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 #define OFD_SOFT_SYNC_LIMIT_DEFAULT 16
58 LPROC_OFD_STATS_READ = 0,
59 LPROC_OFD_STATS_WRITE,
60 LPROC_OFD_STATS_GETATTR,
61 LPROC_OFD_STATS_SETATTR,
62 LPROC_OFD_STATS_PUNCH,
64 LPROC_OFD_STATS_DESTROY,
65 LPROC_OFD_STATS_CREATE,
66 LPROC_OFD_STATS_STATFS,
67 LPROC_OFD_STATS_GET_INFO,
68 LPROC_OFD_STATS_SET_INFO,
69 LPROC_OFD_STATS_QUOTACTL,
73 static inline void ofd_counter_incr(struct obd_export *exp, int opcode,
74 char *jobid, long amount)
76 if (exp->exp_obd && exp->exp_obd->obd_stats)
77 lprocfs_counter_add(exp->exp_obd->obd_stats, opcode, amount);
79 if (exp->exp_obd && exp->exp_obd->u.obt.obt_jobstats.ojs_hash &&
80 (exp_connect_flags(exp) & OBD_CONNECT_JOBSTATS))
81 lprocfs_job_stats_log(exp->exp_obd, jobid, opcode, amount);
83 if (exp->exp_nid_stats != NULL &&
84 exp->exp_nid_stats->nid_stats != NULL) {
85 lprocfs_counter_add(exp->exp_nid_stats->nid_stats, opcode,
91 struct list_head os_list;
93 spinlock_t os_last_oid_lock;
94 struct mutex os_create_lock;
96 atomic_t os_precreate_in_progress;
97 struct dt_object *os_lastid_obj;
98 unsigned long os_destroys_in_progress:1;
102 struct dt_device ofd_dt_dev;
103 struct dt_device *ofd_osd;
104 struct obd_export *ofd_osd_exp;
105 /* DLM name-space for meta-data locks maintained by this server */
106 struct ldlm_namespace *ofd_namespace;
109 struct lu_target ofd_lut;
110 struct dt_object *ofd_health_check_file;
111 struct local_oid_storage *ofd_los;
113 __u64 ofd_inconsistency_self_detected;
114 __u64 ofd_inconsistency_self_repaired;
116 struct list_head ofd_seq_list;
117 rwlock_t ofd_seq_list_lock;
119 int ofd_precreate_batch;
120 spinlock_t ofd_batch_lock;
122 /* preferred BRW size, decided by storage type and capability */
124 /* checksum types supported on this node */
125 enum cksum_types ofd_cksum_types_supported;
127 spinlock_t ofd_flags_lock;
128 unsigned long ofd_raid_degraded:1,
129 /* sync journal on writes */
131 /* Protected by ofd_lastid_rwsem. */
132 ofd_lastid_rebuilding:1,
133 ofd_record_fid_accessed:1,
134 ofd_lfsck_verify_pfid:1,
137 /* Whether to enforce T10PI checksum of RPC */
138 ofd_checksum_t10pi_enforce:1;
139 struct seq_server_site ofd_seq_site;
140 /* the limit of SOFT_SYNC RPCs that will trigger a soft sync */
141 unsigned int ofd_soft_sync_limit;
142 /* Protect ::ofd_lastid_rebuilding */
143 struct rw_semaphore ofd_lastid_rwsem;
144 __u64 ofd_lastid_gen;
145 struct ptlrpc_thread ofd_inconsistency_thread;
146 struct list_head ofd_inconsistency_list;
147 spinlock_t ofd_inconsistency_lock;
150 static inline struct ofd_device *ofd_dev(struct lu_device *d)
152 return container_of0(d, struct ofd_device, ofd_dt_dev.dd_lu_dev);
155 static inline struct obd_device *ofd_obd(struct ofd_device *ofd)
157 return ofd->ofd_dt_dev.dd_lu_dev.ld_obd;
160 static inline struct ofd_device *ofd_exp(struct obd_export *exp)
162 return ofd_dev(exp->exp_obd->obd_lu_dev);
165 static inline char *ofd_name(struct ofd_device *ofd)
167 return ofd->ofd_dt_dev.dd_lu_dev.ld_obd->obd_name;
171 struct lu_object_header ofo_header;
172 struct dt_object ofo_obj;
173 struct filter_fid ofo_ff;
174 unsigned int ofo_pfid_checking:1,
178 static inline struct ofd_object *ofd_obj(struct lu_object *o)
180 return container_of0(o, struct ofd_object, ofo_obj.do_lu);
183 static inline int ofd_object_exists(struct ofd_object *obj)
185 LASSERT(obj != NULL);
186 if (lu_object_is_dying(obj->ofo_obj.do_lu.lo_header))
188 return lu_object_exists(&obj->ofo_obj.do_lu);
191 static inline struct dt_object *fo2dt(struct ofd_object *obj)
193 return &obj->ofo_obj;
196 static inline struct dt_object *ofd_object_child(struct ofd_object *_obj)
198 struct lu_object *lu = &(_obj)->ofo_obj.do_lu;
200 return container_of0(lu_object_next(lu), struct dt_object, do_lu);
203 static inline struct ofd_device *ofd_obj2dev(const struct ofd_object *fo)
205 return ofd_dev(fo->ofo_obj.do_lu.lo_dev);
208 static inline void ofd_read_lock(const struct lu_env *env,
209 struct ofd_object *fo)
211 struct dt_object *next = ofd_object_child(fo);
213 next->do_ops->do_read_lock(env, next, 0);
216 static inline void ofd_read_unlock(const struct lu_env *env,
217 struct ofd_object *fo)
219 struct dt_object *next = ofd_object_child(fo);
221 next->do_ops->do_read_unlock(env, next);
224 static inline void ofd_write_lock(const struct lu_env *env,
225 struct ofd_object *fo)
227 struct dt_object *next = ofd_object_child(fo);
229 next->do_ops->do_write_lock(env, next, 0);
232 static inline void ofd_write_unlock(const struct lu_env *env,
233 struct ofd_object *fo)
235 struct dt_object *next = ofd_object_child(fo);
237 next->do_ops->do_write_unlock(env, next);
241 * Common data shared by obdofd-level handlers. This is allocated per-thread
242 * to reduce stack consumption.
244 struct ofd_thread_info {
245 const struct lu_env *fti_env;
247 struct obd_export *fti_exp;
249 __u64 fti_pre_version;
251 struct lu_fid fti_fid;
252 struct lu_attr fti_attr;
253 struct lu_attr fti_attr2;
254 struct ldlm_res_id fti_resid;
255 struct filter_fid fti_mds_fid;
256 struct ost_id fti_ostid;
257 struct ofd_object *fti_obj;
259 char name[64]; /* for ofd_init0() */
260 struct obd_statfs osfs; /* for obdofd_statfs() */
263 /* Ops object filename */
264 struct lu_name fti_name;
265 struct dt_object_format fti_dof;
266 struct lu_buf fti_buf;
269 struct ost_lvb fti_lvb;
271 struct lfsck_req_local fti_lrl;
272 struct obd_connect_data fti_ocd;
276 extern void target_recovery_fini(struct obd_device *obd);
277 extern void target_recovery_init(struct lu_target *lut, svc_handler_t handler);
280 extern struct lu_context_key ofd_thread_key;
281 int ofd_postrecov(const struct lu_env *env, struct ofd_device *ofd);
282 int ofd_fiemap_get(const struct lu_env *env, struct ofd_device *ofd,
283 struct lu_fid *fid, struct fiemap *fiemap);
286 extern struct obd_ops ofd_obd_ops;
287 int ofd_destroy_by_fid(const struct lu_env *env, struct ofd_device *ofd,
288 const struct lu_fid *fid, int orphan);
289 int ofd_statfs(const struct lu_env *env, struct obd_export *exp,
290 struct obd_statfs *osfs, time64_t max_age, __u32 flags);
291 int ofd_obd_disconnect(struct obd_export *exp);
294 u64 ofd_seq_last_oid(struct ofd_seq *oseq);
295 void ofd_seq_last_oid_set(struct ofd_seq *oseq, u64 id);
296 int ofd_seq_last_oid_write(const struct lu_env *env, struct ofd_device *ofd,
297 struct ofd_seq *oseq);
298 int ofd_seqs_init(const struct lu_env *env, struct ofd_device *ofd);
299 struct ofd_seq *ofd_seq_get(struct ofd_device *ofd, u64 seq);
300 void ofd_seq_put(const struct lu_env *env, struct ofd_seq *oseq);
302 int ofd_fs_setup(const struct lu_env *env, struct ofd_device *ofd,
303 struct obd_device *obd);
304 void ofd_fs_cleanup(const struct lu_env *env, struct ofd_device *ofd);
305 int ofd_precreate_batch(struct ofd_device *ofd, int batch);
306 struct ofd_seq *ofd_seq_load(const struct lu_env *env, struct ofd_device *ofd,
308 void ofd_seqs_fini(const struct lu_env *env, struct ofd_device *ofd);
309 void ofd_seqs_free(const struct lu_env *env, struct ofd_device *ofd);
312 int ofd_start_inconsistency_verification_thread(struct ofd_device *ofd);
313 int ofd_stop_inconsistency_verification_thread(struct ofd_device *ofd);
314 int ofd_verify_ff(const struct lu_env *env, struct ofd_object *fo,
316 int ofd_verify_layout_version(const struct lu_env *env,
317 struct ofd_object *fo, const struct obdo *oa);
318 int ofd_preprw(const struct lu_env *env,int cmd, struct obd_export *exp,
319 struct obdo *oa, int objcount, struct obd_ioobj *obj,
320 struct niobuf_remote *rnb, int *nr_local,
321 struct niobuf_local *lnb);
322 int ofd_commitrw(const struct lu_env *env, int cmd, struct obd_export *exp,
323 struct obdo *oa, int objcount, struct obd_ioobj *obj,
324 struct niobuf_remote *rnb, int npages,
325 struct niobuf_local *lnb, int old_rc);
328 struct thandle *ofd_trans_create(const struct lu_env *env,
329 struct ofd_device *ofd);
330 int ofd_trans_start(const struct lu_env *env,
331 struct ofd_device *ofd, struct ofd_object *fo,
333 int ofd_trans_stop(const struct lu_env *env, struct ofd_device *ofd,
334 struct thandle *th, int rc);
335 int ofd_txn_stop_cb(const struct lu_env *env, struct thandle *txn,
339 int ofd_tunables_init(struct ofd_device *ofd);
340 #ifdef CONFIG_PROC_FS
341 void ofd_stats_counter_init(struct lprocfs_stats *stats);
343 static inline void ofd_stats_counter_init(struct lprocfs_stats *stats) {}
347 struct ofd_object *ofd_object_find(const struct lu_env *env,
348 struct ofd_device *ofd,
349 const struct lu_fid *fid);
350 int ofd_object_ff_load(const struct lu_env *env, struct ofd_object *fo);
351 int ofd_object_ff_update(const struct lu_env *env, struct ofd_object *fo,
352 const struct obdo *oa, struct filter_fid *ff);
353 int ofd_precreate_objects(const struct lu_env *env, struct ofd_device *ofd,
354 u64 id, struct ofd_seq *oseq, int nr, int sync);
356 static inline void ofd_object_put(const struct lu_env *env,
357 struct ofd_object *fo)
359 dt_object_put(env, &fo->ofo_obj);
361 int ofd_attr_set(const struct lu_env *env, struct ofd_object *fo,
362 struct lu_attr *la, struct obdo *oa);
363 int ofd_object_punch(const struct lu_env *env, struct ofd_object *fo,
364 __u64 start, __u64 end, struct lu_attr *la,
366 int ofd_destroy(const struct lu_env *, struct ofd_object *, int);
367 int ofd_attr_get(const struct lu_env *env, struct ofd_object *fo,
369 int ofd_attr_handle_id(const struct lu_env *env, struct ofd_object *fo,
370 struct lu_attr *la, int is_setattr);
373 struct ofd_object *ofd_object_find_exists(const struct lu_env *env,
374 struct ofd_device *ofd,
375 const struct lu_fid *fid)
377 struct ofd_object *fo;
379 fo = ofd_object_find(env, ofd, fid);
380 if (!IS_ERR(fo) && !ofd_object_exists(fo)) {
381 ofd_object_put(env, fo);
382 fo = ERR_PTR(-ENOENT);
388 int ofd_fid_set_index(const struct lu_env *env, struct ofd_device *ofd,
390 int ofd_fid_init(const struct lu_env *env, struct ofd_device *ofd);
391 int ofd_fid_fini(const struct lu_env *env, struct ofd_device *ofd);
394 extern struct ldlm_valblock_ops ofd_lvbo;
397 extern struct kmem_cache *ldlm_glimpse_work_kmem;
399 int ofd_intent_policy(const struct lu_env *env, struct ldlm_namespace *ns,
400 struct ldlm_lock **lockp, void *req_cookie,
401 enum ldlm_mode mode, __u64 flags, void *data);
403 static inline struct ofd_thread_info *ofd_info(const struct lu_env *env)
405 return lu_env_info(env, &ofd_thread_key);
408 static inline struct ofd_thread_info *ofd_info_init(const struct lu_env *env,
409 struct obd_export *exp)
411 struct ofd_thread_info *info;
413 info = ofd_info(env);
414 LASSERT(info->fti_exp == NULL);
415 LASSERT(info->fti_env == NULL);
416 LASSERT(info->fti_attr.la_valid == 0);
420 info->fti_pre_version = 0;
424 static inline struct ofd_thread_info *tsi2ofd_info(struct tgt_session_info *tsi)
426 struct ptlrpc_request *req = tgt_ses_req(tsi);
427 struct ofd_thread_info *info;
429 info = ofd_info(tsi->tsi_env);
430 LASSERT(info->fti_exp == NULL);
431 LASSERT(info->fti_env == NULL);
432 LASSERT(info->fti_attr.la_valid == 0);
434 info->fti_env = tsi->tsi_env;
435 info->fti_exp = tsi->tsi_exp;
437 info->fti_xid = req->rq_xid;
438 /** VBR: take versions from request */
439 if (req->rq_reqmsg != NULL &&
440 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
441 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
443 info->fti_pre_version = pre_version ? pre_version[0] : 0;
448 /* sync on lock cancel is useless when we force a journal flush,
449 * and if we enable async journal commit, we should also turn on
450 * sync on lock cancel if it is not enabled already. */
451 static inline void ofd_slc_set(struct ofd_device *ofd)
453 if (ofd->ofd_sync_journal == 1)
454 ofd->ofd_lut.lut_sync_lock_cancel = SYNC_LOCK_CANCEL_NEVER;
455 else if (ofd->ofd_lut.lut_sync_lock_cancel == SYNC_LOCK_CANCEL_NEVER)
456 ofd->ofd_lut.lut_sync_lock_cancel = SYNC_LOCK_CANCEL_ALWAYS;
459 static inline int ofd_validate_seq(struct obd_export *exp, __u64 seq)
461 struct filter_export_data *fed = &exp->exp_filter_data;
463 if (unlikely(seq == FID_SEQ_OST_MDT0 && fed->fed_group != 0)) {
464 /* IDIF request only operates on MDT0 group */
465 CERROR("%s: Invalid sequence %#llx for group %u\n",
466 exp->exp_obd->obd_name, seq, fed->fed_group);
472 #endif /* _OFD_INTERNAL_H */