ocfs2: Add a mount option "coherency=*" to handle cluster coherency for O_DIRECT writes.

Currently, the default behavior of O_DIRECT writes was allowing concurrent writing among nodes to the same file, with no cluster coherency guaranteed (no EX lock held). This can leave stale data in the cache for buffered reads on other nodes. The new mount option introduce a chance to choose two different behaviors for O_DIRECT writes: * coherency=full, as the default value, will disallow concurrent O_DIRECT writes by taking EX locks. * coherency=buffered, allow concurrent O_DIRECT writes without EX lock among nodes, which gains high performance at risk of getting stale data on other nodes. Signed-off-by: Tristan Ye <tristan.ye@oracle.com> Signed-off-by: Joel Becker <joel.becker@oracle.com>

ocfs2: Add a mount option "coherency=*" to handle cluster coherency for O_DIRECT writes.
Currently, the default behavior of O_DIRECT writes was allowing concurrent writing among nodes to the same file, with no cluster coherency guaranteed (no EX lock held). This can leave stale data in the cache for buffered reads on other nodes. The new mount option introduce a chance to choose two different behaviors for O_DIRECT writes: * coherency=full, as the default value, will disallow concurrent O_DIRECT writes by taking EX locks. * coherency=buffered, allow concurrent O_DIRECT writes without EX lock among nodes, which gains high performance at risk of getting stale data on other nodes. Signed-off-by: Tristan Ye <tristan.ye@oracle.com> Signed-off-by: Joel Becker <joel.becker@oracle.com>
Tristan Ye · Joel Becker
1 parent 75d9bbc738
Showing 4 changed files with 52 additions and 2 deletions Side-by-side Diff
Documentation/filesystems/ocfs2.txt
fs/ocfs2/file.c
fs/ocfs2/ocfs2.h
fs/ocfs2/super.c
@@ -87,4 +87,11 @@
 			reservations - users should rarely need to change this
 			value. If allocation reservations are turned off, this
 			option will have no effect.
+coherency=full  (*)	Disallow concurrent O_DIRECT writes, cluster inode
+			lock will be taken to force other nodes drop cache,
+			therefore full cluster coherency is guaranteed even
+			for O_DIRECT writes.
+coherency=buffered	Allow concurrent O_DIRECT writes without EX lock among
+			nodes, which gains high performance at risk of getting
+			stale data on other nodes.
@@ -2225,6 +2225,8 @@
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	int full_coherency = !(osb->s_mount_opt &
+			       OCFS2_MOUNT_COHERENCY_BUFFERED);
  
 	mlog_entry("(0x%p, %u, '%.*s')\n", file,
 		   (unsigned int)nr_segs,
  
@@ -2248,12 +2250,35 @@
 		have_alloc_sem = 1;
 	}
  
-	/* concurrent O_DIRECT writes are allowed */
-	rw_level = !direct_io;
+	/*
+	 * Concurrent O_DIRECT writes are allowed with
+	 * mount_option "coherency=buffered".
+	 */
+	rw_level = (!direct_io || full_coherency);
+
 	ret = ocfs2_rw_lock(inode, rw_level);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto out_sems;
+	}
+
+	/*
+	 * O_DIRECT writes with "coherency=full" need to take EX cluster
+	 * inode_lock to guarantee coherency.
+	 */
+	if (direct_io && full_coherency) {
+		/*
+		 * We need to take and drop the inode lock to force
+		 * other nodes to drop their caches.  Buffered I/O
+		 * already does this in write_begin().
+		 */
+		ret = ocfs2_inode_lock(inode, NULL, 1);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out_sems;
+		}
+
+		ocfs2_inode_unlock(inode, 1);
 	}
  
 	can_do_direct = direct_io;
@@ -263,6 +263,9 @@
 						   control lists */
 	OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
 	OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
+
+	OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12 /* Allow concurrent O_DIRECT
+						    writes */
 };
  
 #define OCFS2_OSB_SOFT_RO			0x0001
@@ -177,6 +177,8 @@
 	Opt_noacl,
 	Opt_usrquota,
 	Opt_grpquota,
+	Opt_coherency_buffered,
+	Opt_coherency_full,
 	Opt_resv_level,
 	Opt_dir_resv_level,
 	Opt_err,
@@ -205,6 +207,8 @@
 	{Opt_noacl, "noacl"},
 	{Opt_usrquota, "usrquota"},
 	{Opt_grpquota, "grpquota"},
+	{Opt_coherency_buffered, "coherency=buffered"},
+	{Opt_coherency_full, "coherency=full"},
 	{Opt_resv_level, "resv_level=%u"},
 	{Opt_dir_resv_level, "dir_resv_level=%u"},
 	{Opt_err, NULL}
@@ -1452,6 +1456,12 @@
 		case Opt_grpquota:
 			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
 			break;
+		case Opt_coherency_buffered:
+			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
+			break;
+		case Opt_coherency_full:
+			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
+			break;
 		case Opt_acl:
 			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
 			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
@@ -1549,6 +1559,11 @@
 		seq_printf(s, ",usrquota");
 	if (opts & OCFS2_MOUNT_GRPQUOTA)
 		seq_printf(s, ",grpquota");
+
+	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
+		seq_printf(s, ",coherency=buffered");
+	else
+		seq_printf(s, ",coherency=full");
  
 	if (opts & OCFS2_MOUNT_NOUSERXATTR)
 		seq_printf(s, ",nouser_xattr");
...	...	@@ -87,4 +87,11 @@
87	87	reservations - users should rarely need to change this
88	88	value. If allocation reservations are turned off, this
89	89	option will have no effect.
	90	+coherency=full (*) Disallow concurrent O_DIRECT writes, cluster inode
	91	+ lock will be taken to force other nodes drop cache,
	92	+ therefore full cluster coherency is guaranteed even
	93	+ for O_DIRECT writes.
	94	+coherency=buffered Allow concurrent O_DIRECT writes without EX lock among
	95	+ nodes, which gains high performance at risk of getting
	96	+ stale data on other nodes.
...	...	@@ -2225,6 +2225,8 @@
2225	2225	struct file *file = iocb->ki_filp;
2226	2226	struct inode *inode = file->f_path.dentry->d_inode;
2227	2227	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	2228	+ int full_coherency = !(osb->s_mount_opt &
	2229	+ OCFS2_MOUNT_COHERENCY_BUFFERED);
2228	2230
2229	2231	mlog_entry("(0x%p, %u, '%.*s')\n", file,
2230	2232	(unsigned int)nr_segs,
2231	2233
...	...	@@ -2248,12 +2250,35 @@
2248	2250	have_alloc_sem = 1;
2249	2251	}
2250	2252
2251		- /* concurrent O_DIRECT writes are allowed */
2252		- rw_level = !direct_io;
	2253	+ /*
	2254	+ * Concurrent O_DIRECT writes are allowed with
	2255	+ * mount_option "coherency=buffered".
	2256	+ */
	2257	+ rw_level = (!direct_io \|\| full_coherency);
	2258	+
2253	2259	ret = ocfs2_rw_lock(inode, rw_level);
2254	2260	if (ret < 0) {
2255	2261	mlog_errno(ret);
2256	2262	goto out_sems;
	2263	+ }
	2264	+
	2265	+ /*
	2266	+ * O_DIRECT writes with "coherency=full" need to take EX cluster
	2267	+ * inode_lock to guarantee coherency.
	2268	+ */
	2269	+ if (direct_io && full_coherency) {
	2270	+ /*
	2271	+ * We need to take and drop the inode lock to force
	2272	+ * other nodes to drop their caches. Buffered I/O
	2273	+ * already does this in write_begin().
	2274	+ */
	2275	+ ret = ocfs2_inode_lock(inode, NULL, 1);
	2276	+ if (ret < 0) {
	2277	+ mlog_errno(ret);
	2278	+ goto out_sems;
	2279	+ }
	2280	+
	2281	+ ocfs2_inode_unlock(inode, 1);
2257	2282	}
2258	2283
2259	2284	can_do_direct = direct_io;
...	...	@@ -263,6 +263,9 @@
263	263	control lists */
264	264	OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
265	265	OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
	266	+
	267	+ OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12 /* Allow concurrent O_DIRECT
	268	+ writes */
266	269	};
267	270
268	271	#define OCFS2_OSB_SOFT_RO 0x0001
...	...	@@ -177,6 +177,8 @@
177	177	Opt_noacl,
178	178	Opt_usrquota,
179	179	Opt_grpquota,
	180	+ Opt_coherency_buffered,
	181	+ Opt_coherency_full,
180	182	Opt_resv_level,
181	183	Opt_dir_resv_level,
182	184	Opt_err,
...	...	@@ -205,6 +207,8 @@
205	207	{Opt_noacl, "noacl"},
206	208	{Opt_usrquota, "usrquota"},
207	209	{Opt_grpquota, "grpquota"},
	210	+ {Opt_coherency_buffered, "coherency=buffered"},
	211	+ {Opt_coherency_full, "coherency=full"},
208	212	{Opt_resv_level, "resv_level=%u"},
209	213	{Opt_dir_resv_level, "dir_resv_level=%u"},
210	214	{Opt_err, NULL}
...	...	@@ -1452,6 +1456,12 @@
1452	1456	case Opt_grpquota:
1453	1457	mopt->mount_opt \|= OCFS2_MOUNT_GRPQUOTA;
1454	1458	break;
	1459	+ case Opt_coherency_buffered:
	1460	+ mopt->mount_opt \|= OCFS2_MOUNT_COHERENCY_BUFFERED;
	1461	+ break;
	1462	+ case Opt_coherency_full:
	1463	+ mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
	1464	+ break;
1455	1465	case Opt_acl:
1456	1466	mopt->mount_opt \|= OCFS2_MOUNT_POSIX_ACL;
1457	1467	mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
...	...	@@ -1549,6 +1559,11 @@
1549	1559	seq_printf(s, ",usrquota");
1550	1560	if (opts & OCFS2_MOUNT_GRPQUOTA)
1551	1561	seq_printf(s, ",grpquota");
	1562	+
	1563	+ if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
	1564	+ seq_printf(s, ",coherency=buffered");
	1565	+ else
	1566	+ seq_printf(s, ",coherency=full");
1552	1567
1553	1568	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1554	1569	seq_printf(s, ",nouser_xattr");