David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1 | /* Worker thread pool for slow items, such as filesystem lookups or mkdirs |
| 2 | * |
| 3 | * Copyright (C) 2008 Red Hat, Inc. All Rights Reserved. |
| 4 | * Written by David Howells (dhowells@redhat.com) |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU General Public Licence |
| 8 | * as published by the Free Software Foundation; either version |
| 9 | * 2 of the Licence, or (at your option) any later version. |
David Howells | 8f0aa2f | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 10 | * |
| 11 | * See Documentation/slow-work.txt |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 12 | */ |
| 13 | |
| 14 | #include <linux/module.h> |
| 15 | #include <linux/slow-work.h> |
| 16 | #include <linux/kthread.h> |
| 17 | #include <linux/freezer.h> |
| 18 | #include <linux/wait.h> |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 19 | #include <linux/proc_fs.h> |
| 20 | #include "slow-work.h" |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 21 | |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 22 | static void slow_work_cull_timeout(unsigned long); |
| 23 | static void slow_work_oom_timeout(unsigned long); |
| 24 | |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 25 | #ifdef CONFIG_SYSCTL |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 26 | static int slow_work_min_threads_sysctl(struct ctl_table *, int, |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 27 | void __user *, size_t *, loff_t *); |
| 28 | |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 29 | static int slow_work_max_threads_sysctl(struct ctl_table *, int , |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 30 | void __user *, size_t *, loff_t *); |
| 31 | #endif |
| 32 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 33 | /* |
| 34 | * The pool of threads has at least min threads in it as long as someone is |
| 35 | * using the facility, and may have as many as max. |
| 36 | * |
| 37 | * A portion of the pool may be processing very slow operations. |
| 38 | */ |
| 39 | static unsigned slow_work_min_threads = 2; |
| 40 | static unsigned slow_work_max_threads = 4; |
| 41 | static unsigned vslow_work_proportion = 50; /* % of threads that may process |
| 42 | * very slow work */ |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 43 | |
| 44 | #ifdef CONFIG_SYSCTL |
| 45 | static const int slow_work_min_min_threads = 2; |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 46 | static int slow_work_max_max_threads = SLOW_WORK_THREAD_LIMIT; |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 47 | static const int slow_work_min_vslow = 1; |
| 48 | static const int slow_work_max_vslow = 99; |
| 49 | |
| 50 | ctl_table slow_work_sysctls[] = { |
| 51 | { |
| 52 | .ctl_name = CTL_UNNUMBERED, |
| 53 | .procname = "min-threads", |
| 54 | .data = &slow_work_min_threads, |
| 55 | .maxlen = sizeof(unsigned), |
| 56 | .mode = 0644, |
| 57 | .proc_handler = slow_work_min_threads_sysctl, |
| 58 | .extra1 = (void *) &slow_work_min_min_threads, |
| 59 | .extra2 = &slow_work_max_threads, |
| 60 | }, |
| 61 | { |
| 62 | .ctl_name = CTL_UNNUMBERED, |
| 63 | .procname = "max-threads", |
| 64 | .data = &slow_work_max_threads, |
| 65 | .maxlen = sizeof(unsigned), |
| 66 | .mode = 0644, |
| 67 | .proc_handler = slow_work_max_threads_sysctl, |
| 68 | .extra1 = &slow_work_min_threads, |
| 69 | .extra2 = (void *) &slow_work_max_max_threads, |
| 70 | }, |
| 71 | { |
| 72 | .ctl_name = CTL_UNNUMBERED, |
| 73 | .procname = "vslow-percentage", |
| 74 | .data = &vslow_work_proportion, |
| 75 | .maxlen = sizeof(unsigned), |
| 76 | .mode = 0644, |
| 77 | .proc_handler = &proc_dointvec_minmax, |
| 78 | .extra1 = (void *) &slow_work_min_vslow, |
| 79 | .extra2 = (void *) &slow_work_max_vslow, |
| 80 | }, |
| 81 | { .ctl_name = 0 } |
| 82 | }; |
| 83 | #endif |
| 84 | |
| 85 | /* |
| 86 | * The active state of the thread pool |
| 87 | */ |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 88 | static atomic_t slow_work_thread_count; |
| 89 | static atomic_t vslow_work_executing_count; |
| 90 | |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 91 | static bool slow_work_may_not_start_new_thread; |
| 92 | static bool slow_work_cull; /* cull a thread due to lack of activity */ |
| 93 | static DEFINE_TIMER(slow_work_cull_timer, slow_work_cull_timeout, 0, 0); |
| 94 | static DEFINE_TIMER(slow_work_oom_timer, slow_work_oom_timeout, 0, 0); |
| 95 | static struct slow_work slow_work_new_thread; /* new thread starter */ |
| 96 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 97 | /* |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 98 | * slow work ID allocation (use slow_work_queue_lock) |
| 99 | */ |
| 100 | static DECLARE_BITMAP(slow_work_ids, SLOW_WORK_THREAD_LIMIT); |
| 101 | |
| 102 | /* |
| 103 | * Unregistration tracking to prevent put_ref() from disappearing during module |
| 104 | * unload |
| 105 | */ |
| 106 | #ifdef CONFIG_MODULES |
| 107 | static struct module *slow_work_thread_processing[SLOW_WORK_THREAD_LIMIT]; |
| 108 | static struct module *slow_work_unreg_module; |
| 109 | static struct slow_work *slow_work_unreg_work_item; |
| 110 | static DECLARE_WAIT_QUEUE_HEAD(slow_work_unreg_wq); |
| 111 | static DEFINE_MUTEX(slow_work_unreg_sync_lock); |
| 112 | #endif |
| 113 | |
| 114 | /* |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 115 | * Data for tracking currently executing items for indication through /proc |
| 116 | */ |
| 117 | #ifdef CONFIG_SLOW_WORK_PROC |
| 118 | struct slow_work *slow_work_execs[SLOW_WORK_THREAD_LIMIT]; |
| 119 | pid_t slow_work_pids[SLOW_WORK_THREAD_LIMIT]; |
| 120 | DEFINE_RWLOCK(slow_work_execs_lock); |
| 121 | #endif |
| 122 | |
| 123 | /* |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 124 | * The queues of work items and the lock governing access to them. These are |
| 125 | * shared between all the CPUs. It doesn't make sense to have per-CPU queues |
| 126 | * as the number of threads bears no relation to the number of CPUs. |
| 127 | * |
| 128 | * There are two queues of work items: one for slow work items, and one for |
| 129 | * very slow work items. |
| 130 | */ |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 131 | LIST_HEAD(slow_work_queue); |
| 132 | LIST_HEAD(vslow_work_queue); |
| 133 | DEFINE_SPINLOCK(slow_work_queue_lock); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 134 | |
| 135 | /* |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 136 | * The following are two wait queues that get pinged when a work item is placed |
| 137 | * on an empty queue. These allow work items that are hogging a thread by |
| 138 | * sleeping in a way that could be deferred to yield their thread and enqueue |
| 139 | * themselves. |
| 140 | */ |
| 141 | static DECLARE_WAIT_QUEUE_HEAD(slow_work_queue_waits_for_occupation); |
| 142 | static DECLARE_WAIT_QUEUE_HEAD(vslow_work_queue_waits_for_occupation); |
| 143 | |
| 144 | /* |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 145 | * The thread controls. A variable used to signal to the threads that they |
| 146 | * should exit when the queue is empty, a waitqueue used by the threads to wait |
| 147 | * for signals, and a completion set by the last thread to exit. |
| 148 | */ |
| 149 | static bool slow_work_threads_should_exit; |
| 150 | static DECLARE_WAIT_QUEUE_HEAD(slow_work_thread_wq); |
| 151 | static DECLARE_COMPLETION(slow_work_last_thread_exited); |
| 152 | |
| 153 | /* |
| 154 | * The number of users of the thread pool and its lock. Whilst this is zero we |
| 155 | * have no threads hanging around, and when this reaches zero, we wait for all |
| 156 | * active or queued work items to complete and kill all the threads we do have. |
| 157 | */ |
| 158 | static int slow_work_user_count; |
| 159 | static DEFINE_MUTEX(slow_work_user_lock); |
| 160 | |
Jens Axboe | 4d8bb2c | 2009-11-19 18:10:39 +0000 | [diff] [blame] | 161 | static inline int slow_work_get_ref(struct slow_work *work) |
| 162 | { |
| 163 | if (work->ops->get_ref) |
| 164 | return work->ops->get_ref(work); |
| 165 | |
| 166 | return 0; |
| 167 | } |
| 168 | |
| 169 | static inline void slow_work_put_ref(struct slow_work *work) |
| 170 | { |
| 171 | if (work->ops->put_ref) |
| 172 | work->ops->put_ref(work); |
| 173 | } |
| 174 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 175 | /* |
| 176 | * Calculate the maximum number of active threads in the pool that are |
| 177 | * permitted to process very slow work items. |
| 178 | * |
| 179 | * The answer is rounded up to at least 1, but may not equal or exceed the |
| 180 | * maximum number of the threads in the pool. This means we always have at |
| 181 | * least one thread that can process slow work items, and we always have at |
| 182 | * least one thread that won't get tied up doing so. |
| 183 | */ |
| 184 | static unsigned slow_work_calc_vsmax(void) |
| 185 | { |
| 186 | unsigned vsmax; |
| 187 | |
| 188 | vsmax = atomic_read(&slow_work_thread_count) * vslow_work_proportion; |
| 189 | vsmax /= 100; |
| 190 | vsmax = max(vsmax, 1U); |
| 191 | return min(vsmax, slow_work_max_threads - 1); |
| 192 | } |
| 193 | |
| 194 | /* |
| 195 | * Attempt to execute stuff queued on a slow thread. Return true if we managed |
| 196 | * it, false if there was nothing to do. |
| 197 | */ |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 198 | static noinline bool slow_work_execute(int id) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 199 | { |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 200 | #ifdef CONFIG_MODULES |
| 201 | struct module *module; |
| 202 | #endif |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 203 | struct slow_work *work = NULL; |
| 204 | unsigned vsmax; |
| 205 | bool very_slow; |
| 206 | |
| 207 | vsmax = slow_work_calc_vsmax(); |
| 208 | |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 209 | /* see if we can schedule a new thread to be started if we're not |
| 210 | * keeping up with the work */ |
| 211 | if (!waitqueue_active(&slow_work_thread_wq) && |
| 212 | (!list_empty(&slow_work_queue) || !list_empty(&vslow_work_queue)) && |
| 213 | atomic_read(&slow_work_thread_count) < slow_work_max_threads && |
| 214 | !slow_work_may_not_start_new_thread) |
| 215 | slow_work_enqueue(&slow_work_new_thread); |
| 216 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 217 | /* find something to execute */ |
| 218 | spin_lock_irq(&slow_work_queue_lock); |
| 219 | if (!list_empty(&vslow_work_queue) && |
| 220 | atomic_read(&vslow_work_executing_count) < vsmax) { |
| 221 | work = list_entry(vslow_work_queue.next, |
| 222 | struct slow_work, link); |
| 223 | if (test_and_set_bit_lock(SLOW_WORK_EXECUTING, &work->flags)) |
| 224 | BUG(); |
| 225 | list_del_init(&work->link); |
| 226 | atomic_inc(&vslow_work_executing_count); |
| 227 | very_slow = true; |
| 228 | } else if (!list_empty(&slow_work_queue)) { |
| 229 | work = list_entry(slow_work_queue.next, |
| 230 | struct slow_work, link); |
| 231 | if (test_and_set_bit_lock(SLOW_WORK_EXECUTING, &work->flags)) |
| 232 | BUG(); |
| 233 | list_del_init(&work->link); |
| 234 | very_slow = false; |
| 235 | } else { |
| 236 | very_slow = false; /* avoid the compiler warning */ |
| 237 | } |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 238 | |
| 239 | #ifdef CONFIG_MODULES |
| 240 | if (work) |
| 241 | slow_work_thread_processing[id] = work->owner; |
| 242 | #endif |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 243 | if (work) { |
| 244 | slow_work_mark_time(work); |
| 245 | slow_work_begin_exec(id, work); |
| 246 | } |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 247 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 248 | spin_unlock_irq(&slow_work_queue_lock); |
| 249 | |
| 250 | if (!work) |
| 251 | return false; |
| 252 | |
| 253 | if (!test_and_clear_bit(SLOW_WORK_PENDING, &work->flags)) |
| 254 | BUG(); |
| 255 | |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 256 | /* don't execute if the work is in the process of being cancelled */ |
| 257 | if (!test_bit(SLOW_WORK_CANCELLING, &work->flags)) |
| 258 | work->ops->execute(work); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 259 | |
| 260 | if (very_slow) |
| 261 | atomic_dec(&vslow_work_executing_count); |
| 262 | clear_bit_unlock(SLOW_WORK_EXECUTING, &work->flags); |
| 263 | |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 264 | /* wake up anyone waiting for this work to be complete */ |
| 265 | wake_up_bit(&work->flags, SLOW_WORK_EXECUTING); |
| 266 | |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 267 | slow_work_end_exec(id, work); |
| 268 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 269 | /* if someone tried to enqueue the item whilst we were executing it, |
| 270 | * then it'll be left unenqueued to avoid multiple threads trying to |
| 271 | * execute it simultaneously |
| 272 | * |
| 273 | * there is, however, a race between us testing the pending flag and |
| 274 | * getting the spinlock, and between the enqueuer setting the pending |
| 275 | * flag and getting the spinlock, so we use a deferral bit to tell us |
| 276 | * if the enqueuer got there first |
| 277 | */ |
| 278 | if (test_bit(SLOW_WORK_PENDING, &work->flags)) { |
| 279 | spin_lock_irq(&slow_work_queue_lock); |
| 280 | |
| 281 | if (!test_bit(SLOW_WORK_EXECUTING, &work->flags) && |
| 282 | test_and_clear_bit(SLOW_WORK_ENQ_DEFERRED, &work->flags)) |
| 283 | goto auto_requeue; |
| 284 | |
| 285 | spin_unlock_irq(&slow_work_queue_lock); |
| 286 | } |
| 287 | |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 288 | /* sort out the race between module unloading and put_ref() */ |
Jens Axboe | 4d8bb2c | 2009-11-19 18:10:39 +0000 | [diff] [blame] | 289 | slow_work_put_ref(work); |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 290 | |
| 291 | #ifdef CONFIG_MODULES |
| 292 | module = slow_work_thread_processing[id]; |
| 293 | slow_work_thread_processing[id] = NULL; |
| 294 | smp_mb(); |
| 295 | if (slow_work_unreg_work_item == work || |
| 296 | slow_work_unreg_module == module) |
| 297 | wake_up_all(&slow_work_unreg_wq); |
| 298 | #endif |
| 299 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 300 | return true; |
| 301 | |
| 302 | auto_requeue: |
| 303 | /* we must complete the enqueue operation |
| 304 | * - we transfer our ref on the item back to the appropriate queue |
| 305 | * - don't wake another thread up as we're awake already |
| 306 | */ |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 307 | slow_work_mark_time(work); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 308 | if (test_bit(SLOW_WORK_VERY_SLOW, &work->flags)) |
| 309 | list_add_tail(&work->link, &vslow_work_queue); |
| 310 | else |
| 311 | list_add_tail(&work->link, &slow_work_queue); |
| 312 | spin_unlock_irq(&slow_work_queue_lock); |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 313 | slow_work_thread_processing[id] = NULL; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 314 | return true; |
| 315 | } |
| 316 | |
| 317 | /** |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 318 | * slow_work_sleep_till_thread_needed - Sleep till thread needed by other work |
| 319 | * work: The work item under execution that wants to sleep |
| 320 | * _timeout: Scheduler sleep timeout |
| 321 | * |
| 322 | * Allow a requeueable work item to sleep on a slow-work processor thread until |
| 323 | * that thread is needed to do some other work or the sleep is interrupted by |
| 324 | * some other event. |
| 325 | * |
| 326 | * The caller must set up a wake up event before calling this and must have set |
| 327 | * the appropriate sleep mode (such as TASK_UNINTERRUPTIBLE) and tested its own |
| 328 | * condition before calling this function as no test is made here. |
| 329 | * |
| 330 | * False is returned if there is nothing on the queue; true is returned if the |
| 331 | * work item should be requeued |
| 332 | */ |
| 333 | bool slow_work_sleep_till_thread_needed(struct slow_work *work, |
| 334 | signed long *_timeout) |
| 335 | { |
| 336 | wait_queue_head_t *wfo_wq; |
| 337 | struct list_head *queue; |
| 338 | |
| 339 | DEFINE_WAIT(wait); |
| 340 | |
| 341 | if (test_bit(SLOW_WORK_VERY_SLOW, &work->flags)) { |
| 342 | wfo_wq = &vslow_work_queue_waits_for_occupation; |
| 343 | queue = &vslow_work_queue; |
| 344 | } else { |
| 345 | wfo_wq = &slow_work_queue_waits_for_occupation; |
| 346 | queue = &slow_work_queue; |
| 347 | } |
| 348 | |
| 349 | if (!list_empty(queue)) |
| 350 | return true; |
| 351 | |
| 352 | add_wait_queue_exclusive(wfo_wq, &wait); |
| 353 | if (list_empty(queue)) |
| 354 | *_timeout = schedule_timeout(*_timeout); |
| 355 | finish_wait(wfo_wq, &wait); |
| 356 | |
| 357 | return !list_empty(queue); |
| 358 | } |
| 359 | EXPORT_SYMBOL(slow_work_sleep_till_thread_needed); |
| 360 | |
| 361 | /** |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 362 | * slow_work_enqueue - Schedule a slow work item for processing |
| 363 | * @work: The work item to queue |
| 364 | * |
| 365 | * Schedule a slow work item for processing. If the item is already undergoing |
| 366 | * execution, this guarantees not to re-enter the execution routine until the |
| 367 | * first execution finishes. |
| 368 | * |
| 369 | * The item is pinned by this function as it retains a reference to it, managed |
| 370 | * through the item operations. The item is unpinned once it has been |
| 371 | * executed. |
| 372 | * |
| 373 | * An item may hog the thread that is running it for a relatively large amount |
| 374 | * of time, sufficient, for example, to perform several lookup, mkdir, create |
| 375 | * and setxattr operations. It may sleep on I/O and may sleep to obtain locks. |
| 376 | * |
| 377 | * Conversely, if a number of items are awaiting processing, it may take some |
| 378 | * time before any given item is given attention. The number of threads in the |
| 379 | * pool may be increased to deal with demand, but only up to a limit. |
| 380 | * |
| 381 | * If SLOW_WORK_VERY_SLOW is set on the work item, then it will be placed in |
| 382 | * the very slow queue, from which only a portion of the threads will be |
| 383 | * allowed to pick items to execute. This ensures that very slow items won't |
| 384 | * overly block ones that are just ordinarily slow. |
| 385 | * |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 386 | * Returns 0 if successful, -EAGAIN if not (or -ECANCELED if cancelled work is |
| 387 | * attempted queued) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 388 | */ |
| 389 | int slow_work_enqueue(struct slow_work *work) |
| 390 | { |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 391 | wait_queue_head_t *wfo_wq; |
| 392 | struct list_head *queue; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 393 | unsigned long flags; |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 394 | int ret; |
| 395 | |
| 396 | if (test_bit(SLOW_WORK_CANCELLING, &work->flags)) |
| 397 | return -ECANCELED; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 398 | |
| 399 | BUG_ON(slow_work_user_count <= 0); |
| 400 | BUG_ON(!work); |
| 401 | BUG_ON(!work->ops); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 402 | |
| 403 | /* when honouring an enqueue request, we only promise that we will run |
| 404 | * the work function in the future; we do not promise to run it once |
| 405 | * per enqueue request |
| 406 | * |
| 407 | * we use the PENDING bit to merge together repeat requests without |
| 408 | * having to disable IRQs and take the spinlock, whilst still |
| 409 | * maintaining our promise |
| 410 | */ |
| 411 | if (!test_and_set_bit_lock(SLOW_WORK_PENDING, &work->flags)) { |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 412 | if (test_bit(SLOW_WORK_VERY_SLOW, &work->flags)) { |
| 413 | wfo_wq = &vslow_work_queue_waits_for_occupation; |
| 414 | queue = &vslow_work_queue; |
| 415 | } else { |
| 416 | wfo_wq = &slow_work_queue_waits_for_occupation; |
| 417 | queue = &slow_work_queue; |
| 418 | } |
| 419 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 420 | spin_lock_irqsave(&slow_work_queue_lock, flags); |
| 421 | |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 422 | if (unlikely(test_bit(SLOW_WORK_CANCELLING, &work->flags))) |
| 423 | goto cancelled; |
| 424 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 425 | /* we promise that we will not attempt to execute the work |
| 426 | * function in more than one thread simultaneously |
| 427 | * |
| 428 | * this, however, leaves us with a problem if we're asked to |
| 429 | * enqueue the work whilst someone is executing the work |
| 430 | * function as simply queueing the work immediately means that |
| 431 | * another thread may try executing it whilst it is already |
| 432 | * under execution |
| 433 | * |
| 434 | * to deal with this, we set the ENQ_DEFERRED bit instead of |
| 435 | * enqueueing, and the thread currently executing the work |
| 436 | * function will enqueue the work item when the work function |
| 437 | * returns and it has cleared the EXECUTING bit |
| 438 | */ |
| 439 | if (test_bit(SLOW_WORK_EXECUTING, &work->flags)) { |
| 440 | set_bit(SLOW_WORK_ENQ_DEFERRED, &work->flags); |
| 441 | } else { |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 442 | ret = slow_work_get_ref(work); |
| 443 | if (ret < 0) |
| 444 | goto failed; |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 445 | slow_work_mark_time(work); |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 446 | list_add_tail(&work->link, queue); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 447 | wake_up(&slow_work_thread_wq); |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 448 | |
| 449 | /* if someone who could be requeued is sleeping on a |
| 450 | * thread, then ask them to yield their thread */ |
| 451 | if (work->link.prev == queue) |
| 452 | wake_up(wfo_wq); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 453 | } |
| 454 | |
| 455 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
| 456 | } |
| 457 | return 0; |
| 458 | |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 459 | cancelled: |
| 460 | ret = -ECANCELED; |
| 461 | failed: |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 462 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 463 | return ret; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 464 | } |
| 465 | EXPORT_SYMBOL(slow_work_enqueue); |
| 466 | |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 467 | static int slow_work_wait(void *word) |
| 468 | { |
| 469 | schedule(); |
| 470 | return 0; |
| 471 | } |
| 472 | |
| 473 | /** |
| 474 | * slow_work_cancel - Cancel a slow work item |
| 475 | * @work: The work item to cancel |
| 476 | * |
| 477 | * This function will cancel a previously enqueued work item. If we cannot |
| 478 | * cancel the work item, it is guarenteed to have run when this function |
| 479 | * returns. |
| 480 | */ |
| 481 | void slow_work_cancel(struct slow_work *work) |
| 482 | { |
| 483 | bool wait = true, put = false; |
| 484 | |
| 485 | set_bit(SLOW_WORK_CANCELLING, &work->flags); |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 486 | smp_mb(); |
| 487 | |
| 488 | /* if the work item is a delayed work item with an active timer, we |
| 489 | * need to wait for the timer to finish _before_ getting the spinlock, |
| 490 | * lest we deadlock against the timer routine |
| 491 | * |
| 492 | * the timer routine will leave DELAYED set if it notices the |
| 493 | * CANCELLING flag in time |
| 494 | */ |
| 495 | if (test_bit(SLOW_WORK_DELAYED, &work->flags)) { |
| 496 | struct delayed_slow_work *dwork = |
| 497 | container_of(work, struct delayed_slow_work, work); |
| 498 | del_timer_sync(&dwork->timer); |
| 499 | } |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 500 | |
| 501 | spin_lock_irq(&slow_work_queue_lock); |
| 502 | |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 503 | if (test_bit(SLOW_WORK_DELAYED, &work->flags)) { |
| 504 | /* the timer routine aborted or never happened, so we are left |
| 505 | * holding the timer's reference on the item and should just |
| 506 | * drop the pending flag and wait for any ongoing execution to |
| 507 | * finish */ |
| 508 | struct delayed_slow_work *dwork = |
| 509 | container_of(work, struct delayed_slow_work, work); |
| 510 | |
| 511 | BUG_ON(timer_pending(&dwork->timer)); |
| 512 | BUG_ON(!list_empty(&work->link)); |
| 513 | |
| 514 | clear_bit(SLOW_WORK_DELAYED, &work->flags); |
| 515 | put = true; |
| 516 | clear_bit(SLOW_WORK_PENDING, &work->flags); |
| 517 | |
| 518 | } else if (test_bit(SLOW_WORK_PENDING, &work->flags) && |
| 519 | !list_empty(&work->link)) { |
Jens Axboe | 0160950 | 2009-11-19 18:10:43 +0000 | [diff] [blame] | 520 | /* the link in the pending queue holds a reference on the item |
| 521 | * that we will need to release */ |
| 522 | list_del_init(&work->link); |
| 523 | wait = false; |
| 524 | put = true; |
| 525 | clear_bit(SLOW_WORK_PENDING, &work->flags); |
| 526 | |
| 527 | } else if (test_and_clear_bit(SLOW_WORK_ENQ_DEFERRED, &work->flags)) { |
| 528 | /* the executor is holding our only reference on the item, so |
| 529 | * we merely need to wait for it to finish executing */ |
| 530 | clear_bit(SLOW_WORK_PENDING, &work->flags); |
| 531 | } |
| 532 | |
| 533 | spin_unlock_irq(&slow_work_queue_lock); |
| 534 | |
| 535 | /* the EXECUTING flag is set by the executor whilst the spinlock is set |
| 536 | * and before the item is dequeued - so assuming the above doesn't |
| 537 | * actually dequeue it, simply waiting for the EXECUTING flag to be |
| 538 | * released here should be sufficient */ |
| 539 | if (wait) |
| 540 | wait_on_bit(&work->flags, SLOW_WORK_EXECUTING, slow_work_wait, |
| 541 | TASK_UNINTERRUPTIBLE); |
| 542 | |
| 543 | clear_bit(SLOW_WORK_CANCELLING, &work->flags); |
| 544 | if (put) |
| 545 | slow_work_put_ref(work); |
| 546 | } |
| 547 | EXPORT_SYMBOL(slow_work_cancel); |
| 548 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 549 | /* |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 550 | * Handle expiry of the delay timer, indicating that a delayed slow work item |
| 551 | * should now be queued if not cancelled |
| 552 | */ |
| 553 | static void delayed_slow_work_timer(unsigned long data) |
| 554 | { |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 555 | wait_queue_head_t *wfo_wq; |
| 556 | struct list_head *queue; |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 557 | struct slow_work *work = (struct slow_work *) data; |
| 558 | unsigned long flags; |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 559 | bool queued = false, put = false, first = false; |
| 560 | |
| 561 | if (test_bit(SLOW_WORK_VERY_SLOW, &work->flags)) { |
| 562 | wfo_wq = &vslow_work_queue_waits_for_occupation; |
| 563 | queue = &vslow_work_queue; |
| 564 | } else { |
| 565 | wfo_wq = &slow_work_queue_waits_for_occupation; |
| 566 | queue = &slow_work_queue; |
| 567 | } |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 568 | |
| 569 | spin_lock_irqsave(&slow_work_queue_lock, flags); |
| 570 | if (likely(!test_bit(SLOW_WORK_CANCELLING, &work->flags))) { |
| 571 | clear_bit(SLOW_WORK_DELAYED, &work->flags); |
| 572 | |
| 573 | if (test_bit(SLOW_WORK_EXECUTING, &work->flags)) { |
| 574 | /* we discard the reference the timer was holding in |
| 575 | * favour of the one the executor holds */ |
| 576 | set_bit(SLOW_WORK_ENQ_DEFERRED, &work->flags); |
| 577 | put = true; |
| 578 | } else { |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 579 | slow_work_mark_time(work); |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 580 | list_add_tail(&work->link, queue); |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 581 | queued = true; |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 582 | if (work->link.prev == queue) |
| 583 | first = true; |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 584 | } |
| 585 | } |
| 586 | |
| 587 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
| 588 | if (put) |
| 589 | slow_work_put_ref(work); |
David Howells | 3bde31a | 2009-11-19 18:10:57 +0000 | [diff] [blame^] | 590 | if (first) |
| 591 | wake_up(wfo_wq); |
Jens Axboe | 6b8268b | 2009-11-19 18:10:47 +0000 | [diff] [blame] | 592 | if (queued) |
| 593 | wake_up(&slow_work_thread_wq); |
| 594 | } |
| 595 | |
| 596 | /** |
| 597 | * delayed_slow_work_enqueue - Schedule a delayed slow work item for processing |
| 598 | * @dwork: The delayed work item to queue |
| 599 | * @delay: When to start executing the work, in jiffies from now |
| 600 | * |
| 601 | * This is similar to slow_work_enqueue(), but it adds a delay before the work |
| 602 | * is actually queued for processing. |
| 603 | * |
| 604 | * The item can have delayed processing requested on it whilst it is being |
| 605 | * executed. The delay will begin immediately, and if it expires before the |
| 606 | * item finishes executing, the item will be placed back on the queue when it |
| 607 | * has done executing. |
| 608 | */ |
| 609 | int delayed_slow_work_enqueue(struct delayed_slow_work *dwork, |
| 610 | unsigned long delay) |
| 611 | { |
| 612 | struct slow_work *work = &dwork->work; |
| 613 | unsigned long flags; |
| 614 | int ret; |
| 615 | |
| 616 | if (delay == 0) |
| 617 | return slow_work_enqueue(&dwork->work); |
| 618 | |
| 619 | BUG_ON(slow_work_user_count <= 0); |
| 620 | BUG_ON(!work); |
| 621 | BUG_ON(!work->ops); |
| 622 | |
| 623 | if (test_bit(SLOW_WORK_CANCELLING, &work->flags)) |
| 624 | return -ECANCELED; |
| 625 | |
| 626 | if (!test_and_set_bit_lock(SLOW_WORK_PENDING, &work->flags)) { |
| 627 | spin_lock_irqsave(&slow_work_queue_lock, flags); |
| 628 | |
| 629 | if (test_bit(SLOW_WORK_CANCELLING, &work->flags)) |
| 630 | goto cancelled; |
| 631 | |
| 632 | /* the timer holds a reference whilst it is pending */ |
| 633 | ret = work->ops->get_ref(work); |
| 634 | if (ret < 0) |
| 635 | goto cant_get_ref; |
| 636 | |
| 637 | if (test_and_set_bit(SLOW_WORK_DELAYED, &work->flags)) |
| 638 | BUG(); |
| 639 | dwork->timer.expires = jiffies + delay; |
| 640 | dwork->timer.data = (unsigned long) work; |
| 641 | dwork->timer.function = delayed_slow_work_timer; |
| 642 | add_timer(&dwork->timer); |
| 643 | |
| 644 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
| 645 | } |
| 646 | |
| 647 | return 0; |
| 648 | |
| 649 | cancelled: |
| 650 | ret = -ECANCELED; |
| 651 | cant_get_ref: |
| 652 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
| 653 | return ret; |
| 654 | } |
| 655 | EXPORT_SYMBOL(delayed_slow_work_enqueue); |
| 656 | |
| 657 | /* |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 658 | * Schedule a cull of the thread pool at some time in the near future |
| 659 | */ |
| 660 | static void slow_work_schedule_cull(void) |
| 661 | { |
| 662 | mod_timer(&slow_work_cull_timer, |
| 663 | round_jiffies(jiffies + SLOW_WORK_CULL_TIMEOUT)); |
| 664 | } |
| 665 | |
| 666 | /* |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 667 | * Worker thread culling algorithm |
| 668 | */ |
| 669 | static bool slow_work_cull_thread(void) |
| 670 | { |
| 671 | unsigned long flags; |
| 672 | bool do_cull = false; |
| 673 | |
| 674 | spin_lock_irqsave(&slow_work_queue_lock, flags); |
| 675 | |
| 676 | if (slow_work_cull) { |
| 677 | slow_work_cull = false; |
| 678 | |
| 679 | if (list_empty(&slow_work_queue) && |
| 680 | list_empty(&vslow_work_queue) && |
| 681 | atomic_read(&slow_work_thread_count) > |
| 682 | slow_work_min_threads) { |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 683 | slow_work_schedule_cull(); |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 684 | do_cull = true; |
| 685 | } |
| 686 | } |
| 687 | |
| 688 | spin_unlock_irqrestore(&slow_work_queue_lock, flags); |
| 689 | return do_cull; |
| 690 | } |
| 691 | |
| 692 | /* |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 693 | * Determine if there is slow work available for dispatch |
| 694 | */ |
| 695 | static inline bool slow_work_available(int vsmax) |
| 696 | { |
| 697 | return !list_empty(&slow_work_queue) || |
| 698 | (!list_empty(&vslow_work_queue) && |
| 699 | atomic_read(&vslow_work_executing_count) < vsmax); |
| 700 | } |
| 701 | |
| 702 | /* |
| 703 | * Worker thread dispatcher |
| 704 | */ |
| 705 | static int slow_work_thread(void *_data) |
| 706 | { |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 707 | int vsmax, id; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 708 | |
| 709 | DEFINE_WAIT(wait); |
| 710 | |
| 711 | set_freezable(); |
| 712 | set_user_nice(current, -5); |
| 713 | |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 714 | /* allocate ourselves an ID */ |
| 715 | spin_lock_irq(&slow_work_queue_lock); |
| 716 | id = find_first_zero_bit(slow_work_ids, SLOW_WORK_THREAD_LIMIT); |
| 717 | BUG_ON(id < 0 || id >= SLOW_WORK_THREAD_LIMIT); |
| 718 | __set_bit(id, slow_work_ids); |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 719 | slow_work_set_thread_pid(id, current->pid); |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 720 | spin_unlock_irq(&slow_work_queue_lock); |
| 721 | |
| 722 | sprintf(current->comm, "kslowd%03u", id); |
| 723 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 724 | for (;;) { |
| 725 | vsmax = vslow_work_proportion; |
| 726 | vsmax *= atomic_read(&slow_work_thread_count); |
| 727 | vsmax /= 100; |
| 728 | |
Oleg Nesterov | b415c49 | 2009-06-11 13:12:55 +0100 | [diff] [blame] | 729 | prepare_to_wait_exclusive(&slow_work_thread_wq, &wait, |
| 730 | TASK_INTERRUPTIBLE); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 731 | if (!freezing(current) && |
| 732 | !slow_work_threads_should_exit && |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 733 | !slow_work_available(vsmax) && |
| 734 | !slow_work_cull) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 735 | schedule(); |
| 736 | finish_wait(&slow_work_thread_wq, &wait); |
| 737 | |
| 738 | try_to_freeze(); |
| 739 | |
| 740 | vsmax = vslow_work_proportion; |
| 741 | vsmax *= atomic_read(&slow_work_thread_count); |
| 742 | vsmax /= 100; |
| 743 | |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 744 | if (slow_work_available(vsmax) && slow_work_execute(id)) { |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 745 | cond_resched(); |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 746 | if (list_empty(&slow_work_queue) && |
| 747 | list_empty(&vslow_work_queue) && |
| 748 | atomic_read(&slow_work_thread_count) > |
| 749 | slow_work_min_threads) |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 750 | slow_work_schedule_cull(); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 751 | continue; |
| 752 | } |
| 753 | |
| 754 | if (slow_work_threads_should_exit) |
| 755 | break; |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 756 | |
| 757 | if (slow_work_cull && slow_work_cull_thread()) |
| 758 | break; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 759 | } |
| 760 | |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 761 | spin_lock_irq(&slow_work_queue_lock); |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 762 | slow_work_set_thread_pid(id, 0); |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 763 | __clear_bit(id, slow_work_ids); |
| 764 | spin_unlock_irq(&slow_work_queue_lock); |
| 765 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 766 | if (atomic_dec_and_test(&slow_work_thread_count)) |
| 767 | complete_and_exit(&slow_work_last_thread_exited, 0); |
| 768 | return 0; |
| 769 | } |
| 770 | |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 771 | /* |
| 772 | * Handle thread cull timer expiration |
| 773 | */ |
| 774 | static void slow_work_cull_timeout(unsigned long data) |
| 775 | { |
| 776 | slow_work_cull = true; |
| 777 | wake_up(&slow_work_thread_wq); |
| 778 | } |
| 779 | |
| 780 | /* |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 781 | * Start a new slow work thread |
| 782 | */ |
| 783 | static void slow_work_new_thread_execute(struct slow_work *work) |
| 784 | { |
| 785 | struct task_struct *p; |
| 786 | |
| 787 | if (slow_work_threads_should_exit) |
| 788 | return; |
| 789 | |
| 790 | if (atomic_read(&slow_work_thread_count) >= slow_work_max_threads) |
| 791 | return; |
| 792 | |
| 793 | if (!mutex_trylock(&slow_work_user_lock)) |
| 794 | return; |
| 795 | |
| 796 | slow_work_may_not_start_new_thread = true; |
| 797 | atomic_inc(&slow_work_thread_count); |
| 798 | p = kthread_run(slow_work_thread, NULL, "kslowd"); |
| 799 | if (IS_ERR(p)) { |
| 800 | printk(KERN_DEBUG "Slow work thread pool: OOM\n"); |
| 801 | if (atomic_dec_and_test(&slow_work_thread_count)) |
| 802 | BUG(); /* we're running on a slow work thread... */ |
| 803 | mod_timer(&slow_work_oom_timer, |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 804 | round_jiffies(jiffies + SLOW_WORK_OOM_TIMEOUT)); |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 805 | } else { |
| 806 | /* ratelimit the starting of new threads */ |
| 807 | mod_timer(&slow_work_oom_timer, jiffies + 1); |
| 808 | } |
| 809 | |
| 810 | mutex_unlock(&slow_work_user_lock); |
| 811 | } |
| 812 | |
| 813 | static const struct slow_work_ops slow_work_new_thread_ops = { |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 814 | .owner = THIS_MODULE, |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 815 | .execute = slow_work_new_thread_execute, |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 816 | #ifdef CONFIG_SLOW_WORK_PROC |
| 817 | .desc = slow_work_new_thread_desc, |
| 818 | #endif |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 819 | }; |
| 820 | |
| 821 | /* |
| 822 | * post-OOM new thread start suppression expiration |
| 823 | */ |
| 824 | static void slow_work_oom_timeout(unsigned long data) |
| 825 | { |
| 826 | slow_work_may_not_start_new_thread = false; |
| 827 | } |
| 828 | |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 829 | #ifdef CONFIG_SYSCTL |
| 830 | /* |
| 831 | * Handle adjustment of the minimum number of threads |
| 832 | */ |
| 833 | static int slow_work_min_threads_sysctl(struct ctl_table *table, int write, |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 834 | void __user *buffer, |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 835 | size_t *lenp, loff_t *ppos) |
| 836 | { |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 837 | int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 838 | int n; |
| 839 | |
| 840 | if (ret == 0) { |
| 841 | mutex_lock(&slow_work_user_lock); |
| 842 | if (slow_work_user_count > 0) { |
| 843 | /* see if we need to start or stop threads */ |
| 844 | n = atomic_read(&slow_work_thread_count) - |
| 845 | slow_work_min_threads; |
| 846 | |
| 847 | if (n < 0 && !slow_work_may_not_start_new_thread) |
| 848 | slow_work_enqueue(&slow_work_new_thread); |
| 849 | else if (n > 0) |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 850 | slow_work_schedule_cull(); |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 851 | } |
| 852 | mutex_unlock(&slow_work_user_lock); |
| 853 | } |
| 854 | |
| 855 | return ret; |
| 856 | } |
| 857 | |
| 858 | /* |
| 859 | * Handle adjustment of the maximum number of threads |
| 860 | */ |
| 861 | static int slow_work_max_threads_sysctl(struct ctl_table *table, int write, |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 862 | void __user *buffer, |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 863 | size_t *lenp, loff_t *ppos) |
| 864 | { |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 865 | int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 866 | int n; |
| 867 | |
| 868 | if (ret == 0) { |
| 869 | mutex_lock(&slow_work_user_lock); |
| 870 | if (slow_work_user_count > 0) { |
| 871 | /* see if we need to stop threads */ |
| 872 | n = slow_work_max_threads - |
| 873 | atomic_read(&slow_work_thread_count); |
| 874 | |
| 875 | if (n < 0) |
Chris Peterson | 009789f | 2009-06-16 15:33:43 -0700 | [diff] [blame] | 876 | slow_work_schedule_cull(); |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 877 | } |
| 878 | mutex_unlock(&slow_work_user_lock); |
| 879 | } |
| 880 | |
| 881 | return ret; |
| 882 | } |
| 883 | #endif /* CONFIG_SYSCTL */ |
| 884 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 885 | /** |
| 886 | * slow_work_register_user - Register a user of the facility |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 887 | * @module: The module about to make use of the facility |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 888 | * |
| 889 | * Register a user of the facility, starting up the initial threads if there |
| 890 | * aren't any other users at this point. This will return 0 if successful, or |
| 891 | * an error if not. |
| 892 | */ |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 893 | int slow_work_register_user(struct module *module) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 894 | { |
| 895 | struct task_struct *p; |
| 896 | int loop; |
| 897 | |
| 898 | mutex_lock(&slow_work_user_lock); |
| 899 | |
| 900 | if (slow_work_user_count == 0) { |
| 901 | printk(KERN_NOTICE "Slow work thread pool: Starting up\n"); |
| 902 | init_completion(&slow_work_last_thread_exited); |
| 903 | |
| 904 | slow_work_threads_should_exit = false; |
David Howells | 109d927 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 905 | slow_work_init(&slow_work_new_thread, |
| 906 | &slow_work_new_thread_ops); |
| 907 | slow_work_may_not_start_new_thread = false; |
| 908 | slow_work_cull = false; |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 909 | |
| 910 | /* start the minimum number of threads */ |
| 911 | for (loop = 0; loop < slow_work_min_threads; loop++) { |
| 912 | atomic_inc(&slow_work_thread_count); |
| 913 | p = kthread_run(slow_work_thread, NULL, "kslowd"); |
| 914 | if (IS_ERR(p)) |
| 915 | goto error; |
| 916 | } |
| 917 | printk(KERN_NOTICE "Slow work thread pool: Ready\n"); |
| 918 | } |
| 919 | |
| 920 | slow_work_user_count++; |
| 921 | mutex_unlock(&slow_work_user_lock); |
| 922 | return 0; |
| 923 | |
| 924 | error: |
| 925 | if (atomic_dec_and_test(&slow_work_thread_count)) |
| 926 | complete(&slow_work_last_thread_exited); |
| 927 | if (loop > 0) { |
| 928 | printk(KERN_ERR "Slow work thread pool:" |
| 929 | " Aborting startup on ENOMEM\n"); |
| 930 | slow_work_threads_should_exit = true; |
| 931 | wake_up_all(&slow_work_thread_wq); |
| 932 | wait_for_completion(&slow_work_last_thread_exited); |
| 933 | printk(KERN_ERR "Slow work thread pool: Aborted\n"); |
| 934 | } |
| 935 | mutex_unlock(&slow_work_user_lock); |
| 936 | return PTR_ERR(p); |
| 937 | } |
| 938 | EXPORT_SYMBOL(slow_work_register_user); |
| 939 | |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 940 | /* |
| 941 | * wait for all outstanding items from the calling module to complete |
| 942 | * - note that more items may be queued whilst we're waiting |
| 943 | */ |
| 944 | static void slow_work_wait_for_items(struct module *module) |
| 945 | { |
| 946 | DECLARE_WAITQUEUE(myself, current); |
| 947 | struct slow_work *work; |
| 948 | int loop; |
| 949 | |
| 950 | mutex_lock(&slow_work_unreg_sync_lock); |
| 951 | add_wait_queue(&slow_work_unreg_wq, &myself); |
| 952 | |
| 953 | for (;;) { |
| 954 | spin_lock_irq(&slow_work_queue_lock); |
| 955 | |
| 956 | /* first of all, we wait for the last queued item in each list |
| 957 | * to be processed */ |
| 958 | list_for_each_entry_reverse(work, &vslow_work_queue, link) { |
| 959 | if (work->owner == module) { |
| 960 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 961 | slow_work_unreg_work_item = work; |
| 962 | goto do_wait; |
| 963 | } |
| 964 | } |
| 965 | list_for_each_entry_reverse(work, &slow_work_queue, link) { |
| 966 | if (work->owner == module) { |
| 967 | set_current_state(TASK_UNINTERRUPTIBLE); |
| 968 | slow_work_unreg_work_item = work; |
| 969 | goto do_wait; |
| 970 | } |
| 971 | } |
| 972 | |
| 973 | /* then we wait for the items being processed to finish */ |
| 974 | slow_work_unreg_module = module; |
| 975 | smp_mb(); |
| 976 | for (loop = 0; loop < SLOW_WORK_THREAD_LIMIT; loop++) { |
| 977 | if (slow_work_thread_processing[loop] == module) |
| 978 | goto do_wait; |
| 979 | } |
| 980 | spin_unlock_irq(&slow_work_queue_lock); |
| 981 | break; /* okay, we're done */ |
| 982 | |
| 983 | do_wait: |
| 984 | spin_unlock_irq(&slow_work_queue_lock); |
| 985 | schedule(); |
| 986 | slow_work_unreg_work_item = NULL; |
| 987 | slow_work_unreg_module = NULL; |
| 988 | } |
| 989 | |
| 990 | remove_wait_queue(&slow_work_unreg_wq, &myself); |
| 991 | mutex_unlock(&slow_work_unreg_sync_lock); |
| 992 | } |
| 993 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 994 | /** |
| 995 | * slow_work_unregister_user - Unregister a user of the facility |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 996 | * @module: The module whose items should be cleared |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 997 | * |
| 998 | * Unregister a user of the facility, killing all the threads if this was the |
| 999 | * last one. |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 1000 | * |
| 1001 | * This waits for all the work items belonging to the nominated module to go |
| 1002 | * away before proceeding. |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1003 | */ |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 1004 | void slow_work_unregister_user(struct module *module) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1005 | { |
David Howells | 3d7a641 | 2009-11-19 18:10:23 +0000 | [diff] [blame] | 1006 | /* first of all, wait for all outstanding items from the calling module |
| 1007 | * to complete */ |
| 1008 | if (module) |
| 1009 | slow_work_wait_for_items(module); |
| 1010 | |
| 1011 | /* then we can actually go about shutting down the facility if need |
| 1012 | * be */ |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1013 | mutex_lock(&slow_work_user_lock); |
| 1014 | |
| 1015 | BUG_ON(slow_work_user_count <= 0); |
| 1016 | |
| 1017 | slow_work_user_count--; |
| 1018 | if (slow_work_user_count == 0) { |
| 1019 | printk(KERN_NOTICE "Slow work thread pool: Shutting down\n"); |
| 1020 | slow_work_threads_should_exit = true; |
Jonathan Corbet | 418df63c | 2009-04-22 12:01:49 +0100 | [diff] [blame] | 1021 | del_timer_sync(&slow_work_cull_timer); |
| 1022 | del_timer_sync(&slow_work_oom_timer); |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1023 | wake_up_all(&slow_work_thread_wq); |
| 1024 | wait_for_completion(&slow_work_last_thread_exited); |
| 1025 | printk(KERN_NOTICE "Slow work thread pool:" |
| 1026 | " Shut down complete\n"); |
| 1027 | } |
| 1028 | |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1029 | mutex_unlock(&slow_work_user_lock); |
| 1030 | } |
| 1031 | EXPORT_SYMBOL(slow_work_unregister_user); |
| 1032 | |
| 1033 | /* |
| 1034 | * Initialise the slow work facility |
| 1035 | */ |
| 1036 | static int __init init_slow_work(void) |
| 1037 | { |
| 1038 | unsigned nr_cpus = num_possible_cpus(); |
| 1039 | |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1040 | if (slow_work_max_threads < nr_cpus) |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1041 | slow_work_max_threads = nr_cpus; |
David Howells | 12e22c5 | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1042 | #ifdef CONFIG_SYSCTL |
| 1043 | if (slow_work_max_max_threads < nr_cpus * 2) |
| 1044 | slow_work_max_max_threads = nr_cpus * 2; |
| 1045 | #endif |
David Howells | 8fba10a | 2009-11-19 18:10:51 +0000 | [diff] [blame] | 1046 | #ifdef CONFIG_SLOW_WORK_PROC |
| 1047 | proc_create("slow_work_rq", S_IFREG | 0400, NULL, |
| 1048 | &slow_work_runqueue_fops); |
| 1049 | #endif |
David Howells | 07fe7cb | 2009-04-03 16:42:35 +0100 | [diff] [blame] | 1050 | return 0; |
| 1051 | } |
| 1052 | |
| 1053 | subsys_initcall(init_slow_work); |