Re: Problem with compilation prevents me from adding a new device

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

Hi Dominik,

Em Tue, 21 May 2019 22:20:15 +0200
Dominik Danelski <dominik@danelski.pl> escreveu:

> Hello,
> 
> I found this 
> guide:https://www.onetransistor.eu/2017/05/magix-usb-videowandler-2-linux.html 
> which explains how to get Magix Videowandler 2 as an audio-video 
> source. It is achieved by downloading drivers from LinuxTV project, 
> slightly modifying them and installing. I wanted to incorporate changes 
> from the guide to the upstream code, but first I wanted to test 
> everything once again. I tried to follow the described method of 
> compiling drivers (before making any changes) and I got stuck with a 
> following output:
>   CC [M]  /tmp/media_build/v4l/videobuf-core.o
>   CC [M]  /tmp/media_build/v4l/videobuf-dma-sg.o
> /tmp/media_build/v4l/videobuf-dma-sg.c: In function 
> 'videobuf_dma_init_user_locked':
> /tmp/media_build/v4l/videobuf-dma-sg.c:190:17: error: 'FOLL_LONGTERM' 
> undeclared (first use in this function); did you mean 'FOLL_FORCE'?
>          flags | FOLL_LONGTERM, dma->pages, NULL);
>                  ^~~~~~~~~~~~~
>                  FOLL_FORCE

The media_build basically downloads a tarball with the media drivers
from the media_tree.git tree and applies a set of backport patches.

Today, the media_tree.git received an update with Kernel 5.2-rc1
development. It is likely that the media_build tree will require 
some changes after that, as some upstream Kernel APIs may have
changed.

If you can, it would be better if you could just use the git tree
and build a new kernel from it. Otherwise, you may need to wait
a few days for people to add the needed backports from changes that
happened upstream and got reflected at media_build.

For em28xx, you can do whatever it takes for videobuf-dma-sg.c
to build (I would just remove " | FOLL_LONGTERM" from the above),
as em28xx doesn't use videobuf-dma-sg (it uses, instead,
videobuf2-vmalloc.h). This legacy core driver is used only on a couple
of PCI drivers.

I suspect, however, that you may find other breakages over there.

> /tmp/media_build/v4l/videobuf-dma-sg.c:190:17: note: each undeclared 
> identifier is reported only once for each function it appears in
> 
> What could have caused it? I suppose that the code already in the 
> repository has been checked and such thing should not happen if not by 
> my mistake, but I did not change anything after cloning git repo.
> 
> Regards
> Dominik Danelski
> 



Thanks,
Mauro