Skip to contentSkip to conversation

cs5535: add some additional cs5535-specific GPIO functionality

PATCH In review Andres Salomon 3 participants 6 messages Sep 26, 2010 – Nov 12, 2010 GPIO

lore
Share & tools
lore
#0 lore raw

PATCH 3/4 cs5535: add some additional cs5535-specific GPIO functionality

This adds (well, re-adds actually) handling for events/IRQs through cs5535 GPIOs. In the wild and wooly world of CS5535, setup_event() is for assigning an IRQ to a GPIO filter/event pair, and set_irq() sets up the pair to trigger IRQs.

These should really only be used in highly platform-specific drivers (such as OLPC's DCON driver). Sadly, because set_irq() uses MSRs, this causes the driver to become X86-specific.

---
 drivers/gpio/Kconfig       |    2 +-
 drivers/gpio/cs5535-gpio.c |   52 ++++++++++++++++++++++++++++++++++++++++++++
 include/linux/cs5535.h     |    2 +
 3 files changed, 55 insertions(+), 1 deletions(-)
drivers/gpio/Kconfig +1 -1
index 510aa20..80f6702 100644
@@ -271,7 +271,7 @@ comment "PCI GPIO expanders:"
 
 config GPIO_CS5535
 	tristate "AMD CS5535/CS5536 GPIO support"
-	depends on PCI && !CS5535_GPIO
+	depends on PCI && X86 && !CS5535_GPIO
 	help
 	  The AMD CS5535 and CS5536 southbridges support 28 GPIO pins that
 	  can be used for quite a number of things.  The CS5535/6 is found on
drivers/gpio/cs5535-gpio.c +52 -0

64 lines not on this page · raw message

include/linux/cs5535.h +2 -0
index d5a1d48..213cc50 100644
@@ -111,6 +111,8 @@ static inline int cs5535_has_vsa2(void)
 void cs5535_gpio_set(unsigned offset, unsigned int reg);
 void cs5535_gpio_clear(unsigned offset, unsigned int reg);
 int cs5535_gpio_isset(unsigned offset, unsigned int reg);
+int cs5535_gpio_set_irq(unsigned group, unsigned irq);
+void cs5535_gpio_setup_event(unsigned offset, int pair, int pme);
 
 /* MFGPTs */
 
Signature · 1 line
1.5.6.5
#1 lore raw
Quoted 10 lines · Andres Salomon

This adds (well, re-adds actually) handling for events/IRQs through cs5535 GPIOs. In the wild and wooly world of CS5535, setup_event() is for assigning an IRQ to a GPIO filter/event pair, and set_irq() sets up the pair to trigger IRQs.

These should really only be used in highly platform-specific drivers (such as OLPC's DCON driver). Sadly, because set_irq() uses MSRs, this causes the driver to become X86-specific.

Quoted 5 lines
---
 drivers/gpio/Kconfig       |    2 +-
 drivers/gpio/cs5535-gpio.c |   52 ++++++++++++++++++++++++++++++++++++++++++++
 include/linux/cs5535.h     |    2 +

I can't take this change without the gpio subsystem maintainer acking it. Can you get that approval from them?

I'm guessing that your 4/4 patch here depends on it, right? If so, I'll hold off on applying that patch until we can figure out how to handle this, ok?

thanks,

greg k-h

#2 lore raw
Quoted 18 lines · On Wed, 29 Sep 2010 18:05:01 -0700 Greg KH <[email protected]> wrote:
Quoted 10 lines · Andres Salomon

This adds (well, re-adds actually) handling for events/IRQs through cs5535 GPIOs. In the wild and wooly world of CS5535, setup_event() is for assigning an IRQ to a GPIO filter/event pair, and set_irq() sets up the pair to trigger IRQs.

These should really only be used in highly platform-specific drivers (such as OLPC's DCON driver). Sadly, because set_irq() uses MSRs, this causes the driver to become X86-specific.

Quoted 6 lines
---
 drivers/gpio/Kconfig       |    2 +-
 drivers/gpio/cs5535-gpio.c |   52

++++++++++++++++++++++++++++++++++++++++++++
include/linux/cs5535.h | 2 +

Quoted 2 lines

I can't take this change without the gpio subsystem maintainer acking it. Can you get that approval from them?

I believe the GPIO subsystem is unmaintained (please correct me if
I'm wrong). David Brownell had some good input on my original
cs5535-gpio driver patches that I sent, so I'll Cc him. I think that's as close to a maintainer as we're going to get. :)

If I recall correctly, the past few of my (and others') cs5535-gpio patches have gone in by way of the -mm tree.

Quoted 2 lines

I'm guessing that your 4/4 patch here depends on it, right? If so,

Correct.

Quoted 6 lines

I'll hold off on applying that patch until we can figure out how to handle this, ok?

thanks,

greg k-h

#3 lore raw
Quoted 21 lines · Andres Salomon
Quoted 18 lines · On Wed, 29 Sep 2010 18:05:01 -0700 Greg KH <[email protected]> wrote:
Quoted 16 lines · Andres Salomon

This adds (well, re-adds actually) handling for events/IRQs through cs5535 GPIOs. In the wild and wooly world of CS5535, setup_event() is for assigning an IRQ to a GPIO filter/event pair, and set_irq() sets up the pair to trigger IRQs.

These should really only be used in highly platform-specific drivers (such as OLPC's DCON driver). Sadly, because set_irq() uses MSRs, this causes the driver to become X86-specific.

---
 drivers/gpio/Kconfig       |    2 +-
 drivers/gpio/cs5535-gpio.c |   52

++++++++++++++++++++++++++++++++++++++++++++
include/linux/cs5535.h | 2 +

Quoted 8 lines
Quoted 2 lines

I can't take this change without the gpio subsystem maintainer acking it. Can you get that approval from them?

I believe the GPIO subsystem is unmaintained (please correct me if
I'm wrong). David Brownell had some good input on my original
cs5535-gpio driver patches that I sent, so I'll Cc him. I think that's as close to a maintainer as we're going to get. :)

Fair enough, get his ack and agreement that I can take this through the staging tree, and I'll be glad to do it.

Quoted 2 lines

If I recall correctly, the past few of my (and others') cs5535-gpio patches have gone in by way of the -mm tree.

Ah, that means there's no real maintainer :(

thanks,

greg k-h

#4 lore raw
In reply to Andres Salomon

Hi,

Does anyone have any comments on this? Greg, since there's no gpio maintainer, would you mind picking it up for 2.6.38 (along w/ the subsequent patch, which allows olpc_dcon to build)?

Quoted 114 lines · On Sat, 25 Sep 2010 19:07:12 -0700 Andres Salomon <[email protected]> wrote:

This adds (well, re-adds actually) handling for events/IRQs through cs5535 GPIOs. In the wild and wooly world of CS5535, setup_event() is for assigning an IRQ to a GPIO filter/event pair, and set_irq() sets up the pair to trigger IRQs.

These should really only be used in highly platform-specific drivers (such as OLPC's DCON driver). Sadly, because set_irq() uses MSRs, this causes the driver to become X86-specific.

---
 drivers/gpio/Kconfig       |    2 +-
 drivers/gpio/cs5535-gpio.c |   52

++++++++++++++++++++++++++++++++++++++++++++
include/linux/cs5535.h | 2 + 3 files changed, 55
insertions(+), 1 deletions(-)

drivers/gpio/Kconfig +1 -1
index 510aa20..80f6702 100644
@@ -271,7 +271,7 @@ comment "PCI GPIO expanders:"
 
 config GPIO_CS5535
 	tristate "AMD CS5535/CS5536 GPIO support"
-	depends on PCI && !CS5535_GPIO
+	depends on PCI && X86 && !CS5535_GPIO
 	help
 	  The AMD CS5535 and CS5536 southbridges support 28 GPIO

pins that can be used for quite a number of things. The CS5535/6 is found on diff --git a/drivers/gpio/cs5535-gpio.c b/drivers/gpio/cs5535-gpio.c index e23c068..7e5ab74 100644

drivers/gpio/cs5535-gpio.c +52 -0
@@ -15,6 +15,7 @@
 #include <linux/gpio.h>
 #include <linux/io.h>
 #include <linux/cs5535.h>
+#include <asm/msr.h>
 
 #define DRV_NAME "cs5535-gpio"
 #define GPIO_BAR 1
@@ -121,6 +122,57 @@ int cs5535_gpio_isset(unsigned offset, unsigned
 int reg) }
 EXPORT_SYMBOL_GPL(cs5535_gpio_isset);
 
+int cs5535_gpio_set_irq(unsigned group, unsigned irq)
+{
+	uint32_t lo, hi;
+
+	if (group > 7 || irq > 15)
+		return -EINVAL;
+
+	rdmsr(MSR_PIC_ZSEL_HIGH, lo, hi);
+
+	lo &= ~(0xF << (group * 4));
+	lo |= (irq & 0xF) << (group * 4);
+
+	wrmsr(MSR_PIC_ZSEL_HIGH, lo, hi);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cs5535_gpio_set_irq);
+
+void cs5535_gpio_setup_event(unsigned offset, int pair, int pme)
+{
+	struct cs5535_gpio_chip *chip = &cs5535_gpio_chip;
+	uint32_t shift = (offset % 8) * 4;
+	unsigned long flags;
+	uint32_t val;
+
+	if (offset >= 24)
+		offset = GPIO_MAP_W;
+	else if (offset >= 16)
+		offset = GPIO_MAP_Z;
+	else if (offset >= 8)
+		offset = GPIO_MAP_Y;
+	else
+		offset = GPIO_MAP_X;
+
+	spin_lock_irqsave(&chip->lock, flags);
+	val = inl(chip->base + offset);
+
+	/* Clear whatever was there before */
+	val &= ~(0xF << shift);
+
+	/* Set the new value */
+	val |= ((pair & 7) << shift);
+
+	/* Set the PME bit if this is a PME event */
+	if (pme)
+		val |= (1 << (shift + 3));
+
+	outl(val, chip->base + offset);
+	spin_unlock_irqrestore(&chip->lock, flags);
+}
+EXPORT_SYMBOL_GPL(cs5535_gpio_setup_event);
+
 /*
  * Generic gpio_chip API support.
  */
include/linux/cs5535.h +2 -0
index d5a1d48..213cc50 100644
@@ -111,6 +111,8 @@ static inline int cs5535_has_vsa2(void)
 void cs5535_gpio_set(unsigned offset, unsigned int reg);
 void cs5535_gpio_clear(unsigned offset, unsigned int reg);
 int cs5535_gpio_isset(unsigned offset, unsigned int reg);
+int cs5535_gpio_set_irq(unsigned group, unsigned irq);
+void cs5535_gpio_setup_event(unsigned offset, int pair, int pme);
 
 /* MFGPTs */
 
#5 lore raw
Quoted 5 lines · Andres Salomon

Hi,

Does anyone have any comments on this? Greg, since there's no gpio maintainer, would you mind picking it up for 2.6.38 (along w/ the subsequent patch, which allows olpc_dcon to build)?

There's no gpio maintainer? I thought Andrew was doing this for now, or was Grant?

Grant, I thought you were taking gpio stuff now, right?

thanks,

greg k-h