mmc: omap_hsmmc: Fix check for mmc dt node

Proper check should be if the hsmmc driver got its platform_data via devicetree, and not if the devicetree is populated for the platform.

Signed-off-by Jassi Brar <[email protected]>

---

The patch is more than a nit-pick because upstream doesn't yet have dt nodes for mmc on OMAP while the driver wrongly assumes it does. Which isn't a problem until you use some low-end sdhc card (Transcend 8GB Class-4 in my experience) which fails to re-initialize, OCR[30] set, if the first attempt fails due to delay at certain times during the probe (for ex, simply enable CONFIG_REGULATOR_DUMMY). Apparently such cards need proper 0v->1.8v power cycle to recover (another Class-10 card does recover after ~10 attempts by the host).

While this simple patch gets us through the problem, it also tells that adding mmc dt nodes without PBIAS setup callbacks in place (without which we can't switch off the regulator or do 1.8v -> 3.3v switch) is probably not a safe option.

?drivers/mmc/host/omap_hsmmc.c | ? ?6 ++++--
?1 files changed, 4 insertions(+), 2 deletions(-)

drivers/mmc/host/omap_hsmmc.c +1 -0

index c628b95..b5ca4593 100644
@@ -138,6 +138,7 @@ struct omap_hsmmc_next {
 
 ?struct omap_hsmmc_host {
 ? ? ? ?struct ?device ? ? ? ? ?*dev;
+ ? ? ? struct ?of_device_id ? ?*of_dev;
 ? ? ? ?struct ?mmc_host ? ? ? ?*mmc;
 ? ? ? ?struct ?mmc_request ? ? *mrq;
 ? ? ? ?struct ?mmc_command ? ? *cmd;

@@ -249,7 +250,7 @@ static int omap_hsmmc_set_power(struct device *dev, int slot, int power_on,

? ? ? ? * the pbias cell programming support is still missing when ? ? ? ? * booting with Device tree
? ? ? ? */
- ? ? ? if (of_have_populated_dt() && !vdd)
+ ? ? ? if (host->of_dev && !vdd)
? ? ? ? ? ? ? ?return 0;

? ? ? ?if (mmc_slot(host).before_set_reg)

drivers/mmc/host/omap_hsmmc.c +0 -0

@@ -1550,7 +1551,7 @@ static void omap_hsmmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)

? ? ? ? ? ? ? ? ? ? ? ? * can't be allowed when booting with device ? ? ? ? ? ? ? ? ? ? ? ? * tree.
? ? ? ? ? ? ? ? ? ? ? ? */
- ? ? ? ? ? ? ? ? ? ? ? (!of_have_populated_dt())) {
+ ? ? ? ? ? ? ? ? ? ? ? !host->of_dev) {
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?/*
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? * The mmc_select_voltage fn of the core does ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? * not seem to set the power_mode to

drivers/mmc/host/omap_hsmmc.c +1 -1

@@ -1837,6 +1838,7 @@ static int __devinit omap_hsmmc_probe(struct platform_device *pdev)
 ? ? ? ?host ? ? ? ? ? ?= mmc_priv(mmc);
 ? ? ? ?host->mmc ? ? ? = mmc;
 ? ? ? ?host->pdata ? ? = pdata;
+ ? ? ? host->of_dev ? ?= match;
 ? ? ? ?host->dev ? ? ? = &pdev->dev;
 ? ? ? ?host->use_dma ? = 1;
 ? ? ? ?host->dev->dma_mask = &pdata->dma_mask;
--

1.7.4.1

Quoted 10 lines

The patch is more than a nit-pick because upstream doesn't yet have dt nodes for mmc on OMAP while the driver wrongly assumes it does. Which isn't a problem until you use some low-end sdhc card (Transcend 8GB Class-4 in my experience) which fails to re-initialize, OCR[30] set, if the first attempt fails due to delay at certain times during the probe (for ex, simply enable CONFIG_REGULATOR_DUMMY). Apparently such cards need proper 0v->1.8v power cycle to recover (another Class-10 card does recover after ~10 attempts by the host).

Is this issue seen while booting with device tree or non device tree ?

Proper check should be if the hsmmc driver got its platform_data via devicetree, and not if the devicetree is populated for the platform.

Signed-off-by Jassi Brar<[email protected]>

---

The patch is more than a nit-pick because upstream doesn't yet have dt nodes for mmc on OMAP while the driver wrongly assumes it does. Which isn't a problem until you use some low-end sdhc card (Transcend 8GB Class-4 in my experience) which fails to re-initialize, OCR[30] set, if the first attempt fails due to delay at certain times during the probe (for ex, simply enable CONFIG_REGULATOR_DUMMY). Apparently such cards need proper 0v->1.8v power cycle to recover (another Class-10 card does recover after ~10 attempts by the host).

While this simple patch gets us through the problem, it also tells that adding mmc dt nodes without PBIAS setup callbacks in place (without which we can't switch off the regulator or do 1.8v -> 3.3v switch) is probably not a safe option.

  drivers/mmc/host/omap_hsmmc.c |    6 ++++--
  1 files changed, 4 insertions(+), 2 deletions(-)

drivers/mmc/host/omap_hsmmc.c +4 -2

index c628b95..b5ca4593 100644
@@ -138,6 +138,7 @@ struct omap_hsmmc_next {
 
  struct omap_hsmmc_host {
  	struct	device		*dev;
+	struct	of_device_id	*of_dev;
  	struct	mmc_host	*mmc;
  	struct	mmc_request	*mrq;
  	struct	mmc_command	*cmd;

@@ -249,7 +250,7 @@ static int omap_hsmmc_set_power(struct device *dev, int slot, int power_on,
  	 * the pbias cell programming support is still missing when
  	 * booting with Device tree
  	 */
-	if (of_have_populated_dt()&&  !vdd)
+	if (host->of_dev&&  !vdd)
  		return 0;
 
  	if (mmc_slot(host).before_set_reg)

@@ -1550,7 +1551,7 @@ static void omap_hsmmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
  			 * can't be allowed when booting with device
  			 * tree.
  			 */
-			(!of_have_populated_dt())) {
+			!host->of_dev) {
  				/*
  				 * The mmc_select_voltage fn of the core does
  				 * not seem to set the power_mode to

@@ -1837,6 +1838,7 @@ static int __devinit omap_hsmmc_probe(struct platform_device *pdev)
  	host		= mmc_priv(mmc);
  	host->mmc	= mmc;
  	host->pdata	= pdata;
+	host->of_dev	= match;
  	host->dev	=&pdev->dev;
  	host->use_dma	= 1;
  	host->dev->dma_mask =&pdata->dma_mask;

Quoted 4 lines · Jassi Brar

Proper check should be if the hsmmc driver got its platform_data via devicetree, and not if the devicetree is populated for the platform.

This one is already fixed here..
http://www.spinics.net/lists/linux-omap/msg67879.html