aboutsummaryrefslogtreecommitdiffstats
path: root/litmus/locking.c
blob: 4b8382cd77d1084705ecb9055d95d2682c5bd744 (plain) (blame)
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
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
#include <linux/sched.h>
#include <litmus/litmus.h>
#include <litmus/fdso.h>

#ifdef CONFIG_LITMUS_LOCKING

#include <linux/sched.h>
#include <litmus/litmus.h>
#include <litmus/sched_plugin.h>
#include <litmus/trace.h>
#include <litmus/litmus.h>
#include <litmus/wait.h>

#ifdef CONFIG_LITMUS_DGL_SUPPORT
#include <linux/uaccess.h>
#endif

#if defined(CONFIG_LITMUS_AFFINITY_LOCKING) && defined(CONFIG_LITMUS_NVIDIA)
#include <litmus/gpu_affinity.h>
#endif

static int create_generic_lock(void** obj_ref, obj_type_t type, void* __user arg);
static int open_generic_lock(struct od_table_entry* entry, void* __user arg);
static int close_generic_lock(struct od_table_entry* entry);
static void destroy_generic_lock(obj_type_t type, void* sem);

struct fdso_ops generic_lock_ops = {
	.create  = create_generic_lock,
	.open    = open_generic_lock,
	.close   = close_generic_lock,
	.destroy = destroy_generic_lock,
};

static atomic_t lock_id_gen = ATOMIC_INIT(0);


static inline bool is_lock(struct od_table_entry* entry)
{
	return entry->class == &generic_lock_ops;
}

static inline struct litmus_lock* get_lock(struct od_table_entry* entry)
{
	BUG_ON(!is_lock(entry));
	return (struct litmus_lock*) entry->obj->obj;
}

static  int create_generic_lock(void** obj_ref, obj_type_t type, void* __user arg)
{
	struct litmus_lock* lock;
	int err;

	err = litmus->allocate_lock(&lock, type, arg);
	if (err == 0) {
#ifdef CONFIG_LITMUS_NESTED_LOCKING
		lock->nest.lock = lock;
		lock->nest.hp_waiter_eff_prio = NULL;

		INIT_BINHEAP_NODE(&lock->nest.hp_binheap_node);
		if(!lock->nest.hp_waiter_ptr) {
			TRACE_CUR("BEWARE: hp_waiter_ptr should probably not be NULL in "
					  "most cases. (exception: IKGLP donors)\n");
		}
#endif
		lock->type = type;
		lock->ident = atomic_inc_return(&lock_id_gen);
		*obj_ref = lock;

		TRACE_CUR("Lock %d (%p) created. Type = %d\n.", lock->ident, lock, type);

		if (lock->proc && lock->proc->add) {
			lock->proc->add(lock);
		}
    }
	return err;
}

static int open_generic_lock(struct od_table_entry* entry, void* __user arg)
{
	struct litmus_lock* lock = get_lock(entry);
	if (lock->ops->open)
		return lock->ops->open(lock, arg);
	else
		return 0; /* default: any task can open it */
}

static int close_generic_lock(struct od_table_entry* entry)
{
	struct litmus_lock* lock = get_lock(entry);
	if (lock->ops->close) {

		if (lock->proc && lock->proc->remove) {
			lock->proc->remove(lock);
		}

		return lock->ops->close(lock);
	}
	else
		return 0; /* default: closing succeeds */
}

static void destroy_generic_lock(obj_type_t type, void* obj)
{
	struct litmus_lock* lock = (struct litmus_lock*) obj;
	lock->ops->deallocate(lock);
}

asmlinkage long sys_litmus_lock(int lock_od)
{
	long err = -EINVAL;
	struct od_table_entry* entry;
	struct litmus_lock* l;

	TS_SYSCALL_IN_START;

	TS_SYSCALL_IN_END;

	TS_LOCK_START;

	entry = get_entry_for_od(lock_od);
	if (entry && is_lock(entry)) {
		l = get_lock(entry);
		//TRACE_CUR("attempts to lock 0x%p\n", l);
		TRACE_CUR("Attempts to lock %d\n", l->ident);
		err = l->ops->lock(l);
		if (!err) {
			TRACE_CUR("Got lock %d\n", l->ident);
		}
	}

	/* Note: task my have been suspended or preempted in between!  Take
	 * this into account when computing overheads. */
	TS_LOCK_END;

	TS_SYSCALL_OUT_START;

	return err;
}

asmlinkage long sys_litmus_unlock(int lock_od)
{
	long err = -EINVAL;
	struct od_table_entry* entry;
	struct litmus_lock* l;

	TS_SYSCALL_IN_START;

	TS_SYSCALL_IN_END;

	TS_UNLOCK_START;

	entry = get_entry_for_od(lock_od);
	if (entry && is_lock(entry)) {
		l = get_lock(entry);
		//TRACE_CUR("attempts to unlock 0x%p\n", l);
		TRACE_CUR("Attempts to unlock %d\n", l->ident);
		err = l->ops->unlock(l);
		if (!err) {
			TRACE_CUR("Unlocked %d\n", l->ident);
		}
	}

	/* Note: task my have been preempted in between!  Take this into
	 * account when computing overheads. */
	TS_UNLOCK_END;

	TS_SYSCALL_OUT_START;

	return err;
}

struct task_struct* __waitqueue_remove_first(wait_queue_head_t *wq)
{
	wait_queue_t* q;
	struct task_struct* t = NULL;

	if (waitqueue_active(wq)) {
		q = list_entry(wq->task_list.next,
			       wait_queue_t, task_list);
		t = (struct task_struct*) q->private;
		__remove_wait_queue(wq, q);
	}
	return(t);
}

#ifdef CONFIG_LITMUS_NESTED_LOCKING

void print_hp_waiters(struct binheap_node* n, int depth)
{
	struct litmus_lock *l;
	struct nested_info *nest;
	char padding[81] = "                                                                                ";
	struct task_struct *hp = NULL;
	struct task_struct *hp_eff = NULL;
	struct task_struct *node_prio = NULL;


	if(n == NULL) {
		TRACE("+-> %p\n", NULL);
		return;
	}

	nest = binheap_entry(n, struct nested_info, hp_binheap_node);
	l = nest->lock;

	if(depth*2 <= 80)
		padding[depth*2] = '\0';

	if(nest->hp_waiter_ptr && *(nest->hp_waiter_ptr)) {
		hp = *(nest->hp_waiter_ptr);

		if(tsk_rt(hp)->inh_task) {
			hp_eff = tsk_rt(hp)->inh_task;
		}
	}

	node_prio = nest->hp_waiter_eff_prio;

	TRACE("%s+-> %s/%d [waiter = %s/%d] [waiter's inh = %s/%d] (lock = %d)\n",
		  padding,
		  (node_prio) ? node_prio->comm : "null",
		  (node_prio) ? node_prio->pid : 0,
		  (hp) ? hp->comm : "null",
		  (hp) ? hp->pid : 0,
		  (hp_eff) ? hp_eff->comm : "null",
		  (hp_eff) ? hp_eff->pid : 0,
		  l->ident);

    if(n->left) print_hp_waiters(n->left, depth+1);
    if(n->right) print_hp_waiters(n->right, depth+1);
}
#endif


#ifdef CONFIG_LITMUS_DGL_SUPPORT

struct prioq_mutex;

void select_next_lock(dgl_wait_state_t* dgl_wait /*, struct litmus_lock* prev_lock*/)
{
	int start = dgl_wait->last_primary;
	extern void __dump_prioq_lock_info(struct prioq_mutex *mutex);

	/*
	 We pick the next lock in reverse order. This causes inheritance propagation
	 from locks received earlier to flow in the same direction as regular nested
	 locking. This might make fine-grain DGL easier in the future.
	 */

	BUG_ON(tsk_rt(dgl_wait->task)->blocked_lock);

	// note reverse order
	for(dgl_wait->last_primary = (dgl_wait->last_primary != 0) ?  dgl_wait->last_primary - 1 : dgl_wait->size-1;
		dgl_wait->last_primary != start;
		dgl_wait->last_primary = (dgl_wait->last_primary != 0) ?  dgl_wait->last_primary - 1 : dgl_wait->size-1)
	{

		struct litmus_lock *l = dgl_wait->locks[dgl_wait->last_primary];

		if(!l->ops->is_owner(l, dgl_wait->task) &&
		   l->ops->get_owner(l)) {

			tsk_rt(dgl_wait->task)->blocked_lock =
					dgl_wait->locks[dgl_wait->last_primary];
			mb();

			TRACE_TASK(dgl_wait->task, "New blocked lock is %d\n", l->ident);

			l->ops->enable_priority(l, dgl_wait);

			return;
		}
	}

	// There was no one to push on.  This can happen if the blocked task is
	// behind a task that is idling a prioq-mutex.

	// note reverse order
	dgl_wait->last_primary = start;
	for(dgl_wait->last_primary = (dgl_wait->last_primary != 0) ?  dgl_wait->last_primary - 1 : dgl_wait->size-1;
		dgl_wait->last_primary != start;
		dgl_wait->last_primary = (dgl_wait->last_primary != 0) ?  dgl_wait->last_primary - 1 : dgl_wait->size-1)
	{

		struct litmus_lock *l = dgl_wait->locks[dgl_wait->last_primary];

		if(!l->ops->is_owner(l, dgl_wait->task)) {

			tsk_rt(dgl_wait->task)->blocked_lock =
			dgl_wait->locks[dgl_wait->last_primary];
			mb();

			TRACE_TASK(dgl_wait->task, "New blocked lock is %d\n", l->ident);

			l->ops->enable_priority(l, dgl_wait);

			return;
		}
	}

	BUG();
}

int dgl_wake_up(wait_queue_t *wq_node, unsigned mode, int sync, void *key)
{
	// should never be called.
	BUG();
	return 1;
}

struct task_struct* __waitqueue_dgl_remove_first(wait_queue_head_t *wq,
								  dgl_wait_state_t** dgl_wait)
{
	wait_queue_t *q;
	struct task_struct *task = NULL;

	*dgl_wait = NULL;

	if (waitqueue_active(wq)) {
		q = list_entry(wq->task_list.next,
					   wait_queue_t, task_list);

		if(q->func == dgl_wake_up) {
			*dgl_wait = (dgl_wait_state_t*) q->private;
			task = (*dgl_wait)->task;
		}
		else {
			task = (struct task_struct*) q->private;
		}

		__remove_wait_queue(wq, q);
	}
	return task;
}

void init_dgl_waitqueue_entry(wait_queue_t *wq_node, dgl_wait_state_t* dgl_wait)
{
	init_waitqueue_entry(wq_node, dgl_wait->task);
	wq_node->private = dgl_wait;
	wq_node->func = dgl_wake_up;
}

#ifdef CONFIG_SCHED_DEBUG_TRACE
static void snprintf_dgl(char* buf, size_t bsz, struct litmus_lock* dgl_locks[], int sz)
{
	int i;
	char* ptr;

	ptr = buf;
	for(i = 0; i < sz && ptr < buf+bsz; ++i)
	{
		struct litmus_lock *l = dgl_locks[i];
		int remaining = bsz - (ptr-buf);
		int written;

		if(i == 0)
			written = snprintf(ptr, remaining, "%d ", l->ident);
		else if(i == sz - 1)
			written = snprintf(ptr, remaining, " %d", l->ident);
		else
			written = snprintf(ptr, remaining, " %d ", l->ident);
		ptr += written;
	}
}
#endif


static int failed_owner(struct litmus_lock *cur_lock, struct task_struct *t)
{
	struct task_struct *cur_owner = cur_lock->ops->get_owner(cur_lock);
	printk(KERN_EMERG "lock %d expected owner %s/%d but got %s/%d.\n",
					cur_lock->ident,
					(t) ? t->comm : "null",
					(t) ? t->pid : 0,
					(cur_owner) ? cur_owner->comm : "null",
					(cur_owner) ? cur_owner->pid : 0);
	BUG();
}

/* only valid when locks are prioq locks!!!
 * THE BIG DGL LOCK MUST BE HELD! */
int __attempt_atomic_dgl_acquire(struct litmus_lock *cur_lock, dgl_wait_state_t *dgl_wait)
{
	int i;

	/* check to see if we can take all the locks */
	for(i = 0; i < dgl_wait->size; ++i) {
		struct litmus_lock *l = dgl_wait->locks[i];
		if(!l->ops->dgl_can_quick_lock(l, dgl_wait->task))
		{
			return -1;
		}
	}

	/* take the locks */
	for(i = 0; i < dgl_wait->size; ++i) {
		struct litmus_lock *l = dgl_wait->locks[i];

		l->ops->dgl_quick_lock(l, cur_lock, dgl_wait->task, &dgl_wait->wq_nodes[i]);

		if(!(l->ops->is_owner(l, dgl_wait->task)))
			failed_owner(l, dgl_wait->task);
		//BUG_ON(!(l->ops->is_owner(l, dgl_wait->task)));
	}

	return 0; /* success */
}


static long do_litmus_dgl_lock(dgl_wait_state_t *dgl_wait)
{
	int i;
	unsigned long irqflags; //, dummyflags;
	raw_spinlock_t *dgl_lock;

#ifdef CONFIG_SCHED_DEBUG_TRACE
	char dglstr[CONFIG_LITMUS_MAX_DGL_SIZE*5];
	snprintf_dgl(dglstr, sizeof(dglstr), dgl_wait->locks, dgl_wait->size);
	TRACE_CUR("Locking DGL with size %d: %s\n", dgl_wait->size, dglstr);
#endif

	dgl_lock = litmus->get_dgl_spinlock(dgl_wait->task);

	BUG_ON(dgl_wait->task != current);

	raw_spin_lock_irqsave(dgl_lock, irqflags);

	dgl_wait->nr_remaining = dgl_wait->size;

	// try to acquire each lock.  enqueue (non-blocking) if it is unavailable.
	for(i = 0; i < dgl_wait->size; ++i) {
		struct litmus_lock *l = dgl_wait->locks[i];

		// dgl_lock() must set task state to TASK_UNINTERRUPTIBLE if task blocks.

		if(l->ops->dgl_lock(l, dgl_wait, &dgl_wait->wq_nodes[i])) {
			--(dgl_wait->nr_remaining);
			TRACE_CUR("Acquired lock %d immediatly.\n", l->ident);
		}
	}

	if(dgl_wait->nr_remaining == 0) {
		// acquired entire group immediatly
		TRACE_CUR("Acquired all locks in DGL immediatly!\n");
	}
	else {

		TRACE_CUR("As many as %d locks in DGL are pending. Suspending.\n",
				  dgl_wait->nr_remaining);

#if defined(CONFIG_LITMUS_AFFINITY_LOCKING) && defined(CONFIG_LITMUS_NVIDIA)
		// KLUDGE: don't count this suspension as time in the critical gpu
		// critical section
		if(tsk_rt(dgl_wait->task)->held_gpus) {
			tsk_rt(dgl_wait->task)->suspend_gpu_tracker_on_block = 1;
		}
#endif

		// note reverse order.  see comments in select_next_lock for reason.
		for(i = dgl_wait->size - 1; i >= 0; --i) {
			struct litmus_lock *l = dgl_wait->locks[i];
			if(!l->ops->is_owner(l, dgl_wait->task)) {  // double-check to be thread safe

				TRACE_CUR("Activating priority inheritance on lock %d\n",
						  l->ident);

				TS_DGL_LOCK_SUSPEND;

				l->ops->enable_priority(l, dgl_wait);
				dgl_wait->last_primary = i;

				TRACE_CUR("Suspending for lock %d\n", l->ident);

				raw_spin_unlock_irqrestore(dgl_lock, irqflags);  // free dgl_lock before suspending

				suspend_for_lock(); // suspend!!!

				TS_DGL_LOCK_RESUME;

				TRACE_CUR("Woken up from DGL suspension.\n");

				goto all_acquired;  // we should hold all locks when we wake up.
			}
		}

		TRACE_CUR("Didn't have to suspend after all, but calling schedule() anyway.\n");
		//BUG();
	}

	raw_spin_unlock_irqrestore(dgl_lock, irqflags);

all_acquired:

	// FOR SANITY CHECK FOR TESTING
//	for(i = 0; i < dgl_wait->size; ++i) {
//		struct litmus_lock *l = dgl_wait->locks[i];
//		BUG_ON(!l->ops->is_owner(l, dgl_wait->task));
//	}

	TRACE_CUR("Acquired entire DGL\n");

	return 0;
}



static long do_litmus_dgl_atomic_lock(dgl_wait_state_t *dgl_wait)
{
	int i;
	unsigned long irqflags; //, dummyflags;
	raw_spinlock_t *dgl_lock;
	struct litmus_lock *l;

#ifdef CONFIG_SCHED_DEBUG_TRACE
	char dglstr[CONFIG_LITMUS_MAX_DGL_SIZE*5];
	snprintf_dgl(dglstr, sizeof(dglstr), dgl_wait->locks, dgl_wait->size);
	TRACE_CUR("Atomic locking DGL with size %d: %s\n", dgl_wait->size, dglstr);
#endif

	dgl_lock = litmus->get_dgl_spinlock(dgl_wait->task);

	BUG_ON(dgl_wait->task != current);

	raw_spin_lock_irqsave(dgl_lock, irqflags);


	dgl_wait->nr_remaining = dgl_wait->size;

	for(i = 0; i < dgl_wait->size; ++i) {
		struct litmus_lock *l = dgl_wait->locks[i];
		l->ops->dgl_lock(l, dgl_wait, &dgl_wait->wq_nodes[i]);  // this should be a forced enqueue if atomic DGLs are needed.
	}

	if(__attempt_atomic_dgl_acquire(NULL, dgl_wait)) {
		/* Failed to acquire all locks at once.
		 * Pick a lock to push on and suspend. */
		TRACE_CUR("Could not atomically acquire all locks.\n");


#if defined(CONFIG_LITMUS_AFFINITY_LOCKING) && defined(CONFIG_LITMUS_NVIDIA)
		// KLUDGE: don't count this suspension as time in the critical gpu
		// critical section
		if(tsk_rt(dgl_wait->task)->held_gpus) {
			tsk_rt(dgl_wait->task)->suspend_gpu_tracker_on_block = 1;
		}
#endif
		// we are not the owner of any lock, so push on the last one in the DGL
		// by default.

		l = dgl_wait->locks[dgl_wait->size - 1];

		TRACE_CUR("Activating priority inheritance on lock %d\n",
				  l->ident);

		TS_DGL_LOCK_SUSPEND;

		l->ops->enable_priority(l, dgl_wait);
		dgl_wait->last_primary = dgl_wait->size - 1;

		TRACE_CUR("Suspending for lock %d\n", l->ident);

		raw_spin_unlock_irqrestore(dgl_lock, irqflags);  // free dgl_lock before suspending

		suspend_for_lock(); // suspend!!!

		TS_DGL_LOCK_RESUME;

		TRACE_CUR("Woken up from DGL suspension.\n");

		goto all_acquired;  // we should hold all locks when we wake up.
	}
	raw_spin_unlock_irqrestore(dgl_lock, irqflags);

all_acquired:

	dgl_wait->nr_remaining = 0;

	// SANITY CHECK FOR TESTING
	for(i = 0; i < dgl_wait->size; ++i) {
		struct litmus_lock *l = dgl_wait->locks[i];
		if(!(l->ops->is_owner(l, dgl_wait->task)))
			failed_owner(l, dgl_wait->task);
		//BUG_ON(!l->ops->is_owner(l, dgl_wait->task));
	}

	TRACE_CUR("Acquired entire DGL\n");

	return 0;
}




static int supports_dgl(struct litmus_lock *l)
{
	struct litmus_lock_ops* ops = l->ops;
	return (ops->dgl_lock			&&
			ops->is_owner			&&
			ops->get_owner			&&
			ops->enable_priority);
}

static int needs_atomic_dgl(struct litmus_lock *l)
{
	struct litmus_lock_ops* ops = l->ops;
	return (ops->dgl_quick_lock != NULL);
}

asmlinkage long sys_litmus_dgl_lock(void* __user usr_dgl_ods, int dgl_size)
{
	struct task_struct *t = current;
	long err = -EINVAL;
	int dgl_ods[MAX_DGL_SIZE];
	int i;

	int num_need_atomic = 0;

	dgl_wait_state_t dgl_wait_state;  // lives on the stack until all resources in DGL are held.

	if(dgl_size > MAX_DGL_SIZE || dgl_size < 1)
		goto out;

	if(!access_ok(VERIFY_READ, usr_dgl_ods, dgl_size*(sizeof(int))))
		goto out;

	if(__copy_from_user(&dgl_ods, usr_dgl_ods, dgl_size*(sizeof(int))))
		goto out;

	if (!is_realtime(t)) {
		err = -EPERM;
		goto out;
	}

	if (dgl_size == 1) {
		/* DGL size of 1. Just call regular singular lock. */
		TRACE_CUR("DGL lock with size = 1. Treating as regular lock.\n");
		err = sys_litmus_lock(dgl_ods[0]);
	}
	else {
		for(i = 0; i < dgl_size; ++i) {
			struct od_table_entry *entry = get_entry_for_od(dgl_ods[i]);
			if(entry && is_lock(entry)) {
				dgl_wait_state.locks[i] = get_lock(entry);
				if(!supports_dgl(dgl_wait_state.locks[i])) {
					TRACE_CUR("Lock %d does not support all required DGL operations.\n",
							  dgl_wait_state.locks[i]->ident);
					goto out;
				}

				if (needs_atomic_dgl(dgl_wait_state.locks[i])) {
					++num_need_atomic;
				}
			}
			else {
				TRACE_CUR("Invalid lock identifier\n");
				goto out;
			}
		}

		if (num_need_atomic && num_need_atomic != dgl_size) {
			TRACE_CUR("All locks in DGL must support atomic acquire if any one does.\n");
			goto out;
		}

		dgl_wait_state.task = t;
		dgl_wait_state.size = dgl_size;

		TS_DGL_LOCK_START;
		if (!num_need_atomic)
			err = do_litmus_dgl_lock(&dgl_wait_state);
		else
			err = do_litmus_dgl_atomic_lock(&dgl_wait_state);

		/* Note: task my have been suspended or preempted in between!  Take
		 * this into account when computing overheads. */
		TS_DGL_LOCK_END;
	}

out:
	return err;
}

static long do_litmus_dgl_unlock(struct litmus_lock* dgl_locks[], int dgl_size)
{
	int i;
	long err = 0;

#ifdef CONFIG_SCHED_DEBUG_TRACE
	char dglstr[CONFIG_LITMUS_MAX_DGL_SIZE*5];
	snprintf_dgl(dglstr, sizeof(dglstr), dgl_locks, dgl_size);
	TRACE_CUR("Unlocking a DGL with size %d: %s\n", dgl_size, dglstr);
#endif

	for(i = dgl_size - 1; i >= 0; --i) {  // unlock in reverse order

		struct litmus_lock *l = dgl_locks[i];
		long tmp_err;

		TRACE_CUR("Unlocking lock %d of DGL.\n", l->ident);

		tmp_err = l->ops->unlock(l);

		if(tmp_err) {
			TRACE_CUR("There was an error unlocking %d: %d.\n", l->ident, tmp_err);
			err = tmp_err;
		}
	}

	TRACE_CUR("DGL unlocked. err = %d\n", err);

	return err;
}

asmlinkage long sys_litmus_dgl_unlock(void* __user usr_dgl_ods, int dgl_size)
{
	long err = -EINVAL;
	int dgl_ods[MAX_DGL_SIZE];
	struct od_table_entry* entry;
	int i;

	struct litmus_lock* dgl_locks[MAX_DGL_SIZE];

	if(dgl_size > MAX_DGL_SIZE || dgl_size < 1)
		goto out;

	if(!access_ok(VERIFY_READ, usr_dgl_ods, dgl_size*(sizeof(int))))
		goto out;

	if(__copy_from_user(&dgl_ods, usr_dgl_ods, dgl_size*(sizeof(int))))
		goto out;


	if (dgl_size == 1) {
		/* DGL size of 1. Just call regular singular lock. */
		TRACE_CUR("DGL unlock with size = 1. Treating as regular unlock.\n");
		err = sys_litmus_unlock(dgl_ods[0]);
	}
	else {
		for(i = 0; i < dgl_size; ++i) {
			entry = get_entry_for_od(dgl_ods[i]);
			if(entry && is_lock(entry)) {
				dgl_locks[i] = get_lock(entry);
				if(!supports_dgl(dgl_locks[i])) {
					TRACE_CUR("Lock %d does not support all required DGL operations.\n",
							  dgl_locks[i]->ident);
					goto out;
				}
			}
			else {
				TRACE_CUR("Invalid lock identifier\n");
				goto out;
			}
		}

		TS_DGL_UNLOCK_START;
		err = do_litmus_dgl_unlock(dgl_locks, dgl_size);

		/* Note: task my have been suspended or preempted in between!  Take
		 * this into account when computing overheads. */
		TS_DGL_UNLOCK_END;
	}

out:
	return err;
}

#else  // CONFIG_LITMUS_DGL_SUPPORT

asmlinkage long sys_litmus_dgl_lock(void* __user usr_dgl_ods, int dgl_size)
{
	return -ENOSYS;
}

asmlinkage long sys_litmus_dgl_unlock(void* __user usr_dgl_ods, int dgl_size)
{
	return -ENOSYS;
}

#endif

unsigned int __add_wait_queue_prio_exclusive(
	wait_queue_head_t* head,
	prio_wait_queue_t *new)
{
	struct list_head *pos;
	unsigned int passed = 0;

	new->wq.flags |= WQ_FLAG_EXCLUSIVE;

	/* find a spot where the new entry is less than the next */
	list_for_each(pos, &head->task_list) {
		prio_wait_queue_t* queued = list_entry(pos, prio_wait_queue_t,
						       wq.task_list);

		if (unlikely(lt_before(new->priority, queued->priority) ||
			     (new->priority == queued->priority &&
			      new->tie_breaker < queued->tie_breaker))) {
			/* pos is not less than new, thus insert here */
			__list_add(&new->wq.task_list, pos->prev, pos);
			goto out;
		}
		passed++;
	}

	/* if we get to this point either the list is empty or every entry
	 * queued element is less than new.
	 * Let's add new to the end. */
	list_add_tail(&new->wq.task_list, &head->task_list);
out:
	return passed;
}


void suspend_for_lock(void)
{
#if defined(CONFIG_REALTIME_AUX_TASKS) || defined(CONFIG_LITMUS_NVIDIA)
	struct task_struct *t = current;
#endif

#ifdef CONFIG_REALTIME_AUX_TASKS
	unsigned int aux_restore = 0;
	unsigned int aux_hide;
#endif

#ifdef CONFIG_LITMUS_NVIDIA
	unsigned int gpu_restore = 0;
	unsigned int gpu_hide;
#endif

#ifdef CONFIG_REALTIME_AUX_TASKS
	if (tsk_rt(t)->has_aux_tasks) {
		/* hide from aux tasks so they can't inherit our priority when we block
		 * for a litmus lock. inheritance is already going to a litmus lock
		 * holder. */
		aux_hide = tsk_rt(t)->hide_from_aux_tasks;
		aux_restore = 1;
		tsk_rt(t)->hide_from_aux_tasks = 1;
	}
#endif

#ifdef CONFIG_LITMUS_NVIDIA
	if (tsk_rt(t)->held_gpus) {
		gpu_hide = tsk_rt(t)->hide_from_gpu;
		gpu_restore = 1;
		tsk_rt(t)->hide_from_gpu = 1;
	}
#endif

#if defined(CONFIG_LITMUS_AFFINITY_LOCKING) && defined(CONFIG_LITMUS_NVIDIA)
	// disable tracking
	if(tsk_rt(t)->held_gpus) {
		tsk_rt(t)->suspend_gpu_tracker_on_block = 1;
	}
#endif
	
	schedule();

	
	/* TODO: Move the following to wake_up_for_lock()? */
	
#if defined(CONFIG_LITMUS_AFFINITY_LOCKING) && defined(CONFIG_LITMUS_NVIDIA)
	// re-enable tracking
	if(tsk_rt(t)->held_gpus) {
		tsk_rt(t)->suspend_gpu_tracker_on_block = 0;
	}
#endif
	
#ifdef CONFIG_LITMUS_NVIDIA
	if (gpu_restore) {
		/* restore our state */
		tsk_rt(t)->hide_from_gpu = gpu_hide;
	}
#endif

#ifdef CONFIG_REALTIME_AUX_TASKS
	if (aux_restore) {
		/* restore our state */
		tsk_rt(t)->hide_from_aux_tasks = aux_hide;
	}
#endif
}

int wake_up_for_lock(struct task_struct* t)
{
	int ret;
	
	ret = wake_up_process(t);
	
	return ret;
}


#else  // CONFIG_LITMUS_LOCKING

struct fdso_ops generic_lock_ops = {};

asmlinkage long sys_litmus_lock(int sem_od)
{
	return -ENOSYS;
}

asmlinkage long sys_litmus_unlock(int sem_od)
{
	return -ENOSYS;
}

#endif