The patch titled
shmem: don't release lock for hole punching
has been added to the -mm tree. Its filename is
shmem-dont-release-lock-for-hole-punching.patch
*** Remember to use Documentation/SubmitChecklist when testing your code ***
See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
out what to do about this
------------------------------------------------------
Subject: shmem: don't release lock for hole punching
From: Miklos Szeredi <mszeredi@xxxxxxx>
During truncation of shmem page directories, info->lock is released to
improve latency. But this is wrong for hole punching, because the memory
areas being operated on are still in use by shmem_unuse, shmem_getpage and
shmem_writepage.
So for hole punching don't release the lock. Users of MADV_REMOVE likely
don't care about latency anyway. But this function really wants a cleanup,
and with that latency could also be taken care of.
Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxx>
Cc: Hugh Dickins <hugh@xxxxxxxxxxx>
Cc: Badari Pulavarty <pbadari@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/shmem.c | 17 ++++++++++-------
1 file changed, 10 insertions(+), 7 deletions(-)
diff -puN mm/shmem.c~shmem-dont-release-lock-for-hole-punching mm/shmem.c
--- a/mm/shmem.c~shmem-dont-release-lock-for-hole-punching
+++ a/mm/shmem.c
@@ -421,7 +421,7 @@ static int shmem_free_swp(swp_entry_t *d
}
static int shmem_map_and_free_swp(struct page *subdir,
- int offset, int limit, struct page ***dir)
+ int offset, int limit, struct page ***dir, int punch_hole)
{
swp_entry_t *ptr;
int freed = 0;
@@ -429,10 +429,10 @@ static int shmem_map_and_free_swp(struct
ptr = shmem_swp_map(subdir);
for (; offset < limit; offset += LATENCY_LIMIT) {
int size = limit - offset;
- if (size > LATENCY_LIMIT)
+ if (!punch_hole && size > LATENCY_LIMIT)
size = LATENCY_LIMIT;
freed += shmem_free_swp(ptr+offset, ptr+offset+size);
- if (need_resched()) {
+ if (!punch_hole && need_resched()) {
shmem_swp_unmap(ptr);
if (*dir) {
shmem_dir_unmap(*dir);
@@ -506,7 +506,8 @@ static void shmem_truncate_range(struct
nr_pages_to_free++;
list_add(&topdir->lru, &pages_to_free);
}
- spin_unlock(&info->lock);
+ if (!punch_hole)
+ spin_unlock(&info->lock);
if (info->swapped && idx < SHMEM_NR_DIRECT) {
ptr = info->i_direct;
@@ -589,14 +590,14 @@ static void shmem_truncate_range(struct
if (size > ENTRIES_PER_PAGE)
size = ENTRIES_PER_PAGE;
freed = shmem_map_and_free_swp(subdir,
- offset, size, &dir);
+ offset, size, &dir, punch_hole);
if (!dir)
dir = shmem_dir_map(middir);
nr_swaps_freed += freed;
- if (offset)
+ if (offset && !punch_hole)
spin_lock(&info->lock);
set_page_private(subdir, page_private(subdir) - freed);
- if (offset)
+ if (offset && !punch_hole)
spin_unlock(&info->lock);
if (!punch_hole)
BUG_ON(page_private(subdir) > offset);
@@ -613,6 +614,8 @@ static void shmem_truncate_range(struct
done1:
shmem_dir_unmap(dir);
done2:
+ if (punch_hole)
+ spin_unlock(&info->lock);
if (inode->i_mapping->nrpages && (info->flags & SHMEM_PAGEIN)) {
/*
* Call truncate_inode_pages again: racing shmem_unuse_inode
_
Patches currently in -mm which might be from mszeredi@xxxxxxx are
shmem-fix-bug-in-shmem_writepage.patch
shmem-dont-release-lock-for-hole-punching.patch
fix-quadratic-behavior-of-shrink_dcache_parent.patch
mm-shrink-parent-dentries-when-shrinking-slab.patch
add-filesystem-subtype-support.patch
consolidate-generic_writepages-and-mpage_writepages.patch
-
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
