* [PATCH 09/12] mm, thp, tmpfs: huge page support in shmem_fallocate
@ 2013-10-15 0:13 Ning Qu
0 siblings, 0 replies; only message in thread
From: Ning Qu @ 2013-10-15 0:13 UTC (permalink / raw)
To: Andrea Arcangeli, Andrew Morton, Kirill A. Shutemov, Hugh Dickins
Cc: Al Viro, Hugh Dickins, Wu Fengguang, Jan Kara, Mel Gorman,
linux-mm, Andi Kleen, Matthew Wilcox, Hillf Danton, Dave Hansen,
Alexander Shishkin, linux-fsdevel, linux-kernel, Ning Qu
Try to allocate huge page if the range fits, otherwise,
fall back to small pages.
Signed-off-by: Ning Qu <quning@gmail.com>
---
mm/shmem.c | 27 +++++++++++++++++++++++----
1 file changed, 23 insertions(+), 4 deletions(-)
diff --git a/mm/shmem.c b/mm/shmem.c
index 7065ae5..cbf01ce 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2179,8 +2179,11 @@ 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
@@ -2192,8 +2195,16 @@ 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;
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE
+ if ((index == (index & ~HPAGE_CACHE_INDEX_MASK)) &&
+ (index != (end & ~HPAGE_CACHE_INDEX_MASK)))
+ flags |= AOP_FLAG_TRANSHUGE;
+#endif
+
error = shmem_getpage(inode, index, &page, SGP_FALLOC,
- gfp, 0, NULL);
+ gfp, flags, NULL);
}
if (error) {
/* Remove the !PageUptodate pages we added */
@@ -2203,13 +2214,18 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
goto undone;
}
+ nr = hpagecache_nr_pages(page);
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE
+ if (PageTransHugeCache(page))
+ index &= ~HPAGE_CACHE_INDEX_MASK;
+#endif
/*
* 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
@@ -2222,6 +2238,7 @@ 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)
@@ -2232,7 +2249,9 @@ 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
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2013-10-15 0:13 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-10-15 0:13 [PATCH 09/12] mm, thp, tmpfs: huge page support in shmem_fallocate Ning Qu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).