[U-Boot] [PATCH 23/23] ARM: tegra: Enable PCIe on Jetson TK1

[Thierry Reding]

On Wed, Aug 27, 2014 at 03:28:06PM +0200, Thierry Reding wrote:
> On Tue, Aug 26, 2014 at 03:54:50PM +0300, Tuomas Tynkkynen wrote:
> > On 18/08/14 10:16, Thierry Reding wrote:
> > [...]
> > > +static int as3722_gpio_direction_output(u8 gpio, u8 level)
> > > +{
> > > +	u8 value;
> > > +	int err;
> > > +
> > > +	if (gpio > 7)
> > > +		return -EINVAL;
> > > +
> > > +	if (level == 0)
> > > +		value = AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDL;
> > > +	else
> > > +		value = AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDH;
> > > +
> > > +	err = as3722_write(AS3722_GPIO_CONTROL(gpio), value);
> > > +	if (err) {
> > > +		error("as3722: failed to configure GPIO#%u as output: %d\n",
> > > +		      gpio, err);
> > > +		return err;
> > > +	}
> > > +
> > > +	err = as3722_gpio_set(gpio, level);
> > > +	if (err < 0) {
> > > +		error("as3722: failed to set GPIO#%u high: %d\n", gpio, err);
> > > +		return err;
> > > +	}
> > > +
> > > +	return 0;
> > > +}
> > 
> > This function doesn't work correctly if the GPIO was originally configured
> > as inverted and low, which GPIO#2 seems to be.
> > (as3722_read(AS3722_GPIO_CONTROL(2), &value) returns value == 0x87)...
> 
> That should be equivalent to what we're setting but is a somewhat weird
> default. I guess the fact that we're inverting it and then changing the
> value to high in separate transactions makes the output flip twice.
> 
> > > +
> > > +int tegra_pcie_board_init(void)
> > > +{
> > [...]
> > > +
> > > +	err = as3722_gpio_direction_output(2, 1);
> > > +	if (err < 0) {
> > > +		error("as3722: failed to set GPIO#2 high: %d\n", err);
> > > +		return err;
> > > +	}
> > [...]
> > 
> > On my board, this call results in UART corruption, like this:
> > 
> > tegra-pcie:   non-prefetchable memory: 0x13000000-0x20000000
> > tegra-pcie:   prefetchable memory: 0x20000000-0x40000000
> > ¥É½¥¹½bªÍ¥¹b2x1, 1x1 configuration
> > ¹Í5Rþtegra-pcie: probing port 1, using 1 lanes
> > 
> > Likely because GPIO#2 controls the +3.3V_LP0 rail, which powers the UART
> > level shifters. Commenting the function call out fixes the corruption and
> > PCI-E still works fine.
> 
> If I add a udelay(500) after the above I'm not able to reproduce the
> UART breakage anymore. But I guess making the AS3722 GPIO code smarter
> would be helpful. In the kernel this is done by checking the invert bit
> and then setting the value accordingly. I suppose the same could be done
> for the mode bits. I'll see if I can work up a patch.

How about this:
-------------- next part --------------
diff --git a/drivers/power/as3722.c b/drivers/power/as3722.c
index 59d1bf1b50b0..393dc8608d07 100644
--- a/drivers/power/as3722.c
+++ b/drivers/power/as3722.c
@@ -17,6 +17,7 @@
 #define AS3722_GPIO_CONTROL(n) (0x08 + (n))
 #define  AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDH (1 << 0)
 #define  AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDL (7 << 0)
+#define  AS3722_GPIO_CONTROL_MODE_MASK (7 << 0)
 #define  AS3722_GPIO_CONTROL_INVERT (1 << 7)
 #define AS3722_LDO_VOLTAGE(n) (0x10 + (n))
 #define AS3722_GPIO_SIGNAL_OUT 0x20
@@ -220,10 +221,21 @@ int as3722_gpio_direction_output(struct as3722 *pmic, unsigned int gpio,
 	if (gpio > 7)
 		return -EINVAL;
 
+	err = as3722_read(pmic, AS3722_GPIO_CONTROL(gpio), &value);
+	if (err < 0) {
+		error("failed to read GPIO#%u control register: %d", gpio, err);
+		return err;
+	}
+
+	if (value & AS3722_GPIO_CONTROL_INVERT)
+		level = !level;
+
+	value &= ~AS3722_GPIO_CONTROL_MODE_MASK;
+
 	if (level == 0)
-		value = AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDL;
+		value |= AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDL;
 	else
-		value = AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDH;
+		value |= AS3722_GPIO_CONTROL_MODE_OUTPUT_VDDH;
 
 	err = as3722_write(pmic, AS3722_GPIO_CONTROL(gpio), value);
 	if (err) {
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20140827/e771d6cb/attachment.pgp>