Re: [PATCH v11 0/3] media: mediatek: support mdp3 on mt8183 platform

[AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>]

Il 05/01/22 10:37, Moudy Ho ha scritto:
> Change since v10:
> - The routing table needs to be discarded, and the calculation result
>    on the SCP side is used to write a suitable mux setting for
>    1 input port and 2 output ports.
> - Adjust dts parsing flow to remove redundant HW IDs.
> - Fix memory leak caused by no free path information in function "mdp_cmdq_send".
> 
> Change since v9:
> - Keep only the MDP3 driver patches and split the remaining mmsys and
>    mutex patches into another mail.
> - Move mutex mod settings to corresponding driver and make relevant adjustments
>    for this in MDP3 driver.
> - Fix compile warning reported by kernel test robot.
> 
> Change since v8:
> - Rebase on v5.16-rc2.
> - Refer to Angelo's suggestion, adjust the register writing format to increase
>    readability and significance.
> - Refer to Angelo's suggestion, adjust or reduce inappropriate debugging
>    messages.
> - Refer to Rob Herring's suggestion to correct the the binding file
>    to make it with the specification.
> - Fix compile warning reported by kernel test robot.
> 
> Change since v7:
> - Rebase on v5.15-rc6.
> - Revise several V4L2 M2M settings to pass v4l2-compliance test.
> - Integrate those same component dt-binding documents of DRM and MDP, and
>    move them under the MMSYS domain.
> - Split MMSYS and MUTEX into two different files according to
>    their functional properties.
> 
> Changes since v6:
> - Refactor GCE event to corresponding node.
> - Fix dt_binding_check fail.
> - Fix compilation errors.
> 
> Changes since v5:
> - Rebase on v5.14-rc6.
> - Move MMSYS/Mutex settings to corresponding driver.
> - Revise the software license description and copyright.
> - Remove unnecessary enum. or definitions.
> - Optimize platform/chip definition conditions.
> - Use general printing functions instead of MDP3 private ones.
> - Fix compile warning.
> 
> Changes since v4:
> - Rebase on v5.13-rc1.
> - Remove the CMDQ flush flow to match the CMDQ API change.
> - Integrate four of MDP's direct-link subcomponents into MDP controller node
>    from syscon node to avoid illegal clock usage.
> - Rewrite dt-binding in a JSON compatible subset of YAML
> - Fix a bit of macro argument precedence.
> 
> Changes since v3:
> - Rebase on v5.9-rc1.
> - modify code for review comment from Rob Herring, cancel multiple nodes using
>    same register base situation.
> - control IOMMU port through pm runtime get/put to DMA components' device.
> - SCP(VPU) driver revision.
> - stop queuing jobs(remove flush_workqueue()) after mdp_m2m_release().
> - add computation of plane address with data_offset.
> - fix scale ratio check issue.
> - add default v4l2_format setting.
> 
> Changes since v2:
> - modify code for review comment from Tomasz Figa & Alexandre Courbot
> - review comment from Rob Herring will offer code revision in v4, due to
>    it's related to device node modification, will need to modify code
>    architecture
> 
> Changes since v1:
> - modify code for CMDQ v3 API support
> - EC ipi cmd migration
> - fix compliance test fail item (m2m cmd with -f) due to there is two problem in runing all format(-f) cmd:
> 1. out of memory before test complete
>          Due to capture buffer mmap (refcount + 1) after reqbuf but seems
>          no corresponding munmap called before device close.
>          There are total 12XX items(formats) in format test and each format
>          alloc 8 capture/output buffers.
> 2. unceasingly captureBufs() (randomly)
>          Seems the break statement didn't catch the count == 0 situation:
>          In v4l2-test-buffers.cpp, function: captureBufs()
>                          ...
>                          count--;
>                          if (!node->is_m2m && !count)
>                                  break;
>          Log is as attachment
> 
> I will paste the test result with problem part in another e-mail
> 
> Hi,
> 
> This is the first version of RFC patch for Media Data Path 3 (MDP3),
> MDP3 is used for scaling and color format conversion.
> support using GCE to write register in critical time limitation.
> support V4L2 m2m device control.
> 
> Moudy Ho (3):
>    dt-binding: mt8183: add Mediatek MDP3 dt-bindings
>    dts: arm64: mt8183: add Mediatek MDP3 nodes
>    media: platform: mtk-mdp3: add Mediatek MDP3 driver
> 

Hello Moudy,

In next-20220315, all MediaTek media drivers have been moved in a different
subdirectory. Now they are all in drivers/media/platform/mediatek/.

Can you please rebase this series, placing the MDP3 driver in
drivers/media/platform/mediatek/mdp3 ?

While at it, also please include the additional documentation and address the
rdma/rsz (etc) suggestions about using one compatible and a instance-id property
to express the multiple instances (remember: "mediatek,mt8183-mdp3-rdma",
mediatek,instance-id = <0>).... and all the other constructive suggestions that
were given during the extensive review of patch 1/3 of this series.

Looking forward to see a new version!

Regards,
Angelo

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel