Re: [GIT PULL] ftrace: Check if pages were allocated before calling free_pages()

[Linus Torvalds]

On Wed, Mar 31, 2021 at 6:27 AM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
>                 order = get_count_order(pg->size / ENTRIES_PER_PAGE);
> -               free_pages((unsigned long)pg->records, order);
> +               if (order >= 0)
> +                       free_pages((unsigned long)pg->records, order);

Honestly, looking at that code, every single use of
"get_count_order()" seems really really confusing.

And really confused.

The comments are garbage, the code is odd, it's just all very very strange.

See here in ftrace_allocate_records():

        pages = DIV_ROUND_UP(count, ENTRIES_PER_PAGE);
        order = get_count_order(pages);

        /*
         * We want to fill as much as possible. No more than a page
         * may be empty.
         */
        if (!is_power_of_2(pages))
                order--;

can you actually explain what the logic is here?

The 'get_count_order()' function will return the smallest order that
the value fits in. But then if the value wasn't a power of two, you
subtract 1. If I understand the code correctly, that's just the same
as "what is the highest bit set", isn't it?

So afaik, what you *actually* want is just

        pages = DIV_ROUND_UP(count, ENTRIES_PER_PAGE);
        order = fls(pages)-1;

isn't it?

Did I misunderstand what the code wants to do? The "No more than a
page may be empty" seems wrong - you really mean "no empty pages".

I dunno. Maybe I'm wrong, but that code is really confusing (which is
why I may be wrong). It doesn't seem to make any sense at all to use
"get_count_order()" and then modify the end result.

             Linus