Re: [PATCH] mm/page_alloc: let the bulk and folio allocators carry alloc_flags

[Vlastimil Babka (SUSE)]

On 9/23/26 17:24, Gregory Price wrote:
> On Wed, Sep 23, 2026 at 08:32:29AM +0200, Vlastimil Babka (SUSE) wrote:
>> On 9/23/26 03:59, Gregory Price wrote:
>> > On Tue, Sep 22, 2026 at 09:05:58PM +0800, Lance Yang wrote:
>> > 
>> > There is some concern around NOLOCK here yes, in fact sashiko picked
>> > this issue up and I've been poking at it.
>> 
>> It's possible to document it as unsupported, add a VM_WARN_ON_ONCE etc.
>> It's not necessary for bulk alloc to support ALLOC_NOLOCK just because it
>> takes flags. e.g. kmalloc_nolock() supports a subset of gfp flags as well.
>> 
> 
> It seemed simpler to just support it than to create more future
> footguns.  I can RFC my current rework if you like, i attempted
> to deduplicate the logic via a prepare_alloc_flags() called from
> both the bulk and non-bulk contexts.

Ok if it's simple enough, lets see :) Thanks

> But if we prefer the WARNs over adding the new semantics to the bulk
> allocator, that also doesn't seem unreasonable.
> 
> ~Gregory