ksm: shrink 32-bit rmap_item back to 32 bytes
authorHugh Dickins <hughd@google.com>
Sat, 23 Feb 2013 00:36:06 +0000 (16:36 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 24 Feb 2013 01:50:23 +0000 (17:50 -0800)
commitbc56620b493496b8a6962080b644ccc537f4d526
tree46f59d3ac15cac32cbb97b8a48cc4dffb539b3f8
parentb599cbdf1c2d88eac7caed00854ee4eecb119a6b
ksm: shrink 32-bit rmap_item back to 32 bytes

Think of struct rmap_item as an extension of struct page (restricted to
MADV_MERGEABLE areas): there may be a lot of them, we need to keep them
small, especially on 32-bit architectures of limited lowmem.

Siting "int nid" after "unsigned int checksum" works nicely on 64-bit,
making no change to its 64-byte struct rmap_item; but bloats the 32-bit
struct rmap_item from (nicely cache-aligned) 32 bytes to 36 bytes, which
rounds up to 40 bytes once allocated from slab.  We'd better avoid that.

Hey, I only just remembered that the anon_vma pointer in struct
rmap_item has no purpose until the rmap_item is hung from a stable tree
node (which has its own nid field); and rmap_item's nid field no purpose
than to say which tree root to tell rb_erase() when unlinking from an
unstable tree.

Double them up in a union.  There's just one place where we set anon_vma
early (when we already hold mmap_sem): now we must remove tree_rmap_item
from its unstable tree there, before overwriting nid.  No need to
spatter BUG()s around: we'd be seeing oopses if this were wrong.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Petr Holasek <pholasek@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Izik Eidus <izik.eidus@ravellosystems.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/ksm.c