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) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #ifndef _OFD_INTERNAL_H
34 #define _OFD_INTERNAL_H
37 #include <obd_class.h>
38 #include <dt_object.h>
39 #include <md_object.h>
40 #include <lustre_fid.h>
42 #define OFD_INIT_OBJID 0
43 #define OFD_PRECREATE_BATCH_DEFAULT (OBJ_SUBDIR_COUNT * 4)
45 /* on small filesystems we should not precreate too many objects in
46 * a single transaction, otherwise we can overflow transactions */
47 #define OFD_PRECREATE_SMALL_FS (1024ULL * 1024 * 1024)
48 #define OFD_PRECREATE_BATCH_SMALL 8
50 /* Limit the returned fields marked valid to those that we actually might set */
51 #define OFD_VALID_FLAGS (LA_TYPE | LA_MODE | LA_SIZE | LA_BLOCKS | \
52 LA_BLKSIZE | LA_ATIME | LA_MTIME | LA_CTIME)
54 /* FMD tracking data */
56 struct list_head fmd_list; /* linked to tgt_fmd_list */
57 struct lu_fid fmd_fid; /* FID being written to */
58 __u64 fmd_mactime_xid; /* xid highest {m,a,c}time setattr */
59 time64_t fmd_expire; /* time when the fmd should expire */
60 int fmd_refcount; /* reference counter - list holds 1 */
63 #define OFD_SOFT_SYNC_LIMIT_DEFAULT 16
67 LPROC_OFD_STATS_READ = 0,
68 LPROC_OFD_STATS_WRITE,
69 LPROC_OFD_STATS_GETATTR,
70 LPROC_OFD_STATS_SETATTR,
71 LPROC_OFD_STATS_PUNCH,
73 LPROC_OFD_STATS_DESTROY,
74 LPROC_OFD_STATS_CREATE,
75 LPROC_OFD_STATS_STATFS,
76 LPROC_OFD_STATS_GET_INFO,
77 LPROC_OFD_STATS_SET_INFO,
78 LPROC_OFD_STATS_QUOTACTL,
82 static inline void ofd_counter_incr(struct obd_export *exp, int opcode,
83 char *jobid, long amount)
85 if (exp->exp_obd && exp->exp_obd->obd_stats)
86 lprocfs_counter_add(exp->exp_obd->obd_stats, opcode, amount);
88 if (exp->exp_obd && exp->exp_obd->u.obt.obt_jobstats.ojs_hash &&
89 (exp_connect_flags(exp) & OBD_CONNECT_JOBSTATS))
90 lprocfs_job_stats_log(exp->exp_obd, jobid, opcode, amount);
92 if (exp->exp_nid_stats != NULL &&
93 exp->exp_nid_stats->nid_stats != NULL) {
94 lprocfs_counter_add(exp->exp_nid_stats->nid_stats, opcode,
100 struct list_head os_list;
102 spinlock_t os_last_oid_lock;
103 struct mutex os_create_lock;
105 struct dt_object *os_lastid_obj;
106 unsigned long os_destroys_in_progress:1;
110 struct dt_device ofd_dt_dev;
111 struct dt_device *ofd_osd;
112 struct obd_export *ofd_osd_exp;
113 /* DLM name-space for meta-data locks maintained by this server */
114 struct ldlm_namespace *ofd_namespace;
117 struct lu_target ofd_lut;
118 struct dt_object *ofd_health_check_file;
119 struct local_oid_storage *ofd_los;
121 __u64 ofd_inconsistency_self_detected;
122 __u64 ofd_inconsistency_self_repaired;
124 struct list_head ofd_seq_list;
125 rwlock_t ofd_seq_list_lock;
127 int ofd_precreate_batch;
128 spinlock_t ofd_batch_lock;
130 /* preferred BRW size, decided by storage type and capability */
132 /* checksum types supported on this node */
133 enum cksum_types ofd_cksum_types_supported;
135 spinlock_t ofd_flags_lock;
136 unsigned long ofd_raid_degraded:1,
137 /* sync journal on writes */
139 /* Protected by ofd_lastid_rwsem. */
140 ofd_lastid_rebuilding:1,
141 ofd_record_fid_accessed:1,
142 ofd_lfsck_verify_pfid:1,
144 /* Whether to enforce T10PI checksum of RPC */
145 ofd_checksum_t10pi_enforce:1;
146 struct seq_server_site ofd_seq_site;
147 /* the limit of SOFT_SYNC RPCs that will trigger a soft sync */
148 unsigned int ofd_soft_sync_limit;
149 /* Protect ::ofd_lastid_rebuilding */
150 struct rw_semaphore ofd_lastid_rwsem;
151 __u64 ofd_lastid_gen;
152 struct ptlrpc_thread ofd_inconsistency_thread;
153 struct list_head ofd_inconsistency_list;
154 spinlock_t ofd_inconsistency_lock;
157 static inline struct ofd_device *ofd_dev(struct lu_device *d)
159 return container_of0(d, struct ofd_device, ofd_dt_dev.dd_lu_dev);
162 static inline struct obd_device *ofd_obd(struct ofd_device *ofd)
164 return ofd->ofd_dt_dev.dd_lu_dev.ld_obd;
167 static inline struct ofd_device *ofd_exp(struct obd_export *exp)
169 return ofd_dev(exp->exp_obd->obd_lu_dev);
172 static inline char *ofd_name(struct ofd_device *ofd)
174 return ofd->ofd_dt_dev.dd_lu_dev.ld_obd->obd_name;
178 struct lu_object_header ofo_header;
179 struct dt_object ofo_obj;
180 struct filter_fid ofo_ff;
181 unsigned int ofo_pfid_checking:1,
185 static inline struct ofd_object *ofd_obj(struct lu_object *o)
187 return container_of0(o, struct ofd_object, ofo_obj.do_lu);
190 static inline int ofd_object_exists(struct ofd_object *obj)
192 LASSERT(obj != NULL);
193 if (lu_object_is_dying(obj->ofo_obj.do_lu.lo_header))
195 return lu_object_exists(&obj->ofo_obj.do_lu);
198 static inline struct dt_object *fo2dt(struct ofd_object *obj)
200 return &obj->ofo_obj;
203 static inline struct dt_object *ofd_object_child(struct ofd_object *_obj)
205 struct lu_object *lu = &(_obj)->ofo_obj.do_lu;
207 return container_of0(lu_object_next(lu), struct dt_object, do_lu);
210 static inline struct ofd_device *ofd_obj2dev(const struct ofd_object *fo)
212 return ofd_dev(fo->ofo_obj.do_lu.lo_dev);
215 static inline void ofd_read_lock(const struct lu_env *env,
216 struct ofd_object *fo)
218 struct dt_object *next = ofd_object_child(fo);
220 next->do_ops->do_read_lock(env, next, 0);
223 static inline void ofd_read_unlock(const struct lu_env *env,
224 struct ofd_object *fo)
226 struct dt_object *next = ofd_object_child(fo);
228 next->do_ops->do_read_unlock(env, next);
231 static inline void ofd_write_lock(const struct lu_env *env,
232 struct ofd_object *fo)
234 struct dt_object *next = ofd_object_child(fo);
236 next->do_ops->do_write_lock(env, next, 0);
239 static inline void ofd_write_unlock(const struct lu_env *env,
240 struct ofd_object *fo)
242 struct dt_object *next = ofd_object_child(fo);
244 next->do_ops->do_write_unlock(env, next);
248 * Common data shared by obdofd-level handlers. This is allocated per-thread
249 * to reduce stack consumption.
251 struct ofd_thread_info {
252 const struct lu_env *fti_env;
254 struct obd_export *fti_exp;
256 __u64 fti_pre_version;
258 struct lu_fid fti_fid;
259 struct lu_attr fti_attr;
260 struct lu_attr fti_attr2;
261 struct ldlm_res_id fti_resid;
262 struct filter_fid fti_mds_fid;
263 struct ost_id fti_ostid;
264 struct ofd_object *fti_obj;
266 char name[64]; /* for ofd_init0() */
267 struct obd_statfs osfs; /* for obdofd_statfs() */
270 /* Ops object filename */
271 struct lu_name fti_name;
272 struct dt_object_format fti_dof;
273 struct lu_buf fti_buf;
276 struct ost_lvb fti_lvb;
278 struct lfsck_req_local fti_lrl;
279 struct obd_connect_data fti_ocd;
283 extern void target_recovery_fini(struct obd_device *obd);
284 extern void target_recovery_init(struct lu_target *lut, svc_handler_t handler);
287 extern struct lu_context_key ofd_thread_key;
288 int ofd_postrecov(const struct lu_env *env, struct ofd_device *ofd);
289 int ofd_fiemap_get(const struct lu_env *env, struct ofd_device *ofd,
290 struct lu_fid *fid, struct fiemap *fiemap);
293 extern struct obd_ops ofd_obd_ops;
294 int ofd_destroy_by_fid(const struct lu_env *env, struct ofd_device *ofd,
295 const struct lu_fid *fid, int orphan);
296 int ofd_statfs(const struct lu_env *env, struct obd_export *exp,
297 struct obd_statfs *osfs, time64_t max_age, __u32 flags);
298 int ofd_obd_disconnect(struct obd_export *exp);
301 u64 ofd_seq_last_oid(struct ofd_seq *oseq);
302 void ofd_seq_last_oid_set(struct ofd_seq *oseq, u64 id);
303 int ofd_seq_last_oid_write(const struct lu_env *env, struct ofd_device *ofd,
304 struct ofd_seq *oseq);
305 int ofd_seqs_init(const struct lu_env *env, struct ofd_device *ofd);
306 struct ofd_seq *ofd_seq_get(struct ofd_device *ofd, u64 seq);
307 void ofd_seq_put(const struct lu_env *env, struct ofd_seq *oseq);
309 int ofd_fs_setup(const struct lu_env *env, struct ofd_device *ofd,
310 struct obd_device *obd);
311 void ofd_fs_cleanup(const struct lu_env *env, struct ofd_device *ofd);
312 int ofd_precreate_batch(struct ofd_device *ofd, int batch);
313 struct ofd_seq *ofd_seq_load(const struct lu_env *env, struct ofd_device *ofd,
315 void ofd_seqs_fini(const struct lu_env *env, struct ofd_device *ofd);
316 void ofd_seqs_free(const struct lu_env *env, struct ofd_device *ofd);
319 int ofd_start_inconsistency_verification_thread(struct ofd_device *ofd);
320 int ofd_stop_inconsistency_verification_thread(struct ofd_device *ofd);
321 int ofd_verify_ff(const struct lu_env *env, struct ofd_object *fo,
323 int ofd_verify_layout_version(const struct lu_env *env,
324 struct ofd_object *fo, const struct obdo *oa);
325 int ofd_preprw(const struct lu_env *env,int cmd, struct obd_export *exp,
326 struct obdo *oa, int objcount, struct obd_ioobj *obj,
327 struct niobuf_remote *rnb, int *nr_local,
328 struct niobuf_local *lnb);
329 int ofd_commitrw(const struct lu_env *env, int cmd, struct obd_export *exp,
330 struct obdo *oa, int objcount, struct obd_ioobj *obj,
331 struct niobuf_remote *rnb, int npages,
332 struct niobuf_local *lnb, int old_rc);
335 struct thandle *ofd_trans_create(const struct lu_env *env,
336 struct ofd_device *ofd);
337 int ofd_trans_start(const struct lu_env *env,
338 struct ofd_device *ofd, struct ofd_object *fo,
340 int ofd_trans_stop(const struct lu_env *env, struct ofd_device *ofd,
341 struct thandle *th, int rc);
342 int ofd_txn_stop_cb(const struct lu_env *env, struct thandle *txn,
346 int ofd_tunables_init(struct ofd_device *ofd);
347 #ifdef CONFIG_PROC_FS
348 void ofd_stats_counter_init(struct lprocfs_stats *stats);
350 static inline void ofd_stats_counter_init(struct lprocfs_stats *stats) {}
354 struct ofd_object *ofd_object_find(const struct lu_env *env,
355 struct ofd_device *ofd,
356 const struct lu_fid *fid);
357 int ofd_object_ff_load(const struct lu_env *env, struct ofd_object *fo);
358 int ofd_object_ff_update(const struct lu_env *env, struct ofd_object *fo,
359 const struct obdo *oa, struct filter_fid *ff);
360 int ofd_precreate_objects(const struct lu_env *env, struct ofd_device *ofd,
361 u64 id, struct ofd_seq *oseq, int nr, int sync);
363 static inline void ofd_object_put(const struct lu_env *env,
364 struct ofd_object *fo)
366 dt_object_put(env, &fo->ofo_obj);
368 int ofd_attr_set(const struct lu_env *env, struct ofd_object *fo,
369 struct lu_attr *la, struct obdo *oa);
370 int ofd_object_punch(const struct lu_env *env, struct ofd_object *fo,
371 __u64 start, __u64 end, struct lu_attr *la,
373 int ofd_destroy(const struct lu_env *, struct ofd_object *, int);
374 int ofd_attr_get(const struct lu_env *env, struct ofd_object *fo,
376 int ofd_attr_handle_id(const struct lu_env *env, struct ofd_object *fo,
377 struct lu_attr *la, int is_setattr);
380 struct ofd_object *ofd_object_find_exists(const struct lu_env *env,
381 struct ofd_device *ofd,
382 const struct lu_fid *fid)
384 struct ofd_object *fo;
386 fo = ofd_object_find(env, ofd, fid);
387 if (!IS_ERR(fo) && !ofd_object_exists(fo)) {
388 ofd_object_put(env, fo);
389 fo = ERR_PTR(-ENOENT);
395 extern struct kmem_cache *tgt_fmd_kmem;
396 void ofd_fmd_expire(struct obd_export *exp);
397 void ofd_fmd_cleanup(struct obd_export *exp);
399 void ofd_fmd_drop(struct obd_export *exp, const struct lu_fid *fid);
401 #define ofd_fmd_drop(exp, fid) do {} while (0)
405 int ofd_fid_set_index(const struct lu_env *env, struct ofd_device *ofd,
407 int ofd_fid_init(const struct lu_env *env, struct ofd_device *ofd);
408 int ofd_fid_fini(const struct lu_env *env, struct ofd_device *ofd);
411 extern struct ldlm_valblock_ops ofd_lvbo;
414 extern struct kmem_cache *ldlm_glimpse_work_kmem;
416 int ofd_intent_policy(const struct lu_env *env, struct ldlm_namespace *ns,
417 struct ldlm_lock **lockp, void *req_cookie,
418 enum ldlm_mode mode, __u64 flags, void *data);
420 static inline struct ofd_thread_info *ofd_info(const struct lu_env *env)
422 struct ofd_thread_info *info;
424 lu_env_refill((void *)env);
425 info = lu_context_key_get(&env->le_ctx, &ofd_thread_key);
430 static inline struct ofd_thread_info *ofd_info_init(const struct lu_env *env,
431 struct obd_export *exp)
433 struct ofd_thread_info *info;
435 info = ofd_info(env);
436 LASSERT(info->fti_exp == NULL);
437 LASSERT(info->fti_env == NULL);
438 LASSERT(info->fti_attr.la_valid == 0);
442 info->fti_pre_version = 0;
446 static inline struct ofd_thread_info *tsi2ofd_info(struct tgt_session_info *tsi)
448 struct ptlrpc_request *req = tgt_ses_req(tsi);
449 struct ofd_thread_info *info;
451 info = ofd_info(tsi->tsi_env);
452 LASSERT(info->fti_exp == NULL);
453 LASSERT(info->fti_env == NULL);
454 LASSERT(info->fti_attr.la_valid == 0);
456 info->fti_env = tsi->tsi_env;
457 info->fti_exp = tsi->tsi_exp;
459 info->fti_xid = req->rq_xid;
460 /** VBR: take versions from request */
461 if (req->rq_reqmsg != NULL &&
462 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
463 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
465 info->fti_pre_version = pre_version ? pre_version[0] : 0;
470 /* sync on lock cancel is useless when we force a journal flush,
471 * and if we enable async journal commit, we should also turn on
472 * sync on lock cancel if it is not enabled already. */
473 static inline void ofd_slc_set(struct ofd_device *ofd)
475 if (ofd->ofd_sync_journal == 1)
476 ofd->ofd_lut.lut_sync_lock_cancel = NEVER_SYNC_ON_CANCEL;
477 else if (ofd->ofd_lut.lut_sync_lock_cancel == NEVER_SYNC_ON_CANCEL)
478 ofd->ofd_lut.lut_sync_lock_cancel = ALWAYS_SYNC_ON_CANCEL;
481 static inline int ofd_validate_seq(struct obd_export *exp, __u64 seq)
483 struct filter_export_data *fed = &exp->exp_filter_data;
485 if (unlikely(seq == FID_SEQ_OST_MDT0 && fed->fed_group != 0)) {
486 /* IDIF request only operates on MDT0 group */
487 CERROR("%s: Invalid sequence %#llx for group %u\n",
488 exp->exp_obd->obd_name, seq, fed->fed_group);
494 #endif /* _OFD_INTERNAL_H */