| Message ID | 20240424235741.17093-1-laurent.pinchart@ideasonboard.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | media: vimc improvements | expand |
Hello, On Thu, Apr 25, 2024 at 02:57:32AM +0300, Laurent Pinchart wrote: > Hello, > > This patch series cleans up and improves the vimc driver, with the end > goal of converting it to the V4L2 subdev active state API. The goal of > this exercise is to make the API used by a virtual test driver, to > increase test coverage. > > The series starts with 4 random cleanups, to avoid unnecessary > iterations (1/9), constify structures (2/9 and 3/9) and rename a > weirdly-named enum (4/9). Patch 5/9 then centralizes the subdev > internal_ops initialization to prepare for the switch to the active > state API. The remaining patches (6/9 to 9/9) convert the vimc entities > to the new API one by one. > > The result has been tested using the libcamera unit tests, which make > extensive use of the vimc driver, as well as with v4l2-compliance. The > latter reports 4 errors, but they occur already with the latest stage > master branch. Any comment about this patch series ? Shuah, do you plan to give it a look ? If you don't have much time for vimc these days that's OK, I can work with Kieran to get the patches reviewed. On a related note, I would like to modify the MC pipeline of the vimc driver to make it look more like a real ISP. This would help with unit testing and compliance testing in libcamera. I briefly discussed this with Hans on IRC who had no objection. If anyone has any issue or concern with this plan, I would appreciate if you could let me know. > Laurent Pinchart (9): > media: vimc: Don't iterate over single pad > media: vimc: Constify vimc_ent_type structures > media: vimc: Constify the ent_config array > media: vimc: scaler: Rename vic_sca_pad enum to vimc_scaler_pad > media: vimc: Centralize subdev internal_ops initialization > media: vimc: Initialize subdev active state > media: vimc: sensor: Use subdev active state > media: vimc: debayer: Use subdev active state > media: vimc: scaler: Use subdev active state > > .../media/test-drivers/vimc/vimc-capture.c | 2 +- > drivers/media/test-drivers/vimc/vimc-common.c | 25 ++- > drivers/media/test-drivers/vimc/vimc-common.h | 14 +- > drivers/media/test-drivers/vimc/vimc-core.c | 2 +- > .../media/test-drivers/vimc/vimc-debayer.c | 197 ++++++++---------- > drivers/media/test-drivers/vimc/vimc-lens.c | 5 +- > drivers/media/test-drivers/vimc/vimc-scaler.c | 134 +++++------- > drivers/media/test-drivers/vimc/vimc-sensor.c | 125 +++++------ > 8 files changed, 236 insertions(+), 268 deletions(-) > > > base-commit: e42a204f0519a2540f1507ac2798be2aeaa76bee
On 5/5/24 14:36, Laurent Pinchart wrote: > Hello, > > On Thu, Apr 25, 2024 at 02:57:32AM +0300, Laurent Pinchart wrote: >> Hello, >> >> This patch series cleans up and improves the vimc driver, with the end >> goal of converting it to the V4L2 subdev active state API. The goal of >> this exercise is to make the API used by a virtual test driver, to >> increase test coverage. >> >> The series starts with 4 random cleanups, to avoid unnecessary >> iterations (1/9), constify structures (2/9 and 3/9) and rename a >> weirdly-named enum (4/9). Patch 5/9 then centralizes the subdev >> internal_ops initialization to prepare for the switch to the active >> state API. The remaining patches (6/9 to 9/9) convert the vimc entities >> to the new API one by one. >> >> The result has been tested using the libcamera unit tests, which make >> extensive use of the vimc driver, as well as with v4l2-compliance. The >> latter reports 4 errors, but they occur already with the latest stage >> master branch. > > Any comment about this patch series ? Shuah, do you plan to give it a > look ? If you don't have much time for vimc these days that's OK, I can > work with Kieran to get the patches reviewed. > Sorry for the delay. I will take a look and apply them for the next release. > On a related note, I would like to modify the MC pipeline of the vimc > driver to make it look more like a real ISP. This would help with unit > testing and compliance testing in libcamera. I briefly discussed this > with Hans on IRC who had no objection. If anyone has any issue or > concern with this plan, I would appreciate if you could let me know. > I am good with the plan. thanks, -- Shuah
On 4/24/24 17:57, Laurent Pinchart wrote: > Hello, > > This patch series cleans up and improves the vimc driver, with the end > goal of converting it to the V4L2 subdev active state API. The goal of > this exercise is to make the API used by a virtual test driver, to > increase test coverage. > > The series starts with 4 random cleanups, to avoid unnecessary > iterations (1/9), constify structures (2/9 and 3/9) and rename a > weirdly-named enum (4/9). Patch 5/9 then centralizes the subdev > internal_ops initialization to prepare for the switch to the active > state API. The remaining patches (6/9 to 9/9) convert the vimc entities > to the new API one by one. > > The result has been tested using the libcamera unit tests, which make > extensive use of the vimc driver, as well as with v4l2-compliance. The > latter reports 4 errors, but they occur already with the latest stage > master branch. > > Laurent Pinchart (9): > media: vimc: Don't iterate over single pad Please drop the first patch. > media: vimc: Constify vimc_ent_type structures > media: vimc: Constify the ent_config array > media: vimc: scaler: Rename vic_sca_pad enum to vimc_scaler_pad > media: vimc: Centralize subdev internal_ops initialization > media: vimc: Initialize subdev active state > media: vimc: sensor: Use subdev active state > media: vimc: debayer: Use subdev active state > media: vimc: scaler: Use subdev active state > > .../media/test-drivers/vimc/vimc-capture.c | 2 +- > drivers/media/test-drivers/vimc/vimc-common.c | 25 ++- > drivers/media/test-drivers/vimc/vimc-common.h | 14 +- > drivers/media/test-drivers/vimc/vimc-core.c | 2 +- > .../media/test-drivers/vimc/vimc-debayer.c | 197 ++++++++---------- > drivers/media/test-drivers/vimc/vimc-lens.c | 5 +- > drivers/media/test-drivers/vimc/vimc-scaler.c | 134 +++++------- > drivers/media/test-drivers/vimc/vimc-sensor.c | 125 +++++------ > 8 files changed, 236 insertions(+), 268 deletions(-) > > > base-commit: e42a204f0519a2540f1507ac2798be2aeaa76bee Others look good to me. I can apply them all for 6.11-rc1. Please send v2 without the first patch if others depend on it. thanks, -- Shuah
On 4/24/24 17:57, Laurent Pinchart wrote: > Hello, > > This patch series cleans up and improves the vimc driver, with the end > goal of converting it to the V4L2 subdev active state API. The goal of > this exercise is to make the API used by a virtual test driver, to > increase test coverage. > > The series starts with 4 random cleanups, to avoid unnecessary > iterations (1/9), constify structures (2/9 and 3/9) and rename a > weirdly-named enum (4/9). Patch 5/9 then centralizes the subdev > internal_ops initialization to prepare for the switch to the active > state API. The remaining patches (6/9 to 9/9) convert the vimc entities > to the new API one by one. > > The result has been tested using the libcamera unit tests, which make > extensive use of the vimc driver, as well as with v4l2-compliance. The > latter reports 4 errors, but they occur already with the latest stage > master branch. > > Laurent Pinchart (9): > media: vimc: Don't iterate over single pad > media: vimc: Constify vimc_ent_type structures > media: vimc: Constify the ent_config array > media: vimc: scaler: Rename vic_sca_pad enum to vimc_scaler_pad > media: vimc: Centralize subdev internal_ops initialization > media: vimc: Initialize subdev active state > media: vimc: sensor: Use subdev active state > media: vimc: debayer: Use subdev active state > media: vimc: scaler: Use subdev active state > Thank you. Applied the series to https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux.git/log/?h=vimc I will send pull request to Mauro in a week. thanks, -- Shuah
On Thu, Jun 20, 2024 at 09:59:31AM -0600, Shuah Khan wrote: > On 4/24/24 17:57, Laurent Pinchart wrote: > > Hello, > > > > This patch series cleans up and improves the vimc driver, with the end > > goal of converting it to the V4L2 subdev active state API. The goal of > > this exercise is to make the API used by a virtual test driver, to > > increase test coverage. > > > > The series starts with 4 random cleanups, to avoid unnecessary > > iterations (1/9), constify structures (2/9 and 3/9) and rename a > > weirdly-named enum (4/9). Patch 5/9 then centralizes the subdev > > internal_ops initialization to prepare for the switch to the active > > state API. The remaining patches (6/9 to 9/9) convert the vimc entities > > to the new API one by one. > > > > The result has been tested using the libcamera unit tests, which make > > extensive use of the vimc driver, as well as with v4l2-compliance. The > > latter reports 4 errors, but they occur already with the latest stage > > master branch. > > > > Laurent Pinchart (9): > > media: vimc: Don't iterate over single pad > > media: vimc: Constify vimc_ent_type structures > > media: vimc: Constify the ent_config array > > media: vimc: scaler: Rename vic_sca_pad enum to vimc_scaler_pad > > media: vimc: Centralize subdev internal_ops initialization > > media: vimc: Initialize subdev active state > > media: vimc: sensor: Use subdev active state > > media: vimc: debayer: Use subdev active state > > media: vimc: scaler: Use subdev active state > > > > Thank you. Applied the series to > > https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux.git/log/?h=vimc > > I will send pull request to Mauro in a week. Thank you Shuah. Much appreciated.