aboutsummaryrefslogtreecommitdiff
path: root/fs/ocfs2/aops.c
diff options
context:
space:
mode:
authorMark Fasheh <mark.fasheh@oracle.com>2006-05-05 19:04:03 -0700
committerMark Fasheh <mark.fasheh@oracle.com>2006-05-17 14:38:47 -0700
commit53013cba4118a5cfe8f7c7ea5e5bc1c48b160f76 (patch)
tree5170ed12fbe07b5e8557e61952aa27c25034bd7a /fs/ocfs2/aops.c
parent0c056c50a6218e0e577817c16ba8851af593d742 (diff)
ocfs2: take data locks around extend
We need to take a data lock around extends to protect the pages that ocfs2_zero_extend is going to be pulling into the page cache. Otherwise an extend on one node might populate the page cache with data pages that have no lock coverage. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Diffstat (limited to 'fs/ocfs2/aops.c')
-rw-r--r--fs/ocfs2/aops.c46
1 files changed, 39 insertions, 7 deletions
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 0d858d0b25b..47152bf9a7f 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -276,13 +276,29 @@ static int ocfs2_writepage(struct page *page, struct writeback_control *wbc)
return ret;
}
+/* This can also be called from ocfs2_write_zero_page() which has done
+ * it's own cluster locking. */
+int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page,
+ unsigned from, unsigned to)
+{
+ int ret;
+
+ down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+ ret = block_prepare_write(page, from, to, ocfs2_get_block);
+
+ up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+ return ret;
+}
+
/*
* ocfs2_prepare_write() can be an outer-most ocfs2 call when it is called
* from loopback. It must be able to perform its own locking around
* ocfs2_get_block().
*/
-int ocfs2_prepare_write(struct file *file, struct page *page,
- unsigned from, unsigned to)
+static int ocfs2_prepare_write(struct file *file, struct page *page,
+ unsigned from, unsigned to)
{
struct inode *inode = page->mapping->host;
int ret;
@@ -295,11 +311,7 @@ int ocfs2_prepare_write(struct file *file, struct page *page,
goto out;
}
- down_read(&OCFS2_I(inode)->ip_alloc_sem);
-
- ret = block_prepare_write(page, from, to, ocfs2_get_block);
-
- up_read(&OCFS2_I(inode)->ip_alloc_sem);
+ ret = ocfs2_prepare_write_nolock(inode, page, from, to);
ocfs2_meta_unlock(inode, 0);
out:
@@ -625,11 +637,31 @@ static ssize_t ocfs2_direct_IO(int rw,
int ret;
mlog_entry_void();
+
+ /*
+ * We get PR data locks even for O_DIRECT. This allows
+ * concurrent O_DIRECT I/O but doesn't let O_DIRECT with
+ * extending and buffered zeroing writes race. If they did
+ * race then the buffered zeroing could be written back after
+ * the O_DIRECT I/O. It's one thing to tell people not to mix
+ * buffered and O_DIRECT writes, but expecting them to
+ * understand that file extension is also an implicit buffered
+ * write is too much. By getting the PR we force writeback of
+ * the buffered zeroing before proceeding.
+ */
+ ret = ocfs2_data_lock(inode, 0);
+ if (ret < 0) {
+ mlog_errno(ret);
+ goto out;
+ }
+ ocfs2_data_unlock(inode, 0);
+
ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
inode->i_sb->s_bdev, iov, offset,
nr_segs,
ocfs2_direct_IO_get_blocks,
ocfs2_dio_end_io);
+out:
mlog_exit(ret);
return ret;
}