This source file includes following definitions.
- tegra_vde_writel
- tegra_vde_readl
- tegra_vde_set_bits
- tegra_vde_wait_mbe
- tegra_vde_setup_mbe_frame_idx
- tegra_vde_mbe_set_0xa_reg
- tegra_vde_wait_bsev
- tegra_vde_push_to_bsev_icmdqueue
- tegra_vde_setup_frameid
- tegra_setup_frameidx
- tegra_vde_setup_iram_entry
- tegra_vde_setup_iram_tables
- tegra_vde_setup_hw_context
- tegra_vde_decode_frame
- tegra_vde_attach_dmabuf
- tegra_vde_attach_dmabufs_to_frame
- tegra_vde_release_frame_dmabufs
- tegra_vde_validate_frame
- tegra_vde_validate_h264_ctx
- tegra_vde_ioctl_decode_h264
- tegra_vde_unlocked_ioctl
- tegra_vde_release_file
- tegra_vde_isr
- tegra_vde_runtime_suspend
- tegra_vde_runtime_resume
- tegra_vde_probe
- tegra_vde_remove
- tegra_vde_pm_suspend
- tegra_vde_pm_resume
1
2
3
4
5
6
7
8
9 #include <linux/clk.h>
10 #include <linux/dma-buf.h>
11 #include <linux/genalloc.h>
12 #include <linux/interrupt.h>
13 #include <linux/iopoll.h>
14 #include <linux/list.h>
15 #include <linux/miscdevice.h>
16 #include <linux/module.h>
17 #include <linux/of_device.h>
18 #include <linux/pm_runtime.h>
19 #include <linux/reset.h>
20 #include <linux/slab.h>
21 #include <linux/uaccess.h>
22
23 #include <soc/tegra/pmc.h>
24
25 #include "uapi.h"
26 #include "vde.h"
27
28 #define CREATE_TRACE_POINTS
29 #include "trace.h"
30
31 #define ICMDQUE_WR 0x00
32 #define CMDQUE_CONTROL 0x08
33 #define INTR_STATUS 0x18
34 #define BSE_INT_ENB 0x40
35 #define BSE_CONFIG 0x44
36
37 #define BSE_ICMDQUE_EMPTY BIT(3)
38 #define BSE_DMA_BUSY BIT(23)
39
40 struct video_frame {
41 struct dma_buf_attachment *y_dmabuf_attachment;
42 struct dma_buf_attachment *cb_dmabuf_attachment;
43 struct dma_buf_attachment *cr_dmabuf_attachment;
44 struct dma_buf_attachment *aux_dmabuf_attachment;
45 dma_addr_t y_addr;
46 dma_addr_t cb_addr;
47 dma_addr_t cr_addr;
48 dma_addr_t aux_addr;
49 u32 frame_num;
50 u32 flags;
51 };
52
53 static void tegra_vde_writel(struct tegra_vde *vde,
54 u32 value, void __iomem *base, u32 offset)
55 {
56 trace_vde_writel(vde, base, offset, value);
57
58 writel_relaxed(value, base + offset);
59 }
60
61 static u32 tegra_vde_readl(struct tegra_vde *vde,
62 void __iomem *base, u32 offset)
63 {
64 u32 value = readl_relaxed(base + offset);
65
66 trace_vde_readl(vde, base, offset, value);
67
68 return value;
69 }
70
71 static void tegra_vde_set_bits(struct tegra_vde *vde,
72 u32 mask, void __iomem *base, u32 offset)
73 {
74 u32 value = tegra_vde_readl(vde, base, offset);
75
76 tegra_vde_writel(vde, value | mask, base, offset);
77 }
78
79 static int tegra_vde_wait_mbe(struct tegra_vde *vde)
80 {
81 u32 tmp;
82
83 return readl_relaxed_poll_timeout(vde->mbe + 0x8C, tmp,
84 (tmp >= 0x10), 1, 100);
85 }
86
87 static int tegra_vde_setup_mbe_frame_idx(struct tegra_vde *vde,
88 unsigned int refs_nb,
89 bool setup_refs)
90 {
91 u32 frame_idx_enb_mask = 0;
92 u32 value;
93 unsigned int frame_idx;
94 unsigned int idx;
95 int err;
96
97 tegra_vde_writel(vde, 0xD0000000 | (0 << 23), vde->mbe, 0x80);
98 tegra_vde_writel(vde, 0xD0200000 | (0 << 23), vde->mbe, 0x80);
99
100 err = tegra_vde_wait_mbe(vde);
101 if (err)
102 return err;
103
104 if (!setup_refs)
105 return 0;
106
107 for (idx = 0, frame_idx = 1; idx < refs_nb; idx++, frame_idx++) {
108 tegra_vde_writel(vde, 0xD0000000 | (frame_idx << 23),
109 vde->mbe, 0x80);
110 tegra_vde_writel(vde, 0xD0200000 | (frame_idx << 23),
111 vde->mbe, 0x80);
112
113 frame_idx_enb_mask |= frame_idx << (6 * (idx % 4));
114
115 if (idx % 4 == 3 || idx == refs_nb - 1) {
116 value = 0xC0000000;
117 value |= (idx >> 2) << 24;
118 value |= frame_idx_enb_mask;
119
120 tegra_vde_writel(vde, value, vde->mbe, 0x80);
121
122 err = tegra_vde_wait_mbe(vde);
123 if (err)
124 return err;
125
126 frame_idx_enb_mask = 0;
127 }
128 }
129
130 return 0;
131 }
132
133 static void tegra_vde_mbe_set_0xa_reg(struct tegra_vde *vde, int reg, u32 val)
134 {
135 tegra_vde_writel(vde, 0xA0000000 | (reg << 24) | (val & 0xFFFF),
136 vde->mbe, 0x80);
137 tegra_vde_writel(vde, 0xA0000000 | ((reg + 1) << 24) | (val >> 16),
138 vde->mbe, 0x80);
139 }
140
141 static int tegra_vde_wait_bsev(struct tegra_vde *vde, bool wait_dma)
142 {
143 struct device *dev = vde->miscdev.parent;
144 u32 value;
145 int err;
146
147 err = readl_relaxed_poll_timeout(vde->bsev + INTR_STATUS, value,
148 !(value & BIT(2)), 1, 100);
149 if (err) {
150 dev_err(dev, "BSEV unknown bit timeout\n");
151 return err;
152 }
153
154 err = readl_relaxed_poll_timeout(vde->bsev + INTR_STATUS, value,
155 (value & BSE_ICMDQUE_EMPTY), 1, 100);
156 if (err) {
157 dev_err(dev, "BSEV ICMDQUE flush timeout\n");
158 return err;
159 }
160
161 if (!wait_dma)
162 return 0;
163
164 err = readl_relaxed_poll_timeout(vde->bsev + INTR_STATUS, value,
165 !(value & BSE_DMA_BUSY), 1, 100);
166 if (err) {
167 dev_err(dev, "BSEV DMA timeout\n");
168 return err;
169 }
170
171 return 0;
172 }
173
174 static int tegra_vde_push_to_bsev_icmdqueue(struct tegra_vde *vde,
175 u32 value, bool wait_dma)
176 {
177 tegra_vde_writel(vde, value, vde->bsev, ICMDQUE_WR);
178
179 return tegra_vde_wait_bsev(vde, wait_dma);
180 }
181
182 static void tegra_vde_setup_frameid(struct tegra_vde *vde,
183 struct video_frame *frame,
184 unsigned int frameid,
185 u32 mbs_width, u32 mbs_height)
186 {
187 u32 y_addr = frame ? frame->y_addr : 0x6CDEAD00;
188 u32 cb_addr = frame ? frame->cb_addr : 0x6CDEAD00;
189 u32 cr_addr = frame ? frame->cr_addr : 0x6CDEAD00;
190 u32 value1 = frame ? ((mbs_width << 16) | mbs_height) : 0;
191 u32 value2 = frame ? ((((mbs_width + 1) >> 1) << 6) | 1) : 0;
192
193 tegra_vde_writel(vde, y_addr >> 8, vde->frameid, 0x000 + frameid * 4);
194 tegra_vde_writel(vde, cb_addr >> 8, vde->frameid, 0x100 + frameid * 4);
195 tegra_vde_writel(vde, cr_addr >> 8, vde->frameid, 0x180 + frameid * 4);
196 tegra_vde_writel(vde, value1, vde->frameid, 0x080 + frameid * 4);
197 tegra_vde_writel(vde, value2, vde->frameid, 0x280 + frameid * 4);
198 }
199
200 static void tegra_setup_frameidx(struct tegra_vde *vde,
201 struct video_frame *frames,
202 unsigned int frames_nb,
203 u32 mbs_width, u32 mbs_height)
204 {
205 unsigned int idx;
206
207 for (idx = 0; idx < frames_nb; idx++)
208 tegra_vde_setup_frameid(vde, &frames[idx], idx,
209 mbs_width, mbs_height);
210
211 for (; idx < 17; idx++)
212 tegra_vde_setup_frameid(vde, NULL, idx, 0, 0);
213 }
214
215 static void tegra_vde_setup_iram_entry(struct tegra_vde *vde,
216 unsigned int table,
217 unsigned int row,
218 u32 value1, u32 value2)
219 {
220 u32 *iram_tables = vde->iram;
221
222 trace_vde_setup_iram_entry(table, row, value1, value2);
223
224 iram_tables[0x20 * table + row * 2] = value1;
225 iram_tables[0x20 * table + row * 2 + 1] = value2;
226 }
227
228 static void tegra_vde_setup_iram_tables(struct tegra_vde *vde,
229 struct video_frame *dpb_frames,
230 unsigned int ref_frames_nb,
231 unsigned int with_earlier_poc_nb)
232 {
233 struct video_frame *frame;
234 u32 value, aux_addr;
235 int with_later_poc_nb;
236 unsigned int i, k;
237
238 trace_vde_ref_l0(dpb_frames[0].frame_num);
239
240 for (i = 0; i < 16; i++) {
241 if (i < ref_frames_nb) {
242 frame = &dpb_frames[i + 1];
243
244 aux_addr = frame->aux_addr;
245
246 value = (i + 1) << 26;
247 value |= !(frame->flags & FLAG_B_FRAME) << 25;
248 value |= 1 << 24;
249 value |= frame->frame_num;
250 } else {
251 aux_addr = 0x6ADEAD00;
252 value = 0;
253 }
254
255 tegra_vde_setup_iram_entry(vde, 0, i, value, aux_addr);
256 tegra_vde_setup_iram_entry(vde, 1, i, value, aux_addr);
257 tegra_vde_setup_iram_entry(vde, 2, i, value, aux_addr);
258 tegra_vde_setup_iram_entry(vde, 3, i, value, aux_addr);
259 }
260
261 if (!(dpb_frames[0].flags & FLAG_B_FRAME))
262 return;
263
264 if (with_earlier_poc_nb >= ref_frames_nb)
265 return;
266
267 with_later_poc_nb = ref_frames_nb - with_earlier_poc_nb;
268
269 trace_vde_ref_l1(with_later_poc_nb, with_earlier_poc_nb);
270
271 for (i = 0, k = with_earlier_poc_nb; i < with_later_poc_nb; i++, k++) {
272 frame = &dpb_frames[k + 1];
273
274 aux_addr = frame->aux_addr;
275
276 value = (k + 1) << 26;
277 value |= !(frame->flags & FLAG_B_FRAME) << 25;
278 value |= 1 << 24;
279 value |= frame->frame_num;
280
281 tegra_vde_setup_iram_entry(vde, 2, i, value, aux_addr);
282 }
283
284 for (k = 0; i < ref_frames_nb; i++, k++) {
285 frame = &dpb_frames[k + 1];
286
287 aux_addr = frame->aux_addr;
288
289 value = (k + 1) << 26;
290 value |= !(frame->flags & FLAG_B_FRAME) << 25;
291 value |= 1 << 24;
292 value |= frame->frame_num;
293
294 tegra_vde_setup_iram_entry(vde, 2, i, value, aux_addr);
295 }
296 }
297
298 static int tegra_vde_setup_hw_context(struct tegra_vde *vde,
299 struct tegra_vde_h264_decoder_ctx *ctx,
300 struct video_frame *dpb_frames,
301 dma_addr_t bitstream_data_addr,
302 size_t bitstream_data_size,
303 unsigned int macroblocks_nb)
304 {
305 struct device *dev = vde->miscdev.parent;
306 u32 value;
307 int err;
308
309 tegra_vde_set_bits(vde, 0x000A, vde->sxe, 0xF0);
310 tegra_vde_set_bits(vde, 0x000B, vde->bsev, CMDQUE_CONTROL);
311 tegra_vde_set_bits(vde, 0x8002, vde->mbe, 0x50);
312 tegra_vde_set_bits(vde, 0x000A, vde->mbe, 0xA0);
313 tegra_vde_set_bits(vde, 0x000A, vde->ppe, 0x14);
314 tegra_vde_set_bits(vde, 0x000A, vde->ppe, 0x28);
315 tegra_vde_set_bits(vde, 0x0A00, vde->mce, 0x08);
316 tegra_vde_set_bits(vde, 0x000A, vde->tfe, 0x00);
317 tegra_vde_set_bits(vde, 0x0005, vde->vdma, 0x04);
318
319 tegra_vde_writel(vde, 0x00000000, vde->vdma, 0x1C);
320 tegra_vde_writel(vde, 0x00000000, vde->vdma, 0x00);
321 tegra_vde_writel(vde, 0x00000007, vde->vdma, 0x04);
322 tegra_vde_writel(vde, 0x00000007, vde->frameid, 0x200);
323 tegra_vde_writel(vde, 0x00000005, vde->tfe, 0x04);
324 tegra_vde_writel(vde, 0x00000000, vde->mbe, 0x84);
325 tegra_vde_writel(vde, 0x00000010, vde->sxe, 0x08);
326 tegra_vde_writel(vde, 0x00000150, vde->sxe, 0x54);
327 tegra_vde_writel(vde, 0x0000054C, vde->sxe, 0x58);
328 tegra_vde_writel(vde, 0x00000E34, vde->sxe, 0x5C);
329 tegra_vde_writel(vde, 0x063C063C, vde->mce, 0x10);
330 tegra_vde_writel(vde, 0x0003FC00, vde->bsev, INTR_STATUS);
331 tegra_vde_writel(vde, 0x0000150D, vde->bsev, BSE_CONFIG);
332 tegra_vde_writel(vde, 0x00000100, vde->bsev, BSE_INT_ENB);
333 tegra_vde_writel(vde, 0x00000000, vde->bsev, 0x98);
334 tegra_vde_writel(vde, 0x00000060, vde->bsev, 0x9C);
335
336 memset(vde->iram + 128, 0, macroblocks_nb / 2);
337
338 tegra_setup_frameidx(vde, dpb_frames, ctx->dpb_frames_nb,
339 ctx->pic_width_in_mbs, ctx->pic_height_in_mbs);
340
341 tegra_vde_setup_iram_tables(vde, dpb_frames,
342 ctx->dpb_frames_nb - 1,
343 ctx->dpb_ref_frames_with_earlier_poc_nb);
344
345
346
347
348
349 wmb();
350
351 tegra_vde_writel(vde, 0x00000000, vde->bsev, 0x8C);
352 tegra_vde_writel(vde, bitstream_data_addr + bitstream_data_size,
353 vde->bsev, 0x54);
354
355 value = ctx->pic_width_in_mbs << 11 | ctx->pic_height_in_mbs << 3;
356
357 tegra_vde_writel(vde, value, vde->bsev, 0x88);
358
359 err = tegra_vde_wait_bsev(vde, false);
360 if (err)
361 return err;
362
363 err = tegra_vde_push_to_bsev_icmdqueue(vde, 0x800003FC, false);
364 if (err)
365 return err;
366
367 value = 0x01500000;
368 value |= ((vde->iram_lists_addr + 512) >> 2) & 0xFFFF;
369
370 err = tegra_vde_push_to_bsev_icmdqueue(vde, value, true);
371 if (err)
372 return err;
373
374 err = tegra_vde_push_to_bsev_icmdqueue(vde, 0x840F054C, false);
375 if (err)
376 return err;
377
378 err = tegra_vde_push_to_bsev_icmdqueue(vde, 0x80000080, false);
379 if (err)
380 return err;
381
382 value = 0x0E340000 | ((vde->iram_lists_addr >> 2) & 0xFFFF);
383
384 err = tegra_vde_push_to_bsev_icmdqueue(vde, value, true);
385 if (err)
386 return err;
387
388 value = 0x00800005;
389 value |= ctx->pic_width_in_mbs << 11;
390 value |= ctx->pic_height_in_mbs << 3;
391
392 tegra_vde_writel(vde, value, vde->sxe, 0x10);
393
394 value = !ctx->baseline_profile << 17;
395 value |= ctx->level_idc << 13;
396 value |= ctx->log2_max_pic_order_cnt_lsb << 7;
397 value |= ctx->pic_order_cnt_type << 5;
398 value |= ctx->log2_max_frame_num;
399
400 tegra_vde_writel(vde, value, vde->sxe, 0x40);
401
402 value = ctx->pic_init_qp << 25;
403 value |= !!(ctx->deblocking_filter_control_present_flag) << 2;
404 value |= !!ctx->pic_order_present_flag;
405
406 tegra_vde_writel(vde, value, vde->sxe, 0x44);
407
408 value = ctx->chroma_qp_index_offset;
409 value |= ctx->num_ref_idx_l0_active_minus1 << 5;
410 value |= ctx->num_ref_idx_l1_active_minus1 << 10;
411 value |= !!ctx->constrained_intra_pred_flag << 15;
412
413 tegra_vde_writel(vde, value, vde->sxe, 0x48);
414
415 value = 0x0C000000;
416 value |= !!(dpb_frames[0].flags & FLAG_B_FRAME) << 24;
417
418 tegra_vde_writel(vde, value, vde->sxe, 0x4C);
419
420 value = 0x03800000;
421 value |= bitstream_data_size & GENMASK(19, 15);
422
423 tegra_vde_writel(vde, value, vde->sxe, 0x68);
424
425 tegra_vde_writel(vde, bitstream_data_addr, vde->sxe, 0x6C);
426
427 value = 0x10000005;
428 value |= ctx->pic_width_in_mbs << 11;
429 value |= ctx->pic_height_in_mbs << 3;
430
431 tegra_vde_writel(vde, value, vde->mbe, 0x80);
432
433 value = 0x26800000;
434 value |= ctx->level_idc << 4;
435 value |= !ctx->baseline_profile << 1;
436 value |= !!ctx->direct_8x8_inference_flag;
437
438 tegra_vde_writel(vde, value, vde->mbe, 0x80);
439
440 tegra_vde_writel(vde, 0xF4000001, vde->mbe, 0x80);
441 tegra_vde_writel(vde, 0x20000000, vde->mbe, 0x80);
442 tegra_vde_writel(vde, 0xF4000101, vde->mbe, 0x80);
443
444 value = 0x20000000;
445 value |= ctx->chroma_qp_index_offset << 8;
446
447 tegra_vde_writel(vde, value, vde->mbe, 0x80);
448
449 err = tegra_vde_setup_mbe_frame_idx(vde,
450 ctx->dpb_frames_nb - 1,
451 ctx->pic_order_cnt_type == 0);
452 if (err) {
453 dev_err(dev, "MBE frames setup failed %d\n", err);
454 return err;
455 }
456
457 tegra_vde_mbe_set_0xa_reg(vde, 0, 0x000009FC);
458 tegra_vde_mbe_set_0xa_reg(vde, 2, 0x61DEAD00);
459 tegra_vde_mbe_set_0xa_reg(vde, 4, 0x62DEAD00);
460 tegra_vde_mbe_set_0xa_reg(vde, 6, 0x63DEAD00);
461 tegra_vde_mbe_set_0xa_reg(vde, 8, dpb_frames[0].aux_addr);
462
463 value = 0xFC000000;
464 value |= !!(dpb_frames[0].flags & FLAG_B_FRAME) << 2;
465
466 if (!ctx->baseline_profile)
467 value |= !!(dpb_frames[0].flags & FLAG_REFERENCE) << 1;
468
469 tegra_vde_writel(vde, value, vde->mbe, 0x80);
470
471 err = tegra_vde_wait_mbe(vde);
472 if (err) {
473 dev_err(dev, "MBE programming failed %d\n", err);
474 return err;
475 }
476
477 return 0;
478 }
479
480 static void tegra_vde_decode_frame(struct tegra_vde *vde,
481 unsigned int macroblocks_nb)
482 {
483 reinit_completion(&vde->decode_completion);
484
485 tegra_vde_writel(vde, 0x00000001, vde->bsev, 0x8C);
486 tegra_vde_writel(vde, 0x20000000 | (macroblocks_nb - 1),
487 vde->sxe, 0x00);
488 }
489
490 static int tegra_vde_attach_dmabuf(struct tegra_vde *vde,
491 int fd,
492 unsigned long offset,
493 size_t min_size,
494 size_t align_size,
495 struct dma_buf_attachment **a,
496 dma_addr_t *addrp,
497 size_t *size,
498 enum dma_data_direction dma_dir)
499 {
500 struct device *dev = vde->miscdev.parent;
501 struct dma_buf *dmabuf;
502 int err;
503
504 dmabuf = dma_buf_get(fd);
505 if (IS_ERR(dmabuf)) {
506 dev_err(dev, "Invalid dmabuf FD\n");
507 return PTR_ERR(dmabuf);
508 }
509
510 if (dmabuf->size & (align_size - 1)) {
511 dev_err(dev, "Unaligned dmabuf 0x%zX, should be aligned to 0x%zX\n",
512 dmabuf->size, align_size);
513 return -EINVAL;
514 }
515
516 if ((u64)offset + min_size > dmabuf->size) {
517 dev_err(dev, "Too small dmabuf size %zu @0x%lX, should be at least %zu\n",
518 dmabuf->size, offset, min_size);
519 return -EINVAL;
520 }
521
522 err = tegra_vde_dmabuf_cache_map(vde, dmabuf, dma_dir, a, addrp);
523 if (err)
524 goto err_put;
525
526 *addrp = *addrp + offset;
527
528 if (size)
529 *size = dmabuf->size - offset;
530
531 return 0;
532
533 err_put:
534 dma_buf_put(dmabuf);
535
536 return err;
537 }
538
539 static int tegra_vde_attach_dmabufs_to_frame(struct tegra_vde *vde,
540 struct video_frame *frame,
541 struct tegra_vde_h264_frame *src,
542 enum dma_data_direction dma_dir,
543 bool baseline_profile,
544 size_t lsize, size_t csize)
545 {
546 int err;
547
548 err = tegra_vde_attach_dmabuf(vde, src->y_fd,
549 src->y_offset, lsize, SZ_256,
550 &frame->y_dmabuf_attachment,
551 &frame->y_addr,
552 NULL, dma_dir);
553 if (err)
554 return err;
555
556 err = tegra_vde_attach_dmabuf(vde, src->cb_fd,
557 src->cb_offset, csize, SZ_256,
558 &frame->cb_dmabuf_attachment,
559 &frame->cb_addr,
560 NULL, dma_dir);
561 if (err)
562 goto err_release_y;
563
564 err = tegra_vde_attach_dmabuf(vde, src->cr_fd,
565 src->cr_offset, csize, SZ_256,
566 &frame->cr_dmabuf_attachment,
567 &frame->cr_addr,
568 NULL, dma_dir);
569 if (err)
570 goto err_release_cb;
571
572 if (baseline_profile) {
573 frame->aux_addr = 0x64DEAD00;
574 return 0;
575 }
576
577 err = tegra_vde_attach_dmabuf(vde, src->aux_fd,
578 src->aux_offset, csize, SZ_256,
579 &frame->aux_dmabuf_attachment,
580 &frame->aux_addr,
581 NULL, dma_dir);
582 if (err)
583 goto err_release_cr;
584
585 return 0;
586
587 err_release_cr:
588 tegra_vde_dmabuf_cache_unmap(vde, frame->cr_dmabuf_attachment, true);
589 err_release_cb:
590 tegra_vde_dmabuf_cache_unmap(vde, frame->cb_dmabuf_attachment, true);
591 err_release_y:
592 tegra_vde_dmabuf_cache_unmap(vde, frame->y_dmabuf_attachment, true);
593
594 return err;
595 }
596
597 static void tegra_vde_release_frame_dmabufs(struct tegra_vde *vde,
598 struct video_frame *frame,
599 enum dma_data_direction dma_dir,
600 bool baseline_profile,
601 bool release)
602 {
603 if (!baseline_profile)
604 tegra_vde_dmabuf_cache_unmap(vde, frame->aux_dmabuf_attachment,
605 release);
606
607 tegra_vde_dmabuf_cache_unmap(vde, frame->cr_dmabuf_attachment, release);
608 tegra_vde_dmabuf_cache_unmap(vde, frame->cb_dmabuf_attachment, release);
609 tegra_vde_dmabuf_cache_unmap(vde, frame->y_dmabuf_attachment, release);
610 }
611
612 static int tegra_vde_validate_frame(struct device *dev,
613 struct tegra_vde_h264_frame *frame)
614 {
615 if (frame->frame_num > 0x7FFFFF) {
616 dev_err(dev, "Bad frame_num %u\n", frame->frame_num);
617 return -EINVAL;
618 }
619
620 return 0;
621 }
622
623 static int tegra_vde_validate_h264_ctx(struct device *dev,
624 struct tegra_vde_h264_decoder_ctx *ctx)
625 {
626 if (ctx->dpb_frames_nb == 0 || ctx->dpb_frames_nb > 17) {
627 dev_err(dev, "Bad DPB size %u\n", ctx->dpb_frames_nb);
628 return -EINVAL;
629 }
630
631 if (ctx->level_idc > 15) {
632 dev_err(dev, "Bad level value %u\n", ctx->level_idc);
633 return -EINVAL;
634 }
635
636 if (ctx->pic_init_qp > 52) {
637 dev_err(dev, "Bad pic_init_qp value %u\n", ctx->pic_init_qp);
638 return -EINVAL;
639 }
640
641 if (ctx->log2_max_pic_order_cnt_lsb > 16) {
642 dev_err(dev, "Bad log2_max_pic_order_cnt_lsb value %u\n",
643 ctx->log2_max_pic_order_cnt_lsb);
644 return -EINVAL;
645 }
646
647 if (ctx->log2_max_frame_num > 16) {
648 dev_err(dev, "Bad log2_max_frame_num value %u\n",
649 ctx->log2_max_frame_num);
650 return -EINVAL;
651 }
652
653 if (ctx->chroma_qp_index_offset > 31) {
654 dev_err(dev, "Bad chroma_qp_index_offset value %u\n",
655 ctx->chroma_qp_index_offset);
656 return -EINVAL;
657 }
658
659 if (ctx->pic_order_cnt_type > 2) {
660 dev_err(dev, "Bad pic_order_cnt_type value %u\n",
661 ctx->pic_order_cnt_type);
662 return -EINVAL;
663 }
664
665 if (ctx->num_ref_idx_l0_active_minus1 > 15) {
666 dev_err(dev, "Bad num_ref_idx_l0_active_minus1 value %u\n",
667 ctx->num_ref_idx_l0_active_minus1);
668 return -EINVAL;
669 }
670
671 if (ctx->num_ref_idx_l1_active_minus1 > 15) {
672 dev_err(dev, "Bad num_ref_idx_l1_active_minus1 value %u\n",
673 ctx->num_ref_idx_l1_active_minus1);
674 return -EINVAL;
675 }
676
677 if (!ctx->pic_width_in_mbs || ctx->pic_width_in_mbs > 127) {
678 dev_err(dev, "Bad pic_width_in_mbs value %u\n",
679 ctx->pic_width_in_mbs);
680 return -EINVAL;
681 }
682
683 if (!ctx->pic_height_in_mbs || ctx->pic_height_in_mbs > 127) {
684 dev_err(dev, "Bad pic_height_in_mbs value %u\n",
685 ctx->pic_height_in_mbs);
686 return -EINVAL;
687 }
688
689 return 0;
690 }
691
692 static int tegra_vde_ioctl_decode_h264(struct tegra_vde *vde,
693 unsigned long vaddr)
694 {
695 struct device *dev = vde->miscdev.parent;
696 struct tegra_vde_h264_decoder_ctx ctx;
697 struct tegra_vde_h264_frame *frames;
698 struct tegra_vde_h264_frame __user *frames_user;
699 struct video_frame *dpb_frames;
700 struct dma_buf_attachment *bitstream_data_dmabuf_attachment;
701 enum dma_data_direction dma_dir;
702 dma_addr_t bitstream_data_addr;
703 dma_addr_t bsev_ptr;
704 size_t lsize, csize;
705 size_t bitstream_data_size;
706 unsigned int macroblocks_nb;
707 unsigned int read_bytes;
708 unsigned int cstride;
709 unsigned int i;
710 long timeout;
711 int ret, err;
712
713 if (copy_from_user(&ctx, (void __user *)vaddr, sizeof(ctx)))
714 return -EFAULT;
715
716 ret = tegra_vde_validate_h264_ctx(dev, &ctx);
717 if (ret)
718 return ret;
719
720 ret = tegra_vde_attach_dmabuf(vde, ctx.bitstream_data_fd,
721 ctx.bitstream_data_offset,
722 SZ_16K, SZ_16K,
723 &bitstream_data_dmabuf_attachment,
724 &bitstream_data_addr,
725 &bitstream_data_size,
726 DMA_TO_DEVICE);
727 if (ret)
728 return ret;
729
730 frames = kmalloc_array(ctx.dpb_frames_nb, sizeof(*frames), GFP_KERNEL);
731 if (!frames) {
732 ret = -ENOMEM;
733 goto release_bitstream_dmabuf;
734 }
735
736 dpb_frames = kcalloc(ctx.dpb_frames_nb, sizeof(*dpb_frames),
737 GFP_KERNEL);
738 if (!dpb_frames) {
739 ret = -ENOMEM;
740 goto free_frames;
741 }
742
743 macroblocks_nb = ctx.pic_width_in_mbs * ctx.pic_height_in_mbs;
744 frames_user = u64_to_user_ptr(ctx.dpb_frames_ptr);
745
746 if (copy_from_user(frames, frames_user,
747 ctx.dpb_frames_nb * sizeof(*frames))) {
748 ret = -EFAULT;
749 goto free_dpb_frames;
750 }
751
752 cstride = ALIGN(ctx.pic_width_in_mbs * 8, 16);
753 csize = cstride * ctx.pic_height_in_mbs * 8;
754 lsize = macroblocks_nb * 256;
755
756 for (i = 0; i < ctx.dpb_frames_nb; i++) {
757 ret = tegra_vde_validate_frame(dev, &frames[i]);
758 if (ret)
759 goto release_dpb_frames;
760
761 dpb_frames[i].flags = frames[i].flags;
762 dpb_frames[i].frame_num = frames[i].frame_num;
763
764 dma_dir = (i == 0) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
765
766 ret = tegra_vde_attach_dmabufs_to_frame(vde, &dpb_frames[i],
767 &frames[i], dma_dir,
768 ctx.baseline_profile,
769 lsize, csize);
770 if (ret)
771 goto release_dpb_frames;
772 }
773
774 ret = mutex_lock_interruptible(&vde->lock);
775 if (ret)
776 goto release_dpb_frames;
777
778 ret = pm_runtime_get_sync(dev);
779 if (ret < 0)
780 goto unlock;
781
782
783
784
785
786 ret = reset_control_assert(vde->rst_mc);
787 if (ret) {
788 dev_err(dev, "DEC start: Failed to assert MC reset: %d\n",
789 ret);
790 goto put_runtime_pm;
791 }
792
793 ret = reset_control_reset(vde->rst);
794 if (ret) {
795 dev_err(dev, "DEC start: Failed to reset HW: %d\n", ret);
796 goto put_runtime_pm;
797 }
798
799 ret = reset_control_deassert(vde->rst_mc);
800 if (ret) {
801 dev_err(dev, "DEC start: Failed to deassert MC reset: %d\n",
802 ret);
803 goto put_runtime_pm;
804 }
805
806 ret = tegra_vde_setup_hw_context(vde, &ctx, dpb_frames,
807 bitstream_data_addr,
808 bitstream_data_size,
809 macroblocks_nb);
810 if (ret)
811 goto put_runtime_pm;
812
813 tegra_vde_decode_frame(vde, macroblocks_nb);
814
815 timeout = wait_for_completion_interruptible_timeout(
816 &vde->decode_completion, msecs_to_jiffies(1000));
817 if (timeout == 0) {
818 bsev_ptr = tegra_vde_readl(vde, vde->bsev, 0x10);
819 macroblocks_nb = tegra_vde_readl(vde, vde->sxe, 0xC8) & 0x1FFF;
820 read_bytes = bsev_ptr ? bsev_ptr - bitstream_data_addr : 0;
821
822 dev_err(dev, "Decoding failed: read 0x%X bytes, %u macroblocks parsed\n",
823 read_bytes, macroblocks_nb);
824
825 ret = -EIO;
826 } else if (timeout < 0) {
827 ret = timeout;
828 }
829
830
831
832
833
834
835 err = reset_control_assert(vde->rst_mc);
836 if (err)
837 dev_err(dev, "DEC end: Failed to assert MC reset: %d\n", err);
838
839 err = reset_control_assert(vde->rst);
840 if (err)
841 dev_err(dev, "DEC end: Failed to assert HW reset: %d\n", err);
842
843 put_runtime_pm:
844 pm_runtime_mark_last_busy(dev);
845 pm_runtime_put_autosuspend(dev);
846
847 unlock:
848 mutex_unlock(&vde->lock);
849
850 release_dpb_frames:
851 while (i--) {
852 dma_dir = (i == 0) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
853
854 tegra_vde_release_frame_dmabufs(vde, &dpb_frames[i], dma_dir,
855 ctx.baseline_profile, ret != 0);
856 }
857
858 free_dpb_frames:
859 kfree(dpb_frames);
860
861 free_frames:
862 kfree(frames);
863
864 release_bitstream_dmabuf:
865 tegra_vde_dmabuf_cache_unmap(vde, bitstream_data_dmabuf_attachment,
866 ret != 0);
867
868 return ret;
869 }
870
871 static long tegra_vde_unlocked_ioctl(struct file *filp,
872 unsigned int cmd, unsigned long arg)
873 {
874 struct miscdevice *miscdev = filp->private_data;
875 struct tegra_vde *vde = container_of(miscdev, struct tegra_vde,
876 miscdev);
877
878 switch (cmd) {
879 case TEGRA_VDE_IOCTL_DECODE_H264:
880 return tegra_vde_ioctl_decode_h264(vde, arg);
881 }
882
883 dev_err(miscdev->parent, "Invalid IOCTL command %u\n", cmd);
884
885 return -ENOTTY;
886 }
887
888 static int tegra_vde_release_file(struct inode *inode, struct file *filp)
889 {
890 struct miscdevice *miscdev = filp->private_data;
891 struct tegra_vde *vde = container_of(miscdev, struct tegra_vde,
892 miscdev);
893
894 tegra_vde_dmabuf_cache_unmap_sync(vde);
895
896 return 0;
897 }
898
899 static const struct file_operations tegra_vde_fops = {
900 .owner = THIS_MODULE,
901 .unlocked_ioctl = tegra_vde_unlocked_ioctl,
902 .release = tegra_vde_release_file,
903 };
904
905 static irqreturn_t tegra_vde_isr(int irq, void *data)
906 {
907 struct tegra_vde *vde = data;
908
909 if (completion_done(&vde->decode_completion))
910 return IRQ_NONE;
911
912 tegra_vde_set_bits(vde, 0, vde->frameid, 0x208);
913 complete(&vde->decode_completion);
914
915 return IRQ_HANDLED;
916 }
917
918 static int tegra_vde_runtime_suspend(struct device *dev)
919 {
920 struct tegra_vde *vde = dev_get_drvdata(dev);
921 int err;
922
923 err = tegra_powergate_power_off(TEGRA_POWERGATE_VDEC);
924 if (err) {
925 dev_err(dev, "Failed to power down HW: %d\n", err);
926 return err;
927 }
928
929 clk_disable_unprepare(vde->clk);
930
931 return 0;
932 }
933
934 static int tegra_vde_runtime_resume(struct device *dev)
935 {
936 struct tegra_vde *vde = dev_get_drvdata(dev);
937 int err;
938
939 err = tegra_powergate_sequence_power_up(TEGRA_POWERGATE_VDEC,
940 vde->clk, vde->rst);
941 if (err) {
942 dev_err(dev, "Failed to power up HW : %d\n", err);
943 return err;
944 }
945
946 return 0;
947 }
948
949 static int tegra_vde_probe(struct platform_device *pdev)
950 {
951 struct device *dev = &pdev->dev;
952 struct resource *regs;
953 struct tegra_vde *vde;
954 int irq, err;
955
956 vde = devm_kzalloc(dev, sizeof(*vde), GFP_KERNEL);
957 if (!vde)
958 return -ENOMEM;
959
960 platform_set_drvdata(pdev, vde);
961
962 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "sxe");
963 if (!regs)
964 return -ENODEV;
965
966 vde->sxe = devm_ioremap_resource(dev, regs);
967 if (IS_ERR(vde->sxe))
968 return PTR_ERR(vde->sxe);
969
970 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "bsev");
971 if (!regs)
972 return -ENODEV;
973
974 vde->bsev = devm_ioremap_resource(dev, regs);
975 if (IS_ERR(vde->bsev))
976 return PTR_ERR(vde->bsev);
977
978 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "mbe");
979 if (!regs)
980 return -ENODEV;
981
982 vde->mbe = devm_ioremap_resource(dev, regs);
983 if (IS_ERR(vde->mbe))
984 return PTR_ERR(vde->mbe);
985
986 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "ppe");
987 if (!regs)
988 return -ENODEV;
989
990 vde->ppe = devm_ioremap_resource(dev, regs);
991 if (IS_ERR(vde->ppe))
992 return PTR_ERR(vde->ppe);
993
994 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "mce");
995 if (!regs)
996 return -ENODEV;
997
998 vde->mce = devm_ioremap_resource(dev, regs);
999 if (IS_ERR(vde->mce))
1000 return PTR_ERR(vde->mce);
1001
1002 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "tfe");
1003 if (!regs)
1004 return -ENODEV;
1005
1006 vde->tfe = devm_ioremap_resource(dev, regs);
1007 if (IS_ERR(vde->tfe))
1008 return PTR_ERR(vde->tfe);
1009
1010 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "ppb");
1011 if (!regs)
1012 return -ENODEV;
1013
1014 vde->ppb = devm_ioremap_resource(dev, regs);
1015 if (IS_ERR(vde->ppb))
1016 return PTR_ERR(vde->ppb);
1017
1018 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "vdma");
1019 if (!regs)
1020 return -ENODEV;
1021
1022 vde->vdma = devm_ioremap_resource(dev, regs);
1023 if (IS_ERR(vde->vdma))
1024 return PTR_ERR(vde->vdma);
1025
1026 regs = platform_get_resource_byname(pdev, IORESOURCE_MEM, "frameid");
1027 if (!regs)
1028 return -ENODEV;
1029
1030 vde->frameid = devm_ioremap_resource(dev, regs);
1031 if (IS_ERR(vde->frameid))
1032 return PTR_ERR(vde->frameid);
1033
1034 vde->clk = devm_clk_get(dev, NULL);
1035 if (IS_ERR(vde->clk)) {
1036 err = PTR_ERR(vde->clk);
1037 dev_err(dev, "Could not get VDE clk %d\n", err);
1038 return err;
1039 }
1040
1041 vde->rst = devm_reset_control_get(dev, NULL);
1042 if (IS_ERR(vde->rst)) {
1043 err = PTR_ERR(vde->rst);
1044 dev_err(dev, "Could not get VDE reset %d\n", err);
1045 return err;
1046 }
1047
1048 vde->rst_mc = devm_reset_control_get_optional(dev, "mc");
1049 if (IS_ERR(vde->rst_mc)) {
1050 err = PTR_ERR(vde->rst_mc);
1051 dev_err(dev, "Could not get MC reset %d\n", err);
1052 return err;
1053 }
1054
1055 irq = platform_get_irq_byname(pdev, "sync-token");
1056 if (irq < 0)
1057 return irq;
1058
1059 err = devm_request_irq(dev, irq, tegra_vde_isr, 0,
1060 dev_name(dev), vde);
1061 if (err) {
1062 dev_err(dev, "Could not request IRQ %d\n", err);
1063 return err;
1064 }
1065
1066 vde->iram_pool = of_gen_pool_get(dev->of_node, "iram", 0);
1067 if (!vde->iram_pool) {
1068 dev_err(dev, "Could not get IRAM pool\n");
1069 return -EPROBE_DEFER;
1070 }
1071
1072 vde->iram = gen_pool_dma_alloc(vde->iram_pool,
1073 gen_pool_size(vde->iram_pool),
1074 &vde->iram_lists_addr);
1075 if (!vde->iram) {
1076 dev_err(dev, "Could not reserve IRAM\n");
1077 return -ENOMEM;
1078 }
1079
1080 INIT_LIST_HEAD(&vde->map_list);
1081 mutex_init(&vde->map_lock);
1082 mutex_init(&vde->lock);
1083 init_completion(&vde->decode_completion);
1084
1085 vde->miscdev.minor = MISC_DYNAMIC_MINOR;
1086 vde->miscdev.name = "tegra_vde";
1087 vde->miscdev.fops = &tegra_vde_fops;
1088 vde->miscdev.parent = dev;
1089
1090 err = tegra_vde_iommu_init(vde);
1091 if (err) {
1092 dev_err(dev, "Failed to initialize IOMMU: %d\n", err);
1093 goto err_gen_free;
1094 }
1095
1096 err = misc_register(&vde->miscdev);
1097 if (err) {
1098 dev_err(dev, "Failed to register misc device: %d\n", err);
1099 goto err_deinit_iommu;
1100 }
1101
1102 pm_runtime_enable(dev);
1103 pm_runtime_use_autosuspend(dev);
1104 pm_runtime_set_autosuspend_delay(dev, 300);
1105
1106 if (!pm_runtime_enabled(dev)) {
1107 err = tegra_vde_runtime_resume(dev);
1108 if (err)
1109 goto err_misc_unreg;
1110 }
1111
1112 return 0;
1113
1114 err_misc_unreg:
1115 misc_deregister(&vde->miscdev);
1116
1117 err_deinit_iommu:
1118 tegra_vde_iommu_deinit(vde);
1119
1120 err_gen_free:
1121 gen_pool_free(vde->iram_pool, (unsigned long)vde->iram,
1122 gen_pool_size(vde->iram_pool));
1123
1124 return err;
1125 }
1126
1127 static int tegra_vde_remove(struct platform_device *pdev)
1128 {
1129 struct tegra_vde *vde = platform_get_drvdata(pdev);
1130 struct device *dev = &pdev->dev;
1131 int err;
1132
1133 if (!pm_runtime_enabled(dev)) {
1134 err = tegra_vde_runtime_suspend(dev);
1135 if (err)
1136 return err;
1137 }
1138
1139 pm_runtime_dont_use_autosuspend(dev);
1140 pm_runtime_disable(dev);
1141
1142 misc_deregister(&vde->miscdev);
1143
1144 tegra_vde_dmabuf_cache_unmap_all(vde);
1145 tegra_vde_iommu_deinit(vde);
1146
1147 gen_pool_free(vde->iram_pool, (unsigned long)vde->iram,
1148 gen_pool_size(vde->iram_pool));
1149
1150 return 0;
1151 }
1152
1153 #ifdef CONFIG_PM_SLEEP
1154 static int tegra_vde_pm_suspend(struct device *dev)
1155 {
1156 struct tegra_vde *vde = dev_get_drvdata(dev);
1157 int err;
1158
1159 mutex_lock(&vde->lock);
1160
1161 err = pm_runtime_force_suspend(dev);
1162 if (err < 0)
1163 return err;
1164
1165 return 0;
1166 }
1167
1168 static int tegra_vde_pm_resume(struct device *dev)
1169 {
1170 struct tegra_vde *vde = dev_get_drvdata(dev);
1171 int err;
1172
1173 err = pm_runtime_force_resume(dev);
1174 if (err < 0)
1175 return err;
1176
1177 mutex_unlock(&vde->lock);
1178
1179 return 0;
1180 }
1181 #endif
1182
1183 static const struct dev_pm_ops tegra_vde_pm_ops = {
1184 SET_RUNTIME_PM_OPS(tegra_vde_runtime_suspend,
1185 tegra_vde_runtime_resume,
1186 NULL)
1187 SET_SYSTEM_SLEEP_PM_OPS(tegra_vde_pm_suspend,
1188 tegra_vde_pm_resume)
1189 };
1190
1191 static const struct of_device_id tegra_vde_of_match[] = {
1192 { .compatible = "nvidia,tegra20-vde", },
1193 { },
1194 };
1195 MODULE_DEVICE_TABLE(of, tegra_vde_of_match);
1196
1197 static struct platform_driver tegra_vde_driver = {
1198 .probe = tegra_vde_probe,
1199 .remove = tegra_vde_remove,
1200 .driver = {
1201 .name = "tegra-vde",
1202 .of_match_table = tegra_vde_of_match,
1203 .pm = &tegra_vde_pm_ops,
1204 },
1205 };
1206 module_platform_driver(tegra_vde_driver);
1207
1208 MODULE_DESCRIPTION("NVIDIA Tegra Video Decoder driver");
1209 MODULE_AUTHOR("Dmitry Osipenko <digetx@gmail.com>");
1210 MODULE_LICENSE("GPL");