Re: Buggy __free(kfree) usage pattern already in tree

[Peter Zijlstra <peterz@infradead.org>]

On Tue, Sep 19, 2023 at 03:10:38PM +0200, Peter Zijlstra wrote:

> This isn't exactly nice..
> 
> I tried something like:
> 
> 	scoped_cond_guard (mutex_intr, return -EINTR, &task->signal->cred_guard_mutex) {
> 		...
> 	}
> 
> Which I can make work, but then I also tried to capture my other case:
> 
> 	scoped_cond_guard (rwsem_down_intr, if (task) return -EINTR,
> 			   task ? &task->signal->exec_guard_mutex : NULL)  {
> 
> 		...
> 	}
> 
> But I can't get that to work because of that extra if, the not case
> doesn't fall through and do the body.
> 
> Anyway, I'll poke more..


#define scoped_cond_guard(_name, _label, args...)		\
	for (CLASS(_name, scope)(args),				\
	     *done = NULL; !done; done = (void *)1)		\
		if (!__guard_ptr(_name)(&scope)) goto _label;	\
		else

Allows one to write:

	scoped_cond_guard (rwsem_down_intr, no_lock,
 			   task ? &task->signal->exec_guard_mutex : NULL)  {

		if (0) {
no_lock:
			if (task)
				return -EINTR;
		}

		... block that holds exec_guard_mutex if task ...
	}

Still not exactly pretty, but perhaps better than before...