On Fri, Sep 12, 2025 at 2:03 PM Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote: > > On Fri, Sep 12, 2025 at 12:27 PM Shakeel Butt <shakeel.butt@xxxxxxxxx> wrote: > > > > +Suren, Roman > > > > On Mon, Sep 08, 2025 at 06:00:06PM -0700, Alexei Starovoitov wrote: > > > From: Alexei Starovoitov <ast@xxxxxxxxxx> > > > > > > Since the combination of valid upper bits in slab->obj_exts with > > > OBJEXTS_ALLOC_FAIL bit can never happen, > > > use OBJEXTS_ALLOC_FAIL == (1ull << 0) as a magic sentinel > > > instead of (1ull << 2) to free up bit 2. > > > > > > Signed-off-by: Alexei Starovoitov <ast@xxxxxxxxxx> > > > > Are we low on bits that we need to do this or is this good to have > > optimization but not required? > > That's a good question. After this change MEMCG_DATA_OBJEXTS and > OBJEXTS_ALLOC_FAIL will have the same value and they are used with the > same field (page->memcg_data and slab->obj_exts are aliases). Even if > page_memcg_data_flags can never be used for slab pages I think > overlapping these bits is not a good idea and creates additional > risks. Unless there is a good reason to do this I would advise against > it. Completely disagree. You both missed the long discussion during v4. The other alternative was to increase alignment and waste memory. Saving the bit is obviously cleaner. The next patch is using the saved bit.
