Re: build failure of next-20220906 due to 396369d67549 ("drm: vkms: Add support to the RGB565 format")

[Daniel Vetter]

On Tue, Sep 06, 2022 at 08:35:49PM -0300, Igor Matheus Andrade Torrente wrote:
> On 9/6/22 18:26, Sudip Mukherjee wrote:
> > On Tue, Sep 6, 2022 at 4:59 PM Sudip Mukherjee (Codethink)
> > <sudipm.mukherjee@xxxxxxxxx> wrote:
> > > 
> > > Hi All,
> > > 
> > > The builds of next-20220906 fails for mips, xtensa and arm allmodconfig.
> > > 
> > > The errors in mips and xtensa are:
> > > 
> > > ERROR: modpost: "__divdi3" [drivers/gpu/drm/vkms/vkms.ko] undefined!
> > > ERROR: modpost: "__udivdi3" [drivers/gpu/drm/vkms/vkms.ko] undefined!
> > > 
> > > The error in arm is:
> > > 
> > > ERROR: modpost: "__aeabi_uldivmod" [drivers/gpu/drm/vkms/vkms.ko] undefined!
> > > ERROR: modpost: "__aeabi_ldivmod" [drivers/gpu/drm/vkms/vkms.ko] undefined!
> > > 
> > > 
> > > Trying to do a git bisect to find out the offending commit.
> > 
> > git bisect points to 396369d67549 ("drm: vkms: Add support to the
> > RGB565 format")
> 
> Are these architectures incapable of doing 64bits int division?

Yeah 32bit archs in general can't do that, and you have to use the right
macros because otherwise gcc falls back to its own built-ins, and those
don't exist in the kernel since the kernel isn't (cannot!) linked against
any userspace library.

For pretty much this reasons it's really good to build test against 32bit
x86, or probably more relevant these days, 32bit arm.

Cheers, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch