1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2006 Cluster File Systems, Inc.
5 * Author: Nathan Rutman <nathan@clusterfs.com>
7 * This file is part of Lustre, http://www.lustre.org.
9 * Lustre is free software; you can redistribute it and/or
10 * modify it under the terms of version 2 of the GNU General Public
11 * License as published by the Free Software Foundation.
13 * Lustre is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with Lustre; if not, write to the Free Software
20 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23 * Lustre disk format definitions.
26 #ifndef _LUSTRE_DISK_H
27 #define _LUSTRE_DISK_H
29 #include <lnet/types.h>
31 /****************** on-disk files *********************/
33 #define MDT_LOGS_DIR "LOGS" /* COMPAT_146 */
34 #define MOUNT_CONFIGS_DIR "CONFIGS"
35 /* Persistent mount data are stored on the disk in this file. */
36 #define MOUNT_DATA_FILE MOUNT_CONFIGS_DIR"/mountdata"
37 #define LAST_RCVD "last_received"
38 #define LOV_OBJID "lov_objid"
39 #define HEALTH_CHECK "health_check"
40 #define CAPA_KEYS "capa_keys"
42 /****************** persistent mount data *********************/
44 #define LDD_F_SV_TYPE_MDT 0x0001
45 #define LDD_F_SV_TYPE_OST 0x0002
46 #define LDD_F_SV_TYPE_MGS 0x0004
47 #define LDD_F_NEED_INDEX 0x0010 /* need an index assignment */
48 #define LDD_F_VIRGIN 0x0020 /* never registered */
49 #define LDD_F_UPDATE 0x0040 /* update the config logs for this server*/
50 #define LDD_F_REWRITE_LDD 0x0080 /* rewrite the LDD */
51 #define LDD_F_WRITECONF 0x0100 /* regenerate all logs for this fs */
52 #define LDD_F_UPGRADE14 0x0200 /* COMPAT_14 */
53 #define LDD_F_PARAM_FNID 0x0400 /* process failover nids as params */
63 static inline char *mt_str(enum ldd_mount_type mt)
65 static char *mount_type_string[] = {
71 return mount_type_string[mt];
74 #define LDD_INCOMPAT_SUPP 0
75 #define LDD_ROCOMPAT_SUPP 0
77 #define LDD_MAGIC 0x1dd00001
79 /* FIXME does on-disk ldd have to be a fixed endianness? (like last_rcvd) */
80 struct lustre_disk_data {
82 __u32 ldd_feature_compat; /* compatible feature flags */
83 __u32 ldd_feature_rocompat;/* read-only compatible feature flags */
84 __u32 ldd_feature_incompat;/* incompatible feature flags */
86 __u32 ldd_config_ver; /* config rewrite count - not used */
87 __u32 ldd_flags; /* LDD_SV_TYPE */
88 __u32 ldd_svindex; /* server index (0001), must match
90 __u32 ldd_mount_type; /* target fs type LDD_MT_* */
91 char ldd_fsname[64]; /* filesystem this server is part of */
92 char ldd_svname[64]; /* this server's name (lustre-mdt0001)*/
93 __u8 ldd_uuid[40]; /* server UUID (COMPAT_146) */
95 /*200*/ char ldd_userdata[1024 - 200]; /* arbitrary user string */
96 /*1024*/__u8 ldd_padding[4096 - 1024];
97 /*4096*/char ldd_mount_opts[4096]; /* target fs mount opts */
98 /*8192*/char ldd_params[4096]; /* key=value pairs */
101 #define IS_MDT(data) ((data)->ldd_flags & LDD_F_SV_TYPE_MDT)
102 #define IS_OST(data) ((data)->ldd_flags & LDD_F_SV_TYPE_OST)
103 #define IS_MGS(data) ((data)->ldd_flags & LDD_F_SV_TYPE_MGS)
104 #define MT_STR(data) mt_str((data)->ldd_mount_type)
106 /* Make the mdt/ost server obd name based on the filesystem name */
107 static inline int server_make_name(__u32 flags, __u16 index, char *fs,
110 if (flags & (LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_OST)) {
111 sprintf(name, "%.8s-%s%04x", fs,
112 (flags & LDD_F_SV_TYPE_MDT) ? "MDT" : "OST",
114 } else if (flags & LDD_F_SV_TYPE_MGS) {
115 sprintf(name, "MGS");
117 CERROR("unknown server type %#x\n", flags);
123 /* Get the index from the obd name */
124 int server_name2index(char *svname, __u32 *idx, char **endptr);
127 /****************** mount command *********************/
129 /* The lmd is only used internally by Lustre; mount simply passes
130 everything as string options */
132 #define LMD_MAGIC 0xbdacbd03
134 /* gleaned from the mount command - no persistent info here */
135 struct lustre_mount_data {
137 __u32 lmd_flags; /* lustre mount flags */
138 int lmd_mgs_failnodes; /* mgs failover node count */
139 int lmd_exclude_count;
140 char *lmd_dev; /* device name */
141 char *lmd_profile; /* client only */
142 char *lmd_sec_mdt; /* sec from mdt (to ost/mdt) */
143 char *lmd_sec_cli; /* sec from client (to ost/mdt) */
144 uid_t lmd_nllu; /* non-lustre-local-user id */
145 gid_t lmd_nllg; /* non-lustre-local-group id */
146 char *lmd_opts; /* lustre mount options (as opposed to
147 _device_ mount options) */
148 __u32 *lmd_exclude; /* array of OSTs to ignore */
151 #define LMD_FLG_SERVER 0x0001 /* Mounting a server */
152 #define LMD_FLG_CLIENT 0x0002 /* Mounting a client */
153 #define LMD_FLG_ABORT_RECOV 0x0008 /* Abort recovery */
154 #define LMD_FLG_NOSVC 0x0010 /* Only start MGS/MGC for servers,
157 #define lmd_is_client(x) ((x)->lmd_flags & LMD_FLG_CLIENT)
160 /****************** last_rcvd file *********************/
162 #define LR_SERVER_SIZE 512
163 #define LR_CLIENT_START 8192
164 #define LR_CLIENT_SIZE 128
165 #if LR_CLIENT_START < LR_SERVER_SIZE
166 #error "Can't have LR_CLIENT_START < LR_SERVER_SIZE"
168 /* This limit is arbitrary (32k clients on x86), but it is convenient to use
169 * 2^n * PAGE_SIZE * 8 for the number of bits that fit an order-n allocation. */
170 #define LR_MAX_CLIENTS (PAGE_SIZE * 8)
174 #define OBD_COMPAT_OST 0x00000002 /* this is an OST (temporary) */
175 #define OBD_COMPAT_MDT 0x00000004 /* this is an MDT (temporary) */
178 #define OBD_ROCOMPAT_LOVOBJID 0x00000001 /* MDS handles LOV_OBJID file */
179 #define OBD_ROCOMPAT_CROW 0x00000002 /* OST will CROW create objects */
181 #define OBD_INCOMPAT_GROUPS 0x00000001 /* OST handles group subdirs */
182 #define OBD_INCOMPAT_OST 0x00000002 /* this is an OST */
183 #define OBD_INCOMPAT_MDT 0x00000004 /* this is an MDT */
184 #define OBD_INCOMPAT_COMMON_LR 0x00000008 /* common last_rvcd format */
187 /* Data stored per server at the head of the last_rcvd file. In le32 order.
188 This should be common to filter_internal.h, lustre_mds.h */
189 struct lr_server_data {
190 __u8 lsd_uuid[40]; /* server UUID */
191 __u64 lsd_last_transno; /* last completed transaction ID */
192 __u64 lsd_compat14; /* reserved - compat with old last_rcvd */
193 __u64 lsd_mount_count; /* incarnation number */
194 __u32 lsd_feature_compat; /* compatible feature flags */
195 __u32 lsd_feature_rocompat;/* read-only compatible feature flags */
196 __u32 lsd_feature_incompat;/* incompatible feature flags */
197 __u32 lsd_server_size; /* size of server data area */
198 __u32 lsd_client_start; /* start of per-client data area */
199 __u16 lsd_client_size; /* size of per-client data area */
200 __u16 lsd_subdir_count; /* number of subdirectories for objects */
201 __u64 lsd_catalog_oid; /* recovery catalog object id */
202 __u32 lsd_catalog_ogen; /* recovery catalog inode generation */
203 __u8 lsd_peeruuid[40]; /* UUID of MDS associated with this OST */
204 __u32 lsd_ost_index; /* index number of OST in LOV */
205 __u32 lsd_mdt_index; /* index number of MDT in LMV */
206 __u8 lsd_padding[LR_SERVER_SIZE - 148];
209 /* Data stored per client in the last_rcvd file. In le32 order. */
210 struct lsd_client_data {
211 __u8 lcd_uuid[40]; /* client UUID */
212 __u64 lcd_last_transno; /* last completed transaction ID */
213 __u64 lcd_last_xid; /* xid for the last transaction */
214 __u32 lcd_last_result; /* result from last RPC */
215 __u32 lcd_last_data; /* per-op data (disposition for open &c.) */
216 /* for MDS_CLOSE requests */
217 __u64 lcd_last_close_transno; /* last completed transaction ID */
218 __u64 lcd_last_close_xid; /* xid for the last transaction */
219 __u32 lcd_last_close_result; /* result from last RPC */
220 __u32 lcd_last_close_data; /* per-op data */
221 __u8 lcd_padding[LR_CLIENT_SIZE - 88];
226 /****************** superblock additional info *********************/
229 struct lustre_sb_info {
231 struct obd_device *lsi_mgc; /* mgc obd */
232 struct lustre_mount_data *lsi_lmd; /* mount command info */
233 struct lustre_disk_data *lsi_ldd; /* mount info on-disk */
234 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
235 struct vfsmount *lsi_srv_mnt; /* the one server mount */
236 atomic_t lsi_mounts; /* references to the srv_mnt */
239 #define LSI_SERVER 0x00000001
240 #define LSI_UMOUNT_FORCE 0x00000010
241 #define LSI_UMOUNT_FAILOVER 0x00000020
243 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
244 # define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
245 # define s2lsi_nocast(sb) ((sb)->s_fs_info)
247 # define s2lsi(sb) ((struct lustre_sb_info *)((sb)->u.generic_sbp))
248 # define s2lsi_nocast(sb) ((sb)->u.generic_sbp)
251 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
253 #endif /* __KERNEL__ */
255 /****************** mount lookup info *********************/
257 struct lustre_mount_info {
259 struct super_block *lmi_sb;
260 struct vfsmount *lmi_mnt;
261 struct list_head lmi_list_chain;
264 /****************** prototypes *********************/
267 #include <obd_class.h>
270 void lustre_register_client_fill_super(int (*cfs)(struct super_block *sb));
271 int lustre_common_put_super(struct super_block *sb);
272 int lustre_process_log(struct super_block *sb, char *logname,
273 struct config_llog_instance *cfg);
274 int lustre_end_log(struct super_block *sb, char *logname,
275 struct config_llog_instance *cfg);
276 struct lustre_mount_info *server_get_mount(const char *name);
277 int server_put_mount(const char *name, struct vfsmount *mnt);
278 int server_register_target(struct super_block *sb);
279 struct mgs_target_info;
280 int server_mti_print(char *title, struct mgs_target_info *mti);
283 int mgc_logname2resid(char *logname, struct ldlm_res_id *res_id);
287 #endif // _LUSTRE_DISK_H