4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2014, 2017, Intel Corporation.
26 * lustre/osp/osp_trans.c
29 * 1. OSP (Object Storage Proxy) transaction methods
31 * Implement OSP layer transaction related interfaces for the dt_device API
32 * dt_device_operations.
35 * 2. Handle asynchronous idempotent operations
37 * The OSP uses OUT (Object Unified Target) RPC to talk with other server
38 * (MDT or OST) for kinds of operations, such as create, unlink, insert,
39 * delete, lookup, set_(x)attr, get_(x)attr, and etc. To reduce the number
40 * of RPCs, we allow multiple operations to be packaged together in single
43 * For the asynchronous idempotent operations, such as get_(x)attr, related
44 * RPCs will be inserted into an osp_device based shared asynchronous request
45 * queue - osp_device::opd_async_requests. When the queue is full, all the
46 * requests in the queue will be packaged into a single OUT RPC and given to
47 * the ptlrpcd daemon (for sending), then the queue is purged and other new
48 * requests can be inserted into it.
50 * When the asynchronous idempotent operation inserts the request into the
51 * shared queue, it will register an interpreter. When the packaged OUT RPC
52 * is replied (or failed to be sent out), all the registered interpreters
53 * will be called one by one to handle each own result.
56 * There are three kinds of transactions
58 * 1. Local transaction, all of updates of the transaction are in the local MDT.
59 * 2. Remote transaction, all of updates of the transaction are in one remote
60 * MDT, which only happens in LFSCK now.
61 * 3. Distribute transaction, updates for the transaction are in mulitple MDTs.
63 * Author: Di Wang <di.wang@intel.com>
64 * Author: Fan, Yong <fan.yong@intel.com>
67 #define DEBUG_SUBSYSTEM S_MDS
69 #include <lustre_net.h>
70 #include "osp_internal.h"
73 * The argument for the interpreter callback of osp request.
75 struct osp_update_args {
76 struct osp_update_request *oaua_update;
78 wait_queue_head_t *oaua_waitq;
79 bool oaua_flow_control;
80 const struct lu_env *oaua_update_env;
84 * Call back for each update request.
86 struct osp_update_callback {
87 /* list in the osp_update_request::our_cb_items */
88 struct list_head ouc_list;
90 /* The target of the async update request. */
91 struct osp_object *ouc_obj;
93 /* The data used by or_interpreter. */
96 /* The interpreter function called after the async request handled. */
97 osp_update_interpreter_t ouc_interpreter;
101 * Allocate new update request
103 * Allocate new update request and insert it to the req_update_list.
105 * \param [in] our osp_udate_request where to create a new
108 * \retval 0 if creation succeeds.
109 * \retval negative errno if creation fails.
111 int osp_object_update_request_create(struct osp_update_request *our,
114 struct osp_update_request_sub *ours;
115 struct object_update_request *ourq;
121 /* The object update request will be added to an SG list for
122 * bulk transfer. Some IB HW cannot handle partial pages in SG
123 * lists (since they create gaps in memory regions) so we
124 * round the size up to the next multiple of PAGE_SIZE. See
127 size = round_up(size, PAGE_SIZE);
128 OBD_ALLOC_LARGE(ourq, size);
134 ourq->ourq_magic = UPDATE_REQUEST_MAGIC;
135 ourq->ourq_count = 0;
136 ours->ours_req = ourq;
137 ours->ours_req_size = size;
138 INIT_LIST_HEAD(&ours->ours_list);
139 list_add_tail(&ours->ours_list, &our->our_req_list);
146 * Get current update request
148 * Get current object update request from our_req_list in
149 * osp_update_request, because we always insert the new update
150 * request in the last position, so the last update request
151 * in the list will be the current update req.
153 * \param[in] our osp update request where to get the
154 * current object update.
156 * \retval the current object update.
158 struct osp_update_request_sub *
159 osp_current_object_update_request(struct osp_update_request *our)
161 if (list_empty(&our->our_req_list))
164 return list_entry(our->our_req_list.prev, struct osp_update_request_sub,
169 * Allocate and initialize osp_update_request
171 * osp_update_request is being used to track updates being executed on
172 * this dt_device(OSD or OSP). The update buffer will be 4k initially,
173 * and increased if needed.
175 * \param [in] dt dt device
177 * \retval osp_update_request being allocated if succeed
178 * \retval ERR_PTR(errno) if failed
180 struct osp_update_request *osp_update_request_create(struct dt_device *dt)
182 struct osp_update_request *our;
187 return ERR_PTR(-ENOMEM);
189 INIT_LIST_HEAD(&our->our_req_list);
190 INIT_LIST_HEAD(&our->our_cb_items);
191 INIT_LIST_HEAD(&our->our_list);
192 INIT_LIST_HEAD(&our->our_invalidate_cb_list);
193 spin_lock_init(&our->our_list_lock);
195 rc = osp_object_update_request_create(our, PAGE_SIZE);
203 void osp_update_request_destroy(const struct lu_env *env,
204 struct osp_update_request *our)
206 struct osp_update_request_sub *ours;
207 struct osp_update_request_sub *tmp;
212 list_for_each_entry_safe(ours, tmp, &our->our_req_list, ours_list) {
213 list_del(&ours->ours_list);
214 if (ours->ours_req != NULL)
215 OBD_FREE_LARGE(ours->ours_req, ours->ours_req_size);
219 if (!list_empty(&our->our_invalidate_cb_list)) {
221 struct osp_object *obj;
222 struct osp_object *next;
225 lu_env_init(&lenv, LCT_MD_THREAD | LCT_DT_THREAD);
229 list_for_each_entry_safe(obj, next,
230 &our->our_invalidate_cb_list,
231 opo_invalidate_cb_list) {
232 spin_lock(&obj->opo_lock);
233 list_del_init(&obj->opo_invalidate_cb_list);
234 spin_unlock(&obj->opo_lock);
236 dt_object_put(env, &obj->opo_obj);
247 object_update_request_dump(const struct object_update_request *ourq,
251 size_t total_size = 0;
253 for (i = 0; i < ourq->ourq_count; i++) {
254 struct object_update *update;
257 update = object_update_request_get(ourq, i, &size);
258 LASSERT(update != NULL);
259 CDEBUG(mask, "i = %u fid = "DFID" op = %s "
260 "params = %d batchid = %llu size = %zu repsize %u\n",
261 i, PFID(&update->ou_fid),
262 update_op_str(update->ou_type),
263 update->ou_params_count,
264 update->ou_batchid, size,
265 (unsigned)update->ou_result_size);
270 CDEBUG(mask, "updates = %p magic = %x count = %d size = %zu\n", ourq,
271 ourq->ourq_magic, ourq->ourq_count, total_size);
275 * Prepare inline update request
277 * Prepare OUT update ptlrpc inline request, and the request usually includes
278 * one update buffer, which does not need bulk transfer.
280 * \param[in] env execution environment
281 * \param[in] req ptlrpc request
282 * \param[in] ours sub osp_update_request to be packed
284 * \retval 0 if packing succeeds
285 * \retval negative errno if packing fails
287 int osp_prep_inline_update_req(const struct lu_env *env,
288 struct ptlrpc_request *req,
289 struct osp_update_request *our,
292 struct osp_update_request_sub *ours;
293 struct out_update_header *ouh;
294 __u32 update_req_size;
297 ours = list_entry(our->our_req_list.next,
298 struct osp_update_request_sub, ours_list);
299 update_req_size = object_update_request_size(ours->ours_req);
300 req_capsule_set_size(&req->rq_pill, &RMF_OUT_UPDATE_HEADER, RCL_CLIENT,
301 update_req_size + sizeof(*ouh));
303 rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, OUT_UPDATE);
307 ouh = req_capsule_client_get(&req->rq_pill, &RMF_OUT_UPDATE_HEADER);
308 ouh->ouh_magic = OUT_UPDATE_HEADER_MAGIC;
310 ouh->ouh_inline_length = update_req_size;
311 ouh->ouh_reply_size = repsize;
313 memcpy(ouh->ouh_inline_data, ours->ours_req, update_req_size);
315 req_capsule_set_size(&req->rq_pill, &RMF_OUT_UPDATE_REPLY,
316 RCL_SERVER, repsize);
318 ptlrpc_request_set_replen(req);
319 req->rq_request_portal = OUT_PORTAL;
320 req->rq_reply_portal = OSC_REPLY_PORTAL;
326 * Prepare update request.
328 * Prepare OUT update ptlrpc request, and the request usually includes
329 * all of updates (stored in \param ureq) from one operation.
331 * \param[in] env execution environment
332 * \param[in] imp import on which ptlrpc request will be sent
333 * \param[in] ureq hold all of updates which will be packed into the req
334 * \param[in] reqp request to be created
336 * \retval 0 if preparation succeeds.
337 * \retval negative errno if preparation fails.
339 int osp_prep_update_req(const struct lu_env *env, struct obd_import *imp,
340 struct osp_update_request *our,
341 struct ptlrpc_request **reqp)
343 struct ptlrpc_request *req;
344 struct ptlrpc_bulk_desc *desc;
345 struct osp_update_request_sub *ours;
346 const struct object_update_request *ourq;
347 struct out_update_header *ouh;
348 struct out_update_buffer *oub;
351 struct object_update_reply *reply;
356 list_for_each_entry(ours, &our->our_req_list, ours_list) {
357 object_update_request_dump(ours->ours_req, D_INFO);
359 ourq = ours->ours_req;
360 for (i = 0; i < ourq->ourq_count; i++) {
361 struct object_update *update;
365 /* XXX: it's very inefficient to lookup update
366 * this way, iterating from the beginning
368 update = object_update_request_get(ourq, i, &size);
369 LASSERT(update != NULL);
371 repsize += sizeof(reply->ourp_lens[0]);
372 repsize += sizeof(struct object_update_result);
373 repsize += update->ou_result_size;
378 repsize += sizeof(*reply);
379 repsize = (repsize + OUT_UPDATE_REPLY_SIZE - 1) &
380 ~(OUT_UPDATE_REPLY_SIZE - 1);
381 LASSERT(buf_count > 0);
383 req = ptlrpc_request_alloc(imp, &RQF_OUT_UPDATE);
387 if (buf_count == 1) {
388 ours = list_entry(our->our_req_list.next,
389 struct osp_update_request_sub, ours_list);
391 /* Let's check if it can be packed inline */
392 if (object_update_request_size(ours->ours_req) +
393 sizeof(struct out_update_header) <
394 OUT_UPDATE_MAX_INLINE_SIZE) {
395 rc = osp_prep_inline_update_req(env, req, our, repsize);
402 req_capsule_set_size(&req->rq_pill, &RMF_OUT_UPDATE_HEADER, RCL_CLIENT,
403 sizeof(struct osp_update_request));
405 req_capsule_set_size(&req->rq_pill, &RMF_OUT_UPDATE_BUF, RCL_CLIENT,
406 buf_count * sizeof(*oub));
408 rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, OUT_UPDATE);
412 ouh = req_capsule_client_get(&req->rq_pill, &RMF_OUT_UPDATE_HEADER);
413 ouh->ouh_magic = OUT_UPDATE_HEADER_MAGIC;
414 ouh->ouh_count = buf_count;
415 ouh->ouh_inline_length = 0;
416 ouh->ouh_reply_size = repsize;
417 oub = req_capsule_client_get(&req->rq_pill, &RMF_OUT_UPDATE_BUF);
418 list_for_each_entry(ours, &our->our_req_list, ours_list) {
419 oub->oub_size = ours->ours_req_size;
423 req->rq_bulk_write = 1;
424 desc = ptlrpc_prep_bulk_imp(req, buf_count,
425 MD_MAX_BRW_SIZE >> LNET_MTU_BITS,
426 PTLRPC_BULK_GET_SOURCE | PTLRPC_BULK_BUF_KVEC,
427 MDS_BULK_PORTAL, &ptlrpc_bulk_kvec_ops);
429 GOTO(out_req, rc = -ENOMEM);
431 /* NB req now owns desc and will free it when it gets freed */
432 list_for_each_entry(ours, &our->our_req_list, ours_list) {
433 desc->bd_frag_ops->add_iov_frag(desc, ours->ours_req,
434 ours->ours_req_size);
435 total += ours->ours_req_size;
437 CDEBUG(D_OTHER, "total %d in %u\n", total, our->our_update_nr);
439 req_capsule_set_size(&req->rq_pill, &RMF_OUT_UPDATE_REPLY,
440 RCL_SERVER, repsize);
442 ptlrpc_request_set_replen(req);
443 req->rq_request_portal = OUT_PORTAL;
444 req->rq_reply_portal = OSC_REPLY_PORTAL;
449 ptlrpc_req_finished(req);
457 * Send update request to the remote MDT synchronously.
459 * \param[in] env execution environment
460 * \param[in] imp import on which ptlrpc request will be sent
461 * \param[in] our hold all of updates which will be packed into the req
462 * \param[in] reqp request to be created
464 * \retval 0 if RPC succeeds.
465 * \retval negative errno if RPC fails.
467 int osp_remote_sync(const struct lu_env *env, struct osp_device *osp,
468 struct osp_update_request *our,
469 struct ptlrpc_request **reqp)
471 struct obd_import *imp = osp->opd_obd->u.cli.cl_import;
472 struct ptlrpc_request *req = NULL;
476 rc = osp_prep_update_req(env, imp, our, &req);
480 osp_set_req_replay(osp, req);
481 req->rq_allow_intr = 1;
483 /* Note: some dt index api might return non-zero result here, like
484 * osd_index_ea_lookup, so we should only check rc < 0 here */
485 rc = ptlrpc_queue_wait(req);
487 if (rc < 0 || reqp == NULL)
488 ptlrpc_req_finished(req);
496 * Invalidate all objects in the osp thandle
498 * invalidate all of objects in the update request, which will be called
499 * when the transaction is aborted.
501 * \param[in] oth osp thandle.
503 static void osp_thandle_invalidate_object(const struct lu_env *env,
504 struct osp_thandle *oth,
507 struct osp_update_request *our = oth->ot_our;
508 struct osp_object *obj;
509 struct osp_object *next;
514 list_for_each_entry_safe(obj, next, &our->our_invalidate_cb_list,
515 opo_invalidate_cb_list) {
517 osp_invalidate(env, &obj->opo_obj);
519 spin_lock(&obj->opo_lock);
520 list_del_init(&obj->opo_invalidate_cb_list);
521 spin_unlock(&obj->opo_lock);
523 dt_object_put(env, &obj->opo_obj);
527 static void osp_trans_stop_cb(const struct lu_env *env,
528 struct osp_thandle *oth, int result)
530 struct dt_txn_commit_cb *dcb;
531 struct dt_txn_commit_cb *tmp;
533 /* call per-transaction stop callbacks if any */
534 list_for_each_entry_safe(dcb, tmp, &oth->ot_stop_dcb_list,
536 LASSERTF(dcb->dcb_magic == TRANS_COMMIT_CB_MAGIC,
537 "commit callback entry: magic=%x name='%s'\n",
538 dcb->dcb_magic, dcb->dcb_name);
539 list_del_init(&dcb->dcb_linkage);
540 dcb->dcb_func(NULL, &oth->ot_super, dcb, result);
543 osp_thandle_invalidate_object(env, oth, result);
547 * Allocate an osp request and initialize it with the given parameters.
549 * \param[in] obj pointer to the operation target
550 * \param[in] data pointer to the data used by the interpreter
551 * \param[in] interpreter pointer to the interpreter function
553 * \retval pointer to the asychronous request
554 * \retval NULL if the allocation failed
556 static struct osp_update_callback *
557 osp_update_callback_init(struct osp_object *obj, void *data,
558 osp_update_interpreter_t interpreter)
560 struct osp_update_callback *ouc;
566 lu_object_get(osp2lu_obj(obj));
567 INIT_LIST_HEAD(&ouc->ouc_list);
569 ouc->ouc_data = data;
570 ouc->ouc_interpreter = interpreter;
576 * Destroy the osp_update_callback.
578 * \param[in] env pointer to the thread context
579 * \param[in] ouc pointer to osp_update_callback
581 static void osp_update_callback_fini(const struct lu_env *env,
582 struct osp_update_callback *ouc)
584 LASSERT(list_empty(&ouc->ouc_list));
586 lu_object_put(env, osp2lu_obj(ouc->ouc_obj));
591 * Interpret the packaged OUT RPC results.
593 * For every packaged sub-request, call its registered interpreter function.
594 * Then destroy the sub-request.
596 * \param[in] env pointer to the thread context
597 * \param[in] req pointer to the RPC
598 * \param[in] arg pointer to data used by the interpreter
599 * \param[in] rc the RPC return value
601 * \retval 0 for success
602 * \retval negative error number on failure
604 static int osp_update_interpret(const struct lu_env *env,
605 struct ptlrpc_request *req, void *arg, int rc)
607 struct object_update_reply *reply = NULL;
608 struct osp_update_args *oaua = arg;
609 struct osp_update_request *our = oaua->oaua_update;
610 struct osp_thandle *oth;
611 struct osp_update_callback *ouc;
612 struct osp_update_callback *next;
622 /* Sigh env might be NULL in some cases, see
624 * osp_send_update_thread()
625 * ptlrpc_set_wait() ----> null env.
627 * osp_update_interpret()
628 * Let's use env in oaua for this case.
631 env = oaua->oaua_update_env;
633 oaua->oaua_update = NULL;
635 if (oaua->oaua_flow_control) {
636 struct osp_device *osp;
638 LASSERT(oth != NULL);
639 osp = dt2osp_dev(oth->ot_super.th_dev);
640 obd_put_request_slot(&osp->opd_obd->u.cli);
643 /* Unpack the results from the reply message. */
644 if (req->rq_repmsg != NULL && req->rq_replied) {
645 reply = req_capsule_server_sized_get(&req->rq_pill,
646 &RMF_OUT_UPDATE_REPLY,
647 OUT_UPDATE_REPLY_SIZE);
648 if (reply == NULL || reply->ourp_magic != UPDATE_REPLY_MAGIC) {
652 count = reply->ourp_count;
656 list_for_each_entry_safe(ouc, next, &our->our_cb_items, ouc_list) {
657 list_del_init(&ouc->ouc_list);
659 /* The peer may only have handled some requests (indicated
660 * by the 'count') in the packaged OUT RPC, we can only get
661 * results for the handled part. */
662 if (index < count && reply->ourp_lens[index] > 0 && rc >= 0) {
663 struct object_update_result *result;
665 result = object_update_result_get(reply, index, NULL);
669 rc1 = rc = result->our_rc;
670 } else if (rc1 >= 0) {
671 /* The peer did not handle these request, let's return
672 * -EINVAL to update interpret for now */
679 if (ouc->ouc_interpreter != NULL)
680 ouc->ouc_interpreter(env, reply, req, ouc->ouc_obj,
681 ouc->ouc_data, index, rc1);
683 osp_update_callback_fini(env, ouc);
687 if (oaua->oaua_count != NULL && atomic_dec_and_test(oaua->oaua_count))
688 wake_up_all(oaua->oaua_waitq);
691 /* oth and osp_update_requests will be destoryed in
693 osp_trans_stop_cb(env, oth, rc);
694 osp_thandle_put(env, oth);
696 osp_update_request_destroy(env, our);
703 * Pack all the requests in the shared asynchronous idempotent request queue
704 * into a single OUT RPC that will be given to the background ptlrpcd daemon.
706 * \param[in] env pointer to the thread context
707 * \param[in] osp pointer to the OSP device
708 * \param[in] our pointer to the shared queue
710 * \retval 0 for success
711 * \retval negative error number on failure
713 int osp_unplug_async_request(const struct lu_env *env,
714 struct osp_device *osp,
715 struct osp_update_request *our)
717 struct osp_update_args *args;
718 struct ptlrpc_request *req = NULL;
721 rc = osp_prep_update_req(env, osp->opd_obd->u.cli.cl_import,
724 struct osp_update_callback *ouc;
725 struct osp_update_callback *next;
727 list_for_each_entry_safe(ouc, next,
728 &our->our_cb_items, ouc_list) {
729 list_del_init(&ouc->ouc_list);
730 if (ouc->ouc_interpreter != NULL)
731 ouc->ouc_interpreter(env, NULL, NULL,
733 ouc->ouc_data, 0, rc);
734 osp_update_callback_fini(env, ouc);
736 osp_update_request_destroy(env, our);
738 args = ptlrpc_req_async_args(req);
739 args->oaua_update = our;
740 args->oaua_count = NULL;
741 args->oaua_waitq = NULL;
742 /* Note: this is asynchronous call for the request, so the
743 * interrupte cb and current function will be different
744 * thread, so we need use different env */
745 args->oaua_update_env = NULL;
746 args->oaua_flow_control = false;
747 req->rq_interpret_reply = osp_update_interpret;
748 ptlrpcd_add_req(req);
755 * Find or create (if NOT exist or purged) the shared asynchronous idempotent
756 * request queue - osp_device::opd_async_requests.
758 * If the osp_device::opd_async_requests is not NULL, then return it directly;
759 * otherwise create new osp_update_request and attach it to opd_async_requests.
761 * \param[in] osp pointer to the OSP device
763 * \retval pointer to the shared queue
764 * \retval negative error number on failure
766 static struct osp_update_request *
767 osp_find_or_create_async_update_request(struct osp_device *osp)
769 struct osp_update_request *our = osp->opd_async_requests;
774 our = osp_update_request_create(&osp->opd_dt_dev);
778 osp->opd_async_requests = our;
784 * Insert an osp_update_callback into the osp_update_request.
786 * Insert an osp_update_callback to the osp_update_request. Usually each update
787 * in the osp_update_request will have one correspondent callback, and these
788 * callbacks will be called in rq_interpret_reply.
790 * \param[in] env pointer to the thread context
791 * \param[in] obj pointer to the operation target object
792 * \param[in] data pointer to the data used by the interpreter
793 * \param[in] interpreter pointer to the interpreter function
795 * \retval 0 for success
796 * \retval negative error number on failure
798 int osp_insert_update_callback(const struct lu_env *env,
799 struct osp_update_request *our,
800 struct osp_object *obj, void *data,
801 osp_update_interpreter_t interpreter)
803 struct osp_update_callback *ouc;
805 ouc = osp_update_callback_init(obj, data, interpreter);
809 list_add_tail(&ouc->ouc_list, &our->our_cb_items);
815 * Insert an asynchronous idempotent request to the shared request queue that
816 * is attached to the osp_device.
818 * This function generates a new osp_async_request with the given parameters,
819 * then tries to insert the request into the osp_device-based shared request
820 * queue. If the queue is full, then triggers the packaged OUT RPC to purge
821 * the shared queue firstly, and then re-tries.
823 * NOTE: must hold the osp::opd_async_requests_mutex to serialize concurrent
824 * osp_insert_async_request call from others.
826 * \param[in] env pointer to the thread context
827 * \param[in] op operation type, see 'enum update_type'
828 * \param[in] obj pointer to the operation target
829 * \param[in] count array size of the subsequent \a lens and \a bufs
830 * \param[in] lens buffer length array for the subsequent \a bufs
831 * \param[in] bufs the buffers to compose the request
832 * \param[in] data pointer to the data used by the interpreter
833 * \param[in] repsize how many bytes the caller allocated for \a data
834 * \param[in] interpreter pointer to the interpreter function
836 * \retval 0 for success
837 * \retval negative error number on failure
839 int osp_insert_async_request(const struct lu_env *env, enum update_type op,
840 struct osp_object *obj, int count,
841 __u16 *lens, const void **bufs,
842 void *data, __u32 repsize,
843 osp_update_interpreter_t interpreter)
845 struct osp_device *osp;
846 struct osp_update_request *our;
847 struct object_update *object_update;
848 size_t max_update_size;
849 struct object_update_request *ureq;
850 struct osp_update_request_sub *ours;
854 osp = lu2osp_dev(osp2lu_obj(obj)->lo_dev);
855 our = osp_find_or_create_async_update_request(osp);
857 RETURN(PTR_ERR(our));
860 ours = osp_current_object_update_request(our);
862 ureq = ours->ours_req;
863 max_update_size = ours->ours_req_size -
864 object_update_request_size(ureq);
866 object_update = update_buffer_get_update(ureq, ureq->ourq_count);
867 rc = out_update_pack(env, object_update, &max_update_size, op,
868 lu_object_fid(osp2lu_obj(obj)), count, lens, bufs,
870 /* The queue is full. */
872 osp->opd_async_requests = NULL;
873 mutex_unlock(&osp->opd_async_requests_mutex);
875 rc = osp_unplug_async_request(env, osp, our);
876 mutex_lock(&osp->opd_async_requests_mutex);
880 our = osp_find_or_create_async_update_request(osp);
882 RETURN(PTR_ERR(our));
890 our->our_update_nr++;
893 rc = osp_insert_update_callback(env, our, obj, data, interpreter);
898 int osp_trans_update_request_create(struct thandle *th)
900 struct osp_thandle *oth = thandle_to_osp_thandle(th);
901 struct osp_update_request *our;
903 if (oth->ot_our != NULL)
906 our = osp_update_request_create(th->th_dev);
908 th->th_result = PTR_ERR(our);
918 void osp_thandle_destroy(const struct lu_env *env,
919 struct osp_thandle *oth)
921 LASSERT(oth->ot_magic == OSP_THANDLE_MAGIC);
922 LASSERT(list_empty(&oth->ot_commit_dcb_list));
923 LASSERT(list_empty(&oth->ot_stop_dcb_list));
924 if (oth->ot_our != NULL)
925 osp_update_request_destroy(env, oth->ot_our);
930 * The OSP layer dt_device_operations::dt_trans_create() interface
931 * to create a transaction.
933 * There are two kinds of transactions that will involve OSP:
935 * 1) If the transaction only contains the updates on remote server
936 * (MDT or OST), such as re-generating the lost OST-object for
937 * LFSCK, then it is a remote transaction. For remote transaction,
938 * the upper layer caller (such as the LFSCK engine) will call the
939 * dt_trans_create() (with the OSP dt_device as the parameter),
940 * then the call will be directed to the osp_trans_create() that
941 * creates the transaction handler and returns it to the caller.
943 * 2) If the transcation contains both local and remote updates,
944 * such as cross MDTs create under DNE mode, then the upper layer
945 * caller will not trigger osp_trans_create(). Instead, it will
946 * call dt_trans_create() on other dt_device, such as LOD that
947 * will generate the transaction handler. Such handler will be
948 * used by the whole transaction in subsequent sub-operations.
950 * \param[in] env pointer to the thread context
951 * \param[in] d pointer to the OSP dt_device
953 * \retval pointer to the transaction handler
954 * \retval negative error number on failure
956 struct thandle *osp_trans_create(const struct lu_env *env, struct dt_device *d)
958 struct osp_thandle *oth;
959 struct thandle *th = NULL;
963 if (unlikely(oth == NULL))
964 RETURN(ERR_PTR(-ENOMEM));
966 oth->ot_magic = OSP_THANDLE_MAGIC;
970 atomic_set(&oth->ot_refcount, 1);
971 INIT_LIST_HEAD(&oth->ot_commit_dcb_list);
972 INIT_LIST_HEAD(&oth->ot_stop_dcb_list);
978 * Add commit callback to transaction.
980 * Add commit callback to the osp thandle, which will be called
981 * when the thandle is committed remotely.
983 * \param[in] th the thandle
984 * \param[in] dcb commit callback structure
986 * \retval only return 0 for now.
988 int osp_trans_cb_add(struct thandle *th, struct dt_txn_commit_cb *dcb)
990 struct osp_thandle *oth = thandle_to_osp_thandle(th);
992 LASSERT(dcb->dcb_magic == TRANS_COMMIT_CB_MAGIC);
993 LASSERT(&dcb->dcb_func != NULL);
994 if (dcb->dcb_flags & DCB_TRANS_STOP)
995 list_add(&dcb->dcb_linkage, &oth->ot_stop_dcb_list);
997 list_add(&dcb->dcb_linkage, &oth->ot_commit_dcb_list);
1001 static void osp_trans_commit_cb(struct osp_thandle *oth, int result)
1003 struct dt_txn_commit_cb *dcb;
1004 struct dt_txn_commit_cb *tmp;
1006 LASSERT(atomic_read(&oth->ot_refcount) > 0);
1007 /* call per-transaction callbacks if any */
1008 list_for_each_entry_safe(dcb, tmp, &oth->ot_commit_dcb_list,
1010 LASSERTF(dcb->dcb_magic == TRANS_COMMIT_CB_MAGIC,
1011 "commit callback entry: magic=%x name='%s'\n",
1012 dcb->dcb_magic, dcb->dcb_name);
1013 list_del_init(&dcb->dcb_linkage);
1014 dcb->dcb_func(NULL, &oth->ot_super, dcb, result);
1018 static void osp_request_commit_cb(struct ptlrpc_request *req)
1020 struct thandle *th = req->rq_cb_data;
1021 struct osp_thandle *oth;
1022 __u64 last_committed_transno = 0;
1023 int result = req->rq_status;
1029 oth = thandle_to_osp_thandle(th);
1030 if (req->rq_repmsg != NULL &&
1031 lustre_msg_get_last_committed(req->rq_repmsg))
1032 last_committed_transno =
1033 lustre_msg_get_last_committed(req->rq_repmsg);
1035 if (last_committed_transno <
1036 req->rq_import->imp_peer_committed_transno)
1037 last_committed_transno =
1038 req->rq_import->imp_peer_committed_transno;
1040 CDEBUG(D_HA, "trans no %llu committed transno %llu\n",
1041 req->rq_transno, last_committed_transno);
1043 /* If the transaction is not really committed, mark result = 1 */
1044 if (req->rq_transno != 0 &&
1045 (req->rq_transno > last_committed_transno) && result == 0)
1048 osp_trans_commit_cb(oth, result);
1049 req->rq_committed = 1;
1050 osp_thandle_put(NULL, oth);
1055 * callback of osp transaction
1057 * Call all of callbacks for this osp thandle. This will only be
1058 * called in error handler path. In the normal processing path,
1059 * these callback will be called in osp_request_commit_cb() and
1060 * osp_update_interpret().
1062 * \param [in] env execution environment
1063 * \param [in] oth osp thandle
1064 * \param [in] rc result of the osp thandle
1066 void osp_trans_callback(const struct lu_env *env,
1067 struct osp_thandle *oth, int rc)
1069 struct osp_update_callback *ouc;
1070 struct osp_update_callback *next;
1072 if (oth->ot_our != NULL) {
1073 list_for_each_entry_safe(ouc, next,
1074 &oth->ot_our->our_cb_items, ouc_list) {
1075 list_del_init(&ouc->ouc_list);
1076 if (ouc->ouc_interpreter != NULL)
1077 ouc->ouc_interpreter(env, NULL, NULL,
1079 ouc->ouc_data, 0, rc);
1080 osp_update_callback_fini(env, ouc);
1083 osp_trans_stop_cb(env, oth, rc);
1084 osp_trans_commit_cb(oth, rc);
1088 * Send the request for remote updates.
1090 * Send updates to the remote MDT. Prepare the request by osp_update_req
1091 * and send them to remote MDT, for sync request, it will wait
1092 * until the reply return, otherwise hand it to ptlrpcd.
1094 * Please refer to osp_trans_create() for transaction type.
1096 * \param[in] env pointer to the thread context
1097 * \param[in] osp pointer to the OSP device
1098 * \param[in] our pointer to the osp_update_request
1100 * \retval 0 for success
1101 * \retval negative error number on failure
1103 static int osp_send_update_req(const struct lu_env *env,
1104 struct osp_device *osp,
1105 struct osp_update_request *our)
1107 struct osp_update_args *args;
1108 struct ptlrpc_request *req;
1109 struct osp_thandle *oth = our->our_th;
1113 LASSERT(oth != NULL);
1114 rc = osp_prep_update_req(env, osp->opd_obd->u.cli.cl_import,
1117 osp_trans_callback(env, oth, rc);
1121 args = ptlrpc_req_async_args(req);
1122 args->oaua_update = our;
1123 /* set env to NULL, in case the interrupt cb and current function
1124 * are in different thread */
1125 args->oaua_update_env = NULL;
1126 osp_thandle_get(oth); /* hold for update interpret */
1127 req->rq_interpret_reply = osp_update_interpret;
1128 if (!oth->ot_super.th_wait_submit && !oth->ot_super.th_sync) {
1129 if (!osp->opd_imp_active || !osp->opd_imp_connected) {
1130 osp_trans_callback(env, oth, rc);
1131 osp_thandle_put(env, oth);
1132 GOTO(out, rc = -ENOTCONN);
1135 rc = obd_get_request_slot(&osp->opd_obd->u.cli);
1137 osp_trans_callback(env, oth, rc);
1138 osp_thandle_put(env, oth);
1139 GOTO(out, rc = -ENOTCONN);
1141 args->oaua_flow_control = true;
1143 if (!osp->opd_connect_mdt) {
1144 down_read(&osp->opd_async_updates_rwsem);
1145 args->oaua_count = &osp->opd_async_updates_count;
1146 args->oaua_waitq = &osp->opd_sync_barrier_waitq;
1147 up_read(&osp->opd_async_updates_rwsem);
1148 atomic_inc(args->oaua_count);
1151 ptlrpcd_add_req(req);
1154 osp_thandle_get(oth); /* hold for commit callback */
1155 req->rq_commit_cb = osp_request_commit_cb;
1156 req->rq_cb_data = &oth->ot_super;
1157 args->oaua_flow_control = false;
1159 /* If the transaction is created during MDT recoverying
1160 * process, it means this is an recovery update, we need
1161 * to let OSP send it anyway without checking recoverying
1162 * status, in case the other target is being recoveried
1163 * at the same time, and if we wait here for the import
1164 * to be recoveryed, it might cause deadlock */
1165 osp_set_req_replay(osp, req);
1167 /* Because this req will be synchronus, i.e. it will be called
1168 * in the same thread, so it will be safe to use current
1170 args->oaua_update_env = env;
1171 if (osp->opd_connect_mdt)
1172 osp_get_rpc_lock(osp);
1173 rc = ptlrpc_queue_wait(req);
1174 if (osp->opd_connect_mdt)
1175 osp_put_rpc_lock(osp);
1177 /* We use rq_queued_time to distinguish between local
1178 * and remote -ENOMEM. */
1179 if ((rc == -ENOMEM && req->rq_queued_time == 0) ||
1180 (req->rq_transno == 0 && !req->rq_committed)) {
1181 if (args->oaua_update != NULL) {
1182 /* If osp_update_interpret is not being called,
1183 * release the osp_thandle */
1184 args->oaua_update = NULL;
1185 osp_thandle_put(env, oth);
1188 req->rq_cb_data = NULL;
1189 rc = rc == 0 ? req->rq_status : rc;
1190 osp_trans_callback(env, oth, rc);
1191 osp_thandle_put(env, oth);
1197 ptlrpc_req_finished(req);
1203 * Get local thandle for osp_thandle
1205 * Get the local OSD thandle from the OSP thandle. Currently, there
1206 * are a few OSP API (osp_create() and osp_sync_add()) needs
1207 * to update the object on local OSD device.
1209 * If the osp_thandle comes from normal stack (MDD->LOD->OSP), then
1210 * we will get local thandle by thandle_get_sub_by_dt.
1212 * If the osp_thandle is remote thandle (th_top == NULL, only used
1213 * by LFSCK), then it will create a local thandle, and stop it in
1214 * osp_trans_stop(). And this only happens on OSP for OST.
1216 * These are temporary solution, once OSP accessing OSD object is
1217 * being fixed properly, this function should be removed. XXX
1219 * \param[in] env pointer to the thread context
1220 * \param[in] th pointer to the transaction handler
1221 * \param[in] dt pointer to the OSP device
1223 * \retval pointer to the local thandle
1224 * \retval ERR_PTR(errno) if it fails.
1226 struct thandle *osp_get_storage_thandle(const struct lu_env *env,
1228 struct osp_device *osp)
1230 struct osp_thandle *oth;
1231 struct thandle *local_th;
1233 if (th->th_top != NULL)
1234 return thandle_get_sub_by_dt(env, th->th_top,
1237 LASSERT(!osp->opd_connect_mdt);
1238 oth = thandle_to_osp_thandle(th);
1239 if (oth->ot_storage_th != NULL)
1240 return oth->ot_storage_th;
1242 local_th = dt_trans_create(env, osp->opd_storage);
1243 if (IS_ERR(local_th))
1246 oth->ot_storage_th = local_th;
1252 * Set version for the transaction
1254 * Set the version for the transaction and add the request to
1255 * the sending list, then after transaction stop, the request
1256 * will be sent in the order of version by the sending thread.
1258 * \param [in] oth osp thandle to be set version.
1260 * \retval 0 if set version succeeds
1261 * negative errno if set version fails.
1263 int osp_check_and_set_rpc_version(struct osp_thandle *oth,
1264 struct osp_object *obj)
1266 struct osp_device *osp = dt2osp_dev(oth->ot_super.th_dev);
1267 struct osp_updates *ou = osp->opd_update;
1272 if (oth->ot_our->our_version != 0)
1275 spin_lock(&ou->ou_lock);
1276 spin_lock(&oth->ot_our->our_list_lock);
1277 if (obj->opo_stale) {
1278 spin_unlock(&oth->ot_our->our_list_lock);
1279 spin_unlock(&ou->ou_lock);
1283 /* Assign the version and add it to the sending list */
1284 osp_thandle_get(oth);
1285 oth->ot_our->our_version = ou->ou_version++;
1286 oth->ot_our->our_generation = ou->ou_generation;
1287 list_add_tail(&oth->ot_our->our_list,
1288 &osp->opd_update->ou_list);
1289 oth->ot_our->our_req_ready = 0;
1290 spin_unlock(&oth->ot_our->our_list_lock);
1291 spin_unlock(&ou->ou_lock);
1293 LASSERT(oth->ot_super.th_wait_submit == 1);
1294 CDEBUG(D_INFO, "%s: version %llu gen %llu oth:version %p:%llu\n",
1295 osp->opd_obd->obd_name, ou->ou_version, ou->ou_generation, oth,
1296 oth->ot_our->our_version);
1302 * Get next OSP update request in the sending list
1303 * Get next OSP update request in the sending list by version number, next
1305 * 1. transaction which does not have a version number.
1306 * 2. transaction whose version == opd_rpc_version.
1308 * \param [in] ou osp update structure.
1309 * \param [out] ourp the pointer holding the next update request.
1311 * \retval true if getting the next transaction.
1312 * \retval false if not getting the next transaction.
1315 osp_get_next_request(struct osp_updates *ou, struct osp_update_request **ourp)
1317 struct osp_update_request *our;
1318 struct osp_update_request *tmp;
1319 bool got_req = false;
1321 spin_lock(&ou->ou_lock);
1322 list_for_each_entry_safe(our, tmp, &ou->ou_list, our_list) {
1323 LASSERT(our->our_th != NULL);
1324 CDEBUG(D_HA, "ou %p version %llu rpc_version %llu\n",
1325 ou, our->our_version, ou->ou_rpc_version);
1326 spin_lock(&our->our_list_lock);
1327 /* Find next osp_update_request in the list */
1328 if (our->our_version == ou->ou_rpc_version &&
1329 our->our_req_ready) {
1330 list_del_init(&our->our_list);
1331 spin_unlock(&our->our_list_lock);
1336 spin_unlock(&our->our_list_lock);
1338 spin_unlock(&ou->ou_lock);
1344 * Invalidate update request
1346 * Invalidate update request in the OSP sending list, so all of
1347 * requests in the sending list will return error, which happens
1348 * when it finds one update (with writing llog) requests fails or
1349 * the OSP is evicted by remote target. see osp_send_update_thread().
1351 * \param[in] osp OSP device whose update requests will be
1354 void osp_invalidate_request(struct osp_device *osp)
1357 struct osp_updates *ou = osp->opd_update;
1358 struct osp_update_request *our;
1359 struct osp_update_request *tmp;
1367 rc = lu_env_init(&env, osp->opd_dt_dev.dd_lu_dev.ld_type->ldt_ctx_tags);
1369 CERROR("%s: init env error: rc = %d\n", osp->opd_obd->obd_name,
1372 spin_lock(&ou->ou_lock);
1373 ou->ou_generation++;
1374 spin_unlock(&ou->ou_lock);
1379 INIT_LIST_HEAD(&list);
1381 spin_lock(&ou->ou_lock);
1382 /* invalidate all of request in the sending list */
1383 list_for_each_entry_safe(our, tmp, &ou->ou_list, our_list) {
1384 spin_lock(&our->our_list_lock);
1385 if (our->our_req_ready)
1386 list_move(&our->our_list, &list);
1388 list_del_init(&our->our_list);
1390 if (our->our_th->ot_super.th_result == 0)
1391 our->our_th->ot_super.th_result = -EIO;
1393 if (our->our_version >= ou->ou_rpc_version)
1394 ou->ou_rpc_version = our->our_version + 1;
1395 spin_unlock(&our->our_list_lock);
1397 CDEBUG(D_HA, "%s invalidate our %p\n", osp->opd_obd->obd_name,
1401 /* Increase the generation, then the update request with old generation
1402 * will fail with -EIO. */
1403 ou->ou_generation++;
1404 spin_unlock(&ou->ou_lock);
1406 /* invalidate all of request in the sending list */
1407 list_for_each_entry_safe(our, tmp, &list, our_list) {
1408 spin_lock(&our->our_list_lock);
1409 list_del_init(&our->our_list);
1410 spin_unlock(&our->our_list_lock);
1411 osp_trans_callback(&env, our->our_th,
1412 our->our_th->ot_super.th_result);
1413 osp_thandle_put(&env, our->our_th);
1419 * Sending update thread
1421 * Create thread to send update request to other MDTs, this thread will pull
1422 * out update request from the list in OSP by version number, i.e. it will
1423 * make sure the update request with lower version number will be sent first.
1425 * \param[in] arg hold the OSP device.
1427 * \retval 0 if the thread is created successfully.
1428 * \retal negative error if the thread is not created
1431 int osp_send_update_thread(void *arg)
1434 struct osp_device *osp = arg;
1435 struct l_wait_info lwi = { 0 };
1436 struct osp_updates *ou = osp->opd_update;
1437 struct ptlrpc_thread *thread = &osp->opd_update_thread;
1438 struct osp_update_request *our = NULL;
1442 LASSERT(ou != NULL);
1443 rc = lu_env_init(&env, osp->opd_dt_dev.dd_lu_dev.ld_type->ldt_ctx_tags);
1445 CERROR("%s: init env error: rc = %d\n", osp->opd_obd->obd_name,
1450 thread->t_flags = SVC_RUNNING;
1451 wake_up(&thread->t_ctl_waitq);
1454 l_wait_event(ou->ou_waitq,
1455 !osp_send_update_thread_running(osp) ||
1456 osp_get_next_request(ou, &our), &lwi);
1458 if (!osp_send_update_thread_running(osp)) {
1460 osp_trans_callback(&env, our->our_th, -EINTR);
1461 osp_thandle_put(&env, our->our_th);
1466 LASSERT(our->our_th != NULL);
1467 if (our->our_th->ot_super.th_result != 0) {
1468 osp_trans_callback(&env, our->our_th,
1469 our->our_th->ot_super.th_result);
1470 rc = our->our_th->ot_super.th_result;
1471 } else if (ou->ou_generation != our->our_generation ||
1472 OBD_FAIL_CHECK(OBD_FAIL_INVALIDATE_UPDATE)) {
1474 osp_trans_callback(&env, our->our_th, rc);
1476 rc = osp_send_update_req(&env, osp, our);
1479 /* Update the rpc version */
1480 spin_lock(&ou->ou_lock);
1481 if (our->our_version == ou->ou_rpc_version)
1482 ou->ou_rpc_version++;
1483 spin_unlock(&ou->ou_lock);
1485 /* If one update request fails, let's fail all of the requests
1486 * in the sending list, because the request in the sending
1487 * list are dependent on either other, continue sending these
1488 * request might cause llog or filesystem corruption */
1490 osp_invalidate_request(osp);
1492 /* Balanced for thandle_get in osp_check_and_set_rpc_version */
1493 osp_thandle_put(&env, our->our_th);
1496 thread->t_flags = SVC_STOPPED;
1498 wake_up(&thread->t_ctl_waitq);
1504 * The OSP layer dt_device_operations::dt_trans_start() interface
1505 * to start the transaction.
1507 * If the transaction is a remote transaction, then related remote
1508 * updates will be triggered in the osp_trans_stop().
1509 * Please refer to osp_trans_create() for transaction type.
1511 * \param[in] env pointer to the thread context
1512 * \param[in] dt pointer to the OSP dt_device
1513 * \param[in] th pointer to the transaction handler
1515 * \retval 0 for success
1516 * \retval negative error number on failure
1518 int osp_trans_start(const struct lu_env *env, struct dt_device *dt,
1521 struct osp_thandle *oth = thandle_to_osp_thandle(th);
1523 if (oth->ot_super.th_sync)
1524 oth->ot_our->our_flags |= UPDATE_FL_SYNC;
1525 /* For remote thandle, if there are local thandle, start it here*/
1526 if (is_only_remote_trans(th) && oth->ot_storage_th != NULL)
1527 return dt_trans_start(env, oth->ot_storage_th->th_dev,
1528 oth->ot_storage_th);
1533 * The OSP layer dt_device_operations::dt_trans_stop() interface
1534 * to stop the transaction.
1536 * If the transaction is a remote transaction, related remote
1537 * updates will be triggered at the end of this function.
1539 * For synchronous mode update or any failed update, the request
1540 * will be destroyed explicitly when the osp_trans_stop().
1542 * Please refer to osp_trans_create() for transaction type.
1544 * \param[in] env pointer to the thread context
1545 * \param[in] dt pointer to the OSP dt_device
1546 * \param[in] th pointer to the transaction handler
1548 * \retval 0 for success
1549 * \retval negative error number on failure
1551 int osp_trans_stop(const struct lu_env *env, struct dt_device *dt,
1554 struct osp_thandle *oth = thandle_to_osp_thandle(th);
1555 struct osp_update_request *our = oth->ot_our;
1556 struct osp_device *osp = dt2osp_dev(dt);
1560 /* For remote transaction, if there is local storage thandle,
1562 if (oth->ot_storage_th != NULL && th->th_top == NULL) {
1563 dt_trans_stop(env, oth->ot_storage_th->th_dev,
1564 oth->ot_storage_th);
1565 oth->ot_storage_th = NULL;
1568 if (our == NULL || list_empty(&our->our_req_list)) {
1569 osp_trans_callback(env, oth, th->th_result);
1570 GOTO(out, rc = th->th_result);
1573 if (!osp->opd_connect_mdt) {
1574 osp_trans_callback(env, oth, th->th_result);
1575 rc = osp_send_update_req(env, osp, oth->ot_our);
1579 if (osp->opd_update == NULL ||
1580 !osp_send_update_thread_running(osp)) {
1581 osp_trans_callback(env, oth, -EIO);
1582 GOTO(out, rc = -EIO);
1585 CDEBUG(D_HA, "%s: add oth %p with version %llu\n",
1586 osp->opd_obd->obd_name, oth, our->our_version);
1588 LASSERT(our->our_req_ready == 0);
1589 spin_lock(&our->our_list_lock);
1590 if (likely(!list_empty(&our->our_list))) {
1591 /* notify sending thread */
1592 our->our_req_ready = 1;
1593 wake_up(&osp->opd_update->ou_waitq);
1594 spin_unlock(&our->our_list_lock);
1595 } else if (th->th_result == 0) {
1596 /* if the request does not needs to be serialized,
1597 * read-only request etc, let's send it right away */
1598 spin_unlock(&our->our_list_lock);
1599 rc = osp_send_update_req(env, osp, our);
1601 spin_unlock(&our->our_list_lock);
1602 osp_trans_callback(env, oth, th->th_result);
1605 osp_thandle_put(env, oth);