1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18#include <linux/completion.h>
19#include <linux/delay.h>
20#include <linux/dma-mapping.h>
21#include <linux/dmaengine.h>
22#include <linux/errno.h>
23#include <linux/init.h>
24#include <linux/interrupt.h>
25#include <linux/list.h>
26#include <linux/mm.h>
27#include <linux/module.h>
28#include <linux/platform_device.h>
29#include <linux/sched.h>
30#include <linux/seq_file.h>
31#include <linux/slab.h>
32#include <linux/time.h>
33#include <linux/vmalloc.h>
34#include <linux/wait.h>
35
36#include "omap_dmm_tiler.h"
37#include "omap_dmm_priv.h"
38
39#define DMM_DRIVER_NAME "dmm"
40
41
42static struct tcm *containers[TILFMT_NFORMATS];
43static struct dmm *omap_dmm;
44
45#if defined(CONFIG_OF)
46static const struct of_device_id dmm_of_match[];
47#endif
48
49
50static DEFINE_SPINLOCK(list_lock);
51
52
53#define GEOM(xshift, yshift, bytes_per_pixel) { \
54 .x_shft = (xshift), \
55 .y_shft = (yshift), \
56 .cpp = (bytes_per_pixel), \
57 .slot_w = 1 << (SLOT_WIDTH_BITS - (xshift)), \
58 .slot_h = 1 << (SLOT_HEIGHT_BITS - (yshift)), \
59 }
60
61static const struct {
62 u32 x_shft;
63 u32 y_shft;
64 u32 cpp;
65 u32 slot_w;
66 u32 slot_h;
67} geom[TILFMT_NFORMATS] = {
68 [TILFMT_8BIT] = GEOM(0, 0, 1),
69 [TILFMT_16BIT] = GEOM(0, 1, 2),
70 [TILFMT_32BIT] = GEOM(1, 1, 4),
71 [TILFMT_PAGE] = GEOM(SLOT_WIDTH_BITS, SLOT_HEIGHT_BITS, 1),
72};
73
74
75
76static const u32 reg[][4] = {
77 [PAT_STATUS] = {DMM_PAT_STATUS__0, DMM_PAT_STATUS__1,
78 DMM_PAT_STATUS__2, DMM_PAT_STATUS__3},
79 [PAT_DESCR] = {DMM_PAT_DESCR__0, DMM_PAT_DESCR__1,
80 DMM_PAT_DESCR__2, DMM_PAT_DESCR__3},
81};
82
83static int dmm_dma_copy(struct dmm *dmm, dma_addr_t src, dma_addr_t dst)
84{
85 struct dma_async_tx_descriptor *tx;
86 enum dma_status status;
87 dma_cookie_t cookie;
88
89 tx = dmaengine_prep_dma_memcpy(dmm->wa_dma_chan, dst, src, 4, 0);
90 if (!tx) {
91 dev_err(dmm->dev, "Failed to prepare DMA memcpy\n");
92 return -EIO;
93 }
94
95 cookie = tx->tx_submit(tx);
96 if (dma_submit_error(cookie)) {
97 dev_err(dmm->dev, "Failed to do DMA tx_submit\n");
98 return -EIO;
99 }
100
101 status = dma_sync_wait(dmm->wa_dma_chan, cookie);
102 if (status != DMA_COMPLETE)
103 dev_err(dmm->dev, "i878 wa DMA copy failure\n");
104
105 dmaengine_terminate_all(dmm->wa_dma_chan);
106 return 0;
107}
108
109static u32 dmm_read_wa(struct dmm *dmm, u32 reg)
110{
111 dma_addr_t src, dst;
112 int r;
113
114 src = dmm->phys_base + reg;
115 dst = dmm->wa_dma_handle;
116
117 r = dmm_dma_copy(dmm, src, dst);
118 if (r) {
119 dev_err(dmm->dev, "sDMA read transfer timeout\n");
120 return readl(dmm->base + reg);
121 }
122
123
124
125
126
127
128 rmb();
129 return readl(dmm->wa_dma_data);
130}
131
132static void dmm_write_wa(struct dmm *dmm, u32 val, u32 reg)
133{
134 dma_addr_t src, dst;
135 int r;
136
137 writel(val, dmm->wa_dma_data);
138
139
140
141
142
143
144 wmb();
145
146 src = dmm->wa_dma_handle;
147 dst = dmm->phys_base + reg;
148
149 r = dmm_dma_copy(dmm, src, dst);
150 if (r) {
151 dev_err(dmm->dev, "sDMA write transfer timeout\n");
152 writel(val, dmm->base + reg);
153 }
154}
155
156static u32 dmm_read(struct dmm *dmm, u32 reg)
157{
158 if (dmm->dmm_workaround) {
159 u32 v;
160 unsigned long flags;
161
162 spin_lock_irqsave(&dmm->wa_lock, flags);
163 v = dmm_read_wa(dmm, reg);
164 spin_unlock_irqrestore(&dmm->wa_lock, flags);
165
166 return v;
167 } else {
168 return readl(dmm->base + reg);
169 }
170}
171
172static void dmm_write(struct dmm *dmm, u32 val, u32 reg)
173{
174 if (dmm->dmm_workaround) {
175 unsigned long flags;
176
177 spin_lock_irqsave(&dmm->wa_lock, flags);
178 dmm_write_wa(dmm, val, reg);
179 spin_unlock_irqrestore(&dmm->wa_lock, flags);
180 } else {
181 writel(val, dmm->base + reg);
182 }
183}
184
185static int dmm_workaround_init(struct dmm *dmm)
186{
187 dma_cap_mask_t mask;
188
189 spin_lock_init(&dmm->wa_lock);
190
191 dmm->wa_dma_data = dma_alloc_coherent(dmm->dev, sizeof(u32),
192 &dmm->wa_dma_handle, GFP_KERNEL);
193 if (!dmm->wa_dma_data)
194 return -ENOMEM;
195
196 dma_cap_zero(mask);
197 dma_cap_set(DMA_MEMCPY, mask);
198
199 dmm->wa_dma_chan = dma_request_channel(mask, NULL, NULL);
200 if (!dmm->wa_dma_chan) {
201 dma_free_coherent(dmm->dev, 4, dmm->wa_dma_data, dmm->wa_dma_handle);
202 return -ENODEV;
203 }
204
205 return 0;
206}
207
208static void dmm_workaround_uninit(struct dmm *dmm)
209{
210 dma_release_channel(dmm->wa_dma_chan);
211
212 dma_free_coherent(dmm->dev, 4, dmm->wa_dma_data, dmm->wa_dma_handle);
213}
214
215
216static void *alloc_dma(struct dmm_txn *txn, size_t sz, dma_addr_t *pa)
217{
218 void *ptr;
219 struct refill_engine *engine = txn->engine_handle;
220
221
222 txn->current_pa = round_up(txn->current_pa, 16);
223 txn->current_va = (void *)round_up((long)txn->current_va, 16);
224
225 ptr = txn->current_va;
226 *pa = txn->current_pa;
227
228 txn->current_pa += sz;
229 txn->current_va += sz;
230
231 BUG_ON((txn->current_va - engine->refill_va) > REFILL_BUFFER_SIZE);
232
233 return ptr;
234}
235
236
237static int wait_status(struct refill_engine *engine, u32 wait_mask)
238{
239 struct dmm *dmm = engine->dmm;
240 u32 r = 0, err, i;
241
242 i = DMM_FIXED_RETRY_COUNT;
243 while (true) {
244 r = dmm_read(dmm, reg[PAT_STATUS][engine->id]);
245 err = r & DMM_PATSTATUS_ERR;
246 if (err) {
247 dev_err(dmm->dev,
248 "%s: error (engine%d). PAT_STATUS: 0x%08x\n",
249 __func__, engine->id, r);
250 return -EFAULT;
251 }
252
253 if ((r & wait_mask) == wait_mask)
254 break;
255
256 if (--i == 0) {
257 dev_err(dmm->dev,
258 "%s: timeout (engine%d). PAT_STATUS: 0x%08x\n",
259 __func__, engine->id, r);
260 return -ETIMEDOUT;
261 }
262
263 udelay(1);
264 }
265
266 return 0;
267}
268
269static void release_engine(struct refill_engine *engine)
270{
271 unsigned long flags;
272
273 spin_lock_irqsave(&list_lock, flags);
274 list_add(&engine->idle_node, &omap_dmm->idle_head);
275 spin_unlock_irqrestore(&list_lock, flags);
276
277 atomic_inc(&omap_dmm->engine_counter);
278 wake_up_interruptible(&omap_dmm->engine_queue);
279}
280
281static irqreturn_t omap_dmm_irq_handler(int irq, void *arg)
282{
283 struct dmm *dmm = arg;
284 u32 status = dmm_read(dmm, DMM_PAT_IRQSTATUS);
285 int i;
286
287
288 dmm_write(dmm, status, DMM_PAT_IRQSTATUS);
289
290 for (i = 0; i < dmm->num_engines; i++) {
291 if (status & DMM_IRQSTAT_ERR_MASK)
292 dev_err(dmm->dev,
293 "irq error(engine%d): IRQSTAT 0x%02x\n",
294 i, status & 0xff);
295
296 if (status & DMM_IRQSTAT_LST) {
297 if (dmm->engines[i].async)
298 release_engine(&dmm->engines[i]);
299
300 complete(&dmm->engines[i].compl);
301 }
302
303 status >>= 8;
304 }
305
306 return IRQ_HANDLED;
307}
308
309
310
311
312static struct dmm_txn *dmm_txn_init(struct dmm *dmm, struct tcm *tcm)
313{
314 struct dmm_txn *txn = NULL;
315 struct refill_engine *engine = NULL;
316 int ret;
317 unsigned long flags;
318
319
320
321 ret = wait_event_interruptible(omap_dmm->engine_queue,
322 atomic_add_unless(&omap_dmm->engine_counter, -1, 0));
323 if (ret)
324 return ERR_PTR(ret);
325
326
327 spin_lock_irqsave(&list_lock, flags);
328 if (!list_empty(&dmm->idle_head)) {
329 engine = list_entry(dmm->idle_head.next, struct refill_engine,
330 idle_node);
331 list_del(&engine->idle_node);
332 }
333 spin_unlock_irqrestore(&list_lock, flags);
334
335 BUG_ON(!engine);
336
337 txn = &engine->txn;
338 engine->tcm = tcm;
339 txn->engine_handle = engine;
340 txn->last_pat = NULL;
341 txn->current_va = engine->refill_va;
342 txn->current_pa = engine->refill_pa;
343
344 return txn;
345}
346
347
348
349
350
351static void dmm_txn_append(struct dmm_txn *txn, struct pat_area *area,
352 struct page **pages, u32 npages, u32 roll)
353{
354 dma_addr_t pat_pa = 0, data_pa = 0;
355 u32 *data;
356 struct pat *pat;
357 struct refill_engine *engine = txn->engine_handle;
358 int columns = (1 + area->x1 - area->x0);
359 int rows = (1 + area->y1 - area->y0);
360 int i = columns*rows;
361
362 pat = alloc_dma(txn, sizeof(*pat), &pat_pa);
363
364 if (txn->last_pat)
365 txn->last_pat->next_pa = (u32)pat_pa;
366
367 pat->area = *area;
368
369
370 pat->area.y0 += engine->tcm->y_offset;
371 pat->area.y1 += engine->tcm->y_offset;
372
373 pat->ctrl = (struct pat_ctrl){
374 .start = 1,
375 .lut_id = engine->tcm->lut_id,
376 };
377
378 data = alloc_dma(txn, 4*i, &data_pa);
379
380 pat->data_pa = data_pa;
381
382 while (i--) {
383 int n = i + roll;
384 if (n >= npages)
385 n -= npages;
386 data[i] = (pages && pages[n]) ?
387 page_to_phys(pages[n]) : engine->dmm->dummy_pa;
388 }
389
390 txn->last_pat = pat;
391
392 return;
393}
394
395
396
397
398static int dmm_txn_commit(struct dmm_txn *txn, bool wait)
399{
400 int ret = 0;
401 struct refill_engine *engine = txn->engine_handle;
402 struct dmm *dmm = engine->dmm;
403
404 if (!txn->last_pat) {
405 dev_err(engine->dmm->dev, "need at least one txn\n");
406 ret = -EINVAL;
407 goto cleanup;
408 }
409
410 txn->last_pat->next_pa = 0;
411
412 wmb();
413
414
415
416
417
418
419
420
421 readl(&txn->last_pat->next_pa);
422
423
424 dmm_write(dmm, 0x0, reg[PAT_DESCR][engine->id]);
425
426
427 ret = wait_status(engine, DMM_PATSTATUS_READY);
428 if (ret) {
429 ret = -EFAULT;
430 goto cleanup;
431 }
432
433
434 engine->async = wait ? false : true;
435 reinit_completion(&engine->compl);
436
437 smp_mb();
438
439
440 dmm_write(dmm, engine->refill_pa, reg[PAT_DESCR][engine->id]);
441
442 if (wait) {
443 if (!wait_for_completion_timeout(&engine->compl,
444 msecs_to_jiffies(100))) {
445 dev_err(dmm->dev, "timed out waiting for done\n");
446 ret = -ETIMEDOUT;
447 goto cleanup;
448 }
449
450
451 ret = wait_status(engine, DMM_PATSTATUS_READY |
452 DMM_PATSTATUS_VALID | DMM_PATSTATUS_DONE);
453 }
454
455cleanup:
456
457 if (ret || wait)
458 release_engine(engine);
459
460 return ret;
461}
462
463
464
465
466static int fill(struct tcm_area *area, struct page **pages,
467 u32 npages, u32 roll, bool wait)
468{
469 int ret = 0;
470 struct tcm_area slice, area_s;
471 struct dmm_txn *txn;
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486 wait = true;
487
488 txn = dmm_txn_init(omap_dmm, area->tcm);
489 if (IS_ERR_OR_NULL(txn))
490 return -ENOMEM;
491
492 tcm_for_each_slice(slice, *area, area_s) {
493 struct pat_area p_area = {
494 .x0 = slice.p0.x, .y0 = slice.p0.y,
495 .x1 = slice.p1.x, .y1 = slice.p1.y,
496 };
497
498 dmm_txn_append(txn, &p_area, pages, npages, roll);
499
500 roll += tcm_sizeof(slice);
501 }
502
503 ret = dmm_txn_commit(txn, wait);
504
505 return ret;
506}
507
508
509
510
511
512
513
514int tiler_pin(struct tiler_block *block, struct page **pages,
515 u32 npages, u32 roll, bool wait)
516{
517 int ret;
518
519 ret = fill(&block->area, pages, npages, roll, wait);
520
521 if (ret)
522 tiler_unpin(block);
523
524 return ret;
525}
526
527int tiler_unpin(struct tiler_block *block)
528{
529 return fill(&block->area, NULL, 0, 0, false);
530}
531
532
533
534
535struct tiler_block *tiler_reserve_2d(enum tiler_fmt fmt, u16 w,
536 u16 h, u16 align)
537{
538 struct tiler_block *block;
539 u32 min_align = 128;
540 int ret;
541 unsigned long flags;
542 u32 slot_bytes;
543
544 block = kzalloc(sizeof(*block), GFP_KERNEL);
545 if (!block)
546 return ERR_PTR(-ENOMEM);
547
548 BUG_ON(!validfmt(fmt));
549
550
551 w = DIV_ROUND_UP(w, geom[fmt].slot_w);
552 h = DIV_ROUND_UP(h, geom[fmt].slot_h);
553
554
555 slot_bytes = geom[fmt].slot_w * geom[fmt].cpp;
556 min_align = max(min_align, slot_bytes);
557 align = (align > min_align) ? ALIGN(align, min_align) : min_align;
558 align /= slot_bytes;
559
560 block->fmt = fmt;
561
562 ret = tcm_reserve_2d(containers[fmt], w, h, align, -1, slot_bytes,
563 &block->area);
564 if (ret) {
565 kfree(block);
566 return ERR_PTR(-ENOMEM);
567 }
568
569
570 spin_lock_irqsave(&list_lock, flags);
571 list_add(&block->alloc_node, &omap_dmm->alloc_head);
572 spin_unlock_irqrestore(&list_lock, flags);
573
574 return block;
575}
576
577struct tiler_block *tiler_reserve_1d(size_t size)
578{
579 struct tiler_block *block = kzalloc(sizeof(*block), GFP_KERNEL);
580 int num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
581 unsigned long flags;
582
583 if (!block)
584 return ERR_PTR(-ENOMEM);
585
586 block->fmt = TILFMT_PAGE;
587
588 if (tcm_reserve_1d(containers[TILFMT_PAGE], num_pages,
589 &block->area)) {
590 kfree(block);
591 return ERR_PTR(-ENOMEM);
592 }
593
594 spin_lock_irqsave(&list_lock, flags);
595 list_add(&block->alloc_node, &omap_dmm->alloc_head);
596 spin_unlock_irqrestore(&list_lock, flags);
597
598 return block;
599}
600
601
602int tiler_release(struct tiler_block *block)
603{
604 int ret = tcm_free(&block->area);
605 unsigned long flags;
606
607 if (block->area.tcm)
608 dev_err(omap_dmm->dev, "failed to release block\n");
609
610 spin_lock_irqsave(&list_lock, flags);
611 list_del(&block->alloc_node);
612 spin_unlock_irqrestore(&list_lock, flags);
613
614 kfree(block);
615 return ret;
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
641static u32 tiler_get_address(enum tiler_fmt fmt, u32 orient, u32 x, u32 y)
642{
643 u32 x_bits, y_bits, tmp, x_mask, y_mask, alignment;
644
645 x_bits = CONT_WIDTH_BITS - geom[fmt].x_shft;
646 y_bits = CONT_HEIGHT_BITS - geom[fmt].y_shft;
647 alignment = geom[fmt].x_shft + geom[fmt].y_shft;
648
649
650 x_mask = MASK(x_bits);
651 y_mask = MASK(y_bits);
652
653 if (x < 0 || x > x_mask || y < 0 || y > y_mask) {
654 DBG("invalid coords: %u < 0 || %u > %u || %u < 0 || %u > %u",
655 x, x, x_mask, y, y, y_mask);
656 return 0;
657 }
658
659
660 if (orient & MASK_X_INVERT)
661 x ^= x_mask;
662 if (orient & MASK_Y_INVERT)
663 y ^= y_mask;
664
665
666 if (orient & MASK_XY_FLIP)
667 tmp = ((x << y_bits) + y);
668 else
669 tmp = ((y << x_bits) + x);
670
671 return TIL_ADDR((tmp << alignment), orient, fmt);
672}
673
674dma_addr_t tiler_ssptr(struct tiler_block *block)
675{
676 BUG_ON(!validfmt(block->fmt));
677
678 return TILVIEW_8BIT + tiler_get_address(block->fmt, 0,
679 block->area.p0.x * geom[block->fmt].slot_w,
680 block->area.p0.y * geom[block->fmt].slot_h);
681}
682
683dma_addr_t tiler_tsptr(struct tiler_block *block, u32 orient,
684 u32 x, u32 y)
685{
686 struct tcm_pt *p = &block->area.p0;
687 BUG_ON(!validfmt(block->fmt));
688
689 return tiler_get_address(block->fmt, orient,
690 (p->x * geom[block->fmt].slot_w) + x,
691 (p->y * geom[block->fmt].slot_h) + y);
692}
693
694void tiler_align(enum tiler_fmt fmt, u16 *w, u16 *h)
695{
696 BUG_ON(!validfmt(fmt));
697 *w = round_up(*w, geom[fmt].slot_w);
698 *h = round_up(*h, geom[fmt].slot_h);
699}
700
701u32 tiler_stride(enum tiler_fmt fmt, u32 orient)
702{
703 BUG_ON(!validfmt(fmt));
704
705 if (orient & MASK_XY_FLIP)
706 return 1 << (CONT_HEIGHT_BITS + geom[fmt].x_shft);
707 else
708 return 1 << (CONT_WIDTH_BITS + geom[fmt].y_shft);
709}
710
711size_t tiler_size(enum tiler_fmt fmt, u16 w, u16 h)
712{
713 tiler_align(fmt, &w, &h);
714 return geom[fmt].cpp * w * h;
715}
716
717size_t tiler_vsize(enum tiler_fmt fmt, u16 w, u16 h)
718{
719 BUG_ON(!validfmt(fmt));
720 return round_up(geom[fmt].cpp * w, PAGE_SIZE) * h;
721}
722
723u32 tiler_get_cpu_cache_flags(void)
724{
725 return omap_dmm->plat_data->cpu_cache_flags;
726}
727
728bool dmm_is_available(void)
729{
730 return omap_dmm ? true : false;
731}
732
733static int omap_dmm_remove(struct platform_device *dev)
734{
735 struct tiler_block *block, *_block;
736 int i;
737 unsigned long flags;
738
739 if (omap_dmm) {
740
741 dmm_write(omap_dmm, 0x7e7e7e7e, DMM_PAT_IRQENABLE_CLR);
742 free_irq(omap_dmm->irq, omap_dmm);
743
744
745 spin_lock_irqsave(&list_lock, flags);
746 list_for_each_entry_safe(block, _block, &omap_dmm->alloc_head,
747 alloc_node) {
748 list_del(&block->alloc_node);
749 kfree(block);
750 }
751 spin_unlock_irqrestore(&list_lock, flags);
752
753 for (i = 0; i < omap_dmm->num_lut; i++)
754 if (omap_dmm->tcm && omap_dmm->tcm[i])
755 omap_dmm->tcm[i]->deinit(omap_dmm->tcm[i]);
756 kfree(omap_dmm->tcm);
757
758 kfree(omap_dmm->engines);
759 if (omap_dmm->refill_va)
760 dma_free_wc(omap_dmm->dev,
761 REFILL_BUFFER_SIZE * omap_dmm->num_engines,
762 omap_dmm->refill_va, omap_dmm->refill_pa);
763 if (omap_dmm->dummy_page)
764 __free_page(omap_dmm->dummy_page);
765
766 if (omap_dmm->dmm_workaround)
767 dmm_workaround_uninit(omap_dmm);
768
769 iounmap(omap_dmm->base);
770 kfree(omap_dmm);
771 omap_dmm = NULL;
772 }
773
774 return 0;
775}
776
777static int omap_dmm_probe(struct platform_device *dev)
778{
779 int ret = -EFAULT, i;
780 struct tcm_area area = {0};
781 u32 hwinfo, pat_geom;
782 struct resource *mem;
783
784 omap_dmm = kzalloc(sizeof(*omap_dmm), GFP_KERNEL);
785 if (!omap_dmm)
786 goto fail;
787
788
789 INIT_LIST_HEAD(&omap_dmm->alloc_head);
790 INIT_LIST_HEAD(&omap_dmm->idle_head);
791
792 init_waitqueue_head(&omap_dmm->engine_queue);
793
794 if (dev->dev.of_node) {
795 const struct of_device_id *match;
796
797 match = of_match_node(dmm_of_match, dev->dev.of_node);
798 if (!match) {
799 dev_err(&dev->dev, "failed to find matching device node\n");
800 ret = -ENODEV;
801 goto fail;
802 }
803
804 omap_dmm->plat_data = match->data;
805 }
806
807
808 mem = platform_get_resource(dev, IORESOURCE_MEM, 0);
809 if (!mem) {
810 dev_err(&dev->dev, "failed to get base address resource\n");
811 goto fail;
812 }
813
814 omap_dmm->phys_base = mem->start;
815 omap_dmm->base = ioremap(mem->start, SZ_2K);
816
817 if (!omap_dmm->base) {
818 dev_err(&dev->dev, "failed to get dmm base address\n");
819 goto fail;
820 }
821
822 omap_dmm->irq = platform_get_irq(dev, 0);
823 if (omap_dmm->irq < 0) {
824 dev_err(&dev->dev, "failed to get IRQ resource\n");
825 goto fail;
826 }
827
828 omap_dmm->dev = &dev->dev;
829
830 if (of_machine_is_compatible("ti,dra7")) {
831
832
833
834
835
836 if (!dmm_workaround_init(omap_dmm)) {
837 omap_dmm->dmm_workaround = true;
838 dev_info(&dev->dev,
839 "workaround for errata i878 in use\n");
840 } else {
841 dev_warn(&dev->dev,
842 "failed to initialize work-around for i878\n");
843 }
844 }
845
846 hwinfo = dmm_read(omap_dmm, DMM_PAT_HWINFO);
847 omap_dmm->num_engines = (hwinfo >> 24) & 0x1F;
848 omap_dmm->num_lut = (hwinfo >> 16) & 0x1F;
849 omap_dmm->container_width = 256;
850 omap_dmm->container_height = 128;
851
852 atomic_set(&omap_dmm->engine_counter, omap_dmm->num_engines);
853
854
855 pat_geom = dmm_read(omap_dmm, DMM_PAT_GEOMETRY);
856 omap_dmm->lut_width = ((pat_geom >> 16) & 0xF) << 5;
857 omap_dmm->lut_height = ((pat_geom >> 24) & 0xF) << 5;
858
859
860
861 if (omap_dmm->lut_height != omap_dmm->container_height)
862 omap_dmm->num_lut++;
863
864
865 dmm_write(omap_dmm, 0x88888888, DMM_PAT_VIEW__0);
866 dmm_write(omap_dmm, 0x88888888, DMM_PAT_VIEW__1);
867 dmm_write(omap_dmm, 0x80808080, DMM_PAT_VIEW_MAP__0);
868 dmm_write(omap_dmm, 0x80000000, DMM_PAT_VIEW_MAP_BASE);
869 dmm_write(omap_dmm, 0x88888888, DMM_TILER_OR__0);
870 dmm_write(omap_dmm, 0x88888888, DMM_TILER_OR__1);
871
872 omap_dmm->dummy_page = alloc_page(GFP_KERNEL | __GFP_DMA32);
873 if (!omap_dmm->dummy_page) {
874 dev_err(&dev->dev, "could not allocate dummy page\n");
875 ret = -ENOMEM;
876 goto fail;
877 }
878
879
880 ret = dma_set_coherent_mask(&dev->dev, DMA_BIT_MASK(32));
881 if (ret)
882 goto fail;
883
884 omap_dmm->dummy_pa = page_to_phys(omap_dmm->dummy_page);
885
886
887 omap_dmm->refill_va = dma_alloc_wc(&dev->dev,
888 REFILL_BUFFER_SIZE * omap_dmm->num_engines,
889 &omap_dmm->refill_pa, GFP_KERNEL);
890 if (!omap_dmm->refill_va) {
891 dev_err(&dev->dev, "could not allocate refill memory\n");
892 goto fail;
893 }
894
895
896 omap_dmm->engines = kcalloc(omap_dmm->num_engines,
897 sizeof(*omap_dmm->engines), GFP_KERNEL);
898 if (!omap_dmm->engines) {
899 ret = -ENOMEM;
900 goto fail;
901 }
902
903 for (i = 0; i < omap_dmm->num_engines; i++) {
904 omap_dmm->engines[i].id = i;
905 omap_dmm->engines[i].dmm = omap_dmm;
906 omap_dmm->engines[i].refill_va = omap_dmm->refill_va +
907 (REFILL_BUFFER_SIZE * i);
908 omap_dmm->engines[i].refill_pa = omap_dmm->refill_pa +
909 (REFILL_BUFFER_SIZE * i);
910 init_completion(&omap_dmm->engines[i].compl);
911
912 list_add(&omap_dmm->engines[i].idle_node, &omap_dmm->idle_head);
913 }
914
915 omap_dmm->tcm = kcalloc(omap_dmm->num_lut, sizeof(*omap_dmm->tcm),
916 GFP_KERNEL);
917 if (!omap_dmm->tcm) {
918 ret = -ENOMEM;
919 goto fail;
920 }
921
922
923
924
925
926 for (i = 0; i < omap_dmm->num_lut; i++) {
927 omap_dmm->tcm[i] = sita_init(omap_dmm->container_width,
928 omap_dmm->container_height);
929
930 if (!omap_dmm->tcm[i]) {
931 dev_err(&dev->dev, "failed to allocate container\n");
932 ret = -ENOMEM;
933 goto fail;
934 }
935
936 omap_dmm->tcm[i]->lut_id = i;
937 }
938
939
940
941
942 containers[TILFMT_8BIT] = omap_dmm->tcm[0];
943 containers[TILFMT_16BIT] = omap_dmm->tcm[0];
944 containers[TILFMT_32BIT] = omap_dmm->tcm[0];
945
946 if (omap_dmm->container_height != omap_dmm->lut_height) {
947
948
949
950 containers[TILFMT_PAGE] = omap_dmm->tcm[1];
951 omap_dmm->tcm[1]->y_offset = OMAP5_LUT_OFFSET;
952 omap_dmm->tcm[1]->lut_id = 0;
953 } else {
954 containers[TILFMT_PAGE] = omap_dmm->tcm[0];
955 }
956
957 area = (struct tcm_area) {
958 .tcm = NULL,
959 .p1.x = omap_dmm->container_width - 1,
960 .p1.y = omap_dmm->container_height - 1,
961 };
962
963 ret = request_irq(omap_dmm->irq, omap_dmm_irq_handler, IRQF_SHARED,
964 "omap_dmm_irq_handler", omap_dmm);
965
966 if (ret) {
967 dev_err(&dev->dev, "couldn't register IRQ %d, error %d\n",
968 omap_dmm->irq, ret);
969 omap_dmm->irq = -1;
970 goto fail;
971 }
972
973
974
975
976
977
978
979 dmm_write(omap_dmm, 0x7e7e7e7e, DMM_PAT_IRQENABLE_SET);
980
981
982 for (i = 0; i < omap_dmm->num_lut; i++) {
983 area.tcm = omap_dmm->tcm[i];
984 if (fill(&area, NULL, 0, 0, true))
985 dev_err(omap_dmm->dev, "refill failed");
986 }
987
988 dev_info(omap_dmm->dev, "initialized all PAT entries\n");
989
990 return 0;
991
992fail:
993 if (omap_dmm_remove(dev))
994 dev_err(&dev->dev, "cleanup failed\n");
995 return ret;
996}
997
998
999
1000
1001
1002#ifdef CONFIG_DEBUG_FS
1003
1004static const char *alphabet = "abcdefghijklmnopqrstuvwxyz"
1005 "ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789";
1006static const char *special = ".,:;'\"`~!^-+";
1007
1008static void fill_map(char **map, int xdiv, int ydiv, struct tcm_area *a,
1009 char c, bool ovw)
1010{
1011 int x, y;
1012 for (y = a->p0.y / ydiv; y <= a->p1.y / ydiv; y++)
1013 for (x = a->p0.x / xdiv; x <= a->p1.x / xdiv; x++)
1014 if (map[y][x] == ' ' || ovw)
1015 map[y][x] = c;
1016}
1017
1018static void fill_map_pt(char **map, int xdiv, int ydiv, struct tcm_pt *p,
1019 char c)
1020{
1021 map[p->y / ydiv][p->x / xdiv] = c;
1022}
1023
1024static char read_map_pt(char **map, int xdiv, int ydiv, struct tcm_pt *p)
1025{
1026 return map[p->y / ydiv][p->x / xdiv];
1027}
1028
1029static int map_width(int xdiv, int x0, int x1)
1030{
1031 return (x1 / xdiv) - (x0 / xdiv) + 1;
1032}
1033
1034static void text_map(char **map, int xdiv, char *nice, int yd, int x0, int x1)
1035{
1036 char *p = map[yd] + (x0 / xdiv);
1037 int w = (map_width(xdiv, x0, x1) - strlen(nice)) / 2;
1038 if (w >= 0) {
1039 p += w;
1040 while (*nice)
1041 *p++ = *nice++;
1042 }
1043}
1044
1045static void map_1d_info(char **map, int xdiv, int ydiv, char *nice,
1046 struct tcm_area *a)
1047{
1048 sprintf(nice, "%dK", tcm_sizeof(*a) * 4);
1049 if (a->p0.y + 1 < a->p1.y) {
1050 text_map(map, xdiv, nice, (a->p0.y + a->p1.y) / 2 / ydiv, 0,
1051 256 - 1);
1052 } else if (a->p0.y < a->p1.y) {
1053 if (strlen(nice) < map_width(xdiv, a->p0.x, 256 - 1))
1054 text_map(map, xdiv, nice, a->p0.y / ydiv,
1055 a->p0.x + xdiv, 256 - 1);
1056 else if (strlen(nice) < map_width(xdiv, 0, a->p1.x))
1057 text_map(map, xdiv, nice, a->p1.y / ydiv,
1058 0, a->p1.y - xdiv);
1059 } else if (strlen(nice) + 1 < map_width(xdiv, a->p0.x, a->p1.x)) {
1060 text_map(map, xdiv, nice, a->p0.y / ydiv, a->p0.x, a->p1.x);
1061 }
1062}
1063
1064static void map_2d_info(char **map, int xdiv, int ydiv, char *nice,
1065 struct tcm_area *a)
1066{
1067 sprintf(nice, "(%d*%d)", tcm_awidth(*a), tcm_aheight(*a));
1068 if (strlen(nice) + 1 < map_width(xdiv, a->p0.x, a->p1.x))
1069 text_map(map, xdiv, nice, (a->p0.y + a->p1.y) / 2 / ydiv,
1070 a->p0.x, a->p1.x);
1071}
1072
1073int tiler_map_show(struct seq_file *s, void *arg)
1074{
1075 int xdiv = 2, ydiv = 1;
1076 char **map = NULL, *global_map;
1077 struct tiler_block *block;
1078 struct tcm_area a, p;
1079 int i;
1080 const char *m2d = alphabet;
1081 const char *a2d = special;
1082 const char *m2dp = m2d, *a2dp = a2d;
1083 char nice[128];
1084 int h_adj;
1085 int w_adj;
1086 unsigned long flags;
1087 int lut_idx;
1088
1089
1090 if (!omap_dmm) {
1091
1092 return 0;
1093 }
1094
1095 h_adj = omap_dmm->container_height / ydiv;
1096 w_adj = omap_dmm->container_width / xdiv;
1097
1098 map = kmalloc_array(h_adj, sizeof(*map), GFP_KERNEL);
1099 global_map = kmalloc_array(w_adj + 1, h_adj, GFP_KERNEL);
1100
1101 if (!map || !global_map)
1102 goto error;
1103
1104 for (lut_idx = 0; lut_idx < omap_dmm->num_lut; lut_idx++) {
1105 memset(map, 0, h_adj * sizeof(*map));
1106 memset(global_map, ' ', (w_adj + 1) * h_adj);
1107
1108 for (i = 0; i < omap_dmm->container_height; i++) {
1109 map[i] = global_map + i * (w_adj + 1);
1110 map[i][w_adj] = 0;
1111 }
1112
1113 spin_lock_irqsave(&list_lock, flags);
1114
1115 list_for_each_entry(block, &omap_dmm->alloc_head, alloc_node) {
1116 if (block->area.tcm == omap_dmm->tcm[lut_idx]) {
1117 if (block->fmt != TILFMT_PAGE) {
1118 fill_map(map, xdiv, ydiv, &block->area,
1119 *m2dp, true);
1120 if (!*++a2dp)
1121 a2dp = a2d;
1122 if (!*++m2dp)
1123 m2dp = m2d;
1124 map_2d_info(map, xdiv, ydiv, nice,
1125 &block->area);
1126 } else {
1127 bool start = read_map_pt(map, xdiv,
1128 ydiv, &block->area.p0) == ' ';
1129 bool end = read_map_pt(map, xdiv, ydiv,
1130 &block->area.p1) == ' ';
1131
1132 tcm_for_each_slice(a, block->area, p)
1133 fill_map(map, xdiv, ydiv, &a,
1134 '=', true);
1135 fill_map_pt(map, xdiv, ydiv,
1136 &block->area.p0,
1137 start ? '<' : 'X');
1138 fill_map_pt(map, xdiv, ydiv,
1139 &block->area.p1,
1140 end ? '>' : 'X');
1141 map_1d_info(map, xdiv, ydiv, nice,
1142 &block->area);
1143 }
1144 }
1145 }
1146
1147 spin_unlock_irqrestore(&list_lock, flags);
1148
1149 if (s) {
1150 seq_printf(s, "CONTAINER %d DUMP BEGIN\n", lut_idx);
1151 for (i = 0; i < 128; i++)
1152 seq_printf(s, "%03d:%s\n", i, map[i]);
1153 seq_printf(s, "CONTAINER %d DUMP END\n", lut_idx);
1154 } else {
1155 dev_dbg(omap_dmm->dev, "CONTAINER %d DUMP BEGIN\n",
1156 lut_idx);
1157 for (i = 0; i < 128; i++)
1158 dev_dbg(omap_dmm->dev, "%03d:%s\n", i, map[i]);
1159 dev_dbg(omap_dmm->dev, "CONTAINER %d DUMP END\n",
1160 lut_idx);
1161 }
1162 }
1163
1164error:
1165 kfree(map);
1166 kfree(global_map);
1167
1168 return 0;
1169}
1170#endif
1171
1172#ifdef CONFIG_PM_SLEEP
1173static int omap_dmm_resume(struct device *dev)
1174{
1175 struct tcm_area area;
1176 int i;
1177
1178 if (!omap_dmm)
1179 return -ENODEV;
1180
1181 area = (struct tcm_area) {
1182 .tcm = NULL,
1183 .p1.x = omap_dmm->container_width - 1,
1184 .p1.y = omap_dmm->container_height - 1,
1185 };
1186
1187
1188 for (i = 0; i < omap_dmm->num_lut; i++) {
1189 area.tcm = omap_dmm->tcm[i];
1190 if (fill(&area, NULL, 0, 0, true))
1191 dev_err(dev, "refill failed");
1192 }
1193
1194 return 0;
1195}
1196#endif
1197
1198static SIMPLE_DEV_PM_OPS(omap_dmm_pm_ops, NULL, omap_dmm_resume);
1199
1200#if defined(CONFIG_OF)
1201static const struct dmm_platform_data dmm_omap4_platform_data = {
1202 .cpu_cache_flags = OMAP_BO_WC,
1203};
1204
1205static const struct dmm_platform_data dmm_omap5_platform_data = {
1206 .cpu_cache_flags = OMAP_BO_UNCACHED,
1207};
1208
1209static const struct of_device_id dmm_of_match[] = {
1210 {
1211 .compatible = "ti,omap4-dmm",
1212 .data = &dmm_omap4_platform_data,
1213 },
1214 {
1215 .compatible = "ti,omap5-dmm",
1216 .data = &dmm_omap5_platform_data,
1217 },
1218 {},
1219};
1220#endif
1221
1222struct platform_driver omap_dmm_driver = {
1223 .probe = omap_dmm_probe,
1224 .remove = omap_dmm_remove,
1225 .driver = {
1226 .owner = THIS_MODULE,
1227 .name = DMM_DRIVER_NAME,
1228 .of_match_table = of_match_ptr(dmm_of_match),
1229 .pm = &omap_dmm_pm_ops,
1230 },
1231};
1232
1233MODULE_LICENSE("GPL v2");
1234MODULE_AUTHOR("Andy Gross <andy.gross@ti.com>");
1235MODULE_DESCRIPTION("OMAP DMM/Tiler Driver");
1236