]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
22 | * Copyright 2008 Sun Microsystems, Inc. All rights reserved. | |
23 | * Use is subject to license terms. | |
24 | */ | |
25 | ||
26 | /* Portions Copyright 2007 Jeremy Teo */ | |
27 | ||
28 | #pragma ident "@(#)zfs_znode.c 1.34 08/04/27 SMI" | |
29 | ||
30 | #ifdef _KERNEL | |
31 | #include <sys/types.h> | |
32 | #include <sys/param.h> | |
33 | #include <sys/time.h> | |
34 | #include <sys/systm.h> | |
35 | #include <sys/sysmacros.h> | |
36 | #include <sys/resource.h> | |
37 | #include <sys/mntent.h> | |
38 | #include <sys/mkdev.h> | |
39 | #include <sys/u8_textprep.h> | |
40 | #include <sys/dsl_dataset.h> | |
41 | #include <sys/vfs.h> | |
42 | #include <sys/vfs_opreg.h> | |
43 | #include <sys/vnode.h> | |
44 | #include <sys/file.h> | |
45 | #include <sys/kmem.h> | |
46 | #include <sys/errno.h> | |
47 | #include <sys/unistd.h> | |
48 | #include <sys/mode.h> | |
49 | #include <sys/atomic.h> | |
50 | #include <vm/pvn.h> | |
51 | #include "fs/fs_subr.h" | |
52 | #include <sys/zfs_dir.h> | |
53 | #include <sys/zfs_acl.h> | |
54 | #include <sys/zfs_ioctl.h> | |
55 | #include <sys/zfs_rlock.h> | |
56 | #include <sys/zfs_fuid.h> | |
57 | #include <sys/fs/zfs.h> | |
58 | #include <sys/kidmap.h> | |
59 | #endif /* _KERNEL */ | |
60 | ||
61 | #include <sys/dmu.h> | |
62 | #include <sys/refcount.h> | |
63 | #include <sys/stat.h> | |
64 | #include <sys/zap.h> | |
65 | #include <sys/zfs_znode.h> | |
66 | ||
67 | #include "zfs_prop.h" | |
68 | ||
69 | /* | |
70 | * Functions needed for userland (ie: libzpool) are not put under | |
71 | * #ifdef_KERNEL; the rest of the functions have dependencies | |
72 | * (such as VFS logic) that will not compile easily in userland. | |
73 | */ | |
74 | #ifdef _KERNEL | |
75 | struct kmem_cache *znode_cache = NULL; | |
76 | ||
77 | /*ARGSUSED*/ | |
78 | static void | |
79 | znode_evict_error(dmu_buf_t *dbuf, void *user_ptr) | |
80 | { | |
81 | /* | |
82 | * We should never drop all dbuf refs without first clearing | |
83 | * the eviction callback. | |
84 | */ | |
85 | panic("evicting znode %p\n", user_ptr); | |
86 | } | |
87 | ||
88 | /*ARGSUSED*/ | |
89 | static int | |
90 | zfs_znode_cache_constructor(void *buf, void *cdrarg, int kmflags) | |
91 | { | |
92 | znode_t *zp = buf; | |
93 | ||
94 | zp->z_vnode = vn_alloc(KM_SLEEP); | |
95 | zp->z_vnode->v_data = (caddr_t)zp; | |
96 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
97 | rw_init(&zp->z_map_lock, NULL, RW_DEFAULT, NULL); | |
98 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); | |
99 | rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); | |
100 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); | |
101 | ||
102 | mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); | |
103 | avl_create(&zp->z_range_avl, zfs_range_compare, | |
104 | sizeof (rl_t), offsetof(rl_t, r_node)); | |
105 | ||
106 | zp->z_dbuf = NULL; | |
107 | zp->z_dirlocks = 0; | |
108 | return (0); | |
109 | } | |
110 | ||
111 | /*ARGSUSED*/ | |
112 | static void | |
113 | zfs_znode_cache_destructor(void *buf, void *cdarg) | |
114 | { | |
115 | znode_t *zp = buf; | |
116 | ||
117 | ASSERT(zp->z_dirlocks == 0); | |
118 | mutex_destroy(&zp->z_lock); | |
119 | rw_destroy(&zp->z_map_lock); | |
120 | rw_destroy(&zp->z_parent_lock); | |
121 | rw_destroy(&zp->z_name_lock); | |
122 | mutex_destroy(&zp->z_acl_lock); | |
123 | avl_destroy(&zp->z_range_avl); | |
124 | mutex_destroy(&zp->z_range_lock); | |
125 | ||
126 | ASSERT(zp->z_dbuf == NULL); | |
127 | ASSERT(ZTOV(zp)->v_count == 0); | |
128 | vn_free(ZTOV(zp)); | |
129 | } | |
130 | ||
131 | void | |
132 | zfs_znode_init(void) | |
133 | { | |
134 | /* | |
135 | * Initialize zcache | |
136 | */ | |
137 | ASSERT(znode_cache == NULL); | |
138 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
139 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
140 | zfs_znode_cache_destructor, NULL, NULL, NULL, 0); | |
141 | } | |
142 | ||
143 | void | |
144 | zfs_znode_fini(void) | |
145 | { | |
146 | /* | |
147 | * Cleanup vfs & vnode ops | |
148 | */ | |
149 | zfs_remove_op_tables(); | |
150 | ||
151 | /* | |
152 | * Cleanup zcache | |
153 | */ | |
154 | if (znode_cache) | |
155 | kmem_cache_destroy(znode_cache); | |
156 | znode_cache = NULL; | |
157 | } | |
158 | ||
159 | struct vnodeops *zfs_dvnodeops; | |
160 | struct vnodeops *zfs_fvnodeops; | |
161 | struct vnodeops *zfs_symvnodeops; | |
162 | struct vnodeops *zfs_xdvnodeops; | |
163 | struct vnodeops *zfs_evnodeops; | |
164 | ||
165 | void | |
166 | zfs_remove_op_tables() | |
167 | { | |
168 | /* | |
169 | * Remove vfs ops | |
170 | */ | |
171 | ASSERT(zfsfstype); | |
172 | (void) vfs_freevfsops_by_type(zfsfstype); | |
173 | zfsfstype = 0; | |
174 | ||
175 | /* | |
176 | * Remove vnode ops | |
177 | */ | |
178 | if (zfs_dvnodeops) | |
179 | vn_freevnodeops(zfs_dvnodeops); | |
180 | if (zfs_fvnodeops) | |
181 | vn_freevnodeops(zfs_fvnodeops); | |
182 | if (zfs_symvnodeops) | |
183 | vn_freevnodeops(zfs_symvnodeops); | |
184 | if (zfs_xdvnodeops) | |
185 | vn_freevnodeops(zfs_xdvnodeops); | |
186 | if (zfs_evnodeops) | |
187 | vn_freevnodeops(zfs_evnodeops); | |
188 | ||
189 | zfs_dvnodeops = NULL; | |
190 | zfs_fvnodeops = NULL; | |
191 | zfs_symvnodeops = NULL; | |
192 | zfs_xdvnodeops = NULL; | |
193 | zfs_evnodeops = NULL; | |
194 | } | |
195 | ||
196 | extern const fs_operation_def_t zfs_dvnodeops_template[]; | |
197 | extern const fs_operation_def_t zfs_fvnodeops_template[]; | |
198 | extern const fs_operation_def_t zfs_xdvnodeops_template[]; | |
199 | extern const fs_operation_def_t zfs_symvnodeops_template[]; | |
200 | extern const fs_operation_def_t zfs_evnodeops_template[]; | |
201 | ||
202 | int | |
203 | zfs_create_op_tables() | |
204 | { | |
205 | int error; | |
206 | ||
207 | /* | |
208 | * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() | |
209 | * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). | |
210 | * In this case we just return as the ops vectors are already set up. | |
211 | */ | |
212 | if (zfs_dvnodeops) | |
213 | return (0); | |
214 | ||
215 | error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, | |
216 | &zfs_dvnodeops); | |
217 | if (error) | |
218 | return (error); | |
219 | ||
220 | error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, | |
221 | &zfs_fvnodeops); | |
222 | if (error) | |
223 | return (error); | |
224 | ||
225 | error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, | |
226 | &zfs_symvnodeops); | |
227 | if (error) | |
228 | return (error); | |
229 | ||
230 | error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, | |
231 | &zfs_xdvnodeops); | |
232 | if (error) | |
233 | return (error); | |
234 | ||
235 | error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, | |
236 | &zfs_evnodeops); | |
237 | ||
238 | return (error); | |
239 | } | |
240 | ||
241 | /* | |
242 | * zfs_init_fs - Initialize the zfsvfs struct and the file system | |
243 | * incore "master" object. Verify version compatibility. | |
244 | */ | |
245 | int | |
246 | zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp, cred_t *cr) | |
247 | { | |
248 | extern int zfsfstype; | |
249 | ||
250 | objset_t *os = zfsvfs->z_os; | |
251 | int i, error; | |
252 | dmu_object_info_t doi; | |
253 | uint64_t fsid_guid; | |
254 | uint64_t zval; | |
255 | ||
256 | *zpp = NULL; | |
257 | ||
258 | /* | |
259 | * XXX - hack to auto-create the pool root filesystem at | |
260 | * the first attempted mount. | |
261 | */ | |
262 | if (dmu_object_info(os, MASTER_NODE_OBJ, &doi) == ENOENT) { | |
263 | dmu_tx_t *tx = dmu_tx_create(os); | |
264 | uint64_t zpl_version; | |
265 | nvlist_t *zprops; | |
266 | ||
267 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* master */ | |
268 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* del queue */ | |
269 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); /* root node */ | |
270 | error = dmu_tx_assign(tx, TXG_WAIT); | |
271 | ASSERT3U(error, ==, 0); | |
272 | if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID) | |
273 | zpl_version = ZPL_VERSION; | |
274 | else | |
275 | zpl_version = ZPL_VERSION_FUID - 1; | |
276 | ||
277 | VERIFY(nvlist_alloc(&zprops, NV_UNIQUE_NAME, KM_SLEEP) == 0); | |
278 | VERIFY(nvlist_add_uint64(zprops, | |
279 | zfs_prop_to_name(ZFS_PROP_VERSION), zpl_version) == 0); | |
280 | zfs_create_fs(os, cr, zprops, tx); | |
281 | nvlist_free(zprops); | |
282 | dmu_tx_commit(tx); | |
283 | } | |
284 | ||
285 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); | |
286 | if (error) { | |
287 | return (error); | |
288 | } else if (zfsvfs->z_version > ZPL_VERSION) { | |
289 | (void) printf("Mismatched versions: File system " | |
290 | "is version %llu on-disk format, which is " | |
291 | "incompatible with this software version %lld!", | |
292 | (u_longlong_t)zfsvfs->z_version, ZPL_VERSION); | |
293 | return (ENOTSUP); | |
294 | } | |
295 | ||
296 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) | |
297 | return (error); | |
298 | zfsvfs->z_norm = (int)zval; | |
299 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
300 | return (error); | |
301 | zfsvfs->z_utf8 = (zval != 0); | |
302 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
303 | return (error); | |
304 | zfsvfs->z_case = (uint_t)zval; | |
305 | /* | |
306 | * Fold case on file systems that are always or sometimes case | |
307 | * insensitive. | |
308 | */ | |
309 | if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || | |
310 | zfsvfs->z_case == ZFS_CASE_MIXED) | |
311 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; | |
312 | ||
313 | /* | |
314 | * The fsid is 64 bits, composed of an 8-bit fs type, which | |
315 | * separates our fsid from any other filesystem types, and a | |
316 | * 56-bit objset unique ID. The objset unique ID is unique to | |
317 | * all objsets open on this system, provided by unique_create(). | |
318 | * The 8-bit fs type must be put in the low bits of fsid[1] | |
319 | * because that's where other Solaris filesystems put it. | |
320 | */ | |
321 | fsid_guid = dmu_objset_fsid_guid(os); | |
322 | ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); | |
323 | zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid; | |
324 | zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | | |
325 | zfsfstype & 0xFF; | |
326 | ||
327 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
328 | &zfsvfs->z_root); | |
329 | if (error) | |
330 | return (error); | |
331 | ASSERT(zfsvfs->z_root != 0); | |
332 | ||
333 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
334 | &zfsvfs->z_unlinkedobj); | |
335 | if (error) | |
336 | return (error); | |
337 | ||
338 | /* | |
339 | * Initialize zget mutex's | |
340 | */ | |
341 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) | |
342 | mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); | |
343 | ||
344 | error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp); | |
345 | if (error) { | |
346 | /* | |
347 | * On error, we destroy the mutexes here since it's not | |
348 | * possible for the caller to determine if the mutexes were | |
349 | * initialized properly. | |
350 | */ | |
351 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) | |
352 | mutex_destroy(&zfsvfs->z_hold_mtx[i]); | |
353 | return (error); | |
354 | } | |
355 | ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root); | |
356 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
357 | &zfsvfs->z_fuid_obj); | |
358 | if (error == ENOENT) | |
359 | error = 0; | |
360 | ||
361 | return (0); | |
362 | } | |
363 | ||
364 | /* | |
365 | * define a couple of values we need available | |
366 | * for both 64 and 32 bit environments. | |
367 | */ | |
368 | #ifndef NBITSMINOR64 | |
369 | #define NBITSMINOR64 32 | |
370 | #endif | |
371 | #ifndef MAXMAJ64 | |
372 | #define MAXMAJ64 0xffffffffUL | |
373 | #endif | |
374 | #ifndef MAXMIN64 | |
375 | #define MAXMIN64 0xffffffffUL | |
376 | #endif | |
377 | ||
378 | /* | |
379 | * Create special expldev for ZFS private use. | |
380 | * Can't use standard expldev since it doesn't do | |
381 | * what we want. The standard expldev() takes a | |
382 | * dev32_t in LP64 and expands it to a long dev_t. | |
383 | * We need an interface that takes a dev32_t in ILP32 | |
384 | * and expands it to a long dev_t. | |
385 | */ | |
386 | static uint64_t | |
387 | zfs_expldev(dev_t dev) | |
388 | { | |
389 | #ifndef _LP64 | |
390 | major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32; | |
391 | return (((uint64_t)major << NBITSMINOR64) | | |
392 | ((minor_t)dev & MAXMIN32)); | |
393 | #else | |
394 | return (dev); | |
395 | #endif | |
396 | } | |
397 | ||
398 | /* | |
399 | * Special cmpldev for ZFS private use. | |
400 | * Can't use standard cmpldev since it takes | |
401 | * a long dev_t and compresses it to dev32_t in | |
402 | * LP64. We need to do a compaction of a long dev_t | |
403 | * to a dev32_t in ILP32. | |
404 | */ | |
405 | dev_t | |
406 | zfs_cmpldev(uint64_t dev) | |
407 | { | |
408 | #ifndef _LP64 | |
409 | minor_t minor = (minor_t)dev & MAXMIN64; | |
410 | major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64; | |
411 | ||
412 | if (major > MAXMAJ32 || minor > MAXMIN32) | |
413 | return (NODEV32); | |
414 | ||
415 | return (((dev32_t)major << NBITSMINOR32) | minor); | |
416 | #else | |
417 | return (dev); | |
418 | #endif | |
419 | } | |
420 | ||
421 | static void | |
422 | zfs_znode_dmu_init(znode_t *zp, dmu_buf_t *db) | |
423 | { | |
424 | znode_t *nzp; | |
425 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
426 | ||
427 | ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp))); | |
428 | ||
429 | mutex_enter(&zp->z_lock); | |
430 | ||
431 | ASSERT(zp->z_dbuf == NULL); | |
432 | zp->z_dbuf = db; | |
433 | nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error); | |
434 | ||
435 | /* | |
436 | * there should be no | |
437 | * concurrent zgets on this object. | |
438 | */ | |
439 | if (nzp != NULL) | |
440 | panic("existing znode %p for dbuf %p", nzp, db); | |
441 | ||
442 | /* | |
443 | * Slap on VROOT if we are the root znode | |
444 | */ | |
445 | if (zp->z_id == zfsvfs->z_root) | |
446 | ZTOV(zp)->v_flag |= VROOT; | |
447 | ||
448 | mutex_exit(&zp->z_lock); | |
449 | vn_exists(ZTOV(zp)); | |
450 | } | |
451 | ||
452 | void | |
453 | zfs_znode_dmu_fini(znode_t *zp) | |
454 | { | |
455 | dmu_buf_t *db = zp->z_dbuf; | |
456 | ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp)) || zp->z_unlinked || | |
457 | RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); | |
458 | ASSERT(zp->z_dbuf != NULL); | |
459 | zp->z_dbuf = NULL; | |
460 | VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL)); | |
461 | dmu_buf_rele(db, NULL); | |
462 | } | |
463 | ||
464 | /* | |
465 | * Construct a new znode/vnode and intialize. | |
466 | * | |
467 | * This does not do a call to dmu_set_user() that is | |
468 | * up to the caller to do, in case you don't want to | |
469 | * return the znode | |
470 | */ | |
471 | static znode_t * | |
472 | zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz) | |
473 | { | |
474 | znode_t *zp; | |
475 | vnode_t *vp; | |
476 | ||
477 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); | |
478 | ||
479 | ASSERT(zp->z_dirlocks == NULL); | |
480 | ASSERT(zp->z_dbuf == NULL); | |
481 | ||
482 | zp->z_phys = NULL; | |
483 | zp->z_zfsvfs = zfsvfs; | |
484 | zp->z_unlinked = 0; | |
485 | zp->z_atime_dirty = 0; | |
486 | zp->z_mapcnt = 0; | |
487 | zp->z_last_itx = 0; | |
488 | zp->z_id = db->db_object; | |
489 | zp->z_blksz = blksz; | |
490 | zp->z_seq = 0x7A4653; | |
491 | zp->z_sync_cnt = 0; | |
492 | ||
493 | vp = ZTOV(zp); | |
494 | vn_reinit(vp); | |
495 | ||
496 | zfs_znode_dmu_init(zp, db); | |
497 | ||
498 | zp->z_gen = zp->z_phys->zp_gen; | |
499 | ||
500 | mutex_enter(&zfsvfs->z_znodes_lock); | |
501 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
502 | mutex_exit(&zfsvfs->z_znodes_lock); | |
503 | ||
504 | vp->v_vfsp = zfsvfs->z_parent->z_vfs; | |
505 | vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode); | |
506 | ||
507 | switch (vp->v_type) { | |
508 | case VDIR: | |
509 | if (zp->z_phys->zp_flags & ZFS_XATTR) { | |
510 | vn_setops(vp, zfs_xdvnodeops); | |
511 | vp->v_flag |= V_XATTRDIR; | |
512 | } else { | |
513 | vn_setops(vp, zfs_dvnodeops); | |
514 | } | |
515 | zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ | |
516 | break; | |
517 | case VBLK: | |
518 | case VCHR: | |
519 | vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev); | |
520 | /*FALLTHROUGH*/ | |
521 | case VFIFO: | |
522 | case VSOCK: | |
523 | case VDOOR: | |
524 | vn_setops(vp, zfs_fvnodeops); | |
525 | break; | |
526 | case VREG: | |
527 | vp->v_flag |= VMODSORT; | |
528 | vn_setops(vp, zfs_fvnodeops); | |
529 | break; | |
530 | case VLNK: | |
531 | vn_setops(vp, zfs_symvnodeops); | |
532 | break; | |
533 | default: | |
534 | vn_setops(vp, zfs_evnodeops); | |
535 | break; | |
536 | } | |
537 | ||
538 | VFS_HOLD(zfsvfs->z_vfs); | |
539 | return (zp); | |
540 | } | |
541 | ||
542 | /* | |
543 | * Create a new DMU object to hold a zfs znode. | |
544 | * | |
545 | * IN: dzp - parent directory for new znode | |
546 | * vap - file attributes for new znode | |
547 | * tx - dmu transaction id for zap operations | |
548 | * cr - credentials of caller | |
549 | * flag - flags: | |
550 | * IS_ROOT_NODE - new object will be root | |
551 | * IS_XATTR - new object is an attribute | |
552 | * IS_REPLAY - intent log replay | |
553 | * bonuslen - length of bonus buffer | |
554 | * setaclp - File/Dir initial ACL | |
555 | * fuidp - Tracks fuid allocation. | |
556 | * | |
557 | * OUT: zpp - allocated znode | |
558 | * | |
559 | */ | |
560 | void | |
561 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
562 | uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_t *setaclp, | |
563 | zfs_fuid_info_t **fuidp) | |
564 | { | |
565 | dmu_buf_t *db; | |
566 | znode_phys_t *pzp; | |
567 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
568 | timestruc_t now; | |
569 | uint64_t gen, obj; | |
570 | int err; | |
571 | ||
572 | ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); | |
573 | ||
574 | if (zfsvfs->z_assign >= TXG_INITIAL) { /* ZIL replay */ | |
575 | obj = vap->va_nodeid; | |
576 | flag |= IS_REPLAY; | |
577 | now = vap->va_ctime; /* see zfs_replay_create() */ | |
578 | gen = vap->va_nblocks; /* ditto */ | |
579 | } else { | |
580 | obj = 0; | |
581 | gethrestime(&now); | |
582 | gen = dmu_tx_get_txg(tx); | |
583 | } | |
584 | ||
585 | /* | |
586 | * Create a new DMU object. | |
587 | */ | |
588 | /* | |
589 | * There's currently no mechanism for pre-reading the blocks that will | |
590 | * be to needed allocate a new object, so we accept the small chance | |
591 | * that there will be an i/o error and we will fail one of the | |
592 | * assertions below. | |
593 | */ | |
594 | if (vap->va_type == VDIR) { | |
595 | if (flag & IS_REPLAY) { | |
596 | err = zap_create_claim_norm(zfsvfs->z_os, obj, | |
597 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
598 | DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); | |
599 | ASSERT3U(err, ==, 0); | |
600 | } else { | |
601 | obj = zap_create_norm(zfsvfs->z_os, | |
602 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
603 | DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); | |
604 | } | |
605 | } else { | |
606 | if (flag & IS_REPLAY) { | |
607 | err = dmu_object_claim(zfsvfs->z_os, obj, | |
608 | DMU_OT_PLAIN_FILE_CONTENTS, 0, | |
609 | DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); | |
610 | ASSERT3U(err, ==, 0); | |
611 | } else { | |
612 | obj = dmu_object_alloc(zfsvfs->z_os, | |
613 | DMU_OT_PLAIN_FILE_CONTENTS, 0, | |
614 | DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); | |
615 | } | |
616 | } | |
617 | VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db)); | |
618 | dmu_buf_will_dirty(db, tx); | |
619 | ||
620 | /* | |
621 | * Initialize the znode physical data to zero. | |
622 | */ | |
623 | ASSERT(db->db_size >= sizeof (znode_phys_t)); | |
624 | bzero(db->db_data, db->db_size); | |
625 | pzp = db->db_data; | |
626 | ||
627 | /* | |
628 | * If this is the root, fix up the half-initialized parent pointer | |
629 | * to reference the just-allocated physical data area. | |
630 | */ | |
631 | if (flag & IS_ROOT_NODE) { | |
632 | dzp->z_dbuf = db; | |
633 | dzp->z_phys = pzp; | |
634 | dzp->z_id = obj; | |
635 | } | |
636 | ||
637 | /* | |
638 | * If parent is an xattr, so am I. | |
639 | */ | |
640 | if (dzp->z_phys->zp_flags & ZFS_XATTR) | |
641 | flag |= IS_XATTR; | |
642 | ||
643 | if (vap->va_type == VBLK || vap->va_type == VCHR) { | |
644 | pzp->zp_rdev = zfs_expldev(vap->va_rdev); | |
645 | } | |
646 | ||
647 | if (zfsvfs->z_use_fuids) | |
648 | pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; | |
649 | ||
650 | if (vap->va_type == VDIR) { | |
651 | pzp->zp_size = 2; /* contents ("." and "..") */ | |
652 | pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; | |
653 | } | |
654 | ||
655 | pzp->zp_parent = dzp->z_id; | |
656 | if (flag & IS_XATTR) | |
657 | pzp->zp_flags |= ZFS_XATTR; | |
658 | ||
659 | pzp->zp_gen = gen; | |
660 | ||
661 | ZFS_TIME_ENCODE(&now, pzp->zp_crtime); | |
662 | ZFS_TIME_ENCODE(&now, pzp->zp_ctime); | |
663 | ||
664 | if (vap->va_mask & AT_ATIME) { | |
665 | ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); | |
666 | } else { | |
667 | ZFS_TIME_ENCODE(&now, pzp->zp_atime); | |
668 | } | |
669 | ||
670 | if (vap->va_mask & AT_MTIME) { | |
671 | ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); | |
672 | } else { | |
673 | ZFS_TIME_ENCODE(&now, pzp->zp_mtime); | |
674 | } | |
675 | ||
676 | pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode); | |
677 | if (!(flag & IS_ROOT_NODE)) { | |
678 | ZFS_OBJ_HOLD_ENTER(zfsvfs, obj) | |
679 | *zpp = zfs_znode_alloc(zfsvfs, db, 0); | |
680 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); | |
681 | } else { | |
682 | /* | |
683 | * If we are creating the root node, the "parent" we | |
684 | * passed in is the znode for the root. | |
685 | */ | |
686 | *zpp = dzp; | |
687 | } | |
688 | zfs_perm_init(*zpp, dzp, flag, vap, tx, cr, setaclp, fuidp); | |
689 | } | |
690 | ||
691 | void | |
692 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap) | |
693 | { | |
694 | xoptattr_t *xoap; | |
695 | ||
696 | xoap = xva_getxoptattr(xvap); | |
697 | ASSERT(xoap); | |
698 | ||
699 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
700 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime); | |
701 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
702 | } | |
703 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
704 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly); | |
705 | XVA_SET_RTN(xvap, XAT_READONLY); | |
706 | } | |
707 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
708 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden); | |
709 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
710 | } | |
711 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
712 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system); | |
713 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
714 | } | |
715 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
716 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive); | |
717 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
718 | } | |
719 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
720 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable); | |
721 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
722 | } | |
723 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
724 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink); | |
725 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
726 | } | |
727 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
728 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly); | |
729 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
730 | } | |
731 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
732 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump); | |
733 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
734 | } | |
735 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
736 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque); | |
737 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
738 | } | |
739 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
740 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
741 | xoap->xoa_av_quarantined); | |
742 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
743 | } | |
744 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
745 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified); | |
746 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
747 | } | |
748 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
749 | (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp, | |
750 | sizeof (xoap->xoa_av_scanstamp)); | |
751 | zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP; | |
752 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
753 | } | |
754 | } | |
755 | ||
756 | int | |
757 | zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) | |
758 | { | |
759 | dmu_object_info_t doi; | |
760 | dmu_buf_t *db; | |
761 | znode_t *zp; | |
762 | int err; | |
763 | ||
764 | *zpp = NULL; | |
765 | ||
766 | ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); | |
767 | ||
768 | err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); | |
769 | if (err) { | |
770 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
771 | return (err); | |
772 | } | |
773 | ||
774 | dmu_object_info_from_db(db, &doi); | |
775 | if (doi.doi_bonus_type != DMU_OT_ZNODE || | |
776 | doi.doi_bonus_size < sizeof (znode_phys_t)) { | |
777 | dmu_buf_rele(db, NULL); | |
778 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
779 | return (EINVAL); | |
780 | } | |
781 | ||
782 | zp = dmu_buf_get_user(db); | |
783 | if (zp != NULL) { | |
784 | mutex_enter(&zp->z_lock); | |
785 | ||
786 | /* | |
787 | * Since we do immediate eviction of the z_dbuf, we | |
788 | * should never find a dbuf with a znode that doesn't | |
789 | * know about the dbuf. | |
790 | */ | |
791 | ASSERT3P(zp->z_dbuf, ==, db); | |
792 | ASSERT3U(zp->z_id, ==, obj_num); | |
793 | if (zp->z_unlinked) { | |
794 | err = ENOENT; | |
795 | } else { | |
796 | VN_HOLD(ZTOV(zp)); | |
797 | *zpp = zp; | |
798 | err = 0; | |
799 | } | |
800 | dmu_buf_rele(db, NULL); | |
801 | mutex_exit(&zp->z_lock); | |
802 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
803 | return (err); | |
804 | } | |
805 | ||
806 | /* | |
807 | * Not found create new znode/vnode | |
808 | */ | |
809 | zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size); | |
810 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
811 | *zpp = zp; | |
812 | return (0); | |
813 | } | |
814 | ||
815 | int | |
816 | zfs_rezget(znode_t *zp) | |
817 | { | |
818 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
819 | dmu_object_info_t doi; | |
820 | dmu_buf_t *db; | |
821 | uint64_t obj_num = zp->z_id; | |
822 | int err; | |
823 | ||
824 | ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); | |
825 | ||
826 | err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); | |
827 | if (err) { | |
828 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
829 | return (err); | |
830 | } | |
831 | ||
832 | dmu_object_info_from_db(db, &doi); | |
833 | if (doi.doi_bonus_type != DMU_OT_ZNODE || | |
834 | doi.doi_bonus_size < sizeof (znode_phys_t)) { | |
835 | dmu_buf_rele(db, NULL); | |
836 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
837 | return (EINVAL); | |
838 | } | |
839 | ||
840 | if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) { | |
841 | dmu_buf_rele(db, NULL); | |
842 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
843 | return (EIO); | |
844 | } | |
845 | ||
846 | zfs_znode_dmu_init(zp, db); | |
847 | zp->z_unlinked = (zp->z_phys->zp_links == 0); | |
848 | zp->z_blksz = doi.doi_data_block_size; | |
849 | ||
850 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); | |
851 | ||
852 | return (0); | |
853 | } | |
854 | ||
855 | void | |
856 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
857 | { | |
858 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
859 | uint64_t obj = zp->z_id; | |
860 | ||
861 | ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); | |
862 | if (zp->z_phys->zp_acl.z_acl_extern_obj) { | |
863 | VERIFY(0 == dmu_object_free(zfsvfs->z_os, | |
864 | zp->z_phys->zp_acl.z_acl_extern_obj, tx)); | |
865 | } | |
866 | VERIFY(0 == dmu_object_free(zfsvfs->z_os, obj, tx)); | |
867 | zfs_znode_dmu_fini(zp); | |
868 | ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); | |
869 | zfs_znode_free(zp); | |
870 | } | |
871 | ||
872 | void | |
873 | zfs_zinactive(znode_t *zp) | |
874 | { | |
875 | vnode_t *vp = ZTOV(zp); | |
876 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
877 | uint64_t z_id = zp->z_id; | |
878 | ||
879 | ASSERT(zp->z_dbuf && zp->z_phys); | |
880 | ||
881 | /* | |
882 | * Don't allow a zfs_zget() while were trying to release this znode | |
883 | */ | |
884 | ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); | |
885 | ||
886 | mutex_enter(&zp->z_lock); | |
887 | mutex_enter(&vp->v_lock); | |
888 | vp->v_count--; | |
889 | if (vp->v_count > 0 || vn_has_cached_data(vp)) { | |
890 | /* | |
891 | * If the hold count is greater than zero, somebody has | |
892 | * obtained a new reference on this znode while we were | |
893 | * processing it here, so we are done. If we still have | |
894 | * mapped pages then we are also done, since we don't | |
895 | * want to inactivate the znode until the pages get pushed. | |
896 | * | |
897 | * XXX - if vn_has_cached_data(vp) is true, but count == 0, | |
898 | * this seems like it would leave the znode hanging with | |
899 | * no chance to go inactive... | |
900 | */ | |
901 | mutex_exit(&vp->v_lock); | |
902 | mutex_exit(&zp->z_lock); | |
903 | ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); | |
904 | return; | |
905 | } | |
906 | mutex_exit(&vp->v_lock); | |
907 | ||
908 | /* | |
909 | * If this was the last reference to a file with no links, | |
910 | * remove the file from the file system. | |
911 | */ | |
912 | if (zp->z_unlinked) { | |
913 | mutex_exit(&zp->z_lock); | |
914 | ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); | |
915 | zfs_rmnode(zp); | |
916 | return; | |
917 | } | |
918 | mutex_exit(&zp->z_lock); | |
919 | zfs_znode_dmu_fini(zp); | |
920 | ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); | |
921 | zfs_znode_free(zp); | |
922 | } | |
923 | ||
924 | void | |
925 | zfs_znode_free(znode_t *zp) | |
926 | { | |
927 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
928 | ||
929 | vn_invalid(ZTOV(zp)); | |
930 | ||
931 | mutex_enter(&zfsvfs->z_znodes_lock); | |
932 | list_remove(&zfsvfs->z_all_znodes, zp); | |
933 | mutex_exit(&zfsvfs->z_znodes_lock); | |
934 | ||
935 | kmem_cache_free(znode_cache, zp); | |
936 | ||
937 | VFS_RELE(zfsvfs->z_vfs); | |
938 | } | |
939 | ||
940 | void | |
941 | zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx) | |
942 | { | |
943 | timestruc_t now; | |
944 | ||
945 | ASSERT(MUTEX_HELD(&zp->z_lock)); | |
946 | ||
947 | gethrestime(&now); | |
948 | ||
949 | if (tx) { | |
950 | dmu_buf_will_dirty(zp->z_dbuf, tx); | |
951 | zp->z_atime_dirty = 0; | |
952 | zp->z_seq++; | |
953 | } else { | |
954 | zp->z_atime_dirty = 1; | |
955 | } | |
956 | ||
957 | if (flag & AT_ATIME) | |
958 | ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime); | |
959 | ||
960 | if (flag & AT_MTIME) { | |
961 | ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime); | |
962 | if (zp->z_zfsvfs->z_use_fuids) | |
963 | zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED); | |
964 | } | |
965 | ||
966 | if (flag & AT_CTIME) { | |
967 | ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime); | |
968 | if (zp->z_zfsvfs->z_use_fuids) | |
969 | zp->z_phys->zp_flags |= ZFS_ARCHIVE; | |
970 | } | |
971 | } | |
972 | ||
973 | /* | |
974 | * Update the requested znode timestamps with the current time. | |
975 | * If we are in a transaction, then go ahead and mark the znode | |
976 | * dirty in the transaction so the timestamps will go to disk. | |
977 | * Otherwise, we will get pushed next time the znode is updated | |
978 | * in a transaction, or when this znode eventually goes inactive. | |
979 | * | |
980 | * Why is this OK? | |
981 | * 1 - Only the ACCESS time is ever updated outside of a transaction. | |
982 | * 2 - Multiple consecutive updates will be collapsed into a single | |
983 | * znode update by the transaction grouping semantics of the DMU. | |
984 | */ | |
985 | void | |
986 | zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx) | |
987 | { | |
988 | mutex_enter(&zp->z_lock); | |
989 | zfs_time_stamper_locked(zp, flag, tx); | |
990 | mutex_exit(&zp->z_lock); | |
991 | } | |
992 | ||
993 | /* | |
994 | * Grow the block size for a file. | |
995 | * | |
996 | * IN: zp - znode of file to free data in. | |
997 | * size - requested block size | |
998 | * tx - open transaction. | |
999 | * | |
1000 | * NOTE: this function assumes that the znode is write locked. | |
1001 | */ | |
1002 | void | |
1003 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1004 | { | |
1005 | int error; | |
1006 | u_longlong_t dummy; | |
1007 | ||
1008 | if (size <= zp->z_blksz) | |
1009 | return; | |
1010 | /* | |
1011 | * If the file size is already greater than the current blocksize, | |
1012 | * we will not grow. If there is more than one block in a file, | |
1013 | * the blocksize cannot change. | |
1014 | */ | |
1015 | if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz) | |
1016 | return; | |
1017 | ||
1018 | error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, | |
1019 | size, 0, tx); | |
1020 | if (error == ENOTSUP) | |
1021 | return; | |
1022 | ASSERT3U(error, ==, 0); | |
1023 | ||
1024 | /* What blocksize did we actually get? */ | |
1025 | dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy); | |
1026 | } | |
1027 | ||
1028 | /* | |
1029 | * This is a dummy interface used when pvn_vplist_dirty() should *not* | |
1030 | * be calling back into the fs for a putpage(). E.g.: when truncating | |
1031 | * a file, the pages being "thrown away* don't need to be written out. | |
1032 | */ | |
1033 | /* ARGSUSED */ | |
1034 | static int | |
1035 | zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, | |
1036 | int flags, cred_t *cr) | |
1037 | { | |
1038 | ASSERT(0); | |
1039 | return (0); | |
1040 | } | |
1041 | ||
1042 | /* | |
1043 | * Free space in a file. | |
1044 | * | |
1045 | * IN: zp - znode of file to free data in. | |
1046 | * off - start of section to free. | |
1047 | * len - length of section to free (0 => to EOF). | |
1048 | * flag - current file open mode flags. | |
1049 | * | |
1050 | * RETURN: 0 if success | |
1051 | * error code if failure | |
1052 | */ | |
1053 | int | |
1054 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1055 | { | |
1056 | vnode_t *vp = ZTOV(zp); | |
1057 | dmu_tx_t *tx; | |
1058 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
1059 | zilog_t *zilog = zfsvfs->z_log; | |
1060 | rl_t *rl; | |
1061 | uint64_t end = off + len; | |
1062 | uint64_t size, new_blksz; | |
1063 | uint64_t pflags = zp->z_phys->zp_flags; | |
1064 | int error; | |
1065 | ||
1066 | if ((pflags & (ZFS_IMMUTABLE|ZFS_READONLY)) || | |
1067 | off < zp->z_phys->zp_size && (pflags & ZFS_APPENDONLY)) | |
1068 | return (EPERM); | |
1069 | ||
1070 | if (ZTOV(zp)->v_type == VFIFO) | |
1071 | return (0); | |
1072 | ||
1073 | /* | |
1074 | * If we will change zp_size then lock the whole file, | |
1075 | * otherwise just lock the range being freed. | |
1076 | */ | |
1077 | if (len == 0 || off + len > zp->z_phys->zp_size) { | |
1078 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); | |
1079 | } else { | |
1080 | rl = zfs_range_lock(zp, off, len, RL_WRITER); | |
1081 | /* recheck, in case zp_size changed */ | |
1082 | if (off + len > zp->z_phys->zp_size) { | |
1083 | /* lost race: file size changed, lock whole file */ | |
1084 | zfs_range_unlock(rl); | |
1085 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); | |
1086 | } | |
1087 | } | |
1088 | ||
1089 | /* | |
1090 | * Nothing to do if file already at desired length. | |
1091 | */ | |
1092 | size = zp->z_phys->zp_size; | |
1093 | if (len == 0 && size == off && off != 0) { | |
1094 | zfs_range_unlock(rl); | |
1095 | return (0); | |
1096 | } | |
1097 | ||
1098 | /* | |
1099 | * Check for any locks in the region to be freed. | |
1100 | */ | |
1101 | if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) { | |
1102 | uint64_t start = off; | |
1103 | uint64_t extent = len; | |
1104 | ||
1105 | if (off > size) { | |
1106 | start = size; | |
1107 | extent += off - size; | |
1108 | } else if (len == 0) { | |
1109 | extent = size - off; | |
1110 | } | |
1111 | if (error = chklock(vp, FWRITE, start, extent, flag, NULL)) { | |
1112 | zfs_range_unlock(rl); | |
1113 | return (error); | |
1114 | } | |
1115 | } | |
1116 | ||
1117 | tx = dmu_tx_create(zfsvfs->z_os); | |
1118 | dmu_tx_hold_bonus(tx, zp->z_id); | |
1119 | new_blksz = 0; | |
1120 | if (end > size && | |
1121 | (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { | |
1122 | /* | |
1123 | * We are growing the file past the current block size. | |
1124 | */ | |
1125 | if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { | |
1126 | ASSERT(!ISP2(zp->z_blksz)); | |
1127 | new_blksz = MIN(end, SPA_MAXBLOCKSIZE); | |
1128 | } else { | |
1129 | new_blksz = MIN(end, zp->z_zfsvfs->z_max_blksz); | |
1130 | } | |
1131 | dmu_tx_hold_write(tx, zp->z_id, 0, MIN(end, new_blksz)); | |
1132 | } else if (off < size) { | |
1133 | /* | |
1134 | * If len == 0, we are truncating the file. | |
1135 | */ | |
1136 | dmu_tx_hold_free(tx, zp->z_id, off, len ? len : DMU_OBJECT_END); | |
1137 | } | |
1138 | ||
1139 | error = dmu_tx_assign(tx, zfsvfs->z_assign); | |
1140 | if (error) { | |
1141 | if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) | |
1142 | dmu_tx_wait(tx); | |
1143 | dmu_tx_abort(tx); | |
1144 | zfs_range_unlock(rl); | |
1145 | return (error); | |
1146 | } | |
1147 | ||
1148 | if (new_blksz) | |
1149 | zfs_grow_blocksize(zp, new_blksz, tx); | |
1150 | ||
1151 | if (end > size || len == 0) | |
1152 | zp->z_phys->zp_size = end; | |
1153 | ||
1154 | if (off < size) { | |
1155 | objset_t *os = zfsvfs->z_os; | |
1156 | uint64_t rlen = len; | |
1157 | ||
1158 | if (len == 0) | |
1159 | rlen = -1; | |
1160 | else if (end > size) | |
1161 | rlen = size - off; | |
1162 | VERIFY(0 == dmu_free_range(os, zp->z_id, off, rlen, tx)); | |
1163 | } | |
1164 | ||
1165 | if (log) { | |
1166 | zfs_time_stamper(zp, CONTENT_MODIFIED, tx); | |
1167 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); | |
1168 | } | |
1169 | ||
1170 | zfs_range_unlock(rl); | |
1171 | ||
1172 | dmu_tx_commit(tx); | |
1173 | ||
1174 | /* | |
1175 | * Clear any mapped pages in the truncated region. This has to | |
1176 | * happen outside of the transaction to avoid the possibility of | |
1177 | * a deadlock with someone trying to push a page that we are | |
1178 | * about to invalidate. | |
1179 | */ | |
1180 | rw_enter(&zp->z_map_lock, RW_WRITER); | |
1181 | if (off < size && vn_has_cached_data(vp)) { | |
1182 | page_t *pp; | |
1183 | uint64_t start = off & PAGEMASK; | |
1184 | int poff = off & PAGEOFFSET; | |
1185 | ||
1186 | if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) { | |
1187 | /* | |
1188 | * We need to zero a partial page. | |
1189 | */ | |
1190 | pagezero(pp, poff, PAGESIZE - poff); | |
1191 | start += PAGESIZE; | |
1192 | page_unlock(pp); | |
1193 | } | |
1194 | error = pvn_vplist_dirty(vp, start, zfs_no_putpage, | |
1195 | B_INVAL | B_TRUNC, NULL); | |
1196 | ASSERT(error == 0); | |
1197 | } | |
1198 | rw_exit(&zp->z_map_lock); | |
1199 | ||
1200 | return (0); | |
1201 | } | |
1202 | ||
1203 | void | |
1204 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1205 | { | |
1206 | zfsvfs_t zfsvfs; | |
1207 | uint64_t moid, doid; | |
1208 | uint64_t version = 0; | |
1209 | uint64_t sense = ZFS_CASE_SENSITIVE; | |
1210 | uint64_t norm = 0; | |
1211 | nvpair_t *elem; | |
1212 | int error; | |
1213 | znode_t *rootzp = NULL; | |
1214 | vnode_t *vp; | |
1215 | vattr_t vattr; | |
1216 | znode_t *zp; | |
1217 | ||
1218 | /* | |
1219 | * First attempt to create master node. | |
1220 | */ | |
1221 | /* | |
1222 | * In an empty objset, there are no blocks to read and thus | |
1223 | * there can be no i/o errors (which we assert below). | |
1224 | */ | |
1225 | moid = MASTER_NODE_OBJ; | |
1226 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1227 | DMU_OT_NONE, 0, tx); | |
1228 | ASSERT(error == 0); | |
1229 | ||
1230 | /* | |
1231 | * Set starting attributes. | |
1232 | */ | |
1233 | elem = NULL; | |
1234 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1235 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1236 | uint64_t val; | |
1237 | char *name; | |
1238 | ||
1239 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1240 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1241 | name = nvpair_name(elem); | |
1242 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
1243 | version = val; | |
1244 | error = zap_update(os, moid, ZPL_VERSION_STR, | |
1245 | 8, 1, &version, tx); | |
1246 | } else { | |
1247 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1248 | } | |
1249 | ASSERT(error == 0); | |
1250 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1251 | norm = val; | |
1252 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) | |
1253 | sense = val; | |
1254 | } | |
1255 | ASSERT(version != 0); | |
1256 | ||
1257 | /* | |
1258 | * Create a delete queue. | |
1259 | */ | |
1260 | doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); | |
1261 | ||
1262 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx); | |
1263 | ASSERT(error == 0); | |
1264 | ||
1265 | /* | |
1266 | * Create root znode. Create minimal znode/vnode/zfsvfs | |
1267 | * to allow zfs_mknode to work. | |
1268 | */ | |
1269 | vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; | |
1270 | vattr.va_type = VDIR; | |
1271 | vattr.va_mode = S_IFDIR|0755; | |
1272 | vattr.va_uid = crgetuid(cr); | |
1273 | vattr.va_gid = crgetgid(cr); | |
1274 | ||
1275 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); | |
1276 | rootzp->z_zfsvfs = &zfsvfs; | |
1277 | rootzp->z_unlinked = 0; | |
1278 | rootzp->z_atime_dirty = 0; | |
1279 | ||
1280 | vp = ZTOV(rootzp); | |
1281 | vn_reinit(vp); | |
1282 | vp->v_type = VDIR; | |
1283 | ||
1284 | bzero(&zfsvfs, sizeof (zfsvfs_t)); | |
1285 | ||
1286 | zfsvfs.z_os = os; | |
1287 | zfsvfs.z_assign = TXG_NOWAIT; | |
1288 | zfsvfs.z_parent = &zfsvfs; | |
1289 | zfsvfs.z_version = version; | |
1290 | zfsvfs.z_use_fuids = USE_FUIDS(version, os); | |
1291 | zfsvfs.z_norm = norm; | |
1292 | /* | |
1293 | * Fold case on file systems that are always or sometimes case | |
1294 | * insensitive. | |
1295 | */ | |
1296 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) | |
1297 | zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; | |
1298 | ||
1299 | mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); | |
1300 | list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), | |
1301 | offsetof(znode_t, z_link_node)); | |
1302 | ||
1303 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, NULL, NULL); | |
1304 | ASSERT3P(zp, ==, rootzp); | |
1305 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1306 | ASSERT(error == 0); | |
1307 | ||
1308 | ZTOV(rootzp)->v_count = 0; | |
1309 | dmu_buf_rele(rootzp->z_dbuf, NULL); | |
1310 | rootzp->z_dbuf = NULL; | |
1311 | kmem_cache_free(znode_cache, rootzp); | |
1312 | } | |
1313 | ||
1314 | #endif /* _KERNEL */ | |
1315 | /* | |
1316 | * Given an object number, return its parent object number and whether | |
1317 | * or not the object is an extended attribute directory. | |
1318 | */ | |
1319 | static int | |
1320 | zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir) | |
1321 | { | |
1322 | dmu_buf_t *db; | |
1323 | dmu_object_info_t doi; | |
1324 | znode_phys_t *zp; | |
1325 | int error; | |
1326 | ||
1327 | if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0) | |
1328 | return (error); | |
1329 | ||
1330 | dmu_object_info_from_db(db, &doi); | |
1331 | if (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1332 | doi.doi_bonus_size < sizeof (znode_phys_t)) { | |
1333 | dmu_buf_rele(db, FTAG); | |
1334 | return (EINVAL); | |
1335 | } | |
1336 | ||
1337 | zp = db->db_data; | |
1338 | *pobjp = zp->zp_parent; | |
1339 | *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) && | |
1340 | S_ISDIR(zp->zp_mode); | |
1341 | dmu_buf_rele(db, FTAG); | |
1342 | ||
1343 | return (0); | |
1344 | } | |
1345 | ||
1346 | int | |
1347 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
1348 | { | |
1349 | char *path = buf + len - 1; | |
1350 | int error; | |
1351 | ||
1352 | *path = '\0'; | |
1353 | ||
1354 | for (;;) { | |
1355 | uint64_t pobj; | |
1356 | char component[MAXNAMELEN + 2]; | |
1357 | size_t complen; | |
1358 | int is_xattrdir; | |
1359 | ||
1360 | if ((error = zfs_obj_to_pobj(osp, obj, &pobj, | |
1361 | &is_xattrdir)) != 0) | |
1362 | break; | |
1363 | ||
1364 | if (pobj == obj) { | |
1365 | if (path[0] != '/') | |
1366 | *--path = '/'; | |
1367 | break; | |
1368 | } | |
1369 | ||
1370 | component[0] = '/'; | |
1371 | if (is_xattrdir) { | |
1372 | (void) sprintf(component + 1, "<xattrdir>"); | |
1373 | } else { | |
1374 | error = zap_value_search(osp, pobj, obj, | |
1375 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
1376 | if (error != 0) | |
1377 | break; | |
1378 | } | |
1379 | ||
1380 | complen = strlen(component); | |
1381 | path -= complen; | |
1382 | ASSERT(path >= buf); | |
1383 | bcopy(component, path, complen); | |
1384 | obj = pobj; | |
1385 | } | |
1386 | ||
1387 | if (error == 0) | |
1388 | (void) memmove(buf, path, buf + len - path); | |
1389 | return (error); | |
1390 | } |