1This is a place for planning the ongoing long-term work in the GPIO 2subsystem. 3 4 5GPIO descriptors 6 7Starting with commit 79a9becda894 the GPIO subsystem embarked on a journey 8to move away from the global GPIO numberspace and toward a descriptor-based 9approach. This means that GPIO consumers, drivers and machine descriptions 10ideally have no use or idea of the global GPIO numberspace that has/was 11used in the inception of the GPIO subsystem. 12 13The numberspace issue is the same as to why irq is moving away from irq 14numbers to IRQ descriptors. 15 16The underlying motivation for this is that the GPIO numberspace has become 17unmanageable: machine board files tend to become full of macros trying to 18establish the numberspace at compile-time, making it hard to add any numbers 19in the middle (such as if you missed a pin on a chip) without the numberspace 20breaking. 21 22Machine descriptions such as device tree or ACPI does not have a concept of the 23Linux GPIO number as those descriptions are external to the Linux kernel 24and treat GPIO lines as abstract entities. 25 26The runtime-assigned GPIO numberspace (what you get if you assign the GPIO 27base as -1 in struct gpio_chip) has also became unpredictable due to factors 28such as probe ordering and the introduction of -EPROBE_DEFER making probe 29ordering of independent GPIO chips essentially unpredictable, as their base 30number will be assigned on a first come first serve basis. 31 32The best way to get out of the problem is to make the global GPIO numbers 33unimportant by simply not using them. GPIO descriptors deal with this. 34 35Work items: 36 37- Convert all GPIO device drivers to only #include <linux/gpio/driver.h> 38 39- Convert all consumer drivers to only #include <linux/gpio/consumer.h> 40 41- Convert all machine descriptors in "boardfiles" to only 42 #include <linux/gpio/machine.h>, the other option being to convert it 43 to a machine description such as device tree, ACPI or fwnode that 44 implicitly does not use global GPIO numbers. 45 46- When this work is complete (will require some of the items in the 47 following ongoing work as well) we can delete the old global 48 numberspace accessors from <linux/gpio.h> and eventually delete 49 <linux/gpio.h> altogether. 50 51 52Get rid of <linux/of_gpio.h> 53 54This header and helpers appeared at one point when there was no proper 55driver infrastructure for doing simpler MMIO GPIO devices and there was 56no core support for parsing device tree GPIOs from the core library with 57the [devm_]gpiod_get() calls we have today that will implicitly go into 58the device tree back-end. It is legacy and should not be used in new code. 59 60Work items: 61 62- Get rid of struct of_mm_gpio_chip altogether: use the generic MMIO 63 GPIO for all current users (see below). Delete struct of_mm_gpio_chip, 64 to_of_mm_gpio_chip(), of_mm_gpiochip_add_data(), of_mm_gpiochip_add() 65 of_mm_gpiochip_remove() from the kernel. 66 67- Change all consumer drivers that #include <linux/of_gpio.h> to 68 #include <linux/gpio/consumer.h> and stop doing custom parsing of the 69 GPIO lines from the device tree. This can be tricky and often ivolves 70 changing boardfiles, etc. 71 72- Pull semantics for legacy device tree (OF) GPIO lookups into 73 gpiolib-of.c: in some cases subsystems are doing custom flags and 74 lookups for polarity inversion, open drain and what not. As we now 75 handle this with generic OF bindings, pull all legacy handling into 76 gpiolib so the library API becomes narrow and deep and handle all 77 legacy bindings internally. (See e.g. commits 6953c57ab172, 78 6a537d48461d etc) 79 80- Delete <linux/of_gpio.h> when all the above is complete and everything 81 uses <linux/gpio/consumer.h> or <linux/gpio/driver.h> instead. 82 83 84Get rid of <linux/gpio.h> 85 86This legacy header is a one stop shop for anything GPIO is closely tied 87to the global GPIO numberspace. The endgame of the above refactorings will 88be the removal of <linux/gpio.h> and from that point only the specialized 89headers under <linux/gpio/*.h> will be used. This requires all the above to 90be completed and is expected to take a long time. 91 92 93Collect drivers 94 95Collect GPIO drivers from arch/* and other places that should be placed 96in drivers/gpio/gpio-*. Augment platforms to create platform devices or 97similar and probe a proper driver in the gpiolib subsystem. 98 99In some cases it makes sense to create a GPIO chip from the local driver 100for a few GPIOs. Those should stay where they are. 101 102At the same time it makes sense to get rid of code duplication in existing or 103new coming drivers. For example, gpio-ml-ioh should be incorporated into 104gpio-pch. 105 106 107Generic MMIO GPIO 108 109The GPIO drivers can utilize the generic MMIO helper library in many 110cases, and the helper library should be as helpful as possible for MMIO 111drivers. (drivers/gpio/gpio-mmio.c) 112 113Work items: 114 115- Look over and identify any remaining easily converted drivers and 116 dry-code conversions to MMIO GPIO for maintainers to test 117 118- Expand the MMIO GPIO or write a new library for regmap-based I/O 119 helpers for GPIO drivers on regmap that simply use offsets 120 0..n in some register to drive GPIO lines 121 122- Expand the MMIO GPIO or write a new library for port-mapped I/O 123 helpers (x86 inb()/outb()) and convert port-mapped I/O drivers to use 124 this with dry-coding and sending to maintainers to test 125 126 127GPIOLIB irqchip 128 129The GPIOLIB irqchip is a helper irqchip for "simple cases" that should 130try to cover any generic kind of irqchip cascaded from a GPIO. 131 132- Look over and identify any remaining easily converted drivers and 133 dry-code conversions to gpiolib irqchip for maintainers to test 134 135 136Increase integration with pin control 137 138There are already ways to use pin control as back-end for GPIO and 139it may make sense to bring these subsystems closer. One reason for 140creating pin control as its own subsystem was that we could avoid any 141use of the global GPIO numbers. Once the above is complete, it may 142make sense to simply join the subsystems into one and make pin 143multiplexing, pin configuration, GPIO, etc selectable options in one 144and the same pin control and GPIO subsystem. 145 146 147Debugfs in place of sysfs 148 149The old sysfs code that enables simple uses of GPIOs from the 150command line is still popular despite the existance of the proper 151character device. The reason is that it is simple to use on 152root filesystems where you only have a minimal set of tools such 153as "cat", "echo" etc. 154 155The old sysfs still need to be strongly deprecated and removed 156as it relies on the global GPIO numberspace that assume a strict 157order of global GPIO numbers that do not change between boots 158and is independent of probe order. 159 160To solve this and provide an ABI that people can use for hacks 161and development, implement a debugfs interface to manipulate 162GPIO lines that can do everything that sysfs can do today: one 163directory per gpiochip and one file entry per line: 164 165/sys/kernel/debug/gpiochip/gpiochip0 166/sys/kernel/debug/gpiochip/gpiochip0/gpio0 167/sys/kernel/debug/gpiochip/gpiochip0/gpio1 168/sys/kernel/debug/gpiochip/gpiochip0/gpio2 169/sys/kernel/debug/gpiochip/gpiochip0/gpio3 170... 171/sys/kernel/debug/gpiochip/gpiochip1 172/sys/kernel/debug/gpiochip/gpiochip1/gpio0 173/sys/kernel/debug/gpiochip/gpiochip1/gpio1 174... 175 176The exact files and design of the debugfs interface can be 177discussed but the idea is to provide a low-level access point 178for debugging and hacking and to expose all lines without the 179need of any exporting. Also provide ample ammunition to shoot 180oneself in the foot, because this is debugfs after all. 181 182 183Moving over to immutable irq_chip structures 184 185Most of the gpio chips implementing interrupt support rely on gpiolib 186intercepting some of the irq_chip callbacks, preventing the structures 187from being made read-only and forcing duplication of structures that 188should otherwise be unique. 189 190The solution is to call into the gpiolib code when needed (resource 191management, enable/disable or unmask/mask callbacks), and to let the 192core code know about that by exposing a flag (IRQCHIP_IMMUTABLE) in 193the irq_chip structure. The irq_chip structure can then be made unique 194and const. 195 196A small number of drivers have been converted (pl061, tegra186, msm, 197amd, apple), and can be used as examples of how to proceed with this 198conversion. Note that drivers using the generic irqchip framework 199cannot be converted yet, but watch this space! 200