Re: [PATCH v4 07/16] slub: Remove page->counters

[Matthew Wilcox <willy@infradead.org>]

On Thu, May 03, 2018 at 10:03:10AM -0500, Christopher Lameter wrote:
> On Wed, 2 May 2018, Matthew Wilcox wrote:
> 
> > > > Option 2:
> > > > +                       union {
> > > > +                               unsigned long counters;
> > > > +                               struct {
> > > > +                                       unsigned inuse:16;
> > > > +                                       unsigned objects:15;
> > > > +                                       unsigned frozen:1;
> > > > +                               };
> > > > +                       };
> > > >
> > > > Pro: Expresses exactly what we do.
> > > > Con: Back to five levels of indentation in struct page
> 
> I like that better. Improves readability of the code using struct page. I
> think that is more important than the actual definition of struct page.

OK.  Do you want the conversion of slub to using slub_freelist and slub_list
as part of this patch series as well, then?

The end result looks like this, btw:

                struct {        /* slub */ 
                        union {
                                struct list_head slub_list;
                                struct {
                                        struct page *next; /* Next partial */
#ifdef CONFIG_64BIT
                                        int pages;      /* Nr of pages left */
                                        int pobjects;   /* Apprx # of objects */
#else
                                        short int pages;
                                        short int pobjects;
#endif
                                };
                        };
                        struct kmem_cache *slub_cache;  /* shared with slab */
                        /* Double-word boundary */
                        void *slub_freelist;            /* shared with slab */
                        union {
                                unsigned long counters;
                                struct {
                                        unsigned inuse:16;
                                        unsigned objects:15;
                                        unsigned frozen:1;
                                };
                        };
                };

Oh, and what do you want to do about cache_from_obj() in mm/slab.h?
That relies on having slab_cache be in the same location in struct
page as slub_cache.  Maybe something like this?

        page = virt_to_head_page(x);
#ifdef CONFIG_SLUB
        cachep = page->slub_cache;
#else
        cachep = page->slab_cache;
#endif
        if (slab_equal_or_root(cachep, s))
                return cachep;

> Given the overloaded overload situation this will require some deep
> throught for newbies anyways. ;-)

Yes, it's all quite entangled.