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;
58 /* all of sub updates are linked here */
59 struct list_head dtrq_sub_list;
60 spinlock_t dtrq_sub_list_lock;
63 /* Each one represents a sub replay item under a distribute
64 * transaction. A distribute transaction will be operated in
65 * two or more MDTs, and updates on each MDT will be represented
66 * by this structure */
67 struct distribute_txn_replay_req_sub {
68 __u32 dtrqs_mdt_index;
69 struct llog_cookie dtrqs_llog_cookie;
70 struct list_head dtrqs_list;
73 struct target_distribute_txn_data;
74 typedef int (*distribute_txn_replay_handler_t)(struct lu_env *env,
75 struct target_distribute_txn_data *tdtd,
76 struct distribute_txn_replay_req *dtrq);
77 struct target_distribute_txn_data {
78 /* Distribution ID is used to identify updates log on different
79 * MDTs for one operation */
80 spinlock_t tdtd_batchid_lock;
82 struct lu_target *tdtd_lut;
83 struct dt_object *tdtd_batchid_obj;
84 struct dt_device *tdtd_dt;
86 /* Committed batchid for distribute transaction */
87 __u64 tdtd_committed_batchid;
89 /* List for distribute transaction */
90 struct list_head tdtd_list;
92 /* Threads to manage distribute transaction */
93 wait_queue_head_t tdtd_commit_thread_waitq;
94 atomic_t tdtd_refcount;
97 distribute_txn_replay_handler_t tdtd_replay_handler;
98 struct list_head tdtd_replay_list;
99 spinlock_t tdtd_replay_list_lock;
100 /* last replay update transno */
101 __u64 tdtd_last_update_transno;
102 __u32 tdtd_replay_ready:1;
107 struct obd_device *lut_obd;
108 struct dt_device *lut_bottom;
110 struct target_distribute_txn_data *lut_tdtd;
111 struct ptlrpc_thread lut_tdtd_commit_thread;
113 /* supported opcodes and handlers for this target */
114 struct tgt_opc_slice *lut_slice;
115 __u32 lut_reply_fail_id;
116 __u32 lut_request_fail_id;
119 rwlock_t lut_sptlrpc_lock;
120 struct sptlrpc_rule_set lut_sptlrpc_rset;
121 spinlock_t lut_flags_lock;
123 unsigned int lut_syncjournal:1,
124 lut_sync_lock_cancel:2,
126 lut_no_reconstruct:1;
127 /** last_rcvd file */
128 struct dt_object *lut_last_rcvd;
129 /* transaction callbacks */
130 struct dt_txn_callback lut_txn_cb;
131 /** server data in last_rcvd file */
132 struct lr_server_data lut_lsd;
133 /** Server last transaction number */
134 __u64 lut_last_transno;
135 /** Lock protecting last transaction number */
136 spinlock_t lut_translock;
137 /** Lock protecting client bitmap */
138 spinlock_t lut_client_bitmap_lock;
139 /** Bitmap of known clients */
140 unsigned long *lut_client_bitmap;
141 /* Number of clients supporting multiple modify RPCs
142 * recorded in the bitmap */
143 atomic_t lut_num_clients;
144 /* Client generation to identify client slot reuse */
145 atomic_t lut_client_generation;
146 /** reply_data file */
147 struct dt_object *lut_reply_data;
148 /** Bitmap of used slots in the reply data file */
149 unsigned long **lut_reply_bitmap;
152 /* number of slots in reply bitmap */
153 #define LUT_REPLY_SLOTS_PER_CHUNK (1<<20)
154 #define LUT_REPLY_SLOTS_MAX_CHUNKS 16
159 struct tg_reply_data {
160 /** chain of reply data anchored in tg_export_data */
161 struct list_head trd_list;
162 /** copy of on-disk reply data */
163 struct lsd_reply_data trd_reply;
164 /** versions for Version Based Recovery */
165 __u64 trd_pre_versions[4];
166 /** slot index in reply_data file */
168 /** tag the client used */
172 extern struct lu_context_key tgt_session_key;
174 struct tgt_session_info {
176 * The following members will be filled explicitly
177 * with specific data in tgt_ses_init().
179 struct req_capsule *tsi_pill;
182 * Lock request for "habeo clavis" operations.
184 struct ldlm_request *tsi_dlm_req;
186 /* although we have export in req, there are cases when it is not
187 * available, e.g. closing files upon export destroy */
188 struct obd_export *tsi_exp;
189 const struct lu_env *tsi_env;
190 struct lu_target *tsi_tgt;
192 const struct mdt_body *tsi_mdt_body;
193 struct ost_body *tsi_ost_body;
194 struct lu_object *tsi_corpus;
196 struct lu_fid tsi_fid;
197 struct ldlm_res_id tsi_resid;
199 /* object affected by VBR, for last_rcvd_update */
200 struct dt_object *tsi_vbr_obj;
201 /* opdata for mdt_reint_open(), has the same value as
202 * ldlm_reply:lock_policy_res1. The tgt_update_last_rcvd() stores
203 * this value onto disk for recovery when tgt_txn_stop_cb() is called.
208 * Additional fail id that can be set by handler.
210 int tsi_reply_fail_id;
211 bool tsi_preprocessed;
216 static inline struct tgt_session_info *tgt_ses_info(const struct lu_env *env)
218 struct tgt_session_info *tsi;
220 LASSERT(env->le_ses != NULL);
221 tsi = lu_context_key_get(env->le_ses, &tgt_session_key);
226 static inline void tgt_vbr_obj_set(const struct lu_env *env,
227 struct dt_object *obj)
229 struct tgt_session_info *tsi;
231 if (env->le_ses != NULL) {
232 tsi = tgt_ses_info(env);
233 tsi->tsi_vbr_obj = obj;
237 static inline void tgt_opdata_set(const struct lu_env *env, __u64 flags)
239 struct tgt_session_info *tsi;
241 if (env->le_ses != NULL) {
242 tsi = tgt_ses_info(env);
243 tsi->tsi_opdata |= flags;
247 static inline void tgt_opdata_clear(const struct lu_env *env, __u64 flags)
249 struct tgt_session_info *tsi;
251 if (env->le_ses != NULL) {
252 tsi = tgt_ses_info(env);
253 tsi->tsi_opdata &= ~flags;
258 * Generic unified target support.
260 enum tgt_handler_flags {
262 * struct *_body is passed in the incoming message, and object
263 * identified by this fid exists on disk.
265 * "habeo corpus" == "I have a body"
267 HABEO_CORPUS = (1 << 0),
269 * struct ldlm_request is passed in the incoming message.
271 * "habeo clavis" == "I have a key"
273 HABEO_CLAVIS = (1 << 1),
275 * this request has fixed reply format, so that reply message can be
276 * packed by generic code.
278 * "habeo refero" == "I have a reply"
280 HABEO_REFERO = (1 << 2),
282 * this request will modify something, so check whether the file system
283 * is readonly or not, then return -EROFS to client asap if necessary.
285 * "mutabor" == "I shall modify"
291 /* The name of this handler. */
293 /* Fail id, check at the beginning */
297 /* Flags in enum tgt_handler_flags */
299 /* Request version for this opcode */
301 /* Handler function */
302 int (*th_act)(struct tgt_session_info *tsi);
303 /* Handler function for high priority requests */
304 void (*th_hp)(struct tgt_session_info *tsi);
305 /* Request format for this request */
306 const struct req_format *th_fmt;
309 struct tgt_opc_slice {
310 __u32 tos_opc_start; /* First op code */
311 __u32 tos_opc_end; /* Last op code */
312 struct tgt_handler *tos_hs; /* Registered handler */
315 static inline struct ptlrpc_request *tgt_ses_req(struct tgt_session_info *tsi)
317 return tsi->tsi_pill ? tsi->tsi_pill->rc_req : NULL;
320 static inline __u64 tgt_conn_flags(struct tgt_session_info *tsi)
322 LASSERT(tsi->tsi_exp);
323 return exp_connect_flags(tsi->tsi_exp);
326 static inline int req_is_replay(struct ptlrpc_request *req)
328 LASSERT(req->rq_reqmsg);
329 return !!(lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY);
332 static inline bool tgt_is_multimodrpcs_client(struct obd_export *exp)
334 return exp_connect_flags(exp) & OBD_CONNECT_MULTIMODRPCS;
338 /* target/tgt_handler.c */
339 int tgt_request_handle(struct ptlrpc_request *req);
340 char *tgt_name(struct lu_target *tgt);
341 void tgt_counter_incr(struct obd_export *exp, int opcode);
342 int tgt_connect_check_sptlrpc(struct ptlrpc_request *req,
343 struct obd_export *exp);
344 int tgt_adapt_sptlrpc_conf(struct lu_target *tgt, int initial);
345 int tgt_connect(struct tgt_session_info *tsi);
346 int tgt_disconnect(struct tgt_session_info *uti);
347 int tgt_obd_ping(struct tgt_session_info *tsi);
348 int tgt_enqueue(struct tgt_session_info *tsi);
349 int tgt_convert(struct tgt_session_info *tsi);
350 int tgt_bl_callback(struct tgt_session_info *tsi);
351 int tgt_cp_callback(struct tgt_session_info *tsi);
352 int tgt_llog_open(struct tgt_session_info *tsi);
353 int tgt_llog_close(struct tgt_session_info *tsi);
354 int tgt_llog_destroy(struct tgt_session_info *tsi);
355 int tgt_llog_read_header(struct tgt_session_info *tsi);
356 int tgt_llog_next_block(struct tgt_session_info *tsi);
357 int tgt_llog_prev_block(struct tgt_session_info *tsi);
358 int tgt_sec_ctx_init(struct tgt_session_info *tsi);
359 int tgt_sec_ctx_init_cont(struct tgt_session_info *tsi);
360 int tgt_sec_ctx_fini(struct tgt_session_info *tsi);
361 int tgt_sendpage(struct tgt_session_info *tsi, struct lu_rdpg *rdpg, int nob);
362 int tgt_validate_obdo(struct tgt_session_info *tsi, struct obdo *oa);
363 int tgt_sync(const struct lu_env *env, struct lu_target *tgt,
364 struct dt_object *obj, __u64 start, __u64 end);
366 int tgt_io_thread_init(struct ptlrpc_thread *thread);
367 void tgt_io_thread_done(struct ptlrpc_thread *thread);
369 int tgt_extent_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
370 __u64 start, __u64 end, struct lustre_handle *lh,
371 int mode, __u64 *flags);
372 void tgt_extent_unlock(struct lustre_handle *lh, ldlm_mode_t mode);
373 int tgt_brw_lock(struct ldlm_namespace *ns, struct ldlm_res_id *res_id,
374 struct obd_ioobj *obj, struct niobuf_remote *nb,
375 struct lustre_handle *lh, int mode);
376 void tgt_brw_unlock(struct obd_ioobj *obj, struct niobuf_remote *niob,
377 struct lustre_handle *lh, int mode);
378 int tgt_brw_read(struct tgt_session_info *tsi);
379 int tgt_brw_write(struct tgt_session_info *tsi);
380 int tgt_hpreq_handler(struct ptlrpc_request *req);
381 void tgt_register_lfsck_in_notify(int (*notify)(const struct lu_env *,
383 struct lfsck_request *,
385 void tgt_register_lfsck_query(int (*query)(const struct lu_env *,
387 struct lfsck_request *));
388 bool req_can_reconstruct(struct ptlrpc_request *req, struct tg_reply_data *trd);
390 extern struct tgt_handler tgt_lfsck_handlers[];
391 extern struct tgt_handler tgt_obd_handlers[];
392 extern struct tgt_handler tgt_dlm_handlers[];
393 extern struct tgt_handler tgt_llog_handlers[];
394 extern struct tgt_handler tgt_out_handlers[];
395 extern struct tgt_handler fld_handlers[];
396 extern struct tgt_handler seq_handlers[];
398 typedef void (*tgt_cb_t)(struct lu_target *lut, __u64 transno,
399 void *data, int err);
400 struct tgt_commit_cb {
401 tgt_cb_t tgt_cb_func;
405 int tgt_hpreq_handler(struct ptlrpc_request *req);
407 /* target/tgt_main.c */
408 void tgt_boot_epoch_update(struct lu_target *lut);
409 int tgt_last_commit_cb_add(struct thandle *th, struct lu_target *lut,
410 struct obd_export *exp, __u64 transno);
411 int tgt_new_client_cb_add(struct thandle *th, struct obd_export *exp);
412 int tgt_init(const struct lu_env *env, struct lu_target *lut,
413 struct obd_device *obd, struct dt_device *dt,
414 struct tgt_opc_slice *slice,
415 int request_fail_id, int reply_fail_id);
416 void tgt_fini(const struct lu_env *env, struct lu_target *lut);
417 int tgt_client_alloc(struct obd_export *exp);
418 void tgt_client_free(struct obd_export *exp);
419 int tgt_client_del(const struct lu_env *env, struct obd_export *exp);
420 int tgt_client_add(const struct lu_env *env, struct obd_export *exp, int);
421 int tgt_client_new(const struct lu_env *env, struct obd_export *exp);
422 int tgt_client_data_read(const struct lu_env *env, struct lu_target *tg,
423 struct lsd_client_data *lcd, loff_t *off, int index);
424 int tgt_client_data_write(const struct lu_env *env, struct lu_target *tg,
425 struct lsd_client_data *lcd, loff_t *off, struct thandle *th);
426 int tgt_server_data_read(const struct lu_env *env, struct lu_target *tg);
427 int tgt_server_data_write(const struct lu_env *env, struct lu_target *tg,
429 int tgt_server_data_update(const struct lu_env *env, struct lu_target *tg,
431 int tgt_truncate_last_rcvd(const struct lu_env *env, struct lu_target *tg,
433 int tgt_reply_data_init(const struct lu_env *env, struct lu_target *tgt);
434 bool tgt_lookup_reply(struct ptlrpc_request *req, struct tg_reply_data *trd);
436 /* target/update_trans.c */
437 int distribute_txn_init(const struct lu_env *env,
438 struct lu_target *lut,
439 struct target_distribute_txn_data *tdtd,
441 void distribute_txn_fini(const struct lu_env *env,
442 struct target_distribute_txn_data *tdtd);
444 /* target/update_recovery.c */
445 int insert_update_records_to_replay_list(struct target_distribute_txn_data *,
446 struct llog_update_record *,
447 struct llog_cookie *, __u32);
448 void dtrq_list_dump(struct target_distribute_txn_data *tdtd,
450 void dtrq_list_destroy(struct target_distribute_txn_data *tdtd);
451 int distribute_txn_replay_handle(struct lu_env *env,
452 struct target_distribute_txn_data *tdtd,
453 struct distribute_txn_replay_req *dtrq);
454 __u64 distribute_txn_get_next_transno(struct target_distribute_txn_data *tdtd);
455 struct distribute_txn_replay_req *
456 distribute_txn_get_next_req(struct target_distribute_txn_data *tdtd);
457 void dtrq_destroy(struct distribute_txn_replay_req *dtrq);
458 struct distribute_txn_replay_req_sub *
459 dtrq_sub_lookup(struct distribute_txn_replay_req *dtrq, __u32 mdt_index);
465 static inline int err_serious(int rc)
468 return -(-rc | ESERIOUS);
471 static inline int clear_serious(int rc)
474 rc = -(-rc & ~ESERIOUS);
478 static inline int is_serious(int rc)
480 return (rc < 0 && -rc & ESERIOUS);
484 * Do not return server-side uid/gid to remote client
486 static inline void tgt_drop_id(struct obd_export *exp, struct obdo *oa)
488 if (unlikely(exp_connect_rmtclient(exp))) {
491 oa->o_valid &= ~(OBD_MD_FLUID | OBD_MD_FLGID);
496 * Unified target generic handers macros and generic functions.
498 #define TGT_RPC_HANDLER_HP(base, flags, opc, fn, hp, fmt, version) \
501 .th_fail_id = OBD_FAIL_ ## opc ## _NET, \
506 .th_version = version, \
509 #define TGT_RPC_HANDLER(base, flags, opc, fn, fmt, version) \
510 TGT_RPC_HANDLER_HP(base, flags, opc, fn, NULL, fmt, version)
512 /* MDT Request with a format known in advance */
513 #define TGT_MDT_HDL(flags, name, fn) \
514 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
516 /* Request with a format we do not yet know */
517 #define TGT_MDT_HDL_VAR(flags, name, fn) \
518 TGT_RPC_HANDLER(MDS_FIRST_OPC, flags, name, fn, NULL, \
521 /* OST Request with a format known in advance */
522 #define TGT_OST_HDL(flags, name, fn) \
523 TGT_RPC_HANDLER(OST_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
525 #define TGT_OST_HDL_HP(flags, name, fn, hp) \
526 TGT_RPC_HANDLER_HP(OST_FIRST_OPC, flags, name, fn, hp, \
527 &RQF_ ## name, LUSTRE_OST_VERSION)
529 /* MGS request with a format known in advance */
530 #define TGT_MGS_HDL(flags, name, fn) \
531 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
533 #define TGT_MGS_HDL_VAR(flags, name, fn) \
534 TGT_RPC_HANDLER(MGS_FIRST_OPC, flags, name, fn, NULL, \
538 * OBD handler macros and generic functions.
540 #define TGT_OBD_HDL(flags, name, fn) \
541 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
543 #define TGT_OBD_HDL_VAR(flags, name, fn) \
544 TGT_RPC_HANDLER(OBD_FIRST_OPC, flags, name, fn, NULL, \
548 * DLM handler macros and generic functions.
550 #define TGT_DLM_HDL_VAR(flags, name, fn) \
551 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, NULL, \
553 #define TGT_DLM_HDL(flags, name, fn) \
554 TGT_RPC_HANDLER(LDLM_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
558 * LLOG handler macros and generic functions.
560 #define TGT_LLOG_HDL_VAR(flags, name, fn) \
561 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, NULL, \
563 #define TGT_LLOG_HDL(flags, name, fn) \
564 TGT_RPC_HANDLER(LLOG_FIRST_OPC, flags, name, fn, &RQF_ ## name, \
568 * Sec context handler macros and generic functions.
570 #define TGT_SEC_HDL_VAR(flags, name, fn) \
571 TGT_RPC_HANDLER(SEC_FIRST_OPC, flags, name, fn, NULL, \
574 #define TGT_QUOTA_HDL(flags, name, fn) \
575 TGT_RPC_HANDLER(QUOTA_DQACQ, flags, name, fn, &RQF_ ## name, \
578 /* Sequence service handlers */
579 #define TGT_SEQ_HDL(flags, name, fn) \
580 TGT_RPC_HANDLER(SEQ_QUERY, flags, name, fn, &RQF_ ## name, \
583 /* FID Location Database handlers */
584 #define TGT_FLD_HDL_VAR(flags, name, fn) \
585 TGT_RPC_HANDLER(FLD_QUERY, flags, name, fn, NULL, \
589 #define TGT_LFSCK_HDL(flags, name, fn) \
590 TGT_RPC_HANDLER(LFSCK_FIRST_OPC, flags, name, fn, \
591 &RQF_ ## name, LUSTRE_OBD_VERSION)
593 /* Request with a format known in advance */
594 #define TGT_UPDATE_HDL(flags, name, fn) \
595 TGT_RPC_HANDLER(OUT_UPDATE, flags, name, fn, &RQF_ ## name, \
598 #endif /* __LUSTRE_LU_TARGET_H */