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 #include <uapi/linux/lustre/lustre_disk.h>
51 #include <uapi/linux/lustre/lustre_idl.h>
53 #define IS_MDT(data) ((data)->lsi_flags & LDD_F_SV_TYPE_MDT)
54 #define IS_OST(data) ((data)->lsi_flags & LDD_F_SV_TYPE_OST)
55 #define IS_MGS(data) ((data)->lsi_flags & LDD_F_SV_TYPE_MGS)
56 #define IS_SERVER(data) ((data)->lsi_flags & (LDD_F_SV_TYPE_MGS | \
59 #define MT_STR(data) mt_str((data)->ldd_mount_type)
61 /****************** mount command *********************/
63 /* The lmd is only used internally by Lustre; mount simply passes
64 * everything as string options
66 #define LMD_MAGIC 0xbdacbd03
67 #define LMD_PARAMS_MAXLEN 4096
69 /* gleaned from the mount command - no persistent info here */
70 struct lustre_mount_data {
72 u32 lmd_flags; /* lustre mount flags */
73 int lmd_mgs_failnodes; /* mgs failover node count */
74 int lmd_exclude_count;
75 int lmd_recovery_time_soft;
76 int lmd_recovery_time_hard;
77 char *lmd_dev; /* device name */
78 char *lmd_profile; /* client only */
79 char *lmd_fileset; /* mount fileset */
80 char *lmd_mgssec; /* sptlrpc flavor to mgs */
81 char *lmd_opts; /* lustre mount options (as opposed to
82 * device_ mount options) */
83 char *lmd_params; /* lustre params */
84 u32 *lmd_exclude; /* array of OSTs to ignore */
85 char *lmd_mgs; /* MGS nid */
86 char *lmd_osd_type; /* OSD type */
87 char *lmd_nidnet; /* network to restrict this client to */
90 #define LMD_FLG_SERVER 0x0001 /* Mounting a server */
91 #define LMD_FLG_CLIENT 0x0002 /* Mounting a client */
92 #define LMD_FLG_SKIP_LFSCK 0x0004 /* NOT auto resume LFSCK when mount */
93 #define LMD_FLG_ABORT_RECOV 0x0008 /* Abort recovery */
94 #define LMD_FLG_NOSVC 0x0010 /* Only start MGS/MGC for servers,
96 #define LMD_FLG_NOMGS 0x0020 /* Only start target for servers, reusing
97 existing MGS services */
98 #define LMD_FLG_WRITECONF 0x0040 /* Rewrite config log */
99 #define LMD_FLG_NOIR 0x0080 /* NO imperative recovery */
100 #define LMD_FLG_NOSCRUB 0x0100 /* Do not trigger scrub automatically */
101 #define LMD_FLG_MGS 0x0200 /* Also start MGS along with server */
102 #define LMD_FLG_IAM 0x0400 /* IAM dir */
103 #define LMD_FLG_NO_PRIMNODE 0x0800 /* all nodes are service nodes */
104 #define LMD_FLG_VIRGIN 0x1000 /* the service registers first time */
105 #define LMD_FLG_UPDATE 0x2000 /* update parameters */
106 #define LMD_FLG_HSM 0x4000 /* Start coordinator */
107 #define LMD_FLG_DEV_RDONLY 0x8000 /* discard modification quitely */
108 #define LMD_FLG_NO_PRECREATE 0x10000 /* do not allow OST object creation */
109 #define LMD_FLG_LOCAL_RECOV 0x20000 /* force recovery for local clients */
110 #define LMD_FLG_ABORT_RECOV_MDT 0x40000 /* Abort recovery between MDTs */
111 #define LMD_FLG_NO_LOCAL_LOGS 0x80000 /* Use config logs from MGS */
113 #define lmd_is_client(x) ((x)->lmd_flags & LMD_FLG_CLIENT)
115 /****************** superblock additional info *********************/
119 struct lustre_sb_info {
121 struct obd_device *lsi_mgc; /* mgc obd */
122 struct lustre_mount_data *lsi_lmd; /* mount command info */
123 struct ll_sb_info *lsi_llsbi; /* add'l client sbi info */
124 struct dt_device *lsi_dt_dev; /* dt device to access disk fs*/
125 atomic_t lsi_mounts; /* references to the srv_mnt */
126 struct kobject *lsi_kobj;
127 char lsi_svname[MTI_NAME_MAXLEN];
128 /* lsi_osd_obdname format = 'lsi->ls_svname'-osd */
129 char lsi_osd_obdname[MTI_NAME_MAXLEN + 4];
130 /* lsi_osd_uuid format = 'lsi->ls_osd_obdname'_UUID */
131 char lsi_osd_uuid[MTI_NAME_MAXLEN + 9];
132 struct obd_export *lsi_osd_exp;
133 char lsi_osd_type[16];
135 struct backing_dev_info lsi_bdi; /* each client mountpoint needs
136 own backing_dev_info */
137 /* protect lsi_lwp_list */
138 struct mutex lsi_lwp_mutex;
139 struct list_head lsi_lwp_list;
140 unsigned long lsi_lwp_started:1,
141 lsi_server_started:1;
142 #ifdef CONFIG_LL_ENCRYPTION
143 const struct llcrypt_operations *lsi_cop;
144 struct key *lsi_master_keys; /* master crypto keys used */
148 #define LSI_UMOUNT_FAILOVER 0x00200000
150 #define s2lsi(sb) ((struct lustre_sb_info *)((sb)->s_fs_info))
151 #define s2lsi_nocast(sb) ((sb)->s_fs_info)
153 #define get_profile_name(sb) (s2lsi(sb)->lsi_lmd->lmd_profile)
154 #define get_mount_fileset(sb) (s2lsi(sb)->lsi_lmd->lmd_fileset)
156 # ifdef HAVE_SERVER_SUPPORT
157 /* opc for target register */
158 #define LDD_F_OPC_REG 0x10000000
159 #define LDD_F_OPC_UNREG 0x20000000
160 #define LDD_F_OPC_READY 0x40000000
161 #define LDD_F_OPC_MASK 0xf0000000
163 #define LDD_F_MASK 0xFFFF
166 * This limit is arbitrary (131072 clients on x86), but it is convenient to use
167 * 2^n * PAGE_SIZE * 8 for the number of bits that fit an order-n allocation.
168 * If we need more than 131072 clients (order-2 allocation on x86) then this
169 * should become an array of single-page pointers that are allocated on demand.
171 #if (128 * 1024UL) > (PAGE_SIZE * 8)
172 #define LR_MAX_CLIENTS (128 * 1024UL)
174 #define LR_MAX_CLIENTS (PAGE_SIZE * 8)
177 /** COMPAT_146: this is an OST (temporary) */
178 #define OBD_COMPAT_OST 0x00000002
179 /** COMPAT_146: this is an MDT (temporary) */
180 #define OBD_COMPAT_MDT 0x00000004
181 /** 2.0 server, interop flag to show server version is changed */
182 #define OBD_COMPAT_20 0x00000008
184 /** MDS handles LOV_OBJID file */
185 #define OBD_ROCOMPAT_LOVOBJID 0x00000001
186 /** store OST index in the IDIF */
187 #define OBD_ROCOMPAT_IDX_IN_IDIF 0x00000002
189 /** OST handles group subdirs */
190 #define OBD_INCOMPAT_GROUPS 0x00000001
191 /** this is an OST */
192 #define OBD_INCOMPAT_OST 0x00000002
193 /** this is an MDT */
194 #define OBD_INCOMPAT_MDT 0x00000004
195 /** common last_rvcd format */
196 #define OBD_INCOMPAT_COMMON_LR 0x00000008
197 /** FID is enabled */
198 #define OBD_INCOMPAT_FID 0x00000010
199 /** Size-on-MDS is enabled */
200 #define OBD_INCOMPAT_SOM 0x00000020
201 /** filesystem using iam format to store directory entries */
202 #define OBD_INCOMPAT_IAM_DIR 0x00000040
203 /** LMA attribute contains per-inode incompatible flags */
204 #define OBD_INCOMPAT_LMA 0x00000080
205 /** lmm_stripe_count has been shrunk from u32 to u16 and the remaining 16
206 * bits are now used to store a generation. Once we start changing the layout
207 * and bumping the generation, old versions expecting a 32-bit lmm_stripe_count
208 * will be confused by interpreting stripe_count | gen << 16 as the actual
210 #define OBD_INCOMPAT_LMM_VER 0x00000100
211 /** multiple OI files for MDT */
212 #define OBD_INCOMPAT_MULTI_OI 0x00000200
213 /** multiple RPCs in flight */
214 #define OBD_INCOMPAT_MULTI_RPCS 0x00000400
216 /* last_rcvd handling */
217 static inline void lsd_le_to_cpu(struct lr_server_data *buf,
218 struct lr_server_data *lsd)
222 memcpy(lsd->lsd_uuid, buf->lsd_uuid, sizeof(lsd->lsd_uuid));
223 lsd->lsd_last_transno = le64_to_cpu(buf->lsd_last_transno);
224 lsd->lsd_compat14 = le64_to_cpu(buf->lsd_compat14);
225 lsd->lsd_mount_count = le64_to_cpu(buf->lsd_mount_count);
226 lsd->lsd_feature_compat = le32_to_cpu(buf->lsd_feature_compat);
227 lsd->lsd_feature_rocompat = le32_to_cpu(buf->lsd_feature_rocompat);
228 lsd->lsd_feature_incompat = le32_to_cpu(buf->lsd_feature_incompat);
229 lsd->lsd_server_size = le32_to_cpu(buf->lsd_server_size);
230 lsd->lsd_client_start = le32_to_cpu(buf->lsd_client_start);
231 lsd->lsd_client_size = le16_to_cpu(buf->lsd_client_size);
232 lsd->lsd_subdir_count = le16_to_cpu(buf->lsd_subdir_count);
233 lsd->lsd_catalog_oid = le64_to_cpu(buf->lsd_catalog_oid);
234 lsd->lsd_catalog_ogen = le32_to_cpu(buf->lsd_catalog_ogen);
235 memcpy(lsd->lsd_peeruuid, buf->lsd_peeruuid, sizeof(lsd->lsd_peeruuid));
236 lsd->lsd_osd_index = le32_to_cpu(buf->lsd_osd_index);
237 lsd->lsd_padding1 = le32_to_cpu(buf->lsd_padding1);
238 lsd->lsd_start_epoch = le32_to_cpu(buf->lsd_start_epoch);
239 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
240 lsd->lsd_trans_table[i] = le64_to_cpu(buf->lsd_trans_table[i]);
241 lsd->lsd_trans_table_time = le32_to_cpu(buf->lsd_trans_table_time);
242 lsd->lsd_expire_intervals = le32_to_cpu(buf->lsd_expire_intervals);
245 static inline void lsd_cpu_to_le(struct lr_server_data *lsd,
246 struct lr_server_data *buf)
250 memcpy(buf->lsd_uuid, lsd->lsd_uuid, sizeof(buf->lsd_uuid));
251 buf->lsd_last_transno = cpu_to_le64(lsd->lsd_last_transno);
252 buf->lsd_compat14 = cpu_to_le64(lsd->lsd_compat14);
253 buf->lsd_mount_count = cpu_to_le64(lsd->lsd_mount_count);
254 buf->lsd_feature_compat = cpu_to_le32(lsd->lsd_feature_compat);
255 buf->lsd_feature_rocompat = cpu_to_le32(lsd->lsd_feature_rocompat);
256 buf->lsd_feature_incompat = cpu_to_le32(lsd->lsd_feature_incompat);
257 buf->lsd_server_size = cpu_to_le32(lsd->lsd_server_size);
258 buf->lsd_client_start = cpu_to_le32(lsd->lsd_client_start);
259 buf->lsd_client_size = cpu_to_le16(lsd->lsd_client_size);
260 buf->lsd_subdir_count = cpu_to_le16(lsd->lsd_subdir_count);
261 buf->lsd_catalog_oid = cpu_to_le64(lsd->lsd_catalog_oid);
262 buf->lsd_catalog_ogen = cpu_to_le32(lsd->lsd_catalog_ogen);
263 memcpy(buf->lsd_peeruuid, lsd->lsd_peeruuid, sizeof(buf->lsd_peeruuid));
264 buf->lsd_osd_index = cpu_to_le32(lsd->lsd_osd_index);
265 buf->lsd_padding1 = cpu_to_le32(lsd->lsd_padding1);
266 buf->lsd_start_epoch = cpu_to_le32(lsd->lsd_start_epoch);
267 for (i = 0; i < LR_EXPIRE_INTERVALS; i++)
268 buf->lsd_trans_table[i] = cpu_to_le64(lsd->lsd_trans_table[i]);
269 buf->lsd_trans_table_time = cpu_to_le32(lsd->lsd_trans_table_time);
270 buf->lsd_expire_intervals = cpu_to_le32(lsd->lsd_expire_intervals);
273 static inline void lcd_le_to_cpu(struct lsd_client_data *buf,
274 struct lsd_client_data *lcd)
276 memcpy(lcd->lcd_uuid, buf->lcd_uuid, sizeof (lcd->lcd_uuid));
277 lcd->lcd_last_transno = le64_to_cpu(buf->lcd_last_transno);
278 lcd->lcd_last_xid = le64_to_cpu(buf->lcd_last_xid);
279 lcd->lcd_last_result = le32_to_cpu(buf->lcd_last_result);
280 lcd->lcd_last_data = le32_to_cpu(buf->lcd_last_data);
281 lcd->lcd_last_close_transno = le64_to_cpu(buf->lcd_last_close_transno);
282 lcd->lcd_last_close_xid = le64_to_cpu(buf->lcd_last_close_xid);
283 lcd->lcd_last_close_result = le32_to_cpu(buf->lcd_last_close_result);
284 lcd->lcd_last_close_data = le32_to_cpu(buf->lcd_last_close_data);
285 lcd->lcd_pre_versions[0] = le64_to_cpu(buf->lcd_pre_versions[0]);
286 lcd->lcd_pre_versions[1] = le64_to_cpu(buf->lcd_pre_versions[1]);
287 lcd->lcd_pre_versions[2] = le64_to_cpu(buf->lcd_pre_versions[2]);
288 lcd->lcd_pre_versions[3] = le64_to_cpu(buf->lcd_pre_versions[3]);
289 lcd->lcd_last_epoch = le32_to_cpu(buf->lcd_last_epoch);
290 lcd->lcd_generation = le32_to_cpu(buf->lcd_generation);
293 static inline void lcd_cpu_to_le(struct lsd_client_data *lcd,
294 struct lsd_client_data *buf)
296 memcpy(buf->lcd_uuid, lcd->lcd_uuid, sizeof (lcd->lcd_uuid));
297 buf->lcd_last_transno = cpu_to_le64(lcd->lcd_last_transno);
298 buf->lcd_last_xid = cpu_to_le64(lcd->lcd_last_xid);
299 buf->lcd_last_result = cpu_to_le32(lcd->lcd_last_result);
300 buf->lcd_last_data = cpu_to_le32(lcd->lcd_last_data);
301 buf->lcd_last_close_transno = cpu_to_le64(lcd->lcd_last_close_transno);
302 buf->lcd_last_close_xid = cpu_to_le64(lcd->lcd_last_close_xid);
303 buf->lcd_last_close_result = cpu_to_le32(lcd->lcd_last_close_result);
304 buf->lcd_last_close_data = cpu_to_le32(lcd->lcd_last_close_data);
305 buf->lcd_pre_versions[0] = cpu_to_le64(lcd->lcd_pre_versions[0]);
306 buf->lcd_pre_versions[1] = cpu_to_le64(lcd->lcd_pre_versions[1]);
307 buf->lcd_pre_versions[2] = cpu_to_le64(lcd->lcd_pre_versions[2]);
308 buf->lcd_pre_versions[3] = cpu_to_le64(lcd->lcd_pre_versions[3]);
309 buf->lcd_last_epoch = cpu_to_le32(lcd->lcd_last_epoch);
310 buf->lcd_generation = cpu_to_le32(lcd->lcd_generation);
313 static inline u64 lcd_last_transno(struct lsd_client_data *lcd)
315 return (lcd->lcd_last_transno > lcd->lcd_last_close_transno ?
316 lcd->lcd_last_transno : lcd->lcd_last_close_transno);
319 static inline u64 lcd_last_xid(struct lsd_client_data *lcd)
321 return (lcd->lcd_last_xid > lcd->lcd_last_close_xid ?
322 lcd->lcd_last_xid : lcd->lcd_last_close_xid);
325 /****************** mount lookup info *********************/
327 struct lustre_mount_info {
329 struct super_block *lmi_sb;
330 struct list_head lmi_list_chain;
333 /****************** prototypes *********************/
335 /* obd_mount_server.c */
336 int server_fill_super(struct super_block *sb);
337 struct lustre_mount_info *server_get_mount(const char *name);
338 int server_put_mount(const char *name, bool dereg_mnt);
339 struct mgs_target_info;
340 int server_mti_print(const char *title, struct mgs_target_info *mti);
341 void server_calc_timeout(struct lustre_sb_info *lsi, struct obd_device *obd);
344 int server_name2svname(const char *label, char *svname, const char **endptr,
347 int server_name_is_ost(const char *svname);
348 int target_name2index(const char *svname, u32 *idx, const char **endptr);
350 int lustre_put_lsi(struct super_block *sb);
351 int lustre_start_simple(char *obdname, char *type, char *uuid,
352 char *s1, char *s2, char *s3, char *s4);
353 #endif /* HAVE_SERVER_SUPPORT */
354 int server_name2fsname(const char *svname, char *fsname, const char **endptr);
355 void obdname2fsname(const char *tgt, char *fsname, size_t fslen);
357 int lustre_start_mgc(struct super_block *sb);
358 int lustre_common_put_super(struct super_block *sb);
360 struct lustre_sb_info *lustre_init_lsi(struct super_block *sb);
361 int lustre_put_lsi(struct super_block *sb);
362 int lmd_parse(char *options, struct lustre_mount_data *lmd);
365 int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id, int type);
366 int mgc_logname2resid(char *fsname, struct ldlm_res_id *res_id, int type);
370 #endif /* _LUSTRE_DISK_H */