sys/kern/subr_gtaskqueue.c

   1 /*-
   2  * Copyright (c) 2000 Doug Rabson
   3  * Copyright (c) 2014 Jeff Roberson
   4  * Copyright (c) 2016 Matthew Macy
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  26  * SUCH DAMAGE.
  27  */
  28
  29 #include <sys/cdefs.h>
  30 __FBSDID("$FreeBSD$");
  31
  32 #include <sys/param.h>
  33 #include <sys/systm.h>
  34 #include <sys/bus.h>
  35 #include <sys/cpuset.h>
  36 #include <sys/interrupt.h>
  37 #include <sys/kernel.h>
  38 #include <sys/kthread.h>
  39 #include <sys/libkern.h>
  40 #include <sys/limits.h>
  41 #include <sys/lock.h>
  42 #include <sys/malloc.h>
  43 #include <sys/mutex.h>
  44 #include <sys/proc.h>
  45 #include <sys/sched.h>
  46 #include <sys/smp.h>
  47 #include <sys/gtaskqueue.h>
  48 #include <sys/unistd.h>
  49 #include <machine/stdarg.h>
  50
  51 static MALLOC_DEFINE(M_GTASKQUEUE, "gtaskqueue", "Group Task Queues");
  52 static void     gtaskqueue_thread_enqueue(void *);
  53 static void     gtaskqueue_thread_loop(void *arg);
  54 static int      task_is_running(struct gtaskqueue *queue, struct gtask *gtask);
  55 static void     gtaskqueue_drain_locked(struct gtaskqueue *queue, struct gtask *gtask);
  56
  57 TASKQGROUP_DEFINE(softirq, mp_ncpus, 1);
  58
  59 struct gtaskqueue_busy {
  60         struct gtask            *tb_running;
  61         u_int                    tb_seq;
  62         LIST_ENTRY(gtaskqueue_busy) tb_link;
  63 };
  64
  65 typedef void (*gtaskqueue_enqueue_fn)(void *context);
  66
  67 struct gtaskqueue {
  68         STAILQ_HEAD(, gtask)    tq_queue;
  69         LIST_HEAD(, gtaskqueue_busy) tq_active;
  70         u_int                   tq_seq;
  71         int                     tq_callouts;
  72         struct mtx_padalign     tq_mutex;
  73         gtaskqueue_enqueue_fn   tq_enqueue;
  74         void                    *tq_context;
  75         char                    *tq_name;
  76         struct thread           **tq_threads;
  77         int                     tq_tcount;
  78         int                     tq_spin;
  79         int                     tq_flags;
  80         taskqueue_callback_fn   tq_callbacks[TASKQUEUE_NUM_CALLBACKS];
  81         void                    *tq_cb_contexts[TASKQUEUE_NUM_CALLBACKS];
  82 };
  83
  84 #define TQ_FLAGS_ACTIVE         (1 << 0)
  85 #define TQ_FLAGS_BLOCKED        (1 << 1)
  86 #define TQ_FLAGS_UNLOCKED_ENQUEUE       (1 << 2)
  87
  88 #define DT_CALLOUT_ARMED        (1 << 0)
  89
  90 #define TQ_LOCK(tq)                                                     \
  91         do {                                                            \
  92                 if ((tq)->tq_spin)                                      \
  93                         mtx_lock_spin(&(tq)->tq_mutex);                 \
  94                 else                                                    \
  95                         mtx_lock(&(tq)->tq_mutex);                      \
  96         } while (0)
  97 #define TQ_ASSERT_LOCKED(tq)    mtx_assert(&(tq)->tq_mutex, MA_OWNED)
  98
  99 #define TQ_UNLOCK(tq)                                                   \
 100         do {                                                            \
 101                 if ((tq)->tq_spin)                                      \
 102                         mtx_unlock_spin(&(tq)->tq_mutex);               \
 103                 else                                                    \
 104                         mtx_unlock(&(tq)->tq_mutex);                    \
 105         } while (0)
 106 #define TQ_ASSERT_UNLOCKED(tq)  mtx_assert(&(tq)->tq_mutex, MA_NOTOWNED)
 107
 108 #ifdef INVARIANTS
 109 static void
 110 gtask_dump(struct gtask *gtask)
 111 {
 112         printf("gtask: %p ta_flags=%x ta_priority=%d ta_func=%p ta_context=%p\n",
 113                gtask, gtask->ta_flags, gtask->ta_priority, gtask->ta_func, gtask->ta_context);
 114 }
 115 #endif
 116
 117 static __inline int
 118 TQ_SLEEP(struct gtaskqueue *tq, void *p, const char *wm)
 119 {
 120         if (tq->tq_spin)
 121                 return (msleep_spin(p, (struct mtx *)&tq->tq_mutex, wm, 0));
 122         return (msleep(p, &tq->tq_mutex, 0, wm, 0));
 123 }
 124
 125 static struct gtaskqueue *
 126 _gtaskqueue_create(const char *name, int mflags,
 127                  taskqueue_enqueue_fn enqueue, void *context,
 128                  int mtxflags, const char *mtxname __unused)
 129 {
 130         struct gtaskqueue *queue;
 131         char *tq_name;
 132
 133         tq_name = malloc(TASKQUEUE_NAMELEN, M_GTASKQUEUE, mflags | M_ZERO);
 134         if (!tq_name)
 135                 return (NULL);
 136
 137         snprintf(tq_name, TASKQUEUE_NAMELEN, "%s", (name) ? name : "taskqueue");
 138
 139         queue = malloc(sizeof(struct gtaskqueue), M_GTASKQUEUE, mflags | M_ZERO);
 140         if (!queue) {
 141                 free(tq_name, M_GTASKQUEUE);
 142                 return (NULL);
 143         }
 144
 145         STAILQ_INIT(&queue->tq_queue);
 146         LIST_INIT(&queue->tq_active);
 147         queue->tq_enqueue = enqueue;
 148         queue->tq_context = context;
 149         queue->tq_name = tq_name;
 150         queue->tq_spin = (mtxflags & MTX_SPIN) != 0;
 151         queue->tq_flags |= TQ_FLAGS_ACTIVE;
 152         if (enqueue == gtaskqueue_thread_enqueue)
 153                 queue->tq_flags |= TQ_FLAGS_UNLOCKED_ENQUEUE;
 154         mtx_init(&queue->tq_mutex, tq_name, NULL, mtxflags);
 155
 156         return (queue);
 157 }
 158
 159
 160 /*
 161  * Signal a taskqueue thread to terminate.
 162  */
 163 static void
 164 gtaskqueue_terminate(struct thread **pp, struct gtaskqueue *tq)
 165 {
 166
 167         while (tq->tq_tcount > 0 || tq->tq_callouts > 0) {
 168                 wakeup(tq);
 169                 TQ_SLEEP(tq, pp, "gtq_destroy");
 170         }
 171 }
 172
 173 static void
 174 gtaskqueue_free(struct gtaskqueue *queue)
 175 {
 176
 177         TQ_LOCK(queue);
 178         queue->tq_flags &= ~TQ_FLAGS_ACTIVE;
 179         gtaskqueue_terminate(queue->tq_threads, queue);
 180         KASSERT(LIST_EMPTY(&queue->tq_active), ("Tasks still running?"));
 181         KASSERT(queue->tq_callouts == 0, ("Armed timeout tasks"));
 182         mtx_destroy(&queue->tq_mutex);
 183         free(queue->tq_threads, M_GTASKQUEUE);
 184         free(queue->tq_name, M_GTASKQUEUE);
 185         free(queue, M_GTASKQUEUE);
 186 }
 187
 188 /*
 189  * Wait for all to complete, then prevent it from being enqueued
 190  */
 191 void
 192 grouptask_block(struct grouptask *grouptask)
 193 {
 194         struct gtaskqueue *queue = grouptask->gt_taskqueue;
 195         struct gtask *gtask = &grouptask->gt_task;
 196
 197 #ifdef INVARIANTS
 198         if (queue == NULL) {
 199                 gtask_dump(gtask);
 200                 panic("queue == NULL");
 201         }
 202 #endif
 203         TQ_LOCK(queue);
 204         gtask->ta_flags |= TASK_NOENQUEUE;
 205         gtaskqueue_drain_locked(queue, gtask);
 206         TQ_UNLOCK(queue);
 207 }
 208
 209 void
 210 grouptask_unblock(struct grouptask *grouptask)
 211 {
 212         struct gtaskqueue *queue = grouptask->gt_taskqueue;
 213         struct gtask *gtask = &grouptask->gt_task;
 214
 215 #ifdef INVARIANTS
 216         if (queue == NULL) {
 217                 gtask_dump(gtask);
 218                 panic("queue == NULL");
 219         }
 220 #endif
 221         TQ_LOCK(queue);
 222         gtask->ta_flags &= ~TASK_NOENQUEUE;
 223         TQ_UNLOCK(queue);
 224 }
 225
 226 int
 227 grouptaskqueue_enqueue(struct gtaskqueue *queue, struct gtask *gtask)
 228 {
 229 #ifdef INVARIANTS
 230         if (queue == NULL) {
 231                 gtask_dump(gtask);
 232                 panic("queue == NULL");
 233         }
 234 #endif
 235         TQ_LOCK(queue);
 236         if (gtask->ta_flags & TASK_ENQUEUED) {
 237                 TQ_UNLOCK(queue);
 238                 return (0);
 239         }
 240         if (gtask->ta_flags & TASK_NOENQUEUE) {
 241                 TQ_UNLOCK(queue);
 242                 return (EAGAIN);
 243         }
 244         STAILQ_INSERT_TAIL(&queue->tq_queue, gtask, ta_link);
 245         gtask->ta_flags |= TASK_ENQUEUED;
 246         TQ_UNLOCK(queue);
 247         if ((queue->tq_flags & TQ_FLAGS_BLOCKED) == 0)
 248                 queue->tq_enqueue(queue->tq_context);
 249         return (0);
 250 }
 251
 252 static void
 253 gtaskqueue_task_nop_fn(void *context)
 254 {
 255 }
 256
 257 /*
 258  * Block until all currently queued tasks in this taskqueue
 259  * have begun execution.  Tasks queued during execution of
 260  * this function are ignored.
 261  */
 262 static void
 263 gtaskqueue_drain_tq_queue(struct gtaskqueue *queue)
 264 {
 265         struct gtask t_barrier;
 266
 267         if (STAILQ_EMPTY(&queue->tq_queue))
 268                 return;
 269
 270         /*
 271          * Enqueue our barrier after all current tasks, but with
 272          * the highest priority so that newly queued tasks cannot
 273          * pass it.  Because of the high priority, we can not use
 274          * taskqueue_enqueue_locked directly (which drops the lock
 275          * anyway) so just insert it at tail while we have the
 276          * queue lock.
 277          */
 278         GTASK_INIT(&t_barrier, 0, USHRT_MAX, gtaskqueue_task_nop_fn, &t_barrier);
 279         STAILQ_INSERT_TAIL(&queue->tq_queue, &t_barrier, ta_link);
 280         t_barrier.ta_flags |= TASK_ENQUEUED;
 281
 282         /*
 283          * Once the barrier has executed, all previously queued tasks
 284          * have completed or are currently executing.
 285          */
 286         while (t_barrier.ta_flags & TASK_ENQUEUED)
 287                 TQ_SLEEP(queue, &t_barrier, "gtq_qdrain");
 288 }
 289
 290 /*
 291  * Block until all currently executing tasks for this taskqueue
 292  * complete.  Tasks that begin execution during the execution
 293  * of this function are ignored.
 294  */
 295 static void
 296 gtaskqueue_drain_tq_active(struct gtaskqueue *queue)
 297 {
 298         struct gtaskqueue_busy *tb;
 299         u_int seq;
 300
 301         if (LIST_EMPTY(&queue->tq_active))
 302                 return;
 303
 304         /* Block taskq_terminate().*/
 305         queue->tq_callouts++;
 306
 307         /* Wait for any active task with sequence from the past. */
 308         seq = queue->tq_seq;
 309 restart:
 310         LIST_FOREACH(tb, &queue->tq_active, tb_link) {
 311                 if ((int)(tb->tb_seq - seq) <= 0) {
 312                         TQ_SLEEP(queue, tb->tb_running, "gtq_adrain");
 313                         goto restart;
 314                 }
 315         }
 316
 317         /* Release taskqueue_terminate(). */
 318         queue->tq_callouts--;
 319         if ((queue->tq_flags & TQ_FLAGS_ACTIVE) == 0)
 320                 wakeup_one(queue->tq_threads);
 321 }
 322
 323 void
 324 gtaskqueue_block(struct gtaskqueue *queue)
 325 {
 326
 327         TQ_LOCK(queue);
 328         queue->tq_flags |= TQ_FLAGS_BLOCKED;
 329         TQ_UNLOCK(queue);
 330 }
 331
 332 void
 333 gtaskqueue_unblock(struct gtaskqueue *queue)
 334 {
 335
 336         TQ_LOCK(queue);
 337         queue->tq_flags &= ~TQ_FLAGS_BLOCKED;
 338         if (!STAILQ_EMPTY(&queue->tq_queue))
 339                 queue->tq_enqueue(queue->tq_context);
 340         TQ_UNLOCK(queue);
 341 }
 342
 343 static void
 344 gtaskqueue_run_locked(struct gtaskqueue *queue)
 345 {
 346         struct gtaskqueue_busy tb;
 347         struct gtask *gtask;
 348
 349         KASSERT(queue != NULL, ("tq is NULL"));
 350         TQ_ASSERT_LOCKED(queue);
 351         tb.tb_running = NULL;
 352         LIST_INSERT_HEAD(&queue->tq_active, &tb, tb_link);
 353
 354         while ((gtask = STAILQ_FIRST(&queue->tq_queue)) != NULL) {
 355                 STAILQ_REMOVE_HEAD(&queue->tq_queue, ta_link);
 356                 gtask->ta_flags &= ~TASK_ENQUEUED;
 357                 tb.tb_running = gtask;
 358                 tb.tb_seq = ++queue->tq_seq;
 359                 TQ_UNLOCK(queue);
 360
 361                 KASSERT(gtask->ta_func != NULL, ("task->ta_func is NULL"));
 362                 gtask->ta_func(gtask->ta_context);
 363
 364                 TQ_LOCK(queue);
 365                 wakeup(gtask);
 366         }
 367         LIST_REMOVE(&tb, tb_link);
 368 }
 369
 370 static int
 371 task_is_running(struct gtaskqueue *queue, struct gtask *gtask)
 372 {
 373         struct gtaskqueue_busy *tb;
 374
 375         TQ_ASSERT_LOCKED(queue);
 376         LIST_FOREACH(tb, &queue->tq_active, tb_link) {
 377                 if (tb->tb_running == gtask)
 378                         return (1);
 379         }
 380         return (0);
 381 }
 382
 383 static int
 384 gtaskqueue_cancel_locked(struct gtaskqueue *queue, struct gtask *gtask)
 385 {
 386
 387         if (gtask->ta_flags & TASK_ENQUEUED)
 388                 STAILQ_REMOVE(&queue->tq_queue, gtask, gtask, ta_link);
 389         gtask->ta_flags &= ~TASK_ENQUEUED;
 390         return (task_is_running(queue, gtask) ? EBUSY : 0);
 391 }
 392
 393 int
 394 gtaskqueue_cancel(struct gtaskqueue *queue, struct gtask *gtask)
 395 {
 396         int error;
 397
 398         TQ_LOCK(queue);
 399         error = gtaskqueue_cancel_locked(queue, gtask);
 400         TQ_UNLOCK(queue);
 401
 402         return (error);
 403 }
 404
 405 static void
 406 gtaskqueue_drain_locked(struct gtaskqueue *queue, struct gtask *gtask)
 407 {
 408         while ((gtask->ta_flags & TASK_ENQUEUED) || task_is_running(queue, gtask))
 409                 TQ_SLEEP(queue, gtask, "gtq_drain");
 410 }
 411
 412 void
 413 gtaskqueue_drain(struct gtaskqueue *queue, struct gtask *gtask)
 414 {
 415
 416         if (!queue->tq_spin)
 417                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, __func__);
 418
 419         TQ_LOCK(queue);
 420         gtaskqueue_drain_locked(queue, gtask);
 421         TQ_UNLOCK(queue);
 422 }
 423
 424 void
 425 gtaskqueue_drain_all(struct gtaskqueue *queue)
 426 {
 427
 428         if (!queue->tq_spin)
 429                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, __func__);
 430
 431         TQ_LOCK(queue);
 432         gtaskqueue_drain_tq_queue(queue);
 433         gtaskqueue_drain_tq_active(queue);
 434         TQ_UNLOCK(queue);
 435 }
 436
 437 static int
 438 _gtaskqueue_start_threads(struct gtaskqueue **tqp, int count, int pri,
 439     cpuset_t *mask, const char *name, va_list ap)
 440 {
 441         char ktname[MAXCOMLEN + 1];
 442         struct thread *td;
 443         struct gtaskqueue *tq;
 444         int i, error;
 445
 446         if (count <= 0)
 447                 return (EINVAL);
 448
 449         vsnprintf(ktname, sizeof(ktname), name, ap);
 450         tq = *tqp;
 451
 452         tq->tq_threads = malloc(sizeof(struct thread *) * count, M_GTASKQUEUE,
 453             M_NOWAIT | M_ZERO);
 454         if (tq->tq_threads == NULL) {
 455                 printf("%s: no memory for %s threads\n", __func__, ktname);
 456                 return (ENOMEM);
 457         }
 458
 459         for (i = 0; i < count; i++) {
 460                 if (count == 1)
 461                         error = kthread_add(gtaskqueue_thread_loop, tqp, NULL,
 462                             &tq->tq_threads[i], RFSTOPPED, 0, "%s", ktname);
 463                 else
 464                         error = kthread_add(gtaskqueue_thread_loop, tqp, NULL,
 465                             &tq->tq_threads[i], RFSTOPPED, 0,
 466                             "%s_%d", ktname, i);
 467                 if (error) {
 468                         /* should be ok to continue, taskqueue_free will dtrt */
 469                         printf("%s: kthread_add(%s): error %d", __func__,
 470                             ktname, error);
 471                         tq->tq_threads[i] = NULL;               /* paranoid */
 472                 } else
 473                         tq->tq_tcount++;
 474         }
 475         for (i = 0; i < count; i++) {
 476                 if (tq->tq_threads[i] == NULL)
 477                         continue;
 478                 td = tq->tq_threads[i];
 479                 if (mask) {
 480                         error = cpuset_setthread(td->td_tid, mask);
 481                         /*
 482                          * Failing to pin is rarely an actual fatal error;
 483                          * it'll just affect performance.
 484                          */
 485                         if (error)
 486                                 printf("%s: curthread=%llu: can't pin; "
 487                                     "error=%d\n",
 488                                     __func__,
 489                                     (unsigned long long) td->td_tid,
 490                                     error);
 491                 }
 492                 thread_lock(td);
 493                 sched_prio(td, pri);
 494                 sched_add(td, SRQ_BORING);
 495                 thread_unlock(td);
 496         }
 497
 498         return (0);
 499 }
 500
 501 static int
 502 gtaskqueue_start_threads(struct gtaskqueue **tqp, int count, int pri,
 503     const char *name, ...)
 504 {
 505         va_list ap;
 506         int error;
 507
 508         va_start(ap, name);
 509         error = _gtaskqueue_start_threads(tqp, count, pri, NULL, name, ap);
 510         va_end(ap);
 511         return (error);
 512 }
 513
 514 static inline void
 515 gtaskqueue_run_callback(struct gtaskqueue *tq,
 516     enum taskqueue_callback_type cb_type)
 517 {
 518         taskqueue_callback_fn tq_callback;
 519
 520         TQ_ASSERT_UNLOCKED(tq);
 521         tq_callback = tq->tq_callbacks[cb_type];
 522         if (tq_callback != NULL)
 523                 tq_callback(tq->tq_cb_contexts[cb_type]);
 524 }
 525
 526 static void
 527 gtaskqueue_thread_loop(void *arg)
 528 {
 529         struct gtaskqueue **tqp, *tq;
 530
 531         tqp = arg;
 532         tq = *tqp;
 533         gtaskqueue_run_callback(tq, TASKQUEUE_CALLBACK_TYPE_INIT);
 534         TQ_LOCK(tq);
 535         while ((tq->tq_flags & TQ_FLAGS_ACTIVE) != 0) {
 536                 /* XXX ? */
 537                 gtaskqueue_run_locked(tq);
 538                 /*
 539                  * Because taskqueue_run() can drop tq_mutex, we need to
 540                  * check if the TQ_FLAGS_ACTIVE flag wasn't removed in the
 541                  * meantime, which means we missed a wakeup.
 542                  */
 543                 if ((tq->tq_flags & TQ_FLAGS_ACTIVE) == 0)
 544                         break;
 545                 TQ_SLEEP(tq, tq, "-");
 546         }
 547         gtaskqueue_run_locked(tq);
 548         /*
 549          * This thread is on its way out, so just drop the lock temporarily
 550          * in order to call the shutdown callback.  This allows the callback
 551          * to look at the taskqueue, even just before it dies.
 552          */
 553         TQ_UNLOCK(tq);
 554         gtaskqueue_run_callback(tq, TASKQUEUE_CALLBACK_TYPE_SHUTDOWN);
 555         TQ_LOCK(tq);
 556
 557         /* rendezvous with thread that asked us to terminate */
 558         tq->tq_tcount--;
 559         wakeup_one(tq->tq_threads);
 560         TQ_UNLOCK(tq);
 561         kthread_exit();
 562 }
 563
 564 static void
 565 gtaskqueue_thread_enqueue(void *context)
 566 {
 567         struct gtaskqueue **tqp, *tq;
 568
 569         tqp = context;
 570         tq = *tqp;
 571         wakeup_any(tq);
 572 }
 573
 574
 575 static struct gtaskqueue *
 576 gtaskqueue_create_fast(const char *name, int mflags,
 577                  taskqueue_enqueue_fn enqueue, void *context)
 578 {
 579         return _gtaskqueue_create(name, mflags, enqueue, context,
 580                         MTX_SPIN, "fast_taskqueue");
 581 }
 582
 583
 584 struct taskqgroup_cpu {
 585         LIST_HEAD(, grouptask)  tgc_tasks;
 586         struct gtaskqueue       *tgc_taskq;
 587         int     tgc_cnt;
 588         int     tgc_cpu;
 589 };
 590
 591 struct taskqgroup {
 592         struct taskqgroup_cpu tqg_queue[MAXCPU];
 593         struct mtx      tqg_lock;
 594         const char *    tqg_name;
 595         int             tqg_adjusting;
 596         int             tqg_stride;
 597         int             tqg_cnt;
 598 };
 599
 600 struct taskq_bind_task {
 601         struct gtask bt_task;
 602         int     bt_cpuid;
 603 };
 604
 605 static void
 606 taskqgroup_cpu_create(struct taskqgroup *qgroup, int idx, int cpu)
 607 {
 608         struct taskqgroup_cpu *qcpu;
 609
 610         qcpu = &qgroup->tqg_queue[idx];
 611         LIST_INIT(&qcpu->tgc_tasks);
 612         qcpu->tgc_taskq = gtaskqueue_create_fast(NULL, M_WAITOK,
 613             taskqueue_thread_enqueue, &qcpu->tgc_taskq);
 614         gtaskqueue_start_threads(&qcpu->tgc_taskq, 1, PI_SOFT,
 615             "%s_%d", qgroup->tqg_name, idx);
 616         qcpu->tgc_cpu = cpu;
 617 }
 618
 619 static void
 620 taskqgroup_cpu_remove(struct taskqgroup *qgroup, int idx)
 621 {
 622
 623         gtaskqueue_free(qgroup->tqg_queue[idx].tgc_taskq);
 624 }
 625
 626 /*
 627  * Find the taskq with least # of tasks that doesn't currently have any
 628  * other queues from the uniq identifier.
 629  */
 630 static int
 631 taskqgroup_find(struct taskqgroup *qgroup, void *uniq)
 632 {
 633         struct grouptask *n;
 634         int i, idx, mincnt;
 635         int strict;
 636
 637         mtx_assert(&qgroup->tqg_lock, MA_OWNED);
 638         if (qgroup->tqg_cnt == 0)
 639                 return (0);
 640         idx = -1;
 641         mincnt = INT_MAX;
 642         /*
 643          * Two passes;  First scan for a queue with the least tasks that
 644          * does not already service this uniq id.  If that fails simply find
 645          * the queue with the least total tasks;
 646          */
 647         for (strict = 1; mincnt == INT_MAX; strict = 0) {
 648                 for (i = 0; i < qgroup->tqg_cnt; i++) {
 649                         if (qgroup->tqg_queue[i].tgc_cnt > mincnt)
 650                                 continue;
 651                         if (strict) {
 652                                 LIST_FOREACH(n,
 653                                     &qgroup->tqg_queue[i].tgc_tasks, gt_list)
 654                                         if (n->gt_uniq == uniq)
 655                                                 break;
 656                                 if (n != NULL)
 657                                         continue;
 658                         }
 659                         mincnt = qgroup->tqg_queue[i].tgc_cnt;
 660                         idx = i;
 661                 }
 662         }
 663         if (idx == -1)
 664                 panic("%s: failed to pick a qid.", __func__);
 665
 666         return (idx);
 667 }
 668
 669 /*
 670  * smp_started is unusable since it is not set for UP kernels or even for
 671  * SMP kernels when there is 1 CPU.  This is usually handled by adding a
 672  * (mp_ncpus == 1) test, but that would be broken here since we need to
 673  * to synchronize with the SI_SUB_SMP ordering.  Even in the pure SMP case
 674  * smp_started only gives a fuzzy ordering relative to SI_SUB_SMP.
 675  *
 676  * So maintain our own flag.  It must be set after all CPUs are started
 677  * and before SI_SUB_SMP:SI_ORDER_ANY so that the SYSINIT for delayed
 678  * adjustment is properly delayed.  SI_ORDER_FOURTH is clearly before
 679  * SI_ORDER_ANY and unclearly after the CPUs are started.  It would be
 680  * simpler for adjustment to pass a flag indicating if it is delayed.
 681  */
 682
 683 static int tqg_smp_started;
 684
 685 static void
 686 tqg_record_smp_started(void *arg)
 687 {
 688         tqg_smp_started = 1;
 689 }
 690
 691 SYSINIT(tqg_record_smp_started, SI_SUB_SMP, SI_ORDER_FOURTH,
 692         tqg_record_smp_started, NULL);
 693
 694 void
 695 taskqgroup_attach(struct taskqgroup *qgroup, struct grouptask *gtask,
 696     void *uniq, int irq, const char *name)
 697 {
 698         cpuset_t mask;
 699         int qid, error;
 700
 701         gtask->gt_uniq = uniq;
 702         snprintf(gtask->gt_name, GROUPTASK_NAMELEN, "%s", name ? name : "grouptask");
 703         gtask->gt_irq = irq;
 704         gtask->gt_cpu = -1;
 705         mtx_lock(&qgroup->tqg_lock);
 706         qid = taskqgroup_find(qgroup, uniq);
 707         qgroup->tqg_queue[qid].tgc_cnt++;
 708         LIST_INSERT_HEAD(&qgroup->tqg_queue[qid].tgc_tasks, gtask, gt_list);
 709         gtask->gt_taskqueue = qgroup->tqg_queue[qid].tgc_taskq;
 710         if (irq != -1 && tqg_smp_started) {
 711                 gtask->gt_cpu = qgroup->tqg_queue[qid].tgc_cpu;
 712                 CPU_ZERO(&mask);
 713                 CPU_SET(qgroup->tqg_queue[qid].tgc_cpu, &mask);
 714                 mtx_unlock(&qgroup->tqg_lock);
 715                 error = intr_setaffinity(irq, CPU_WHICH_IRQ, &mask);
 716                 if (error)
 717                         printf("%s: binding interrupt failed for %s: %d\n",
 718                             __func__, gtask->gt_name, error);
 719         } else
 720                 mtx_unlock(&qgroup->tqg_lock);
 721 }
 722
 723 static void
 724 taskqgroup_attach_deferred(struct taskqgroup *qgroup, struct grouptask *gtask)
 725 {
 726         cpuset_t mask;
 727         int qid, cpu, error;
 728
 729         mtx_lock(&qgroup->tqg_lock);
 730         qid = taskqgroup_find(qgroup, gtask->gt_uniq);
 731         cpu = qgroup->tqg_queue[qid].tgc_cpu;
 732         if (gtask->gt_irq != -1) {
 733                 mtx_unlock(&qgroup->tqg_lock);
 734
 735                 CPU_ZERO(&mask);
 736                 CPU_SET(cpu, &mask);
 737                 error = intr_setaffinity(gtask->gt_irq, CPU_WHICH_IRQ, &mask);
 738                 mtx_lock(&qgroup->tqg_lock);
 739                 if (error)
 740                         printf("%s: binding interrupt failed for %s: %d\n",
 741                             __func__, gtask->gt_name, error);
 742
 743         }
 744         qgroup->tqg_queue[qid].tgc_cnt++;
 745         LIST_INSERT_HEAD(&qgroup->tqg_queue[qid].tgc_tasks, gtask, gt_list);
 746         MPASS(qgroup->tqg_queue[qid].tgc_taskq != NULL);
 747         gtask->gt_taskqueue = qgroup->tqg_queue[qid].tgc_taskq;
 748         mtx_unlock(&qgroup->tqg_lock);
 749 }
 750
 751 int
 752 taskqgroup_attach_cpu(struct taskqgroup *qgroup, struct grouptask *gtask,
 753     void *uniq, int cpu, int irq, const char *name)
 754 {
 755         cpuset_t mask;
 756         int i, qid, error;
 757
 758         qid = -1;
 759         gtask->gt_uniq = uniq;
 760         snprintf(gtask->gt_name, GROUPTASK_NAMELEN, "%s", name ? name : "grouptask");
 761         gtask->gt_irq = irq;
 762         gtask->gt_cpu = cpu;
 763         mtx_lock(&qgroup->tqg_lock);
 764         if (tqg_smp_started) {
 765                 for (i = 0; i < qgroup->tqg_cnt; i++)
 766                         if (qgroup->tqg_queue[i].tgc_cpu == cpu) {
 767                                 qid = i;
 768                                 break;
 769                         }
 770                 if (qid == -1) {
 771                         mtx_unlock(&qgroup->tqg_lock);
 772                         printf("%s: qid not found for %s cpu=%d\n", __func__, gtask->gt_name, cpu);
 773                         return (EINVAL);
 774                 }
 775         } else
 776                 qid = 0;
 777         qgroup->tqg_queue[qid].tgc_cnt++;
 778         LIST_INSERT_HEAD(&qgroup->tqg_queue[qid].tgc_tasks, gtask, gt_list);
 779         gtask->gt_taskqueue = qgroup->tqg_queue[qid].tgc_taskq;
 780         cpu = qgroup->tqg_queue[qid].tgc_cpu;
 781         mtx_unlock(&qgroup->tqg_lock);
 782
 783         CPU_ZERO(&mask);
 784         CPU_SET(cpu, &mask);
 785         if (irq != -1 && tqg_smp_started) {
 786                 error = intr_setaffinity(irq, CPU_WHICH_IRQ, &mask);
 787                 if (error)
 788                         printf("%s: binding interrupt failed for %s: %d\n",
 789                             __func__, gtask->gt_name, error);
 790         }
 791         return (0);
 792 }
 793
 794 static int
 795 taskqgroup_attach_cpu_deferred(struct taskqgroup *qgroup, struct grouptask *gtask)
 796 {
 797         cpuset_t mask;
 798         int i, qid, irq, cpu, error;
 799
 800         qid = -1;
 801         irq = gtask->gt_irq;
 802         cpu = gtask->gt_cpu;
 803         MPASS(tqg_smp_started);
 804         mtx_lock(&qgroup->tqg_lock);
 805         for (i = 0; i < qgroup->tqg_cnt; i++)
 806                 if (qgroup->tqg_queue[i].tgc_cpu == cpu) {
 807                         qid = i;
 808                         break;
 809                 }
 810         if (qid == -1) {
 811                 mtx_unlock(&qgroup->tqg_lock);
 812                 printf("%s: qid not found for %s cpu=%d\n", __func__, gtask->gt_name, cpu);
 813                 return (EINVAL);
 814         }
 815         qgroup->tqg_queue[qid].tgc_cnt++;
 816         LIST_INSERT_HEAD(&qgroup->tqg_queue[qid].tgc_tasks, gtask, gt_list);
 817         MPASS(qgroup->tqg_queue[qid].tgc_taskq != NULL);
 818         gtask->gt_taskqueue = qgroup->tqg_queue[qid].tgc_taskq;
 819         mtx_unlock(&qgroup->tqg_lock);
 820
 821         CPU_ZERO(&mask);
 822         CPU_SET(cpu, &mask);
 823
 824         if (irq != -1) {
 825                 error = intr_setaffinity(irq, CPU_WHICH_IRQ, &mask);
 826                 if (error)
 827                         printf("%s: binding interrupt failed for %s: %d\n",
 828                             __func__, gtask->gt_name, error);
 829         }
 830         return (0);
 831 }
 832
 833 void
 834 taskqgroup_detach(struct taskqgroup *qgroup, struct grouptask *gtask)
 835 {
 836         int i;
 837
 838         grouptask_block(gtask);
 839         mtx_lock(&qgroup->tqg_lock);
 840         for (i = 0; i < qgroup->tqg_cnt; i++)
 841                 if (qgroup->tqg_queue[i].tgc_taskq == gtask->gt_taskqueue)
 842                         break;
 843         if (i == qgroup->tqg_cnt)
 844                 panic("%s: task %s not in group", __func__, gtask->gt_name);
 845         qgroup->tqg_queue[i].tgc_cnt--;
 846         LIST_REMOVE(gtask, gt_list);
 847         mtx_unlock(&qgroup->tqg_lock);
 848         gtask->gt_taskqueue = NULL;
 849         gtask->gt_task.ta_flags &= ~TASK_NOENQUEUE;
 850 }
 851
 852 static void
 853 taskqgroup_binder(void *ctx)
 854 {
 855         struct taskq_bind_task *gtask = (struct taskq_bind_task *)ctx;
 856         cpuset_t mask;
 857         int error;
 858
 859         CPU_ZERO(&mask);
 860         CPU_SET(gtask->bt_cpuid, &mask);
 861         error = cpuset_setthread(curthread->td_tid, &mask);
 862         thread_lock(curthread);
 863         sched_bind(curthread, gtask->bt_cpuid);
 864         thread_unlock(curthread);
 865
 866         if (error)
 867                 printf("%s: binding curthread failed: %d\n", __func__, error);
 868         free(gtask, M_DEVBUF);
 869 }
 870
 871 static void
 872 taskqgroup_bind(struct taskqgroup *qgroup)
 873 {
 874         struct taskq_bind_task *gtask;
 875         int i;
 876
 877         /*
 878          * Bind taskqueue threads to specific CPUs, if they have been assigned
 879          * one.
 880          */
 881         if (qgroup->tqg_cnt == 1)
 882                 return;
 883
 884         for (i = 0; i < qgroup->tqg_cnt; i++) {
 885                 gtask = malloc(sizeof (*gtask), M_DEVBUF, M_WAITOK);
 886                 GTASK_INIT(&gtask->bt_task, 0, 0, taskqgroup_binder, gtask);
 887                 gtask->bt_cpuid = qgroup->tqg_queue[i].tgc_cpu;
 888                 grouptaskqueue_enqueue(qgroup->tqg_queue[i].tgc_taskq,
 889                     &gtask->bt_task);
 890         }
 891 }
 892
 893 static int
 894 _taskqgroup_adjust(struct taskqgroup *qgroup, int cnt, int stride)
 895 {
 896         LIST_HEAD(, grouptask) gtask_head = LIST_HEAD_INITIALIZER(NULL);
 897         struct grouptask *gtask;
 898         int i, k, old_cnt, old_cpu, cpu;
 899
 900         mtx_assert(&qgroup->tqg_lock, MA_OWNED);
 901
 902         if (cnt < 1 || cnt * stride > mp_ncpus || !tqg_smp_started) {
 903                 printf("%s: failed cnt: %d stride: %d "
 904                     "mp_ncpus: %d tqg_smp_started: %d\n",
 905                     __func__, cnt, stride, mp_ncpus, tqg_smp_started);
 906                 return (EINVAL);
 907         }
 908         if (qgroup->tqg_adjusting) {
 909                 printf("%s failed: adjusting\n", __func__);
 910                 return (EBUSY);
 911         }
 912         qgroup->tqg_adjusting = 1;
 913         old_cnt = qgroup->tqg_cnt;
 914         old_cpu = 0;
 915         if (old_cnt < cnt)
 916                 old_cpu = qgroup->tqg_queue[old_cnt].tgc_cpu;
 917         mtx_unlock(&qgroup->tqg_lock);
 918         /*
 919          * Set up queue for tasks added before boot.
 920          */
 921         if (old_cnt == 0) {
 922                 LIST_SWAP(&gtask_head, &qgroup->tqg_queue[0].tgc_tasks,
 923                     grouptask, gt_list);
 924                 qgroup->tqg_queue[0].tgc_cnt = 0;
 925         }
 926
 927         /*
 928          * If new taskq threads have been added.
 929          */
 930         cpu = old_cpu;
 931         for (i = old_cnt; i < cnt; i++) {
 932                 taskqgroup_cpu_create(qgroup, i, cpu);
 933
 934                 for (k = 0; k < stride; k++)
 935                         cpu = CPU_NEXT(cpu);
 936         }
 937         mtx_lock(&qgroup->tqg_lock);
 938         qgroup->tqg_cnt = cnt;
 939         qgroup->tqg_stride = stride;
 940
 941         /*
 942          * Adjust drivers to use new taskqs.
 943          */
 944         for (i = 0; i < old_cnt; i++) {
 945                 while ((gtask = LIST_FIRST(&qgroup->tqg_queue[i].tgc_tasks))) {
 946                         LIST_REMOVE(gtask, gt_list);
 947                         qgroup->tqg_queue[i].tgc_cnt--;
 948                         LIST_INSERT_HEAD(&gtask_head, gtask, gt_list);
 949                 }
 950         }
 951         mtx_unlock(&qgroup->tqg_lock);
 952
 953         while ((gtask = LIST_FIRST(&gtask_head))) {
 954                 LIST_REMOVE(gtask, gt_list);
 955                 if (gtask->gt_cpu == -1)
 956                         taskqgroup_attach_deferred(qgroup, gtask);
 957                 else if (taskqgroup_attach_cpu_deferred(qgroup, gtask))
 958                         taskqgroup_attach_deferred(qgroup, gtask);
 959         }
 960
 961 #ifdef INVARIANTS
 962         mtx_lock(&qgroup->tqg_lock);
 963         for (i = 0; i < qgroup->tqg_cnt; i++) {
 964                 MPASS(qgroup->tqg_queue[i].tgc_taskq != NULL);
 965                 LIST_FOREACH(gtask, &qgroup->tqg_queue[i].tgc_tasks, gt_list)
 966                         MPASS(gtask->gt_taskqueue != NULL);
 967         }
 968         mtx_unlock(&qgroup->tqg_lock);
 969 #endif
 970         /*
 971          * If taskq thread count has been reduced.
 972          */
 973         for (i = cnt; i < old_cnt; i++)
 974                 taskqgroup_cpu_remove(qgroup, i);
 975
 976         taskqgroup_bind(qgroup);
 977
 978         mtx_lock(&qgroup->tqg_lock);
 979         qgroup->tqg_adjusting = 0;
 980
 981         return (0);
 982 }
 983
 984 int
 985 taskqgroup_adjust(struct taskqgroup *qgroup, int cnt, int stride)
 986 {
 987         int error;
 988
 989         mtx_lock(&qgroup->tqg_lock);
 990         error = _taskqgroup_adjust(qgroup, cnt, stride);
 991         mtx_unlock(&qgroup->tqg_lock);
 992
 993         return (error);
 994 }
 995
 996 struct taskqgroup *
 997 taskqgroup_create(const char *name)
 998 {
 999         struct taskqgroup *qgroup;
1000
1001         qgroup = malloc(sizeof(*qgroup), M_GTASKQUEUE, M_WAITOK | M_ZERO);
1002         mtx_init(&qgroup->tqg_lock, "taskqgroup", NULL, MTX_DEF);
1003         qgroup->tqg_name = name;
1004         LIST_INIT(&qgroup->tqg_queue[0].tgc_tasks);
1005
1006         return (qgroup);
1007 }
1008
1009 void
1010 taskqgroup_destroy(struct taskqgroup *qgroup)
1011 {
1012 }
1013