gpiolib: introduce for_each_gpio_in_chip macro
PATCH 1/1 gpiolib: introduce for_each_gpio_in_chip macro
gpiolib: introduce for_each_gpio_in_chip macro
There are a number of places in gpiolib where all the gpio's handled by a chip are walked thru using a for() loop. This introduces a for_each_* macro to clarify the code.
---
drivers/gpio/gpiolib.c +9 -4
@@ -56,6 +56,11 @@ struct gpio_desc { };static struct gpio_desc gpio_desc[ARCH_NR_GPIOS ]; +#define for_each_gpio_in_chip (__gpio, __chip) \ + for ((__gpio) = (__chip)->base; \ + (__gpio) < (__chip)->base + (__chip)->ngpio; \ + (__gpio)++) +static inline void desc_set_label (struct gpio_desc *d,const char *label) {#ifdef CONFIG_DEBUG_FS
@@ -694,14 +699,14 @@ int gpiochip_add(struct gpio_chip *chip) }/* these GPIO numbers must not be managed by another gpio_chip */ -for (id = base; id < base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) {if (gpio_desc[id].chip !=NULL ) { status = -EBUSY ;break ; } }if (status ==0 ) { -for (id = base; id < base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) { gpio_desc[id].chip = chip;/* REVISIT: most hardware initializes GPIOs as
@@ -744,14 +749,14 @@ int gpiochip_remove(struct gpio_chip *chip)spin_lock_irqsave (&gpio_lock, flags); -for (id = chip->base; id < chip->base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) {if (test_bit (FLAG_REQUESTED , &gpio_desc[id].flags)) { status = -EBUSY ;break ; } }if (status ==0 ) { -for (id = chip->base; id < chip->base + chip->ngpio; id++) +for_each_gpio_in_chip (id, chip) gpio_desc[id].chip =NULL ; }
David Brownell
NAKQuoted 5 lines · H Hartley Sweeten
gpiolib: introduce
for_each_gpio_in_chipmacroThere are a number of places in gpiolib where all the gpio's handled by a chip are walked thru using a
for()loop. This introduces a for_each_* macro to clarify the code.
I'd rather not. There are four such loops, and these are really simple iterators. Such a macro IMO just obfuscates, when it's just hiding such trivial index ops.
NAK.
Quoted 56 lines
Signed-off-by H Hartley Sweeten <[email protected]>---
drivers/gpio/gpiolib.c +9
-4@@ -56,6 +56,11 @@ struct gpio_desc { };static struct gpio_desc gpio_desc[ARCH_NR_GPIOS ]; +#define for_each_gpio_in_chip (__gpio, __chip) \ + for ((__gpio) = (__chip)->base; \ + (__gpio) < (__chip)->base + (__chip)->ngpio; \ + (__gpio)++) +static inline void desc_set_label (struct gpio_desc *d,const char *label) {#ifdef CONFIG_DEBUG_FS @@ -694,14 +699,14 @@ int gpiochip_add(struct gpio_chip *chip) }/* these GPIO numbers must not be managed by another gpio_chip */ -for (id = base; id < base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) {if (gpio_desc[id].chip !=NULL ) { status = -EBUSY ;break ; } }if (status ==0 ) { -for (id = base; id < base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) { gpio_desc[id].chip = chip;/* REVISIT: most hardware initializes GPIOs as @@ -744,14 +749,14 @@ int gpiochip_remove(struct gpio_chip *chip)spin_lock_irqsave (&gpio_lock, flags); -for (id = chip->base; id < chip->base + chip->ngpio; id++) { +for_each_gpio_in_chip (id, chip) {if (test_bit (FLAG_REQUESTED , &gpio_desc[id].flags)) { status = -EBUSY ;break ; } }if (status ==0 ) { -for (id = chip->base; id < chip->base + chip->ngpio; id++) +for_each_gpio_in_chip (id, chip) gpio_desc[id].chip =NULL ; }
Quoted 7 lines · David Brownell
Quoted 5 lines · H Hartley Sweeten
gpiolib: introduce
for_each_gpio_in_chipmacroThere are a number of places in gpiolib where all the gpio's handled by a chip are walked thru using a
for()loop. This introduces a for_each_* macro to clarify the code.
Quoted 5 lines
I'd rather not. There are four such loops, and these are really simple iterators. Such a macro IMO just obfuscates, when it's just hiding such trivial index ops.
NAK.
Ok.
For the record. The reason I sent this is I'm trying to work out an extension to gpiolib that adds gpio_port_* access to the API. Most of the gpiolib drivers already the necessary logic since the raw I/O is performed on the entire 'chip'. The API just needs the extensions added to request/free the port, set the direction and get/set the value.
Is this a worthwhile addition?
Regards,
Hartley
Quoted 7 lines · H Hartley Sweeten
For the record. The reason I sent this is I'm trying to work out an extension to gpiolib that adds gpio_port_* access to the API. Most of the gpiolib drivers already the necessary logic since the raw I/O is performed on the entire 'chip'. The API just needs the extensions added to request/free the port, set the direction and get/set the value.
Is this a worthwhile addition?
Plenty of people seem to think so. Personally I haven't seen a great use case except "'coz I can", but if you've got one I'd love to hear.
Have you seen http://lkml.org/lkml/2009/1/25/10 ? Donno what ended up happening to that patchset..
--Ben.
Quoted 8 lines · Ben Nizette
Quoted 6 lines · H Hartley Sweeten
For the record. The reason I sent this is I'm trying to work out an extension to gpiolib that adds gpio_port_* access to the API. Most of the gpiolib drivers already the necessary logic since the raw I/O is performed on the entire 'chip'. The API just needs the extensions added to request/free the port, set the direction and get/set the value.
Quoted 4 lines
Quoted 1 line
Is this a worthwhile addition?
Plenty of people seem to think so. Personally I haven't seen a great use case except "'coz I can", but if you've got one I'd love to hear.
Yes, you're right that there has been no major demand for it. There are (luckily?) only a moderate number of devices that are using gpio as their parallel bus interface. I've been supporting the batch-gpio patchset below out-of-the-tree because it has come in handy with a few e-paper display controllers and LCD 8080-IO that I've been developing with.
Quoted 4 lines
Have you seen http://lkml.org/lkml/2009/1/25/10 ? Donno what ended up happening to that patchset..
I didn't pursue it further and have maintained it out-of-tree. I felt that David had concerns about the API I implemented so it was unlikely to get merged and I didn't have the motivation to implement another. :-)
Thanks,
jaya
ps: I'm in Portland for the festival of linux conferences this week and would be happy to work on this/discuss alternate APIs if it is of interest.
Quoted 14 lines · Jaya Kumar
Quoted 8 lines · Ben Nizette
Quoted 6 lines · H Hartley Sweeten
For the record. The reason I sent this is I'm trying to work out an extension to gpiolib that adds gpio_port_* access to the API. Most of the gpiolib drivers already the necessary logic since the raw I/O is performed on the entire 'chip'. The API just needs the extensions added to request/free the port, set the direction and get/set the value.
Quoted 4 lines
Quoted 1 line
Is this a worthwhile addition?
Plenty of people seem to think so. Personally I haven't seen a great use case except "'coz I can", but if you've got one I'd love to hear.
Quoted 6 lines
Yes, you're right that there has been no major demand for it. There are (luckily?) only a moderate number of devices that are using gpio as their parallel bus interface. I've been supporting the batch-gpio patchset below out-of-the-tree because it has come in handy with a few e-paper display controllers and LCD 8080-IO that I've been developing with.
With the abundant number of GPIO's available on many of the ARM chips it seemed to me a "port" extension would be worthwhile. It would allow a side-band bus from the chip to access low-speed devices like a character LCD as an example.
Quoted 9 lines
Quoted 4 lines
Have you seen http://lkml.org/lkml/2009/1/25/10 ? Donno what ended up happening to that patchset..
I didn't pursue it further and have maintained it out-of-tree. I felt that David had concerns about the API I implemented so it was unlikely to get merged and I didn't have the motivation to implement another. :-)
Hmm.. That patchset is a lot different than what I was thinking of. Your patch allows a variable width to the number of gpio's in the "port". But it also still gets/sets the "port" by individual bit accesses to the gpio_chip. By doing this I don't see how you could get a performance increase.
The extension I was working on just allowed accessing the native width of the gpio chip. Most of the gpiolib drivers read/write the bits in a native width, the individual gpio pin is masked in/out. My patch just allows access to the raw data without masking anything.
Take the .get method in pca953x.c driver as an example.
static int pca953x_gpio_get_value(struct gpio_chip *gc, unsigned off)
{
struct pca953x_chip *chip;
uint16_t reg_val;
int ret;
chip = container_of(gc, struct pca953x_chip, gpio_chip);
ret = pca953x_read_reg(chip, PCA953X_INPUT, ®_val);
if (ret < 0) {
/* NOTE: diagnostic already emitted; that's all we should
* do unless gpio_*_value_cansleep() calls become different
* from their nonsleeping siblings (and report faults).
*/
return 0;
}
return (reg_val & (1u << off)) ? 1 : 0;
}The native width of the device is either 8 or 16 bits. To get a gpio value all of the bits are read then the desired gpio is masked out.
My thought was to just add the following methods to struct gpio_chip:
int (*port_direction_input)(struct gpio_chip *chip);
unsigned int (*port_get)(struct gpio_chip *chip);
int (*port_direction_output)(struct gpio_chip *chip, unsigned int value);
void (*port_set)(struct gpio_chip *chip, unsigned int value);
I basically stopped working on this after Ben's comment and getting no other feedback. If this appears useful I can look at it again.
Quoted 6 lines
Thanks,
jayaps: I'm in Portland for the festival of linux conferences this week and would be happy to work on this/discuss alternate APIs if it is of interest.
Regards,
Hartley
Quoted 7 lines · H Hartley Sweeten
On Friday, September 18, 2009 5:03 PM, Jaya Kumar wrote:
Hmm.. That patchset is a lot different than what I was thinking of. Your patch allows a variable width to the number of gpio's in the "port". But it also still gets/sets the "port" by individual bit accesses to the
gpio_chip. By doing this I don't see how you could get a performance increase.
It only does individual bit set/get if the specific arch does not offer a multi bit set/get. See __gpio_set_batch_generic which calls chip->set versus __gpio_set_batch which calls chip->set_batch so that if the underlying architecture supports it then we do multi-bit access. I provided one for pxa here:
http://marc.info/?l=linux-kernel&m=123287743104542&w=2 . So, the following scenarios should all be optimized for:
a) if the user calls gpio_s/get_batch with a width that fits the native register width and all those bits fit into the said register, then it results in a call directly to the multibit register set/get function/inline. For example, on pxa with above implementation, gpio_set_batch(1, mask=0xFFFFFFF, value=0xc001f00d), at compile time, that will result in a single GPSR/GPCR.
b) if the user calls gpio_s/get_batch with a width that more than one register, the implementation walks the bits and splits it across the registers as needed. This is done in __gpio_s/get_batch
c) if the user calls the gpio_s/get_batch with non-consecutive bits using the 32-bit mask, then the implementation walks the bits and splits it across the registers as needed.
So, it is fairly optimized. If I understood David's suggested design correctly, it is one where the user does not need to worry about masks or widths at all. Instead, you register with the library, a list of bits you want to make a bus out of, and then the library is responsible for taking that and optimizing that further as much as possible. I agree that that would be a more elegant API. What I was hoping for is that my implementation above could be a starting point for that by first providing optimized multi-bit access that could be used by the more elegant API.
Thanks,
jaya