From mboxrd@z Thu Jan 1 00:00:00 1970 From: linus.walleij@linaro.org (Linus Walleij) Date: Sun, 15 May 2011 19:50:26 +0200 Subject: [PATCH 0/4] Pinmux subsystem In-Reply-To: <20110515133312.GM4071@lunn.ch> References: <1304363768-30338-1-git-send-email-linus.walleij@stericsson.com> <20110503172712.GE6538@lunn.ch> <20110515133312.GM4071@lunn.ch> Message-ID: To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org 2011/5/15 Andrew Lunn : > With a bit of scripting, came to the following results. Each line is > on pinmux function description what would be needed in the driver. The > number at the front says how many boards would use it. In total there > are 33 descriptions needed. Of these, 18 are used only once. This to > me suggests the board needs to be able to specify a pin description > when there is no reuse with other boards. As indicated elsewhere in this thread, a pinmux driver can pass in platform/package/board/system data just like any other platform driver in the kernel tree can. You will have to specify it somehow in a custom header file, just like with any other platform driver. Example from: arch/arm/mach-orion5x/ls-chl-setup.c: static struct gpio_led lschl_led_pins[] = { { .name = "alarm:red", .gpio = LSCHL_GPIO_LED_ALARM, .active_low = 1, }, { .name = "info:amber", .gpio = LSCHL_GPIO_LED_INFO, .active_low = 1, }, { .name = "func:blue:top", .gpio = LSCHL_GPIO_LED_FUNC, .active_low = 1, }, { .name = "power:blue:bottom", .gpio = LSCHL_GPIO_LED_PWR, }, }; static struct gpio_led_platform_data lschl_led_data = { .leds = lschl_led_pins, .num_leds = ARRAY_SIZE(lschl_led_pins), }; static struct platform_device lschl_leds = { .name = "leds-gpio", .id = -1, .dev = { .platform_data = &lschl_led_data, }, }; Still, the driver for these LEDs is in drivers/leds/leds-gpio.c and the above structure is described in include/linux/leds.h There is no difference between this and letting drivers/pinmux/pinmux-orion.c have a header file in include/linux/pinmux/orion.h and pass in necessary board data that way. Are we now in agreement that this will work just fine? Yours, Linus Walleij