Commit | Line | Data |
---|---|---|
d7e09d03 PT |
1 | /* |
2 | * GPL HEADER START | |
3 | * | |
4 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. | |
5 | * | |
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. | |
9 | * | |
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). | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License | |
17 | * version 2 along with this program; If not, see | |
18 | * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf | |
19 | * | |
20 | * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, | |
21 | * CA 95054 USA or visit www.sun.com if you need additional information or | |
22 | * have any questions. | |
23 | * | |
24 | * GPL HEADER END | |
25 | */ | |
26 | /* | |
27 | * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved. | |
28 | * Use is subject to license terms. | |
29 | * | |
30 | * Copyright (c) 2011, 2012, Intel Corporation. | |
31 | */ | |
32 | /* | |
33 | * This file is part of Lustre, http://www.lustre.org/ | |
34 | * Lustre is a trademark of Sun Microsystems, Inc. | |
35 | */ | |
36 | ||
37 | #include <linux/fs.h> | |
38 | #include <linux/sched.h> | |
39 | #include <linux/mm.h> | |
40 | #include <linux/quotaops.h> | |
41 | #include <linux/highmem.h> | |
42 | #include <linux/pagemap.h> | |
43 | #include <linux/security.h> | |
44 | ||
45 | #define DEBUG_SUBSYSTEM S_LLITE | |
46 | ||
67a235f5 GKH |
47 | #include "../include/obd_support.h" |
48 | #include "../include/lustre_fid.h" | |
49 | #include "../include/lustre_lite.h" | |
50 | #include "../include/lustre_dlm.h" | |
51 | #include "../include/lustre_ver.h" | |
d7e09d03 PT |
52 | #include "llite_internal.h" |
53 | ||
54 | static int ll_create_it(struct inode *, struct dentry *, | |
55 | int, struct lookup_intent *); | |
56 | ||
57 | /* | |
58 | * Check if we have something mounted at the named dchild. | |
59 | * In such a case there would always be dentry present. | |
60 | */ | |
61 | static int ll_d_mountpoint(struct dentry *dparent, struct dentry *dchild, | |
62 | struct qstr *name) | |
63 | { | |
64 | int mounted = 0; | |
65 | ||
66 | if (unlikely(dchild)) { | |
67 | mounted = d_mountpoint(dchild); | |
68 | } else if (dparent) { | |
69 | dchild = d_lookup(dparent, name); | |
70 | if (dchild) { | |
71 | mounted = d_mountpoint(dchild); | |
72 | dput(dchild); | |
73 | } | |
74 | } | |
75 | return mounted; | |
76 | } | |
77 | ||
5a3cd992 | 78 | /* called from iget5_locked->find_inode() under inode_hash_lock spinlock */ |
d7e09d03 PT |
79 | static int ll_test_inode(struct inode *inode, void *opaque) |
80 | { | |
81 | struct ll_inode_info *lli = ll_i2info(inode); | |
82 | struct lustre_md *md = opaque; | |
83 | ||
84 | if (unlikely(!(md->body->valid & OBD_MD_FLID))) { | |
85 | CERROR("MDS body missing FID\n"); | |
86 | return 0; | |
87 | } | |
88 | ||
89 | if (!lu_fid_eq(&lli->lli_fid, &md->body->fid1)) | |
90 | return 0; | |
91 | ||
92 | return 1; | |
93 | } | |
94 | ||
95 | static int ll_set_inode(struct inode *inode, void *opaque) | |
96 | { | |
97 | struct ll_inode_info *lli = ll_i2info(inode); | |
98 | struct mdt_body *body = ((struct lustre_md *)opaque)->body; | |
99 | ||
100 | if (unlikely(!(body->valid & OBD_MD_FLID))) { | |
101 | CERROR("MDS body missing FID\n"); | |
102 | return -EINVAL; | |
103 | } | |
104 | ||
105 | lli->lli_fid = body->fid1; | |
106 | if (unlikely(!(body->valid & OBD_MD_FLTYPE))) { | |
107 | CERROR("Can not initialize inode "DFID" without object type: " | |
55f5a824 | 108 | "valid = %#llx\n", PFID(&lli->lli_fid), body->valid); |
d7e09d03 PT |
109 | return -EINVAL; |
110 | } | |
111 | ||
112 | inode->i_mode = (inode->i_mode & ~S_IFMT) | (body->mode & S_IFMT); | |
113 | if (unlikely(inode->i_mode == 0)) { | |
114 | CERROR("Invalid inode "DFID" type\n", PFID(&lli->lli_fid)); | |
115 | return -EINVAL; | |
116 | } | |
117 | ||
118 | ll_lli_init(lli); | |
119 | ||
120 | return 0; | |
121 | } | |
122 | ||
123 | ||
124 | /* | |
125 | * Get an inode by inode number (already instantiated by the intent lookup). | |
126 | * Returns inode or NULL | |
127 | */ | |
128 | struct inode *ll_iget(struct super_block *sb, ino_t hash, | |
129 | struct lustre_md *md) | |
130 | { | |
131 | struct inode *inode; | |
d7e09d03 PT |
132 | |
133 | LASSERT(hash != 0); | |
134 | inode = iget5_locked(sb, hash, ll_test_inode, ll_set_inode, md); | |
135 | ||
136 | if (inode) { | |
137 | if (inode->i_state & I_NEW) { | |
138 | int rc = 0; | |
139 | ||
140 | ll_read_inode2(inode, md); | |
141 | if (S_ISREG(inode->i_mode) && | |
142 | ll_i2info(inode)->lli_clob == NULL) { | |
143 | CDEBUG(D_INODE, | |
144 | "%s: apply lsm %p to inode "DFID".\n", | |
145 | ll_get_fsname(sb, NULL, 0), md->lsm, | |
146 | PFID(ll_inode2fid(inode))); | |
147 | rc = cl_file_inode_init(inode, md); | |
148 | } | |
149 | if (rc != 0) { | |
150 | make_bad_inode(inode); | |
151 | unlock_new_inode(inode); | |
152 | iput(inode); | |
153 | inode = ERR_PTR(rc); | |
154 | } else | |
155 | unlock_new_inode(inode); | |
156 | } else if (!(inode->i_state & (I_FREEING | I_CLEAR))) | |
157 | ll_update_inode(inode, md); | |
158 | CDEBUG(D_VFSTRACE, "got inode: %p for "DFID"\n", | |
159 | inode, PFID(&md->body->fid1)); | |
160 | } | |
0a3bdb00 | 161 | return inode; |
d7e09d03 PT |
162 | } |
163 | ||
164 | static void ll_invalidate_negative_children(struct inode *dir) | |
165 | { | |
166 | struct dentry *dentry, *tmp_subdir; | |
167 | struct ll_d_hlist_node *p; | |
168 | ||
169 | ll_lock_dcache(dir); | |
946e51f2 | 170 | ll_d_hlist_for_each_entry(dentry, p, &dir->i_dentry, d_u.d_alias) { |
d7e09d03 PT |
171 | spin_lock(&dentry->d_lock); |
172 | if (!list_empty(&dentry->d_subdirs)) { | |
173 | struct dentry *child; | |
174 | ||
175 | list_for_each_entry_safe(child, tmp_subdir, | |
176 | &dentry->d_subdirs, | |
946e51f2 | 177 | d_child) { |
d7e09d03 | 178 | if (child->d_inode == NULL) |
b1d2a127 | 179 | d_lustre_invalidate(child, 1); |
d7e09d03 PT |
180 | } |
181 | } | |
182 | spin_unlock(&dentry->d_lock); | |
183 | } | |
184 | ll_unlock_dcache(dir); | |
185 | } | |
186 | ||
187 | int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, | |
188 | void *data, int flag) | |
189 | { | |
d7e09d03 | 190 | struct lustre_handle lockh; |
45b2a010 | 191 | int rc; |
d7e09d03 PT |
192 | |
193 | switch (flag) { | |
194 | case LDLM_CB_BLOCKING: | |
195 | ldlm_lock2handle(lock, &lockh); | |
196 | rc = ldlm_cli_cancel(&lockh, LCF_ASYNC); | |
197 | if (rc < 0) { | |
45b2a010 | 198 | CDEBUG(D_INODE, "ldlm_cli_cancel: rc = %d\n", rc); |
0a3bdb00 | 199 | return rc; |
d7e09d03 PT |
200 | } |
201 | break; | |
202 | case LDLM_CB_CANCELING: { | |
203 | struct inode *inode = ll_inode_from_resource_lock(lock); | |
d7e09d03 | 204 | __u64 bits = lock->l_policy_data.l_inodebits.bits; |
d7e09d03 PT |
205 | |
206 | /* Inode is set to lock->l_resource->lr_lvb_inode | |
207 | * for mdc - bug 24555 */ | |
208 | LASSERT(lock->l_ast_data == NULL); | |
209 | ||
d7e09d03 PT |
210 | if (inode == NULL) |
211 | break; | |
212 | ||
45b2a010 | 213 | /* Invalidate all dentries associated with this inode */ |
d7e09d03 | 214 | LASSERT(lock->l_flags & LDLM_FL_CANCELING); |
7fc1f831 | 215 | |
45b2a010 JH |
216 | if (!fid_res_name_eq(ll_inode2fid(inode), |
217 | &lock->l_resource->lr_name)) { | |
218 | LDLM_ERROR(lock, "data mismatch with object "DFID"(%p)", | |
219 | PFID(ll_inode2fid(inode)), inode); | |
220 | LBUG(); | |
221 | } | |
222 | ||
223 | if (bits & MDS_INODELOCK_XATTR) { | |
7fc1f831 | 224 | ll_xattr_cache_destroy(inode); |
45b2a010 JH |
225 | bits &= ~MDS_INODELOCK_XATTR; |
226 | } | |
7fc1f831 | 227 | |
d7e09d03 PT |
228 | /* For OPEN locks we differentiate between lock modes |
229 | * LCK_CR, LCK_CW, LCK_PR - bug 22891 */ | |
d7e09d03 | 230 | if (bits & MDS_INODELOCK_OPEN) |
45b2a010 | 231 | ll_have_md_lock(inode, &bits, lock->l_req_mode); |
d7e09d03 PT |
232 | |
233 | if (bits & MDS_INODELOCK_OPEN) { | |
45b2a010 JH |
234 | fmode_t fmode; |
235 | ||
d7e09d03 PT |
236 | switch (lock->l_req_mode) { |
237 | case LCK_CW: | |
45b2a010 | 238 | fmode = FMODE_WRITE; |
d7e09d03 PT |
239 | break; |
240 | case LCK_PR: | |
45b2a010 | 241 | fmode = FMODE_EXEC; |
d7e09d03 PT |
242 | break; |
243 | case LCK_CR: | |
45b2a010 | 244 | fmode = FMODE_READ; |
d7e09d03 PT |
245 | break; |
246 | default: | |
45b2a010 JH |
247 | LDLM_ERROR(lock, "bad lock mode for OPEN lock"); |
248 | LBUG(); | |
d7e09d03 | 249 | } |
45b2a010 JH |
250 | |
251 | ll_md_real_close(inode, fmode); | |
d7e09d03 PT |
252 | } |
253 | ||
45b2a010 JH |
254 | if (bits & (MDS_INODELOCK_LOOKUP | MDS_INODELOCK_UPDATE | |
255 | MDS_INODELOCK_LAYOUT | MDS_INODELOCK_PERM)) | |
256 | ll_have_md_lock(inode, &bits, LCK_MINMODE); | |
257 | ||
d7e09d03 | 258 | if (bits & MDS_INODELOCK_LAYOUT) { |
45b2a010 JH |
259 | struct cl_object_conf conf = { |
260 | .coc_opc = OBJECT_CONF_INVALIDATE, | |
261 | .coc_inode = inode, | |
262 | }; | |
d7e09d03 | 263 | |
d7e09d03 | 264 | rc = ll_layout_conf(inode, &conf); |
45b2a010 JH |
265 | if (rc < 0) |
266 | CDEBUG(D_INODE, "cannot invalidate layout of " | |
267 | DFID": rc = %d\n", | |
268 | PFID(ll_inode2fid(inode)), rc); | |
d7e09d03 PT |
269 | } |
270 | ||
ae5ef67b | 271 | if (bits & MDS_INODELOCK_UPDATE) { |
45b2a010 JH |
272 | struct ll_inode_info *lli = ll_i2info(inode); |
273 | ||
ae5ef67b | 274 | spin_lock(&lli->lli_lock); |
d7e09d03 | 275 | lli->lli_flags &= ~LLIF_MDS_SIZE_LOCK; |
ae5ef67b SB |
276 | spin_unlock(&lli->lli_lock); |
277 | } | |
d7e09d03 | 278 | |
45b2a010 | 279 | if ((bits & MDS_INODELOCK_UPDATE) && S_ISDIR(inode->i_mode)) { |
d7e09d03 PT |
280 | CDEBUG(D_INODE, "invalidating inode %lu\n", |
281 | inode->i_ino); | |
282 | truncate_inode_pages(inode->i_mapping, 0); | |
283 | ll_invalidate_negative_children(inode); | |
284 | } | |
285 | ||
45b2a010 JH |
286 | if ((bits & (MDS_INODELOCK_LOOKUP | MDS_INODELOCK_PERM)) && |
287 | inode->i_sb->s_root != NULL && | |
288 | inode != inode->i_sb->s_root->d_inode) | |
d7e09d03 | 289 | ll_invalidate_aliases(inode); |
45b2a010 | 290 | |
d7e09d03 PT |
291 | iput(inode); |
292 | break; | |
293 | } | |
294 | default: | |
295 | LBUG(); | |
296 | } | |
297 | ||
0a3bdb00 | 298 | return 0; |
d7e09d03 PT |
299 | } |
300 | ||
301 | __u32 ll_i2suppgid(struct inode *i) | |
302 | { | |
4b1a25f0 PT |
303 | if (in_group_p(i->i_gid)) |
304 | return (__u32)from_kgid(&init_user_ns, i->i_gid); | |
d7e09d03 PT |
305 | else |
306 | return (__u32)(-1); | |
307 | } | |
308 | ||
309 | /* Pack the required supplementary groups into the supplied groups array. | |
310 | * If we don't need to use the groups from the target inode(s) then we | |
311 | * instead pack one or more groups from the user's supplementary group | |
312 | * array in case it might be useful. Not needed if doing an MDS-side upcall. */ | |
313 | void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2) | |
314 | { | |
315 | #if 0 | |
316 | int i; | |
317 | #endif | |
318 | ||
319 | LASSERT(i1 != NULL); | |
320 | LASSERT(suppgids != NULL); | |
321 | ||
322 | suppgids[0] = ll_i2suppgid(i1); | |
323 | ||
324 | if (i2) | |
325 | suppgids[1] = ll_i2suppgid(i2); | |
326 | else | |
327 | suppgids[1] = -1; | |
328 | ||
329 | #if 0 | |
330 | for (i = 0; i < current_ngroups; i++) { | |
331 | if (suppgids[0] == -1) { | |
332 | if (current_groups[i] != suppgids[1]) | |
333 | suppgids[0] = current_groups[i]; | |
334 | continue; | |
335 | } | |
336 | if (suppgids[1] == -1) { | |
337 | if (current_groups[i] != suppgids[0]) | |
338 | suppgids[1] = current_groups[i]; | |
339 | continue; | |
340 | } | |
341 | break; | |
342 | } | |
343 | #endif | |
344 | } | |
345 | ||
346 | /* | |
347 | * try to reuse three types of dentry: | |
348 | * 1. unhashed alias, this one is unhashed by d_invalidate (but it may be valid | |
349 | * by concurrent .revalidate). | |
350 | * 2. INVALID alias (common case for no valid ldlm lock held, but this flag may | |
351 | * be cleared by others calling d_lustre_revalidate). | |
352 | * 3. DISCONNECTED alias. | |
353 | */ | |
354 | static struct dentry *ll_find_alias(struct inode *inode, struct dentry *dentry) | |
355 | { | |
356 | struct dentry *alias, *discon_alias, *invalid_alias; | |
357 | struct ll_d_hlist_node *p; | |
358 | ||
359 | if (ll_d_hlist_empty(&inode->i_dentry)) | |
360 | return NULL; | |
361 | ||
362 | discon_alias = invalid_alias = NULL; | |
363 | ||
364 | ll_lock_dcache(inode); | |
946e51f2 | 365 | ll_d_hlist_for_each_entry(alias, p, &inode->i_dentry, d_u.d_alias) { |
d7e09d03 PT |
366 | LASSERT(alias != dentry); |
367 | ||
368 | spin_lock(&alias->d_lock); | |
369 | if (alias->d_flags & DCACHE_DISCONNECTED) | |
370 | /* LASSERT(last_discon == NULL); LU-405, bz 20055 */ | |
371 | discon_alias = alias; | |
372 | else if (alias->d_parent == dentry->d_parent && | |
373 | alias->d_name.hash == dentry->d_name.hash && | |
374 | alias->d_name.len == dentry->d_name.len && | |
375 | memcmp(alias->d_name.name, dentry->d_name.name, | |
376 | dentry->d_name.len) == 0) | |
377 | invalid_alias = alias; | |
378 | spin_unlock(&alias->d_lock); | |
379 | ||
380 | if (invalid_alias) | |
381 | break; | |
382 | } | |
383 | alias = invalid_alias ?: discon_alias ?: NULL; | |
384 | if (alias) { | |
385 | spin_lock(&alias->d_lock); | |
386 | dget_dlock(alias); | |
387 | spin_unlock(&alias->d_lock); | |
388 | } | |
389 | ll_unlock_dcache(inode); | |
390 | ||
391 | return alias; | |
392 | } | |
393 | ||
394 | /* | |
395 | * Similar to d_splice_alias(), but lustre treats invalid alias | |
396 | * similar to DCACHE_DISCONNECTED, and tries to use it anyway. | |
397 | */ | |
398 | struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de) | |
399 | { | |
400 | struct dentry *new; | |
3ea8f3bc | 401 | int rc; |
d7e09d03 PT |
402 | |
403 | if (inode) { | |
404 | new = ll_find_alias(inode, de); | |
405 | if (new) { | |
3ea8f3bc LS |
406 | rc = ll_d_init(new); |
407 | if (rc < 0) { | |
408 | dput(new); | |
409 | return ERR_PTR(rc); | |
410 | } | |
d7e09d03 PT |
411 | d_move(new, de); |
412 | iput(inode); | |
413 | CDEBUG(D_DENTRY, | |
414 | "Reuse dentry %p inode %p refc %d flags %#x\n", | |
193deee1 | 415 | new, new->d_inode, d_count(new), new->d_flags); |
d7e09d03 PT |
416 | return new; |
417 | } | |
418 | } | |
3ea8f3bc LS |
419 | rc = ll_d_init(de); |
420 | if (rc < 0) | |
421 | return ERR_PTR(rc); | |
d7e09d03 PT |
422 | d_add(de, inode); |
423 | CDEBUG(D_DENTRY, "Add dentry %p inode %p refc %d flags %#x\n", | |
193deee1 | 424 | de, de->d_inode, d_count(de), de->d_flags); |
d7e09d03 PT |
425 | return de; |
426 | } | |
427 | ||
2d95f10e JH |
428 | static int ll_lookup_it_finish(struct ptlrpc_request *request, |
429 | struct lookup_intent *it, | |
430 | struct inode *parent, struct dentry **de) | |
d7e09d03 | 431 | { |
d7e09d03 PT |
432 | struct inode *inode = NULL; |
433 | __u64 bits = 0; | |
434 | int rc; | |
d7e09d03 PT |
435 | |
436 | /* NB 1 request reference will be taken away by ll_intent_lock() | |
437 | * when I return */ | |
438 | CDEBUG(D_DENTRY, "it %p it_disposition %x\n", it, | |
439 | it->d.lustre.it_disposition); | |
440 | if (!it_disposition(it, DISP_LOOKUP_NEG)) { | |
441 | rc = ll_prep_inode(&inode, request, (*de)->d_sb, it); | |
442 | if (rc) | |
0a3bdb00 | 443 | return rc; |
d7e09d03 PT |
444 | |
445 | ll_set_lock_data(ll_i2sbi(parent)->ll_md_exp, inode, it, &bits); | |
446 | ||
447 | /* We used to query real size from OSTs here, but actually | |
448 | this is not needed. For stat() calls size would be updated | |
449 | from subsequent do_revalidate()->ll_inode_revalidate_it() in | |
450 | 2.4 and | |
451 | vfs_getattr_it->ll_getattr()->ll_inode_revalidate_it() in 2.6 | |
452 | Everybody else who needs correct file size would call | |
453 | ll_glimpse_size or some equivalent themselves anyway. | |
454 | Also see bug 7198. */ | |
455 | } | |
456 | ||
457 | /* Only hash *de if it is unhashed (new dentry). | |
d0a0acc3 | 458 | * Atoimc_open may passing hashed dentries for open. |
d7e09d03 | 459 | */ |
3ea8f3bc | 460 | if (d_unhashed(*de)) { |
7486bc06 SP |
461 | struct dentry *alias; |
462 | ||
463 | alias = ll_splice_alias(inode, *de); | |
464 | if (IS_ERR(alias)) | |
465 | return PTR_ERR(alias); | |
466 | *de = alias; | |
08a78a27 | 467 | } else if (!it_disposition(it, DISP_LOOKUP_NEG) && |
468 | !it_disposition(it, DISP_OPEN_CREATE)) { | |
469 | /* With DISP_OPEN_CREATE dentry will | |
470 | instantiated in ll_create_it. */ | |
471 | LASSERT((*de)->d_inode == NULL); | |
472 | d_instantiate(*de, inode); | |
3ea8f3bc | 473 | } |
d7e09d03 PT |
474 | |
475 | if (!it_disposition(it, DISP_LOOKUP_NEG)) { | |
476 | /* we have lookup look - unhide dentry */ | |
477 | if (bits & MDS_INODELOCK_LOOKUP) | |
478 | d_lustre_revalidate(*de); | |
479 | } else if (!it_disposition(it, DISP_OPEN_CREATE)) { | |
480 | /* If file created on server, don't depend on parent UPDATE | |
481 | * lock to unhide it. It is left hidden and next lookup can | |
482 | * find it in ll_splice_alias. | |
483 | */ | |
484 | /* Check that parent has UPDATE lock. */ | |
485 | struct lookup_intent parent_it = { | |
486 | .it_op = IT_GETATTR, | |
487 | .d.lustre.it_lock_handle = 0 }; | |
488 | ||
489 | if (md_revalidate_lock(ll_i2mdexp(parent), &parent_it, | |
490 | &ll_i2info(parent)->lli_fid, NULL)) { | |
491 | d_lustre_revalidate(*de); | |
492 | ll_intent_release(&parent_it); | |
493 | } | |
494 | } | |
495 | ||
0a3bdb00 | 496 | return 0; |
d7e09d03 PT |
497 | } |
498 | ||
499 | static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry, | |
500 | struct lookup_intent *it, int lookup_flags) | |
501 | { | |
502 | struct lookup_intent lookup_it = { .it_op = IT_LOOKUP }; | |
503 | struct dentry *save = dentry, *retval; | |
504 | struct ptlrpc_request *req = NULL; | |
505 | struct md_op_data *op_data; | |
d7e09d03 PT |
506 | __u32 opc; |
507 | int rc; | |
d7e09d03 PT |
508 | |
509 | if (dentry->d_name.len > ll_i2sbi(parent)->ll_namelen) | |
0a3bdb00 | 510 | return ERR_PTR(-ENAMETOOLONG); |
d7e09d03 | 511 | |
09561a53 AV |
512 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p),intent=%s\n", |
513 | dentry, parent->i_ino, | |
d7e09d03 PT |
514 | parent->i_generation, parent, LL_IT2STR(it)); |
515 | ||
516 | if (d_mountpoint(dentry)) | |
517 | CERROR("Tell Peter, lookup on mtpt, it %s\n", LL_IT2STR(it)); | |
518 | ||
2d95f10e JH |
519 | if (it == NULL || it->it_op == IT_GETXATTR) |
520 | it = &lookup_it; | |
d7e09d03 | 521 | |
d7e09d03 PT |
522 | if (it->it_op == IT_GETATTR) { |
523 | rc = ll_statahead_enter(parent, &dentry, 0); | |
524 | if (rc == 1) { | |
525 | if (dentry == save) | |
34e1f2bb JL |
526 | retval = NULL; |
527 | else | |
528 | retval = dentry; | |
529 | goto out; | |
d7e09d03 PT |
530 | } |
531 | } | |
532 | ||
1e8a576e | 533 | if (it->it_op & IT_CREAT) |
d7e09d03 PT |
534 | opc = LUSTRE_OPC_CREATE; |
535 | else | |
536 | opc = LUSTRE_OPC_ANY; | |
537 | ||
538 | op_data = ll_prep_md_op_data(NULL, parent, NULL, dentry->d_name.name, | |
539 | dentry->d_name.len, lookup_flags, opc, | |
540 | NULL); | |
541 | if (IS_ERR(op_data)) | |
0a3bdb00 | 542 | return (void *)op_data; |
d7e09d03 PT |
543 | |
544 | /* enforce umask if acl disabled or MDS doesn't support umask */ | |
545 | if (!IS_POSIXACL(parent) || !exp_connect_umask(ll_i2mdexp(parent))) | |
546 | it->it_create_mode &= ~current_umask(); | |
547 | ||
548 | rc = md_intent_lock(ll_i2mdexp(parent), op_data, NULL, 0, it, | |
549 | lookup_flags, &req, ll_md_blocking_ast, 0); | |
550 | ll_finish_md_op_data(op_data); | |
34e1f2bb JL |
551 | if (rc < 0) { |
552 | retval = ERR_PTR(rc); | |
553 | goto out; | |
554 | } | |
d7e09d03 | 555 | |
2d95f10e | 556 | rc = ll_lookup_it_finish(req, it, parent, &dentry); |
d7e09d03 PT |
557 | if (rc != 0) { |
558 | ll_intent_release(it); | |
34e1f2bb JL |
559 | retval = ERR_PTR(rc); |
560 | goto out; | |
d7e09d03 PT |
561 | } |
562 | ||
563 | if ((it->it_op & IT_OPEN) && dentry->d_inode && | |
564 | !S_ISREG(dentry->d_inode->i_mode) && | |
565 | !S_ISDIR(dentry->d_inode->i_mode)) { | |
e22fdcc8 | 566 | ll_release_openhandle(dentry->d_inode, it); |
d7e09d03 PT |
567 | } |
568 | ll_lookup_finish_locks(it, dentry); | |
569 | ||
570 | if (dentry == save) | |
34e1f2bb | 571 | retval = NULL; |
d7e09d03 | 572 | else |
34e1f2bb JL |
573 | retval = dentry; |
574 | goto out; | |
d7e09d03 PT |
575 | out: |
576 | if (req) | |
577 | ptlrpc_req_finished(req); | |
578 | if (it->it_op == IT_GETATTR && (retval == NULL || retval == dentry)) | |
579 | ll_statahead_mark(parent, dentry); | |
580 | return retval; | |
581 | } | |
582 | ||
583 | static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, | |
584 | unsigned int flags) | |
585 | { | |
586 | struct lookup_intent *itp, it = { .it_op = IT_GETATTR }; | |
587 | struct dentry *de; | |
588 | ||
09561a53 AV |
589 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p),flags=%u\n", |
590 | dentry, parent->i_ino, | |
d7e09d03 PT |
591 | parent->i_generation, parent, flags); |
592 | ||
593 | /* Optimize away (CREATE && !OPEN). Let .create handle the race. */ | |
4f2fb455 | 594 | if ((flags & LOOKUP_CREATE) && !(flags & LOOKUP_OPEN)) |
d7e09d03 | 595 | return NULL; |
d7e09d03 PT |
596 | |
597 | if (flags & (LOOKUP_PARENT|LOOKUP_OPEN|LOOKUP_CREATE)) | |
598 | itp = NULL; | |
599 | else | |
600 | itp = ⁢ | |
601 | de = ll_lookup_it(parent, dentry, itp, 0); | |
602 | ||
603 | if (itp != NULL) | |
604 | ll_intent_release(itp); | |
605 | ||
606 | return de; | |
607 | } | |
608 | ||
609 | /* | |
610 | * For cached negative dentry and new dentry, handle lookup/create/open | |
611 | * together. | |
612 | */ | |
613 | static int ll_atomic_open(struct inode *dir, struct dentry *dentry, | |
614 | struct file *file, unsigned open_flags, | |
615 | umode_t mode, int *opened) | |
616 | { | |
617 | struct lookup_intent *it; | |
618 | struct dentry *de; | |
619 | long long lookup_flags = LOOKUP_OPEN; | |
620 | int rc = 0; | |
d7e09d03 | 621 | |
09561a53 | 622 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p),file %p," |
d7e09d03 | 623 | "open_flags %x,mode %x opened %d\n", |
09561a53 | 624 | dentry, dir->i_ino, |
d7e09d03 PT |
625 | dir->i_generation, dir, file, open_flags, mode, *opened); |
626 | ||
496a51bd | 627 | it = kzalloc(sizeof(*it), GFP_NOFS); |
d7e09d03 | 628 | if (!it) |
0a3bdb00 | 629 | return -ENOMEM; |
d7e09d03 PT |
630 | |
631 | it->it_op = IT_OPEN; | |
1e8a576e | 632 | if (open_flags & O_CREAT) { |
d7e09d03 PT |
633 | it->it_op |= IT_CREAT; |
634 | lookup_flags |= LOOKUP_CREATE; | |
635 | } | |
636 | it->it_create_mode = (mode & S_IALLUGO) | S_IFREG; | |
637 | it->it_flags = (open_flags & ~O_ACCMODE) | OPEN_FMODE(open_flags); | |
638 | ||
639 | /* Dentry added to dcache tree in ll_lookup_it */ | |
640 | de = ll_lookup_it(dir, dentry, it, lookup_flags); | |
641 | if (IS_ERR(de)) | |
642 | rc = PTR_ERR(de); | |
643 | else if (de != NULL) | |
644 | dentry = de; | |
645 | ||
646 | if (!rc) { | |
647 | if (it_disposition(it, DISP_OPEN_CREATE)) { | |
648 | /* Dentry instantiated in ll_create_it. */ | |
649 | rc = ll_create_it(dir, dentry, mode, it); | |
650 | if (rc) { | |
651 | /* We dget in ll_splice_alias. */ | |
652 | if (de != NULL) | |
653 | dput(de); | |
654 | goto out_release; | |
655 | } | |
656 | ||
657 | *opened |= FILE_CREATED; | |
658 | } | |
659 | if (dentry->d_inode && it_disposition(it, DISP_OPEN_OPEN)) { | |
660 | /* Open dentry. */ | |
661 | if (S_ISFIFO(dentry->d_inode->i_mode)) { | |
662 | /* We cannot call open here as it would | |
663 | * deadlock. | |
664 | */ | |
665 | if (it_disposition(it, DISP_ENQ_OPEN_REF)) | |
666 | ptlrpc_req_finished( | |
667 | (struct ptlrpc_request *) | |
668 | it->d.lustre.it_data); | |
669 | rc = finish_no_open(file, de); | |
670 | } else { | |
671 | file->private_data = it; | |
672 | rc = finish_open(file, dentry, NULL, opened); | |
673 | /* We dget in ll_splice_alias. finish_open takes | |
674 | * care of dget for fd open. | |
675 | */ | |
676 | if (de != NULL) | |
677 | dput(de); | |
678 | } | |
679 | } else { | |
680 | rc = finish_no_open(file, de); | |
681 | } | |
682 | } | |
683 | ||
684 | out_release: | |
685 | ll_intent_release(it); | |
686 | OBD_FREE(it, sizeof(*it)); | |
687 | ||
0a3bdb00 | 688 | return rc; |
d7e09d03 PT |
689 | } |
690 | ||
691 | ||
692 | /* We depend on "mode" being set with the proper file type/umask by now */ | |
2d95f10e | 693 | static struct inode *ll_create_node(struct inode *dir, struct lookup_intent *it) |
d7e09d03 PT |
694 | { |
695 | struct inode *inode = NULL; | |
696 | struct ptlrpc_request *request = NULL; | |
697 | struct ll_sb_info *sbi = ll_i2sbi(dir); | |
698 | int rc; | |
d7e09d03 PT |
699 | |
700 | LASSERT(it && it->d.lustre.it_disposition); | |
701 | ||
702 | LASSERT(it_disposition(it, DISP_ENQ_CREATE_REF)); | |
703 | request = it->d.lustre.it_data; | |
704 | it_clear_disposition(it, DISP_ENQ_CREATE_REF); | |
705 | rc = ll_prep_inode(&inode, request, dir->i_sb, it); | |
34e1f2bb JL |
706 | if (rc) { |
707 | inode = ERR_PTR(rc); | |
708 | goto out; | |
709 | } | |
d7e09d03 PT |
710 | |
711 | LASSERT(ll_d_hlist_empty(&inode->i_dentry)); | |
712 | ||
713 | /* We asked for a lock on the directory, but were granted a | |
714 | * lock on the inode. Since we finally have an inode pointer, | |
715 | * stuff it in the lock. */ | |
716 | CDEBUG(D_DLMTRACE, "setting l_ast_data to inode %p (%lu/%u)\n", | |
717 | inode, inode->i_ino, inode->i_generation); | |
718 | ll_set_lock_data(sbi->ll_md_exp, inode, it, NULL); | |
d7e09d03 PT |
719 | out: |
720 | ptlrpc_req_finished(request); | |
721 | return inode; | |
722 | } | |
723 | ||
724 | /* | |
725 | * By the time this is called, we already have created the directory cache | |
726 | * entry for the new file, but it is so far negative - it has no inode. | |
727 | * | |
728 | * We defer creating the OBD object(s) until open, to keep the intent and | |
729 | * non-intent code paths similar, and also because we do not have the MDS | |
730 | * inode number before calling ll_create_node() (which is needed for LOV), | |
731 | * so we would need to do yet another RPC to the MDS to store the LOV EA | |
732 | * data on the MDS. If needed, we would pass the PACKED lmm as data and | |
733 | * lmm_size in datalen (the MDS still has code which will handle that). | |
734 | * | |
735 | * If the create succeeds, we fill in the inode information | |
736 | * with d_instantiate(). | |
737 | */ | |
738 | static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, | |
739 | struct lookup_intent *it) | |
740 | { | |
741 | struct inode *inode; | |
742 | int rc = 0; | |
d7e09d03 | 743 | |
09561a53 AV |
744 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p),intent=%s\n", |
745 | dentry, dir->i_ino, | |
d7e09d03 PT |
746 | dir->i_generation, dir, LL_IT2STR(it)); |
747 | ||
748 | rc = it_open_error(DISP_OPEN_CREATE, it); | |
749 | if (rc) | |
0a3bdb00 | 750 | return rc; |
d7e09d03 | 751 | |
2d95f10e | 752 | inode = ll_create_node(dir, it); |
d7e09d03 | 753 | if (IS_ERR(inode)) |
0a3bdb00 | 754 | return PTR_ERR(inode); |
d7e09d03 | 755 | |
d7e09d03 | 756 | d_instantiate(dentry, inode); |
0a3bdb00 | 757 | return 0; |
d7e09d03 PT |
758 | } |
759 | ||
760 | static void ll_update_times(struct ptlrpc_request *request, | |
761 | struct inode *inode) | |
762 | { | |
763 | struct mdt_body *body = req_capsule_server_get(&request->rq_pill, | |
764 | &RMF_MDT_BODY); | |
765 | ||
766 | LASSERT(body); | |
767 | if (body->valid & OBD_MD_FLMTIME && | |
768 | body->mtime > LTIME_S(inode->i_mtime)) { | |
b0f5aad5 | 769 | CDEBUG(D_INODE, "setting ino %lu mtime from %lu to %llu\n", |
d7e09d03 PT |
770 | inode->i_ino, LTIME_S(inode->i_mtime), body->mtime); |
771 | LTIME_S(inode->i_mtime) = body->mtime; | |
772 | } | |
773 | if (body->valid & OBD_MD_FLCTIME && | |
774 | body->ctime > LTIME_S(inode->i_ctime)) | |
775 | LTIME_S(inode->i_ctime) = body->ctime; | |
776 | } | |
777 | ||
b2037bb6 | 778 | static int ll_new_node(struct inode *dir, struct dentry *dentry, |
d7e09d03 | 779 | const char *tgt, int mode, int rdev, |
b2037bb6 | 780 | __u32 opc) |
d7e09d03 PT |
781 | { |
782 | struct ptlrpc_request *request = NULL; | |
783 | struct md_op_data *op_data; | |
784 | struct inode *inode = NULL; | |
785 | struct ll_sb_info *sbi = ll_i2sbi(dir); | |
786 | int tgt_len = 0; | |
787 | int err; | |
788 | ||
d7e09d03 PT |
789 | if (unlikely(tgt != NULL)) |
790 | tgt_len = strlen(tgt) + 1; | |
791 | ||
b2037bb6 AV |
792 | op_data = ll_prep_md_op_data(NULL, dir, NULL, |
793 | dentry->d_name.name, | |
794 | dentry->d_name.len, | |
795 | 0, opc, NULL); | |
34e1f2bb JL |
796 | if (IS_ERR(op_data)) { |
797 | err = PTR_ERR(op_data); | |
798 | goto err_exit; | |
799 | } | |
d7e09d03 PT |
800 | |
801 | err = md_create(sbi->ll_md_exp, op_data, tgt, tgt_len, mode, | |
4b1a25f0 PT |
802 | from_kuid(&init_user_ns, current_fsuid()), |
803 | from_kgid(&init_user_ns, current_fsgid()), | |
d7e09d03 PT |
804 | cfs_curproc_cap_pack(), rdev, &request); |
805 | ll_finish_md_op_data(op_data); | |
806 | if (err) | |
34e1f2bb | 807 | goto err_exit; |
d7e09d03 PT |
808 | |
809 | ll_update_times(request, dir); | |
810 | ||
b2037bb6 AV |
811 | err = ll_prep_inode(&inode, request, dir->i_sb, NULL); |
812 | if (err) | |
813 | goto err_exit; | |
d7e09d03 | 814 | |
b2037bb6 | 815 | d_instantiate(dentry, inode); |
d7e09d03 PT |
816 | err_exit: |
817 | ptlrpc_req_finished(request); | |
818 | ||
819 | return err; | |
820 | } | |
821 | ||
d6689e5f AV |
822 | static int ll_mknod(struct inode *dir, struct dentry *dchild, |
823 | umode_t mode, dev_t rdev) | |
d7e09d03 PT |
824 | { |
825 | int err; | |
d7e09d03 | 826 | |
d6689e5f AV |
827 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p) mode %o dev %x\n", |
828 | dchild, dir->i_ino, dir->i_generation, dir, | |
829 | mode, old_encode_dev(rdev)); | |
d7e09d03 PT |
830 | |
831 | if (!IS_POSIXACL(dir) || !exp_connect_umask(ll_i2mdexp(dir))) | |
832 | mode &= ~current_umask(); | |
833 | ||
834 | switch (mode & S_IFMT) { | |
835 | case 0: | |
836 | mode |= S_IFREG; /* for mode = 0 case, fallthrough */ | |
837 | case S_IFREG: | |
838 | case S_IFCHR: | |
839 | case S_IFBLK: | |
840 | case S_IFIFO: | |
841 | case S_IFSOCK: | |
b2037bb6 AV |
842 | err = ll_new_node(dir, dchild, NULL, mode, |
843 | old_encode_dev(rdev), | |
d7e09d03 PT |
844 | LUSTRE_OPC_MKNOD); |
845 | break; | |
846 | case S_IFDIR: | |
847 | err = -EPERM; | |
848 | break; | |
849 | default: | |
850 | err = -EINVAL; | |
851 | } | |
852 | ||
853 | if (!err) | |
854 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_MKNOD, 1); | |
855 | ||
0a3bdb00 | 856 | return err; |
d7e09d03 PT |
857 | } |
858 | ||
859 | /* | |
860 | * Plain create. Intent create is handled in atomic_open. | |
861 | */ | |
862 | static int ll_create_nd(struct inode *dir, struct dentry *dentry, | |
863 | umode_t mode, bool want_excl) | |
864 | { | |
865 | int rc; | |
866 | ||
09561a53 | 867 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p)," |
d7e09d03 | 868 | "flags=%u, excl=%d\n", |
09561a53 | 869 | dentry, dir->i_ino, |
d7e09d03 PT |
870 | dir->i_generation, dir, mode, want_excl); |
871 | ||
d6689e5f | 872 | rc = ll_mknod(dir, dentry, mode, 0); |
d7e09d03 PT |
873 | |
874 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_CREATE, 1); | |
875 | ||
09561a53 AV |
876 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd, unhashed %d\n", |
877 | dentry, d_unhashed(dentry)); | |
d7e09d03 PT |
878 | |
879 | return rc; | |
880 | } | |
881 | ||
d7e09d03 PT |
882 | static int ll_link_generic(struct inode *src, struct inode *dir, |
883 | struct qstr *name, struct dentry *dchild) | |
884 | { | |
885 | struct ll_sb_info *sbi = ll_i2sbi(dir); | |
886 | struct ptlrpc_request *request = NULL; | |
887 | struct md_op_data *op_data; | |
888 | int err; | |
889 | ||
d7e09d03 PT |
890 | CDEBUG(D_VFSTRACE, |
891 | "VFS Op: inode=%lu/%u(%p), dir=%lu/%u(%p), target=%.*s\n", | |
892 | src->i_ino, src->i_generation, src, dir->i_ino, | |
893 | dir->i_generation, dir, name->len, name->name); | |
894 | ||
895 | op_data = ll_prep_md_op_data(NULL, src, dir, name->name, name->len, | |
896 | 0, LUSTRE_OPC_ANY, NULL); | |
897 | if (IS_ERR(op_data)) | |
0a3bdb00 | 898 | return PTR_ERR(op_data); |
d7e09d03 PT |
899 | |
900 | err = md_link(sbi->ll_md_exp, op_data, &request); | |
901 | ll_finish_md_op_data(op_data); | |
902 | if (err) | |
34e1f2bb | 903 | goto out; |
d7e09d03 PT |
904 | |
905 | ll_update_times(request, dir); | |
906 | ll_stats_ops_tally(sbi, LPROC_LL_LINK, 1); | |
d7e09d03 PT |
907 | out: |
908 | ptlrpc_req_finished(request); | |
0a3bdb00 | 909 | return err; |
d7e09d03 PT |
910 | } |
911 | ||
d7e09d03 PT |
912 | /* Try to find the child dentry by its name. |
913 | If found, put the result fid into @fid. */ | |
914 | static void ll_get_child_fid(struct inode * dir, struct qstr *name, | |
915 | struct lu_fid *fid) | |
916 | { | |
917 | struct dentry *parent, *child; | |
918 | ||
946e51f2 | 919 | parent = ll_d_hlist_entry(dir->i_dentry, struct dentry, d_u.d_alias); |
d7e09d03 PT |
920 | child = d_lookup(parent, name); |
921 | if (child) { | |
922 | if (child->d_inode) | |
923 | *fid = *ll_inode2fid(child->d_inode); | |
924 | dput(child); | |
925 | } | |
926 | } | |
927 | ||
d7e09d03 PT |
928 | /** |
929 | * Remove dir entry | |
930 | **/ | |
931 | int ll_rmdir_entry(struct inode *dir, char *name, int namelen) | |
932 | { | |
933 | struct ptlrpc_request *request = NULL; | |
934 | struct md_op_data *op_data; | |
935 | int rc; | |
d7e09d03 PT |
936 | |
937 | CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", | |
938 | namelen, name, dir->i_ino, dir->i_generation, dir); | |
939 | ||
940 | op_data = ll_prep_md_op_data(NULL, dir, NULL, name, strlen(name), | |
941 | S_IFDIR, LUSTRE_OPC_ANY, NULL); | |
942 | if (IS_ERR(op_data)) | |
0a3bdb00 | 943 | return PTR_ERR(op_data); |
d7e09d03 PT |
944 | op_data->op_cli_flags |= CLI_RM_ENTRY; |
945 | rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); | |
946 | ll_finish_md_op_data(op_data); | |
947 | if (rc == 0) { | |
948 | ll_update_times(request, dir); | |
949 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_RMDIR, 1); | |
950 | } | |
951 | ||
952 | ptlrpc_req_finished(request); | |
0a3bdb00 | 953 | return rc; |
d7e09d03 PT |
954 | } |
955 | ||
956 | int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir) | |
957 | { | |
958 | struct mdt_body *body; | |
959 | struct lov_mds_md *eadata; | |
960 | struct lov_stripe_md *lsm = NULL; | |
961 | struct obd_trans_info oti = { 0 }; | |
962 | struct obdo *oa; | |
963 | struct obd_capa *oc = NULL; | |
964 | int rc; | |
d7e09d03 PT |
965 | |
966 | /* req is swabbed so this is safe */ | |
967 | body = req_capsule_server_get(&request->rq_pill, &RMF_MDT_BODY); | |
968 | if (!(body->valid & OBD_MD_FLEASIZE)) | |
0a3bdb00 | 969 | return 0; |
d7e09d03 PT |
970 | |
971 | if (body->eadatasize == 0) { | |
972 | CERROR("OBD_MD_FLEASIZE set but eadatasize zero\n"); | |
34e1f2bb JL |
973 | rc = -EPROTO; |
974 | goto out; | |
d7e09d03 PT |
975 | } |
976 | ||
977 | /* The MDS sent back the EA because we unlinked the last reference | |
978 | * to this file. Use this EA to unlink the objects on the OST. | |
979 | * It's opaque so we don't swab here; we leave it to obd_unpackmd() to | |
980 | * check it is complete and sensible. */ | |
981 | eadata = req_capsule_server_sized_get(&request->rq_pill, &RMF_MDT_MD, | |
982 | body->eadatasize); | |
983 | LASSERT(eadata != NULL); | |
984 | ||
985 | rc = obd_unpackmd(ll_i2dtexp(dir), &lsm, eadata, body->eadatasize); | |
986 | if (rc < 0) { | |
987 | CERROR("obd_unpackmd: %d\n", rc); | |
34e1f2bb | 988 | goto out; |
d7e09d03 PT |
989 | } |
990 | LASSERT(rc >= sizeof(*lsm)); | |
991 | ||
992 | OBDO_ALLOC(oa); | |
34e1f2bb JL |
993 | if (oa == NULL) { |
994 | rc = -ENOMEM; | |
995 | goto out_free_memmd; | |
996 | } | |
d7e09d03 PT |
997 | |
998 | oa->o_oi = lsm->lsm_oi; | |
999 | oa->o_mode = body->mode & S_IFMT; | |
1000 | oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLGROUP; | |
1001 | ||
1002 | if (body->valid & OBD_MD_FLCOOKIE) { | |
1003 | oa->o_valid |= OBD_MD_FLCOOKIE; | |
1004 | oti.oti_logcookies = | |
1005 | req_capsule_server_sized_get(&request->rq_pill, | |
1006 | &RMF_LOGCOOKIES, | |
1007 | sizeof(struct llog_cookie) * | |
1008 | lsm->lsm_stripe_count); | |
1009 | if (oti.oti_logcookies == NULL) { | |
1010 | oa->o_valid &= ~OBD_MD_FLCOOKIE; | |
1011 | body->valid &= ~OBD_MD_FLCOOKIE; | |
1012 | } | |
1013 | } | |
1014 | ||
1015 | if (body->valid & OBD_MD_FLOSSCAPA) { | |
1016 | rc = md_unpack_capa(ll_i2mdexp(dir), request, &RMF_CAPA2, &oc); | |
1017 | if (rc) | |
34e1f2bb | 1018 | goto out_free_memmd; |
d7e09d03 PT |
1019 | } |
1020 | ||
1021 | rc = obd_destroy(NULL, ll_i2dtexp(dir), oa, lsm, &oti, | |
1022 | ll_i2mdexp(dir), oc); | |
1023 | capa_put(oc); | |
1024 | if (rc) | |
1025 | CERROR("obd destroy objid "DOSTID" error %d\n", | |
1026 | POSTID(&lsm->lsm_oi), rc); | |
1027 | out_free_memmd: | |
1028 | obd_free_memmd(ll_i2dtexp(dir), &lsm); | |
1029 | OBDO_FREE(oa); | |
1030 | out: | |
1031 | return rc; | |
1032 | } | |
1033 | ||
1034 | /* ll_unlink_generic() doesn't update the inode with the new link count. | |
1035 | * Instead, ll_ddelete() and ll_d_iput() will update it based upon if there | |
1036 | * is any lock existing. They will recycle dentries and inodes based upon locks | |
1037 | * too. b=20433 */ | |
1038 | static int ll_unlink_generic(struct inode *dir, struct dentry *dparent, | |
1039 | struct dentry *dchild, struct qstr *name) | |
1040 | { | |
1041 | struct ptlrpc_request *request = NULL; | |
1042 | struct md_op_data *op_data; | |
1043 | int rc; | |
d7e09d03 PT |
1044 | CDEBUG(D_VFSTRACE, "VFS Op:name=%.*s,dir=%lu/%u(%p)\n", |
1045 | name->len, name->name, dir->i_ino, dir->i_generation, dir); | |
1046 | ||
1047 | /* | |
1048 | * XXX: unlink bind mountpoint maybe call to here, | |
1049 | * just check it as vfs_unlink does. | |
1050 | */ | |
1051 | if (unlikely(ll_d_mountpoint(dparent, dchild, name))) | |
0a3bdb00 | 1052 | return -EBUSY; |
d7e09d03 PT |
1053 | |
1054 | op_data = ll_prep_md_op_data(NULL, dir, NULL, name->name, | |
1055 | name->len, 0, LUSTRE_OPC_ANY, NULL); | |
1056 | if (IS_ERR(op_data)) | |
0a3bdb00 | 1057 | return PTR_ERR(op_data); |
d7e09d03 PT |
1058 | |
1059 | ll_get_child_fid(dir, name, &op_data->op_fid3); | |
1060 | op_data->op_fid2 = op_data->op_fid3; | |
1061 | rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); | |
1062 | ll_finish_md_op_data(op_data); | |
1063 | if (rc) | |
34e1f2bb | 1064 | goto out; |
d7e09d03 PT |
1065 | |
1066 | ll_update_times(request, dir); | |
1067 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_UNLINK, 1); | |
1068 | ||
1069 | rc = ll_objects_destroy(request, dir); | |
1070 | out: | |
1071 | ptlrpc_req_finished(request); | |
0a3bdb00 | 1072 | return rc; |
d7e09d03 PT |
1073 | } |
1074 | ||
1075 | static int ll_rename_generic(struct inode *src, struct dentry *src_dparent, | |
1076 | struct dentry *src_dchild, struct qstr *src_name, | |
1077 | struct inode *tgt, struct dentry *tgt_dparent, | |
1078 | struct dentry *tgt_dchild, struct qstr *tgt_name) | |
1079 | { | |
1080 | struct ptlrpc_request *request = NULL; | |
1081 | struct ll_sb_info *sbi = ll_i2sbi(src); | |
1082 | struct md_op_data *op_data; | |
1083 | int err; | |
29aaf496 | 1084 | |
1d8cb70c GD |
1085 | CDEBUG(D_VFSTRACE, |
1086 | "VFS Op:oldname=%.*s,src_dir=%lu/%u(%p),newname=%.*s," | |
d7e09d03 PT |
1087 | "tgt_dir=%lu/%u(%p)\n", src_name->len, src_name->name, |
1088 | src->i_ino, src->i_generation, src, tgt_name->len, | |
1089 | tgt_name->name, tgt->i_ino, tgt->i_generation, tgt); | |
1090 | ||
1091 | if (unlikely(ll_d_mountpoint(src_dparent, src_dchild, src_name) || | |
1092 | ll_d_mountpoint(tgt_dparent, tgt_dchild, tgt_name))) | |
0a3bdb00 | 1093 | return -EBUSY; |
d7e09d03 PT |
1094 | |
1095 | op_data = ll_prep_md_op_data(NULL, src, tgt, NULL, 0, 0, | |
1096 | LUSTRE_OPC_ANY, NULL); | |
1097 | if (IS_ERR(op_data)) | |
0a3bdb00 | 1098 | return PTR_ERR(op_data); |
d7e09d03 PT |
1099 | |
1100 | ll_get_child_fid(src, src_name, &op_data->op_fid3); | |
1101 | ll_get_child_fid(tgt, tgt_name, &op_data->op_fid4); | |
1102 | err = md_rename(sbi->ll_md_exp, op_data, | |
1103 | src_name->name, src_name->len, | |
1104 | tgt_name->name, tgt_name->len, &request); | |
1105 | ll_finish_md_op_data(op_data); | |
1106 | if (!err) { | |
1107 | ll_update_times(request, src); | |
1108 | ll_update_times(request, tgt); | |
1109 | ll_stats_ops_tally(sbi, LPROC_LL_RENAME, 1); | |
1110 | err = ll_objects_destroy(request, src); | |
1111 | } | |
1112 | ||
1113 | ptlrpc_req_finished(request); | |
1114 | ||
0a3bdb00 | 1115 | return err; |
d7e09d03 PT |
1116 | } |
1117 | ||
d7e09d03 PT |
1118 | static int ll_unlink(struct inode * dir, struct dentry *dentry) |
1119 | { | |
1120 | return ll_unlink_generic(dir, NULL, dentry, &dentry->d_name); | |
1121 | } | |
1122 | ||
1123 | static int ll_mkdir(struct inode *dir, struct dentry *dentry, ll_umode_t mode) | |
1124 | { | |
7c2f9094 AV |
1125 | int err; |
1126 | ||
1127 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p)\n", | |
1128 | dentry, dir->i_ino, dir->i_generation, dir); | |
1129 | ||
1130 | if (!IS_POSIXACL(dir) || !exp_connect_umask(ll_i2mdexp(dir))) | |
1131 | mode &= ~current_umask(); | |
1132 | mode = (mode & (S_IRWXUGO|S_ISVTX)) | S_IFDIR; | |
b2037bb6 | 1133 | err = ll_new_node(dir, dentry, NULL, mode, 0, LUSTRE_OPC_MKDIR); |
7c2f9094 AV |
1134 | |
1135 | if (!err) | |
1136 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_MKDIR, 1); | |
1137 | ||
1138 | return err; | |
d7e09d03 PT |
1139 | } |
1140 | ||
1141 | static int ll_rmdir(struct inode *dir, struct dentry *dentry) | |
1142 | { | |
55dec617 AV |
1143 | struct ptlrpc_request *request = NULL; |
1144 | struct md_op_data *op_data; | |
1145 | int rc; | |
1146 | ||
1147 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p)\n", | |
1148 | dentry, dir->i_ino, dir->i_generation, dir); | |
1149 | ||
1150 | if (unlikely(ll_d_mountpoint(NULL, dentry, &dentry->d_name))) | |
1151 | return -EBUSY; | |
1152 | ||
1153 | op_data = ll_prep_md_op_data(NULL, dir, NULL, | |
1154 | dentry->d_name.name, | |
1155 | dentry->d_name.len, | |
1156 | S_IFDIR, LUSTRE_OPC_ANY, NULL); | |
1157 | if (IS_ERR(op_data)) | |
1158 | return PTR_ERR(op_data); | |
1159 | ||
1160 | ll_get_child_fid(dir, &dentry->d_name, &op_data->op_fid3); | |
1161 | op_data->op_fid2 = op_data->op_fid3; | |
1162 | rc = md_unlink(ll_i2sbi(dir)->ll_md_exp, op_data, &request); | |
1163 | ll_finish_md_op_data(op_data); | |
1164 | if (rc == 0) { | |
1165 | ll_update_times(request, dir); | |
1166 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_RMDIR, 1); | |
1167 | } | |
1168 | ||
1169 | ptlrpc_req_finished(request); | |
1170 | return rc; | |
d7e09d03 PT |
1171 | } |
1172 | ||
1173 | static int ll_symlink(struct inode *dir, struct dentry *dentry, | |
1174 | const char *oldname) | |
1175 | { | |
60dd654e AV |
1176 | int err; |
1177 | ||
1178 | CDEBUG(D_VFSTRACE, "VFS Op:name=%pd,dir=%lu/%u(%p),target=%.*s\n", | |
1179 | dentry, dir->i_ino, dir->i_generation, | |
1180 | dir, 3000, oldname); | |
1181 | ||
b2037bb6 AV |
1182 | err = ll_new_node(dir, dentry, oldname, S_IFLNK | S_IRWXUGO, |
1183 | 0, LUSTRE_OPC_SYMLINK); | |
60dd654e AV |
1184 | |
1185 | if (!err) | |
1186 | ll_stats_ops_tally(ll_i2sbi(dir), LPROC_LL_SYMLINK, 1); | |
1187 | ||
1188 | return err; | |
d7e09d03 PT |
1189 | } |
1190 | ||
1191 | static int ll_link(struct dentry *old_dentry, struct inode *dir, | |
1192 | struct dentry *new_dentry) | |
1193 | { | |
1194 | return ll_link_generic(old_dentry->d_inode, dir, &new_dentry->d_name, | |
1195 | new_dentry); | |
1196 | } | |
1197 | ||
1198 | static int ll_rename(struct inode *old_dir, struct dentry *old_dentry, | |
1199 | struct inode *new_dir, struct dentry *new_dentry) | |
1200 | { | |
1201 | int err; | |
1202 | err = ll_rename_generic(old_dir, NULL, | |
1203 | old_dentry, &old_dentry->d_name, | |
1204 | new_dir, NULL, new_dentry, | |
1205 | &new_dentry->d_name); | |
1206 | if (!err) { | |
1207 | d_move(old_dentry, new_dentry); | |
1208 | } | |
1209 | return err; | |
1210 | } | |
1211 | ||
2d95f10e | 1212 | const struct inode_operations ll_dir_inode_operations = { |
d7e09d03 PT |
1213 | .mknod = ll_mknod, |
1214 | .atomic_open = ll_atomic_open, | |
1215 | .lookup = ll_lookup_nd, | |
1216 | .create = ll_create_nd, | |
1217 | /* We need all these non-raw things for NFSD, to not patch it. */ | |
1218 | .unlink = ll_unlink, | |
1219 | .mkdir = ll_mkdir, | |
1220 | .rmdir = ll_rmdir, | |
1221 | .symlink = ll_symlink, | |
1222 | .link = ll_link, | |
1223 | .rename = ll_rename, | |
1224 | .setattr = ll_setattr, | |
1225 | .getattr = ll_getattr, | |
1226 | .permission = ll_inode_permission, | |
1227 | .setxattr = ll_setxattr, | |
1228 | .getxattr = ll_getxattr, | |
1229 | .listxattr = ll_listxattr, | |
1230 | .removexattr = ll_removexattr, | |
1231 | .get_acl = ll_get_acl, | |
1232 | }; | |
1233 | ||
2d95f10e | 1234 | const struct inode_operations ll_special_inode_operations = { |
d7e09d03 PT |
1235 | .setattr = ll_setattr, |
1236 | .getattr = ll_getattr, | |
1237 | .permission = ll_inode_permission, | |
1238 | .setxattr = ll_setxattr, | |
1239 | .getxattr = ll_getxattr, | |
1240 | .listxattr = ll_listxattr, | |
1241 | .removexattr = ll_removexattr, | |
1242 | .get_acl = ll_get_acl, | |
1243 | }; |