We always go through i_size_read(), and we rarely end up needing it. Push
the read to down where we need to check it, which avoids it for most
cases.
It looks like we can even remove this check entirely, which might be
worth pursuing. But at least this takes it out of the hot path.
Acked-by: Chris Mason <clm@xxxxxx>
Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
---
I came across this and wrote the patch the other day, then Pavel pointed
me at his original posting of a very similar patch back in August.
Discussed it with Chris, and it sure _seems_ like this would be fine.
In an attempt to move the original discussion forward, here's this
posting.
diff --git a/mm/filemap.c b/mm/filemap.c
index 44b4b551e430..850920276846 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2736,9 +2736,7 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
- loff_t size;
- size = i_size_read(inode);
if (iocb->ki_flags & IOCB_NOWAIT) {
if (filemap_range_needs_writeback(mapping, iocb->ki_pos,
iocb->ki_pos + count - 1))
@@ -2770,8 +2768,9 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
* the rest of the read. Buffered reads will not work for
* DAX files, so don't bother trying.
*/
- if (retval < 0 || !count || iocb->ki_pos >= size ||
- IS_DAX(inode))
+ if (retval < 0 || !count || IS_DAX(inode))
+ return retval;
+ if (iocb->ki_pos >= i_size_read(inode))
return retval;
}
--
Jens Axboe
