Re: [tip:x86/tdx 8/12] vmlinux.o: warning: objtool: __tdx_hypercall+0x128: __tdx_hypercall_failed() is missing a __noreturn annotation

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Sep 14, 2023 at 10:02:40AM +0000, Huang, Kai wrote:
> On Thu, 2023-09-14 at 10:59 +0200, Peter Zijlstra wrote:
> > On Thu, Sep 14, 2023 at 07:54:10AM +0000, Huang, Kai wrote:
> > 
> > > > The point of noreturn is that the caller should know to stop generating
> > > > code. For that the declaration needs the attribute, because call sites
> > > > typically do not have access to the function definition in C.
> > > 
> > > Ah that makes perfect sense.  Thanks!
> > > 
> > > Then I assume we don't need to annotate __noreturn in the function body, but
> > > only in the declaration?  Because the compiler must already have seen the
> > > declaration when it generates the code for the function body.
> > 
> > I think so, I'm sure it'll tell you if it disagrees :-)
> > 
> > > Btw, I happened to notice that the objtool documentation suggests that we should
> > > also add the the function to tools/objtool/noreturns.h:
> > > 
> > > 3. file.o: warning: objtool: foo+0x48c: bar() is missing a __noreturn annotation
> > > 
> > >    The call from foo() to bar() doesn't return, but bar() is missing the
> > >    __noreturn annotation.  NOTE: In addition to annotating the function
> > >    with __noreturn, please also add it to tools/objtool/noreturns.h.
> > > 
> > > Is it a behaviour that we still need to follow?
> > 
> > Yes. objtool has some heuristics to detect noreturn, but is is very
> > difficult. Sadly noreturn is not something that is reflected in the ELF
> > object file so we have to guess.
> > 
> > For now manually adding it to the objtool list is required, we're trying
> > to get to the point where it is generated/validated by the compiler,
> > perhaps with a plugin, but we're not there yet.
> 
> Sorry one more question, do you know what Kconfig option triggers this
> __noreturn objtool check?  I couldn't reproduce it on my own machine but have to
> depend on LKP for now.

Nope, sorry. Could be very GCC version specific through, this is all
about a specific code-gen pattern.

AFAICT we emit this warning when objtool finds this function is a
noreturn but code-gen continued.