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_fid.h
33 * Author: Yury Umanets <umka@clusterfs.com>
36 #ifndef __LUSTRE_FID_H
37 #define __LUSTRE_FID_H
43 * http://wiki.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
44 * describes the FID namespace and interoperability requirements for FIDs.
45 * The important parts of that document are included here for reference.
48 * File IDentifier generated by client from range allocated by the SEQuence
49 * service and stored in struct lu_fid. The FID is composed of three parts:
50 * SEQuence, ObjectID, and VERsion. The SEQ component is a filesystem
51 * unique 64-bit integer, and only one client is ever assigned any SEQ value.
52 * The first 0x400 FID_SEQ_NORMAL [2^33, 2^33 + 0x400] values are reserved
53 * for system use. The OID component is a 32-bit value generated by the
54 * client on a per-SEQ basis to allow creating many unique FIDs without
55 * communication with the server. The VER component is a 32-bit value that
56 * distinguishes between different FID instantiations, such as snapshots or
57 * separate subtrees within the filesystem. FIDs with the same VER field
58 * are considered part of the same namespace.
60 * OLD filesystems are those upgraded from Lustre 1.x that predate FIDs, and
61 * MDTs use 32-bit ldiskfs internal inode/generation numbers (IGIFs), while
62 * OSTs use 64-bit Lustre object IDs and generation numbers.
64 * NEW filesystems are those formatted since the introduction of FIDs.
67 * Inode and Generation In FID, a surrogate FID used to globally identify
68 * an existing object on OLD formatted MDT file system. This would only be
69 * used on MDT0 in a DNE filesystem, because there cannot be more than one
70 * MDT in an OLD formatted filesystem. Belongs to sequence in [12, 2^32 - 1]
71 * range, where inode number is stored in SEQ, and inode generation is in OID.
72 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
73 * which is the maximum possible for an ldiskfs backend. It also assumes
74 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
75 * to clients, which has always been true.
78 * object ID In FID, a surrogate FID used to globally identify an existing
79 * OST object on OLD formatted OST file system. Belongs to a sequence in
80 * [2^32, 2^33 - 1]. Sequence number is calculated as:
82 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
84 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object
85 * ID. The generation of unique SEQ values per OST allows the IDIF FIDs to
86 * be identified in the FLD correctly. The OID field is calculated as:
90 * that is, it consists of lower 32 bits of object ID. For objects within
91 * the IDIF range, object ID extraction will be:
93 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
94 * o_seq = 0; // formerly group number
96 * NOTE: This assumes that no more than 2^48-1 objects have ever been created
97 * on any OST, and that no more than 65535 OSTs are in use. Both are very
98 * reasonable assumptions, i.e. an IDIF can uniquely map all objects assuming
99 * a maximum creation rate of 1M objects per second for a maximum of 9 years,
100 * or combinations thereof.
103 * Surrogate FID used to identify an existing object on OLD formatted OST
104 * filesystem. Belongs to the reserved SEQuence 0, and is used prior to
105 * the introduction of FID-on-OST, at which point IDIF will be used to
106 * identify objects as residing on a specific OST.
109 * For Lustre Log objects the object sequence 1 is used. This is compatible
110 * with both OLD and NEW namespaces, as this SEQ number is in the
111 * ext3/ldiskfs reserved inode range and does not conflict with IGIF
115 * For testing OST IO performance the object sequence 2 is used. This is
116 * compatible with both OLD and NEW namespaces, as this SEQ number is in
117 * the ext3/ldiskfs reserved inode range and does not conflict with IGIF
120 * OST_MDT1 .. OST_MAX
121 * For testing with multiple MDTs the object sequence 3 through 9 is used,
122 * allowing direct mapping of MDTs 1 through 7 respectively, for a total
123 * of 8 MDTs including OST_MDT0. This matches the legacy CMD project "group"
124 * mappings. However, this SEQ range is only for testing prior to any
125 * production DNE release, as the objects in this range conflict across all
126 * OSTs, as the OST index is not part of the FID. For production DNE usage,
127 * OST objects created by MDT1+ will use FID_SEQ_NORMAL FIDs.
129 * DLM OST objid to IDIF mapping
130 * For compatibility with existing OLD OST network protocol structures, the
131 * FID must map onto the o_id and o_seq in a manner that ensures existing
132 * objects are identified consistently for IO, as well as onto the LDLM
133 * namespace to ensure IDIFs there is only a single resource name for any
134 * object in the DLM. The OLD OST object DLM resource mapping is:
136 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
138 * The NEW OST object DLM resource mapping is the same for both MDT and OST:
140 * resource[] = {SEQ, OID, VER, HASH};
142 * NOTE: for mapping IDIF values to DLM resource names the o_id may be
143 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
144 * for the o_id numbers to overlap FID SEQ numbers in the resource. However,
145 * in all production releases the OLD o_seq field is always zero, and all
146 * valid FID OID values are non-zero, so the lock resources will not collide.
147 * Even so, the MDT and OST resources are also in different LDLM namespaces.
150 #include <libcfs/libcfs.h>
151 #include <lu_object.h>
152 #include <uapi/linux/lustre/lustre_fid.h>
153 #include <uapi/linux/lustre/lustre_idl.h>
154 #include <uapi/linux/lustre/lustre_ostid.h>
156 /* Lustre service names are following the format
157 * service name + MDT + seq name
159 #define LUSTRE_MDT_MAXNAMELEN 80
167 /* Whole sequences space range and zero range definitions */
168 extern const struct lu_seq_range LUSTRE_SEQ_SPACE_RANGE;
169 extern const struct lu_seq_range LUSTRE_SEQ_ZERO_RANGE;
170 extern const struct lu_fid LUSTRE_BFL_FID;
171 extern const struct lu_fid LU_LPF_FID;
172 extern const struct lu_fid LU_BACKEND_LPF_FID;
175 /* Max Number metadata FIDs allocated in one sequence(128k) */
176 LUSTRE_METADATA_SEQ_MAX_WIDTH = 0x0000000000020000ULL,
178 /* Max Number data FIDs allocated in one sequence(32M - 1) */
179 LUSTRE_DATA_SEQ_MAX_WIDTH = 0x0000000001FFFFFFULL,
181 /* How many sequences to allocate to a client at once */
182 LUSTRE_SEQ_META_WIDTH = 0x0000000000000001ULL,
184 /* seq allocation pool size. */
185 LUSTRE_SEQ_BATCH_WIDTH = LUSTRE_SEQ_META_WIDTH * 1000,
187 /* Max number sequences in one super-sequence allocated to MDTs */
188 LUSTRE_SEQ_SUPER_WIDTH = ((1ULL << 30ULL) * LUSTRE_SEQ_META_WIDTH)
191 /** special OID for local objects */
193 /** \see fld_mod_init */
195 /** \see fid_mod_init */
196 FID_SEQ_CTL_OID = 4UL,
197 FID_SEQ_SRV_OID = 5UL,
198 /** \see mdd_mod_init */
199 MDD_ROOT_INDEX_OID = 6UL, /* deprecated in 2.4 */
200 MDD_ORPHAN_OID = 7UL, /* deprecated in 2.4 */
201 MDD_LOV_OBJ_OID = 8UL,
202 MDD_CAPA_KEYS_OID = 9UL,
203 /** \see mdt_mod_init */
204 LAST_RECV_OID = 11UL,
205 OSD_FS_ROOT_OID = 13UL,
206 ACCT_USER_OID = 15UL,
207 ACCT_GROUP_OID = 16UL,
208 LFSCK_BOOKMARK_OID = 17UL,
209 OTABLE_IT_OID = 18UL,
211 REPLY_DATA_OID = 21UL,
212 ACCT_PROJECT_OID = 22UL,
213 INDEX_BACKUP_OID = 4116UL,
214 OFD_LAST_GROUP_OID = 4117UL,
215 LLOG_CATALOGS_OID = 4118UL,
216 MGS_CONFIGS_OID = 4119UL,
217 OFD_HEALTH_CHECK_OID = 4120UL,
218 MDD_LOV_OBJ_OSEQ = 4121UL,
219 LFSCK_NAMESPACE_OID = 4122UL,
220 REMOTE_PARENT_DIR_OID = 4123UL,
221 /* This definition is obsolete
222 * SLAVE_LLOG_CATALOGS_OID = 4124UL,
224 BATCHID_COMMITTED_OID = 4125UL,
227 static inline void lu_local_obj_fid(struct lu_fid *fid, __u32 oid)
229 fid->f_seq = FID_SEQ_LOCAL_FILE;
234 static inline void lu_local_name_obj_fid(struct lu_fid *fid, __u32 oid)
236 fid->f_seq = FID_SEQ_LOCAL_NAME;
241 /* For new FS (>= 2.4), the root FID will be changed to
242 * [FID_SEQ_ROOT:1:0], for existing FS, (upgraded to 2.4),
243 * the root FID will still be IGIF
245 static inline int fid_is_root(const struct lu_fid *fid)
247 return unlikely(lu_fid_eq(fid, &LU_ROOT_FID));
250 static inline int fid_is_dot_lustre(const struct lu_fid *fid)
252 return unlikely(lu_fid_eq(fid, &LU_DOT_LUSTRE_FID));
255 static inline int fid_is_obf(const struct lu_fid *fid)
257 return unlikely(lu_fid_eq(fid, &LU_OBF_FID));
260 static inline int fid_is_otable_it(const struct lu_fid *fid)
262 return unlikely(fid_seq(fid) == FID_SEQ_LOCAL_FILE &&
263 fid_oid(fid) == OTABLE_IT_OID);
266 static inline int fid_oid_is_quota(const struct lu_fid *fid)
268 switch (fid_oid(fid)) {
271 case ACCT_PROJECT_OID:
278 static inline int fid_is_acct(const struct lu_fid *fid)
280 return fid_seq(fid) == FID_SEQ_LOCAL_FILE &&
281 fid_oid_is_quota(fid);
284 static inline int fid_is_quota(const struct lu_fid *fid)
286 return fid_seq(fid) == FID_SEQ_QUOTA ||
287 fid_seq(fid) == FID_SEQ_QUOTA_GLB;
290 static inline int fid_is_name_llog(const struct lu_fid *fid)
292 return fid_seq(fid) == FID_SEQ_LLOG_NAME;
295 static inline int fid_seq_in_fldb(u64 seq)
297 return fid_seq_is_igif(seq) || fid_seq_is_norm(seq) ||
298 fid_seq_is_root(seq) || fid_seq_is_dot(seq);
301 #ifdef HAVE_SERVER_SUPPORT
302 static inline int fid_is_namespace_visible(const struct lu_fid *fid)
304 const __u64 seq = fid_seq(fid);
306 /* Here, we cannot distinguish whether the normal FID is for OST
307 * object or not. It is caller's duty to check more if needed.
309 return (!fid_is_last_id(fid) &&
310 (fid_seq_is_norm(seq) || fid_seq_is_igif(seq))) ||
311 fid_is_root(fid) || fid_seq_is_dot(seq);
314 static inline void ost_layout_cpu_to_le(struct ost_layout *dst,
315 const struct ost_layout *src)
317 dst->ol_stripe_size = __cpu_to_le32(src->ol_stripe_size);
318 dst->ol_stripe_count = __cpu_to_le32(src->ol_stripe_count);
319 dst->ol_comp_start = __cpu_to_le64(src->ol_comp_start);
320 dst->ol_comp_end = __cpu_to_le64(src->ol_comp_end);
321 dst->ol_comp_id = __cpu_to_le32(src->ol_comp_id);
324 static inline void ost_layout_le_to_cpu(struct ost_layout *dst,
325 const struct ost_layout *src)
327 dst->ol_stripe_size = __le32_to_cpu(src->ol_stripe_size);
328 dst->ol_stripe_count = __le32_to_cpu(src->ol_stripe_count);
329 dst->ol_comp_start = __le64_to_cpu(src->ol_comp_start);
330 dst->ol_comp_end = __le64_to_cpu(src->ol_comp_end);
331 dst->ol_comp_id = __le32_to_cpu(src->ol_comp_id);
334 static inline void filter_fid_cpu_to_le(struct filter_fid *dst,
335 const struct filter_fid *src, int size)
337 fid_cpu_to_le(&dst->ff_parent, &src->ff_parent);
339 if (size < sizeof(struct filter_fid)) {
340 memset(&dst->ff_layout, 0, sizeof(dst->ff_layout));
342 ost_layout_cpu_to_le(&dst->ff_layout, &src->ff_layout);
343 dst->ff_layout_version = cpu_to_le32(src->ff_layout_version);
344 dst->ff_range = cpu_to_le32(src->ff_range);
347 /* XXX: Add more if filter_fid is enlarged in the future. */
350 static inline void filter_fid_le_to_cpu(struct filter_fid *dst,
351 const struct filter_fid *src, int size)
353 fid_le_to_cpu(&dst->ff_parent, &src->ff_parent);
355 if (size < sizeof(struct filter_fid)) {
356 memset(&dst->ff_layout, 0, sizeof(dst->ff_layout));
358 ost_layout_le_to_cpu(&dst->ff_layout, &src->ff_layout);
359 dst->ff_layout_version = le32_to_cpu(src->ff_layout_version);
360 dst->ff_range = le32_to_cpu(src->ff_range);
363 /* XXX: Add more if filter_fid is enlarged in the future. */
365 #endif /* HAVE_SERVER_SUPPORT */
367 static inline void lu_last_id_fid(struct lu_fid *fid, __u64 seq, __u32 ost_idx)
369 if (fid_seq_is_mdt0(seq)) {
370 fid->f_seq = fid_idif_seq(0, ost_idx);
372 LASSERTF(fid_seq_is_norm(seq) || fid_seq_is_echo(seq) ||
373 fid_seq_is_idif(seq), "%#llx\n", seq);
380 static inline bool fid_is_md_operative(const struct lu_fid *fid)
382 return fid_is_mdt0(fid) || fid_is_igif(fid) ||
383 fid_is_norm(fid) || fid_is_root(fid);
386 /* seq client type */
388 LUSTRE_SEQ_METADATA = 1,
394 LUSTRE_SEQ_CONTROLLER
397 struct lu_server_seq;
399 /* Client sequence manager interface. */
400 struct lu_client_seq {
401 /* Sequence-controller export. */
402 struct obd_export *lcs_exp;
403 struct mutex lcs_mutex;
406 * Range of allowed for allocation sequeces. When using lu_client_seq on
407 * clients, this contains meta-sequence range. And for servers this
408 * contains super-sequence range.
410 struct lu_seq_range lcs_space;
412 /* Seq related debugfs */
413 struct dentry *lcs_debugfs_entry;
415 /* This holds last allocated fid in last obtained seq */
416 struct lu_fid lcs_fid;
418 /* LUSTRE_SEQ_METADATA or LUSTRE_SEQ_DATA */
419 enum lu_cli_type lcs_type;
422 * Service uuid, passed from MDT + seq name to form unique seq name to
423 * use it with debugfs.
425 char lcs_name[LUSTRE_MDT_MAXNAMELEN];
428 * Sequence width, that is how many objects may be allocated in one
429 * sequence. Default value for it is LUSTRE_SEQ_MAX_WIDTH.
433 /* Seq-server for direct talking */
434 struct lu_server_seq *lcs_srv;
437 /* server sequence manager interface */
438 struct lu_server_seq {
439 /* Available sequences space */
440 struct lu_seq_range lss_space;
442 /* keeps highwater in lsr_end for seq allocation algorithm */
443 struct lu_seq_range lss_lowater_set;
444 struct lu_seq_range lss_hiwater_set;
446 /* Device for server side seq manager need saving seq to backingstore */
447 struct dt_device *lss_dev;
449 /* /seq file object device */
450 struct dt_object *lss_obj;
452 /* Seq related debugfs */
453 struct dentry *lss_debugfs_entry;
455 /* LUSTRE_SEQ_SERVER or LUSTRE_SEQ_CONTROLLER */
456 enum lu_mgr_type lss_type;
458 /* Client interface to request controller */
459 struct lu_client_seq *lss_cli;
461 /* Mutex for protecting allocation */
462 struct mutex lss_mutex;
465 * Service uuid, passed from MDT + seq name to form unique seq name to
466 * use it with debugfs.
468 char lss_name[LUSTRE_MDT_MAXNAMELEN];
471 * Allocation chunks for super and meta sequences. Default values are
472 * LUSTRE_SEQ_SUPER_WIDTH and LUSTRE_SEQ_META_WIDTH.
476 /* minimum lss_alloc_set size that should be allocated from lss_space */
479 /* sync is needed for update operation */
482 /* Pointer to site object, required to access site fld */
483 struct seq_server_site *lss_site;
486 struct seq_server_site {
487 struct lu_site *ss_lu;
488 /* mds number of this site */
490 /* Fid location database */
491 struct lu_server_fld *ss_server_fld;
492 struct lu_client_fld *ss_client_fld;
494 /* Server Seq Manager */
495 struct lu_server_seq *ss_server_seq;
497 /* Controller Seq Manager */
498 struct lu_server_seq *ss_control_seq;
499 struct obd_export *ss_control_exp;
501 /* Client Seq Manager */
502 struct lu_client_seq *ss_client_seq;
507 int seq_server_init(const struct lu_env *env,
508 struct lu_server_seq *seq,
509 struct dt_device *dev,
511 enum lu_mgr_type type,
512 struct seq_server_site *ss);
514 void seq_server_fini(struct lu_server_seq *seq,
515 const struct lu_env *env);
517 int seq_server_alloc_super(struct lu_server_seq *seq,
518 struct lu_seq_range *out,
519 const struct lu_env *env);
521 int seq_server_alloc_meta(struct lu_server_seq *seq,
522 struct lu_seq_range *out, const struct lu_env *env);
524 int seq_server_set_cli(const struct lu_env *env,
525 struct lu_server_seq *seq,
526 struct lu_client_seq *cli);
528 int seq_server_check_and_alloc_super(const struct lu_env *env,
529 struct lu_server_seq *seq);
531 int fid_alloc_generic(const struct lu_env *env, struct lu_device *lu,
532 struct lu_fid *fid, struct lu_object *parent,
533 const struct lu_name *name);
535 int seq_target_init(const struct lu_env *env,
536 struct dt_device *dt, char *svname,
539 void seq_target_fini(const struct lu_env *env,
540 struct dt_device *dt);
543 void seq_client_init(struct lu_client_seq *seq,
544 struct obd_export *exp,
545 enum lu_cli_type type,
547 struct lu_server_seq *srv);
549 void seq_client_fini(struct lu_client_seq *seq);
551 void seq_client_flush(struct lu_client_seq *seq);
553 int seq_client_alloc_fid(const struct lu_env *env, struct lu_client_seq *seq,
555 int seq_client_get_seq(const struct lu_env *env, struct lu_client_seq *seq,
557 int seq_site_fini(const struct lu_env *env, struct seq_server_site *ss);
558 /* Fids common stuff */
559 int fid_is_local(const struct lu_env *env, struct lu_site *site,
560 const struct lu_fid *fid);
563 int client_fid_init(struct obd_device *obd, struct obd_export *exp,
564 enum lu_cli_type type);
565 int client_fid_fini(struct obd_device *obd);
569 struct ldlm_namespace;
572 * Build (DLM) resource name from FID.
574 * NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
575 * but was moved into name[1] along with the OID to avoid consuming the
576 * renaming name[2,3] fields that need to be used for the quota identifier.
579 fid_build_reg_res_name(const struct lu_fid *fid, struct ldlm_res_id *res)
581 memset(res, 0, sizeof(*res));
582 res->name[LUSTRE_RES_ID_SEQ_OFF] = fid_seq(fid);
583 res->name[LUSTRE_RES_ID_VER_OID_OFF] = fid_ver_oid(fid);
587 * Return true if resource is for object identified by FID.
589 static inline int fid_res_name_eq(const struct lu_fid *fid,
590 const struct ldlm_res_id *res)
592 return res->name[LUSTRE_RES_ID_SEQ_OFF] == fid_seq(fid) &&
593 res->name[LUSTRE_RES_ID_VER_OID_OFF] == fid_ver_oid(fid);
597 * Extract FID from LDLM resource. Reverse of fid_build_reg_res_name().
600 fid_extract_from_res_name(struct lu_fid *fid, const struct ldlm_res_id *res)
602 fid->f_seq = res->name[LUSTRE_RES_ID_SEQ_OFF];
603 fid->f_oid = (__u32)(res->name[LUSTRE_RES_ID_VER_OID_OFF]);
604 fid->f_ver = (__u32)(res->name[LUSTRE_RES_ID_VER_OID_OFF] >> 32);
605 LASSERT(fid_res_name_eq(fid, res));
609 * Build (DLM) resource identifier from global quota FID and quota ID.
612 fid_build_quota_res_name(const struct lu_fid *glb_fid, union lquota_id *qid,
613 struct ldlm_res_id *res)
615 fid_build_reg_res_name(glb_fid, res);
616 res->name[LUSTRE_RES_ID_QUOTA_SEQ_OFF] = fid_seq(&qid->qid_fid);
617 res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF] = fid_ver_oid(&qid->qid_fid);
621 * Extract global FID and quota ID from resource name
623 static inline void fid_extract_from_quota_res(struct lu_fid *glb_fid,
624 union lquota_id *qid,
625 const struct ldlm_res_id *res)
627 fid_extract_from_res_name(glb_fid, res);
628 qid->qid_fid.f_seq = res->name[LUSTRE_RES_ID_QUOTA_SEQ_OFF];
629 qid->qid_fid.f_oid = (__u32)res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF];
631 (__u32)(res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF] >> 32);
635 fid_build_pdo_res_name(const struct lu_fid *fid, unsigned int hash,
636 struct ldlm_res_id *res)
638 fid_build_reg_res_name(fid, res);
639 res->name[LUSTRE_RES_ID_HSH_OFF] = hash;
643 * Build DLM resource name from object id & seq, which will be removed
644 * finally, when we replace ost_id with FID in data stack.
646 * Currently, resid from the old client, whose res[0] = object_id,
647 * res[1] = object_seq, is just oposite with Metatdata
648 * resid, where, res[0] = fid->f_seq, res[1] = fid->f_oid.
649 * To unifiy the resid identification, we will reverse the data
650 * resid to keep it same with Metadata resid, i.e.
652 * For resid from the old client,
653 * res[0] = objid, res[1] = 0, still keep the original order,
657 * res will be built from normal FID directly, i.e. res[0] = f_seq,
658 * res[1] = f_oid + f_ver.
660 static inline void ostid_build_res_name(const struct ost_id *oi,
661 struct ldlm_res_id *name)
663 memset(name, 0, sizeof(*name));
664 if (fid_seq_is_mdt0(ostid_seq(oi))) {
665 name->name[LUSTRE_RES_ID_SEQ_OFF] = ostid_id(oi);
666 name->name[LUSTRE_RES_ID_VER_OID_OFF] = ostid_seq(oi);
668 fid_build_reg_res_name(&oi->oi_fid, name);
673 * Return true if the resource is for the object identified by this id & group.
675 static inline bool ostid_res_name_eq(const struct ost_id *oi,
676 const struct ldlm_res_id *name)
678 /* Note: it is just a trick here to save some effort, probably the
679 * correct way would be turn them into the FID and compare
681 if (fid_seq_is_mdt0(ostid_seq(oi))) {
682 return name->name[LUSTRE_RES_ID_SEQ_OFF] == ostid_id(oi) &&
683 name->name[LUSTRE_RES_ID_VER_OID_OFF] == ostid_seq(oi);
685 return name->name[LUSTRE_RES_ID_SEQ_OFF] == ostid_seq(oi) &&
686 name->name[LUSTRE_RES_ID_VER_OID_OFF] == ostid_id(oi);
691 * Note: we need check oi_seq to decide where to set oi_id,
692 * so oi_seq should always be set ahead of oi_id.
694 static inline int ostid_set_id(struct ost_id *oi, __u64 oid)
696 if (fid_seq_is_mdt0(oi->oi.oi_seq)) {
697 if (oid > IDIF_MAX_OID)
700 } else if (fid_is_idif(&oi->oi_fid)) {
701 if (oid > IDIF_MAX_OID)
703 oi->oi_fid.f_seq = fid_idif_seq(oid,
704 fid_idif_ost_idx(&oi->oi_fid));
705 oi->oi_fid.f_oid = oid;
706 oi->oi_fid.f_ver = oid >> 48;
708 if (oid > OBIF_MAX_OID)
710 oi->oi_fid.f_oid = oid;
715 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
716 static inline int fid_to_ostid(const struct lu_fid *fid, struct ost_id *ostid)
720 if (fid_seq_is_igif(fid->f_seq))
723 if (fid_is_idif(fid)) {
724 ostid_set_seq_mdt0(ostid);
725 rc = ostid_set_id(ostid, fid_idif_id(fid_seq(fid),
726 fid_oid(fid), fid_ver(fid)));
728 ostid->oi_fid = *fid;
734 /* The same as osc_build_res_name() */
735 static inline void ost_fid_build_resid(const struct lu_fid *fid,
736 struct ldlm_res_id *resname)
738 if (fid_is_mdt0(fid) || fid_is_idif(fid)) {
741 oi.oi.oi_id = 0; /* gcc 4.7.2 complains otherwise */
742 if (fid_to_ostid(fid, &oi) != 0)
744 ostid_build_res_name(&oi, resname);
746 fid_build_reg_res_name(fid, resname);
750 static inline void ost_fid_from_resid(struct lu_fid *fid,
751 const struct ldlm_res_id *name,
754 if (fid_seq_is_mdt0(name->name[LUSTRE_RES_ID_VER_OID_OFF])) {
758 memset(&oi, 0, sizeof(oi));
759 ostid_set_seq(&oi, name->name[LUSTRE_RES_ID_VER_OID_OFF]);
760 if (ostid_set_id(&oi, name->name[LUSTRE_RES_ID_SEQ_OFF])) {
761 CERROR("Bad %llu to set " DOSTID "\n",
762 name->name[LUSTRE_RES_ID_SEQ_OFF], POSTID(&oi));
764 ostid_to_fid(fid, &oi, ost_idx);
767 fid_extract_from_res_name(fid, name);
771 static inline __u32 fid_hash(const struct lu_fid *f, int bits)
774 * All objects with same id and different versions will belong to same
777 return cfs_hash_long(fid_flatten64(f), bits);
780 u32 lu_fid_hash(const void *data, u32 len, u32 seed);
782 static inline int fid_set_id(struct lu_fid *fid, u64 oid)
784 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
785 CERROR("bad IGIF, "DFID"\n", PFID(fid));
789 if (fid_is_idif(fid)) {
790 if (oid > IDIF_MAX_OID) {
791 CERROR("Too large OID %#llx to set IDIF "DFID"\n",
792 (unsigned long long)oid, PFID(fid));
795 fid->f_seq = fid_idif_seq(oid, fid_idif_ost_idx(fid));
797 fid->f_ver = oid >> 48;
799 if (oid > OBIF_MAX_OID) {
800 CERROR("Too large OID %#llx to set REG "DFID"\n",
801 (unsigned long long)oid, PFID(fid));
809 #define LUSTRE_SEQ_SRV_NAME "seq_srv"
810 #define LUSTRE_SEQ_CTL_NAME "seq_ctl"
812 /* Range common stuff */
814 range_cpu_to_le(struct lu_seq_range *dst, const struct lu_seq_range *src)
816 dst->lsr_start = cpu_to_le64(src->lsr_start);
817 dst->lsr_end = cpu_to_le64(src->lsr_end);
818 dst->lsr_index = cpu_to_le32(src->lsr_index);
819 dst->lsr_flags = cpu_to_le32(src->lsr_flags);
823 range_le_to_cpu(struct lu_seq_range *dst, const struct lu_seq_range *src)
825 dst->lsr_start = le64_to_cpu(src->lsr_start);
826 dst->lsr_end = le64_to_cpu(src->lsr_end);
827 dst->lsr_index = le32_to_cpu(src->lsr_index);
828 dst->lsr_flags = le32_to_cpu(src->lsr_flags);
832 range_cpu_to_be(struct lu_seq_range *dst, const struct lu_seq_range *src)
834 dst->lsr_start = cpu_to_be64(src->lsr_start);
835 dst->lsr_end = cpu_to_be64(src->lsr_end);
836 dst->lsr_index = cpu_to_be32(src->lsr_index);
837 dst->lsr_flags = cpu_to_be32(src->lsr_flags);
841 range_be_to_cpu(struct lu_seq_range *dst, const struct lu_seq_range *src)
843 dst->lsr_start = be64_to_cpu(src->lsr_start);
844 dst->lsr_end = be64_to_cpu(src->lsr_end);
845 dst->lsr_index = be32_to_cpu(src->lsr_index);
846 dst->lsr_flags = be32_to_cpu(src->lsr_flags);
849 static inline void range_array_cpu_to_le(struct lu_seq_range_array *dst,
850 const struct lu_seq_range_array *src)
854 for (i = 0; i < src->lsra_count; i++)
855 range_cpu_to_le(&dst->lsra_lsr[i], &src->lsra_lsr[i]);
857 dst->lsra_count = cpu_to_le32(src->lsra_count);
860 static inline void range_array_le_to_cpu(struct lu_seq_range_array *dst,
861 const struct lu_seq_range_array *src)
865 dst->lsra_count = le32_to_cpu(src->lsra_count);
866 for (i = 0; i < dst->lsra_count; i++)
867 range_le_to_cpu(&dst->lsra_lsr[i], &src->lsra_lsr[i]);
872 #endif /* __LUSTRE_FID_H */