1 files changed, 108 insertions, 29 deletions

diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
index 4dfc246c5f95..943940b44e83 100644
--- a/drivers/mmc/host/mtk-sd.c
+++ b/drivers/mmc/host/mtk-sd.c
@@ -8,6 +8,7 @@
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
+#include <linux/iopoll.h>
 #include <linux/ioport.h>
 #include <linux/irq.h>
 #include <linux/of_address.h>
@@ -258,6 +259,7 @@
 #define MSDC_PAD_TUNE_RD_SEL	  (0x1 << 13)   /* RW */
 #define MSDC_PAD_TUNE_CMD_SEL	  (0x1 << 21)   /* RW */
 
+#define PAD_DS_TUNE_DLY_SEL       (0x1 << 0)	/* RW */
 #define PAD_DS_TUNE_DLY1	  (0x1f << 2)   /* RW */
 #define PAD_DS_TUNE_DLY2	  (0x1f << 7)   /* RW */
 #define PAD_DS_TUNE_DLY3	  (0x1f << 12)  /* RW */
@@ -301,6 +303,11 @@
 #define PAD_CMD_RD_RXDLY_SEL    (0x1 << 11)     /* RW */
 #define PAD_CMD_TX_DLY          (0x1f << 12)    /* RW */
 
+/* EMMC50_PAD_DS_TUNE mask */
+#define PAD_DS_DLY_SEL		(0x1 << 16)	/* RW */
+#define PAD_DS_DLY1		(0x1f << 10)	/* RW */
+#define PAD_DS_DLY3		(0x1f << 0)	/* RW */
+
 #define REQ_CMD_EIO  (0x1 << 0)
 #define REQ_CMD_TMO  (0x1 << 1)
 #define REQ_DAT_ERR  (0x1 << 2)
@@ -448,11 +455,13 @@ struct msdc_host {
 	bool vqmmc_enabled;
 	u32 latch_ck;
 	u32 hs400_ds_delay;
+	u32 hs400_ds_dly3;
 	u32 hs200_cmd_int_delay; /* cmd internal delay for HS200/SDR104 */
 	u32 hs400_cmd_int_delay; /* cmd internal delay for HS400 */
 	bool hs400_cmd_resp_sel_rising;
 				 /* cmd response sample selection for HS400 */
 	bool hs400_mode;	/* current eMMC will run at hs400 mode */
+	bool hs400_tuning;	/* hs400 mode online tuning */
 	bool internal_cd;	/* Use internal card-detect logic */
 	bool cqhci;		/* support eMMC hw cmdq */
 	struct msdc_save_para save_para; /* used when gate HCLK */
@@ -961,7 +970,7 @@ static void msdc_set_mclk(struct msdc_host *host, unsigned char timing, u32 hz)
 }
 
 static inline u32 msdc_cmd_find_resp(struct msdc_host *host,
-		struct mmc_request *mrq, struct mmc_command *cmd)
+		struct mmc_command *cmd)
 {
 	u32 resp;
 
@@ -997,7 +1006,7 @@ static inline u32 msdc_cmd_prepare_raw_cmd(struct msdc_host *host,
 	 * stop << 14 | rw << 13 | dtype << 11 | rsptyp << 7 | brk << 6 | opcode
 	 */
 	u32 opcode = cmd->opcode;
-	u32 resp = msdc_cmd_find_resp(host, mrq, cmd);
+	u32 resp = msdc_cmd_find_resp(host, cmd);
 	u32 rawcmd = (opcode & 0x3f) | ((resp & 0x7) << 7);
 
 	host->cmd_rsp = resp;
@@ -1043,8 +1052,8 @@ static inline u32 msdc_cmd_prepare_raw_cmd(struct msdc_host *host,
 	return rawcmd;
 }
 
-static void msdc_start_data(struct msdc_host *host, struct mmc_request *mrq,
-			    struct mmc_command *cmd, struct mmc_data *data)
+static void msdc_start_data(struct msdc_host *host, struct mmc_command *cmd,
+		struct mmc_data *data)
 {
 	bool read;
 
@@ -1112,8 +1121,7 @@ static void msdc_recheck_sdio_irq(struct msdc_host *host)
 	}
 }
 
-static void msdc_track_cmd_data(struct msdc_host *host,
-				struct mmc_command *cmd, struct mmc_data *data)
+static void msdc_track_cmd_data(struct msdc_host *host, struct mmc_command *cmd)
 {
 	if (host->error)
 		dev_dbg(host->dev, "%s: cmd=%d arg=%08X; host->error=0x%08X\n",
@@ -1134,7 +1142,7 @@ static void msdc_request_done(struct msdc_host *host, struct mmc_request *mrq)
 	host->mrq = NULL;
 	spin_unlock_irqrestore(&host->lock, flags);
 
-	msdc_track_cmd_data(host, mrq->cmd, mrq->data);
+	msdc_track_cmd_data(host, mrq->cmd);
 	if (mrq->data)
 		msdc_unprepare_data(host, mrq->data);
 	if (host->error)
@@ -1190,7 +1198,8 @@ static bool msdc_cmd_done(struct msdc_host *host, int events,
 	if (!sbc_error && !(events & MSDC_INT_CMDRDY)) {
 		if (events & MSDC_INT_CMDTMO ||
 		    (cmd->opcode != MMC_SEND_TUNING_BLOCK &&
-		     cmd->opcode != MMC_SEND_TUNING_BLOCK_HS200))
+		     cmd->opcode != MMC_SEND_TUNING_BLOCK_HS200 &&
+		     !host->hs400_tuning))
 			/*
 			 * should not clear fifo/interrupt as the tune data
 			 * may have alreay come when cmd19/cmd21 gets response
@@ -1287,7 +1296,8 @@ static void msdc_cmd_next(struct msdc_host *host,
 	if ((cmd->error &&
 	    !(cmd->error == -EILSEQ &&
 	      (cmd->opcode == MMC_SEND_TUNING_BLOCK ||
-	       cmd->opcode == MMC_SEND_TUNING_BLOCK_HS200))) ||
+	       cmd->opcode == MMC_SEND_TUNING_BLOCK_HS200 ||
+	       host->hs400_tuning))) ||
 	    (mrq->sbc && mrq->sbc->error))
 		msdc_request_done(host, mrq);
 	else if (cmd == mrq->sbc)
@@ -1295,7 +1305,7 @@ static void msdc_cmd_next(struct msdc_host *host,
 	else if (!cmd->data)
 		msdc_request_done(host, mrq);
 	else
-		msdc_start_data(host, mrq, cmd, cmd->data);
+		msdc_start_data(host, cmd, cmd->data);
 }
 
 static void msdc_ops_request(struct mmc_host *mmc, struct mmc_request *mrq)
@@ -2251,6 +2261,67 @@ static int msdc_prepare_hs400_tuning(struct mmc_host *mmc, struct mmc_ios *ios)
 	return 0;
 }
 
+static int msdc_execute_hs400_tuning(struct mmc_host *mmc, struct mmc_card *card)
+{
+	struct msdc_host *host = mmc_priv(mmc);
+	struct msdc_delay_phase dly1_delay;
+	u32 val, result_dly1 = 0;
+	u8 *ext_csd;
+	int i, ret;
+
+	if (host->top_base) {
+		sdr_set_bits(host->top_base + EMMC50_PAD_DS_TUNE,
+			     PAD_DS_DLY_SEL);
+		if (host->hs400_ds_dly3)
+			sdr_set_field(host->top_base + EMMC50_PAD_DS_TUNE,
+				      PAD_DS_DLY3, host->hs400_ds_dly3);
+	} else {
+		sdr_set_bits(host->base + PAD_DS_TUNE, PAD_DS_TUNE_DLY_SEL);
+		if (host->hs400_ds_dly3)
+			sdr_set_field(host->base + PAD_DS_TUNE,
+				      PAD_DS_TUNE_DLY3, host->hs400_ds_dly3);
+	}
+
+	host->hs400_tuning = true;
+	for (i = 0; i < PAD_DELAY_MAX; i++) {
+		if (host->top_base)
+			sdr_set_field(host->top_base + EMMC50_PAD_DS_TUNE,
+				      PAD_DS_DLY1, i);
+		else
+			sdr_set_field(host->base + PAD_DS_TUNE,
+				      PAD_DS_TUNE_DLY1, i);
+		ret = mmc_get_ext_csd(card, &ext_csd);
+		if (!ret)
+			result_dly1 |= (1 << i);
+	}
+	host->hs400_tuning = false;
+
+	dly1_delay = get_best_delay(host, result_dly1);
+	if (dly1_delay.maxlen == 0) {
+		dev_err(host->dev, "Failed to get DLY1 delay!\n");
+		goto fail;
+	}
+	if (host->top_base)
+		sdr_set_field(host->top_base + EMMC50_PAD_DS_TUNE,
+			      PAD_DS_DLY1, dly1_delay.final_phase);
+	else
+		sdr_set_field(host->base + PAD_DS_TUNE,
+			      PAD_DS_TUNE_DLY1, dly1_delay.final_phase);
+
+	if (host->top_base)
+		val = readl(host->top_base + EMMC50_PAD_DS_TUNE);
+	else
+		val = readl(host->base + PAD_DS_TUNE);
+
+	dev_info(host->dev, "Fianl PAD_DS_TUNE: 0x%x\n", val);
+
+	return 0;
+
+fail:
+	dev_err(host->dev, "Failed to tuning DS pin delay!\n");
+	return -EIO;
+}
+
 static void msdc_hw_reset(struct mmc_host *mmc)
 {
 	struct msdc_host *host = mmc_priv(mmc);
@@ -2330,6 +2401,7 @@ static void msdc_cqe_enable(struct mmc_host *mmc)
 static void msdc_cqe_disable(struct mmc_host *mmc, bool recovery)
 {
 	struct msdc_host *host = mmc_priv(mmc);
+	unsigned int val = 0;
 
 	/* disable cmdq irq */
 	sdr_clr_bits(host->base + MSDC_INTEN, MSDC_INT_CMDQ);
@@ -2339,6 +2411,9 @@ static void msdc_cqe_disable(struct mmc_host *mmc, bool recovery)
 	if (recovery) {
 		sdr_set_field(host->base + MSDC_DMA_CTRL,
 			      MSDC_DMA_CTRL_STOP, 1);
+		if (WARN_ON(readl_poll_timeout(host->base + MSDC_DMA_CFG, val,
+			!(val & MSDC_DMA_CFG_STS), 1, 3000)))
+			return;
 		msdc_reset_hw(host);
 	}
 }
@@ -2377,6 +2452,7 @@ static const struct mmc_host_ops mt_msdc_ops = {
 	.card_busy = msdc_card_busy,
 	.execute_tuning = msdc_execute_tuning,
 	.prepare_hs400_tuning = msdc_prepare_hs400_tuning,
+	.execute_hs400_tuning = msdc_execute_hs400_tuning,
 	.hw_reset = msdc_hw_reset,
 };
 
@@ -2396,6 +2472,9 @@ static void msdc_of_property_parse(struct platform_device *pdev,
 	of_property_read_u32(pdev->dev.of_node, "hs400-ds-delay",
 			     &host->hs400_ds_delay);
 
+	of_property_read_u32(pdev->dev.of_node, "mediatek,hs400-ds-dly3",
+			     &host->hs400_ds_dly3);
+
 	of_property_read_u32(pdev->dev.of_node, "mediatek,hs200-cmd-int-delay",
 			     &host->hs200_cmd_int_delay);
 
@@ -2577,6 +2656,25 @@ static int msdc_drv_probe(struct platform_device *pdev)
 		host->dma_mask = DMA_BIT_MASK(32);
 	mmc_dev(mmc)->dma_mask = &host->dma_mask;
 
+	host->timeout_clks = 3 * 1048576;
+	host->dma.gpd = dma_alloc_coherent(&pdev->dev,
+				2 * sizeof(struct mt_gpdma_desc),
+				&host->dma.gpd_addr, GFP_KERNEL);
+	host->dma.bd = dma_alloc_coherent(&pdev->dev,
+				MAX_BD_NUM * sizeof(struct mt_bdma_desc),
+				&host->dma.bd_addr, GFP_KERNEL);
+	if (!host->dma.gpd || !host->dma.bd) {
+		ret = -ENOMEM;
+		goto release_mem;
+	}
+	msdc_init_gpd_bd(host, &host->dma);
+	INIT_DELAYED_WORK(&host->req_timeout, msdc_request_timeout);
+	spin_lock_init(&host->lock);
+
+	platform_set_drvdata(pdev, mmc);
+	msdc_ungate_clock(host);
+	msdc_init_hw(host);
+
 	if (mmc->caps2 & MMC_CAP2_CQE) {
 		host->cq_host = devm_kzalloc(mmc->parent,
 					     sizeof(*host->cq_host),
@@ -2597,25 +2695,6 @@ static int msdc_drv_probe(struct platform_device *pdev)
 		mmc->max_seg_size = 64 * 1024;
 	}
 
-	host->timeout_clks = 3 * 1048576;
-	host->dma.gpd = dma_alloc_coherent(&pdev->dev,
-				2 * sizeof(struct mt_gpdma_desc),
-				&host->dma.gpd_addr, GFP_KERNEL);
-	host->dma.bd = dma_alloc_coherent(&pdev->dev,
-				MAX_BD_NUM * sizeof(struct mt_bdma_desc),
-				&host->dma.bd_addr, GFP_KERNEL);
-	if (!host->dma.gpd || !host->dma.bd) {
-		ret = -ENOMEM;
-		goto release_mem;
-	}
-	msdc_init_gpd_bd(host, &host->dma);
-	INIT_DELAYED_WORK(&host->req_timeout, msdc_request_timeout);
-	spin_lock_init(&host->lock);
-
-	platform_set_drvdata(pdev, mmc);
-	msdc_ungate_clock(host);
-	msdc_init_hw(host);
-
 	ret = devm_request_irq(&pdev->dev, host->irq, msdc_irq,
 			       IRQF_TRIGGER_NONE, pdev->name, host);
 	if (ret)