Skip to contentSkip to conversation

watchdog: add f71862fg support

PATCH Superseded Lutz Ballaschke 3 participants 5 messages Sep 6, 2010 – Sep 10, 2010 Watchdog device drivers

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 watchdog: add f71862fg support

From: Lutz Ballaschke <[email protected]>

Fintek Super-I/O f71862fg watchdog support added to f71808e_wdt driver. Furthermore the WDIOC_GETTIMELEFT ioctl is added to the driver.

---

drivers/watchdog/f71808e_wdt.c +40 -5
diff -Nurp linux-2.6.36-rc1-orig/drivers/watchdog/f71808e_wdt.c linux-2.6.36-rc1/drivers/watchdog/f71808e_wdt.c
@@ -42,6 +42,11 @@
 #define SIO_REG_DEVID		0x20	/* Device ID (2 bytes) */
 #define SIO_REG_DEVREV		0x22	/* Device revision */
 #define SIO_REG_MANID		0x23	/* Fintek ID (2 bytes) */
+#define SIO_REG_ROM_ADDR_SEL	0x27	/* ROM address select */
+#define SIO_REG_MFUNCT1		0x29	/* Multi function select 1 */
+#define SIO_REG_MFUNCT2		0x2a	/* Multi function select 2 */
+#define SIO_REG_MFUNCT3		0x2b	/* Multi function select 3 */
+#define SIO_REG_MFUNCT4		0x2c	/* Multi function select 4 */
 #define SIO_REG_ENABLE		0x30	/* Logical device enable */
 #define SIO_REG_ADDR		0x60	/* Logical device address (2 bytes) */
 
@@ -52,8 +57,6 @@
 #define SIO_F71882_ID		0x0541	/* Chipset ID */
 #define SIO_F71889_ID		0x0723	/* Chipset ID */
 
-#define	F71882FG_REG_START		0x01
-
 #define F71808FG_REG_WDO_CONF		0xf0
 #define F71808FG_REG_WDT_CONF		0xf5
 #define F71808FG_REG_WD_TIME		0xf6
@@ -252,6 +255,24 @@ exit_unlock:
 	return err;
 }
 
+static int watchdog_time_left(void)
+{
+	int ret = 0;
+
+	mutex_lock(&watchdog.lock);
+	ret = superio_enter(watchdog.sioaddr);
+	if (ret)
+		goto exit_unlock;
+
+	superio_select(watchdog.sioaddr, SIO_F71808FG_LD_WDT);
+	ret = (int) superio_inb(watchdog.sioaddr, F71808FG_REG_WD_TIME);
+	superio_exit(watchdog.sioaddr);
+
+exit_unlock:
+	mutex_unlock(&watchdog.lock);
+	return ret;
+}
+
 static int watchdog_keepalive(void)
 {
 	int err = 0;
@@ -299,13 +320,19 @@ static int watchdog_start(void)
 	switch (watchdog.type) {
 	case f71808fg:
 		/* Set pin 21 to GPIO23/WDTRST#, then to WDTRST# */
-		superio_clear_bit(watchdog.sioaddr, 0x2a, 3);
-		superio_clear_bit(watchdog.sioaddr, 0x2b, 3);
+		superio_clear_bit(watchdog.sioaddr, SIO_REG_MFUNCT2, 3);
+		superio_clear_bit(watchdog.sioaddr, SIO_REG_MFUNCT3, 3);
 		break;
 
 	case f71882fg:
 		/* Set pin 56 to WDTRST# */
-		superio_set_bit(watchdog.sioaddr, 0x29, 1);
+		superio_set_bit(watchdog.sioaddr, SIO_REG_MFUNCT1, 1);
+		break;
+
+	case f71862fg:
+		/* Set pin 63 to WDTRST# (SPI must be disabled first!) */
+		superio_clear_bit(watchdog.sioaddr, SIO_REG_ROM_ADDR_SEL, 6);
+		superio_set_bit(watchdog.sioaddr, SIO_REG_MFUNCT3, 4);
 		break;
 
 	default:
@@ -549,6 +576,12 @@ static long watchdog_ioctl(struct file *
 	case WDIOC_GETTIMEOUT:
 		return put_user(watchdog.timeout, uarg.i);
 
+	case WDIOC_GETTIMELEFT:
+		status = watchdog_time_left();
+		if (status < 0)
+			return status;
+		return put_user(status, uarg.i);
+
 	default:
 		return -ENOTTY;
 
@@ -709,6 +742,8 @@ static int __init f71808e_find(int sioad
 		watchdog.type = f71882fg;
 		break;
 	case SIO_F71862_ID:
+		watchdog.type = f71862fg;
+		break;
 	case SIO_F71889_ID:
 		/* These have a watchdog, though it isn't implemented (yet). */
 		err = -ENOSYS;
#1 lore raw
Quoted 2 lines · Lutz Ballaschke

Fintek Super-I/O f71862fg watchdog support added to f71808e_wdt driver. Furthermore the WDIOC_GETTIMELEFT ioctl is added to the driver.

You're patch makes three changes:
* Add f71862 support
* Add the WDIOC_GETTIMELEFT ioctl
* Change some config-register (magic) numbers into constants

Please split these things into separate patches.

Quoted 18 lines

diff -Nurp linux-2.6.36-rc1-orig/drivers/watchdog/f71808e_wdt.c linux-2.6.36-rc1/drivers/watchdog/f71808e_wdt.c

drivers/watchdog/f71808e_wdt.c +11 -0
@@ -252,6 +255,24 @@ exit_unlock:
 	return err;
 }
 
+static int watchdog_time_left(void)
+{
+	int ret = 0;
+
+	mutex_lock(&watchdog.lock);
+	ret = superio_enter(watchdog.sioaddr);
+	if (ret)
+		goto exit_unlock;
+
+	superio_select(watchdog.sioaddr, SIO_F71808FG_LD_WDT);
+	ret = (int) superio_inb(watchdog.sioaddr, F71808FG_REG_WD_TIME);

drivers/watchdog/f71808e_wdt.c exit_unlock: · line 278

Quoted 10 lines
drivers/watchdog/f71808e_wdt.c +7 -0
+	superio_exit(watchdog.sioaddr);
+
+exit_unlock:
+	mutex_unlock(&watchdog.lock);
+	return ret;
+}
+
 static int watchdog_keepalive(void)
 {
 	int err = 0;

This function doesn't perform any unit-conversion. The ioctl expects to receive seconds as the result, while this function can either return minutes or seconds depending on the value of WD_UNIT (see datasheet and/or uses of minutes_mode in the driver).

Further, why are you casting superio_inb()'s return value to (int) while that's already the return type of that function?

drivers/watchdog/f71808e_wdt.c static int watchdog_start(void) · line 338

Quoted 11 lines
drivers/watchdog/f71808e_wdt.c +6 -0
@@ -308,4 +329,10 @@ static int watchdog_start(void)
 		superio_set_bit(watchdog.sioaddr, SIO_REG_MFUNCT1, 1);
 		break;
 
+	case f71862fg:
+		/* Set pin 63 to WDTRST# (SPI must be disabled first!) */
+		superio_clear_bit(watchdog.sioaddr, SIO_REG_ROM_ADDR_SEL, 6);
+		superio_set_bit(watchdog.sioaddr, SIO_REG_MFUNCT3, 4);
+		break;
+
 	default:

Given that the F71862 has *two* pins assignable to the WDTRST# signal, please document that fact, and the reason for choosing this one in the code.

Perhaps it might even be desirable to insert a printk there. This to make the user aware of the fact that this driver might not work if the pin configuration of their hardware doesn't match the assumption you make about it in this code.

Signature · 3 lines
Met vriendelijke groet,
With kind regards,
Giel van Schijndel
#2 lore raw

Hi All,

Quoted 8 lines

Given that the F71862 has *two* pins assignable to the WDTRST# signal, please document that fact, and the reason for choosing this one in the code.

Perhaps it might even be desirable to insert a printk there. This to make the user aware of the fact that this driver might not work if the pin configuration of their hardware doesn't match the assumption you make about it in this code.

I propose to Document it in the code and add a module-parameter so that it is selectable.

Kind regards,
Wim.

#3 lore raw
Quoted 5 lines · Wim Van Sebroeck
Quoted 5 lines

Given that the F71862 has *two* pins assignable to the WDTRST# signal, please document that fact, and the reason for choosing this one in the code.

Quoted 7 lines
Quoted 4 lines

Perhaps it might even be desirable to insert a printk there. This to make the user aware of the fact that this driver might not work if the pin configuration of their hardware doesn't match the assumption you make about it in this code.

I propose to Document it in the code and add a module-parameter so that it is selectable.

I agree. I will post the patch again (missing parts will follow soon). I placed printk in watchdog_start(). Since this means printk output everytime /dev/watchdog is opened I'm not sure whether i like it there. Maybe it fits better in f71808e_find(). But this would require additional if / else and stuff.

Regards,
Lutz Ballaschke

#4 lore raw
Quoted 10 lines · Lutz Ballaschke
Quoted 9 lines · Wim Van Sebroeck
Quoted 8 lines

Given that the F71862 has *two* pins assignable to the WDTRST# signal, please document that fact, and the reason for choosing this one in the code.

Perhaps it might even be desirable to insert a printk there. This to make the user aware of the fact that this driver might not work if the pin configuration of their hardware doesn't match the assumption you make about it in this code.

Quoted 8 lines
Quoted 2 lines

I propose to Document it in the code and add a module-parameter so that it is selectable.

I agree. I will post the patch again (missing parts will follow soon). I placed printk in watchdog_start(). Since this means printk output everytime /dev/watchdog is opened I'm not sure whether i like it there. Maybe it fits better in f71808e_find(). But this would require additional if / else and stuff.

Can't you fit that printk in the switch(devid) block? That wouldn't require additional logic. In fact, if you make the selected pin configurable I wouldn't consider a printk explaining the current policy/config assumptions necessary because that way there aren't any policy-related assumptions made in the driver anymore.

Signature · 7 lines
Met vriendelijke groet,
With kind regards,
Giel van Schijndel
--
"It would seem that perfection is attained not when no more can be
 added, but when no more can be removed."
  -- Antoine de Saint Exupéry