ocfs2: Negotiate locking protocol versions.

Currently, when ocfs2 nodes connect via TCP, they advertise their
compatibility level.  If the versions do not match, two nodes cannot speak
to each other and they disconnect. As a result, this provides no forward or
backwards compatibility.

This patch implements a simple protocol negotiation at the dlm level by
introducing a major/minor version number scheme for entities that
communicate.  Specifically, o2dlm has a major/minor version for interaction
with o2dlm on other nodes, and ocfs2 itself has a major/minor version for
interacting with the filesystem on other nodes.

This will allow rolling upgrades of ocfs2 clusters when changes to the
locking or network protocols can be done in a backwards compatible manner.
In those cases, only the minor number is changed and the negotatied protocol
minor is returned from dlm join. In the far less likely event that a
required protocol change makes backwards compatibility impossible, we simply
bump the major number.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>

fs/ocfs2/cluster/tcp_internal.h

@@ -38,6 +38,15 @@
  * locking semantics of the file system using the protocol.  It should 
  * be somewhere else, I'm sure, but right now it isn't.
  *
+ * With version 11, we separate out the filesystem locking portion.  The
+ * filesystem now has a major.minor version it negotiates.  Version 11
+ * introduces this negotiation to the o2dlm protocol, and as such the
+ * version here in tcp_internal.h should not need to be bumped for
+ * filesystem locking changes.
+ *
+ * New in version 11
+ * 	- Negotiation of filesystem locking in the dlm join.
+ *
  * New in version 10:
  * 	- Meta/data locks combined
  *
@@ -66,7 +75,7 @@
  * 	- full 64 bit i_size in the metadata lock lvbs
  * 	- introduction of "rw" lock and pushing meta/data locking down
  */
-#define O2NET_PROTOCOL_VERSION 10ULL
+#define O2NET_PROTOCOL_VERSION 11ULL
 struct o2net_handshake {
 	__be64	protocol_version;
 	__be64	connector_id;

fs/ocfs2/dlm/dlmapi.h

@@ -193,7 +193,12 @@ enum dlm_status dlmunlock(struct dlm_ctxt *dlm,
 			  dlm_astunlockfunc_t *unlockast,
 			  void *data);
 
-struct dlm_ctxt * dlm_register_domain(const char *domain, u32 key);
+struct dlm_protocol_version {
+	u8 pv_major;
+	u8 pv_minor;
+};
+struct dlm_ctxt * dlm_register_domain(const char *domain, u32 key,
+				      struct dlm_protocol_version *fs_proto);
 
 void dlm_unregister_domain(struct dlm_ctxt *dlm);
 

fs/ocfs2/dlm/dlmcommon.h

@@ -142,6 +142,12 @@ struct dlm_ctxt
 	spinlock_t work_lock;
 	struct list_head dlm_domain_handlers;
 	struct list_head	dlm_eviction_callbacks;
+
+	/* The filesystem specifies this at domain registration.  We
+	 * cache it here to know what to tell other nodes. */
+	struct dlm_protocol_version fs_locking_proto;
+	/* This is the inter-dlm communication version */
+	struct dlm_protocol_version dlm_locking_proto;
 };
 
 static inline struct hlist_head *dlm_lockres_hash(struct dlm_ctxt *dlm, unsigned i)
@@ -589,10 +595,24 @@ struct dlm_proxy_ast
 #define DLM_PROXY_AST_MAX_LEN  (sizeof(struct dlm_proxy_ast)+DLM_LVB_LEN)
 
 #define DLM_MOD_KEY (0x666c6172)
-enum dlm_query_join_response {
+enum dlm_query_join_response_code {
 	JOIN_DISALLOW = 0,
 	JOIN_OK,
 	JOIN_OK_NO_MAP,
+	JOIN_PROTOCOL_MISMATCH,
+};
+
+union dlm_query_join_response {
+	u32 intval;
+	struct {
+		u8 code;	/* Response code.  dlm_minor and fs_minor
+				   are only valid if this is JOIN_OK */
+		u8 dlm_minor;	/* The minor version of the protocol the
+				   dlm is speaking. */
+		u8 fs_minor;	/* The minor version of the protocol the
+				   filesystem is speaking. */
+		u8 reserved;
+	} packet;
 };
 
 struct dlm_lock_request
@@ -633,6 +653,8 @@ struct dlm_query_join_request
 	u8 node_idx;
 	u8 pad1[2];
 	u8 name_len;
+	struct dlm_protocol_version dlm_proto;
+	struct dlm_protocol_version fs_proto;
 	u8 domain[O2NM_MAX_NAME_LEN];
 	u8 node_map[BITS_TO_BYTES(O2NM_MAX_NODES)];
 };

fs/ocfs2/dlm/dlmfs.c

@@ -60,6 +60,8 @@
 #define MLOG_MASK_PREFIX ML_DLMFS
 #include "cluster/masklog.h"
 
+#include "ocfs2_lockingver.h"
+
 static const struct super_operations dlmfs_ops;
 static const struct file_operations dlmfs_file_operations;
 static const struct inode_operations dlmfs_dir_inode_operations;
@@ -69,6 +71,16 @@ static struct kmem_cache *dlmfs_inode_cache;
 
 struct workqueue_struct *user_dlm_worker;
 
+/*
+ * This is the userdlmfs locking protocol version.
+ *
+ * See fs/ocfs2/dlmglue.c for more details on locking versions.
+ */
+static const struct dlm_protocol_version user_locking_protocol = {
+	.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
+	.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
+};
+
 /*
  * decodes a set of open flags into a valid lock level and a set of flags.
  * returns < 0 if we have invalid flags
@@ -416,6 +428,7 @@ static int dlmfs_mkdir(struct inode * dir,
 	struct qstr *domain = &dentry->d_name;
 	struct dlmfs_inode_private *ip;
 	struct dlm_ctxt *dlm;
+	struct dlm_protocol_version proto = user_locking_protocol;
 
 	mlog(0, "mkdir %.*s\n", domain->len, domain->name);
 
@@ -435,7 +448,7 @@ static int dlmfs_mkdir(struct inode * dir,
 
 	ip = DLMFS_I(inode);
 
-	dlm = user_dlm_register_context(domain);
+	dlm = user_dlm_register_context(domain, &proto);
 	if (IS_ERR(dlm)) {
 		status = PTR_ERR(dlm);
 		mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n",

fs/ocfs2/dlm/userdlm.c

@@ -645,7 +645,8 @@ bail:
 	return status;
 }
 
-struct dlm_ctxt *user_dlm_register_context(struct qstr *name)
+struct dlm_ctxt *user_dlm_register_context(struct qstr *name,
+					   struct dlm_protocol_version *proto)
 {
 	struct dlm_ctxt *dlm;
 	u32 dlm_key;
@@ -661,7 +662,7 @@ struct dlm_ctxt *user_dlm_register_context(struct qstr *name)
 
 	snprintf(domain, name->len + 1, "%.*s", name->len, name->name);
 
-	dlm = dlm_register_domain(domain, dlm_key);
+	dlm = dlm_register_domain(domain, dlm_key, proto);
 	if (IS_ERR(dlm))
 		mlog_errno(PTR_ERR(dlm));
 

fs/ocfs2/dlm/userdlm.h

@@ -83,7 +83,8 @@ void user_dlm_write_lvb(struct inode *inode,
 void user_dlm_read_lvb(struct inode *inode,
 		       char *val,
 		       unsigned int len);
-struct dlm_ctxt *user_dlm_register_context(struct qstr *name);
+struct dlm_ctxt *user_dlm_register_context(struct qstr *name,
+					   struct dlm_protocol_version *proto);
 void user_dlm_unregister_context(struct dlm_ctxt *dlm);
 
 struct dlmfs_inode_private {

fs/ocfs2/dlmglue.c

@@ -43,6 +43,7 @@
 #include <cluster/masklog.h>
 
 #include "ocfs2.h"
+#include "ocfs2_lockingver.h"
 
 #include "alloc.h"
 #include "dcache.h"
@@ -258,6 +259,31 @@ static struct ocfs2_lock_res_ops ocfs2_flock_lops = {
 	.flags		= 0,
 };
 
+/*
+ * This is the filesystem locking protocol version.
+ *
+ * Whenever the filesystem does new things with locks (adds or removes a
+ * lock, orders them differently, does different things underneath a lock),
+ * the version must be changed.  The protocol is negotiated when joining
+ * the dlm domain.  A node may join the domain if its major version is
+ * identical to all other nodes and its minor version is greater than
+ * or equal to all other nodes.  When its minor version is greater than
+ * the other nodes, it will run at the minor version specified by the
+ * other nodes.
+ *
+ * If a locking change is made that will not be compatible with older
+ * versions, the major number must be increased and the minor version set
+ * to zero.  If a change merely adds a behavior that can be disabled when
+ * speaking to older versions, the minor version must be increased.  If a
+ * change adds a fully backwards compatible change (eg, LVB changes that
+ * are just ignored by older versions), the version does not need to be
+ * updated.
+ */
+const struct dlm_protocol_version ocfs2_locking_protocol = {
+	.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
+	.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
+};
+
 static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
 {
 	return lockres->l_type == OCFS2_LOCK_TYPE_META ||
@@ -2506,7 +2532,8 @@ int ocfs2_dlm_init(struct ocfs2_super *osb)
 	dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str));
 
 	/* for now, uuid == domain */
-	dlm = dlm_register_domain(osb->uuid_str, dlm_key);
+	dlm = dlm_register_domain(osb->uuid_str, dlm_key,
+				  &osb->osb_locking_proto);
 	if (IS_ERR(dlm)) {
 		status = PTR_ERR(dlm);
 		mlog_errno(status);

fs/ocfs2/dlmglue.h

@@ -116,4 +116,5 @@ void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb);
 struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void);
 void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug);
 
+extern const struct dlm_protocol_version ocfs2_locking_protocol;
 #endif	/* DLMGLUE_H */

fs/ocfs2/ocfs2.h

@@ -251,6 +251,7 @@ struct ocfs2_super
 	struct ocfs2_lock_res osb_rename_lockres;
 	struct dlm_eviction_cb osb_eviction_cb;
 	struct ocfs2_dlm_debug *osb_dlm_debug;
+	struct dlm_protocol_version osb_locking_proto;
 
 	struct dentry *osb_debug_root;
 

fs/ocfs2/ocfs2_lockingver.h

+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_lockingver.h
+ *
+ * Defines OCFS2 Locking version values.
+ *
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#ifndef OCFS2_LOCKINGVER_H
+#define OCFS2_LOCKINGVER_H
+
+/*
+ * The protocol version for ocfs2 cluster locking.  See dlmglue.c for
+ * more details.
+ */
+#define OCFS2_LOCKING_PROTOCOL_MAJOR 1
+#define OCFS2_LOCKING_PROTOCOL_MINOR 0
+
+#endif  /* OCFS2_LOCKINGVER_H */

fs/ocfs2/super.c

@@ -1355,6 +1355,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
 	sb->s_fs_info = osb;
 	sb->s_op = &ocfs2_sops;
 	sb->s_export_op = &ocfs2_export_ops;
+	osb->osb_locking_proto = ocfs2_locking_protocol;
 	sb->s_time_gran = 1;
 	sb->s_flags |= MS_NOATIME;
 	/* this is needed to support O_LARGEFILE */