Re: atomisp kernel driver(s)

[Mauro Carvalho Chehab <mchehab+huawei@kernel.org>]

Em Thu, 30 Apr 2020 17:09:48 +0200
Patrik Gfeller <patrik.gfeller@gmail.com> escreveu:

> Am 4/30/2020 um 12:55 PM schrieb Mauro Carvalho Chehab:
> > Em Thu, 30 Apr 2020 09:56:53 +0200
> > Patrik Gfeller<patrik.gfeller@gmail.com>  escreveu:
> >  
> >> For my first test tried to re-compile to module, without the whole
> >> kernel. That was a mistake, as I mixed something up, probably it loaded
> >> an old version of the module ... to be on the save side the steps I used
> >> this time (in case we see something unexpected and for my later reference):
> >>
> >> $ git log --oneline
> >> 4c922df10252 (HEAD -> atomisp, origin/atomisp) media: atomisp2: get rid
> >> of ia_css_sc_param.h version dependency
> >> ...
> >>
> >> $ make -j4 clean
> >> $ make -j4
> >> $ sudo make modules_install INSTALL_MOD_STRIP=1
> >> $ sudo make install  
> > Please try to build from this branch:
> >
> > 	https://git.linuxtv.org/mchehab/experimental.git/log/?h=atomisp_v2
> >
> > You'll need to setup a new config var there. So, please run this before
> > make clean. So, for building it, you will do:
> >
> > 	$ ./scripts/config -e CONFIG_VIDEO_ATOMISP_ISP2401 && make -j modules_prepare
> > 	$ make -j4 clean && make -j4
> > 	$ sudo make modules_install INSTALL_MOD_STRIP=1 && sudo make install
> >
> > This won't change the regulator detection, but it should hopefully use
> > the ISP2401-specific code, with seems to be needed for your device.  
> 
> I've updated to the latest source (git checkout atomisp_v2 && git pull) 
> and compiled using the instructions above. Compilation worked well, but 
> the linker had some problems (full log attached):
> 
> ...
> ld: 
> drivers/staging/media/atomisp/pci/css_2401_system/hive_isp_css_2401_system_generated/ia_css_isp_states.o:(.data+0x0): 
> multiple definition of `ia_css_kernel_init_state'; 
> drivers/staging/media/atomisp/pci/css_2401_csi2p_system/hive_isp_css_2401_system_csi2p_generated/ia_css_isp_states.o:(.data+0x0): 
> first defined here
> ...

Ok. That's because there are two "hive" variants. the building system
should use either one of them (but not both at the same time).

I didn't get the error before because I was just building a module
(that speeds up the development). Such errors only happen on a full 
build.

Fixed.

As I did a git rebase (in order to have something that we could
upstream later), you'll need to use this procedure to update:

	$ git remote update
	$ git reset --hard origin/atomisp_v2

There's no need to clean your last build. Just run:

	$ make -j4 

And it should build fine this time.

> 
> Not sure if I can help with that. Sounds like we have to remove 
> definitions - which I might be able to do. But I would need to know 
> where the right place is to keep the definitions.

> If a code generator is 
> involved (one of the paths looks like it) it will be more difficult for 
> me.

Intel seems to use some code generator internally. Basically, for each
specific device, it can remove/add/change things. Don't mind about that.

For us, we're seeing just the generated code and working to simplify it
and making it more generic.

> But with some hints I'm of course willing to give it a shot. Please 
> give me an example of a definition) and a hint in case we deal with 
> generated code.
> 
> > Ah, when replying, could you please use an emailer that won't be breaking
> > long lines? Your emailer is currently breaking lines longer than 80 columns,
> > with is quite annoying when checking log results ;-)  
> I changed the configuration of my mail client; a test message looked ok. 
> Let me know if the problem persists.

Yeah, is is fine now. Thanks!

Thanks,
Mauro