summaryrefslogtreecommitdiff
path: root/arch/powerpc/platforms/cell/spufs/sched.c
blob: 2f25e68b4bac1c5c8eb95160ba73d1ebfec44b81 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
/* sched.c - SPU scheduler.
 *
 * Copyright (C) IBM 2005
 * Author: Mark Nutter <mnutter@us.ibm.com>
 *
 * 2006-03-31	NUMA domains added.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#undef DEBUG

#include <linux/module.h>
#include <linux/errno.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/completion.h>
#include <linux/vmalloc.h>
#include <linux/smp.h>
#include <linux/smp_lock.h>
#include <linux/stddef.h>
#include <linux/unistd.h>
#include <linux/numa.h>
#include <linux/mutex.h>
#include <linux/notifier.h>

#include <asm/io.h>
#include <asm/mmu_context.h>
#include <asm/spu.h>
#include <asm/spu_csa.h>
#include <asm/spu_priv1.h>
#include "spufs.h"

#define SPU_TIMESLICE	(HZ)

struct spu_prio_array {
	DECLARE_BITMAP(bitmap, MAX_PRIO);
	struct list_head runq[MAX_PRIO];
	spinlock_t runq_lock;
	struct list_head active_list[MAX_NUMNODES];
	struct mutex active_mutex[MAX_NUMNODES];
};

static struct spu_prio_array *spu_prio;
static struct workqueue_struct *spu_sched_wq;

static inline int node_allowed(int node)
{
	cpumask_t mask;

	if (!nr_cpus_node(node))
		return 0;
	mask = node_to_cpumask(node);
	if (!cpus_intersects(mask, current->cpus_allowed))
		return 0;
	return 1;
}

void spu_start_tick(struct spu_context *ctx)
{
	if (ctx->policy == SCHED_RR)
		queue_delayed_work(spu_sched_wq, &ctx->sched_work, SPU_TIMESLICE);
}

void spu_stop_tick(struct spu_context *ctx)
{
	if (ctx->policy == SCHED_RR)
		cancel_delayed_work(&ctx->sched_work);
}

void spu_sched_tick(struct work_struct *work)
{
	struct spu_context *ctx =
		container_of(work, struct spu_context, sched_work.work);
	struct spu *spu;
	int rearm = 1;

	mutex_lock(&ctx->state_mutex);
	spu = ctx->spu;
	if (spu) {
		int best = sched_find_first_bit(spu_prio->bitmap);
		if (best <= ctx->prio) {
			spu_deactivate(ctx);
			rearm = 0;
		}
	}
	mutex_unlock(&ctx->state_mutex);

	if (rearm)
		spu_start_tick(ctx);
}

/**
 * spu_add_to_active_list - add spu to active list
 * @spu:	spu to add to the active list
 */
static void spu_add_to_active_list(struct spu *spu)
{
	mutex_lock(&spu_prio->active_mutex[spu->node]);
	list_add_tail(&spu->list, &spu_prio->active_list[spu->node]);
	mutex_unlock(&spu_prio->active_mutex[spu->node]);
}

/**
 * spu_remove_from_active_list - remove spu from active list
 * @spu:       spu to remove from the active list
 */
static void spu_remove_from_active_list(struct spu *spu)
{
	int node = spu->node;

	mutex_lock(&spu_prio->active_mutex[node]);
	list_del_init(&spu->list);
	mutex_unlock(&spu_prio->active_mutex[node]);
}

static inline void mm_needs_global_tlbie(struct mm_struct *mm)
{
	int nr = (NR_CPUS > 1) ? NR_CPUS : NR_CPUS + 1;

	/* Global TLBIE broadcast required with SPEs. */
	__cpus_setall(&mm->cpu_vm_mask, nr);
}

static BLOCKING_NOTIFIER_HEAD(spu_switch_notifier);

static void spu_switch_notify(struct spu *spu, struct spu_context *ctx)
{
	blocking_notifier_call_chain(&spu_switch_notifier,
			    ctx ? ctx->object_id : 0, spu);
}

int spu_switch_event_register(struct notifier_block * n)
{
	return blocking_notifier_chain_register(&spu_switch_notifier, n);
}

int spu_switch_event_unregister(struct notifier_block * n)
{
	return blocking_notifier_chain_unregister(&spu_switch_notifier, n);
}

/**
 * spu_bind_context - bind spu context to physical spu
 * @spu:	physical spu to bind to
 * @ctx:	context to bind
 */
static void spu_bind_context(struct spu *spu, struct spu_context *ctx)
{
	pr_debug("%s: pid=%d SPU=%d NODE=%d\n", __FUNCTION__, current->pid,
		 spu->number, spu->node);
	spu->ctx = ctx;
	spu->flags = 0;
	ctx->spu = spu;
	ctx->ops = &spu_hw_ops;
	spu->pid = current->pid;
	spu->mm = ctx->owner;
	mm_needs_global_tlbie(spu->mm);
	spu->ibox_callback = spufs_ibox_callback;
	spu->wbox_callback = spufs_wbox_callback;
	spu->stop_callback = spufs_stop_callback;
	spu->mfc_callback = spufs_mfc_callback;
	spu->dma_callback = spufs_dma_callback;
	mb();
	spu_unmap_mappings(ctx);
	spu_restore(&ctx->csa, spu);
	spu->timestamp = jiffies;
	spu_cpu_affinity_set(spu, raw_smp_processor_id());
	spu_switch_notify(spu, ctx);
	spu_add_to_active_list(spu);
	ctx->state = SPU_STATE_RUNNABLE;
}

/**
 * spu_unbind_context - unbind spu context from physical spu
 * @spu:	physical spu to unbind from
 * @ctx:	context to unbind
 */
static void spu_unbind_context(struct spu *spu, struct spu_context *ctx)
{
	pr_debug("%s: unbind pid=%d SPU=%d NODE=%d\n", __FUNCTION__,
		 spu->pid, spu->number, spu->node);

	spu_remove_from_active_list(spu);
	spu_switch_notify(spu, NULL);
	spu_unmap_mappings(ctx);
	spu_save(&ctx->csa, spu);
	spu->timestamp = jiffies;
	ctx->state = SPU_STATE_SAVED;
	spu->ibox_callback = NULL;
	spu->wbox_callback = NULL;
	spu->stop_callback = NULL;
	spu->mfc_callback = NULL;
	spu->dma_callback = NULL;
	spu->mm = NULL;
	spu->pid = 0;
	ctx->ops = &spu_backing_ops;
	ctx->spu = NULL;
	spu->flags = 0;
	spu->ctx = NULL;
}

/**
 * spu_add_to_rq - add a context to the runqueue
 * @ctx:       context to add
 */
static void spu_add_to_rq(struct spu_context *ctx)
{
	spin_lock(&spu_prio->runq_lock);
	list_add_tail(&ctx->rq, &spu_prio->runq[ctx->prio]);
	set_bit(ctx->prio, spu_prio->bitmap);
	spin_unlock(&spu_prio->runq_lock);
}

/**
 * spu_del_from_rq - remove a context from the runqueue
 * @ctx:       context to remove
 */
static void spu_del_from_rq(struct spu_context *ctx)
{
	spin_lock(&spu_prio->runq_lock);
	list_del_init(&ctx->rq);
	if (list_empty(&spu_prio->runq[ctx->prio]))
		clear_bit(ctx->prio, spu_prio->bitmap);
	spin_unlock(&spu_prio->runq_lock);
}

/**
 * spu_grab_context - remove one context from the runqueue
 * @prio:      priority of the context to be removed
 *
 * This function removes one context from the runqueue for priority @prio.
 * If there is more than one context with the given priority the first
 * task on the runqueue will be taken.
 *
 * Returns the spu_context it just removed.
 *
 * Must be called with spu_prio->runq_lock held.
 */
static struct spu_context *spu_grab_context(int prio)
{
	struct list_head *rq = &spu_prio->runq[prio];

	if (list_empty(rq))
		return NULL;
	return list_entry(rq->next, struct spu_context, rq);
}

static void spu_prio_wait(struct spu_context *ctx)
{
	DEFINE_WAIT(wait);

	set_bit(SPU_SCHED_WAKE, &ctx->sched_flags);
	prepare_to_wait_exclusive(&ctx->stop_wq, &wait, TASK_INTERRUPTIBLE);
	if (!signal_pending(current)) {
		mutex_unlock(&ctx->state_mutex);
		schedule();
		mutex_lock(&ctx->state_mutex);
	}
	__set_current_state(TASK_RUNNING);
	remove_wait_queue(&ctx->stop_wq, &wait);
	clear_bit(SPU_SCHED_WAKE, &ctx->sched_flags);
}

/**
 * spu_reschedule - try to find a runnable context for a spu
 * @spu:       spu available
 *
 * This function is called whenever a spu becomes idle.  It looks for the
 * most suitable runnable spu context and schedules it for execution.
 */
static void spu_reschedule(struct spu *spu)
{
	int best;

	spu_free(spu);

	spin_lock(&spu_prio->runq_lock);
	best = sched_find_first_bit(spu_prio->bitmap);
	if (best < MAX_PRIO) {
		struct spu_context *ctx = spu_grab_context(best);
		if (ctx && test_bit(SPU_SCHED_WAKE, &ctx->sched_flags))
			wake_up(&ctx->stop_wq);
	}
	spin_unlock(&spu_prio->runq_lock);
}

static struct spu *spu_get_idle(struct spu_context *ctx)
{
	struct spu *spu = NULL;
	int node = cpu_to_node(raw_smp_processor_id());
	int n;

	for (n = 0; n < MAX_NUMNODES; n++, node++) {
		node = (node < MAX_NUMNODES) ? node : 0;
		if (!node_allowed(node))
			continue;
		spu = spu_alloc_node(node);
		if (spu)
			break;
	}
	return spu;
}

/**
 * find_victim - find a lower priority context to preempt
 * @ctx:	canidate context for running
 *
 * Returns the freed physical spu to run the new context on.
 */
static struct spu *find_victim(struct spu_context *ctx)
{
	struct spu_context *victim = NULL;
	struct spu *spu;
	int node, n;

	/*
	 * Look for a possible preemption candidate on the local node first.
	 * If there is no candidate look at the other nodes.  This isn't
	 * exactly fair, but so far the whole spu schedule tries to keep
	 * a strong node affinity.  We might want to fine-tune this in
	 * the future.
	 */
 restart:
	node = cpu_to_node(raw_smp_processor_id());
	for (n = 0; n < MAX_NUMNODES; n++, node++) {
		node = (node < MAX_NUMNODES) ? node : 0;
		if (!node_allowed(node))
			continue;

		mutex_lock(&spu_prio->active_mutex[node]);
		list_for_each_entry(spu, &spu_prio->active_list[node], list) {
			struct spu_context *tmp = spu->ctx;

			if (tmp->rt_priority < ctx->rt_priority &&
			    (!victim || tmp->rt_priority < victim->rt_priority))
				victim = spu->ctx;
		}
		mutex_unlock(&spu_prio->active_mutex[node]);

		if (victim) {
			/*
			 * This nests ctx->state_mutex, but we always lock
			 * higher priority contexts before lower priority
			 * ones, so this is safe until we introduce
			 * priority inheritance schemes.
			 */
			if (!mutex_trylock(&victim->state_mutex)) {
				victim = NULL;
				goto restart;
			}

			spu = victim->spu;
			if (!spu) {
				/*
				 * This race can happen because we've dropped
				 * the active list mutex.  No a problem, just
				 * restart the search.
				 */
				mutex_unlock(&victim->state_mutex);
				victim = NULL;
				goto restart;
			}
			spu_unbind_context(spu, victim);
			mutex_unlock(&victim->state_mutex);
			return spu;
		}
	}

	return NULL;
}

/**
 * spu_activate - find a free spu for a context and execute it
 * @ctx:	spu context to schedule
 * @flags:	flags (currently ignored)
 *
 * Tries to find a free spu to run @ctx.  If no free spu is availble
 * add the context to the runqueue so it gets woken up once an spu
 * is available.
 */
int spu_activate(struct spu_context *ctx, unsigned long flags)
{

	if (ctx->spu)
		return 0;

	do {
		struct spu *spu;

		spu = spu_get_idle(ctx);
		/*
		 * If this is a realtime thread we try to get it running by
		 * preempting a lower priority thread.
		 */
		if (!spu && ctx->rt_priority)
			spu = find_victim(ctx);
		if (spu) {
			spu_bind_context(spu, ctx);
			return 0;
		}

		spu_add_to_rq(ctx);
		if (!(flags & SPU_ACTIVATE_NOWAKE))
			spu_prio_wait(ctx);
		spu_del_from_rq(ctx);
	} while (!signal_pending(current));

	return -ERESTARTSYS;
}

/**
 * spu_deactivate - unbind a context from it's physical spu
 * @ctx:	spu context to unbind
 *
 * Unbind @ctx from the physical spu it is running on and schedule
 * the highest priority context to run on the freed physical spu.
 */
void spu_deactivate(struct spu_context *ctx)
{
	struct spu *spu = ctx->spu;

	if (spu) {
		spu_unbind_context(spu, ctx);
		spu_reschedule(spu);
	}
}

/**
 * spu_yield -  yield a physical spu if others are waiting
 * @ctx:	spu context to yield
 *
 * Check if there is a higher priority context waiting and if yes
 * unbind @ctx from the physical spu and schedule the highest
 * priority context to run on the freed physical spu instead.
 */
void spu_yield(struct spu_context *ctx)
{
	struct spu *spu;
	int need_yield = 0;

	if (mutex_trylock(&ctx->state_mutex)) {
		if ((spu = ctx->spu) != NULL) {
			int best = sched_find_first_bit(spu_prio->bitmap);
			if (best < MAX_PRIO) {
				pr_debug("%s: yielding SPU %d NODE %d\n",
					 __FUNCTION__, spu->number, spu->node);
				spu_deactivate(ctx);
				need_yield = 1;
			}
		}
		mutex_unlock(&ctx->state_mutex);
	}
	if (unlikely(need_yield))
		yield();
}

int __init spu_sched_init(void)
{
	int i;

	spu_sched_wq = create_singlethread_workqueue("spusched");
	if (!spu_sched_wq)
		return 1;

	spu_prio = kzalloc(sizeof(struct spu_prio_array), GFP_KERNEL);
	if (!spu_prio) {
		printk(KERN_WARNING "%s: Unable to allocate priority queue.\n",
		       __FUNCTION__);
		       destroy_workqueue(spu_sched_wq);
		return 1;
	}
	for (i = 0; i < MAX_PRIO; i++) {
		INIT_LIST_HEAD(&spu_prio->runq[i]);
		__clear_bit(i, spu_prio->bitmap);
	}
	__set_bit(MAX_PRIO, spu_prio->bitmap);
	for (i = 0; i < MAX_NUMNODES; i++) {
		mutex_init(&spu_prio->active_mutex[i]);
		INIT_LIST_HEAD(&spu_prio->active_list[i]);
	}
	spin_lock_init(&spu_prio->runq_lock);
	return 0;
}

void __exit spu_sched_exit(void)
{
	struct spu *spu, *tmp;
	int node;

	for (node = 0; node < MAX_NUMNODES; node++) {
		mutex_lock(&spu_prio->active_mutex[node]);
		list_for_each_entry_safe(spu, tmp, &spu_prio->active_list[node],
					 list) {
			list_del_init(&spu->list);
			spu_free(spu);
		}
		mutex_unlock(&spu_prio->active_mutex[node]);
	}
	kfree(spu_prio);
	destroy_workqueue(spu_sched_wq);
}