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 (c) 2003, 2010, Oracle and/or its affiliates. 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 struct mdc_getattr_args {
60 struct obd_export *ga_exp;
61 struct md_enqueue_info *ga_minfo;
62 struct ldlm_enqueue_info *ga_einfo;
65 int it_disposition(struct lookup_intent *it, int flag)
67 return it->d.lustre.it_disposition & flag;
69 EXPORT_SYMBOL(it_disposition);
71 void it_set_disposition(struct lookup_intent *it, int flag)
73 it->d.lustre.it_disposition |= flag;
75 EXPORT_SYMBOL(it_set_disposition);
77 void it_clear_disposition(struct lookup_intent *it, int flag)
79 it->d.lustre.it_disposition &= ~flag;
81 EXPORT_SYMBOL(it_clear_disposition);
83 int it_open_error(int phase, struct lookup_intent *it)
85 if (it_disposition(it, DISP_OPEN_OPEN)) {
86 if (phase >= DISP_OPEN_OPEN)
87 return it->d.lustre.it_status;
92 if (it_disposition(it, DISP_OPEN_CREATE)) {
93 if (phase >= DISP_OPEN_CREATE)
94 return it->d.lustre.it_status;
99 if (it_disposition(it, DISP_LOOKUP_EXECD)) {
100 if (phase >= DISP_LOOKUP_EXECD)
101 return it->d.lustre.it_status;
106 if (it_disposition(it, DISP_IT_EXECD)) {
107 if (phase >= DISP_IT_EXECD)
108 return it->d.lustre.it_status;
112 CERROR("it disp: %X, status: %d\n", it->d.lustre.it_disposition,
113 it->d.lustre.it_status);
117 EXPORT_SYMBOL(it_open_error);
119 /* this must be called on a lockh that is known to have a referenced lock */
120 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
123 struct ldlm_lock *lock;
132 lock = ldlm_handle2lock((struct lustre_handle *)lockh);
134 LASSERT(lock != NULL);
135 lock_res_and_lock(lock);
137 if (lock->l_ast_data && lock->l_ast_data != data) {
138 struct inode *new_inode = data;
139 struct inode *old_inode = lock->l_ast_data;
140 LASSERTF(old_inode->i_state & I_FREEING,
141 "Found existing inode %p/%lu/%u state %lu in lock: "
142 "setting data to %p/%lu/%u\n", old_inode,
143 old_inode->i_ino, old_inode->i_generation,
145 new_inode, new_inode->i_ino, new_inode->i_generation);
148 lock->l_ast_data = data;
150 *bits = lock->l_policy_data.l_inodebits.bits;
152 unlock_res_and_lock(lock);
158 ldlm_mode_t mdc_lock_match(struct obd_export *exp, int flags,
159 const struct lu_fid *fid, ldlm_type_t type,
160 ldlm_policy_data_t *policy, ldlm_mode_t mode,
161 struct lustre_handle *lockh)
163 struct ldlm_res_id res_id;
167 fid_build_reg_res_name(fid, &res_id);
168 rc = ldlm_lock_match(class_exp2obd(exp)->obd_namespace, flags,
169 &res_id, type, policy, mode, lockh, 0);
173 int mdc_cancel_unused(struct obd_export *exp,
174 const struct lu_fid *fid,
175 ldlm_policy_data_t *policy,
177 ldlm_cancel_flags_t flags,
180 struct ldlm_res_id res_id;
181 struct obd_device *obd = class_exp2obd(exp);
186 fid_build_reg_res_name(fid, &res_id);
187 rc = ldlm_cli_cancel_unused_resource(obd->obd_namespace, &res_id,
188 policy, mode, flags, opaque);
192 int mdc_change_cbdata(struct obd_export *exp,
193 const struct lu_fid *fid,
194 ldlm_iterator_t it, void *data)
196 struct ldlm_res_id res_id;
199 fid_build_reg_res_name(fid, &res_id);
200 ldlm_resource_iterate(class_exp2obd(exp)->obd_namespace,
207 /* find any ldlm lock of the inode in mdc
211 int mdc_find_cbdata(struct obd_export *exp,
212 const struct lu_fid *fid,
213 ldlm_iterator_t it, void *data)
215 struct ldlm_res_id res_id;
219 fid_build_reg_res_name((struct lu_fid*)fid, &res_id);
220 rc = ldlm_resource_iterate(class_exp2obd(exp)->obd_namespace, &res_id,
222 if (rc == LDLM_ITER_STOP)
224 else if (rc == LDLM_ITER_CONTINUE)
229 static inline void mdc_clear_replay_flag(struct ptlrpc_request *req, int rc)
231 /* Don't hold error requests for replay. */
232 if (req->rq_replay) {
233 cfs_spin_lock(&req->rq_lock);
235 cfs_spin_unlock(&req->rq_lock);
237 if (rc && req->rq_transno != 0) {
238 DEBUG_REQ(D_ERROR, req, "transno returned on error rc %d", rc);
243 /* Save a large LOV EA into the request buffer so that it is available
244 * for replay. We don't do this in the initial request because the
245 * original request doesn't need this buffer (at most it sends just the
246 * lov_mds_md) and it is a waste of RAM/bandwidth to send the empty
247 * buffer and may also be difficult to allocate and save a very large
248 * request buffer for each open. (bug 5707)
250 * OOM here may cause recovery failure if lmm is needed (only for the
251 * original open if the MDS crashed just when this client also OOM'd)
252 * but this is incredibly unlikely, and questionable whether the client
253 * could do MDS recovery under OOM anyways... */
254 static void mdc_realloc_openmsg(struct ptlrpc_request *req,
255 struct mdt_body *body)
259 /* FIXME: remove this explicit offset. */
260 rc = sptlrpc_cli_enlarge_reqbuf(req, DLM_INTENT_REC_OFF + 4,
263 CERROR("Can't enlarge segment %d size to %d\n",
264 DLM_INTENT_REC_OFF + 4, body->eadatasize);
265 body->valid &= ~OBD_MD_FLEASIZE;
266 body->eadatasize = 0;
270 static struct ptlrpc_request *mdc_intent_open_pack(struct obd_export *exp,
271 struct lookup_intent *it,
272 struct md_op_data *op_data,
273 void *lmm, int lmmsize,
276 struct ptlrpc_request *req;
277 struct obd_device *obddev = class_exp2obd(exp);
278 struct ldlm_intent *lit;
279 CFS_LIST_HEAD(cancels);
285 it->it_create_mode = (it->it_create_mode & ~S_IFMT) | S_IFREG;
287 /* XXX: openlock is not cancelled for cross-refs. */
288 /* If inode is known, cancel conflicting OPEN locks. */
289 if (fid_is_sane(&op_data->op_fid2)) {
290 if (it->it_flags & (FMODE_WRITE|MDS_OPEN_TRUNC))
293 else if (it->it_flags & FMODE_EXEC)
298 count = mdc_resource_get_unused(exp, &op_data->op_fid2,
303 /* If CREATE, cancel parent's UPDATE lock. */
304 if (it->it_op & IT_CREAT)
308 count += mdc_resource_get_unused(exp, &op_data->op_fid1,
310 MDS_INODELOCK_UPDATE);
312 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
313 &RQF_LDLM_INTENT_OPEN);
315 ldlm_lock_list_put(&cancels, l_bl_ast, count);
316 RETURN(ERR_PTR(-ENOMEM));
319 /* parent capability */
320 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
321 /* child capability, reserve the size according to parent capa, it will
322 * be filled after we get the reply */
323 mdc_set_capa_size(req, &RMF_CAPA2, op_data->op_capa1);
325 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
326 op_data->op_namelen + 1);
327 req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
328 max(lmmsize, obddev->u.cli.cl_default_mds_easize));
330 rc = ldlm_prep_enqueue_req(exp, req, &cancels, count);
332 ptlrpc_request_free(req);
336 cfs_spin_lock(&req->rq_lock);
337 req->rq_replay = req->rq_import->imp_replayable;
338 cfs_spin_unlock(&req->rq_lock);
340 /* pack the intent */
341 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
342 lit->opc = (__u64)it->it_op;
344 /* pack the intended request */
345 mdc_open_pack(req, op_data, it->it_create_mode, 0, it->it_flags, lmm,
348 /* for remote client, fetch remote perm for current user */
349 if (client_is_remote(exp))
350 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
351 sizeof(struct mdt_remote_perm));
352 ptlrpc_request_set_replen(req);
356 static struct ptlrpc_request *mdc_intent_unlink_pack(struct obd_export *exp,
357 struct lookup_intent *it,
358 struct md_op_data *op_data)
360 struct ptlrpc_request *req;
361 struct obd_device *obddev = class_exp2obd(exp);
362 struct ldlm_intent *lit;
366 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
367 &RQF_LDLM_INTENT_UNLINK);
369 RETURN(ERR_PTR(-ENOMEM));
371 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
372 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
373 op_data->op_namelen + 1);
375 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
377 ptlrpc_request_free(req);
381 /* pack the intent */
382 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
383 lit->opc = (__u64)it->it_op;
385 /* pack the intended request */
386 mdc_unlink_pack(req, op_data);
388 req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
389 obddev->u.cli.cl_max_mds_easize);
390 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
391 obddev->u.cli.cl_max_mds_cookiesize);
392 ptlrpc_request_set_replen(req);
396 static struct ptlrpc_request *mdc_intent_getattr_pack(struct obd_export *exp,
397 struct lookup_intent *it,
398 struct md_op_data *op_data)
400 struct ptlrpc_request *req;
401 struct obd_device *obddev = class_exp2obd(exp);
402 obd_valid valid = OBD_MD_FLGETATTR | OBD_MD_FLEASIZE |
403 OBD_MD_FLMODEASIZE | OBD_MD_FLDIREA |
404 OBD_MD_FLMDSCAPA | OBD_MD_MEA |
405 (client_is_remote(exp) ?
406 OBD_MD_FLRMTPERM : OBD_MD_FLACL);
407 struct ldlm_intent *lit;
411 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
412 &RQF_LDLM_INTENT_GETATTR);
414 RETURN(ERR_PTR(-ENOMEM));
416 mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
417 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
418 op_data->op_namelen + 1);
420 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
422 ptlrpc_request_free(req);
426 /* pack the intent */
427 lit = req_capsule_client_get(&req->rq_pill, &RMF_LDLM_INTENT);
428 lit->opc = (__u64)it->it_op;
430 /* pack the intended request */
431 mdc_getattr_pack(req, valid, it->it_flags, op_data,
432 obddev->u.cli.cl_max_mds_easize);
434 req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
435 obddev->u.cli.cl_max_mds_easize);
436 if (client_is_remote(exp))
437 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
438 sizeof(struct mdt_remote_perm));
439 ptlrpc_request_set_replen(req);
443 static struct ptlrpc_request *ldlm_enqueue_pack(struct obd_export *exp)
445 struct ptlrpc_request *req;
449 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_LDLM_ENQUEUE);
451 RETURN(ERR_PTR(-ENOMEM));
453 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
455 ptlrpc_request_free(req);
459 ptlrpc_request_set_replen(req);
463 static int mdc_finish_enqueue(struct obd_export *exp,
464 struct ptlrpc_request *req,
465 struct ldlm_enqueue_info *einfo,
466 struct lookup_intent *it,
467 struct lustre_handle *lockh,
470 struct req_capsule *pill = &req->rq_pill;
471 struct ldlm_request *lockreq;
472 struct ldlm_reply *lockrep;
476 /* Similarly, if we're going to replay this request, we don't want to
477 * actually get a lock, just perform the intent. */
478 if (req->rq_transno || req->rq_replay) {
479 lockreq = req_capsule_client_get(pill, &RMF_DLM_REQ);
480 lockreq->lock_flags |= LDLM_FL_INTENT_ONLY;
483 if (rc == ELDLM_LOCK_ABORTED) {
485 memset(lockh, 0, sizeof(*lockh));
487 } else { /* rc = 0 */
488 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
491 /* If the server gave us back a different lock mode, we should
492 * fix up our variables. */
493 if (lock->l_req_mode != einfo->ei_mode) {
494 ldlm_lock_addref(lockh, lock->l_req_mode);
495 ldlm_lock_decref(lockh, einfo->ei_mode);
496 einfo->ei_mode = lock->l_req_mode;
501 lockrep = req_capsule_server_get(pill, &RMF_DLM_REP);
502 LASSERT(lockrep != NULL); /* checked by ldlm_cli_enqueue() */
504 it->d.lustre.it_disposition = (int)lockrep->lock_policy_res1;
505 it->d.lustre.it_status = (int)lockrep->lock_policy_res2;
506 it->d.lustre.it_lock_mode = einfo->ei_mode;
507 it->d.lustre.it_lock_handle = lockh->cookie;
508 it->d.lustre.it_data = req;
510 if (it->d.lustre.it_status < 0 && req->rq_replay)
511 mdc_clear_replay_flag(req, it->d.lustre.it_status);
513 /* If we're doing an IT_OPEN which did not result in an actual
514 * successful open, then we need to remove the bit which saves
515 * this request for unconditional replay.
517 * It's important that we do this first! Otherwise we might exit the
518 * function without doing so, and try to replay a failed create
520 if (it->it_op & IT_OPEN && req->rq_replay &&
521 (!it_disposition(it, DISP_OPEN_OPEN) ||it->d.lustre.it_status != 0))
522 mdc_clear_replay_flag(req, it->d.lustre.it_status);
524 DEBUG_REQ(D_RPCTRACE, req, "op: %d disposition: %x, status: %d",
525 it->it_op,it->d.lustre.it_disposition,it->d.lustre.it_status);
527 /* We know what to expect, so we do any byte flipping required here */
528 if (it->it_op & (IT_OPEN | IT_UNLINK | IT_LOOKUP | IT_GETATTR)) {
529 struct mdt_body *body;
531 body = req_capsule_server_get(pill, &RMF_MDT_BODY);
533 CERROR ("Can't swab mdt_body\n");
537 if (it_disposition(it, DISP_OPEN_OPEN) &&
538 !it_open_error(DISP_OPEN_OPEN, it)) {
540 * If this is a successful OPEN request, we need to set
541 * replay handler and data early, so that if replay
542 * happens immediately after swabbing below, new reply
543 * is swabbed by that handler correctly.
545 mdc_set_open_replay_data(NULL, NULL, req);
548 if ((body->valid & (OBD_MD_FLDIREA | OBD_MD_FLEASIZE)) != 0) {
551 mdc_update_max_ea_from_body(exp, body);
554 * The eadata is opaque; just check that it is there.
555 * Eventually, obd_unpackmd() will check the contents.
557 eadata = req_capsule_server_sized_get(pill, &RMF_MDT_MD,
563 * We save the reply LOV EA in case we have to replay a
564 * create for recovery. If we didn't allocate a large
565 * enough request buffer above we need to reallocate it
566 * here to hold the actual LOV EA.
568 * To not save LOV EA if request is not going to replay
569 * (for example error one).
571 if ((it->it_op & IT_OPEN) && req->rq_replay) {
573 if (req_capsule_get_size(pill, &RMF_EADATA,
576 mdc_realloc_openmsg(req, body);
578 req_capsule_shrink(pill, &RMF_EADATA,
582 req_capsule_set_size(pill, &RMF_EADATA,
586 lmm = req_capsule_client_get(pill, &RMF_EADATA);
588 memcpy(lmm, eadata, body->eadatasize);
592 if (body->valid & OBD_MD_FLRMTPERM) {
593 struct mdt_remote_perm *perm;
595 LASSERT(client_is_remote(exp));
596 perm = req_capsule_server_swab_get(pill, &RMF_ACL,
597 lustre_swab_mdt_remote_perm);
601 if (body->valid & OBD_MD_FLMDSCAPA) {
602 struct lustre_capa *capa, *p;
604 capa = req_capsule_server_get(pill, &RMF_CAPA1);
608 if (it->it_op & IT_OPEN) {
609 /* client fid capa will be checked in replay */
610 p = req_capsule_client_get(pill, &RMF_CAPA2);
615 if (body->valid & OBD_MD_FLOSSCAPA) {
616 struct lustre_capa *capa;
618 capa = req_capsule_server_get(pill, &RMF_CAPA2);
627 /* We always reserve enough space in the reply packet for a stripe MD, because
628 * we don't know in advance the file type. */
629 int mdc_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
630 struct lookup_intent *it, struct md_op_data *op_data,
631 struct lustre_handle *lockh, void *lmm, int lmmsize,
632 struct ptlrpc_request **reqp, int extra_lock_flags)
634 struct obd_device *obddev = class_exp2obd(exp);
635 struct ptlrpc_request *req = NULL;
636 int flags = extra_lock_flags;
638 struct ldlm_res_id res_id;
639 static const ldlm_policy_data_t lookup_policy =
640 { .l_inodebits = { MDS_INODELOCK_LOOKUP } };
641 static const ldlm_policy_data_t update_policy =
642 { .l_inodebits = { MDS_INODELOCK_UPDATE } };
643 ldlm_policy_data_t const *policy = &lookup_policy;
646 LASSERTF(!it || einfo->ei_type == LDLM_IBITS, "lock type %d\n",
649 fid_build_reg_res_name(&op_data->op_fid1, &res_id);
652 flags |= LDLM_FL_HAS_INTENT;
653 if (it && it->it_op & (IT_UNLINK | IT_GETATTR | IT_READDIR))
654 policy = &update_policy;
660 /* The only way right now is FLOCK, in this case we hide flock
661 policy as lmm, but lmmsize is 0 */
662 LASSERT(lmm && lmmsize == 0);
663 LASSERTF(einfo->ei_type == LDLM_FLOCK, "lock type %d\n",
665 policy = (ldlm_policy_data_t *)lmm;
666 res_id.name[3] = LDLM_FLOCK;
667 } else if (it->it_op & IT_OPEN) {
668 req = mdc_intent_open_pack(exp, it, op_data, lmm, lmmsize,
670 policy = &update_policy;
671 einfo->ei_cbdata = NULL;
673 } else if (it->it_op & IT_UNLINK)
674 req = mdc_intent_unlink_pack(exp, it, op_data);
675 else if (it->it_op & (IT_GETATTR | IT_LOOKUP))
676 req = mdc_intent_getattr_pack(exp, it, op_data);
677 else if (it->it_op == IT_READDIR)
678 req = ldlm_enqueue_pack(exp);
685 RETURN(PTR_ERR(req));
687 /* It is important to obtain rpc_lock first (if applicable), so that
688 * threads that are serialised with rpc_lock are not polluting our
689 * rpcs in flight counter. We do not do flock request limiting, though*/
691 mdc_get_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
692 rc = mdc_enter_request(&obddev->u.cli);
694 mdc_put_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
695 mdc_clear_replay_flag(req, 0);
696 ptlrpc_req_finished(req);
701 rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, policy, &flags, NULL,
707 mdc_exit_request(&obddev->u.cli);
708 mdc_put_rpc_lock(obddev->u.cli.cl_rpc_lock, it);
711 /* For flock requests we immediatelly return without further
712 delay and let caller deal with the rest, since rest of
713 this function metadata processing makes no sense for flock
719 CERROR("ldlm_cli_enqueue: %d\n", rc);
720 mdc_clear_replay_flag(req, rc);
721 ptlrpc_req_finished(req);
724 rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
729 static int mdc_finish_intent_lock(struct obd_export *exp,
730 struct ptlrpc_request *request,
731 struct md_op_data *op_data,
732 struct lookup_intent *it,
733 struct lustre_handle *lockh)
735 struct lustre_handle old_lock;
736 struct mdt_body *mdt_body;
737 struct ldlm_lock *lock;
741 LASSERT(request != NULL);
742 LASSERT(request != LP_POISON);
743 LASSERT(request->rq_repmsg != LP_POISON);
745 if (!it_disposition(it, DISP_IT_EXECD)) {
746 /* The server failed before it even started executing the
747 * intent, i.e. because it couldn't unpack the request. */
748 LASSERT(it->d.lustre.it_status != 0);
749 RETURN(it->d.lustre.it_status);
751 rc = it_open_error(DISP_IT_EXECD, it);
755 mdt_body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY);
756 LASSERT(mdt_body != NULL); /* mdc_enqueue checked */
758 /* If we were revalidating a fid/name pair, mark the intent in
759 * case we fail and get called again from lookup */
760 if (fid_is_sane(&op_data->op_fid2) &&
761 it->it_create_mode & M_CHECK_STALE &&
762 it->it_op != IT_GETATTR) {
763 it_set_disposition(it, DISP_ENQ_COMPLETE);
765 /* Also: did we find the same inode? */
766 /* sever can return one of two fids:
767 * op_fid2 - new allocated fid - if file is created.
768 * op_fid3 - existent fid - if file only open.
769 * op_fid3 is saved in lmv_intent_open */
770 if ((!lu_fid_eq(&op_data->op_fid2, &mdt_body->fid1)) &&
771 (!lu_fid_eq(&op_data->op_fid3, &mdt_body->fid1))) {
772 CDEBUG(D_DENTRY, "Found stale data "DFID"("DFID")/"DFID
773 "\n", PFID(&op_data->op_fid2),
774 PFID(&op_data->op_fid2), PFID(&mdt_body->fid1));
779 rc = it_open_error(DISP_LOOKUP_EXECD, it);
783 /* keep requests around for the multiple phases of the call
784 * this shows the DISP_XX must guarantee we make it into the call
786 if (!it_disposition(it, DISP_ENQ_CREATE_REF) &&
787 it_disposition(it, DISP_OPEN_CREATE) &&
788 !it_open_error(DISP_OPEN_CREATE, it)) {
789 it_set_disposition(it, DISP_ENQ_CREATE_REF);
790 ptlrpc_request_addref(request); /* balanced in ll_create_node */
792 if (!it_disposition(it, DISP_ENQ_OPEN_REF) &&
793 it_disposition(it, DISP_OPEN_OPEN) &&
794 !it_open_error(DISP_OPEN_OPEN, it)) {
795 it_set_disposition(it, DISP_ENQ_OPEN_REF);
796 ptlrpc_request_addref(request); /* balanced in ll_file_open */
797 /* BUG 11546 - eviction in the middle of open rpc processing */
798 OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_ENQUEUE_PAUSE, obd_timeout);
801 if (it->it_op & IT_CREAT) {
802 /* XXX this belongs in ll_create_it */
803 } else if (it->it_op == IT_OPEN) {
804 LASSERT(!it_disposition(it, DISP_OPEN_CREATE));
806 LASSERT(it->it_op & (IT_GETATTR | IT_LOOKUP));
809 /* If we already have a matching lock, then cancel the new
810 * one. We have to set the data here instead of in
811 * mdc_enqueue, because we need to use the child's inode as
812 * the l_ast_data to match, and that's not available until
813 * intent_finish has performed the iget().) */
814 lock = ldlm_handle2lock(lockh);
816 ldlm_policy_data_t policy = lock->l_policy_data;
817 LDLM_DEBUG(lock, "matching against this");
819 LASSERTF(fid_res_name_eq(&mdt_body->fid1,
820 &lock->l_resource->lr_name),
821 "Lock res_id: %lu/%lu/%lu, fid: %lu/%lu/%lu.\n",
822 (unsigned long)lock->l_resource->lr_name.name[0],
823 (unsigned long)lock->l_resource->lr_name.name[1],
824 (unsigned long)lock->l_resource->lr_name.name[2],
825 (unsigned long)fid_seq(&mdt_body->fid1),
826 (unsigned long)fid_oid(&mdt_body->fid1),
827 (unsigned long)fid_ver(&mdt_body->fid1));
830 memcpy(&old_lock, lockh, sizeof(*lockh));
831 if (ldlm_lock_match(NULL, LDLM_FL_BLOCK_GRANTED, NULL,
832 LDLM_IBITS, &policy, LCK_NL, &old_lock, 0)) {
833 ldlm_lock_decref_and_cancel(lockh,
834 it->d.lustre.it_lock_mode);
835 memcpy(lockh, &old_lock, sizeof(old_lock));
836 it->d.lustre.it_lock_handle = lockh->cookie;
839 CDEBUG(D_DENTRY,"D_IT dentry %.*s intent: %s status %d disp %x rc %d\n",
840 op_data->op_namelen, op_data->op_name, ldlm_it2str(it->it_op),
841 it->d.lustre.it_status, it->d.lustre.it_disposition, rc);
845 int mdc_revalidate_lock(struct obd_export *exp, struct lookup_intent *it,
846 struct lu_fid *fid, __u64 *bits)
848 /* We could just return 1 immediately, but since we should only
849 * be called in revalidate_it if we already have a lock, let's
851 struct ldlm_res_id res_id;
852 struct lustre_handle lockh;
853 ldlm_policy_data_t policy;
857 if (it->d.lustre.it_lock_handle) {
858 lockh.cookie = it->d.lustre.it_lock_handle;
859 mode = ldlm_revalidate_lock_handle(&lockh, bits);
861 fid_build_reg_res_name(fid, &res_id);
862 policy.l_inodebits.bits = (it->it_op == IT_GETATTR) ?
863 MDS_INODELOCK_UPDATE : MDS_INODELOCK_LOOKUP;
864 mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
865 LDLM_FL_BLOCK_GRANTED, &res_id,
867 LCK_CR|LCK_CW|LCK_PR|LCK_PW, &lockh, 0);
871 it->d.lustre.it_lock_handle = lockh.cookie;
872 it->d.lustre.it_lock_mode = mode;
874 it->d.lustre.it_lock_handle = 0;
875 it->d.lustre.it_lock_mode = 0;
882 * This long block is all about fixing up the lock and request state
883 * so that it is correct as of the moment _before_ the operation was
884 * applied; that way, the VFS will think that everything is normal and
885 * call Lustre's regular VFS methods.
887 * If we're performing a creation, that means that unless the creation
888 * failed with EEXIST, we should fake up a negative dentry.
890 * For everything else, we want to lookup to succeed.
892 * One additional note: if CREATE or OPEN succeeded, we add an extra
893 * reference to the request because we need to keep it around until
894 * ll_create/ll_open gets called.
896 * The server will return to us, in it_disposition, an indication of
897 * exactly what d.lustre.it_status refers to.
899 * If DISP_OPEN_OPEN is set, then d.lustre.it_status refers to the open() call,
900 * otherwise if DISP_OPEN_CREATE is set, then it status is the
901 * creation failure mode. In either case, one of DISP_LOOKUP_NEG or
902 * DISP_LOOKUP_POS will be set, indicating whether the child lookup
905 * Else, if DISP_LOOKUP_EXECD then d.lustre.it_status is the rc of the
908 int mdc_intent_lock(struct obd_export *exp, struct md_op_data *op_data,
909 void *lmm, int lmmsize, struct lookup_intent *it,
910 int lookup_flags, struct ptlrpc_request **reqp,
911 ldlm_blocking_callback cb_blocking,
912 int extra_lock_flags)
914 struct lustre_handle lockh;
919 CDEBUG(D_DLMTRACE, "(name: %.*s,"DFID") in obj "DFID
920 ", intent: %s flags %#o\n", op_data->op_namelen,
921 op_data->op_name, PFID(&op_data->op_fid2),
922 PFID(&op_data->op_fid1), ldlm_it2str(it->it_op),
926 if (fid_is_sane(&op_data->op_fid2) &&
927 (it->it_op & (IT_LOOKUP | IT_GETATTR))) {
928 /* We could just return 1 immediately, but since we should only
929 * be called in revalidate_it if we already have a lock, let's
931 it->d.lustre.it_lock_handle = 0;
932 rc = mdc_revalidate_lock(exp, it, &op_data->op_fid2, NULL);
933 /* Only return failure if it was not GETATTR by cfid
934 (from inode_revalidate) */
935 if (rc || op_data->op_namelen != 0)
939 /* lookup_it may be called only after revalidate_it has run, because
940 * revalidate_it cannot return errors, only zero. Returning zero causes
941 * this call to lookup, which *can* return an error.
943 * We only want to execute the request associated with the intent one
944 * time, however, so don't send the request again. Instead, skip past
945 * this and use the request from revalidate. In this case, revalidate
946 * never dropped its reference, so the refcounts are all OK */
947 if (!it_disposition(it, DISP_ENQ_COMPLETE)) {
948 struct ldlm_enqueue_info einfo =
949 { LDLM_IBITS, it_to_lock_mode(it), cb_blocking,
950 ldlm_completion_ast, NULL, NULL, NULL };
952 /* For case if upper layer did not alloc fid, do it now. */
953 if (!fid_is_sane(&op_data->op_fid2) && it->it_op & IT_CREAT) {
954 rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
956 CERROR("Can't alloc new fid, rc %d\n", rc);
960 rc = mdc_enqueue(exp, &einfo, it, op_data, &lockh,
961 lmm, lmmsize, NULL, extra_lock_flags);
964 } else if (!fid_is_sane(&op_data->op_fid2) ||
965 !(it->it_create_mode & M_CHECK_STALE)) {
966 /* DISP_ENQ_COMPLETE set means there is extra reference on
967 * request referenced from this intent, saved for subsequent
968 * lookup. This path is executed when we proceed to this
969 * lookup, so we clear DISP_ENQ_COMPLETE */
970 it_clear_disposition(it, DISP_ENQ_COMPLETE);
972 *reqp = it->d.lustre.it_data;
973 rc = mdc_finish_intent_lock(exp, *reqp, op_data, it, &lockh);
977 static int mdc_intent_getattr_async_interpret(const struct lu_env *env,
978 struct ptlrpc_request *req,
981 struct mdc_getattr_args *ga = args;
982 struct obd_export *exp = ga->ga_exp;
983 struct md_enqueue_info *minfo = ga->ga_minfo;
984 struct ldlm_enqueue_info *einfo = ga->ga_einfo;
985 struct lookup_intent *it;
986 struct lustre_handle *lockh;
987 struct obd_device *obddev;
988 int flags = LDLM_FL_HAS_INTENT;
992 lockh = &minfo->mi_lockh;
994 obddev = class_exp2obd(exp);
996 mdc_exit_request(&obddev->u.cli);
997 if (OBD_FAIL_CHECK(OBD_FAIL_MDC_GETATTR_ENQUEUE))
1000 rc = ldlm_cli_enqueue_fini(exp, req, einfo->ei_type, 1, einfo->ei_mode,
1001 &flags, NULL, 0, lockh, rc);
1003 CERROR("ldlm_cli_enqueue_fini: %d\n", rc);
1004 mdc_clear_replay_flag(req, rc);
1008 rc = mdc_finish_enqueue(exp, req, einfo, it, lockh, rc);
1012 rc = mdc_finish_intent_lock(exp, req, &minfo->mi_data, it, lockh);
1016 OBD_FREE_PTR(einfo);
1017 minfo->mi_cb(req, minfo, rc);
1021 int mdc_intent_getattr_async(struct obd_export *exp,
1022 struct md_enqueue_info *minfo,
1023 struct ldlm_enqueue_info *einfo)
1025 struct md_op_data *op_data = &minfo->mi_data;
1026 struct lookup_intent *it = &minfo->mi_it;
1027 struct ptlrpc_request *req;
1028 struct mdc_getattr_args *ga;
1029 struct obd_device *obddev = class_exp2obd(exp);
1030 struct ldlm_res_id res_id;
1031 /*XXX: Both MDS_INODELOCK_LOOKUP and MDS_INODELOCK_UPDATE are needed
1032 * for statahead currently. Consider CMD in future, such two bits
1033 * maybe managed by different MDS, should be adjusted then. */
1034 ldlm_policy_data_t policy = {
1035 .l_inodebits = { MDS_INODELOCK_LOOKUP |
1036 MDS_INODELOCK_UPDATE }
1039 int flags = LDLM_FL_HAS_INTENT;
1042 CDEBUG(D_DLMTRACE,"name: %.*s in inode "DFID", intent: %s flags %#o\n",
1043 op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
1044 ldlm_it2str(it->it_op), it->it_flags);
1046 fid_build_reg_res_name(&op_data->op_fid1, &res_id);
1047 req = mdc_intent_getattr_pack(exp, it, op_data);
1051 rc = mdc_enter_request(&obddev->u.cli);
1053 ptlrpc_req_finished(req);
1057 rc = ldlm_cli_enqueue(exp, &req, einfo, &res_id, &policy, &flags, NULL,
1058 0, &minfo->mi_lockh, 1);
1060 mdc_exit_request(&obddev->u.cli);
1061 ptlrpc_req_finished(req);
1065 CLASSERT(sizeof(*ga) <= sizeof(req->rq_async_args));
1066 ga = ptlrpc_req_async_args(req);
1068 ga->ga_minfo = minfo;
1069 ga->ga_einfo = einfo;
1071 req->rq_interpret_reply = mdc_intent_getattr_async_interpret;
1072 ptlrpcd_add_req(req, PDL_POLICY_LOCAL, -1);