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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2015, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef _LUSTRE_LU_TARGET_H
38 #define _LUSTRE_LU_TARGET_H
40 #include <dt_object.h>
41 #include <lustre_export.h>
42 #include <lustre_update.h>
43 #include <lustre_disk.h>
44 #include <lustre_lfsck.h>
46 /* Each one represents a distribute transaction replay
47 * operation, and updates on each MDTs are linked to
49 struct distribute_txn_replay_req {
50 /* update record, may be vmalloc'd */
51 struct llog_update_record *dtrq_lur;
54 /* linked to the distribute transaction replay
55 * list (tdtd_replay_list) */
56 struct list_head dtrq_list;
57 __u64 dtrq_master_transno;
61 /* all of sub updates are linked here */
62 struct list_head dtrq_sub_list;
63 spinlock_t dtrq_sub_list_lock;
65 /* If the local update has been executed during replay */
66 __u32 dtrq_local_update_executed:1;
69 /* Each one represents a sub replay item under a distribute
70 * transaction. A distribute transaction will be operated in
71 * two or more MDTs, and updates on each MDT will be represented
72 * by this structure */
73 struct distribute_txn_replay_req_sub {
74 __u32 dtrqs_mdt_index;
76 /* All of cookies for the update will be linked here */
77 spinlock_t dtrqs_cookie_list_lock;
78 struct list_head dtrqs_cookie_list;
79 struct list_head dtrqs_list;
82 struct target_distribute_txn_data;
83 typedef int (*distribute_txn_replay_handler_t)(struct lu_env *env,
84 struct target_distribute_txn_data *tdtd,
85 struct distribute_txn_replay_req *dtrq);
86 typedef char *(*target_show_update_logs_retrievers_t)(void *data, int *size,
88 struct target_distribute_txn_data {
89 /* Distribution ID is used to identify updates log on different
90 * MDTs for one operation */
91 spinlock_t tdtd_batchid_lock;
93 struct lu_target *tdtd_lut;
94 struct dt_object *tdtd_batchid_obj;
95 struct dt_device *tdtd_dt;
97 /* Committed batchid for distribute transaction */
98 __u64 tdtd_committed_batchid;
100 /* List for distribute transaction */
101 struct list_head tdtd_list;
103 /* Threads to manage distribute transaction */
104 wait_queue_head_t tdtd_commit_thread_waitq;
105 atomic_t tdtd_refcount;
107 /* recovery update */
108 distribute_txn_replay_handler_t tdtd_replay_handler;
109 struct list_head tdtd_replay_list;
110 struct list_head tdtd_replay_finish_list;
111 spinlock_t tdtd_replay_list_lock;
112 /* last replay update transno */
113 __u32 tdtd_replay_ready:1;
115 /* Manage the llog recovery threads */
116 atomic_t tdtd_recovery_threads_count;
117 wait_queue_head_t tdtd_recovery_threads_waitq;
118 target_show_update_logs_retrievers_t
119 tdtd_show_update_logs_retrievers;
120 void *tdtd_show_retrievers_cbdata;
124 struct obd_device *lut_obd;
125 struct dt_device *lut_bottom;
127 struct target_distribute_txn_data *lut_tdtd;
128 struct ptlrpc_thread lut_tdtd_commit_thread;
130 /* supported opcodes and handlers for this target */
131 struct tgt_opc_slice *lut_slice;
132 __u32 lut_reply_fail_id;
133 __u32 lut_request_fail_id;
136 rwlock_t lut_sptlrpc_lock;
137 struct sptlrpc_rule_set lut_sptlrpc_rset;
138 spinlock_t lut_flags_lock;
139 unsigned int lut_syncjournal:1,
140 lut_sync_lock_cancel:2,
142 lut_no_reconstruct:1;
143 /** last_rcvd file */
144 struct dt_object *lut_last_rcvd;
145 /* transaction callbacks */
146 struct dt_txn_callback lut_txn_cb;
147 /** server data in last_rcvd file */
148 struct lr_server_data lut_lsd;
149 /** Server last transaction number */
150 __u64 lut_last_transno;
151 /** Lock protecting last transaction number */
152 spinlock_t lut_translock;
153 /** Lock protecting client bitmap */
154 spinlock_t lut_client_bitmap_lock;
155 /** Bitmap of known clients */
156 unsigned long *lut_client_bitmap;
157 /* Number of clients supporting multiple modify RPCs
158 * recorded in the bitmap */
159 atomic_t lut_num_clients;
160 /* Client generation to identify client slot reuse */
161 atomic_t lut_client_generation;
162 /** reply_data file */
163 struct dt_object *lut_reply_data;
164 /** Bitmap of used slots in the reply data file */
165 unsigned long **lut_reply_bitmap;
166 /** target sync count, used for debug & test */
167 atomic_t lut_sync_count;
169 /** cross MDT locks which should trigger Sync-on-Lock-Cancel */
170 spinlock_t lut_slc_locks_guard;
171 struct list_head lut_slc_locks;
174 /* number of slots in reply bitmap */
175 #define LUT_REPLY_SLOTS_PER_CHUNK (1<<20)
176 #define LUT_REPLY_SLOTS_MAX_CHUNKS 16
181 struct tg_reply_data {
182 /** chain of reply data anchored in tg_export_data */
183 struct list_head trd_list;
184 /** copy of on-disk reply data */
185 struct lsd_reply_data trd_reply;
186 /** versions for Version Based Recovery */
187 __u64 trd_pre_versions[4];
188 /** slot index in reply_data file */
190 /** tag the client used */
194 extern struct lu_context_key tgt_session_key;
196 struct tgt_session_info {
198 * The following members will be filled explicitly
199 * with specific data in tgt_ses_init().
201 struct req_capsule *tsi_pill;
204 * Lock request for "habeo clavis" operations.
206 struct ldlm_request *tsi_dlm_req;
208 /* although we have export in req, there are cases when it is not
209 * available, e.g. closing files upon export destroy */
210 struct obd_export *tsi_exp;
211 const struct lu_env *tsi_env;
212 struct lu_target *tsi_tgt;
214 const struct mdt_body *tsi_mdt_body;
215 struct ost_body *tsi_ost_body;
216 struct lu_object *tsi_corpus;
218 struct lu_fid tsi_fid;
219 struct ldlm_res_id tsi_resid;
221 /* object affected by VBR, for last_rcvd_update */
222 struct dt_object *tsi_vbr_obj;
223 /* opdata for mdt_reint_open(), has the same value as
224 * ldlm_reply:lock_policy_res1. The tgt_update_last_rcvd() stores
225 * this value onto disk for recovery when tgt_txn_stop_cb() is called.
230 * Additional fail id that can be set by handler.
232 int tsi_reply_fail_id;
233 bool tsi_preprocessed;
240 __u32 tsi_client_gen;
243 static inline struct tgt_session_info *tgt_ses_info(const struct lu_env *env)
245 struct tgt_session_info *tsi;
247 LASSERT(env->le_ses != NULL);
248 tsi = lu_context_key_get(env->le_ses, &tgt_session_key);
253 static inline void tgt_vbr_obj_set(const struct lu_env *env,
254 struct dt_object *obj)
256 struct tgt_session_info *tsi;
258 if (env->le_ses != NULL) {
259 tsi = tgt_ses_info(env);
260 tsi->tsi_vbr_obj = obj;
264 static inline void tgt_opdata_set(const struct lu_env *env, __u64 flags)
266 struct tgt_session_info *tsi;
268 if (env->le_ses != NULL) {
269 tsi = tgt_ses_info(env);
270 tsi->tsi_opdata |= flags;
274 static inline void tgt_opdata_clear(const struct lu_env *env, __u64 flags)
276 struct tgt_session_info *tsi;
278 if (env->le_ses != NULL) {
279 tsi = tgt_ses_info(env);
280 tsi->tsi_opdata &= ~flags;
285 * Generic unified target support.
287 enum tgt_handler_flags {
289 * struct *_body is passed in the incoming message, and object
290 * identified by this fid exists on disk.
292 * "habeo corpus" == "I have a body"
294 HABEO_CORPUS = (1 << 0),
296 * struct ldlm_request is passed in the incoming message.
298 * "habeo clavis" == "I have a key"
300 HABEO_CLAVIS = (1 << 1),
302 * this request has fixed reply format, so that reply message can be
303 * packed by generic code.
305 * "habeo refero" == "I have a reply"
307 HABEO_REFERO = (1 << 2),
309 * this request will modify something, so check whether the file system
310 * is readonly or not, then return -EROFS to client asap if necessary.
312 * "mutabor" == "I shall modify"
318 /* The name of this handler. */
320 /* Fail id, check at the beginning */
324 /* Flags in enum tgt_handler_flags */
326 /* Request version for this opcode */
328 /* Handler function */
329 int (*th_act)(struct tgt_session_info *tsi);
330 /* Handler function for high priority requests */
331 void (*th_hp)(struct tgt_session_info *tsi);
332 /* Request format for this request */
333 const struct req_format *th_fmt;
336 struct tgt_opc_slice {
337 __u32 tos_opc_start; /* First op code */
338 __u32 tos_opc_end; /* Last op code */
339 struct tgt_handler *tos_hs; /* Registered handler */
342 static inline struct ptlrpc_request *tgt_ses_req(struct tgt_session_info *tsi)
344 return tsi->tsi_pill ? tsi->tsi_pill->rc_req : NULL;
347 static inline __u64 tgt_conn_flags(struct tgt_session_info *tsi)
349 LASSERT(tsi->tsi_exp);
350 return exp_connect_flags(tsi->tsi_exp);
353 static inline int req_is_replay(struct ptlrpc_request *req)
355 LASSERT(req->rq_reqmsg);
356 return !!(lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY);
359 static inline bool tgt_is_multimodrpcs_client(struct obd_export *exp)
361 return exp_connect_flags(exp) & OBD_CONNECT_MULTIMODRPCS;
365 /* target/tgt_handler.c */
366 int tgt_request_handle(struct ptlrpc_request *req);
367 char *tgt_name(struct lu_target *tgt);
368 void tgt_counter_incr(struct obd_export *exp, int opcode);
369 int tgt_connect_check_sptlrpc(struct ptlrpc_request *req,
370 struct obd_export *exp);
371 int tgt_adapt_sptlrpc_conf(struct lu_target *tgt, int initial);
372 int tgt_connect(struct tgt_session_info *tsi);
373 int tgt_disconnect(struct tgt_session_info *uti);
374 int tgt_obd_ping(struct tgt_session_info *tsi);
375 int tgt_enqueue(struct tgt_session_info *tsi);
376 int tgt_convert(struct tgt_session_info *tsi);
377 int tgt_bl_callback(struct tgt_session_info *tsi);
378 int tgt_cp_callback(struct tgt_session_info *tsi);
379 int tgt_llog_open(struct tgt_session_info *tsi);
380 int tgt_llog_close(struct tgt_session_info *tsi);
381 int tgt_llog_destroy(struct tgt_session_info *tsi);
382 int tgt_llog_read_header(struct tgt_session_info *tsi);
383 int tgt_llog_next_block(struct tgt_session_info *tsi);
384 int tgt_llog_prev_block(struct tgt_session_info *tsi);
385 int tgt_sec_ctx_init(struct tgt_session_info *tsi);
386 int tgt_sec_ctx_init_cont(struct tgt_session_info *tsi);
387 int tgt_sec_ctx_fini(struct tgt_session_info *tsi);
388 int tgt_sendpage(struct tgt_session_info *tsi, struct lu_rdpg *rdpg, int nob);
389 int tgt_send_buffer(struct tgt_session_info *tsi, struct lu_rdbuf *rdbuf);
390 int tgt_validate_obdo(struct tgt_session_info *tsi, struct obdo *oa);
391 int tgt_sync(const struct lu_env *env, struct lu_target *tgt,
392 struct dt_object *obj, __u64 start, __u64 end);
394 int tgt_io_thread_init(struct ptlrpc_thread *thread);
395 void tgt_io_thread_done(struct ptlrpc_thread *thread);
397 int tgt_extent_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
398 __u64 start, __u64 end, struct lustre_handle *lh,
399 int mode, __u64 *flags);
400 void tgt_extent_unlock(struct lustre_handle *lh, enum ldlm_mode mode);
401 int tgt_brw_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
402 struct obd_ioobj *obj, struct niobuf_remote *nb,
403 struct lustre_handle *lh, enum ldlm_mode mode);
404 void tgt_brw_unlock(struct obd_ioobj *obj, struct niobuf_remote *niob,
405 struct lustre_handle *lh, enum ldlm_mode mode);
406 int tgt_brw_read(struct tgt_session_info *tsi);
407 int tgt_brw_write(struct tgt_session_info *tsi);
408 int tgt_hpreq_handler(struct ptlrpc_request *req);
409 void tgt_register_lfsck_in_notify(int (*notify)(const struct lu_env *,
411 struct lfsck_request *,
413 void tgt_register_lfsck_query(int (*query)(const struct lu_env *,
415 struct lfsck_request *,
416 struct lfsck_reply *,
417 struct lfsck_query *));
418 bool req_can_reconstruct(struct ptlrpc_request *req, struct tg_reply_data *trd);
420 extern struct tgt_handler tgt_sec_ctx_handlers[];
421 extern struct tgt_handler tgt_lfsck_handlers[];
422 extern struct tgt_handler tgt_obd_handlers[];
423 extern struct tgt_handler tgt_dlm_handlers[];
424 extern struct tgt_handler tgt_llog_handlers[];
425 extern struct tgt_handler tgt_out_handlers[];
426 extern struct tgt_handler fld_handlers[];
427 extern struct tgt_handler seq_handlers[];
429 typedef void (*tgt_cb_t)(struct lu_target *lut, __u64 transno,
430 void *data, int err);
431 struct tgt_commit_cb {
432 tgt_cb_t tgt_cb_func;
436 int tgt_hpreq_handler(struct ptlrpc_request *req);
438 /* target/tgt_main.c */
439 void tgt_boot_epoch_update(struct lu_target *lut);
440 void tgt_save_slc_lock(struct lu_target *lut, struct ldlm_lock *lock,
442 void tgt_discard_slc_lock(struct lu_target *lut, struct ldlm_lock *lock);
443 int tgt_last_commit_cb_add(struct thandle *th, struct lu_target *lut,
444 struct obd_export *exp, __u64 transno);
445 int tgt_new_client_cb_add(struct thandle *th, struct obd_export *exp);
446 int tgt_init(const struct lu_env *env, struct lu_target *lut,
447 struct obd_device *obd, struct dt_device *dt,
448 struct tgt_opc_slice *slice,
449 int request_fail_id, int reply_fail_id);
450 void tgt_fini(const struct lu_env *env, struct lu_target *lut);
451 int tgt_client_alloc(struct obd_export *exp);
452 void tgt_client_free(struct obd_export *exp);
453 int tgt_client_del(const struct lu_env *env, struct obd_export *exp);
454 int tgt_client_add(const struct lu_env *env, struct obd_export *exp, int);
455 int tgt_client_new(const struct lu_env *env, struct obd_export *exp);
456 int tgt_client_data_read(const struct lu_env *env, struct lu_target *tg,
457 struct lsd_client_data *lcd, loff_t *off, int index);
458 int tgt_client_data_write(const struct lu_env *env, struct lu_target *tg,
459 struct lsd_client_data *lcd, loff_t *off, struct thandle *th);
460 int tgt_server_data_read(const struct lu_env *env, struct lu_target *tg);
461 int tgt_server_data_write(const struct lu_env *env, struct lu_target *tg,
463 int tgt_server_data_update(const struct lu_env *env, struct lu_target *tg,
465 int tgt_truncate_last_rcvd(const struct lu_env *env, struct lu_target *tg,
467 int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt);
468 bool tgt_lookup_reply(struct ptlrpc_request *req, struct tg_reply_data *trd);
469 int tgt_add_reply_data(const struct lu_env *env, struct lu_target *tgt,
470 struct tg_export_data *ted, struct tg_reply_data *trd,
471 struct thandle *th, bool update_lrd_file);
472 struct tg_reply_data *tgt_lookup_reply_by_xid(struct tg_export_data *ted,
475 /* target/update_trans.c */
476 int distribute_txn_init(const struct lu_env *env,
477 struct lu_target *lut,
478 struct target_distribute_txn_data *tdtd,
480 void distribute_txn_fini(const struct lu_env *env,
481 struct target_distribute_txn_data *tdtd);
483 /* target/update_recovery.c */
484 int insert_update_records_to_replay_list(struct target_distribute_txn_data *,
485 struct llog_update_record *,
486 struct llog_cookie *, __u32);
487 void dtrq_list_dump(struct target_distribute_txn_data *tdtd,
489 void dtrq_list_destroy(struct target_distribute_txn_data *tdtd);
490 int distribute_txn_replay_handle(struct lu_env *env,
491 struct target_distribute_txn_data *tdtd,
492 struct distribute_txn_replay_req *dtrq);
493 __u64 distribute_txn_get_next_transno(struct target_distribute_txn_data *tdtd);
494 struct distribute_txn_replay_req *
495 distribute_txn_get_next_req(struct target_distribute_txn_data *tdtd);
496 void dtrq_destroy(struct distribute_txn_replay_req *dtrq);
497 struct distribute_txn_replay_req_sub *
498 dtrq_sub_lookup(struct distribute_txn_replay_req *dtrq, __u32 mdt_index);
499 struct distribute_txn_replay_req *
500 distribute_txn_lookup_finish_list(struct target_distribute_txn_data *tdtd,
502 bool is_req_replayed_by_update(struct ptlrpc_request *req);
507 static inline int err_serious(int rc)
510 return -(-rc | ESERIOUS);
513 static inline int clear_serious(int rc)
516 rc = -(-rc & ~ESERIOUS);
520 static inline int is_serious(int rc)
522 return (rc < 0 && -rc & ESERIOUS);
526 * Unified target generic handers macros and generic functions.
528 #define TGT_RPC_HANDLER_HP(base, flags, opc, fn, hp, fmt, version) \
531 .th_fail_id = OBD_FAIL_ ## opc ## _NET, \
536 .th_version = version, \
539 #define TGT_RPC_HANDLER(base, flags, opc, fn, fmt, version) \
540 TGT_RPC_HANDLER_HP(base, flags, opc, fn, NULL, fmt, version)
542 /* MDT Request with a format known in advance */
543 #define TGT_MDT_HDL(flags, name, fn) \
544 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
546 /* Request with a format we do not yet know */
547 #define TGT_MDT_HDL_VAR(flags, name, fn) \
548 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, NULL, \
551 /* OST Request with a format known in advance */
552 #define TGT_OST_HDL(flags, name, fn) \
553 TGT_RPC_HANDLER(OST_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
555 #define TGT_OST_HDL_HP(flags, name, fn, hp) \
556 TGT_RPC_HANDLER_HP(OST_FIRST_OPC, flags, name, fn, hp, \
557 &RQF_ ## name, LUSTRE_OST_VERSION)
559 /* MGS request with a format known in advance */
560 #define TGT_MGS_HDL(flags, name, fn) \
561 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
563 #define TGT_MGS_HDL_VAR(flags, name, fn) \
564 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, NULL, \
568 * OBD handler macros and generic functions.
570 #define TGT_OBD_HDL(flags, name, fn) \
571 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
573 #define TGT_OBD_HDL_VAR(flags, name, fn) \
574 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, NULL, \
578 * DLM handler macros and generic functions.
580 #define TGT_DLM_HDL_VAR(flags, name, fn) \
581 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, NULL, \
583 #define TGT_DLM_HDL(flags, name, fn) \
584 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
588 * LLOG handler macros and generic functions.
590 #define TGT_LLOG_HDL_VAR(flags, name, fn) \
591 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, NULL, \
593 #define TGT_LLOG_HDL(flags, name, fn) \
594 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
598 * Sec context handler macros and generic functions.
600 #define TGT_SEC_HDL_VAR(flags, name, fn) \
601 TGT_RPC_HANDLER(SEC_FIRST_OPC, flags, name, fn, NULL, \
604 #define TGT_QUOTA_HDL(flags, name, fn) \
605 TGT_RPC_HANDLER(QUOTA_DQACQ, flags, name, fn, &RQF_ ## name, \
608 /* Sequence service handlers */
609 #define TGT_SEQ_HDL(flags, name, fn) \
610 TGT_RPC_HANDLER(SEQ_QUERY, flags, name, fn, &RQF_ ## name, \
613 /* FID Location Database handlers */
614 #define TGT_FLD_HDL_VAR(flags, name, fn) \
615 TGT_RPC_HANDLER(FLD_QUERY, flags, name, fn, NULL, \
619 #define TGT_LFSCK_HDL(flags, name, fn) \
620 TGT_RPC_HANDLER(LFSCK_FIRST_OPC, flags, name, fn, \
621 &RQF_ ## name, LUSTRE_OBD_VERSION)
623 /* Request with a format known in advance */
624 #define TGT_UPDATE_HDL(flags, name, fn) \
625 TGT_RPC_HANDLER(OUT_UPDATE, flags, name, fn, &RQF_ ## name, \
628 #endif /* __LUSTRE_LU_TARGET_H */