1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre_disk.h
38 * Lustre disk format definitions.
40 * Author: Nathan Rutman <nathan@clusterfs.com>
43 #ifndef _LUSTRE_DISK_H
44 #define _LUSTRE_DISK_H
46 /** \defgroup disk disk
51 #include <libcfs/libcfs.h>
52 #include <lnet/types.h>
54 /****************** on-disk files *********************/
56 #define MDT_LOGS_DIR "LOGS" /* COMPAT_146 */
57 #define MOUNT_CONFIGS_DIR "CONFIGS"
58 #define CONFIGS_FILE "mountdata"
59 /** Persistent mount data are stored on the disk in this file. */
60 #define MOUNT_DATA_FILE MOUNT_CONFIGS_DIR"/"CONFIGS_FILE
61 #define LAST_RCVD "last_rcvd"
62 #define LOV_OBJID "lov_objid"
63 #define HEALTH_CHECK "health_check"
64 #define CAPA_KEYS "capa_keys"
65 #define CHANGELOG_USERS "changelog_users"
68 /****************** persistent mount data *********************/
70 /* First 4 bits reserved for SVTYPEs */
71 /** need an index assignment */
72 #define LDD_F_NEED_INDEX 0x0010
73 /** never registered */
74 #define LDD_F_VIRGIN 0x0020
75 /** update the config logs for this server */
76 #define LDD_F_UPDATE 0x0040
77 /** rewrite the LDD */
78 #define LDD_F_REWRITE_LDD 0x0080
79 /** regenerate config logs for this fs or server */
80 #define LDD_F_WRITECONF 0x0100
82 #define LDD_F_UPGRADE14 0x0200
83 /** process as lctl conf_param */
84 #define LDD_F_PARAM 0x0400
85 /** backend fs make use of IAM directory format. */
86 #define LDD_F_IAM_DIR 0x0800
97 static inline char *mt_str(enum ldd_mount_type mt)
99 static char *mount_type_string[] = {
106 return mount_type_string[mt];
109 #define LDD_INCOMPAT_SUPP 0
110 #define LDD_ROCOMPAT_SUPP 0
112 #define LDD_MAGIC 0x1dd00001
114 /* On-disk configuration file. In host-endian order. */
115 struct lustre_disk_data {
117 __u32 ldd_feature_compat; /* compatible feature flags */
118 __u32 ldd_feature_rocompat;/* read-only compatible feature flags */
119 __u32 ldd_feature_incompat;/* incompatible feature flags */
121 __u32 ldd_config_ver; /* config rewrite count - not used */
122 __u32 ldd_flags; /* SVTYPE */
123 __u32 ldd_svindex; /* server index (0001), must match
125 __u32 ldd_mount_type; /* target fs type LDD_MT_* */
126 char ldd_fsname[64]; /* filesystem this server is part of,
128 char ldd_svname[64]; /* this server's name (lustre-mdt0001)*/
129 __u8 ldd_uuid[40]; /* server UUID (COMPAT_146) */
131 /*200*/ char ldd_userdata[1024 - 200]; /* arbitrary user string */
132 /*1024*/__u8 ldd_padding[4096 - 1024];
133 /*4096*/char ldd_mount_opts[4096]; /* target fs mount opts */
134 /*8192*/char ldd_params[4096]; /* key=value pairs */
137 #define IS_MDT(data) ((data)->ldd_flags & SVTYPE_MDT)
138 #define IS_OST(data) ((data)->ldd_flags & SVTYPE_OST)
139 #define IS_MGS(data) ((data)->ldd_flags & SVTYPE_MGS)
140 #define MT_STR(data) mt_str((data)->ldd_mount_type)
142 /* Make the mdt/ost server obd name based on the filesystem name */
143 static inline int server_make_name(__u32 flags, __u16 index, char *fs,
146 if (flags & (SVTYPE_MDT | SVTYPE_OST)) {
147 if (!(flags & SVTYPE_ALL))
148 sprintf(name, "%.8s-%s%04x", fs,
149 (flags & SVTYPE_MDT) ? "MDT" : "OST",
151 } else if (flags & SVTYPE_MGS) {
152 sprintf(name, "MGS");
154 CERROR("unknown server type %#x\n", flags);
161 /****************** mount command *********************/
163 /* The lmd is only used internally by Lustre; mount simply passes
164 everything as string options */
166 #define LMD_MAGIC 0xbdacbd03
168 /* gleaned from the mount command - no persistent info here */
169 struct lustre_mount_data {
171 __u32 lmd_flags; /* lustre mount flags */
172 int lmd_mgs_failnodes; /* mgs failover node count */
173 int lmd_exclude_count;
174 int lmd_recovery_time_soft;
175 int lmd_recovery_time_hard;
176 char *lmd_dev; /* device name */
177 char *lmd_profile; /* client only */
178 char *lmd_mgssec; /* sptlrpc flavor to mgs */
179 char *lmd_opts; /* lustre mount options (as opposed to
180 _device_ mount options) */
181 __u32 *lmd_exclude; /* array of OSTs to ignore */
184 #define LMD_FLG_SERVER 0x0001 /* Mounting a server */
185 #define LMD_FLG_CLIENT 0x0002 /* Mounting a client */
186 #define LMD_FLG_ABORT_RECOV 0x0008 /* Abort recovery */
187 #define LMD_FLG_NOSVC 0x0010 /* Only start MGS/MGC for servers,
189 #define LMD_FLG_NOMGS 0x0020 /* Only start target for servers, reusing
190 existing MGS services */
191 #define LMD_FLG_WRITECONF 0x0040 /* Rewrite config log */
193 #define lmd_is_client(x) ((x)->lmd_flags & LMD_FLG_CLIENT)
196 /****************** last_rcvd file *********************/
198 /** version recovery epoch */
199 #define LR_EPOCH_BITS 32
200 #define lr_epoch(a) ((a) >> LR_EPOCH_BITS)
201 #define LR_EXPIRE_INTERVALS 16 /**< number of intervals to track transno */
202 #define ENOENT_VERSION 1 /** 'virtual' version of non-existent object */
204 #define LR_SERVER_SIZE 512
205 #define LR_CLIENT_START 8192
206 #define LR_CLIENT_SIZE 128
207 #if LR_CLIENT_START < LR_SERVER_SIZE
208 #error "Can't have LR_CLIENT_START < LR_SERVER_SIZE"
212 * This limit is arbitrary (131072 clients on x86), but it is convenient to use
213 * 2^n * CFS_PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
214 * If we need more than 131072 clients (order-2 allocation on x86) then this
215 * should become an array of single-page pointers that are allocated on demand.
217 #if (128 * 1024UL) > (CFS_PAGE_SIZE * 8)
218 #define LR_MAX_CLIENTS (128 * 1024UL)
220 #define LR_MAX_CLIENTS (CFS_PAGE_SIZE * 8)
223 /** COMPAT_146: this is an OST (temporary) */
224 #define OBD_COMPAT_OST 0x00000002
225 /** COMPAT_146: this is an MDT (temporary) */
226 #define OBD_COMPAT_MDT 0x00000004
227 /** 2.0 server, interop flag to show server version is changed */
228 #define OBD_COMPAT_20 0x00000008
230 /** MDS handles LOV_OBJID file */
231 #define OBD_ROCOMPAT_LOVOBJID 0x00000001
233 /** OST handles group subdirs */
234 #define OBD_INCOMPAT_GROUPS 0x00000001
235 /** this is an OST */
236 #define OBD_INCOMPAT_OST 0x00000002
237 /** this is an MDT */
238 #define OBD_INCOMPAT_MDT 0x00000004
239 /** common last_rvcd format */
240 #define OBD_INCOMPAT_COMMON_LR 0x00000008
241 /** FID is enabled */
242 #define OBD_INCOMPAT_FID 0x00000010
243 /** Size-on-MDS is enabled */
244 #define OBD_INCOMPAT_SOM 0x00000020
245 /** filesystem using iam format to store directory entries */
246 #define OBD_INCOMPAT_IAM_DIR 0x00000040
247 /** LMA attribute contains per-inode incompatible flags */
248 #define OBD_INCOMPAT_LMA 0x00000080
251 /* Data stored per server at the head of the last_rcvd file. In le32 order.
252 This should be common to filter_internal.h, lustre_mds.h */
253 struct lr_server_data {
254 __u8 lsd_uuid[40]; /* server UUID */
255 __u64 lsd_last_transno; /* last completed transaction ID */
256 __u64 lsd_compat14; /* reserved - compat with old last_rcvd */
257 __u64 lsd_mount_count; /* incarnation number */
258 __u32 lsd_feature_compat; /* compatible feature flags */
259 __u32 lsd_feature_rocompat;/* read-only compatible feature flags */
260 __u32 lsd_feature_incompat;/* incompatible feature flags */
261 __u32 lsd_server_size; /* size of server data area */
262 __u32 lsd_client_start; /* start of per-client data area */
263 __u16 lsd_client_size; /* size of per-client data area */
264 __u16 lsd_subdir_count; /* number of subdirectories for objects */
265 __u64 lsd_catalog_oid; /* recovery catalog object id */
266 __u32 lsd_catalog_ogen; /* recovery catalog inode generation */
267 __u8 lsd_peeruuid[40]; /* UUID of MDS associated with this OST */
268 __u32 lsd_ost_index; /* index number of OST in LOV */
269 __u32 lsd_mdt_index; /* index number of MDT in LMV */
270 __u32 lsd_start_epoch; /* VBR: start epoch from last boot */
271 /** transaction values since lsd_trans_table_time */
272 __u64 lsd_trans_table[LR_EXPIRE_INTERVALS];
273 /** start point of transno table below */
274 __u32 lsd_trans_table_time; /* time of first slot in table above */
275 __u32 lsd_expire_intervals; /* LR_EXPIRE_INTERVALS */
276 __u8 lsd_padding[LR_SERVER_SIZE - 288];
279 /* Data stored per client in the last_rcvd file. In le32 order. */
280 struct lsd_client_data {
281 __u8 lcd_uuid[40]; /* client UUID */
282 __u64 lcd_last_transno; /* last completed transaction ID */
283 __u64 lcd_last_xid; /* xid for the last transaction */
284 __u32 lcd_last_result; /* result from last RPC */
285 __u32 lcd_last_data; /* per-op data (disposition for open &c.) */
286 /* for MDS_CLOSE requests */
287 __u64 lcd_last_close_transno; /* last completed transaction ID */
288 __u64 lcd_last_close_xid; /* xid for the last transaction */
289 __u32 lcd_last_close_result; /* result from last RPC */
290 __u32 lcd_last_close_data; /* per-op data */
291 /* VBR: last versions */
292 __u64 lcd_pre_versions[4];
293 __u32 lcd_last_epoch;
294 /** orphans handling for delayed export rely on that */
295 __u32 lcd_first_epoch;
296 __u8 lcd_padding[LR_CLIENT_SIZE - 128];
299 /* bug20354: the lcd_uuid for export of clients may be wrong */
300 static inline void check_lcd(char *obd_name, int index,
301 struct lsd_client_data *lcd)
303 int length = sizeof(lcd->lcd_uuid);
304 if (strnlen((char*)lcd->lcd_uuid, length) == length) {
305 lcd->lcd_uuid[length - 1] = '\0';
307 LCONSOLE_ERROR("the client UUID (%s) on %s for exports"
308 "stored in last_rcvd(index = %d) is bad!\n",
309 lcd->lcd_uuid, obd_name, index);
313 /* last_rcvd handling */
314 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
315 struct lr_server_data *lsd)
318 memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof (lsd->lsd_uuid));
319 lsd->lsd_last_transno = le64_to_cpu(buf->lsd_last_transno);
320 lsd->lsd_compat14 = le64_to_cpu(buf->lsd_compat14);
321 lsd->lsd_mount_count = le64_to_cpu(buf->lsd_mount_count);
322 lsd->lsd_feature_compat = le32_to_cpu(buf->lsd_feature_compat);
323 lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
324 lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
325 lsd->lsd_server_size = le32_to_cpu(buf->lsd_server_size);
326 lsd->lsd_client_start = le32_to_cpu(buf->lsd_client_start);
327 lsd->lsd_client_size = le16_to_cpu(buf->lsd_client_size);
328 lsd->lsd_subdir_count = le16_to_cpu(buf->lsd_subdir_count);
329 lsd->lsd_catalog_oid = le64_to_cpu(buf->lsd_catalog_oid);
330 lsd->lsd_catalog_ogen = le32_to_cpu(buf->lsd_catalog_ogen);
331 memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
332 lsd->lsd_ost_index = le32_to_cpu(buf->lsd_ost_index);
333 lsd->lsd_mdt_index = le32_to_cpu(buf->lsd_mdt_index);
334 lsd->lsd_start_epoch = le32_to_cpu(buf->lsd_start_epoch);
335 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
336 lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
337 lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
338 lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
341 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
342 struct lr_server_data *buf)
345 memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof (buf->lsd_uuid));
346 buf->lsd_last_transno = cpu_to_le64(lsd->lsd_last_transno);
347 buf->lsd_compat14 = cpu_to_le64(lsd->lsd_compat14);
348 buf->lsd_mount_count = cpu_to_le64(lsd->lsd_mount_count);
349 buf->lsd_feature_compat = cpu_to_le32(lsd->lsd_feature_compat);
350 buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
351 buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
352 buf->lsd_server_size = cpu_to_le32(lsd->lsd_server_size);
353 buf->lsd_client_start = cpu_to_le32(lsd->lsd_client_start);
354 buf->lsd_client_size = cpu_to_le16(lsd->lsd_client_size);
355 buf->lsd_subdir_count = cpu_to_le16(lsd->lsd_subdir_count);
356 buf->lsd_catalog_oid = cpu_to_le64(lsd->lsd_catalog_oid);
357 buf->lsd_catalog_ogen = cpu_to_le32(lsd->lsd_catalog_ogen);
358 memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
359 buf->lsd_ost_index = cpu_to_le32(lsd->lsd_ost_index);
360 buf->lsd_mdt_index = cpu_to_le32(lsd->lsd_mdt_index);
361 buf->lsd_start_epoch = cpu_to_le32(lsd->lsd_start_epoch);
362 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
363 buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
364 buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
365 buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
368 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
369 struct lsd_client_data *lcd)
371 memcpy(lcd->lcd_uuid, buf->lcd_uuid, sizeof (lcd->lcd_uuid));
372 lcd->lcd_last_transno = le64_to_cpu(buf->lcd_last_transno);
373 lcd->lcd_last_xid = le64_to_cpu(buf->lcd_last_xid);
374 lcd->lcd_last_result = le32_to_cpu(buf->lcd_last_result);
375 lcd->lcd_last_data = le32_to_cpu(buf->lcd_last_data);
376 lcd->lcd_last_close_transno = le64_to_cpu(buf->lcd_last_close_transno);
377 lcd->lcd_last_close_xid = le64_to_cpu(buf->lcd_last_close_xid);
378 lcd->lcd_last_close_result = le32_to_cpu(buf->lcd_last_close_result);
379 lcd->lcd_last_close_data = le32_to_cpu(buf->lcd_last_close_data);
380 lcd->lcd_pre_versions[0] = le64_to_cpu(buf->lcd_pre_versions[0]);
381 lcd->lcd_pre_versions[1] = le64_to_cpu(buf->lcd_pre_versions[1]);
382 lcd->lcd_pre_versions[2] = le64_to_cpu(buf->lcd_pre_versions[2]);
383 lcd->lcd_pre_versions[3] = le64_to_cpu(buf->lcd_pre_versions[3]);
384 lcd->lcd_last_epoch = le32_to_cpu(buf->lcd_last_epoch);
385 lcd->lcd_first_epoch = le32_to_cpu(buf->lcd_first_epoch);
388 static inline void lcd_cpu_to_le(struct lsd_client_data *lcd,
389 struct lsd_client_data *buf)
391 memcpy(buf->lcd_uuid, lcd->lcd_uuid, sizeof (lcd->lcd_uuid));
392 buf->lcd_last_transno = cpu_to_le64(lcd->lcd_last_transno);
393 buf->lcd_last_xid = cpu_to_le64(lcd->lcd_last_xid);
394 buf->lcd_last_result = cpu_to_le32(lcd->lcd_last_result);
395 buf->lcd_last_data = cpu_to_le32(lcd->lcd_last_data);
396 buf->lcd_last_close_transno = cpu_to_le64(lcd->lcd_last_close_transno);
397 buf->lcd_last_close_xid = cpu_to_le64(lcd->lcd_last_close_xid);
398 buf->lcd_last_close_result = cpu_to_le32(lcd->lcd_last_close_result);
399 buf->lcd_last_close_data = cpu_to_le32(lcd->lcd_last_close_data);
400 buf->lcd_pre_versions[0] = cpu_to_le64(lcd->lcd_pre_versions[0]);
401 buf->lcd_pre_versions[1] = cpu_to_le64(lcd->lcd_pre_versions[1]);
402 buf->lcd_pre_versions[2] = cpu_to_le64(lcd->lcd_pre_versions[2]);
403 buf->lcd_pre_versions[3] = cpu_to_le64(lcd->lcd_pre_versions[3]);
404 buf->lcd_last_epoch = cpu_to_le32(lcd->lcd_last_epoch);
405 buf->lcd_first_epoch = cpu_to_le32(lcd->lcd_first_epoch);
408 static inline __u64 lcd_last_transno(struct lsd_client_data *lcd)
410 return (lcd->lcd_last_transno > lcd->lcd_last_close_transno ?
411 lcd->lcd_last_transno : lcd->lcd_last_close_transno);
414 static inline __u64 lcd_last_xid(struct lsd_client_data *lcd)
416 return (lcd->lcd_last_xid > lcd->lcd_last_close_xid ?
417 lcd->lcd_last_xid : lcd->lcd_last_close_xid);
420 /****************** superblock additional info *********************/
425 struct lustre_sb_info {
427 struct obd_device *lsi_mgc; /* mgc obd */
428 struct lustre_mount_data *lsi_lmd; /* mount command info */
429 struct lustre_disk_data *lsi_ldd; /* mount info on-disk */
430 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
431 struct vfsmount *lsi_srv_mnt; /* the one server mount */
432 cfs_atomic_t lsi_mounts; /* references to the srv_mnt */
435 #define LSI_SERVER 0x00000001
436 #define LSI_UMOUNT_FORCE 0x00000010
437 #define LSI_UMOUNT_FAILOVER 0x00000020
439 #define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
440 #define s2lsi_nocast(sb) ((sb)->s_fs_info)
442 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
444 #endif /* __KERNEL__ */
446 /****************** mount lookup info *********************/
448 struct lustre_mount_info {
450 struct super_block *lmi_sb;
451 struct vfsmount *lmi_mnt;
452 cfs_list_t lmi_list_chain;
455 /****************** prototypes *********************/
460 void lustre_register_client_fill_super(int (*cfs)(struct super_block *sb));
461 void lustre_register_kill_super_cb(void (*cfs)(struct super_block *sb));
464 int lustre_common_put_super(struct super_block *sb);
465 struct lustre_mount_info *server_find_mount_locked(const char *name);
466 struct lustre_mount_info *server_get_mount(const char *name);
467 struct lustre_mount_info *server_get_mount_2(const char *name);
468 int server_put_mount(const char *name, struct vfsmount *mnt);
469 int server_put_mount_2(const char *name, struct vfsmount *mnt);
470 int server_register_target(struct super_block *sb);
471 struct mgs_target_info;
472 int server_mti_print(char *title, struct mgs_target_info *mti);
475 int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id);
481 #endif // _LUSTRE_DISK_H