Hi Paul On 14/08/2021 21:56, Laurent Pinchart wrote: > Hi Daniel, > > On Fri, Aug 13, 2021 at 10:45:48AM +0100, Daniel Scally wrote: >> On 13/08/2021 04:05, Laurent Pinchart wrote: >>> On Tue, Aug 10, 2021 at 11:07:22PM +0100, Daniel Scally wrote: >>>> On 10/08/2021 15:29, Sakari Ailus wrote: >>>>> On Mon, Aug 09, 2021 at 11:58:41PM +0100, Daniel Scally wrote: >>>>>> @@ -2542,6 +2544,13 @@ static int ov8865_ctrls_init(struct ov8865_sensor *sensor) >>>>>> 0, 0, ov8865_test_pattern_menu); >>>>>> >>>>>> /* Blanking */ >>>>>> + hblank = mode->hts < mode->output_size_x ? 0 : mode->hts - mode->output_size_x; >>>>> Is the result in relation with the analogue crop size? Based on the above >>>>> it wouldn't seem like that. >>>> This was a weird one actually. My understanding was that HTS should >>>> always be >= the horizontal crop plus hblank...but that doesn't hold >>>> true for some of this driver's modes and nor does it hold true when >>>> running the camera in Windows (I checked the registers whilst running >>>> it). So I went with setting hblank to 0 if the mode's HTS exceeded the >>>> horizontal crop as the only way I could see to reconcile that. >>> There's something very fishy here, HTS is, by definition, equal to the >>> analog crop width plus the horizontal blanking. I suspect that the >>> values in ov8865_modes are wrong. >> I thought that initially too but confirming that the same thing happened >> running windows switched me into "you're probably wrong" mode. If we're >> confident that the HTS is likely wrong though I can add an extra patch >> to bring those into lining with that definition. > I think it's worth investigating this. The hblank computed here is > clearly incorrect, and would thus be useless for all practical purposes. > As usual with OmniVision, the datasheet is also quite useless. > > Paul, do you have any information about this ? A gentle ping on this...I played around setting HTS / VTS values whilst the camera was running windows; and it behaves as you'd expect it to (raising/lowering the frame rate), so as far as I can tell the sensor itself isn't doing anything unusual... >>>>>> + ctrls->hblank = v4l2_ctrl_new_std(handler, ops, V4L2_CID_HBLANK, hblank, >>>>>> + hblank, 1, hblank); >>>>>> + >>>>>> + if (ctrls->hblank) >>>>>> + ctrls->hblank->flags |= V4L2_CTRL_FLAG_READ_ONLY; >>>>>> + >>>>>> vblank_max = OV8865_TIMING_MAX_VTS - mode->output_size_y; >>>>>> vblank_def = mode->vts - mode->output_size_y; >>>>>> ctrls->vblank = v4l2_ctrl_new_std(handler, ops, V4L2_CID_VBLANK,
