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, 2014, 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 {
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;
60 /* all of sub updates are linked here */
61 struct list_head dtrq_sub_list;
62 spinlock_t dtrq_sub_list_lock;
65 /* Each one represents a sub replay item under a distribute
66 * transaction. A distribute transaction will be operated in
67 * two or more MDTs, and updates on each MDT will be represented
68 * by this structure */
69 struct distribute_txn_replay_req_sub {
70 __u32 dtrqs_mdt_index;
72 /* All of cookies for the update will be linked here */
73 spinlock_t dtrqs_cookie_list_lock;
74 struct list_head dtrqs_cookie_list;
75 struct list_head dtrqs_list;
78 struct target_distribute_txn_data;
79 typedef int (*distribute_txn_replay_handler_t)(struct lu_env *env,
80 struct target_distribute_txn_data *tdtd,
81 struct distribute_txn_replay_req *dtrq);
82 struct target_distribute_txn_data {
83 /* Distribution ID is used to identify updates log on different
84 * MDTs for one operation */
85 spinlock_t tdtd_batchid_lock;
87 struct lu_target *tdtd_lut;
88 struct dt_object *tdtd_batchid_obj;
89 struct dt_device *tdtd_dt;
91 /* Committed batchid for distribute transaction */
92 __u64 tdtd_committed_batchid;
94 /* List for distribute transaction */
95 struct list_head tdtd_list;
97 /* Threads to manage distribute transaction */
98 wait_queue_head_t tdtd_commit_thread_waitq;
99 atomic_t tdtd_refcount;
101 /* recovery update */
102 distribute_txn_replay_handler_t tdtd_replay_handler;
103 struct list_head tdtd_replay_list;
104 spinlock_t tdtd_replay_list_lock;
105 /* last replay update transno */
106 __u64 tdtd_last_update_transno;
107 __u32 tdtd_replay_ready:1;
112 struct obd_device *lut_obd;
113 struct dt_device *lut_bottom;
115 struct target_distribute_txn_data *lut_tdtd;
116 struct ptlrpc_thread lut_tdtd_commit_thread;
118 /* supported opcodes and handlers for this target */
119 struct tgt_opc_slice *lut_slice;
120 __u32 lut_reply_fail_id;
121 __u32 lut_request_fail_id;
124 rwlock_t lut_sptlrpc_lock;
125 struct sptlrpc_rule_set lut_sptlrpc_rset;
126 spinlock_t lut_flags_lock;
128 unsigned int lut_syncjournal:1,
129 lut_sync_lock_cancel:2,
131 lut_no_reconstruct:1;
132 /** last_rcvd file */
133 struct dt_object *lut_last_rcvd;
134 /* transaction callbacks */
135 struct dt_txn_callback lut_txn_cb;
136 /** server data in last_rcvd file */
137 struct lr_server_data lut_lsd;
138 /** Server last transaction number */
139 __u64 lut_last_transno;
140 /** Lock protecting last transaction number */
141 spinlock_t lut_translock;
142 /** Lock protecting client bitmap */
143 spinlock_t lut_client_bitmap_lock;
144 /** Bitmap of known clients */
145 unsigned long *lut_client_bitmap;
146 /* Number of clients supporting multiple modify RPCs
147 * recorded in the bitmap */
148 atomic_t lut_num_clients;
149 /* Client generation to identify client slot reuse */
150 atomic_t lut_client_generation;
151 /** reply_data file */
152 struct dt_object *lut_reply_data;
153 /** Bitmap of used slots in the reply data file */
154 unsigned long **lut_reply_bitmap;
157 /* number of slots in reply bitmap */
158 #define LUT_REPLY_SLOTS_PER_CHUNK (1<<20)
159 #define LUT_REPLY_SLOTS_MAX_CHUNKS 16
164 struct tg_reply_data {
165 /** chain of reply data anchored in tg_export_data */
166 struct list_head trd_list;
167 /** copy of on-disk reply data */
168 struct lsd_reply_data trd_reply;
169 /** versions for Version Based Recovery */
170 __u64 trd_pre_versions[4];
171 /** slot index in reply_data file */
173 /** tag the client used */
177 extern struct lu_context_key tgt_session_key;
179 struct tgt_session_info {
181 * The following members will be filled explicitly
182 * with specific data in tgt_ses_init().
184 struct req_capsule *tsi_pill;
187 * Lock request for "habeo clavis" operations.
189 struct ldlm_request *tsi_dlm_req;
191 /* although we have export in req, there are cases when it is not
192 * available, e.g. closing files upon export destroy */
193 struct obd_export *tsi_exp;
194 const struct lu_env *tsi_env;
195 struct lu_target *tsi_tgt;
197 const struct mdt_body *tsi_mdt_body;
198 struct ost_body *tsi_ost_body;
199 struct lu_object *tsi_corpus;
201 struct lu_fid tsi_fid;
202 struct ldlm_res_id tsi_resid;
204 /* object affected by VBR, for last_rcvd_update */
205 struct dt_object *tsi_vbr_obj;
206 /* opdata for mdt_reint_open(), has the same value as
207 * ldlm_reply:lock_policy_res1. The tgt_update_last_rcvd() stores
208 * this value onto disk for recovery when tgt_txn_stop_cb() is called.
213 * Additional fail id that can be set by handler.
215 int tsi_reply_fail_id;
216 bool tsi_preprocessed;
221 static inline struct tgt_session_info *tgt_ses_info(const struct lu_env *env)
223 struct tgt_session_info *tsi;
225 LASSERT(env->le_ses != NULL);
226 tsi = lu_context_key_get(env->le_ses, &tgt_session_key);
231 static inline void tgt_vbr_obj_set(const struct lu_env *env,
232 struct dt_object *obj)
234 struct tgt_session_info *tsi;
236 if (env->le_ses != NULL) {
237 tsi = tgt_ses_info(env);
238 tsi->tsi_vbr_obj = obj;
242 static inline void tgt_opdata_set(const struct lu_env *env, __u64 flags)
244 struct tgt_session_info *tsi;
246 if (env->le_ses != NULL) {
247 tsi = tgt_ses_info(env);
248 tsi->tsi_opdata |= flags;
252 static inline void tgt_opdata_clear(const struct lu_env *env, __u64 flags)
254 struct tgt_session_info *tsi;
256 if (env->le_ses != NULL) {
257 tsi = tgt_ses_info(env);
258 tsi->tsi_opdata &= ~flags;
263 * Generic unified target support.
265 enum tgt_handler_flags {
267 * struct *_body is passed in the incoming message, and object
268 * identified by this fid exists on disk.
270 * "habeo corpus" == "I have a body"
272 HABEO_CORPUS = (1 << 0),
274 * struct ldlm_request is passed in the incoming message.
276 * "habeo clavis" == "I have a key"
278 HABEO_CLAVIS = (1 << 1),
280 * this request has fixed reply format, so that reply message can be
281 * packed by generic code.
283 * "habeo refero" == "I have a reply"
285 HABEO_REFERO = (1 << 2),
287 * this request will modify something, so check whether the file system
288 * is readonly or not, then return -EROFS to client asap if necessary.
290 * "mutabor" == "I shall modify"
296 /* The name of this handler. */
298 /* Fail id, check at the beginning */
302 /* Flags in enum tgt_handler_flags */
304 /* Request version for this opcode */
306 /* Handler function */
307 int (*th_act)(struct tgt_session_info *tsi);
308 /* Handler function for high priority requests */
309 void (*th_hp)(struct tgt_session_info *tsi);
310 /* Request format for this request */
311 const struct req_format *th_fmt;
314 struct tgt_opc_slice {
315 __u32 tos_opc_start; /* First op code */
316 __u32 tos_opc_end; /* Last op code */
317 struct tgt_handler *tos_hs; /* Registered handler */
320 static inline struct ptlrpc_request *tgt_ses_req(struct tgt_session_info *tsi)
322 return tsi->tsi_pill ? tsi->tsi_pill->rc_req : NULL;
325 static inline __u64 tgt_conn_flags(struct tgt_session_info *tsi)
327 LASSERT(tsi->tsi_exp);
328 return exp_connect_flags(tsi->tsi_exp);
331 static inline int req_is_replay(struct ptlrpc_request *req)
333 LASSERT(req->rq_reqmsg);
334 return !!(lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY);
337 static inline bool tgt_is_multimodrpcs_client(struct obd_export *exp)
339 return exp_connect_flags(exp) & OBD_CONNECT_MULTIMODRPCS;
343 /* target/tgt_handler.c */
344 int tgt_request_handle(struct ptlrpc_request *req);
345 char *tgt_name(struct lu_target *tgt);
346 void tgt_counter_incr(struct obd_export *exp, int opcode);
347 int tgt_connect_check_sptlrpc(struct ptlrpc_request *req,
348 struct obd_export *exp);
349 int tgt_adapt_sptlrpc_conf(struct lu_target *tgt, int initial);
350 int tgt_connect(struct tgt_session_info *tsi);
351 int tgt_disconnect(struct tgt_session_info *uti);
352 int tgt_obd_ping(struct tgt_session_info *tsi);
353 int tgt_enqueue(struct tgt_session_info *tsi);
354 int tgt_convert(struct tgt_session_info *tsi);
355 int tgt_bl_callback(struct tgt_session_info *tsi);
356 int tgt_cp_callback(struct tgt_session_info *tsi);
357 int tgt_llog_open(struct tgt_session_info *tsi);
358 int tgt_llog_close(struct tgt_session_info *tsi);
359 int tgt_llog_destroy(struct tgt_session_info *tsi);
360 int tgt_llog_read_header(struct tgt_session_info *tsi);
361 int tgt_llog_next_block(struct tgt_session_info *tsi);
362 int tgt_llog_prev_block(struct tgt_session_info *tsi);
363 int tgt_sec_ctx_init(struct tgt_session_info *tsi);
364 int tgt_sec_ctx_init_cont(struct tgt_session_info *tsi);
365 int tgt_sec_ctx_fini(struct tgt_session_info *tsi);
366 int tgt_sendpage(struct tgt_session_info *tsi, struct lu_rdpg *rdpg, int nob);
367 int tgt_validate_obdo(struct tgt_session_info *tsi, struct obdo *oa);
368 int tgt_sync(const struct lu_env *env, struct lu_target *tgt,
369 struct dt_object *obj, __u64 start, __u64 end);
371 int tgt_io_thread_init(struct ptlrpc_thread *thread);
372 void tgt_io_thread_done(struct ptlrpc_thread *thread);
374 int tgt_extent_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
375 __u64 start, __u64 end, struct lustre_handle *lh,
376 int mode, __u64 *flags);
377 void tgt_extent_unlock(struct lustre_handle *lh, ldlm_mode_t mode);
378 int tgt_brw_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
379 struct obd_ioobj *obj, struct niobuf_remote *nb,
380 struct lustre_handle *lh, int mode);
381 void tgt_brw_unlock(struct obd_ioobj *obj, struct niobuf_remote *niob,
382 struct lustre_handle *lh, int mode);
383 int tgt_brw_read(struct tgt_session_info *tsi);
384 int tgt_brw_write(struct tgt_session_info *tsi);
385 int tgt_hpreq_handler(struct ptlrpc_request *req);
386 void tgt_register_lfsck_in_notify(int (*notify)(const struct lu_env *,
388 struct lfsck_request *,
390 void tgt_register_lfsck_query(int (*query)(const struct lu_env *,
392 struct lfsck_request *));
393 bool req_can_reconstruct(struct ptlrpc_request *req, struct tg_reply_data *trd);
395 extern struct tgt_handler tgt_lfsck_handlers[];
396 extern struct tgt_handler tgt_obd_handlers[];
397 extern struct tgt_handler tgt_dlm_handlers[];
398 extern struct tgt_handler tgt_llog_handlers[];
399 extern struct tgt_handler tgt_out_handlers[];
400 extern struct tgt_handler fld_handlers[];
401 extern struct tgt_handler seq_handlers[];
403 typedef void (*tgt_cb_t)(struct lu_target *lut, __u64 transno,
404 void *data, int err);
405 struct tgt_commit_cb {
406 tgt_cb_t tgt_cb_func;
410 int tgt_hpreq_handler(struct ptlrpc_request *req);
412 /* target/tgt_main.c */
413 void tgt_boot_epoch_update(struct lu_target *lut);
414 int tgt_last_commit_cb_add(struct thandle *th, struct lu_target *lut,
415 struct obd_export *exp, __u64 transno);
416 int tgt_new_client_cb_add(struct thandle *th, struct obd_export *exp);
417 int tgt_init(const struct lu_env *env, struct lu_target *lut,
418 struct obd_device *obd, struct dt_device *dt,
419 struct tgt_opc_slice *slice,
420 int request_fail_id, int reply_fail_id);
421 void tgt_fini(const struct lu_env *env, struct lu_target *lut);
422 int tgt_client_alloc(struct obd_export *exp);
423 void tgt_client_free(struct obd_export *exp);
424 int tgt_client_del(const struct lu_env *env, struct obd_export *exp);
425 int tgt_client_add(const struct lu_env *env, struct obd_export *exp, int);
426 int tgt_client_new(const struct lu_env *env, struct obd_export *exp);
427 int tgt_client_data_read(const struct lu_env *env, struct lu_target *tg,
428 struct lsd_client_data *lcd, loff_t *off, int index);
429 int tgt_client_data_write(const struct lu_env *env, struct lu_target *tg,
430 struct lsd_client_data *lcd, loff_t *off, struct thandle *th);
431 int tgt_server_data_read(const struct lu_env *env, struct lu_target *tg);
432 int tgt_server_data_write(const struct lu_env *env, struct lu_target *tg,
434 int tgt_server_data_update(const struct lu_env *env, struct lu_target *tg,
436 int tgt_truncate_last_rcvd(const struct lu_env *env, struct lu_target *tg,
438 int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt);
439 bool tgt_lookup_reply(struct ptlrpc_request *req, struct tg_reply_data *trd);
441 /* target/update_trans.c */
442 int distribute_txn_init(const struct lu_env *env,
443 struct lu_target *lut,
444 struct target_distribute_txn_data *tdtd,
446 void distribute_txn_fini(const struct lu_env *env,
447 struct target_distribute_txn_data *tdtd);
449 /* target/update_recovery.c */
450 int insert_update_records_to_replay_list(struct target_distribute_txn_data *,
451 struct llog_update_record *,
452 struct llog_cookie *, __u32);
453 void dtrq_list_dump(struct target_distribute_txn_data *tdtd,
455 void dtrq_list_destroy(struct target_distribute_txn_data *tdtd);
456 int distribute_txn_replay_handle(struct lu_env *env,
457 struct target_distribute_txn_data *tdtd,
458 struct distribute_txn_replay_req *dtrq);
459 __u64 distribute_txn_get_next_transno(struct target_distribute_txn_data *tdtd);
460 struct distribute_txn_replay_req *
461 distribute_txn_get_next_req(struct target_distribute_txn_data *tdtd);
462 void dtrq_destroy(struct distribute_txn_replay_req *dtrq);
463 struct distribute_txn_replay_req_sub *
464 dtrq_sub_lookup(struct distribute_txn_replay_req *dtrq, __u32 mdt_index);
470 static inline int err_serious(int rc)
473 return -(-rc | ESERIOUS);
476 static inline int clear_serious(int rc)
479 rc = -(-rc & ~ESERIOUS);
483 static inline int is_serious(int rc)
485 return (rc < 0 && -rc & ESERIOUS);
489 * Do not return server-side uid/gid to remote client
491 static inline void tgt_drop_id(struct obd_export *exp, struct obdo *oa)
493 if (unlikely(exp_connect_rmtclient(exp))) {
496 oa->o_valid &= ~(OBD_MD_FLUID | OBD_MD_FLGID);
501 * Unified target generic handers macros and generic functions.
503 #define TGT_RPC_HANDLER_HP(base, flags, opc, fn, hp, fmt, version) \
506 .th_fail_id = OBD_FAIL_ ## opc ## _NET, \
511 .th_version = version, \
514 #define TGT_RPC_HANDLER(base, flags, opc, fn, fmt, version) \
515 TGT_RPC_HANDLER_HP(base, flags, opc, fn, NULL, fmt, version)
517 /* MDT Request with a format known in advance */
518 #define TGT_MDT_HDL(flags, name, fn) \
519 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
521 /* Request with a format we do not yet know */
522 #define TGT_MDT_HDL_VAR(flags, name, fn) \
523 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, NULL, \
526 /* OST Request with a format known in advance */
527 #define TGT_OST_HDL(flags, name, fn) \
528 TGT_RPC_HANDLER(OST_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
530 #define TGT_OST_HDL_HP(flags, name, fn, hp) \
531 TGT_RPC_HANDLER_HP(OST_FIRST_OPC, flags, name, fn, hp, \
532 &RQF_ ## name, LUSTRE_OST_VERSION)
534 /* MGS request with a format known in advance */
535 #define TGT_MGS_HDL(flags, name, fn) \
536 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
538 #define TGT_MGS_HDL_VAR(flags, name, fn) \
539 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, NULL, \
543 * OBD handler macros and generic functions.
545 #define TGT_OBD_HDL(flags, name, fn) \
546 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
548 #define TGT_OBD_HDL_VAR(flags, name, fn) \
549 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, NULL, \
553 * DLM handler macros and generic functions.
555 #define TGT_DLM_HDL_VAR(flags, name, fn) \
556 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, NULL, \
558 #define TGT_DLM_HDL(flags, name, fn) \
559 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
563 * LLOG handler macros and generic functions.
565 #define TGT_LLOG_HDL_VAR(flags, name, fn) \
566 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, NULL, \
568 #define TGT_LLOG_HDL(flags, name, fn) \
569 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
573 * Sec context handler macros and generic functions.
575 #define TGT_SEC_HDL_VAR(flags, name, fn) \
576 TGT_RPC_HANDLER(SEC_FIRST_OPC, flags, name, fn, NULL, \
579 #define TGT_QUOTA_HDL(flags, name, fn) \
580 TGT_RPC_HANDLER(QUOTA_DQACQ, flags, name, fn, &RQF_ ## name, \
583 /* Sequence service handlers */
584 #define TGT_SEQ_HDL(flags, name, fn) \
585 TGT_RPC_HANDLER(SEQ_QUERY, flags, name, fn, &RQF_ ## name, \
588 /* FID Location Database handlers */
589 #define TGT_FLD_HDL_VAR(flags, name, fn) \
590 TGT_RPC_HANDLER(FLD_QUERY, flags, name, fn, NULL, \
594 #define TGT_LFSCK_HDL(flags, name, fn) \
595 TGT_RPC_HANDLER(LFSCK_FIRST_OPC, flags, name, fn, \
596 &RQF_ ## name, LUSTRE_OBD_VERSION)
598 /* Request with a format known in advance */
599 #define TGT_UPDATE_HDL(flags, name, fn) \
600 TGT_RPC_HANDLER(OUT_UPDATE, flags, name, fn, &RQF_ ## name, \
603 #endif /* __LUSTRE_LU_TARGET_H */