On Sat, Nov 21, 2020 at 04:23:05PM +0800, kernel test robot wrote:
> tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> head: 27bba9c532a8d21050b94224ffd310ad0058c353
> commit: 7b285f41f7376dc37e7fad1e803995fd39f42848 media: ipu3-cio2: Introduce
> date: 2 months ago
> config: arm64-randconfig-r031-20201121 (attached as .config)
> compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project
> reproduce (this is a W=1 build):
> chmod +x ~/bin/make.cross
> # install arm64 cross compiling tool for clang build
> # apt-get install binutils-aarch64-linux-gnu
> git remote add linus
> git fetch --no-tags linus master
> git checkout 7b285f41f7376dc37e7fad1e803995fd39f42848
> # save the attached .config to linux build tree
> COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=arm64
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp(a)intel.com>
> All warnings (new ones prefixed by >>):
> >> drivers/media/pci/intel/ipu3/ipu3-cio2.c:163:56: warning: implicit
conversion from 'unsigned long' to 'u16' (aka 'unsigned short')
changes value from 131072 to 0 [-Wconstant-conversion]
> entry.second_entry.num_of_pages = CIO2_LOP_ENTRIES * CIO2_MAX_LOPS;
> ~ ~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~
> 1 warning generated.
Okay, now we have an interesting case. The IP is quite unlikely be used on
ARM64, but my patches made the clear picture about use of PAGE_SIZE here.
So, I see at least the following options to mitigate the above, i.e.:
1/ reduce driver scope to X86
2/ fix the variables to be wider type to be able to hold PAGE_SIZE > 4k
3/ switch to custom PAGE_SIZE / _SHIFT / _MASK and accompanying macros
And I still consider 3/ is silly move because as we see the driver was
never assumed to work with big page sizes (besides unsigned short type
here, PAGE_SHIFT and PAGE_MASK in the original code was as is and on ARM64
they compiled to 0 values w/o warnings, effectively make the driver
improperly functioning anyway).
Apologies for the late answer.
I think I'd favour the first option. It's not really useful to be able to
compile this elsewhere; as such the driver doesn't do anything special that
would make it prone to breakage through changes elsewhere.
Would you like to send a patch? :-)