async.c 10.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
/*
 * 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>
52 53
#include <linux/atomic.h>
#include <linux/ktime.h>
54
#include <linux/export.h>
55 56
#include <linux/wait.h>
#include <linux/sched.h>
57
#include <linux/slab.h>
58
#include <linux/workqueue.h>
59

60 61
#include "workqueue_internal.h"

62 63
static async_cookie_t next_cookie = 1;

64 65
#define MAX_WORK		32768
#define ASYNC_COOKIE_MAX	ULLONG_MAX	/* infinity cookie */
66

67
static LIST_HEAD(async_global_pending);	/* pending from all registered doms */
68
static ASYNC_DOMAIN(async_dfl_domain);
69 70 71
static DEFINE_SPINLOCK(async_lock);

struct async_entry {
72 73
	struct list_head	domain_list;
	struct list_head	global_list;
74 75
	struct work_struct	work;
	async_cookie_t		cookie;
76
	async_func_t		func;
77
	void			*data;
78
	struct async_domain	*domain;
79 80 81 82 83 84
};

static DECLARE_WAIT_QUEUE_HEAD(async_done);

static atomic_t entry_count;

85
static async_cookie_t lowest_in_progress(struct async_domain *domain)
86
{
87
	struct async_entry *first = NULL;
88
	async_cookie_t ret = ASYNC_COOKIE_MAX;
89 90 91
	unsigned long flags;

	spin_lock_irqsave(&async_lock, flags);
92

93 94 95 96 97 98 99 100 101
	if (domain) {
		if (!list_empty(&domain->pending))
			first = list_first_entry(&domain->pending,
					struct async_entry, domain_list);
	} else {
		if (!list_empty(&async_global_pending))
			first = list_first_entry(&async_global_pending,
					struct async_entry, global_list);
	}
102

103 104
	if (first)
		ret = first->cookie;
105

106 107 108
	spin_unlock_irqrestore(&async_lock, flags);
	return ret;
}
109

110 111 112
/*
 * pick the first pending entry and run it
 */
113
static void async_run_entry_fn(struct work_struct *work)
114
{
115 116
	struct async_entry *entry =
		container_of(work, struct async_entry, work);
117
	unsigned long flags;
118
	ktime_t uninitialized_var(calltime), delta, rettime;
119

120
	/* 1) run (and print duration) */
121
	if (initcall_debug && system_state < SYSTEM_RUNNING) {
122
		pr_debug("calling  %lli_%pF @ %i\n",
123
			(long long)entry->cookie,
124
			entry->func, task_pid_nr(current));
125 126 127
		calltime = ktime_get();
	}
	entry->func(entry->data, entry->cookie);
128
	if (initcall_debug && system_state < SYSTEM_RUNNING) {
129 130
		rettime = ktime_get();
		delta = ktime_sub(rettime, calltime);
131
		pr_debug("initcall %lli_%pF returned 0 after %lld usecs\n",
132 133 134
			(long long)entry->cookie,
			entry->func,
			(long long)ktime_to_ns(delta) >> 10);
135 136
	}

137
	/* 2) remove self from the pending queues */
138
	spin_lock_irqsave(&async_lock, flags);
139 140
	list_del_init(&entry->domain_list);
	list_del_init(&entry->global_list);
141

142
	/* 3) free the entry */
143 144 145 146 147
	kfree(entry);
	atomic_dec(&entry_count);

	spin_unlock_irqrestore(&async_lock, flags);

148
	/* 4) wake up any waiters */
149 150 151
	wake_up(&async_done);
}

152
static async_cookie_t __async_schedule(async_func_t func, void *data, struct async_domain *domain)
153 154 155 156 157 158 159 160 161 162 163 164
{
	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.
	 */
165
	if (!entry || atomic_read(&entry_count) > MAX_WORK) {
166 167 168 169 170 171
		kfree(entry);
		spin_lock_irqsave(&async_lock, flags);
		newcookie = next_cookie++;
		spin_unlock_irqrestore(&async_lock, flags);

		/* low on memory.. run synchronously */
172
		func(data, newcookie);
173 174
		return newcookie;
	}
175 176
	INIT_LIST_HEAD(&entry->domain_list);
	INIT_LIST_HEAD(&entry->global_list);
177
	INIT_WORK(&entry->work, async_run_entry_fn);
178
	entry->func = func;
179
	entry->data = data;
180
	entry->domain = domain;
181 182

	spin_lock_irqsave(&async_lock, flags);
183 184

	/* allocate cookie and queue */
185
	newcookie = entry->cookie = next_cookie++;
186 187 188 189 190

	list_add_tail(&entry->domain_list, &domain->pending);
	if (domain->registered)
		list_add_tail(&entry->global_list, &async_global_pending);

191 192
	atomic_inc(&entry_count);
	spin_unlock_irqrestore(&async_lock, flags);
193

194 195 196
	/* mark that this task has queued an async job, used by module init */
	current->flags |= PF_USED_ASYNC;

197 198 199
	/* schedule for execution */
	queue_work(system_unbound_wq, &entry->work);

200 201 202
	return newcookie;
}

203 204
/**
 * async_schedule - schedule a function for asynchronous execution
205
 * @func: function to execute asynchronously
206 207 208 209 210
 * @data: data pointer to pass to the function
 *
 * Returns an async_cookie_t that may be used for checkpointing later.
 * Note: This function may be called from atomic or non-atomic contexts.
 */
211
async_cookie_t async_schedule(async_func_t func, void *data)
212
{
213
	return __async_schedule(func, data, &async_dfl_domain);
214 215 216
}
EXPORT_SYMBOL_GPL(async_schedule);

217
/**
218
 * async_schedule_domain - schedule a function for asynchronous execution within a certain domain
219
 * @func: function to execute asynchronously
220
 * @data: data pointer to pass to the function
221
 * @domain: the domain
222 223
 *
 * Returns an async_cookie_t that may be used for checkpointing later.
224 225 226 227
 * @domain may be used in the async_synchronize_*_domain() functions to
 * wait within a certain synchronization domain rather than globally.  A
 * synchronization domain is specified via @domain.  Note: This function
 * may be called from atomic or non-atomic contexts.
228
 */
229
async_cookie_t async_schedule_domain(async_func_t func, void *data,
230
				     struct async_domain *domain)
231
{
232
	return __async_schedule(func, data, domain);
233
}
234
EXPORT_SYMBOL_GPL(async_schedule_domain);
235

236 237 238 239 240
/**
 * async_synchronize_full - synchronize all asynchronous function calls
 *
 * This function waits until all asynchronous function calls have been done.
 */
241 242
void async_synchronize_full(void)
{
243
	async_synchronize_full_domain(NULL);
244 245 246
}
EXPORT_SYMBOL_GPL(async_synchronize_full);

247 248 249 250 251 252 253 254 255 256 257 258
/**
 * async_unregister_domain - ensure no more anonymous waiters on this domain
 * @domain: idle domain to flush out of any async_synchronize_full instances
 *
 * async_synchronize_{cookie|full}_domain() are not flushed since callers
 * of these routines should know the lifetime of @domain
 *
 * Prefer ASYNC_DOMAIN_EXCLUSIVE() declarations over flushing
 */
void async_unregister_domain(struct async_domain *domain)
{
	spin_lock_irq(&async_lock);
259
	WARN_ON(!domain->registered || !list_empty(&domain->pending));
260 261 262 263 264
	domain->registered = 0;
	spin_unlock_irq(&async_lock);
}
EXPORT_SYMBOL_GPL(async_unregister_domain);

265
/**
266
 * async_synchronize_full_domain - synchronize all asynchronous function within a certain domain
267
 * @domain: the domain to synchronize
268
 *
269
 * This function waits until all asynchronous function calls for the
270
 * synchronization domain specified by @domain have been done.
271
 */
272
void async_synchronize_full_domain(struct async_domain *domain)
273
{
274
	async_synchronize_cookie_domain(ASYNC_COOKIE_MAX, domain);
275
}
276
EXPORT_SYMBOL_GPL(async_synchronize_full_domain);
277

278
/**
279
 * async_synchronize_cookie_domain - synchronize asynchronous function calls within a certain domain with cookie checkpointing
280
 * @cookie: async_cookie_t to use as checkpoint
281
 * @domain: the domain to synchronize (%NULL for all registered domains)
282
 *
283
 * This function waits until all asynchronous function calls for the
284 285
 * synchronization domain specified by @domain submitted prior to @cookie
 * have been done.
286
 */
287
void async_synchronize_cookie_domain(async_cookie_t cookie, struct async_domain *domain)
288
{
289
	ktime_t uninitialized_var(starttime), delta, endtime;
290

291
	if (initcall_debug && system_state < SYSTEM_RUNNING) {
292
		pr_debug("async_waiting @ %i\n", task_pid_nr(current));
293 294 295
		starttime = ktime_get();
	}

296
	wait_event(async_done, lowest_in_progress(domain) >= cookie);
297

298
	if (initcall_debug && system_state < SYSTEM_RUNNING) {
299 300 301
		endtime = ktime_get();
		delta = ktime_sub(endtime, starttime);

302
		pr_debug("async_continuing @ %i after %lli usec\n",
303 304
			task_pid_nr(current),
			(long long)ktime_to_ns(delta) >> 10);
305 306
	}
}
307
EXPORT_SYMBOL_GPL(async_synchronize_cookie_domain);
308

309 310 311 312 313 314 315
/**
 * async_synchronize_cookie - synchronize asynchronous function calls with cookie checkpointing
 * @cookie: async_cookie_t to use as checkpoint
 *
 * This function waits until all asynchronous function calls prior to @cookie
 * have been done.
 */
316 317
void async_synchronize_cookie(async_cookie_t cookie)
{
318
	async_synchronize_cookie_domain(cookie, &async_dfl_domain);
319 320
}
EXPORT_SYMBOL_GPL(async_synchronize_cookie);
321 322 323 324 325 326 327 328 329 330 331 332

/**
 * current_is_async - is %current an async worker task?
 *
 * Returns %true if %current is an async worker task.
 */
bool current_is_async(void)
{
	struct worker *worker = current_wq_worker();

	return worker && worker->current_func == async_run_entry_fn;
}
333
EXPORT_SYMBOL_GPL(current_is_async);