| Message ID | 20230905185309.131295-1-brgl@bgdev.pl |
|---|---|
| Headers | show |
| Series | gpio: convert users to gpio_device_find() and remove gpiochip_find() | expand |
On Tue, Sep 5, 2023 at 8:53 PM Bartosz Golaszewski <brgl@bgdev.pl> wrote: > The GPIO subsystem does not handle hot-unplug events very well. Yeah :/ it was never designed for this, and I've seen the discussions. The person who made the biggest effort to make this sort-of work was actually Johan Hovold so I added him to the mail so you can include him in upcoming iterations. I think he was working with GPIO on greybus at the time. Maybe he want to take a look! > Before we can even get to fixing the locking, we need to address a serious > abuse of the GPIO driver API - accessing struct gpio_chip by anyone who isn't > the driver owning this object. This structure is owned by the GPIO provider > and its lifetime is tied to that of that provider. It is destroyed when the > device is unregistered and this may happen at any moment. struct gpio_device > is the opaque, reference counted interface to struct gpio_chip (which is the > low-level implementation) and all access should pass through it. Thanks for looking into this. As I remember I have tried to bring down this abuse over the years and IIRC it used to be even worse, it came from the fact that all GPIO drivers used to be under some arch/* tree and often loosely using the kernel GPIO API but in addition providing a custom API... > The end-goal is to make all gpio_device manipulators check the existence of > gdev->chip and then lock it for the duration of any of the calls using SRCU. Excellent! > This series starts the process by replacing gpiochip_find() with > gpio_device_find(). This is in line with other device_find type interfaces and > returns a reference to the GPIO device that is guaranteed to remain valid > until it is put. I agree with the direction and I see no major problem with the patches other than some testing and cosmetics. The kernel sure as hell looks better *after* this than *before* so once you have rough confidence in the patches I think they should be merged and any issuse fixed up in-tree so we get wider audience and can continue the planned refactorings. So: Acked-by: Linus Walleij <linus.walleij@linaro.org> I'll try to provide some detailed reviews if something stands out. Yours, Linus Walleij
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> The GPIO subsystem does not handle hot-unplug events very well. We have recently patched the user-space part of it so that at least a rouge user cannot crash the kernel but in-kernel users are still affected by a lot of issues: from incorrect locking or lack thereof to using structures that are private to GPIO drivers. Since almost all GPIO controllers can be unbound, not to mention that we have USB devices registering GPIO expanders as well as I2C-on-USB HID devices on which I2C GPIO expanders can live, various media gadgets etc., we really need to make GPIO hotplug/unplug friendly. Before we can even get to fixing the locking, we need to address a serious abuse of the GPIO driver API - accessing struct gpio_chip by anyone who isn't the driver owning this object. This structure is owned by the GPIO provider and its lifetime is tied to that of that provider. It is destroyed when the device is unregistered and this may happen at any moment. struct gpio_device is the opaque, reference counted interface to struct gpio_chip (which is the low-level implementation) and all access should pass through it. The end-goal is to make all gpio_device manipulators check the existence of gdev->chip and then lock it for the duration of any of the calls using SRCU. Before we can get there, we need to first provide a set of functions that will replace any gpio_chip functions and convert all in-kernel users. This series starts the process by replacing gpiochip_find() with gpio_device_find(). This is in line with other device_find type interfaces and returns a reference to the GPIO device that is guaranteed to remain valid until it is put. Note that this does not make everything correct just yet. Especially the GPIOLIB internal users release the reference returned by the lookup function after getting the descriptor of interest but before requesting it. This will eventually be addressed. This is not a regression either. First we add a bunch of new APIs that are needed to start replacing calls to gpiochip_find. We then use them first in external users and then locally in GPIOLIB core. Finally we remove gpiochip_find(). Bartosz Golaszewski (21): gpiolib: make gpio_device_get() and gpio_device_put() public gpiolib: provide gpio_device_find() gpiolib: provide gpio_device_find_by_label() gpiolib: provide gpio_device_get_desc() gpiolib: add support for scope-based management to gpio_device gpiolib: provide gpiod_to_device() gpiolib: provide gpio_device_get_base() gpio: acpi: provide acpi_gpio_device_free_interrupts() gpiolib: reluctantly provide gpio_device_get_chip() gpiolib: replace find_chip_by_name() with gpio_device_find_by_label() platform: x86: android-tablets: don't access GPIOLIB private members hte: allow building modules with COMPILE_TEST enabled hte: tegra194: improve the GPIO-related comment hte: tegra194: don't access struct gpio_chip arm: omap1: ams-delta: stop using gpiochip_find() gpio: of: correct notifier return codes gpio: of: replace gpiochip_find_* with gpio_device_find_* gpio: acpi: replace gpiochip_find() with gpio_device_find() gpio: swnode: replace gpiochip_find() with gpio_device_find_by_label() gpio: sysfs: drop the mention of gpiochip_find() from sysfs code gpiolib: remove gpiochip_find() arch/arm/mach-omap1/board-ams-delta.c | 36 ++-- drivers/gpio/gpiolib-acpi.c | 37 +++- drivers/gpio/gpiolib-of.c | 48 ++--- drivers/gpio/gpiolib-swnode.c | 29 ++- drivers/gpio/gpiolib-sysfs.c | 2 +- drivers/gpio/gpiolib.c | 203 +++++++++++++----- drivers/gpio/gpiolib.h | 10 - drivers/hte/Kconfig | 4 +- drivers/hte/hte-tegra194.c | 49 +++-- .../platform/x86/x86-android-tablets/core.c | 38 ++-- include/linux/gpio/driver.h | 30 ++- 11 files changed, 316 insertions(+), 170 deletions(-)