1/*
2 * omap_vout_vrfb.c
3 *
4 * Copyright (C) 2010 Texas Instruments.
5 *
6 * This file is licensed under the terms of the GNU General Public License
7 * version 2. This program is licensed "as is" without any warranty of any
8 * kind, whether express or implied.
9 *
10 */
11
12#include <linux/sched.h>
13#include <linux/platform_device.h>
14#include <linux/videodev2.h>
15#include <linux/slab.h>
16
17#include <media/v4l2-device.h>
18
19#include <video/omapvrfb.h>
20
21#include "omap_voutdef.h"
22#include "omap_voutlib.h"
23#include "omap_vout_vrfb.h"
24
25#define OMAP_DMA_NO_DEVICE	0
26
27/*
28 * Function for allocating video buffers
29 */
30static int omap_vout_allocate_vrfb_buffers(struct omap_vout_device *vout,
31		unsigned int *count, int startindex)
32{
33	int i, j;
34
35	for (i = 0; i < *count; i++) {
36		if (!vout->smsshado_virt_addr[i]) {
37			vout->smsshado_virt_addr[i] =
38				omap_vout_alloc_buffer(vout->smsshado_size,
39						&vout->smsshado_phy_addr[i]);
40		}
41		if (!vout->smsshado_virt_addr[i] && startindex != -1) {
42			if (vout->vq.memory == V4L2_MEMORY_MMAP && i >= startindex)
43				break;
44		}
45		if (!vout->smsshado_virt_addr[i]) {
46			for (j = 0; j < i; j++) {
47				omap_vout_free_buffer(
48						vout->smsshado_virt_addr[j],
49						vout->smsshado_size);
50				vout->smsshado_virt_addr[j] = 0;
51				vout->smsshado_phy_addr[j] = 0;
52			}
53			*count = 0;
54			return -ENOMEM;
55		}
56		memset((void *)(long)vout->smsshado_virt_addr[i], 0,
57		       vout->smsshado_size);
58	}
59	return 0;
60}
61
62/*
63 * Wakes up the application once the DMA transfer to VRFB space is completed.
64 */
65static void omap_vout_vrfb_dma_tx_callback(void *data)
66{
67	struct vid_vrfb_dma *t = (struct vid_vrfb_dma *) data;
68
69	t->tx_status = 1;
70	wake_up_interruptible(&t->wait);
71}
72
73/*
74 * Free VRFB buffers
75 */
76void omap_vout_free_vrfb_buffers(struct omap_vout_device *vout)
77{
78	int j;
79
80	for (j = 0; j < VRFB_NUM_BUFS; j++) {
81		if (vout->smsshado_virt_addr[j]) {
82			omap_vout_free_buffer(vout->smsshado_virt_addr[j],
83					      vout->smsshado_size);
84			vout->smsshado_virt_addr[j] = 0;
85			vout->smsshado_phy_addr[j] = 0;
86		}
87	}
88}
89
90int omap_vout_setup_vrfb_bufs(struct platform_device *pdev, int vid_num,
91			      bool static_vrfb_allocation)
92{
93	int ret = 0, i, j;
94	struct omap_vout_device *vout;
95	struct video_device *vfd;
96	dma_cap_mask_t mask;
97	int image_width, image_height;
98	int vrfb_num_bufs = VRFB_NUM_BUFS;
99	struct v4l2_device *v4l2_dev = platform_get_drvdata(pdev);
100	struct omap2video_device *vid_dev =
101		container_of(v4l2_dev, struct omap2video_device, v4l2_dev);
102
103	vout = vid_dev->vouts[vid_num];
104	vfd = vout->vfd;
105
106	for (i = 0; i < VRFB_NUM_BUFS; i++) {
107		if (omap_vrfb_request_ctx(&vout->vrfb_context[i])) {
108			dev_info(&pdev->dev, ": VRFB allocation failed\n");
109			for (j = 0; j < i; j++)
110				omap_vrfb_release_ctx(&vout->vrfb_context[j]);
111			return -ENOMEM;
112		}
113	}
114
115	/* Calculate VRFB memory size */
116	/* allocate for worst case size */
117	image_width = VID_MAX_WIDTH / TILE_SIZE;
118	if (VID_MAX_WIDTH % TILE_SIZE)
119		image_width++;
120
121	image_width = image_width * TILE_SIZE;
122	image_height = VID_MAX_HEIGHT / TILE_SIZE;
123
124	if (VID_MAX_HEIGHT % TILE_SIZE)
125		image_height++;
126
127	image_height = image_height * TILE_SIZE;
128	vout->smsshado_size = PAGE_ALIGN(image_width * image_height * 2 * 2);
129
130	/*
131	 * Request and Initialize DMA, for DMA based VRFB transfer
132	 */
133	dma_cap_zero(mask);
134	dma_cap_set(DMA_INTERLEAVE, mask);
135	vout->vrfb_dma_tx.chan = dma_request_chan_by_mask(&mask);
136	if (IS_ERR(vout->vrfb_dma_tx.chan)) {
137		vout->vrfb_dma_tx.req_status = DMA_CHAN_NOT_ALLOTED;
138	} else {
139		size_t xt_size = sizeof(struct dma_interleaved_template) +
140				 sizeof(struct data_chunk);
141
142		vout->vrfb_dma_tx.xt = kzalloc(xt_size, GFP_KERNEL);
143		if (!vout->vrfb_dma_tx.xt) {
144			dma_release_channel(vout->vrfb_dma_tx.chan);
145			vout->vrfb_dma_tx.req_status = DMA_CHAN_NOT_ALLOTED;
146		}
147	}
148
149	if (vout->vrfb_dma_tx.req_status == DMA_CHAN_NOT_ALLOTED)
150		dev_info(&pdev->dev,
151			 ": failed to allocate DMA Channel for video%d\n",
152			 vfd->minor);
153
154	init_waitqueue_head(&vout->vrfb_dma_tx.wait);
155
156	/*
157	 * statically allocated the VRFB buffer is done through
158	 * command line arguments
159	 */
160	if (static_vrfb_allocation) {
161		if (omap_vout_allocate_vrfb_buffers(vout, &vrfb_num_bufs, -1)) {
162			ret =  -ENOMEM;
163			goto release_vrfb_ctx;
164		}
165		vout->vrfb_static_allocation = true;
166	}
167	return 0;
168
169release_vrfb_ctx:
170	for (j = 0; j < VRFB_NUM_BUFS; j++)
171		omap_vrfb_release_ctx(&vout->vrfb_context[j]);
172	return ret;
173}
174
175/*
176 * Release the VRFB context once the module exits
177 */
178void omap_vout_release_vrfb(struct omap_vout_device *vout)
179{
180	int i;
181
182	for (i = 0; i < VRFB_NUM_BUFS; i++)
183		omap_vrfb_release_ctx(&vout->vrfb_context[i]);
184
185	if (vout->vrfb_dma_tx.req_status == DMA_CHAN_ALLOTED) {
186		vout->vrfb_dma_tx.req_status = DMA_CHAN_NOT_ALLOTED;
187		kfree(vout->vrfb_dma_tx.xt);
188		dmaengine_terminate_sync(vout->vrfb_dma_tx.chan);
189		dma_release_channel(vout->vrfb_dma_tx.chan);
190	}
191}
192
193/*
194 * Allocate the buffers for the VRFB space.  Data is copied from V4L2
195 * buffers to the VRFB buffers using the DMA engine.
196 */
197int omap_vout_vrfb_buffer_setup(struct omap_vout_device *vout,
198			  unsigned int *count, unsigned int startindex)
199{
200	int i;
201	bool yuv_mode;
202
203	if (!is_rotation_enabled(vout))
204		return 0;
205
206	/* If rotation is enabled, allocate memory for VRFB space also */
207	*count = *count > VRFB_NUM_BUFS ? VRFB_NUM_BUFS : *count;
208
209	/* Allocate the VRFB buffers only if the buffers are not
210	 * allocated during init time.
211	 */
212	if (!vout->vrfb_static_allocation)
213		if (omap_vout_allocate_vrfb_buffers(vout, count, startindex))
214			return -ENOMEM;
215
216	if (vout->dss_mode == OMAP_DSS_COLOR_YUV2 ||
217			vout->dss_mode == OMAP_DSS_COLOR_UYVY)
218		yuv_mode = true;
219	else
220		yuv_mode = false;
221
222	for (i = 0; i < *count; i++)
223		omap_vrfb_setup(&vout->vrfb_context[i],
224				vout->smsshado_phy_addr[i], vout->pix.width,
225				vout->pix.height, vout->bpp, yuv_mode);
226
227	return 0;
228}
229
230int omap_vout_prepare_vrfb(struct omap_vout_device *vout,
231			   struct vb2_buffer *vb)
232{
233	struct dma_async_tx_descriptor *tx;
234	enum dma_ctrl_flags flags = DMA_PREP_INTERRUPT | DMA_CTRL_ACK;
235	struct dma_chan *chan = vout->vrfb_dma_tx.chan;
236	struct dma_interleaved_template *xt = vout->vrfb_dma_tx.xt;
237	dma_cookie_t cookie;
238	dma_addr_t buf_phy_addr = vb2_dma_contig_plane_dma_addr(vb, 0);
239	enum dma_status status;
240	enum dss_rotation rotation;
241	size_t dst_icg;
242	u32 pixsize;
243
244	if (!is_rotation_enabled(vout))
245		return 0;
246
247	/* If rotation is enabled, copy input buffer into VRFB
248	 * memory space using DMA. We are copying input buffer
249	 * into VRFB memory space of desired angle and DSS will
250	 * read image VRFB memory for 0 degree angle
251	 */
252
253	pixsize = vout->bpp * vout->vrfb_bpp;
254	dst_icg = MAX_PIXELS_PER_LINE * pixsize - vout->pix.width * vout->bpp;
255
256	xt->src_start = buf_phy_addr;
257	xt->dst_start = vout->vrfb_context[vb->index].paddr[0];
258
259	xt->numf = vout->pix.height;
260	xt->frame_size = 1;
261	xt->sgl[0].size = vout->pix.width * vout->bpp;
262	xt->sgl[0].icg = dst_icg;
263
264	xt->dir = DMA_MEM_TO_MEM;
265	xt->src_sgl = false;
266	xt->src_inc = true;
267	xt->dst_sgl = true;
268	xt->dst_inc = true;
269
270	tx = dmaengine_prep_interleaved_dma(chan, xt, flags);
271	if (tx == NULL) {
272		pr_err("%s: DMA interleaved prep error\n", __func__);
273		return -EINVAL;
274	}
275
276	tx->callback = omap_vout_vrfb_dma_tx_callback;
277	tx->callback_param = &vout->vrfb_dma_tx;
278
279	cookie = dmaengine_submit(tx);
280	if (dma_submit_error(cookie)) {
281		pr_err("%s: dmaengine_submit failed (%d)\n", __func__, cookie);
282		return -EINVAL;
283	}
284
285	vout->vrfb_dma_tx.tx_status = 0;
286	dma_async_issue_pending(chan);
287
288	wait_event_interruptible_timeout(vout->vrfb_dma_tx.wait,
289					 vout->vrfb_dma_tx.tx_status == 1,
290					 VRFB_TX_TIMEOUT);
291
292	status = dma_async_is_tx_complete(chan, cookie, NULL, NULL);
293
294	if (vout->vrfb_dma_tx.tx_status == 0) {
295		pr_err("%s: Timeout while waiting for DMA\n", __func__);
296		dmaengine_terminate_sync(chan);
297		return -EINVAL;
298	} else if (status != DMA_COMPLETE) {
299		pr_err("%s: DMA completion %s status\n", __func__,
300		       status == DMA_ERROR ? "error" : "busy");
301		dmaengine_terminate_sync(chan);
302		return -EINVAL;
303	}
304
305	/* Store buffers physical address into an array. Addresses
306	 * from this array will be used to configure DSS */
307	rotation = calc_rotation(vout);
308	vout->queued_buf_addr[vb->index] =
309		vout->vrfb_context[vb->index].paddr[rotation];
310	return 0;
311}
312
313/*
314 * Calculate the buffer offsets from which the streaming should
315 * start. This offset calculation is mainly required because of
316 * the VRFB 32 pixels alignment with rotation.
317 */
318void omap_vout_calculate_vrfb_offset(struct omap_vout_device *vout)
319{
320	enum dss_rotation rotation;
321	bool mirroring = vout->mirror;
322	struct v4l2_rect *crop = &vout->crop;
323	struct v4l2_pix_format *pix = &vout->pix;
324	int *cropped_offset = &vout->cropped_offset;
325	int vr_ps = 1, ps = 2, temp_ps = 2;
326	int offset = 0, ctop = 0, cleft = 0, line_length = 0;
327
328	rotation = calc_rotation(vout);
329
330	if (V4L2_PIX_FMT_YUYV == pix->pixelformat ||
331			V4L2_PIX_FMT_UYVY == pix->pixelformat) {
332		if (is_rotation_enabled(vout)) {
333			/*
334			 * ps    - Actual pixel size for YUYV/UYVY for
335			 *         VRFB/Mirroring is 4 bytes
336			 * vr_ps - Virtually pixel size for YUYV/UYVY is
337			 *         2 bytes
338			 */
339			ps = 4;
340			vr_ps = 2;
341		} else {
342			ps = 2;	/* otherwise the pixel size is 2 byte */
343		}
344	} else if (V4L2_PIX_FMT_RGB32 == pix->pixelformat) {
345		ps = 4;
346	} else if (V4L2_PIX_FMT_RGB24 == pix->pixelformat) {
347		ps = 3;
348	}
349	vout->ps = ps;
350	vout->vr_ps = vr_ps;
351
352	if (is_rotation_enabled(vout)) {
353		line_length = MAX_PIXELS_PER_LINE;
354		ctop = (pix->height - crop->height) - crop->top;
355		cleft = (pix->width - crop->width) - crop->left;
356	} else {
357		line_length = pix->width;
358	}
359	vout->line_length = line_length;
360	switch (rotation) {
361	case dss_rotation_90_degree:
362		offset = vout->vrfb_context[0].yoffset *
363			vout->vrfb_context[0].bytespp;
364		temp_ps = ps / vr_ps;
365		if (!mirroring) {
366			*cropped_offset = offset + line_length *
367				temp_ps * cleft + crop->top * temp_ps;
368		} else {
369			*cropped_offset = offset + line_length * temp_ps *
370				cleft + crop->top * temp_ps + (line_length *
371				((crop->width / (vr_ps)) - 1) * ps);
372		}
373		break;
374	case dss_rotation_180_degree:
375		offset = ((MAX_PIXELS_PER_LINE * vout->vrfb_context[0].yoffset *
376			vout->vrfb_context[0].bytespp) +
377			(vout->vrfb_context[0].xoffset *
378			vout->vrfb_context[0].bytespp));
379		if (!mirroring) {
380			*cropped_offset = offset + (line_length * ps * ctop) +
381				(cleft / vr_ps) * ps;
382
383		} else {
384			*cropped_offset = offset + (line_length * ps * ctop) +
385				(cleft / vr_ps) * ps + (line_length *
386				(crop->height - 1) * ps);
387		}
388		break;
389	case dss_rotation_270_degree:
390		offset = MAX_PIXELS_PER_LINE * vout->vrfb_context[0].xoffset *
391			vout->vrfb_context[0].bytespp;
392		temp_ps = ps / vr_ps;
393		if (!mirroring) {
394			*cropped_offset = offset + line_length *
395			    temp_ps * crop->left + ctop * ps;
396		} else {
397			*cropped_offset = offset + line_length *
398				temp_ps * crop->left + ctop * ps +
399				(line_length * ((crop->width / vr_ps) - 1) *
400				 ps);
401		}
402		break;
403	case dss_rotation_0_degree:
404		if (!mirroring) {
405			*cropped_offset = (line_length * ps) *
406				crop->top + (crop->left / vr_ps) * ps;
407		} else {
408			*cropped_offset = (line_length * ps) *
409				crop->top + (crop->left / vr_ps) * ps +
410				(line_length * (crop->height - 1) * ps);
411		}
412		break;
413	default:
414		*cropped_offset = (line_length * ps * crop->top) /
415			vr_ps + (crop->left * ps) / vr_ps +
416			((crop->width / vr_ps) - 1) * ps;
417		break;
418	}
419}
420