aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/kernel/ds.c
blob: d6938d9351cfefa4afaaf5b78eb3b584111709e5 (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
/*
 * Debug Store support
 *
 * This provides a low-level interface to the hardware's Debug Store
 * feature that is used for branch trace store (BTS) and
 * precise-event based sampling (PEBS).
 *
 * It manages:
 * - per-thread and per-cpu allocation of BTS and PEBS
 * - buffer memory allocation (optional)
 * - buffer overflow handling
 * - buffer access
 *
 * It assumes:
 * - get_task_struct on all parameter tasks
 * - current is allowed to trace parameter tasks
 *
 *
 * Copyright (C) 2007-2008 Intel Corporation.
 * Markus Metzger <markus.t.metzger@intel.com>, 2007-2008
 */


#include <asm/ds.h>

#include <linux/errno.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/mm.h>


/*
 * The configuration for a particular DS hardware implementation.
 */
struct ds_configuration {
	/* the size of the DS structure in bytes */
	unsigned char  sizeof_ds;
	/* the size of one pointer-typed field in the DS structure in bytes;
	   this covers the first 8 fields related to buffer management. */
	unsigned char  sizeof_field;
	/* the size of a BTS/PEBS record in bytes */
	unsigned char  sizeof_rec[2];
};
static struct ds_configuration ds_cfg;


/*
 * Debug Store (DS) save area configuration (see Intel64 and IA32
 * Architectures Software Developer's Manual, section 18.5)
 *
 * The DS configuration consists of the following fields; different
 * architetures vary in the size of those fields.
 * - double-word aligned base linear address of the BTS buffer
 * - write pointer into the BTS buffer
 * - end linear address of the BTS buffer (one byte beyond the end of
 *   the buffer)
 * - interrupt pointer into BTS buffer
 *   (interrupt occurs when write pointer passes interrupt pointer)
 * - double-word aligned base linear address of the PEBS buffer
 * - write pointer into the PEBS buffer
 * - end linear address of the PEBS buffer (one byte beyond the end of
 *   the buffer)
 * - interrupt pointer into PEBS buffer
 *   (interrupt occurs when write pointer passes interrupt pointer)
 * - value to which counter is reset following counter overflow
 *
 * Later architectures use 64bit pointers throughout, whereas earlier
 * architectures use 32bit pointers in 32bit mode.
 *
 *
 * We compute the base address for the first 8 fields based on:
 * - the field size stored in the DS configuration
 * - the relative field position
 * - an offset giving the start of the respective region
 *
 * This offset is further used to index various arrays holding
 * information for BTS and PEBS at the respective index.
 *
 * On later 32bit processors, we only access the lower 32bit of the
 * 64bit pointer fields. The upper halves will be zeroed out.
 */

enum ds_field {
	ds_buffer_base = 0,
	ds_index,
	ds_absolute_maximum,
	ds_interrupt_threshold,
};

enum ds_qualifier {
	ds_bts  = 0,
	ds_pebs
};

static inline unsigned long ds_get(const unsigned char *base,
				   enum ds_qualifier qual, enum ds_field field)
{
	base += (ds_cfg.sizeof_field * (field + (4 * qual)));
	return *(unsigned long *)base;
}

static inline void ds_set(unsigned char *base, enum ds_qualifier qual,
			  enum ds_field field, unsigned long value)
{
	base += (ds_cfg.sizeof_field * (field + (4 * qual)));
	(*(unsigned long *)base) = value;
}


/*
 * Locking is done only for allocating BTS or PEBS resources and for
 * guarding context and buffer memory allocation.
 *
 * Most functions require the current task to own the ds context part
 * they are going to access. All the locking is done when validating
 * access to the context.
 */
static spinlock_t ds_lock = __SPIN_LOCK_UNLOCKED(ds_lock);

/*
 * Validate that the current task is allowed to access the BTS/PEBS
 * buffer of the parameter task.
 *
 * Returns 0, if access is granted; -Eerrno, otherwise.
 */
static inline int ds_validate_access(struct ds_context *context,
				     enum ds_qualifier qual)
{
	if (!context)
		return -EPERM;

	if (context->owner[qual] == current)
		return 0;

	return -EPERM;
}


/*
 * We either support (system-wide) per-cpu or per-thread allocation.
 * We distinguish the two based on the task_struct pointer, where a
 * NULL pointer indicates per-cpu allocation for the current cpu.
 *
 * Allocations are use-counted. As soon as resources are allocated,
 * further allocations must be of the same type (per-cpu or
 * per-thread). We model this by counting allocations (i.e. the number
 * of tracers of a certain type) for one type negatively:
 *   =0  no tracers
 *   >0  number of per-thread tracers
 *   <0  number of per-cpu tracers
 *
 * The below functions to get and put tracers and to check the
 * allocation type require the ds_lock to be held by the caller.
 *
 * Tracers essentially gives the number of ds contexts for a certain
 * type of allocation.
 */
static long tracers;

static inline void get_tracer(struct task_struct *task)
{
	tracers += (task ? 1 : -1);
}

static inline void put_tracer(struct task_struct *task)
{
	tracers -= (task ? 1 : -1);
}

static inline int check_tracer(struct task_struct *task)
{
	return (task ? (tracers >= 0) : (tracers <= 0));
}


/*
 * The DS context is either attached to a thread or to a cpu:
 * - in the former case, the thread_struct contains a pointer to the
 *   attached context.
 * - in the latter case, we use a static array of per-cpu context
 *   pointers.
 *
 * Contexts are use-counted. They are allocated on first access and
 * deallocated when the last user puts the context.
 *
 * We distinguish between an allocating and a non-allocating get of a
 * context:
 * - the allocating get is used for requesting BTS/PEBS resources. It
 *   requires the caller to hold the global ds_lock.
 * - the non-allocating get is used for all other cases. A
 *   non-existing context indicates an error. It acquires and releases
 *   the ds_lock itself for obtaining the context.
 *
 * A context and its DS configuration are allocated and deallocated
 * together. A context always has a DS configuration of the
 * appropriate size.
 */
static DEFINE_PER_CPU(struct ds_context *, system_context);

#define this_system_context per_cpu(system_context, smp_processor_id())

/*
 * Returns the pointer to the parameter task's context or to the
 * system-wide context, if task is NULL.
 *
 * Increases the use count of the returned context, if not NULL.
 */
static inline struct ds_context *ds_get_context(struct task_struct *task)
{
	struct ds_context *context;
	unsigned long irq;

	spin_lock_irqsave(&ds_lock, irq);

	context = (task ? task->thread.ds_ctx : this_system_context);
	if (context)
		context->count++;

	spin_unlock_irqrestore(&ds_lock, irq);

	return context;
}

/*
 * Same as ds_get_context, but allocates the context and it's DS
 * structure, if necessary; returns NULL; if out of memory.
 */
static inline struct ds_context *ds_alloc_context(struct task_struct *task)
{
	struct ds_context **p_context =
		(task ? &task->thread.ds_ctx : &this_system_context);
	struct ds_context *context = *p_context;
	unsigned long irq;

	if (!context) {
		context = kzalloc(sizeof(*context), GFP_KERNEL);
		if (!context)
			return NULL;

		context->ds = kzalloc(ds_cfg.sizeof_ds, GFP_KERNEL);
		if (!context->ds) {
			kfree(context);
			return NULL;
		}

		spin_lock_irqsave(&ds_lock, irq);

		if (*p_context) {
			kfree(context->ds);
			kfree(context);

			context = *p_context;
		} else {
			*p_context = context;

			context->this = p_context;
			context->task = task;

			if (task)
				set_tsk_thread_flag(task, TIF_DS_AREA_MSR);

			if (!task || (task == current))
				wrmsrl(MSR_IA32_DS_AREA,
				       (unsigned long)context->ds);
		}
		spin_unlock_irqrestore(&ds_lock, irq);
	}

	context->count++;

	return context;
}

/*
 * Decreases the use count of the parameter context, if not NULL.
 * Deallocates the context, if the use count reaches zero.
 */
static inline void ds_put_context(struct ds_context *context)
{
	unsigned long irq;

	if (!context)
		return;

	spin_lock_irqsave(&ds_lock, irq);

	if (--context->count)
		goto out;

	*(context->this) = NULL;

	if (context->task)
		clear_tsk_thread_flag(context->task, TIF_DS_AREA_MSR);

	if (!context->task || (context->task == current))
		wrmsrl(MSR_IA32_DS_AREA, 0);

	put_tracer(context->task);

	/* free any leftover buffers from tracers that did not
	 * deallocate them properly. */
	kfree(context->buffer[ds_bts]);
	kfree(context->buffer[ds_pebs]);
	kfree(context->ds);
	kfree(context);
 out:
	spin_unlock_irqrestore(&ds_lock, irq);
}


/*
 * Handle a buffer overflow
 *
 * task: the task whose buffers are overflowing;
 *       NULL for a buffer overflow on the current cpu
 * context: the ds context
 * qual: the buffer type
 */
static void ds_overflow(struct task_struct *task, struct ds_context *context,
			enum ds_qualifier qual)
{
	if (!context)
		return;

	if (context->callback[qual])
		(*context->callback[qual])(task);

	/* todo: do some more overflow handling */
}


/*
 * Allocate a non-pageable buffer of the parameter size.
 * Checks the memory and the locked memory rlimit.
 *
 * Returns the buffer, if successful;
 *         NULL, if out of memory or rlimit exceeded.
 *
 * size: the requested buffer size in bytes
 * pages (out): if not NULL, contains the number of pages reserved
 */
static inline void *ds_allocate_buffer(size_t size, unsigned int *pages)
{
	unsigned long rlim, vm, pgsz;
	void *buffer;

	pgsz = PAGE_ALIGN(size) >> PAGE_SHIFT;

	rlim = current->signal->rlim[RLIMIT_AS].rlim_cur >> PAGE_SHIFT;
	vm   = current->mm->total_vm  + pgsz;
	if (rlim < vm)
		return NULL;

	rlim = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >> PAGE_SHIFT;
	vm   = current->mm->locked_vm  + pgsz;
	if (rlim < vm)
		return NULL;

	buffer = kzalloc(size, GFP_KERNEL);
	if (!buffer)
		return NULL;

	current->mm->total_vm  += pgsz;
	current->mm->locked_vm += pgsz;

	if (pages)
		*pages = pgsz;

	return buffer;
}

static int ds_request(struct task_struct *task, void *base, size_t size,
		      ds_ovfl_callback_t ovfl, enum ds_qualifier qual)
{
	struct ds_context *context;
	unsigned long buffer, adj;
	const unsigned long alignment = (1 << 3);
	unsigned long irq;
	int error = 0;

	if (!ds_cfg.sizeof_ds)
		return -EOPNOTSUPP;

	/* we require some space to do alignment adjustments below */
	if (size < (alignment + ds_cfg.sizeof_rec[qual]))
		return -EINVAL;

	/* buffer overflow notification is not yet implemented */
	if (ovfl)
		return -EOPNOTSUPP;


	context = ds_alloc_context(task);
	if (!context)
		return -ENOMEM;

	spin_lock_irqsave(&ds_lock, irq);

	error = -EPERM;
	if (!check_tracer(task))
		goto out_unlock;

	get_tracer(task);

	error = -EALREADY;
	if (context->owner[qual] == current)
		goto out_put_tracer;
	error = -EPERM;
	if (context->owner[qual] != NULL)
		goto out_put_tracer;
	context->owner[qual] = current;

	spin_unlock_irqrestore(&ds_lock, irq);


	error = -ENOMEM;
	if (!base) {
		base = ds_allocate_buffer(size, &context->pages[qual]);
		if (!base)
			goto out_release;

		context->buffer[qual]   = base;
	}
	error = 0;

	context->callback[qual] = ovfl;

	/* adjust the buffer address and size to meet alignment
	 * constraints:
	 * - buffer is double-word aligned
	 * - size is multiple of record size
	 *
	 * We checked the size at the very beginning; we have enough
	 * space to do the adjustment.
	 */
	buffer = (unsigned long)base;

	adj = ALIGN(buffer, alignment) - buffer;
	buffer += adj;
	size   -= adj;

	size /= ds_cfg.sizeof_rec[qual];
	size *= ds_cfg.sizeof_rec[qual];

	ds_set(context->ds, qual, ds_buffer_base, buffer);
	ds_set(context->ds, qual, ds_index, buffer);
	ds_set(context->ds, qual, ds_absolute_maximum, buffer + size);

	if (ovfl) {
		/* todo: select a suitable interrupt threshold */
	} else
		ds_set(context->ds, qual,
		       ds_interrupt_threshold, buffer + size + 1);

	/* we keep the context until ds_release */
	return error;

 out_release:
	context->owner[qual] = NULL;
	ds_put_context(context);
	put_tracer(task);
	return error;

 out_put_tracer:
	spin_unlock_irqrestore(&ds_lock, irq);
	ds_put_context(context);
	put_tracer(task);
	return error;

 out_unlock:
	spin_unlock_irqrestore(&ds_lock, irq);
	ds_put_context(context);
	return error;
}

int ds_request_bts(struct task_struct *task, void *base, size_t size,
		   ds_ovfl_callback_t ovfl)
{
	return ds_request(task, base, size, ovfl, ds_bts);
}

int ds_request_pebs(struct task_struct *task, void *base, size_t size,
		    ds_ovfl_callback_t ovfl)
{
	return ds_request(task, base, size, ovfl, ds_pebs);
}

static int ds_release(struct task_struct *task, enum ds_qualifier qual)
{
	struct ds_context *context;
	int error;

	context = ds_get_context(task);
	error = ds_validate_access(context, qual);
	if (error < 0)
		goto out;

	kfree(context->buffer[qual]);
	context->buffer[qual] = NULL;

	current->mm->total_vm  -= context->pages[qual];
	current->mm->locked_vm -= context->pages[qual];
	context->pages[qual] = 0;
	context->owner[qual] = NULL;

	/*
	 * we put the context twice:
	 *   once for the ds_get_context
	 *   once for the corresponding ds_request
	 */
	ds_put_context(context);
 out:
	ds_put_context(context);
	return error;
}

int ds_release_bts(struct task_struct *task)
{
	return ds_release(task, ds_bts);
}

int ds_release_pebs(struct task_struct *task)
{
	return ds_release(task, ds_pebs);
}

static int ds_get_index(struct task_struct *task, size_t *pos,
			enum ds_qualifier qual)
{
	struct ds_context *context;
	unsigned long base, index;
	int error;

	context = ds_get_context(task);
	error = ds_validate_access(context, qual);
	if (error < 0)
		goto out;

	base  = ds_get(context->ds, qual, ds_buffer_base);
	index = ds_get(context->ds, qual, ds_index);

	error = ((index - base) / ds_cfg.sizeof_rec[qual]);
	if (pos)
		*pos = error;
 out:
	ds_put_context(context);
	return error;
}

int ds_get_bts_index(struct task_struct *task, size_t *pos)
{
	return ds_get_index(task, pos, ds_bts);
}

int ds_get_pebs_index(struct task_struct *task, size_t *pos)
{
	return ds_get_index(task, pos, ds_pebs);
}

static int ds_get_end(struct task_struct *task, size_t *pos,
		      enum ds_qualifier qual)
{
	struct ds_context *context;
	unsigned long base, end;
	int error;

	context = ds_get_context(task);
	error = ds_validate_access(context, qual);
	if (error < 0)
		goto out;

	base = ds_get(context->ds, qual, ds_buffer_base);
	end  = ds_get(context->ds, qual, ds_absolute_maximum);

	error = ((end - base) / ds_cfg.sizeof_rec[qual]);
	if (pos)
		*pos = error;
 out:
	ds_put_context(context);
	return error;
}

int ds_get_bts_end(struct task_struct *task, size_t *pos)
{
	return ds_get_end(task, pos, ds_bts);
}

int ds_get_pebs_end(struct task_struct *task, size_t *pos)
{
	return ds_get_end(task, pos, ds_pebs);
}

static int ds_access(struct task_struct *task, size_t index,
		     const void **record, enum ds_qualifier qual)
{
	struct ds_context *context;
	unsigned long base, idx;
	int error;

	if (!record)
		return -EINVAL;

	context = ds_get_context(task);
	error = ds_validate_access(context, qual);
	if (error < 0)
		goto out;

	base = ds_get(context->ds, qual, ds_buffer_base);
	idx = base + (index * ds_cfg.sizeof_rec[qual]);

	error = -EINVAL;
	if (idx > ds_get(context->ds, qual, ds_absolute_maximum))
		goto out;

	*record = (const void *)idx;
	error = ds_cfg.sizeof_rec[qual];
 out:
	ds_put_context(context);
	return error;
}

int ds_access_bts(struct task_struct *task, size_t index, const void **record)
{
	return ds_access(task, index, record, ds_bts);
}

int ds_access_pebs(struct task_struct *task, size_t index, const void **record)
{
	return ds_access(task, index, record, ds_pebs);
}

static int ds_write(struct task_struct *task, const void *record, size_t size,
		    enum ds_qualifier qual, int force)
{
	struct ds_context *context;
	int error;

	if (!record)
		return -EINVAL;

	error = -EPERM;
	context = ds_get_context(task);
	if (!context)
		goto out;

	if (!force) {
		error = ds_validate_access(context, qual);
		if (error < 0)
			goto out;
	}

	error = 0;
	while (size) {
		unsigned long base, index, end, write_end, int_th;
		unsigned long write_size, adj_write_size;

		/*
		 * write as much as possible without producing an
		 * overflow interrupt.
		 *
		 * interrupt_threshold must either be
		 * - bigger than absolute_maximum or
		 * - point to a record between buffer_base and absolute_maximum
		 *
		 * index points to a valid record.
		 */
		base   = ds_get(context->ds, qual, ds_buffer_base);
		index  = ds_get(context->ds, qual, ds_index);
		end    = ds_get(context->ds, qual, ds_absolute_maximum);
		int_th = ds_get(context->ds, qual, ds_interrupt_threshold);

		write_end = min(end, int_th);

		/* if we are already beyond the interrupt threshold,
		 * we fill the entire buffer */
		if (write_end <= index)
			write_end = end;

		if (write_end <= index)
			goto out;

		write_size = min((unsigned long) size, write_end - index);
		memcpy((void *)index, record, write_size);

		record = (const char *)record + write_size;
		size  -= write_size;
		error += write_size;

		adj_write_size = write_size / ds_cfg.sizeof_rec[qual];
		adj_write_size *= ds_cfg.sizeof_rec[qual];

		/* zero out trailing bytes */
		memset((char *)index + write_size, 0,
		       adj_write_size - write_size);
		index += adj_write_size;

		if (index >= end)
			index = base;
		ds_set(context->ds, qual, ds_index, index);

		if (index >= int_th)
			ds_overflow(task, context, qual);
	}

 out:
	ds_put_context(context);
	return error;
}

int ds_write_bts(struct task_struct *task, const void *record, size_t size)
{
	return ds_write(task, record, size, ds_bts, /* force = */ 0);
}

int ds_write_pebs(struct task_struct *task, const void *record, size_t size)
{
	return ds_write(task, record, size, ds_pebs, /* force = */ 0);
}

int ds_unchecked_write_bts(struct task_struct *task,
			   const void *record, size_t size)
{
	return ds_write(task, record, size, ds_bts, /* force = */ 1);
}

int ds_unchecked_write_pebs(struct task_struct *task,
			    const void *record, size_t size)
{
	return ds_write(task, record, size, ds_pebs, /* force = */ 1);
}

static int ds_reset_or_clear(struct task_struct *task,
			     enum ds_qualifier qual, int clear)
{
	struct ds_context *context;
	unsigned long base, end;
	int error;

	context = ds_get_context(task);
	error = ds_validate_access(context, qual);
	if (error < 0)
		goto out;

	base = ds_get(context->ds, qual, ds_buffer_base);
	end  = ds_get(context->ds, qual, ds_absolute_maximum);

	if (clear)
		memset((void *)base, 0, end - base);

	ds_set(context->ds, qual, ds_index, base);

	error = 0;
 out:
	ds_put_context(context);
	return error;
}

int ds_reset_bts(struct task_struct *task)
{
	return ds_reset_or_clear(task, ds_bts, /* clear = */ 0);
}

int ds_reset_pebs(struct task_struct *task)
{
	return ds_reset_or_clear(task, ds_pebs, /* clear = */ 0);
}

int ds_clear_bts(struct task_struct *task)
{
	return ds_reset_or_clear(task, ds_bts, /* clear = */ 1);
}

int ds_clear_pebs(struct task_struct *task)
{
	return ds_reset_or_clear(task, ds_pebs, /* clear = */ 1);
}

int ds_get_pebs_reset(struct task_struct *task, u64 *value)
{
	struct ds_context *context;
	int error;

	if (!value)
		return -EINVAL;

	context = ds_get_context(task);
	error = ds_validate_access(context, ds_pebs);
	if (error < 0)
		goto out;

	*value = *(u64 *)(context->ds + (ds_cfg.sizeof_field * 8));

	error = 0;
 out:
	ds_put_context(context);
	return error;
}

int ds_set_pebs_reset(struct task_struct *task, u64 value)
{
	struct ds_context *context;
	int error;

	context = ds_get_context(task);
	error = ds_validate_access(context, ds_pebs);
	if (error < 0)
		goto out;

	*(u64 *)(context->ds + (ds_cfg.sizeof_field * 8)) = value;

	error = 0;
 out:
	ds_put_context(context);
	return error;
}

static const struct ds_configuration ds_cfg_var = {
	.sizeof_ds    = sizeof(long) * 12,
	.sizeof_field = sizeof(long),
	.sizeof_rec[ds_bts]   = sizeof(long) * 3,
#ifdef __i386__
	.sizeof_rec[ds_pebs]  = sizeof(long) * 10
#else
	.sizeof_rec[ds_pebs]  = sizeof(long) * 18
#endif
};
static const struct ds_configuration ds_cfg_64 = {
	.sizeof_ds    = 8 * 12,
	.sizeof_field = 8,
	.sizeof_rec[ds_bts]   = 8 * 3,
#ifdef __i386__
	.sizeof_rec[ds_pebs]  = 8 * 10
#else
	.sizeof_rec[ds_pebs]  = 8 * 18
#endif
};

static inline void
ds_configure(const struct ds_configuration *cfg)
{
	ds_cfg = *cfg;
}

void __cpuinit ds_init_intel(struct cpuinfo_x86 *c)
{
	switch (c->x86) {
	case 0x6:
		switch (c->x86_model) {
		case 0 ... 0xC:
			/* sorry, don't know about them */
			break;
		case 0xD:
		case 0xE: /* Pentium M */
			ds_configure(&ds_cfg_var);
			break;
		default: /* Core2, Atom, ... */
			ds_configure(&ds_cfg_64);
			break;
		}
		break;
	case 0xF:
		switch (c->x86_model) {
		case 0x0:
		case 0x1:
		case 0x2: /* Netburst */
			ds_configure(&ds_cfg_var);
			break;
		default:
			/* sorry, don't know about them */
			break;
		}
		break;
	default:
		/* sorry, don't know about them */
		break;
	}
}

void ds_free(struct ds_context *context)
{
	/* This is called when the task owning the parameter context
	 * is dying. There should not be any user of that context left
	 * to disturb us, anymore. */
	unsigned long leftovers = context->count;
	while (leftovers--)
		ds_put_context(context);
}