Re: [PATCH v5 1/1] mm: enumerate all gfp flags

[Michal Hocko]

On Sun 25-02-24 01:12:46, Suren Baghdasaryan wrote:
> On Sat, Feb 24, 2024 at 7:03 AM Christophe JAILLET
> <christophe.jaillet@xxxxxxxxxx> wrote:
> >
> > Le 24/02/2024 à 02:58, Suren Baghdasaryan a écrit :
> > > Introduce GFP bits enumeration to let compiler track the number of used
> > > bits (which depends on the config options) instead of hardcoding them.
> > > That simplifies __GFP_BITS_SHIFT calculation.
> > >
> > > Suggested-by: Petr Tesařík <petr@xxxxxxxxxxx>
> > > Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
> > > Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > > Reviewed-by: Pasha Tatashin <pasha.tatashin@xxxxxxxxxx>
> > > Acked-by: Michal Hocko <mhocko@xxxxxxxx>
> > > ---
> > > Changes from v4 [1]:
> > > - Split from the series [2] as a stand-alone patch, per Michal Hocko
> > > - Added Reviewed-by, per Pasha Tatashin
> > > - Added Acked-by, per Michal Hocko
> > >
> > > [1] https://lore.kernel.org/all/20240221194052.927623-7-surenb@xxxxxxxxxx/
> > > [2] https://lore.kernel.org/all/20240221194052.927623-1-surenb@xxxxxxxxxx/
> > >
> > >   include/linux/gfp_types.h | 90 +++++++++++++++++++++++++++------------
> > >   1 file changed, 62 insertions(+), 28 deletions(-)
> > >
> > > diff --git a/include/linux/gfp_types.h b/include/linux/gfp_types.h
> > > index 1b6053da8754..868c8fb1bbc1 100644
> > > --- a/include/linux/gfp_types.h
> > > +++ b/include/linux/gfp_types.h
> > > @@ -21,44 +21,78 @@ typedef unsigned int __bitwise gfp_t;
> > >    * include/trace/events/mmflags.h and tools/perf/builtin-kmem.c
> > >    */
> > >
> > > +enum {
> > > +     ___GFP_DMA_BIT,
> > > +     ___GFP_HIGHMEM_BIT,
> > > +     ___GFP_DMA32_BIT,
> > > +     ___GFP_MOVABLE_BIT,
> > > +     ___GFP_RECLAIMABLE_BIT,
> > > +     ___GFP_HIGH_BIT,
> > > +     ___GFP_IO_BIT,
> > > +     ___GFP_FS_BIT,
> > > +     ___GFP_ZERO_BIT,
> > > +     ___GFP_UNUSED_BIT,      /* 0x200u unused */
> >
> > Hi,
> >
> > what is the need to have this ___GFP_UNUSED_BIT now?
> 
> Hi!
> We can remove it but then all values will shift. That should be safe
> to do now but I prefer one patch to do only one thing. We can add a
> separate patch to do further cleanup of unused values.

Agreed!

> > > +     ___GFP_DIRECT_RECLAIM_BIT,
> > > +     ___GFP_KSWAPD_RECLAIM_BIT,
> > > +     ___GFP_WRITE_BIT,
> > > +     ___GFP_NOWARN_BIT,
> > > +     ___GFP_RETRY_MAYFAIL_BIT,
> > > +     ___GFP_NOFAIL_BIT,
> > > +     ___GFP_NORETRY_BIT,
> > > +     ___GFP_MEMALLOC_BIT,
> > > +     ___GFP_COMP_BIT,
> > > +     ___GFP_NOMEMALLOC_BIT,
> > > +     ___GFP_HARDWALL_BIT,
> > > +     ___GFP_THISNODE_BIT,
> > > +     ___GFP_ACCOUNT_BIT,
> > > +     ___GFP_ZEROTAGS_BIT,
> > > +#ifdef CONFIG_KASAN_HW_TAGS
> > > +     ___GFP_SKIP_ZERO_BIT,
> > > +     ___GFP_SKIP_KASAN_BIT,
> > > +#endif
> > > +#ifdef CONFIG_LOCKDEP
> > > +     ___GFP_NOLOCKDEP_BIT,
> > > +#endif
> > > +     ___GFP_LAST_BIT
> > > +};
> >
> > Does it make sense to have something like:
> >    BUILD_BUG_ON(___GFP_LAST_BIT > BITS_PER_LONG, "blah");
> 
> I suppose that would not hurt, except gfp_t is unsigned int, not long.
> Something like this would work I think:
> 
> BUILD_BUG_ON_MSG(___GFP_LAST_BIT > BITS_PER_TYPE(gfp_t), "GFP bit overflow");
> 
> except I'm not sure where to put this check. One of the __init
> functions in page_alloc.c would probably work but none seem to be
> appropriate. mm_core_init() perhaps? Other ideas?

Would that check add much? We currently cannot use the full width of the
gfp_t because radix tree code needs to fit also its own tag into the
same word (see radix_tree_init). If the radix tree constrain is lifted
then we should add something like the above.
-- 
Michal Hocko
SUSE Labs