1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
38 # define EXPORT_SYMTAB
40 #define DEBUG_SUBSYSTEM S_MDC
43 # include <linux/module.h>
44 # include <linux/pagemap.h>
45 # include <linux/miscdevice.h>
46 # include <linux/init.h>
48 # include <liblustre.h>
51 #include <lustre_acl.h>
52 #include <obd_class.h>
53 #include <lustre_dlm.h>
54 /* fid_res_name_eq() */
55 #include <lustre_fid.h>
56 #include <lprocfs_status.h>
57 #include "mdc_internal.h"
59 int it_disposition(struct lookup_intent *it, int flag)
61 return it->d.lustre.it_disposition & flag;
63 EXPORT_SYMBOL(it_disposition);
65 void it_set_disposition(struct lookup_intent *it, int flag)
67 it->d.lustre.it_disposition |= flag;
69 EXPORT_SYMBOL(it_set_disposition);
71 void it_clear_disposition(struct lookup_intent *it, int flag)
73 it->d.lustre.it_disposition &= ~flag;
75 EXPORT_SYMBOL(it_clear_disposition);
77 int it_open_error(int phase, struct lookup_intent *it)
79 if (it_disposition(it, DISP_OPEN_OPEN)) {
80 if (phase >= DISP_OPEN_OPEN)
81 return it->d.lustre.it_status;
86 if (it_disposition(it, DISP_OPEN_CREATE)) {
87 if (phase >= DISP_OPEN_CREATE)
88 return it->d.lustre.it_status;
93 if (it_disposition(it, DISP_LOOKUP_EXECD)) {
94 if (phase >= DISP_LOOKUP_EXECD)
95 return it->d.lustre.it_status;
100 if (it_disposition(it, DISP_IT_EXECD)) {
101 if (phase >= DISP_IT_EXECD)
102 return it->d.lustre.it_status;
106 CERROR("it disp: %X, status: %d\n", it->d.lustre.it_disposition,
107 it->d.lustre.it_status);
111 EXPORT_SYMBOL(it_open_error);
113 /* this must be called on a lockh that is known to have a referenced lock */
114 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
117 struct ldlm_lock *lock;
128 lock = ldlm_handle2lock((struct lustre_handle *)lockh);
130 LASSERT(lock != NULL);
131 lock_res_and_lock(lock);
133 if (lock->l_ast_data && lock->l_ast_data != data) {
134 struct inode *new_inode = data;
135 struct inode *old_inode = lock->l_ast_data;
136 LASSERTF(old_inode->i_state & I_FREEING,
137 "Found existing inode %p/%lu/%u state %lu in lock: "
138 "setting data to %p/%lu/%u\n", old_inode,
139 old_inode->i_ino, old_inode->i_generation,
141 new_inode, new_inode->i_ino, new_inode->i_generation);
144 lock->l_ast_data = data;
146 *bits = lock->l_policy_data.l_inodebits.bits;
148 unlock_res_and_lock(lock);
154 ldlm_mode_t mdc_lock_match(struct obd_export *exp, int flags,
155 const struct lu_fid *fid, ldlm_type_t type,
156 ldlm_policy_data_t *policy, ldlm_mode_t mode,
157 struct lustre_handle *lockh)
159 struct ldlm_res_id res_id;
163 fid_build_reg_res_name(fid, &res_id);
164 rc = ldlm_lock_match(class_exp2obd(exp)->obd_namespace, flags,
165 &res_id, type, policy, mode, lockh, 0);
169 int mdc_cancel_unused(struct obd_export *exp,
170 const struct lu_fid *fid,
171 ldlm_policy_data_t *policy,
172 ldlm_mode_t mode, int flags, void *opaque)
174 struct ldlm_res_id res_id;
175 struct obd_device *obd = class_exp2obd(exp);
180 fid_build_reg_res_name(fid, &res_id);
181 rc = ldlm_cli_cancel_unused_resource(obd->obd_namespace, &res_id,
182 policy, mode, flags, opaque);
186 int mdc_change_cbdata(struct obd_export *exp,
187 const struct lu_fid *fid,
188 ldlm_iterator_t it, void *data)
190 struct ldlm_res_id res_id;
193 fid_build_reg_res_name(fid, &res_id);
194 ldlm_resource_iterate(class_exp2obd(exp)->obd_namespace,
201 static inline void mdc_clear_replay_flag(struct ptlrpc_request *req, int rc)
203 /* Don't hold error requests for replay. */
204 if (req->rq_replay) {
205 spin_lock(&req->rq_lock);
207 spin_unlock(&req->rq_lock);
209 if (rc && req->rq_transno != 0) {
210 DEBUG_REQ(D_ERROR, req, "transno returned on error rc %d", rc);
215 /* Save a large LOV EA into the request buffer so that it is available
216 * for replay. We don't do this in the initial request because the
217 * original request doesn't need this buffer (at most it sends just the
218 * lov_mds_md) and it is a waste of RAM/bandwidth to send the empty
219 * buffer and may also be difficult to allocate and save a very large
220 * request buffer for each open. (bug 5707)
222 * OOM here may cause recovery failure if lmm is needed (only for the
223 * original open if the MDS crashed just when this client also OOM'd)
224 * but this is incredibly unlikely, and questionable whether the client
225 * could do MDS recovery under OOM anyways... */
226 static void mdc_realloc_openmsg(struct ptlrpc_request *req,
227 struct mdt_body *body)
231 /* FIXME: remove this explicit offset. */
232 rc = sptlrpc_cli_enlarge_reqbuf(req, DLM_INTENT_REC_OFF + 4,
235 CERROR("Can't enlarge segment %d size to %d\n",
236 DLM_INTENT_REC_OFF + 4, body->eadatasize);
237 body->valid &= ~OBD_MD_FLEASIZE;
238 body->eadatasize = 0;
242 static struct ptlrpc_request *mdc_intent_open_pack(struct obd_export *exp,
243 struct lookup_intent *it,
244 struct md_op_data *op_data,
245 void *lmm, int lmmsize,
248 struct ptlrpc_request *req;
249 struct obd_device *obddev = class_exp2obd(exp);
250 struct ldlm_intent *lit;
251 CFS_LIST_HEAD(cancels);
257 it->it_create_mode = (it->it_create_mode & ~S_IFMT) | S_IFREG;
259 /* XXX: openlock is not cancelled for cross-refs. */
260 /* If inode is known, cancel conflicting OPEN locks. */
261 if (fid_is_sane(&op_data->op_fid2)) {
262 if (it->it_flags & (FMODE_WRITE|MDS_OPEN_TRUNC))
265 else if (it->it_flags & FMODE_EXEC)
270 count = mdc_resource_get_unused(exp, &op_data->op_fid2,
275 /* If CREATE, cancel parent's UPDATE lock. */
276 if (it->it_op & IT_CREAT)
280 count += mdc_resource_get_unused(exp, &op_data->op_fid1,
282 MDS_INODELOCK_UPDATE);
284 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
285 &RQF_LDLM_INTENT_OPEN);
287 ldlm_lock_list_put(&cancels, l_bl_ast, count);
288 RETURN(ERR_PTR(-ENOMEM));
291 /* parent capability */
292 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
293 /* child capability, reserve the size according to parent capa, it will
294 * be filled after we get the reply */
295 mdc_set_capa_size(req, &RMF_CAPA2, op_data->op_capa1);
297 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
298 op_data->op_namelen + 1);
299 req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
300 max(lmmsize, obddev->u.cli.cl_default_mds_easize));
302 rc = ldlm_prep_enqueue_req(exp, req, &cancels, count);
304 ptlrpc_request_free(req);
308 spin_lock(&req->rq_lock);
309 req->rq_replay = req->rq_import->imp_replayable;
310 spin_unlock(&req->rq_lock);
312 /* pack the intent */
313 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
314 lit->opc = (__u64)it->it_op;
316 /* pack the intended request */
317 mdc_open_pack(req, op_data, it->it_create_mode, 0, it->it_flags, lmm,
320 /* for remote client, fetch remote perm for current user */
321 if (client_is_remote(exp))
322 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
323 sizeof(struct mdt_remote_perm));
324 ptlrpc_request_set_replen(req);
328 static struct ptlrpc_request *mdc_intent_unlink_pack(struct obd_export *exp,
329 struct lookup_intent *it,
330 struct md_op_data *op_data)
332 struct ptlrpc_request *req;
333 struct obd_device *obddev = class_exp2obd(exp);
334 struct ldlm_intent *lit;
338 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
339 &RQF_LDLM_INTENT_UNLINK);
341 RETURN(ERR_PTR(-ENOMEM));
343 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
344 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
345 op_data->op_namelen + 1);
347 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
349 ptlrpc_request_free(req);
353 /* pack the intent */
354 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
355 lit->opc = (__u64)it->it_op;
357 /* pack the intended request */
358 mdc_unlink_pack(req, op_data);
360 req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
361 obddev->u.cli.cl_max_mds_easize);
362 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
363 obddev->u.cli.cl_max_mds_cookiesize);
364 ptlrpc_request_set_replen(req);
368 static struct ptlrpc_request *mdc_intent_getattr_pack(struct obd_export *exp,
369 struct lookup_intent *it,
370 struct md_op_data *op_data)
372 struct ptlrpc_request *req;
373 struct obd_device *obddev = class_exp2obd(exp);
374 obd_valid valid = OBD_MD_FLGETATTR | OBD_MD_FLEASIZE |
375 OBD_MD_FLMODEASIZE | OBD_MD_FLDIREA |
376 OBD_MD_FLMDSCAPA | OBD_MD_MEA |
377 (client_is_remote(exp) ?
378 OBD_MD_FLRMTPERM : OBD_MD_FLACL);
379 struct ldlm_intent *lit;
383 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
384 &RQF_LDLM_INTENT_GETATTR);
386 RETURN(ERR_PTR(-ENOMEM));
388 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
389 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
390 op_data->op_namelen + 1);
392 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
394 ptlrpc_request_free(req);
398 /* pack the intent */
399 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
400 lit->opc = (__u64)it->it_op;
402 /* pack the intended request */
403 mdc_getattr_pack(req, valid, it->it_flags, op_data);
405 req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
406 obddev->u.cli.cl_max_mds_easize);
407 if (client_is_remote(exp))
408 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
409 sizeof(struct mdt_remote_perm));
410 ptlrpc_request_set_replen(req);
414 static struct ptlrpc_request *ldlm_enqueue_pack(struct obd_export *exp)
416 struct ptlrpc_request *req;
420 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_LDLM_ENQUEUE);
422 RETURN(ERR_PTR(-ENOMEM));
424 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
426 ptlrpc_request_free(req);
430 ptlrpc_request_set_replen(req);
434 static int mdc_finish_enqueue(struct obd_export *exp,
435 struct ptlrpc_request *req,
436 struct ldlm_enqueue_info *einfo,
437 struct lookup_intent *it,
438 struct lustre_handle *lockh,
441 struct req_capsule *pill = &req->rq_pill;
442 struct ldlm_request *lockreq;
443 struct ldlm_reply *lockrep;
447 /* Similarly, if we're going to replay this request, we don't want to
448 * actually get a lock, just perform the intent. */
449 if (req->rq_transno || req->rq_replay) {
450 lockreq = req_capsule_client_get(pill, &RMF_DLM_REQ);
451 lockreq->lock_flags |= LDLM_FL_INTENT_ONLY;
454 if (rc == ELDLM_LOCK_ABORTED) {
456 memset(lockh, 0, sizeof(*lockh));
458 } else { /* rc = 0 */
459 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
462 /* If the server gave us back a different lock mode, we should
463 * fix up our variables. */
464 if (lock->l_req_mode != einfo->ei_mode) {
465 ldlm_lock_addref(lockh, lock->l_req_mode);
466 ldlm_lock_decref(lockh, einfo->ei_mode);
467 einfo->ei_mode = lock->l_req_mode;
472 lockrep = req_capsule_server_get(pill, &RMF_DLM_REP);
473 LASSERT(lockrep != NULL); /* checked by ldlm_cli_enqueue() */
475 it->d.lustre.it_disposition = (int)lockrep->lock_policy_res1;
476 it->d.lustre.it_status = (int)lockrep->lock_policy_res2;
477 it->d.lustre.it_lock_mode = einfo->ei_mode;
478 it->d.lustre.it_lock_handle = lockh->cookie;
479 it->d.lustre.it_data = req;
481 if (it->d.lustre.it_status < 0 && req->rq_replay)
482 mdc_clear_replay_flag(req, it->d.lustre.it_status);
484 /* If we're doing an IT_OPEN which did not result in an actual
485 * successful open, then we need to remove the bit which saves
486 * this request for unconditional replay.
488 * It's important that we do this first! Otherwise we might exit the
489 * function without doing so, and try to replay a failed create
491 if (it->it_op & IT_OPEN && req->rq_replay &&
492 (!it_disposition(it, DISP_OPEN_OPEN) ||it->d.lustre.it_status != 0))
493 mdc_clear_replay_flag(req, it->d.lustre.it_status);
495 DEBUG_REQ(D_RPCTRACE, req, "op: %d disposition: %x, status: %d",
496 it->it_op,it->d.lustre.it_disposition,it->d.lustre.it_status);
498 /* We know what to expect, so we do any byte flipping required here */
499 if (it->it_op & (IT_OPEN | IT_UNLINK | IT_LOOKUP | IT_GETATTR)) {
500 struct mdt_body *body;
502 body = req_capsule_server_get(pill, &RMF_MDT_BODY);
504 CERROR ("Can't swab mdt_body\n");
508 if (it_disposition(it, DISP_OPEN_OPEN) &&
509 !it_open_error(DISP_OPEN_OPEN, it)) {
511 * If this is a successful OPEN request, we need to set
512 * replay handler and data early, so that if replay
513 * happens immediately after swabbing below, new reply
514 * is swabbed by that handler correctly.
516 mdc_set_open_replay_data(NULL, NULL, req);
519 if ((body->valid & (OBD_MD_FLDIREA | OBD_MD_FLEASIZE)) != 0) {
522 mdc_update_max_ea_from_body(exp, body);
525 * The eadata is opaque; just check that it is there.
526 * Eventually, obd_unpackmd() will check the contents.
528 eadata = req_capsule_server_sized_get(pill, &RMF_MDT_MD,
534 * We save the reply LOV EA in case we have to replay a
535 * create for recovery. If we didn't allocate a large
536 * enough request buffer above we need to reallocate it
537 * here to hold the actual LOV EA.
539 * To not save LOV EA if request is not going to replay
540 * (for example error one).
542 if ((it->it_op & IT_OPEN) && req->rq_replay) {
544 if (req_capsule_get_size(pill, &RMF_EADATA,
547 mdc_realloc_openmsg(req, body);
549 req_capsule_shrink(pill, &RMF_EADATA,
553 req_capsule_set_size(pill, &RMF_EADATA,
557 lmm = req_capsule_client_get(pill, &RMF_EADATA);
559 memcpy(lmm, eadata, body->eadatasize);
563 if (body->valid & OBD_MD_FLRMTPERM) {
564 struct mdt_remote_perm *perm;
566 LASSERT(client_is_remote(exp));
567 perm = req_capsule_server_swab_get(pill, &RMF_ACL,
568 lustre_swab_mdt_remote_perm);
572 if (body->valid & OBD_MD_FLMDSCAPA) {
573 struct lustre_capa *capa, *p;
575 capa = req_capsule_server_get(pill, &RMF_CAPA1);
579 if (it->it_op & IT_OPEN) {
580 /* client fid capa will be checked in replay */
581 p = req_capsule_client_get(pill, &RMF_CAPA2);
586 if (body->valid & OBD_MD_FLOSSCAPA) {
587 struct lustre_capa *capa;
589 capa = req_capsule_server_get(pill, &RMF_CAPA2);
598 /* We always reserve enough space in the reply packet for a stripe MD, because
599 * we don't know in advance the file type. */
600 int mdc_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
601 struct lookup_intent *it, struct md_op_data *op_data,
602 struct lustre_handle *lockh, void *lmm, int lmmsize,
603 struct ptlrpc_request **reqp, int extra_lock_flags)
605 struct obd_device *obddev = class_exp2obd(exp);
606 struct ptlrpc_request *req = NULL;
607 struct req_capsule *pill;
608 int flags = extra_lock_flags;
610 struct ldlm_res_id res_id;
611 ldlm_policy_data_t policy = { .l_inodebits = { MDS_INODELOCK_LOOKUP } };
614 LASSERTF(!it || einfo->ei_type == LDLM_IBITS, "lock type %d\n",
617 fid_build_reg_res_name(&op_data->op_fid1, &res_id);
620 flags |= LDLM_FL_HAS_INTENT;
621 if (it && it->it_op & (IT_UNLINK | IT_GETATTR | IT_READDIR))
622 policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
628 /* The only way right now is FLOCK, in this case we hide flock
629 policy as lmm, but lmmsize is 0 */
630 LASSERT(lmm && lmmsize == 0);
631 LASSERTF(einfo->ei_type == LDLM_FLOCK, "lock type %d\n",
633 policy = *(ldlm_policy_data_t *)lmm;
634 res_id.name[3] = LDLM_FLOCK;
635 } else if (it->it_op & IT_OPEN) {
636 req = mdc_intent_open_pack(exp, it, op_data, lmm, lmmsize,
638 policy.l_inodebits.bits = MDS_INODELOCK_UPDATE;
639 einfo->ei_cbdata = NULL;
641 } else if (it->it_op & IT_UNLINK)
642 req = mdc_intent_unlink_pack(exp, it, op_data);
643 else if (it->it_op & (IT_GETATTR | IT_LOOKUP))
644 req = mdc_intent_getattr_pack(exp, it, op_data);
645 else if (it->it_op == IT_READDIR)
646 req = ldlm_enqueue_pack(exp);
653 RETURN(PTR_ERR(req));
654 pill = &req->rq_pill;
656 /* It is important to obtain rpc_lock first (if applicable), so that
657 * threads that are serialised with rpc_lock are not polluting our
658 * rpcs in flight counter. We do not do flock request limiting, though*/
660 mdc_get_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
661 mdc_enter_request(&obddev->u.cli);
663 rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
669 mdc_exit_request(&obddev->u.cli);
670 mdc_put_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
673 /* For flock requests we immediatelly return without further
674 delay and let caller deal with the rest, since rest of
675 this function metadata processing makes no sense for flock
681 CERROR("ldlm_cli_enqueue: %d\n", rc);
682 mdc_clear_replay_flag(req, rc);
683 ptlrpc_req_finished(req);
686 rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
691 static int mdc_finish_intent_lock(struct obd_export *exp,
692 struct ptlrpc_request *request,
693 struct md_op_data *op_data,
694 struct lookup_intent *it,
695 struct lustre_handle *lockh)
697 struct lustre_handle old_lock;
698 struct mdt_body *mdt_body;
699 struct ldlm_lock *lock;
703 LASSERT(request != NULL);
704 LASSERT(request != LP_POISON);
705 LASSERT(request->rq_repmsg != LP_POISON);
707 if (!it_disposition(it, DISP_IT_EXECD)) {
708 /* The server failed before it even started executing the
709 * intent, i.e. because it couldn't unpack the request. */
710 LASSERT(it->d.lustre.it_status != 0);
711 RETURN(it->d.lustre.it_status);
713 rc = it_open_error(DISP_IT_EXECD, it);
717 mdt_body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
718 LASSERT(mdt_body != NULL); /* mdc_enqueue checked */
720 /* If we were revalidating a fid/name pair, mark the intent in
721 * case we fail and get called again from lookup */
722 if (fid_is_sane(&op_data->op_fid2) &&
723 it->it_create_mode & M_CHECK_STALE &&
724 it->it_op != IT_GETATTR) {
725 it_set_disposition(it, DISP_ENQ_COMPLETE);
727 /* Also: did we find the same inode? */
728 /* sever can return one of two fids:
729 * op_fid2 - new allocated fid - if file is created.
730 * op_fid3 - existent fid - if file only open.
731 * op_fid3 is saved in lmv_intent_open */
732 if ((!lu_fid_eq(&op_data->op_fid2, &mdt_body->fid1)) &&
733 (!lu_fid_eq(&op_data->op_fid3, &mdt_body->fid1))) {
734 CDEBUG(D_DENTRY, "Found stale data "DFID"("DFID")/"DFID
735 "\n", PFID(&op_data->op_fid2),
736 PFID(&op_data->op_fid2), PFID(&mdt_body->fid1));
741 rc = it_open_error(DISP_LOOKUP_EXECD, it);
745 /* keep requests around for the multiple phases of the call
746 * this shows the DISP_XX must guarantee we make it into the call
748 if (!it_disposition(it, DISP_ENQ_CREATE_REF) &&
749 it_disposition(it, DISP_OPEN_CREATE) &&
750 !it_open_error(DISP_OPEN_CREATE, it)) {
751 it_set_disposition(it, DISP_ENQ_CREATE_REF);
752 ptlrpc_request_addref(request); /* balanced in ll_create_node */
754 if (!it_disposition(it, DISP_ENQ_OPEN_REF) &&
755 it_disposition(it, DISP_OPEN_OPEN) &&
756 !it_open_error(DISP_OPEN_OPEN, it)) {
757 it_set_disposition(it, DISP_ENQ_OPEN_REF);
758 ptlrpc_request_addref(request); /* balanced in ll_file_open */
759 /* BUG 11546 - eviction in the middle of open rpc processing */
760 OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_ENQUEUE_PAUSE, obd_timeout);
763 if (it->it_op & IT_CREAT) {
764 /* XXX this belongs in ll_create_it */
765 } else if (it->it_op == IT_OPEN) {
766 LASSERT(!it_disposition(it, DISP_OPEN_CREATE));
768 LASSERT(it->it_op & (IT_GETATTR | IT_LOOKUP));
771 /* If we already have a matching lock, then cancel the new
772 * one. We have to set the data here instead of in
773 * mdc_enqueue, because we need to use the child's inode as
774 * the l_ast_data to match, and that's not available until
775 * intent_finish has performed the iget().) */
776 lock = ldlm_handle2lock(lockh);
778 ldlm_policy_data_t policy = lock->l_policy_data;
779 LDLM_DEBUG(lock, "matching against this");
781 LASSERTF(fid_res_name_eq(&mdt_body->fid1,
782 &lock->l_resource->lr_name),
783 "Lock res_id: %lu/%lu/%lu, fid: %lu/%lu/%lu.\n",
784 (unsigned long)lock->l_resource->lr_name.name[0],
785 (unsigned long)lock->l_resource->lr_name.name[1],
786 (unsigned long)lock->l_resource->lr_name.name[2],
787 (unsigned long)fid_seq(&mdt_body->fid1),
788 (unsigned long)fid_oid(&mdt_body->fid1),
789 (unsigned long)fid_ver(&mdt_body->fid1));
792 memcpy(&old_lock, lockh, sizeof(*lockh));
793 if (ldlm_lock_match(NULL, LDLM_FL_BLOCK_GRANTED, NULL,
794 LDLM_IBITS, &policy, LCK_NL, &old_lock, 0)) {
795 ldlm_lock_decref_and_cancel(lockh,
796 it->d.lustre.it_lock_mode);
797 memcpy(lockh, &old_lock, sizeof(old_lock));
798 it->d.lustre.it_lock_handle = lockh->cookie;
801 CDEBUG(D_DENTRY,"D_IT dentry %.*s intent: %s status %d disp %x rc %d\n",
802 op_data->op_namelen, op_data->op_name, ldlm_it2str(it->it_op),
803 it->d.lustre.it_status, it->d.lustre.it_disposition, rc);
808 * This long block is all about fixing up the lock and request state
809 * so that it is correct as of the moment _before_ the operation was
810 * applied; that way, the VFS will think that everything is normal and
811 * call Lustre's regular VFS methods.
813 * If we're performing a creation, that means that unless the creation
814 * failed with EEXIST, we should fake up a negative dentry.
816 * For everything else, we want to lookup to succeed.
818 * One additional note: if CREATE or OPEN succeeded, we add an extra
819 * reference to the request because we need to keep it around until
820 * ll_create/ll_open gets called.
822 * The server will return to us, in it_disposition, an indication of
823 * exactly what d.lustre.it_status refers to.
825 * If DISP_OPEN_OPEN is set, then d.lustre.it_status refers to the open() call,
826 * otherwise if DISP_OPEN_CREATE is set, then it status is the
827 * creation failure mode. In either case, one of DISP_LOOKUP_NEG or
828 * DISP_LOOKUP_POS will be set, indicating whether the child lookup
831 * Else, if DISP_LOOKUP_EXECD then d.lustre.it_status is the rc of the
834 int mdc_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
835 void *lmm, int lmmsize, struct lookup_intent *it,
836 int lookup_flags, struct ptlrpc_request **reqp,
837 ldlm_blocking_callback cb_blocking,
838 int extra_lock_flags)
840 struct lustre_handle lockh;
845 CDEBUG(D_DLMTRACE, "(name: %.*s,"DFID") in obj "DFID
846 ", intent: %s flags %#o\n", op_data->op_namelen,
847 op_data->op_name, PFID(&op_data->op_fid2),
848 PFID(&op_data->op_fid1), ldlm_it2str(it->it_op),
852 if (fid_is_sane(&op_data->op_fid2) &&
853 (it->it_op & (IT_LOOKUP | IT_GETATTR))) {
854 /* We could just return 1 immediately, but since we should only
855 * be called in revalidate_it if we already have a lock, let's
857 ldlm_policy_data_t policy;
860 /* As not all attributes are kept under update lock, e.g.
861 owner/group/acls are under lookup lock, we need both
862 ibits for GETATTR. */
864 /* For CMD, UPDATE lock and LOOKUP lock can not be got
865 * at the same for cross-object, so we can not match
866 * the 2 lock at the same time FIXME: but how to handle
867 * the above situation */
868 policy.l_inodebits.bits = (it->it_op == IT_GETATTR) ?
869 MDS_INODELOCK_UPDATE : MDS_INODELOCK_LOOKUP;
871 mode = mdc_lock_match(exp, LDLM_FL_BLOCK_GRANTED,
872 &op_data->op_fid2, LDLM_IBITS, &policy,
873 LCK_CR|LCK_CW|LCK_PR|LCK_PW, &lockh);
875 it->d.lustre.it_lock_handle = lockh.cookie;
876 it->d.lustre.it_lock_mode = mode;
879 /* Only return failure if it was not GETATTR by cfid
880 (from inode_revalidate) */
881 if (mode || op_data->op_namelen != 0)
885 /* lookup_it may be called only after revalidate_it has run, because
886 * revalidate_it cannot return errors, only zero. Returning zero causes
887 * this call to lookup, which *can* return an error.
889 * We only want to execute the request associated with the intent one
890 * time, however, so don't send the request again. Instead, skip past
891 * this and use the request from revalidate. In this case, revalidate
892 * never dropped its reference, so the refcounts are all OK */
893 if (!it_disposition(it, DISP_ENQ_COMPLETE)) {
894 struct ldlm_enqueue_info einfo =
895 { LDLM_IBITS, it_to_lock_mode(it), cb_blocking,
896 ldlm_completion_ast, NULL, NULL, NULL };
898 /* For case if upper layer did not alloc fid, do it now. */
899 if (!fid_is_sane(&op_data->op_fid2) && it->it_op & IT_CREAT) {
900 rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
902 CERROR("Can't alloc new fid, rc %d\n", rc);
906 rc = mdc_enqueue(exp, &einfo, it, op_data, &lockh,
907 lmm, lmmsize, NULL, extra_lock_flags);
910 } else if (!fid_is_sane(&op_data->op_fid2) ||
911 !(it->it_create_mode & M_CHECK_STALE)) {
912 /* DISP_ENQ_COMPLETE set means there is extra reference on
913 * request referenced from this intent, saved for subsequent
914 * lookup. This path is executed when we proceed to this
915 * lookup, so we clear DISP_ENQ_COMPLETE */
916 it_clear_disposition(it, DISP_ENQ_COMPLETE);
918 *reqp = it->d.lustre.it_data;
919 rc = mdc_finish_intent_lock(exp, *reqp, op_data, it, &lockh);
923 static int mdc_intent_getattr_async_interpret(const struct lu_env *env,
924 struct ptlrpc_request *req,
925 void *unused, int rc)
927 struct obd_export *exp = req->rq_async_args.pointer_arg[0];
928 struct md_enqueue_info *minfo = req->rq_async_args.pointer_arg[1];
929 struct ldlm_enqueue_info *einfo = req->rq_async_args.pointer_arg[2];
930 struct lookup_intent *it;
931 struct lustre_handle *lockh;
932 struct obd_device *obddev;
933 int flags = LDLM_FL_HAS_INTENT;
937 lockh = &minfo->mi_lockh;
939 obddev = class_exp2obd(exp);
941 mdc_exit_request(&obddev->u.cli);
942 if (OBD_FAIL_CHECK(OBD_FAIL_MDC_GETATTR_ENQUEUE))
945 rc = ldlm_cli_enqueue_fini(exp, req, einfo->ei_type, 1, einfo->ei_mode,
946 &flags, NULL, 0, lockh, rc);
948 CERROR("ldlm_cli_enqueue_fini: %d\n", rc);
949 mdc_clear_replay_flag(req, rc);
953 rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
957 rc = mdc_finish_intent_lock(exp, req, &minfo->mi_data, it, lockh);
962 minfo->mi_cb(req, minfo, rc);
966 int mdc_intent_getattr_async(struct obd_export *exp,
967 struct md_enqueue_info *minfo,
968 struct ldlm_enqueue_info *einfo)
970 struct md_op_data *op_data = &minfo->mi_data;
971 struct lookup_intent *it = &minfo->mi_it;
972 struct ptlrpc_request *req;
973 struct obd_device *obddev = class_exp2obd(exp);
974 struct ldlm_res_id res_id;
975 ldlm_policy_data_t policy = {
976 .l_inodebits = { MDS_INODELOCK_LOOKUP }
979 int flags = LDLM_FL_HAS_INTENT;
982 CDEBUG(D_DLMTRACE,"name: %.*s in inode "DFID", intent: %s flags %#o\n",
983 op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
984 ldlm_it2str(it->it_op), it->it_flags);
986 fid_build_reg_res_name(&op_data->op_fid1, &res_id);
987 req = mdc_intent_getattr_pack(exp, it, op_data);
991 mdc_enter_request(&obddev->u.cli);
992 rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
993 0, &minfo->mi_lockh, 1);
995 mdc_exit_request(&obddev->u.cli);
999 req->rq_async_args.pointer_arg[0] = exp;
1000 req->rq_async_args.pointer_arg[1] = minfo;
1001 req->rq_async_args.pointer_arg[2] = einfo;
1002 req->rq_interpret_reply = mdc_intent_getattr_async_interpret;
1003 ptlrpcd_add_req(req, PSCOPE_OTHER);
1008 int mdc_revalidate_lock(struct obd_export *exp,
1009 struct lookup_intent *it,
1012 /* We could just return 1 immediately, but since we should only
1013 * be called in revalidate_it if we already have a lock, let's
1015 struct ldlm_res_id res_id;
1016 struct lustre_handle lockh;
1017 ldlm_policy_data_t policy;
1021 fid_build_reg_res_name(fid, &res_id);
1022 /* As not all attributes are kept under update lock, e.g.
1023 owner/group/acls are under lookup lock, we need both
1024 ibits for GETATTR. */
1025 policy.l_inodebits.bits = (it->it_op == IT_GETATTR) ?
1026 MDS_INODELOCK_UPDATE : MDS_INODELOCK_LOOKUP;
1028 mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
1029 LDLM_FL_BLOCK_GRANTED, &res_id, LDLM_IBITS,
1030 &policy, LCK_CR|LCK_CW|LCK_PR|LCK_PW, &lockh, 0);
1032 it->d.lustre.it_lock_handle = lockh.cookie;
1033 it->d.lustre.it_lock_mode = mode;