On Fri, Jun 27, 2025 at 08:58:11AM +0200, Benjamin Berg wrote:
> On Fri, 2025-06-27 at 06:20 +0200, Thomas Weißschuh wrote:
> > On Thu, Jun 26, 2025 at 08:11:17PM +0200, Benjamin Berg wrote:
> > > I ran into two minor issues trying out the patches, see inline.
> >
> > Thanks for testing the series.
> >
> > > On Thu, 2025-06-26 at 08:10 +0200, Thomas Weißschuh wrote:
> > > > Enable running UAPI tests as part of kunit.
> > > > The selftests are embedded into the kernel image and their output is
> > > > forwarded to kunit for unified reporting.
> > > >
> > > > The implementation reuses parts of usermode drivers and usermode
> > > > helpers. However these frameworks are not used directly as they make it
> > > > impossible to retrieve a thread's exit code.
> > > >
> > > > Signed-off-by: Thomas Weißschuh <thomas.weisssc...@linutronix.de>
> > > >
> > > > [SNIP]
> > > > +/**
> > > > + * KUNIT_UAPI_EMBED_BLOB() - Embed another build artifact into the
> > > > kernel
> > > > + * @_name: The name of symbol under which the artifact is embedded.
> > > > + * @_path: Path to the artifact on disk.
> > > > + *
> > > > + * Embeds a build artifact like a userspace executable into the kernel
> > > > or current module.
> > > > + * The build artifact is read from disk and needs to be already built.
> > > > + */
> > > > +#define KUNIT_UAPI_EMBED_BLOB(_name, _path)
> > > > \
> > > > + asm (
> > > > \
> > > > + " .pushsection .rodata, \"a\"
> > > > \n" \
> > > > + " .global " __stringify(CONCATENATE(_name, _data)) "
> > > > \n" \
> > > > + __stringify(CONCATENATE(_name, _data)) ":
> > > > \n" \
> > > > + " .incbin " __stringify(_path) "
> > > > \n" \
> > > > + " .size " __stringify(CONCATENATE(_name, _data)) ", "
> > > > \
> > > > + ". - " __stringify(CONCATENATE(_name, _data)) "
> > > > \n" \
> > > > + " .global " __stringify(CONCATENATE(_name, _end)) "
> > > > \n" \
> > > > + __stringify(CONCATENATE(_name, _end)) ":
> > > > \n" \
> > > > + " .popsection
> > > > \n" \
> > > > + );
> > > > \
> > > > +
> > > > \
> > > > + extern const char CONCATENATE(_name, _data)[];
> > > > \
> > > > + extern const char CONCATENATE(_name, _end)[];
> > > > \
> > > > +
> > > > \
> > > > + static const struct kunit_uapi_blob _name = {
> > > > \
> > > > + .path = _path,
> > > > \
> > > > + .data = CONCATENATE(_name, _data),
> > > > \
> > > > + .end = CONCATENATE(_name, _end),
> > > > \
> > > > + }
> > > > \
> > >
> > > For me, the compiler could not find the files for the ".incbin" unless
> > > I added an include path. i.e. adding
> > > ccflags-y := -I$(obj)
> > > to lib/kunit/Makefile fixed the problem for me.
> >
> > Can you share some more details on your build setup?
> > This worked for me as-is and also passed 0day build testing.
>
> Funny, I ran this on a Fedora 41 with gcc --version saying
> gcc (GCC) 14.3.1 20250523 (Red Hat 14.3.1-1)
>
> I tried both 32 and 64 bit builds for ARCH=um.
>
> Attaching my current kernel configuration and the last few lines of a
> V=1 build.
>
> The kernel I used is a bit newer than what you had. Applied on top of
> ee88bddf7f2f5d1f1da87dd7bedc734048b70e88 (bpf-fixes merge).
So this happens because you are building inside the source tree.
scripts/Makefile.lib has this block:
# $(src) for including checkin headers from generated source files
# $(obj) for including generated headers from checkin source files
ifdef building_out_of_srctree
_c_flags += $(addprefix -I, $(src) $(obj))
_a_flags += $(addprefix -I, $(src) $(obj))
_cpp_flags += $(addprefix -I, $(src) $(obj))
endif
Apparently GNU/clang assemblers don't look for .incbin/.include files next to
the including files [0].
In contrast, the ARM compiler does (at least according to its docs) [1].
Maybe we can work around this in the macro, but I assume this will become even
uglier. So for the next revision I'll use your proposal of explicit cflags.
Or if Masahiro prefers to have a more global and generic solution, we can do
that.
[0] https://sourceware.org/binutils/docs/as/I.html
[1]
https://developer.arm.com/documentation/dui0801/l/Directives-Reference/INCBIN
Thomas
