[media] pci: Add tw5864 driver

Support for boards based on Techwell TW5864 chip which provides
multichannel video & audio grabbing and encoding (H.264, MJPEG,
ADPCM G.726).

This submission implements only H.264 encoding of all channels at D1
resolution.

Thanks to Mark Thompson <sw@jkqxz.net> for help, and for contribution of
H.264 startcode emulation prevention code.
Signed-off-by: Andrey Utkin <andrey.utkin@corp.bluecherry.net>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>

MAINTAINERS

@@ -11839,6 +11839,14 @@ T:	git git://linuxtv.org/media_tree.git
 S:	Odd fixes
 F:	drivers/media/usb/tm6000/
 
+TW5864 VIDEO4LINUX DRIVER
+M:	Bluecherry Maintainers <maintainers@bluecherrydvr.com>
+M:	Andrey Utkin <andrey.utkin@corp.bluecherry.net>
+M:	Andrey Utkin <andrey_utkin@fastmail.com>
+L:	linux-media@vger.kernel.org
+S:	Supported
+F:	drivers/media/pci/tw5864/
+
 TW68 VIDEO4LINUX DRIVER
 M:	Hans Verkuil <hverkuil@xs4all.nl>
 L:	linux-media@vger.kernel.org

drivers/media/pci/Kconfig

@@ -13,6 +13,7 @@ if MEDIA_CAMERA_SUPPORT
 source "drivers/media/pci/meye/Kconfig"
 source "drivers/media/pci/solo6x10/Kconfig"
 source "drivers/media/pci/sta2x11/Kconfig"
+source "drivers/media/pci/tw5864/Kconfig"
 source "drivers/media/pci/tw68/Kconfig"
 source "drivers/media/pci/tw686x/Kconfig"
 source "drivers/media/pci/zoran/Kconfig"

drivers/media/pci/Makefile

@@ -31,3 +31,4 @@ obj-$(CONFIG_VIDEO_MEYE) += meye/
 obj-$(CONFIG_STA2X11_VIP) += sta2x11/
 obj-$(CONFIG_VIDEO_SOLO6X10) += solo6x10/
 obj-$(CONFIG_VIDEO_COBALT) += cobalt/
+obj-$(CONFIG_VIDEO_TW5864) += tw5864/

drivers/media/pci/tw5864/Kconfig

+config VIDEO_TW5864
+	tristate "Techwell TW5864 video/audio grabber and encoder"
+	depends on VIDEO_DEV && PCI && VIDEO_V4L2
+	select VIDEOBUF2_DMA_CONTIG
+	---help---
+	  Support for boards based on Techwell TW5864 chip which provides
+	  multichannel video & audio grabbing and encoding (H.264, MJPEG,
+	  ADPCM G.726).
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called tw5864.

drivers/media/pci/tw5864/Makefile

+tw5864-objs := tw5864-core.o tw5864-video.o tw5864-h264.o tw5864-util.o
+
+obj-$(CONFIG_VIDEO_TW5864) += tw5864.o

drivers/media/pci/tw5864/tw5864-h264.c

+/*
+ *  TW5864 driver - H.264 headers generation functions
+ *
+ *  Copyright (C) 2016 Bluecherry, LLC <maintainers@bluecherrydvr.com>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 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/log2.h>
+
+#include "tw5864.h"
+
+static u8 marker[] = { 0x00, 0x00, 0x00, 0x01 };
+
+/*
+ * Exponential-Golomb coding functions
+ *
+ * These functions are used for generation of H.264 bitstream headers.
+ *
+ * This code is derived from tw5864 reference driver by manufacturers, which
+ * itself apparently was derived from x264 project.
+ */
+
+/* Bitstream writing context */
+struct bs {
+	u8 *buf; /* pointer to buffer beginning */
+	u8 *buf_end; /* pointer to buffer end */
+	u8 *ptr; /* pointer to current byte in buffer */
+	unsigned int bits_left; /* number of available bits in current byte */
+};
+
+static void bs_init(struct bs *s, void *buf, int size)
+{
+	s->buf = buf;
+	s->ptr = buf;
+	s->buf_end = s->ptr + size;
+	s->bits_left = 8;
+}
+
+static int bs_len(struct bs *s)
+{
+	return s->ptr - s->buf;
+}
+
+static void bs_write(struct bs *s, int count, u32 bits)
+{
+	if (s->ptr >= s->buf_end - 4)
+		return;
+	while (count > 0) {
+		if (count < 32)
+			bits &= (1 << count) - 1;
+		if (count < s->bits_left) {
+			*s->ptr = (*s->ptr << count) | bits;
+			s->bits_left -= count;
+			break;
+		}
+		*s->ptr = (*s->ptr << s->bits_left) |
+			(bits >> (count - s->bits_left));
+		count -= s->bits_left;
+		s->ptr++;
+		s->bits_left = 8;
+	}
+}
+
+static void bs_write1(struct bs *s, u32 bit)
+{
+	if (s->ptr < s->buf_end) {
+		*s->ptr <<= 1;
+		*s->ptr |= bit;
+		s->bits_left--;
+		if (s->bits_left == 0) {
+			s->ptr++;
+			s->bits_left = 8;
+		}
+	}
+}
+
+static void bs_write_ue(struct bs *s, u32 val)
+{
+	if (val == 0) {
+		bs_write1(s, 1);
+	} else {
+		val++;
+		bs_write(s, 2 * fls(val) - 1, val);
+	}
+}
+
+static void bs_write_se(struct bs *s, int val)
+{
+	bs_write_ue(s, val <= 0 ? -val * 2 : val * 2 - 1);
+}
+
+static void bs_rbsp_trailing(struct bs *s)
+{
+	bs_write1(s, 1);
+	if (s->bits_left != 8)
+		bs_write(s, s->bits_left, 0x00);
+}
+
+/* H.264 headers generation functions */
+
+static int tw5864_h264_gen_sps_rbsp(u8 *buf, size_t size, int width, int height)
+{
+	struct bs bs, *s;
+
+	s = &bs;
+	bs_init(s, buf, size);
+	bs_write(s, 8, 0x42); /* profile_idc, baseline */
+	bs_write(s, 1, 1); /* constraint_set0_flag */
+	bs_write(s, 1, 1); /* constraint_set1_flag */
+	bs_write(s, 1, 0); /* constraint_set2_flag */
+	bs_write(s, 5, 0); /* reserved_zero_5bits */
+	bs_write(s, 8, 0x1e); /* level_idc */
+	bs_write_ue(s, 0); /* seq_parameter_set_id */
+	bs_write_ue(s, ilog2(MAX_GOP_SIZE) - 4); /* log2_max_frame_num_minus4 */
+	bs_write_ue(s, 0); /* pic_order_cnt_type */
+	/* log2_max_pic_order_cnt_lsb_minus4 */
+	bs_write_ue(s, ilog2(MAX_GOP_SIZE) - 4);
+	bs_write_ue(s, 1); /* num_ref_frames */
+	bs_write(s, 1, 0); /* gaps_in_frame_num_value_allowed_flag */
+	bs_write_ue(s, width / 16 - 1); /* pic_width_in_mbs_minus1 */
+	bs_write_ue(s, height / 16 - 1); /* pic_height_in_map_units_minus1 */
+	bs_write(s, 1, 1); /* frame_mbs_only_flag */
+	bs_write(s, 1, 0); /* direct_8x8_inference_flag */
+	bs_write(s, 1, 0); /* frame_cropping_flag */
+	bs_write(s, 1, 0); /* vui_parameters_present_flag */
+	bs_rbsp_trailing(s);
+	return bs_len(s);
+}
+
+static int tw5864_h264_gen_pps_rbsp(u8 *buf, size_t size, int qp)
+{
+	struct bs bs, *s;
+
+	s = &bs;
+	bs_init(s, buf, size);
+	bs_write_ue(s, 0); /* pic_parameter_set_id */
+	bs_write_ue(s, 0); /* seq_parameter_set_id */
+	bs_write(s, 1, 0); /* entropy_coding_mode_flag */
+	bs_write(s, 1, 0); /* pic_order_present_flag */
+	bs_write_ue(s, 0); /* num_slice_groups_minus1 */
+	bs_write_ue(s, 0); /* i_num_ref_idx_l0_active_minus1 */
+	bs_write_ue(s, 0); /* i_num_ref_idx_l1_active_minus1 */
+	bs_write(s, 1, 0); /* weighted_pred_flag */
+	bs_write(s, 2, 0); /* weighted_bipred_idc */
+	bs_write_se(s, qp - 26); /* pic_init_qp_minus26 */
+	bs_write_se(s, qp - 26); /* pic_init_qs_minus26 */
+	bs_write_se(s, 0); /* chroma_qp_index_offset */
+	bs_write(s, 1, 0); /* deblocking_filter_control_present_flag */
+	bs_write(s, 1, 0); /* constrained_intra_pred_flag */
+	bs_write(s, 1, 0); /* redundant_pic_cnt_present_flag */
+	bs_rbsp_trailing(s);
+	return bs_len(s);
+}
+
+static int tw5864_h264_gen_slice_head(u8 *buf, size_t size,
+				      unsigned int idr_pic_id,
+				      unsigned int frame_gop_seqno,
+				      int *tail_nb_bits, u8 *tail)
+{
+	struct bs bs, *s;
+	int is_i_frame = frame_gop_seqno == 0;
+
+	s = &bs;
+	bs_init(s, buf, size);
+	bs_write_ue(s, 0); /* first_mb_in_slice */
+	bs_write_ue(s, is_i_frame ? 2 : 5); /* slice_type - I or P */
+	bs_write_ue(s, 0); /* pic_parameter_set_id */
+	bs_write(s, ilog2(MAX_GOP_SIZE), frame_gop_seqno); /* frame_num */
+	if (is_i_frame)
+		bs_write_ue(s, idr_pic_id);
+
+	/* pic_order_cnt_lsb */
+	bs_write(s, ilog2(MAX_GOP_SIZE), frame_gop_seqno);
+
+	if (is_i_frame) {
+		bs_write1(s, 0); /* no_output_of_prior_pics_flag */
+		bs_write1(s, 0); /* long_term_reference_flag */
+	} else {
+		bs_write1(s, 0); /* num_ref_idx_active_override_flag */
+		bs_write1(s, 0); /* ref_pic_list_reordering_flag_l0 */
+		bs_write1(s, 0); /* adaptive_ref_pic_marking_mode_flag */
+	}
+
+	bs_write_se(s, 0); /* slice_qp_delta */
+
+	if (s->bits_left != 8) {
+		*tail = ((s->ptr[0]) << s->bits_left);
+		*tail_nb_bits = 8 - s->bits_left;
+	} else {
+		*tail = 0;
+		*tail_nb_bits = 0;
+	}
+
+	return bs_len(s);
+}
+
+void tw5864_h264_put_stream_header(u8 **buf, size_t *space_left, int qp,
+				   int width, int height)
+{
+	int nal_len;
+
+	/* SPS */
+	memcpy(*buf, marker, sizeof(marker));
+	*buf += 4;
+	*space_left -= 4;
+
+	**buf = 0x67; /* SPS NAL header */
+	*buf += 1;
+	*space_left -= 1;
+
+	nal_len = tw5864_h264_gen_sps_rbsp(*buf, *space_left, width, height);
+	*buf += nal_len;
+	*space_left -= nal_len;
+
+	/* PPS */
+	memcpy(*buf, marker, sizeof(marker));
+	*buf += 4;
+	*space_left -= 4;
+
+	**buf = 0x68; /* PPS NAL header */
+	*buf += 1;
+	*space_left -= 1;
+
+	nal_len = tw5864_h264_gen_pps_rbsp(*buf, *space_left, qp);
+	*buf += nal_len;
+	*space_left -= nal_len;
+}
+
+void tw5864_h264_put_slice_header(u8 **buf, size_t *space_left,
+				  unsigned int idr_pic_id,
+				  unsigned int frame_gop_seqno,
+				  int *tail_nb_bits, u8 *tail)
+{
+	int nal_len;
+
+	memcpy(*buf, marker, sizeof(marker));
+	*buf += 4;
+	*space_left -= 4;
+
+	/* Frame NAL header */
+	**buf = (frame_gop_seqno == 0) ? 0x25 : 0x21;
+	*buf += 1;
+	*space_left -= 1;
+
+	nal_len = tw5864_h264_gen_slice_head(*buf, *space_left, idr_pic_id,
+					     frame_gop_seqno, tail_nb_bits,
+					     tail);
+	*buf += nal_len;
+	*space_left -= nal_len;
+}

drivers/media/pci/tw5864/tw5864-util.c

+#include "tw5864.h"
+
+void tw5864_indir_writeb(struct tw5864_dev *dev, u16 addr, u8 data)
+{
+	int retries = 30000;
+
+	while (tw_readl(TW5864_IND_CTL) & BIT(31) && --retries)
+		;
+	if (!retries)
+		dev_err(&dev->pci->dev,
+			"tw_indir_writel() retries exhausted before writing\n");
+
+	tw_writel(TW5864_IND_DATA, data);
+	tw_writel(TW5864_IND_CTL, addr << 2 | TW5864_RW | TW5864_ENABLE);
+}
+
+u8 tw5864_indir_readb(struct tw5864_dev *dev, u16 addr)
+{
+	int retries = 30000;
+
+	while (tw_readl(TW5864_IND_CTL) & BIT(31) && --retries)
+		;
+	if (!retries)
+		dev_err(&dev->pci->dev,
+			"tw_indir_readl() retries exhausted before reading\n");
+
+	tw_writel(TW5864_IND_CTL, addr << 2 | TW5864_ENABLE);
+
+	retries = 30000;
+	while (tw_readl(TW5864_IND_CTL) & BIT(31) && --retries)
+		;
+	if (!retries)
+		dev_err(&dev->pci->dev,
+			"tw_indir_readl() retries exhausted at reading\n");
+
+	return tw_readl(TW5864_IND_DATA);
+}

drivers/media/pci/tw5864/tw5864.h

+/*
+ *  TW5864 driver  - common header file
+ *
+ *  Copyright (C) 2016 Bluecherry, LLC <maintainers@bluecherrydvr.com>
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 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/pci.h>
+#include <linux/videodev2.h>
+#include <linux/notifier.h>
+#include <linux/delay.h>
+#include <linux/mutex.h>
+#include <linux/io.h>
+#include <linux/interrupt.h>
+
+#include <media/v4l2-common.h>
+#include <media/v4l2-ioctl.h>
+#include <media/v4l2-ctrls.h>
+#include <media/v4l2-device.h>
+#include <media/videobuf2-dma-sg.h>
+
+#include "tw5864-reg.h"
+
+#define PCI_DEVICE_ID_TECHWELL_5864 0x5864
+
+#define TW5864_NORMS V4L2_STD_ALL
+
+/* ----------------------------------------------------------- */
+/* card configuration   */
+
+#define TW5864_INPUTS 4
+
+/* The TW5864 uses 192 (16x12) detection cells in full screen for motion
+ * detection. Each detection cell is composed of 44 pixels and 20 lines for
+ * NTSC and 24 lines for PAL.
+ */
+#define MD_CELLS_HOR 16
+#define MD_CELLS_VERT 12
+#define MD_CELLS (MD_CELLS_HOR * MD_CELLS_VERT)
+
+#define H264_VLC_BUF_SIZE 0x80000
+#define H264_MV_BUF_SIZE 0x2000 /* device writes 5396 bytes */
+#define QP_VALUE 28
+#define MAX_GOP_SIZE 255
+#define GOP_SIZE MAX_GOP_SIZE
+
+enum resolution {
+	D1 = 1,
+	HD1 = 2, /* half d1 - 360x(240|288) */
+	CIF = 3,
+	QCIF = 4,
+};
+
+/* ----------------------------------------------------------- */
+/* device / file handle status                                 */
+
+struct tw5864_dev; /* forward delclaration */
+
+/* buffer for one video/vbi/ts frame */
+struct tw5864_buf {
+	struct vb2_v4l2_buffer vb;
+	struct list_head list;
+
+	unsigned int size;
+};
+
+struct tw5864_dma_buf {
+	void *addr;
+	dma_addr_t dma_addr;
+};
+
+enum tw5864_vid_std {
+	STD_NTSC = 0, /* NTSC (M) */
+	STD_PAL = 1, /* PAL (B, D, G, H, I) */
+	STD_SECAM = 2, /* SECAM */
+	STD_NTSC443 = 3, /* NTSC4.43 */
+	STD_PAL_M = 4, /* PAL (M) */
+	STD_PAL_CN = 5, /* PAL (CN) */
+	STD_PAL_60 = 6, /* PAL 60 */
+	STD_INVALID = 7,
+	STD_AUTO = 7,
+};
+
+struct tw5864_input {
+	int nr; /* input number */
+	struct tw5864_dev *root;
+	struct mutex lock; /* used for vidq and vdev */
+	spinlock_t slock; /* used for sync between ISR, tasklet & V4L2 API */
+	struct video_device vdev;
+	struct v4l2_ctrl_handler hdl;
+	struct vb2_queue vidq;
+	struct list_head active;
+	enum resolution resolution;
+	unsigned int width, height;
+	unsigned int frame_seqno;
+	unsigned int frame_gop_seqno;
+	unsigned int h264_idr_pic_id;
+	int enabled;
+	enum tw5864_vid_std std;
+	v4l2_std_id v4l2_std;
+	int tail_nb_bits;
+	u8 tail;
+	u8 *buf_cur_ptr;
+	int buf_cur_space_left;
+
+	u32 reg_interlacing;
+	u32 reg_vlc;
+	u32 reg_dsp_codec;
+	u32 reg_dsp;
+	u32 reg_emu;
+	u32 reg_dsp_qp;
+	u32 reg_dsp_ref_mvp_lambda;
+	u32 reg_dsp_i4x4_weight;
+	u32 buf_id;
+
+	struct tw5864_buf *vb;
+
+	struct v4l2_ctrl *md_threshold_grid_ctrl;
+	u16 md_threshold_grid_values[12 * 16];
+	int qp;
+	int gop;
+
+	/*
+	 * In (1/MAX_FPS) units.
+	 * For max FPS (default), set to 1.
+	 * For 1 FPS, set to e.g. 32.
+	 */
+	int frame_interval;
+	unsigned long new_frame_deadline;
+};
+
+struct tw5864_h264_frame {
+	struct tw5864_dma_buf vlc;
+	struct tw5864_dma_buf mv;
+	int vlc_len;
+	u32 checksum;
+	struct tw5864_input *input;
+	u64 timestamp;
+	unsigned int seqno;
+	unsigned int gop_seqno;
+};
+
+/* global device status */
+struct tw5864_dev {
+	spinlock_t slock; /* used for sync between ISR, tasklet & V4L2 API */
+	struct v4l2_device v4l2_dev;
+	struct tw5864_input inputs[TW5864_INPUTS];
+#define H264_BUF_CNT 4
+	struct tw5864_h264_frame h264_buf[H264_BUF_CNT];
+	int h264_buf_r_index;
+	int h264_buf_w_index;
+
+	struct tasklet_struct tasklet;
+
+	int encoder_busy;
+	/* Input number to check next for ready raw picture (in RR fashion) */
+	int next_input;
+
+	/* pci i/o */
+	char name[64];
+	struct pci_dev *pci;
+	void __iomem *mmio;
+	u32 irqmask;
+};
+
+#define tw_readl(reg) readl(dev->mmio + reg)
+#define tw_mask_readl(reg, mask) \
+	(tw_readl(reg) & (mask))
+#define tw_mask_shift_readl(reg, mask, shift) \
+	(tw_mask_readl((reg), ((mask) << (shift))) >> (shift))
+
+#define tw_writel(reg, value) writel((value), dev->mmio + reg)
+#define tw_mask_writel(reg, mask, value) \
+	tw_writel(reg, (tw_readl(reg) & ~(mask)) | ((value) & (mask)))
+#define tw_mask_shift_writel(reg, mask, shift, value) \
+	tw_mask_writel((reg), ((mask) << (shift)), ((value) << (shift)))
+
+#define tw_setl(reg, bit) tw_writel((reg), tw_readl(reg) | (bit))
+#define tw_clearl(reg, bit) tw_writel((reg), tw_readl(reg) & ~(bit))
+
+u8 tw5864_indir_readb(struct tw5864_dev *dev, u16 addr);
+#define tw_indir_readb(addr) tw5864_indir_readb(dev, addr)
+void tw5864_indir_writeb(struct tw5864_dev *dev, u16 addr, u8 data);
+#define tw_indir_writeb(addr, data) tw5864_indir_writeb(dev, addr, data)
+
+void tw5864_irqmask_apply(struct tw5864_dev *dev);
+int tw5864_video_init(struct tw5864_dev *dev, int *video_nr);
+void tw5864_video_fini(struct tw5864_dev *dev);
+void tw5864_prepare_frame_headers(struct tw5864_input *input);
+void tw5864_h264_put_stream_header(u8 **buf, size_t *space_left, int qp,
+				   int width, int height);
+void tw5864_h264_put_slice_header(u8 **buf, size_t *space_left,
+				  unsigned int idr_pic_id,
+				  unsigned int frame_gop_seqno,
+				  int *tail_nb_bits, u8 *tail);
+void tw5864_request_encoded_frame(struct tw5864_input *input);