blob: 8355e55b4e8ed75fa94d0f087f5908b95e78e12c [file] [log] [blame]
/*
* tw68 functions to handle video data
*
* Much of this code is derived from the cx88 and sa7134 drivers, which
* were in turn derived from the bt87x driver. The original work was by
* Gerd Knorr; more recently the code was enhanced by Mauro Carvalho Chehab,
* Hans Verkuil, Andy Walls and many others. Their work is gratefully
* acknowledged. Full credit goes to them - any problems within this code
* are mine.
*
* Copyright (C) 2009 William M. Brack
*
* Refactored and updated to the latest v4l core frameworks:
*
* Copyright (C) 2014 Hans Verkuil <hverkuil@xs4all.nl>
*
* 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 of the License, 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.
*/
#include <linux/module.h>
#include <media/v4l2-common.h>
#include <media/v4l2-event.h>
#include <media/videobuf2-dma-sg.h>
#include "tw68.h"
#include "tw68-reg.h"
/* ------------------------------------------------------------------ */
/* data structs for video */
/*
* FIXME -
* Note that the saa7134 has formats, e.g. YUV420, which are classified
* as "planar". These affect overlay mode, and are flagged with a field
* ".planar" in the format. Do we need to implement this in this driver?
*/
static const struct tw68_format formats[] = {
{
.name = "15 bpp RGB, le",
.fourcc = V4L2_PIX_FMT_RGB555,
.depth = 16,
.twformat = ColorFormatRGB15,
}, {
.name = "15 bpp RGB, be",
.fourcc = V4L2_PIX_FMT_RGB555X,
.depth = 16,
.twformat = ColorFormatRGB15 | ColorFormatBSWAP,
}, {
.name = "16 bpp RGB, le",
.fourcc = V4L2_PIX_FMT_RGB565,
.depth = 16,
.twformat = ColorFormatRGB16,
}, {
.name = "16 bpp RGB, be",
.fourcc = V4L2_PIX_FMT_RGB565X,
.depth = 16,
.twformat = ColorFormatRGB16 | ColorFormatBSWAP,
}, {
.name = "24 bpp RGB, le",
.fourcc = V4L2_PIX_FMT_BGR24,
.depth = 24,
.twformat = ColorFormatRGB24,
}, {
.name = "24 bpp RGB, be",
.fourcc = V4L2_PIX_FMT_RGB24,
.depth = 24,
.twformat = ColorFormatRGB24 | ColorFormatBSWAP,
}, {
.name = "32 bpp RGB, le",
.fourcc = V4L2_PIX_FMT_BGR32,
.depth = 32,
.twformat = ColorFormatRGB32,
}, {
.name = "32 bpp RGB, be",
.fourcc = V4L2_PIX_FMT_RGB32,
.depth = 32,
.twformat = ColorFormatRGB32 | ColorFormatBSWAP |
ColorFormatWSWAP,
}, {
.name = "4:2:2 packed, YUYV",
.fourcc = V4L2_PIX_FMT_YUYV,
.depth = 16,
.twformat = ColorFormatYUY2,
}, {
.name = "4:2:2 packed, UYVY",
.fourcc = V4L2_PIX_FMT_UYVY,
.depth = 16,
.twformat = ColorFormatYUY2 | ColorFormatBSWAP,
}
};
#define FORMATS ARRAY_SIZE(formats)
#define NORM_625_50 \
.h_delay = 3, \
.h_delay0 = 133, \
.h_start = 0, \
.h_stop = 719, \
.v_delay = 24, \
.vbi_v_start_0 = 7, \
.vbi_v_stop_0 = 22, \
.video_v_start = 24, \
.video_v_stop = 311, \
.vbi_v_start_1 = 319
#define NORM_525_60 \
.h_delay = 8, \
.h_delay0 = 138, \
.h_start = 0, \
.h_stop = 719, \
.v_delay = 22, \
.vbi_v_start_0 = 10, \
.vbi_v_stop_0 = 21, \
.video_v_start = 22, \
.video_v_stop = 262, \
.vbi_v_start_1 = 273
/*
* The following table is searched by tw68_s_std, first for a specific
* match, then for an entry which contains the desired id. The table
* entries should therefore be ordered in ascending order of specificity.
*/
static const struct tw68_tvnorm tvnorms[] = {
{
.name = "PAL", /* autodetect */
.id = V4L2_STD_PAL,
NORM_625_50,
.sync_control = 0x18,
.luma_control = 0x40,
.chroma_ctrl1 = 0x81,
.chroma_gain = 0x2a,
.chroma_ctrl2 = 0x06,
.vgate_misc = 0x1c,
.format = VideoFormatPALBDGHI,
}, {
.name = "NTSC",
.id = V4L2_STD_NTSC,
NORM_525_60,
.sync_control = 0x59,
.luma_control = 0x40,
.chroma_ctrl1 = 0x89,
.chroma_gain = 0x2a,
.chroma_ctrl2 = 0x0e,
.vgate_misc = 0x18,
.format = VideoFormatNTSC,
}, {
.name = "SECAM",
.id = V4L2_STD_SECAM,
NORM_625_50,
.sync_control = 0x18,
.luma_control = 0x1b,
.chroma_ctrl1 = 0xd1,
.chroma_gain = 0x80,
.chroma_ctrl2 = 0x00,
.vgate_misc = 0x1c,
.format = VideoFormatSECAM,
}, {
.name = "PAL-M",
.id = V4L2_STD_PAL_M,
NORM_525_60,
.sync_control = 0x59,
.luma_control = 0x40,
.chroma_ctrl1 = 0xb9,
.chroma_gain = 0x2a,
.chroma_ctrl2 = 0x0e,
.vgate_misc = 0x18,
.format = VideoFormatPALM,
}, {
.name = "PAL-Nc",
.id = V4L2_STD_PAL_Nc,
NORM_625_50,
.sync_control = 0x18,
.luma_control = 0x40,
.chroma_ctrl1 = 0xa1,
.chroma_gain = 0x2a,
.chroma_ctrl2 = 0x06,
.vgate_misc = 0x1c,
.format = VideoFormatPALNC,
}, {
.name = "PAL-60",
.id = V4L2_STD_PAL_60,
.h_delay = 186,
.h_start = 0,
.h_stop = 719,
.v_delay = 26,
.video_v_start = 23,
.video_v_stop = 262,
.vbi_v_start_0 = 10,
.vbi_v_stop_0 = 21,
.vbi_v_start_1 = 273,
.sync_control = 0x18,
.luma_control = 0x40,
.chroma_ctrl1 = 0x81,
.chroma_gain = 0x2a,
.chroma_ctrl2 = 0x06,
.vgate_misc = 0x1c,
.format = VideoFormatPAL60,
}
};
#define TVNORMS ARRAY_SIZE(tvnorms)
static const struct tw68_format *format_by_fourcc(unsigned int fourcc)
{
unsigned int i;
for (i = 0; i < FORMATS; i++)
if (formats[i].fourcc == fourcc)
return formats+i;
return NULL;
}
/* ------------------------------------------------------------------ */
/*
* Note that the cropping rectangles are described in terms of a single
* frame, i.e. line positions are only 1/2 the interlaced equivalent
*/
static void set_tvnorm(struct tw68_dev *dev, const struct tw68_tvnorm *norm)
{
if (norm != dev->tvnorm) {
dev->width = 720;
dev->height = (norm->id & V4L2_STD_525_60) ? 480 : 576;
dev->tvnorm = norm;
tw68_set_tvnorm_hw(dev);
}
}
/*
* tw68_set_scale
*
* Scaling and Cropping for video decoding
*
* We are working with 3 values for horizontal and vertical - scale,
* delay and active.
*
* HACTIVE represent the actual number of pixels in the "usable" image,
* before scaling. HDELAY represents the number of pixels skipped
* between the start of the horizontal sync and the start of the image.
* HSCALE is calculated using the formula
* HSCALE = (HACTIVE / (#pixels desired)) * 256
*
* The vertical registers are similar, except based upon the total number
* of lines in the image, and the first line of the image (i.e. ignoring
* vertical sync and VBI).
*
* Note that the number of bytes reaching the FIFO (and hence needing
* to be processed by the DMAP program) is completely dependent upon
* these values, especially HSCALE.
*
* Parameters:
* @dev pointer to the device structure, needed for
* getting current norm (as well as debug print)
* @width actual image width (from user buffer)
* @height actual image height
* @field indicates Top, Bottom or Interlaced
*/
static int tw68_set_scale(struct tw68_dev *dev, unsigned int width,
unsigned int height, enum v4l2_field field)
{
const struct tw68_tvnorm *norm = dev->tvnorm;
/* set individually for debugging clarity */
int hactive, hdelay, hscale;
int vactive, vdelay, vscale;
int comb;
if (V4L2_FIELD_HAS_BOTH(field)) /* if field is interlaced */
height /= 2; /* we must set for 1-frame */
pr_debug("%s: width=%d, height=%d, both=%d\n"
" tvnorm h_delay=%d, h_start=%d, h_stop=%d, "
"v_delay=%d, v_start=%d, v_stop=%d\n" , __func__,
width, height, V4L2_FIELD_HAS_BOTH(field),
norm->h_delay, norm->h_start, norm->h_stop,
norm->v_delay, norm->video_v_start,
norm->video_v_stop);
switch (dev->vdecoder) {
case TW6800:
hdelay = norm->h_delay0;
break;
default:
hdelay = norm->h_delay;
break;
}
hdelay += norm->h_start;
hactive = norm->h_stop - norm->h_start + 1;
hscale = (hactive * 256) / (width);
vdelay = norm->v_delay;
vactive = ((norm->id & V4L2_STD_525_60) ? 524 : 624) / 2 - norm->video_v_start;
vscale = (vactive * 256) / height;
pr_debug("%s: %dx%d [%s%s,%s]\n", __func__,
width, height,
V4L2_FIELD_HAS_TOP(field) ? "T" : "",
V4L2_FIELD_HAS_BOTTOM(field) ? "B" : "",
v4l2_norm_to_name(dev->tvnorm->id));
pr_debug("%s: hactive=%d, hdelay=%d, hscale=%d; "
"vactive=%d, vdelay=%d, vscale=%d\n", __func__,
hactive, hdelay, hscale, vactive, vdelay, vscale);
comb = ((vdelay & 0x300) >> 2) |
((vactive & 0x300) >> 4) |
((hdelay & 0x300) >> 6) |
((hactive & 0x300) >> 8);
pr_debug("%s: setting CROP_HI=%02x, VDELAY_LO=%02x, "
"VACTIVE_LO=%02x, HDELAY_LO=%02x, HACTIVE_LO=%02x\n",
__func__, comb, vdelay, vactive, hdelay, hactive);
tw_writeb(TW68_CROP_HI, comb);
tw_writeb(TW68_VDELAY_LO, vdelay & 0xff);
tw_writeb(TW68_VACTIVE_LO, vactive & 0xff);
tw_writeb(TW68_HDELAY_LO, hdelay & 0xff);
tw_writeb(TW68_HACTIVE_LO, hactive & 0xff);
comb = ((vscale & 0xf00) >> 4) | ((hscale & 0xf00) >> 8);
pr_debug("%s: setting SCALE_HI=%02x, VSCALE_LO=%02x, "
"HSCALE_LO=%02x\n", __func__, comb, vscale, hscale);
tw_writeb(TW68_SCALE_HI, comb);
tw_writeb(TW68_VSCALE_LO, vscale);
tw_writeb(TW68_HSCALE_LO, hscale);
return 0;
}
/* ------------------------------------------------------------------ */
int tw68_video_start_dma(struct tw68_dev *dev, struct tw68_buf *buf)
{
/* Set cropping and scaling */
tw68_set_scale(dev, dev->width, dev->height, dev->field);
/*
* Set start address for RISC program. Note that if the DMAP
* processor is currently running, it must be stopped before
* a new address can be set.
*/
tw_clearl(TW68_DMAC, TW68_DMAP_EN);
tw_writel(TW68_DMAP_SA, buf->dma);
/* Clear any pending interrupts */
tw_writel(TW68_INTSTAT, dev->board_virqmask);
/* Enable the risc engine and the fifo */
tw_andorl(TW68_DMAC, 0xff, dev->fmt->twformat |
ColorFormatGamma | TW68_DMAP_EN | TW68_FIFO_EN);
dev->pci_irqmask |= dev->board_virqmask;
tw_setl(TW68_INTMASK, dev->pci_irqmask);
return 0;
}
/* ------------------------------------------------------------------ */
/* calc max # of buffers from size (must not exceed the 4MB virtual
* address space per DMA channel) */
static int tw68_buffer_count(unsigned int size, unsigned int count)
{
unsigned int maxcount;
maxcount = (4 * 1024 * 1024) / roundup(size, PAGE_SIZE);
if (count > maxcount)
count = maxcount;
return count;
}
/* ------------------------------------------------------------- */
/* vb2 queue operations */
static int tw68_queue_setup(struct vb2_queue *q, const struct v4l2_format *fmt,
unsigned int *num_buffers, unsigned int *num_planes,
unsigned int sizes[], void *alloc_ctxs[])
{
struct tw68_dev *dev = vb2_get_drv_priv(q);
unsigned tot_bufs = q->num_buffers + *num_buffers;
sizes[0] = (dev->fmt->depth * dev->width * dev->height) >> 3;
alloc_ctxs[0] = dev->alloc_ctx;
/*
* We allow create_bufs, but only if the sizeimage is the same as the
* current sizeimage. The tw68_buffer_count calculation becomes quite
* difficult otherwise.
*/
if (fmt && fmt->fmt.pix.sizeimage < sizes[0])
return -EINVAL;
*num_planes = 1;
if (tot_bufs < 2)
tot_bufs = 2;
tot_bufs = tw68_buffer_count(sizes[0], tot_bufs);
*num_buffers = tot_bufs - q->num_buffers;
return 0;
}
/*
* The risc program for each buffers works as follows: it starts with a simple
* 'JUMP to addr + 8', which is effectively a NOP. Then the program to DMA the
* buffer follows and at the end we have a JUMP back to the start + 8 (skipping
* the initial JUMP).
*
* This is the program of the first buffer to be queued if the active list is
* empty and it just keeps DMAing this buffer without generating any interrupts.
*
* If a new buffer is added then the initial JUMP in the program generates an
* interrupt as well which signals that the previous buffer has been DMAed
* successfully and that it can be returned to userspace.
*
* It also sets the final jump of the previous buffer to the start of the new
* buffer, thus chaining the new buffer into the DMA chain. This is a single
* atomic u32 write, so there is no race condition.
*
* The end-result of all this that you only get an interrupt when a buffer
* is ready, so the control flow is very easy.
*/
static void tw68_buf_queue(struct vb2_buffer *vb)
{
struct vb2_queue *vq = vb->vb2_queue;
struct tw68_dev *dev = vb2_get_drv_priv(vq);
struct tw68_buf *buf = container_of(vb, struct tw68_buf, vb);
struct tw68_buf *prev;
unsigned long flags;
spin_lock_irqsave(&dev->slock, flags);
/* append a 'JUMP to start of buffer' to the buffer risc program */
buf->jmp[0] = cpu_to_le32(RISC_JUMP);
buf->jmp[1] = cpu_to_le32(buf->dma + 8);
if (!list_empty(&dev->active)) {
prev = list_entry(dev->active.prev, struct tw68_buf, list);
buf->cpu[0] |= cpu_to_le32(RISC_INT_BIT);
prev->jmp[1] = cpu_to_le32(buf->dma);
}
list_add_tail(&buf->list, &dev->active);
spin_unlock_irqrestore(&dev->slock, flags);
}
/*
* buffer_prepare
*
* Set the ancilliary information into the buffer structure. This
* includes generating the necessary risc program if it hasn't already
* been done for the current buffer format.
* The structure fh contains the details of the format requested by the
* user - type, width, height and #fields. This is compared with the
* last format set for the current buffer. If they differ, the risc
* code (which controls the filling of the buffer) is (re-)generated.
*/
static int tw68_buf_prepare(struct vb2_buffer *vb)
{
struct vb2_queue *vq = vb->vb2_queue;
struct tw68_dev *dev = vb2_get_drv_priv(vq);
struct tw68_buf *buf = container_of(vb, struct tw68_buf, vb);
struct sg_table *dma = vb2_dma_sg_plane_desc(vb, 0);
unsigned size, bpl;
size = (dev->width * dev->height * dev->fmt->depth) >> 3;
if (vb2_plane_size(vb, 0) < size)
return -EINVAL;
vb2_set_plane_payload(vb, 0, size);
bpl = (dev->width * dev->fmt->depth) >> 3;
switch (dev->field) {
case V4L2_FIELD_TOP:
tw68_risc_buffer(dev->pci, buf, dma->sgl,
0, UNSET, bpl, 0, dev->height);
break;
case V4L2_FIELD_BOTTOM:
tw68_risc_buffer(dev->pci, buf, dma->sgl,
UNSET, 0, bpl, 0, dev->height);
break;
case V4L2_FIELD_SEQ_TB:
tw68_risc_buffer(dev->pci, buf, dma->sgl,
0, bpl * (dev->height >> 1),
bpl, 0, dev->height >> 1);
break;
case V4L2_FIELD_SEQ_BT:
tw68_risc_buffer(dev->pci, buf, dma->sgl,
bpl * (dev->height >> 1), 0,
bpl, 0, dev->height >> 1);
break;
case V4L2_FIELD_INTERLACED:
default:
tw68_risc_buffer(dev->pci, buf, dma->sgl,
0, bpl, bpl, bpl, dev->height >> 1);
break;
}
return 0;
}
static void tw68_buf_finish(struct vb2_buffer *vb)
{
struct vb2_queue *vq = vb->vb2_queue;
struct tw68_dev *dev = vb2_get_drv_priv(vq);
struct tw68_buf *buf = container_of(vb, struct tw68_buf, vb);
pci_free_consistent(dev->pci, buf->size, buf->cpu, buf->dma);
}
static int tw68_start_streaming(struct vb2_queue *q, unsigned int count)
{
struct tw68_dev *dev = vb2_get_drv_priv(q);
struct tw68_buf *buf =
container_of(dev->active.next, struct tw68_buf, list);
dev->seqnr = 0;
tw68_video_start_dma(dev, buf);
return 0;
}
static void tw68_stop_streaming(struct vb2_queue *q)
{
struct tw68_dev *dev = vb2_get_drv_priv(q);
/* Stop risc & fifo */
tw_clearl(TW68_DMAC, TW68_DMAP_EN | TW68_FIFO_EN);
while (!list_empty(&dev->active)) {
struct tw68_buf *buf =
container_of(dev->active.next, struct tw68_buf, list);
list_del(&buf->list);
vb2_buffer_done(&buf->vb, VB2_BUF_STATE_ERROR);
}
}
static struct vb2_ops tw68_video_qops = {
.queue_setup = tw68_queue_setup,
.buf_queue = tw68_buf_queue,
.buf_prepare = tw68_buf_prepare,
.buf_finish = tw68_buf_finish,
.start_streaming = tw68_start_streaming,
.stop_streaming = tw68_stop_streaming,
.wait_prepare = vb2_ops_wait_prepare,
.wait_finish = vb2_ops_wait_finish,
};
/* ------------------------------------------------------------------ */
static int tw68_s_ctrl(struct v4l2_ctrl *ctrl)
{
struct tw68_dev *dev =
container_of(ctrl->handler, struct tw68_dev, hdl);
switch (ctrl->id) {
case V4L2_CID_BRIGHTNESS:
tw_writeb(TW68_BRIGHT, ctrl->val);
break;
case V4L2_CID_HUE:
tw_writeb(TW68_HUE, ctrl->val);
break;
case V4L2_CID_CONTRAST:
tw_writeb(TW68_CONTRAST, ctrl->val);
break;
case V4L2_CID_SATURATION:
tw_writeb(TW68_SAT_U, ctrl->val);
tw_writeb(TW68_SAT_V, ctrl->val);
break;
case V4L2_CID_COLOR_KILLER:
if (ctrl->val)
tw_andorb(TW68_MISC2, 0xe0, 0xe0);
else
tw_andorb(TW68_MISC2, 0xe0, 0x00);
break;
case V4L2_CID_CHROMA_AGC:
if (ctrl->val)
tw_andorb(TW68_LOOP, 0x30, 0x20);
else
tw_andorb(TW68_LOOP, 0x30, 0x00);
break;
}
return 0;
}
/* ------------------------------------------------------------------ */
/*
* Note that this routine returns what is stored in the fh structure, and
* does not interrogate any of the device registers.
*/
static int tw68_g_fmt_vid_cap(struct file *file, void *priv,
struct v4l2_format *f)
{
struct tw68_dev *dev = video_drvdata(file);
f->fmt.pix.width = dev->width;
f->fmt.pix.height = dev->height;
f->fmt.pix.field = dev->field;
f->fmt.pix.pixelformat = dev->fmt->fourcc;
f->fmt.pix.bytesperline =
(f->fmt.pix.width * (dev->fmt->depth)) >> 3;
f->fmt.pix.sizeimage =
f->fmt.pix.height * f->fmt.pix.bytesperline;
f->fmt.pix.colorspace = V4L2_COLORSPACE_SMPTE170M;
f->fmt.pix.priv = 0;
return 0;
}
static int tw68_try_fmt_vid_cap(struct file *file, void *priv,
struct v4l2_format *f)
{
struct tw68_dev *dev = video_drvdata(file);
const struct tw68_format *fmt;
enum v4l2_field field;
unsigned int maxh;
fmt = format_by_fourcc(f->fmt.pix.pixelformat);
if (NULL == fmt)
return -EINVAL;
field = f->fmt.pix.field;
maxh = (dev->tvnorm->id & V4L2_STD_525_60) ? 480 : 576;
switch (field) {
case V4L2_FIELD_TOP:
case V4L2_FIELD_BOTTOM:
break;
case V4L2_FIELD_INTERLACED:
case V4L2_FIELD_SEQ_BT:
case V4L2_FIELD_SEQ_TB:
maxh = maxh * 2;
break;
default:
field = (f->fmt.pix.height > maxh / 2)
? V4L2_FIELD_INTERLACED
: V4L2_FIELD_BOTTOM;
break;
}
f->fmt.pix.field = field;
if (f->fmt.pix.width < 48)
f->fmt.pix.width = 48;
if (f->fmt.pix.height < 32)
f->fmt.pix.height = 32;
if (f->fmt.pix.width > 720)
f->fmt.pix.width = 720;
if (f->fmt.pix.height > maxh)
f->fmt.pix.height = maxh;
f->fmt.pix.width &= ~0x03;
f->fmt.pix.bytesperline =
(f->fmt.pix.width * (fmt->depth)) >> 3;
f->fmt.pix.sizeimage =
f->fmt.pix.height * f->fmt.pix.bytesperline;
f->fmt.pix.colorspace = V4L2_COLORSPACE_SMPTE170M;
return 0;
}
/*
* Note that tw68_s_fmt_vid_cap sets the information into the fh structure,
* and it will be used for all future new buffers. However, there could be
* some number of buffers on the "active" chain which will be filled before
* the change takes place.
*/
static int tw68_s_fmt_vid_cap(struct file *file, void *priv,
struct v4l2_format *f)
{
struct tw68_dev *dev = video_drvdata(file);
int err;
err = tw68_try_fmt_vid_cap(file, priv, f);
if (0 != err)
return err;
dev->fmt = format_by_fourcc(f->fmt.pix.pixelformat);
dev->width = f->fmt.pix.width;
dev->height = f->fmt.pix.height;
dev->field = f->fmt.pix.field;
return 0;
}
static int tw68_enum_input(struct file *file, void *priv,
struct v4l2_input *i)
{
struct tw68_dev *dev = video_drvdata(file);
unsigned int n;
n = i->index;
if (n >= TW68_INPUT_MAX)
return -EINVAL;
i->index = n;
i->type = V4L2_INPUT_TYPE_CAMERA;
snprintf(i->name, sizeof(i->name), "Composite %d", n);
/* If the query is for the current input, get live data */
if (n == dev->input) {
int v1 = tw_readb(TW68_STATUS1);
int v2 = tw_readb(TW68_MVSN);
if (0 != (v1 & (1 << 7)))
i->status |= V4L2_IN_ST_NO_SYNC;
if (0 != (v1 & (1 << 6)))
i->status |= V4L2_IN_ST_NO_H_LOCK;
if (0 != (v1 & (1 << 2)))
i->status |= V4L2_IN_ST_NO_SIGNAL;
if (0 != (v1 & 1 << 1))
i->status |= V4L2_IN_ST_NO_COLOR;
if (0 != (v2 & (1 << 2)))
i->status |= V4L2_IN_ST_MACROVISION;
}
i->std = video_devdata(file)->tvnorms;
return 0;
}
static int tw68_g_input(struct file *file, void *priv, unsigned int *i)
{
struct tw68_dev *dev = video_drvdata(file);
*i = dev->input;
return 0;
}
static int tw68_s_input(struct file *file, void *priv, unsigned int i)
{
struct tw68_dev *dev = video_drvdata(file);
if (i >= TW68_INPUT_MAX)
return -EINVAL;
dev->input = i;
tw_andorb(TW68_INFORM, 0x03 << 2, dev->input << 2);
return 0;
}
static int tw68_querycap(struct file *file, void *priv,
struct v4l2_capability *cap)
{
struct tw68_dev *dev = video_drvdata(file);
strcpy(cap->driver, "tw68");
strlcpy(cap->card, "Techwell Capture Card",
sizeof(cap->card));
sprintf(cap->bus_info, "PCI:%s", pci_name(dev->pci));
cap->device_caps =
V4L2_CAP_VIDEO_CAPTURE |
V4L2_CAP_READWRITE |
V4L2_CAP_STREAMING;
cap->capabilities = cap->device_caps | V4L2_CAP_DEVICE_CAPS;
return 0;
}
static int tw68_s_std(struct file *file, void *priv, v4l2_std_id id)
{
struct tw68_dev *dev = video_drvdata(file);
unsigned int i;
if (vb2_is_busy(&dev->vidq))
return -EBUSY;
/* Look for match on complete norm id (may have mult bits) */
for (i = 0; i < TVNORMS; i++) {
if (id == tvnorms[i].id)
break;
}
/* If no exact match, look for norm which contains this one */
if (i == TVNORMS) {
for (i = 0; i < TVNORMS; i++)
if (id & tvnorms[i].id)
break;
}
/* If still not matched, give up */
if (i == TVNORMS)
return -EINVAL;
set_tvnorm(dev, &tvnorms[i]); /* do the actual setting */
return 0;
}
static int tw68_g_std(struct file *file, void *priv, v4l2_std_id *id)
{
struct tw68_dev *dev = video_drvdata(file);
*id = dev->tvnorm->id;
return 0;
}
static int tw68_enum_fmt_vid_cap(struct file *file, void *priv,
struct v4l2_fmtdesc *f)
{
if (f->index >= FORMATS)
return -EINVAL;
strlcpy(f->description, formats[f->index].name,
sizeof(f->description));
f->pixelformat = formats[f->index].fourcc;
return 0;
}
/*
* Used strictly for internal development and debugging, this routine
* prints out the current register contents for the tw68xx device.
*/
static void tw68_dump_regs(struct tw68_dev *dev)
{
unsigned char line[80];
int i, j, k;
unsigned char *cptr;
pr_info("Full dump of TW68 registers:\n");
/* First we do the PCI regs, 8 4-byte regs per line */
for (i = 0; i < 0x100; i += 32) {
cptr = line;
cptr += sprintf(cptr, "%03x ", i);
/* j steps through the next 4 words */
for (j = i; j < i + 16; j += 4)
cptr += sprintf(cptr, "%08x ", tw_readl(j));
*cptr++ = ' ';
for (; j < i + 32; j += 4)
cptr += sprintf(cptr, "%08x ", tw_readl(j));
*cptr++ = '\n';
*cptr = 0;
pr_info("%s", line);
}
/* Next the control regs, which are single-byte, address mod 4 */
while (i < 0x400) {
cptr = line;
cptr += sprintf(cptr, "%03x ", i);
/* Print out 4 groups of 4 bytes */
for (j = 0; j < 4; j++) {
for (k = 0; k < 4; k++) {
cptr += sprintf(cptr, "%02x ",
tw_readb(i));
i += 4;
}
*cptr++ = ' ';
}
*cptr++ = '\n';
*cptr = 0;
pr_info("%s", line);
}
}
static int vidioc_log_status(struct file *file, void *priv)
{
struct tw68_dev *dev = video_drvdata(file);
tw68_dump_regs(dev);
return v4l2_ctrl_log_status(file, priv);
}
#ifdef CONFIG_VIDEO_ADV_DEBUG
static int vidioc_g_register(struct file *file, void *priv,
struct v4l2_dbg_register *reg)
{
struct tw68_dev *dev = video_drvdata(file);
if (reg->size == 1)
reg->val = tw_readb(reg->reg);
else
reg->val = tw_readl(reg->reg);
return 0;
}
static int vidioc_s_register(struct file *file, void *priv,
const struct v4l2_dbg_register *reg)
{
struct tw68_dev *dev = video_drvdata(file);
if (reg->size == 1)
tw_writeb(reg->reg, reg->val);
else
tw_writel(reg->reg & 0xffff, reg->val);
return 0;
}
#endif
static const struct v4l2_ctrl_ops tw68_ctrl_ops = {
.s_ctrl = tw68_s_ctrl,
};
static const struct v4l2_file_operations video_fops = {
.owner = THIS_MODULE,
.open = v4l2_fh_open,
.release = vb2_fop_release,
.read = vb2_fop_read,
.poll = vb2_fop_poll,
.mmap = vb2_fop_mmap,
.unlocked_ioctl = video_ioctl2,
};
static const struct v4l2_ioctl_ops video_ioctl_ops = {
.vidioc_querycap = tw68_querycap,
.vidioc_enum_fmt_vid_cap = tw68_enum_fmt_vid_cap,
.vidioc_reqbufs = vb2_ioctl_reqbufs,
.vidioc_create_bufs = vb2_ioctl_create_bufs,
.vidioc_querybuf = vb2_ioctl_querybuf,
.vidioc_qbuf = vb2_ioctl_qbuf,
.vidioc_dqbuf = vb2_ioctl_dqbuf,
.vidioc_s_std = tw68_s_std,
.vidioc_g_std = tw68_g_std,
.vidioc_enum_input = tw68_enum_input,
.vidioc_g_input = tw68_g_input,
.vidioc_s_input = tw68_s_input,
.vidioc_streamon = vb2_ioctl_streamon,
.vidioc_streamoff = vb2_ioctl_streamoff,
.vidioc_g_fmt_vid_cap = tw68_g_fmt_vid_cap,
.vidioc_try_fmt_vid_cap = tw68_try_fmt_vid_cap,
.vidioc_s_fmt_vid_cap = tw68_s_fmt_vid_cap,
.vidioc_log_status = vidioc_log_status,
.vidioc_subscribe_event = v4l2_ctrl_subscribe_event,
.vidioc_unsubscribe_event = v4l2_event_unsubscribe,
#ifdef CONFIG_VIDEO_ADV_DEBUG
.vidioc_g_register = vidioc_g_register,
.vidioc_s_register = vidioc_s_register,
#endif
};
static struct video_device tw68_video_template = {
.name = "tw68_video",
.fops = &video_fops,
.ioctl_ops = &video_ioctl_ops,
.release = video_device_release_empty,
.tvnorms = TW68_NORMS,
};
/* ------------------------------------------------------------------ */
/* exported stuff */
void tw68_set_tvnorm_hw(struct tw68_dev *dev)
{
tw_andorb(TW68_SDT, 0x07, dev->tvnorm->format);
}
int tw68_video_init1(struct tw68_dev *dev)
{
struct v4l2_ctrl_handler *hdl = &dev->hdl;
v4l2_ctrl_handler_init(hdl, 6);
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_BRIGHTNESS, -128, 127, 1, 20);
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_CONTRAST, 0, 255, 1, 100);
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_SATURATION, 0, 255, 1, 128);
/* NTSC only */
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_HUE, -128, 127, 1, 0);
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_COLOR_KILLER, 0, 1, 1, 0);
v4l2_ctrl_new_std(hdl, &tw68_ctrl_ops,
V4L2_CID_CHROMA_AGC, 0, 1, 1, 1);
if (hdl->error) {
v4l2_ctrl_handler_free(hdl);
return hdl->error;
}
dev->v4l2_dev.ctrl_handler = hdl;
v4l2_ctrl_handler_setup(hdl);
return 0;
}
int tw68_video_init2(struct tw68_dev *dev, int video_nr)
{
int ret;
set_tvnorm(dev, &tvnorms[0]);
dev->fmt = format_by_fourcc(V4L2_PIX_FMT_BGR24);
dev->width = 720;
dev->height = 576;
dev->field = V4L2_FIELD_INTERLACED;
INIT_LIST_HEAD(&dev->active);
dev->vidq.type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
dev->vidq.timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
dev->vidq.io_modes = VB2_MMAP | VB2_USERPTR | VB2_READ | VB2_DMABUF;
dev->vidq.ops = &tw68_video_qops;
dev->vidq.mem_ops = &vb2_dma_sg_memops;
dev->vidq.drv_priv = dev;
dev->vidq.gfp_flags = __GFP_DMA32;
dev->vidq.buf_struct_size = sizeof(struct tw68_buf);
dev->vidq.lock = &dev->lock;
dev->vidq.min_buffers_needed = 2;
ret = vb2_queue_init(&dev->vidq);
if (ret)
return ret;
dev->vdev = tw68_video_template;
dev->vdev.v4l2_dev = &dev->v4l2_dev;
dev->vdev.lock = &dev->lock;
dev->vdev.queue = &dev->vidq;
video_set_drvdata(&dev->vdev, dev);
return video_register_device(&dev->vdev, VFL_TYPE_GRABBER, video_nr);
}
/*
* tw68_irq_video_done
*/
void tw68_irq_video_done(struct tw68_dev *dev, unsigned long status)
{
__u32 reg;
/* reset interrupts handled by this routine */
tw_writel(TW68_INTSTAT, status);
/*
* Check most likely first
*
* DMAPI shows we have reached the end of the risc code
* for the current buffer.
*/
if (status & TW68_DMAPI) {
struct tw68_buf *buf;
spin_lock(&dev->slock);
buf = list_entry(dev->active.next, struct tw68_buf, list);
list_del(&buf->list);
spin_unlock(&dev->slock);
v4l2_get_timestamp(&buf->vb.v4l2_buf.timestamp);
buf->vb.v4l2_buf.field = dev->field;
buf->vb.v4l2_buf.sequence = dev->seqnr++;
vb2_buffer_done(&buf->vb, VB2_BUF_STATE_DONE);
status &= ~(TW68_DMAPI);
if (0 == status)
return;
}
if (status & (TW68_VLOCK | TW68_HLOCK))
dev_dbg(&dev->pci->dev, "Lost sync\n");
if (status & TW68_PABORT)
dev_err(&dev->pci->dev, "PABORT interrupt\n");
if (status & TW68_DMAPERR)
dev_err(&dev->pci->dev, "DMAPERR interrupt\n");
/*
* On TW6800, FDMIS is apparently generated if video input is switched
* during operation. Therefore, it is not enabled for that chip.
*/
if (status & TW68_FDMIS)
dev_dbg(&dev->pci->dev, "FDMIS interrupt\n");
if (status & TW68_FFOF) {
/* probably a logic error */
reg = tw_readl(TW68_DMAC) & TW68_FIFO_EN;
tw_clearl(TW68_DMAC, TW68_FIFO_EN);
dev_dbg(&dev->pci->dev, "FFOF interrupt\n");
tw_setl(TW68_DMAC, reg);
}
if (status & TW68_FFERR)
dev_dbg(&dev->pci->dev, "FFERR interrupt\n");
}