Print this page
OS-5343 dockerinit gets stuck in zfd_ready, zfd devices not showing up after > 1 hour
Reviewed by: Patrick Mooney <patrick.mooney@joyent.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/dev/sdev_vnops.c
+++ new/usr/src/uts/common/fs/dev/sdev_vnops.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
|
↓ open down ↓ |
14 lines elided |
↑ open up ↑ |
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
23 23 */
24 24 /*
25 - * Copyright (c) 2013, Joyent, Inc. All rights reserved.
25 + * Copyright 2016, Joyent, Inc.
26 26 */
27 27
28 28 /*
29 29 * vnode ops for the /dev filesystem
30 30 *
31 31 * - VDIR, VCHR, CBLK, and VLNK are considered must supported files
32 32 * - VREG and VDOOR are used for some internal implementations in
33 33 * the global zone, e.g. devname and devfsadm communication
34 34 * - other file types are unusual in this namespace and
35 35 * not supported for now
36 36 */
37 37
38 38 /*
39 39 * sdev has a few basic goals:
40 40 * o Provide /dev for the global zone as well as various non-global zones.
41 41 * o Provide the basic functionality that devfsadm might need (mknod,
42 42 * symlinks, etc.)
43 43 * o Allow persistent permissions on files in /dev.
44 44 * o Allow for dynamic directories and nodes for use by various services (pts,
45 45 * zvol, net, etc.)
46 46 *
47 47 * The sdev file system is primarily made up of sdev_node_t's which is sdev's
48 48 * counterpart to the vnode_t. There are two different classes of sdev_node_t's
49 49 * that we generally care about, dynamic and otherwise.
50 50 *
51 51 * Persisting Information
52 52 * ----------------------
53 53 *
54 54 * When sdev is mounted, it keeps track of the underlying file system it is
55 55 * mounted over. In certain situations, sdev will go and create entries in that
56 56 * underlying file system. These underlying 'back end' nodes are used as proxies
57 57 * for various changes in permissions. While specific sets of nodes, such as
58 58 * dynamic ones, are exempt, this process stores permission changes against
59 59 * these back end nodes. The point of all of this is to allow for these settings
60 60 * to persist across host and zone reboots. As an example, consider the entry
61 61 * /dev/dsk/c0t0d0 which is a character device and that / is in UFS. Upon
62 62 * changing the permissions on c0t0d0 you'd have the following logical
63 63 * relationships:
64 64 *
65 65 * +------------------+ sdev_vnode +--------------+
66 66 * | sdev_node_t |<---------------->| vnode_t |
67 67 * | /dev/dsk/c0t0d0 |<---------------->| for sdev |
68 68 * +------------------+ +--------------+
69 69 * |
70 70 * | sdev_attrvp
71 71 * |
72 72 * | +---------------------+
73 73 * +--->| vnode_t for UFS|ZFS |
74 74 * | /dev/dsk/c0t0d0 |
75 75 * +---------------------+
76 76 *
77 77 * sdev is generally in memory. Therefore when a lookup happens and there is no
78 78 * entry already inside of a directory cache, it will next check the backing
79 79 * store. If the backing store exists, we will reconstitute the sdev_node based
80 80 * on the information that we persisted. When we create the backing store node,
81 81 * we use the struct vattr information that we already have in sdev_node_t.
82 82 * Because of this, we already know if the entry was previously a symlink,
83 83 * directory, or some other kind of type. Note that not all types of nodes are
84 84 * supported. Currently only VDIR, VCHR, VBLK, VREG, VDOOR, and VLNK are
85 85 * eligible to be persisted.
86 86 *
87 87 * When the sdev_node is created and the lookup is done, we grab a hold on the
88 88 * underlying vnode as part of the call to VOP_LOOKUP. That reference is held
89 89 * until the sdev_node becomes inactive. Once its reference count reaches one
90 90 * and the VOP_INACTIVE callback fires leading to the destruction of the node,
91 91 * the reference on the underlying vnode will be released.
92 92 *
93 93 * The backing store node will be deleted only when the node itself is deleted
94 94 * through the means of a VOP_REMOVE, VOP_RMDIR, or similar call.
95 95 *
96 96 * Not everything can be persisted, see The Rules section for more details.
97 97 *
98 98 * Dynamic Nodes
99 99 * -------------
100 100 *
101 101 * Dynamic nodes allow for specific interactions with various kernel subsystems
102 102 * when looking up directory entries. This allows the lookup and readdir
103 103 * functions to check against the kernel subsystem's for validity. eg. does a
104 104 * zvol or nic still exist.
105 105 *
106 106 * More specifically, when we create various directories we check if the
107 107 * directory name matches that of one of the names in the vtab[] (sdev_subr.c).
108 108 * If it does, we swap out the vnode operations into a new set which combine the
109 109 * normal sdev vnode operations with the dynamic set here.
110 110 *
111 111 * In addition, various dynamic nodes implement a verification entry point. This
112 112 * verification entry is used as a part of lookup and readdir. The goal for
113 113 * these dynamic nodes is to allow them to check with the underlying subsystems
114 114 * to ensure that these devices are still present, or if they have gone away, to
115 115 * remove them from the results. This is indicated by using the SDEV_VTOR flag
116 116 * in vtab[].
117 117 *
118 118 * Dynamic nodes have additional restrictions placed upon them. They may only
119 119 * appear at the top level directory of the file system. In addition, users
120 120 * cannot create dirents below any leve of a dynamic node aside from its special
121 121 * vnops.
122 122 *
123 123 * Profiles
124 124 * --------
125 125 *
126 126 * Profiles exist for the purpose of non-global zones. They work with the zone
127 127 * brands and zoneadmd to set up a filter of allowed devices that can appear in
128 128 * a non-global zone's /dev. These are sent to sdev by means of libdevinfo and a
129 129 * modctl system call. Specifically it allows one to add patterns of device
130 130 * paths to include and exclude. It allows for a collection of symlinks to be
131 131 * added and it allows for remapping names.
132 132 *
133 133 * When operating in a non-global zone, several of the sdev vnops are redirected
134 134 * to the profile versions. These impose additional restrictions such as
135 135 * enforcing that a non-global zone's /dev is read only.
136 136 *
137 137 * sdev_node_t States
138 138 * ------------------
139 139 *
140 140 * A given sdev_node_t has a field called the sdev_state which describes where
141 141 * in the sdev life cycle it is. There are three primary states: SDEV_INIT,
142 142 * SDEV_READY, and SDEV_ZOMBIE.
143 143 *
144 144 * SDEV_INIT: When a new /dev file is first looked up, a sdev_node
145 145 * is allocated, initialized and added to the directory's
146 146 * sdev_node cache. A node at this state will also
147 147 * have the SDEV_LOOKUP flag set.
148 148 *
149 149 * Other threads that are trying to look up a node at
150 150 * this state will be blocked until the SDEV_LOOKUP flag
151 151 * is cleared.
152 152 *
153 153 * When the SDEV_LOOKUP flag is cleared, the node may
154 154 * transition into the SDEV_READY state for a successful
155 155 * lookup or the node is removed from the directory cache
156 156 * and destroyed if the named node can not be found.
157 157 * An ENOENT error is returned for the second case.
158 158 *
159 159 * SDEV_READY: A /dev file has been successfully looked up and
160 160 * associated with a vnode. The /dev file is available
161 161 * for the supported /dev file system operations.
162 162 *
163 163 * SDEV_ZOMBIE: Deletion of a /dev file has been explicitly issued
164 164 * to an SDEV_READY node. The node is transitioned into
165 165 * the SDEV_ZOMBIE state if the vnode reference count
166 166 * is still held. A SDEV_ZOMBIE node does not support
167 167 * any of the /dev file system operations. A SDEV_ZOMBIE
168 168 * node is immediately removed from the directory cache
169 169 * and destroyed once the reference count reaches zero.
170 170 *
171 171 * Historically nodes that were marked SDEV_ZOMBIE were not removed from the
172 172 * underlying directory caches. This has been the source of numerous bugs and
173 173 * thus to better mimic what happens on a real file system, it is no longer the
174 174 * case.
175 175 *
176 176 * The following state machine describes the life cycle of a given node and its
177 177 * associated states:
178 178 *
179 179 * node is . . . . .
180 180 * allocated via . +-------------+ . . . . . . . vnode_t refcount
181 181 * sdev_nodeinit() . | Unallocated | . reaches zero and
182 182 * +--------*-----| Memory |<--------*---+ sdev_inactive is
183 183 * | +-------------+ | called.
184 184 * | +------------^ | called.
185 185 * v | |
186 186 * +-----------+ * . . sdev_nodeready() +-------------+
187 187 * | SDEV_INIT | | or related setup | SDEV_ZOMBIE |
188 188 * +-----------+ | failure +-------------+
189 189 * | | ^
190 190 * | | +------------+ |
191 191 * +-*----------->| SDEV_READY |--------*-----+
192 192 * . +------------+ . The node is no longer
193 193 * . . node successfully . . . . . valid or we've been
194 194 * inserted into the asked to remove it.
195 195 * directory cache This happens via
196 196 * and sdev_nodready() sdev_dirdelete().
197 197 * call successful.
198 198 *
199 199 * Adding and Removing Dirents, Zombie Nodes
200 200 * -----------------------------------------
201 201 *
202 202 * As part of doing a lookup, readdir, or an explicit creation operation like
203 203 * mkdir or create, nodes may be created. Every directory has an avl tree which
204 204 * contains its children, the sdev_entries tree. This is only used if the type
205 205 * is VDIR. Access to this is controlled by the sdev_node_t's contents_lock and
206 206 * it is managed through sdev_cache_update().
207 207 *
208 208 * Every sdev_node_t has a field sdev_state, which describes the current state
209 209 * of the node. A node is generally speaking in the SDEV_READY state. When it is
210 210 * there, it can be looked up, accessed, and operations performed on it. When a
211 211 * node is going to be removed from the directory cache it is marked as a
212 212 * zombie. Once a node becomes a zombie, no other file system operations will
213 213 * succeed and it will continue to exist as a node until the vnode count on the
214 214 * node reaches zero. At that point, the node will be freed. However, once a
215 215 * node has been marked as a zombie, it will be removed immediately from the
216 216 * directory cache such that no one else may find it again. This means that
217 217 * someone else can insert a new entry into that directory with the same name
218 218 * and without a problem.
219 219 *
220 220 * To remove a node, see the section on that in The Rules.
221 221 *
222 222 * The Rules
223 223 * ---------
224 224 * These are the rules to live by when working in sdev. These are not
225 225 * exhaustive.
226 226 *
227 227 * - Set 1: Working with Backing Nodes
228 228 * o If there is a SDEV_READY sdev_node_t, it knows about its backing node.
229 229 * o If we find a backing node when looking up an sdev_node_t for the first
230 230 * time, we use its attributes to build our sdev_node_t.
231 231 * o If there is a found backing node, or we create a backing node, that's
232 232 * when we grab the hold on its vnode.
233 233 * o If we mark an sdev_node_t a ZOMBIE, we must remove its backing node from
234 234 * the underlying file system. It must not be searchable or findable.
235 235 * o We release our hold on the backing node vnode when we destroy the
236 236 * sdev_node_t.
237 237 *
238 238 * - Set 2: Locking rules for sdev (not exhaustive)
239 239 * o The majority of nodes contain an sdev_contents rw lock. You must hold it
240 240 * for read or write if manipulating its contents appropriately.
241 241 * o You must lock your parent before yourself.
242 242 * o If you need your vnode's v_lock and the sdev_contents rw lock, you must
243 243 * grab the v_lock before the sdev_contents rw_lock.
244 244 * o If you release a lock on the node as a part of upgrading it, you must
245 245 * verify that the node has not become a zombie as a part of this process.
246 246 *
247 247 * - Set 3: Zombie Status and What it Means
248 248 * o If you encounter a node that is a ZOMBIE, that means that it has been
249 249 * unlinked from the backing store.
250 250 * o If you release your contents lock and acquire it again (say as part of
251 251 * trying to grab a write lock) you must check that the node has not become
252 252 * a zombie.
253 253 * o You should VERIFY that a looked up node is not a zombie. This follows
254 254 * from the following logic. To mark something as a zombie means that it is
255 255 * removed from the parents directory cache. To do that, you must have a
256 256 * write lock on the parent's sdev_contents. To lookup through that
257 257 * directory you must have a read lock. This then becomes a simple ordering
258 258 * problem. If you've been granted the lock then the other operation cannot
259 259 * be in progress or must have already succeeded.
260 260 *
261 261 * - Set 4: Removing Directory Entries (aka making nodes Zombies)
262 262 * o Write lock must be held on the directory
263 263 * o Write lock must be held on the node
264 264 * o Remove the sdev_node_t from its parent cache
265 265 * o Remove the corresponding backing store node, if it exists, eg. use
266 266 * VOP_REMOVE or VOP_RMDIR.
267 267 * o You must NOT make any change in the vnode reference count! Nodes should
268 268 * only be cleaned up through VOP_INACTIVE callbacks.
269 269 * o VOP_INACTIVE is the only one responsible for doing the final vn_rele of
270 270 * the backing store vnode that was grabbed during lookup.
271 271 *
272 272 * - Set 5: What Nodes may be Persisted
273 273 * o The root, /dev is always persisted
274 274 * o Any node in vtab which is marked SDEV_DYNAMIC, may not be persisted
275 275 * unless it is also marked SDEV_PERSIST
276 276 * o Anything whose parent directory is marked SDEV_PERSIST will pass that
277 277 * along to the child as long as it does not contradict the above rules
278 278 */
279 279
280 280 #include <sys/types.h>
281 281 #include <sys/param.h>
282 282 #include <sys/t_lock.h>
283 283 #include <sys/systm.h>
284 284 #include <sys/sysmacros.h>
285 285 #include <sys/user.h>
286 286 #include <sys/time.h>
287 287 #include <sys/vfs.h>
288 288 #include <sys/vnode.h>
289 289 #include <sys/vfs_opreg.h>
290 290 #include <sys/file.h>
291 291 #include <sys/fcntl.h>
292 292 #include <sys/flock.h>
293 293 #include <sys/kmem.h>
294 294 #include <sys/uio.h>
295 295 #include <sys/errno.h>
296 296 #include <sys/stat.h>
297 297 #include <sys/cred.h>
298 298 #include <sys/dirent.h>
299 299 #include <sys/pathname.h>
300 300 #include <sys/cmn_err.h>
301 301 #include <sys/debug.h>
302 302 #include <sys/policy.h>
303 303 #include <vm/hat.h>
304 304 #include <vm/seg_vn.h>
305 305 #include <vm/seg_map.h>
306 306 #include <vm/seg.h>
307 307 #include <vm/as.h>
308 308 #include <vm/page.h>
309 309 #include <sys/proc.h>
310 310 #include <sys/mode.h>
311 311 #include <sys/sunndi.h>
312 312 #include <sys/ptms.h>
313 313 #include <fs/fs_subr.h>
314 314 #include <sys/fs/dv_node.h>
315 315 #include <sys/fs/sdev_impl.h>
316 316
317 317 /*ARGSUSED*/
318 318 static int
319 319 sdev_open(struct vnode **vpp, int flag, struct cred *cred, caller_context_t *ct)
320 320 {
321 321 struct sdev_node *dv = VTOSDEV(*vpp);
322 322 struct sdev_node *ddv = dv->sdev_dotdot;
323 323 int error = 0;
324 324
325 325 if ((*vpp)->v_type == VDIR)
326 326 return (0);
327 327
328 328 if (!SDEV_IS_GLOBAL(dv))
329 329 return (ENOTSUP);
330 330
331 331 if ((*vpp)->v_type == VLNK)
332 332 return (ENOENT);
333 333 ASSERT((*vpp)->v_type == VREG);
334 334 if ((*vpp)->v_type != VREG)
335 335 return (ENOTSUP);
336 336
337 337 ASSERT(ddv);
338 338 rw_enter(&ddv->sdev_contents, RW_READER);
339 339 if (dv->sdev_attrvp == NULL) {
340 340 rw_exit(&ddv->sdev_contents);
341 341 return (ENOENT);
342 342 }
343 343 error = VOP_OPEN(&(dv->sdev_attrvp), flag, cred, ct);
344 344 rw_exit(&ddv->sdev_contents);
345 345 return (error);
346 346 }
347 347
348 348 /*ARGSUSED1*/
349 349 static int
350 350 sdev_close(struct vnode *vp, int flag, int count,
351 351 offset_t offset, struct cred *cred, caller_context_t *ct)
352 352 {
353 353 struct sdev_node *dv = VTOSDEV(vp);
354 354
355 355 if (vp->v_type == VDIR) {
356 356 cleanlocks(vp, ttoproc(curthread)->p_pid, 0);
357 357 cleanshares(vp, ttoproc(curthread)->p_pid);
358 358 return (0);
359 359 }
360 360
361 361 if (!SDEV_IS_GLOBAL(dv))
362 362 return (ENOTSUP);
363 363
364 364 ASSERT(vp->v_type == VREG);
365 365 if (vp->v_type != VREG)
366 366 return (ENOTSUP);
367 367
368 368 ASSERT(dv->sdev_attrvp);
369 369 return (VOP_CLOSE(dv->sdev_attrvp, flag, count, offset, cred, ct));
370 370 }
371 371
372 372 /*ARGSUSED*/
373 373 static int
374 374 sdev_read(struct vnode *vp, struct uio *uio, int ioflag, struct cred *cred,
375 375 struct caller_context *ct)
376 376 {
377 377 struct sdev_node *dv = (struct sdev_node *)VTOSDEV(vp);
378 378 int error;
379 379
380 380 if (!SDEV_IS_GLOBAL(dv))
381 381 return (EINVAL);
382 382
383 383 if (vp->v_type == VDIR)
384 384 return (EISDIR);
385 385
386 386 /* only supporting regular files in /dev */
387 387 ASSERT(vp->v_type == VREG);
388 388 if (vp->v_type != VREG)
389 389 return (EINVAL);
390 390
391 391 ASSERT(RW_READ_HELD(&VTOSDEV(vp)->sdev_contents));
392 392 ASSERT(dv->sdev_attrvp);
393 393 (void) VOP_RWLOCK(dv->sdev_attrvp, 0, ct);
394 394 error = VOP_READ(dv->sdev_attrvp, uio, ioflag, cred, ct);
395 395 VOP_RWUNLOCK(dv->sdev_attrvp, 0, ct);
396 396 return (error);
397 397 }
398 398
399 399 /*ARGSUSED*/
400 400 static int
401 401 sdev_write(struct vnode *vp, struct uio *uio, int ioflag, struct cred *cred,
402 402 struct caller_context *ct)
403 403 {
404 404 struct sdev_node *dv = VTOSDEV(vp);
405 405 int error = 0;
406 406
407 407 if (!SDEV_IS_GLOBAL(dv))
408 408 return (EINVAL);
409 409
410 410 if (vp->v_type == VDIR)
411 411 return (EISDIR);
412 412
413 413 /* only supporting regular files in /dev */
414 414 ASSERT(vp->v_type == VREG);
415 415 if (vp->v_type != VREG)
416 416 return (EINVAL);
417 417
418 418 ASSERT(dv->sdev_attrvp);
419 419
420 420 (void) VOP_RWLOCK(dv->sdev_attrvp, 1, ct);
421 421 error = VOP_WRITE(dv->sdev_attrvp, uio, ioflag, cred, ct);
422 422 VOP_RWUNLOCK(dv->sdev_attrvp, 1, ct);
423 423 if (error == 0) {
424 424 sdev_update_timestamps(dv->sdev_attrvp, kcred,
425 425 AT_MTIME);
426 426 }
427 427 return (error);
428 428 }
429 429
430 430 /*ARGSUSED*/
431 431 static int
432 432 sdev_ioctl(struct vnode *vp, int cmd, intptr_t arg, int flag,
433 433 struct cred *cred, int *rvalp, caller_context_t *ct)
434 434 {
435 435 struct sdev_node *dv = VTOSDEV(vp);
436 436
437 437 if (!SDEV_IS_GLOBAL(dv) || (vp->v_type == VDIR))
438 438 return (ENOTTY);
439 439
440 440 ASSERT(vp->v_type == VREG);
441 441 if (vp->v_type != VREG)
442 442 return (EINVAL);
443 443
444 444 ASSERT(dv->sdev_attrvp);
445 445 return (VOP_IOCTL(dv->sdev_attrvp, cmd, arg, flag, cred, rvalp, ct));
446 446 }
447 447
448 448 static int
449 449 sdev_getattr(struct vnode *vp, struct vattr *vap, int flags,
450 450 struct cred *cr, caller_context_t *ct)
451 451 {
452 452 int error = 0;
453 453 struct sdev_node *dv = VTOSDEV(vp);
454 454 struct sdev_node *parent = dv->sdev_dotdot;
455 455
456 456 ASSERT(parent);
457 457
458 458 rw_enter(&parent->sdev_contents, RW_READER);
459 459 ASSERT(dv->sdev_attr || dv->sdev_attrvp);
460 460
461 461 /*
462 462 * search order:
463 463 * - for persistent nodes (SDEV_PERSIST): backstore
464 464 * - for non-persistent nodes: module ops if global, then memory
465 465 */
466 466 if (dv->sdev_attrvp) {
467 467 rw_exit(&parent->sdev_contents);
468 468 error = VOP_GETATTR(dv->sdev_attrvp, vap, flags, cr, ct);
469 469 sdev_vattr_merge(dv, vap);
470 470 } else {
471 471 ASSERT(dv->sdev_attr);
472 472 *vap = *dv->sdev_attr;
473 473 sdev_vattr_merge(dv, vap);
474 474 rw_exit(&parent->sdev_contents);
475 475 }
476 476
477 477 return (error);
478 478 }
479 479
480 480 /*ARGSUSED4*/
481 481 static int
482 482 sdev_setattr(struct vnode *vp, struct vattr *vap, int flags,
483 483 struct cred *cred, caller_context_t *ctp)
484 484 {
485 485 return (devname_setattr_func(vp, vap, flags, cred, NULL, 0));
486 486 }
487 487
488 488 static int
489 489 sdev_getsecattr(struct vnode *vp, struct vsecattr *vsap, int flags,
490 490 struct cred *cr, caller_context_t *ct)
491 491 {
492 492 int error;
493 493 struct sdev_node *dv = VTOSDEV(vp);
494 494 struct vnode *avp = dv->sdev_attrvp;
495 495
496 496 if (avp == NULL) {
497 497 /* return fs_fab_acl() if flavor matches, else do nothing */
498 498 if ((SDEV_ACL_FLAVOR(vp) == _ACL_ACLENT_ENABLED &&
499 499 (vsap->vsa_mask & (VSA_ACLCNT | VSA_DFACLCNT))) ||
500 500 (SDEV_ACL_FLAVOR(vp) == _ACL_ACE_ENABLED &&
501 501 (vsap->vsa_mask & (VSA_ACECNT | VSA_ACE))))
502 502 return (fs_fab_acl(vp, vsap, flags, cr, ct));
503 503
504 504 return (ENOSYS);
505 505 }
506 506
507 507 (void) VOP_RWLOCK(avp, 1, ct);
508 508 error = VOP_GETSECATTR(avp, vsap, flags, cr, ct);
509 509 VOP_RWUNLOCK(avp, 1, ct);
510 510 return (error);
511 511 }
512 512
513 513 static int
514 514 sdev_setsecattr(struct vnode *vp, struct vsecattr *vsap, int flags,
515 515 struct cred *cr, caller_context_t *ct)
516 516 {
517 517 int error;
518 518 struct sdev_node *dv = VTOSDEV(vp);
519 519 struct vnode *avp = dv->sdev_attrvp;
520 520
521 521 if (dv->sdev_state == SDEV_ZOMBIE)
522 522 return (0);
523 523
524 524 if (avp == NULL) {
525 525 if (SDEV_IS_GLOBAL(dv) && !SDEV_IS_PERSIST(dv))
526 526 return (fs_nosys());
527 527 ASSERT(dv->sdev_attr);
528 528 /*
529 529 * if coming in directly, the acl system call will
530 530 * have held the read-write lock via VOP_RWLOCK()
531 531 * If coming in via specfs, specfs will have
532 532 * held the rw lock on the realvp i.e. us.
533 533 */
534 534 ASSERT(RW_WRITE_HELD(&dv->sdev_contents));
535 535 sdev_vattr_merge(dv, dv->sdev_attr);
536 536 error = sdev_shadow_node(dv, cr);
537 537 if (error) {
538 538 return (fs_nosys());
539 539 }
540 540
541 541 ASSERT(dv->sdev_attrvp);
542 542 /* clean out the memory copy if any */
543 543 if (dv->sdev_attr) {
544 544 kmem_free(dv->sdev_attr, sizeof (struct vattr));
545 545 dv->sdev_attr = NULL;
546 546 }
547 547 avp = dv->sdev_attrvp;
548 548 }
549 549 ASSERT(avp);
550 550
551 551 (void) VOP_RWLOCK(avp, V_WRITELOCK_TRUE, ct);
552 552 error = VOP_SETSECATTR(avp, vsap, flags, cr, ct);
553 553 VOP_RWUNLOCK(avp, V_WRITELOCK_TRUE, ct);
554 554 return (error);
555 555 }
556 556
557 557 /*
558 558 * There are two different unlocked routines. This one is not static as it is
559 559 * used as part of the secpolicy_vnode_setattr calls in sdev_subr.c. Because it
560 560 * is used in that function it has to have a specific signature.
561 561 */
562 562 int
563 563 sdev_unlocked_access(void *vdv, int mode, struct cred *cr)
564 564 {
565 565 struct sdev_node *dv = vdv;
566 566 int shift = 0;
567 567 uid_t owner = dv->sdev_attr->va_uid;
568 568
569 569 if (crgetuid(cr) != owner) {
570 570 shift += 3;
571 571 if (groupmember(dv->sdev_attr->va_gid, cr) == 0)
572 572 shift += 3;
573 573 }
574 574
575 575 return (secpolicy_vnode_access2(cr, SDEVTOV(dv), owner,
576 576 dv->sdev_attr->va_mode << shift, mode));
577 577 }
578 578
579 579 static int
580 580 sdev_self_access(sdev_node_t *dv, int mode, int flags, struct cred *cr,
581 581 caller_context_t *ct)
582 582 {
583 583 int ret;
584 584
585 585 ASSERT(dv->sdev_attr || dv->sdev_attrvp);
586 586 if (dv->sdev_attrvp) {
587 587 ret = VOP_ACCESS(dv->sdev_attrvp, mode, flags, cr, ct);
588 588 } else if (dv->sdev_attr) {
589 589 ret = sdev_unlocked_access(dv, mode, cr);
590 590 if (ret)
591 591 ret = EACCES;
592 592 }
593 593
594 594 return (ret);
595 595 }
596 596
597 597 static int
598 598 sdev_access(struct vnode *vp, int mode, int flags, struct cred *cr,
599 599 caller_context_t *ct)
600 600 {
601 601 struct sdev_node *dv = VTOSDEV(vp);
602 602 int ret;
603 603
604 604 rw_enter(&dv->sdev_contents, RW_READER);
605 605 ret = sdev_self_access(dv, mode, flags, cr, ct);
606 606 rw_exit(&dv->sdev_contents);
607 607
608 608 return (ret);
609 609 }
610 610
611 611 /*
612 612 * Lookup
613 613 */
614 614 /*ARGSUSED3*/
615 615 static int
616 616 sdev_lookup(struct vnode *dvp, char *nm, struct vnode **vpp,
617 617 struct pathname *pnp, int flags, struct vnode *rdir, struct cred *cred,
618 618 caller_context_t *ct, int *direntflags, pathname_t *realpnp)
619 619 {
620 620 struct sdev_node *parent;
621 621 int error;
622 622
623 623 parent = VTOSDEV(dvp);
624 624 ASSERT(parent);
625 625
626 626 /* execute access is required to search the directory */
627 627 if ((error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct)) != 0)
628 628 return (error);
629 629
630 630 if (!SDEV_IS_GLOBAL(parent))
631 631 return (prof_lookup(dvp, nm, vpp, cred));
632 632 return (devname_lookup_func(parent, nm, vpp, cred, NULL, 0));
633 633 }
634 634
635 635 /*ARGSUSED2*/
636 636 static int
637 637 sdev_create(struct vnode *dvp, char *nm, struct vattr *vap, vcexcl_t excl,
638 638 int mode, struct vnode **vpp, struct cred *cred, int flag,
639 639 caller_context_t *ct, vsecattr_t *vsecp)
640 640 {
641 641 struct vnode *vp = NULL;
642 642 struct vnode *avp;
643 643 struct sdev_node *parent;
644 644 struct sdev_node *self = NULL;
645 645 int error = 0;
646 646 vtype_t type = vap->va_type;
647 647
648 648 ASSERT(type != VNON && type != VBAD);
649 649
650 650 if ((type == VFIFO) || (type == VSOCK) ||
651 651 (type == VPROC) || (type == VPORT))
652 652 return (ENOTSUP);
653 653
654 654 parent = VTOSDEV(dvp);
655 655 ASSERT(parent);
656 656
657 657 rw_enter(&parent->sdev_dotdot->sdev_contents, RW_READER);
658 658 if (parent->sdev_state == SDEV_ZOMBIE) {
659 659 rw_exit(&parent->sdev_dotdot->sdev_contents);
660 660 return (ENOENT);
661 661 }
662 662
663 663 /* non-global do not allow pure node creation */
664 664 if (!SDEV_IS_GLOBAL(parent)) {
665 665 rw_exit(&parent->sdev_dotdot->sdev_contents);
666 666 return (prof_lookup(dvp, nm, vpp, cred));
667 667 }
668 668 rw_exit(&parent->sdev_dotdot->sdev_contents);
669 669
670 670 /* execute access is required to search the directory */
671 671 if ((error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct)) != 0)
672 672 return (error);
673 673
674 674 /* check existing name */
675 675 /* XXXci - We may need to translate the C-I flags on VOP_LOOKUP */
676 676 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cred, ct, NULL, NULL);
677 677
678 678 /* name found */
679 679 if (error == 0) {
680 680 ASSERT(vp);
681 681 if (excl == EXCL) {
682 682 error = EEXIST;
683 683 } else if ((vp->v_type == VDIR) && (mode & VWRITE)) {
684 684 /* allowing create/read-only an existing directory */
685 685 error = EISDIR;
686 686 } else {
687 687 error = VOP_ACCESS(vp, mode, 0, cred, ct);
688 688 }
689 689
690 690 if (error) {
691 691 VN_RELE(vp);
692 692 return (error);
693 693 }
694 694
695 695 /* truncation first */
696 696 if ((vp->v_type == VREG) && (vap->va_mask & AT_SIZE) &&
697 697 (vap->va_size == 0)) {
698 698 ASSERT(parent->sdev_attrvp);
699 699 error = VOP_CREATE(parent->sdev_attrvp,
700 700 nm, vap, excl, mode, &avp, cred, flag, ct, vsecp);
701 701
702 702 if (error) {
703 703 VN_RELE(vp);
704 704 return (error);
705 705 }
706 706 }
707 707
708 708 sdev_update_timestamps(vp, kcred,
709 709 AT_CTIME|AT_MTIME|AT_ATIME);
710 710 *vpp = vp;
711 711 return (0);
712 712 }
713 713
714 714 /* bail out early */
715 715 if (error != ENOENT)
716 716 return (error);
717 717
718 718 /* verify write access - compliance specifies ENXIO */
719 719 if ((error = VOP_ACCESS(dvp, VEXEC|VWRITE, 0, cred, ct)) != 0) {
720 720 if (error == EACCES)
721 721 error = ENXIO;
722 722 return (error);
723 723 }
724 724
725 725 /*
726 726 * For memory-based (ROFS) directory:
727 727 * - either disallow node creation;
728 728 * - or implement VOP_CREATE of its own
729 729 */
730 730 rw_enter(&parent->sdev_contents, RW_WRITER);
731 731 if (!SDEV_IS_PERSIST(parent)) {
732 732 rw_exit(&parent->sdev_contents);
733 733 return (ENOTSUP);
734 734 }
735 735 ASSERT(parent->sdev_attrvp);
736 736 error = sdev_mknode(parent, nm, &self, vap, NULL, NULL,
737 737 cred, SDEV_READY);
738 738 if (error) {
739 739 rw_exit(&parent->sdev_contents);
740 740 if (self)
741 741 SDEV_RELE(self);
742 742 return (error);
743 743 }
744 744 rw_exit(&parent->sdev_contents);
745 745
746 746 ASSERT(self);
747 747 /* take care the timestamps for the node and its parent */
748 748 sdev_update_timestamps(SDEVTOV(self), kcred,
749 749 AT_CTIME|AT_MTIME|AT_ATIME);
750 750 sdev_update_timestamps(dvp, kcred, AT_MTIME|AT_ATIME);
751 751 if (SDEV_IS_GLOBAL(parent))
752 752 atomic_inc_ulong(&parent->sdev_gdir_gen);
753 753
754 754 /* wake up other threads blocked on looking up this node */
755 755 mutex_enter(&self->sdev_lookup_lock);
756 756 SDEV_UNBLOCK_OTHERS(self, SDEV_LOOKUP);
757 757 mutex_exit(&self->sdev_lookup_lock);
758 758 error = sdev_to_vp(self, vpp);
759 759 return (error);
760 760 }
761 761
762 762 static int
763 763 sdev_remove(struct vnode *dvp, char *nm, struct cred *cred,
764 764 caller_context_t *ct, int flags)
765 765 {
766 766 int error;
767 767 struct sdev_node *parent = (struct sdev_node *)VTOSDEV(dvp);
768 768 struct vnode *vp = NULL;
769 769 struct sdev_node *dv = NULL;
770 770 int len;
771 771 int bkstore;
772 772
773 773 /* bail out early */
774 774 len = strlen(nm);
775 775 if (nm[0] == '.') {
776 776 if (len == 1) {
777 777 return (EINVAL);
778 778 } else if (len == 2 && nm[1] == '.') {
779 779 return (EEXIST);
780 780 }
781 781 }
782 782
783 783 ASSERT(parent);
784 784 rw_enter(&parent->sdev_contents, RW_READER);
785 785 if (!SDEV_IS_GLOBAL(parent)) {
786 786 rw_exit(&parent->sdev_contents);
787 787 return (ENOTSUP);
788 788 }
789 789
790 790 /* execute access is required to search the directory */
791 791 if ((error = sdev_self_access(parent, VEXEC, 0, cred, ct)) != 0) {
792 792 rw_exit(&parent->sdev_contents);
793 793 return (error);
794 794 }
795 795
796 796 /* check existence first */
797 797 dv = sdev_cache_lookup(parent, nm);
798 798 if (dv == NULL) {
799 799 rw_exit(&parent->sdev_contents);
800 800 return (ENOENT);
801 801 }
802 802
803 803 vp = SDEVTOV(dv);
804 804 if ((dv->sdev_state == SDEV_INIT) ||
805 805 (dv->sdev_state == SDEV_ZOMBIE)) {
806 806 rw_exit(&parent->sdev_contents);
807 807 VN_RELE(vp);
808 808 return (ENOENT);
809 809 }
810 810
811 811 /* write access is required to remove an entry */
812 812 if ((error = sdev_self_access(parent, VWRITE, 0, cred, ct)) != 0) {
813 813 rw_exit(&parent->sdev_contents);
814 814 VN_RELE(vp);
815 815 return (error);
816 816 }
817 817
818 818 bkstore = SDEV_IS_PERSIST(dv) ? 1 : 0;
819 819 if (!rw_tryupgrade(&parent->sdev_contents)) {
820 820 rw_exit(&parent->sdev_contents);
821 821 rw_enter(&parent->sdev_contents, RW_WRITER);
822 822 /* Make sure we didn't become a zombie */
823 823 if (parent->sdev_state == SDEV_ZOMBIE) {
824 824 rw_exit(&parent->sdev_contents);
825 825 VN_RELE(vp);
826 826 return (ENOENT);
827 827 }
828 828 }
829 829
830 830 /* we do not support unlinking a non-empty directory */
831 831 if (vp->v_type == VDIR && dv->sdev_nlink > 2) {
832 832 rw_exit(&parent->sdev_contents);
833 833 VN_RELE(vp);
834 834 return (EBUSY);
835 835 }
836 836
837 837 /*
838 838 * sdev_dirdelete does the real job of:
839 839 * - make sure no open ref count
840 840 * - destroying the sdev_node
841 841 * - releasing the hold on attrvp
842 842 */
843 843 sdev_cache_update(parent, &dv, nm, SDEV_CACHE_DELETE);
844 844 VN_RELE(vp);
845 845 rw_exit(&parent->sdev_contents);
846 846
847 847 /*
848 848 * best efforts clean up the backing store
849 849 */
850 850 if (bkstore) {
851 851 ASSERT(parent->sdev_attrvp);
852 852 error = VOP_REMOVE(parent->sdev_attrvp, nm, cred,
853 853 ct, flags);
854 854 /*
855 855 * do not report BUSY error
856 856 * because the backing store ref count is released
|
↓ open down ↓ |
821 lines elided |
↑ open up ↑ |
857 857 * when the last ref count on the sdev_node is
858 858 * released.
859 859 */
860 860 if (error == EBUSY) {
861 861 sdcmn_err2(("sdev_remove: device %s is still on"
862 862 "disk %s\n", nm, parent->sdev_path));
863 863 error = 0;
864 864 }
865 865 }
866 866
867 + if (error == 0)
868 + i_ddi_di_cache_invalidate();
869 +
867 870 return (error);
868 871 }
869 872
870 873 /*
871 874 * Some restrictions for this file system:
872 875 * - both oldnm and newnm are in the scope of /dev file system,
873 876 * to simply the namespace management model.
874 877 */
875 878 /*ARGSUSED6*/
876 879 static int
877 880 sdev_rename(struct vnode *odvp, char *onm, struct vnode *ndvp, char *nnm,
878 881 struct cred *cred, caller_context_t *ct, int flags)
879 882 {
880 883 struct sdev_node *fromparent = NULL;
881 884 struct vattr vattr;
882 885 struct sdev_node *toparent;
883 886 struct sdev_node *fromdv = NULL; /* source node */
884 887 struct vnode *ovp = NULL; /* source vnode */
885 888 struct sdev_node *todv = NULL; /* destination node */
886 889 struct vnode *nvp = NULL; /* destination vnode */
887 890 int samedir = 0; /* set if odvp == ndvp */
888 891 struct vnode *realvp;
889 892 int error = 0;
890 893 dev_t fsid;
891 894 int bkstore = 0;
892 895 vtype_t type;
893 896
894 897 /* prevent modifying "." and ".." */
895 898 if ((onm[0] == '.' &&
896 899 (onm[1] == '\0' || (onm[1] == '.' && onm[2] == '\0'))) ||
897 900 (nnm[0] == '.' &&
898 901 (nnm[1] == '\0' || (nnm[1] == '.' && nnm[2] == '\0')))) {
899 902 return (EINVAL);
900 903 }
901 904
902 905 fromparent = VTOSDEV(odvp);
903 906 toparent = VTOSDEV(ndvp);
904 907
905 908 /* ZOMBIE parent doesn't allow new node creation */
906 909 rw_enter(&fromparent->sdev_dotdot->sdev_contents, RW_READER);
907 910 if (fromparent->sdev_state == SDEV_ZOMBIE) {
908 911 rw_exit(&fromparent->sdev_dotdot->sdev_contents);
909 912 return (ENOENT);
910 913 }
911 914
912 915 /* renaming only supported for global device nodes */
913 916 if (!SDEV_IS_GLOBAL(fromparent)) {
914 917 rw_exit(&fromparent->sdev_dotdot->sdev_contents);
915 918 return (ENOTSUP);
916 919 }
917 920 rw_exit(&fromparent->sdev_dotdot->sdev_contents);
918 921
919 922 rw_enter(&toparent->sdev_dotdot->sdev_contents, RW_READER);
920 923 if (toparent->sdev_state == SDEV_ZOMBIE) {
921 924 rw_exit(&toparent->sdev_dotdot->sdev_contents);
922 925 return (ENOENT);
923 926 }
924 927 rw_exit(&toparent->sdev_dotdot->sdev_contents);
925 928
926 929 /*
927 930 * acquire the global lock to prevent
928 931 * mount/unmount/other rename activities.
929 932 */
930 933 mutex_enter(&sdev_lock);
931 934
932 935 /* check existence of the source node */
933 936 /* XXXci - We may need to translate the C-I flags on VOP_LOOKUP */
934 937 error = VOP_LOOKUP(odvp, onm, &ovp, NULL, 0, NULL, cred, ct,
935 938 NULL, NULL);
936 939 if (error) {
937 940 sdcmn_err2(("sdev_rename: the source node %s exists\n",
938 941 onm));
939 942 mutex_exit(&sdev_lock);
940 943 return (error);
941 944 }
942 945
943 946 if (VOP_REALVP(ovp, &realvp, ct) == 0) {
944 947 VN_HOLD(realvp);
945 948 VN_RELE(ovp);
946 949 ovp = realvp;
947 950 }
948 951
949 952 /* check existence of destination */
950 953 /* XXXci - We may need to translate the C-I flags on VOP_LOOKUP */
951 954 error = VOP_LOOKUP(ndvp, nnm, &nvp, NULL, 0, NULL, cred, ct,
952 955 NULL, NULL);
953 956 if (error && (error != ENOENT)) {
954 957 mutex_exit(&sdev_lock);
955 958 VN_RELE(ovp);
956 959 return (error);
957 960 }
958 961
959 962 if (nvp && (VOP_REALVP(nvp, &realvp, ct) == 0)) {
960 963 VN_HOLD(realvp);
961 964 VN_RELE(nvp);
962 965 nvp = realvp;
963 966 }
964 967
965 968 /*
966 969 * make sure the source and the destination are
967 970 * in the same dev filesystem
968 971 */
969 972 if (odvp != ndvp) {
970 973 vattr.va_mask = AT_FSID;
971 974 if (error = VOP_GETATTR(odvp, &vattr, 0, cred, ct)) {
972 975 mutex_exit(&sdev_lock);
973 976 VN_RELE(ovp);
974 977 if (nvp != NULL)
975 978 VN_RELE(nvp);
976 979 return (error);
977 980 }
978 981 fsid = vattr.va_fsid;
979 982 vattr.va_mask = AT_FSID;
980 983 if (error = VOP_GETATTR(ndvp, &vattr, 0, cred, ct)) {
981 984 mutex_exit(&sdev_lock);
982 985 VN_RELE(ovp);
983 986 if (nvp != NULL)
984 987 VN_RELE(nvp);
985 988 return (error);
986 989 }
987 990 if (fsid != vattr.va_fsid) {
988 991 mutex_exit(&sdev_lock);
989 992 VN_RELE(ovp);
990 993 if (nvp != NULL)
991 994 VN_RELE(nvp);
992 995 return (EXDEV);
993 996 }
994 997 }
995 998
996 999 /* make sure the old entry can be deleted */
997 1000 error = VOP_ACCESS(odvp, VWRITE, 0, cred, ct);
998 1001 if (error) {
999 1002 mutex_exit(&sdev_lock);
1000 1003 VN_RELE(ovp);
1001 1004 if (nvp != NULL)
1002 1005 VN_RELE(nvp);
1003 1006 return (error);
1004 1007 }
1005 1008
1006 1009 /* make sure the destination allows creation */
1007 1010 samedir = (fromparent == toparent);
1008 1011 if (!samedir) {
1009 1012 error = VOP_ACCESS(ndvp, VEXEC|VWRITE, 0, cred, ct);
1010 1013 if (error) {
1011 1014 mutex_exit(&sdev_lock);
1012 1015 VN_RELE(ovp);
1013 1016 if (nvp != NULL)
1014 1017 VN_RELE(nvp);
1015 1018 return (error);
1016 1019 }
1017 1020 }
1018 1021
1019 1022 fromdv = VTOSDEV(ovp);
1020 1023 ASSERT(fromdv);
1021 1024
1022 1025 /* destination file exists */
1023 1026 if (nvp != NULL) {
1024 1027 todv = VTOSDEV(nvp);
1025 1028 ASSERT(todv);
1026 1029 }
1027 1030
1028 1031 if ((fromdv->sdev_flags & SDEV_DYNAMIC) != 0 ||
1029 1032 (todv != NULL && (todv->sdev_flags & SDEV_DYNAMIC) != 0)) {
1030 1033 mutex_exit(&sdev_lock);
1031 1034 if (nvp != NULL)
1032 1035 VN_RELE(nvp);
1033 1036 VN_RELE(ovp);
1034 1037 return (EACCES);
1035 1038 }
1036 1039
1037 1040 /*
1038 1041 * link source to new target in the memory. Regardless of failure, we
1039 1042 * must rele our hold on nvp.
1040 1043 */
1041 1044 error = sdev_rnmnode(fromparent, fromdv, toparent, &todv, nnm, cred);
1042 1045 if (nvp != NULL)
1043 1046 VN_RELE(nvp);
1044 1047 if (error) {
1045 1048 sdcmn_err2(("sdev_rename: renaming %s to %s failed "
1046 1049 " with error %d\n", onm, nnm, error));
1047 1050 mutex_exit(&sdev_lock);
1048 1051 VN_RELE(ovp);
1049 1052 return (error);
1050 1053 }
1051 1054
1052 1055 /*
1053 1056 * unlink from source
1054 1057 */
1055 1058 rw_enter(&fromparent->sdev_contents, RW_READER);
1056 1059 fromdv = sdev_cache_lookup(fromparent, onm);
1057 1060 if (fromdv == NULL) {
1058 1061 rw_exit(&fromparent->sdev_contents);
1059 1062 mutex_exit(&sdev_lock);
1060 1063 VN_RELE(ovp);
1061 1064 sdcmn_err2(("sdev_rename: the source is deleted already\n"));
1062 1065 return (0);
1063 1066 }
1064 1067
1065 1068 if (fromdv->sdev_state == SDEV_ZOMBIE) {
1066 1069 rw_exit(&fromparent->sdev_contents);
1067 1070 mutex_exit(&sdev_lock);
1068 1071 VN_RELE(SDEVTOV(fromdv));
1069 1072 VN_RELE(ovp);
1070 1073 sdcmn_err2(("sdev_rename: the source is being deleted\n"));
1071 1074 return (0);
1072 1075 }
1073 1076 rw_exit(&fromparent->sdev_contents);
1074 1077 ASSERT(SDEVTOV(fromdv) == ovp);
1075 1078 VN_RELE(ovp);
1076 1079
1077 1080 /* clean out the directory contents before it can be removed */
1078 1081 type = SDEVTOV(fromdv)->v_type;
1079 1082 if (type == VDIR) {
1080 1083 error = sdev_cleandir(fromdv, NULL, 0);
1081 1084 sdcmn_err2(("sdev_rename: cleandir finished with %d\n",
1082 1085 error));
1083 1086 if (error == EBUSY)
1084 1087 error = 0;
1085 1088 }
1086 1089
1087 1090 rw_enter(&fromparent->sdev_contents, RW_WRITER);
1088 1091 bkstore = SDEV_IS_PERSIST(fromdv) ? 1 : 0;
1089 1092 sdev_cache_update(fromparent, &fromdv, onm,
1090 1093 SDEV_CACHE_DELETE);
1091 1094 VN_RELE(SDEVTOV(fromdv));
1092 1095
1093 1096 /* best effforts clean up the backing store */
1094 1097 if (bkstore) {
1095 1098 ASSERT(fromparent->sdev_attrvp);
1096 1099 if (type != VDIR) {
1097 1100 /* XXXci - We may need to translate the C-I flags on VOP_REMOVE */
1098 1101 error = VOP_REMOVE(fromparent->sdev_attrvp,
1099 1102 onm, kcred, ct, 0);
1100 1103 } else {
1101 1104 /* XXXci - We may need to translate the C-I flags on VOP_RMDIR */
1102 1105 error = VOP_RMDIR(fromparent->sdev_attrvp,
1103 1106 onm, fromparent->sdev_attrvp, kcred, ct, 0);
1104 1107 }
1105 1108
1106 1109 if (error) {
1107 1110 sdcmn_err2(("sdev_rename: device %s is "
1108 1111 "still on disk %s\n", onm,
1109 1112 fromparent->sdev_path));
1110 1113 error = 0;
1111 1114 }
1112 1115 }
1113 1116 rw_exit(&fromparent->sdev_contents);
1114 1117 mutex_exit(&sdev_lock);
1115 1118
1116 1119 /* once reached to this point, the rename is regarded successful */
1117 1120 return (0);
1118 1121 }
1119 1122
1120 1123 /*
1121 1124 * dev-fs version of "ln -s path dev-name"
1122 1125 * tnm - path, e.g. /devices/... or /dev/...
1123 1126 * lnm - dev_name
1124 1127 */
1125 1128 /*ARGSUSED6*/
1126 1129 static int
1127 1130 sdev_symlink(struct vnode *dvp, char *lnm, struct vattr *tva,
1128 1131 char *tnm, struct cred *cred, caller_context_t *ct, int flags)
1129 1132 {
1130 1133 int error;
1131 1134 struct vnode *vp = NULL;
1132 1135 struct sdev_node *parent = (struct sdev_node *)VTOSDEV(dvp);
1133 1136 struct sdev_node *self = (struct sdev_node *)NULL;
1134 1137
1135 1138 ASSERT(parent);
1136 1139 rw_enter(&parent->sdev_dotdot->sdev_contents, RW_READER);
1137 1140 if (parent->sdev_state == SDEV_ZOMBIE) {
1138 1141 rw_exit(&parent->sdev_dotdot->sdev_contents);
1139 1142 sdcmn_err2(("sdev_symlink: parent %s is ZOMBIED \n",
1140 1143 parent->sdev_name));
1141 1144 return (ENOENT);
1142 1145 }
1143 1146
1144 1147 if (!SDEV_IS_GLOBAL(parent)) {
1145 1148 rw_exit(&parent->sdev_dotdot->sdev_contents);
1146 1149 return (ENOTSUP);
1147 1150 }
1148 1151 rw_exit(&parent->sdev_dotdot->sdev_contents);
1149 1152
1150 1153 /* execute access is required to search a directory */
1151 1154 if ((error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct)) != 0)
1152 1155 return (error);
1153 1156
1154 1157 /* find existing name */
1155 1158 /* XXXci - We may need to translate the C-I flags here */
1156 1159 error = VOP_LOOKUP(dvp, lnm, &vp, NULL, 0, NULL, cred, ct, NULL, NULL);
1157 1160 if (error == 0) {
1158 1161 ASSERT(vp);
1159 1162 VN_RELE(vp);
1160 1163 sdcmn_err2(("sdev_symlink: node %s already exists\n", lnm));
1161 1164 return (EEXIST);
1162 1165 }
1163 1166 if (error != ENOENT)
1164 1167 return (error);
1165 1168
1166 1169 /* write access is required to create a symlink */
1167 1170 if ((error = VOP_ACCESS(dvp, VWRITE, 0, cred, ct)) != 0)
1168 1171 return (error);
1169 1172
1170 1173 /* put it into memory cache */
1171 1174 rw_enter(&parent->sdev_contents, RW_WRITER);
1172 1175 error = sdev_mknode(parent, lnm, &self, tva, NULL, (void *)tnm,
1173 1176 cred, SDEV_READY);
1174 1177 if (error) {
1175 1178 rw_exit(&parent->sdev_contents);
1176 1179 sdcmn_err2(("sdev_symlink: node %s creation failed\n", lnm));
1177 1180 if (self)
1178 1181 SDEV_RELE(self);
1179 1182
1180 1183 return (error);
|
↓ open down ↓ |
304 lines elided |
↑ open up ↑ |
1181 1184 }
1182 1185 ASSERT(self && (self->sdev_state == SDEV_READY));
1183 1186 rw_exit(&parent->sdev_contents);
1184 1187
1185 1188 /* take care the timestamps for the node and its parent */
1186 1189 sdev_update_timestamps(SDEVTOV(self), kcred,
1187 1190 AT_CTIME|AT_MTIME|AT_ATIME);
1188 1191 sdev_update_timestamps(dvp, kcred, AT_MTIME|AT_ATIME);
1189 1192 if (SDEV_IS_GLOBAL(parent))
1190 1193 atomic_inc_ulong(&parent->sdev_gdir_gen);
1194 + i_ddi_di_cache_invalidate();
1191 1195
1192 1196 /* wake up other threads blocked on looking up this node */
1193 1197 mutex_enter(&self->sdev_lookup_lock);
1194 1198 SDEV_UNBLOCK_OTHERS(self, SDEV_LOOKUP);
1195 1199 mutex_exit(&self->sdev_lookup_lock);
1196 1200 SDEV_RELE(self); /* don't return with vnode held */
1197 1201 return (0);
1198 1202 }
1199 1203
1200 1204 /*ARGSUSED6*/
1201 1205 static int
1202 1206 sdev_mkdir(struct vnode *dvp, char *nm, struct vattr *va, struct vnode **vpp,
1203 1207 struct cred *cred, caller_context_t *ct, int flags, vsecattr_t *vsecp)
1204 1208 {
1205 1209 int error;
1206 1210 struct sdev_node *parent = (struct sdev_node *)VTOSDEV(dvp);
1207 1211 struct sdev_node *self = NULL;
1208 1212 struct vnode *vp = NULL;
1209 1213
1210 1214 ASSERT(parent && parent->sdev_dotdot);
1211 1215 rw_enter(&parent->sdev_dotdot->sdev_contents, RW_READER);
1212 1216 if (parent->sdev_state == SDEV_ZOMBIE) {
1213 1217 rw_exit(&parent->sdev_dotdot->sdev_contents);
1214 1218 return (ENOENT);
1215 1219 }
1216 1220
1217 1221 /* non-global do not allow pure directory creation */
1218 1222 if (!SDEV_IS_GLOBAL(parent)) {
1219 1223 rw_exit(&parent->sdev_dotdot->sdev_contents);
1220 1224 return (prof_lookup(dvp, nm, vpp, cred));
1221 1225 }
1222 1226 rw_exit(&parent->sdev_dotdot->sdev_contents);
1223 1227
1224 1228 /* execute access is required to search the directory */
1225 1229 if ((error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct)) != 0) {
1226 1230 return (error);
1227 1231 }
1228 1232
1229 1233 /* find existing name */
1230 1234 /* XXXci - We may need to translate the C-I flags on VOP_LOOKUP */
1231 1235 error = VOP_LOOKUP(dvp, nm, &vp, NULL, 0, NULL, cred, ct, NULL, NULL);
1232 1236 if (error == 0) {
1233 1237 VN_RELE(vp);
1234 1238 return (EEXIST);
1235 1239 }
1236 1240 if (error != ENOENT)
1237 1241 return (error);
1238 1242
1239 1243 /* require write access to create a directory */
1240 1244 if ((error = VOP_ACCESS(dvp, VWRITE, 0, cred, ct)) != 0) {
1241 1245 return (error);
1242 1246 }
1243 1247
1244 1248 /* put it into memory */
1245 1249 rw_enter(&parent->sdev_contents, RW_WRITER);
1246 1250 error = sdev_mknode(parent, nm, &self,
1247 1251 va, NULL, NULL, cred, SDEV_READY);
1248 1252 if (error) {
1249 1253 rw_exit(&parent->sdev_contents);
1250 1254 if (self)
1251 1255 SDEV_RELE(self);
1252 1256 return (error);
|
↓ open down ↓ |
52 lines elided |
↑ open up ↑ |
1253 1257 }
1254 1258 ASSERT(self && (self->sdev_state == SDEV_READY));
1255 1259 rw_exit(&parent->sdev_contents);
1256 1260
1257 1261 /* take care the timestamps for the node and its parent */
1258 1262 sdev_update_timestamps(SDEVTOV(self), kcred,
1259 1263 AT_CTIME|AT_MTIME|AT_ATIME);
1260 1264 sdev_update_timestamps(dvp, kcred, AT_MTIME|AT_ATIME);
1261 1265 if (SDEV_IS_GLOBAL(parent))
1262 1266 atomic_inc_ulong(&parent->sdev_gdir_gen);
1267 + i_ddi_di_cache_invalidate();
1263 1268
1264 1269 /* wake up other threads blocked on looking up this node */
1265 1270 mutex_enter(&self->sdev_lookup_lock);
1266 1271 SDEV_UNBLOCK_OTHERS(self, SDEV_LOOKUP);
1267 1272 mutex_exit(&self->sdev_lookup_lock);
1268 1273 *vpp = SDEVTOV(self);
1269 1274 return (0);
1270 1275 }
1271 1276
1272 1277 /*
1273 1278 * allowing removing an empty directory under /dev
1274 1279 */
1275 1280 /*ARGSUSED*/
1276 1281 static int
1277 1282 sdev_rmdir(struct vnode *dvp, char *nm, struct vnode *cdir, struct cred *cred,
1278 1283 caller_context_t *ct, int flags)
1279 1284 {
1280 1285 int error = 0;
1281 1286 struct sdev_node *parent = (struct sdev_node *)VTOSDEV(dvp);
1282 1287 struct sdev_node *self = NULL;
1283 1288 struct vnode *vp = NULL;
1284 1289
1285 1290 /* bail out early */
1286 1291 if (strcmp(nm, ".") == 0)
1287 1292 return (EINVAL);
1288 1293 if (strcmp(nm, "..") == 0)
1289 1294 return (EEXIST); /* should be ENOTEMPTY */
1290 1295
1291 1296 /* no destruction of non-global node */
1292 1297 ASSERT(parent && parent->sdev_dotdot);
1293 1298 rw_enter(&parent->sdev_dotdot->sdev_contents, RW_READER);
1294 1299 if (!SDEV_IS_GLOBAL(parent)) {
1295 1300 rw_exit(&parent->sdev_dotdot->sdev_contents);
1296 1301 return (ENOTSUP);
1297 1302 }
1298 1303 rw_exit(&parent->sdev_dotdot->sdev_contents);
1299 1304
1300 1305 /* execute access is required to search the directory */
1301 1306 if ((error = VOP_ACCESS(dvp, VEXEC|VWRITE, 0, cred, ct)) != 0)
1302 1307 return (error);
1303 1308
1304 1309 /* check existing name */
1305 1310 rw_enter(&parent->sdev_contents, RW_WRITER);
1306 1311 self = sdev_cache_lookup(parent, nm);
1307 1312 if (self == NULL) {
1308 1313 rw_exit(&parent->sdev_contents);
1309 1314 return (ENOENT);
1310 1315 }
1311 1316
1312 1317 vp = SDEVTOV(self);
1313 1318 if ((self->sdev_state == SDEV_INIT) ||
1314 1319 (self->sdev_state == SDEV_ZOMBIE)) {
1315 1320 rw_exit(&parent->sdev_contents);
1316 1321 VN_RELE(vp);
1317 1322 return (ENOENT);
1318 1323 }
1319 1324
1320 1325 /* some sanity checks */
1321 1326 if (vp == dvp || vp == cdir) {
1322 1327 rw_exit(&parent->sdev_contents);
1323 1328 VN_RELE(vp);
1324 1329 return (EINVAL);
1325 1330 }
1326 1331
1327 1332 if (vp->v_type != VDIR) {
1328 1333 rw_exit(&parent->sdev_contents);
1329 1334 VN_RELE(vp);
1330 1335 return (ENOTDIR);
1331 1336 }
1332 1337
1333 1338 if (vn_vfswlock(vp)) {
1334 1339 rw_exit(&parent->sdev_contents);
1335 1340 VN_RELE(vp);
1336 1341 return (EBUSY);
1337 1342 }
1338 1343
1339 1344 if (vn_mountedvfs(vp) != NULL) {
1340 1345 rw_exit(&parent->sdev_contents);
1341 1346 vn_vfsunlock(vp);
1342 1347 VN_RELE(vp);
1343 1348 return (EBUSY);
1344 1349 }
1345 1350
1346 1351 self = VTOSDEV(vp);
1347 1352 /* bail out on a non-empty directory */
1348 1353 rw_enter(&self->sdev_contents, RW_READER);
1349 1354 if (self->sdev_nlink > 2) {
1350 1355 rw_exit(&self->sdev_contents);
1351 1356 rw_exit(&parent->sdev_contents);
1352 1357 vn_vfsunlock(vp);
1353 1358 VN_RELE(vp);
1354 1359 return (ENOTEMPTY);
1355 1360 }
1356 1361 rw_exit(&self->sdev_contents);
1357 1362
1358 1363 /* unlink it from the directory cache */
1359 1364 sdev_cache_update(parent, &self, nm, SDEV_CACHE_DELETE);
1360 1365 rw_exit(&parent->sdev_contents);
1361 1366 vn_vfsunlock(vp);
1362 1367 VN_RELE(vp);
1363 1368
1364 1369 /* best effort to clean up the backing store */
1365 1370 if (SDEV_IS_PERSIST(parent)) {
1366 1371 ASSERT(parent->sdev_attrvp);
1367 1372 error = VOP_RMDIR(parent->sdev_attrvp, nm,
|
↓ open down ↓ |
95 lines elided |
↑ open up ↑ |
1368 1373 parent->sdev_attrvp, kcred, ct, flags);
1369 1374
1370 1375 if (error)
1371 1376 sdcmn_err2(("sdev_rmdir: cleaning device %s is on"
1372 1377 " disk error %d\n", parent->sdev_path, error));
1373 1378 if (error == EBUSY)
1374 1379 error = 0;
1375 1380
1376 1381 }
1377 1382
1383 + if (error == 0)
1384 + i_ddi_di_cache_invalidate();
1385 +
1378 1386 return (error);
1379 1387 }
1380 1388
1381 1389 /*
1382 1390 * read the contents of a symbolic link
1383 1391 */
1384 1392 static int
1385 1393 sdev_readlink(struct vnode *vp, struct uio *uiop, struct cred *cred,
1386 1394 caller_context_t *ct)
1387 1395 {
1388 1396 struct sdev_node *dv;
1389 1397 int error = 0;
1390 1398
1391 1399 ASSERT(vp->v_type == VLNK);
1392 1400
1393 1401 dv = VTOSDEV(vp);
1394 1402
1395 1403 if (dv->sdev_attrvp) {
1396 1404 /* non-NULL attrvp implys a persisted node at READY state */
1397 1405 return (VOP_READLINK(dv->sdev_attrvp, uiop, cred, ct));
1398 1406 } else if (dv->sdev_symlink != NULL) {
1399 1407 /* memory nodes, e.g. local nodes */
1400 1408 rw_enter(&dv->sdev_contents, RW_READER);
1401 1409 sdcmn_err2(("sdev_readlink link is %s\n", dv->sdev_symlink));
1402 1410 error = uiomove(dv->sdev_symlink, strlen(dv->sdev_symlink),
1403 1411 UIO_READ, uiop);
1404 1412 rw_exit(&dv->sdev_contents);
1405 1413 return (error);
1406 1414 }
1407 1415
1408 1416 return (ENOENT);
1409 1417 }
1410 1418
1411 1419 /*ARGSUSED4*/
1412 1420 static int
1413 1421 sdev_readdir(struct vnode *dvp, struct uio *uiop, struct cred *cred, int *eofp,
1414 1422 caller_context_t *ct, int flags)
1415 1423 {
1416 1424 struct sdev_node *parent = VTOSDEV(dvp);
1417 1425 int error;
1418 1426
1419 1427 /*
1420 1428 * We must check that we have execute access to search the directory --
1421 1429 * but because our sdev_contents lock is already held as a reader (the
1422 1430 * caller must have done a VOP_RWLOCK()), we call directly into the
1423 1431 * underlying access routine if sdev_attr is non-NULL.
1424 1432 */
1425 1433 if (parent->sdev_attr != NULL) {
1426 1434 VERIFY(RW_READ_HELD(&parent->sdev_contents));
1427 1435
1428 1436 if (sdev_unlocked_access(parent, VEXEC, cred) != 0)
1429 1437 return (EACCES);
1430 1438 } else {
1431 1439 if ((error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct)) != 0)
1432 1440 return (error);
1433 1441 }
1434 1442
1435 1443 ASSERT(parent);
1436 1444 if (!SDEV_IS_GLOBAL(parent))
1437 1445 prof_filldir(parent);
1438 1446 return (devname_readdir_func(dvp, uiop, cred, eofp, SDEV_BROWSE));
1439 1447 }
1440 1448
1441 1449 /*ARGSUSED1*/
1442 1450 static void
1443 1451 sdev_inactive(struct vnode *vp, struct cred *cred, caller_context_t *ct)
1444 1452 {
1445 1453 devname_inactive_func(vp, cred, NULL);
1446 1454 }
1447 1455
1448 1456 /*ARGSUSED2*/
1449 1457 static int
1450 1458 sdev_fid(struct vnode *vp, struct fid *fidp, caller_context_t *ct)
1451 1459 {
1452 1460 struct sdev_node *dv = VTOSDEV(vp);
1453 1461 struct sdev_fid *sdev_fid;
1454 1462
1455 1463 if (fidp->fid_len < (sizeof (struct sdev_fid) - sizeof (ushort_t))) {
1456 1464 fidp->fid_len = sizeof (struct sdev_fid) - sizeof (ushort_t);
1457 1465 return (ENOSPC);
1458 1466 }
1459 1467
1460 1468 sdev_fid = (struct sdev_fid *)fidp;
1461 1469 bzero(sdev_fid, sizeof (struct sdev_fid));
1462 1470 sdev_fid->sdevfid_len =
1463 1471 (int)sizeof (struct sdev_fid) - sizeof (ushort_t);
1464 1472 sdev_fid->sdevfid_ino = dv->sdev_ino;
1465 1473
1466 1474 return (0);
1467 1475 }
1468 1476
1469 1477 /*
1470 1478 * This pair of routines bracket all VOP_READ, VOP_WRITE
1471 1479 * and VOP_READDIR requests. The contents lock stops things
1472 1480 * moving around while we're looking at them.
1473 1481 */
1474 1482 /*ARGSUSED2*/
1475 1483 static int
1476 1484 sdev_rwlock(struct vnode *vp, int write_flag, caller_context_t *ctp)
1477 1485 {
1478 1486 rw_enter(&VTOSDEV(vp)->sdev_contents,
1479 1487 write_flag ? RW_WRITER : RW_READER);
1480 1488 return (write_flag ? V_WRITELOCK_TRUE : V_WRITELOCK_FALSE);
1481 1489 }
1482 1490
1483 1491 /*ARGSUSED1*/
1484 1492 static void
1485 1493 sdev_rwunlock(struct vnode *vp, int write_flag, caller_context_t *ctp)
1486 1494 {
1487 1495 rw_exit(&VTOSDEV(vp)->sdev_contents);
1488 1496 }
1489 1497
1490 1498 /*ARGSUSED1*/
1491 1499 static int
1492 1500 sdev_seek(struct vnode *vp, offset_t ooff, offset_t *noffp,
1493 1501 caller_context_t *ct)
1494 1502 {
1495 1503 struct vnode *attrvp = VTOSDEV(vp)->sdev_attrvp;
1496 1504
1497 1505 ASSERT(vp->v_type != VCHR &&
1498 1506 vp->v_type != VBLK && vp->v_type != VLNK);
1499 1507
1500 1508 if (vp->v_type == VDIR)
1501 1509 return (fs_seek(vp, ooff, noffp, ct));
1502 1510
1503 1511 ASSERT(attrvp);
1504 1512 return (VOP_SEEK(attrvp, ooff, noffp, ct));
1505 1513 }
1506 1514
1507 1515 /*ARGSUSED1*/
1508 1516 static int
1509 1517 sdev_frlock(struct vnode *vp, int cmd, struct flock64 *bfp, int flag,
1510 1518 offset_t offset, struct flk_callback *flk_cbp, struct cred *cr,
1511 1519 caller_context_t *ct)
1512 1520 {
1513 1521 int error;
1514 1522 struct sdev_node *dv = VTOSDEV(vp);
1515 1523
1516 1524 ASSERT(dv);
1517 1525 ASSERT(dv->sdev_attrvp);
1518 1526 error = VOP_FRLOCK(dv->sdev_attrvp, cmd, bfp, flag, offset,
1519 1527 flk_cbp, cr, ct);
1520 1528
1521 1529 return (error);
1522 1530 }
1523 1531
1524 1532 static int
1525 1533 sdev_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
1526 1534 caller_context_t *ct)
1527 1535 {
1528 1536 switch (cmd) {
1529 1537 case _PC_ACL_ENABLED:
1530 1538 *valp = SDEV_ACL_FLAVOR(vp);
1531 1539 return (0);
1532 1540 }
1533 1541
1534 1542 return (fs_pathconf(vp, cmd, valp, cr, ct));
1535 1543 }
1536 1544
1537 1545 vnodeops_t *sdev_vnodeops;
1538 1546
1539 1547 const fs_operation_def_t sdev_vnodeops_tbl[] = {
1540 1548 VOPNAME_OPEN, { .vop_open = sdev_open },
1541 1549 VOPNAME_CLOSE, { .vop_close = sdev_close },
1542 1550 VOPNAME_READ, { .vop_read = sdev_read },
1543 1551 VOPNAME_WRITE, { .vop_write = sdev_write },
1544 1552 VOPNAME_IOCTL, { .vop_ioctl = sdev_ioctl },
1545 1553 VOPNAME_GETATTR, { .vop_getattr = sdev_getattr },
1546 1554 VOPNAME_SETATTR, { .vop_setattr = sdev_setattr },
1547 1555 VOPNAME_ACCESS, { .vop_access = sdev_access },
1548 1556 VOPNAME_LOOKUP, { .vop_lookup = sdev_lookup },
1549 1557 VOPNAME_CREATE, { .vop_create = sdev_create },
1550 1558 VOPNAME_RENAME, { .vop_rename = sdev_rename },
1551 1559 VOPNAME_REMOVE, { .vop_remove = sdev_remove },
1552 1560 VOPNAME_MKDIR, { .vop_mkdir = sdev_mkdir },
1553 1561 VOPNAME_RMDIR, { .vop_rmdir = sdev_rmdir },
1554 1562 VOPNAME_READDIR, { .vop_readdir = sdev_readdir },
1555 1563 VOPNAME_SYMLINK, { .vop_symlink = sdev_symlink },
1556 1564 VOPNAME_READLINK, { .vop_readlink = sdev_readlink },
1557 1565 VOPNAME_INACTIVE, { .vop_inactive = sdev_inactive },
1558 1566 VOPNAME_FID, { .vop_fid = sdev_fid },
1559 1567 VOPNAME_RWLOCK, { .vop_rwlock = sdev_rwlock },
1560 1568 VOPNAME_RWUNLOCK, { .vop_rwunlock = sdev_rwunlock },
1561 1569 VOPNAME_SEEK, { .vop_seek = sdev_seek },
1562 1570 VOPNAME_FRLOCK, { .vop_frlock = sdev_frlock },
1563 1571 VOPNAME_PATHCONF, { .vop_pathconf = sdev_pathconf },
1564 1572 VOPNAME_SETSECATTR, { .vop_setsecattr = sdev_setsecattr },
1565 1573 VOPNAME_GETSECATTR, { .vop_getsecattr = sdev_getsecattr },
1566 1574 NULL, NULL
1567 1575 };
1568 1576
1569 1577 int sdev_vnodeops_tbl_size = sizeof (sdev_vnodeops_tbl);
|
↓ open down ↓ |
182 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX