Re: [PATCH 1/2] mm/memcontrol: return 1 from cgroup.memory __setup() handler

[Randy Dunlap <rdunlap@infradead.org>]

On 3/3/22 14:32, Michal Koutný wrote:
> On Thu, Mar 03, 2022 at 01:53:03PM -0800, Randy Dunlap <rdunlap@infradead.org> wrote:
>>> Isn't mere presence of the handler sufficient to filter those out? [1]
>>
>> What is [1] here?
> 
> Please ignore, too much editing on my side.
> 
>> I don't know of any case where "foo=2" should be passed to init if
>> there is a setup function for "foo=" defined.
> 
> Good. I was asking because of the following semantics:
> - absent handler -- pass to init,

Ack: if the handler code is not built, it is an Unknown boot option
and is passed to init.

> - returns 0 -- filter out,
> - returns negative -- filter out, print message.

Currently setup functions should return 1 (or any non-zero value)
to indicate "handled" or should return 0 to indicate "not handled".

Andrew has a patch in mmotm: include/linux/init.h so that the comment
before __setup() says:

/*
 * NOTE: __setup functions return values:
 * @fn returns 1 (or non-zero) if the option argument is "handled"
 * and returns 0 if the option argument is "not handled".
 */

>>> (Richer reporting or -EINVAL is by my understanding now a different
>>> problem.)

-- 
~Randy