From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sasha Levin Subject: [PATCH AUTOSEL 4.14 008/123] gpio: gpio-omap: fix level interrupt idling Date: Wed, 27 Mar 2019 14:14:32 -0400 Message-ID: <20190327181628.15899-8-sashal@kernel.org> References: <20190327181628.15899-1-sashal@kernel.org> Mime-Version: 1.0 Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: <20190327181628.15899-1-sashal@kernel.org> Sender: linux-kernel-owner@vger.kernel.org To: linux-kernel@vger.kernel.org, stable@vger.kernel.org Cc: Russell King , Aaro Koskinen , Keerthy , Peter Ujfalusi , Tony Lindgren , Linus Walleij , Sasha Levin , linux-omap@vger.kernel.org, linux-gpio@vger.kernel.org List-Id: linux-gpio@vger.kernel.org From: Russell King [ Upstream commit d01849f7deba81f4959fd9e51bf20dbf46987d1c ] Tony notes that the GPIO module does not idle when level interrupts are in use, as the wakeup appears to get stuck. After extensive investigation, it appears that the wakeup will only be cleared if the interrupt status register is cleared while the interrupt is enabled. However, we are currently clearing it with the interrupt disabled for level-based interrupts. It is acknowledged that this observed behaviour conflicts with a statement in the TRM: CAUTION After servicing the interrupt, the status bit in the interrupt status register (GPIOi.GPIO_IRQSTATUS_0 or GPIOi.GPIO_IRQSTATUS_1) must be reset and the interrupt line released (by setting the corresponding bit of the interrupt status register to 1) before enabling an interrupt for the GPIO channel in the interrupt-enable register (GPIOi.GPIO_IRQSTATUS_SET_0 or GPIOi.GPIO_IRQSTATUS_SET_1) to prevent the occurrence of unexpected interrupts when enabling an interrupt for the GPIO channel. However, this does not appear to be a practical problem. Further, as reported by Grygorii Strashko , the TI Android kernel tree has an earlier similar patch as "GPIO: OMAP: Fix the sequence to clear the IRQ status" saying: if the status is cleared after disabling the IRQ then sWAKEUP will not be cleared and gates the module transition When we unmask the level interrupt after the interrupt has been handled, enable the interrupt and only then clear the interrupt. If the interrupt is still pending, the hardware will re-assert the interrupt status. Should the caution note in the TRM prove to be a problem, we could use a clear-enable-clear sequence instead. Cc: Aaro Koskinen Cc: Keerthy Cc: Peter Ujfalusi Signed-off-by: Russell King [tony@atomide.com: updated comments based on an earlier TI patch] Signed-off-by: Tony Lindgren Acked-by: Grygorii Strashko Signed-off-by: Linus Walleij Signed-off-by: Sasha Levin --- drivers/gpio/gpio-omap.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/drivers/gpio/gpio-omap.c b/drivers/gpio/gpio-omap.c index 3233b72b6828..148e81eea35a 100644 --- a/drivers/gpio/gpio-omap.c +++ b/drivers/gpio/gpio-omap.c @@ -841,14 +841,16 @@ static void omap_gpio_unmask_irq(struct irq_data *d) if (trigger) omap_set_gpio_triggering(bank, offset, trigger); - /* For level-triggered GPIOs, the clearing must be done after - * the HW source is cleared, thus after the handler has run */ - if (bank->level_mask & BIT(offset)) { - omap_set_gpio_irqenable(bank, offset, 0); + omap_set_gpio_irqenable(bank, offset, 1); + + /* + * For level-triggered GPIOs, clearing must be done after the source + * is cleared, thus after the handler has run. OMAP4 needs this done + * after enabing the interrupt to clear the wakeup status. + */ + if (bank->level_mask & BIT(offset)) omap_clear_gpio_irqstatus(bank, offset); - } - omap_set_gpio_irqenable(bank, offset, 1); raw_spin_unlock_irqrestore(&bank->lock, flags); } -- 2.19.1