Re: [PATCH v4 06/14] x86: Add support for suppressing warning backtraces

Josh Poimboeuf <[email protected]> Tue, 1 Apr 2025 13:00:55 -0700

Newsgroups	gmane.linux.documentation,gmane.comp.video.dri.devel,gmane.linux.kernel.cross-arch,gmane.linux.ports.arm.kernel,gmane.linux.kernel,gmane.linux.ports.parisc,gmane.linux.ports.ppc64.devel,gmane.linux.ports.riscv,gmane.linux.ports.sh.devel
Message-ID	<rkuxvq6pkha6pixz5rtu327ujt7ism5f4wgrak7egcecuxwe42@qkn5ewdzwhvo>

On Tue, Apr 01, 2025 at 10:53:46AM -0700, Guenter Roeck wrote:
> On 4/1/25 10:08, Peter Zijlstra wrote:
> > > +	if (!KUNIT_IS_SUPPRESSED_WARNING(__func__))			\
> > > +		_BUG_FLAGS(ASM_UD2, __flags, ANNOTATE_REACHABLE(1b));	\
> > >   	instrumentation_end();					\
> > >   } while (0)
> > 
> > NAK, this grows the BUG site for now appreciable reason.
> 
> Only if CONFIG_KUNIT_SUPPRESS_BACKTRACE is enabled. Why does that
> warrant a NACK ?

I agree with Peter, this bloats the code around thousands of UD2 sites.

It would be much better to do the checking after the exception.  In fact
it looks like you're already doing that in report_bug()?

	if (warning && KUNIT_IS_SUPPRESSED_WARNING(function))
		return BUG_TRAP_TYPE_WARN;

Why check it twice?

-- 
Josh