This concerns gpio edge detection for GPIO IRQs used from
userspace for GPIO event listeners.
Trying to work out the right event if it's not sure that the
examined gpio actually moved is impossible.
Consider two gpios "gpioA" and "gpioB" that share an interrupt.
gpioA's irq should trigger on any edge, gpioB's on a falling edge.
If now the common irq fires and both gpio lines are high, there
are several possibilities that could have happend:
a) gpioA just had a low-to-high edge
b) gpioB just had a high-to-low-to-high spike
c) a combination of both a) and b)
While c) is unlikely (in most setups) a) and b) alone are bad
enough. Currently the code assumes case a) unconditionally and
doesn't report an event for gpioB. Note that even if there is no
irq sharing involved a spike for a gpio might not result in an
event if it's configured to trigger for a single edge only.
The only way to improve this is to drop support for interrupt
sharing. This way a spike results in an event for the right gpio
at least. Note that apart from dropping IRQF_SHARED this
effectively undoes commit df1e76f28f
("gpiolib: skip unwanted events, don't convert them to opposite edge").
This obviously breaks setups that rely on interrupt sharing,
but given that this cannot be reliable, this is probably an
acceptable trade-off.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
[Assuming there are no users of interrupt sharing yet]
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This is a GPIO driver so only include <linux/gpio/driver.h>.
Acked-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The SPI chipselects are assumed to be active low in the current
binding, so when we want to use GPIO descriptors and handle
the active low/high semantics in gpiolib, we need a special
parsing quirk to deal with this.
We check for the property "spi-cs-high" and if that is
NOT present we assume the CS line is active low.
If the line is tagged as active low in the device tree and
has no "spi-cs-high" property all is fine, the device
tree and the SPI bindings are in agreement.
If the line is tagged as active high in the device tree with
the second cell flag and has no "spi-cs-high" property we
enforce active low semantics (as this is the exception we can
just tag on the flag).
If the line is tagged as active low with the second cell flag
AND tagged with "spi-cs-high" the SPI active high property
takes precedence and we print a warning.
Cc: Mark Brown <broonie@kernel.org>
Cc: linux-spi@vger.kernel.org
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Since this driver does not use the gpiolib irqchip helpers it will have to
allocate the irq resources and irq_en/disable itself.
Use the new gpiochip_req/relres_irq helpers to request/release all the
resources.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Cc: Ray Jui <rjui@broadcom.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
When using the gpiolib irqchip helpers install irq_enable/disable
hooks for the irqchip to ensure that gpiolib knows when the irq
is enabled or disabled, allowing drivers to disable the irq and then
use it as an output pin, and later switch the direction to input and
re-enable the irq.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
GPIO drivers call gpiochip_(un)lock_as_irq whenever they want to use a gpio
as an interrupt. This is done when the irq is requested and it marks the
gpio as in use by an interrupt.
This is problematic for cases where a gpio pin is used as an interrupt
pin, then, after the irq is disabled, is used as a regular gpio pin.
Currently it is not possible to do this other than by first freeing
the interrupt so gpiochip_unlock_as_irq is called, since an attempt to
switch the gpio direction for output will fail since gpiolib believes
that the gpio is in use for an interrupt and it does not know that it
the irq is actually disabled.
There are currently two drivers that would like to be able to do this:
the tda998x_drv.c driver where a regular gpio pin needs to be temporarily
reconfigured as an interrupt pin during CEC calibration, and the cec-gpio
driver where you want to configure the gpio pin as an interrupt while
waiting for traffic over the CEC bus, or as a regular pin when receiving or
transmitting a CEC message.
The solution is to add a new flag that is set when the irq is enabled,
and have gpiod_direction_output check for that flag.
We also add functions that drivers that do not use GPIOLIB_IRQCHIP
can call when they enable/disable the irq.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Centralize setting the irq_request/release_resources callbacks
in one function since we'll be adding more callbacks to that.
Also fix the removal of the callback overrides: this should
only be done if we actually installed our own callback there.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
GPIO drivers that do not use GPIOLIB_IRQCHIP can hook these into
the irq_request_resource and irq_release_resource callbacks of the
irq_chip so they correctly 'get' the module and lock the gpio line
for IRQ use.
This will simplify driver code.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The problem is that if port == ARRAY_SIZE() and "gc == &epg->gc[port]"
then that should be treated as invalid.
Fixes: fd935fc421 ("gpio: ep93xx: Do not pingpong irq numbers")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Currently the while loop checks for the end of the array using
the size of egp->gc rather that the number of elements in the array,
so fix this. Also, perform the array size check first as stylistically
it is always good to bounds check on an array first before referencing
the array (in this case, we're just computing the address of an
element in an array so this is a moot point).
Fixes: fd935fc421 ("gpio: ep93xx: Do not pingpong irq numbers")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The gpiolib cannot deduce the fact that every line is output
by itself, implement a .get_direction() callback so we can
inspect this.
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
It's nice to use BIT() macros rather than open coding the same.
It's good practice as sometimes people use BIT(31) and forget
that the constant must be cast unsigned long.
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
It's nice to be able to read back the direction of the GPIO
line from the hardware so implement .get_direction() for
twl4030.
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
The MXS driver was calling back into the GPIO API from
its irqchip. This is not very elegant, as we are a driver,
let's just shortcut back into the gpio_chip .get() function
instead.
This is a tricky case since the .get() callback is not in
this file, instead assigned by bgpio_init(). Calling the
function direcly in the gpio_chip is however the lesser
evil.
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Janusz Uzycki <j.uzycki@elproma.com.pl>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Janusz Krzysztofik
Uwe Kleine-König
Linus Walleij
Hans Verkuil
Dan Carpenter
Colin Ian King