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 #define LDD_F_SV_TYPE_MDT 0x0001
71 #define LDD_F_SV_TYPE_OST 0x0002
72 #define LDD_F_SV_TYPE_MGS 0x0004
73 #define LDD_F_SV_ALL 0x0008
74 /** need an index assignment */
75 #define LDD_F_NEED_INDEX 0x0010
76 /** never registered */
77 #define LDD_F_VIRGIN 0x0020
78 /** update the config logs for this server */
79 #define LDD_F_UPDATE 0x0040
80 /** rewrite the LDD */
81 #define LDD_F_REWRITE_LDD 0x0080
82 /** regenerate config logs for this fs or server */
83 #define LDD_F_WRITECONF 0x0100
85 #define LDD_F_UPGRADE14 0x0200
86 /** process as lctl conf_param */
87 #define LDD_F_PARAM 0x0400
88 /** backend fs make use of IAM directory format. */
89 #define LDD_F_IAM_DIR 0x0800
100 static inline char *mt_str(enum ldd_mount_type mt)
102 static char *mount_type_string[] = {
109 return mount_type_string[mt];
112 #define LDD_INCOMPAT_SUPP 0
113 #define LDD_ROCOMPAT_SUPP 0
115 #define LDD_MAGIC 0x1dd00001
117 /* On-disk configuration file. In host-endian order. */
118 struct lustre_disk_data {
120 __u32 ldd_feature_compat; /* compatible feature flags */
121 __u32 ldd_feature_rocompat;/* read-only compatible feature flags */
122 __u32 ldd_feature_incompat;/* incompatible feature flags */
124 __u32 ldd_config_ver; /* config rewrite count - not used */
125 __u32 ldd_flags; /* LDD_SV_TYPE */
126 __u32 ldd_svindex; /* server index (0001), must match
128 __u32 ldd_mount_type; /* target fs type LDD_MT_* */
129 char ldd_fsname[64]; /* filesystem this server is part of,
131 char ldd_svname[64]; /* this server's name (lustre-mdt0001)*/
132 __u8 ldd_uuid[40]; /* server UUID (COMPAT_146) */
134 /*200*/ char ldd_userdata[1024 - 200]; /* arbitrary user string */
135 /*1024*/__u8 ldd_padding[4096 - 1024];
136 /*4096*/char ldd_mount_opts[4096]; /* target fs mount opts */
137 /*8192*/char ldd_params[4096]; /* key=value pairs */
140 #define IS_MDT(data) ((data)->ldd_flags & LDD_F_SV_TYPE_MDT)
141 #define IS_OST(data) ((data)->ldd_flags & LDD_F_SV_TYPE_OST)
142 #define IS_MGS(data) ((data)->ldd_flags & LDD_F_SV_TYPE_MGS)
143 #define MT_STR(data) mt_str((data)->ldd_mount_type)
145 /* Make the mdt/ost server obd name based on the filesystem name */
146 static inline int server_make_name(__u32 flags, __u16 index, char *fs,
149 if (flags & (LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_OST)) {
150 if (!(flags & LDD_F_SV_ALL))
151 sprintf(name, "%.8s-%s%04x", fs,
152 (flags & LDD_F_SV_TYPE_MDT) ? "MDT" : "OST",
154 } else if (flags & LDD_F_SV_TYPE_MGS) {
155 sprintf(name, "MGS");
157 CERROR("unknown server type %#x\n", flags);
163 /* Get the index from the obd name */
164 int server_name2index(char *svname, __u32 *idx, char **endptr);
167 /****************** mount command *********************/
169 /* The lmd is only used internally by Lustre; mount simply passes
170 everything as string options */
172 #define LMD_MAGIC 0xbdacbd03
174 /* gleaned from the mount command - no persistent info here */
175 struct lustre_mount_data {
177 __u32 lmd_flags; /* lustre mount flags */
178 int lmd_mgs_failnodes; /* mgs failover node count */
179 int lmd_exclude_count;
180 int lmd_recovery_time_soft;
181 int lmd_recovery_time_hard;
182 char *lmd_dev; /* device name */
183 char *lmd_profile; /* client only */
184 char *lmd_mgssec; /* sptlrpc flavor to mgs */
185 char *lmd_opts; /* lustre mount options (as opposed to
186 _device_ mount options) */
187 __u32 *lmd_exclude; /* array of OSTs to ignore */
190 #define LMD_FLG_SERVER 0x0001 /* Mounting a server */
191 #define LMD_FLG_CLIENT 0x0002 /* Mounting a client */
192 #define LMD_FLG_ABORT_RECOV 0x0008 /* Abort recovery */
193 #define LMD_FLG_NOSVC 0x0010 /* Only start MGS/MGC for servers,
195 #define LMD_FLG_NOMGS 0x0020 /* Only start target for servers, reusing
196 existing MGS services */
197 #define LMD_FLG_WRITECONF 0x0040 /* Rewrite config log */
199 #define lmd_is_client(x) ((x)->lmd_flags & LMD_FLG_CLIENT)
202 /****************** last_rcvd file *********************/
204 /** version recovery epoch */
205 #define LR_EPOCH_BITS 32
206 #define lr_epoch(a) ((a) >> LR_EPOCH_BITS)
207 #define LR_EXPIRE_INTERVALS 16 /**< number of intervals to track transno */
208 #define ENOENT_VERSION 1 /** 'virtual' version of non-existent object */
210 #define LR_SERVER_SIZE 512
211 #define LR_CLIENT_START 8192
212 #define LR_CLIENT_SIZE 128
213 #if LR_CLIENT_START < LR_SERVER_SIZE
214 #error "Can't have LR_CLIENT_START < LR_SERVER_SIZE"
218 * This limit is arbitrary (131072 clients on x86), but it is convenient to use
219 * 2^n * CFS_PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
220 * If we need more than 131072 clients (order-2 allocation on x86) then this
221 * should become an array of single-page pointers that are allocated on demand.
223 #if (128 * 1024UL) > (CFS_PAGE_SIZE * 8)
224 #define LR_MAX_CLIENTS (128 * 1024UL)
226 #define LR_MAX_CLIENTS (CFS_PAGE_SIZE * 8)
229 /** COMPAT_146: this is an OST (temporary) */
230 #define OBD_COMPAT_OST 0x00000002
231 /** COMPAT_146: this is an MDT (temporary) */
232 #define OBD_COMPAT_MDT 0x00000004
233 /** 2.0 server, interop flag to show server version is changed */
234 #define OBD_COMPAT_20 0x00000008
236 /** MDS handles LOV_OBJID file */
237 #define OBD_ROCOMPAT_LOVOBJID 0x00000001
239 /** OST handles group subdirs */
240 #define OBD_INCOMPAT_GROUPS 0x00000001
241 /** this is an OST */
242 #define OBD_INCOMPAT_OST 0x00000002
243 /** this is an MDT */
244 #define OBD_INCOMPAT_MDT 0x00000004
245 /** common last_rvcd format */
246 #define OBD_INCOMPAT_COMMON_LR 0x00000008
247 /** FID is enabled */
248 #define OBD_INCOMPAT_FID 0x00000010
249 /** Size-on-MDS is enabled */
250 #define OBD_INCOMPAT_SOM 0x00000020
251 /** filesystem using iam format to store directory entries */
252 #define OBD_INCOMPAT_IAM_DIR 0x00000040
253 /** LMA attribute contains per-inode incompatible flags */
254 #define OBD_INCOMPAT_LMA 0x00000080
257 /* Data stored per server at the head of the last_rcvd file. In le32 order.
258 This should be common to filter_internal.h, lustre_mds.h */
259 struct lr_server_data {
260 __u8 lsd_uuid[40]; /* server UUID */
261 __u64 lsd_last_transno; /* last completed transaction ID */
262 __u64 lsd_compat14; /* reserved - compat with old last_rcvd */
263 __u64 lsd_mount_count; /* incarnation number */
264 __u32 lsd_feature_compat; /* compatible feature flags */
265 __u32 lsd_feature_rocompat;/* read-only compatible feature flags */
266 __u32 lsd_feature_incompat;/* incompatible feature flags */
267 __u32 lsd_server_size; /* size of server data area */
268 __u32 lsd_client_start; /* start of per-client data area */
269 __u16 lsd_client_size; /* size of per-client data area */
270 __u16 lsd_subdir_count; /* number of subdirectories for objects */
271 __u64 lsd_catalog_oid; /* recovery catalog object id */
272 __u32 lsd_catalog_ogen; /* recovery catalog inode generation */
273 __u8 lsd_peeruuid[40]; /* UUID of MDS associated with this OST */
274 __u32 lsd_ost_index; /* index number of OST in LOV */
275 __u32 lsd_mdt_index; /* index number of MDT in LMV */
276 __u32 lsd_start_epoch; /* VBR: start epoch from last boot */
277 /** transaction values since lsd_trans_table_time */
278 __u64 lsd_trans_table[LR_EXPIRE_INTERVALS];
279 /** start point of transno table below */
280 __u32 lsd_trans_table_time; /* time of first slot in table above */
281 __u32 lsd_expire_intervals; /* LR_EXPIRE_INTERVALS */
282 __u8 lsd_padding[LR_SERVER_SIZE - 288];
285 /* Data stored per client in the last_rcvd file. In le32 order. */
286 struct lsd_client_data {
287 __u8 lcd_uuid[40]; /* client UUID */
288 __u64 lcd_last_transno; /* last completed transaction ID */
289 __u64 lcd_last_xid; /* xid for the last transaction */
290 __u32 lcd_last_result; /* result from last RPC */
291 __u32 lcd_last_data; /* per-op data (disposition for open &c.) */
292 /* for MDS_CLOSE requests */
293 __u64 lcd_last_close_transno; /* last completed transaction ID */
294 __u64 lcd_last_close_xid; /* xid for the last transaction */
295 __u32 lcd_last_close_result; /* result from last RPC */
296 __u32 lcd_last_close_data; /* per-op data */
297 /* VBR: last versions */
298 __u64 lcd_pre_versions[4];
299 __u32 lcd_last_epoch;
300 /** orphans handling for delayed export rely on that */
301 __u32 lcd_first_epoch;
302 __u8 lcd_padding[LR_CLIENT_SIZE - 128];
305 /* bug20354: the lcd_uuid for export of clients may be wrong */
306 static inline void check_lcd(char *obd_name, int index,
307 struct lsd_client_data *lcd)
309 int length = sizeof(lcd->lcd_uuid);
310 if (strnlen((char*)lcd->lcd_uuid, length) == length) {
311 lcd->lcd_uuid[length - 1] = '\0';
313 LCONSOLE_ERROR("the client UUID (%s) on %s for exports"
314 "stored in last_rcvd(index = %d) is bad!\n",
315 lcd->lcd_uuid, obd_name, index);
319 /* last_rcvd handling */
320 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
321 struct lr_server_data *lsd)
324 memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof (lsd->lsd_uuid));
325 lsd->lsd_last_transno = le64_to_cpu(buf->lsd_last_transno);
326 lsd->lsd_compat14 = le64_to_cpu(buf->lsd_compat14);
327 lsd->lsd_mount_count = le64_to_cpu(buf->lsd_mount_count);
328 lsd->lsd_feature_compat = le32_to_cpu(buf->lsd_feature_compat);
329 lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
330 lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
331 lsd->lsd_server_size = le32_to_cpu(buf->lsd_server_size);
332 lsd->lsd_client_start = le32_to_cpu(buf->lsd_client_start);
333 lsd->lsd_client_size = le16_to_cpu(buf->lsd_client_size);
334 lsd->lsd_subdir_count = le16_to_cpu(buf->lsd_subdir_count);
335 lsd->lsd_catalog_oid = le64_to_cpu(buf->lsd_catalog_oid);
336 lsd->lsd_catalog_ogen = le32_to_cpu(buf->lsd_catalog_ogen);
337 memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
338 lsd->lsd_ost_index = le32_to_cpu(buf->lsd_ost_index);
339 lsd->lsd_mdt_index = le32_to_cpu(buf->lsd_mdt_index);
340 lsd->lsd_start_epoch = le32_to_cpu(buf->lsd_start_epoch);
341 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
342 lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
343 lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
344 lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
347 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
348 struct lr_server_data *buf)
351 memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof (buf->lsd_uuid));
352 buf->lsd_last_transno = cpu_to_le64(lsd->lsd_last_transno);
353 buf->lsd_compat14 = cpu_to_le64(lsd->lsd_compat14);
354 buf->lsd_mount_count = cpu_to_le64(lsd->lsd_mount_count);
355 buf->lsd_feature_compat = cpu_to_le32(lsd->lsd_feature_compat);
356 buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
357 buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
358 buf->lsd_server_size = cpu_to_le32(lsd->lsd_server_size);
359 buf->lsd_client_start = cpu_to_le32(lsd->lsd_client_start);
360 buf->lsd_client_size = cpu_to_le16(lsd->lsd_client_size);
361 buf->lsd_subdir_count = cpu_to_le16(lsd->lsd_subdir_count);
362 buf->lsd_catalog_oid = cpu_to_le64(lsd->lsd_catalog_oid);
363 buf->lsd_catalog_ogen = cpu_to_le32(lsd->lsd_catalog_ogen);
364 memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
365 buf->lsd_ost_index = cpu_to_le32(lsd->lsd_ost_index);
366 buf->lsd_mdt_index = cpu_to_le32(lsd->lsd_mdt_index);
367 buf->lsd_start_epoch = cpu_to_le32(lsd->lsd_start_epoch);
368 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
369 buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
370 buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
371 buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
374 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
375 struct lsd_client_data *lcd)
377 memcpy(lcd->lcd_uuid, buf->lcd_uuid, sizeof (lcd->lcd_uuid));
378 lcd->lcd_last_transno = le64_to_cpu(buf->lcd_last_transno);
379 lcd->lcd_last_xid = le64_to_cpu(buf->lcd_last_xid);
380 lcd->lcd_last_result = le32_to_cpu(buf->lcd_last_result);
381 lcd->lcd_last_data = le32_to_cpu(buf->lcd_last_data);
382 lcd->lcd_last_close_transno = le64_to_cpu(buf->lcd_last_close_transno);
383 lcd->lcd_last_close_xid = le64_to_cpu(buf->lcd_last_close_xid);
384 lcd->lcd_last_close_result = le32_to_cpu(buf->lcd_last_close_result);
385 lcd->lcd_last_close_data = le32_to_cpu(buf->lcd_last_close_data);
386 lcd->lcd_pre_versions[0] = le64_to_cpu(buf->lcd_pre_versions[0]);
387 lcd->lcd_pre_versions[1] = le64_to_cpu(buf->lcd_pre_versions[1]);
388 lcd->lcd_pre_versions[2] = le64_to_cpu(buf->lcd_pre_versions[2]);
389 lcd->lcd_pre_versions[3] = le64_to_cpu(buf->lcd_pre_versions[3]);
390 lcd->lcd_last_epoch = le32_to_cpu(buf->lcd_last_epoch);
391 lcd->lcd_first_epoch = le32_to_cpu(buf->lcd_first_epoch);
394 static inline void lcd_cpu_to_le(struct lsd_client_data *lcd,
395 struct lsd_client_data *buf)
397 memcpy(buf->lcd_uuid, lcd->lcd_uuid, sizeof (lcd->lcd_uuid));
398 buf->lcd_last_transno = cpu_to_le64(lcd->lcd_last_transno);
399 buf->lcd_last_xid = cpu_to_le64(lcd->lcd_last_xid);
400 buf->lcd_last_result = cpu_to_le32(lcd->lcd_last_result);
401 buf->lcd_last_data = cpu_to_le32(lcd->lcd_last_data);
402 buf->lcd_last_close_transno = cpu_to_le64(lcd->lcd_last_close_transno);
403 buf->lcd_last_close_xid = cpu_to_le64(lcd->lcd_last_close_xid);
404 buf->lcd_last_close_result = cpu_to_le32(lcd->lcd_last_close_result);
405 buf->lcd_last_close_data = cpu_to_le32(lcd->lcd_last_close_data);
406 buf->lcd_pre_versions[0] = cpu_to_le64(lcd->lcd_pre_versions[0]);
407 buf->lcd_pre_versions[1] = cpu_to_le64(lcd->lcd_pre_versions[1]);
408 buf->lcd_pre_versions[2] = cpu_to_le64(lcd->lcd_pre_versions[2]);
409 buf->lcd_pre_versions[3] = cpu_to_le64(lcd->lcd_pre_versions[3]);
410 buf->lcd_last_epoch = cpu_to_le32(lcd->lcd_last_epoch);
411 buf->lcd_first_epoch = cpu_to_le32(lcd->lcd_first_epoch);
414 static inline __u64 lcd_last_transno(struct lsd_client_data *lcd)
416 return (lcd->lcd_last_transno > lcd->lcd_last_close_transno ?
417 lcd->lcd_last_transno : lcd->lcd_last_close_transno);
420 static inline __u64 lcd_last_xid(struct lsd_client_data *lcd)
422 return (lcd->lcd_last_xid > lcd->lcd_last_close_xid ?
423 lcd->lcd_last_xid : lcd->lcd_last_close_xid);
426 /****************** superblock additional info *********************/
431 struct lustre_sb_info {
433 struct obd_device *lsi_mgc; /* mgc obd */
434 struct lustre_mount_data *lsi_lmd; /* mount command info */
435 struct lustre_disk_data *lsi_ldd; /* mount info on-disk */
436 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
437 struct vfsmount *lsi_srv_mnt; /* the one server mount */
438 cfs_atomic_t lsi_mounts; /* references to the srv_mnt */
439 struct backing_dev_info bdi; /* Each client mountpoint needs own backing_dev_info */
442 #define LSI_SERVER 0x00000001
443 #define LSI_UMOUNT_FORCE 0x00000010
444 #define LSI_UMOUNT_FAILOVER 0x00000020
446 #define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
447 #define s2lsi_nocast(sb) ((sb)->s_fs_info)
449 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
451 #endif /* __KERNEL__ */
453 /****************** mount lookup info *********************/
455 struct lustre_mount_info {
457 struct super_block *lmi_sb;
458 struct vfsmount *lmi_mnt;
459 cfs_list_t lmi_list_chain;
462 /****************** prototypes *********************/
467 void lustre_register_client_fill_super(int (*cfs)(struct super_block *sb));
468 void lustre_register_kill_super_cb(void (*cfs)(struct super_block *sb));
471 int lustre_common_put_super(struct super_block *sb);
472 struct lustre_mount_info *server_find_mount_locked(const char *name);
473 struct lustre_mount_info *server_get_mount(const char *name);
474 struct lustre_mount_info *server_get_mount_2(const char *name);
475 int server_put_mount(const char *name, struct vfsmount *mnt);
476 int server_put_mount_2(const char *name, struct vfsmount *mnt);
477 int server_register_target(struct super_block *sb);
478 struct mgs_target_info;
479 int server_mti_print(char *title, struct mgs_target_info *mti);
482 int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id);
488 #endif // _LUSTRE_DISK_H