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_CLIENT, /* Mounting a client */
87 LMD_FLG_SKIP_LFSCK, /* NOT auto resume LFSCK when mount */
88 LMD_FLG_ABORT_RECOV, /* Abort recovery */
89 LMD_FLG_NOSVC, /* Only start MGS/MGC for servers,
92 LMD_FLG_NOMGS, /* Only start target for servers,
93 * reusing existing MGS services
95 LMD_FLG_WRITECONF, /* Rewrite config log */
96 LMD_FLG_NOIR, /* NO imperative recovery */
97 LMD_FLG_NOSCRUB, /* Do not trigger scrub automatically */
98 LMD_FLG_MGS, /* Also start MGS along with server */
99 LMD_FLG_NO_PRIMNODE, /* all nodes are service nodes */
100 LMD_FLG_VIRGIN, /* the service registers first time */
101 LMD_FLG_UPDATE, /* update parameters */
102 LMD_FLG_HSM, /* Start coordinator */
103 LMD_FLG_DEV_RDONLY, /* discard modification quitely */
104 LMD_FLG_NO_CREATE, /* prevent MDT/OST object creation */
105 LMD_FLG_LOCAL_RECOV, /* force recovery for local clients */
106 LMD_FLG_ABORT_RECOV_MDT, /* Abort recovery between MDTs */
107 LMD_FLG_NO_LOCAL_LOGS, /* Use config logs from MGS */
111 /* gleaned from the mount command - no persistent info here */
112 struct lustre_mount_data {
114 DECLARE_BITMAP(lmd_flags, LMD_FLG_NUM_FLAGS); /* lustre mount flags */
115 int lmd_mgs_failnodes; /* mgs failover node count */
116 int lmd_exclude_count;
117 int lmd_recovery_time_soft;
118 int lmd_recovery_time_hard;
119 char *lmd_dev; /* device name */
120 char *lmd_profile; /* client only */
121 char *lmd_fileset; /* mount fileset */
122 char *lmd_mgssec; /* sptlrpc flavor to mgs */
123 char *lmd_opts; /* lustre mount options (as opposed to
124 * device_ mount options) */
125 char *lmd_params; /* lustre params */
126 u32 *lmd_exclude; /* array of OSTs to ignore */
127 char *lmd_mgs; /* MGS nid */
128 char *lmd_osd_type; /* OSD type */
129 char *lmd_nidnet; /* network to restrict this client to */
132 #define lmd_is_client(x) (test_bit(LMD_FLG_CLIENT, (x)->lmd_flags))
134 /****************** superblock additional info *********************/
138 struct lustre_sb_info {
140 struct obd_device *lsi_mgc; /* mgc obd */
141 struct lustre_mount_data *lsi_lmd; /* mount command info */
142 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
143 struct dt_device *lsi_dt_dev; /* dt device to access disk fs*/
144 struct kref lsi_mounts; /* references to the srv_mnt */
145 struct kobject *lsi_kobj;
146 char lsi_svname[MTI_NAME_MAXLEN];
147 /* lsi_osd_obdname format = 'lsi->ls_svname'-osd */
148 char lsi_osd_obdname[MTI_NAME_MAXLEN + 4];
149 /* lsi_osd_uuid format = 'lsi->ls_osd_obdname'_UUID */
150 char lsi_osd_uuid[MTI_NAME_MAXLEN + 9];
151 struct obd_export *lsi_osd_exp;
152 char lsi_osd_type[16];
154 struct backing_dev_info lsi_bdi; /* each client mountpoint needs
155 own backing_dev_info */
156 /* protect lsi_lwp_list */
157 struct mutex lsi_lwp_mutex;
158 struct list_head lsi_lwp_list;
159 unsigned long lsi_lwp_started:1,
160 lsi_server_started:1;
161 #ifdef CONFIG_LL_ENCRYPTION
162 const struct llcrypt_operations *lsi_cop;
163 struct key *lsi_master_keys; /* master crypto keys used */
164 #elif defined(HAVE_LUSTRE_CRYPTO) && !defined(HAVE_FSCRYPT_DUMMY_CONTEXT_ENABLED)
165 /* Dummy Encryption policy for '-o test_dummy_encryption' */
166 struct llcrypt_dummy_policy lsi_dummy_enc_policy;
170 #define LSI_UMOUNT_FAILOVER 0x00200000
171 #ifndef HAVE_SUPER_SETUP_BDI_NAME
172 #define LSI_BDI_INITIALIZED 0x00400000
174 #ifdef CONFIG_LL_ENCRYPTION
175 #define LSI_FILENAME_ENC 0x00800000 /* enable name encryption */
177 #define LSI_FILENAME_ENC_B64_OLD_CLI 0x01000000 /* use old style base64 */
179 #define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
180 #define s2lsi_nocast(sb) ((sb)->s_fs_info)
182 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
183 #define get_mount_fileset(sb) (s2lsi(sb)->lsi_lmd->lmd_fileset)
185 /* opc for target register, see also uapi/linux/lustre/lustre_disk.h.
186 * For mti_flags the lower 16 bits are used for mount options so these
187 * have to be masked out with LDD_F_MASK. Otherwise these values will
188 * be seen as unsupported mount options. Bit 16 is already used by
189 * LDD_F_NO_LOCAL_LOGS so 17 is next free bit.
191 enum ldd_target_flags {
192 LDD_F_LARGE_NID = BIT(17), /* 0x20000 */
195 static inline bool target_supports_large_nid(struct mgs_target_info *mti)
197 return mti->mti_flags & LDD_F_LARGE_NID;
200 # ifdef HAVE_SERVER_SUPPORT
201 /* opc for target register */
202 #define LDD_F_OPC_REG 0x10000000 /* bit 28 */
203 #define LDD_F_OPC_UNREG 0x20000000 /* bit 29 */
204 #define LDD_F_OPC_READY 0x40000000 /* bit 30 */
205 #define LDD_F_OPC_MASK 0xf0000000
207 #define LDD_F_MASK 0xFFFF
210 * This limit is arbitrary (131072 clients on x86), but it is convenient to use
211 * 2^n * PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
212 * If we need more than 131072 clients (order-2 allocation on x86) then this
213 * should become an array of single-page pointers that are allocated on demand.
215 #if (128 * 1024UL) > (PAGE_SIZE * 8)
216 #define LR_MAX_CLIENTS (128 * 1024UL)
218 #define LR_MAX_CLIENTS (PAGE_SIZE * 8)
221 /** COMPAT_146: this is an OST (temporary) */
222 #define OBD_COMPAT_OST 0x00000002
223 /** COMPAT_146: this is an MDT (temporary) */
224 #define OBD_COMPAT_MDT 0x00000004
225 /** 2.0 server, interop flag to show server version is changed */
226 #define OBD_COMPAT_20 0x00000008
228 /** MDS handles LOV_OBJID file */
229 #define OBD_ROCOMPAT_LOVOBJID 0x00000001
230 /** store OST index in the IDIF */
231 #define OBD_ROCOMPAT_IDX_IN_IDIF 0x00000002
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
249 /** lmm_stripe_count has been shrunk from u32 to u16 and the remaining 16
250 * bits are now used to store a generation. Once we start changing the layout
251 * and bumping the generation, old versions expecting a 32-bit lmm_stripe_count
252 * will be confused by interpreting stripe_count | gen << 16 as the actual
254 #define OBD_INCOMPAT_LMM_VER 0x00000100
255 /** multiple OI files for MDT */
256 #define OBD_INCOMPAT_MULTI_OI 0x00000200
257 /** multiple RPCs in flight */
258 #define OBD_INCOMPAT_MULTI_RPCS 0x00000400
260 /* last_rcvd handling */
261 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
262 struct lr_server_data *lsd)
266 memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof(lsd->lsd_uuid));
267 lsd->lsd_last_transno = le64_to_cpu(buf->lsd_last_transno);
268 lsd->lsd_compat14 = le64_to_cpu(buf->lsd_compat14);
269 lsd->lsd_mount_count = le64_to_cpu(buf->lsd_mount_count);
270 lsd->lsd_feature_compat = le32_to_cpu(buf->lsd_feature_compat);
271 lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
272 lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
273 lsd->lsd_server_size = le32_to_cpu(buf->lsd_server_size);
274 lsd->lsd_client_start = le32_to_cpu(buf->lsd_client_start);
275 lsd->lsd_client_size = le16_to_cpu(buf->lsd_client_size);
276 lsd->lsd_subdir_count = le16_to_cpu(buf->lsd_subdir_count);
277 lsd->lsd_catalog_oid = le64_to_cpu(buf->lsd_catalog_oid);
278 lsd->lsd_catalog_ogen = le32_to_cpu(buf->lsd_catalog_ogen);
279 memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
280 lsd->lsd_osd_index = le32_to_cpu(buf->lsd_osd_index);
281 lsd->lsd_padding1 = le32_to_cpu(buf->lsd_padding1);
282 lsd->lsd_start_epoch = le32_to_cpu(buf->lsd_start_epoch);
283 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
284 lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
285 lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
286 lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
289 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
290 struct lr_server_data *buf)
294 memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof(buf->lsd_uuid));
295 buf->lsd_last_transno = cpu_to_le64(lsd->lsd_last_transno);
296 buf->lsd_compat14 = cpu_to_le64(lsd->lsd_compat14);
297 buf->lsd_mount_count = cpu_to_le64(lsd->lsd_mount_count);
298 buf->lsd_feature_compat = cpu_to_le32(lsd->lsd_feature_compat);
299 buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
300 buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
301 buf->lsd_server_size = cpu_to_le32(lsd->lsd_server_size);
302 buf->lsd_client_start = cpu_to_le32(lsd->lsd_client_start);
303 buf->lsd_client_size = cpu_to_le16(lsd->lsd_client_size);
304 buf->lsd_subdir_count = cpu_to_le16(lsd->lsd_subdir_count);
305 buf->lsd_catalog_oid = cpu_to_le64(lsd->lsd_catalog_oid);
306 buf->lsd_catalog_ogen = cpu_to_le32(lsd->lsd_catalog_ogen);
307 memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
308 buf->lsd_osd_index = cpu_to_le32(lsd->lsd_osd_index);
309 buf->lsd_padding1 = cpu_to_le32(lsd->lsd_padding1);
310 buf->lsd_start_epoch = cpu_to_le32(lsd->lsd_start_epoch);
311 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
312 buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
313 buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
314 buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
317 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
318 struct lsd_client_data *lcd)
320 memcpy(lcd->lcd_uuid, buf->lcd_uuid, sizeof (lcd->lcd_uuid));
321 lcd->lcd_last_transno = le64_to_cpu(buf->lcd_last_transno);
322 lcd->lcd_last_xid = le64_to_cpu(buf->lcd_last_xid);
323 lcd->lcd_last_result = le32_to_cpu(buf->lcd_last_result);
324 lcd->lcd_last_data = le32_to_cpu(buf->lcd_last_data);
325 lcd->lcd_last_close_transno = le64_to_cpu(buf->lcd_last_close_transno);
326 lcd->lcd_last_close_xid = le64_to_cpu(buf->lcd_last_close_xid);
327 lcd->lcd_last_close_result = le32_to_cpu(buf->lcd_last_close_result);
328 lcd->lcd_last_close_data = le32_to_cpu(buf->lcd_last_close_data);
329 lcd->lcd_pre_versions[0] = le64_to_cpu(buf->lcd_pre_versions[0]);
330 lcd->lcd_pre_versions[1] = le64_to_cpu(buf->lcd_pre_versions[1]);
331 lcd->lcd_pre_versions[2] = le64_to_cpu(buf->lcd_pre_versions[2]);
332 lcd->lcd_pre_versions[3] = le64_to_cpu(buf->lcd_pre_versions[3]);
333 lcd->lcd_last_epoch = le32_to_cpu(buf->lcd_last_epoch);
334 lcd->lcd_generation = le32_to_cpu(buf->lcd_generation);
337 static inline void lcd_cpu_to_le(struct lsd_client_data *lcd,
338 struct lsd_client_data *buf)
340 memcpy(buf->lcd_uuid, lcd->lcd_uuid, sizeof (lcd->lcd_uuid));
341 buf->lcd_last_transno = cpu_to_le64(lcd->lcd_last_transno);
342 buf->lcd_last_xid = cpu_to_le64(lcd->lcd_last_xid);
343 buf->lcd_last_result = cpu_to_le32(lcd->lcd_last_result);
344 buf->lcd_last_data = cpu_to_le32(lcd->lcd_last_data);
345 buf->lcd_last_close_transno = cpu_to_le64(lcd->lcd_last_close_transno);
346 buf->lcd_last_close_xid = cpu_to_le64(lcd->lcd_last_close_xid);
347 buf->lcd_last_close_result = cpu_to_le32(lcd->lcd_last_close_result);
348 buf->lcd_last_close_data = cpu_to_le32(lcd->lcd_last_close_data);
349 buf->lcd_pre_versions[0] = cpu_to_le64(lcd->lcd_pre_versions[0]);
350 buf->lcd_pre_versions[1] = cpu_to_le64(lcd->lcd_pre_versions[1]);
351 buf->lcd_pre_versions[2] = cpu_to_le64(lcd->lcd_pre_versions[2]);
352 buf->lcd_pre_versions[3] = cpu_to_le64(lcd->lcd_pre_versions[3]);
353 buf->lcd_last_epoch = cpu_to_le32(lcd->lcd_last_epoch);
354 buf->lcd_generation = cpu_to_le32(lcd->lcd_generation);
357 static inline u64 lcd_last_transno(struct lsd_client_data *lcd)
359 return (lcd->lcd_last_transno > lcd->lcd_last_close_transno ?
360 lcd->lcd_last_transno : lcd->lcd_last_close_transno);
363 static inline u64 lcd_last_xid(struct lsd_client_data *lcd)
365 return (lcd->lcd_last_xid > lcd->lcd_last_close_xid ?
366 lcd->lcd_last_xid : lcd->lcd_last_close_xid);
369 /****************** mount lookup info *********************/
371 struct lustre_mount_info {
373 struct super_block *lmi_sb;
374 struct list_head lmi_list_chain;
377 /****************** prototypes *********************/
379 /* obd_mount_server.c */
380 int server_fill_super(struct super_block *sb);
381 struct lustre_mount_info *server_get_mount(const char *name);
382 int server_put_mount(const char *name, bool dereg_mnt);
383 struct mgs_target_info;
384 int server_mti_print(const char *title, struct mgs_target_info *mti);
385 void server_calc_timeout(struct lustre_sb_info *lsi, struct obd_device *obd);
388 int server_name2svname(const char *label, char *svname, const char **endptr,
391 int server_name_is_ost(const char *svname);
392 int target_name2index(const char *svname, u32 *idx, const char **endptr);
394 int lustre_put_lsi(struct super_block *sb);
395 int lustre_start_simple(char *obdname, char *type, char *uuid,
396 char *s1, char *s2, char *s3, char *s4);
397 int lustre_stop_mgc(struct super_block *sb);
398 #endif /* HAVE_SERVER_SUPPORT */
399 int server_name2fsname(const char *svname, char *fsname, const char **endptr);
400 void obdname2fsname(const char *tgt, char *fsname, size_t fslen);
402 int lustre_start_mgc(struct super_block *sb);
403 int lustre_common_put_super(struct super_block *sb);
405 struct lustre_sb_info *lustre_init_lsi(struct super_block *sb);
406 int lustre_put_lsi(struct super_block *sb);
407 int lmd_parse(char *options, struct lustre_mount_data *lmd);
410 int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id,
411 enum mgs_cfg_type type);
412 int mgc_logname2resid(char *fsname, struct ldlm_res_id *res_id,
413 enum mgs_cfg_type type);
417 #endif /* _LUSTRE_DISK_H */