Re: [RFC PATCH 0/5] add process_madvise() flags to modify behaviour

[David Hildenbrand <david@redhat.com>]

On 20.05.25 20:25, Lorenzo Stoakes wrote:
> On Tue, May 20, 2025 at 06:47:48PM +0100, Lorenzo Stoakes wrote:
>> On Tue, May 20, 2025 at 05:28:35PM +0200, David Hildenbrand wrote:
> [snip]
>>>> 3. PMADV_SET_FORK_EXEC_DEFAULT
>>>>
>>>> It may be desirable for a user to specify that all VMAs mapped in a process
>>>> address space default to having an madvise() behaviour established by
>>>> default, in such a fashion as that this persists across fork/exec.
>>>
>>> This is very specific for MADV_HUGEPAGE only, so I wonder how we could
>>> either avoid that flag or just make it clear that it shall stick around ...
>>>
> 
> Sorry missed this bit.
> 
> I don't really like the idea of only for MADV_HUGEPAGE (and MADV_NOHUGEPAGE)
> defaulting to doing this, I think that's far less clear than a user explicitly
> asking for it, plus most users using process_madvise() wouldn't expect it.

The thing is that PMADV_ENTIRE_ADDRESS_SPACE already adds something 
unexpected: suddenly also *new* mappings will be affected.

But maybe that'd be covered by the PMADV_SET_DEFAULT idea.

Maybe

PMADV_ENTIRE_ADDRESS_SPACE | PMADV_SET_DEFAULT | PMADV_SET_EXEC_DEFAULT

... hm, but "PMADV_ENTIRE_ADDRESS_SPACE" can then simply be "pass in the 
entire address space as a range" ... and ignore errors.

I'd probably have to see a revised proposal after the current discussions.


Anyhow, limiting the flags to a bare minimum for now usually makes 
sense, as long as it is expandable.

-- 
Cheers,

David / dhildenb