[ofa-general] Re: [PATCH 08 of 11] anon-vma-rwsem

[Andrea Arcangeli]

On Wed, May 07, 2008 at 09:14:45PM -0700, Linus Torvalds wrote:
> IOW, you didn't even look at it, did you?

Actually I looked both at the struct and at the slab alignment just in
case it was changed recently. Now after reading your mail I also
compiled it just in case.

2.6.26-rc1

# name            <active_objs> <num_objs> <objsize> <objperslab> <pagesperslab> : tunables <limit> <batchcount> <sharedfactor> : slabdata <active_slabs> <num_slabs> <sharedavail>
anon_vma             260    576     24  144    1 : tunables  120   60    8 : slabdata      4      4      0
                                    ^^  ^^^

2.6.26-rc1 + below patch

diff --git a/include/linux/rmap.h b/include/linux/rmap.h
--- a/include/linux/rmap.h
+++ b/include/linux/rmap.h
@@ -27,6 +27,7 @@ struct anon_vma {
 struct anon_vma {
 	spinlock_t lock;	/* Serialize access to vma list */
 	struct list_head head;	/* List of private "related" vmas */
+	int flag:1;
 };
 
 #ifdef CONFIG_MMU

# name            <active_objs> <num_objs> <objsize> <objperslab> <pagesperslab> : tunables <limit> <batchcount> <sharedfactor> : slabdata <active_slabs> <num_slabs> <sharedavail>
anon_vma             250    560     32  112    1 : tunables  120   60    8 : slabdata      5      5      0
                                    ^^  ^^^

Not a big deal sure to grow it 33%, it's so small anyway, but I don't
see the point in growing it. sort() can be interrupted by signals, and
until it can we can cap it to 512 vmas making the worst case taking
dozen usecs, I fail to see what you have against sort().

Again: if a vma bitflag + global lock could have avoided sort and run
O(N) instead of current O(N*log(N)) I would have done that
immediately, infact I was in the process of doing it when you posted
the followup. Nothing personal here, just staying technical. Hope you
too.