WARNING: fs/proc/generic.c:173 __xlate_proc_name

WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Stefan Wahren]

Hi,
we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
platform. Recently we implemented an application which waits for GPIO
interrupts and we were able to trigger a warning by naming the owner of
the GPIO as "R1/S1":

WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
__xlate_proc_name+0x78/0x98 name 'R1/S1'
CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
6.1.49-00019-g9dbc76303a17 #147
Hardware name: Freescale i.MX6 Ultralite (Device Tree)
unwind_backtrace from show_stack+0x10/0x14
show_stack from dump_stack_lvl+0x24/0x2c
dump_stack_lvl from __warn+0x74/0xbc
__warn from warn_slowpath_fmt+0xc8/0x120
warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
__xlate_proc_name from __proc_create+0x3c/0x284
__proc_create from _proc_mkdir+0x2c/0x70
_proc_mkdir from proc_mkdir_data+0x10/0x18
proc_mkdir_data from register_handler_proc+0xc8/0x118
register_handler_proc from __setup_irq+0x554/0x664
__setup_irq from request_threaded_irq+0xac/0x13c
request_threaded_irq from edge_detector_setup+0xc0/0x1f8
edge_detector_setup from linereq_create+0x30c/0x384
linereq_create from vfs_ioctl+0x20/0x38
vfs_ioctl from sys_ioctl+0xbc/0x8b0
sys_ioctl from ret_fast_syscall+0x0/0x54
Exception stack(0xe0b61fa8 to 0xe0b61ff0)
1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
beeae880
1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
beeaeb60
1fe0: 00000036 beeae868 b6a88569 b6a01ae6
---[ end trace 0000000000000000 ]---

I'm not sure where this should be fixed.

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Bartosz Golaszewski]

On Sat, Feb 10, 2024 at 11:06 AM Stefan Wahren <wahrenst@gmx.net> wrote:
>
> Hi,
> we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
> platform. Recently we implemented an application which waits for GPIO
> interrupts and we were able to trigger a warning by naming the owner of
> the GPIO as "R1/S1":
>
> WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> __xlate_proc_name+0x78/0x98 name 'R1/S1'
> CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
> 6.1.49-00019-g9dbc76303a17 #147
> Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> unwind_backtrace from show_stack+0x10/0x14
> show_stack from dump_stack_lvl+0x24/0x2c
> dump_stack_lvl from __warn+0x74/0xbc
> __warn from warn_slowpath_fmt+0xc8/0x120
> warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
> __xlate_proc_name from __proc_create+0x3c/0x284
> __proc_create from _proc_mkdir+0x2c/0x70
> _proc_mkdir from proc_mkdir_data+0x10/0x18
> proc_mkdir_data from register_handler_proc+0xc8/0x118
> register_handler_proc from __setup_irq+0x554/0x664
> __setup_irq from request_threaded_irq+0xac/0x13c
> request_threaded_irq from edge_detector_setup+0xc0/0x1f8
> edge_detector_setup from linereq_create+0x30c/0x384
> linereq_create from vfs_ioctl+0x20/0x38
> vfs_ioctl from sys_ioctl+0xbc/0x8b0
> sys_ioctl from ret_fast_syscall+0x0/0x54
> Exception stack(0xe0b61fa8 to 0xe0b61ff0)
> 1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
> beeae880
> 1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
> beeaeb60
> 1fe0: 00000036 beeae868 b6a88569 b6a01ae6
> ---[ end trace 0000000000000000 ]---
>
> I'm not sure where this should be fixed.
>

Interesting. Seems as if the interrupt subsystem cannot handle '/' in
the interrupt consumer name. I'm wondering if we should filter out any
'/' in GPIO names when using the lines for interrupts? What do we
replace them with?

Bart

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Linus Walleij]

Hi,

looping in Geert becaus I think he wants to look into this!

On Sat, Feb 10, 2024 at 11:06 AM Stefan Wahren <wahrenst@gmx.net> wrote:

> we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
> platform. Recently we implemented an application which waits for GPIO
> interrupts and we were able to trigger a warning by naming the owner of
> the GPIO as "R1/S1":
>
> WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> __xlate_proc_name+0x78/0x98 name 'R1/S1'
> CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
> 6.1.49-00019-g9dbc76303a17 #147
> Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> unwind_backtrace from show_stack+0x10/0x14
> show_stack from dump_stack_lvl+0x24/0x2c
> dump_stack_lvl from __warn+0x74/0xbc
> __warn from warn_slowpath_fmt+0xc8/0x120
> warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
> __xlate_proc_name from __proc_create+0x3c/0x284
> __proc_create from _proc_mkdir+0x2c/0x70
> _proc_mkdir from proc_mkdir_data+0x10/0x18
> proc_mkdir_data from register_handler_proc+0xc8/0x118
> register_handler_proc from __setup_irq+0x554/0x664
> __setup_irq from request_threaded_irq+0xac/0x13c
> request_threaded_irq from edge_detector_setup+0xc0/0x1f8
> edge_detector_setup from linereq_create+0x30c/0x384
> linereq_create from vfs_ioctl+0x20/0x38
> vfs_ioctl from sys_ioctl+0xbc/0x8b0
> sys_ioctl from ret_fast_syscall+0x0/0x54
> Exception stack(0xe0b61fa8 to 0xe0b61ff0)
> 1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
> beeae880
> 1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
> beeaeb60
> 1fe0: 00000036 beeae868 b6a88569 b6a01ae6
> ---[ end trace 0000000000000000 ]---
>
> I'm not sure where this should be fixed.

Yours,
Linus Walleij

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Geert Uytterhoeven]

Hi Linus,

On Mon, Feb 12, 2024 at 2:35 PM Linus Walleij <linus.walleij@linaro.org> wrote:
> looping in Geert becaus I think he wants to look into this!

Oh no ;-)

> On Sat, Feb 10, 2024 at 11:06 AM Stefan Wahren <wahrenst@gmx.net> wrote:
> > we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
> > platform. Recently we implemented an application which waits for GPIO
> > interrupts and we were able to trigger a warning by naming the owner of
> > the GPIO as "R1/S1":
> >
> > WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> > __xlate_proc_name+0x78/0x98 name 'R1/S1'
> > CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
> > 6.1.49-00019-g9dbc76303a17 #147
> > Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> > unwind_backtrace from show_stack+0x10/0x14
> > show_stack from dump_stack_lvl+0x24/0x2c
> > dump_stack_lvl from __warn+0x74/0xbc
> > __warn from warn_slowpath_fmt+0xc8/0x120
> > warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
> > __xlate_proc_name from __proc_create+0x3c/0x284
> > __proc_create from _proc_mkdir+0x2c/0x70
> > _proc_mkdir from proc_mkdir_data+0x10/0x18
> > proc_mkdir_data from register_handler_proc+0xc8/0x118
> > register_handler_proc from __setup_irq+0x554/0x664
> > __setup_irq from request_threaded_irq+0xac/0x13c
> > request_threaded_irq from edge_detector_setup+0xc0/0x1f8
> > edge_detector_setup from linereq_create+0x30c/0x384
> > linereq_create from vfs_ioctl+0x20/0x38
> > vfs_ioctl from sys_ioctl+0xbc/0x8b0
> > sys_ioctl from ret_fast_syscall+0x0/0x54
> > Exception stack(0xe0b61fa8 to 0xe0b61ff0)
> > 1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
> > beeae880
> > 1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
> > beeaeb60
> > 1fe0: 00000036 beeae868 b6a88569 b6a01ae6
> > ---[ end trace 0000000000000000 ]---
> >
> > I'm not sure where this should be fixed.

Any names ending up in sysfs cannot contain a slash?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Bartosz Golaszewski]

On Mon, Feb 12, 2024 at 2:47 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>
> Hi Linus,
>
> On Mon, Feb 12, 2024 at 2:35 PM Linus Walleij <linus.walleij@linaro.org> wrote:
> > looping in Geert becaus I think he wants to look into this!
>
> Oh no ;-)
>
> > On Sat, Feb 10, 2024 at 11:06 AM Stefan Wahren <wahrenst@gmx.net> wrote:
> > > we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
> > > platform. Recently we implemented an application which waits for GPIO
> > > interrupts and we were able to trigger a warning by naming the owner of
> > > the GPIO as "R1/S1":
> > >
> > > WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> > > __xlate_proc_name+0x78/0x98 name 'R1/S1'
> > > CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
> > > 6.1.49-00019-g9dbc76303a17 #147
> > > Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> > > unwind_backtrace from show_stack+0x10/0x14
> > > show_stack from dump_stack_lvl+0x24/0x2c
> > > dump_stack_lvl from __warn+0x74/0xbc
> > > __warn from warn_slowpath_fmt+0xc8/0x120
> > > warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
> > > __xlate_proc_name from __proc_create+0x3c/0x284
> > > __proc_create from _proc_mkdir+0x2c/0x70
> > > _proc_mkdir from proc_mkdir_data+0x10/0x18
> > > proc_mkdir_data from register_handler_proc+0xc8/0x118
> > > register_handler_proc from __setup_irq+0x554/0x664
> > > __setup_irq from request_threaded_irq+0xac/0x13c
> > > request_threaded_irq from edge_detector_setup+0xc0/0x1f8
> > > edge_detector_setup from linereq_create+0x30c/0x384
> > > linereq_create from vfs_ioctl+0x20/0x38
> > > vfs_ioctl from sys_ioctl+0xbc/0x8b0
> > > sys_ioctl from ret_fast_syscall+0x0/0x54
> > > Exception stack(0xe0b61fa8 to 0xe0b61ff0)
> > > 1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
> > > beeae880
> > > 1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
> > > beeaeb60
> > > 1fe0: 00000036 beeae868 b6a88569 b6a01ae6
> > > ---[ end trace 0000000000000000 ]---
> > >
> > > I'm not sure where this should be fixed.
>
> Any names ending up in sysfs cannot contain a slash?

It's actually procfs. Every irq has an entry under /proc/irq/xyz/
containing a directory named after the name of the interrupt. In this
case the path of the directory would have been /proc/irq/xyz/R1/S1 but
of course this wouldn't work. Should we replace every `/` in GPIO line
names with `:` for interrupts?

Bart

>
> Gr{oetje,eeting}s,
>
>                         Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Kent Gibson]

On Mon, Feb 12, 2024 at 03:11:43PM +0100, Bartosz Golaszewski wrote:
> On Mon, Feb 12, 2024 at 2:47 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:

...

> > > >
> > > > I'm not sure where this should be fixed.
> >
> > Any names ending up in sysfs cannot contain a slash?
>
> It's actually procfs. Every irq has an entry under /proc/irq/xyz/
> containing a directory named after the name of the interrupt. In this
> case the path of the directory would have been /proc/irq/xyz/R1/S1 but
> of course this wouldn't work. Should we replace every `/` in GPIO line
> names with `:` for interrupts?
>

Agreed that the consumer label (not the GPIO name, btw), needs to be
sanitized before being used to create the procfs path, what I am uncertain
about is whether that should be performed in cdev, irq or even procfs.
If cdev then I would also like to see irq document that the devname cannot
contain slashes - it currently only states:

 *	@devname: An ascii name for the claiming device

Sanitizing with ":" works for me.

Cheers,
Kent.

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Stefan Wahren]

Hi,

[add lkml and linux-fsdevel]

Am 10.02.24 um 11:06 schrieb Stefan Wahren:
> Hi,
> we are using libgpiod-2.0.1 with Linux 6.1.49 on our Tarragon hardware
> platform. Recently we implemented an application which waits for GPIO
> interrupts and we were able to trigger a warning by naming the owner
> of the GPIO as "R1/S1":
>
> WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> __xlate_proc_name+0x78/0x98 name 'R1/S1'
> CPU: 0 PID: 429 Comm: cb_tarragon_dri Not tainted
> 6.1.49-00019-g9dbc76303a17 #147
> Hardware name: Freescale i.MX6 Ultralite (Device Tree)
> unwind_backtrace from show_stack+0x10/0x14
> show_stack from dump_stack_lvl+0x24/0x2c
> dump_stack_lvl from __warn+0x74/0xbc
> __warn from warn_slowpath_fmt+0xc8/0x120
> warn_slowpath_fmt from __xlate_proc_name+0x78/0x98
> __xlate_proc_name from __proc_create+0x3c/0x284
> __proc_create from _proc_mkdir+0x2c/0x70
> _proc_mkdir from proc_mkdir_data+0x10/0x18
> proc_mkdir_data from register_handler_proc+0xc8/0x118
> register_handler_proc from __setup_irq+0x554/0x664
> __setup_irq from request_threaded_irq+0xac/0x13c
> request_threaded_irq from edge_detector_setup+0xc0/0x1f8
> edge_detector_setup from linereq_create+0x30c/0x384
> linereq_create from vfs_ioctl+0x20/0x38
> vfs_ioctl from sys_ioctl+0xbc/0x8b0
> sys_ioctl from ret_fast_syscall+0x0/0x54
> Exception stack(0xe0b61fa8 to 0xe0b61ff0)
> 1fa0:                   01b019b8 01a9f428 0000000d c250b407 beeae888
> beeae880
> 1fc0: 01b019b8 01a9f428 01af7e40 00000036 beeaeb88 beeaeb80 beeaeb58
> beeaeb60
> 1fe0: 00000036 beeae868 b6a88569 b6a01ae6
> ---[ end trace 0000000000000000 ]---
>
> I'm not sure where this should be fixed.
>
since the discussion seems to stuck here [1], i reposted my original
mail to a wider audience.

Regards

[1] - https://lore.kernel.org/linux-gpio/20240213030409.GA35527@rigel/

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Alexey Dobriyan]

> WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> __xlate_proc_name+0x78/0x98 name 'R1/S1'

proc_mkdir() didn't find 'R1' directory.

In other words, you can't have slashes in irq names.

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Stefan Wahren]

Hi Alexey,

Am 25.02.24 um 11:37 schrieb Alexey Dobriyan:
>> WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
>> __xlate_proc_name+0x78/0x98 name 'R1/S1'
> proc_mkdir() didn't find 'R1' directory.
>
> In other words, you can't have slashes in irq names.
we already came to the point in the discussion before (link in my last
mail). The problem is no libgpiod user (userspace) is aware of this. So
the next question is where it should be fixed?

Sorry, i took my original message because posting the last message to
linux-fsdevel without context would be pointless.

Regards

Re: WARNING: fs/proc/generic.c:173 __xlate_proc_name

[Alexey Dobriyan]

On Sun, Feb 25, 2024 at 11:47:30AM +0100, Stefan Wahren wrote:
> Hi Alexey,
> 
> Am 25.02.24 um 11:37 schrieb Alexey Dobriyan:
> > > WARNING: CPU: 0 PID: 429 at fs/proc/generic.c:173
> > > __xlate_proc_name+0x78/0x98 name 'R1/S1'
> > proc_mkdir() didn't find 'R1' directory.
> > 
> > In other words, you can't have slashes in irq names.
> we already came to the point in the discussion before (link in my last
> mail). The problem is no libgpiod user (userspace) is aware of this.

Changing '/' to '-' or '.' should be OK. Whatever you do, don't change
to '!' because it will make these paths annoying to use in bash.

> So the next question is where it should be fixed?

libgpiod silently weren't getting directory inside /proc/irq/${irq}
and nobody noticed until now.

I think returning error may be too harsh but replacing characters is not.

> Sorry, i took my original message because posting the last message to
> linux-fsdevel without context would be pointless.