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 https://opensource.org/licenses/CDDL-1.0.
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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
24 * Copyright (c) 2014 Integros [integros.com]
25 */
26
27 /* Portions Copyright 2007 Jeremy Teo */
28 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
29
30 #include <sys/types.h>
31 #include <sys/param.h>
32 #include <sys/time.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
36 #include <sys/resourcevar.h>
37 #include <sys/mntent.h>
38 #include <sys/u8_textprep.h>
39 #include <sys/dsl_dataset.h>
40 #include <sys/vfs.h>
41 #include <sys/vnode.h>
42 #include <sys/file.h>
43 #include <sys/kmem.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/atomic.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/zfs_ioctl.h>
50 #include <sys/zfs_rlock.h>
51 #include <sys/zfs_fuid.h>
52 #include <sys/dnode.h>
53 #include <sys/fs/zfs.h>
54 #include <sys/dmu.h>
55 #include <sys/dmu_objset.h>
56 #include <sys/dmu_tx.h>
57 #include <sys/zfs_refcount.h>
58 #include <sys/stat.h>
59 #include <sys/zap.h>
60 #include <sys/zfs_znode.h>
61 #include <sys/sa.h>
62 #include <sys/zfs_sa.h>
63 #include <sys/zfs_stat.h>
64
65 #include "zfs_prop.h"
66 #include "zfs_comutil.h"
67
68 /* Used by fstat(1). */
69 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
70 SYSCTL_NULL_INT_PTR, sizeof (znode_t), "sizeof(znode_t)");
71
72 /*
73 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
74 * turned on when DEBUG is also defined.
75 */
76 #ifdef ZFS_DEBUG
77 #define ZNODE_STATS
78 #endif /* DEBUG */
79
80 #ifdef ZNODE_STATS
81 #define ZNODE_STAT_ADD(stat) ((stat)++)
82 #else
83 #define ZNODE_STAT_ADD(stat) /* nothing */
84 #endif /* ZNODE_STATS */
85
86 #if !defined(KMEM_DEBUG)
87 #define _ZFS_USE_SMR
88 static uma_zone_t znode_uma_zone;
89 #else
90 static kmem_cache_t *znode_cache = NULL;
91 #endif
92
93 extern struct vop_vector zfs_vnodeops;
94 extern struct vop_vector zfs_fifoops;
95 extern struct vop_vector zfs_shareops;
96
97
98 /*
99 * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on
100 * z_rangelock. It will modify the offset and length of the lock to reflect
101 * znode-specific information, and convert RL_APPEND to RL_WRITER. This is
102 * called with the rangelock_t's rl_lock held, which avoids races.
103 */
104 static void
zfs_rangelock_cb(zfs_locked_range_t * new,void * arg)105 zfs_rangelock_cb(zfs_locked_range_t *new, void *arg)
106 {
107 znode_t *zp = arg;
108
109 /*
110 * If in append mode, convert to writer and lock starting at the
111 * current end of file.
112 */
113 if (new->lr_type == RL_APPEND) {
114 new->lr_offset = zp->z_size;
115 new->lr_type = RL_WRITER;
116 }
117
118 /*
119 * If we need to grow the block size then lock the whole file range.
120 */
121 uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length);
122 if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
123 zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) {
124 new->lr_offset = 0;
125 new->lr_length = UINT64_MAX;
126 }
127 }
128
129 static int
zfs_znode_cache_constructor(void * buf,void * arg,int kmflags)130 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
131 {
132 znode_t *zp = buf;
133
134 POINTER_INVALIDATE(&zp->z_zfsvfs);
135
136 list_link_init(&zp->z_link_node);
137
138 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
139 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
140 rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
141
142 zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp);
143
144 zp->z_acl_cached = NULL;
145 zp->z_xattr_cached = NULL;
146 zp->z_xattr_parent = 0;
147 zp->z_vnode = NULL;
148 zp->z_sync_writes_cnt = 0;
149 zp->z_async_writes_cnt = 0;
150
151 return (0);
152 }
153
154 static void
zfs_znode_cache_destructor(void * buf,void * arg)155 zfs_znode_cache_destructor(void *buf, void *arg)
156 {
157 (void) arg;
158 znode_t *zp = buf;
159
160 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
161 ASSERT3P(zp->z_vnode, ==, NULL);
162 ASSERT(!list_link_active(&zp->z_link_node));
163 mutex_destroy(&zp->z_lock);
164 mutex_destroy(&zp->z_acl_lock);
165 rw_destroy(&zp->z_xattr_lock);
166 zfs_rangelock_fini(&zp->z_rangelock);
167
168 ASSERT3P(zp->z_acl_cached, ==, NULL);
169 ASSERT3P(zp->z_xattr_cached, ==, NULL);
170
171 ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt));
172 ASSERT0(atomic_load_32(&zp->z_async_writes_cnt));
173 }
174
175
176 #ifdef _ZFS_USE_SMR
177 VFS_SMR_DECLARE;
178
179 static int
zfs_znode_cache_constructor_smr(void * mem,int size __unused,void * private,int flags)180 zfs_znode_cache_constructor_smr(void *mem, int size __unused, void *private,
181 int flags)
182 {
183 return (zfs_znode_cache_constructor(mem, private, flags));
184 }
185
186 static void
zfs_znode_cache_destructor_smr(void * mem,int size __unused,void * private)187 zfs_znode_cache_destructor_smr(void *mem, int size __unused, void *private)
188 {
189 zfs_znode_cache_destructor(mem, private);
190 }
191
192 void
zfs_znode_init(void)193 zfs_znode_init(void)
194 {
195 /*
196 * Initialize zcache
197 */
198 ASSERT3P(znode_uma_zone, ==, NULL);
199 znode_uma_zone = uma_zcreate("zfs_znode_cache",
200 sizeof (znode_t), zfs_znode_cache_constructor_smr,
201 zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0);
202 VFS_SMR_ZONE_SET(znode_uma_zone);
203 }
204
205 static znode_t *
zfs_znode_alloc_kmem(int flags)206 zfs_znode_alloc_kmem(int flags)
207 {
208 return (uma_zalloc_smr(znode_uma_zone, flags));
209 }
210
211 static void
zfs_znode_free_kmem(znode_t * zp)212 zfs_znode_free_kmem(znode_t *zp)
213 {
214 if (zp->z_xattr_cached) {
215 nvlist_free(zp->z_xattr_cached);
216 zp->z_xattr_cached = NULL;
217 }
218 uma_zfree_smr(znode_uma_zone, zp);
219 }
220 #else
221 void
zfs_znode_init(void)222 zfs_znode_init(void)
223 {
224 /*
225 * Initialize zcache
226 */
227 ASSERT3P(znode_cache, ==, NULL);
228 znode_cache = kmem_cache_create("zfs_znode_cache",
229 sizeof (znode_t), 0, zfs_znode_cache_constructor,
230 zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_RECLAIMABLE);
231 }
232
233 static znode_t *
zfs_znode_alloc_kmem(int flags)234 zfs_znode_alloc_kmem(int flags)
235 {
236 return (kmem_cache_alloc(znode_cache, flags));
237 }
238
239 static void
zfs_znode_free_kmem(znode_t * zp)240 zfs_znode_free_kmem(znode_t *zp)
241 {
242 if (zp->z_xattr_cached) {
243 nvlist_free(zp->z_xattr_cached);
244 zp->z_xattr_cached = NULL;
245 }
246 kmem_cache_free(znode_cache, zp);
247 }
248 #endif
249
250 void
zfs_znode_fini(void)251 zfs_znode_fini(void)
252 {
253 /*
254 * Cleanup zcache
255 */
256 #ifdef _ZFS_USE_SMR
257 if (znode_uma_zone) {
258 uma_zdestroy(znode_uma_zone);
259 znode_uma_zone = NULL;
260 }
261 #else
262 if (znode_cache) {
263 kmem_cache_destroy(znode_cache);
264 znode_cache = NULL;
265 }
266 #endif
267 }
268
269
270 static int
zfs_create_share_dir(zfsvfs_t * zfsvfs,dmu_tx_t * tx)271 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
272 {
273 zfs_acl_ids_t acl_ids;
274 vattr_t vattr;
275 znode_t *sharezp;
276 znode_t *zp;
277 int error;
278
279 vattr.va_mask = AT_MODE|AT_UID|AT_GID;
280 vattr.va_type = VDIR;
281 vattr.va_mode = S_IFDIR|0555;
282 vattr.va_uid = crgetuid(kcred);
283 vattr.va_gid = crgetgid(kcred);
284
285 sharezp = zfs_znode_alloc_kmem(KM_SLEEP);
286 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
287 sharezp->z_unlinked = 0;
288 sharezp->z_atime_dirty = 0;
289 sharezp->z_zfsvfs = zfsvfs;
290 sharezp->z_is_sa = zfsvfs->z_use_sa;
291
292 VERIFY0(zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
293 kcred, NULL, &acl_ids, NULL));
294 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
295 ASSERT3P(zp, ==, sharezp);
296 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
297 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
298 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
299 zfsvfs->z_shares_dir = sharezp->z_id;
300
301 zfs_acl_ids_free(&acl_ids);
302 sa_handle_destroy(sharezp->z_sa_hdl);
303 zfs_znode_free_kmem(sharezp);
304
305 return (error);
306 }
307
308 /*
309 * define a couple of values we need available
310 * for both 64 and 32 bit environments.
311 */
312 #ifndef NBITSMINOR64
313 #define NBITSMINOR64 32
314 #endif
315 #ifndef MAXMAJ64
316 #define MAXMAJ64 0xffffffffUL
317 #endif
318 #ifndef MAXMIN64
319 #define MAXMIN64 0xffffffffUL
320 #endif
321
322 /*
323 * Create special expldev for ZFS private use.
324 * Can't use standard expldev since it doesn't do
325 * what we want. The standard expldev() takes a
326 * dev32_t in LP64 and expands it to a long dev_t.
327 * We need an interface that takes a dev32_t in ILP32
328 * and expands it to a long dev_t.
329 */
330 static uint64_t
zfs_expldev(dev_t dev)331 zfs_expldev(dev_t dev)
332 {
333 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
334 }
335 /*
336 * Special cmpldev for ZFS private use.
337 * Can't use standard cmpldev since it takes
338 * a long dev_t and compresses it to dev32_t in
339 * LP64. We need to do a compaction of a long dev_t
340 * to a dev32_t in ILP32.
341 */
342 dev_t
zfs_cmpldev(uint64_t dev)343 zfs_cmpldev(uint64_t dev)
344 {
345 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
346 }
347
348 static void
zfs_znode_sa_init(zfsvfs_t * zfsvfs,znode_t * zp,dmu_buf_t * db,dmu_object_type_t obj_type,sa_handle_t * sa_hdl)349 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
350 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
351 {
352 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
353 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
354
355 ASSERT3P(zp->z_sa_hdl, ==, NULL);
356 ASSERT3P(zp->z_acl_cached, ==, NULL);
357 if (sa_hdl == NULL) {
358 VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp,
359 SA_HDL_SHARED, &zp->z_sa_hdl));
360 } else {
361 zp->z_sa_hdl = sa_hdl;
362 sa_set_userp(sa_hdl, zp);
363 }
364
365 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
366
367 /*
368 * Slap on VROOT if we are the root znode unless we are the root
369 * node of a snapshot mounted under .zfs.
370 */
371 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
372 ZTOV(zp)->v_flag |= VROOT;
373 }
374
375 void
zfs_znode_dmu_fini(znode_t * zp)376 zfs_znode_dmu_fini(znode_t *zp)
377 {
378 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
379 ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zp->z_zfsvfs));
380
381 sa_handle_destroy(zp->z_sa_hdl);
382 zp->z_sa_hdl = NULL;
383 }
384
385 static void
zfs_vnode_forget(vnode_t * vp)386 zfs_vnode_forget(vnode_t *vp)
387 {
388
389 /* copied from insmntque_stddtr */
390 vp->v_data = NULL;
391 vp->v_op = &dead_vnodeops;
392 vgone(vp);
393 vput(vp);
394 }
395
396 /*
397 * Construct a new znode/vnode and initialize.
398 *
399 * This does not do a call to dmu_set_user() that is
400 * up to the caller to do, in case you don't want to
401 * return the znode
402 */
403 static znode_t *
zfs_znode_alloc(zfsvfs_t * zfsvfs,dmu_buf_t * db,int blksz,dmu_object_type_t obj_type,sa_handle_t * hdl)404 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
405 dmu_object_type_t obj_type, sa_handle_t *hdl)
406 {
407 znode_t *zp;
408 vnode_t *vp;
409 uint64_t mode;
410 uint64_t parent;
411 #ifdef notyet
412 uint64_t mtime[2], ctime[2];
413 #endif
414 uint64_t projid = ZFS_DEFAULT_PROJID;
415 sa_bulk_attr_t bulk[9];
416 int count = 0;
417 int error;
418
419 zp = zfs_znode_alloc_kmem(KM_SLEEP);
420
421 #ifndef _ZFS_USE_SMR
422 KASSERT((zfsvfs->z_parent->z_vfs->mnt_kern_flag & MNTK_FPLOOKUP) == 0,
423 ("%s: fast path lookup enabled without smr", __func__));
424 #endif
425
426 KASSERT(curthread->td_vp_reserved != NULL,
427 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
428 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
429 if (error != 0) {
430 zfs_znode_free_kmem(zp);
431 return (NULL);
432 }
433 zp->z_vnode = vp;
434 vp->v_data = zp;
435
436 /*
437 * Acquire the vnode lock before any possible interaction with the
438 * outside world. Specifically, there is an error path that calls
439 * zfs_vnode_forget() and the vnode should be exclusively locked.
440 */
441 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
442
443 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
444
445 zp->z_sa_hdl = NULL;
446 zp->z_unlinked = 0;
447 zp->z_atime_dirty = 0;
448 zp->z_mapcnt = 0;
449 zp->z_id = db->db_object;
450 zp->z_blksz = blksz;
451 zp->z_seq = 0x7A4653;
452 zp->z_sync_cnt = 0;
453 zp->z_sync_writes_cnt = 0;
454 zp->z_async_writes_cnt = 0;
455 atomic_store_ptr(&zp->z_cached_symlink, NULL);
456
457 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
458
459 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
460 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
461 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
462 &zp->z_size, 8);
463 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
464 &zp->z_links, 8);
465 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
466 &zp->z_pflags, 8);
467 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
468 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
469 &zp->z_atime, 16);
470 #ifdef notyet
471 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
472 &mtime, 16);
473 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
474 &ctime, 16);
475 #endif
476 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
477 &zp->z_uid, 8);
478 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
479 &zp->z_gid, 8);
480
481 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0 ||
482 (dmu_objset_projectquota_enabled(zfsvfs->z_os) &&
483 (zp->z_pflags & ZFS_PROJID) &&
484 sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) {
485 if (hdl == NULL)
486 sa_handle_destroy(zp->z_sa_hdl);
487 zfs_vnode_forget(vp);
488 zp->z_vnode = NULL;
489 zfs_znode_free_kmem(zp);
490 return (NULL);
491 }
492
493 zp->z_projid = projid;
494 zp->z_mode = mode;
495
496 /* Cache the xattr parent id */
497 if (zp->z_pflags & ZFS_XATTR)
498 zp->z_xattr_parent = parent;
499
500 vp->v_type = IFTOVT((mode_t)mode);
501
502 switch (vp->v_type) {
503 case VDIR:
504 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
505 break;
506 case VFIFO:
507 vp->v_op = &zfs_fifoops;
508 break;
509 case VREG:
510 if (parent == zfsvfs->z_shares_dir) {
511 ASSERT0(zp->z_uid);
512 ASSERT0(zp->z_gid);
513 vp->v_op = &zfs_shareops;
514 }
515 break;
516 default:
517 break;
518 }
519
520 mutex_enter(&zfsvfs->z_znodes_lock);
521 list_insert_tail(&zfsvfs->z_all_znodes, zp);
522 zp->z_zfsvfs = zfsvfs;
523 mutex_exit(&zfsvfs->z_znodes_lock);
524
525 #if __FreeBSD_version >= 1400077
526 vn_set_state(vp, VSTATE_CONSTRUCTED);
527 #endif
528 VN_LOCK_AREC(vp);
529 if (vp->v_type != VFIFO)
530 VN_LOCK_ASHARE(vp);
531
532 return (zp);
533 }
534
535 static uint64_t empty_xattr;
536 static uint64_t pad[4];
537 static zfs_acl_phys_t acl_phys;
538 /*
539 * Create a new DMU object to hold a zfs znode.
540 *
541 * IN: dzp - parent directory for new znode
542 * vap - file attributes for new znode
543 * tx - dmu transaction id for zap operations
544 * cr - credentials of caller
545 * flag - flags:
546 * IS_ROOT_NODE - new object will be root
547 * IS_XATTR - new object is an attribute
548 * bonuslen - length of bonus buffer
549 * setaclp - File/Dir initial ACL
550 * fuidp - Tracks fuid allocation.
551 *
552 * OUT: zpp - allocated znode
553 *
554 */
555 void
zfs_mknode(znode_t * dzp,vattr_t * vap,dmu_tx_t * tx,cred_t * cr,uint_t flag,znode_t ** zpp,zfs_acl_ids_t * acl_ids)556 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
557 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
558 {
559 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
560 uint64_t mode, size, links, parent, pflags;
561 uint64_t dzp_pflags = 0;
562 uint64_t rdev = 0;
563 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
564 dmu_buf_t *db;
565 timestruc_t now;
566 uint64_t gen, obj;
567 int bonuslen;
568 int dnodesize;
569 sa_handle_t *sa_hdl;
570 dmu_object_type_t obj_type;
571 sa_bulk_attr_t *sa_attrs;
572 int cnt = 0;
573 zfs_acl_locator_cb_t locate = { 0 };
574
575 ASSERT3P(vap, !=, NULL);
576 ASSERT3U((vap->va_mask & AT_MODE), ==, AT_MODE);
577
578 if (zfsvfs->z_replay) {
579 obj = vap->va_nodeid;
580 now = vap->va_ctime; /* see zfs_replay_create() */
581 gen = vap->va_nblocks; /* ditto */
582 dnodesize = vap->va_fsid; /* ditto */
583 } else {
584 obj = 0;
585 vfs_timestamp(&now);
586 gen = dmu_tx_get_txg(tx);
587 dnodesize = dmu_objset_dnodesize(zfsvfs->z_os);
588 }
589
590 if (dnodesize == 0)
591 dnodesize = DNODE_MIN_SIZE;
592
593 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
594 bonuslen = (obj_type == DMU_OT_SA) ?
595 DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE;
596
597 /*
598 * Create a new DMU object.
599 */
600 /*
601 * There's currently no mechanism for pre-reading the blocks that will
602 * be needed to allocate a new object, so we accept the small chance
603 * that there will be an i/o error and we will fail one of the
604 * assertions below.
605 */
606 if (vap->va_type == VDIR) {
607 if (zfsvfs->z_replay) {
608 VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj,
609 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
610 obj_type, bonuslen, dnodesize, tx));
611 } else {
612 obj = zap_create_norm_dnsize(zfsvfs->z_os,
613 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
614 obj_type, bonuslen, dnodesize, tx);
615 }
616 } else {
617 if (zfsvfs->z_replay) {
618 VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj,
619 DMU_OT_PLAIN_FILE_CONTENTS, 0,
620 obj_type, bonuslen, dnodesize, tx));
621 } else {
622 obj = dmu_object_alloc_dnsize(zfsvfs->z_os,
623 DMU_OT_PLAIN_FILE_CONTENTS, 0,
624 obj_type, bonuslen, dnodesize, tx);
625 }
626 }
627
628 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
629 VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
630
631 /*
632 * If this is the root, fix up the half-initialized parent pointer
633 * to reference the just-allocated physical data area.
634 */
635 if (flag & IS_ROOT_NODE) {
636 dzp->z_id = obj;
637 } else {
638 dzp_pflags = dzp->z_pflags;
639 }
640
641 /*
642 * If parent is an xattr, so am I.
643 */
644 if (dzp_pflags & ZFS_XATTR) {
645 flag |= IS_XATTR;
646 }
647
648 if (zfsvfs->z_use_fuids)
649 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
650 else
651 pflags = 0;
652
653 if (vap->va_type == VDIR) {
654 size = 2; /* contents ("." and "..") */
655 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
656 } else {
657 size = links = 0;
658 }
659
660 if (vap->va_type == VBLK || vap->va_type == VCHR) {
661 rdev = zfs_expldev(vap->va_rdev);
662 }
663
664 parent = dzp->z_id;
665 mode = acl_ids->z_mode;
666 if (flag & IS_XATTR)
667 pflags |= ZFS_XATTR;
668
669 /*
670 * No execs denied will be determined when zfs_mode_compute() is called.
671 */
672 pflags |= acl_ids->z_aclp->z_hints &
673 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
674 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
675
676 ZFS_TIME_ENCODE(&now, crtime);
677 ZFS_TIME_ENCODE(&now, ctime);
678
679 if (vap->va_mask & AT_ATIME) {
680 ZFS_TIME_ENCODE(&vap->va_atime, atime);
681 } else {
682 ZFS_TIME_ENCODE(&now, atime);
683 }
684
685 if (vap->va_mask & AT_MTIME) {
686 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
687 } else {
688 ZFS_TIME_ENCODE(&now, mtime);
689 }
690
691 /* Now add in all of the "SA" attributes */
692 VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
693 &sa_hdl));
694
695 /*
696 * Setup the array of attributes to be replaced/set on the new file
697 *
698 * order for DMU_OT_ZNODE is critical since it needs to be constructed
699 * in the old znode_phys_t format. Don't change this ordering
700 */
701 sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
702
703 if (obj_type == DMU_OT_ZNODE) {
704 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
705 NULL, &atime, 16);
706 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
707 NULL, &mtime, 16);
708 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
709 NULL, &ctime, 16);
710 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
711 NULL, &crtime, 16);
712 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
713 NULL, &gen, 8);
714 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
715 NULL, &mode, 8);
716 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
717 NULL, &size, 8);
718 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
719 NULL, &parent, 8);
720 } else {
721 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
722 NULL, &mode, 8);
723 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
724 NULL, &size, 8);
725 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
726 NULL, &gen, 8);
727 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs),
728 NULL, &acl_ids->z_fuid, 8);
729 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs),
730 NULL, &acl_ids->z_fgid, 8);
731 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
732 NULL, &parent, 8);
733 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
734 NULL, &pflags, 8);
735 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
736 NULL, &atime, 16);
737 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
738 NULL, &mtime, 16);
739 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
740 NULL, &ctime, 16);
741 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
742 NULL, &crtime, 16);
743 }
744
745 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
746
747 if (obj_type == DMU_OT_ZNODE) {
748 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
749 &empty_xattr, 8);
750 }
751 if (obj_type == DMU_OT_ZNODE ||
752 (vap->va_type == VBLK || vap->va_type == VCHR)) {
753 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
754 NULL, &rdev, 8);
755
756 }
757 if (obj_type == DMU_OT_ZNODE) {
758 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
759 NULL, &pflags, 8);
760 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
761 &acl_ids->z_fuid, 8);
762 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
763 &acl_ids->z_fgid, 8);
764 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
765 sizeof (uint64_t) * 4);
766 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
767 &acl_phys, sizeof (zfs_acl_phys_t));
768 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
769 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
770 &acl_ids->z_aclp->z_acl_count, 8);
771 locate.cb_aclp = acl_ids->z_aclp;
772 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
773 zfs_acl_data_locator, &locate,
774 acl_ids->z_aclp->z_acl_bytes);
775 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
776 acl_ids->z_fuid, acl_ids->z_fgid);
777 }
778
779 VERIFY0(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx));
780
781 if (!(flag & IS_ROOT_NODE)) {
782 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
783 ASSERT3P(*zpp, !=, NULL);
784 } else {
785 /*
786 * If we are creating the root node, the "parent" we
787 * passed in is the znode for the root.
788 */
789 *zpp = dzp;
790
791 (*zpp)->z_sa_hdl = sa_hdl;
792 }
793
794 (*zpp)->z_pflags = pflags;
795 (*zpp)->z_mode = mode;
796 (*zpp)->z_dnodesize = dnodesize;
797
798 if (vap->va_mask & AT_XVATTR)
799 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
800
801 if (obj_type == DMU_OT_ZNODE ||
802 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
803 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
804 }
805 if (!(flag & IS_ROOT_NODE)) {
806 vnode_t *vp = ZTOV(*zpp);
807 vp->v_vflag |= VV_FORCEINSMQ;
808 int err = insmntque(vp, zfsvfs->z_vfs);
809 vp->v_vflag &= ~VV_FORCEINSMQ;
810 (void) err;
811 KASSERT(err == 0, ("insmntque() failed: error %d", err));
812 }
813 kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
814 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
815 }
816
817 /*
818 * Update in-core attributes. It is assumed the caller will be doing an
819 * sa_bulk_update to push the changes out.
820 */
821 void
zfs_xvattr_set(znode_t * zp,xvattr_t * xvap,dmu_tx_t * tx)822 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
823 {
824 xoptattr_t *xoap;
825
826 xoap = xva_getxoptattr(xvap);
827 ASSERT3P(xoap, !=, NULL);
828
829 if (zp->z_zfsvfs->z_replay == B_FALSE) {
830 ASSERT_VOP_IN_SEQC(ZTOV(zp));
831 }
832
833 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
834 uint64_t times[2];
835 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
836 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
837 ×, sizeof (times), tx);
838 XVA_SET_RTN(xvap, XAT_CREATETIME);
839 }
840 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
841 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
842 zp->z_pflags, tx);
843 XVA_SET_RTN(xvap, XAT_READONLY);
844 }
845 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
846 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
847 zp->z_pflags, tx);
848 XVA_SET_RTN(xvap, XAT_HIDDEN);
849 }
850 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
851 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
852 zp->z_pflags, tx);
853 XVA_SET_RTN(xvap, XAT_SYSTEM);
854 }
855 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
856 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
857 zp->z_pflags, tx);
858 XVA_SET_RTN(xvap, XAT_ARCHIVE);
859 }
860 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
861 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
862 zp->z_pflags, tx);
863 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
864 }
865 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
866 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
867 zp->z_pflags, tx);
868 XVA_SET_RTN(xvap, XAT_NOUNLINK);
869 }
870 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
871 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
872 zp->z_pflags, tx);
873 XVA_SET_RTN(xvap, XAT_APPENDONLY);
874 }
875 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
876 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
877 zp->z_pflags, tx);
878 XVA_SET_RTN(xvap, XAT_NODUMP);
879 }
880 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
881 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
882 zp->z_pflags, tx);
883 XVA_SET_RTN(xvap, XAT_OPAQUE);
884 }
885 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
886 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
887 xoap->xoa_av_quarantined, zp->z_pflags, tx);
888 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
889 }
890 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
891 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
892 zp->z_pflags, tx);
893 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
894 }
895 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
896 zfs_sa_set_scanstamp(zp, xvap, tx);
897 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
898 }
899 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
900 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
901 zp->z_pflags, tx);
902 XVA_SET_RTN(xvap, XAT_REPARSE);
903 }
904 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
905 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
906 zp->z_pflags, tx);
907 XVA_SET_RTN(xvap, XAT_OFFLINE);
908 }
909 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
910 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
911 zp->z_pflags, tx);
912 XVA_SET_RTN(xvap, XAT_SPARSE);
913 }
914 }
915
916 int
zfs_zget(zfsvfs_t * zfsvfs,uint64_t obj_num,znode_t ** zpp)917 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
918 {
919 dmu_object_info_t doi;
920 dmu_buf_t *db;
921 znode_t *zp;
922 vnode_t *vp;
923 sa_handle_t *hdl;
924 int locked;
925 int err;
926
927 getnewvnode_reserve();
928 again:
929 *zpp = NULL;
930 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
931
932 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
933 if (err) {
934 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
935 getnewvnode_drop_reserve();
936 return (err);
937 }
938
939 dmu_object_info_from_db(db, &doi);
940 if (doi.doi_bonus_type != DMU_OT_SA &&
941 (doi.doi_bonus_type != DMU_OT_ZNODE ||
942 (doi.doi_bonus_type == DMU_OT_ZNODE &&
943 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
944 sa_buf_rele(db, NULL);
945 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
946 getnewvnode_drop_reserve();
947 return (SET_ERROR(EINVAL));
948 }
949
950 hdl = dmu_buf_get_user(db);
951 if (hdl != NULL) {
952 zp = sa_get_userdata(hdl);
953
954 /*
955 * Since "SA" does immediate eviction we
956 * should never find a sa handle that doesn't
957 * know about the znode.
958 */
959 ASSERT3P(zp, !=, NULL);
960 ASSERT3U(zp->z_id, ==, obj_num);
961 if (zp->z_unlinked) {
962 err = SET_ERROR(ENOENT);
963 } else {
964 vp = ZTOV(zp);
965 /*
966 * Don't let the vnode disappear after
967 * ZFS_OBJ_HOLD_EXIT.
968 */
969 VN_HOLD(vp);
970 *zpp = zp;
971 err = 0;
972 }
973
974 sa_buf_rele(db, NULL);
975 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
976
977 if (err) {
978 getnewvnode_drop_reserve();
979 return (err);
980 }
981
982 locked = VOP_ISLOCKED(vp);
983 VI_LOCK(vp);
984 if (VN_IS_DOOMED(vp) && locked != LK_EXCLUSIVE) {
985 /*
986 * The vnode is doomed and this thread doesn't
987 * hold the exclusive lock on it, so the vnode
988 * must be being reclaimed by another thread.
989 * Otherwise the doomed vnode is being reclaimed
990 * by this thread and zfs_zget is called from
991 * ZIL internals.
992 */
993 VI_UNLOCK(vp);
994
995 /*
996 * XXX vrele() locks the vnode when the last reference
997 * is dropped. Although in this case the vnode is
998 * doomed / dead and so no inactivation is required,
999 * the vnode lock is still acquired. That could result
1000 * in a LOR with z_teardown_lock if another thread holds
1001 * the vnode's lock and tries to take z_teardown_lock.
1002 * But that is only possible if the other thread peforms
1003 * a ZFS vnode operation on the vnode. That either
1004 * should not happen if the vnode is dead or the thread
1005 * should also have a reference to the vnode and thus
1006 * our reference is not last.
1007 */
1008 VN_RELE(vp);
1009 goto again;
1010 }
1011 VI_UNLOCK(vp);
1012 getnewvnode_drop_reserve();
1013 return (err);
1014 }
1015
1016 /*
1017 * Not found create new znode/vnode
1018 * but only if file exists.
1019 *
1020 * There is a small window where zfs_vget() could
1021 * find this object while a file create is still in
1022 * progress. This is checked for in zfs_znode_alloc()
1023 *
1024 * if zfs_znode_alloc() fails it will drop the hold on the
1025 * bonus buffer.
1026 */
1027 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1028 doi.doi_bonus_type, NULL);
1029 if (zp == NULL) {
1030 err = SET_ERROR(ENOENT);
1031 } else {
1032 *zpp = zp;
1033 }
1034 if (err == 0) {
1035 vnode_t *vp = ZTOV(zp);
1036
1037 err = insmntque(vp, zfsvfs->z_vfs);
1038 if (err == 0) {
1039 vp->v_hash = obj_num;
1040 VOP_UNLOCK(vp);
1041 } else {
1042 zp->z_vnode = NULL;
1043 zfs_znode_dmu_fini(zp);
1044 zfs_znode_free(zp);
1045 *zpp = NULL;
1046 }
1047 }
1048 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1049 getnewvnode_drop_reserve();
1050 return (err);
1051 }
1052
1053 int
zfs_rezget(znode_t * zp)1054 zfs_rezget(znode_t *zp)
1055 {
1056 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1057 dmu_object_info_t doi;
1058 dmu_buf_t *db;
1059 vnode_t *vp;
1060 uint64_t obj_num = zp->z_id;
1061 uint64_t mode, size;
1062 sa_bulk_attr_t bulk[8];
1063 int err;
1064 int count = 0;
1065 uint64_t gen;
1066
1067 /*
1068 * Remove cached pages before reloading the znode, so that they are not
1069 * lingering after we run into any error. Ideally, we should vgone()
1070 * the vnode in case of error, but currently we cannot do that
1071 * because of the LOR between the vnode lock and z_teardown_lock.
1072 * So, instead, we have to "doom" the znode in the illumos style.
1073 *
1074 * Ignore invalid pages during the scan. This is to avoid deadlocks
1075 * between page busying and the teardown lock, as pages are busied prior
1076 * to a VOP_GETPAGES operation, which acquires the teardown read lock.
1077 * Such pages will be invalid and can safely be skipped here.
1078 */
1079 vp = ZTOV(zp);
1080 #if __FreeBSD_version >= 1400042
1081 vn_pages_remove_valid(vp, 0, 0);
1082 #else
1083 vn_pages_remove(vp, 0, 0);
1084 #endif
1085
1086 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1087
1088 mutex_enter(&zp->z_acl_lock);
1089 if (zp->z_acl_cached) {
1090 zfs_acl_free(zp->z_acl_cached);
1091 zp->z_acl_cached = NULL;
1092 }
1093 mutex_exit(&zp->z_acl_lock);
1094
1095 rw_enter(&zp->z_xattr_lock, RW_WRITER);
1096 if (zp->z_xattr_cached) {
1097 nvlist_free(zp->z_xattr_cached);
1098 zp->z_xattr_cached = NULL;
1099 }
1100 rw_exit(&zp->z_xattr_lock);
1101
1102 ASSERT3P(zp->z_sa_hdl, ==, NULL);
1103 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1104 if (err) {
1105 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1106 return (err);
1107 }
1108
1109 dmu_object_info_from_db(db, &doi);
1110 if (doi.doi_bonus_type != DMU_OT_SA &&
1111 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1112 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1113 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1114 sa_buf_rele(db, NULL);
1115 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1116 return (SET_ERROR(EINVAL));
1117 }
1118
1119 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1120 size = zp->z_size;
1121
1122 /* reload cached values */
1123 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1124 &gen, sizeof (gen));
1125 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1126 &zp->z_size, sizeof (zp->z_size));
1127 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1128 &zp->z_links, sizeof (zp->z_links));
1129 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1130 &zp->z_pflags, sizeof (zp->z_pflags));
1131 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1132 &zp->z_atime, sizeof (zp->z_atime));
1133 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1134 &zp->z_uid, sizeof (zp->z_uid));
1135 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1136 &zp->z_gid, sizeof (zp->z_gid));
1137 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1138 &mode, sizeof (mode));
1139
1140 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1141 zfs_znode_dmu_fini(zp);
1142 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1143 return (SET_ERROR(EIO));
1144 }
1145
1146 zp->z_mode = mode;
1147
1148 if (gen != zp->z_gen) {
1149 zfs_znode_dmu_fini(zp);
1150 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1151 return (SET_ERROR(EIO));
1152 }
1153
1154 /*
1155 * It is highly improbable but still quite possible that two
1156 * objects in different datasets are created with the same
1157 * object numbers and in transaction groups with the same
1158 * numbers. znodes corresponding to those objects would
1159 * have the same z_id and z_gen, but their other attributes
1160 * may be different.
1161 * zfs recv -F may replace one of such objects with the other.
1162 * As a result file properties recorded in the replaced
1163 * object's vnode may no longer match the received object's
1164 * properties. At present the only cached property is the
1165 * files type recorded in v_type.
1166 * So, handle this case by leaving the old vnode and znode
1167 * disassociated from the actual object. A new vnode and a
1168 * znode will be created if the object is accessed
1169 * (e.g. via a look-up). The old vnode and znode will be
1170 * recycled when the last vnode reference is dropped.
1171 */
1172 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1173 zfs_znode_dmu_fini(zp);
1174 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1175 return (SET_ERROR(EIO));
1176 }
1177
1178 /*
1179 * If the file has zero links, then it has been unlinked on the send
1180 * side and it must be in the received unlinked set.
1181 * We call zfs_znode_dmu_fini() now to prevent any accesses to the
1182 * stale data and to prevent automatically removal of the file in
1183 * zfs_zinactive(). The file will be removed either when it is removed
1184 * on the send side and the next incremental stream is received or
1185 * when the unlinked set gets processed.
1186 */
1187 zp->z_unlinked = (zp->z_links == 0);
1188 if (zp->z_unlinked) {
1189 zfs_znode_dmu_fini(zp);
1190 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1191 return (0);
1192 }
1193
1194 zp->z_blksz = doi.doi_data_block_size;
1195 if (zp->z_size != size)
1196 vnode_pager_setsize(vp, zp->z_size);
1197
1198 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1199
1200 return (0);
1201 }
1202
1203 void
zfs_znode_delete(znode_t * zp,dmu_tx_t * tx)1204 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1205 {
1206 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1207 objset_t *os = zfsvfs->z_os;
1208 uint64_t obj = zp->z_id;
1209 uint64_t acl_obj = zfs_external_acl(zp);
1210
1211 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1212 if (acl_obj) {
1213 VERIFY(!zp->z_is_sa);
1214 VERIFY0(dmu_object_free(os, acl_obj, tx));
1215 }
1216 VERIFY0(dmu_object_free(os, obj, tx));
1217 zfs_znode_dmu_fini(zp);
1218 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1219 }
1220
1221 void
zfs_zinactive(znode_t * zp)1222 zfs_zinactive(znode_t *zp)
1223 {
1224 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1225 uint64_t z_id = zp->z_id;
1226
1227 ASSERT3P(zp->z_sa_hdl, !=, NULL);
1228
1229 /*
1230 * Don't allow a zfs_zget() while were trying to release this znode
1231 */
1232 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1233
1234 /*
1235 * If this was the last reference to a file with no links, remove
1236 * the file from the file system unless the file system is mounted
1237 * read-only. That can happen, for example, if the file system was
1238 * originally read-write, the file was opened, then unlinked and
1239 * the file system was made read-only before the file was finally
1240 * closed. The file will remain in the unlinked set.
1241 */
1242 if (zp->z_unlinked) {
1243 ASSERT(!zfsvfs->z_issnap);
1244 if ((zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) == 0) {
1245 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1246 zfs_rmnode(zp);
1247 return;
1248 }
1249 }
1250
1251 zfs_znode_dmu_fini(zp);
1252 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1253 zfs_znode_free(zp);
1254 }
1255
1256 void
zfs_znode_free(znode_t * zp)1257 zfs_znode_free(znode_t *zp)
1258 {
1259 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1260 char *symlink;
1261
1262 ASSERT3P(zp->z_sa_hdl, ==, NULL);
1263 zp->z_vnode = NULL;
1264 mutex_enter(&zfsvfs->z_znodes_lock);
1265 POINTER_INVALIDATE(&zp->z_zfsvfs);
1266 list_remove(&zfsvfs->z_all_znodes, zp);
1267 mutex_exit(&zfsvfs->z_znodes_lock);
1268
1269 symlink = atomic_load_ptr(&zp->z_cached_symlink);
1270 if (symlink != NULL) {
1271 atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink,
1272 (uintptr_t)NULL);
1273 cache_symlink_free(symlink, strlen(symlink) + 1);
1274 }
1275
1276 if (zp->z_acl_cached) {
1277 zfs_acl_free(zp->z_acl_cached);
1278 zp->z_acl_cached = NULL;
1279 }
1280
1281 zfs_znode_free_kmem(zp);
1282 }
1283
1284 void
zfs_tstamp_update_setup_ext(znode_t * zp,uint_t flag,uint64_t mtime[2],uint64_t ctime[2],boolean_t have_tx)1285 zfs_tstamp_update_setup_ext(znode_t *zp, uint_t flag, uint64_t mtime[2],
1286 uint64_t ctime[2], boolean_t have_tx)
1287 {
1288 timestruc_t now;
1289
1290 vfs_timestamp(&now);
1291
1292 if (have_tx) { /* will sa_bulk_update happen really soon? */
1293 zp->z_atime_dirty = 0;
1294 zp->z_seq++;
1295 } else {
1296 zp->z_atime_dirty = 1;
1297 }
1298
1299 if (flag & AT_ATIME) {
1300 ZFS_TIME_ENCODE(&now, zp->z_atime);
1301 }
1302
1303 if (flag & AT_MTIME) {
1304 ZFS_TIME_ENCODE(&now, mtime);
1305 if (zp->z_zfsvfs->z_use_fuids) {
1306 zp->z_pflags |= (ZFS_ARCHIVE |
1307 ZFS_AV_MODIFIED);
1308 }
1309 }
1310
1311 if (flag & AT_CTIME) {
1312 ZFS_TIME_ENCODE(&now, ctime);
1313 if (zp->z_zfsvfs->z_use_fuids)
1314 zp->z_pflags |= ZFS_ARCHIVE;
1315 }
1316 }
1317
1318
1319 void
zfs_tstamp_update_setup(znode_t * zp,uint_t flag,uint64_t mtime[2],uint64_t ctime[2])1320 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1321 uint64_t ctime[2])
1322 {
1323 zfs_tstamp_update_setup_ext(zp, flag, mtime, ctime, B_TRUE);
1324 }
1325 /*
1326 * Grow the block size for a file.
1327 *
1328 * IN: zp - znode of file to free data in.
1329 * size - requested block size
1330 * tx - open transaction.
1331 *
1332 * NOTE: this function assumes that the znode is write locked.
1333 */
1334 void
zfs_grow_blocksize(znode_t * zp,uint64_t size,dmu_tx_t * tx)1335 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1336 {
1337 int error;
1338 u_longlong_t dummy;
1339
1340 if (size <= zp->z_blksz)
1341 return;
1342 /*
1343 * If the file size is already greater than the current blocksize,
1344 * we will not grow. If there is more than one block in a file,
1345 * the blocksize cannot change.
1346 */
1347 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1348 return;
1349
1350 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1351 size, 0, tx);
1352
1353 if (error == ENOTSUP)
1354 return;
1355 ASSERT0(error);
1356
1357 /* What blocksize did we actually get? */
1358 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1359 }
1360
1361 /*
1362 * Increase the file length
1363 *
1364 * IN: zp - znode of file to free data in.
1365 * end - new end-of-file
1366 *
1367 * RETURN: 0 on success, error code on failure
1368 */
1369 static int
zfs_extend(znode_t * zp,uint64_t end)1370 zfs_extend(znode_t *zp, uint64_t end)
1371 {
1372 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1373 dmu_tx_t *tx;
1374 zfs_locked_range_t *lr;
1375 uint64_t newblksz;
1376 int error;
1377
1378 /*
1379 * We will change zp_size, lock the whole file.
1380 */
1381 lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
1382
1383 /*
1384 * Nothing to do if file already at desired length.
1385 */
1386 if (end <= zp->z_size) {
1387 zfs_rangelock_exit(lr);
1388 return (0);
1389 }
1390 tx = dmu_tx_create(zfsvfs->z_os);
1391 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1392 zfs_sa_upgrade_txholds(tx, zp);
1393 if (end > zp->z_blksz &&
1394 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1395 /*
1396 * We are growing the file past the current block size.
1397 */
1398 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1399 /*
1400 * File's blocksize is already larger than the
1401 * "recordsize" property. Only let it grow to
1402 * the next power of 2.
1403 */
1404 ASSERT(!ISP2(zp->z_blksz));
1405 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1406 } else {
1407 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1408 }
1409 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1410 } else {
1411 newblksz = 0;
1412 }
1413
1414 error = dmu_tx_assign(tx, TXG_WAIT);
1415 if (error) {
1416 dmu_tx_abort(tx);
1417 zfs_rangelock_exit(lr);
1418 return (error);
1419 }
1420
1421 if (newblksz)
1422 zfs_grow_blocksize(zp, newblksz, tx);
1423
1424 zp->z_size = end;
1425
1426 VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1427 &zp->z_size, sizeof (zp->z_size), tx));
1428
1429 vnode_pager_setsize(ZTOV(zp), end);
1430
1431 zfs_rangelock_exit(lr);
1432
1433 dmu_tx_commit(tx);
1434
1435 return (0);
1436 }
1437
1438 /*
1439 * Free space in a file.
1440 *
1441 * IN: zp - znode of file to free data in.
1442 * off - start of section to free.
1443 * len - length of section to free.
1444 *
1445 * RETURN: 0 on success, error code on failure
1446 */
1447 static int
zfs_free_range(znode_t * zp,uint64_t off,uint64_t len)1448 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1449 {
1450 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1451 zfs_locked_range_t *lr;
1452 int error;
1453
1454 /*
1455 * Lock the range being freed.
1456 */
1457 lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER);
1458
1459 /*
1460 * Nothing to do if file already at desired length.
1461 */
1462 if (off >= zp->z_size) {
1463 zfs_rangelock_exit(lr);
1464 return (0);
1465 }
1466
1467 if (off + len > zp->z_size)
1468 len = zp->z_size - off;
1469
1470 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1471
1472 if (error == 0) {
1473 #if __FreeBSD_version >= 1400032
1474 vnode_pager_purge_range(ZTOV(zp), off, off + len);
1475 #else
1476 /*
1477 * Before __FreeBSD_version 1400032 we cannot free block in the
1478 * middle of a file, but only at the end of a file, so this code
1479 * path should never happen.
1480 */
1481 vnode_pager_setsize(ZTOV(zp), off);
1482 #endif
1483 }
1484
1485 zfs_rangelock_exit(lr);
1486
1487 return (error);
1488 }
1489
1490 /*
1491 * Truncate a file
1492 *
1493 * IN: zp - znode of file to free data in.
1494 * end - new end-of-file.
1495 *
1496 * RETURN: 0 on success, error code on failure
1497 */
1498 static int
zfs_trunc(znode_t * zp,uint64_t end)1499 zfs_trunc(znode_t *zp, uint64_t end)
1500 {
1501 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1502 vnode_t *vp = ZTOV(zp);
1503 dmu_tx_t *tx;
1504 zfs_locked_range_t *lr;
1505 int error;
1506 sa_bulk_attr_t bulk[2];
1507 int count = 0;
1508
1509 /*
1510 * We will change zp_size, lock the whole file.
1511 */
1512 lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);
1513
1514 /*
1515 * Nothing to do if file already at desired length.
1516 */
1517 if (end >= zp->z_size) {
1518 zfs_rangelock_exit(lr);
1519 return (0);
1520 }
1521
1522 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,
1523 DMU_OBJECT_END);
1524 if (error) {
1525 zfs_rangelock_exit(lr);
1526 return (error);
1527 }
1528 tx = dmu_tx_create(zfsvfs->z_os);
1529 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1530 zfs_sa_upgrade_txholds(tx, zp);
1531 dmu_tx_mark_netfree(tx);
1532 error = dmu_tx_assign(tx, TXG_WAIT);
1533 if (error) {
1534 dmu_tx_abort(tx);
1535 zfs_rangelock_exit(lr);
1536 return (error);
1537 }
1538
1539 zp->z_size = end;
1540 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1541 NULL, &zp->z_size, sizeof (zp->z_size));
1542
1543 if (end == 0) {
1544 zp->z_pflags &= ~ZFS_SPARSE;
1545 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1546 NULL, &zp->z_pflags, 8);
1547 }
1548 VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
1549
1550 dmu_tx_commit(tx);
1551
1552 /*
1553 * Clear any mapped pages in the truncated region. This has to
1554 * happen outside of the transaction to avoid the possibility of
1555 * a deadlock with someone trying to push a page that we are
1556 * about to invalidate.
1557 */
1558 vnode_pager_setsize(vp, end);
1559
1560 zfs_rangelock_exit(lr);
1561
1562 return (0);
1563 }
1564
1565 /*
1566 * Free space in a file
1567 *
1568 * IN: zp - znode of file to free data in.
1569 * off - start of range
1570 * len - end of range (0 => EOF)
1571 * flag - current file open mode flags.
1572 * log - TRUE if this action should be logged
1573 *
1574 * RETURN: 0 on success, error code on failure
1575 */
1576 int
zfs_freesp(znode_t * zp,uint64_t off,uint64_t len,int flag,boolean_t log)1577 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1578 {
1579 dmu_tx_t *tx;
1580 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1581 zilog_t *zilog = zfsvfs->z_log;
1582 uint64_t mode;
1583 uint64_t mtime[2], ctime[2];
1584 sa_bulk_attr_t bulk[3];
1585 int count = 0;
1586 int error;
1587
1588 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1589 sizeof (mode))) != 0)
1590 return (error);
1591
1592 if (off > zp->z_size) {
1593 error = zfs_extend(zp, off+len);
1594 if (error == 0 && log)
1595 goto log;
1596 else
1597 return (error);
1598 }
1599
1600 if (len == 0) {
1601 error = zfs_trunc(zp, off);
1602 } else {
1603 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1604 off + len > zp->z_size)
1605 error = zfs_extend(zp, off+len);
1606 }
1607 if (error || !log)
1608 return (error);
1609 log:
1610 tx = dmu_tx_create(zfsvfs->z_os);
1611 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1612 zfs_sa_upgrade_txholds(tx, zp);
1613 error = dmu_tx_assign(tx, TXG_WAIT);
1614 if (error) {
1615 dmu_tx_abort(tx);
1616 return (error);
1617 }
1618
1619 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1620 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1621 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1622 NULL, &zp->z_pflags, 8);
1623 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
1624 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1625 ASSERT0(error);
1626
1627 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1628
1629 dmu_tx_commit(tx);
1630 return (0);
1631 }
1632
1633 void
zfs_create_fs(objset_t * os,cred_t * cr,nvlist_t * zplprops,dmu_tx_t * tx)1634 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1635 {
1636 uint64_t moid, obj, sa_obj, version;
1637 uint64_t sense = ZFS_CASE_SENSITIVE;
1638 uint64_t norm = 0;
1639 nvpair_t *elem;
1640 int error;
1641 int i;
1642 znode_t *rootzp = NULL;
1643 zfsvfs_t *zfsvfs;
1644 vattr_t vattr;
1645 znode_t *zp;
1646 zfs_acl_ids_t acl_ids;
1647
1648 /*
1649 * First attempt to create master node.
1650 */
1651 /*
1652 * In an empty objset, there are no blocks to read and thus
1653 * there can be no i/o errors (which we assert below).
1654 */
1655 moid = MASTER_NODE_OBJ;
1656 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1657 DMU_OT_NONE, 0, tx);
1658 ASSERT0(error);
1659
1660 /*
1661 * Set starting attributes.
1662 */
1663 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1664 elem = NULL;
1665 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1666 /* For the moment we expect all zpl props to be uint64_ts */
1667 uint64_t val;
1668 const char *name;
1669
1670 ASSERT3S(nvpair_type(elem), ==, DATA_TYPE_UINT64);
1671 val = fnvpair_value_uint64(elem);
1672 name = nvpair_name(elem);
1673 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1674 if (val < version)
1675 version = val;
1676 } else {
1677 error = zap_update(os, moid, name, 8, 1, &val, tx);
1678 }
1679 ASSERT0(error);
1680 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1681 norm = val;
1682 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1683 sense = val;
1684 }
1685 ASSERT3U(version, !=, 0);
1686 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1687 ASSERT0(error);
1688
1689 /*
1690 * Create zap object used for SA attribute registration
1691 */
1692
1693 if (version >= ZPL_VERSION_SA) {
1694 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1695 DMU_OT_NONE, 0, tx);
1696 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1697 ASSERT0(error);
1698 } else {
1699 sa_obj = 0;
1700 }
1701 /*
1702 * Create a delete queue.
1703 */
1704 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1705
1706 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1707 ASSERT0(error);
1708
1709 /*
1710 * Create root znode. Create minimal znode/vnode/zfsvfs
1711 * to allow zfs_mknode to work.
1712 */
1713 VATTR_NULL(&vattr);
1714 vattr.va_mask = AT_MODE|AT_UID|AT_GID;
1715 vattr.va_type = VDIR;
1716 vattr.va_mode = S_IFDIR|0755;
1717 vattr.va_uid = crgetuid(cr);
1718 vattr.va_gid = crgetgid(cr);
1719
1720 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
1721
1722 rootzp = zfs_znode_alloc_kmem(KM_SLEEP);
1723 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1724 rootzp->z_unlinked = 0;
1725 rootzp->z_atime_dirty = 0;
1726 rootzp->z_is_sa = USE_SA(version, os);
1727
1728 zfsvfs->z_os = os;
1729 zfsvfs->z_parent = zfsvfs;
1730 zfsvfs->z_version = version;
1731 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
1732 zfsvfs->z_use_sa = USE_SA(version, os);
1733 zfsvfs->z_norm = norm;
1734
1735 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1736 &zfsvfs->z_attr_table);
1737
1738 ASSERT0(error);
1739
1740 /*
1741 * Fold case on file systems that are always or sometimes case
1742 * insensitive.
1743 */
1744 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1745 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
1746
1747 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1748 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
1749 offsetof(znode_t, z_link_node));
1750
1751 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1752 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1753
1754 rootzp->z_zfsvfs = zfsvfs;
1755 VERIFY0(zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1756 cr, NULL, &acl_ids, NULL));
1757 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1758 ASSERT3P(zp, ==, rootzp);
1759 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1760 ASSERT0(error);
1761 zfs_acl_ids_free(&acl_ids);
1762 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1763
1764 sa_handle_destroy(rootzp->z_sa_hdl);
1765 zfs_znode_free_kmem(rootzp);
1766
1767 /*
1768 * Create shares directory
1769 */
1770
1771 error = zfs_create_share_dir(zfsvfs, tx);
1772
1773 ASSERT0(error);
1774
1775 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1776 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1777 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1778 }
1779
1780 void
zfs_znode_update_vfs(znode_t * zp)1781 zfs_znode_update_vfs(znode_t *zp)
1782 {
1783 vm_object_t object;
1784
1785 if ((object = ZTOV(zp)->v_object) == NULL ||
1786 zp->z_size == object->un_pager.vnp.vnp_size)
1787 return;
1788
1789 vnode_pager_setsize(ZTOV(zp), zp->z_size);
1790 }
1791
1792 int
zfs_znode_parent_and_name(znode_t * zp,znode_t ** dzpp,char * buf,uint64_t buflen)1793 zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf,
1794 uint64_t buflen)
1795 {
1796 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1797 uint64_t parent;
1798 int is_xattrdir;
1799 int err;
1800
1801 /* Extended attributes should not be visible as regular files. */
1802 if ((zp->z_pflags & ZFS_XATTR) != 0)
1803 return (SET_ERROR(EINVAL));
1804
1805 err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table,
1806 &parent, &is_xattrdir);
1807 if (err != 0)
1808 return (err);
1809 ASSERT0(is_xattrdir);
1810
1811 /* No name as this is a root object. */
1812 if (parent == zp->z_id)
1813 return (SET_ERROR(EINVAL));
1814
1815 err = zap_value_search(zfsvfs->z_os, parent, zp->z_id,
1816 ZFS_DIRENT_OBJ(-1ULL), buf, buflen);
1817 if (err != 0)
1818 return (err);
1819 err = zfs_zget(zfsvfs, parent, dzpp);
1820 return (err);
1821 }
1822
1823 int
zfs_rlimit_fsize(off_t fsize)1824 zfs_rlimit_fsize(off_t fsize)
1825 {
1826 struct thread *td = curthread;
1827 off_t lim;
1828
1829 if (td == NULL)
1830 return (0);
1831
1832 lim = lim_cur(td, RLIMIT_FSIZE);
1833 if (__predict_true((uoff_t)fsize <= lim))
1834 return (0);
1835
1836 /*
1837 * The limit is reached.
1838 */
1839 PROC_LOCK(td->td_proc);
1840 kern_psignal(td->td_proc, SIGXFSZ);
1841 PROC_UNLOCK(td->td_proc);
1842
1843 return (EFBIG);
1844 }
1845