Re: [PATCH v1] rust: bug: Add bug_on macro

[Greg KH <gregkh@linuxfoundation.org>]

On Wed, Jul 08, 2026 at 07:48:59AM +0900, FUJITA Tomonori wrote:
> On Tue, 7 Jul 2026 17:20:10 +0200
> Greg KH <gregkh@linuxfoundation.org> wrote:
> 
> > On Tue, Jul 07, 2026 at 04:51:40PM +0200, Danilo Krummrich wrote:
> >> On Tue Jul 7, 2026 at 4:30 PM CEST, FUJITA Tomonori wrote:
> >> > Use of this macro is heavily discouraged. Prefer warn_on() macro with
> >> > recovery code where feasible. See the documentation for C's BUG_ON()
> >> > for guidance on when it is rarely justified.
> >> 
> >> Given that, can you please spell out the motivation for this patch?
> > 
> > I agree, this should probably NOT be added unless there is a real user
> > for it that requires it and we all agree that crashing the system is the
> > only way out.
> 
> bindings::BUG() is already in use today, indirectly, from the kernel
> crate itself and from drivers such as binder and nova-core, via
> panic!()/.exepct().unwrap() on invariants that "can't happen". All of
> those funnel through panic() in kernel/lib.rs (#[panic_handler]),
> which itself calls BUG().
> 
> The motivation for this patch is narrow: replace that bindings::BUG()
> call in lib.rs's panic() with bug_on!(), for a somewhat better
> trace. It isn't adding a new decision to crash the system -- that
> decision is already made at each of the existing panic!()/.expect()
> call sites.

That wasn't obvious that this was the reason why this change was being
introduced, perhaps the changelog should have said that?

thanks,

greg k-h