[CODA] Info about internal buffers in CodaDX6

[CODA] Info about internal buffers in CodaDX6

[Gaëtan Carlier]

[-- Attachment #1: Type: text/plain, Size: 843 bytes --]

Dear,
I am back to add support of h.264 decoding using Coda DX6 on i.MX27 
(after long months of inactivity).
I base my work on driver from linux 2.6.22 (libvpu) and last coda.c from 
linux-next/master.

When I send DEC_SEQ_INIT command, it fails but I don't know why.
1) Which internal buffers do Coda DX6 really have/used for decoding 
PARABUF, WORKBUF, PSBUF, ...) ?
2) What is their role ?
3) I see in some code that there is a command 
CODA_RET_DEC_SEQ_ERR_REASON (0x1E0), which has the same opcode has 
RET_DEC_SEQ_NEXT_FRAME_NUM, but when I run this command after 
DEC_SEQ_INIT, it returns 1 that does not seems to be correct error 
(regarding RetCode enum in vpu_lib.h in libvpu)

Code is based on 3.6.0 kernel revision with some backport from more 
recent version of coda.c

Thanks a lot for your help.
Best regards,
Gaëtan Carlier.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: coda.patch --]
[-- Type: text/x-patch; name="coda.patch", Size: 78670 bytes --]

--- /dev/null	2014-06-03 09:49:18.243206630 +0200
+++ coda.h	2014-06-03 10:11:53.536413773 +0200
@@ -0,0 +1,285 @@
+/*
+ * linux/drivers/media/platform/coda/coda_regs.h
+ *
+ * Copyright (C) 2012 Vista Silicon SL
+ *    Javier Martin <javier.martin@vista-silicon.com>
+ *    Xavier Duret
+ *
+ * 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.
+ */
+
+#ifndef _REGS_CODA_H_
+#define _REGS_CODA_H_
+
+/* HW registers */
+#define CODA_REG_BIT_CODE_RUN			0x000
+#define		CODA_REG_RUN_ENABLE		(1 << 0)
+#define CODA_REG_BIT_CODE_DOWN			0x004
+#define		CODA_DOWN_ADDRESS_SET(x)	(((x) & 0xffff) << 16)
+#define		CODA_DOWN_DATA_SET(x)		((x) & 0xffff)
+#define CODA_REG_BIT_HOST_IN_REQ		0x008
+#define CODA_REG_BIT_INT_CLEAR			0x00c
+#define		CODA_REG_BIT_INT_CLEAR_SET	0x1
+#define CODA_REG_BIT_INT_STATUS		0x010
+#define CODA_REG_BIT_CODE_RESET		0x014
+#define		CODA_REG_RESET_ENABLE		(1 << 0)
+#define CODA_REG_BIT_CUR_PC			0x018
+
+/* Static SW registers */
+#define CODA_REG_BIT_CODE_BUF_ADDR		0x100
+#define CODA_REG_BIT_WORK_BUF_ADDR		0x104
+#define CODA_REG_BIT_PARA_BUF_ADDR		0x108
+#define CODA_REG_BIT_STREAM_CTRL		0x10c
+#define		CODA7_STREAM_BUF_PIC_RESET	(1 << 4)
+#define		CODADX6_STREAM_BUF_PIC_RESET	(1 << 3)
+#define		CODA7_STREAM_BUF_PIC_FLUSH	(1 << 3)
+#define		CODADX6_STREAM_BUF_PIC_FLUSH	(1 << 2)
+#define		CODA7_STREAM_BUF_DYNALLOC_EN	(1 << 5)
+#define		CODADX6_STREAM_BUF_DYNALLOC_EN	(1 << 4)
+#define 	CODA_STREAM_CHKDIS_OFFSET	(1 << 1)
+#define		CODA_STREAM_ENDIAN_SELECT	(1 << 0)
+#define CODA_REG_BIT_FRAME_MEM_CTRL		0x110
+#define		CODA_IMAGE_ENDIAN_SELECT	(1 << 0)
+#define CODA_REG_BIT_RD_PTR(x)			(0x120 + 8 * (x))
+#define CODA_REG_BIT_WR_PTR(x)			(0x124 + 8 * (x))
+#define CODADX6_REG_BIT_SEARCH_RAM_BASE_ADDR	0x140
+#define CODA7_REG_BIT_AXI_SRAM_USE		0x140
+#define		CODA7_USE_BIT_ENABLE		(1 << 0)
+#define		CODA7_USE_HOST_BIT_ENABLE	(1 << 7)
+#define		CODA7_USE_ME_ENABLE		(1 << 4)
+#define		CODA7_USE_HOST_ME_ENABLE	(1 << 11)
+#define CODA_REG_BIT_BUSY			0x160
+#define		CODA_REG_BIT_BUSY_FLAG		1
+#define CODA_REG_BIT_RUN_COMMAND		0x164
+#define		CODA_COMMAND_SEQ_INIT		1
+#define		CODA_COMMAND_SEQ_END		2
+#define		CODA_COMMAND_PIC_RUN		3
+#define		CODA_COMMAND_SET_FRAME_BUF	4
+#define		CODA_COMMAND_ENCODE_HEADER	5
+#define		CODA_COMMAND_ENC_PARA_SET	6
+#define		CODA_COMMAND_DEC_PARA_SET	7
+#define		CODA_COMMAND_DEC_BUF_FLUSH	8
+#define		CODA_COMMAND_RC_CHANGE_PARAMETER 9
+#define		CODA_COMMAND_FIRMWARE_GET	0xf
+#define CODA_REG_BIT_RUN_INDEX			0x168
+#define		CODA_INDEX_SET(x)		((x) & 0x3)
+#define CODA_REG_BIT_RUN_COD_STD		0x16c
+#define		CODADX6_MODE_DECODE_MP4		0
+#define		CODADX6_MODE_ENCODE_MP4		1
+#define		CODADX6_MODE_DECODE_H264	2
+#define		CODADX6_MODE_ENCODE_H264	3
+#define		CODA7_MODE_DECODE_H264		0
+#define		CODA7_MODE_DECODE_VC1		1
+#define		CODA7_MODE_DECODE_MP2		2
+#define		CODA7_MODE_DECODE_MP4		3
+#define		CODA7_MODE_DECODE_DV3		3
+#define		CODA7_MODE_DECODE_RV		4
+#define		CODA7_MODE_DECODE_MJPG		5
+#define		CODA7_MODE_ENCODE_H264		8
+#define		CODA7_MODE_ENCODE_MP4		11
+#define		CODA7_MODE_ENCODE_MJPG		13
+#define 	CODA_MODE_INVALID		0xffff
+#define CODA_REG_BIT_INT_ENABLE		0x170
+#define		CODA_INT_INTERRUPT_ENABLE	(1 << 3)
+
+/*
+ * Commands' mailbox:
+ * registers with offsets in the range 0x180-0x1d0
+ * have different meaning depending on the command being
+ * issued.
+ */
+
+/* Encoder Sequence Initialization */
+#define CODA_CMD_ENC_SEQ_BB_START				0x180
+#define CODA_CMD_ENC_SEQ_BB_SIZE				0x184
+#define CODA_CMD_ENC_SEQ_OPTION				0x188
+#define		CODA7_OPTION_GAMMA_OFFSET			8
+#define		CODADX6_OPTION_GAMMA_OFFSET			7
+#define		CODA_OPTION_LIMITQP_OFFSET			6
+#define		CODA_OPTION_RCINTRAQP_OFFSET			5
+#define		CODA_OPTION_FMO_OFFSET				4
+#define		CODA_OPTION_SLICEREPORT_OFFSET			1
+#define CODA_CMD_ENC_SEQ_COD_STD				0x18c
+#define		CODA_STD_MPEG4					0
+#define		CODA_STD_H263					1
+#define		CODA_STD_H264					2
+#define		CODA_STD_MJPG					3
+#define CODA_CMD_ENC_SEQ_SRC_SIZE				0x190
+#define		CODA7_PICWIDTH_OFFSET				16
+#define		CODA7_PICWIDTH_MASK				0xffff
+#define		CODADX6_PICWIDTH_OFFSET				10
+#define		CODADX6_PICWIDTH_MASK				0x3ff
+#define		CODA_PICHEIGHT_OFFSET				0
+#define		CODADX6_PICHEIGHT_MASK				0x3ff
+#define		CODA7_PICHEIGHT_MASK				0xffff
+#define CODA_CMD_ENC_SEQ_SRC_F_RATE				0x194
+#define CODA_CMD_ENC_SEQ_MP4_PARA				0x198
+#define		CODA_MP4PARAM_VERID_OFFSET			6
+#define		CODA_MP4PARAM_VERID_MASK			0x01
+#define		CODA_MP4PARAM_INTRADCVLCTHR_OFFSET		2
+#define		CODA_MP4PARAM_INTRADCVLCTHR_MASK		0x07
+#define		CODA_MP4PARAM_REVERSIBLEVLCENABLE_OFFSET	1
+#define		CODA_MP4PARAM_REVERSIBLEVLCENABLE_MASK		0x01
+#define		CODA_MP4PARAM_DATAPARTITIONENABLE_OFFSET	0
+#define		CODA_MP4PARAM_DATAPARTITIONENABLE_MASK		0x01
+#define CODA_CMD_ENC_SEQ_263_PARA				0x19c
+#define		CODA_263PARAM_ANNEXJENABLE_OFFSET		2
+#define		CODA_263PARAM_ANNEXJENABLE_MASK		0x01
+#define		CODA_263PARAM_ANNEXKENABLE_OFFSET		1
+#define		CODA_263PARAM_ANNEXKENABLE_MASK		0x01
+#define		CODA_263PARAM_ANNEXTENABLE_OFFSET		0
+#define		CODA_263PARAM_ANNEXTENABLE_MASK		0x01
+#define CODA_CMD_ENC_SEQ_264_PARA				0x1a0
+#define		CODA_264PARAM_DEBLKFILTEROFFSETBETA_OFFSET	12
+#define		CODA_264PARAM_DEBLKFILTEROFFSETBETA_MASK	0x0f
+#define		CODA_264PARAM_DEBLKFILTEROFFSETALPHA_OFFSET	8
+#define		CODA_264PARAM_DEBLKFILTEROFFSETALPHA_MASK	0x0f
+#define		CODA_264PARAM_DISABLEDEBLK_OFFSET		6
+#define		CODA_264PARAM_DISABLEDEBLK_MASK		0x01
+#define		CODA_264PARAM_CONSTRAINEDINTRAPREDFLAG_OFFSET	5
+#define		CODA_264PARAM_CONSTRAINEDINTRAPREDFLAG_MASK	0x01
+#define		CODA_264PARAM_CHROMAQPOFFSET_OFFSET		0
+#define		CODA_264PARAM_CHROMAQPOFFSET_MASK		0x1f
+#define CODA_CMD_ENC_SEQ_SLICE_MODE				0x1a4
+#define		CODA_SLICING_SIZE_OFFSET			2
+#define		CODA_SLICING_SIZE_MASK				0x3fffffff
+#define		CODA_SLICING_UNIT_OFFSET			1
+#define		CODA_SLICING_UNIT_MASK				0x01
+#define		CODA_SLICING_MODE_OFFSET			0
+#define		CODA_SLICING_MODE_MASK				0x01
+#define CODA_CMD_ENC_SEQ_GOP_SIZE				0x1a8
+#define		CODA_GOP_SIZE_OFFSET				0
+#define		CODA_GOP_SIZE_MASK				0x3f
+#define CODA_CMD_ENC_SEQ_RC_PARA				0x1ac
+#define		CODA_RATECONTROL_AUTOSKIP_OFFSET		31
+#define		CODA_RATECONTROL_AUTOSKIP_MASK			0x01
+#define		CODA_RATECONTROL_INITIALDELAY_OFFSET		16
+#define		CODA_RATECONTROL_INITIALDELAY_MASK		0x7f
+#define		CODA_RATECONTROL_BITRATE_OFFSET		1
+#define		CODA_RATECONTROL_BITRATE_MASK			0x7f
+#define		CODA_RATECONTROL_ENABLE_OFFSET			0
+#define		CODA_RATECONTROL_ENABLE_MASK			0x01
+#define CODA_CMD_ENC_SEQ_RC_BUF_SIZE				0x1b0
+#define CODA_CMD_ENC_SEQ_INTRA_REFRESH				0x1b4
+#define CODADX6_CMD_ENC_SEQ_FMO					0x1b8
+#define		CODA_FMOPARAM_TYPE_OFFSET			4
+#define		CODA_FMOPARAM_TYPE_MASK				1
+#define		CODA_FMOPARAM_SLICENUM_OFFSET			0
+#define		CODA_FMOPARAM_SLICENUM_MASK			0x0f
+#define CODA7_CMD_ENC_SEQ_SEARCH_BASE				0x1b8
+#define CODA7_CMD_ENC_SEQ_SEARCH_SIZE				0x1bc
+#define CODA_CMD_ENC_SEQ_RC_QP_MAX				0x1c8
+#define		CODA_QPMAX_OFFSET				0
+#define		CODA_QPMAX_MASK					0x3f
+#define CODA_CMD_ENC_SEQ_RC_GAMMA				0x1cc
+#define		CODA_GAMMA_OFFSET				0
+#define		CODA_GAMMA_MASK					0xffff
+#define CODA_RET_ENC_SEQ_SUCCESS				0x1c0
+
+/* Encoder Picture Run */
+#define CODA_CMD_ENC_PIC_SRC_ADDR_Y	0x180
+#define CODA_CMD_ENC_PIC_SRC_ADDR_CB	0x184
+#define CODA_CMD_ENC_PIC_SRC_ADDR_CR	0x188
+#define CODA_CMD_ENC_PIC_QS		0x18c
+#define CODA_CMD_ENC_PIC_ROT_MODE	0x190
+#define		CODA_ROT_MIR_ENABLE				(1 << 4)
+#define		CODA_ROT_0					(0x0 << 0)
+#define		CODA_ROT_90					(0x1 << 0)
+#define		CODA_ROT_180					(0x2 << 0)
+#define		CODA_ROT_270					(0x3 << 0)
+#define		CODA_MIR_NONE					(0x0 << 2)
+#define		CODA_MIR_VER					(0x1 << 2)
+#define		CODA_MIR_HOR					(0x2 << 2)
+#define		CODA_MIR_VER_HOR				(0x3 << 2)
+#define CODA_CMD_ENC_PIC_OPTION	0x194
+#define CODA_CMD_ENC_PIC_BB_START	0x198
+#define CODA_CMD_ENC_PIC_BB_SIZE	0x19c
+#define CODA_RET_ENC_PIC_TYPE		0x1c4
+#define CODA_RET_ENC_PIC_SLICE_NUM	0x1cc
+#define CODA_RET_ENC_PIC_FLAG		0x1d0
+
+/* Decoder Sequence Initialization */
+#define CODA_CMD_DEC_SEQ_BB_START				0x180
+#define CODA_CMD_DEC_SEQ_BB_SIZE				0x184
+#define CODA_CMD_DEC_SEQ_OPTION					0x188
+#define		CODADX6_DEC_OPTION_QPREPORT			1
+#define		CODA7_DEC_OPTION_MP4DEBLOCK			1
+#define CODA_CMD_DEC_SEQ_SRC_SIZE				0x18c
+#define		CODA7_PICWIDTH_OFFSET				16
+#define		CODA7_PICWIDTH_MASK				0xffff
+#define		CODADX6_PICWIDTH_OFFSET				10
+#define		CODADX6_PICWIDTH_MASK				0x3ff
+#define		CODA_PICHEIGHT_OFFSET				0
+#define		CODADX6_PICHEIGHT_MASK				0x3ff
+#define		CODA7_PICHEIGHT_MASK				0xffff
+#define CODA_CMD_DEC_SEQ_START_BYTE				0x190
+#define CODA_CMD_DEC_SEQ_PS_BB_START				0x194
+#define CODA_CMD_DEC_SEQ_PS_BB_SIZE				0x198
+#define CODA_CMD_DEC_SEQ_INIT_ESCAPE				0x114
+#define CODA_RET_DEC_SEQ_SUCCESS				0x1c0
+#define 	RET_DEC_SEQ_SRC_FMT				0x1c4
+#define 	RET_DEC_SEQ_SRC_SIZE				0x1c4
+#define 	RET_DEC_SEQ_SRC_F_RATE				0x1c8
+#define 	RET_DEC_SEQ_FRAME_NEED				0x1cc
+#define 	RET_DEC_SEQ_FRAME_DELAY				0x1d0
+#define 	RET_DEC_SEQ_INFO				0x1d4
+#define 	RET_DEC_SEQ_CROP_LEFT_RIGHT			0x1d8
+#define 	RET_DEC_SEQ_CROP_TOP_BOTTOM			0x1dc
+#define 	RET_DEC_SEQ_NEXT_FRAME_NUM			0x1e0
+#define CODA_RET_DEC_SEQ_ERR_REASON		0x1e0
+
+
+/* Set Frame Buffer */
+#define CODA_CMD_SET_FRAME_BUF_NUM		0x180
+#define CODA_CMD_SET_FRAME_BUF_STRIDE		0x184
+#define CODA7_CMD_SET_FRAME_AXI_BIT_ADDR	0x190
+#define CODA7_CMD_SET_FRAME_AXI_IPACDC_ADDR	0x194
+#define CODA7_CMD_SET_FRAME_AXI_DBKY_ADDR	0x198
+#define CODA7_CMD_SET_FRAME_AXI_DBKC_ADDR	0x19c
+#define CODA7_CMD_SET_FRAME_AXI_OVL_ADDR	0x1a0
+#define CODA7_CMD_SET_FRAME_SOURCE_BUF_STRIDE	0x1a8
+
+/* Encoder Header */
+#define CODA_CMD_ENC_HEADER_CODE	0x180
+#define		CODA_GAMMA_OFFSET	0
+#define		CODA_HEADER_H264_SPS	0
+#define		CODA_HEADER_H264_PPS	1
+#define		CODA_HEADER_MP4V_VOL	0
+#define		CODA_HEADER_MP4V_VOS	1
+#define		CODA_HEADER_MP4V_VIS	2
+#define CODA_CMD_ENC_HEADER_BB_START	0x184
+#define CODA_CMD_ENC_HEADER_BB_SIZE	0x188
+
+/* Get Version */
+#define CODA_CMD_FIRMWARE_VERNUM		0x1c0
+#define		CODA_FIRMWARE_PRODUCT(x)	(((x) >> 16) & 0xffff)
+#define		CODA_FIRMWARE_MAJOR(x)		(((x) >> 12) & 0x0f)
+#define		CODA_FIRMWARE_MINOR(x)		(((x) >> 8) & 0x0f)
+#define		CODA_FIRMWARE_RELEASE(x)	((x) & 0xff)
+#define		CODA_FIRMWARE_VERNUM(product, major, minor, release)	\
+			((product) << 16 | ((major) << 12) |		\
+			((minor) << 8) | (release))
+
+typedef enum {
+	RETCODE_SUCCESS = 0,
+	RETCODE_FAILURE = -1,
+	RETCODE_INVALID_HANDLE = -2,
+	RETCODE_INVALID_PARAM = -3,
+	RETCODE_INVALID_COMMAND = -4,
+	RETCODE_ROTATOR_OUTPUT_NOT_SET = -5,
+	RETCODE_ROTATOR_STRIDE_NOT_SET = -11,
+	RETCODE_FRAME_NOT_COMPLETE = -6,
+	RETCODE_INVALID_FRAME_BUFFER = -7,
+	RETCODE_INSUFFICIENT_FRAME_BUFFERS = -8,
+	RETCODE_INVALID_STRIDE = -9,
+	RETCODE_WRONG_CALL_SEQUENCE = -10,
+	RETCODE_CALLED_BEFORE = -12,
+	RETCODE_NOT_INITIALIZED = -13,
+	RETCODE_DEBLOCKING_OUTPUT_NOT_SET = -14,
+	RETCODE_NOT_SUPPORTED = -15
+} RetCode;
+
+#endif
--- /dev/null	2014-06-03 09:49:18.243206630 +0200
+++ coda.c	2014-06-03 10:11:25.158412400 +0200
@@ -0,0 +1,2423 @@
+/*
+ * Coda multi-standard codec IP
+ *
+ * Copyright (C) 2012 Vista Silicon S.L.
+ *    Javier Martin, <javier.martin@vista-silicon.com>
+ *    Xavier Duret
+ *
+ * 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.
+ */
+
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/firmware.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/irq.h>
+#ifdef USE_KFIFO
+#include <linux/kfifo.h>
+#endif
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/videodev2.h>
+#include <linux/of.h>
+
+#include <mach/iram.h>
+#include <media/v4l2-ctrls.h>
+#include <media/v4l2-device.h>
+#include <media/v4l2-ioctl.h>
+#include <media/v4l2-mem2mem.h>
+#include <media/videobuf2-core.h>
+#include <media/videobuf2-dma-contig.h>
+
+#include "coda.h"
+
+#define CODA_NAME		"coda"
+
+#define CODADX6_MAX_INSTANCES	4
+
+#define CODA_FMO_BUF_SIZE	32
+#define CODADX6_WORK_BUF_SIZE	(288 * 1024 + CODA_FMO_BUF_SIZE * 8 * 1024)
+#define CODA7_WORK_BUF_SIZE	(512 * 1024 + CODA_FMO_BUF_SIZE * 8 * 1024)
+#define CODA_PARA_BUF_SIZE	(10 * 1024)
+#define CODA_ISRAM_SIZE	(2048 * 2)
+#define CODA7_IRAM_SIZE		0x14000 /* 81920 bytes */
+
+#define CODA7_PS_BUF_SIZE	0x28000
+
+#define CODA_MAX_FRAMEBUFFERS	2
+
+#define MAX_W		8192
+#define MAX_H		8192
+#define CODA_MAX_FRAME_SIZE	0x100000
+#define FMO_SLICE_SAVE_BUF_SIZE         (32)
+#define CODA_DEFAULT_GAMMA		4096
+
+#define MIN_W 176
+#define MIN_H 144
+
+#define S_ALIGN		1 /* multiple of 2 */
+#define W_ALIGN		1 /* multiple of 2 */
+#define H_ALIGN		1 /* multiple of 2 */
+
+#define fh_to_ctx(__fh)	container_of(__fh, struct coda_ctx, fh)
+
+static int coda_debug;
+module_param(coda_debug, int, 0644);
+MODULE_PARM_DESC(coda_debug, "Debug level (0-1)");
+
+enum {
+	V4L2_M2M_SRC = 0,
+	V4L2_M2M_DST = 1,
+};
+
+enum coda_inst_type {
+	CODA_INST_ENCODER,
+	CODA_INST_DECODER,
+};
+
+enum coda_product {
+	CODA_DX6 = 0xf001,
+	CODA_7541 = 0xf012,
+};
+
+struct coda_fmt {
+	char *name;
+	u32 fourcc;
+};
+
+struct coda_codec {
+	u32 mode;
+	u32 src_fourcc;
+	u32 dst_fourcc;
+	u32 max_w;
+	u32 max_h;
+};
+
+struct coda_devtype {
+	char			*firmware;
+	enum coda_product	product;
+	struct coda_codec	*codecs;
+	unsigned int		num_codecs;
+	size_t			workbuf_size;
+};
+
+/* Per-queue, driver-specific private data */
+struct coda_q_data {
+	unsigned int		width;
+	unsigned int		height;
+	unsigned int		sizeimage;
+	unsigned int		fourcc;
+};
+
+struct coda_aux_buf {
+	void			*vaddr;
+	dma_addr_t		paddr;
+	u32			size;
+};
+
+struct coda_dev {
+	struct v4l2_device	v4l2_dev;
+	struct video_device	vfd;
+	struct platform_device	*plat_dev;
+	const struct coda_devtype *devtype;
+
+	void __iomem		*regs_base;
+	struct clk		*clk_per;
+	struct clk		*clk_ahb;
+
+	struct coda_aux_buf	codebuf;
+	struct coda_aux_buf	workbuf;
+	long unsigned int	iram_paddr;
+
+	spinlock_t		irqlock;
+	struct mutex		dev_mutex;
+	struct mutex		coda_mutex;
+	struct v4l2_m2m_dev	*m2m_dev;
+	struct vb2_alloc_ctx	*alloc_ctx;
+	struct list_head	instances;
+	unsigned long		instance_mask;
+	struct delayed_work	timeout;
+};
+
+struct coda_params {
+	u8			rot_mode;
+	u8			h264_intra_qp;
+	u8			h264_inter_qp;
+	u8			mpeg4_intra_qp;
+	u8			mpeg4_inter_qp;
+	u8			gop_size;
+	int			codec_mode;
+	enum v4l2_mpeg_video_multi_slice_mode slice_mode;
+	u32			framerate;
+	u16			bitrate;
+	u32			slice_max_bits;
+	u32			slice_max_mb;
+};
+
+struct coda_ctx {
+	struct coda_dev			*dev;
+	struct mutex			buffer_mutex;
+	struct list_head		list;
+	int				aborting;
+	int				streamon_out;
+	int				streamon_cap;
+	u32				isequence;
+	struct coda_q_data		q_data[2];
+	enum coda_inst_type		inst_type;
+	struct coda_codec		*codec;
+	enum v4l2_colorspace		colorspace;
+	struct coda_params		params;
+	struct v4l2_m2m_ctx		*m2m_ctx;
+	struct v4l2_ctrl_handler	ctrls;
+	struct v4l2_fh			fh;
+	int				gopcounter;
+	char				vpu_header[3][64];
+	int				vpu_header_size[3];
+#ifdef USE_KFIFO
+	struct kfifo			bitstream_fifo;
+#endif
+	struct mutex			bitstream_mutex;
+	struct coda_aux_buf		bitstream;
+	struct coda_aux_buf		parabuf;
+	struct coda_aux_buf		psbuf;
+	struct coda_aux_buf		slicebuf;
+	struct coda_aux_buf		internal_frames[CODA_MAX_FRAMEBUFFERS];
+	struct coda_aux_buf		workbuf;
+	int				num_internal_frames;
+	int				idx;
+	int				reg_idx;
+};
+
+static const u8 coda_filler_nal[14] = { 0x00, 0x00, 0x00, 0x01, 0x0c, 0xff,
+			0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x80 };
+static const u8 coda_filler_size[8] = { 0, 7, 14, 13, 12, 11, 10, 9 };
+
+static inline void coda_write(struct coda_dev *dev, u32 data, u32 reg)
+{
+	v4l2_dbg(1, coda_debug, &dev->v4l2_dev,
+		 "%s: data=0x%x, reg=0x%x\n", __func__, data, reg);
+	writel(data, dev->regs_base + reg);
+}
+
+static inline unsigned int coda_read(struct coda_dev *dev, u32 reg)
+{
+	u32 data;
+	data = readl(dev->regs_base + reg);
+	v4l2_dbg(1, coda_debug, &dev->v4l2_dev,
+		 "%s: data=0x%x, reg=0x%x\n", __func__, data, reg);
+	return data;
+}
+
+static inline unsigned long coda_isbusy(struct coda_dev *dev)
+{
+	return coda_read(dev, CODA_REG_BIT_BUSY);
+}
+
+static inline int coda_is_initialized(struct coda_dev *dev)
+{
+	return (coda_read(dev, CODA_REG_BIT_CUR_PC) != 0);
+}
+
+static int coda_wait_timeout(struct coda_dev *dev)
+{
+	unsigned long timeout = jiffies + msecs_to_jiffies(1000);
+
+	while (coda_isbusy(dev)) {
+		if (time_after(jiffies, timeout))
+			return -ETIMEDOUT;
+	}
+	return 0;
+}
+
+static void coda_command_async(struct coda_ctx *ctx, int cmd)
+{
+	struct coda_dev *dev = ctx->dev;
+	coda_write(dev, CODA_REG_BIT_BUSY_FLAG, CODA_REG_BIT_BUSY);
+
+	coda_write(dev, ctx->idx, CODA_REG_BIT_RUN_INDEX);
+	coda_write(dev, ctx->params.codec_mode, CODA_REG_BIT_RUN_COD_STD);
+	coda_write(dev, cmd, CODA_REG_BIT_RUN_COMMAND);
+}
+
+static int coda_command_sync(struct coda_ctx *ctx, int cmd)
+{
+	struct coda_dev *dev = ctx->dev;
+
+	coda_command_async(ctx, cmd);
+	return coda_wait_timeout(dev);
+}
+
+static struct coda_q_data *get_q_data(struct coda_ctx *ctx,
+					 enum v4l2_buf_type type)
+{
+	switch (type) {
+	case V4L2_BUF_TYPE_VIDEO_OUTPUT:
+		return &(ctx->q_data[V4L2_M2M_SRC]);
+	case V4L2_BUF_TYPE_VIDEO_CAPTURE:
+		return &(ctx->q_data[V4L2_M2M_DST]);
+	default:
+		BUG();
+	}
+	return NULL;
+}
+
+/*
+ * Array of all formats supported by any version of Coda:
+ */
+static struct coda_fmt coda_formats[] = {
+	{
+		.name = "YUV 4:2:0 Planar, YCbCr",
+		.fourcc = V4L2_PIX_FMT_YUV420,
+	},
+	{
+		.name = "YUV 4:2:0 Planar, YCrCb",
+		.fourcc = V4L2_PIX_FMT_YVU420,
+	},
+	{
+		.name = "H264 Encoded Stream",
+		.fourcc = V4L2_PIX_FMT_H264,
+	},
+	{
+		.name = "MPEG4 Encoded Stream",
+		.fourcc = V4L2_PIX_FMT_MPEG4,
+	},
+};
+
+#define CODA_CODEC(mode, src_fourcc, dst_fourcc, max_w, max_h) \
+	{ mode, src_fourcc, dst_fourcc, max_w, max_h }
+
+/*
+ * Arrays of codecs supported by each given version of Coda:
+ *  i.MX27 -> codadx6
+ *  i.MX5x -> coda7
+ *  i.MX6  -> coda960
+ * Use V4L2_PIX_FMT_YUV420 as placeholder for all supported YUV 4:2:0 variants
+ */
+static struct coda_codec codadx6_codecs[] = {
+	CODA_CODEC(CODADX6_MODE_ENCODE_H264, V4L2_PIX_FMT_YUV420, V4L2_PIX_FMT_H264,  720, 576),
+	CODA_CODEC(CODADX6_MODE_ENCODE_MP4,  V4L2_PIX_FMT_YUV420, V4L2_PIX_FMT_MPEG4, 720, 576),
+	CODA_CODEC(CODADX6_MODE_DECODE_H264, V4L2_PIX_FMT_H264, V4L2_PIX_FMT_YUV420,  720, 576),
+};
+
+static struct coda_codec coda7_codecs[] = {
+	CODA_CODEC(CODA7_MODE_ENCODE_H264, V4L2_PIX_FMT_YUV420, V4L2_PIX_FMT_H264,   1280, 720),
+	CODA_CODEC(CODA7_MODE_ENCODE_MP4,  V4L2_PIX_FMT_YUV420, V4L2_PIX_FMT_MPEG4,  1280, 720),
+};
+
+static bool coda_format_is_yuv(u32 fourcc)
+{
+	switch (fourcc) {
+	case V4L2_PIX_FMT_YUV420:
+	case V4L2_PIX_FMT_YVU420:
+		return true;
+	default:
+		return false;
+	}
+}
+
+/*
+ * Normalize all supported YUV 4:2:0 formats to the value used in the codec
+ * tables.
+ */
+static u32 coda_format_normalize_yuv(u32 fourcc)
+{
+	return coda_format_is_yuv(fourcc) ? V4L2_PIX_FMT_YUV420 : fourcc;
+}
+
+static struct coda_codec *coda_find_codec(struct coda_dev *dev, int src_fourcc,
+					  int dst_fourcc)
+{
+	struct coda_codec *codecs = dev->devtype->codecs;
+	int num_codecs = dev->devtype->num_codecs;
+	int k;
+
+	src_fourcc = coda_format_normalize_yuv(src_fourcc);
+	dst_fourcc = coda_format_normalize_yuv(dst_fourcc);
+	if (src_fourcc == dst_fourcc)
+		return NULL;
+
+	for (k = 0; k < num_codecs; k++) {
+		if (codecs[k].src_fourcc == src_fourcc &&
+		    codecs[k].dst_fourcc == dst_fourcc)
+			break;
+	}
+
+	if (k == num_codecs)
+		return NULL;
+
+	return &codecs[k];
+}
+
+/*
+ * V4L2 ioctl() operations.
+ */
+static int vidioc_querycap(struct file *file, void *priv,
+			   struct v4l2_capability *cap)
+{
+	strlcpy(cap->driver, CODA_NAME, sizeof(cap->driver));
+	strlcpy(cap->card, CODA_NAME, sizeof(cap->card));
+	strlcpy(cap->bus_info, "platform:" CODA_NAME, sizeof(cap->bus_info));
+	/*
+	 * This is only a mem-to-mem video device. The capture and output
+	 * device capability flags are left only for backward compatibility
+	 * and are scheduled for removal.
+	 */
+	cap->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_VIDEO_OUTPUT |
+			   V4L2_CAP_VIDEO_M2M | V4L2_CAP_STREAMING;
+	cap->capabilities = cap->device_caps | V4L2_CAP_DEVICE_CAPS;
+
+	return 0;
+}
+
+static int enum_fmt(void *priv, struct v4l2_fmtdesc *f,
+			enum v4l2_buf_type type)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+	struct coda_codec *codecs = ctx->dev->devtype->codecs;
+	struct coda_fmt *formats = coda_formats;
+	struct coda_fmt *fmt;
+	int num_codecs = ctx->dev->devtype->num_codecs;
+	int num_formats = ARRAY_SIZE(coda_formats);
+	int i, k, num = 0;
+
+	for (i = 0; i < num_formats; i++) {
+		/* Both uncompressed formats are always supported */
+		if (coda_format_is_yuv(formats[i].fourcc)) {
+			if (num == f->index)
+				break;
+			++num;
+			continue;
+		}
+		/* Compressed formats may be supported, check the codec list */
+		for (k = 0; k < num_codecs; k++) {
+			if (type == V4L2_BUF_TYPE_VIDEO_CAPTURE &&
+			    formats[i].fourcc == codecs[k].dst_fourcc)
+				break;
+			if (type == V4L2_BUF_TYPE_VIDEO_OUTPUT &&
+			    formats[i].fourcc == codecs[k].src_fourcc)
+				break;
+		}
+		if (k < num_codecs) {
+			if (num == f->index)
+				break;
+			++num;
+		}
+	}
+
+	if (i < num_formats) {
+		fmt = &formats[i];
+		strlcpy(f->description, fmt->name, sizeof(f->description));
+		f->pixelformat = fmt->fourcc;
+		return 0;
+	}
+
+	/* Format not found */
+	return -EINVAL;
+}
+
+static int vidioc_enum_fmt_vid_cap(struct file *file, void *priv,
+				   struct v4l2_fmtdesc *f)
+{
+	return enum_fmt(priv, f, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+}
+
+static int vidioc_enum_fmt_vid_out(struct file *file, void *priv,
+				   struct v4l2_fmtdesc *f)
+{
+	return enum_fmt(priv, f, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+}
+
+static int vidioc_g_fmt(struct file *file, void *priv, struct v4l2_format *f)
+{
+	struct vb2_queue *vq;
+	struct coda_q_data *q_data;
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	vq = v4l2_m2m_get_vq(ctx->m2m_ctx, f->type);
+	if (!vq)
+		return -EINVAL;
+
+	q_data = get_q_data(ctx, f->type);
+
+	f->fmt.pix.field	= V4L2_FIELD_NONE;
+	f->fmt.pix.pixelformat	= q_data->fourcc;
+	f->fmt.pix.width	= q_data->width;
+	f->fmt.pix.height	= q_data->height;
+	if (coda_format_is_yuv(f->fmt.pix.pixelformat))
+		f->fmt.pix.bytesperline = round_up(f->fmt.pix.width, 2);
+	else /* encoded formats h.264/mpeg4 */
+		f->fmt.pix.bytesperline = 0;
+
+	f->fmt.pix.sizeimage	= q_data->sizeimage;
+	f->fmt.pix.colorspace	= ctx->colorspace;
+
+	return 0;
+}
+
+static int vidioc_try_fmt(struct coda_codec *codec, struct v4l2_format *f)
+{
+	unsigned int max_w, max_h;
+	enum v4l2_field field;
+
+	field = f->fmt.pix.field;
+	if (field == V4L2_FIELD_ANY)
+		field = V4L2_FIELD_NONE;
+	else if (V4L2_FIELD_NONE != field)
+		return -EINVAL;
+
+	/* V4L2 specification suggests the driver corrects the format struct
+	 * if any of the dimensions is unsupported */
+	f->fmt.pix.field = field;
+
+	if (codec) {
+		max_w = codec->max_w;
+		max_h = codec->max_h;
+	} else {
+		max_w = MAX_W;
+		max_h = MAX_H;
+	}
+	v4l_bound_align_image(&f->fmt.pix.width, MIN_W, max_w,
+			      W_ALIGN, &f->fmt.pix.height,
+			      MIN_H, max_h, H_ALIGN, S_ALIGN);
+
+	if (coda_format_is_yuv(f->fmt.pix.pixelformat)) {
+		/* Frame stride must be multiple of 8 */
+		f->fmt.pix.bytesperline = round_up(f->fmt.pix.width, 8);
+		f->fmt.pix.sizeimage = f->fmt.pix.bytesperline *
+					f->fmt.pix.height * 3 / 2;
+	} else { /*encoded formats h.264/mpeg4 */
+		f->fmt.pix.bytesperline = 0;
+		f->fmt.pix.sizeimage = CODA_MAX_FRAME_SIZE;
+	}
+
+	f->fmt.pix.priv = 0;
+
+	return 0;
+}
+
+static int vidioc_try_fmt_vid_cap(struct file *file, void *priv,
+				  struct v4l2_format *f)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+	struct coda_codec *codec = NULL;
+
+	/* Determine codec by the encoded format */
+	codec = coda_find_codec(ctx->dev, V4L2_PIX_FMT_YUV420,
+				f->fmt.pix.pixelformat);
+	f->fmt.pix.colorspace = ctx->colorspace;
+
+	return vidioc_try_fmt(codec, f);
+}
+
+static int vidioc_try_fmt_vid_out(struct file *file, void *priv,
+				  struct v4l2_format *f)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+	struct coda_codec *codec;
+
+	/* Determine codec by encoded format, returns NULL if raw or invalid */
+	codec = coda_find_codec(ctx->dev, f->fmt.pix.pixelformat,
+				V4L2_PIX_FMT_YUV420);
+
+	if (!f->fmt.pix.colorspace)
+		f->fmt.pix.colorspace = V4L2_COLORSPACE_REC709;
+
+	return vidioc_try_fmt(codec, f);
+}
+
+static int vidioc_s_fmt(struct coda_ctx *ctx, struct v4l2_format *f)
+{
+	struct coda_q_data *q_data;
+	struct vb2_queue *vq;
+
+	vq = v4l2_m2m_get_vq(ctx->m2m_ctx, f->type);
+	if (!vq)
+		return -EINVAL;
+
+	q_data = get_q_data(ctx, f->type);
+	if (!q_data)
+		return -EINVAL;
+
+	if (vb2_is_busy(vq)) {
+		v4l2_err(&ctx->dev->v4l2_dev, "%s queue busy\n", __func__);
+		return -EBUSY;
+	}
+
+	q_data->fourcc = f->fmt.pix.pixelformat;
+	q_data->width = f->fmt.pix.width;
+	q_data->height = f->fmt.pix.height;
+	q_data->sizeimage = f->fmt.pix.sizeimage;
+
+	/*v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,*/
+	v4l2_err(&ctx->dev->v4l2_dev,
+		"Setting format for type %d, wxh: %dx%d, fmt: %08x\n",
+		f->type, q_data->width, q_data->height, q_data->fourcc);
+
+	return 0;
+}
+
+static int vidioc_s_fmt_vid_cap(struct file *file, void *priv,
+				struct v4l2_format *f)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+	int ret;
+
+	ret = vidioc_try_fmt_vid_cap(file, priv, f);
+	if (ret)
+		return ret;
+
+	return vidioc_s_fmt(ctx, f);
+}
+
+static int vidioc_s_fmt_vid_out(struct file *file, void *priv,
+				struct v4l2_format *f)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+	int ret;
+
+	ret = vidioc_try_fmt_vid_out(file, priv, f);
+	if (ret)
+		return ret;
+
+	ret = vidioc_s_fmt(ctx, f);
+	if (ret)
+		ctx->colorspace = f->fmt.pix.colorspace;
+
+	return ret;
+}
+
+static int vidioc_reqbufs(struct file *file, void *priv,
+			  struct v4l2_requestbuffers *reqbufs)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_reqbufs(file, ctx->m2m_ctx, reqbufs);
+}
+
+static int vidioc_querybuf(struct file *file, void *priv,
+			   struct v4l2_buffer *buf)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_querybuf(file, ctx->m2m_ctx, buf);
+}
+
+static int vidioc_qbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_qbuf(file, ctx->m2m_ctx, buf);
+}
+
+static int vidioc_dqbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_dqbuf(file, ctx->m2m_ctx, buf);
+}
+
+static int vidioc_streamon(struct file *file, void *priv,
+			   enum v4l2_buf_type type)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_streamon(file, ctx->m2m_ctx, type);
+}
+
+static int vidioc_streamoff(struct file *file, void *priv,
+			    enum v4l2_buf_type type)
+{
+	struct coda_ctx *ctx = fh_to_ctx(priv);
+
+	return v4l2_m2m_streamoff(file, ctx->m2m_ctx, type);
+}
+
+static const struct v4l2_ioctl_ops coda_ioctl_ops = {
+	.vidioc_querycap	= vidioc_querycap,
+
+	.vidioc_enum_fmt_vid_cap = vidioc_enum_fmt_vid_cap,
+	.vidioc_g_fmt_vid_cap	= vidioc_g_fmt,
+	.vidioc_try_fmt_vid_cap	= vidioc_try_fmt_vid_cap,
+	.vidioc_s_fmt_vid_cap	= vidioc_s_fmt_vid_cap,
+
+	.vidioc_enum_fmt_vid_out = vidioc_enum_fmt_vid_out,
+	.vidioc_g_fmt_vid_out	= vidioc_g_fmt,
+	.vidioc_try_fmt_vid_out	= vidioc_try_fmt_vid_out,
+	.vidioc_s_fmt_vid_out	= vidioc_s_fmt_vid_out,
+
+	.vidioc_reqbufs		= vidioc_reqbufs,
+	.vidioc_querybuf	= vidioc_querybuf,
+
+	.vidioc_qbuf		= vidioc_qbuf,
+	.vidioc_dqbuf		= vidioc_dqbuf,
+
+	.vidioc_streamon	= vidioc_streamon,
+	.vidioc_streamoff	= vidioc_streamoff,
+};
+
+#ifdef USE_KFIFO
+static inline int coda_get_bitstream_payload(struct coda_ctx *ctx)
+{
+	return kfifo_len(&ctx->bitstream_fifo);
+}
+#else
+#define coda_get_bitstream_payload(x) (-1)
+#endif
+
+/*
+ * Mem-to-mem operations.
+ */
+static void coda_device_run(void *m2m_priv)
+{
+	struct coda_ctx *ctx = m2m_priv;
+	struct coda_q_data *q_data_src, *q_data_dst;
+	struct vb2_buffer *src_buf, *dst_buf;
+	struct coda_dev *dev = ctx->dev;
+	int force_ipicture;
+	int quant_param = 0;
+	u32 picture_y, picture_cb, picture_cr;
+	u32 pic_stream_buffer_addr, pic_stream_buffer_size;
+	u32 dst_fourcc;
+
+	mutex_lock(&dev->coda_mutex);
+
+	src_buf = v4l2_m2m_next_src_buf(ctx->m2m_ctx);
+	dst_buf = v4l2_m2m_next_dst_buf(ctx->m2m_ctx);
+	q_data_src = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+	q_data_dst = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+	dst_fourcc = q_data_dst->fourcc;
+
+	src_buf->v4l2_buf.sequence = ctx->isequence;
+	dst_buf->v4l2_buf.sequence = ctx->isequence;
+	ctx->isequence++;
+
+	/*
+	 * Workaround coda firmware BUG that only marks the first
+	 * frame as IDR. This is a problem for some decoders that can't
+	 * recover when a frame is lost.
+	 */
+	if (src_buf->v4l2_buf.sequence % ctx->params.gop_size) {
+		src_buf->v4l2_buf.flags |= V4L2_BUF_FLAG_PFRAME;
+		src_buf->v4l2_buf.flags &= ~V4L2_BUF_FLAG_KEYFRAME;
+	} else {
+		src_buf->v4l2_buf.flags |= V4L2_BUF_FLAG_KEYFRAME;
+		src_buf->v4l2_buf.flags &= ~V4L2_BUF_FLAG_PFRAME;
+	}
+
+	/*
+	 * Copy headers at the beginning of the first frame for H.264 only.
+	 * In MPEG4 they are already copied by the coda.
+	 */
+	if (src_buf->v4l2_buf.sequence == 0) {
+		pic_stream_buffer_addr =
+			vb2_dma_contig_plane_dma_addr(dst_buf, 0) +
+			ctx->vpu_header_size[0] +
+			ctx->vpu_header_size[1] +
+			ctx->vpu_header_size[2];
+		pic_stream_buffer_size = CODA_MAX_FRAME_SIZE -
+			ctx->vpu_header_size[0] -
+			ctx->vpu_header_size[1] -
+			ctx->vpu_header_size[2];
+		memcpy(vb2_plane_vaddr(dst_buf, 0),
+		       &ctx->vpu_header[0][0], ctx->vpu_header_size[0]);
+		memcpy(vb2_plane_vaddr(dst_buf, 0) + ctx->vpu_header_size[0],
+		       &ctx->vpu_header[1][0], ctx->vpu_header_size[1]);
+		memcpy(vb2_plane_vaddr(dst_buf, 0) + ctx->vpu_header_size[0] +
+			ctx->vpu_header_size[1], &ctx->vpu_header[2][0],
+			ctx->vpu_header_size[2]);
+	} else {
+		pic_stream_buffer_addr =
+			vb2_dma_contig_plane_dma_addr(dst_buf, 0);
+		pic_stream_buffer_size = CODA_MAX_FRAME_SIZE;
+	}
+
+	if (src_buf->v4l2_buf.flags & V4L2_BUF_FLAG_KEYFRAME) {
+		force_ipicture = 1;
+		switch (dst_fourcc) {
+		case V4L2_PIX_FMT_H264:
+			quant_param = ctx->params.h264_intra_qp;
+			break;
+		case V4L2_PIX_FMT_MPEG4:
+			quant_param = ctx->params.mpeg4_intra_qp;
+			break;
+		default:
+			v4l2_warn(&ctx->dev->v4l2_dev,
+				"cannot set intra qp, fmt not supported\n");
+			break;
+		}
+	} else {
+		force_ipicture = 0;
+		switch (dst_fourcc) {
+		case V4L2_PIX_FMT_H264:
+			quant_param = ctx->params.h264_inter_qp;
+			break;
+		case V4L2_PIX_FMT_MPEG4:
+			quant_param = ctx->params.mpeg4_inter_qp;
+			break;
+		default:
+			v4l2_warn(&ctx->dev->v4l2_dev,
+				"cannot set inter qp, fmt not supported\n");
+			break;
+		}
+	}
+
+	/* submit */
+	coda_write(dev, CODA_ROT_MIR_ENABLE | ctx->params.rot_mode, CODA_CMD_ENC_PIC_ROT_MODE);
+	coda_write(dev, quant_param, CODA_CMD_ENC_PIC_QS);
+
+
+	picture_y = vb2_dma_contig_plane_dma_addr(src_buf, 0);
+	switch (q_data_src->fourcc) {
+	case V4L2_PIX_FMT_YVU420:
+		/* Switch Cb and Cr for YVU420 format */
+		picture_cr = picture_y + q_data_src->width * q_data_src->height;
+		picture_cb = picture_cr + q_data_src->width / 2 *
+				q_data_src->height / 2;
+		break;
+	case V4L2_PIX_FMT_YUV420:
+	default:
+		picture_cb = picture_y + q_data_src->width * q_data_src->height;
+		picture_cr = picture_cb + q_data_src->width / 2 *
+				q_data_src->height / 2;
+		break;
+	}
+
+	coda_write(dev, picture_y, CODA_CMD_ENC_PIC_SRC_ADDR_Y);
+	coda_write(dev, picture_cb, CODA_CMD_ENC_PIC_SRC_ADDR_CB);
+	coda_write(dev, picture_cr, CODA_CMD_ENC_PIC_SRC_ADDR_CR);
+	coda_write(dev, force_ipicture << 1 & 0x2,
+		   CODA_CMD_ENC_PIC_OPTION);
+
+	coda_write(dev, pic_stream_buffer_addr, CODA_CMD_ENC_PIC_BB_START);
+	coda_write(dev, pic_stream_buffer_size / 1024,
+		   CODA_CMD_ENC_PIC_BB_SIZE);
+
+	if (dev->devtype->product == CODA_7541) {
+		coda_write(dev, CODA7_USE_BIT_ENABLE | CODA7_USE_HOST_BIT_ENABLE |
+				CODA7_USE_ME_ENABLE | CODA7_USE_HOST_ME_ENABLE,
+				CODA7_REG_BIT_AXI_SRAM_USE);
+	}
+
+	/* 1 second timeout in case CODA locks up */
+	schedule_delayed_work(&dev->timeout, HZ);
+
+	coda_command_async(ctx, CODA_COMMAND_PIC_RUN);
+}
+
+static int coda_job_ready(void *m2m_priv)
+{
+	struct coda_ctx *ctx = m2m_priv;
+
+	/*
+	 * For both 'P' and 'key' frame cases 1 picture
+	 * and 1 frame are needed.
+	 */
+	if (!v4l2_m2m_num_src_bufs_ready(ctx->m2m_ctx) ||
+		!v4l2_m2m_num_dst_bufs_ready(ctx->m2m_ctx)) {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "not ready: not enough video buffers.\n");
+		return 0;
+	}
+
+	if (ctx->aborting) {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "not ready: aborting\n");
+		return 0;
+	}
+
+	v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			"job ready\n");
+	return 1;
+}
+
+static void coda_job_abort(void *priv)
+{
+	struct coda_ctx *ctx = priv;
+
+	ctx->aborting = 1;
+
+	v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+		 "Aborting task\n");
+}
+
+static void coda_lock(void *m2m_priv)
+{
+	struct coda_ctx *ctx = m2m_priv;
+	struct coda_dev *pcdev = ctx->dev;
+	mutex_lock(&pcdev->dev_mutex);
+}
+
+static void coda_unlock(void *m2m_priv)
+{
+	struct coda_ctx *ctx = m2m_priv;
+	struct coda_dev *pcdev = ctx->dev;
+	mutex_unlock(&pcdev->dev_mutex);
+}
+
+static struct v4l2_m2m_ops coda_m2m_ops = {
+	.device_run	= coda_device_run,
+	.job_ready	= coda_job_ready,
+	.job_abort	= coda_job_abort,
+	.lock		= coda_lock,
+	.unlock		= coda_unlock,
+};
+
+static void set_default_params(struct coda_ctx *ctx)
+{
+	int max_w;
+	int max_h;
+
+	ctx->codec = &ctx->dev->devtype->codecs[0];
+	max_w = ctx->codec->max_w;
+	max_h = ctx->codec->max_h;
+
+	ctx->params.codec_mode = CODA_MODE_INVALID;
+	ctx->colorspace = V4L2_COLORSPACE_REC709;
+	ctx->params.framerate = 30;
+	ctx->aborting = 0;
+
+	/* Default formats for output and input queues */
+	ctx->q_data[V4L2_M2M_SRC].fourcc = ctx->codec->src_fourcc;
+	ctx->q_data[V4L2_M2M_DST].fourcc = ctx->codec->dst_fourcc;
+	ctx->q_data[V4L2_M2M_SRC].width = max_w;
+	ctx->q_data[V4L2_M2M_SRC].height = max_h;
+	ctx->q_data[V4L2_M2M_SRC].sizeimage = (max_w * max_h * 3) / 2;
+	ctx->q_data[V4L2_M2M_DST].width = max_w;
+	ctx->q_data[V4L2_M2M_DST].height = max_h;
+	ctx->q_data[V4L2_M2M_DST].sizeimage = CODA_MAX_FRAME_SIZE;
+}
+
+/*
+ * Queue operations
+ */
+static int coda_queue_setup(struct vb2_queue *vq,
+				const struct v4l2_format *fmt,
+				unsigned int *nbuffers, unsigned int *nplanes,
+				unsigned int sizes[], void *alloc_ctxs[])
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(vq);
+	struct coda_q_data *q_data;
+	unsigned int size;
+
+	q_data = get_q_data(ctx, vq->type);
+	size = q_data->sizeimage;
+
+	*nplanes = 1;
+	sizes[0] = size;
+
+	alloc_ctxs[0] = ctx->dev->alloc_ctx;
+
+	v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+		 "get %d buffer(s) of size %d each.\n", *nbuffers, size);
+
+	return 0;
+}
+
+static int coda_buf_prepare(struct vb2_buffer *vb)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
+	struct coda_q_data *q_data;
+
+	q_data = get_q_data(ctx, vb->vb2_queue->type);
+
+	if (vb2_plane_size(vb, 0) < q_data->sizeimage) {
+		v4l2_warn(&ctx->dev->v4l2_dev,
+			  "%s data will not fit into plane (%lu < %lu)\n",
+			  __func__, vb2_plane_size(vb, 0),
+			  (long)q_data->sizeimage);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static void coda_buf_queue(struct vb2_buffer *vb)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(vb->vb2_queue);
+	v4l2_m2m_buf_queue(ctx->m2m_ctx, vb);
+}
+
+static void coda_wait_prepare(struct vb2_queue *q)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	coda_unlock(ctx);
+}
+
+static void coda_wait_finish(struct vb2_queue *q)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	coda_lock(ctx);
+}
+
+static void coda_free_framebuffers(struct coda_ctx *ctx)
+{
+	int i;
+
+	for (i = 0; i < CODA_MAX_FRAMEBUFFERS; i++) {
+		if (ctx->internal_frames[i].vaddr) {
+			dma_free_coherent(&ctx->dev->plat_dev->dev,
+				ctx->internal_frames[i].size,
+				ctx->internal_frames[i].vaddr,
+				ctx->internal_frames[i].paddr);
+			ctx->internal_frames[i].vaddr = NULL;
+		}
+	}
+}
+
+static void coda_parabuf_write(struct coda_ctx *ctx, int index, u32 value)
+{
+	struct coda_dev *dev = ctx->dev;
+	u32 *p = ctx->parabuf.vaddr;
+
+	if (dev->devtype->product == CODA_DX6)
+		p[index] = value;
+	else
+		p[index ^ 1] = value;
+}
+
+static int coda_alloc_aux_buf(struct coda_dev *dev,
+			      struct coda_aux_buf *buf, size_t size)
+{
+	buf->vaddr = dma_alloc_coherent(&dev->plat_dev->dev, size, &buf->paddr,
+					GFP_KERNEL);
+	if (!buf->vaddr)
+		return -ENOMEM;
+
+	buf->size = size;
+
+	return 0;
+}
+
+static inline int coda_alloc_context_buf(struct coda_ctx *ctx,
+					 struct coda_aux_buf *buf, size_t size)
+{
+	return coda_alloc_aux_buf(ctx->dev, buf, size);
+}
+
+static void coda_free_aux_buf(struct coda_dev *dev,
+			      struct coda_aux_buf *buf)
+{
+	if (buf->vaddr) {
+		dma_free_coherent(&dev->plat_dev->dev, buf->size,
+				  buf->vaddr, buf->paddr);
+		buf->vaddr = NULL;
+		buf->size = 0;
+	}
+}
+
+static int coda_alloc_framebuffers(struct coda_ctx *ctx, struct coda_q_data *q_data, u32 fourcc)
+{
+	struct coda_dev *dev = ctx->dev;
+
+	int height = q_data->height;
+	dma_addr_t paddr;
+	int ysize;
+	int i;
+
+	ysize = round_up(q_data->width, 8) * height;
+
+	/* Allocate frame buffers */
+	ctx->num_internal_frames = CODA_MAX_FRAMEBUFFERS;
+	for (i = 0; i < ctx->num_internal_frames; i++) {
+		ctx->internal_frames[i].size = q_data->sizeimage;
+		if (fourcc == V4L2_PIX_FMT_H264 && dev->devtype->product != CODA_DX6)
+			ctx->internal_frames[i].size += ysize/4;
+		ctx->internal_frames[i].vaddr = dma_alloc_coherent(
+				&dev->plat_dev->dev, ctx->internal_frames[i].size,
+				&ctx->internal_frames[i].paddr, GFP_KERNEL);
+		if (!ctx->internal_frames[i].vaddr) {
+			coda_free_framebuffers(ctx);
+			return -ENOMEM;
+		}
+	}
+
+	/* Register frame buffers in the parameter buffer */
+	for (i = 0; i < ctx->num_internal_frames; i++) {
+		paddr = ctx->internal_frames[i].paddr;
+		coda_parabuf_write(ctx, i * 3 + 0, paddr); /* Y */
+		coda_parabuf_write(ctx, i * 3 + 1, paddr + ysize); /* Cb */
+		coda_parabuf_write(ctx, i * 3 + 2, paddr + ysize + ysize/4); /* Cr */
+
+		if (dev->devtype->product != CODA_DX6 && fourcc == V4L2_PIX_FMT_H264)
+			coda_parabuf_write(ctx, 96 + i, ctx->internal_frames[i].paddr + ysize + ysize/4 + ysize/4);
+	}
+
+	return 0;
+}
+
+static int coda_h264_padding(int size, char *p)
+{
+	int nal_size;
+	int diff;
+
+	diff = size - (size & ~0x7);
+	if (diff == 0)
+		return 0;
+
+	nal_size = coda_filler_size[diff];
+	memcpy(p, coda_filler_nal, nal_size);
+
+	/* Add rbsp stop bit and trailing at the end */
+	*(p + nal_size - 1) = 0x80;
+
+	return nal_size;
+}
+
+static void coda_free_context_buffers(struct coda_ctx *ctx)
+{
+	struct coda_dev *dev = ctx->dev;
+
+	if (dev->devtype->product == CODA_DX6) {
+		return;
+	} else {
+		coda_free_aux_buf(dev, &ctx->slicebuf);
+		coda_free_aux_buf(dev, &ctx->workbuf);
+		if (dev->devtype->product == CODA_7541)
+			coda_free_aux_buf(dev, &ctx->psbuf);
+	}
+}
+
+static int coda_alloc_context_buffers(struct coda_ctx *ctx,
+				      struct coda_q_data *q_data)
+{
+	struct coda_dev *dev = ctx->dev;
+	size_t size;
+	int ret;
+
+	switch (dev->devtype->product) {
+	case CODA_7541:
+		size = CODA7_WORK_BUF_SIZE;
+		break;
+	default:
+		return 0;
+	}
+
+	if (ctx->psbuf.vaddr) {
+		v4l2_err(&dev->v4l2_dev, "psmembuf still allocated\n");
+		return -EBUSY;
+	}
+	if (ctx->slicebuf.vaddr) {
+		v4l2_err(&dev->v4l2_dev, "slicebuf still allocated\n");
+		return -EBUSY;
+	}
+	if (ctx->workbuf.vaddr) {
+		v4l2_err(&dev->v4l2_dev, "context buffer still allocated\n");
+		ret = -EBUSY;
+		return -ENOMEM;
+	}
+
+	if (q_data->fourcc == V4L2_PIX_FMT_H264) {
+		/* worst case slice size */
+		size = (DIV_ROUND_UP(q_data->width, 16) *
+			DIV_ROUND_UP(q_data->height, 16)) * 3200 / 8 + 512;
+		ret = coda_alloc_context_buf(ctx, &ctx->slicebuf, size);
+		if (ret < 0) {
+			v4l2_err(&dev->v4l2_dev, "failed to allocate %d byte slice buffer",
+				 ctx->slicebuf.size);
+			return ret;
+		}
+	}
+
+	if (dev->devtype->product == CODA_7541) {
+		ret = coda_alloc_context_buf(ctx, &ctx->psbuf, CODA7_PS_BUF_SIZE);
+		if (ret < 0) {
+			v4l2_err(&dev->v4l2_dev, "failed to allocate psmem buffer");
+			goto err;
+		}
+	}
+
+	ret = coda_alloc_context_buf(ctx, &ctx->workbuf, size);
+	if (ret < 0) {
+		v4l2_err(&dev->v4l2_dev, "failed to allocate %d byte context buffer",
+			 ctx->workbuf.size);
+		goto err;
+	}
+
+	return 0;
+
+err:
+	coda_free_context_buffers(ctx);
+	return ret;
+}
+
+static int coda_encode_header(struct coda_ctx *ctx, struct vb2_buffer *buf,
+			      int header_code, u8 *header, int *size)
+{
+	struct coda_dev *dev = ctx->dev;
+	int ret;
+
+	coda_write(dev, vb2_dma_contig_plane_dma_addr(buf, 0),
+		   CODA_CMD_ENC_HEADER_BB_START);
+	coda_write(dev, vb2_plane_size(buf, 0), CODA_CMD_ENC_HEADER_BB_SIZE);
+	coda_write(dev, header_code, CODA_CMD_ENC_HEADER_CODE);
+	ret = coda_command_sync(ctx, CODA_COMMAND_ENCODE_HEADER);
+	if (ret < 0) {
+		v4l2_err(&dev->v4l2_dev, "CODA_COMMAND_ENCODE_HEADER timeout\n");
+		return ret;
+	}
+	*size = coda_read(dev, CODA_REG_BIT_WR_PTR(ctx->idx)) -
+		coda_read(dev, CODA_CMD_ENC_HEADER_BB_START);
+	memcpy(header, vb2_plane_vaddr(buf, 0), *size);
+
+	return 0;
+}
+
+static int coda_start_streaming_encoder(struct vb2_queue *q, unsigned int count)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	struct v4l2_device *v4l2_dev = &ctx->dev->v4l2_dev;
+	u32 bitstream_buf, bitstream_size;
+	struct coda_dev *dev = ctx->dev;
+	struct coda_q_data *q_data_src, *q_data_dst;
+	struct vb2_buffer *buf;
+	u32 dst_fourcc;
+	u32 value;
+	int ret = 0;
+	q_data_src = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+	q_data_dst = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+	buf = v4l2_m2m_next_dst_buf(ctx->m2m_ctx);
+	bitstream_buf = vb2_dma_contig_plane_dma_addr(buf, 0);
+	bitstream_size = q_data_dst->sizeimage;
+	dst_fourcc = q_data_dst->fourcc;
+	/* Could set rotation here if needed */
+	switch (dev->devtype->product) {
+	case CODA_DX6:
+		value = (q_data_src->width & CODADX6_PICWIDTH_MASK) << CODADX6_PICWIDTH_OFFSET;
+		value |= (q_data_src->height & CODADX6_PICHEIGHT_MASK) << CODA_PICHEIGHT_OFFSET;
+		break;
+	default:
+		value = (q_data_src->width & CODA7_PICWIDTH_MASK) << CODA7_PICWIDTH_OFFSET;
+		value |= (q_data_src->height & CODA7_PICHEIGHT_MASK) << CODA_PICHEIGHT_OFFSET;
+	}
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_SRC_SIZE);
+	coda_write(dev, ctx->params.framerate,
+		   CODA_CMD_ENC_SEQ_SRC_F_RATE);
+
+	switch (dst_fourcc) {
+	case V4L2_PIX_FMT_MPEG4:
+		coda_write(dev, CODA_STD_MPEG4, CODA_CMD_ENC_SEQ_COD_STD);
+		coda_write(dev, 0, CODA_CMD_ENC_SEQ_MP4_PARA);
+		break;
+	case V4L2_PIX_FMT_H264:
+		coda_write(dev, CODA_STD_H264, CODA_CMD_ENC_SEQ_COD_STD);
+		coda_write(dev, 0, CODA_CMD_ENC_SEQ_264_PARA);
+		break;
+	default:
+		v4l2_err(v4l2_dev,
+			 "dst format (0x%08x) invalid.\n", dst_fourcc);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	switch (ctx->params.slice_mode) {
+	case V4L2_MPEG_VIDEO_MULTI_SLICE_MODE_SINGLE:
+		value = 0;
+		break;
+	case V4L2_MPEG_VIDEO_MULTI_SICE_MODE_MAX_MB:
+		value  = (ctx->params.slice_max_mb & CODA_SLICING_SIZE_MASK) << CODA_SLICING_SIZE_OFFSET;
+		value |= (1 & CODA_SLICING_UNIT_MASK) << CODA_SLICING_UNIT_OFFSET;
+		value |=  1 & CODA_SLICING_MODE_MASK;
+		break;
+	case V4L2_MPEG_VIDEO_MULTI_SICE_MODE_MAX_BYTES:
+		value  = (ctx->params.slice_max_bits & CODA_SLICING_SIZE_MASK) << CODA_SLICING_SIZE_OFFSET;
+		value |= (0 & CODA_SLICING_UNIT_MASK) << CODA_SLICING_UNIT_OFFSET;
+		value |=  1 & CODA_SLICING_MODE_MASK;
+		break;
+	}
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_SLICE_MODE);
+	value = ctx->params.gop_size & CODA_GOP_SIZE_MASK;
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_GOP_SIZE);
+
+	if (ctx->params.bitrate) {
+		/* Rate control enabled */
+		value = (ctx->params.bitrate & CODA_RATECONTROL_BITRATE_MASK) << CODA_RATECONTROL_BITRATE_OFFSET;
+		value |=  1 & CODA_RATECONTROL_ENABLE_MASK;
+	} else {
+		value = 0;
+	}
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_RC_PARA);
+
+	coda_write(dev, 0, CODA_CMD_ENC_SEQ_RC_BUF_SIZE);
+	coda_write(dev, 0, CODA_CMD_ENC_SEQ_INTRA_REFRESH);
+
+	coda_write(dev, bitstream_buf, CODA_CMD_ENC_SEQ_BB_START);
+	coda_write(dev, bitstream_size / 1024, CODA_CMD_ENC_SEQ_BB_SIZE);
+
+	/* set default gamma */
+	value = (CODA_DEFAULT_GAMMA & CODA_GAMMA_MASK) << CODA_GAMMA_OFFSET;
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_RC_GAMMA);
+
+	if (CODA_DEFAULT_GAMMA > 0) {
+		if (dev->devtype->product == CODA_DX6)
+			value  = 1 << CODADX6_OPTION_GAMMA_OFFSET;
+		else
+			value  = 1 << CODA7_OPTION_GAMMA_OFFSET;
+	} else {
+		value = 0;
+	}
+	coda_write(dev, value, CODA_CMD_ENC_SEQ_OPTION);
+
+	if (dst_fourcc == V4L2_PIX_FMT_H264) {
+		value  = (FMO_SLICE_SAVE_BUF_SIZE << 7);
+		value |= (0 & CODA_FMOPARAM_TYPE_MASK) << CODA_FMOPARAM_TYPE_OFFSET;
+		value |=  0 & CODA_FMOPARAM_SLICENUM_MASK;
+		if (dev->devtype->product == CODA_DX6) {
+			coda_write(dev, value, CODADX6_CMD_ENC_SEQ_FMO);
+		} else {
+			coda_write(dev, dev->iram_paddr, CODA7_CMD_ENC_SEQ_SEARCH_BASE);
+			coda_write(dev, 48 * 1024, CODA7_CMD_ENC_SEQ_SEARCH_SIZE);
+		}
+	}
+
+	ret = coda_command_sync(ctx, CODA_COMMAND_SEQ_INIT);
+	if (ret < 0) {
+		v4l2_err(v4l2_dev, "CODA_COMMAND_SEQ_INIT timeout\n");
+		goto out;
+	}
+
+	if (coda_read(dev, CODA_RET_ENC_SEQ_SUCCESS) == 0) {
+		v4l2_err(v4l2_dev, "CODA_COMMAND_SEQ_INIT failed\n");
+		ret = -EFAULT;
+		goto out;
+	}
+
+	ret = coda_alloc_framebuffers(ctx, q_data_src, dst_fourcc);
+	if (ret < 0) {
+		v4l2_err(v4l2_dev, "failed to allocate framebuffers\n");
+		goto out;
+	}
+
+	coda_write(dev, ctx->num_internal_frames, CODA_CMD_SET_FRAME_BUF_NUM);
+	coda_write(dev, round_up(q_data_src->width, 8), CODA_CMD_SET_FRAME_BUF_STRIDE);
+	if (dev->devtype->product != CODA_DX6) {
+		coda_write(dev, round_up(q_data_src->width, 8), CODA7_CMD_SET_FRAME_SOURCE_BUF_STRIDE);
+		coda_write(dev, dev->iram_paddr + 48 * 1024, CODA7_CMD_SET_FRAME_AXI_DBKY_ADDR);
+		coda_write(dev, dev->iram_paddr + 53 * 1024, CODA7_CMD_SET_FRAME_AXI_DBKC_ADDR);
+		coda_write(dev, dev->iram_paddr + 58 * 1024, CODA7_CMD_SET_FRAME_AXI_BIT_ADDR);
+		coda_write(dev, dev->iram_paddr + 68 * 1024, CODA7_CMD_SET_FRAME_AXI_IPACDC_ADDR);
+		coda_write(dev, 0x0, CODA7_CMD_SET_FRAME_AXI_OVL_ADDR);
+	}
+	ret = coda_command_sync(ctx, CODA_COMMAND_SET_FRAME_BUF);
+	if (ret < 0) {
+		v4l2_err(v4l2_dev, "CODA_COMMAND_SET_FRAME_BUF timeout\n");
+		goto out;
+	}
+
+	/* Save stream headers */
+	buf = v4l2_m2m_next_dst_buf(ctx->m2m_ctx);
+	switch (dst_fourcc) {
+	case V4L2_PIX_FMT_H264:
+		/*
+		 * Get SPS in the first frame and copy it to an
+		 * intermediate buffer.
+		 */
+		ret = coda_encode_header(ctx, buf, CODA_HEADER_H264_SPS,
+					 &ctx->vpu_header[0][0],
+					 &ctx->vpu_header_size[0]);
+		if (ret < 0)
+			goto out;
+
+		/*
+		 * Get PPS in the first frame and copy it to an
+		 * intermediate buffer.
+		 */
+		ret = coda_encode_header(ctx, buf, CODA_HEADER_H264_PPS,
+					 &ctx->vpu_header[1][0],
+					 &ctx->vpu_header_size[1]);
+		if (ret < 0)
+			goto out;
+
+		/*
+		 * Length of H.264 headers is variable and thus it might not be
+		 * aligned for the coda to append the encoded frame. In that is
+		 * the case a filler NAL must be added to header 2.
+		 */
+		ctx->vpu_header_size[2] = coda_h264_padding(
+					(ctx->vpu_header_size[0] +
+					 ctx->vpu_header_size[1]),
+					 ctx->vpu_header[2]);
+		break;
+	case V4L2_PIX_FMT_MPEG4:
+		/*
+		 * Get VOS in the first frame and copy it to an
+		 * intermediate buffer
+		 */
+		ret = coda_encode_header(ctx, buf, CODA_HEADER_MP4V_VOS,
+					 &ctx->vpu_header[0][0],
+					 &ctx->vpu_header_size[0]);
+		if (ret < 0)
+			goto out;
+
+		ret = coda_encode_header(ctx, buf, CODA_HEADER_MP4V_VIS,
+					 &ctx->vpu_header[1][0],
+					 &ctx->vpu_header_size[1]);
+		if (ret < 0)
+			goto out;
+
+		ret = coda_encode_header(ctx, buf, CODA_HEADER_MP4V_VOL,
+					 &ctx->vpu_header[2][0],
+					 &ctx->vpu_header_size[2]);
+		if (ret < 0)
+			goto out;
+		break;
+	default:
+		/* No more formats need to save headers at the moment */
+		break;
+	}
+
+out:
+	return ret;
+}
+
+static int coda_start_streaming_decoder(struct vb2_queue *q, unsigned int count)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	struct v4l2_device *v4l2_dev = &ctx->dev->v4l2_dev;
+	u32 bitstream_buf, bitstream_size;
+	struct coda_dev *dev = ctx->dev;
+	struct coda_q_data *q_data_src, *q_data_dst;
+	struct vb2_buffer *buf;
+	u32 dst_fourcc;
+	u32 value;
+	int ret = 0;
+	q_data_src = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+	q_data_dst = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+	buf = v4l2_m2m_next_dst_buf(ctx->m2m_ctx);
+	bitstream_buf = vb2_dma_contig_plane_dma_addr(buf, 0);
+	bitstream_size = q_data_dst->sizeimage;
+	dst_fourcc = q_data_dst->fourcc;
+	/* First put Decoder in escape init mode */
+	coda_write(dev, 1, CODA_CMD_DEC_SEQ_INIT_ESCAPE);
+	/* Set encoded buffer address/size */
+	coda_write(dev, bitstream_buf, CODA_CMD_DEC_SEQ_BB_START);
+	coda_write(dev, bitstream_size / 1024, CODA_CMD_DEC_SEQ_BB_SIZE);
+	/* Coda will be used with live stream, set start byte to CODA_CMD_DEC_SEQ_BB_START value */
+	//coda_write(dev, bitstream_buf, CODA_CMD_DEC_SEQ_START_BYTE);
+printk("%s\n", __FUNCTION__);
+
+	/* Normally at this stage WrPtr and RdPtr must not be equal */
+	//if (DecBitstreamBufEmpty(pDecInfo)) return RETCODE_WRONG_CALL_SEQUENCE;
+
+	/* Disable options (CODADX6_DEC_OPTION_QPREPORT) */
+	coda_write(dev, 0, CODA_CMD_DEC_SEQ_OPTION);
+
+	coda_write(dev, 0/*bitstream_buf*/, CODA_CMD_DEC_SEQ_PS_BB_START);
+	coda_write(dev, 0/*bitstream_size / 1024*/, CODA_CMD_DEC_SEQ_PS_BB_SIZE);
+
+	/* Could set rotation here if needed */
+	switch (dev->devtype->product) {
+	case CODA_DX6:
+		value = (q_data_src->width & CODADX6_PICWIDTH_MASK) << CODADX6_PICWIDTH_OFFSET;
+		value |= (q_data_src->height & CODADX6_PICHEIGHT_MASK) << CODA_PICHEIGHT_OFFSET;
+		break;
+	default:
+		/* Only Coda DX6 is handled */
+		ret = -EINVAL;
+		goto out;
+	}
+	coda_write(dev, value, CODA_CMD_DEC_SEQ_SRC_SIZE);
+
+	ret = coda_command_sync(ctx, CODA_COMMAND_SEQ_INIT);
+	if (ret < 0) {
+		v4l2_err(v4l2_dev, "CODA_COMMAND_SEQ_INIT timeout\n");
+		goto out;
+	}
+
+	if (coda_read(dev, CODA_RET_DEC_SEQ_SUCCESS) == 0) {
+		v4l2_err(v4l2_dev, "CODA_COMMAND_SEQ_INIT failed %d\n", coda_read(dev, CODA_RET_DEC_SEQ_ERR_REASON));
+		ret = -EFAULT;
+		goto out;
+	}
+
+out:
+	return ret;
+}
+
+static int coda_start_streaming(struct vb2_queue *q, unsigned int count)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	struct v4l2_device *v4l2_dev = &ctx->dev->v4l2_dev;
+	u32 bitstream_buf, bitstream_size;
+	struct coda_dev *dev = ctx->dev;
+	struct coda_q_data *q_data_src, *q_data_dst;
+	struct vb2_buffer *buf;
+	u32 dst_fourcc;
+	u32 value;
+	int ret = 0;
+
+	q_data_src = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+	if (q->type == V4L2_BUF_TYPE_VIDEO_OUTPUT) {
+		if (q_data_src->fourcc == V4L2_PIX_FMT_H264) {
+#if 0
+			if (coda_get_bitstream_payload(ctx) < 512)
+				return -EINVAL;
+#endif
+		} else {
+			if (count < 1)
+				return -EINVAL;
+		}
+
+		ctx->streamon_out = 1;
+
+		if (coda_format_is_yuv(q_data_src->fourcc))
+			ctx->inst_type = CODA_INST_ENCODER;
+		else
+			ctx->inst_type = CODA_INST_DECODER;
+	} else {
+		if (count < 1)
+			return -EINVAL;
+
+		ctx->streamon_cap = 1;
+	}
+
+	/* Don't start the coda unless both queues are on */
+	if (!(ctx->streamon_out & ctx->streamon_cap))
+		return 0;
+
+	ctx->gopcounter = ctx->params.gop_size - 1;
+	buf = v4l2_m2m_next_dst_buf(ctx->m2m_ctx);
+	bitstream_buf = vb2_dma_contig_plane_dma_addr(buf, 0);
+	q_data_dst = get_q_data(ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+	bitstream_size = q_data_dst->sizeimage;
+	dst_fourcc = q_data_dst->fourcc;
+
+	ctx->codec = coda_find_codec(ctx->dev, q_data_src->fourcc,
+				     q_data_dst->fourcc);
+	if (!ctx->codec) {
+		v4l2_err(v4l2_dev, "couldn't tell instance type.\n");
+		return -EINVAL;
+	}
+	/* Save encoder mode into param. Will be used by coda_command_sync() */
+	ctx->params.codec_mode = ctx->codec->mode;
+
+v4l2_err(v4l2_dev, "Codec mode %d.\n", ctx->params.codec_mode);
+
+	if (!coda_is_initialized(dev)) {
+		v4l2_err(v4l2_dev, "coda is not initialized.\n");
+		return -EFAULT;
+	}
+
+	mutex_lock(&dev->coda_mutex);
+
+	coda_write(dev, ctx->parabuf.paddr, CODA_REG_BIT_PARA_BUF_ADDR);
+	coda_write(dev, bitstream_buf, CODA_REG_BIT_RD_PTR(ctx->idx));
+	coda_write(dev, bitstream_buf, CODA_REG_BIT_WR_PTR(ctx->idx));
+	switch (dev->devtype->product) {
+	case CODA_DX6:
+		coda_write(dev, CODADX6_STREAM_BUF_DYNALLOC_EN |
+			CODADX6_STREAM_BUF_PIC_RESET, CODA_REG_BIT_STREAM_CTRL);
+		break;
+	default:
+		coda_write(dev, CODA7_STREAM_BUF_DYNALLOC_EN |
+			CODA7_STREAM_BUF_PIC_RESET, CODA_REG_BIT_STREAM_CTRL);
+	}
+
+	/* Now do specific encoder/decoder stuff */
+	switch (ctx->inst_type) {
+	case CODA_INST_ENCODER:
+		if (dev->devtype->product == CODA_DX6) {
+			/* Configure the coda */
+			coda_write(dev, dev->iram_paddr, CODADX6_REG_BIT_SEARCH_RAM_BASE_ADDR);
+		}
+
+		ret = coda_start_streaming_encoder(q, count);
+		if (ret < 0)
+			goto out;
+		break;
+	case CODA_INST_DECODER:
+		ret = coda_start_streaming_decoder(q, count);
+		if (ret < 0)
+			goto out;
+		break;
+	}
+out:
+	mutex_unlock(&dev->coda_mutex);
+	return ret;
+}
+
+static int coda_stop_streaming(struct vb2_queue *q)
+{
+	struct coda_ctx *ctx = vb2_get_drv_priv(q);
+	struct coda_dev *dev = ctx->dev;
+
+	if (q->type == V4L2_BUF_TYPE_VIDEO_OUTPUT) {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "%s: output\n", __func__);
+		ctx->streamon_out = 0;
+	} else {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "%s: capture\n", __func__);
+		ctx->streamon_cap = 0;
+	}
+
+	/* Don't stop the coda unless both queues are off */
+	if (ctx->streamon_out || ctx->streamon_cap)
+		return 0;
+
+	cancel_delayed_work(&dev->timeout);
+
+	mutex_lock(&dev->coda_mutex);
+	v4l2_dbg(1, coda_debug, &dev->v4l2_dev,
+		 "%s: sent command 'SEQ_END' to coda\n", __func__);
+	if (coda_command_sync(ctx, CODA_COMMAND_SEQ_END)) {
+		v4l2_err(&dev->v4l2_dev,
+			 "CODA_COMMAND_SEQ_END failed\n");
+		return -ETIMEDOUT;
+	}
+	mutex_unlock(&dev->coda_mutex);
+
+	coda_free_framebuffers(ctx);
+
+	return 0;
+}
+
+static struct vb2_ops coda_qops = {
+	.queue_setup		= coda_queue_setup,
+	.buf_prepare		= coda_buf_prepare,
+	.buf_queue		= coda_buf_queue,
+	.wait_prepare		= coda_wait_prepare,
+	.wait_finish		= coda_wait_finish,
+	.start_streaming	= coda_start_streaming,
+	.stop_streaming		= coda_stop_streaming,
+};
+
+static int coda_s_ctrl(struct v4l2_ctrl *ctrl)
+{
+	struct coda_ctx *ctx =
+			container_of(ctrl->handler, struct coda_ctx, ctrls);
+
+	v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+		 "s_ctrl: id = %d, val = %d\n", ctrl->id, ctrl->val);
+
+	switch (ctrl->id) {
+	case V4L2_CID_HFLIP:
+		if (ctrl->val)
+			ctx->params.rot_mode |= CODA_MIR_HOR;
+		else
+			ctx->params.rot_mode &= ~CODA_MIR_HOR;
+		break;
+	case V4L2_CID_VFLIP:
+		if (ctrl->val)
+			ctx->params.rot_mode |= CODA_MIR_VER;
+		else
+			ctx->params.rot_mode &= ~CODA_MIR_VER;
+		break;
+	case V4L2_CID_MPEG_VIDEO_BITRATE:
+		ctx->params.bitrate = ctrl->val / 1000;
+		break;
+	case V4L2_CID_MPEG_VIDEO_GOP_SIZE:
+		ctx->params.gop_size = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_H264_I_FRAME_QP:
+		ctx->params.h264_intra_qp = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_H264_P_FRAME_QP:
+		ctx->params.h264_inter_qp = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_MPEG4_I_FRAME_QP:
+		ctx->params.mpeg4_intra_qp = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_MPEG4_P_FRAME_QP:
+		ctx->params.mpeg4_inter_qp = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MODE:
+		ctx->params.slice_mode = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MAX_MB:
+		ctx->params.slice_max_mb = ctrl->val;
+		break;
+	case V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MAX_BYTES:
+		ctx->params.slice_max_bits = ctrl->val * 8;
+		break;
+	case V4L2_CID_MPEG_VIDEO_HEADER_MODE:
+		break;
+	default:
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			"Invalid control, id=%d, val=%d\n",
+			ctrl->id, ctrl->val);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static struct v4l2_ctrl_ops coda_ctrl_ops = {
+	.s_ctrl = coda_s_ctrl,
+};
+
+static int coda_ctrls_setup(struct coda_ctx *ctx)
+{
+	v4l2_ctrl_handler_init(&ctx->ctrls, 9);
+
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_HFLIP, 0, 1, 1, 0);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_VFLIP, 0, 1, 1, 0);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_BITRATE, 0, 32767000, 1, 0);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_GOP_SIZE, 1, 60, 1, 16);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_H264_I_FRAME_QP, 1, 51, 1, 25);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_H264_P_FRAME_QP, 1, 51, 1, 25);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_MPEG4_I_FRAME_QP, 1, 31, 1, 2);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_MPEG4_P_FRAME_QP, 1, 31, 1, 2);
+	v4l2_ctrl_new_std_menu(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MODE,
+		V4L2_MPEG_VIDEO_MULTI_SICE_MODE_MAX_BYTES, 0x0,
+		V4L2_MPEG_VIDEO_MULTI_SLICE_MODE_SINGLE);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MAX_MB, 1, 0x3fffffff, 1, 1);
+	v4l2_ctrl_new_std(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_MULTI_SLICE_MAX_BYTES, 1, 0x3fffffff, 1, 500);
+	v4l2_ctrl_new_std_menu(&ctx->ctrls, &coda_ctrl_ops,
+		V4L2_CID_MPEG_VIDEO_HEADER_MODE,
+		V4L2_MPEG_VIDEO_HEADER_MODE_JOINED_WITH_1ST_FRAME,
+		(1 << V4L2_MPEG_VIDEO_HEADER_MODE_SEPARATE),
+		V4L2_MPEG_VIDEO_HEADER_MODE_JOINED_WITH_1ST_FRAME);
+
+	if (ctx->ctrls.error) {
+		v4l2_err(&ctx->dev->v4l2_dev, "control initialization error (%d)",
+			ctx->ctrls.error);
+		return -EINVAL;
+	}
+
+	return v4l2_ctrl_handler_setup(&ctx->ctrls);
+}
+
+static int coda_queue_init(void *priv, struct vb2_queue *src_vq,
+		      struct vb2_queue *dst_vq)
+{
+	struct coda_ctx *ctx = priv;
+	int ret;
+
+	src_vq->type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
+	src_vq->io_modes = VB2_MMAP | VB2_USERPTR;
+	src_vq->drv_priv = ctx;
+	src_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
+	src_vq->ops = &coda_qops;
+	src_vq->mem_ops = &vb2_dma_contig_memops;
+
+	ret = vb2_queue_init(src_vq);
+	if (ret)
+		return ret;
+
+	dst_vq->type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
+	dst_vq->io_modes = VB2_MMAP | VB2_USERPTR;
+	dst_vq->drv_priv = ctx;
+	dst_vq->buf_struct_size = sizeof(struct v4l2_m2m_buffer);
+	dst_vq->ops = &coda_qops;
+	dst_vq->mem_ops = &vb2_dma_contig_memops;
+
+	return vb2_queue_init(dst_vq);
+}
+
+static int coda_next_free_instance(struct coda_dev *dev)
+{
+	int idx = ffz(dev->instance_mask);
+
+	if ((idx < 0) ||
+	    (dev->devtype->product == CODA_DX6 && idx > CODADX6_MAX_INSTANCES))
+		return -EBUSY;
+
+	return idx;
+}
+
+static int coda_open(struct file *file)
+{
+	struct coda_dev *dev = video_drvdata(file);
+	struct coda_ctx *ctx = NULL;
+	int ret = 0;
+	int idx;
+
+	ctx = kzalloc(sizeof *ctx, GFP_KERNEL);
+	if (!ctx) {
+		return -ENOMEM;
+	}
+
+	idx = coda_next_free_instance(dev);
+	if (idx < 0) {
+		ret = idx;
+		goto err_coda_max;
+	}
+	set_bit(idx, &dev->instance_mask);
+
+	v4l2_fh_init(&ctx->fh, video_devdata(file));
+	file->private_data = &ctx->fh;
+	v4l2_fh_add(&ctx->fh);
+	ctx->dev = dev;
+	ctx->idx = idx;
+	switch (dev->devtype->product) {
+	case CODA_7541:
+		ctx->reg_idx = 0;
+		break;
+	default:
+		ctx->reg_idx = idx;
+	}
+
+	ret = clk_prepare_enable(dev->clk_per);
+	if (ret)
+		goto err_clk_per;
+
+	ret = clk_prepare_enable(dev->clk_ahb);
+	if (ret)
+		goto err_clk_ahb;
+
+	set_default_params(ctx);
+	ctx->m2m_ctx = v4l2_m2m_ctx_init(dev->m2m_dev, ctx,
+					 &coda_queue_init);
+	if (IS_ERR(ctx->m2m_ctx)) {
+		ret = PTR_ERR(ctx->m2m_ctx);
+
+		v4l2_err(&dev->v4l2_dev, "%s return error (%d)\n",
+			 __func__, ret);
+		goto err_ctx_init;
+	}
+	ret = coda_ctrls_setup(ctx);
+	if (ret) {
+		v4l2_err(&dev->v4l2_dev, "failed to setup coda controls\n");
+		goto err_ctrls_setup;
+	}
+
+	ctx->fh.ctrl_handler = &ctx->ctrls;
+
+	ret = coda_alloc_context_buf(ctx, &ctx->parabuf, CODA_PARA_BUF_SIZE);
+	if (ret < 0) {
+		v4l2_err(&dev->v4l2_dev, "failed to allocate parabuf");
+		goto err_dma_alloc;
+	}
+
+	ctx->bitstream.size = CODA_MAX_FRAME_SIZE;
+	ctx->bitstream.vaddr = dma_alloc_writecombine(&dev->plat_dev->dev,
+			ctx->bitstream.size, &ctx->bitstream.paddr, GFP_KERNEL);
+	if (!ctx->bitstream.vaddr) {
+		v4l2_err(&dev->v4l2_dev, "failed to allocate bitstream ringbuffer");
+		ret = -ENOMEM;
+		goto err_dma_writecombine;
+	}
+#ifdef USE_KFIFO
+	kfifo_init(&ctx->bitstream_fifo,
+		ctx->bitstream.vaddr, ctx->bitstream.size);
+#endif
+	mutex_init(&ctx->bitstream_mutex);
+	mutex_init(&ctx->buffer_mutex);
+
+	coda_lock(ctx);
+	list_add(&ctx->list, &dev->instances);
+	coda_unlock(ctx);
+
+	v4l2_dbg(1, coda_debug, &dev->v4l2_dev, "Created instance %d (%p)\n",
+		 ctx->idx, ctx);
+
+	return 0;
+
+err_dma_writecombine:
+	coda_free_context_buffers(ctx);
+	if (ctx->dev->devtype->product == CODA_DX6)
+		coda_free_aux_buf(dev, &ctx->workbuf);
+	coda_free_aux_buf(dev, &ctx->parabuf);
+err_dma_alloc:
+	v4l2_ctrl_handler_free(&ctx->ctrls);
+err_ctrls_setup:
+	v4l2_m2m_ctx_release(ctx->m2m_ctx);
+err_ctx_init:
+	clk_disable_unprepare(dev->clk_ahb);
+err_clk_ahb:
+	clk_disable_unprepare(dev->clk_per);
+err_clk_per:
+/*err:*/
+	v4l2_fh_del(&ctx->fh);
+	v4l2_fh_exit(&ctx->fh);
+	clear_bit(ctx->idx, &dev->instance_mask);
+err_coda_max:
+	kfree(ctx);
+	return ret;
+}
+
+static int coda_release(struct file *file)
+{
+	struct coda_dev *dev = video_drvdata(file);
+	struct coda_ctx *ctx = fh_to_ctx(file->private_data);
+
+	v4l2_dbg(1, coda_debug, &dev->v4l2_dev, "Releasing instance %p\n",
+		 ctx);
+
+	coda_lock(ctx);
+	list_del(&ctx->list);
+	coda_unlock(ctx);
+
+	dma_free_coherent(&dev->plat_dev->dev, CODA_PARA_BUF_SIZE,
+		ctx->parabuf.vaddr, ctx->parabuf.paddr);
+	v4l2_m2m_ctx_release(ctx->m2m_ctx);
+	v4l2_ctrl_handler_free(&ctx->ctrls);
+	clk_disable_unprepare(dev->clk_per);
+	clk_disable_unprepare(dev->clk_ahb);
+	v4l2_fh_del(&ctx->fh);
+	v4l2_fh_exit(&ctx->fh);
+	clear_bit(ctx->idx, &dev->instance_mask);
+	kfree(ctx);
+
+	return 0;
+}
+
+static unsigned int coda_poll(struct file *file,
+				 struct poll_table_struct *wait)
+{
+	struct coda_ctx *ctx = fh_to_ctx(file->private_data);
+	int ret;
+
+	coda_lock(ctx);
+	ret = v4l2_m2m_poll(file, ctx->m2m_ctx, wait);
+	coda_unlock(ctx);
+	return ret;
+}
+
+static int coda_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	struct coda_ctx *ctx = fh_to_ctx(file->private_data);
+
+	return v4l2_m2m_mmap(file, ctx->m2m_ctx, vma);
+}
+
+static const struct v4l2_file_operations coda_fops = {
+	.owner		= THIS_MODULE,
+	.open		= coda_open,
+	.release	= coda_release,
+	.poll		= coda_poll,
+	.unlocked_ioctl	= video_ioctl2,
+	.mmap		= coda_mmap,
+};
+
+static irqreturn_t coda_irq_handler(int irq, void *data)
+{
+	struct vb2_buffer *src_buf, *dst_buf;
+	struct coda_dev *dev = data;
+	u32 wr_ptr, start_ptr;
+	struct coda_ctx *ctx;
+
+	cancel_delayed_work(&dev->timeout);
+
+	/* read status register to attend the IRQ */
+	coda_read(dev, CODA_REG_BIT_INT_STATUS);
+	coda_write(dev, CODA_REG_BIT_INT_CLEAR_SET,
+		      CODA_REG_BIT_INT_CLEAR);
+
+	ctx = v4l2_m2m_get_curr_priv(dev->m2m_dev);
+	if (ctx == NULL) {
+		v4l2_err(&dev->v4l2_dev, "Instance released before the end of transaction\n");
+		mutex_unlock(&dev->coda_mutex);
+		return IRQ_HANDLED;
+	}
+
+	if (ctx->aborting) {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "task has been aborted\n");
+		mutex_unlock(&dev->coda_mutex);
+		return IRQ_HANDLED;
+	}
+
+	if (coda_isbusy(ctx->dev)) {
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev,
+			 "coda is still busy!!!!\n");
+		return IRQ_NONE;
+	}
+
+	src_buf = v4l2_m2m_src_buf_remove(ctx->m2m_ctx);
+	dst_buf = v4l2_m2m_dst_buf_remove(ctx->m2m_ctx);
+
+	switch (ctx->inst_type) {
+	case CODA_INST_DECODER:
+		break;
+	case CODA_INST_ENCODER:
+		/* Get results from the coda */
+		start_ptr = coda_read(dev, CODA_CMD_ENC_PIC_BB_START);
+		wr_ptr = coda_read(dev, CODA_REG_BIT_WR_PTR(ctx->idx));
+		/* Calculate bytesused field */
+		if (dst_buf->v4l2_buf.sequence == 0) {
+			vb2_set_plane_payload(dst_buf, 0, wr_ptr - start_ptr +
+					ctx->vpu_header_size[0] +
+					ctx->vpu_header_size[1] +
+					ctx->vpu_header_size[2]);
+		} else {
+			vb2_set_plane_payload(dst_buf, 0, wr_ptr - start_ptr);
+		}
+
+		v4l2_dbg(1, coda_debug, &ctx->dev->v4l2_dev, "frame size = %u\n",
+				wr_ptr - start_ptr);
+
+		coda_read(dev, CODA_RET_ENC_PIC_SLICE_NUM);
+		coda_read(dev, CODA_RET_ENC_PIC_FLAG);
+
+		if (coda_read(dev, CODA_RET_ENC_PIC_TYPE) == 0) {
+			dst_buf->v4l2_buf.flags |= V4L2_BUF_FLAG_KEYFRAME;
+			dst_buf->v4l2_buf.flags &= ~V4L2_BUF_FLAG_PFRAME;
+		} else {
+			dst_buf->v4l2_buf.flags |= V4L2_BUF_FLAG_PFRAME;
+			dst_buf->v4l2_buf.flags &= ~V4L2_BUF_FLAG_KEYFRAME;
+		}
+
+		v4l2_m2m_buf_done(src_buf, VB2_BUF_STATE_DONE);
+		v4l2_m2m_buf_done(dst_buf, VB2_BUF_STATE_DONE);
+
+		ctx->gopcounter--;
+		if (ctx->gopcounter < 0)
+			ctx->gopcounter = ctx->params.gop_size - 1;
+
+		v4l2_dbg(1, coda_debug, &dev->v4l2_dev,
+				"job finished: encoding frame (%d) (%s)\n",
+				dst_buf->v4l2_buf.sequence,
+				(dst_buf->v4l2_buf.flags & V4L2_BUF_FLAG_KEYFRAME) ?
+				"KEYFRAME" : "PFRAME");
+		break;
+	default:
+		break;
+	}
+	mutex_unlock(&dev->coda_mutex);
+
+	v4l2_m2m_job_finish(ctx->dev->m2m_dev, ctx->m2m_ctx);
+
+	return IRQ_HANDLED;
+}
+
+static void coda_timeout(struct work_struct *work)
+{
+	struct coda_ctx *ctx;
+	struct coda_dev *dev = container_of(to_delayed_work(work),
+					    struct coda_dev, timeout);
+
+	dev_err(&dev->plat_dev->dev, "CODA PIC_RUN timeout, stopping all streams\n");
+
+	mutex_lock(&dev->dev_mutex);
+	list_for_each_entry(ctx, &dev->instances, list) {
+		v4l2_m2m_streamoff(NULL, ctx->m2m_ctx, V4L2_BUF_TYPE_VIDEO_OUTPUT);
+		v4l2_m2m_streamoff(NULL, ctx->m2m_ctx, V4L2_BUF_TYPE_VIDEO_CAPTURE);
+	}
+	mutex_unlock(&dev->dev_mutex);
+
+	mutex_unlock(&dev->coda_mutex);
+	ctx = v4l2_m2m_get_curr_priv(dev->m2m_dev);
+	v4l2_m2m_job_finish(ctx->dev->m2m_dev, ctx->m2m_ctx);
+}
+
+static u32 coda_supported_firmwares[] = {
+	CODA_FIRMWARE_VERNUM(CODA_DX6, 2, 2, 5),
+	CODA_FIRMWARE_VERNUM(CODA_7541, 13, 4, 29),
+};
+
+static bool coda_firmware_supported(u32 vernum)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(coda_supported_firmwares); i++)
+		if (vernum == coda_supported_firmwares[i])
+			return true;
+	return false;
+}
+
+static char *coda_product_name(int product)
+{
+	static char buf[9];
+
+	switch (product) {
+	case CODA_DX6:
+		return "CodaDx6";
+	case CODA_7541:
+		return "CODA7541";
+	default:
+		snprintf(buf, sizeof(buf), "(0x%04x)", product);
+		return buf;
+	}
+}
+
+static int coda_hw_init(struct coda_dev *dev)
+{
+	u16 product, major, minor, release;
+	u32 data;
+	u16 *p;
+	int i;
+
+	clk_prepare_enable(dev->clk_per);
+	clk_prepare_enable(dev->clk_ahb);
+
+	/*
+	 * Copy the first CODA_ISRAM_SIZE in the internal SRAM.
+	 * The 16-bit chars in the code buffer are in memory access
+	 * order, re-sort them to CODA order for register download.
+	 * Data in this SRAM survives a reboot.
+	 */
+	p = (u16 *)dev->codebuf.vaddr;
+	if (dev->devtype->product == CODA_DX6) {
+		for (i = 0; i < (CODA_ISRAM_SIZE / 2); i++)  {
+			data = CODA_DOWN_ADDRESS_SET(i) |
+				CODA_DOWN_DATA_SET(p[i ^ 1]);
+			coda_write(dev, data, CODA_REG_BIT_CODE_DOWN);
+		}
+	} else {
+		for (i = 0; i < (CODA_ISRAM_SIZE / 2); i++) {
+			data = CODA_DOWN_ADDRESS_SET(i) |
+				CODA_DOWN_DATA_SET(p[round_down(i, 4) +
+							3 - (i % 4)]);
+			coda_write(dev, data, CODA_REG_BIT_CODE_DOWN);
+		}
+	}
+
+	/* Clear registers */
+	for (i = 0; i < 64; i++)
+		coda_write(dev, 0, CODA_REG_BIT_CODE_BUF_ADDR + i * 4);
+
+	/* Tell the BIT where to find everything it needs */
+	coda_write(dev, dev->workbuf.paddr,
+		      CODA_REG_BIT_WORK_BUF_ADDR);
+	coda_write(dev, dev->codebuf.paddr,
+		      CODA_REG_BIT_CODE_BUF_ADDR);
+	coda_write(dev, 0, CODA_REG_BIT_CODE_RUN);
+
+	/* Set default values */
+	switch (dev->devtype->product) {
+	case CODA_DX6:
+		coda_write(dev, CODADX6_STREAM_BUF_PIC_FLUSH, CODA_REG_BIT_STREAM_CTRL);
+		break;
+	default:
+		coda_write(dev, CODA7_STREAM_BUF_PIC_FLUSH, CODA_REG_BIT_STREAM_CTRL);
+	}
+	coda_write(dev, 0, CODA_REG_BIT_FRAME_MEM_CTRL);
+
+	if (dev->devtype->product != CODA_DX6)
+		coda_write(dev, 0, CODA7_REG_BIT_AXI_SRAM_USE);
+
+	coda_write(dev, CODA_INT_INTERRUPT_ENABLE,
+		      CODA_REG_BIT_INT_ENABLE);
+
+	/* Reset VPU and start processor */
+	data = coda_read(dev, CODA_REG_BIT_CODE_RESET);
+	data |= CODA_REG_RESET_ENABLE;
+	coda_write(dev, data, CODA_REG_BIT_CODE_RESET);
+	udelay(10);
+	data &= ~CODA_REG_RESET_ENABLE;
+	coda_write(dev, data, CODA_REG_BIT_CODE_RESET);
+	coda_write(dev, CODA_REG_RUN_ENABLE, CODA_REG_BIT_CODE_RUN);
+
+	/* Load firmware */
+	coda_write(dev, 0, CODA_CMD_FIRMWARE_VERNUM);
+	coda_write(dev, CODA_REG_BIT_BUSY_FLAG, CODA_REG_BIT_BUSY);
+	coda_write(dev, 0, CODA_REG_BIT_RUN_INDEX);
+	coda_write(dev, 0, CODA_REG_BIT_RUN_COD_STD);
+	coda_write(dev, CODA_COMMAND_FIRMWARE_GET, CODA_REG_BIT_RUN_COMMAND);
+	if (coda_wait_timeout(dev)) {
+		clk_disable_unprepare(dev->clk_per);
+		clk_disable_unprepare(dev->clk_ahb);
+		v4l2_err(&dev->v4l2_dev, "firmware get command error\n");
+		return -EIO;
+	}
+
+	/* Check we are compatible with the loaded firmware */
+	data = coda_read(dev, CODA_CMD_FIRMWARE_VERNUM);
+	product = CODA_FIRMWARE_PRODUCT(data);
+	major = CODA_FIRMWARE_MAJOR(data);
+	minor = CODA_FIRMWARE_MINOR(data);
+	release = CODA_FIRMWARE_RELEASE(data);
+
+	clk_disable_unprepare(dev->clk_per);
+	clk_disable_unprepare(dev->clk_ahb);
+
+	if (product != dev->devtype->product) {
+		v4l2_err(&dev->v4l2_dev, "Wrong firmware. Hw: %s, Fw: %s,"
+			 " Version: %u.%u.%u\n",
+			 coda_product_name(dev->devtype->product),
+			 coda_product_name(product), major, minor, release);
+		return -EINVAL;
+	}
+
+	v4l2_info(&dev->v4l2_dev, "Initialized %s.\n",
+		  coda_product_name(product));
+
+	if (coda_firmware_supported(data)) {
+		v4l2_info(&dev->v4l2_dev, "Firmware version: %u.%u.%u\n",
+			  major, minor, release);
+	} else {
+		v4l2_warn(&dev->v4l2_dev, "Unsupported firmware version: "
+			  "%u.%u.%u\n", major, minor, release);
+	}
+
+	return 0;
+}
+
+static void coda_fw_callback(const struct firmware *fw, void *context)
+{
+	struct coda_dev *dev = context;
+	struct platform_device *pdev = dev->plat_dev;
+	int ret;
+
+	if (!fw) {
+		v4l2_err(&dev->v4l2_dev, "firmware request failed\n");
+		return;
+	}
+
+	/* allocate auxiliary per-device code buffer for the BIT processor */
+	dev->codebuf.size = fw->size;
+	dev->codebuf.vaddr = dma_alloc_coherent(&pdev->dev, fw->size,
+						    &dev->codebuf.paddr,
+						    GFP_KERNEL);
+	if (!dev->codebuf.vaddr) {
+		dev_err(&pdev->dev, "failed to allocate code buffer\n");
+		return;
+	}
+
+	/* Copy the whole firmware image to the code buffer */
+	memcpy(dev->codebuf.vaddr, fw->data, fw->size);
+	release_firmware(fw);
+
+	ret = coda_hw_init(dev);
+	if (ret) {
+		v4l2_err(&dev->v4l2_dev, "HW initialization failed\n");
+		return;
+	}
+
+	dev->vfd.fops	= &coda_fops,
+	dev->vfd.ioctl_ops	= &coda_ioctl_ops;
+	dev->vfd.release	= video_device_release_empty,
+	dev->vfd.lock	= &dev->dev_mutex;
+	dev->vfd.v4l2_dev	= &dev->v4l2_dev;
+	dev->vfd.vfl_dir	= VFL_DIR_M2M;
+	snprintf(dev->vfd.name, sizeof(dev->vfd.name), "%s", CODA_NAME);
+	video_set_drvdata(&dev->vfd, dev);
+
+	dev->alloc_ctx = vb2_dma_contig_init_ctx(&pdev->dev);
+	if (IS_ERR(dev->alloc_ctx)) {
+		v4l2_err(&dev->v4l2_dev, "Failed to alloc vb2 context\n");
+		return;
+	}
+
+	dev->m2m_dev = v4l2_m2m_init(&coda_m2m_ops);
+	if (IS_ERR(dev->m2m_dev)) {
+		v4l2_err(&dev->v4l2_dev, "Failed to init mem2mem device\n");
+		goto rel_ctx;
+	}
+
+	ret = video_register_device(&dev->vfd, VFL_TYPE_GRABBER, 0);
+	if (ret) {
+		v4l2_err(&dev->v4l2_dev, "Failed to register video device\n");
+		goto rel_m2m;
+	}
+	v4l2_info(&dev->v4l2_dev, "codec registered as /dev/video%d\n",
+		  dev->vfd.num);
+
+	return;
+
+rel_m2m:
+	v4l2_m2m_release(dev->m2m_dev);
+rel_ctx:
+	vb2_dma_contig_cleanup_ctx(dev->alloc_ctx);
+}
+
+static int coda_firmware_request(struct coda_dev *dev)
+{
+	char *fw = dev->devtype->firmware;
+
+	dev_dbg(&dev->plat_dev->dev, "requesting firmware '%s' for %s\n", fw,
+		coda_product_name(dev->devtype->product));
+
+	return request_firmware_nowait(THIS_MODULE, true,
+		fw, &dev->plat_dev->dev, GFP_KERNEL, dev, coda_fw_callback);
+}
+
+enum coda_platform {
+	CODA_IMX27,
+	CODA_IMX53,
+};
+
+static const struct coda_devtype coda_devdata[] = {
+	[CODA_IMX27] = {
+		.firmware   = "v4l-codadx6-imx27.bin",
+		.product    = CODA_DX6,
+		.codecs     = codadx6_codecs,
+		.num_codecs = ARRAY_SIZE(codadx6_codecs),
+	},
+	[CODA_IMX53] = {
+		.firmware   = "v4l-coda7541-imx53.bin",
+		.product    = CODA_7541,
+		.codecs     = coda7_codecs,
+		.num_codecs = ARRAY_SIZE(coda7_codecs),
+	},
+};
+
+static struct platform_device_id coda_platform_ids[] = {
+	{ .name = "coda-imx27", .driver_data = CODA_IMX27 },
+	{ .name = "coda-imx53", .driver_data = CODA_IMX53 },
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(platform, coda_platform_ids);
+
+#ifdef CONFIG_OF
+static const struct of_device_id coda_dt_ids[] = {
+	{ .compatible = "fsl,imx27-vpu", .data = &coda_platform_ids[CODA_IMX27] },
+	{ .compatible = "fsl,imx53-vpu", .data = &coda_devdata[CODA_IMX53] },
+	{ /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(of, coda_dt_ids);
+#endif
+
+static int coda_probe(struct platform_device *pdev)
+{
+	const struct of_device_id *of_id =
+			of_match_device(of_match_ptr(coda_dt_ids), &pdev->dev);
+	const struct platform_device_id *pdev_id;
+	struct coda_dev *dev;
+	struct resource *res;
+	int ret, irq;
+
+	dev = devm_kzalloc(&pdev->dev, sizeof *dev, GFP_KERNEL);
+	if (!dev) {
+		dev_err(&pdev->dev, "Not enough memory for %s\n",
+			CODA_NAME);
+		return -ENOMEM;
+	}
+
+	spin_lock_init(&dev->irqlock);
+	INIT_LIST_HEAD(&dev->instances);
+	INIT_DELAYED_WORK(&dev->timeout, coda_timeout);
+
+	dev->plat_dev = pdev;
+	dev->clk_per = devm_clk_get(&pdev->dev, "per");
+	if (IS_ERR(dev->clk_per)) {
+		dev_err(&pdev->dev, "Could not get per clock\n");
+		return PTR_ERR(dev->clk_per);
+	}
+
+	dev->clk_ahb = devm_clk_get(&pdev->dev, "ahb");
+	if (IS_ERR(dev->clk_ahb)) {
+		dev_err(&pdev->dev, "Could not get ahb clock\n");
+		return PTR_ERR(dev->clk_ahb);
+	}
+
+	/* Get  memory for physical registers */
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (res == NULL) {
+		dev_err(&pdev->dev, "failed to get memory region resource\n");
+		return -ENOENT;
+	}
+
+	if (devm_request_mem_region(&pdev->dev, res->start,
+			resource_size(res), CODA_NAME) == NULL) {
+		dev_err(&pdev->dev, "failed to request memory region\n");
+		return -ENOENT;
+	}
+	dev->regs_base = devm_ioremap(&pdev->dev, res->start,
+				      resource_size(res));
+	if (!dev->regs_base) {
+		dev_err(&pdev->dev, "failed to ioremap address region\n");
+		return -ENOENT;
+	}
+
+	/* IRQ */
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0) {
+		dev_err(&pdev->dev, "failed to get irq resource\n");
+		return -ENOENT;
+	}
+
+	if (devm_request_irq(&pdev->dev, irq, coda_irq_handler,
+		0, CODA_NAME, dev) < 0) {
+		dev_err(&pdev->dev, "failed to request irq\n");
+		return -ENOENT;
+	}
+
+	ret = v4l2_device_register(&pdev->dev, &dev->v4l2_dev);
+	if (ret)
+		return ret;
+
+	mutex_init(&dev->dev_mutex);
+	mutex_init(&dev->coda_mutex);
+
+	pdev_id = of_id ? of_id->data : platform_get_device_id(pdev);
+
+	if (of_id) {
+		dev->devtype = of_id->data;
+	} else if (pdev_id) {
+		dev->devtype = &coda_devdata[pdev_id->driver_data];
+	} else {
+		v4l2_device_unregister(&dev->v4l2_dev);
+		return -EINVAL;
+	}
+
+	/* allocate auxiliary per-device buffers for the BIT processor */
+	switch (dev->devtype->product) {
+	case CODA_DX6:
+		dev->workbuf.size = CODADX6_WORK_BUF_SIZE;
+		break;
+	default:
+		dev->workbuf.size = CODA7_WORK_BUF_SIZE;
+	}
+	dev->workbuf.vaddr = dma_alloc_coherent(&pdev->dev, dev->workbuf.size,
+						    &dev->workbuf.paddr,
+						    GFP_KERNEL);
+	if (!dev->workbuf.vaddr) {
+		dev_err(&pdev->dev, "failed to allocate work buffer\n");
+		v4l2_device_unregister(&dev->v4l2_dev);
+		return -ENOMEM;
+	}
+
+	if (dev->devtype->product == CODA_DX6) {
+		dev->iram_paddr = 0xffff4c00;
+	} else {
+		void __iomem *iram_vaddr;
+
+		iram_vaddr = iram_alloc(CODA7_IRAM_SIZE,
+					&dev->iram_paddr);
+		if (!iram_vaddr) {
+			dev_err(&pdev->dev, "unable to alloc iram\n");
+			return -ENOMEM;
+		}
+	}
+
+	platform_set_drvdata(pdev, dev);
+
+	return coda_firmware_request(dev);
+}
+
+static int coda_remove(struct platform_device *pdev)
+{
+	struct coda_dev *dev = platform_get_drvdata(pdev);
+
+	video_unregister_device(&dev->vfd);
+	if (dev->m2m_dev)
+		v4l2_m2m_release(dev->m2m_dev);
+	if (dev->alloc_ctx)
+		vb2_dma_contig_cleanup_ctx(dev->alloc_ctx);
+	v4l2_device_unregister(&dev->v4l2_dev);
+	if (dev->iram_paddr)
+		iram_free(dev->iram_paddr, CODA7_IRAM_SIZE);
+	if (dev->codebuf.vaddr)
+		dma_free_coherent(&pdev->dev, dev->codebuf.size,
+				  &dev->codebuf.vaddr, dev->codebuf.paddr);
+	if (dev->workbuf.vaddr)
+		dma_free_coherent(&pdev->dev, dev->workbuf.size, &dev->workbuf.vaddr,
+			  dev->workbuf.paddr);
+	return 0;
+}
+
+static struct platform_driver coda_driver = {
+	.probe	= coda_probe,
+	.remove	= coda_remove,
+	.driver	= {
+		.name	= CODA_NAME,
+		.owner	= THIS_MODULE,
+		.of_match_table = of_match_ptr(coda_dt_ids),
+	},
+	.id_table = coda_platform_ids,
+};
+
+module_platform_driver(coda_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Javier Martin <javier.martin@vista-silicon.com>");
+MODULE_DESCRIPTION("Coda multi-standard codec V4L2 driver");

Re: [CODA] Info about internal buffers in CodaDX6

[Gaëtan Carlier]

Hello,
No one can give me more information about these registers ?
Thank you.
Bets regards,
Gaëtan Carlier.

On 06/03/2014 10:32 AM, Gaëtan Carlier wrote:
> Dear,
> I am back to add support of h.264 decoding using Coda DX6 on i.MX27
> (after long months of inactivity).
> I base my work on driver from linux 2.6.22 (libvpu) and last coda.c from
> linux-next/master.
>
> When I send DEC_SEQ_INIT command, it fails but I don't know why.
> 1) Which internal buffers do Coda DX6 really have/used for decoding
> PARABUF, WORKBUF, PSBUF, ...) ?
> 2) What is their role ?
> 3) I see in some code that there is a command
> CODA_RET_DEC_SEQ_ERR_REASON (0x1E0), which has the same opcode has
> RET_DEC_SEQ_NEXT_FRAME_NUM, but when I run this command after
> DEC_SEQ_INIT, it returns 1 that does not seems to be correct error
> (regarding RetCode enum in vpu_lib.h in libvpu)
>
> Code is based on 3.6.0 kernel revision with some backport from more
> recent version of coda.c
>
> Thanks a lot for your help.
> Best regards,
> Gaëtan Carlier.