Re: [PATCH v3 2/5] tracing: Use __free() in trace_probe for cleanup

[Steven Rostedt]

On Wed, 8 Jan 2025 09:38:43 +0900
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx> wrote:

> > I don't get this? You are telling the compiler not to free tmp, because you
> > decided to free it yourself? Why not just remove the kfree() here altogether?  
> 
> In the for-loop block, the __free() work only when we exit the loop, not
> each iteration. In each iteration, kstrdup() is assigned to the 'tmp',
> so we need to kfree() each time.

Really? It doesn't trigger for each iteration? That's rather unintuitive. :-/
And sounds buggy, as wouldn't that then cause a memory leak?

I would say not to use __free() for tmp at all. Because now it's just
getting confusing.

-- Steve


> 
> Hmm, maybe this is a sign that I should not use __free() for the 'tmp',
> or I should call kfree(tmp) right before kstrdup(), like below.
> 
>  	for (i = 0; i < argc; i++) {
> 		char *tmp __free(kfree) = NULL;
> 		...
> 		kfree(tmp);
> 		tmp = kstrdup(argv[i], GFP_KERNEL);
> 	}
> 
> Does this make sense?