include <linux/async.h>
	52	+#include <linux/module.h>
	53	+#include <linux/wait.h>
	54	+#include <linux/sched.h>
	55	+#include <linux/init.h>
	56	+#include <linux/kthread.h>
	57	+#include <asm/atomic.h>
	58	+
	59	+static async_cookie_t next_cookie = 1;
	60	+
	61	+#define MAX_THREADS 256
	62	+#define MAX_WORK 32768
	63	+
	64	+static LIST_HEAD(async_pending);
	65	+static LIST_HEAD(async_running);
	66	+static DEFINE_SPINLOCK(async_lock);
	67	+
	68	+struct async_entry {
	69	+ struct list_head list;
	70	+ async_cookie_t cookie;
	71	+ async_func_ptr *func;
	72	+ void *data;
	73	+ struct list_head *running;
	74	+};
	75	+
	76	+static DECLARE_WAIT_QUEUE_HEAD(async_done);
	77	+static DECLARE_WAIT_QUEUE_HEAD(async_new);
	78	+
	79	+static atomic_t entry_count;
	80	+static atomic_t thread_count;
	81	+
	82	+extern int initcall_debug;
	83	+
	84	+
	85	+/*
	86	+ * MUST be called with the lock held!
	87	+ */
	88	+static async_cookie_t __lowest_in_progress(struct list_head *running)
	89	+{
	90	+ struct async_entry *entry;
	91	+ if (!list_empty(&async_pending)) {
	92	+ entry = list_first_entry(&async_pending,
	93	+ struct async_entry, list);
	94	+ return entry->cookie;
	95	+ } else if (!list_empty(running)) {
	96	+ entry = list_first_entry(running,
	97	+ struct async_entry, list);
	98	+ return entry->cookie;
	99	+ } else {
	100	+ /* nothing in progress... next_cookie is "infinity" */
	101	+ return next_cookie;
	102	+ }
	103	+
	104	+}
	105	+/*
	106	+ * pick the first pending entry and run it
	107	+ */
	108	+static void run_one_entry(void)
	109	+{
	110	+ unsigned long flags;
	111	+ struct async_entry *entry;
	112	+ ktime_t calltime, delta, rettime;
	113	+
	114	+ /* 1) pick one task from the pending queue */
	115	+
	116	+ spin_lock_irqsave(&async_lock, flags);
	117	+ if (list_empty(&async_pending))
	118	+ goto out;
	119	+ entry = list_first_entry(&async_pending, struct async_entry, list);
	120	+
	121	+ /* 2) move it to the running queue */
	122	+ list_del(&entry->list);
	123	+ list_add_tail(&entry->list, &async_running);
	124	+ spin_unlock_irqrestore(&async_lock, flags);
	125	+
	126	+ /* 3) run it (and print duration)*/
	127	+ if (initcall_debug) {
	128	+ printk("calling %lli_%pF @ %i\n", entry->cookie, entry->func, task_pid_nr(current));
	129	+ calltime = ktime_get();
	130	+ }
	131	+ entry->func(entry->data, entry->cookie);
	132	+ if (initcall_debug) {
	133	+ rettime = ktime_get();
	134	+ delta = ktime_sub(rettime, calltime);
	135	+ printk("initcall %lli_%pF returned 0 after %lld usecs\n", entry->cookie,
	136	+ entry->func, ktime_to_ns(delta) >> 10);
	137	+ }
	138	+
	139	+ /* 4) remove it from the running queue */
	140	+ spin_lock_irqsave(&async_lock, flags);
	141	+ list_del(&entry->list);
	142	+
	143	+ /* 5) free the entry */
	144	+ kfree(entry);
	145	+ atomic_dec(&entry_count);
	146	+
	147	+ spin_unlock_irqrestore(&async_lock, flags);
	148	+
	149	+ /* 6) wake up any waiters. */
	150	+ wake_up(&async_done);
	151	+ return;
	152	+
	153	+out:
	154	+ spin_unlock_irqrestore(&async_lock, flags);
	155	+}
	156	+
	157	+
	158	+static async_cookie_t __async_schedule(async_func_ptr ptr, void data, struct list_head *running)
	159	+{
	160	+ struct async_entry *entry;
	161	+ unsigned long flags;
	162	+ async_cookie_t newcookie;
	163	+
	164	+
	165	+ /* allow irq-off callers */
	166	+ entry = kzalloc(sizeof(struct async_entry), GFP_ATOMIC);
	167	+
	168	+ /*
	169	+ * If we're out of memory or if there's too much work
	170	+ * pending already, we execute synchronously.
	171	+ */
	172	+ if (!entry \|\| atomic_read(&entry_count) > MAX_WORK) {
	173	+ kfree(entry);
	174	+ spin_lock_irqsave(&async_lock, flags);
	175	+ newcookie = next_cookie++;
	176	+ spin_unlock_irqrestore(&async_lock, flags);
	177	+
	178	+ /* low on memory.. run synchronously */
	179	+ ptr(data, newcookie);
	180	+ return newcookie;
	181	+ }
	182	+ entry->func = ptr;
	183	+ entry->data = data;
	184	+ entry->running = running;
	185	+
	186	+ spin_lock_irqsave(&async_lock, flags);
	187	+ newcookie = entry->cookie = next_cookie++;
	188	+ list_add_tail(&entry->list, &async_pending);
	189	+ atomic_inc(&entry_count);
	190	+ spin_unlock_irqrestore(&async_lock, flags);
	191	+ wake_up(&async_new);
	192	+ return newcookie;
	193	+}
	194	+
	195	+async_cookie_t async_schedule(async_func_ptr ptr, void data)
	196	+{
	197	+ return __async_schedule(ptr, data, &async_pending);
	198	+}
	199	+EXPORT_SYMBOL_GPL(async_schedule);
	200	+
	201	+async_cookie_t async_schedule_special(async_func_ptr ptr, void data, struct list_head *running)
	202	+{
	203	+ return __async_schedule(ptr, data, running);
	204	+}
	205	+EXPORT_SYMBOL_GPL(async_schedule_special);
	206	+
	207	+void async_synchronize_full(void)
	208	+{
	209	+ async_synchronize_cookie(next_cookie);
	210	+}
	211	+EXPORT_SYMBOL_GPL(async_synchronize_full);
	212	+
	213	+void async_synchronize_full_special(struct list_head *list)
	214	+{
	215	+ async_synchronize_cookie_special(next_cookie, list);
	216	+}
	217	+EXPORT_SYMBOL_GPL(async_synchronize_full_special);
	218	+
	219	+void async_synchronize_cookie_special(async_cookie_t cookie, struct list_head *running)
	220	+{
	221	+ ktime_t starttime, delta, endtime;
	222	+
	223	+ if (initcall_debug) {
	224	+ printk("async_waiting @ %i\n", task_pid_nr(current));
	225	+ starttime = ktime_get();
	226	+ }
	227	+
	228	+ wait_event(async_done, __lowest_in_progress(running) >= cookie);
	229	+
	230	+ if (initcall_debug) {
	231	+ endtime = ktime_get();
	232	+ delta = ktime_sub(endtime, starttime);
	233	+
	234	+ printk("async_continuing @ %i after %lli usec\n",
	235	+ task_pid_nr(current), ktime_to_ns(delta) >> 10);
	236	+ }
	237	+}
	238	+EXPORT_SYMBOL_GPL(async_synchronize_cookie_special);
	239	+
	240	+void async_synchronize_cookie(async_cookie_t cookie)
	241	+{
	242	+ async_synchronize_cookie_special(cookie, &async_running);
	243	+}
	244	+EXPORT_SYMBOL_GPL(async_synchronize_cookie);
	245	+
	246	+
	247	+static int async_thread(void *unused)
	248	+{
	249	+ DECLARE_WAITQUEUE(wq, current);
	250	+ add_wait_queue(&async_new, &wq);
	251	+
	252	+ while (!kthread_should_stop()) {
	253	+ int ret = HZ;
	254	+ set_current_state(TASK_INTERRUPTIBLE);
	255	+ /*
	256	+ * check the list head without lock.. false positives
	257	+ * are dealt with inside run_one_entry() while holding
	258	+ * the lock.
	259	+ */
	260	+ rmb();
	261	+ if (!list_empty(&async_pending))
	262	+ run_one_entry();
	263	+ else
	264	+ ret = schedule_timeout(HZ);
	265	+
	266	+ if (ret == 0) {
	267	+ /*
	268	+ * we timed out, this means we as thread are redundant.
	269	+ * we sign off and die, but we to avoid any races there
	270	+ * is a last-straw check to see if work snuck in.
	271	+ */
	272	+ atomic_dec(&thread_count);
	273	+ wmb(); /* manager must see our departure first */
	274	+ if (list_empty(&async_pending))
	275	+ break;
	276	+ /*
	277	+ * woops work came in between us timing out and us
	278	+ * signing off; we need to stay alive and keep working.
	279	+ */
	280	+ atomic_inc(&thread_count);
	281	+ }
	282	+ }
	283	+ remove_wait_queue(&async_new, &wq);
	284	+
	285	+ return 0;
	286	+}
	287	+
	288	+static int async_manager_thread(void *unused)
	289	+{
	290	+ DECLARE_WAITQUEUE(wq, current);
	291	+ add_wait_queue(&async_new, &wq);
	292	+
	293	+ while (!kthread_should_stop()) {
	294	+ int tc, ec;
	295	+
	296	+ set_current_state(TASK_INTERRUPTIBLE);
	297	+
	298	+ tc = atomic_read(&thread_count);
	299	+ rmb();
	300	+ ec = atomic_read(&entry_count);
	301	+
	302	+ while (tc < ec && tc < MAX_THREADS) {
	303	+ kthread_run(async_thread, NULL, "async/%i", tc);
	304	+ atomic_inc(&thread_count);
	305	+ tc++;
	306	+ }
	307	+
	308	+ schedule();
	309	+ }
	310	+ remove_wait_queue(&async_new, &wq);
	311	+
	312	+ return 0;
	313	+}
	314	+
	315	+static int __init async_init(void)
	316	+{
	317	+ kthread_run(async_manager_thread, NULL, "async/mgr");
	318	+ return 0;
	319	+}
	320	+
	321	+core_initcall(async_init);
...	...	@@ -10,6 +10,7 @@
10	10	#include <linux/module.h>
11	11	#include <linux/interrupt.h>
12	12	#include <linux/delay.h>
	13	+#include <linux/async.h>
13	14
14	15	#include "internals.h"
15	16
...	...	@@ -34,6 +35,10 @@
34	35	unsigned int status;
35	36	int i;
36	37
	38	+ /*
	39	+ * quiesce the kernel, or at least the asynchronous portion
	40	+ */
	41	+ async_synchronize_full();
37	42	mutex_lock(&probing_active);
38	43	/*
39	44	* something may have generated an irq long ago and we want to
...	...	@@ -50,6 +50,7 @@
50	50	#include <asm/sections.h>
51	51	#include <linux/tracepoint.h>
52	52	#include <linux/ftrace.h>
	53	+#include <linux/async.h>
53	54
54	55	#if 0
55	56	#define DEBUGP printk
...	...	@@ -816,6 +817,7 @@
816	817	mod->exit();
817	818	blocking_notifier_call_chain(&module_notify_list,
818	819	MODULE_STATE_GOING, mod);
	820	+ async_synchronize_full();
819	821	mutex_lock(&module_mutex);
820	822	/* Store the name of the last unloaded module for diagnostic purposes */
821	823	strlcpy(last_unloaded_module, mod->name, sizeof(last_unloaded_module));