Re: [PATCH v4 6/8] drm/mediatek: add dsi transfer function

Hi, YT:

Some comments inline.

On Fri, 2016-07-15 at 18:07 +0800, YT Shen wrote:

From: shaoming chen shaoming.chen@mediatek.com

add dsi read/write commands for transfer function

Signed-off-by: shaoming chen shaoming.chen@mediatek.com

drivers/gpu/drm/mediatek/mtk_dsi.c | 322 ++++++++++++++++++++++++++++++++++++ 1 file changed, 322 insertions(+)

diff --git a/drivers/gpu/drm/mediatek/mtk_dsi.c b/drivers/gpu/drm/mediatek/mtk_dsi.c index de5ad7f..1f99894 100644 --- a/drivers/gpu/drm/mediatek/mtk_dsi.c +++ b/drivers/gpu/drm/mediatek/mtk_dsi.c @@ -24,6 +24,7 @@ #include <linux/of_graph.h> #include <linux/phy/phy.h> #include <linux/platform_device.h> +#include <video/mipi_display.h> #include <video/videomode.h>

#include "mtk_drm_ddp_comp.h" @@ -80,8 +81,16 @@ #define DSI_HBP_WC 0x54 #define DSI_HFP_WC 0x58

+#define DSI_CMDQ_SIZE 0x60 +#define CMDQ_SIZE 0x3f

#define DSI_HSTX_CKL_WC 0x64

+#define DSI_RX_DATA0 0x74 +#define DSI_RX_DATA1 0x78 +#define DSI_RX_DATA2 0x7c +#define DSI_RX_DATA3 0x80

#define DSI_RACK 0x84 #define RACK BIT(0)

@@ -117,8 +126,25 @@ #define CLK_HS_POST (0xff << 8) #define CLK_HS_EXIT (0xff << 16)

+#define DSI_CMDQ0 0x180

#define NS_TO_CYCLE(n, c) ((n) / (c) + (((n) % (c)) ? 1 : 0))

+#define MTK_DSI_HOST_IS_READ(type) \

• ((type == MIPI_DSI_GENERIC_READ_REQUEST_0_PARAM) || \
• (type == MIPI_DSI_GENERIC_READ_REQUEST_1_PARAM) || \
• (type == MIPI_DSI_GENERIC_READ_REQUEST_2_PARAM) || \
• (type == MIPI_DSI_DCS_READ))

+#define MTK_DSI_HOST_IS_WRITE(type) \

• ((type == MIPI_DSI_GENERIC_SHORT_WRITE_0_PARAM) || \
• (type == MIPI_DSI_GENERIC_SHORT_WRITE_1_PARAM) || \
• (type == MIPI_DSI_GENERIC_SHORT_WRITE_2_PARAM) || \
• (type == MIPI_DSI_DCS_SHORT_WRITE) || \
• (type == MIPI_DSI_DCS_SHORT_WRITE_PARAM) || \
• (type == MIPI_DSI_GENERIC_LONG_WRITE) || \
• (type == MIPI_DSI_DCS_LONG_WRITE))

struct phy;

struct mtk_dsi { @@ -148,6 +174,38 @@ struct mtk_dsi { int irq_num, irq_data; };

+struct dsi_cmd_t0 {

• u8 config;
• u8 type;
• u8 data0;
• u8 data1;

+};

+struct dsi_cmd_t2 {

• u8 config;
• u8 type;
• u16 wc16;
• u8 *pdata;

+};

+struct dsi_rx_data {

• u8 byte0;
• u8 byte1;
• u8 byte2;
• u8 byte3;

+};

+struct dsi_tx_cmdq {

• u8 byte0;
• u8 byte1;
• u8 byte2;
• u8 byte3;

+};

+struct dsi_tx_cmdq_regs {

• struct dsi_tx_cmdq data[128];

+};

enum { DSI_INT_SLEEPOUT_DONE_FLAG = BIT(6), DSI_INT_VM_CMD_DONE_FLAG = BIT(5), @@ -858,9 +916,273 @@ static int mtk_dsi_host_detach(struct mipi_dsi_host *host, return 0; }

+static void mtk_dsi_set_cmdq(void __iomem *reg, u32 mask, u32 data) +{

• u32 temp = readl(reg);
• writel((temp & ~mask) | (data & mask), reg);

+}

+static void mtk_dsi_wait_for_idle(struct mtk_dsi *dsi) +{

• u32 timeout_ms = 500000; /* total 1s ~ 2s timeout */
• while (timeout_ms--) {

• if (!(readl(dsi->regs + DSI_INTSTA) & DSI_BUSY))
  

• 	break;
  

• usleep_range(2, 4);
  

• }
• if (timeout_ms == 0) {

• dev_info(dsi->dev, "polling dsi wait not busy timeout!\n");
  

• mtk_dsi_enable(dsi);
  

• mtk_dsi_reset_engine(dsi);
  

• }

+}

+static ssize_t mtk_dsi_host_read_cmd(struct mtk_dsi *dsi,

• 		     const struct mipi_dsi_msg *msg)
  

+{

• u8 max_try_count = 5;
• u32 recv_data_cnt, tmp_val;
• u32 recv_data0, recv_data1, recv_data2, recv_data3;
• struct dsi_rx_data read_data0, read_data1, read_data2, read_data3;
• struct dsi_cmd_t0 t0;
• s32 ret;
• u8 *buffer = msg->rx_buf;
• u8 buffer_size = msg->rx_len;
• u8 packet_type;
• if (readl(dsi->regs + DSI_MODE_CTRL) & 0x03) {

• dev_info(dsi->dev, "dsi engine is not command mode\n");
  

• return -1;
  

• }
• if (!buffer) {

• dev_info(dsi->dev, "dsi receive buffer size may be NULL\n");
  

• return -1;
  

• }
• do {

• if (max_try_count == 0) {
  

• 	dev_info(dsi->dev, "dsi engine read counter has been maxinum\n");
  

• 	return -1;
  

• }
  

• max_try_count--;
  

• recv_data_cnt = 0;
  

• mtk_dsi_wait_for_idle(dsi);
  

• t0.config = 0x04;
  

• t0.data0 = *((u8 *)(msg->tx_buf));
  

• if (buffer_size < 0x3)
  

It's better to use '3' instead of '0x3'.

• 	t0.type = MIPI_DSI_DCS_READ;
  

• else
  

• 	t0.type = MIPI_DSI_GENERIC_READ_REQUEST_1_PARAM;
  

• t0.data1 = 0;
  

• tmp_val = (t0.data1 << 24) | (t0.data0 << 16) | (t0.type << 8) |
  

• 	t0.config;
  

Why need 'struct dsi_cmd_t0 t0'? I think code is more simple when get rid of this structure. The code looks like:

tmp_val = (data1 << 24) | (data0 << 16) | (type << 8) | config;

• writel(tmp_val, dsi->regs + DSI_CMDQ0);
  

• mtk_dsi_mask(dsi, DSI_CMDQ_SIZE, CMDQ_SIZE, 1);
  

• mtk_dsi_mask(dsi, DSI_RACK, RACK, RACK);
  

• mtk_dsi_mask(dsi, DSI_INTSTA, LPRX_RD_RDY_INT_FLAG,
  

• 	     ~LPRX_RD_RDY_INT_FLAG);
  

• mtk_dsi_mask(dsi, DSI_INTSTA, CMD_DONE_INT_FLAG,
  

• 	     ~CMD_DONE_INT_FLAG);
  

I think interrupt status should be cleared in irq handler. So you need not to clear it here.

• mtk_dsi_mask(dsi, DSI_INTEN, DSI_INT_LPRX_RD_RDY_FLAG,
  

• 	     DSI_INT_LPRX_RD_RDY_FLAG);
  

• mtk_dsi_mask(dsi, DSI_INTEN, DSI_INT_CMD_DONE_FLAG,
  

• 	     DSI_INT_CMD_DONE_FLAG);
  

Why enable cmd_done interrupt? You just need interrupt of LPRX_RD_RDY.

• mtk_dsi_start(dsi);
  

• dev_info(dsi->dev, "Start polling DSI read ready!!!\n");
  

•  /* 2s timeout*/
  

• ret = mtk_dsi_wait_for_irq_timeout(dsi,
  

• 				   DSI_INT_LPRX_RD_RDY_FLAG,
  

• 				   2000);
  

• if (ret) {
  

• 	dev_info(dsi->dev, "Polling DSI read ready timeout!!!\n");
  

• 	mtk_dsi_enable(dsi);
  

• 	mtk_dsi_reset_engine(dsi);
  

• 	return ret;
  

• }
  

• dev_info(dsi->dev, "End polling DSI read ready!!!\n");
  

• mtk_dsi_mask(dsi, DSI_RACK, RACK, RACK);
  

• mtk_dsi_mask(dsi, DSI_INTSTA, LPRX_RD_RDY_INT_FLAG,
  

• 	     ~LPRX_RD_RDY_INT_FLAG);
  

• recv_data0 = readl(dsi->regs + DSI_RX_DATA0);
  

• recv_data1 = readl(dsi->regs + DSI_RX_DATA1);
  

• recv_data2 = readl(dsi->regs + DSI_RX_DATA2);
  

• recv_data3 = readl(dsi->regs + DSI_RX_DATA3);
  

• read_data0 = *((struct dsi_rx_data *)(&recv_data0));
  

• read_data1 = *((struct dsi_rx_data *)(&recv_data1));
  

• read_data2 = *((struct dsi_rx_data *)(&recv_data2));
  

• read_data3 = *((struct dsi_rx_data *)(&recv_data3));
  

You only use read_data0. read_data1, read_data2, and read_data3 are just for debug. Remove the debug only modification if they are not necessary. Or move them to another patch to make this patch clear.

• ret = readl(dsi->regs + DSI_CMDQ_SIZE);
  

• dev_info(dsi->dev, "DSI_CMDQ_SIZE : 0x%x\n", ret & CMDQ_SIZE);
  

• ret = readl(dsi->regs + DSI_CMDQ0);
  

• dev_info(dsi->dev, "DSI_CMDQ_DATA0 : 0x%x\n",
  

• 	 ret & 0xff);
  

• dev_info(dsi->dev, "DSI_CMDQ_DATA1 : 0x%x\n",
  

• 	 (ret >> 8) & 0xff);
  

• dev_info(dsi->dev, "DSI_CMDQ_DATA2 : 0x%x\n",
  

• 	 (ret >> 16) & 0xff);
  

• dev_info(dsi->dev, "DSI_CMDQ_DATA3 : 0x%x\n",
  

• 	 (ret >> 24) & 0xff);
  

• dev_info(dsi->dev, "DSI_RX_DATA0: 0x%x\n", recv_data0);
  

• dev_info(dsi->dev, "DSI_RX_DATA1: 0x%x\n", recv_data1);
  

• dev_info(dsi->dev, "DSI_RX_DATA2: 0x%x\n", recv_data2);
  

• dev_info(dsi->dev, "DSI_RX_DATA3: 0x%x\n", recv_data3);
  

• dev_info(dsi->dev, "read_data0: %x,%x,%x,%x\n",
  

• 	 read_data0.byte0, read_data0.byte1, read_data0.byte2,
  

• 	 read_data0.byte3);
  

• dev_info(dsi->dev, "read_data1: %x,%x,%x,%x\n",
  

• 	 read_data1.byte0, read_data1.byte1, read_data1.byte2,
  

• 	 read_data1.byte3);
  

• dev_info(dsi->dev, "read_data2: %x,%x,%x,%x\n",
  

• 	 read_data2.byte0, read_data2.byte1, read_data2.byte2,
  

• 	 read_data2.byte3);
  

• dev_info(dsi->dev, "read_data3: %x,%x,%x,%x\n",
  

• 	 read_data3.byte0, read_data3.byte1, read_data3.byte2,
  

• 	 read_data3.byte3);
  

• packet_type = read_data0.byte0;
  

• dev_info(dsi->dev, "DSI read packet_type is 0x%x\n",
  

• 	 packet_type);
  

• if (packet_type == 0x1a || packet_type == 0x1c) {
  

• 	void *read_tmp = (void *)&recv_data1;
  

• 	recv_data_cnt = read_data0.byte1 +
  

• 			read_data0.byte2 * 16;
  

• 	if (recv_data_cnt > 10)
  

• 		recv_data_cnt = 10;
  

Why do you drop data over size of 10? Isn't it an error?

• 	if (recv_data_cnt > buffer_size)
  

• 		recv_data_cnt = buffer_size;
  

Isn't it an error while you drop data over the buffer size?

• 	memcpy(buffer, read_tmp, recv_data_cnt);
  

• } else {
  

• 	/* short  packet */
  

• 	recv_data_cnt = 2;
  

• 	if (recv_data_cnt > buffer_size)
  

• 		recv_data_cnt = buffer_size;
  

Isn't it an error while you drop data over the buffer size?

• 	memcpy(buffer, &read_data0.byte1, 2);
  

If buffer_size is 1, why do you still copy 2 bytes?

• }
  

• } while (packet_type != 0x1c && packet_type != 0x21 &&

• packet_type != 0x22 && packet_type != 0x1a);
  

Could you give these magic number a descriptive name?

• dev_info(dsi->dev, "dsi get %d byte data from the panel address(0x%x)\n",

•  recv_data_cnt, *((u8 *)(msg->tx_buf)));
  

• return recv_data_cnt;

+}

+static ssize_t mtk_dsi_host_write_cmd(struct mtk_dsi *dsi,

• 		      const struct mipi_dsi_msg *msg)
  

+{

• u32 i;
• u32 goto_addr, mask_para, set_para, reg_val;
• void __iomem *cmdq_reg;
• struct dsi_cmd_t0 t0;
• struct dsi_cmd_t2 t2;
• const char *tx_buf = msg->tx_buf;
• struct dsi_tx_cmdq_regs *dsi_cmd_reg;
• dsi_cmd_reg = (struct dsi_tx_cmdq_regs *)(dsi->regs + DSI_CMDQ0);
• mtk_dsi_wait_for_idle(dsi);
• if (msg->tx_len > 2) {

• t2.config = 2;
  

• t2.type = msg->type;
  

• t2.wc16 = msg->tx_len;
  

• reg_val = (t2.wc16 << 16) | (t2.type << 8) | t2.config;
  

Why need 'struct dsi_cmd_t2 t2'? I think code is more simple when get rid of this structure. The code looks like:

reg_val = (wc16 << 16) | (type << 8) | config;

• writel(reg_val, &dsi_cmd_reg->data[0]);
  

• goto_addr = (u32)(&dsi_cmd_reg->data[1].byte0);
  

• mask_para = (0xff << ((goto_addr & 0x3) * 8));
  

• set_para = (tx_buf[0] << ((goto_addr & 0x3) * 8));
  

• cmdq_reg = (void __iomem *)(goto_addr & (~0x3));
  

• mtk_dsi_set_cmdq(cmdq_reg, mask_para, set_para);
  

I think you can merge this setting into 'for loop' and the loop look like this:

for (i = 0; i < msg->tx_len; i++) { goto_addr = (u32)(&dsi_cmd_reg->data[1].byte0) + i; mask_para = (0xff << ((goto_addr & 0x3) * 8)); set_para = (tx_buf[i] << ((goto_addr & 0x3) * 8)); cmdq_reg = (void __iomem *)(goto_addr & (~0x3)); mtk_dsi_set_cmdq(cmdq_reg, mask_para, set_para); }

• for (i = 1; i < msg->tx_len; i++) {
  

• 	goto_addr = (u32)(&dsi_cmd_reg->data[1].byte1) + i - 1;
  

• 	mask_para = (0xff << ((goto_addr & 0x3) * 8));
  

• 	set_para = (tx_buf[i] << ((goto_addr & 0x3) * 8));
  

• 	cmdq_reg = (void __iomem *)(goto_addr & (~0x3));
  

• 	mtk_dsi_set_cmdq(cmdq_reg, mask_para, set_para);
  

• }
  

• mtk_dsi_mask(dsi, DSI_CMDQ_SIZE, CMDQ_SIZE,
  

• 	     2 + (msg->tx_len - 1) / 4);
  

I think it is better to change formula of DSI_CMDQ_SIZE to '1 + (msg->tx_len + 3) / 4'. The '1' means the first cmd which describe the size of tx_buf, and the rest means the other cmds for body of tx_buf.

• } else {

• t0.config = 0;
  

• t0.data0 = tx_buf[0];
  

• if (msg->tx_len == 2) {
  

• 	t0.type = MIPI_DSI_DCS_SHORT_WRITE_PARAM;
  

• 	t0.data1 = tx_buf[1];
  

• } else {
  

• 	t0.type = MIPI_DSI_DCS_SHORT_WRITE;
  

• 	t0.data1 = 0;
  

• }
  

• reg_val = (t0.data1 << 24) | (t0.data0 << 16) | (t0.type << 8) |
  

• 	   t0.config;
  

• writel(reg_val, &dsi_cmd_reg->data[0]);
  

• mtk_dsi_mask(dsi, DSI_CMDQ_SIZE, CMDQ_SIZE, 1);
  

• }
• mtk_dsi_start(dsi);
• mtk_dsi_wait_for_idle(dsi);
• return 0;

+}

+static ssize_t mtk_dsi_host_transfer(struct mipi_dsi_host *host,

• 		     const struct mipi_dsi_msg *msg)
  

+{

• struct mtk_dsi *dsi = host_to_dsi(host);
• u8 type = msg->type;
• ssize_t ret = 0;
• if (MTK_DSI_HOST_IS_READ(type))

• ret = mtk_dsi_host_read_cmd(dsi, msg);
  

• else if (MTK_DSI_HOST_IS_WRITE(type))

• ret = mtk_dsi_host_write_cmd(dsi, msg);
  

• return ret;

+}

static const struct mipi_dsi_host_ops mtk_dsi_ops = { .attach = mtk_dsi_host_attach, .detach = mtk_dsi_host_detach,

• .transfer = mtk_dsi_host_transfer,

};

static int mtk_dsi_bind(struct device *dev, struct device *master, void *data)

Regards, CK