debug-resume-hang.patch

Weeks of frantic effort to control Glamo, traced the issue to two outcomes:
nWAIT is forced down and the device is hard locked, or we survive immediate
Glamo resume and die again with nWAIT forced down when the framebuffer driver
tries to flash the soft cursor.

Signed-off-by: Andy Green <andy@openmoko.com>

1 files changed, 39 insertions, 4 deletions

diff --git a/drivers/mfd/glamo/glamo-mci.c b/drivers/mfd/glamo/glamo-mci.c
index 13175c4a240..b5d6ad10dd9 100644
--- a/drivers/mfd/glamo/glamo-mci.c
+++ b/drivers/mfd/glamo/glamo-mci.c
@@ -341,6 +341,12 @@ static void glamo_mci_irq(unsigned int irq, struct irq_desc *desc)
 
 	if (!host)
 		return;
+
+	if (host->suspending) { /* bad news, dangerous time */
+		dev_err(&host->pdev->dev, "****glamo_mci_irq before resumed\n");
+		return;
+	}
+
 	if (!host->mrq)
 		return;
 	cmd = host->mrq->cmd;
@@ -499,7 +505,7 @@ static int glamo_mci_send_command(struct glamo_mci_host *host,
 			 /* multiblock with stop */
 			fire |= GLAMO_FIRE_MMC_CC_MBWS;
 		else
-			 /* multiblock NO stop-- 'RESERVED'? */
+// 			 /* multiblock NO stop-- 'RESERVED'? */
 			fire |= GLAMO_FIRE_MMC_CC_MBWNS;
 		break;
 	case MMC_STOP_TRANSMISSION:
@@ -573,7 +579,14 @@ static void glamo_mci_send_request(struct mmc_host *mmc)
 	u16 * reg_resp = (u16 *)(host->base + GLAMO_REG_MMC_CMD_RSP1);
 	u16 status;
 	int n;
-	int timeout = 100000000;
+	int timeout = 10000000;
+	int insanity_timeout = 10000000;
+
+	if (host->suspending) {
+		dev_err(&host->pdev->dev, "IGNORING glamo_mci_send_request while "
+								 "suspended\n");
+		return;
+	}
 
 	host->ccnt++;
 	/*
@@ -620,7 +633,12 @@ static void glamo_mci_send_request(struct mmc_host *mmc)
 			     GLAMO_STAT1_MMC_RTOUT |
 			     GLAMO_STAT1_MMC_DTOUT |
 			     GLAMO_STAT1_MMC_BWERR |
-			     GLAMO_STAT1_MMC_BRERR))));
+			     GLAMO_STAT1_MMC_BRERR))) && (insanity_timeout--));
+
+	if (insanity_timeout < 0) {
+		cmd->error = -ETIMEDOUT;
+		dev_err(&host->pdev->dev, "****** insanity timeout\n");
+	}
 
 	if (status & (GLAMO_STAT1_MMC_RTOUT |
 		      GLAMO_STAT1_MMC_DTOUT))
@@ -716,13 +734,18 @@ static void glamo_mci_request(struct mmc_host *mmc, struct mmc_request *mrq)
 #if 0
 static void glamo_mci_reset(struct glamo_mci_host *host)
 {
+	if (host->suspending) {
+		dev_err(&host->pdev->dev, "IGNORING glamo_mci_reset while "
+								 "suspended\n");
+		return;
+	}
 	dev_err(&host->pdev->dev, "******* glamo_mci_reset\n");
 	/* reset MMC controller */
 	writew(GLAMO_CLOCK_MMC_RESET | GLAMO_CLOCK_MMC_DG_TCLK |
 		   GLAMO_CLOCK_MMC_EN_TCLK | GLAMO_CLOCK_MMC_DG_M9CLK |
 		   GLAMO_CLOCK_MMC_EN_M9CLK,
 		   glamo_mci_def_pdata.pglamo->base + GLAMO_REG_CLOCK_MMC);
-	msleep(1);
+	udelay(10);
 	/* and disable reset */
 	writew(GLAMO_CLOCK_MMC_DG_TCLK |
 		   GLAMO_CLOCK_MMC_EN_TCLK | GLAMO_CLOCK_MMC_DG_M9CLK |
@@ -738,6 +761,12 @@ static void glamo_mci_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 	int div;
 	int powering = 0;
 
+	if (host->suspending) {
+		dev_err(&host->pdev->dev, "IGNORING glamo_mci_set_ios while "
+								 "suspended\n");
+		return;
+	}
+
 	/* Set power */
 	switch(ios->power_mode) {
 	case MMC_POWER_ON:
@@ -1002,6 +1031,7 @@ static int glamo_mci_suspend(struct platform_device *dev, pm_message_t state)
 
 	ret = mmc_suspend_host(mmc, state);
 
+	host->suspending++;
 	/* so that when we resume, we use any modified max rate */
 	mmc->f_max = sd_max_clk;
 
@@ -1014,6 +1044,11 @@ int glamo_mci_resume(struct platform_device *dev)
 	struct glamo_mci_host 	*host = mmc_priv(mmc);
 	int ret;
 
+	sd_idleclk = 1;
+
+	glamo_engine_enable(host->pdata->pglamo, GLAMO_ENGINE_MMC);
+	glamo_mci_reset(host);
+
 	host->suspending--;
 
 	ret = mmc_resume_host(mmc);