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) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
32 * Copyright (c) 2011 Whamcloud, Inc.
36 * This file is part of Lustre, http://www.lustre.org/
37 * Lustre is a trademark of Sun Microsystems, Inc.
41 #include <linux/sched.h>
42 #include <linux/smp_lock.h>
43 #include <linux/quotaops.h>
45 #define DEBUG_SUBSYSTEM S_LLITE
47 #include <obd_support.h>
48 #include <lustre_lite.h>
49 #include <lustre/lustre_idl.h>
50 #include <lustre_dlm.h>
51 #include <lustre_mdc.h>
52 //#include <lustre_ver.h>
53 //#include <lustre_version.h>
55 #include "llite_internal.h"
57 cfs_spinlock_t ll_lookup_lock = CFS_SPIN_LOCK_UNLOCKED;
59 /* should NOT be called with the dcache lock, see fs/dcache.c */
60 static void ll_release(struct dentry *de)
62 struct ll_dentry_data *lld;
66 if (lld == NULL) { /* NFS copies the de->d_op methods (bug 4655) */
71 ll_intent_release(lld->lld_it);
72 OBD_FREE(lld->lld_it, sizeof(*lld->lld_it));
74 LASSERT(lld->lld_cwd_count == 0);
75 LASSERT(lld->lld_mnt_count == 0);
76 OBD_FREE(de->d_fsdata, sizeof(*lld));
81 /* Compare if two dentries are the same. Don't match if the existing dentry
82 * is marked DCACHE_LUSTRE_INVALID. Returns 1 if different, 0 if the same.
84 * This avoids a race where ll_lookup_it() instantiates a dentry, but we get
85 * an AST before calling d_revalidate_it(). The dentry still exists (marked
86 * INVALID) so d_lookup() matches it, but we have no lock on it (so
87 * lock_match() fails) and we spin around real_lookup(). */
88 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name)
90 struct dentry *dchild;
93 if (d_name->len != name->len)
96 if (memcmp(d_name->name, name->name, name->len))
99 /* XXX: d_name must be in-dentry structure */
100 dchild = container_of(d_name, struct dentry, d_name); /* ugh */
102 CDEBUG(D_DENTRY,"found name %.*s(%p) - flags %d/%x - refc %d\n",
103 name->len, name->name, dchild,
104 d_mountpoint(dchild), dchild->d_flags & DCACHE_LUSTRE_INVALID,
105 atomic_read(&dchild->d_count));
107 /* mountpoint is always valid */
108 if (d_mountpoint(dchild))
111 if (dchild->d_flags & DCACHE_LUSTRE_INVALID)
117 static inline int return_if_equal(struct ldlm_lock *lock, void *data)
120 (LDLM_FL_CANCELING | LDLM_FL_DISCARD_DATA)) ==
121 (LDLM_FL_CANCELING | LDLM_FL_DISCARD_DATA))
122 return LDLM_ITER_CONTINUE;
123 return LDLM_ITER_STOP;
126 /* find any ldlm lock of the inode in mdc and lov
130 static int find_cbdata(struct inode *inode)
132 struct ll_inode_info *lli = ll_i2info(inode);
133 struct ll_sb_info *sbi = ll_i2sbi(inode);
138 rc = md_find_cbdata(sbi->ll_md_exp, ll_inode2fid(inode),
139 return_if_equal, NULL);
144 rc = obd_find_cbdata(sbi->ll_dt_exp, lli->lli_smd,
145 return_if_equal, NULL);
151 * Called when last reference to a dentry is dropped and dcache wants to know
152 * whether or not it should cache it:
153 * - return 1 to delete the dentry immediately
154 * - return 0 to cache the dentry
155 * Should NOT be called with the dcache lock, see fs/dcache.c
157 static int ll_ddelete(struct dentry *de)
162 CDEBUG(D_DENTRY, "%s dentry %.*s (%p, parent %p, inode %p) %s%s\n",
163 (de->d_flags & DCACHE_LUSTRE_INVALID ? "deleting" : "keeping"),
164 de->d_name.len, de->d_name.name, de, de->d_parent, de->d_inode,
165 d_unhashed(de) ? "" : "hashed,",
166 list_empty(&de->d_subdirs) ? "" : "subdirs");
168 /* if not ldlm lock for this inode, set i_nlink to 0 so that
169 * this inode can be recycled later b=20433 */
170 LASSERT(atomic_read(&de->d_count) == 0);
171 if (de->d_inode && !find_cbdata(de->d_inode))
172 de->d_inode->i_nlink = 0;
174 if (de->d_flags & DCACHE_LUSTRE_INVALID)
180 static int ll_set_dd(struct dentry *de)
185 CDEBUG(D_DENTRY, "ldd on dentry %.*s (%p) parent %p inode %p refc %d\n",
186 de->d_name.len, de->d_name.name, de, de->d_parent, de->d_inode,
187 atomic_read(&de->d_count));
189 if (de->d_fsdata == NULL) {
190 struct ll_dentry_data *lld;
193 if (likely(lld != NULL)) {
195 if (likely(de->d_fsdata == NULL))
208 int ll_dops_init(struct dentry *de, int block, int init_sa)
210 struct ll_dentry_data *lld = ll_d2d(de);
213 if (lld == NULL && block != 0) {
221 if (lld != NULL && init_sa != 0)
222 lld->lld_sa_generation = 0;
224 de->d_op = &ll_d_ops;
228 void ll_intent_drop_lock(struct lookup_intent *it)
230 struct lustre_handle *handle;
232 if (it->it_op && it->d.lustre.it_lock_mode) {
233 handle = (struct lustre_handle *)&it->d.lustre.it_lock_handle;
234 CDEBUG(D_DLMTRACE, "releasing lock with cookie "LPX64
235 " from it %p\n", handle->cookie, it);
236 ldlm_lock_decref(handle, it->d.lustre.it_lock_mode);
238 /* bug 494: intent_release may be called multiple times, from
239 * this thread and we don't want to double-decref this lock */
240 it->d.lustre.it_lock_mode = 0;
244 void ll_intent_release(struct lookup_intent *it)
248 CDEBUG(D_INFO, "intent %p released\n", it);
249 ll_intent_drop_lock(it);
250 /* We are still holding extra reference on a request, need to free it */
251 if (it_disposition(it, DISP_ENQ_OPEN_REF))
252 ptlrpc_req_finished(it->d.lustre.it_data); /* ll_file_open */
253 if (it_disposition(it, DISP_ENQ_CREATE_REF)) /* create rec */
254 ptlrpc_req_finished(it->d.lustre.it_data);
255 if (it_disposition(it, DISP_ENQ_COMPLETE)) /* saved req from revalidate
257 ptlrpc_req_finished(it->d.lustre.it_data);
259 it->d.lustre.it_disposition = 0;
260 it->d.lustre.it_data = NULL;
264 /* Drop dentry if it is not used already, unhash otherwise.
265 Should be called with dcache lock held!
266 Returns: 1 if dentry was dropped, 0 if unhashed. */
267 int ll_drop_dentry(struct dentry *dentry)
270 if (atomic_read(&dentry->d_count) == 0) {
271 CDEBUG(D_DENTRY, "deleting dentry %.*s (%p) parent %p "
272 "inode %p\n", dentry->d_name.len,
273 dentry->d_name.name, dentry, dentry->d_parent,
277 unlock_dentry(dentry);
278 spin_unlock(&dcache_lock);
279 cfs_spin_unlock(&ll_lookup_lock);
281 cfs_spin_lock(&ll_lookup_lock);
282 spin_lock(&dcache_lock);
285 /* disconected dentry can not be find without lookup, because we
286 * not need his to unhash or mark invalid. */
287 if (dentry->d_flags & DCACHE_DISCONNECTED) {
288 unlock_dentry(dentry);
292 if (!(dentry->d_flags & DCACHE_LUSTRE_INVALID)) {
293 CDEBUG(D_DENTRY, "unhashing dentry %.*s (%p) parent %p "
294 "inode %p refc %d\n", dentry->d_name.len,
295 dentry->d_name.name, dentry, dentry->d_parent,
296 dentry->d_inode, atomic_read(&dentry->d_count));
297 /* actually we don't unhash the dentry, rather just
298 * mark it inaccessible for to __d_lookup(). otherwise
299 * sys_getcwd() could return -ENOENT -bzzz */
300 dentry->d_flags |= DCACHE_LUSTRE_INVALID;
301 if (!dentry->d_inode || !S_ISDIR(dentry->d_inode->i_mode))
304 unlock_dentry(dentry);
308 void ll_unhash_aliases(struct inode *inode)
310 struct list_head *tmp, *head;
314 CERROR("unexpected NULL inode, tell phil\n");
318 CDEBUG(D_INODE, "marking dentries for ino %lu/%u(%p) invalid\n",
319 inode->i_ino, inode->i_generation, inode);
321 head = &inode->i_dentry;
322 cfs_spin_lock(&ll_lookup_lock);
323 spin_lock(&dcache_lock);
326 while ((tmp = tmp->next) != head) {
327 struct dentry *dentry = list_entry(tmp, struct dentry, d_alias);
329 CDEBUG(D_DENTRY, "dentry in drop %.*s (%p) parent %p "
330 "inode %p flags %d\n", dentry->d_name.len,
331 dentry->d_name.name, dentry, dentry->d_parent,
332 dentry->d_inode, dentry->d_flags);
334 if (dentry->d_name.len == 1 && dentry->d_name.name[0] == '/') {
335 CERROR("called on root (?) dentry=%p, inode=%p "
336 "ino=%lu\n", dentry, inode, inode->i_ino);
337 lustre_dump_dentry(dentry, 1);
338 libcfs_debug_dumpstack(NULL);
341 if (ll_drop_dentry(dentry))
344 spin_unlock(&dcache_lock);
345 cfs_spin_unlock(&ll_lookup_lock);
350 int ll_revalidate_it_finish(struct ptlrpc_request *request,
351 struct lookup_intent *it,
360 if (it_disposition(it, DISP_LOOKUP_NEG))
363 rc = ll_prep_inode(&de->d_inode, request, NULL);
368 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry)
371 LASSERT(dentry != NULL);
373 if (it->d.lustre.it_lock_mode && dentry->d_inode != NULL) {
374 struct inode *inode = dentry->d_inode;
375 struct ll_sb_info *sbi = ll_i2sbi(dentry->d_inode);
377 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
378 inode, inode->i_ino, inode->i_generation);
379 md_set_lock_data(sbi->ll_md_exp, &it->d.lustre.it_lock_handle,
383 /* drop lookup or getattr locks immediately */
384 if (it->it_op == IT_LOOKUP || it->it_op == IT_GETATTR) {
385 /* on 2.6 there are situation when several lookups and
386 * revalidations may be requested during single operation.
387 * therefore, we don't release intent here -bzzz */
388 ll_intent_drop_lock(it);
392 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft)
394 struct lookup_intent *it = *itp;
396 if (!it || it->it_op == IT_GETXATTR)
401 int ll_revalidate_it(struct dentry *de, int lookup_flags,
402 struct lookup_intent *it)
404 struct md_op_data *op_data;
405 struct ptlrpc_request *req = NULL;
406 struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
407 struct obd_export *exp;
408 struct inode *parent = de->d_parent->d_inode;
412 CDEBUG(D_VFSTRACE, "VFS Op:name=%s,intent=%s\n", de->d_name.name,
415 if (de->d_inode == NULL) {
418 /* We can only use negative dentries if this is stat or lookup,
419 for opens and stuff we do need to query server. */
420 /* If there is IT_CREAT in intent op set, then we must throw
421 away this negative dentry and actually do the request to
422 kernel to create whatever needs to be created (if possible)*/
423 if (it && (it->it_op & IT_CREAT))
426 if (de->d_flags & DCACHE_LUSTRE_INVALID)
429 ibits = MDS_INODELOCK_UPDATE;
430 rc = ll_have_md_lock(parent, &ibits, LCK_MINMODE);
434 /* Never execute intents for mount points.
435 * Attributes will be fixed up in ll_inode_revalidate_it */
436 if (d_mountpoint(de))
437 GOTO(out_sa, rc = 1);
439 /* need to get attributes in case root got changed from other client */
440 if (de == de->d_sb->s_root) {
441 rc = __ll_inode_revalidate_it(de, it, MDS_INODELOCK_LOOKUP);
447 exp = ll_i2mdexp(de->d_inode);
449 OBD_FAIL_TIMEOUT(OBD_FAIL_MDC_REVALIDATE_PAUSE, 5);
450 ll_frob_intent(&it, &lookup_it);
453 if (it->it_op == IT_LOOKUP && !(de->d_flags & DCACHE_LUSTRE_INVALID))
454 GOTO(out_sa, rc = 1);
456 op_data = ll_prep_md_op_data(NULL, parent, de->d_inode,
457 de->d_name.name, de->d_name.len,
458 0, LUSTRE_OPC_ANY, NULL);
460 RETURN(PTR_ERR(op_data));
462 if ((it->it_op == IT_OPEN) && de->d_inode) {
463 struct inode *inode = de->d_inode;
464 struct ll_inode_info *lli = ll_i2info(inode);
465 struct obd_client_handle **och_p;
470 * We used to check for MDS_INODELOCK_OPEN here, but in fact
471 * just having LOOKUP lock is enough to justify inode is the
472 * same. And if inode is the same and we have suitable
473 * openhandle, then there is no point in doing another OPEN RPC
474 * just to throw away newly received openhandle. There are no
475 * security implications too, if file owner or access mode is
476 * change, LOOKUP lock is revoked.
480 if (it->it_flags & FMODE_WRITE) {
481 och_p = &lli->lli_mds_write_och;
482 och_usecount = &lli->lli_open_fd_write_count;
483 } else if (it->it_flags & FMODE_EXEC) {
484 och_p = &lli->lli_mds_exec_och;
485 och_usecount = &lli->lli_open_fd_exec_count;
487 och_p = &lli->lli_mds_read_och;
488 och_usecount = &lli->lli_open_fd_read_count;
490 /* Check for the proper lock. */
491 ibits = MDS_INODELOCK_LOOKUP;
492 if (!ll_have_md_lock(inode, &ibits, LCK_MINMODE))
494 cfs_down(&lli->lli_och_sem);
495 if (*och_p) { /* Everything is open already, do nothing */
496 /*(*och_usecount)++; Do not let them steal our open
497 handle from under us */
498 /* XXX The code above was my original idea, but in case
499 we have the handle, but we cannot use it due to later
500 checks (e.g. O_CREAT|O_EXCL flags set), nobody
501 would decrement counter increased here. So we just
502 hope the lock won't be invalidated in between. But
503 if it would be, we'll reopen the open request to
504 MDS later during file open path */
505 cfs_up(&lli->lli_och_sem);
506 ll_finish_md_op_data(op_data);
509 cfs_up(&lli->lli_och_sem);
513 if (it->it_op == IT_GETATTR)
514 first = ll_statahead_enter(parent, &de, 0);
517 it->it_create_mode &= ~cfs_curproc_umask();
518 it->it_create_mode |= M_CHECK_STALE;
519 rc = md_intent_lock(exp, op_data, NULL, 0, it,
521 &req, ll_md_blocking_ast, 0);
522 it->it_create_mode &= ~M_CHECK_STALE;
523 ll_finish_md_op_data(op_data);
524 if (it->it_op == IT_GETATTR && !first)
525 /* If there are too many locks on client-side, then some
526 * locks taken by statahead maybe dropped automatically
527 * before the real "revalidate" using them. */
528 ll_statahead_exit(parent, de, req == NULL ? rc : 0);
529 else if (first == -EEXIST)
530 ll_statahead_mark(parent, de);
532 /* If req is NULL, then md_intent_lock only tried to do a lock match;
533 * if all was well, it will return 1 if it found locks, 0 otherwise. */
534 if (req == NULL && rc >= 0) {
542 CDEBUG(D_INFO, "ll_intent_lock: rc %d : it->it_status "
543 "%d\n", rc, it->d.lustre.it_status);
549 rc = ll_revalidate_it_finish(req, it, de);
551 if (rc != -ESTALE && rc != -ENOENT)
552 ll_intent_release(it);
556 if ((it->it_op & IT_OPEN) && de->d_inode &&
557 !S_ISREG(de->d_inode->i_mode) &&
558 !S_ISDIR(de->d_inode->i_mode)) {
559 ll_release_openhandle(de, it);
563 /* unfortunately ll_intent_lock may cause a callback and revoke our
565 cfs_spin_lock(&ll_lookup_lock);
566 spin_lock(&dcache_lock);
570 d_rehash_cond(de, 0);
571 spin_unlock(&dcache_lock);
572 cfs_spin_unlock(&ll_lookup_lock);
575 /* We do not free request as it may be reused during following lookup
576 * (see comment in mdc/mdc_locks.c::mdc_intent_lock()), request will
577 * be freed in ll_lookup_it or in ll_intent_release. But if
578 * request was not completed, we need to free it. (bug 5154, 9903) */
579 if (req != NULL && !it_disposition(it, DISP_ENQ_COMPLETE))
580 ptlrpc_req_finished(req);
582 ll_unhash_aliases(de->d_inode);
583 /* done in ll_unhash_aliases()
584 dentry->d_flags |= DCACHE_LUSTRE_INVALID; */
586 CDEBUG(D_DENTRY, "revalidated dentry %.*s (%p) parent %p "
587 "inode %p refc %d\n", de->d_name.len,
588 de->d_name.name, de, de->d_parent, de->d_inode,
589 atomic_read(&de->d_count));
590 if (de->d_flags & DCACHE_LUSTRE_INVALID) {
592 de->d_flags &= ~DCACHE_LUSTRE_INVALID;
595 ll_lookup_finish_locks(it, de);
600 * This part is here to combat evil-evil race in real_lookup on 2.6
601 * kernels. The race details are: We enter do_lookup() looking for some
602 * name, there is nothing in dcache for this name yet and d_lookup()
603 * returns NULL. We proceed to real_lookup(), and while we do this,
604 * another process does open on the same file we looking up (most simple
605 * reproducer), open succeeds and the dentry is added. Now back to
606 * us. In real_lookup() we do d_lookup() again and suddenly find the
607 * dentry, so we call d_revalidate on it, but there is no lock, so
608 * without this code we would return 0, but unpatched real_lookup just
609 * returns -ENOENT in such a case instead of retrying the lookup. Once
610 * this is dealt with in real_lookup(), all of this ugly mess can go and
611 * we can just check locks in ->d_revalidate without doing any RPCs
615 if (it != &lookup_it) {
616 /* MDS_INODELOCK_UPDATE needed for IT_GETATTR case. */
617 if (it->it_op == IT_GETATTR)
618 lookup_it.it_op = IT_GETATTR;
619 ll_lookup_finish_locks(it, de);
623 /* Do real lookup here. */
624 op_data = ll_prep_md_op_data(NULL, parent, NULL, de->d_name.name,
625 de->d_name.len, 0, (it->it_op & IT_CREAT ?
627 LUSTRE_OPC_ANY), NULL);
629 RETURN(PTR_ERR(op_data));
631 rc = md_intent_lock(exp, op_data, NULL, 0, it, 0, &req,
632 ll_md_blocking_ast, 0);
634 struct mdt_body *mdt_body;
635 struct lu_fid fid = {.f_seq = 0, .f_oid = 0, .f_ver = 0};
636 mdt_body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
639 fid = *ll_inode2fid(de->d_inode);
641 /* see if we got same inode, if not - return error */
642 if (lu_fid_eq(&fid, &mdt_body->fid1)) {
643 ll_finish_md_op_data(op_data);
645 goto revalidate_finish;
647 ll_intent_release(it);
649 ll_finish_md_op_data(op_data);
654 * For rc == 1 case, should not return directly to prevent losing
655 * statahead windows; for rc == 0 case, the "lookup" will be done later.
657 if (it && it->it_op == IT_GETATTR && rc == 1) {
658 first = ll_statahead_enter(parent, &de, 0);
660 ll_statahead_exit(parent, de, 1);
661 else if (first == -EEXIST)
662 ll_statahead_mark(parent, de);
668 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd)
673 if (nd && !(nd->flags & (LOOKUP_CONTINUE|LOOKUP_PARENT))) {
674 struct lookup_intent *it;
676 it = ll_convert_intent(&nd->intent.open, nd->flags);
680 if (it->it_op == (IT_OPEN|IT_CREAT) &&
681 nd->intent.open.flags & O_EXCL) {
682 CDEBUG(D_VFSTRACE, "create O_EXCL, returning 0\n");
687 rc = ll_revalidate_it(dentry, nd->flags, it);
689 if (rc && (nd->flags & LOOKUP_OPEN) &&
690 it_disposition(it, DISP_OPEN_OPEN)) {/*Open*/
691 #ifdef HAVE_FILE_IN_STRUCT_INTENT
692 // XXX Code duplication with ll_lookup_nd
693 if (S_ISFIFO(dentry->d_inode->i_mode)) {
694 // We cannot call open here as it would
697 (struct ptlrpc_request *)
698 it->d.lustre.it_data);
700 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17))
701 /* 2.6.1[456] have a bug in open_namei() that forgets to check
702 * nd->intent.open.file for error, so we need to return it as lookup's result
706 nd->intent.open.file->private_data = it;
707 filp = lookup_instantiate_filp(nd, dentry,NULL);
712 nd->intent.open.file->private_data = it;
713 (void)lookup_instantiate_filp(nd, dentry,NULL);
717 ll_release_openhandle(dentry, it);
718 #endif /* HAVE_FILE_IN_STRUCT_INTENT */
720 if (!rc && (nd->flags & LOOKUP_CREATE) &&
721 it_disposition(it, DISP_OPEN_CREATE)) {
722 /* We created something but we may only return
723 * negative dentry here, so save request in dentry,
724 * if lookup will be called later on, it will
725 * pick the request, otherwise it would be freed
727 ll_d2d(dentry)->lld_it = it;
728 it = NULL; /* avoid freeing */
733 ll_intent_release(it);
734 OBD_FREE(it, sizeof(*it));
737 rc = ll_revalidate_it(dentry, 0, NULL);
743 void ll_d_iput(struct dentry *de, struct inode *inode)
746 if (!find_cbdata(inode))
751 struct dentry_operations ll_d_ops = {
752 .d_revalidate = ll_revalidate_nd,
753 .d_release = ll_release,
754 .d_delete = ll_ddelete,
756 .d_compare = ll_dcompare,