Re: hooking Sony 47MPixel sensor to NXP imx8m-mini MIPI CSI2

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Petko,

On Wed, Jun 29, 2022 at 11:45:56AM +0300, Petko Manolov wrote:
> On 22-06-29 11:28:26, Laurent Pinchart wrote:
> > On Wed, Jun 29, 2022 at 10:55:43AM +0300, Petko Manolov wrote:
> > > 	Hi guys,
> > > 
> > > First, i'd like to apologize for the noise and sending this email directly.
> > 
> > Your support request is directed to the linux-media mailing list, CC'ing
> > appropriate developers is a good practice to make sure we notice, there's
> > nothing to apologize for.
> 
> Well, it could still be annoyance.
> 
> > > I am trying to hook Sony IMX492 sensor to v5.10 imx8m-mini based platform.
> > > The capture driver does not support Bayer GBRG 10 and 12bit pixel format, so
> > > i had to add these.  I also stumbled upon a bug in mx6s_capture.c which, i
> > > assume, is completely out of interest here as it's related to a vendor tree.
> > 
> > I'm surprised you stumbled upon one bug only :-) I always advise to stay away
> > from BSP code as much as possible.
> 
> I am certain there is a bunch of these, but this particular one was hard not to
> notice immediately, being NULL pointer dereference. :)
> 
> > > I got as far as sensor is registering with the capture mipi csi driver and,
> > > if HW is properly set up on both sides, should be streaming.  While my
> > > confidence is pretty high about the IMX492 setup i can't say the same about
> > > the recipient part.
> > > 
> > > So here is my question: how stable are NXP's MIPI CSI2 and CSI2 bridge
> > > drivers in v5.18?  Would it be better to switch to this version instead of
> > > hacking madly at the v5.10 vendor code?
> > 
> > I consider them more stable than the vendor code. If there are issues there,
> > we can help fixing them.
> 
> I might be mistaken, but at least in terms of media drivers code in NXP's 5.18
> seems either pretty close to mainline or maybe the same.  What's your advice,
> shall i use their (5.18.x+fslc) branch or take only the device-tree, if needed?

For an i.MX8MM I'd try running the mainline kernel directly. If that
doesn't work, I would backport the camera drivers from mainline to the
v5.15 NXP kernel. As far as I know, the v5.18 branch isn't an official
BSP release (I'm actually not sure what it's for).

> > > I've already published the code here.  Once the development is done the
> > > intent is to mainline Sony IMX492 sensor driver.
> > 
> > That's great ! I'm looking forward to that.
> 
> Sure, that's the goal here.

-- 
Regards,

Laurent Pinchart