Hi Tomasz, -----Original Message----- From: Tomasz Figa [mailto:tfiga@xxxxxxxxxxxx] Sent: Thursday, May 10, 2018 3:04 PM To: Zheng, Jian Xu <jian.xu.zheng@xxxxxxxxx> Cc: Chen, JasonX Z <jasonx.z.chen@xxxxxxxxx>; Yeh, Andy <andy.yeh@xxxxxxxxx>; Linux Media Mailing List <linux-media@xxxxxxxxxxxxxxx>; Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx>; Chiang, AlanX <alanx.chiang@xxxxxxxxx> Subject: Re: [PATCH v11] media: imx258: Add imx258 camera sensor driver On Thu, May 10, 2018 at 3:56 PM Zheng, Jian Xu <jian.xu.zheng@xxxxxxxxx> wrote: > Hi Tomasz, > > -----Original Message----- > > From: linux-media-owner@xxxxxxxxxxxxxxx [mailto:linux-media- > > owner@xxxxxxxxxxxxxxx] On Behalf Of Tomasz Figa > > Sent: Wednesday, May 9, 2018 6:05 PM > > To: Chen, JasonX Z <jasonx.z.chen@xxxxxxxxx> > > Cc: Yeh, Andy <andy.yeh@xxxxxxxxx>; Linux Media Mailing List <linux- > > media@xxxxxxxxxxxxxxx>; Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx>; Chiang, > > AlanX <alanx.chiang@xxxxxxxxx> > > Subject: Re: [PATCH v11] media: imx258: Add imx258 camera sensor > > driver > > > > Hi Jason, > > > > > IPU3 HAL has a handler to bind test_pattern mode. > > > The COLOR BAR MODE in HAL has been configured to 1 when APP > > > requests to > > output color bar image. > > > However Sony sensor's COLOR BAR MODE is designed as 2 in register table. > > (grey color bars as 1). > > > When HAL sends handler to driver to switch test pattern mode (to > > > COLOR > > BAR - val: 1), it will be grey color, since driver still set TEST_PATTERN_MODE > > reg value to 1, those it is not what we expected. > > > > > That is why we have to make an array with index to arrange the > > > order of > > the test pattern items, so driver will choose COLOR BAR correctly > > when HAL > > send test_pattern message (with 1). > > > The concept is the test_pattern_menu could be listed in driver per > > > real > > requirement, no matter how the sensor register is designed. > > > > > > V4L2 specification does not define any particular order of menu > > entries in > > V4L2_CID_TEST_PATTERN. The application should query the strings in > > the menu and determine the option it needs based on that. If it > > hardcodes particular index, it's a bug. > Is there any reason that there is no certain macro define for > different type of test pattern in v4l2? > So App will not depend on any strings where could be different on different sensor drivers. > Yes. Available patterns differ significantly between one sensor and another, so the menu positions are considered hardware-specific. In current design, application can still query the available patterns and select per their needs. Although the menu position doesn't exactly the same as the datasheet, it still complies with the V4L2 rule. Because it will not introduce bug for any application, we think it is a feasible approach. Best regards, Tomasz
