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) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2012, 2015, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/osp/osp_sync.c
38 * Lustre OST Proxy Device
40 * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
41 * Author: Mikhail Pershin <mike.pershin@intel.com>
44 #define DEBUG_SUBSYSTEM S_MDS
46 #include <linux/kthread.h>
47 #include <lustre_log.h>
48 #include <lustre_update.h>
49 #include "osp_internal.h"
51 static int osp_sync_id_traction_init(struct osp_device *d);
52 static void osp_sync_id_traction_fini(struct osp_device *d);
53 static __u32 osp_sync_id_get(struct osp_device *d, __u32 id);
54 static void osp_sync_remove_from_tracker(struct osp_device *d);
57 * this is a components of OSP implementing synchronization between MDS and OST
58 * it llogs all interesting changes (currently it's uig/gid change and object
59 * destroy) atomically, then makes sure changes hit OST storage
61 * we have 4 queues of work:
63 * the first queue is llog itself, once read a change is stored in 2nd queue
64 * in form of RPC (but RPC isn't fired yet).
66 * the second queue (opd_syn_waiting_for_commit) holds changes awaiting local
67 * commit. once change is committed locally it migrates onto 3rd queue.
69 * the third queue (opd_syn_committed_here) holds changes committed locally,
70 * but not sent to OST (as the pipe can be full). once pipe becomes non-full
71 * we take a change from the queue and fire corresponded RPC.
73 * once RPC is reported committed by OST (using regular last_committed mech.)
74 * the change jumps into 4th queue (opd_syn_committed_there), now we can
75 * cancel corresponded llog record and release RPC
77 * opd_syn_changes is a number of unread llog records (to be processed).
78 * notice this number doesn't include llog records from previous boots.
79 * with OSP_SYN_THRESHOLD we try to batch processing a bit (TO BE IMPLEMENTED)
81 * opd_syn_rpc_in_progress is a number of requests in 2-4 queues.
82 * we control this with OSP_MAX_IN_PROGRESS so that OSP don't consume
83 * too much memory -- how to deal with 1000th OSTs ? batching could help?
85 * opd_syn_rpc_in_flight is a number of RPC in flight.
86 * we control this with OSP_MAX_IN_FLIGHT
89 /* XXX: do math to learn reasonable threshold
90 * should it be ~ number of changes fitting bulk? */
92 #define OSP_SYN_THRESHOLD 10
93 #define OSP_MAX_IN_FLIGHT 8
94 #define OSP_MAX_IN_PROGRESS 4096
96 #define OSP_JOB_MAGIC 0x26112005
98 struct osp_job_req_args {
99 /** bytes reserved for ptlrpc_replay_req() */
100 struct ptlrpc_replay_async_args jra_raa;
101 struct list_head jra_committed_link;
102 struct list_head jra_inflight_link;
106 static inline int osp_sync_running(struct osp_device *d)
108 return !!(d->opd_syn_thread.t_flags & SVC_RUNNING);
112 * Check status: whether OSP thread has stopped
114 * \param[in] d OSP device
116 * \retval 0 still running
119 static inline int osp_sync_stopped(struct osp_device *d)
121 return !!(d->opd_syn_thread.t_flags & SVC_STOPPED);
125 ** Check for new changes to sync
127 * \param[in] d OSP device
129 * \retval 1 there are changes
130 * \retval 0 there are no changes
132 static inline int osp_sync_has_new_job(struct osp_device *d)
134 return ((d->opd_syn_last_processed_id < d->opd_syn_last_used_id) &&
135 (d->opd_syn_last_processed_id < d->opd_syn_last_committed_id))
136 || (d->opd_syn_prev_done == 0);
139 static inline int osp_sync_inflight_conflict(struct osp_device *d,
140 struct llog_rec_hdr *h)
142 struct osp_job_req_args *jra;
146 if (h == NULL || h->lrh_type == LLOG_GEN_REC ||
147 list_empty(&d->opd_syn_inflight_list))
150 memset(&ostid, 0, sizeof(ostid));
151 switch (h->lrh_type) {
153 ostid_set_seq(&ostid, ((struct llog_unlink_rec *)h)->lur_oseq);
154 ostid_set_id(&ostid, ((struct llog_unlink_rec *)h)->lur_oid);
156 case MDS_UNLINK64_REC:
157 fid_to_ostid(&((struct llog_unlink64_rec *)h)->lur_fid, &ostid);
159 case MDS_SETATTR64_REC:
160 ostid = ((struct llog_setattr64_rec *)h)->lsr_oi;
166 spin_lock(&d->opd_syn_lock);
167 list_for_each_entry(jra, &d->opd_syn_inflight_list, jra_inflight_link) {
168 struct ptlrpc_request *req;
169 struct ost_body *body;
171 LASSERT(jra->jra_magic == OSP_JOB_MAGIC);
173 req = container_of((void *)jra, struct ptlrpc_request,
175 body = req_capsule_client_get(&req->rq_pill,
179 if (memcmp(&ostid, &body->oa.o_oi, sizeof(ostid)) == 0) {
184 spin_unlock(&d->opd_syn_lock);
189 static inline int osp_sync_low_in_progress(struct osp_device *d)
191 return d->opd_syn_rpc_in_progress < d->opd_syn_max_rpc_in_progress;
195 * Check for room in the network pipe to OST
197 * \param[in] d OSP device
199 * \retval 1 there is room
200 * \retval 0 no room, the pipe is full
202 static inline int osp_sync_low_in_flight(struct osp_device *d)
204 return d->opd_syn_rpc_in_flight < d->opd_syn_max_rpc_in_flight;
208 * Wake up check for the main sync thread
210 * \param[in] d OSP device
212 * \retval 1 time to wake up
213 * \retval 0 no need to wake up
215 static inline int osp_sync_has_work(struct osp_device *d)
217 /* has new/old changes and low in-progress? */
218 if (osp_sync_has_new_job(d) && osp_sync_low_in_progress(d) &&
219 osp_sync_low_in_flight(d) && d->opd_imp_connected)
222 /* has remotely committed? */
223 if (!list_empty(&d->opd_syn_committed_there))
229 #define osp_sync_check_for_work(d) \
231 if (osp_sync_has_work(d)) { \
232 wake_up(&d->opd_syn_waitq); \
236 void __osp_sync_check_for_work(struct osp_device *d)
238 osp_sync_check_for_work(d);
242 * Check and return ready-for-new status.
244 * The thread processing llog record uses this function to check whether
245 * it's time to take another record and process it. The number of conditions
246 * must be met: the connection should be ready, RPCs in flight not exceeding
247 * the limit, the record is committed locally, etc (see the lines below).
249 * \param[in] d OSP device
250 * \param[in] rec next llog record to process
252 * \retval 0 not ready
255 static inline int osp_sync_can_process_new(struct osp_device *d,
256 struct llog_rec_hdr *rec)
260 if (unlikely(atomic_read(&d->opd_syn_barrier) > 0))
262 if (unlikely(osp_sync_inflight_conflict(d, rec)))
264 if (!osp_sync_low_in_progress(d))
266 if (!osp_sync_low_in_flight(d))
268 if (!d->opd_imp_connected)
270 if (d->opd_syn_prev_done == 0)
272 if (d->opd_syn_changes == 0)
274 if (rec == NULL || rec->lrh_id <= d->opd_syn_last_committed_id)
280 * Declare intention to add a new change.
282 * With regard to OSD API, we have to declare any changes ahead. In this
283 * case we declare an intention to add a llog record representing the
284 * change on the local storage.
286 * \param[in] env LU environment provided by the caller
287 * \param[in] o OSP object
288 * \param[in] type type of change: MDS_UNLINK64_REC or MDS_SETATTR64_REC
289 * \param[in] th transaction handle (local)
291 * \retval 0 on success
292 * \retval negative negated errno on error
294 int osp_sync_declare_add(const struct lu_env *env, struct osp_object *o,
295 llog_op_type type, struct thandle *th)
297 struct osp_thread_info *osi = osp_env_info(env);
298 struct osp_device *d = lu2osp_dev(o->opo_obj.do_lu.lo_dev);
299 struct llog_ctxt *ctxt;
300 struct thandle *storage_th;
305 /* it's a layering violation, to access internals of th,
306 * but we can do this as a sanity check, for a while */
307 LASSERT(th->th_top != NULL);
308 storage_th = thandle_get_sub_by_dt(env, th->th_top, d->opd_storage);
309 if (IS_ERR(storage_th))
310 RETURN(PTR_ERR(storage_th));
313 case MDS_UNLINK64_REC:
314 osi->osi_hdr.lrh_len = sizeof(struct llog_unlink64_rec);
316 case MDS_SETATTR64_REC:
317 osi->osi_hdr.lrh_len = sizeof(struct llog_setattr64_rec);
323 /* we want ->dt_trans_start() to allocate per-thandle structure */
324 storage_th->th_tags |= LCT_OSP_THREAD;
326 ctxt = llog_get_context(d->opd_obd, LLOG_MDS_OST_ORIG_CTXT);
329 rc = llog_declare_add(env, ctxt->loc_handle, &osi->osi_hdr,
337 * Generate a llog record for a given change.
339 * Generates a llog record for the change passed. The change can be of two
340 * types: unlink and setattr. The record gets an ID which later will be
341 * used to track commit status of the change. For unlink changes, the caller
342 * can supply a starting FID and the count of the objects to destroy. For
343 * setattr the caller should apply attributes to apply.
346 * \param[in] env LU environment provided by the caller
347 * \param[in] d OSP device
348 * \param[in] fid fid of the object the change should be applied to
349 * \param[in] type type of change: MDS_UNLINK64_REC or MDS_SETATTR64_REC
350 * \param[in] count count of objects to destroy
351 * \param[in] th transaction handle (local)
352 * \param[in] attr attributes for setattr
354 * \retval 0 on success
355 * \retval negative negated errno on error
357 static int osp_sync_add_rec(const struct lu_env *env, struct osp_device *d,
358 const struct lu_fid *fid, llog_op_type type,
359 int count, struct thandle *th,
360 const struct lu_attr *attr)
362 struct osp_thread_info *osi = osp_env_info(env);
363 struct llog_ctxt *ctxt;
364 struct osp_txn_info *txn;
365 struct thandle *storage_th;
370 /* it's a layering violation, to access internals of th,
371 * but we can do this as a sanity check, for a while */
372 LASSERT(th->th_top != NULL);
373 storage_th = thandle_get_sub_by_dt(env, th->th_top, d->opd_storage);
374 if (IS_ERR(storage_th))
375 RETURN(PTR_ERR(storage_th));
378 case MDS_UNLINK64_REC:
379 osi->osi_hdr.lrh_len = sizeof(osi->osi_unlink);
380 osi->osi_hdr.lrh_type = MDS_UNLINK64_REC;
381 osi->osi_unlink.lur_fid = *fid;
382 osi->osi_unlink.lur_count = count;
384 case MDS_SETATTR64_REC:
385 rc = fid_to_ostid(fid, &osi->osi_oi);
387 osi->osi_hdr.lrh_len = sizeof(osi->osi_setattr);
388 osi->osi_hdr.lrh_type = MDS_SETATTR64_REC;
389 osi->osi_setattr.lsr_oi = osi->osi_oi;
391 osi->osi_setattr.lsr_uid = attr->la_uid;
392 osi->osi_setattr.lsr_gid = attr->la_gid;
393 osi->osi_setattr.lsr_valid =
394 ((attr->la_valid & LA_UID) ? OBD_MD_FLUID : 0) |
395 ((attr->la_valid & LA_GID) ? OBD_MD_FLGID : 0);
401 txn = osp_txn_info(&storage_th->th_ctx);
404 txn->oti_current_id = osp_sync_id_get(d, txn->oti_current_id);
405 osi->osi_hdr.lrh_id = txn->oti_current_id;
407 ctxt = llog_get_context(d->opd_obd, LLOG_MDS_OST_ORIG_CTXT);
411 rc = llog_add(env, ctxt->loc_handle, &osi->osi_hdr, &osi->osi_cookie,
415 if (likely(rc >= 0)) {
416 CDEBUG(D_OTHER, "%s: new record "DOSTID":%lu/%lu: %d\n",
417 d->opd_obd->obd_name,
418 POSTID(&osi->osi_cookie.lgc_lgl.lgl_oi),
419 (unsigned long)osi->osi_cookie.lgc_lgl.lgl_ogen,
420 (unsigned long)osi->osi_cookie.lgc_index, rc);
421 spin_lock(&d->opd_syn_lock);
422 d->opd_syn_changes++;
423 spin_unlock(&d->opd_syn_lock);
425 /* return 0 always here, error case just cause no llog record */
429 int osp_sync_add(const struct lu_env *env, struct osp_object *o,
430 llog_op_type type, struct thandle *th,
431 const struct lu_attr *attr)
433 return osp_sync_add_rec(env, lu2osp_dev(o->opo_obj.do_lu.lo_dev),
434 lu_object_fid(&o->opo_obj.do_lu), type, 1,
438 int osp_sync_gap(const struct lu_env *env, struct osp_device *d,
439 struct lu_fid *fid, int lost, struct thandle *th)
441 return osp_sync_add_rec(env, d, fid, MDS_UNLINK64_REC, lost, th, NULL);
445 * it's quite obvious we can't maintain all the structures in the memory:
446 * while OST is down, MDS can be processing thousands and thousands of unlinks
447 * filling persistent llogs and in-core respresentation
449 * this doesn't scale at all. so we need basically the following:
450 * a) destroy/setattr append llog records
451 * b) once llog has grown to X records, we process first Y committed records
453 * once record R is found via llog_process(), it becomes committed after any
454 * subsequent commit callback (at the most)
458 * ptlrpc commit callback.
460 * The callback is called by PTLRPC when a RPC is reported committed by the
461 * target (OST). We register the callback for the every RPC applying a change
462 * from the llog. This way we know then the llog records can be cancelled.
463 * Notice the callback can be called when OSP is finishing. We can detect this
464 * checking that actual transno in the request is less or equal of known
465 * committed transno (see osp_sync_process_committed() for the details).
466 * XXX: this is pretty expensive and can be improved later using batching.
468 * \param[in] req request
470 static void osp_sync_request_commit_cb(struct ptlrpc_request *req)
472 struct osp_device *d = req->rq_cb_data;
473 struct osp_job_req_args *jra;
475 CDEBUG(D_HA, "commit req %p, transno "LPU64"\n", req, req->rq_transno);
477 if (unlikely(req->rq_transno == 0))
480 /* do not do any opd_dyn_rpc_* accounting here
481 * it's done in osp_sync_interpret sooner or later */
484 jra = ptlrpc_req_async_args(req);
485 LASSERT(jra->jra_magic == OSP_JOB_MAGIC);
486 LASSERT(list_empty(&jra->jra_committed_link));
488 ptlrpc_request_addref(req);
490 spin_lock(&d->opd_syn_lock);
491 list_add(&jra->jra_committed_link, &d->opd_syn_committed_there);
492 spin_unlock(&d->opd_syn_lock);
494 /* XXX: some batching wouldn't hurt */
495 wake_up(&d->opd_syn_waitq);
499 * RPC interpretation callback.
501 * The callback is called by ptlrpc when RPC is replied. Now we have to decide
503 * - put request on a special list to wait until it's committed by the target,
504 * if the request is successful
505 * - schedule llog record cancel if no target object is found
506 * - try later (essentially after reboot) in case of unexpected error
508 * \param[in] env LU environment provided by the caller
509 * \param[in] req request replied
510 * \param[in] aa callback data
511 * \param[in] rc result of RPC
515 static int osp_sync_interpret(const struct lu_env *env,
516 struct ptlrpc_request *req, void *aa, int rc)
518 struct osp_device *d = req->rq_cb_data;
519 struct osp_job_req_args *jra = aa;
521 if (jra->jra_magic != OSP_JOB_MAGIC) {
522 DEBUG_REQ(D_ERROR, req, "bad magic %u\n", jra->jra_magic);
527 CDEBUG(D_HA, "reply req %p/%d, rc %d, transno %u\n", req,
528 atomic_read(&req->rq_refcount),
529 rc, (unsigned) req->rq_transno);
530 LASSERT(rc || req->rq_transno);
534 * we tried to destroy object or update attributes,
535 * but object doesn't exist anymore - cancell llog record
537 LASSERT(req->rq_transno == 0);
538 LASSERT(list_empty(&jra->jra_committed_link));
540 ptlrpc_request_addref(req);
542 spin_lock(&d->opd_syn_lock);
543 list_add(&jra->jra_committed_link, &d->opd_syn_committed_there);
544 spin_unlock(&d->opd_syn_lock);
546 wake_up(&d->opd_syn_waitq);
548 struct obd_import *imp = req->rq_import;
550 * error happened, we'll try to repeat on next boot ?
552 LASSERTF(req->rq_transno == 0 ||
553 req->rq_import_generation < imp->imp_generation,
554 "transno "LPU64", rc %d, gen: req %d, imp %d\n",
555 req->rq_transno, rc, req->rq_import_generation,
556 imp->imp_generation);
557 if (req->rq_transno == 0) {
558 /* this is the last time we see the request
559 * if transno is not zero, then commit cb
560 * will be called at some point */
561 LASSERT(d->opd_syn_rpc_in_progress > 0);
562 spin_lock(&d->opd_syn_lock);
563 d->opd_syn_rpc_in_progress--;
564 spin_unlock(&d->opd_syn_lock);
567 wake_up(&d->opd_syn_waitq);
568 } else if (d->opd_pre != NULL &&
569 unlikely(d->opd_pre_status == -ENOSPC)) {
571 * if current status is -ENOSPC (lack of free space on OST)
572 * then we should poll OST immediately once object destroy
575 osp_statfs_need_now(d);
578 LASSERT(d->opd_syn_rpc_in_flight > 0);
579 spin_lock(&d->opd_syn_lock);
580 d->opd_syn_rpc_in_flight--;
581 list_del_init(&jra->jra_inflight_link);
582 spin_unlock(&d->opd_syn_lock);
583 if (unlikely(atomic_read(&d->opd_syn_barrier) > 0))
584 wake_up(&d->opd_syn_barrier_waitq);
585 CDEBUG(D_OTHER, "%s: %d in flight, %d in progress\n",
586 d->opd_obd->obd_name, d->opd_syn_rpc_in_flight,
587 d->opd_syn_rpc_in_progress);
589 osp_sync_check_for_work(d);
595 ** Add request to ptlrpc queue.
597 * This is just a tiny helper function to put the request on the sending list
599 * \param[in] d OSP device
600 * \param[in] req request
602 static void osp_sync_send_new_rpc(struct osp_device *d,
603 struct ptlrpc_request *req)
605 struct osp_job_req_args *jra;
607 LASSERT(d->opd_syn_rpc_in_flight <= d->opd_syn_max_rpc_in_flight);
609 jra = ptlrpc_req_async_args(req);
610 jra->jra_magic = OSP_JOB_MAGIC;
611 INIT_LIST_HEAD(&jra->jra_committed_link);
612 spin_lock(&d->opd_syn_lock);
613 list_add_tail(&jra->jra_inflight_link, &d->opd_syn_inflight_list);
614 spin_unlock(&d->opd_syn_lock);
616 ptlrpcd_add_req(req);
621 * Allocate and prepare RPC for a new change.
623 * The function allocates and initializes an RPC which will be sent soon to
624 * apply the change to the target OST. The request is initialized from the
625 * llog record passed. Notice only the fields common to all type of changes
628 * \param[in] d OSP device
629 * \param[in] llh llog handle where the record is stored
630 * \param[in] h llog record
631 * \param[in] op type of the change
632 * \param[in] format request format to be used
634 * \retval pointer new request on success
635 * \retval ERR_PTR(errno) on error
637 static struct ptlrpc_request *osp_sync_new_job(struct osp_device *d,
638 struct llog_handle *llh,
639 struct llog_rec_hdr *h,
641 const struct req_format *format)
643 struct ptlrpc_request *req;
644 struct ost_body *body;
645 struct obd_import *imp;
648 /* Prepare the request */
649 imp = d->opd_obd->u.cli.cl_import;
652 if (OBD_FAIL_CHECK(OBD_FAIL_OSP_CHECK_ENOMEM))
653 RETURN(ERR_PTR(-ENOMEM));
655 req = ptlrpc_request_alloc(imp, format);
657 RETURN(ERR_PTR(-ENOMEM));
659 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, op);
661 ptlrpc_req_finished(req);
666 * this is a trick: to save on memory allocations we put cookie
667 * into the request, but don't set corresponded flag in o_valid
668 * so that OST doesn't interpret this cookie. once the request
669 * is committed on OST we take cookie from the request and cancel
671 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
673 body->oa.o_lcookie.lgc_lgl = llh->lgh_id;
674 body->oa.o_lcookie.lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
675 body->oa.o_lcookie.lgc_index = h->lrh_index;
677 req->rq_interpret_reply = osp_sync_interpret;
678 req->rq_commit_cb = osp_sync_request_commit_cb;
681 ptlrpc_request_set_replen(req);
687 * Generate a request for setattr change.
689 * The function prepares a new RPC, initializes it with setattr specific
690 * bits and send the RPC.
692 * \param[in] d OSP device
693 * \param[in] llh llog handle where the record is stored
694 * \param[in] h llog record
696 * \retval 0 on success
697 * \retval 1 on invalid record
698 * \retval negative negated errno on error
700 static int osp_sync_new_setattr_job(struct osp_device *d,
701 struct llog_handle *llh,
702 struct llog_rec_hdr *h)
704 struct llog_setattr64_rec *rec = (struct llog_setattr64_rec *)h;
705 struct ptlrpc_request *req;
706 struct ost_body *body;
709 LASSERT(h->lrh_type == MDS_SETATTR64_REC);
711 if (OBD_FAIL_CHECK(OBD_FAIL_OSP_CHECK_INVALID_REC))
713 /* lsr_valid can only be 0 or have OBD_MD_{FLUID,FLGID} set,
714 * so no bits other than these should be set. */
715 if ((rec->lsr_valid & ~(OBD_MD_FLUID | OBD_MD_FLGID)) != 0) {
716 CERROR("%s: invalid setattr record, lsr_valid:"LPU64"\n",
717 d->opd_obd->obd_name, rec->lsr_valid);
718 /* return 1 on invalid record */
722 req = osp_sync_new_job(d, llh, h, OST_SETATTR, &RQF_OST_SETATTR);
724 RETURN(PTR_ERR(req));
726 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
728 body->oa.o_oi = rec->lsr_oi;
729 body->oa.o_uid = rec->lsr_uid;
730 body->oa.o_gid = rec->lsr_gid;
731 body->oa.o_valid = OBD_MD_FLGROUP | OBD_MD_FLID;
732 /* old setattr record (prior 2.6.0) doesn't have 'valid' stored,
733 * we assume that both UID and GID are valid in that case. */
734 if (rec->lsr_valid == 0)
735 body->oa.o_valid |= (OBD_MD_FLUID | OBD_MD_FLGID);
737 body->oa.o_valid |= rec->lsr_valid;
739 osp_sync_send_new_rpc(d, req);
744 * Generate a request for unlink change.
746 * The function prepares a new RPC, initializes it with unlink(destroy)
747 * specific bits and sends the RPC. The function is used to handle
748 * llog_unlink_rec which were used in the older versions of Lustre.
749 * Current version uses llog_unlink_rec64.
751 * \param[in] d OSP device
752 * \param[in] llh llog handle where the record is stored
753 * \param[in] h llog record
755 * \retval 0 on success
756 * \retval negative negated errno on error
758 static int osp_sync_new_unlink_job(struct osp_device *d,
759 struct llog_handle *llh,
760 struct llog_rec_hdr *h)
762 struct llog_unlink_rec *rec = (struct llog_unlink_rec *)h;
763 struct ptlrpc_request *req;
764 struct ost_body *body;
767 LASSERT(h->lrh_type == MDS_UNLINK_REC);
769 req = osp_sync_new_job(d, llh, h, OST_DESTROY, &RQF_OST_DESTROY);
771 RETURN(PTR_ERR(req));
773 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
775 ostid_set_seq(&body->oa.o_oi, rec->lur_oseq);
776 ostid_set_id(&body->oa.o_oi, rec->lur_oid);
777 body->oa.o_misc = rec->lur_count;
778 body->oa.o_valid = OBD_MD_FLGROUP | OBD_MD_FLID;
780 body->oa.o_valid |= OBD_MD_FLOBJCOUNT;
782 osp_sync_send_new_rpc(d, req);
787 * Generate a request for unlink change.
789 * The function prepares a new RPC, initializes it with unlink(destroy)
790 * specific bits and sends the RPC. Depending on the target (MDT or OST)
791 * two different protocols are used. For MDT we use OUT (basically OSD API
792 * updates transferred via a network). For OST we still use the old
793 * protocol (OBD?), originally for compatibility. Later we can start to
794 * use OUT for OST as well, this will allow batching and better code
797 * \param[in] d OSP device
798 * \param[in] llh llog handle where the record is stored
799 * \param[in] h llog record
801 * \retval 0 on success
802 * \retval negative negated errno on error
804 static int osp_sync_new_unlink64_job(struct osp_device *d,
805 struct llog_handle *llh,
806 struct llog_rec_hdr *h)
808 struct llog_unlink64_rec *rec = (struct llog_unlink64_rec *)h;
809 struct ptlrpc_request *req = NULL;
810 struct ost_body *body;
814 LASSERT(h->lrh_type == MDS_UNLINK64_REC);
815 req = osp_sync_new_job(d, llh, h, OST_DESTROY,
818 RETURN(PTR_ERR(req));
820 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
823 rc = fid_to_ostid(&rec->lur_fid, &body->oa.o_oi);
826 body->oa.o_misc = rec->lur_count;
827 body->oa.o_valid = OBD_MD_FLGROUP | OBD_MD_FLID |
829 osp_sync_send_new_rpc(d, req);
834 * Process llog records.
836 * This function is called to process the llog records committed locally.
837 * In the recovery model used by OSP we can apply a change to a remote
838 * target once corresponding transaction (like posix unlink) is committed
839 * locally so can't revert.
840 * Depending on the llog record type, a given handler is called that is
841 * responsible for preparing and sending the RPC to apply the change.
842 * Special record type LLOG_GEN_REC marking a reboot is cancelled right away.
844 * \param[in] env LU environment provided by the caller
845 * \param[in] d OSP device
846 * \param[in] llh llog handle where the record is stored
847 * \param[in] rec llog record
849 static void osp_sync_process_record(const struct lu_env *env,
850 struct osp_device *d,
851 struct llog_handle *llh,
852 struct llog_rec_hdr *rec)
854 struct llog_handle *cathandle = llh->u.phd.phd_cat_handle;
855 struct llog_cookie cookie;
860 cookie.lgc_lgl = llh->lgh_id;
861 cookie.lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
862 cookie.lgc_index = rec->lrh_index;
864 if (unlikely(rec->lrh_type == LLOG_GEN_REC)) {
865 struct llog_gen_rec *gen = (struct llog_gen_rec *)rec;
867 /* we're waiting for the record generated by this instance */
868 LASSERT(d->opd_syn_prev_done == 0);
869 if (!memcmp(&d->opd_syn_generation, &gen->lgr_gen,
870 sizeof(gen->lgr_gen))) {
871 CDEBUG(D_HA, "processed all old entries\n");
872 d->opd_syn_prev_done = 1;
875 /* cancel any generation record */
876 rc = llog_cat_cancel_records(env, cathandle, 1, &cookie);
882 * now we prepare and fill requests to OST, put them on the queue
883 * and fire after next commit callback
886 /* notice we increment counters before sending RPC, to be consistent
887 * in RPC interpret callback which may happen very quickly */
888 spin_lock(&d->opd_syn_lock);
889 d->opd_syn_rpc_in_flight++;
890 d->opd_syn_rpc_in_progress++;
891 spin_unlock(&d->opd_syn_lock);
893 switch (rec->lrh_type) {
894 /* case MDS_UNLINK_REC is kept for compatibility */
896 rc = osp_sync_new_unlink_job(d, llh, rec);
898 case MDS_UNLINK64_REC:
899 rc = osp_sync_new_unlink64_job(d, llh, rec);
901 case MDS_SETATTR64_REC:
902 rc = osp_sync_new_setattr_job(d, llh, rec);
905 CERROR("%s: unknown record type: %x\n", d->opd_obd->obd_name,
907 /* treat "unknown record type" as "invalid" */
912 spin_lock(&d->opd_syn_lock);
914 /* For all kinds of records, not matter successful or not,
915 * we should decrease changes and bump last_processed_id.
917 if (d->opd_syn_prev_done) {
918 LASSERT(d->opd_syn_changes > 0);
919 LASSERT(rec->lrh_id <= d->opd_syn_last_committed_id);
920 /* NOTE: it's possible to meet same id if
921 * OST stores few stripes of same file
923 if (rec->lrh_id > d->opd_syn_last_processed_id) {
924 d->opd_syn_last_processed_id = rec->lrh_id;
925 wake_up(&d->opd_syn_barrier_waitq);
927 d->opd_syn_changes--;
930 d->opd_syn_rpc_in_flight--;
931 d->opd_syn_rpc_in_progress--;
933 CDEBUG(D_OTHER, "%s: %d in flight, %d in progress\n",
934 d->opd_obd->obd_name, d->opd_syn_rpc_in_flight,
935 d->opd_syn_rpc_in_progress);
937 spin_unlock(&d->opd_syn_lock);
939 /* Delete the invalid record */
941 rc = llog_cat_cancel_records(env, cathandle, 1, &cookie);
943 CERROR("%s: can't delete invalid record: "
944 "fid = "DFID", rec_id = %u, rc = %d\n",
945 d->opd_obd->obd_name,
946 PFID(lu_object_fid(&cathandle->lgh_obj->do_lu)),
950 CDEBUG(D_HA, "found record %x, %d, idx %u, id %u\n",
951 rec->lrh_type, rec->lrh_len, rec->lrh_index, rec->lrh_id);
957 * Cancel llog records for the committed changes.
959 * The function walks through the list of the committed RPCs and cancels
960 * corresponding llog records. see osp_sync_request_commit_cb() for the
963 * \param[in] env LU environment provided by the caller
964 * \param[in] d OSP device
966 static void osp_sync_process_committed(const struct lu_env *env,
967 struct osp_device *d)
969 struct obd_device *obd = d->opd_obd;
970 struct obd_import *imp = obd->u.cli.cl_import;
971 struct ost_body *body;
972 struct ptlrpc_request *req;
973 struct llog_ctxt *ctxt;
974 struct llog_handle *llh;
975 struct list_head list;
980 if (list_empty(&d->opd_syn_committed_there))
984 * if current status is -ENOSPC (lack of free space on OST)
985 * then we should poll OST immediately once object destroy
987 * notice: we do this upon commit as well because some backends
988 * (like DMU) do not release space right away.
990 if (d->opd_pre != NULL && unlikely(d->opd_pre_status == -ENOSPC))
991 osp_statfs_need_now(d);
994 * now cancel them all
995 * XXX: can we improve this using some batching?
996 * with batch RPC that'll happen automatically?
997 * XXX: can we store ctxt in lod_device and save few cycles ?
999 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
1002 llh = ctxt->loc_handle;
1005 INIT_LIST_HEAD(&list);
1006 spin_lock(&d->opd_syn_lock);
1007 list_splice(&d->opd_syn_committed_there, &list);
1008 INIT_LIST_HEAD(&d->opd_syn_committed_there);
1009 spin_unlock(&d->opd_syn_lock);
1011 while (!list_empty(&list)) {
1012 struct osp_job_req_args *jra;
1014 jra = list_entry(list.next, struct osp_job_req_args,
1015 jra_committed_link);
1016 LASSERT(jra->jra_magic == OSP_JOB_MAGIC);
1017 list_del_init(&jra->jra_committed_link);
1019 req = container_of((void *)jra, struct ptlrpc_request,
1021 body = req_capsule_client_get(&req->rq_pill,
1024 /* import can be closing, thus all commit cb's are
1025 * called we can check committness directly */
1026 if (req->rq_import_generation == imp->imp_generation) {
1027 rc = llog_cat_cancel_records(env, llh, 1,
1028 &body->oa.o_lcookie);
1030 CERROR("%s: can't cancel record: %d\n",
1033 DEBUG_REQ(D_OTHER, req, "imp_committed = "LPU64,
1034 imp->imp_peer_committed_transno);
1036 ptlrpc_req_finished(req);
1040 llog_ctxt_put(ctxt);
1042 LASSERT(d->opd_syn_rpc_in_progress >= done);
1043 spin_lock(&d->opd_syn_lock);
1044 d->opd_syn_rpc_in_progress -= done;
1045 spin_unlock(&d->opd_syn_lock);
1046 CDEBUG(D_OTHER, "%s: %d in flight, %d in progress\n",
1047 d->opd_obd->obd_name, d->opd_syn_rpc_in_flight,
1048 d->opd_syn_rpc_in_progress);
1050 osp_sync_check_for_work(d);
1052 /* wake up the thread if requested to stop:
1053 * it might be waiting for in-progress to complete */
1054 if (unlikely(osp_sync_running(d) == 0))
1055 wake_up(&d->opd_syn_waitq);
1061 * The core of the syncing mechanism.
1063 * This is a callback called by the llog processing function. Essentially it
1064 * suspends llog processing until there is a record to process (it's supposed
1065 * to be committed locally). The function handles RPCs committed by the target
1066 * and cancels corresponding llog records.
1068 * \param[in] env LU environment provided by the caller
1069 * \param[in] llh llog handle we're processing
1070 * \param[in] rec current llog record
1071 * \param[in] data callback data containing a pointer to the device
1073 * \retval 0 to ask the caller (llog_process()) to continue
1074 * \retval LLOG_PROC_BREAK to ask the caller to break
1076 static int osp_sync_process_queues(const struct lu_env *env,
1077 struct llog_handle *llh,
1078 struct llog_rec_hdr *rec,
1081 struct osp_device *d = data;
1084 struct l_wait_info lwi = { 0 };
1086 if (!osp_sync_running(d)) {
1087 CDEBUG(D_HA, "stop llog processing\n");
1088 return LLOG_PROC_BREAK;
1091 /* process requests committed by OST */
1092 osp_sync_process_committed(env, d);
1094 /* if we there are changes to be processed and we have
1095 * resources for this ... do now */
1096 if (osp_sync_can_process_new(d, rec)) {
1098 /* ask llog for another record */
1099 CDEBUG(D_HA, "%lu changes, %u in progress,"
1102 d->opd_syn_rpc_in_progress,
1103 d->opd_syn_rpc_in_flight);
1106 osp_sync_process_record(env, d, llh, rec);
1111 if (d->opd_syn_last_processed_id == d->opd_syn_last_used_id)
1112 osp_sync_remove_from_tracker(d);
1114 l_wait_event(d->opd_syn_waitq,
1115 !osp_sync_running(d) ||
1116 osp_sync_can_process_new(d, rec) ||
1117 !list_empty(&d->opd_syn_committed_there),
1125 * This thread runs llog_cat_process() scanner calling our callback
1126 * to process llog records. in the callback we implement tricky
1127 * state machine as we don't want to start scanning of the llog again
1128 * and again, also we don't want to process too many records and send
1129 * too many RPCs a time. so, depending on current load (num of changes
1130 * being synced to OST) the callback can suspend awaiting for some
1131 * new conditions, like syncs completed.
1133 * In order to process llog records left by previous boots and to allow
1134 * llog_process_thread() to find something (otherwise it'd just exit
1135 * immediately) we add a special GENERATATION record on each boot.
1137 * \param[in] _arg a pointer to thread's arguments
1139 * \retval 0 on success
1140 * \retval negative negated errno on error
1142 static int osp_sync_thread(void *_arg)
1144 struct osp_device *d = _arg;
1145 struct ptlrpc_thread *thread = &d->opd_syn_thread;
1146 struct l_wait_info lwi = { 0 };
1147 struct llog_ctxt *ctxt;
1148 struct obd_device *obd = d->opd_obd;
1149 struct llog_handle *llh;
1155 rc = lu_env_init(&env, LCT_LOCAL);
1157 CERROR("%s: can't initialize env: rc = %d\n",
1162 spin_lock(&d->opd_syn_lock);
1163 thread->t_flags = SVC_RUNNING;
1164 spin_unlock(&d->opd_syn_lock);
1165 wake_up(&thread->t_ctl_waitq);
1167 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
1169 CERROR("can't get appropriate context\n");
1170 GOTO(out, rc = -EINVAL);
1173 llh = ctxt->loc_handle;
1175 CERROR("can't get llh\n");
1176 llog_ctxt_put(ctxt);
1177 GOTO(out, rc = -EINVAL);
1180 rc = llog_cat_process(&env, llh, osp_sync_process_queues, d, 0, 0);
1181 LASSERTF(rc == 0 || rc == LLOG_PROC_BREAK,
1182 "%lu changes, %u in progress, %u in flight: %d\n",
1183 d->opd_syn_changes, d->opd_syn_rpc_in_progress,
1184 d->opd_syn_rpc_in_flight, rc);
1186 /* we don't expect llog_process_thread() to exit till umount */
1187 LASSERTF(thread->t_flags != SVC_RUNNING,
1188 "%lu changes, %u in progress, %u in flight\n",
1189 d->opd_syn_changes, d->opd_syn_rpc_in_progress,
1190 d->opd_syn_rpc_in_flight);
1192 /* wait till all the requests are completed */
1194 while (d->opd_syn_rpc_in_progress > 0) {
1195 osp_sync_process_committed(&env, d);
1197 lwi = LWI_TIMEOUT(cfs_time_seconds(5), NULL, NULL);
1198 rc = l_wait_event(d->opd_syn_waitq,
1199 d->opd_syn_rpc_in_progress == 0,
1201 if (rc == -ETIMEDOUT)
1203 LASSERTF(count < 10, "%s: %d %d %sempty\n",
1204 d->opd_obd->obd_name, d->opd_syn_rpc_in_progress,
1205 d->opd_syn_rpc_in_flight,
1206 list_empty(&d->opd_syn_committed_there) ? "" : "!");
1210 llog_cat_close(&env, llh);
1211 rc = llog_cleanup(&env, ctxt);
1213 CERROR("can't cleanup llog: %d\n", rc);
1215 LASSERTF(d->opd_syn_rpc_in_progress == 0,
1216 "%s: %d %d %sempty\n",
1217 d->opd_obd->obd_name, d->opd_syn_rpc_in_progress,
1218 d->opd_syn_rpc_in_flight,
1219 list_empty(&d->opd_syn_committed_there) ? "" : "!");
1221 thread->t_flags = SVC_STOPPED;
1223 wake_up(&thread->t_ctl_waitq);
1233 * Initializes the llog. Specific llog to be used depends on the type of the
1234 * target OSP represents (OST or MDT). The function adds appends a new llog
1235 * record to mark the place where the records associated with this boot
1238 * \param[in] env LU environment provided by the caller
1239 * \param[in] d OSP device
1241 * \retval 0 on success
1242 * \retval negative negated errno on error
1244 static int osp_sync_llog_init(const struct lu_env *env, struct osp_device *d)
1246 struct osp_thread_info *osi = osp_env_info(env);
1247 struct lu_fid *fid = &osi->osi_fid;
1248 struct llog_handle *lgh = NULL;
1249 struct obd_device *obd = d->opd_obd;
1250 struct llog_ctxt *ctxt;
1258 * open llog corresponding to our OST
1260 OBD_SET_CTXT_MAGIC(&obd->obd_lvfs_ctxt);
1261 obd->obd_lvfs_ctxt.dt = d->opd_storage;
1263 lu_local_obj_fid(fid, LLOG_CATALOGS_OID);
1265 rc = llog_osd_get_cat_list(env, d->opd_storage, d->opd_index, 1,
1266 &osi->osi_cid, fid);
1268 if (rc != -EFAULT) {
1269 CERROR("%s: can't get id from catalogs: rc = %d\n",
1274 /* After sparse OST indices is supported, the CATALOG file
1275 * may become a sparse file that results in failure on
1276 * reading. Skip this error as the llog will be created
1278 memset(&osi->osi_cid, 0, sizeof(osi->osi_cid));
1282 CDEBUG(D_INFO, "%s: Init llog for %d - catid "DOSTID":%x\n",
1283 obd->obd_name, d->opd_index,
1284 POSTID(&osi->osi_cid.lci_logid.lgl_oi),
1285 osi->osi_cid.lci_logid.lgl_ogen);
1287 rc = llog_setup(env, obd, &obd->obd_olg, LLOG_MDS_OST_ORIG_CTXT,
1288 d->opd_storage->dd_lu_dev.ld_obd,
1289 &osp_mds_ost_orig_logops);
1293 ctxt = llog_get_context(obd, LLOG_MDS_OST_ORIG_CTXT);
1296 if (likely(logid_id(&osi->osi_cid.lci_logid) != 0)) {
1297 rc = llog_open(env, ctxt, &lgh, &osi->osi_cid.lci_logid, NULL,
1299 /* re-create llog if it is missing */
1301 logid_set_id(&osi->osi_cid.lci_logid, 0);
1303 GOTO(out_cleanup, rc);
1306 if (unlikely(logid_id(&osi->osi_cid.lci_logid) == 0)) {
1307 rc = llog_open_create(env, ctxt, &lgh, NULL, NULL);
1309 GOTO(out_cleanup, rc);
1310 osi->osi_cid.lci_logid = lgh->lgh_id;
1313 LASSERT(lgh != NULL);
1314 ctxt->loc_handle = lgh;
1316 rc = llog_init_handle(env, lgh, LLOG_F_IS_CAT, NULL);
1318 GOTO(out_close, rc);
1320 rc = llog_osd_put_cat_list(env, d->opd_storage, d->opd_index, 1,
1321 &osi->osi_cid, fid);
1323 GOTO(out_close, rc);
1326 * put a mark in the llog till which we'll be processing
1327 * old records restless
1329 d->opd_syn_generation.mnt_cnt = cfs_time_current();
1330 d->opd_syn_generation.conn_cnt = cfs_time_current();
1332 osi->osi_hdr.lrh_type = LLOG_GEN_REC;
1333 osi->osi_hdr.lrh_len = sizeof(osi->osi_gen);
1335 memcpy(&osi->osi_gen.lgr_gen, &d->opd_syn_generation,
1336 sizeof(osi->osi_gen.lgr_gen));
1338 rc = llog_cat_add(env, lgh, &osi->osi_gen.lgr_hdr, &osi->osi_cookie);
1340 GOTO(out_close, rc);
1341 llog_ctxt_put(ctxt);
1344 llog_cat_close(env, lgh);
1346 llog_cleanup(env, ctxt);
1351 * Cleanup llog used for syncing.
1353 * Closes and cleanups the llog. The function is called when the device is
1356 * \param[in] env LU environment provided by the caller
1357 * \param[in] d OSP device
1359 static void osp_sync_llog_fini(const struct lu_env *env, struct osp_device *d)
1361 struct llog_ctxt *ctxt;
1363 ctxt = llog_get_context(d->opd_obd, LLOG_MDS_OST_ORIG_CTXT);
1365 llog_cat_close(env, ctxt->loc_handle);
1366 llog_cleanup(env, ctxt);
1370 * Initialization of the sync component of OSP.
1372 * Initializes the llog and starts a new thread to handle the changes to
1373 * the remote target (OST or MDT).
1375 * \param[in] env LU environment provided by the caller
1376 * \param[in] d OSP device
1378 * \retval 0 on success
1379 * \retval negative negated errno on error
1381 int osp_sync_init(const struct lu_env *env, struct osp_device *d)
1383 struct l_wait_info lwi = { 0 };
1384 struct task_struct *task;
1389 rc = osp_sync_id_traction_init(d);
1394 * initialize llog storing changes
1396 rc = osp_sync_llog_init(env, d);
1398 CERROR("%s: can't initialize llog: rc = %d\n",
1399 d->opd_obd->obd_name, rc);
1404 * Start synchronization thread
1406 d->opd_syn_max_rpc_in_flight = OSP_MAX_IN_FLIGHT;
1407 d->opd_syn_max_rpc_in_progress = OSP_MAX_IN_PROGRESS;
1408 spin_lock_init(&d->opd_syn_lock);
1409 init_waitqueue_head(&d->opd_syn_waitq);
1410 init_waitqueue_head(&d->opd_syn_barrier_waitq);
1411 init_waitqueue_head(&d->opd_syn_thread.t_ctl_waitq);
1412 INIT_LIST_HEAD(&d->opd_syn_inflight_list);
1413 INIT_LIST_HEAD(&d->opd_syn_committed_there);
1415 task = kthread_run(osp_sync_thread, d, "osp-syn-%u-%u",
1416 d->opd_index, d->opd_group);
1419 CERROR("%s: cannot start sync thread: rc = %d\n",
1420 d->opd_obd->obd_name, rc);
1424 l_wait_event(d->opd_syn_thread.t_ctl_waitq,
1425 osp_sync_running(d) || osp_sync_stopped(d), &lwi);
1429 osp_sync_llog_fini(env, d);
1431 osp_sync_id_traction_fini(d);
1436 * Stop the syncing thread.
1438 * Asks the syncing thread to stop and wait until it's stopped.
1440 * \param[in] d OSP device
1444 int osp_sync_fini(struct osp_device *d)
1446 struct ptlrpc_thread *thread = &d->opd_syn_thread;
1450 thread->t_flags = SVC_STOPPING;
1451 wake_up(&d->opd_syn_waitq);
1452 wait_event(thread->t_ctl_waitq, thread->t_flags & SVC_STOPPED);
1455 * unregister transaction callbacks only when sync thread
1456 * has finished operations with llog
1458 osp_sync_id_traction_fini(d);
1463 static DEFINE_MUTEX(osp_id_tracker_sem);
1464 static struct list_head osp_id_tracker_list =
1465 LIST_HEAD_INIT(osp_id_tracker_list);
1468 * OSD commit callback.
1470 * The function is used as a local OSD commit callback to track the highest
1471 * committed llog record id. see osp_sync_id_traction_init() for the details.
1473 * \param[in] th local transaction handle committed
1474 * \param[in] cookie commit callback data (our private structure)
1476 static void osp_sync_tracker_commit_cb(struct thandle *th, void *cookie)
1478 struct osp_id_tracker *tr = cookie;
1479 struct osp_device *d;
1480 struct osp_txn_info *txn;
1484 txn = osp_txn_info(&th->th_ctx);
1485 if (txn == NULL || txn->oti_current_id < tr->otr_committed_id)
1488 spin_lock(&tr->otr_lock);
1489 if (likely(txn->oti_current_id > tr->otr_committed_id)) {
1490 CDEBUG(D_OTHER, "committed: %u -> %u\n",
1491 tr->otr_committed_id, txn->oti_current_id);
1492 tr->otr_committed_id = txn->oti_current_id;
1494 list_for_each_entry(d, &tr->otr_wakeup_list,
1496 d->opd_syn_last_committed_id = tr->otr_committed_id;
1497 wake_up(&d->opd_syn_waitq);
1500 spin_unlock(&tr->otr_lock);
1504 * Initialize commit tracking mechanism.
1506 * Some setups may have thousands of OSTs and each will be represented by OSP.
1507 * Meaning order of magnitute many more changes to apply every second. In order
1508 * to keep the number of commit callbacks low this mechanism was introduced.
1509 * The mechanism is very similar to transno used by MDT service: it's an single
1510 * ID stream which can be assigned by any OSP to its llog records. The tricky
1511 * part is that ID is stored in per-transaction data and re-used by all the OSPs
1512 * involved in that transaction. Then all these OSPs are woken up utilizing a single OSD commit callback.
1514 * The function initializes the data used by the tracker described above.
1515 * A singler tracker per OSD device is created.
1517 * \param[in] d OSP device
1519 * \retval 0 on success
1520 * \retval negative negated errno on error
1522 static int osp_sync_id_traction_init(struct osp_device *d)
1524 struct osp_id_tracker *tr, *found = NULL;
1528 LASSERT(d->opd_storage);
1529 LASSERT(d->opd_syn_tracker == NULL);
1530 INIT_LIST_HEAD(&d->opd_syn_ontrack);
1532 mutex_lock(&osp_id_tracker_sem);
1533 list_for_each_entry(tr, &osp_id_tracker_list, otr_list) {
1534 if (tr->otr_dev == d->opd_storage) {
1535 LASSERT(atomic_read(&tr->otr_refcount));
1536 atomic_inc(&tr->otr_refcount);
1537 d->opd_syn_tracker = tr;
1543 if (found == NULL) {
1547 d->opd_syn_tracker = tr;
1548 spin_lock_init(&tr->otr_lock);
1549 tr->otr_dev = d->opd_storage;
1550 tr->otr_next_id = 1;
1551 tr->otr_committed_id = 0;
1552 atomic_set(&tr->otr_refcount, 1);
1553 INIT_LIST_HEAD(&tr->otr_wakeup_list);
1554 list_add(&tr->otr_list, &osp_id_tracker_list);
1555 tr->otr_tx_cb.dtc_txn_commit =
1556 osp_sync_tracker_commit_cb;
1557 tr->otr_tx_cb.dtc_cookie = tr;
1558 tr->otr_tx_cb.dtc_tag = LCT_MD_THREAD;
1559 dt_txn_callback_add(d->opd_storage, &tr->otr_tx_cb);
1563 mutex_unlock(&osp_id_tracker_sem);
1569 * Release commit tracker.
1571 * Decrease a refcounter on the tracker used by the given OSP device \a d.
1572 * If no more users left, then the tracker is released.
1574 * \param[in] d OSP device
1576 static void osp_sync_id_traction_fini(struct osp_device *d)
1578 struct osp_id_tracker *tr;
1583 tr = d->opd_syn_tracker;
1589 osp_sync_remove_from_tracker(d);
1591 mutex_lock(&osp_id_tracker_sem);
1592 if (atomic_dec_and_test(&tr->otr_refcount)) {
1593 dt_txn_callback_del(d->opd_storage, &tr->otr_tx_cb);
1594 LASSERT(list_empty(&tr->otr_wakeup_list));
1595 list_del(&tr->otr_list);
1597 d->opd_syn_tracker = NULL;
1599 mutex_unlock(&osp_id_tracker_sem);
1605 * Generate a new ID on a tracker.
1607 * Generates a new ID using the tracker associated with the given OSP device
1608 * \a d, if the given ID \a id is non-zero. Unconditially adds OSP device to
1609 * the wakeup list, so OSP won't miss when a transaction using the ID is
1610 * committed. Notice ID is 32bit, but llog doesn't support >2^32 records anyway.
1612 * \param[in] d OSP device
1613 * \param[in] id 0 or ID generated previously
1615 * \retval ID the caller should use
1617 static __u32 osp_sync_id_get(struct osp_device *d, __u32 id)
1619 struct osp_id_tracker *tr;
1621 tr = d->opd_syn_tracker;
1624 /* XXX: we can improve this introducing per-cpu preallocated ids? */
1625 spin_lock(&tr->otr_lock);
1626 if (unlikely(tr->otr_next_id <= d->opd_syn_last_used_id)) {
1627 spin_unlock(&tr->otr_lock);
1628 CERROR("%s: next %u, last synced %lu\n",
1629 d->opd_obd->obd_name, tr->otr_next_id,
1630 d->opd_syn_last_used_id);
1635 id = tr->otr_next_id++;
1636 if (id > d->opd_syn_last_used_id)
1637 d->opd_syn_last_used_id = id;
1638 if (list_empty(&d->opd_syn_ontrack))
1639 list_add(&d->opd_syn_ontrack, &tr->otr_wakeup_list);
1640 spin_unlock(&tr->otr_lock);
1641 CDEBUG(D_OTHER, "new id %u\n", (unsigned) id);
1647 * Stop to propagate commit status to OSP.
1649 * If the OSP does not have any llog records she's waiting to commit, then
1650 * it is possible to unsubscribe from wakeups from the tracking using this
1653 * \param[in] d OSP device not willing to wakeup
1655 static void osp_sync_remove_from_tracker(struct osp_device *d)
1657 struct osp_id_tracker *tr;
1659 tr = d->opd_syn_tracker;
1662 if (list_empty(&d->opd_syn_ontrack))
1665 spin_lock(&tr->otr_lock);
1666 list_del_init(&d->opd_syn_ontrack);
1667 spin_unlock(&tr->otr_lock);