On Sat 03-08-24 15:15:56, Davidlohr Bueso wrote:
> On Mon, 29 Jul 2024, Michal Hocko wrote:
> > Because it is really hard to figure out what it is supposed to mean.
> > If the caller uses __GFP_NOFAIL then it is (should be) impossible and if
> > NOFAIL is not used then why does it need to check for
> > (gfp & ~__GFP_NOFAIL) != GFP_KERNEL?
>
> Agreed, this is pointless - and cannot recall why it was justified to have
> in the first place.
>
> But I think we should revert back to the original check then, which is there
> to distinguish failure cases between normal (GFP_KERNEL) and nested (GFP_ATOMIC)
> contexts. Removing the check altogether would change the fallback for regular
> allocations.
>
> So this would be:
>
> - if (tbl == NULL && (gfp & ~__GFP_NOFAIL) != GFP_KERNEL) {
> + if (tbl == NULL && gfp != GFP_KERNEL) {
If you want to tell between sleeping and atomic allocations then already
mentioned gfpflags_allow_blocking would be more readable IMHO but the
above is much better already.
--
Michal Hocko
SUSE Labs
