Re: [PATCH 1/1] lib/stackdepot: print disabled message only if truly disabled

[Andrey Konovalov]

On Fri, Oct 27, 2023 at 2:54 PM Marco Elver <elver@xxxxxxxxxx> wrote:
>
> stack_bucket_number_order seems like a redundant variable, that should
> at least be __ro_after_init. All code that does "if
> (stack_bucket_number_order) ..." could just do "if (kasan_enabled())
> ..." and use STACK_BUCKET_NUMBER_ORDER_MAX constant directly instead.
>
> The code here could be simplified if it was removed. No idea why it
> was introduced in the first place. I think f9987921cb541 introduced it
> and there it said "complemented with a boot-time kernel parameter",
> but that never happened.
>
> So I'd be in favor of removing that variable, which will also simplify
> this code.

On the first thought, this seems reasonable with the current code.

On the second though, I think I will soon add a command-line parameter
to allow controlling how much memory is used for the stack depot hash
table.

I propose we keep the code as is for now, but I've taken a note to
either drop this variable or mark it as __ro_after_init when
implementing memory bounding controls for stack depot.

Thanks!