4 files changed, 40 insertions, 9 deletions

diff --git a/drivers/media/platform/amphion/vdec.c b/drivers/media/platform/amphion/vdec.c
index 6a38a0fa0e2d..85d518823159 100644
--- a/drivers/media/platform/amphion/vdec.c
+++ b/drivers/media/platform/amphion/vdec.c
@@ -805,7 +805,7 @@ static void vdec_buf_done(struct vpu_inst *inst, struct vpu_frame_info *frame)
 	cur_fmt = vpu_get_format(inst, inst->cap_format.type);
 	vbuf = &vpu_buf->m2m_buf.vb;
 	if (vbuf->vb2_buf.index != frame->id)
-		dev_err(inst->dev, "[%d] buffer id(%d, %d) dismatch\n",
+		dev_err(inst->dev, "[%d] buffer id(%d, %d) mismatch\n",
 			inst->id, vbuf->vb2_buf.index, frame->id);
 
 	if (vpu_get_buffer_state(vbuf) == VPU_BUF_STATE_READY && vdec->params.display_delay_enable)
diff --git a/drivers/media/platform/amphion/vpu.h b/drivers/media/platform/amphion/vpu.h
index 22f0da26ccec..1451549c9dd2 100644
--- a/drivers/media/platform/amphion/vpu.h
+++ b/drivers/media/platform/amphion/vpu.h
@@ -162,7 +162,6 @@ struct vpu_core {
 	struct delayed_work msg_delayed_work;
 	struct kfifo msg_fifo;
 	void *msg_buffer;
-	unsigned int msg_buffer_size;
 
 	struct vpu_dev *vpu;
 	void *iface;
diff --git a/drivers/media/platform/amphion/vpu_core.c b/drivers/media/platform/amphion/vpu_core.c
index 8df85c14ab3f..da00f5fc0e5d 100644
--- a/drivers/media/platform/amphion/vpu_core.c
+++ b/drivers/media/platform/amphion/vpu_core.c
@@ -250,6 +250,7 @@ static void vpu_core_get_vpu(struct vpu_core *core)
 static int vpu_core_register(struct device *dev, struct vpu_core *core)
 {
 	struct vpu_dev *vpu = dev_get_drvdata(dev);
+	unsigned int buffer_size;
 	int ret = 0;
 
 	dev_dbg(core->dev, "register core %s\n", vpu_core_type_desc(core->type));
@@ -263,14 +264,14 @@ static int vpu_core_register(struct device *dev, struct vpu_core *core)
 	}
 	INIT_WORK(&core->msg_work, vpu_msg_run_work);
 	INIT_DELAYED_WORK(&core->msg_delayed_work, vpu_msg_delayed_work);
-	core->msg_buffer_size = roundup_pow_of_two(VPU_MSG_BUFFER_SIZE);
-	core->msg_buffer = vzalloc(core->msg_buffer_size);
+	buffer_size = roundup_pow_of_two(VPU_MSG_BUFFER_SIZE);
+	core->msg_buffer = vzalloc(buffer_size);
 	if (!core->msg_buffer) {
 		dev_err(core->dev, "failed allocate buffer for fifo\n");
 		ret = -ENOMEM;
 		goto error;
 	}
-	ret = kfifo_init(&core->msg_fifo, core->msg_buffer, core->msg_buffer_size);
+	ret = kfifo_init(&core->msg_fifo, core->msg_buffer, buffer_size);
 	if (ret) {
 		dev_err(core->dev, "failed init kfifo\n");
 		goto error;
diff --git a/drivers/media/platform/amphion/vpu_malone.c b/drivers/media/platform/amphion/vpu_malone.c
index 4769c053c6c2..feca7d4220ed 100644
--- a/drivers/media/platform/amphion/vpu_malone.c
+++ b/drivers/media/platform/amphion/vpu_malone.c
@@ -3,6 +3,7 @@
  * Copyright 2020-2021 NXP
  */
 
+#include <linux/bitfield.h>
 #include <linux/init.h>
 #include <linux/interconnect.h>
 #include <linux/ioctl.h>
@@ -25,6 +26,10 @@
 #include "vpu_imx8q.h"
 #include "vpu_malone.h"
 
+static bool low_latency;
+module_param(low_latency, bool, 0644);
+MODULE_PARM_DESC(low_latency, "Set low latency frame flush mode: 0 (disable) or 1 (enable)");
+
 #define CMD_SIZE			25600
 #define MSG_SIZE			25600
 #define CODEC_SIZE			0x1000
@@ -68,6 +73,12 @@
 
 #define MALONE_DEC_FMT_RV_MASK			BIT(21)
 
+#define MALONE_VERSION_MASK			0xFFFFF
+#define MALONE_VERSION(maj, min, inc)		\
+		(FIELD_PREP(0xF0000, maj) | FIELD_PREP(0xFF00, min) | FIELD_PREP(0xFF, inc))
+#define CHECK_VERSION(iface, maj, min)		\
+		(FIELD_GET(MALONE_VERSION_MASK, (iface)->fw_version) >= MALONE_VERSION(maj, min, 0))
+
 enum vpu_malone_stream_input_mode {
 	INVALID_MODE = 0,
 	FRAME_LVL,
@@ -332,6 +343,8 @@ struct vpu_dec_ctrl {
 	u32 buf_addr[VID_API_NUM_STREAMS];
 };
 
+static const struct malone_padding_scode *get_padding_scode(u32 type, u32 fmt);
+
 u32 vpu_malone_get_data_size(void)
 {
 	return sizeof(struct vpu_dec_ctrl);
@@ -654,9 +667,15 @@ static int vpu_malone_set_params(struct vpu_shared_addr *shared,
 		hc->jpg[instance].jpg_mjpeg_interlaced = 0;
 	}
 
-	hc->codec_param[instance].disp_imm = params->display_delay_enable ? 1 : 0;
-	if (malone_format != MALONE_FMT_AVC)
+	if (params->display_delay_enable &&
+	    get_padding_scode(SCODE_PADDING_BUFFLUSH, params->codec_format))
+		hc->codec_param[instance].disp_imm = 1;
+	else
+		hc->codec_param[instance].disp_imm = 0;
+
+	if (params->codec_format == V4L2_PIX_FMT_HEVC && !CHECK_VERSION(iface, 1, 9))
 		hc->codec_param[instance].disp_imm = 0;
+
 	hc->codec_param[instance].dbglog_enable = 0;
 	iface->dbglog_desc.level = 0;
 
@@ -1023,6 +1042,7 @@ static const struct malone_padding_scode padding_scodes[] = {
 	{SCODE_PADDING_EOS,      V4L2_PIX_FMT_JPEG,        {0x0, 0x0}},
 	{SCODE_PADDING_BUFFLUSH, V4L2_PIX_FMT_H264,        {0x15010000, 0x0}},
 	{SCODE_PADDING_BUFFLUSH, V4L2_PIX_FMT_H264_MVC,    {0x15010000, 0x0}},
+	{SCODE_PADDING_BUFFLUSH, V4L2_PIX_FMT_HEVC,        {0x3e010000, 0x20}},
 };
 
 static const struct malone_padding_scode padding_scode_dft = {0x0, 0x0};
@@ -1057,8 +1077,11 @@ static int vpu_malone_add_padding_scode(struct vpu_buffer *stream_buffer,
 	int ret;
 
 	ps = get_padding_scode(scode_type, pixelformat);
-	if (!ps)
+	if (!ps) {
+		if (scode_type == SCODE_PADDING_BUFFLUSH)
+			return 0;
 		return -EINVAL;
+	}
 
 	wptr = readl(&str_buf->wptr);
 	if (wptr < stream_buffer->phys || wptr > stream_buffer->phys + stream_buffer->length)
@@ -1562,7 +1585,15 @@ static int vpu_malone_input_frame_data(struct vpu_malone_str_buffer __iomem *str
 
 	vpu_malone_update_wptr(str_buf, wptr);
 
-	if (disp_imm && !vpu_vb_is_codecconfig(vbuf)) {
+	/*
+	 * Enable the low latency flush mode if display delay is set to 0
+	 * or the low latency frame flush mode if it is set to 1.
+	 * The low latency flush mode requires some padding data to be appended to each frame,
+	 * but there must not be any padding data between the sequence header and the frame.
+	 * This module is currently only supported for the H264 and HEVC formats,
+	 * for other formats, vpu_malone_add_scode() will return 0.
+	 */
+	if ((disp_imm || low_latency) && !vpu_vb_is_codecconfig(vbuf)) {
 		ret = vpu_malone_add_scode(inst->core->iface,
 					   inst->id,
 					   &inst->stream_buffer,