[BUG] -next objtool build failure (bisected)

[BUG] -next objtool build failure (bisected)

[Paul E. McKenney]

Hello!

The next-20241011 release gets me build errors like the following:

	vmlinux.o: warning: objtool: fetch_item() falls through to next function

Bisection leads me here:

	61595012f280 ("HID: simplify code in fetch_item()")

This diff looks inoffensive to me, but I get this error on this commit
and not on its predecessor.

This build failure happens on quite a few different kernel configurations.
I bisected using this particular reproducer:

tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --buildonly --configs SRCU-T

Thoughts?

							Thanx, Paul

Re: [BUG] -next objtool build failure (bisected)

[Dmitry Torokhov]

Hi Paul,

On Tue, Oct 15, 2024 at 11:10:11AM -0700, Paul E. McKenney wrote:
> Hello!
> 
> The next-20241011 release gets me build errors like the following:
> 
> 	vmlinux.o: warning: objtool: fetch_item() falls through to next function
> 
> Bisection leads me here:
> 
> 	61595012f280 ("HID: simplify code in fetch_item()")
> 
> This diff looks inoffensive to me, but I get this error on this commit
> and not on its predecessor.
> 
> This build failure happens on quite a few different kernel configurations.
> I bisected using this particular reproducer:
> 
> tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --buildonly --configs SRCU-T
> 
> Thoughts?

I just added you to another thread where Nathan proposed a couple of
solutions:

https://lore.kernel.org/r/20241010222451.GA3571761@thelio-3990X

Thanks.

-- 
Dmitry

Re: [BUG] -next objtool build failure (bisected)

[Paul E. McKenney]

On Tue, Oct 15, 2024 at 11:33:11AM -0700, Dmitry Torokhov wrote:
> Hi Paul,
> 
> On Tue, Oct 15, 2024 at 11:10:11AM -0700, Paul E. McKenney wrote:
> > Hello!
> > 
> > The next-20241011 release gets me build errors like the following:
> > 
> > 	vmlinux.o: warning: objtool: fetch_item() falls through to next function
> > 
> > Bisection leads me here:
> > 
> > 	61595012f280 ("HID: simplify code in fetch_item()")
> > 
> > This diff looks inoffensive to me, but I get this error on this commit
> > and not on its predecessor.
> > 
> > This build failure happens on quite a few different kernel configurations.
> > I bisected using this particular reproducer:
> > 
> > tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --buildonly --configs SRCU-T
> > 
> > Thoughts?
> 
> I just added you to another thread where Nathan proposed a couple of
> solutions:
> 
> https://lore.kernel.org/r/20241010222451.GA3571761@thelio-3990X

Thank you, Dmitry!  Glad that it is not just me?  ;-)

							Thanx, Paul

Re: [BUG] -next objtool build failure (bisected)

[Josh Poimboeuf]

On Tue, Oct 15, 2024 at 11:10:11AM -0700, Paul E. McKenney wrote:
> Hello!
> 
> The next-20241011 release gets me build errors like the following:
> 
> 	vmlinux.o: warning: objtool: fetch_item() falls through to next function
> 
> Bisection leads me here:
> 
> 	61595012f280 ("HID: simplify code in fetch_item()")
> 
> This diff looks inoffensive to me, but I get this error on this commit
> and not on its predecessor.
> 
> This build failure happens on quite a few different kernel configurations.
> I bisected using this particular reproducer:
> 
> tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --buildonly --configs SRCU-T

The problem is likely this:

+	default:
+		unreachable();

The unreachable() annotation by itself produces undefined behavior, as
it will just fall off the edge of the function if the default case is
ever executed.

Better to do

	default:
		BUG();

which makes the behavior defined.

-- 
Josh

Re: [BUG] -next objtool build failure (bisected)

[Paul E. McKenney]

On Tue, Oct 15, 2024 at 01:45:07PM -0500, Josh Poimboeuf wrote:
> On Tue, Oct 15, 2024 at 11:10:11AM -0700, Paul E. McKenney wrote:
> > Hello!
> > 
> > The next-20241011 release gets me build errors like the following:
> > 
> > 	vmlinux.o: warning: objtool: fetch_item() falls through to next function
> > 
> > Bisection leads me here:
> > 
> > 	61595012f280 ("HID: simplify code in fetch_item()")
> > 
> > This diff looks inoffensive to me, but I get this error on this commit
> > and not on its predecessor.
> > 
> > This build failure happens on quite a few different kernel configurations.
> > I bisected using this particular reproducer:
> > 
> > tools/testing/selftests/rcutorture/bin/kvm.sh --allcpus --buildonly --configs SRCU-T
> 
> The problem is likely this:
> 
> +	default:
> +		unreachable();
> 
> The unreachable() annotation by itself produces undefined behavior, as
> it will just fall off the edge of the function if the default case is
> ever executed.
> 
> Better to do
> 
> 	default:
> 		BUG();
> 
> which makes the behavior defined.

And this fixes things for me, thank you!

							Thanx, Paul