Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Date:   Fri, 22 Apr 2022 15:03:52 +0200
Message-ID: <s5ha6cdtglj.wl-tiwai@suse.de>
From:   Takashi Iwai <tiwai@suse.de>
To:     Linus Walleij <linus.walleij@linaro.org>,
        Bartosz Golaszewski <brgl@bgdev.pl>
Cc:     Shreeya Patel <shreeya.patel@collabora.com>,
        linux-gpio@vger.kernel.org, linux-kernel@vger.kernel.org,
        regressions@lists.linux.dev
Subject: [REGRESSION] Missing IRQ via amd_gpio
User-Agent: Wanderlust/2.15.9 (Almost Unreal) SEMI/1.14.6 (Maruoka)
 FLIM/1.14.9 (=?UTF-8?B?R29qxY0=?=) APEL/10.8 Emacs/25.3
 (x86_64-suse-linux-gnu) MULE/6.0 (HANACHIRUSATO)
MIME-Version: 1.0 (generated by SEMI 1.14.6 - "Maruoka")
Content-Type: text/plain; charset=US-ASCII
Precedence: bulk

Hi,

we received a bug report for 5.17.3 kernel showing a new error:

 amd_gpio AMDI0030:00: Failed to translate GPIO pin 0x003D to IRQ, err -517

Not only an error message but in practice this leads to a missing IRQ
assignment; the IRQ 27 is no longer assigned to amd_gpio driver.

As the error number (EPROBE_DEFER) indicates, this seems to be the
side-effect of the recent fix, the upstream commit 5467801f1fcb
("gpio: Restrict usage of GPIO chip irq members before
initialization").  As far as I understand, the problem is in
acpi_gpiochip_request_interrupts() call that is called from
gpiochip_add_irqchip() itself.  Since it's called before the
initialized flag set, it always fails now.

Below is a temporary quick fix and it seems working.  But I'm not sure
whether I overlooked something obvious...


thanks,

Takashi

-- 8< --
From: Takashi Iwai <tiwai@suse.de>
Subject: [PATCH] gpio: Fix missing IRQ assginment for ACPI gpiochip

The recent fix for gpiolib caused an error like:
  amd_gpio AMDI0030:00: Failed to translate GPIO pin 0x003D to IRQ, err -517
It indicates -EPROBE_DEFER, and since the function
acpi_gpiochip_request_interrupts() doesn't handle the deferred probe,
this leads to the missing IRQs. 

The problem is that acpi_gpiochip_request_interrupts() itself is
called from gpiochip_add_irqchip() but before gc->irq.initialized flag
is set.  For fixing the regression, let's move the call of
acpi_gpiochip_request_interrupts() after the initialized flag setup.

Fixes: 5467801f1fcb ("gpio: Restrict usage of GPIO chip irq members before initialization")
BugLink: https://bugzilla.suse.com/show_bug.cgi?id=1198697
Signed-off-by: Takashi Iwai <tiwai@suse.de>
---
 drivers/gpio/gpiolib.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index 085348e08986..b7694171655c 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -1601,8 +1601,6 @@ static int gpiochip_add_irqchip(struct gpio_chip *gc,
 
 	gpiochip_set_irq_hooks(gc);
 
-	acpi_gpiochip_request_interrupts(gc);
-
 	/*
 	 * Using barrier() here to prevent compiler from reordering
 	 * gc->irq.initialized before initialization of above
@@ -1612,6 +1610,8 @@ static int gpiochip_add_irqchip(struct gpio_chip *gc,
 
 	gc->irq.initialized = true;
 
+	acpi_gpiochip_request_interrupts(gc);
+
 	return 0;
 }
 
-- 
2.31.1