[PATCHv2 10/13] mm, thp, tmpfs: huge page support in shmem_fallocate

From: Ning Qu
Date: Mon Oct 21 2013 - 17:48:28 EST


Try to allocate huge page if the range fits, otherwise,
fall back to small pages.

Signed-off-by: Ning Qu <quning@xxxxxxxxx>
---
mm/shmem.c | 24 ++++++++++++++++++++----
1 file changed, 20 insertions(+), 4 deletions(-)

diff --git a/mm/shmem.c b/mm/shmem.c
index 1764a29..48b1d84 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2156,8 +2156,11 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
inode->i_private = &shmem_falloc;
spin_unlock(&inode->i_lock);

- for (index = start; index < end; index++) {
+ i_split_down_read(inode);
+ index = start;
+ while (index < end) {
struct page *page;
+ int nr = 1;

/*
* Good, the fallocate(2) manpage permits EINTR: we may have
@@ -2169,8 +2172,15 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
error = -ENOMEM;
else {
gfp_t gfp = mapping_gfp_mask(inode->i_mapping);
+ int flags = 0;
+
+ if (mapping_can_have_hugepages(inode->i_mapping) &&
+ ((index == (index & ~HPAGE_CACHE_INDEX_MASK)) &&
+ (index != (end & ~HPAGE_CACHE_INDEX_MASK))))
+ flags |= AOP_FLAG_TRANSHUGE;
+
error = shmem_getpage(inode, index, &page, SGP_FALLOC,
- gfp, 0, NULL);
+ gfp, flags, NULL);
}
if (error) {
/* Remove the !PageUptodate pages we added */
@@ -2180,13 +2190,16 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
goto undone;
}

+ nr = hpagecache_nr_pages(page);
+ if (PageTransHugeCache(page))
+ index &= ~HPAGE_CACHE_INDEX_MASK;
/*
* Inform shmem_writepage() how far we have reached.
* No need for lock or barrier: we have the page lock.
*/
- shmem_falloc.next++;
+ shmem_falloc.next += nr;
if (!PageUptodate(page))
- shmem_falloc.nr_falloced++;
+ shmem_falloc.nr_falloced += nr;

/*
* If !PageUptodate, leave it that way so that freeable pages
@@ -2199,6 +2212,7 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
unlock_page(page);
page_cache_release(page);
cond_resched();
+ index += nr;
}

if (!(mode & FALLOC_FL_KEEP_SIZE) && offset + len > inode->i_size)
@@ -2209,7 +2223,9 @@ undone:
inode->i_private = NULL;
spin_unlock(&inode->i_lock);
out:
+ i_split_up_read(inode);
mutex_unlock(&inode->i_mutex);
+
return error;
}

--
1.8.4

--
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/