linux-next-20251029 - build error in amdgpu

linux-next-20251029 - build error in amdgpu

[Valdis Klētnieks]

Hit this error while building next-20251029 with W=1 and gcc (GCC) 15.2.1 20251022 (Red Hat 15.2.1-3

  CC [M]  drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.o
drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.c:1571:69: error: positional initialization of field in 'struct' declared with 'designated_init' attribute [-Werror=designated-init]
 1571 | static const struct stream_encoder_funcs dce110_an_str_enc_funcs = {0};
      |                                                                     ^
drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.c:1571:69: note: (near initialization for 'dce110_an_str_enc_funcs')
cc1: some warnings being treated as errors

Caused by:
commit deb072d241c906ae378bc88ff91dfcdc34ebe71a
Author: Timur Krist?f <timur.kristof@gmail.com>
Date:   Fri Sep 26 20:01:51 2025 +0200

    drm/amd/display: Implement DCE analog stream encoders


\0\0\0\0\0\0\0\0

Re: linux-next-20251029 - build error in amdgpu

[Timur Kristóf]

On Thu, 2025-10-30 at 13:20 -0400, Valdis Klētnieks wrote:
> Hit this error while building next-20251029 with W=1 and gcc (GCC)
> 15.2.1 20251022 (Red Hat 15.2.1-3

Hi,

Can you say how to reproduce that?
I use the same version of GCC but it hasn't given me any warning or
error for that patch.

Thanks & best regards,
Timur

> 
>   CC [M] 
> drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.o
> drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.c:157
> 1:69: error: positional initialization of field in 'struct' declared
> with 'designated_init' attribute [-Werror=designated-init]
>  1571 | static const struct stream_encoder_funcs
> dce110_an_str_enc_funcs = {0};
>      
> |                                                                    
> ^
> drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_stream_encoder.c:157
> 1:69: note: (near initialization for 'dce110_an_str_enc_funcs')
> cc1: some warnings being treated as errors
> 
> Caused by:
> commit deb072d241c906ae378bc88ff91dfcdc34ebe71a
> Author: Timur Krist?f <timur.kristof@gmail.com>
> Date:   Fri Sep 26 20:01:51 2025 +0200
> 
>     drm/amd/display: Implement DCE analog stream encoders
> 
> 
> > > > > > > > > 

Re: linux-next-20251029 - build error in amdgpu

[Valdis Klētnieks]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=utf-8, Size: 738 bytes --]

On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:

> Can you say how to reproduce that?
> I use the same version of GCC but it hasn't given me any warning or
> error for that patch.

Upon further testing, 

[/usr/src/linux-next] make 

is sufficient on my system. Turns out that it errors out even without W=1. My
next guess was that it had to do with -Werror, which would cause warnings to
be treated as errors, but my .config has

# CONFIG_WERROR is not set
# CONFIG_DRM_WERROR is not set
# CONFIG_DRM_AMDGPU_WERROR is not set

Do you perhaps have patches in your tree that aren't in next-20251029?

I wonder if Fedora's build of gcc 15.2.1 is different somehow from the
build you're using....


\0\0\0\0\0\0\0\0

Re: linux-next-20251029 - build error in amdgpu

[Timur Kristóf]

[-- Attachment #1: Type: text/plain, Size: 1239 bytes --]

Hello Valdis,

I use Fedora and I developed that patch on Fedora, using the same GCC
version. It did not give me an error or warning when I wrote the patch. It
is still not giving me an error or warning about that code now.

I am working with the amd-staging-drm-next branch.

What do I need to do to trigger the warning?

Thanks & best regards,
Timur

Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont: 2025. nov.
2., Vas 1:59):

> On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
>
> > Can you say how to reproduce that?
> > I use the same version of GCC but it hasn't given me any warning or
> > error for that patch.
>
> Upon further testing,
>
> [/usr/src/linux-next] make
>
> is sufficient on my system. Turns out that it errors out even without W=1.
> My
> next guess was that it had to do with -Werror, which would cause warnings
> to
> be treated as errors, but my .config has
>
> # CONFIG_WERROR is not set
> # CONFIG_DRM_WERROR is not set
> # CONFIG_DRM_AMDGPU_WERROR is not set
>
> Do you perhaps have patches in your tree that aren't in next-20251029?
>
> I wonder if Fedora's build of gcc 15.2.1 is different somehow from the
> build you're using....
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 1797 bytes --]

Re: linux-next-20251029 - build error in amdgpu

[Amit Dhingra]

On Sun, Nov 02, 2025 at 10:27:54AM +0100, Timur Kristóf wrote:
> Hello Valdis,
>
> I use Fedora and I developed that patch on Fedora, using the same GCC
> version. It did not give me an error or warning when I wrote the patch. It
> is still not giving me an error or warning about that code now.
>
> I am working with the amd-staging-drm-next branch.
>
> What do I need to do to trigger the warning?
>
I have the same problem.

linux-next on Arch Linux
GCC : 15.2.1 20250813


Steps to reproduce the error
- make mrproper
- make allmodconfig
- make drivers/gpu/drm/amd/amdgpu/

Turns out its due to randconfig, See [1]

Setting RANDSTRUCT_NONE=y seems to stop the error.

[1] https://lore.kernel.org/all/2025062439-tamer-diner-68e9@gregkh/

- Amit

> Thanks & best regards,
> Timur
>
> Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont: 2025. nov.
> 2., Vas 1:59):
>
> > On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
> >
> > > Can you say how to reproduce that?
> > > I use the same version of GCC but it hasn't given me any warning or
> > > error for that patch.
> >
> > Upon further testing,
> >
> > [/usr/src/linux-next] make
> >
> > is sufficient on my system. Turns out that it errors out even without W=1.
> > My
> > next guess was that it had to do with -Werror, which would cause warnings
> > to
> > be treated as errors, but my .config has
> >
> > # CONFIG_WERROR is not set
> > # CONFIG_DRM_WERROR is not set
> > # CONFIG_DRM_AMDGPU_WERROR is not set
> >
> > Do you perhaps have patches in your tree that aren't in next-20251029?
> >
> > I wonder if Fedora's build of gcc 15.2.1 is different somehow from the
> > build you're using....
> >
> >
> >
> >

Re: linux-next-20251029 - build error in amdgpu

[Timur Kristóf]

On Wed, 2025-11-05 at 18:34 -0800, Amit Dhingra wrote:
> 
> > 
> > What do I need to do to trigger the warning?
> > 
> I have the same problem.
> 
> linux-next on Arch Linux
> GCC : 15.2.1 20250813
> 
> 
> Steps to reproduce the error
> - make mrproper
> - make allmodconfig
> - make drivers/gpu/drm/amd/amdgpu/
> 
> Turns out its due to randconfig, See [1]
> 
> Setting RANDSTRUCT_NONE=y seems to stop the error.

Can you please try and see if this fixes the warning?
https://pastebin.com/raw/b8j3UABj

Alternatively, if anyone has other recommendations on how to fix it,
I'm also open to suggestions.

Thanks,
Timur


> > 
> > Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont: 2025.
> > nov.
> > 2., Vas 1:59):
> > 
> > > On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
> > > 
> > > > Can you say how to reproduce that?
> > > > I use the same version of GCC but it hasn't given me any
> > > > warning or
> > > > error for that patch.
> > > 
> > > Upon further testing,
> > > 
> > > [/usr/src/linux-next] make
> > > 
> > > is sufficient on my system. Turns out that it errors out even
> > > without W=1.
> > > My
> > > next guess was that it had to do with -Werror, which would cause
> > > warnings
> > > to
> > > be treated as errors, but my .config has
> > > 
> > > # CONFIG_WERROR is not set
> > > # CONFIG_DRM_WERROR is not set
> > > # CONFIG_DRM_AMDGPU_WERROR is not set
> > > 
> > > Do you perhaps have patches in your tree that aren't in next-
> > > 20251029?
> > > 
> > > I wonder if Fedora's build of gcc 15.2.1 is different somehow
> > > from the
> > > build you're using....
> > > 
> > > 
> > > 
> > > 

Re: linux-next-20251029 - build error in amdgpu

[Valdis Klētnieks]

On Wed, 05 Nov 2025 18:34:05 -0800, Amit Dhingra said:

> I have the same problem.
>  ...

> Setting RANDSTRUCT_NONE=y seems to stop the error.
>
> [1] https://lore.kernel.org/all/2025062439-tamer-diner-68e9@gregkh/

Wow.  I wouldn't have guessed that RANDSTRUCT would be the cause...
However, RANDSTRUCT_NONE=y does make dce_stream_encoder.c compile cleanly.

<insert Twilight Zone theme music here>

Adding Kees Cook and the linux-hardening list to the cc:, hope somebody has an
idea what's going on.

\0\0\0\0\0\0\0\0

Re: linux-next-20251029 - build error in amdgpu

[Amit Dhingra]

On Thu, Nov 06, 2025 at 03:40:52AM +0100, Timur Kristóf wrote:
> On Wed, 2025-11-05 at 18:34 -0800, Amit Dhingra wrote:
> >
> > >
> > > What do I need to do to trigger the warning?
> > >
> > I have the same problem.
> >
> > linux-next on Arch Linux
> > GCC : 15.2.1 20250813
> >
> >
> > Steps to reproduce the error
> > - make mrproper
> > - make allmodconfig
> > - make drivers/gpu/drm/amd/amdgpu/
> >
> > Turns out its due to randconfig, See [1]
> >
> > Setting RANDSTRUCT_NONE=y seems to stop the error.
>
> Can you please try and see if this fixes the warning?
> https://pastebin.com/raw/b8j3UABj

The patch above defining all fields seems to fix the error.

>
> Alternatively, if anyone has other recommendations on how to fix it,
> I'm also open to suggestions.
>

What seemed to also work for me is to defining the struct as below
without an initialization.

static const struct stream_encoder_funcs dce110_an_str_enc_funcs;

Its static so should be initialized zeroed out. objdump shows correct size
allocation.

objdump -x dce_stream_encoder.o

  0000000000000b60 l     O .rodata	0000000000000130 dce110_str_enc_funcs
  0000000000000a00 l     O .rodata	0000000000000130 dce110_an_str_enc_funcs


> Thanks,
> Timur
>
>
> > >
> > > Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont: 2025.
> > > nov.
> > > 2., Vas 1:59):
> > >
> > > > On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
> > > >
> > > > > Can you say how to reproduce that?
> > > > > I use the same version of GCC but it hasn't given me any
> > > > > warning or
> > > > > error for that patch.
> > > >
> > > > Upon further testing,
> > > >
> > > > [/usr/src/linux-next] make
> > > >
> > > > is sufficient on my system. Turns out that it errors out even
> > > > without W=1.
> > > > My
> > > > next guess was that it had to do with -Werror, which would cause
> > > > warnings
> > > > to
> > > > be treated as errors, but my .config has
> > > >
> > > > # CONFIG_WERROR is not set
> > > > # CONFIG_DRM_WERROR is not set
> > > > # CONFIG_DRM_AMDGPU_WERROR is not set
> > > >
> > > > Do you perhaps have patches in your tree that aren't in next-
> > > > 20251029?
> > > >
> > > > I wonder if Fedora's build of gcc 15.2.1 is different somehow
> > > > from the
> > > > build you're using....
> > > >
> > > >
> > > >
> > > >

Re: linux-next-20251029 - build error in amdgpu

[Timur Kristóf]

On Wed, 2025-11-05 at 19:24 -0800, Amit Dhingra wrote:
> On Thu, Nov 06, 2025 at 03:40:52AM +0100, Timur Kristóf wrote:
> > On Wed, 2025-11-05 at 18:34 -0800, Amit Dhingra wrote:
> > > 
> > > > 
> > > > What do I need to do to trigger the warning?
> > > > 
> > > I have the same problem.
> > > 
> > > linux-next on Arch Linux
> > > GCC : 15.2.1 20250813
> > > 
> > > 
> > > Steps to reproduce the error
> > > - make mrproper
> > > - make allmodconfig
> > > - make drivers/gpu/drm/amd/amdgpu/
> > > 
> > > Turns out its due to randconfig, See [1]
> > > 
> > > Setting RANDSTRUCT_NONE=y seems to stop the error.
> > 
> > Can you please try and see if this fixes the warning?
> > https://pastebin.com/raw/b8j3UABj
> 
> The patch above defining all fields seems to fix the error.
> 
> > 
> > Alternatively, if anyone has other recommendations on how to fix
> > it,
> > I'm also open to suggestions.
> > 
> 
> What seemed to also work for me is to defining the struct as below
> without an initialization.
> 
> static const struct stream_encoder_funcs dce110_an_str_enc_funcs;
> 
> Its static so should be initialized zeroed out. objdump shows correct
> size
> allocation.
> 
> objdump -x dce_stream_encoder.o
> 
>   0000000000000b60 l     O .rodata	0000000000000130
> dce110_str_enc_funcs
>   0000000000000a00 l     O .rodata	0000000000000130
> dce110_an_str_enc_funcs
> 

Thanks! This looks even cleaner.

I will submit a patch to fix the warning next week.
Can I add your name and email to the Suggested-by tag?

Thanks,
Timur



> 
> > Thanks,
> > Timur
> > 
> > 
> > > > 
> > > > Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont:
> > > > 2025.
> > > > nov.
> > > > 2., Vas 1:59):
> > > > 
> > > > > On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
> > > > > 
> > > > > > Can you say how to reproduce that?
> > > > > > I use the same version of GCC but it hasn't given me any
> > > > > > warning or
> > > > > > error for that patch.
> > > > > 
> > > > > Upon further testing,
> > > > > 
> > > > > [/usr/src/linux-next] make
> > > > > 
> > > > > is sufficient on my system. Turns out that it errors out even
> > > > > without W=1.
> > > > > My
> > > > > next guess was that it had to do with -Werror, which would
> > > > > cause
> > > > > warnings
> > > > > to
> > > > > be treated as errors, but my .config has
> > > > > 
> > > > > # CONFIG_WERROR is not set
> > > > > # CONFIG_DRM_WERROR is not set
> > > > > # CONFIG_DRM_AMDGPU_WERROR is not set
> > > > > 
> > > > > Do you perhaps have patches in your tree that aren't in next-
> > > > > 20251029?
> > > > > 
> > > > > I wonder if Fedora's build of gcc 15.2.1 is different somehow
> > > > > from the
> > > > > build you're using....
> > > > > 
> > > > > 
> > > > > 
> > > > > 

Re: linux-next-20251029 - build error in amdgpu

[Valdis Klētnieks]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=utf-8, Size: 407 bytes --]

On Wed, 05 Nov 2025 19:24:22 -0800, Amit Dhingra said:
> On Thu, Nov 06, 2025 at 03:40:52AM +0100, Timur Kristóf wrote:

> The patch above defining all fields seems to fix the error.

The bigger question is whether that patch merely papers over a bug in the gcc
plugin. Basically, why does the struct being one layout compile cleanly, but
throws an error if the layout is randomized...

\0m>
\0\0\0\0

Re: linux-next-20251029 - build error in amdgpu

[Amit Dhingra]

On Thu, Nov 06, 2025 at 04:29:05AM +0100, Timur Kristóf wrote:
> On Wed, 2025-11-05 at 19:24 -0800, Amit Dhingra wrote:
> > On Thu, Nov 06, 2025 at 03:40:52AM +0100, Timur Kristóf wrote:
> > > On Wed, 2025-11-05 at 18:34 -0800, Amit Dhingra wrote:
> > > >
> > > > >
> > > > > What do I need to do to trigger the warning?
> > > > >
> > > > I have the same problem.
> > > >
> > > > linux-next on Arch Linux
> > > > GCC : 15.2.1 20250813
> > > >
> > > >
> > > > Steps to reproduce the error
> > > > - make mrproper
> > > > - make allmodconfig
> > > > - make drivers/gpu/drm/amd/amdgpu/
> > > >
> > > > Turns out its due to randconfig, See [1]
> > > >
> > > > Setting RANDSTRUCT_NONE=y seems to stop the error.
> > >
> > > Can you please try and see if this fixes the warning?
> > > https://pastebin.com/raw/b8j3UABj
> >
> > The patch above defining all fields seems to fix the error.
> >
> > >
> > > Alternatively, if anyone has other recommendations on how to fix
> > > it,
> > > I'm also open to suggestions.
> > >
> >
> > What seemed to also work for me is to defining the struct as below
> > without an initialization.
> >
> > static const struct stream_encoder_funcs dce110_an_str_enc_funcs;
> >
> > Its static so should be initialized zeroed out. objdump shows correct
> > size
> > allocation.
> >
> > objdump -x dce_stream_encoder.o
> >
> >   0000000000000b60 l     O .rodata	0000000000000130
> > dce110_str_enc_funcs
> >   0000000000000a00 l     O .rodata	0000000000000130
> > dce110_an_str_enc_funcs
> >
>
> Thanks! This looks even cleaner.
>
> I will submit a patch to fix the warning next week.
> Can I add your name and email to the Suggested-by tag?

Certainly. Thank you!!

>
> Thanks,
> Timur
>
>
>
> >
> > > Thanks,
> > > Timur
> > >
> > >
> > > > >
> > > > > Valdis Klētnieks <valdis.kletnieks@vt.edu> ezt írta (időpont:
> > > > > 2025.
> > > > > nov.
> > > > > 2., Vas 1:59):
> > > > >
> > > > > > On Thu, 30 Oct 2025 19:05:30 +0100, Timur Kristóf said:
> > > > > >
> > > > > > > Can you say how to reproduce that?
> > > > > > > I use the same version of GCC but it hasn't given me any
> > > > > > > warning or
> > > > > > > error for that patch.
> > > > > >
> > > > > > Upon further testing,
> > > > > >
> > > > > > [/usr/src/linux-next] make
> > > > > >
> > > > > > is sufficient on my system. Turns out that it errors out even
> > > > > > without W=1.
> > > > > > My
> > > > > > next guess was that it had to do with -Werror, which would
> > > > > > cause
> > > > > > warnings
> > > > > > to
> > > > > > be treated as errors, but my .config has
> > > > > >
> > > > > > # CONFIG_WERROR is not set
> > > > > > # CONFIG_DRM_WERROR is not set
> > > > > > # CONFIG_DRM_AMDGPU_WERROR is not set
> > > > > >
> > > > > > Do you perhaps have patches in your tree that aren't in next-
> > > > > > 20251029?
> > > > > >
> > > > > > I wonder if Fedora's build of gcc 15.2.1 is different somehow
> > > > > > from the
> > > > > > build you're using....
> > > > > >
> > > > > >
> > > > > >
> > > > > >

Re: linux-next-20251029 - build error in amdgpu

[Kees Cook]

On November 5, 2025 7:18:51 PM PST, "Valdis Klētnieks" <valdis.kletnieks@vt.edu> wrote:
>On Wed, 05 Nov 2025 18:34:05 -0800, Amit Dhingra said:
>
>> I have the same problem.
>>  ...
>
>> Setting RANDSTRUCT_NONE=y seems to stop the error.
>>
>> [1] https://lore.kernel.org/all/2025062439-tamer-diner-68e9@gregkh/
>
>Wow.  I wouldn't have guessed that RANDSTRUCT would be the cause...
>However, RANDSTRUCT_NONE=y does make dce_stream_encoder.c compile cleanly.
>
><insert Twilight Zone theme music here>
>
>Adding Kees Cook and the linux-hardening list to the cc:, hope somebody has an
>idea what's going on.

I'm surprised the stable team didn't find the associated fix in your linked thread. It had the right Fixes tag:
https://lore.kernel.org/all/20250502224156.work.617-kees@kernel.org/

Regardless, just remove the "0" from the dce110_an_str_enc_funcs initializer; that's an older code pattern that isn't needed any more.

Randstruct requires designated initializers, though I thought the macro turned that on universally, so the fact that the error is only present with randstruct seems like a separate bug.

-Kees

-- 
Kees Cook