[9f8745c5] | 1 | /*
|
---|
| 2 | * Copyright (c) 2012 Adam Hraska
|
---|
| 3 | * All rights reserved.
|
---|
| 4 | *
|
---|
| 5 | * Redistribution and use in source and binary forms, with or without
|
---|
| 6 | * modification, are permitted provided that the following conditions
|
---|
| 7 | * are met:
|
---|
| 8 | *
|
---|
| 9 | * - Redistributions of source code must retain the above copyright
|
---|
| 10 | * notice, this list of conditions and the following disclaimer.
|
---|
| 11 | * - Redistributions in binary form must reproduce the above copyright
|
---|
| 12 | * notice, this list of conditions and the following disclaimer in the
|
---|
| 13 | * documentation and/or other materials provided with the distribution.
|
---|
| 14 | * - The name of the author may not be used to endorse or promote products
|
---|
| 15 | * derived from this software without specific prior written permission.
|
---|
| 16 | *
|
---|
| 17 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
|
---|
| 18 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
|
---|
| 19 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
|
---|
| 20 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
|
---|
| 21 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
|
---|
| 22 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
---|
| 23 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
---|
| 24 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
---|
| 25 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
|
---|
| 26 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
---|
| 27 | */
|
---|
| 28 |
|
---|
| 29 | /** @addtogroup generic
|
---|
| 30 | * @{
|
---|
| 31 | */
|
---|
| 32 |
|
---|
| 33 | /**
|
---|
| 34 | * @file
|
---|
| 35 | * @brief Work queue/thread pool that automatically adjusts its size
|
---|
| 36 | * depending on the current load. Queued work functions may sleep..
|
---|
| 37 | */
|
---|
| 38 |
|
---|
[8a64e81e] | 39 | #include <synch/workqueue.h>
|
---|
| 40 | #include <synch/spinlock.h>
|
---|
| 41 | #include <synch/condvar.h>
|
---|
| 42 | #include <synch/mutex.h>
|
---|
| 43 | #include <proc/thread.h>
|
---|
| 44 | #include <config.h>
|
---|
| 45 | #include <arch.h>
|
---|
| 46 | #include <cpu.h>
|
---|
| 47 | #include <macros.h>
|
---|
| 48 |
|
---|
[0d56712] | 49 | #define WORKQ_MAGIC 0xf00c1333U
|
---|
| 50 | #define WORK_ITEM_MAGIC 0xfeec1777U
|
---|
| 51 |
|
---|
[8a64e81e] | 52 |
|
---|
| 53 | struct work_queue {
|
---|
| 54 | /*
|
---|
| 55 | * Protects everything except activate_worker.
|
---|
| 56 | * Must be acquired after any thread->locks.
|
---|
| 57 | */
|
---|
| 58 | IRQ_SPINLOCK_DECLARE(lock);
|
---|
| 59 |
|
---|
| 60 | /* Activates a worker if new work arrives or if shutting down the queue. */
|
---|
| 61 | condvar_t activate_worker;
|
---|
| 62 |
|
---|
| 63 | /* Queue of work_items ready to be dispatched. */
|
---|
| 64 | list_t queue;
|
---|
| 65 |
|
---|
| 66 | /* List of worker threads. */
|
---|
| 67 | list_t workers;
|
---|
| 68 |
|
---|
| 69 | /* Number of work items queued. */
|
---|
| 70 | size_t item_cnt;
|
---|
| 71 |
|
---|
| 72 | /* Indicates the work queue is shutting down. */
|
---|
| 73 | bool stopping;
|
---|
| 74 | const char *name;
|
---|
| 75 |
|
---|
| 76 | /* Total number of created worker threads. */
|
---|
| 77 | size_t cur_worker_cnt;
|
---|
| 78 | /* Number of workers waiting for work to arrive. */
|
---|
| 79 | size_t idle_worker_cnt;
|
---|
| 80 | /* Number of idle workers signaled that have not yet been woken up. */
|
---|
| 81 | size_t activate_pending;
|
---|
| 82 | /* Number of blocked workers sleeping in work func() (ie not idle). */
|
---|
| 83 | size_t blocked_worker_cnt;
|
---|
| 84 |
|
---|
| 85 | /* Number of pending signal_worker_op() operations. */
|
---|
| 86 | size_t pending_op_cnt;
|
---|
| 87 |
|
---|
| 88 | link_t nb_link;
|
---|
[0d56712] | 89 |
|
---|
| 90 | #ifdef CONFIG_DEBUG
|
---|
| 91 | /* Magic cookie for integrity checks. Immutable. Accessed without lock. */
|
---|
| 92 | uint32_t cookie;
|
---|
| 93 | #endif
|
---|
[8a64e81e] | 94 | };
|
---|
| 95 |
|
---|
| 96 |
|
---|
| 97 | /** Min number of idle workers to keep. */
|
---|
| 98 | static size_t min_worker_cnt;
|
---|
| 99 | /** Max total number of workers - be it blocked, idle, or active. */
|
---|
| 100 | static size_t max_worker_cnt;
|
---|
| 101 | /** Max number of concurrently running active workers, ie not blocked nor idle. */
|
---|
| 102 | static size_t max_concurrent_workers;
|
---|
| 103 | /** Max number of work items per active worker before a new worker is activated.*/
|
---|
| 104 | static const size_t max_items_per_worker = 8;
|
---|
| 105 |
|
---|
| 106 | /** System wide work queue. */
|
---|
| 107 | static struct work_queue g_work_queue;
|
---|
| 108 |
|
---|
| 109 | static int booting = true;
|
---|
| 110 |
|
---|
| 111 |
|
---|
| 112 | typedef struct {
|
---|
| 113 | IRQ_SPINLOCK_DECLARE(lock);
|
---|
| 114 | condvar_t req_cv;
|
---|
| 115 | thread_t *thread;
|
---|
| 116 | list_t work_queues;
|
---|
| 117 | } nonblock_adder_t;
|
---|
| 118 |
|
---|
| 119 | static nonblock_adder_t nonblock_adder;
|
---|
| 120 |
|
---|
| 121 |
|
---|
| 122 |
|
---|
| 123 | /** Typedef a worker thread signaling operation prototype. */
|
---|
| 124 | typedef void (*signal_op_t)(struct work_queue *workq);
|
---|
| 125 |
|
---|
| 126 |
|
---|
| 127 | /* Fwd decl. */
|
---|
| 128 | static void workq_preinit(struct work_queue *workq, const char *name);
|
---|
| 129 | static bool add_worker(struct work_queue *workq);
|
---|
| 130 | static void interrupt_workers(struct work_queue *workq);
|
---|
| 131 | static void wait_for_workers(struct work_queue *workq);
|
---|
| 132 | static int _workq_enqueue(struct work_queue *workq, work_t *work_item,
|
---|
| 133 | work_func_t func, bool can_block);
|
---|
[0d56712] | 134 | static void init_work_item(work_t *work_item, work_func_t func);
|
---|
[8a64e81e] | 135 | static signal_op_t signal_worker_logic(struct work_queue *workq, bool can_block);
|
---|
| 136 | static void worker_thread(void *arg);
|
---|
| 137 | static bool dequeue_work(struct work_queue *workq, work_t **pwork_item);
|
---|
| 138 | static bool worker_unnecessary(struct work_queue *workq);
|
---|
| 139 | static void cv_wait(struct work_queue *workq);
|
---|
| 140 | static void nonblock_init(void);
|
---|
[0d56712] | 141 | static bool workq_corrupted(struct work_queue *workq);
|
---|
| 142 | static bool work_item_corrupted(work_t *work_item);
|
---|
| 143 |
|
---|
[8a64e81e] | 144 |
|
---|
| 145 | /** Creates worker thread for the system-wide worker queue. */
|
---|
| 146 | void workq_global_worker_init(void)
|
---|
| 147 | {
|
---|
| 148 | /*
|
---|
| 149 | * No need for additional synchronization. Stores to word-sized
|
---|
| 150 | * variables are atomic and the change will eventually propagate.
|
---|
| 151 | * Moreover add_worker() includes the necessary memory barriers
|
---|
| 152 | * in spinlock lock/unlock().
|
---|
| 153 | */
|
---|
| 154 | booting = false;
|
---|
| 155 |
|
---|
| 156 | nonblock_init();
|
---|
| 157 |
|
---|
| 158 | if (!add_worker(&g_work_queue))
|
---|
| 159 | panic("Could not create a single global work queue worker!\n");
|
---|
| 160 |
|
---|
| 161 | }
|
---|
| 162 |
|
---|
| 163 | /** Initializes the system wide work queue and support for other work queues. */
|
---|
| 164 | void workq_global_init(void)
|
---|
| 165 | {
|
---|
| 166 | /* Keep idle workers on 1/4-th of cpus, but at least 2 threads. */
|
---|
| 167 | min_worker_cnt = max(2, config.cpu_count / 4);
|
---|
| 168 | /* Allow max 8 sleeping work items per cpu. */
|
---|
| 169 | max_worker_cnt = max(32, 8 * config.cpu_count);
|
---|
| 170 | /* Maximum concurrency without slowing down the system. */
|
---|
| 171 | max_concurrent_workers = max(2, config.cpu_count);
|
---|
| 172 |
|
---|
| 173 | workq_preinit(&g_work_queue, "kworkq");
|
---|
| 174 | }
|
---|
| 175 |
|
---|
| 176 | /** Stops the system global work queue and waits for all work items to complete.*/
|
---|
| 177 | void workq_global_stop(void)
|
---|
| 178 | {
|
---|
| 179 | workq_stop(&g_work_queue);
|
---|
| 180 | }
|
---|
| 181 |
|
---|
| 182 | /** Creates and initializes a work queue. Returns NULL upon failure. */
|
---|
| 183 | struct work_queue * workq_create(const char *name)
|
---|
| 184 | {
|
---|
| 185 | struct work_queue *workq = malloc(sizeof(struct work_queue), 0);
|
---|
| 186 |
|
---|
| 187 | if (workq) {
|
---|
| 188 | if (workq_init(workq, name)) {
|
---|
[0d56712] | 189 | ASSERT(!workq_corrupted(workq));
|
---|
[8a64e81e] | 190 | return workq;
|
---|
| 191 | }
|
---|
| 192 |
|
---|
| 193 | free(workq);
|
---|
| 194 | }
|
---|
| 195 |
|
---|
| 196 | return NULL;
|
---|
| 197 | }
|
---|
| 198 |
|
---|
| 199 | /** Frees work queue resources and stops it if it had not been done so already.*/
|
---|
| 200 | void workq_destroy(struct work_queue *workq)
|
---|
| 201 | {
|
---|
[0d56712] | 202 | ASSERT(!workq_corrupted(workq));
|
---|
[8a64e81e] | 203 |
|
---|
| 204 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 205 | bool stopped = workq->stopping;
|
---|
| 206 | size_t running_workers = workq->cur_worker_cnt;
|
---|
| 207 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 208 |
|
---|
| 209 | if (!stopped) {
|
---|
| 210 | workq_stop(workq);
|
---|
| 211 | } else {
|
---|
| 212 | ASSERT(0 == running_workers);
|
---|
| 213 | }
|
---|
| 214 |
|
---|
[0d56712] | 215 | #ifdef CONFIG_DEBUG
|
---|
| 216 | workq->cookie = 0;
|
---|
| 217 | #endif
|
---|
| 218 |
|
---|
[8a64e81e] | 219 | free(workq);
|
---|
| 220 | }
|
---|
| 221 |
|
---|
| 222 | /** Initializes workq structure without creating any workers. */
|
---|
| 223 | static void workq_preinit(struct work_queue *workq, const char *name)
|
---|
| 224 | {
|
---|
[0d56712] | 225 | #ifdef CONFIG_DEBUG
|
---|
| 226 | workq->cookie = WORKQ_MAGIC;
|
---|
| 227 | #endif
|
---|
| 228 |
|
---|
[8a64e81e] | 229 | irq_spinlock_initialize(&workq->lock, name);
|
---|
| 230 | condvar_initialize(&workq->activate_worker);
|
---|
| 231 |
|
---|
| 232 | list_initialize(&workq->queue);
|
---|
| 233 | list_initialize(&workq->workers);
|
---|
| 234 |
|
---|
| 235 | workq->item_cnt = 0;
|
---|
| 236 | workq->stopping = false;
|
---|
| 237 | workq->name = name;
|
---|
| 238 |
|
---|
| 239 | workq->cur_worker_cnt = 1;
|
---|
| 240 | workq->idle_worker_cnt = 0;
|
---|
| 241 | workq->activate_pending = 0;
|
---|
| 242 | workq->blocked_worker_cnt = 0;
|
---|
| 243 |
|
---|
| 244 | workq->pending_op_cnt = 0;
|
---|
| 245 | link_initialize(&workq->nb_link);
|
---|
| 246 | }
|
---|
| 247 |
|
---|
| 248 | /** Initializes a work queue. Returns true if successful.
|
---|
| 249 | *
|
---|
| 250 | * Before destroying a work queue it must be stopped via
|
---|
| 251 | * workq_stop().
|
---|
| 252 | */
|
---|
| 253 | int workq_init(struct work_queue *workq, const char *name)
|
---|
| 254 | {
|
---|
| 255 | workq_preinit(workq, name);
|
---|
| 256 | return add_worker(workq);
|
---|
| 257 | }
|
---|
| 258 |
|
---|
| 259 | /** Add a new worker thread. Returns false if the thread could not be created. */
|
---|
| 260 | static bool add_worker(struct work_queue *workq)
|
---|
| 261 | {
|
---|
[0d56712] | 262 | ASSERT(!workq_corrupted(workq));
|
---|
| 263 |
|
---|
[8a64e81e] | 264 | thread_t *thread = thread_create(worker_thread, workq, TASK,
|
---|
| 265 | THREAD_FLAG_NONE, workq->name);
|
---|
| 266 |
|
---|
| 267 | if (!thread) {
|
---|
| 268 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 269 |
|
---|
| 270 | /* cur_worker_cnt proactively increased in signal_worker_logic() .*/
|
---|
| 271 | ASSERT(0 < workq->cur_worker_cnt);
|
---|
| 272 | --workq->cur_worker_cnt;
|
---|
| 273 |
|
---|
| 274 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 275 | return false;
|
---|
| 276 | }
|
---|
| 277 |
|
---|
| 278 | /* Respect lock ordering. */
|
---|
| 279 | irq_spinlock_lock(&thread->lock, true);
|
---|
| 280 | irq_spinlock_lock(&workq->lock, false);
|
---|
| 281 |
|
---|
| 282 | bool success;
|
---|
| 283 |
|
---|
| 284 | if (!workq->stopping) {
|
---|
| 285 | success = true;
|
---|
| 286 |
|
---|
| 287 | /* Try to distribute workers among cpus right away. */
|
---|
| 288 | unsigned int cpu_id = (workq->cur_worker_cnt) % config.cpu_active;
|
---|
| 289 |
|
---|
| 290 | if (!cpus[cpu_id].active)
|
---|
| 291 | cpu_id = CPU->id;
|
---|
| 292 |
|
---|
| 293 | thread->workq = workq;
|
---|
| 294 | thread->cpu = &cpus[cpu_id];
|
---|
| 295 | thread->workq_blocked = false;
|
---|
| 296 | thread->workq_idling = false;
|
---|
| 297 | link_initialize(&thread->workq_link);
|
---|
| 298 |
|
---|
| 299 | list_append(&thread->workq_link, &workq->workers);
|
---|
| 300 | } else {
|
---|
| 301 | /*
|
---|
| 302 | * Work queue is shutting down - we must not add the worker
|
---|
| 303 | * and we cannot destroy it without ready-ing it. Mark it
|
---|
| 304 | * interrupted so the worker exits right away without even
|
---|
| 305 | * touching workq.
|
---|
| 306 | */
|
---|
| 307 | success = false;
|
---|
| 308 |
|
---|
| 309 | /* cur_worker_cnt proactively increased in signal_worker() .*/
|
---|
| 310 | ASSERT(0 < workq->cur_worker_cnt);
|
---|
| 311 | --workq->cur_worker_cnt;
|
---|
| 312 | }
|
---|
| 313 |
|
---|
| 314 | irq_spinlock_unlock(&workq->lock, false);
|
---|
| 315 | irq_spinlock_unlock(&thread->lock, true);
|
---|
| 316 |
|
---|
| 317 | if (!success) {
|
---|
| 318 | thread_interrupt(thread);
|
---|
| 319 | }
|
---|
| 320 |
|
---|
| 321 | thread_ready(thread);
|
---|
| 322 |
|
---|
| 323 | return success;
|
---|
| 324 | }
|
---|
| 325 |
|
---|
| 326 | /** Shuts down the work queue. Waits for all pending work items to complete.
|
---|
| 327 | *
|
---|
| 328 | * workq_stop() may only be run once.
|
---|
| 329 | */
|
---|
| 330 | void workq_stop(struct work_queue *workq)
|
---|
| 331 | {
|
---|
[0d56712] | 332 | ASSERT(!workq_corrupted(workq));
|
---|
| 333 |
|
---|
[8a64e81e] | 334 | interrupt_workers(workq);
|
---|
| 335 | wait_for_workers(workq);
|
---|
| 336 | }
|
---|
| 337 |
|
---|
| 338 | /** Notifies worker threads the work queue is shutting down. */
|
---|
| 339 | static void interrupt_workers(struct work_queue *workq)
|
---|
| 340 | {
|
---|
| 341 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 342 |
|
---|
| 343 | /* workq_stop() may only be called once. */
|
---|
| 344 | ASSERT(!workq->stopping);
|
---|
| 345 | workq->stopping = true;
|
---|
| 346 |
|
---|
| 347 | /* Respect lock ordering - do not hold workq->lock during broadcast. */
|
---|
| 348 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 349 |
|
---|
| 350 | condvar_broadcast(&workq->activate_worker);
|
---|
| 351 | }
|
---|
| 352 |
|
---|
| 353 | /** Waits for all worker threads to exit. */
|
---|
| 354 | static void wait_for_workers(struct work_queue *workq)
|
---|
| 355 | {
|
---|
| 356 | ASSERT(!PREEMPTION_DISABLED);
|
---|
| 357 |
|
---|
| 358 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 359 |
|
---|
| 360 | list_foreach_safe(workq->workers, cur_worker, next_worker) {
|
---|
| 361 | thread_t *worker = list_get_instance(cur_worker, thread_t, workq_link);
|
---|
| 362 | list_remove(cur_worker);
|
---|
| 363 |
|
---|
| 364 | /* Wait without the lock. */
|
---|
| 365 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 366 |
|
---|
| 367 | thread_join(worker);
|
---|
| 368 | thread_detach(worker);
|
---|
| 369 |
|
---|
| 370 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 371 | }
|
---|
| 372 |
|
---|
| 373 | ASSERT(list_empty(&workq->workers));
|
---|
| 374 |
|
---|
| 375 | /* Wait for deferred add_worker_op(), signal_worker_op() to finish. */
|
---|
| 376 | while (0 < workq->cur_worker_cnt || 0 < workq->pending_op_cnt) {
|
---|
| 377 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 378 |
|
---|
| 379 | scheduler();
|
---|
| 380 |
|
---|
| 381 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 382 | }
|
---|
| 383 |
|
---|
| 384 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 385 | }
|
---|
| 386 |
|
---|
| 387 | /** Queues a function into the global wait queue without blocking.
|
---|
| 388 | *
|
---|
| 389 | * See workq_enqueue_noblock() for more details.
|
---|
| 390 | */
|
---|
| 391 | int workq_global_enqueue_noblock(work_t *work_item, work_func_t func)
|
---|
| 392 | {
|
---|
| 393 | return workq_enqueue_noblock(&g_work_queue, work_item, func);
|
---|
| 394 | }
|
---|
| 395 |
|
---|
| 396 | /** Queues a function into the global wait queue; may block.
|
---|
| 397 | *
|
---|
| 398 | * See workq_enqueue() for more details.
|
---|
| 399 | */
|
---|
| 400 | int workq_global_enqueue(work_t *work_item, work_func_t func)
|
---|
| 401 | {
|
---|
| 402 | return workq_enqueue(&g_work_queue, work_item, func);
|
---|
| 403 | }
|
---|
| 404 |
|
---|
| 405 | /** Adds a function to be invoked in a separate thread without blocking.
|
---|
| 406 | *
|
---|
| 407 | * workq_enqueue_noblock() is guaranteed not to block. It is safe
|
---|
| 408 | * to invoke from interrupt handlers.
|
---|
| 409 | *
|
---|
| 410 | * Consider using workq_enqueue() instead if at all possible. Otherwise,
|
---|
| 411 | * your work item may have to wait for previously enqueued sleeping
|
---|
| 412 | * work items to complete if you are unlucky.
|
---|
| 413 | *
|
---|
| 414 | * @param workq Work queue where to queue the work item.
|
---|
| 415 | * @param work_item Work item bookkeeping structure. Must be valid
|
---|
| 416 | * until func() is entered.
|
---|
| 417 | * @param func User supplied function to invoke in a worker thread.
|
---|
| 418 |
|
---|
| 419 | * @return false if work queue is shutting down; function is not
|
---|
| 420 | * queued for further processing.
|
---|
| 421 | * @return true Otherwise. func() will be invoked in a separate thread.
|
---|
| 422 | */
|
---|
| 423 | int workq_enqueue_noblock(struct work_queue *workq, work_t *work_item,
|
---|
| 424 | work_func_t func)
|
---|
| 425 | {
|
---|
| 426 | return _workq_enqueue(workq, work_item, func, false);
|
---|
| 427 | }
|
---|
| 428 |
|
---|
| 429 | /** Adds a function to be invoked in a separate thread; may blocking.
|
---|
| 430 | *
|
---|
| 431 | * While the workq_enqueue() is unlikely to block, it may do so if too
|
---|
| 432 | * many previous work items blocked sleeping.
|
---|
| 433 | *
|
---|
| 434 | * @param workq Work queue where to queue the work item.
|
---|
| 435 | * @param work_item Work item bookkeeping structure. Must be valid
|
---|
| 436 | * until func() is entered.
|
---|
| 437 | * @param func User supplied function to invoke in a worker thread.
|
---|
| 438 |
|
---|
| 439 | * @return false if work queue is shutting down; function is not
|
---|
| 440 | * queued for further processing.
|
---|
| 441 | * @return true Otherwise. func() will be invoked in a separate thread.
|
---|
| 442 | */
|
---|
| 443 | int workq_enqueue(struct work_queue *workq, work_t *work_item, work_func_t func)
|
---|
| 444 | {
|
---|
| 445 | return _workq_enqueue(workq, work_item, func, true);
|
---|
| 446 | }
|
---|
| 447 |
|
---|
| 448 | /** Adds a work item that will be processed by a separate worker thread.
|
---|
| 449 | *
|
---|
| 450 | * func() will be invoked in another kernel thread and may block.
|
---|
| 451 | *
|
---|
| 452 | * Prefer to call _workq_enqueue() with can_block set. Otherwise
|
---|
| 453 | * your work item may have to wait for sleeping work items to complete.
|
---|
| 454 | * If all worker threads are blocked/sleeping a new worker thread cannot
|
---|
| 455 | * be create without can_block set because creating a thread might
|
---|
| 456 | * block due to low memory conditions.
|
---|
| 457 | *
|
---|
| 458 | * @param workq Work queue where to queue the work item.
|
---|
| 459 | * @param work_item Work item bookkeeping structure. Must be valid
|
---|
| 460 | * until func() is entered.
|
---|
| 461 | * @param func User supplied function to invoke in a worker thread.
|
---|
| 462 | * @param can_block May adding this work item block?
|
---|
| 463 |
|
---|
| 464 | * @return false if work queue is shutting down; function is not
|
---|
| 465 | * queued for further processing.
|
---|
| 466 | * @return true Otherwise.
|
---|
| 467 | */
|
---|
| 468 | static int _workq_enqueue(struct work_queue *workq, work_t *work_item,
|
---|
| 469 | work_func_t func, bool can_block)
|
---|
| 470 | {
|
---|
[0d56712] | 471 | ASSERT(!workq_corrupted(workq));
|
---|
| 472 |
|
---|
[8a64e81e] | 473 | bool success = true;
|
---|
| 474 | signal_op_t signal_op = NULL;
|
---|
| 475 |
|
---|
| 476 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 477 |
|
---|
| 478 | if (workq->stopping) {
|
---|
| 479 | success = false;
|
---|
| 480 | } else {
|
---|
[0d56712] | 481 | init_work_item(work_item, func);
|
---|
[8a64e81e] | 482 | list_append(&work_item->queue_link, &workq->queue);
|
---|
| 483 | ++workq->item_cnt;
|
---|
| 484 | success = true;
|
---|
| 485 |
|
---|
| 486 | if (!booting) {
|
---|
| 487 | signal_op = signal_worker_logic(workq, can_block);
|
---|
| 488 | } else {
|
---|
| 489 | /*
|
---|
| 490 | * During boot there are no workers to signal. Just queue
|
---|
| 491 | * the work and let future workers take care of it.
|
---|
| 492 | */
|
---|
| 493 | }
|
---|
| 494 | }
|
---|
| 495 |
|
---|
| 496 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 497 |
|
---|
| 498 | if (signal_op) {
|
---|
| 499 | signal_op(workq);
|
---|
| 500 | }
|
---|
| 501 |
|
---|
| 502 | return success;
|
---|
| 503 | }
|
---|
| 504 |
|
---|
[0d56712] | 505 | /** Prepare an item to be added to the work item queue. */
|
---|
| 506 | static void init_work_item(work_t *work_item, work_func_t func)
|
---|
| 507 | {
|
---|
| 508 | #ifdef CONFIG_DEBUG
|
---|
| 509 | work_item->cookie = WORK_ITEM_MAGIC;
|
---|
| 510 | #endif
|
---|
| 511 |
|
---|
| 512 | link_initialize(&work_item->queue_link);
|
---|
| 513 | work_item->func = func;
|
---|
| 514 | }
|
---|
| 515 |
|
---|
[8a64e81e] | 516 | /** Returns the number of workers running work func() that are not blocked. */
|
---|
| 517 | static size_t active_workers_now(struct work_queue *workq)
|
---|
| 518 | {
|
---|
| 519 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 520 |
|
---|
| 521 | /* Workers blocked are sleeping in the work function (ie not idle). */
|
---|
| 522 | ASSERT(workq->blocked_worker_cnt <= workq->cur_worker_cnt);
|
---|
| 523 | /* Idle workers are waiting for more work to arrive in condvar_wait. */
|
---|
| 524 | ASSERT(workq->idle_worker_cnt <= workq->cur_worker_cnt);
|
---|
| 525 |
|
---|
| 526 | /* Idle + blocked workers == sleeping worker threads. */
|
---|
| 527 | size_t sleeping_workers = workq->blocked_worker_cnt + workq->idle_worker_cnt;
|
---|
| 528 |
|
---|
| 529 | ASSERT(sleeping_workers <= workq->cur_worker_cnt);
|
---|
| 530 | /* Workers pending activation are idle workers not yet given a time slice. */
|
---|
| 531 | ASSERT(workq->activate_pending <= workq->idle_worker_cnt);
|
---|
| 532 |
|
---|
| 533 | /*
|
---|
| 534 | * Workers actively running the work func() this very moment and
|
---|
| 535 | * are neither blocked nor idle. Exclude ->activate_pending workers
|
---|
| 536 | * since they will run their work func() once they get a time slice
|
---|
| 537 | * and are not running it right now.
|
---|
| 538 | */
|
---|
| 539 | return workq->cur_worker_cnt - sleeping_workers;
|
---|
| 540 | }
|
---|
| 541 |
|
---|
| 542 | /**
|
---|
| 543 | * Returns the number of workers that are running or are about to run work
|
---|
| 544 | * func() and that are not blocked.
|
---|
| 545 | */
|
---|
| 546 | static size_t active_workers(struct work_queue *workq)
|
---|
| 547 | {
|
---|
| 548 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 549 |
|
---|
| 550 | /*
|
---|
| 551 | * Workers actively running the work func() and are neither blocked nor
|
---|
| 552 | * idle. ->activate_pending workers will run their work func() once they
|
---|
| 553 | * get a time slice after waking from a condvar wait, so count them
|
---|
| 554 | * as well.
|
---|
| 555 | */
|
---|
| 556 | return active_workers_now(workq) + workq->activate_pending;
|
---|
| 557 | }
|
---|
| 558 |
|
---|
| 559 | static void add_worker_noblock_op(struct work_queue *workq)
|
---|
| 560 | {
|
---|
| 561 | condvar_signal(&nonblock_adder.req_cv);
|
---|
| 562 | }
|
---|
| 563 |
|
---|
| 564 | static void add_worker_op(struct work_queue *workq)
|
---|
| 565 | {
|
---|
| 566 | add_worker(workq);
|
---|
| 567 | }
|
---|
| 568 |
|
---|
| 569 | static void signal_worker_op(struct work_queue *workq)
|
---|
| 570 | {
|
---|
[0d56712] | 571 | ASSERT(!workq_corrupted(workq));
|
---|
| 572 |
|
---|
[8a64e81e] | 573 | condvar_signal(&workq->activate_worker);
|
---|
| 574 |
|
---|
| 575 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 576 | ASSERT(0 < workq->pending_op_cnt);
|
---|
| 577 | --workq->pending_op_cnt;
|
---|
| 578 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 579 | }
|
---|
| 580 |
|
---|
| 581 | /** Determines how to signal workers if at all.
|
---|
| 582 | *
|
---|
| 583 | * @param workq Work queue where a new work item was queued.
|
---|
| 584 | * @param can_block True if we may block while signaling a worker or creating
|
---|
| 585 | * a new worker.
|
---|
| 586 | *
|
---|
| 587 | * @return Function that will notify workers or NULL if no action is needed.
|
---|
| 588 | */
|
---|
| 589 | static signal_op_t signal_worker_logic(struct work_queue *workq, bool can_block)
|
---|
| 590 | {
|
---|
[0d56712] | 591 | ASSERT(!workq_corrupted(workq));
|
---|
[8a64e81e] | 592 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 593 |
|
---|
| 594 | /* Only signal workers if really necessary. */
|
---|
| 595 | signal_op_t signal_op = NULL;
|
---|
| 596 |
|
---|
| 597 | /*
|
---|
| 598 | * Workers actively running the work func() and neither blocked nor idle.
|
---|
| 599 | * Including ->activate_pending workers that will run their work func()
|
---|
| 600 | * once they get a time slice.
|
---|
| 601 | */
|
---|
| 602 | size_t active = active_workers(workq);
|
---|
| 603 | /* Max total allowed number of work items queued for active workers. */
|
---|
| 604 | size_t max_load = active * max_items_per_worker;
|
---|
| 605 |
|
---|
| 606 | /* Active workers are getting overwhelmed - activate another. */
|
---|
| 607 | if (max_load < workq->item_cnt) {
|
---|
| 608 |
|
---|
| 609 | size_t remaining_idle =
|
---|
| 610 | workq->idle_worker_cnt - workq->activate_pending;
|
---|
| 611 |
|
---|
| 612 | /* Idle workers still exist - activate one. */
|
---|
| 613 | if (remaining_idle > 0) {
|
---|
| 614 | /*
|
---|
| 615 | * Directly changing idle_worker_cnt here would not allow
|
---|
| 616 | * workers to recognize spurious wake-ups. Change
|
---|
| 617 | * activate_pending instead.
|
---|
| 618 | */
|
---|
| 619 | ++workq->activate_pending;
|
---|
| 620 | ++workq->pending_op_cnt;
|
---|
| 621 | signal_op = signal_worker_op;
|
---|
| 622 | } else {
|
---|
| 623 | /* No idle workers remain. Request that a new one be created. */
|
---|
| 624 | bool need_worker = (active < max_concurrent_workers)
|
---|
| 625 | && (workq->cur_worker_cnt < max_worker_cnt);
|
---|
| 626 |
|
---|
| 627 | if (need_worker && can_block) {
|
---|
| 628 | signal_op = add_worker_op;
|
---|
| 629 | /*
|
---|
| 630 | * It may take some time to actually create the worker.
|
---|
| 631 | * We don't want to swamp the thread pool with superfluous
|
---|
| 632 | * worker creation requests so pretend it was already
|
---|
| 633 | * created and proactively increase the worker count.
|
---|
| 634 | */
|
---|
| 635 | ++workq->cur_worker_cnt;
|
---|
| 636 | }
|
---|
| 637 |
|
---|
| 638 | /*
|
---|
| 639 | * We cannot create a new worker but we need one desperately
|
---|
| 640 | * because all workers are blocked in their work functions.
|
---|
| 641 | */
|
---|
| 642 | if (need_worker && !can_block && 0 == active) {
|
---|
| 643 | ASSERT(0 == workq->idle_worker_cnt);
|
---|
| 644 |
|
---|
| 645 | irq_spinlock_lock(&nonblock_adder.lock, true);
|
---|
| 646 |
|
---|
| 647 | if (nonblock_adder.thread && !link_used(&workq->nb_link)) {
|
---|
| 648 | signal_op = add_worker_noblock_op;
|
---|
| 649 | ++workq->cur_worker_cnt;
|
---|
| 650 | list_append(&workq->nb_link, &nonblock_adder.work_queues);
|
---|
| 651 | }
|
---|
| 652 |
|
---|
| 653 | irq_spinlock_unlock(&nonblock_adder.lock, true);
|
---|
| 654 | }
|
---|
| 655 | }
|
---|
| 656 | } else {
|
---|
| 657 | /*
|
---|
| 658 | * There are enough active/running workers to process the queue.
|
---|
| 659 | * No need to signal/activate any new workers.
|
---|
| 660 | */
|
---|
| 661 | signal_op = NULL;
|
---|
| 662 | }
|
---|
| 663 |
|
---|
| 664 | return signal_op;
|
---|
| 665 | }
|
---|
| 666 |
|
---|
| 667 | /** Executes queued work items. */
|
---|
| 668 | static void worker_thread(void *arg)
|
---|
| 669 | {
|
---|
| 670 | /*
|
---|
| 671 | * The thread has been created after the work queue was ordered to stop.
|
---|
| 672 | * Do not access the work queue and return immediately.
|
---|
| 673 | */
|
---|
| 674 | if (thread_interrupted(THREAD)) {
|
---|
| 675 | thread_detach(THREAD);
|
---|
| 676 | return;
|
---|
| 677 | }
|
---|
| 678 |
|
---|
| 679 | ASSERT(arg != NULL);
|
---|
| 680 |
|
---|
| 681 | struct work_queue *workq = arg;
|
---|
| 682 | work_t *work_item;
|
---|
| 683 |
|
---|
| 684 | while (dequeue_work(workq, &work_item)) {
|
---|
| 685 | /* Copy the func field so func() can safely free work_item. */
|
---|
| 686 | work_func_t func = work_item->func;
|
---|
[0d56712] | 687 |
|
---|
[8a64e81e] | 688 | func(work_item);
|
---|
| 689 | }
|
---|
| 690 | }
|
---|
| 691 |
|
---|
| 692 | /** Waits and retrieves a work item. Returns false if the worker should exit. */
|
---|
| 693 | static bool dequeue_work(struct work_queue *workq, work_t **pwork_item)
|
---|
| 694 | {
|
---|
[0d56712] | 695 | ASSERT(!workq_corrupted(workq));
|
---|
| 696 |
|
---|
[8a64e81e] | 697 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 698 |
|
---|
| 699 | /* Check if we should exit if load is low. */
|
---|
| 700 | if (!workq->stopping && worker_unnecessary(workq)) {
|
---|
| 701 | /* There are too many workers for this load. Exit. */
|
---|
| 702 | ASSERT(0 < workq->cur_worker_cnt);
|
---|
| 703 | --workq->cur_worker_cnt;
|
---|
| 704 | list_remove(&THREAD->workq_link);
|
---|
| 705 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 706 |
|
---|
| 707 | thread_detach(THREAD);
|
---|
| 708 | return false;
|
---|
| 709 | }
|
---|
| 710 |
|
---|
| 711 | bool stop = false;
|
---|
| 712 |
|
---|
| 713 | /* Wait for work to arrive. */
|
---|
| 714 | while (list_empty(&workq->queue) && !workq->stopping) {
|
---|
| 715 | cv_wait(workq);
|
---|
| 716 |
|
---|
| 717 | if (0 < workq->activate_pending)
|
---|
| 718 | --workq->activate_pending;
|
---|
| 719 | }
|
---|
| 720 |
|
---|
| 721 | /* Process remaining work even if requested to stop. */
|
---|
| 722 | if (!list_empty(&workq->queue)) {
|
---|
| 723 | link_t *work_link = list_first(&workq->queue);
|
---|
| 724 | *pwork_item = list_get_instance(work_link, work_t, queue_link);
|
---|
[0d56712] | 725 |
|
---|
| 726 | #ifdef CONFIG_DEBUG
|
---|
| 727 | ASSERT(!work_item_corrupted(*pwork_item));
|
---|
| 728 | (*pwork_item)->cookie = 0;
|
---|
| 729 | #endif
|
---|
[8a64e81e] | 730 | list_remove(work_link);
|
---|
| 731 | --workq->item_cnt;
|
---|
| 732 |
|
---|
| 733 | stop = false;
|
---|
| 734 | } else {
|
---|
| 735 | /* Requested to stop and no more work queued. */
|
---|
| 736 | ASSERT(workq->stopping);
|
---|
| 737 | --workq->cur_worker_cnt;
|
---|
| 738 | stop = true;
|
---|
| 739 | }
|
---|
| 740 |
|
---|
| 741 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 742 |
|
---|
| 743 | return !stop;
|
---|
| 744 | }
|
---|
| 745 |
|
---|
| 746 | /** Returns true if for the given load there are too many workers. */
|
---|
| 747 | static bool worker_unnecessary(struct work_queue *workq)
|
---|
| 748 | {
|
---|
| 749 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 750 |
|
---|
| 751 | /* No work is pending. We don't need too many idle threads. */
|
---|
| 752 | if (list_empty(&workq->queue)) {
|
---|
| 753 | /* There are too many idle workers. Exit. */
|
---|
| 754 | return (min_worker_cnt <= workq->idle_worker_cnt);
|
---|
| 755 | } else {
|
---|
| 756 | /*
|
---|
| 757 | * There is work but we are swamped with too many active workers
|
---|
| 758 | * that were woken up from sleep at around the same time. We
|
---|
| 759 | * don't need another worker fighting for cpu time.
|
---|
| 760 | */
|
---|
| 761 | size_t active = active_workers_now(workq);
|
---|
| 762 | return (max_concurrent_workers < active);
|
---|
| 763 | }
|
---|
| 764 | }
|
---|
| 765 |
|
---|
| 766 | /** Waits for a signal to activate_worker. Thread marked idle while waiting. */
|
---|
| 767 | static void cv_wait(struct work_queue *workq)
|
---|
| 768 | {
|
---|
| 769 | ++workq->idle_worker_cnt;
|
---|
| 770 | THREAD->workq_idling = true;
|
---|
| 771 |
|
---|
| 772 | /* Ignore lock ordering just here. */
|
---|
| 773 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 774 |
|
---|
| 775 | _condvar_wait_timeout_irq_spinlock(&workq->activate_worker,
|
---|
| 776 | &workq->lock, SYNCH_NO_TIMEOUT, SYNCH_FLAGS_NONE);
|
---|
| 777 |
|
---|
[0d56712] | 778 | ASSERT(!workq_corrupted(workq));
|
---|
[8a64e81e] | 779 | ASSERT(irq_spinlock_locked(&workq->lock));
|
---|
| 780 |
|
---|
| 781 | THREAD->workq_idling = false;
|
---|
| 782 | --workq->idle_worker_cnt;
|
---|
| 783 | }
|
---|
| 784 |
|
---|
| 785 |
|
---|
| 786 | /** Invoked from thread_ready() right before the thread is woken up. */
|
---|
| 787 | void workq_before_thread_is_ready(thread_t *thread)
|
---|
| 788 | {
|
---|
| 789 | ASSERT(thread);
|
---|
| 790 | ASSERT(irq_spinlock_locked(&thread->lock));
|
---|
| 791 |
|
---|
| 792 | /* Worker's work func() is about to wake up from sleeping. */
|
---|
| 793 | if (thread->workq && thread->workq_blocked) {
|
---|
| 794 | /* Must be blocked in user work func() and not be waiting for work. */
|
---|
| 795 | ASSERT(!thread->workq_idling);
|
---|
| 796 | ASSERT(thread->state == Sleeping);
|
---|
[0d56712] | 797 | ASSERT(THREAD != thread);
|
---|
| 798 | ASSERT(!workq_corrupted(thread->workq));
|
---|
[8a64e81e] | 799 |
|
---|
| 800 | /* Protected by thread->lock */
|
---|
| 801 | thread->workq_blocked = false;
|
---|
| 802 |
|
---|
| 803 | irq_spinlock_lock(&thread->workq->lock, true);
|
---|
| 804 | --thread->workq->blocked_worker_cnt;
|
---|
| 805 | irq_spinlock_unlock(&thread->workq->lock, true);
|
---|
| 806 | }
|
---|
| 807 | }
|
---|
| 808 |
|
---|
| 809 | /** Invoked from scheduler() before switching away from a thread. */
|
---|
| 810 | void workq_after_thread_ran(void)
|
---|
| 811 | {
|
---|
| 812 | ASSERT(THREAD);
|
---|
| 813 | ASSERT(irq_spinlock_locked(&THREAD->lock));
|
---|
| 814 |
|
---|
| 815 | /* Worker's work func() is about to sleep/block. */
|
---|
| 816 | if (THREAD->workq && THREAD->state == Sleeping && !THREAD->workq_idling) {
|
---|
| 817 | ASSERT(!THREAD->workq_blocked);
|
---|
[0d56712] | 818 | ASSERT(!workq_corrupted(THREAD->workq));
|
---|
| 819 |
|
---|
[8a64e81e] | 820 | THREAD->workq_blocked = true;
|
---|
| 821 |
|
---|
| 822 | irq_spinlock_lock(&THREAD->workq->lock, false);
|
---|
| 823 |
|
---|
| 824 | ++THREAD->workq->blocked_worker_cnt;
|
---|
| 825 |
|
---|
| 826 | bool can_block = false;
|
---|
| 827 | signal_op_t op = signal_worker_logic(THREAD->workq, can_block);
|
---|
| 828 |
|
---|
| 829 | irq_spinlock_unlock(&THREAD->workq->lock, false);
|
---|
| 830 |
|
---|
| 831 | if (op) {
|
---|
| 832 | ASSERT(add_worker_noblock_op == op || signal_worker_op == op);
|
---|
| 833 | op(THREAD->workq);
|
---|
| 834 | }
|
---|
| 835 | }
|
---|
| 836 | }
|
---|
| 837 |
|
---|
| 838 | /** Prints stats of the work queue to the kernel console. */
|
---|
| 839 | void workq_print_info(struct work_queue *workq)
|
---|
| 840 | {
|
---|
| 841 | irq_spinlock_lock(&workq->lock, true);
|
---|
| 842 |
|
---|
| 843 | size_t total = workq->cur_worker_cnt;
|
---|
| 844 | size_t blocked = workq->blocked_worker_cnt;
|
---|
| 845 | size_t idle = workq->idle_worker_cnt;
|
---|
| 846 | size_t active = active_workers(workq);
|
---|
| 847 | size_t items = workq->item_cnt;
|
---|
| 848 | bool stopping = workq->stopping;
|
---|
| 849 | bool worker_surplus = worker_unnecessary(workq);
|
---|
| 850 | const char *load_str = worker_surplus ? "decreasing" :
|
---|
| 851 | (0 < workq->activate_pending) ? "increasing" : "stable";
|
---|
| 852 |
|
---|
| 853 | irq_spinlock_unlock(&workq->lock, true);
|
---|
| 854 |
|
---|
| 855 | printf(
|
---|
[0d56712] | 856 | "Configuration: max_worker_cnt=%zu, min_worker_cnt=%zu,\n"
|
---|
| 857 | " max_concurrent_workers=%zu, max_items_per_worker=%zu\n"
|
---|
[8a64e81e] | 858 | "Workers: %zu\n"
|
---|
| 859 | "Active: %zu (workers currently processing work)\n"
|
---|
| 860 | "Blocked: %zu (work functions sleeping/blocked)\n"
|
---|
| 861 | "Idle: %zu (idle workers waiting for more work)\n"
|
---|
| 862 | "Items: %zu (queued not yet dispatched work)\n"
|
---|
| 863 | "Stopping: %d\n"
|
---|
| 864 | "Load: %s\n",
|
---|
| 865 | max_worker_cnt, min_worker_cnt,
|
---|
| 866 | max_concurrent_workers, max_items_per_worker,
|
---|
| 867 | total,
|
---|
| 868 | active,
|
---|
| 869 | blocked,
|
---|
| 870 | idle,
|
---|
| 871 | items,
|
---|
| 872 | stopping,
|
---|
| 873 | load_str
|
---|
| 874 | );
|
---|
| 875 | }
|
---|
| 876 |
|
---|
| 877 | /** Prints stats of the global work queue. */
|
---|
| 878 | void workq_global_print_info(void)
|
---|
| 879 | {
|
---|
| 880 | workq_print_info(&g_work_queue);
|
---|
| 881 | }
|
---|
| 882 |
|
---|
| 883 |
|
---|
| 884 | static bool dequeue_add_req(nonblock_adder_t *info, struct work_queue **pworkq)
|
---|
| 885 | {
|
---|
| 886 | bool stop = false;
|
---|
| 887 |
|
---|
| 888 | irq_spinlock_lock(&info->lock, true);
|
---|
| 889 |
|
---|
| 890 | while (list_empty(&info->work_queues) && !stop) {
|
---|
| 891 | int ret = _condvar_wait_timeout_irq_spinlock(&info->req_cv,
|
---|
| 892 | &info->lock, SYNCH_NO_TIMEOUT, SYNCH_FLAGS_INTERRUPTIBLE);
|
---|
| 893 |
|
---|
| 894 | stop = (ret == ESYNCH_INTERRUPTED);
|
---|
| 895 | }
|
---|
| 896 |
|
---|
| 897 | if (!stop) {
|
---|
| 898 | *pworkq = list_get_instance(list_first(&info->work_queues),
|
---|
| 899 | struct work_queue, nb_link);
|
---|
[0d56712] | 900 |
|
---|
| 901 | ASSERT(!workq_corrupted(*pworkq));
|
---|
[8a64e81e] | 902 |
|
---|
| 903 | list_remove(&(*pworkq)->nb_link);
|
---|
| 904 | }
|
---|
| 905 |
|
---|
| 906 | irq_spinlock_unlock(&info->lock, true);
|
---|
| 907 |
|
---|
| 908 | return !stop;
|
---|
| 909 | }
|
---|
| 910 |
|
---|
| 911 | static void thr_nonblock_add_worker(void *arg)
|
---|
| 912 | {
|
---|
| 913 | nonblock_adder_t *info = arg;
|
---|
| 914 | struct work_queue *workq;
|
---|
| 915 |
|
---|
| 916 | while (dequeue_add_req(info, &workq)) {
|
---|
| 917 | add_worker(workq);
|
---|
| 918 | }
|
---|
| 919 | }
|
---|
| 920 |
|
---|
| 921 |
|
---|
| 922 | static void nonblock_init(void)
|
---|
| 923 | {
|
---|
[0d56712] | 924 | irq_spinlock_initialize(&nonblock_adder.lock, "kworkq-nb.lock");
|
---|
[8a64e81e] | 925 | condvar_initialize(&nonblock_adder.req_cv);
|
---|
| 926 | list_initialize(&nonblock_adder.work_queues);
|
---|
| 927 |
|
---|
| 928 | nonblock_adder.thread = thread_create(thr_nonblock_add_worker,
|
---|
[0d56712] | 929 | &nonblock_adder, TASK, THREAD_FLAG_NONE, "kworkq-nb");
|
---|
[8a64e81e] | 930 |
|
---|
| 931 | if (nonblock_adder.thread) {
|
---|
| 932 | thread_ready(nonblock_adder.thread);
|
---|
| 933 | } else {
|
---|
| 934 | /*
|
---|
| 935 | * We won't be able to add workers without blocking if all workers
|
---|
| 936 | * sleep, but at least boot the system.
|
---|
| 937 | */
|
---|
[0d56712] | 938 | printf("Failed to create kworkq-nb. Sleeping work may stall the workq.\n");
|
---|
[8a64e81e] | 939 | }
|
---|
| 940 | }
|
---|
| 941 |
|
---|
[0d56712] | 942 | /** Returns true if the workq is definitely corrupted; false if not sure.
|
---|
| 943 | *
|
---|
| 944 | * Can be used outside of any locks.
|
---|
| 945 | */
|
---|
| 946 | static bool workq_corrupted(struct work_queue *workq)
|
---|
| 947 | {
|
---|
| 948 | #ifdef CONFIG_DEBUG
|
---|
| 949 | /*
|
---|
| 950 | * Needed to make the most current cookie value set by workq_preinit()
|
---|
| 951 | * visible even if we access the workq right after it is created but
|
---|
| 952 | * on a different cpu. Otherwise, workq_corrupted() would not work
|
---|
| 953 | * outside a lock.
|
---|
| 954 | */
|
---|
| 955 | memory_barrier();
|
---|
| 956 | return NULL == workq || workq->cookie != WORKQ_MAGIC;
|
---|
| 957 | #else
|
---|
| 958 | return false;
|
---|
| 959 | #endif
|
---|
| 960 | }
|
---|
[8a64e81e] | 961 |
|
---|
[0d56712] | 962 | /** Returns true if the work_item is definitely corrupted; false if not sure.
|
---|
| 963 | *
|
---|
| 964 | * Must be used with the work queue protecting spinlock locked.
|
---|
| 965 | */
|
---|
| 966 | static bool work_item_corrupted(work_t *work_item)
|
---|
| 967 | {
|
---|
| 968 | #ifdef CONFIG_DEBUG
|
---|
| 969 | return NULL == work_item || work_item->cookie != WORK_ITEM_MAGIC;
|
---|
| 970 | #else
|
---|
| 971 | return false;
|
---|
| 972 | #endif
|
---|
| 973 | }
|
---|
[9f8745c5] | 974 |
|
---|
| 975 | /** @}
|
---|
| 976 | */
|
---|