Changing the enable/disable sequence in commit c9b1150a68d9
("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.

This is an alternate series to Linus' series:

https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/

This series first reverts the original commit and reverts a fix for
mediatek which is no longer needed. It then exposes helper functions
from DRM core, and finally implements the new sequence only in the tidss
driver.

There is one more fix in upstream for the original commit, commit
5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
runtime PM calls"), but I have not reverted that one as it looks like a
valid patch in its own.

I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
I wasn't sure what should they point to. But I could perhaps add Fixes:
<original commit> to all of these.

Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
---
Linus Walleij (1):
      drm/atomic-helper: Export and namespace some functions

Tomi Valkeinen (3):
      Revert "drm/atomic-helper: Re-order bridge chain pre-enable and post-disable"
      Revert "drm/mediatek: dsi: Fix DSI host and panel bridge pre-enable order"
      drm/tidss: Fix enable/disable order

 drivers/gpu/drm/drm_atomic_helper.c | 122 ++++++++++++++----
 drivers/gpu/drm/mediatek/mtk_dsi.c  |   6 -
 drivers/gpu/drm/tidss/tidss_kms.c   |  30 ++++-
 include/drm/drm_atomic_helper.h     |  22 ++++
 include/drm/drm_bridge.h            | 249 ++++++++++--------------------------
 5 files changed, 214 insertions(+), 215 deletions(-)
---
base-commit: 88e721ab978a86426aa08da520de77430fa7bb84
change-id: 20251205-drm-seq-fix-b4ed1f56604b

Best regards,
-- 
Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

On Fri, 5 Dec 2025 11:51:47 +0200, Tomi Valkeinen wrote:
> Changing the enable/disable sequence in commit c9b1150a68d9
> ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
> has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
> 
> This is an alternate series to Linus' series:
> 
> [ ... ]

Reviewed-by: Maxime Ripard <mripard@kernel.org>

Thanks!
Maxime

On 05.12.2025 10:51, Tomi Valkeinen wrote:
> Changing the enable/disable sequence in commit c9b1150a68d9
> ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
> has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
>
> This is an alternate series to Linus' series:
>
> https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/
>
> This series first reverts the original commit and reverts a fix for
> mediatek which is no longer needed. It then exposes helper functions
> from DRM core, and finally implements the new sequence only in the tidss
> driver.
>
> There is one more fix in upstream for the original commit, commit
> 5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
> runtime PM calls"), but I have not reverted that one as it looks like a
> valid patch in its own.

Right, that patch is a fix on its own, the changes in the atomic 
sequence just revealed the issue.

> I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
> I wasn't sure what should they point to. But I could perhaps add Fixes:
> <original commit> to all of these.
>
> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
> ---
> Linus Walleij (1):
>        drm/atomic-helper: Export and namespace some functions
>
> Tomi Valkeinen (3):
>        Revert "drm/atomic-helper: Re-order bridge chain pre-enable and post-disable"
>        Revert "drm/mediatek: dsi: Fix DSI host and panel bridge pre-enable order"
>        drm/tidss: Fix enable/disable order
>
>   drivers/gpu/drm/drm_atomic_helper.c | 122 ++++++++++++++----
>   drivers/gpu/drm/mediatek/mtk_dsi.c  |   6 -
>   drivers/gpu/drm/tidss/tidss_kms.c   |  30 ++++-
>   include/drm/drm_atomic_helper.h     |  22 ++++
>   include/drm/drm_bridge.h            | 249 ++++++++++--------------------------
>   5 files changed, 214 insertions(+), 215 deletions(-)
> ---
> base-commit: 88e721ab978a86426aa08da520de77430fa7bb84
> change-id: 20251205-drm-seq-fix-b4ed1f56604b

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland

On Fri, Dec 5, 2025 at 10:52 AM Tomi Valkeinen
<tomi.valkeinen@ideasonboard.com> wrote:

> Changing the enable/disable sequence in commit c9b1150a68d9
> ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
> has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
>
> This is an alternate series to Linus' series:
>
> https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/
>
> This series first reverts the original commit and reverts a fix for
> mediatek which is no longer needed. It then exposes helper functions
> from DRM core, and finally implements the new sequence only in the tidss
> driver.
>
> There is one more fix in upstream for the original commit, commit
> 5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
> runtime PM calls"), but I have not reverted that one as it looks like a
> valid patch in its own.
>
> I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
> I wasn't sure what should they point to. But I could perhaps add Fixes:
> <original commit> to all of these.
>
> Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

The series:
Reviewed-by: Linus Walleij <linusw@kernel.org>
Tested-by: Linus Walleij <linusw@kernel.org>

In my opinion Laurent is the more senior maintainer and my
trust level in him is the highest of all the maintainers, so
I think we should apply this revert series if
Laurent ACKs that this is what he thinks is best for DRI.

Yours,
Linus Walleij

Hi,

On 05/12/2025 11:51, Tomi Valkeinen wrote:
> Changing the enable/disable sequence in commit c9b1150a68d9
> ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
> has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
> 
> This is an alternate series to Linus' series:
> 
> https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/
> 
> This series first reverts the original commit and reverts a fix for
> mediatek which is no longer needed. It then exposes helper functions
> from DRM core, and finally implements the new sequence only in the tidss
> driver.
> 
> There is one more fix in upstream for the original commit, commit
> 5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
> runtime PM calls"), but I have not reverted that one as it looks like a
> valid patch in its own.
> 
> I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
> I wasn't sure what should they point to. But I could perhaps add Fixes:
> <original commit> to all of these.
There has been no comments, so I assume this is the way to go.

Should we merge this series as a fix for 6.18 rcs?

 Tomi

On Fri, Dec 12, 2025 at 3:21 PM Tomi Valkeinen
<tomi.valkeinen@ideasonboard.com> wrote:
> On 05/12/2025 11:51, Tomi Valkeinen wrote:
> > Changing the enable/disable sequence in commit c9b1150a68d9
> > ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
> > has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
> >
> > This is an alternate series to Linus' series:
> >
> > https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/
> >
> > This series first reverts the original commit and reverts a fix for
> > mediatek which is no longer needed. It then exposes helper functions
> > from DRM core, and finally implements the new sequence only in the tidss
> > driver.
> >
> > There is one more fix in upstream for the original commit, commit
> > 5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
> > runtime PM calls"), but I have not reverted that one as it looks like a
> > valid patch in its own.
> >
> > I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
> > I wasn't sure what should they point to. But I could perhaps add Fixes:
> > <original commit> to all of these.

> There has been no comments, so I assume this is the way to go.
>
> Should we merge this series as a fix for 6.18 rcs?

Too late now, so let's merge it as a fix for v6.19 rcs!

Yours,
Linus Walleij

Hi,

On 14/12/2025 13:14, Linus Walleij wrote:
> On Fri, Dec 12, 2025 at 3:21 PM Tomi Valkeinen
> <tomi.valkeinen@ideasonboard.com> wrote:
>> On 05/12/2025 11:51, Tomi Valkeinen wrote:
>>> Changing the enable/disable sequence in commit c9b1150a68d9
>>> ("drm/atomic-helper: Re-order bridge chain pre-enable and post-disable")
>>> has caused regressions on multiple platforms: R-Car, MCDE, Rockchip.
>>>
>>> This is an alternate series to Linus' series:
>>>
>>> https://lore.kernel.org/all/20251202-mcde-drm-regression-thirdfix-v6-0-f1bffd4ec0fa%40kernel.org/
>>>
>>> This series first reverts the original commit and reverts a fix for
>>> mediatek which is no longer needed. It then exposes helper functions
>>> from DRM core, and finally implements the new sequence only in the tidss
>>> driver.
>>>
>>> There is one more fix in upstream for the original commit, commit
>>> 5d91394f2361 ("drm/exynos: fimd: Guard display clock control with
>>> runtime PM calls"), but I have not reverted that one as it looks like a
>>> valid patch in its own.
>>>
>>> I added Cc stable v6.17+ to all patches, but I didn't add Fixes tags, as
>>> I wasn't sure what should they point to. But I could perhaps add Fixes:
>>> <original commit> to all of these.
> 
>> There has been no comments, so I assume this is the way to go.
>>
>> Should we merge this series as a fix for 6.18 rcs?
> 
> Too late now, so let's merge it as a fix for v6.19 rcs!
Ah, right. Indeed, I meant v6.19 rcs.

 Tomi