On Wed, 16 Oct 2013 17:43:57 +0900 Joonsoo Kim <iamjoonsoo.kim@xxxxxxx> wrote: > There is two main topics in this patchset. One is to reduce memory usage > and the other is to change a management method of free objects of a slab. > > The SLAB allocate a struct slab for each slab. The size of this structure > except bufctl array is 40 bytes on 64 bits machine. We can reduce memory > waste and cache footprint if we overload struct slab over struct page. Seems a good idea from a quick look. A thought: when we do things like this - adding additional interpretations to `struct page', we need to bear in mind that other unrelated code can inspect that pageframe. It is not correct to assume that because slab "owns" this page, no other code will be looking at it and interpreting its contents. One example is mm/memory-failure.c:memory_failure(). It starts with a raw pfn, uses that to get at the `struct page', then starts playing around with it. Will that code still work correctly when some of the page's fields have been overlayed with slab-specific contents? And memory_failure() is just one example - another is compact_zone() and there may well be others. This issue hasn't been well thought through. Given a random struct page, there isn't any protocol to determine what it actually *is*. It's a plain old variant record, but it lacks the agreed-upon tag field which tells users which variant is currently in use. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>