linux-next: build failure in final build

linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 1592 bytes --]

Hi all,

After merging all trees, today's final linux-next build (arm64
allyesconfig) failed like this:

error[E0560]: struct `drm_panthor_gpu_info` has no field named `pad0`
  --> /tmp/next/build/drivers/gpu/drm/tyr/gpu.rs:75:13
   |
75 |             pad0: 0,
   |             ^^^^ `drm_panthor_gpu_info` does not have this field
   |
   = note: available fields are: `selected_coherency`

Caused by commit

   8304c44631c37 (drm/tyr: use generated bindings for GpuInfo)

I tried reverting that but got several other build errors:

error: field `device` is never read
  --> /tmp/next/build/drivers/gpu/drm/tyr/driver.rs:36:5
   |
35 | pub(crate) struct TyrDriver {
   |                   --------- field in this struct
36 |     device: ARef<TyrDevice>,
   |     ^^^^^^
   |
   = note: `-D dead-code` implied by `-D warnings`
   = help: to override `-D warnings` add `#[allow(dead_code)]`

error: fields `mali` and `sram` are never read
   --> /tmp/next/build/drivers/gpu/drm/tyr/driver.rs:202:5
    |
201 | struct Regulators {
    |        ---------- fields in this struct
202 |     mali: Regulator<regulator::Enabled>,
    |     ^^^^
203 |     sram: Regulator<regulator::Enabled>,
    |     ^^^^

at which point I gave up and marked the driver as BROKEN for today, I
might take another look tomorrow.

I note that the dependency that the rust DRM drivers have on DRM=y means
that they evade allmodconfig build coverage which is not playing well
here.  I will move the drm-rust and drm-nova trees to be the last trees
merged to mitigate this issue, their fixes trees will still be affected.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure in final build

[Miguel Ojeda]

On Mon, Jan 19, 2026 at 11:30 PM Mark Brown <broonie@kernel.org> wrote:
>
> at which point I gave up and marked the driver as BROKEN for today, I
> might take another look tomorrow.

If they are "just" warnings like the latter ones (i.e. CONFIG_WERROR
is enabled), then I guess one option is to leave them there for a
cycle or two, but I don't know that can be easily tracked to avoid
trees accumulating them... :(

> I note that the dependency that the rust DRM drivers have on DRM=y means
> that they evade allmodconfig build coverage which is not playing well
> here.  I will move the drm-rust and drm-nova trees to be the last trees
> merged to mitigate this issue, their fixes trees will still be affected.

Yeah, sorry about that -- it is a requirement at the moment, but it
will be lifted with build system changes soon.

Cheers,
Miguel

Re: linux-next: build failure in final build

[Danilo Krummrich]

(Cc: Boris)

On Mon Jan 19, 2026 at 11:30 PM CET, Mark Brown wrote:
> Hi all,
>
> After merging all trees, today's final linux-next build (arm64
> allyesconfig) failed like this:
>
> error[E0560]: struct `drm_panthor_gpu_info` has no field named `pad0`
>   --> /tmp/next/build/drivers/gpu/drm/tyr/gpu.rs:75:13
>    |
> 75 |             pad0: 0,
>    |             ^^^^ `drm_panthor_gpu_info` does not have this field
>    |
>    = note: available fields are: `selected_coherency`
>
> Caused by commit
>
>    8304c44631c37 (drm/tyr: use generated bindings for GpuInfo)

I had a quick look and the problem is that another tree (drm-misc-next) changed
the corresponding uAPI struct in commit ea78ec982653 ("drm/panthor: Expose the
selected coherency protocol to the UMD") without also changing the Tyr driver.

This diff in Tyr should fix the problem:

diff --git a/drivers/gpu/drm/tyr/gpu.rs b/drivers/gpu/drm/tyr/gpu.rs
index 3072562e36e5..0c85f03b8a7d 100644
--- a/drivers/gpu/drm/tyr/gpu.rs
+++ b/drivers/gpu/drm/tyr/gpu.rs
@@ -72,7 +72,7 @@ pub(crate) fn new(dev: &Device<Bound>, iomem: &Devres<IoMem>) -> Result<Self> {
             // TODO: Add texture_features_{1,2,3}.
             texture_features: [texture_features, 0, 0, 0],
             as_present,
-            pad0: 0,
+            selected_coherency: 0, // Some variant of `enum drm_panthor_gpu_coherency`.
             shader_present,
             l2_present,
             tiler_present,

Re: linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 779 bytes --]

On Mon, Jan 19, 2026 at 11:37:26PM +0100, Miguel Ojeda wrote:
> On Mon, Jan 19, 2026 at 11:30 PM Mark Brown <broonie@kernel.org> wrote:

> > at which point I gave up and marked the driver as BROKEN for today, I
> > might take another look tomorrow.

> If they are "just" warnings like the latter ones (i.e. CONFIG_WERROR
> is enabled), then I guess one option is to leave them there for a
> cycle or two, but I don't know that can be easily tracked to avoid
> trees accumulating them... :(

Given that all*config (and x86 defconfig) enable CONFIG_WERROR I don't
think this is sustainable, warnings are fatal as far as -next and
general build coverage are concerned.  The tooling can't tell if a build
failure is generated by -Werror or if it's a standard error.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure in final build

[Alice Ryhl]

On Mon, Jan 19, 2026 at 11:49:10PM +0100, Danilo Krummrich wrote:
> (Cc: Boris)
> 
> On Mon Jan 19, 2026 at 11:30 PM CET, Mark Brown wrote:
> > Hi all,
> >
> > After merging all trees, today's final linux-next build (arm64
> > allyesconfig) failed like this:
> >
> > error[E0560]: struct `drm_panthor_gpu_info` has no field named `pad0`
> >   --> /tmp/next/build/drivers/gpu/drm/tyr/gpu.rs:75:13
> >    |
> > 75 |             pad0: 0,
> >    |             ^^^^ `drm_panthor_gpu_info` does not have this field
> >    |
> >    = note: available fields are: `selected_coherency`
> >
> > Caused by commit
> >
> >    8304c44631c37 (drm/tyr: use generated bindings for GpuInfo)
> 
> I had a quick look and the problem is that another tree (drm-misc-next) changed
> the corresponding uAPI struct in commit ea78ec982653 ("drm/panthor: Expose the
> selected coherency protocol to the UMD") without also changing the Tyr driver.
> 
> This diff in Tyr should fix the problem:
> 
> diff --git a/drivers/gpu/drm/tyr/gpu.rs b/drivers/gpu/drm/tyr/gpu.rs
> index 3072562e36e5..0c85f03b8a7d 100644
> --- a/drivers/gpu/drm/tyr/gpu.rs
> +++ b/drivers/gpu/drm/tyr/gpu.rs
> @@ -72,7 +72,7 @@ pub(crate) fn new(dev: &Device<Bound>, iomem: &Devres<IoMem>) -> Result<Self> {
>              // TODO: Add texture_features_{1,2,3}.
>              texture_features: [texture_features, 0, 0, 0],
>              as_present,
> -            pad0: 0,
> +            selected_coherency: 0, // Some variant of `enum drm_panthor_gpu_coherency`.
>              shader_present,
>              l2_present,
>              tiler_present,

Yeah, if that diff can be made in the merge commit, it should solve the
issue.

If it's easier, we could merge a commit into drm-misc-next that renames
pad0 to selected_coherency. That would trigger a merge conflict on the
relevant lines of code.

Alice

Re: linux-next: build failure in final build

[Danilo Krummrich]

On Tue Jan 20, 2026 at 9:36 AM CET, Alice Ryhl wrote:
> On Mon, Jan 19, 2026 at 11:49:10PM +0100, Danilo Krummrich wrote:
>> (Cc: Boris)
>> 
>> On Mon Jan 19, 2026 at 11:30 PM CET, Mark Brown wrote:
>> > Hi all,
>> >
>> > After merging all trees, today's final linux-next build (arm64
>> > allyesconfig) failed like this:
>> >
>> > error[E0560]: struct `drm_panthor_gpu_info` has no field named `pad0`
>> >   --> /tmp/next/build/drivers/gpu/drm/tyr/gpu.rs:75:13
>> >    |
>> > 75 |             pad0: 0,
>> >    |             ^^^^ `drm_panthor_gpu_info` does not have this field
>> >    |
>> >    = note: available fields are: `selected_coherency`
>> >
>> > Caused by commit
>> >
>> >    8304c44631c37 (drm/tyr: use generated bindings for GpuInfo)
>> 
>> I had a quick look and the problem is that another tree (drm-misc-next) changed
>> the corresponding uAPI struct in commit ea78ec982653 ("drm/panthor: Expose the
>> selected coherency protocol to the UMD") without also changing the Tyr driver.
>> 
>> This diff in Tyr should fix the problem:
>> 
>> diff --git a/drivers/gpu/drm/tyr/gpu.rs b/drivers/gpu/drm/tyr/gpu.rs
>> index 3072562e36e5..0c85f03b8a7d 100644
>> --- a/drivers/gpu/drm/tyr/gpu.rs
>> +++ b/drivers/gpu/drm/tyr/gpu.rs
>> @@ -72,7 +72,7 @@ pub(crate) fn new(dev: &Device<Bound>, iomem: &Devres<IoMem>) -> Result<Self> {
>>              // TODO: Add texture_features_{1,2,3}.
>>              texture_features: [texture_features, 0, 0, 0],
>>              as_present,
>> -            pad0: 0,
>> +            selected_coherency: 0, // Some variant of `enum drm_panthor_gpu_coherency`.
>>              shader_present,
>>              l2_present,
>>              tiler_present,
>
> Yeah, if that diff can be made in the merge commit, it should solve the
> issue.
>
> If it's easier, we could merge a commit into drm-misc-next that renames
> pad0 to selected_coherency. That would trigger a merge conflict on the
> relevant lines of code.

I think that would make sense as this way it also becomes obvious when the
changes are pulled into the parent DRM tree eventually.

linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 4078 bytes --]

Hi all,

During the final builds, today's linux-next build (arm64 allyesconfig)
failed like this:

error[E0405]: cannot find trait `AlwaysRefCounted` in module
`$crate::types`
   --> /tmp/next/build/rust/kernel/drm/gem/mod.rs:42:62
    |
42 | unsafe impl $( <$( $tparam_id ),+> )? $crate::types::AlwaysRefCounted
for $type
    |                                                              ^^^^^^^^^^^^^^^^
    |                                                              not
    |                                                              found in
    |                                                              `$crate::types`
...  308 | impl_aref_for_gem_obj!(impl<T> for Object<T> where T:
DriverObject);
    | -------------------------------------------------------------------
    | in this macro invocation
    |
    = note: this error originates in the macro `impl_aref_for_gem_obj` (in Nightly builds, run with -Z macro-backtrace for more info)
help: consider importing this trait
    |
7 + use crate::sync::aref::AlwaysRefCounted;
    |

error[E0277]: the trait bound `Object<T>: aref::AlwaysRefCounted` is
not satisfied
   --> /tmp/next/build/rust/kernel/drm/gem/mod.rs:142:41
    |
142 | impl<T: DriverObject> IntoGEMObject for Object<T> {
    |                                         ^^^^^^^^^ the trait
    |                                         `aref::AlwaysRefCounted`
    |                                         is not implemented for
    |                                         `Object<T>`
    |
    = help: the following other types implement trait `aref::AlwaysRefCounted`:
              Chip<T>
              Credential
              FwNode
              I2cAdapter
              I2cClient
              Interface
              LocalFile
              Mm
            and 12 others
note: required by a bound in `IntoGEMObject`
   --> /tmp/next/build/rust/kernel/drm/gem/mod.rs:96:59
    |
96 | pub trait IntoGEMObject: Sized + super::private::Sealed +
AlwaysRefCounted {
    |                                                           ^^^^^^^^^^^^^^^^
    |                                                           required
    |                                                           by this
    |                                                           bound in
    |                                                           `IntoGEMObject`

error[E0277]: the trait bound `Object<T>: aref::AlwaysRefCounted` is
not satisfied
   --> /tmp/next/build/rust/kernel/drm/gem/mod.rs:320:37
    |
320 | impl<T: DriverObject> AllocImpl for Object<T> {
    |                                     ^^^^^^^^^ the trait
    |                                     `aref::AlwaysRefCounted` is
    |                                     not implemented for `Object<T>`
    |
    = help: the following other types implement trait `aref::AlwaysRefCounted`:
              Chip<T>
              Credential
              FwNode
              I2cAdapter
              I2cClient
              Interface
              LocalFile
              Mm
            and 12 others
note: required for `Object<T>` to implement `IntoGEMObject`
   --> /tmp/next/build/rust/kernel/drm/gem/mod.rs:96:11
    |
96 | pub trait IntoGEMObject: Sized + super::private::Sealed +
AlwaysRefCounted {
    |           ^^^^^^^^^^^^^
note: required by a bound in `AllocImpl`
   --> /tmp/next/build/rust/kernel/drm/driver.rs:90:47
    |
90 | pub trait AllocImpl: super::private::Sealed + drm::gem::IntoGEMObject
{
    |                                               ^^^^^^^^^^^^^^^^^^^^^^^
    |                                               required by this
    |                                               bound in `AllocImpl`

error: aborting due to 3 previous errors

Some errors have detailed explanations: E0277, E0405.  For more
information about an error, try `rustc --explain E0277`.

Presumably caused by one of these commits:

  442ba16a5a513 (rust: gem: Introduce DriverObject::Args)
  e64b9cc293ae7 (rust: drm: Add gem::impl_aref_for_gem_obj!)

I have ignored this for today.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure in final build

[Danilo Krummrich]

On Thu Mar 26, 2026 at 5:04 PM CET, Mark Brown wrote:
> Hi all,
>
> During the final builds, today's linux-next build (arm64 allyesconfig)
> failed like this:
>
> error[E0405]: cannot find trait `AlwaysRefCounted` in module

This is due to a conflict with rust-next, which eventually removed the old
import path of AlwaysRefCounted (kernel::types::AlwaysRefCounted).

We've already fixed this up in [1] and applied it to the drm-rust tree, so those
should be gone for the next round.

Thanks,
Danilo

[1] https://lore.kernel.org/all/20260326-drm-rust-next-fix-aref-v1-0-7f6f58d2828a@google.com/

linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 503 bytes --]

Hi all,

After merging the kthread tree, today's linux-next build (i386
defconfig) failed like this:

/tmp/next/build/kernel/trace/trace.c:820:5: error: no previous prototype for 'tracing_alloc_snapshot' [-Werror=missing-prototypes]
  820 | int tracing_alloc_snapshot(void)
      |     ^~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors

Caused by commit

  bade44fe546212 (tracing: Move snapshot code out of trace.c and into trace_snapshot.c)
   
I've left this for today.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 491 bytes --]

Hi all,

In the final build, today's linux-next build (i386 defconfig) failed
like this:

/tmp/next/build/kernel/trace/trace.c:820:5: error: no previous prototype for 'tracing_alloc_snapshot' [-Werror=missing-prototypes]
  820 | int tracing_alloc_snapshot(void)
      |     ^~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors

Caused by commit

  bade44fe546212 (tracing: Move snapshot code out of trace.c and into trace_snapshot.c)
   
I've left this for today.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: linux-next: build failure in final build

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 267 bytes --]

On Fri, Mar 27, 2026 at 10:01:06PM +0000, Mark Brown wrote:
> Hi all,
> 
> After merging the kthread tree, today's linux-next build (i386
> defconfig) failed like this:

Sorry, didn't finish updating the template before I gut send - the
kthread tree is fine.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]