4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
31 * lustre/include/lustre_disk.h
33 * Lustre disk format definitions.
35 * Author: Nathan Rutman <nathan@clusterfs.com>
38 #ifndef _LUSTRE_DISK_H
39 #define _LUSTRE_DISK_H
41 /** \defgroup disk disk
45 #include <asm/byteorder.h>
46 #include <linux/types.h>
47 #include <linux/backing-dev.h>
48 #include <linux/list.h>
49 #include <libcfs/libcfs.h>
50 #if !defined(CONFIG_LL_ENCRYPTION) && defined(HAVE_LUSTRE_CRYPTO)
51 #include <lustre_crypto.h>
53 #include <uapi/linux/lustre/lustre_idl.h>
54 #ifdef HAVE_SERVER_SUPPORT
55 #include <uapi/linux/lustre/lustre_disk.h>
56 #define IS_MDT(data) ((data)->lsi_flags & LDD_F_SV_TYPE_MDT)
57 #define IS_OST(data) ((data)->lsi_flags & LDD_F_SV_TYPE_OST)
58 #define IS_MGS(data) ((data)->lsi_flags & LDD_F_SV_TYPE_MGS)
59 #define IS_SERVER(data) ((data) && \
60 (data)->lsi_flags & (LDD_F_SV_TYPE_MGS | \
64 #define LDD_F_SV_TYPE_MDT 0x0001
65 #define LDD_F_SV_TYPE_OST 0x0002
66 #define LDD_F_SV_TYPE_MGS 0x0004
67 #define LDD_F_SV_ALL 0x0008
69 #define IS_MDT(data) (0)
70 #define IS_OST(data) (0)
71 #define IS_MGS(data) (0)
72 #define IS_SERVER(data) (0)
75 #define MT_STR(data) mt_str((data)->ldd_mount_type)
77 /****************** mount command *********************/
79 /* The lmd is only used internally by Lustre; mount simply passes
80 * everything as string options
82 #define LMD_MAGIC 0xbdacbd03
83 #define LMD_PARAMS_MAXLEN 4096
86 LMD_FLG_SERVER = 0, /* Mounting a server */
87 LMD_FLG_CLIENT, /* Mounting a client */
88 LMD_FLG_SKIP_LFSCK, /* NOT auto resume LFSCK when mount */
89 LMD_FLG_ABORT_RECOV, /* Abort recovery */
90 LMD_FLG_NOSVC, /* Only start MGS/MGC for servers,
93 LMD_FLG_NOMGS, /* Only start target for servers,
94 * reusing existing MGS services
96 LMD_FLG_WRITECONF, /* Rewrite config log */
97 LMD_FLG_NOIR, /* NO imperative recovery */
98 LMD_FLG_NOSCRUB, /* Do not trigger scrub automatically */
99 LMD_FLG_MGS, /* Also start MGS along with server */
100 LMD_FLG_IAM, /* IAM dir */
101 LMD_FLG_NO_PRIMNODE, /* all nodes are service nodes */
102 LMD_FLG_VIRGIN, /* the service registers first time */
103 LMD_FLG_UPDATE, /* update parameters */
104 LMD_FLG_HSM, /* Start coordinator */
105 LMD_FLG_DEV_RDONLY, /* discard modification quitely */
106 LMD_FLG_NO_PRECREATE, /* do not allow OST object creation */
107 LMD_FLG_LOCAL_RECOV, /* force recovery for local clients */
108 LMD_FLG_ABORT_RECOV_MDT, /* Abort recovery between MDTs */
109 LMD_FLG_NO_LOCAL_LOGS, /* Use config logs from MGS */
113 /* gleaned from the mount command - no persistent info here */
114 struct lustre_mount_data {
116 DECLARE_BITMAP(lmd_flags, LMD_FLG_NUM_FLAGS); /* lustre mount flags */
117 int lmd_mgs_failnodes; /* mgs failover node count */
118 int lmd_exclude_count;
119 int lmd_recovery_time_soft;
120 int lmd_recovery_time_hard;
121 char *lmd_dev; /* device name */
122 char *lmd_profile; /* client only */
123 char *lmd_fileset; /* mount fileset */
124 char *lmd_mgssec; /* sptlrpc flavor to mgs */
125 char *lmd_opts; /* lustre mount options (as opposed to
126 * device_ mount options) */
127 char *lmd_params; /* lustre params */
128 u32 *lmd_exclude; /* array of OSTs to ignore */
129 char *lmd_mgs; /* MGS nid */
130 char *lmd_osd_type; /* OSD type */
131 char *lmd_nidnet; /* network to restrict this client to */
134 #define lmd_is_client(x) (test_bit(LMD_FLG_CLIENT, (x)->lmd_flags))
136 /****************** superblock additional info *********************/
140 struct lustre_sb_info {
142 struct obd_device *lsi_mgc; /* mgc obd */
143 struct lustre_mount_data *lsi_lmd; /* mount command info */
144 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
145 struct dt_device *lsi_dt_dev; /* dt device to access disk fs*/
146 atomic_t lsi_mounts; /* references to the srv_mnt */
147 struct kobject *lsi_kobj;
148 char lsi_svname[MTI_NAME_MAXLEN];
149 /* lsi_osd_obdname format = 'lsi->ls_svname'-osd */
150 char lsi_osd_obdname[MTI_NAME_MAXLEN + 4];
151 /* lsi_osd_uuid format = 'lsi->ls_osd_obdname'_UUID */
152 char lsi_osd_uuid[MTI_NAME_MAXLEN + 9];
153 struct obd_export *lsi_osd_exp;
154 char lsi_osd_type[16];
156 struct backing_dev_info lsi_bdi; /* each client mountpoint needs
157 own backing_dev_info */
158 /* protect lsi_lwp_list */
159 struct mutex lsi_lwp_mutex;
160 struct list_head lsi_lwp_list;
161 unsigned long lsi_lwp_started:1,
162 lsi_server_started:1;
163 #ifdef CONFIG_LL_ENCRYPTION
164 const struct llcrypt_operations *lsi_cop;
165 struct key *lsi_master_keys; /* master crypto keys used */
166 #elif defined(HAVE_LUSTRE_CRYPTO) && !defined(HAVE_FSCRYPT_DUMMY_CONTEXT_ENABLED)
167 /* Dummy Encryption policy for '-o test_dummy_encryption' */
168 struct llcrypt_dummy_policy lsi_dummy_enc_policy;
172 #define LSI_UMOUNT_FAILOVER 0x00200000
173 #ifndef HAVE_SUPER_SETUP_BDI_NAME
174 #define LSI_BDI_INITIALIZED 0x00400000
176 #ifdef CONFIG_LL_ENCRYPTION
177 #define LSI_FILENAME_ENC 0x00800000 /* enable name encryption */
179 #define LSI_FILENAME_ENC_B64_OLD_CLI 0x01000000 /* use old style base64 */
181 #define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
182 #define s2lsi_nocast(sb) ((sb)->s_fs_info)
184 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
185 #define get_mount_fileset(sb) (s2lsi(sb)->lsi_lmd->lmd_fileset)
187 # ifdef HAVE_SERVER_SUPPORT
188 /* opc for target register */
189 #define LDD_F_OPC_REG 0x10000000
190 #define LDD_F_OPC_UNREG 0x20000000
191 #define LDD_F_OPC_READY 0x40000000
192 #define LDD_F_OPC_MASK 0xf0000000
194 #define LDD_F_MASK 0xFFFF
197 * This limit is arbitrary (131072 clients on x86), but it is convenient to use
198 * 2^n * PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
199 * If we need more than 131072 clients (order-2 allocation on x86) then this
200 * should become an array of single-page pointers that are allocated on demand.
202 #if (128 * 1024UL) > (PAGE_SIZE * 8)
203 #define LR_MAX_CLIENTS (128 * 1024UL)
205 #define LR_MAX_CLIENTS (PAGE_SIZE * 8)
208 /** COMPAT_146: this is an OST (temporary) */
209 #define OBD_COMPAT_OST 0x00000002
210 /** COMPAT_146: this is an MDT (temporary) */
211 #define OBD_COMPAT_MDT 0x00000004
212 /** 2.0 server, interop flag to show server version is changed */
213 #define OBD_COMPAT_20 0x00000008
215 /** MDS handles LOV_OBJID file */
216 #define OBD_ROCOMPAT_LOVOBJID 0x00000001
217 /** store OST index in the IDIF */
218 #define OBD_ROCOMPAT_IDX_IN_IDIF 0x00000002
220 /** OST handles group subdirs */
221 #define OBD_INCOMPAT_GROUPS 0x00000001
222 /** this is an OST */
223 #define OBD_INCOMPAT_OST 0x00000002
224 /** this is an MDT */
225 #define OBD_INCOMPAT_MDT 0x00000004
226 /** common last_rvcd format */
227 #define OBD_INCOMPAT_COMMON_LR 0x00000008
228 /** FID is enabled */
229 #define OBD_INCOMPAT_FID 0x00000010
230 /** Size-on-MDS is enabled */
231 #define OBD_INCOMPAT_SOM 0x00000020
232 /** filesystem using iam format to store directory entries */
233 #define OBD_INCOMPAT_IAM_DIR 0x00000040
234 /** LMA attribute contains per-inode incompatible flags */
235 #define OBD_INCOMPAT_LMA 0x00000080
236 /** lmm_stripe_count has been shrunk from u32 to u16 and the remaining 16
237 * bits are now used to store a generation. Once we start changing the layout
238 * and bumping the generation, old versions expecting a 32-bit lmm_stripe_count
239 * will be confused by interpreting stripe_count | gen << 16 as the actual
241 #define OBD_INCOMPAT_LMM_VER 0x00000100
242 /** multiple OI files for MDT */
243 #define OBD_INCOMPAT_MULTI_OI 0x00000200
244 /** multiple RPCs in flight */
245 #define OBD_INCOMPAT_MULTI_RPCS 0x00000400
247 /* last_rcvd handling */
248 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
249 struct lr_server_data *lsd)
253 memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof(lsd->lsd_uuid));
254 lsd->lsd_last_transno = le64_to_cpu(buf->lsd_last_transno);
255 lsd->lsd_compat14 = le64_to_cpu(buf->lsd_compat14);
256 lsd->lsd_mount_count = le64_to_cpu(buf->lsd_mount_count);
257 lsd->lsd_feature_compat = le32_to_cpu(buf->lsd_feature_compat);
258 lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
259 lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
260 lsd->lsd_server_size = le32_to_cpu(buf->lsd_server_size);
261 lsd->lsd_client_start = le32_to_cpu(buf->lsd_client_start);
262 lsd->lsd_client_size = le16_to_cpu(buf->lsd_client_size);
263 lsd->lsd_subdir_count = le16_to_cpu(buf->lsd_subdir_count);
264 lsd->lsd_catalog_oid = le64_to_cpu(buf->lsd_catalog_oid);
265 lsd->lsd_catalog_ogen = le32_to_cpu(buf->lsd_catalog_ogen);
266 memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
267 lsd->lsd_osd_index = le32_to_cpu(buf->lsd_osd_index);
268 lsd->lsd_padding1 = le32_to_cpu(buf->lsd_padding1);
269 lsd->lsd_start_epoch = le32_to_cpu(buf->lsd_start_epoch);
270 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
271 lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
272 lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
273 lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
276 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
277 struct lr_server_data *buf)
281 memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof(buf->lsd_uuid));
282 buf->lsd_last_transno = cpu_to_le64(lsd->lsd_last_transno);
283 buf->lsd_compat14 = cpu_to_le64(lsd->lsd_compat14);
284 buf->lsd_mount_count = cpu_to_le64(lsd->lsd_mount_count);
285 buf->lsd_feature_compat = cpu_to_le32(lsd->lsd_feature_compat);
286 buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
287 buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
288 buf->lsd_server_size = cpu_to_le32(lsd->lsd_server_size);
289 buf->lsd_client_start = cpu_to_le32(lsd->lsd_client_start);
290 buf->lsd_client_size = cpu_to_le16(lsd->lsd_client_size);
291 buf->lsd_subdir_count = cpu_to_le16(lsd->lsd_subdir_count);
292 buf->lsd_catalog_oid = cpu_to_le64(lsd->lsd_catalog_oid);
293 buf->lsd_catalog_ogen = cpu_to_le32(lsd->lsd_catalog_ogen);
294 memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
295 buf->lsd_osd_index = cpu_to_le32(lsd->lsd_osd_index);
296 buf->lsd_padding1 = cpu_to_le32(lsd->lsd_padding1);
297 buf->lsd_start_epoch = cpu_to_le32(lsd->lsd_start_epoch);
298 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
299 buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
300 buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
301 buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
304 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
305 struct lsd_client_data *lcd)
307 memcpy(lcd->lcd_uuid, buf->lcd_uuid, sizeof (lcd->lcd_uuid));
308 lcd->lcd_last_transno = le64_to_cpu(buf->lcd_last_transno);
309 lcd->lcd_last_xid = le64_to_cpu(buf->lcd_last_xid);
310 lcd->lcd_last_result = le32_to_cpu(buf->lcd_last_result);
311 lcd->lcd_last_data = le32_to_cpu(buf->lcd_last_data);
312 lcd->lcd_last_close_transno = le64_to_cpu(buf->lcd_last_close_transno);
313 lcd->lcd_last_close_xid = le64_to_cpu(buf->lcd_last_close_xid);
314 lcd->lcd_last_close_result = le32_to_cpu(buf->lcd_last_close_result);
315 lcd->lcd_last_close_data = le32_to_cpu(buf->lcd_last_close_data);
316 lcd->lcd_pre_versions[0] = le64_to_cpu(buf->lcd_pre_versions[0]);
317 lcd->lcd_pre_versions[1] = le64_to_cpu(buf->lcd_pre_versions[1]);
318 lcd->lcd_pre_versions[2] = le64_to_cpu(buf->lcd_pre_versions[2]);
319 lcd->lcd_pre_versions[3] = le64_to_cpu(buf->lcd_pre_versions[3]);
320 lcd->lcd_last_epoch = le32_to_cpu(buf->lcd_last_epoch);
321 lcd->lcd_generation = le32_to_cpu(buf->lcd_generation);
324 static inline void lcd_cpu_to_le(struct lsd_client_data *lcd,
325 struct lsd_client_data *buf)
327 memcpy(buf->lcd_uuid, lcd->lcd_uuid, sizeof (lcd->lcd_uuid));
328 buf->lcd_last_transno = cpu_to_le64(lcd->lcd_last_transno);
329 buf->lcd_last_xid = cpu_to_le64(lcd->lcd_last_xid);
330 buf->lcd_last_result = cpu_to_le32(lcd->lcd_last_result);
331 buf->lcd_last_data = cpu_to_le32(lcd->lcd_last_data);
332 buf->lcd_last_close_transno = cpu_to_le64(lcd->lcd_last_close_transno);
333 buf->lcd_last_close_xid = cpu_to_le64(lcd->lcd_last_close_xid);
334 buf->lcd_last_close_result = cpu_to_le32(lcd->lcd_last_close_result);
335 buf->lcd_last_close_data = cpu_to_le32(lcd->lcd_last_close_data);
336 buf->lcd_pre_versions[0] = cpu_to_le64(lcd->lcd_pre_versions[0]);
337 buf->lcd_pre_versions[1] = cpu_to_le64(lcd->lcd_pre_versions[1]);
338 buf->lcd_pre_versions[2] = cpu_to_le64(lcd->lcd_pre_versions[2]);
339 buf->lcd_pre_versions[3] = cpu_to_le64(lcd->lcd_pre_versions[3]);
340 buf->lcd_last_epoch = cpu_to_le32(lcd->lcd_last_epoch);
341 buf->lcd_generation = cpu_to_le32(lcd->lcd_generation);
344 static inline u64 lcd_last_transno(struct lsd_client_data *lcd)
346 return (lcd->lcd_last_transno > lcd->lcd_last_close_transno ?
347 lcd->lcd_last_transno : lcd->lcd_last_close_transno);
350 static inline u64 lcd_last_xid(struct lsd_client_data *lcd)
352 return (lcd->lcd_last_xid > lcd->lcd_last_close_xid ?
353 lcd->lcd_last_xid : lcd->lcd_last_close_xid);
356 /****************** mount lookup info *********************/
358 struct lustre_mount_info {
360 struct super_block *lmi_sb;
361 struct list_head lmi_list_chain;
364 /****************** prototypes *********************/
366 /* obd_mount_server.c */
367 int server_fill_super(struct super_block *sb);
368 struct lustre_mount_info *server_get_mount(const char *name);
369 int server_put_mount(const char *name, bool dereg_mnt);
370 struct mgs_target_info;
371 int server_mti_print(const char *title, struct mgs_target_info *mti);
372 void server_calc_timeout(struct lustre_sb_info *lsi, struct obd_device *obd);
375 int server_name2svname(const char *label, char *svname, const char **endptr,
378 int server_name_is_ost(const char *svname);
379 int target_name2index(const char *svname, u32 *idx, const char **endptr);
381 int lustre_put_lsi(struct super_block *sb);
382 int lustre_start_simple(char *obdname, char *type, char *uuid,
383 char *s1, char *s2, char *s3, char *s4);
384 int lustre_stop_mgc(struct super_block *sb);
385 #endif /* HAVE_SERVER_SUPPORT */
386 int server_name2fsname(const char *svname, char *fsname, const char **endptr);
387 void obdname2fsname(const char *tgt, char *fsname, size_t fslen);
389 int lustre_start_mgc(struct super_block *sb);
390 int lustre_common_put_super(struct super_block *sb);
392 struct lustre_sb_info *lustre_init_lsi(struct super_block *sb);
393 int lustre_put_lsi(struct super_block *sb);
394 int lmd_parse(char *options, struct lustre_mount_data *lmd);
397 int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id,
398 enum mgs_cfg_type type);
399 int mgc_logname2resid(char *fsname, struct ldlm_res_id *res_id,
400 enum mgs_cfg_type type);
404 #endif /* _LUSTRE_DISK_H */