Re: [RFC PATCH 4/5] mm/madvise: add PMADV_SET_FORK_EXEC_DEFAULT process_madvise() flag

[David Hildenbrand <david@redhat.com>]

On 20.05.25 18:19, Lorenzo Stoakes wrote:
> On Tue, May 20, 2025 at 06:11:49PM +0200, Jann Horn wrote:
>> On Tue, May 20, 2025 at 10:38 AM Pedro Falcato <pfalcato@suse.de> wrote:
>>> - PMADV_INHERIT_FORK. This makes it so the flag is propagated to child processes (does not imply PMADV_FUTURE)
>>
>> Do we think there will be flags settable through this API that we
>> explicitly _don't_ want to inherit on fork()? My understanding is that
>> sort of the default for fork() is to inherit everything, and things
>> that don't get inherited are weird special cases (like mlock() state).
>> (While the default for execve() is to inherit nothing about the MM.)
> 
> Yeah this is true. It is the exception rather than the rule...
> 
>>
>> (I guess you could make a case that in a fork+exec sequence, the child
>> might not want to create hugepage between fork and exec... but this is
>> probably not the right place to control that?)
> 
>  From my point of view it simply reads better :)
> 
> But perhaps we can drop the fork bit and leave that implied.

Yes, we should. Exec is where the "fun" begins.

-- 
Cheers,

David / dhildenb