aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/platforms/cell/spu_base.c
blob: fec51525252e2dfba33d8bb0c860bbbc0630d974 (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
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
/*
 * Low-level SPU handling
 *
 * (C) Copyright IBM Deutschland Entwicklung GmbH 2005
 *
 * Author: Arnd Bergmann <arndb@de.ibm.com>
 *
 * 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/interrupt.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/ptrace.h>
#include <linux/slab.h>
#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/io.h>
#include <linux/mutex.h>
#include <asm/spu.h>
#include <asm/spu_priv1.h>
#include <asm/xmon.h>

const struct spu_management_ops *spu_management_ops;
EXPORT_SYMBOL_GPL(spu_management_ops);

const struct spu_priv1_ops *spu_priv1_ops;

static struct list_head spu_list[MAX_NUMNODES];
static LIST_HEAD(spu_full_list);
static DEFINE_MUTEX(spu_mutex);
static DEFINE_SPINLOCK(spu_list_lock);

EXPORT_SYMBOL_GPL(spu_priv1_ops);

void spu_invalidate_slbs(struct spu *spu)
{
	struct spu_priv2 __iomem *priv2 = spu->priv2;

	if (spu_mfc_sr1_get(spu) & MFC_STATE1_RELOCATE_MASK)
		out_be64(&priv2->slb_invalidate_all_W, 0UL);
}
EXPORT_SYMBOL_GPL(spu_invalidate_slbs);

/* This is called by the MM core when a segment size is changed, to
 * request a flush of all the SPEs using a given mm
 */
void spu_flush_all_slbs(struct mm_struct *mm)
{
	struct spu *spu;
	unsigned long flags;

	spin_lock_irqsave(&spu_list_lock, flags);
	list_for_each_entry(spu, &spu_full_list, full_list) {
		if (spu->mm == mm)
			spu_invalidate_slbs(spu);
	}
	spin_unlock_irqrestore(&spu_list_lock, flags);
}

/* The hack below stinks... try to do something better one of
 * these days... Does it even work properly with NR_CPUS == 1 ?
 */
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);
}

void spu_associate_mm(struct spu *spu, struct mm_struct *mm)
{
	unsigned long flags;

	spin_lock_irqsave(&spu_list_lock, flags);
	spu->mm = mm;
	spin_unlock_irqrestore(&spu_list_lock, flags);
	if (mm)
		mm_needs_global_tlbie(mm);
}
EXPORT_SYMBOL_GPL(spu_associate_mm);

static int __spu_trap_invalid_dma(struct spu *spu)
{
	pr_debug("%s\n", __FUNCTION__);
	spu->dma_callback(spu, SPE_EVENT_INVALID_DMA);
	return 0;
}

static int __spu_trap_dma_align(struct spu *spu)
{
	pr_debug("%s\n", __FUNCTION__);
	spu->dma_callback(spu, SPE_EVENT_DMA_ALIGNMENT);
	return 0;
}

static int __spu_trap_error(struct spu *spu)
{
	pr_debug("%s\n", __FUNCTION__);
	spu->dma_callback(spu, SPE_EVENT_SPE_ERROR);
	return 0;
}

static void spu_restart_dma(struct spu *spu)
{
	struct spu_priv2 __iomem *priv2 = spu->priv2;

	if (!test_bit(SPU_CONTEXT_SWITCH_PENDING, &spu->flags))
		out_be64(&priv2->mfc_control_RW, MFC_CNTL_RESTART_DMA_COMMAND);
}

static int __spu_trap_data_seg(struct spu *spu, unsigned long ea)
{
	struct spu_priv2 __iomem *priv2 = spu->priv2;
	struct mm_struct *mm = spu->mm;
	u64 esid, vsid, llp;
	int psize;

	pr_debug("%s\n", __FUNCTION__);

	if (test_bit(SPU_CONTEXT_SWITCH_ACTIVE, &spu->flags)) {
		/* SLBs are pre-loaded for context switch, so
		 * we should never get here!
		 */
		printk("%s: invalid access during switch!\n", __func__);
		return 1;
	}
	esid = (ea & ESID_MASK) | SLB_ESID_V;

	switch(REGION_ID(ea)) {
	case USER_REGION_ID:
#ifdef CONFIG_HUGETLB_PAGE
		if (in_hugepage_area(mm->context, ea))
			psize = mmu_huge_psize;
		else
#endif
			psize = mm->context.user_psize;
		vsid = (get_vsid(mm->context.id, ea) << SLB_VSID_SHIFT) |
				SLB_VSID_USER;
		break;
	case VMALLOC_REGION_ID:
		if (ea < VMALLOC_END)
			psize = mmu_vmalloc_psize;
		else
			psize = mmu_io_psize;
		vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
			SLB_VSID_KERNEL;
		break;
	case KERNEL_REGION_ID:
		psize = mmu_linear_psize;
		vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
			SLB_VSID_KERNEL;
		break;
	default:
		/* Future: support kernel segments so that drivers
		 * can use SPUs.
		 */
		pr_debug("invalid region access at %016lx\n", ea);
		return 1;
	}
	llp = mmu_psize_defs[psize].sllp;

	out_be64(&priv2->slb_index_W, spu->slb_replace);
	out_be64(&priv2->slb_vsid_RW, vsid | llp);
	out_be64(&priv2->slb_esid_RW, esid);

	spu->slb_replace++;
	if (spu->slb_replace >= 8)
		spu->slb_replace = 0;

	spu_restart_dma(spu);

	return 0;
}

extern int hash_page(unsigned long ea, unsigned long access, unsigned long trap); //XXX
static int __spu_trap_data_map(struct spu *spu, unsigned long ea, u64 dsisr)
{
	pr_debug("%s, %lx, %lx\n", __FUNCTION__, dsisr, ea);

	/* Handle kernel space hash faults immediately.
	   User hash faults need to be deferred to process context. */
	if ((dsisr & MFC_DSISR_PTE_NOT_FOUND)
	    && REGION_ID(ea) != USER_REGION_ID
	    && hash_page(ea, _PAGE_PRESENT, 0x300) == 0) {
		spu_restart_dma(spu);
		return 0;
	}

	if (test_bit(SPU_CONTEXT_SWITCH_ACTIVE, &spu->flags)) {
		printk("%s: invalid access during switch!\n", __func__);
		return 1;
	}

	spu->dar = ea;
	spu->dsisr = dsisr;
	mb();
	spu->stop_callback(spu);
	return 0;
}

static irqreturn_t
spu_irq_class_0(int irq, void *data)
{
	struct spu *spu;

	spu = data;
	spu->class_0_pending = 1;
	spu->stop_callback(spu);

	return IRQ_HANDLED;
}

int
spu_irq_class_0_bottom(struct spu *spu)
{
	unsigned long stat, mask;
	unsigned long flags;

	spu->class_0_pending = 0;

	spin_lock_irqsave(&spu->register_lock, flags);
	mask = spu_int_mask_get(spu, 0);
	stat = spu_int_stat_get(spu, 0);

	stat &= mask;

	if (stat & 1) /* invalid DMA alignment */
		__spu_trap_dma_align(spu);

	if (stat & 2) /* invalid MFC DMA */
		__spu_trap_invalid_dma(spu);

	if (stat & 4) /* error on SPU */
		__spu_trap_error(spu);

	spu_int_stat_clear(spu, 0, stat);
	spin_unlock_irqrestore(&spu->register_lock, flags);

	return (stat & 0x7) ? -EIO : 0;
}
EXPORT_SYMBOL_GPL(spu_irq_class_0_bottom);

static irqreturn_t
spu_irq_class_1(int irq, void *data)
{
	struct spu *spu;
	unsigned long stat, mask, dar, dsisr;

	spu = data;

	/* atomically read & clear class1 status. */
	spin_lock(&spu->register_lock);
	mask  = spu_int_mask_get(spu, 1);
	stat  = spu_int_stat_get(spu, 1) & mask;
	dar   = spu_mfc_dar_get(spu);
	dsisr = spu_mfc_dsisr_get(spu);
	if (stat & 2) /* mapping fault */
		spu_mfc_dsisr_set(spu, 0ul);
	spu_int_stat_clear(spu, 1, stat);
	spin_unlock(&spu->register_lock);
	pr_debug("%s: %lx %lx %lx %lx\n", __FUNCTION__, mask, stat,
			dar, dsisr);

	if (stat & 1) /* segment fault */
		__spu_trap_data_seg(spu, dar);

	if (stat & 2) { /* mapping fault */
		__spu_trap_data_map(spu, dar, dsisr);
	}

	if (stat & 4) /* ls compare & suspend on get */
		;

	if (stat & 8) /* ls compare & suspend on put */
		;

	return stat ? IRQ_HANDLED : IRQ_NONE;
}

static irqreturn_t
spu_irq_class_2(int irq, void *data)
{
	struct spu *spu;
	unsigned long stat;
	unsigned long mask;

	spu = data;
	spin_lock(&spu->register_lock);
	stat = spu_int_stat_get(spu, 2);
	mask = spu_int_mask_get(spu, 2);
	/* ignore interrupts we're not waiting for */
	stat &= mask;
	/*
	 * mailbox interrupts (0x1 and 0x10) are level triggered.
	 * mask them now before acknowledging.
	 */
	if (stat & 0x11)
		spu_int_mask_and(spu, 2, ~(stat & 0x11));
	/* acknowledge all interrupts before the callbacks */
	spu_int_stat_clear(spu, 2, stat);
	spin_unlock(&spu->register_lock);

	pr_debug("class 2 interrupt %d, %lx, %lx\n", irq, stat, mask);

	if (stat & 1)  /* PPC core mailbox */
		spu->ibox_callback(spu);

	if (stat & 2) /* SPU stop-and-signal */
		spu->stop_callback(spu);

	if (stat & 4) /* SPU halted */
		spu->stop_callback(spu);

	if (stat & 8) /* DMA tag group complete */
		spu->mfc_callback(spu);

	if (stat & 0x10) /* SPU mailbox threshold */
		spu->wbox_callback(spu);

	return stat ? IRQ_HANDLED : IRQ_NONE;
}

static int spu_request_irqs(struct spu *spu)
{
	int ret = 0;

	if (spu->irqs[0] != NO_IRQ) {
		snprintf(spu->irq_c0, sizeof (spu->irq_c0), "spe%02d.0",
			 spu->number);
		ret = request_irq(spu->irqs[0], spu_irq_class_0,
				  IRQF_DISABLED,
				  spu->irq_c0, spu);
		if (ret)
			goto bail0;
	}
	if (spu->irqs[1] != NO_IRQ) {
		snprintf(spu->irq_c1, sizeof (spu->irq_c1), "spe%02d.1",
			 spu->number);
		ret = request_irq(spu->irqs[1], spu_irq_class_1,
				  IRQF_DISABLED,
				  spu->irq_c1, spu);
		if (ret)
			goto bail1;
	}
	if (spu->irqs[2] != NO_IRQ) {
		snprintf(spu->irq_c2, sizeof (spu->irq_c2), "spe%02d.2",
			 spu->number);
		ret = request_irq(spu->irqs[2], spu_irq_class_2,
				  IRQF_DISABLED,
				  spu->irq_c2, spu);
		if (ret)
			goto bail2;
	}
	return 0;

bail2:
	if (spu->irqs[1] != NO_IRQ)
		free_irq(spu->irqs[1], spu);
bail1:
	if (spu->irqs[0] != NO_IRQ)
		free_irq(spu->irqs[0], spu);
bail0:
	return ret;
}

static void spu_free_irqs(struct spu *spu)
{
	if (spu->irqs[0] != NO_IRQ)
		free_irq(spu->irqs[0], spu);
	if (spu->irqs[1] != NO_IRQ)
		free_irq(spu->irqs[1], spu);
	if (spu->irqs[2] != NO_IRQ)
		free_irq(spu->irqs[2], spu);
}

static void spu_init_channels(struct spu *spu)
{
	static const struct {
		 unsigned channel;
		 unsigned count;
	} zero_list[] = {
		{ 0x00, 1, }, { 0x01, 1, }, { 0x03, 1, }, { 0x04, 1, },
		{ 0x18, 1, }, { 0x19, 1, }, { 0x1b, 1, }, { 0x1d, 1, },
	}, count_list[] = {
		{ 0x00, 0, }, { 0x03, 0, }, { 0x04, 0, }, { 0x15, 16, },
		{ 0x17, 1, }, { 0x18, 0, }, { 0x19, 0, }, { 0x1b, 0, },
		{ 0x1c, 1, }, { 0x1d, 0, }, { 0x1e, 1, },
	};
	struct spu_priv2 __iomem *priv2;
	int i;

	priv2 = spu->priv2;

	/* initialize all channel data to zero */
	for (i = 0; i < ARRAY_SIZE(zero_list); i++) {
		int count;

		out_be64(&priv2->spu_chnlcntptr_RW, zero_list[i].channel);
		for (count = 0; count < zero_list[i].count; count++)
			out_be64(&priv2->spu_chnldata_RW, 0);
	}

	/* initialize channel counts to meaningful values */
	for (i = 0; i < ARRAY_SIZE(count_list); i++) {
		out_be64(&priv2->spu_chnlcntptr_RW, count_list[i].channel);
		out_be64(&priv2->spu_chnlcnt_RW, count_list[i].count);
	}
}

struct spu *spu_alloc_node(int node)
{
	struct spu *spu = NULL;

	mutex_lock(&spu_mutex);
	if (!list_empty(&spu_list[node])) {
		spu = list_entry(spu_list[node].next, struct spu, list);
		list_del_init(&spu->list);
		pr_debug("Got SPU %d %d\n", spu->number, spu->node);
	}
	mutex_unlock(&spu_mutex);

	if (spu)
		spu_init_channels(spu);
	return spu;
}
EXPORT_SYMBOL_GPL(spu_alloc_node);

struct spu *spu_alloc(void)
{
	struct spu *spu = NULL;
	int node;

	for (node = 0; node < MAX_NUMNODES; node++) {
		spu = spu_alloc_node(node);
		if (spu)
			break;
	}

	return spu;
}

void spu_free(struct spu *spu)
{
	mutex_lock(&spu_mutex);
	list_add_tail(&spu->list, &spu_list[spu->node]);
	mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_free);

struct sysdev_class spu_sysdev_class = {
	set_kset_name("spu")
};

int spu_add_sysdev_attr(struct sysdev_attribute *attr)
{
	struct spu *spu;
	mutex_lock(&spu_mutex);

	list_for_each_entry(spu, &spu_full_list, full_list)
		sysdev_create_file(&spu->sysdev, attr);

	mutex_unlock(&spu_mutex);
	return 0;
}
EXPORT_SYMBOL_GPL(spu_add_sysdev_attr);

int spu_add_sysdev_attr_group(struct attribute_group *attrs)
{
	struct spu *spu;
	mutex_lock(&spu_mutex);

	list_for_each_entry(spu, &spu_full_list, full_list)
		sysfs_create_group(&spu->sysdev.kobj, attrs);

	mutex_unlock(&spu_mutex);
	return 0;
}
EXPORT_SYMBOL_GPL(spu_add_sysdev_attr_group);


void spu_remove_sysdev_attr(struct sysdev_attribute *attr)
{
	struct spu *spu;
	mutex_lock(&spu_mutex);

	list_for_each_entry(spu, &spu_full_list, full_list)
		sysdev_remove_file(&spu->sysdev, attr);

	mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_remove_sysdev_attr);

void spu_remove_sysdev_attr_group(struct attribute_group *attrs)
{
	struct spu *spu;
	mutex_lock(&spu_mutex);

	list_for_each_entry(spu, &spu_full_list, full_list)
		sysfs_remove_group(&spu->sysdev.kobj, attrs);

	mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_remove_sysdev_attr_group);

static int spu_create_sysdev(struct spu *spu)
{
	int ret;

	spu->sysdev.id = spu->number;
	spu->sysdev.cls = &spu_sysdev_class;
	ret = sysdev_register(&spu->sysdev);
	if (ret) {
		printk(KERN_ERR "Can't register SPU %d with sysfs\n",
				spu->number);
		return ret;
	}

	sysfs_add_device_to_node(&spu->sysdev, spu->node);

	return 0;
}

static int __init create_spu(void *data)
{
	struct spu *spu;
	int ret;
	static int number;
	unsigned long flags;

	ret = -ENOMEM;
	spu = kzalloc(sizeof (*spu), GFP_KERNEL);
	if (!spu)
		goto out;

	spin_lock_init(&spu->register_lock);
	mutex_lock(&spu_mutex);
	spu->number = number++;
	mutex_unlock(&spu_mutex);

	ret = spu_create_spu(spu, data);

	if (ret)
		goto out_free;

	spu_mfc_sdr_setup(spu);
	spu_mfc_sr1_set(spu, 0x33);
	ret = spu_request_irqs(spu);
	if (ret)
		goto out_destroy;

	ret = spu_create_sysdev(spu);
	if (ret)
		goto out_free_irqs;

	mutex_lock(&spu_mutex);
	spin_lock_irqsave(&spu_list_lock, flags);
	list_add(&spu->list, &spu_list[spu->node]);
	list_add(&spu->full_list, &spu_full_list);
	spin_unlock_irqrestore(&spu_list_lock, flags);
	mutex_unlock(&spu_mutex);

	goto out;

out_free_irqs:
	spu_free_irqs(spu);
out_destroy:
	spu_destroy_spu(spu);
out_free:
	kfree(spu);
out:
	return ret;
}

static int __init init_spu_base(void)
{
	int i, ret = 0;

	for (i = 0; i < MAX_NUMNODES; i++)
		INIT_LIST_HEAD(&spu_list[i]);

	if (!spu_management_ops)
		goto out;

	/* create sysdev class for spus */
	ret = sysdev_class_register(&spu_sysdev_class);
	if (ret)
		goto out;

	ret = spu_enumerate_spus(create_spu);

	if (ret) {
		printk(KERN_WARNING "%s: Error initializing spus\n",
			__FUNCTION__);
		goto out_unregister_sysdev_class;
	}

	xmon_register_spus(&spu_full_list);

	return 0;

 out_unregister_sysdev_class:
	sysdev_class_unregister(&spu_sysdev_class);
 out:

	return ret;
}
module_init(init_spu_base);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Arnd Bergmann <arndb@de.ibm.com>");