ocfs2: Add the USERSPACE_STACK incompat bit.

The filesystem gains the USERSPACE_STACK incomat bit and the s_cluster_info field on the superblock. When a userspace stack is in use, the name of the stack is stored on-disk for mount-time verification. The "cluster_stack" option is added to mount(2) processing. The mount process needs to pass the matching stack name. If the passed name and the on-disk name do not match, the mount is failed. When using the classic o2cb stack, the incompat bit is *not* set and no mount option is used other than the usual heartbeat=local. Thus, the filesystem is compatible with older tools. Signed-off-by: Joel Becker <joel.becker@oracle.com> Signed-off-by: Mark Fasheh <mfasheh@suse.com>

ocfs2: Add the USERSPACE_STACK incompat bit.
The filesystem gains the USERSPACE_STACK incomat bit and the s_cluster_info field on the superblock. When a userspace stack is in use, the name of the stack is stored on-disk for mount-time verification. The "cluster_stack" option is added to mount(2) processing. The mount process needs to pass the matching stack name. If the passed name and the on-disk name do not match, the mount is failed. When using the classic o2cb stack, the incompat bit is *not* set and no mount option is used other than the usual heartbeat=local. Thus, the filesystem is compatible with older tools. Signed-off-by: Joel Becker <joel.becker@oracle.com> Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Joel Becker · Mark Fasheh
1 parent 74ae4e104d
Showing 3 changed files with 134 additions and 3 deletions Side-by-side Diff
fs/ocfs2/ocfs2.h
fs/ocfs2/ocfs2_fs.h
fs/ocfs2/super.c
@@ -248,6 +248,7 @@
 	struct ocfs2_alloc_stats alloc_stats;
 	char dev_str[20];		/* "major,minor" of the device */
  
+	char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 	struct ocfs2_cluster_connection *cconn;
 	struct ocfs2_lock_res osb_super_lockres;
 	struct ocfs2_lock_res osb_rename_lockres;
@@ -366,6 +367,12 @@
 	spin_unlock(&osb->osb_lock);
  
 	return ret;
+}
+
+static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
+{
+	return (osb->s_feature_incompat &
+		OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
 }
  
 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
@@ -89,7 +89,8 @@
 #define OCFS2_FEATURE_INCOMPAT_SUPP	(OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
 					 | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
 					 | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
-					 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP)
+					 | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
+					 | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK)
 #define OCFS2_FEATURE_RO_COMPAT_SUPP	OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
  
 /*
@@ -131,6 +132,17 @@
  
  
 /*
+ * Support for alternate, userspace cluster stacks.  If set, the superblock
+ * field s_cluster_info contains a tag for the alternate stack in use as
+ * well as the name of the cluster being joined.
+ * mount.ocfs2 must pass in a matching stack name.
+ *
+ * If not set, the classic stack will be used.  This is compatbile with
+ * all older versions.
+ */
+#define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK	0x0080
+
+/*
  * backup superblock flag is used to indicate that this volume
  * has backup superblocks.
  */
@@ -272,6 +284,10 @@
 #define OCFS2_VOL_UUID_LEN		16
 #define OCFS2_MAX_VOL_LABEL_LEN		64
  
+/* The alternate, userspace stack fields */
+#define OCFS2_STACK_LABEL_LEN		4
+#define OCFS2_CLUSTER_NAME_LEN		16
+
 /* Journal limits (in bytes) */
 #define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
  
@@ -513,6 +529,13 @@
  */
 };
  
+struct ocfs2_cluster_info {
+/*00*/	__u8   ci_stack[OCFS2_STACK_LABEL_LEN];
+	__le32 ci_reserved;
+/*08*/	__u8   ci_cluster[OCFS2_CLUSTER_NAME_LEN];
+/*18*/
+};
+
 /*
  * On disk superblock for OCFS2
  * Note that it is contained inside an ocfs2_dinode, so all offsets
@@ -545,7 +568,20 @@
 					 * group header */
 /*50*/	__u8  s_label[OCFS2_MAX_VOL_LABEL_LEN];	/* Label for mounting, etc. */
 /*90*/	__u8  s_uuid[OCFS2_VOL_UUID_LEN];	/* 128-bit uuid */
-/*A0*/
+/*A0*/  struct ocfs2_cluster_info s_cluster_info; /* Selected userspace
+						     stack.  Only valid
+						     with INCOMPAT flag. */
+/*B8*/  __le64 s_reserved2[17];		/* Fill out superblock */
+/*140*/
+
+	/*
+	 * NOTE: As stated above, all offsets are relative to
+	 * ocfs2_dinode.id2, which is at 0xC0 in the inode.
+	 * 0xC0 + 0x140 = 0x200 or 512 bytes.  A superblock must fit within
+	 * our smallest blocksize, which is 512 bytes.  To ensure this,
+	 * we reserve the space in s_reserved2.  Anything past s_reserved2
+	 * will not be available on the smallest blocksize.
+	 */
 };
  
 /*
@@ -87,6 +87,7 @@
 	unsigned int	atime_quantum;
 	signed short	slot;
 	unsigned int	localalloc_opt;
+	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 };
  
 static int ocfs2_parse_options(struct super_block *sb, char *options,
@@ -152,6 +153,7 @@
 	Opt_commit,
 	Opt_localalloc,
 	Opt_localflocks,
+	Opt_stack,
 	Opt_err,
 };
  
@@ -170,6 +172,7 @@
 	{Opt_commit, "commit=%u"},
 	{Opt_localalloc, "localalloc=%d"},
 	{Opt_localflocks, "localflocks"},
+	{Opt_stack, "cluster_stack=%s"},
 	{Opt_err, NULL}
 };
  
  
@@ -549,8 +552,17 @@
 		}
 	}
  
+	if (ocfs2_userspace_stack(osb)) {
+		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
+			mlog(ML_ERROR, "Userspace stack expected, but "
+			     "o2cb heartbeat arguments passed to mount\n");
+			return -EINVAL;
+		}
+	}
+
 	if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
-		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
+		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
+		    !ocfs2_userspace_stack(osb)) {
 			mlog(ML_ERROR, "Heartbeat has to be started to mount "
 			     "a read-write clustered device.\n");
 			return -EINVAL;
@@ -560,6 +572,35 @@
 	return 0;
 }
  
+/*
+ * If we're using a userspace stack, mount should have passed
+ * a name that matches the disk.  If not, mount should not
+ * have passed a stack.
+ */
+static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
+					struct mount_options *mopt)
+{
+	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
+		mlog(ML_ERROR,
+		     "cluster stack passed to mount, but this filesystem "
+		     "does not support it\n");
+		return -EINVAL;
+	}
+
+	if (ocfs2_userspace_stack(osb) &&
+	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
+		    OCFS2_STACK_LABEL_LEN)) {
+		mlog(ML_ERROR,
+		     "cluster stack passed to mount (\"%s\") does not "
+		     "match the filesystem (\"%s\")\n",
+		     mopt->cluster_stack,
+		     osb->osb_cluster_stack);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct dentry *root;
@@ -598,6 +639,10 @@
 	osb->osb_commit_interval = parsed_options.commit_interval;
 	osb->local_alloc_size = parsed_options.localalloc_opt;
  
+	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
+	if (status)
+		goto read_super_error;
+
 	sb->s_magic = OCFS2_SUPER_MAGIC;
  
 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
@@ -752,6 +797,7 @@
 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
 	mopt->slot = OCFS2_INVALID_SLOT;
 	mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
+	mopt->cluster_stack[0] = '\0';
  
 	if (!options) {
 		status = 1;
@@ -853,6 +899,25 @@
 			if (!is_remount)
 				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
 			break;
+		case Opt_stack:
+			/* Check both that the option we were passed
+			 * is of the right length and that it is a proper
+			 * string of the right length.
+			 */
+			if (((args[0].to - args[0].from) !=
+			     OCFS2_STACK_LABEL_LEN) ||
+			    (strnlen(args[0].from,
+				     OCFS2_STACK_LABEL_LEN) !=
+			     OCFS2_STACK_LABEL_LEN)) {
+				mlog(ML_ERROR,
+				     "Invalid cluster_stack option\n");
+				status = 0;
+				goto bail;
+			}
+			memcpy(mopt->cluster_stack, args[0].from,
+			       OCFS2_STACK_LABEL_LEN);
+			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
+			break;
 		default:
 			mlog(ML_ERROR,
 			     "Unrecognized mount option \"%s\" "
@@ -911,6 +976,10 @@
 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
 		seq_printf(s, ",localflocks,");
  
+	if (osb->osb_cluster_stack[0])
+		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
+			   osb->osb_cluster_stack);
+
 	return 0;
 }
  
@@ -1401,6 +1470,25 @@
 		     "unsupported optional features (%x).\n", i);
 		status = -EINVAL;
 		goto bail;
+	}
+
+	if (ocfs2_userspace_stack(osb)) {
+		memcpy(osb->osb_cluster_stack,
+		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
+		       OCFS2_STACK_LABEL_LEN);
+		osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
+		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
+			mlog(ML_ERROR,
+			     "couldn't mount because of an invalid "
+			     "cluster stack label (%s) \n",
+			     osb->osb_cluster_stack);
+			status = -EINVAL;
+			goto bail;
+		}
+	} else {
+		/* The empty string is identical with classic tools that
+		 * don't know about s_cluster_info. */
+		osb->osb_cluster_stack[0] = '\0';
 	}
  
 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
...	...	@@ -248,6 +248,7 @@
248	248	struct ocfs2_alloc_stats alloc_stats;
249	249	char dev_str[20]; /* "major,minor" of the device */
250	250
	251	+ char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
251	252	struct ocfs2_cluster_connection *cconn;
252	253	struct ocfs2_lock_res osb_super_lockres;
253	254	struct ocfs2_lock_res osb_rename_lockres;
...	...	@@ -366,6 +367,12 @@
366	367	spin_unlock(&osb->osb_lock);
367	368
368	369	return ret;
	370	+}
	371	+
	372	+static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
	373	+{
	374	+ return (osb->s_feature_incompat &
	375	+ OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
369	376	}
370	377
371	378	static inline int ocfs2_mount_local(struct ocfs2_super *osb)
...	...	@@ -89,7 +89,8 @@
89	89	#define OCFS2_FEATURE_INCOMPAT_SUPP (OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
90	90	\| OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
91	91	\| OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
92		- \| OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP)
	92	+ \| OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
	93	+ \| OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK)
93	94	#define OCFS2_FEATURE_RO_COMPAT_SUPP OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
94	95
95	96	/*
...	...	@@ -131,6 +132,17 @@
131	132
132	133
133	134	/*
	135	+ * Support for alternate, userspace cluster stacks. If set, the superblock
	136	+ * field s_cluster_info contains a tag for the alternate stack in use as
	137	+ * well as the name of the cluster being joined.
	138	+ * mount.ocfs2 must pass in a matching stack name.
	139	+ *
	140	+ * If not set, the classic stack will be used. This is compatbile with
	141	+ * all older versions.
	142	+ */
	143	+#define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK 0x0080
	144	+
	145	+/*
134	146	* backup superblock flag is used to indicate that this volume
135	147	* has backup superblocks.
136	148	*/
...	...	@@ -272,6 +284,10 @@
272	284	#define OCFS2_VOL_UUID_LEN 16
273	285	#define OCFS2_MAX_VOL_LABEL_LEN 64
274	286
	287	+/* The alternate, userspace stack fields */
	288	+#define OCFS2_STACK_LABEL_LEN 4
	289	+#define OCFS2_CLUSTER_NAME_LEN 16
	290	+
275	291	/* Journal limits (in bytes) */
276	292	#define OCFS2_MIN_JOURNAL_SIZE (4 * 1024 * 1024)
277	293
...	...	@@ -513,6 +529,13 @@
513	529	*/
514	530	};
515	531
	532	+struct ocfs2_cluster_info {
	533	+/00/ __u8 ci_stack[OCFS2_STACK_LABEL_LEN];
	534	+ __le32 ci_reserved;
	535	+/08/ __u8 ci_cluster[OCFS2_CLUSTER_NAME_LEN];
	536	+/18/
	537	+};
	538	+
516	539	/*
517	540	* On disk superblock for OCFS2
518	541	* Note that it is contained inside an ocfs2_dinode, so all offsets
...	...	@@ -545,7 +568,20 @@
545	568	* group header */
546	569	/50/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */
547	570	/90/ __u8 s_uuid[OCFS2_VOL_UUID_LEN]; /* 128-bit uuid */
548		-/A0/
	571	+/A0/ struct ocfs2_cluster_info s_cluster_info; /* Selected userspace
	572	+ stack. Only valid
	573	+ with INCOMPAT flag. */
	574	+/B8/ __le64 s_reserved2[17]; /* Fill out superblock */
	575	+/140/
	576	+
	577	+ /*
	578	+ * NOTE: As stated above, all offsets are relative to
	579	+ * ocfs2_dinode.id2, which is at 0xC0 in the inode.
	580	+ * 0xC0 + 0x140 = 0x200 or 512 bytes. A superblock must fit within
	581	+ * our smallest blocksize, which is 512 bytes. To ensure this,
	582	+ * we reserve the space in s_reserved2. Anything past s_reserved2
	583	+ * will not be available on the smallest blocksize.
	584	+ */
549	585	};
550	586
551	587	/*
...	...	@@ -87,6 +87,7 @@
87	87	unsigned int atime_quantum;
88	88	signed short slot;
89	89	unsigned int localalloc_opt;
	90	+ char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
90	91	};
91	92
92	93	static int ocfs2_parse_options(struct super_block sb, char options,
...	...	@@ -152,6 +153,7 @@
152	153	Opt_commit,
153	154	Opt_localalloc,
154	155	Opt_localflocks,
	156	+ Opt_stack,
155	157	Opt_err,
156	158	};
157	159
...	...	@@ -170,6 +172,7 @@
170	172	{Opt_commit, "commit=%u"},
171	173	{Opt_localalloc, "localalloc=%d"},
172	174	{Opt_localflocks, "localflocks"},
	175	+ {Opt_stack, "cluster_stack=%s"},
173	176	{Opt_err, NULL}
174	177	};
175	178
176	179
...	...	@@ -549,8 +552,17 @@
549	552	}
550	553	}
551	554
	555	+ if (ocfs2_userspace_stack(osb)) {
	556	+ if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
	557	+ mlog(ML_ERROR, "Userspace stack expected, but "
	558	+ "o2cb heartbeat arguments passed to mount\n");
	559	+ return -EINVAL;
	560	+ }
	561	+ }
	562	+
552	563	if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
553		- if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
	564	+ if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
	565	+ !ocfs2_userspace_stack(osb)) {
554	566	mlog(ML_ERROR, "Heartbeat has to be started to mount "
555	567	"a read-write clustered device.\n");
556	568	return -EINVAL;
...	...	@@ -560,6 +572,35 @@
560	572	return 0;
561	573	}
562	574
	575	+/*
	576	+ * If we're using a userspace stack, mount should have passed
	577	+ * a name that matches the disk. If not, mount should not
	578	+ * have passed a stack.
	579	+ */
	580	+static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
	581	+ struct mount_options *mopt)
	582	+{
	583	+ if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
	584	+ mlog(ML_ERROR,
	585	+ "cluster stack passed to mount, but this filesystem "
	586	+ "does not support it\n");
	587	+ return -EINVAL;
	588	+ }
	589	+
	590	+ if (ocfs2_userspace_stack(osb) &&
	591	+ strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
	592	+ OCFS2_STACK_LABEL_LEN)) {
	593	+ mlog(ML_ERROR,
	594	+ "cluster stack passed to mount (\"%s\") does not "
	595	+ "match the filesystem (\"%s\")\n",
	596	+ mopt->cluster_stack,
	597	+ osb->osb_cluster_stack);
	598	+ return -EINVAL;
	599	+ }
	600	+
	601	+ return 0;
	602	+}
	603	+
563	604	static int ocfs2_fill_super(struct super_block sb, void data, int silent)
564	605	{
565	606	struct dentry *root;
...	...	@@ -598,6 +639,10 @@
598	639	osb->osb_commit_interval = parsed_options.commit_interval;
599	640	osb->local_alloc_size = parsed_options.localalloc_opt;
600	641
	642	+ status = ocfs2_verify_userspace_stack(osb, &parsed_options);
	643	+ if (status)
	644	+ goto read_super_error;
	645	+
601	646	sb->s_magic = OCFS2_SUPER_MAGIC;
602	647
603	648	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
...	...	@@ -752,6 +797,7 @@
752	797	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
753	798	mopt->slot = OCFS2_INVALID_SLOT;
754	799	mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
	800	+ mopt->cluster_stack[0] = '\0';
755	801
756	802	if (!options) {
757	803	status = 1;
...	...	@@ -853,6 +899,25 @@
853	899	if (!is_remount)
854	900	mopt->mount_opt \|= OCFS2_MOUNT_LOCALFLOCKS;
855	901	break;
	902	+ case Opt_stack:
	903	+ /* Check both that the option we were passed
	904	+ * is of the right length and that it is a proper
	905	+ * string of the right length.
	906	+ */
	907	+ if (((args[0].to - args[0].from) !=
	908	+ OCFS2_STACK_LABEL_LEN) \|\|
	909	+ (strnlen(args[0].from,
	910	+ OCFS2_STACK_LABEL_LEN) !=
	911	+ OCFS2_STACK_LABEL_LEN)) {
	912	+ mlog(ML_ERROR,
	913	+ "Invalid cluster_stack option\n");
	914	+ status = 0;
	915	+ goto bail;
	916	+ }
	917	+ memcpy(mopt->cluster_stack, args[0].from,
	918	+ OCFS2_STACK_LABEL_LEN);
	919	+ mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
	920	+ break;
856	921	default:
857	922	mlog(ML_ERROR,
858	923	"Unrecognized mount option \"%s\" "
...	...	@@ -911,6 +976,10 @@
911	976	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
912	977	seq_printf(s, ",localflocks,");
913	978
	979	+ if (osb->osb_cluster_stack[0])
	980	+ seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
	981	+ osb->osb_cluster_stack);
	982	+
914	983	return 0;
915	984	}
916	985
...	...	@@ -1401,6 +1470,25 @@
1401	1470	"unsupported optional features (%x).\n", i);
1402	1471	status = -EINVAL;
1403	1472	goto bail;
	1473	+ }
	1474	+
	1475	+ if (ocfs2_userspace_stack(osb)) {
	1476	+ memcpy(osb->osb_cluster_stack,
	1477	+ OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
	1478	+ OCFS2_STACK_LABEL_LEN);
	1479	+ osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
	1480	+ if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
	1481	+ mlog(ML_ERROR,
	1482	+ "couldn't mount because of an invalid "
	1483	+ "cluster stack label (%s) \n",
	1484	+ osb->osb_cluster_stack);
	1485	+ status = -EINVAL;
	1486	+ goto bail;
	1487	+ }
	1488	+ } else {
	1489	+ /* The empty string is identical with classic tools that
	1490	+ * don't know about s_cluster_info. */
	1491	+ osb->osb_cluster_stack[0] = '\0';
1404	1492	}
1405	1493
1406	1494	get_random_bytes(&osb->s_next_generation, sizeof(u32));