Re: [PATCH 5/5] types: Add standard __ob_trap and __ob_wrap scalar types

[Kees Cook <kees@kernel.org>]

On Wed, Apr 01, 2026 at 11:38:20AM +0200, Peter Zijlstra wrote:
> On Tue, Mar 31, 2026 at 01:31:16PM -0700, Kees Cook wrote:
> 
> > int func()
> > {
> > 	...
> > 	u8 __ob_trap product = 5;
> > 	...
> > 	product = a * b; // if store is truncated, goto __overflow
> > 	...
> > 	return product;
> > 
> > __overflow:
> > 	pr_info("%u\n", product); // shows "5"
> > 	return -1;
> > }
> > 
> > (Isn't this just an implicit "try"?)
> 
> So I like this implicit try with a default label, and mostly I expect
> this will be fine.
> 
> But as Linus already mentioned, sometimes you might want more. Could we
> perhaps also have an explicit version, something along the lines of:
> 
> int func()
> {
> 	int __ob_trap size;
> 
> 	size = try(count * flex_size, __mul_overflow);
> 	size = try(size + base_size, __add_overflow);
> 
> 	obj = kzalloc(size,...);
> 
> }
> 
> where we have something like:
> 
> #define try(stmt, _label) ({		\
> 	__label __overflow; 		\
> 	if (0) {			\
> __overflow:				\
> 		goto _label;		\
> 	}				\
> 	stmt; })
> 
> That is, have the overflow trapped and confined in the
> statement-expression by using the overflow label as a local label and
> use this little trampoline to re-direct to a custom label.

Yeah, that should work, and gives us a nice way to create handler
overrides. We've have to make sure the "locally defined" labels (with
__label__) and __ob_trap worked together sanely.

-- 
Kees Cook