Thread (10 messages) 10 messages, 3 authors, 2012-12-11

Re: [PATCH 1/1] gpio: add Lynxpoint chipset gpio driver.

From: Grant Likely <hidden>
Date: 2012-12-11 16:52:22
Also in: lkml

On Tue, 11 Dec 2012 13:40:51 +0200, Mathias Nyman [off-list ref] wrote:
On 12/11/2012 01:07 AM, Grant Likely wrote:
quoted
On Fri,  7 Dec 2012 16:01:39 +0200, Mathias Nyman[off-list ref]  wrote:
quoted
Add gpio support for Intel Lynxpoint chipset.
Lynxpoint supports 94 gpio pins which can generate interrupts.
Driver will fail requests for pins that are marked as owned by ACPI, or
set in an alternate mode (non-gpio).

Signed-off-by: Mathias Nyman<redacted>
---
+static void lp_gpio_set(struct gpio_chip *chip, unsigned gpio, int value)
+{
+	struct lp_gpio *lg = container_of(chip, struct lp_gpio, chip);
+	unsigned long reg = gpio_reg(chip, gpio, LP_CONFIG1);
+	unsigned long flags;
+
+	spin_lock_irqsave(&lg->lock, flags);
+
+	if (value)
+		outl(inl(reg) | OUT_LVL_BIT, reg);
+	else
+		outl(inl(reg)&  ~OUT_LVL_BIT, reg);
+
+	spin_unlock_irqrestore(&lg->lock, flags);
+}
A *lot* of drivers implement their own GPIO ops like this, and they all
end up looking the same. Please take a look at
drivers/gpio/gpio-generic.c and see if you can use the stock operations
provided there.

g.
Look quite similar but turns out the only potential generic ones which 
could be used are the bgpio_set() and bgpio_get() funtions. Using them 
would require custom pin2mask, write_reg, read_reg, etc functions.
bgpio_set_direction
You would need new write_reg and read_reg, but they'd just become part
of the library. Ah, but I see now that each GPIO has a separate
register, and not a bitmask for multiple gpio lines. Fair enough.

g.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help