Re: [PATCH 00/11] drm: Restore helper usability

Re: [PATCH 00/11] drm: Restore helper usability

[Geert Uytterhoeven]

Hi Arnd,

CC kbuild

On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
> On Mon, Apr 22, 2024, at 15:28, Jani Nikula wrote:
> > On Mon, 22 Apr 2024, "Arnd Bergmann" <arnd@arndb.de> wrote:
> >> 2. Using "select" on user visible symbols that have dependencies
> >>    is a common source for bugs, and this is is a problem in
> >>    drivers/gpu/drm more than elsewhere in the kernel, as these
> >>    drivers traditionally select entire subsystems or drivers
> >>    (I2C, VIRTIO, INPUT, ACPI_WMI, BACKLIGHT_CLASS_DEVICE,
> >>    POWER_SUPPLY, SND_PCM, INTERCONNECT, ...). This regularly
> >>    leads to circular dependencies and we should fix all of them.
> >
> > What annoys me is that the fixes tend to fall in two categories:
> >
> > - Play catch with selecting the dependencies of the selected
> >   symbols. "depends on" handles this recursively, while select does
> >   not.
>
> I'm not sure where this misunderstanding comes from, as you
> seem to be repeating the same incorrect assumption about
> how select works that Maxime wrote in his changelog. To clarify,
> this works exactly as one would expect:
>
> config HELPER_A
>        tristate
>
> config HELPER_B
>        tristate
>        select HELPER_A
>
> config DRIVER
>        tristate "Turn on the driver and the helpers it uses"
>        select HELPER_B # this recursively selects HELPER_A
>
> Whereas this one is broken:
>
> config FEATURE_A
>        tristate "user visible if I2C is enabled"
>        depends on I2C
>
> config HELPER_B
>        tristate # hidden
>        select FEATURE_A
>
> config DRIVER
>        tristate "This driver is broken if I2C is disabled"
>        select HELPER_B

So the DRIVER section should gain a "depends on I2C" statement.

Yamada-san: would it be difficult to modify Kconfig to ignore symbols
like DRIVER that select other symbols with unmet dependencies?
Currently it already warns about that.

Handling this implicitly (instead of the current explict "depends
on") would have the disadvantage though: a user who is not aware of
the implicit dependency may wonder why DRIVER is invisible in his
config interface.

>
> >   There is no end to this, it just goes on and on, as the
> >   dependencies of the selected symbols change over time. Often the
> >   selects require unintuitive if patterns that are about the
> >   implementation details of the symbol being selected.
>
> Agreed, that is the problem I frequently face with drivers/gpu/drm,
> and most of the time it can only be solved by rewriting the whole
> system to not select user-visible symbol at all.

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: [PATCH 00/11] drm: Restore helper usability

[Jani Nikula]

On Mon, 22 Apr 2024, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> Hi Arnd,
>
> CC kbuild
>
> On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
>> I'm not sure where this misunderstanding comes from, as you
>> seem to be repeating the same incorrect assumption about
>> how select works that Maxime wrote in his changelog. To clarify,
>> this works exactly as one would expect:
>>
>> config HELPER_A
>>        tristate
>>
>> config HELPER_B
>>        tristate
>>        select HELPER_A
>>
>> config DRIVER
>>        tristate "Turn on the driver and the helpers it uses"
>>        select HELPER_B # this recursively selects HELPER_A
>>
>> Whereas this one is broken:
>>
>> config FEATURE_A
>>        tristate "user visible if I2C is enabled"
>>        depends on I2C
>>
>> config HELPER_B
>>        tristate # hidden
>>        select FEATURE_A
>>
>> config DRIVER
>>        tristate "This driver is broken if I2C is disabled"
>>        select HELPER_B
>
> So the DRIVER section should gain a "depends on I2C" statement.

Why should DRIVER have to care that HELPER_B needs either FEATURE_A or
I2C? It should only have to care about HELPER_B. And if the dependencies
of FEATURE_A or HELPER_B later change, that's their business, not
DRIVER's.


BR,
Jani.

>
> Yamada-san: would it be difficult to modify Kconfig to ignore symbols
> like DRIVER that select other symbols with unmet dependencies?
> Currently it already warns about that.
>
> Handling this implicitly (instead of the current explict "depends
> on") would have the disadvantage though: a user who is not aware of
> the implicit dependency may wonder why DRIVER is invisible in his
> config interface.

-- 
Jani Nikula, Intel

Re: [PATCH 00/11] drm: Restore helper usability

[Geert Uytterhoeven]

Hi Jani,

On Mon, Apr 22, 2024 at 7:15 PM Jani Nikula <jani.nikula@linux.intel.com> wrote:
> On Mon, 22 Apr 2024, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
> >> I'm not sure where this misunderstanding comes from, as you
> >> seem to be repeating the same incorrect assumption about
> >> how select works that Maxime wrote in his changelog. To clarify,
> >> this works exactly as one would expect:
> >>
> >> config HELPER_A
> >>        tristate
> >>
> >> config HELPER_B
> >>        tristate
> >>        select HELPER_A
> >>
> >> config DRIVER
> >>        tristate "Turn on the driver and the helpers it uses"
> >>        select HELPER_B # this recursively selects HELPER_A
> >>
> >> Whereas this one is broken:
> >>
> >> config FEATURE_A
> >>        tristate "user visible if I2C is enabled"
> >>        depends on I2C
> >>
> >> config HELPER_B
> >>        tristate # hidden
> >>        select FEATURE_A
> >>
> >> config DRIVER
> >>        tristate "This driver is broken if I2C is disabled"
> >>        select HELPER_B
> >
> > So the DRIVER section should gain a "depends on I2C" statement.
>
> Why should DRIVER have to care that HELPER_B needs either FEATURE_A or
> I2C? It should only have to care about HELPER_B. And if the dependencies
> of FEATURE_A or HELPER_B later change, that's their business, not
> DRIVER's.

That's correct. But currently the dependency on I2C is not handled
automatically.

> > Yamada-san: would it be difficult to modify Kconfig to ignore symbols
> > like DRIVER that select other symbols with unmet dependencies?
> > Currently it already warns about that.
> >
> > Handling this implicitly (instead of the current explict "depends
> > on") would have the disadvantage though: a user who is not aware of
> > the implicit dependency may wonder why DRIVER is invisible in his
> > config interface.

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: [PATCH 00/11] drm: Restore helper usability

[Geert Uytterhoeven]

Hi Jani,

CC kbuild

On Mon, Apr 22, 2024 at 7:00 PM Jani Nikula <jani.nikula@linux.intel.com> wrote:
> On Mon, 22 Apr 2024, "Arnd Bergmann" <arnd@arndb.de> wrote:
> > I'm not sure where this misunderstanding comes from, as you
> > seem to be repeating the same incorrect assumption about
> > how select works that Maxime wrote in his changelog. To clarify,
> > this works exactly as one would expect:
> >
> > config HELPER_A
> >        tristate
> >
> > config HELPER_B
> >        tristate
> >        select HELPER_A
> >
> > config DRIVER
> >        tristate "Turn on the driver and the helpers it uses"
> >        select HELPER_B # this recursively selects HELPER_A
> >
> > Whereas this one is broken:
> >
> > config FEATURE_A
> >        tristate "user visible if I2C is enabled"
> >        depends on I2C
> >
> > config HELPER_B
> >        tristate # hidden
> >        select FEATURE_A
> >
> > config DRIVER
> >        tristate "This driver is broken if I2C is disabled"
> >        select HELPER_B
>
> This case is really what I was referring to, although I was sloppy with
> words there. I understand that select does work recursively for selects.
>
> >>   There is no end to this, it just goes on and on, as the
> >>   dependencies of the selected symbols change over time. Often the
> >>   selects require unintuitive if patterns that are about the
> >>   implementation details of the symbol being selected.
> >
> > Agreed, that is the problem I frequently face with drivers/gpu/drm,
> > and most of the time it can only be solved by rewriting the whole
> > system to not select user-visible symbol at all.
> >
> > Using 'depends on' by itself is unfortunately not enough to
> > avoid /all/ the problems. See e.g. today's failure
> >
> > config DRM_DISPLAY_HELPER
> >        tristate "DRM Display Helpers"
> >        default y
> >
> > config DRM_DISPLAY_DP_HELPER
> >        bool "DRM DisplayPort Helpers"
> >        depends on DRM_DISPLAY_HELPER
> >
> > config DRM_PANEL_LG_SW43408
> >        tristate "LG SW43408 panel"
> >        depends on DRM_DISPLAY_DP_HELPER
> >
> > This version is still broken for DRM_DISPLAY_HELPER=m,
> > DRM_DISPLAY_DP_HELPER=m, DRM_PANEL_LG_SW43408=y because
> > the dependency on the bool symbol is not enough to
> > ensure that DRM_DISPLAY_HELPER is also built-in, so you
> > still need explicit dependencies on both
> > DRM_DISPLAY_HELPER and DRM_DISPLAY_DP_HELPER in the users.
> >
> > This can be solved by making DRM_DISPLAY_DP_HELPER a
> > tristate symbol and adjusting the #ifdef checks and
> > Makefile logic accordingly, which is exactly what you'd
> > need to do to make it work with 'select' as well.
>
> So bool is kind of problematic for depends on and select even when it's
> not really used for describing builtin vs. no, but rather yes vs. no?

Yes, the underlying issue is that bool is used for two different things:
  A. To enable a driver module that can be only built-in,
  B. To enable an option or feature of a driver or subsystem.

Without this distinction, dependencies cannot be auto-propagated 100%
correctly.  Fixing that would require introducing a third type (and possibly
renaming the existing ones to end up with 3 good names).

Actually two types could work:
  1. driver,
  2. option,
as case A is just a driver that can only be built-in (i.e. "depends on y",
which is similar to the behavior with CONFIG_MODULES=n).

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: [PATCH 00/11] drm: Restore helper usability

[Arnd Bergmann]

On Mon, Apr 22, 2024, at 18:58, Geert Uytterhoeven wrote:
> On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
>> On Mon, Apr 22, 2024, at 15:28, Jani Nikula wrote:
>> Whereas this one is broken:
>>
>> config FEATURE_A
>>        tristate "user visible if I2C is enabled"
>>        depends on I2C
>>
>> config HELPER_B
>>        tristate # hidden
>>        select FEATURE_A
>>
>> config DRIVER
>>        tristate "This driver is broken if I2C is disabled"
>>        select HELPER_B
>
> So the DRIVER section should gain a "depends on I2C" statement.

That is of course the common workaround, but my point was
that nothing should ever 'select I2C' or any of the other
subsystems that are user visible.

> Yamada-san: would it be difficult to modify Kconfig to ignore symbols
> like DRIVER that select other symbols with unmet dependencies?
> Currently it already warns about that.
>
> Handling this implicitly (instead of the current explict "depends
> on") would have the disadvantage though: a user who is not aware of
> the implicit dependency may wonder why DRIVER is invisible in his
> config interface.

I think hiding this would make it much harder to get anything
right. The symbols in question are almost all ones that should
be enabled in normal configs, and the 'make menuconfig' help
doesn't make it too hard to figure things out normally, we just
have to find a way to avoid regressions when converting things
to 'depends on' that used an incorrect 'select'.

     Arnd

Re: [PATCH 00/11] drm: Restore helper usability

[Masahiro Yamada]

On Tue, Apr 23, 2024 at 3:24 AM Arnd Bergmann <arnd@arndb.de> wrote:
>
> On Mon, Apr 22, 2024, at 18:58, Geert Uytterhoeven wrote:
> > On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
> >> On Mon, Apr 22, 2024, at 15:28, Jani Nikula wrote:
> >> Whereas this one is broken:
> >>
> >> config FEATURE_A
> >>        tristate "user visible if I2C is enabled"
> >>        depends on I2C
> >>
> >> config HELPER_B
> >>        tristate # hidden
> >>        select FEATURE_A
> >>
> >> config DRIVER
> >>        tristate "This driver is broken if I2C is disabled"
> >>        select HELPER_B
> >
> > So the DRIVER section should gain a "depends on I2C" statement.
>
> That is of course the common workaround, but my point was
> that nothing should ever 'select I2C' or any of the other
> subsystems that are user visible.
>
> > Yamada-san: would it be difficult to modify Kconfig to ignore symbols
> > like DRIVER that select other symbols with unmet dependencies?
> > Currently it already warns about that.
> >
> > Handling this implicitly (instead of the current explict "depends
> > on") would have the disadvantage though: a user who is not aware of
> > the implicit dependency may wonder why DRIVER is invisible in his
> > config interface.
>
> I think hiding this would make it much harder to get anything
> right. The symbols in question are almost all ones that should
> be enabled in normal configs, and the 'make menuconfig' help
> doesn't make it too hard to figure things out normally, we just
> have to find a way to avoid regressions when converting things
> to 'depends on' that used an incorrect 'select'.
>
>      Arnd



I am confused because you repeatedly discussed
the missing I2C dependency.


Are you talking about DRM drivers,
or is it just "an example" in general?



DRM selects I2C.

https://github.com/torvalds/linux/blob/v6.9-rc4/drivers/gpu/drm/Kconfig#L16



If you make sure individual DRM drivers depend on DRM,
none of them can be enabled without I2C.



Currently, this is not guaranteed just because
DRM folks do not know how to use the "menuconfig" syntax.



The "menuconfig" makes sense only when it is
followed by "if".




diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
index 5a0c476361c3..6984b3fea271 100644
--- a/drivers/gpu/drm/Kconfig
+++ b/drivers/gpu/drm/Kconfig
@@ -29,6 +29,8 @@ menuconfig DRM
          details.  You should also select and configure AGP
          (/dev/agpgart) support if it is available for your platform.

+if DRM
+
 config DRM_MIPI_DBI
        tristate
        depends on DRM
@@ -414,3 +416,5 @@ config DRM_LIB_RANDOM
 config DRM_PRIVACY_SCREEN
        bool
        default n
+
+endif










--
Best Regards
Masahiro Yamada

Re: [PATCH 00/11] drm: Restore helper usability

[Arnd Bergmann]

On Mon, Apr 22, 2024, at 21:42, Masahiro Yamada wrote:
> On Tue, Apr 23, 2024 at 3:24 AM Arnd Bergmann <arnd@arndb.de> wrote:
>> On Mon, Apr 22, 2024, at 18:58, Geert Uytterhoeven wrote:
>> > On Mon, Apr 22, 2024 at 3:55 PM Arnd Bergmann <arnd@arndb.de> wrote:
>> >> On Mon, Apr 22, 2024, at 15:28, Jani Nikula wrote:
>>
>> I think hiding this would make it much harder to get anything
>> right. The symbols in question are almost all ones that should
>> be enabled in normal configs, and the 'make menuconfig' help
>> doesn't make it too hard to figure things out normally, we just
>> have to find a way to avoid regressions when converting things
>> to 'depends on' that used an incorrect 'select'.
>
> I am confused because you repeatedly discussed
> the missing I2C dependency.
>
>
> Are you talking about DRM drivers,
> or is it just "an example" in general?
>
>
>
> DRM selects I2C.

It's a prominent example because I2C ends up showing
up in most circular dependencies. I forgot that CONFIG_DRM
itself selects this one, but this is clearly part of the
overall problem:

$ git grep -w 'select I2C' | wc -l
35
$ git grep -w 'depends on I2C' | wc -l
1068

Attempting to clean up some of the incorrect 'select'
statements, such as changing DRM_NOUVEAU to 'depends on
ACPI_VIDEO' instead of 'select' tends to run into
another 'select I2C' such as this one:

drivers/i2c/Kconfig:8:	symbol I2C is selected by DRM_NOUVEAU
drivers/gpu/drm/nouveau/Kconfig:2:	symbol DRM_NOUVEAU depends on ACPI_VIDEO
drivers/acpi/Kconfig:214:	symbol ACPI_VIDEO depends on BACKLIGHT_CLASS_DEVICE
drivers/video/backlight/Kconfig:136:	symbol BACKLIGHT_CLASS_DEVICE is selected by FB_BACKLIGHT
drivers/video/fbdev/core/Kconfig:184:	symbol FB_BACKLIGHT is selected by HT16K33
drivers/auxdisplay/Kconfig:490:	symbol HT16K33 depends on I2C

Again, I2C was probably not the best example for an urgent problem
as it ends up being selected unconditionally anyway, but
I think ACPI_VIDEO and BACKLIGHT_CLASS_DEVICE are the ones that
we should stop selecting.

> diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig
> index 5a0c476361c3..6984b3fea271 100644
> --- a/drivers/gpu/drm/Kconfig
> +++ b/drivers/gpu/drm/Kconfig
> @@ -29,6 +29,8 @@ menuconfig DRM
>           details.  You should also select and configure AGP
>           (/dev/agpgart) support if it is available for your platform.
> 
> +if DRM
> +
>  config DRM_MIPI_DBI
>         tristate
>         depends on DRM
> @@ -414,3 +416,5 @@ config DRM_LIB_RANDOM
>  config DRM_PRIVACY_SCREEN
>         bool
>        default n
> +
> +endif

This is a probably good idea (aside from DRM_PANEL_ORIENTATION_QUIRKS,
which needs to be moved out of the section), but seems completely
unrelated to the issue of selecting too many symbols.

     Arnd