ocfs2: Commit journal on sync writes

We're missing a meta data commit for extending sync writes. In thoery, write could return with the meta data required to read the data uncommitted to disk. Fix that by detecting an allocating write and forcing a journal commit in the sync case. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>

ocfs2: Commit journal on sync writes
We're missing a meta data commit for extending sync writes. In thoery, write could return with the meta data required to read the data uncommitted to disk. Fix that by detecting an allocating write and forcing a journal commit in the sync case. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Mark Fasheh
1 parent 9f70968af3
Showing 1 changed file with 25 additions and 1 deletions Side-by-side Diff
fs/ocfs2/file.c
@@ -1891,9 +1891,11 @@
 	ssize_t written = 0;
 	size_t ocount;		/* original count */
 	size_t count;		/* after file limit checks */
-	loff_t *ppos = &iocb->ki_pos;
+	loff_t old_size, *ppos = &iocb->ki_pos;
+	u32 old_clusters;
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_path.dentry->d_inode;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  
 	mlog_entry("(0x%p, %u, '%.*s')\n", file,
 		   (unsigned int)nr_segs,
@@ -1949,6 +1951,13 @@
 		goto relock;
 	}
  
+	/*
+	 * To later detect whether a journal commit for sync writes is
+	 * necessary, we sample i_size, and cluster count here.
+	 */
+	old_size = i_size_read(inode);
+	old_clusters = OCFS2_I(inode)->ip_clusters;
+
 	/* communicate with ocfs2_dio_end_io */
 	ocfs2_iocb_set_rw_locked(iocb, rw_level);
  
@@ -1977,6 +1986,21 @@
 out_dio:
 	/* buffered aio wouldn't have proper lock coverage today */
 	BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT));
+
+	if ((file->f_flags & O_SYNC && !direct_io) || IS_SYNC(inode)) {
+		/*
+		 * The generic write paths have handled getting data
+		 * to disk, but since we don't make use of the dirty
+		 * inode list, a manual journal commit is necessary
+		 * here.
+		 */
+		if (old_size != i_size_read(inode) ||
+		    old_clusters != OCFS2_I(inode)->ip_clusters) {
+			ret = journal_force_commit(osb->journal->j_journal);
+			if (ret < 0)
+				written = ret;
+		}
+	}
  
 	/* 
 	 * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io
...	...	@@ -1891,9 +1891,11 @@
1891	1891	ssize_t written = 0;
1892	1892	size_t ocount; /* original count */
1893	1893	size_t count; /* after file limit checks */
1894		- loff_t *ppos = &iocb->ki_pos;
	1894	+ loff_t old_size, *ppos = &iocb->ki_pos;
	1895	+ u32 old_clusters;
1895	1896	struct file *file = iocb->ki_filp;
1896	1897	struct inode *inode = file->f_path.dentry->d_inode;
	1898	+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1897	1899
1898	1900	mlog_entry("(0x%p, %u, '%.*s')\n", file,
1899	1901	(unsigned int)nr_segs,
...	...	@@ -1949,6 +1951,13 @@
1949	1951	goto relock;
1950	1952	}
1951	1953
	1954	+ /*
	1955	+ * To later detect whether a journal commit for sync writes is
	1956	+ * necessary, we sample i_size, and cluster count here.
	1957	+ */
	1958	+ old_size = i_size_read(inode);
	1959	+ old_clusters = OCFS2_I(inode)->ip_clusters;
	1960	+
1952	1961	/* communicate with ocfs2_dio_end_io */
1953	1962	ocfs2_iocb_set_rw_locked(iocb, rw_level);
1954	1963
...	...	@@ -1977,6 +1986,21 @@
1977	1986	out_dio:
1978	1987	/* buffered aio wouldn't have proper lock coverage today */
1979	1988	BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT));
	1989	+
	1990	+ if ((file->f_flags & O_SYNC && !direct_io) \|\| IS_SYNC(inode)) {
	1991	+ /*
	1992	+ * The generic write paths have handled getting data
	1993	+ * to disk, but since we don't make use of the dirty
	1994	+ * inode list, a manual journal commit is necessary
	1995	+ * here.
	1996	+ */
	1997	+ if (old_size != i_size_read(inode) \|\|
	1998	+ old_clusters != OCFS2_I(inode)->ip_clusters) {
	1999	+ ret = journal_force_commit(osb->journal->j_journal);
	2000	+ if (ret < 0)
	2001	+ written = ret;
	2002	+ }
	2003	+ }
1980	2004
1981	2005	/*
1982	2006	* deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io