media: i2c: ov5647: handle V4L2_CID_LINK_FREQ in s_ctrl
PATCH 1/1 media: i2c: ov5647: handle V4L2_CID_LINK_FREQ in s_ctrl
Handle V4L2_CID_LINK_FREQ in ov5647_s_ctrl().
Currently this control is defined but not handled in s_ctrl(), so V4L2 falls back to estimating link frequency from pixel rate and prints warning like:
v4l2_get_link_freq: Link frequency estimated using pixel rate:
result might be inaccurate
v4l2_get_link_freq: Consider implementing support for V4L2_CID_LINK_FREQ in the transmitter driver
Handle it as no-op since link frequency is fixed per mode and not meant to be changed at runtime.
Avoid these warnings when control is queried.
--- drivers/media/i2c/ov5647.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/media/i2c/ov5647.c +3 -0
@@ -999,6 +999,9 @@ static int ov5647_s_ctrl(struct v4l2_ctrl *ctrl) ret =cci_write (sensor->regmap,OV5647_REG_HTS , sensor->mode->format.width + ctrl->val, &ret);break ; +case V4L2_CID_LINK_FREQ : + ret =0 ; +break ;case V4L2_CID_TEST_PATTERN : ret =cci_write (sensor->regmap,OV5647_REG_ISPCTRL3D , ov5647_test_pattern_val[ctrl->val],NULL );
Signature · 1 line
2.34.1
Sakari Ailus
maintainerHi Suraj,
Thanks for the patch.
Quoted 26 lines · Suraj Sonawane
Handle
V4L2_CID_LINK_FREQinov5647_s_ctrl().Currently this control is defined but not handled in
s_ctrl(), so V4L2 falls back to estimating link frequency from pixel rate and prints warning like:
v4l2_get_link_freq: Link frequency estimated using pixel rate:
result might be inaccurate
v4l2_get_link_freq: Consider implementing support forV4L2_CID_LINK_FREQin the transmitter driverHandle it as no-op since link frequency is fixed per mode and not meant to be changed at runtime.
Avoid these warnings when control is queried.
Signed-off-by Suraj Sonawane <[email protected]>--- drivers/media/i2c/ov5647.c | 3 +++ 1 file changed, 3 insertions(+)drivers/media/i2c/ov5647.c +0
-0
drivers/media/i2c/ov5647.c static int ov5647_s_ctrl(struct v4l2_ctrl *ctrl) · line 1007
Quoted 9 lines
drivers/media/i2c/ov5647.c +3
-0@@ -999,6 +999,9 @@ static int ov5647_s_ctrl(struct v4l2_ctrl *ctrl) ret =cci_write (sensor->regmap,OV5647_REG_HTS , sensor->mode->format.width + ctrl->val, &ret);break ; +case V4L2_CID_LINK_FREQ : + ret =0 ; +break ;case V4L2_CID_TEST_PATTERN : ret =cci_write (sensor->regmap,OV5647_REG_ISPCTRL3D , ov5647_test_pattern_val[ctrl->val],NULL );
Can you instead turn the first if () in the function into a switch and handle this there, too, returning zero on the LINK_FREQ control?
Signature · 3 lines
Regards, Sakari Ailus
Dave Stevenson
maintainerHi Suraj
Quoted 7 lines · Suraj Sonawane
Handle
V4L2_CID_LINK_FREQinov5647_s_ctrl().Currently this control is defined but not handled in
s_ctrl(), so V4L2 falls back to estimating link frequency from pixel rate and prints warning like:
Quoted 7 lines
v4l2_get_link_freq: Link frequency estimated using pixel rate:
result might be inaccurate
v4l2_get_link_freq: Consider implementing support forV4L2_CID_LINK_FREQin the transmitter driverHandle it as no-op since link frequency is fixed per mode and not meant to be changed at runtime.
I'm confused by this description compared to the patch.
v4l2_get_link_freq searches for the V4L2_CID_LINK_FREQ control, and if found then it calls g_ctrl (not s_ctrl).
If it can't find the control then it searches for V4L2_CID_PIXEL_RATE and will log the error message quoted.
The control is registered by the ov5647 driver, therefore it should never go into that second clause, so how have you got that error message logged?
AFAIK no part of that code path will result in a call to ov5647_s_ctrl that you're patching.
I've just run with the ov5647 driver on a Pi5 (which uses v4l2_get_link_freq) running 7.0.0-rc5, and I can't get an error logged. v4l2_get_link_freq finds V4L2_CID_LINK_FREQ and uses the value it reports.
You are right that ov5647_s_ctrl doesn't handle V4L2_CID_LINK_FREQ, which could mean that an error is returned from the __v4l2_ctrl_s_ctrl calls from within the driver to change the link frequency and lead to the dev_info in the driver being logged iff the sensor was powered up at the time (otherwise pm_runtime_get_if_in_use will fail). Generally the sensor won't be powered on as the pad format is set before enable_streams, and it shouldn't be possible to change it whilst streaming.
However, as I understand it, the current preferred way to handle this case of read only controls where the value is changed by the driver is to pass NULL as the ops for the ctrl when registering. That is already the case looking at c6e115144b50 ("media: i2c: ov5647: Add V4L2_CID_LINK_FREQUENCY control"). So how are you managing to get ov5647_s_ctrl called for control V4L2_CID_LINK_FREQ at all when it has no s_ctrl op?
Have I totally missed something here?
Dave
Quoted 24 lines
Avoid these warnings when control is queried.
Signed-off-by Suraj Sonawane <[email protected]>--- drivers/media/i2c/ov5647.c | 3 +++ 1 file changed, 3 insertions(+)drivers/media/i2c/ov5647.c +3
-1@@ -999,6 +999,9 @@ static int ov5647_s_ctrl(struct v4l2_ctrl *ctrl) ret =cci_write (sensor->regmap,OV5647_REG_HTS , sensor->mode->format.width + ctrl->val, &ret);break ; +case V4L2_CID_LINK_FREQ : + ret =0 ; +break ;case V4L2_CID_TEST_PATTERN : ret =cci_write (sensor->regmap,OV5647_REG_ISPCTRL3D , ov5647_test_pattern_val[ctrl->val],NULL ); --2.34.1
Hi Dave
Quoted 47 lines · Dave Stevenson
Hi Suraj
Quoted 7 lines · Suraj Sonawane
Handle
V4L2_CID_LINK_FREQinov5647_s_ctrl().Currently this control is defined but not handled in
s_ctrl(), so V4L2 falls back to estimating link frequency from pixel rate and prints warning like:Quoted 7 lines
v4l2_get_link_freq: Link frequency estimated using pixel rate:
result might be inaccurate
v4l2_get_link_freq: Consider implementing support forV4L2_CID_LINK_FREQin the transmitter driverHandle it as no-op since link frequency is fixed per mode and not meant to be changed at runtime.
I'm confused by this description compared to the patch.
v4l2_get_link_freqsearches for theV4L2_CID_LINK_FREQcontrol, and if found then it callsg_ctrl(nots_ctrl).
If it can't find the control then it searches forV4L2_CID_PIXEL_RATEand will log the error message quoted.The control is registered by the ov5647 driver, therefore it should never go into that second clause, so how have you got that error message logged?
AFAIK no part of that code path will result in a call to
ov5647_s_ctrlthat you're patching.
I've just run with the ov5647 driver on a Pi5 (which usesv4l2_get_link_freq) running 7.0.0-rc5, and I can't get an error logged.v4l2_get_link_freqfindsV4L2_CID_LINK_FREQand uses the value it reports.You are right that
ov5647_s_ctrldoesn't handleV4L2_CID_LINK_FREQ, which could mean that an error is returned from the__v4l2_ctrl_s_ctrlcalls from within the driver to change the link frequency and lead to thedev_infoin the driver being logged iff the sensor was powered up at the time (otherwisepm_runtime_get_if_in_usewill fail). Generally the sensor won't be powered on as the pad format is set beforeenable_streams, and it shouldn't be possible to change it whilst streaming.
Quoted 5 lines
However, as I understand it, the current preferred way to handle this case of read only controls where the value is changed by the driver is to pass NULL as the ops for the ctrl when registering. That is already the case looking at
c6e115144b50("media: i2c: ov5647: AddV4L2_CID_LINK_FREQUENCYcontrol"). So how are you managing to get
I was about to mention the same. Setting the ctrl ops to NULL is preferred for controls not handled by the driver.
I'm surprised I didn't see any merged patch that address the many faulty i2c drivers we have which manually handle the read-only controls in their s_ctrl implementation!
We should WARN if a Read-Only control is registered with a valid ops pointer maybe
Quoted 31 lines
ov5647_s_ctrlcalled for controlV4L2_CID_LINK_FREQat all when it has nos_ctrlop?Have I totally missed something here?
Dave
Quoted 24 lines
Avoid these warnings when control is queried.
Signed-off-by Suraj Sonawane <[email protected]>--- drivers/media/i2c/ov5647.c | 3 +++ 1 file changed, 3 insertions(+)drivers/media/i2c/ov5647.c +3
-1@@ -999,6 +999,9 @@ static int ov5647_s_ctrl(struct v4l2_ctrl *ctrl) ret =cci_write (sensor->regmap,OV5647_REG_HTS , sensor->mode->format.width + ctrl->val, &ret);break ; +case V4L2_CID_LINK_FREQ : + ret =0 ; +break ;case V4L2_CID_TEST_PATTERN : ret =cci_write (sensor->regmap,OV5647_REG_ISPCTRL3D , ov5647_test_pattern_val[ctrl->val],NULL ); --2.34.1
Sakari Ailus
maintainerHi Jacopo,
Quoted 50 lines · Jacopo Mondi
Hi Dave
Quoted 47 lines · Dave Stevenson
Hi Suraj
Quoted 7 lines · Suraj Sonawane
Handle
V4L2_CID_LINK_FREQinov5647_s_ctrl().Currently this control is defined but not handled in
s_ctrl(), so V4L2 falls back to estimating link frequency from pixel rate and prints warning like:Quoted 7 lines
v4l2_get_link_freq: Link frequency estimated using pixel rate:
result might be inaccurate
v4l2_get_link_freq: Consider implementing support forV4L2_CID_LINK_FREQin the transmitter driverHandle it as no-op since link frequency is fixed per mode and not meant to be changed at runtime.
I'm confused by this description compared to the patch.
v4l2_get_link_freqsearches for theV4L2_CID_LINK_FREQcontrol, and if found then it callsg_ctrl(nots_ctrl).
If it can't find the control then it searches forV4L2_CID_PIXEL_RATEand will log the error message quoted.The control is registered by the ov5647 driver, therefore it should never go into that second clause, so how have you got that error message logged?
AFAIK no part of that code path will result in a call to
ov5647_s_ctrlthat you're patching.
I've just run with the ov5647 driver on a Pi5 (which usesv4l2_get_link_freq) running 7.0.0-rc5, and I can't get an error logged.v4l2_get_link_freqfindsV4L2_CID_LINK_FREQand uses the value it reports.You are right that
ov5647_s_ctrldoesn't handleV4L2_CID_LINK_FREQ, which could mean that an error is returned from the__v4l2_ctrl_s_ctrlcalls from within the driver to change the link frequency and lead to thedev_infoin the driver being logged iff the sensor was powered up at the time (otherwisepm_runtime_get_if_in_usewill fail). Generally the sensor won't be powered on as the pad format is set beforeenable_streams, and it shouldn't be possible to change it whilst streaming.
Quoted 8 lines
Quoted 5 lines
However, as I understand it, the current preferred way to handle this case of read only controls where the value is changed by the driver is to pass NULL as the ops for the ctrl when registering. That is already the case looking at
c6e115144b50("media: i2c: ov5647: AddV4L2_CID_LINK_FREQUENCYcontrol"). So how are you managing to getI was about to mention the same. Setting the ctrl ops to NULL is preferred for controls not handled by the driver.
Yes, indeed that's the other option. It makes sense in this case.
Quoted 7 lines
I'm surprised I didn't see any merged patch that address the many faulty i2c drivers we have which manually handle the read-only controls in their
s_ctrlimplementation!We should WARN if a Read-Only control is registered with a valid ops pointer maybe
There may still be side effects from the value change the driver should handle. I guess it could take place through other means, too, but I think I'd keep it as-is.
Signature · 3 lines
Regards, Sakari Ailus