Re: [PATCH] gpiolib: Remove big fat warning in gpiochip_get_direction()

[Christophe Leroy (CS GROUP)]

Le 02/06/2026 à 18:13, Bartosz Golaszewski a écrit :
> On Tue, Jun 2, 2026 at 1:31 PM Christophe Leroy (CS GROUP)
> <chleroy@xxxxxxxxxx> wrote:
> > Since kernel v6.15 the following big fat warning is encountered when
> > reading /sys/kernel/debug/gpio, leading to kernel latency while
> > emiting the warning, and panicing on systems configured to panic on
> > warnings.
> >
> >    ------------[ cut here ]------------
> >    WARNING: drivers/gpio/gpiolib.c:423 at gpiochip_get_direction+0x3c/0x48, CPU#0: cat/12531
> >    CPU: 0 UID: 0 PID: 12531 Comm: cat Tainted: G        W           7.0.10-gitc72c39~-01802-g28c351659258 #27 PREEMPT
> >    Tainted: [W]=WARN
> >    Hardware name: MIAE 8xx 0x500000 CMPC885
> >    NIP:  c043c2f8 LR: c043d740 CTR: 00000000
> >    REGS: ca89bc20 TRAP: 0700   Tainted: G        W            (7.0.10-gitc72c39~-01802-g28c351659258)
> >    MSR:  00029032 <EE,ME,IR,DR,RI>  CR: 24004884  XER: 00000302
> >
> >    GPR00: c043f3f0 ca89bce0 c3278000 c20b5f20 0000000d 00000002 00000000 c0a76208
> >    GPR08: 00000001 00000000 cccccccd c313d830 84004884 100d815e c0a76208 c0a761f8
> >    GPR16: c0a761f4 c0a76278 1048834c 10488350 c21c0b04 c21c0d93 c0a5fb74 c313d848
> >    GPR24: c20b5f20 c21c0d94 00000000 00000000 c21c0d94 00000000 c21c0c00 c21c0b04
> >    NIP [c043c2f8] gpiochip_get_direction+0x3c/0x48
> >    LR [c043d740] gpiod_get_direction+0xa0/0x170
> >    Call Trace:
> >    [ca89bce0] [c28157e8] 0xc28157e8 (unreliable)
> >    [ca89bd10] [c043f3f0] gpiolib_seq_show+0x370/0x524
> >    [ca89bd90] [c021dd1c] seq_read_iter+0x174/0x618
> >    [ca89bdd0] [c021e260] seq_read+0xa0/0xd0
> >    [ca89be40] [c031063c] full_proxy_read+0x80/0xc4
> >    [ca89be70] [c01df3e0] vfs_read+0xb4/0x35c
> >    [ca89bee0] [c01e0180] ksys_read+0x8c/0x15c
> >    [ca89bf10] [c000dc94] system_call_exception+0x88/0x154
> >    [ca89bf30] [c00110a8] ret_from_syscall+0x0/0x28
> >    ---- interrupt: c00 at 0xfc629e8
> >    NIP:  0fc629e8 LR: 0fc62a34 CTR: 00000000
> >    REGS: ca89bf40 TRAP: 0c00   Tainted: G        W            (7.0.10-gitc72c39~-01802-g28c351659258)
> >    MSR:  0000d032 <EE,PR,ME,IR,DR,RI>  CR: 28004884  XER: 00000302
> >
> >    GPR00: 00000003 7f8df6a0 77e37540 00000003 7f8df6e8 00001000 00000000 00000000
> >    GPR08: 00000000 7f8e3efc 00000000 7f8e06f0 7f8e3efc 100d815e 7fe70e10 100d0000
> >    GPR16: 100d0000 00000001 1048834c 10488350 22000882 77e3fe68 1000596c 77e40b28
> >    GPR24: 00000000 28004884 01000000 00001000 7f8df6e8 00000003 0fde36a0 00000000
> >    NIP [0fc629e8] 0xfc629e8
> >    LR [0fc62a34] 0xfc62a34
> >    ---- interrupt: c00
> >    Code: 9421fff0 7c0802a6 90010014 7d2903a6 4e800421 2c030001 40810008 3860ffcc 80010014 38210010 7c0803a6 4e800020 <0fe00000> 3860ffa1 4e800020 81230020
> >    ---[ end trace 0000000000000000 ]---
> >
> > This is due to a WARN_ON() added by commit ec2cceadfae7 ("gpiolib:
> > normalize the return value of gc->get() on behalf of buggy drivers")
> > when the gpiochip doesn't implement get_direction() function. But
> > according to the documentation in <linux/gpio/driver.h> implementing
> > get_direction() is only a recommendation, not a requirement. And
> > regarless, WARN_ON() has no added value here, dumping all CPU
> > registers doesn't give any useful information for that case.
> >
> > Lower it to a simple warn_on_once() message.
> >
> > Fixes: ec2cceadfae7 ("gpiolib: normalize the return value of gc->get() on behalf of buggy drivers")
> > Signed-off-by: Christophe Leroy (CS GROUP) <chleroy@xxxxxxxxxx>
> > ---
>
> Please use scripts/get_maintainers.pl.

That's what I did, what is the problem ?

$ ./scripts/get_maintainer.pl --nom --nor --nomultiline --norolestats 
--nogit --nogit-fallback 
tmp/0001-gpiolib-Remove-big-fat-warning-in-gpiochip_get_direc.patch
Dmitry Torokhov <dmitry.torokhov@xxxxxxxxx>, Linus Walleij 
<linusw@xxxxxxxxxx>, Bartosz Golaszewski 
<bartosz.golaszewski@xxxxxxxxxxxxxxxx>, linux-gpio@xxxxxxxxxxxxxxx, 
linux-kernel@xxxxxxxxxxxxxxx

> What driver are you using?

I have the following ones declared in the device tree:

			CPM1_PIO_A: gpio-controller@950 {
				#gpio-cells = <2>;
				compatible = "fsl,cpm1-pario-bank-a";
				reg = <0x950 0x10>;
				gpio-controller;
			};

			CPM1_PIO_B: gpio-controller@ab8 {
				#gpio-cells = <2>;
				compatible = "fsl,cpm1-pario-bank-b";
				reg = <0xab8 0x10>;
				gpio-controller;
			};

			CPM1_PIO_C: gpio-controller@960 {
				#gpio-cells = <2>;
				compatible = "fsl,cpm1-pario-bank-c";
				reg = <0x960 0x10>;
				interrupts = <1 2 6 9 10 11 14 15 23 24 26 31>;
				fsl,cpm1-gpio-irq-mask = <0x0fff>;
				interrupt-parent = <&CPM_PIC>;
				gpio-controller;
			};

			CPM1_PIO_D: gpio-controller@970 {
				#gpio-cells = <2>;
				compatible = "fsl,cpm1-pario-bank-d";
				reg = <0x970 0x10>;
				gpio-controller;
			};


				codec_fav: peb2466@3 {
					compatible = "infineon,peb2466";
					reg = <3>;
					clocks = <&clk_mclk>;
					clock-names = "mclk";
					spi-max-frequency = <1000000>;
					spi-cs-high;
					reset-gpios = <&rst 6 GPIO_ACTIVE_LOW>;
					#sound-dai-cells = <0>;
					sound-name-prefix = "C2";
					#gpio-cells = <2>;
					gpio-controller;
					firmware-name = "mcr-peb2466-coeffs.bin";
				};
				codec_radio: peb2466@7 {
					compatible = "infineon,peb2466";
					reg = <7>;
					clocks = <&clk_mclk>;
					clock-names = "mclk";
					spi-max-frequency = <1000000>;
					spi-cs-high;
					reset-gpios = <&rst 5 GPIO_ACTIVE_LOW>;
					#sound-dai-cells = <0>;
					sound-name-prefix = "C1";
					#gpio-cells = <2>;
					gpio-controller;
					firmware-name = "mcr-peb2466-coeffs.bin";
				};


Christophe