[PATCH 0/6] vmtrunc: lowlat vmtruncate drop i_mmap_lock

[Hugh Dickins]

Here's a set of patches against 2.6.9-rc3-mm1, or Ingo's -S7 VP tree:
allow lower latency in vmtruncate (more generally, unmap_mapping_range)
by dropping the i_mmap_lock when required.  Though I've been focussing
on correctness rather than latency myself: do these work for you, Ingo?

1/6 undoes something of Andrea's, 2/6 undoes something of Ingo's:
I thought I might as well provide patches as ask the questions, please
let us know why if you disagree with them.  3/6 is cosmetic, 4/6 is
the real work, 5/6 corrects some loose ends, 6/6 is for testing.

Worth the effort, and ~700 bytes bloat, and additional int in each vma?
If only root could run-bash-shared-mappings, I'd say not worth it; but
it seems a little like prio_tree itself, we're not sure we need it,
but feel more secure to have the solution.

Signed-off-by: Hugh Dickins <hugh@xxxxxxxxxxx>
---

 drivers/mtd/devices/blkmtd.c |    1 
 fs/inode.c                   |    1 
 include/linux/fs.h           |    2 
 include/linux/mm.h           |    8 +
 kernel/fork.c                |    1 
 mm/memory.c                  |  223 ++++++++++++++++++++++++++++++++++---------
 mm/mmap.c                    |   14 ++
 mm/mremap.c                  |   16 +--
 mm/rmap.c                    |    9 -
 mm/truncate.c                |    1 
 10 files changed, 215 insertions(+), 61 deletions(-)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/