Support huge page in do_shmem_file_read when possible.
Still have room to improve, since we re-search the page in
page cache everytime, but for huge page, we might save some
searches and reuse the huge page for the next read across
page boundary.
Signed-off-by: Ning Qu <quning@xxxxxxxxx>
---
mm/shmem.c | 52 ++++++++++++++++++++++++++++++++++++++--------------
1 file changed, 38 insertions(+), 14 deletions(-)
diff --git a/mm/shmem.c b/mm/shmem.c
index 90f2e0e..7065ae5 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -1765,13 +1765,25 @@ shmem_write_end(struct file *file, struct address_space *mapping,
return copied;
}
+static unsigned long page_cache_to_mask(struct page *page)
+{
+ if (page && PageTransHugeCache(page))
+ return HPAGE_PMD_MASK;
+ else
+ return PAGE_CACHE_MASK;
+}
+
+static unsigned long pos_to_off(struct page *page, loff_t pos)
+{
+ return pos & ~page_cache_to_mask(page);
+}
+
static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_t *desc, read_actor_t actor)
{
struct inode *inode = file_inode(filp);
gfp_t gfp = mapping_gfp_mask(inode->i_mapping);
struct address_space *mapping = inode->i_mapping;
pgoff_t index;
- unsigned long offset;
enum sgp_type sgp = SGP_READ;
/*
@@ -1783,25 +1795,29 @@ static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_
sgp = SGP_DIRTY;
index = *ppos >> PAGE_CACHE_SHIFT;
- offset = *ppos & ~PAGE_CACHE_MASK;
+ i_split_down_read(inode);
for (;;) {
struct page *page = NULL;
pgoff_t end_index;
unsigned long nr, ret;
loff_t i_size = i_size_read(inode);
+ int flags = 0;
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE
+ flags |= AOP_FLAG_TRANSHUGE;
+#endif
end_index = i_size >> PAGE_CACHE_SHIFT;
if (index > end_index)
break;
if (index == end_index) {
nr = i_size & ~PAGE_CACHE_MASK;
- if (nr <= offset)
+ if (nr <= pos_to_off(page, *ppos))
break;
}
desc->error = shmem_getpage(inode, index, &page, sgp, gfp,
- 0, NULL);
+ flags, NULL);
if (desc->error) {
if (desc->error == -EINVAL)
desc->error = 0;
@@ -1814,18 +1830,27 @@ static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_
* We must evaluate after, since reads (unlike writes)
* are called without i_mutex protection against truncate
*/
- nr = PAGE_CACHE_SIZE;
i_size = i_size_read(inode);
end_index = i_size >> PAGE_CACHE_SHIFT;
+
+ nr = PAGE_CACHE_SIZE;
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE
+ if (page && PageTransHugeCache(page)) {
+ index &= ~HPAGE_CACHE_INDEX_MASK;
+ end_index &= ~HPAGE_CACHE_INDEX_MASK;
+ nr = PAGE_CACHE_SIZE << compound_order(page);
+ }
+#endif
+
if (index == end_index) {
- nr = i_size & ~PAGE_CACHE_MASK;
- if (nr <= offset) {
+ nr = ((i_size - 1) & ~page_cache_to_mask(page)) + 1;
+ if (nr <= pos_to_off(page, *ppos)) {
if (page)
page_cache_release(page);
break;
}
}
- nr -= offset;
+ nr = nr - pos_to_off(page, *ppos);
if (page) {
/*
@@ -1838,7 +1863,7 @@ static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_
/*
* Mark the page accessed if we read the beginning.
*/
- if (!offset)
+ if (!pos_to_off(page, *ppos))
mark_page_accessed(page);
} else {
page = ZERO_PAGE(0);
@@ -1855,10 +1880,9 @@ static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_
* "pos" here (the actor routine has to update the user buffer
* pointers and the remaining count).
*/
- ret = actor(desc, page, offset, nr);
- offset += ret;
- index += offset >> PAGE_CACHE_SHIFT;
- offset &= ~PAGE_CACHE_MASK;
+ ret = actor(desc, page, pos_to_off(page, *ppos), nr);
+ *ppos += ret;
+ index = *ppos >> PAGE_CACHE_SHIFT;
page_cache_release(page);
if (ret != nr || !desc->count)
@@ -1867,7 +1891,7 @@ static void do_shmem_file_read(struct file *filp, loff_t *ppos, read_descriptor_
cond_resched();
}
- *ppos = ((loff_t) index << PAGE_CACHE_SHIFT) + offset;
+ i_split_up_read(inode);
file_accessed(filp);
}
--
1.8.4
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>