The new challenge is to remove VLAs from the kernel
(see https://lkml.org/lkml/2018/3/7/621) to eventually
turn on -Wvla.
Using a kmalloc array is the easy way to fix this but kmalloc is still
more expensive than stack allocation. Introduce a fast path with a
fixed size stack array to cover most chip with gpios below some fixed
amount. The slow path dynamically allocates an array to cover those
chips with a large number of gpios.
Reviewed-by: Geert Uytterhoeven <[email protected]>
Reviewed-by: Phil Reid <[email protected]>
Reviewed-and-tested-by: Lukas Wunner <[email protected]>
Signed-off-by: Lukas Wunner <[email protected]>
Signed-off-by: Laura Abbott <[email protected]>
---
v8: Adjust lower bound
---
drivers/gpio/Kconfig | 12 ++++++
drivers/gpio/gpiolib.c | 76 +++++++++++++++++++++++++++--------
drivers/gpio/gpiolib.h | 2 +-
include/linux/gpio/consumer.h | 10 +++--
4 files changed, 78 insertions(+), 22 deletions(-)
diff --git a/drivers/gpio/Kconfig b/drivers/gpio/Kconfig
index b960f6f35abd..71c0ab46f216 100644
--- a/drivers/gpio/Kconfig
+++ b/drivers/gpio/Kconfig
@@ -22,6 +22,18 @@ menuconfig GPIOLIB
if GPIOLIB
+config GPIOLIB_FASTPATH_LIMIT
+ int "Maximum number of GPIOs for fast path"
+ range 32 512
+ default 512
+ help
+ This adjusts the point at which certain APIs will switch from
+ using a stack allocated buffer to a dynamically allocated buffer.
+
+ You shouldn't need to change this unless you really need to
+ optimize either stack space or performance. Change this carefully
+ since setting an incorrect value could cause stack corruption.
+
config OF_GPIO
def_bool y
depends on OF
diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index d8ccb500872f..d29916cd3b74 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -61,6 +61,11 @@ static struct bus_type gpio_bus_type = {
.name = "gpio",
};
+/*
+ * Number of GPIOs to use for the fast path in set array
+ */
+#define FASTPATH_NGPIO CONFIG_GPIOLIB_FASTPATH_LIMIT
+
/* gpio_lock prevents conflicts during gpio_desc[] table updates.
* While any GPIO is requested, its gpio_chip is not removable;
* each GPIO's "requested" flag serves as a lock and refcount.
@@ -450,12 +455,11 @@ static long linehandle_ioctl(struct file *filep, unsigned int cmd,
vals[i] = !!ghd.values[i];
/* Reuse the array setting function */
- gpiod_set_array_value_complex(false,
+ return gpiod_set_array_value_complex(false,
true,
lh->numdescs,
lh->descs,
vals);
- return 0;
}
return -EINVAL;
}
@@ -1244,6 +1248,10 @@ int gpiochip_add_data_with_key(struct gpio_chip *chip, void *data,
goto err_free_descs;
}
+ if (chip->ngpio > FASTPATH_NGPIO)
+ chip_warn(chip, "line cnt %u is greater than fast path cnt %u\n",
+ chip->ngpio, FASTPATH_NGPIO);
+
gdev->label = kstrdup_const(chip->label ?: "unknown", GFP_KERNEL);
if (!gdev->label) {
status = -ENOMEM;
@@ -2719,16 +2727,28 @@ int gpiod_get_array_value_complex(bool raw, bool can_sleep,
while (i < array_size) {
struct gpio_chip *chip = desc_array[i]->gdev->chip;
- unsigned long mask[BITS_TO_LONGS(chip->ngpio)];
- unsigned long bits[BITS_TO_LONGS(chip->ngpio)];
+ unsigned long fastpath[2 * BITS_TO_LONGS(FASTPATH_NGPIO)];
+ unsigned long *mask, *bits;
int first, j, ret;
+ if (likely(chip->ngpio <= FASTPATH_NGPIO)) {
+ mask = fastpath;
+ } else {
+ mask = kmalloc_array(2 * BITS_TO_LONGS(chip->ngpio),
+ sizeof(*mask),
+ can_sleep ? GFP_KERNEL : GFP_ATOMIC);
+ if (!mask)
+ return -ENOMEM;
+ }
+
+ bits = mask + BITS_TO_LONGS(chip->ngpio);
+ memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
+
if (!can_sleep)
WARN_ON(chip->can_sleep);
/* collect all inputs belonging to the same chip */
first = i;
- memset(mask, 0, sizeof(mask));
do {
const struct gpio_desc *desc = desc_array[i];
int hwgpio = gpio_chip_hwgpio(desc);
@@ -2739,8 +2759,11 @@ int gpiod_get_array_value_complex(bool raw, bool can_sleep,
(desc_array[i]->gdev->chip == chip));
ret = gpio_chip_get_multiple(chip, mask, bits);
- if (ret)
+ if (ret) {
+ if (mask != fastpath)
+ kfree(mask);
return ret;
+ }
for (j = first; j < i; j++) {
const struct gpio_desc *desc = desc_array[j];
@@ -2752,6 +2775,9 @@ int gpiod_get_array_value_complex(bool raw, bool can_sleep,
value_array[j] = value;
trace_gpio_value(desc_to_gpio(desc), 1, value);
}
+
+ if (mask != fastpath)
+ kfree(mask);
}
return 0;
}
@@ -2935,7 +2961,7 @@ static void gpio_chip_set_multiple(struct gpio_chip *chip,
}
}
-void gpiod_set_array_value_complex(bool raw, bool can_sleep,
+int gpiod_set_array_value_complex(bool raw, bool can_sleep,
unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array)
@@ -2944,14 +2970,26 @@ void gpiod_set_array_value_complex(bool raw, bool can_sleep,
while (i < array_size) {
struct gpio_chip *chip = desc_array[i]->gdev->chip;
- unsigned long mask[BITS_TO_LONGS(chip->ngpio)];
- unsigned long bits[BITS_TO_LONGS(chip->ngpio)];
+ unsigned long fastpath[2 * BITS_TO_LONGS(FASTPATH_NGPIO)];
+ unsigned long *mask, *bits;
int count = 0;
+ if (likely(chip->ngpio <= FASTPATH_NGPIO)) {
+ mask = fastpath;
+ } else {
+ mask = kmalloc_array(2 * BITS_TO_LONGS(chip->ngpio),
+ sizeof(*mask),
+ can_sleep ? GFP_KERNEL : GFP_ATOMIC);
+ if (!mask)
+ return -ENOMEM;
+ }
+
+ bits = mask + BITS_TO_LONGS(chip->ngpio);
+ memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
+
if (!can_sleep)
WARN_ON(chip->can_sleep);
- memset(mask, 0, sizeof(mask));
do {
struct gpio_desc *desc = desc_array[i];
int hwgpio = gpio_chip_hwgpio(desc);
@@ -2982,7 +3020,11 @@ void gpiod_set_array_value_complex(bool raw, bool can_sleep,
/* push collected bits to outputs */
if (count != 0)
gpio_chip_set_multiple(chip, mask, bits);
+
+ if (mask != fastpath)
+ kfree(mask);
}
+ return 0;
}
/**
@@ -3057,13 +3099,13 @@ EXPORT_SYMBOL_GPL(gpiod_set_value);
* This function should be called from contexts where we cannot sleep, and will
* complain if the GPIO chip functions potentially sleep.
*/
-void gpiod_set_raw_array_value(unsigned int array_size,
+int gpiod_set_raw_array_value(unsigned int array_size,
struct gpio_desc **desc_array, int *value_array)
{
if (!desc_array)
- return;
- gpiod_set_array_value_complex(true, false, array_size, desc_array,
- value_array);
+ return -EINVAL;
+ return gpiod_set_array_value_complex(true, false, array_size,
+ desc_array, value_array);
}
EXPORT_SYMBOL_GPL(gpiod_set_raw_array_value);
@@ -3383,14 +3425,14 @@ EXPORT_SYMBOL_GPL(gpiod_set_value_cansleep);
*
* This function is to be called from contexts that can sleep.
*/
-void gpiod_set_raw_array_value_cansleep(unsigned int array_size,
+int gpiod_set_raw_array_value_cansleep(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array)
{
might_sleep_if(extra_checks);
if (!desc_array)
- return;
- gpiod_set_array_value_complex(true, true, array_size, desc_array,
+ return -EINVAL;
+ return gpiod_set_array_value_complex(true, true, array_size, desc_array,
value_array);
}
EXPORT_SYMBOL_GPL(gpiod_set_raw_array_value_cansleep);
diff --git a/drivers/gpio/gpiolib.h b/drivers/gpio/gpiolib.h
index ad456b6f9d8b..1a8e20363861 100644
--- a/drivers/gpio/gpiolib.h
+++ b/drivers/gpio/gpiolib.h
@@ -188,7 +188,7 @@ int gpiod_get_array_value_complex(bool raw, bool can_sleep,
unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
-void gpiod_set_array_value_complex(bool raw, bool can_sleep,
+int gpiod_set_array_value_complex(bool raw, bool can_sleep,
unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
diff --git a/include/linux/gpio/consumer.h b/include/linux/gpio/consumer.h
index dbd065963296..243112c7fa7d 100644
--- a/include/linux/gpio/consumer.h
+++ b/include/linux/gpio/consumer.h
@@ -116,7 +116,7 @@ int gpiod_get_raw_array_value(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
void gpiod_set_raw_value(struct gpio_desc *desc, int value);
-void gpiod_set_raw_array_value(unsigned int array_size,
+int gpiod_set_raw_array_value(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
@@ -134,7 +134,7 @@ int gpiod_get_raw_array_value_cansleep(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
void gpiod_set_raw_value_cansleep(struct gpio_desc *desc, int value);
-void gpiod_set_raw_array_value_cansleep(unsigned int array_size,
+int gpiod_set_raw_array_value_cansleep(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array);
@@ -369,12 +369,13 @@ static inline void gpiod_set_raw_value(struct gpio_desc *desc, int value)
/* GPIO can never have been requested */
WARN_ON(1);
}
-static inline void gpiod_set_raw_array_value(unsigned int array_size,
+static inline int gpiod_set_raw_array_value(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array)
{
/* GPIO can never have been requested */
WARN_ON(1);
+ return 0;
}
static inline int gpiod_get_value_cansleep(const struct gpio_desc *desc)
@@ -423,12 +424,13 @@ static inline void gpiod_set_raw_value_cansleep(struct gpio_desc *desc,
/* GPIO can never have been requested */
WARN_ON(1);
}
-static inline void gpiod_set_raw_array_value_cansleep(unsigned int array_size,
+static inline int gpiod_set_raw_array_value_cansleep(unsigned int array_size,
struct gpio_desc **desc_array,
int *value_array)
{
/* GPIO can never have been requested */
WARN_ON(1);
+ return 0;
}
static inline int gpiod_set_debounce(struct gpio_desc *desc, unsigned debounce)
--
2.17.0
On Fri, May 18, 2018 at 8:53 PM, Laura Abbott <[email protected]> wrote:
> The new challenge is to remove VLAs from the kernel
> (see https://lkml.org/lkml/2018/3/7/621) to eventually
> turn on -Wvla.
>
> Using a kmalloc array is the easy way to fix this but kmalloc is still
> more expensive than stack allocation. Introduce a fast path with a
> fixed size stack array to cover most chip with gpios below some fixed
> amount. The slow path dynamically allocates an array to cover those
> chips with a large number of gpios.
> + unsigned long fastpath[2 * BITS_TO_LONGS(FASTPATH_NGPIO)];
> + unsigned long *mask, *bits;
> int first, j, ret;
>
> + if (likely(chip->ngpio <= FASTPATH_NGPIO)) {
> + mask = fastpath;
> + } else {
> + mask = kmalloc_array(2 * BITS_TO_LONGS(chip->ngpio),
> + sizeof(*mask),
> + can_sleep ? GFP_KERNEL : GFP_ATOMIC);
> + if (!mask)
> + return -ENOMEM;
> + }
> +
> + bits = mask + BITS_TO_LONGS(chip->ngpio);
> + memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
Wouldn't be better
bitmap_zero(mask, chip->ngpio);
?
> + bits = mask + BITS_TO_LONGS(chip->ngpio);
> + memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
Ditto.
--
With Best Regards,
Andy Shevchenko
Hi Andy,
On Fri, May 18, 2018 at 10:07 PM, Andy Shevchenko
<[email protected]> wrote:
> On Fri, May 18, 2018 at 8:53 PM, Laura Abbott <[email protected]> wrote:
>> The new challenge is to remove VLAs from the kernel
>> (see https://lkml.org/lkml/2018/3/7/621) to eventually
>> turn on -Wvla.
>>
>> Using a kmalloc array is the easy way to fix this but kmalloc is still
>> more expensive than stack allocation. Introduce a fast path with a
>> fixed size stack array to cover most chip with gpios below some fixed
>> amount. The slow path dynamically allocates an array to cover those
>> chips with a large number of gpios.
>
>> + unsigned long fastpath[2 * BITS_TO_LONGS(FASTPATH_NGPIO)];
>> + unsigned long *mask, *bits;
>> int first, j, ret;
>>
>> + if (likely(chip->ngpio <= FASTPATH_NGPIO)) {
>> + mask = fastpath;
>> + } else {
>> + mask = kmalloc_array(2 * BITS_TO_LONGS(chip->ngpio),
>> + sizeof(*mask),
>> + can_sleep ? GFP_KERNEL : GFP_ATOMIC);
>> + if (!mask)
>> + return -ENOMEM;
>> + }
>> +
>> + bits = mask + BITS_TO_LONGS(chip->ngpio);
>> + memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
>
> Wouldn't be better
>
> bitmap_zero(mask, chip->ngpio);
>
> ?
chip->ngpio is never const, so bitmap_zero() would always fall back to
plain memset().
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Sat, May 19, 2018 at 12:05 AM, Geert Uytterhoeven
<[email protected]> wrote:
> On Fri, May 18, 2018 at 10:07 PM, Andy Shevchenko
> <[email protected]> wrote:
>> On Fri, May 18, 2018 at 8:53 PM, Laura Abbott <[email protected]> wrote:
>>> + bits = mask + BITS_TO_LONGS(chip->ngpio);
>>> + memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
>>
>> Wouldn't be better
>>
>> bitmap_zero(mask, chip->ngpio);
>>
>> ?
>
> chip->ngpio is never const, so bitmap_zero() would always fall back to
> plain memset().
Yes. But it also much more readable and understandable here.
--
With Best Regards,
Andy Shevchenko
On 05/18/2018 02:12 PM, Andy Shevchenko wrote:
> On Sat, May 19, 2018 at 12:05 AM, Geert Uytterhoeven
> <[email protected]> wrote:
>> On Fri, May 18, 2018 at 10:07 PM, Andy Shevchenko
>> <[email protected]> wrote:
>>> On Fri, May 18, 2018 at 8:53 PM, Laura Abbott <[email protected]> wrote:
>
>>>> + bits = mask + BITS_TO_LONGS(chip->ngpio);
>>>> + memset(mask, 0, BITS_TO_LONGS(chip->ngpio) * sizeof(*mask));
>>>
>>> Wouldn't be better
>>>
>>> bitmap_zero(mask, chip->ngpio);
>>>
>>> ?
>>
>> chip->ngpio is never const, so bitmap_zero() would always fall back to
>> plain memset().
>
> Yes. But it also much more readable and understandable here.
>
I wasn't aware of that API. I'll switch to that.