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

[Peter Zijlstra]

On Fri, Sep 15, 2023 at 01:40:25PM -0700, Linus Torvalds wrote:

> Not because I think it's necessarily any kind of final rule, but
> because I think our whole cleanup thing is new enough that I think
> we're better off being a bit inflexible, and having a syntax where a
> simple "grep" ends up showing pretty much exactly what is going on wrt
> the pairing.

So in the perf-event conversion patches I do have this: 

	struct task_struct *task __free(put_task) = NULL;

	...

	if (pid != -1) {
		task = find_lively_task_by_vpid(pid);
		if (!task)
			return -ESRCH;
	}

	...

pattern. The having of task is fully optional in the code-flow.

I suppose I can try and rewrite that a little something like:

	...

	struct task_struct *task __free(put_task) =
		find_lively_task_by_vpid(pid); /* ensure pid==-1 returns NULL */

	if (!task && pid > 0)
		return -ESRCH;

	...


But a little later in that same function I then have:

	do {
		struct rw_semaphore *exec_update_lock __free(up_read) = NULL;
		if (task) {
			err = down_read_interruptible(&task->signal->exec_update_lock);
			if (err)
				return err;

			exec_update_lock = &task->signal->exec_update_lock;

			if (!perf_check_permissions(&attr, task))
				return -EACCESS;
		}

		... stuff serialized against exec *if* this is a task event ...

	} while (0);


And that might be a little harder to 'fix'.


I suppose I'm saying that when thing truly are conditional, this is a
useful pattern, but avoid where reasonably possible.