1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3 *
4 * Copyright (c) 2010-2012 Semihalf
5 * Copyright (c) 2008, 2009 Reinoud Zandijk
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * From: NetBSD: nilfs_vnops.c,v 1.2 2009/08/26 03:40:48 elad
29 */
30
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD: stable/12/sys/fs/nandfs/nandfs_vnops.c 345972 2019-04-05 23:34:46Z sjg $");
33
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/conf.h>
37 #include <sys/kernel.h>
38 #include <sys/lock.h>
39 #include <sys/lockf.h>
40 #include <sys/malloc.h>
41 #include <sys/mount.h>
42 #include <sys/mutex.h>
43 #include <sys/namei.h>
44 #include <sys/sysctl.h>
45 #include <sys/unistd.h>
46 #include <sys/vnode.h>
47 #include <sys/buf.h>
48 #include <sys/bio.h>
49 #include <sys/fcntl.h>
50 #include <sys/dirent.h>
51 #include <sys/rwlock.h>
52 #include <sys/stat.h>
53 #include <sys/priv.h>
54
55 #include <vm/vm.h>
56 #include <vm/vm_extern.h>
57 #include <vm/vm_object.h>
58 #include <vm/vnode_pager.h>
59
60 #include <machine/_inttypes.h>
61
62 #include <fs/nandfs/nandfs_mount.h>
63 #include <fs/nandfs/nandfs.h>
64 #include <fs/nandfs/nandfs_subr.h>
65
66 extern uma_zone_t nandfs_node_zone;
67 static void nandfs_read_filebuf(struct nandfs_node *, struct buf *);
68 static void nandfs_itimes_locked(struct vnode *);
69 static int nandfs_truncate(struct vnode *, uint64_t);
70
71 static vop_pathconf_t nandfs_pathconf;
72
73 #define UPDATE_CLOSE 0
74 #define UPDATE_WAIT 0
75
76 static int
nandfs_inactive(struct vop_inactive_args * ap)77 nandfs_inactive(struct vop_inactive_args *ap)
78 {
79 struct vnode *vp = ap->a_vp;
80 struct nandfs_node *node = VTON(vp);
81 int error = 0;
82
83 DPRINTF(VNCALL, ("%s: vp:%p node:%p\n", __func__, vp, node));
84
85 if (node == NULL) {
86 DPRINTF(NODE, ("%s: inactive NULL node\n", __func__));
87 return (0);
88 }
89
90 if (node->nn_inode.i_mode != 0 && !(node->nn_inode.i_links_count)) {
91 nandfs_truncate(vp, 0);
92 error = nandfs_node_destroy(node);
93 if (error)
94 nandfs_error("%s: destroy node: %p\n", __func__, node);
95 node->nn_flags = 0;
96 vrecycle(vp);
97 }
98
99 return (error);
100 }
101
102 static int
nandfs_reclaim(struct vop_reclaim_args * ap)103 nandfs_reclaim(struct vop_reclaim_args *ap)
104 {
105 struct vnode *vp = ap->a_vp;
106 struct nandfs_node *nandfs_node = VTON(vp);
107 struct nandfs_device *fsdev = nandfs_node->nn_nandfsdev;
108 uint64_t ino = nandfs_node->nn_ino;
109
110 DPRINTF(VNCALL, ("%s: vp:%p node:%p\n", __func__, vp, nandfs_node));
111
112 /* Invalidate all entries to a particular vnode. */
113 cache_purge(vp);
114
115 /* Destroy the vm object and flush associated pages. */
116 vnode_destroy_vobject(vp);
117
118 /* Remove from vfs hash if not system vnode */
119 if (!NANDFS_SYS_NODE(nandfs_node->nn_ino))
120 vfs_hash_remove(vp);
121
122 /* Dispose all node knowledge */
123 nandfs_dispose_node(&nandfs_node);
124
125 if (!NANDFS_SYS_NODE(ino))
126 NANDFS_WRITEUNLOCK(fsdev);
127
128 return (0);
129 }
130
131 static int
nandfs_read(struct vop_read_args * ap)132 nandfs_read(struct vop_read_args *ap)
133 {
134 struct vnode *vp = ap->a_vp;
135 struct nandfs_node *node = VTON(vp);
136 struct nandfs_device *nandfsdev = node->nn_nandfsdev;
137 struct uio *uio = ap->a_uio;
138 struct buf *bp;
139 uint64_t size;
140 uint32_t blocksize;
141 off_t bytesinfile;
142 ssize_t toread, off;
143 daddr_t lbn;
144 ssize_t resid;
145 int error = 0;
146
147 if (uio->uio_resid == 0)
148 return (0);
149
150 size = node->nn_inode.i_size;
151 if (uio->uio_offset >= size)
152 return (0);
153
154 blocksize = nandfsdev->nd_blocksize;
155 bytesinfile = size - uio->uio_offset;
156
157 resid = omin(uio->uio_resid, bytesinfile);
158
159 while (resid) {
160 lbn = uio->uio_offset / blocksize;
161 off = uio->uio_offset & (blocksize - 1);
162
163 toread = omin(resid, blocksize - off);
164
165 DPRINTF(READ, ("nandfs_read bn: 0x%jx toread: 0x%zx (0x%x)\n",
166 (uintmax_t)lbn, toread, blocksize));
167
168 error = nandfs_bread(node, lbn, NOCRED, 0, &bp);
169 if (error) {
170 brelse(bp);
171 break;
172 }
173
174 error = uiomove(bp->b_data + off, toread, uio);
175 if (error) {
176 brelse(bp);
177 break;
178 }
179
180 brelse(bp);
181 resid -= toread;
182 }
183
184 return (error);
185 }
186
187 static int
nandfs_write(struct vop_write_args * ap)188 nandfs_write(struct vop_write_args *ap)
189 {
190 struct nandfs_device *fsdev;
191 struct nandfs_node *node;
192 struct vnode *vp;
193 struct uio *uio;
194 struct buf *bp;
195 uint64_t file_size, vblk;
196 uint32_t blocksize;
197 ssize_t towrite, off;
198 daddr_t lbn;
199 ssize_t resid;
200 int error, ioflag, modified;
201
202 vp = ap->a_vp;
203 uio = ap->a_uio;
204 ioflag = ap->a_ioflag;
205 node = VTON(vp);
206 fsdev = node->nn_nandfsdev;
207
208 if (nandfs_fs_full(fsdev))
209 return (ENOSPC);
210
211 DPRINTF(WRITE, ("nandfs_write called %#zx at %#jx\n",
212 uio->uio_resid, (uintmax_t)uio->uio_offset));
213
214 if (uio->uio_offset < 0)
215 return (EINVAL);
216 if (uio->uio_resid == 0)
217 return (0);
218
219 blocksize = fsdev->nd_blocksize;
220 file_size = node->nn_inode.i_size;
221
222 switch (vp->v_type) {
223 case VREG:
224 if (ioflag & IO_APPEND)
225 uio->uio_offset = file_size;
226 break;
227 case VDIR:
228 return (EISDIR);
229 case VLNK:
230 break;
231 default:
232 panic("%s: bad file type vp: %p", __func__, vp);
233 }
234
235 /* If explicitly asked to append, uio_offset can be wrong? */
236 if (ioflag & IO_APPEND)
237 uio->uio_offset = file_size;
238
239 resid = uio->uio_resid;
240 modified = error = 0;
241
242 while (uio->uio_resid) {
243 lbn = uio->uio_offset / blocksize;
244 off = uio->uio_offset & (blocksize - 1);
245
246 towrite = omin(uio->uio_resid, blocksize - off);
247
248 DPRINTF(WRITE, ("%s: lbn: 0x%jd toread: 0x%zx (0x%x)\n",
249 __func__, (uintmax_t)lbn, towrite, blocksize));
250
251 error = nandfs_bmap_lookup(node, lbn, &vblk);
252 if (error)
253 break;
254
255 DPRINTF(WRITE, ("%s: lbn: 0x%jd toread: 0x%zx (0x%x) "
256 "vblk=%jx\n", __func__, (uintmax_t)lbn, towrite, blocksize,
257 vblk));
258
259 if (vblk != 0)
260 error = nandfs_bread(node, lbn, NOCRED, 0, &bp);
261 else
262 error = nandfs_bcreate(node, lbn, NOCRED, 0, &bp);
263
264 DPRINTF(WRITE, ("%s: vp %p bread bp %p lbn %#jx\n", __func__,
265 vp, bp, (uintmax_t)lbn));
266 if (error) {
267 if (bp)
268 brelse(bp);
269 break;
270 }
271
272 error = uiomove((char *)bp->b_data + off, (int)towrite, uio);
273 if (error)
274 break;
275
276 error = nandfs_dirty_buf(bp, 0);
277 if (error)
278 break;
279
280 modified++;
281 }
282
283 /* XXX proper handling when only part of file was properly written */
284 if (modified) {
285 if (resid > uio->uio_resid && ap->a_cred &&
286 ap->a_cred->cr_uid != 0)
287 node->nn_inode.i_mode &= ~(ISUID | ISGID);
288
289 if (file_size < uio->uio_offset + uio->uio_resid) {
290 node->nn_inode.i_size = uio->uio_offset +
291 uio->uio_resid;
292 node->nn_flags |= IN_CHANGE | IN_UPDATE;
293 vnode_pager_setsize(vp, uio->uio_offset +
294 uio->uio_resid);
295 nandfs_itimes(vp);
296 }
297 }
298
299 DPRINTF(WRITE, ("%s: return:%d\n", __func__, error));
300
301 return (error);
302 }
303
304 static int
nandfs_lookup(struct vop_cachedlookup_args * ap)305 nandfs_lookup(struct vop_cachedlookup_args *ap)
306 {
307 struct vnode *dvp, **vpp;
308 struct componentname *cnp;
309 struct ucred *cred;
310 struct thread *td;
311 struct nandfs_node *dir_node, *node;
312 struct nandfsmount *nmp;
313 uint64_t ino, off;
314 const char *name;
315 int namelen, nameiop, islastcn, mounted_ro;
316 int error, found;
317
318 DPRINTF(VNCALL, ("%s\n", __func__));
319
320 dvp = ap->a_dvp;
321 vpp = ap->a_vpp;
322 *vpp = NULL;
323
324 cnp = ap->a_cnp;
325 cred = cnp->cn_cred;
326 td = cnp->cn_thread;
327
328 dir_node = VTON(dvp);
329 nmp = dir_node->nn_nmp;
330
331 /* Simplify/clarification flags */
332 nameiop = cnp->cn_nameiop;
333 islastcn = cnp->cn_flags & ISLASTCN;
334 mounted_ro = dvp->v_mount->mnt_flag & MNT_RDONLY;
335
336 /*
337 * If requesting a modify on the last path element on a read-only
338 * filingsystem, reject lookup;
339 */
340 if (islastcn && mounted_ro && (nameiop == DELETE || nameiop == RENAME))
341 return (EROFS);
342
343 if (dir_node->nn_inode.i_links_count == 0)
344 return (ENOENT);
345
346 /*
347 * Obviously, the file is not (anymore) in the namecache, we have to
348 * search for it. There are three basic cases: '.', '..' and others.
349 *
350 * Following the guidelines of VOP_LOOKUP manpage and tmpfs.
351 */
352 error = 0;
353 if ((cnp->cn_namelen == 1) && (cnp->cn_nameptr[0] == '.')) {
354 DPRINTF(LOOKUP, ("\tlookup '.'\n"));
355 /* Special case 1 '.' */
356 VREF(dvp);
357 *vpp = dvp;
358 /* Done */
359 } else if (cnp->cn_flags & ISDOTDOT) {
360 /* Special case 2 '..' */
361 DPRINTF(LOOKUP, ("\tlookup '..'\n"));
362
363 /* Get our node */
364 name = "..";
365 namelen = 2;
366 error = nandfs_lookup_name_in_dir(dvp, name, namelen, &ino,
367 &found, &off);
368 if (error)
369 goto out;
370 if (!found)
371 error = ENOENT;
372
373 /* First unlock parent */
374 VOP_UNLOCK(dvp, 0);
375
376 if (error == 0) {
377 DPRINTF(LOOKUP, ("\tfound '..'\n"));
378 /* Try to create/reuse the node */
379 error = nandfs_get_node(nmp, ino, &node);
380
381 if (!error) {
382 DPRINTF(LOOKUP,
383 ("\tnode retrieved/created OK\n"));
384 *vpp = NTOV(node);
385 }
386 }
387
388 /* Try to relock parent */
389 vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
390 } else {
391 DPRINTF(LOOKUP, ("\tlookup file\n"));
392 /* All other files */
393 /* Look up filename in the directory returning its inode */
394 name = cnp->cn_nameptr;
395 namelen = cnp->cn_namelen;
396 error = nandfs_lookup_name_in_dir(dvp, name, namelen,
397 &ino, &found, &off);
398 if (error)
399 goto out;
400 if (!found) {
401 DPRINTF(LOOKUP, ("\tNOT found\n"));
402 /*
403 * UGH, didn't find name. If we're creating or
404 * renaming on the last name this is OK and we ought
405 * to return EJUSTRETURN if its allowed to be created.
406 */
407 error = ENOENT;
408 if ((nameiop == CREATE || nameiop == RENAME) &&
409 islastcn) {
410 error = VOP_ACCESS(dvp, VWRITE, cred, td);
411 if (!error) {
412 /* keep the component name */
413 cnp->cn_flags |= SAVENAME;
414 error = EJUSTRETURN;
415 }
416 }
417 /* Done */
418 } else {
419 if (ino == NANDFS_WHT_INO)
420 cnp->cn_flags |= ISWHITEOUT;
421
422 if ((cnp->cn_flags & ISWHITEOUT) &&
423 (nameiop == LOOKUP))
424 return (ENOENT);
425
426 if ((nameiop == DELETE) && islastcn) {
427 if ((cnp->cn_flags & ISWHITEOUT) &&
428 (cnp->cn_flags & DOWHITEOUT)) {
429 cnp->cn_flags |= SAVENAME;
430 dir_node->nn_diroff = off;
431 return (EJUSTRETURN);
432 }
433
434 error = VOP_ACCESS(dvp, VWRITE, cred,
435 cnp->cn_thread);
436 if (error)
437 return (error);
438
439 /* Try to create/reuse the node */
440 error = nandfs_get_node(nmp, ino, &node);
441 if (!error) {
442 *vpp = NTOV(node);
443 node->nn_diroff = off;
444 }
445
446 if ((dir_node->nn_inode.i_mode & ISVTX) &&
447 cred->cr_uid != 0 &&
448 cred->cr_uid != dir_node->nn_inode.i_uid &&
449 node->nn_inode.i_uid != cred->cr_uid) {
450 vput(*vpp);
451 *vpp = NULL;
452 return (EPERM);
453 }
454 } else if ((nameiop == RENAME) && islastcn) {
455 error = VOP_ACCESS(dvp, VWRITE, cred,
456 cnp->cn_thread);
457 if (error)
458 return (error);
459
460 /* Try to create/reuse the node */
461 error = nandfs_get_node(nmp, ino, &node);
462 if (!error) {
463 *vpp = NTOV(node);
464 node->nn_diroff = off;
465 }
466 } else {
467 /* Try to create/reuse the node */
468 error = nandfs_get_node(nmp, ino, &node);
469 if (!error) {
470 *vpp = NTOV(node);
471 node->nn_diroff = off;
472 }
473 }
474 }
475 }
476
477 out:
478 /*
479 * Store result in the cache if requested. If we are creating a file,
480 * the file might not be found and thus putting it into the namecache
481 * might be seen as negative caching.
482 */
483 if ((cnp->cn_flags & MAKEENTRY) != 0)
484 cache_enter(dvp, *vpp, cnp);
485
486 return (error);
487
488 }
489
490 static int
nandfs_getattr(struct vop_getattr_args * ap)491 nandfs_getattr(struct vop_getattr_args *ap)
492 {
493 struct vnode *vp = ap->a_vp;
494 struct vattr *vap = ap->a_vap;
495 struct nandfs_node *node = VTON(vp);
496 struct nandfs_inode *inode = &node->nn_inode;
497
498 DPRINTF(VNCALL, ("%s: vp: %p\n", __func__, vp));
499 nandfs_itimes(vp);
500
501 /* Basic info */
502 VATTR_NULL(vap);
503 vap->va_atime.tv_sec = inode->i_mtime;
504 vap->va_atime.tv_nsec = inode->i_mtime_nsec;
505 vap->va_mtime.tv_sec = inode->i_mtime;
506 vap->va_mtime.tv_nsec = inode->i_mtime_nsec;
507 vap->va_ctime.tv_sec = inode->i_ctime;
508 vap->va_ctime.tv_nsec = inode->i_ctime_nsec;
509 vap->va_type = IFTOVT(inode->i_mode);
510 vap->va_mode = inode->i_mode & ~S_IFMT;
511 vap->va_nlink = inode->i_links_count;
512 vap->va_uid = inode->i_uid;
513 vap->va_gid = inode->i_gid;
514 vap->va_rdev = inode->i_special;
515 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
516 vap->va_fileid = node->nn_ino;
517 vap->va_size = inode->i_size;
518 vap->va_blocksize = node->nn_nandfsdev->nd_blocksize;
519 vap->va_gen = 0;
520 vap->va_flags = inode->i_flags;
521 vap->va_bytes = inode->i_blocks * vap->va_blocksize;
522 vap->va_filerev = 0;
523 vap->va_vaflags = 0;
524
525 return (0);
526 }
527
528 static int
nandfs_vtruncbuf(struct vnode * vp,uint64_t nblks)529 nandfs_vtruncbuf(struct vnode *vp, uint64_t nblks)
530 {
531 struct nandfs_device *nffsdev;
532 struct bufobj *bo;
533 struct buf *bp, *nbp;
534
535 bo = &vp->v_bufobj;
536 nffsdev = VTON(vp)->nn_nandfsdev;
537
538 ASSERT_VOP_LOCKED(vp, "nandfs_truncate");
539 restart:
540 BO_LOCK(bo);
541 restart_locked:
542 TAILQ_FOREACH_SAFE(bp, &bo->bo_clean.bv_hd, b_bobufs, nbp) {
543 if (bp->b_lblkno < nblks)
544 continue;
545 if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL))
546 goto restart_locked;
547
548 bremfree(bp);
549 bp->b_flags |= (B_INVAL | B_RELBUF);
550 bp->b_flags &= ~(B_ASYNC | B_MANAGED);
551 BO_UNLOCK(bo);
552 brelse(bp);
553 BO_LOCK(bo);
554 }
555
556 TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
557 if (bp->b_lblkno < nblks)
558 continue;
559 if (BUF_LOCK(bp,
560 LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK,
561 BO_LOCKPTR(bo)) == ENOLCK)
562 goto restart;
563 bp->b_flags |= (B_INVAL | B_RELBUF);
564 bp->b_flags &= ~(B_ASYNC | B_MANAGED);
565 brelse(bp);
566 nandfs_dirty_bufs_decrement(nffsdev);
567 BO_LOCK(bo);
568 }
569
570 BO_UNLOCK(bo);
571
572 return (0);
573 }
574
575 static int
nandfs_truncate(struct vnode * vp,uint64_t newsize)576 nandfs_truncate(struct vnode *vp, uint64_t newsize)
577 {
578 struct nandfs_device *nffsdev;
579 struct nandfs_node *node;
580 struct nandfs_inode *inode;
581 struct buf *bp = NULL;
582 uint64_t oblks, nblks, vblk, size, rest;
583 int error;
584
585 node = VTON(vp);
586 nffsdev = node->nn_nandfsdev;
587 inode = &node->nn_inode;
588
589 /* Calculate end of file */
590 size = inode->i_size;
591
592 if (newsize == size) {
593 node->nn_flags |= IN_CHANGE | IN_UPDATE;
594 nandfs_itimes(vp);
595 return (0);
596 }
597
598 if (newsize > size) {
599 inode->i_size = newsize;
600 vnode_pager_setsize(vp, newsize);
601 node->nn_flags |= IN_CHANGE | IN_UPDATE;
602 nandfs_itimes(vp);
603 return (0);
604 }
605
606 nblks = howmany(newsize, nffsdev->nd_blocksize);
607 oblks = howmany(size, nffsdev->nd_blocksize);
608 rest = newsize % nffsdev->nd_blocksize;
609
610 if (rest) {
611 error = nandfs_bmap_lookup(node, nblks - 1, &vblk);
612 if (error)
613 return (error);
614
615 if (vblk != 0)
616 error = nandfs_bread(node, nblks - 1, NOCRED, 0, &bp);
617 else
618 error = nandfs_bcreate(node, nblks - 1, NOCRED, 0, &bp);
619
620 if (error) {
621 if (bp)
622 brelse(bp);
623 return (error);
624 }
625
626 bzero((char *)bp->b_data + rest,
627 (u_int)(nffsdev->nd_blocksize - rest));
628 error = nandfs_dirty_buf(bp, 0);
629 if (error)
630 return (error);
631 }
632
633 DPRINTF(VNCALL, ("%s: vp %p oblks %jx nblks %jx\n", __func__, vp, oblks,
634 nblks));
635
636 error = nandfs_bmap_truncate_mapping(node, oblks - 1, nblks - 1);
637 if (error) {
638 if (bp)
639 nandfs_undirty_buf(bp);
640 return (error);
641 }
642
643 error = nandfs_vtruncbuf(vp, nblks);
644 if (error) {
645 if (bp)
646 nandfs_undirty_buf(bp);
647 return (error);
648 }
649
650 inode->i_size = newsize;
651 vnode_pager_setsize(vp, newsize);
652 node->nn_flags |= IN_CHANGE | IN_UPDATE;
653 nandfs_itimes(vp);
654
655 return (error);
656 }
657
658 static void
nandfs_itimes_locked(struct vnode * vp)659 nandfs_itimes_locked(struct vnode *vp)
660 {
661 struct nandfs_node *node;
662 struct nandfs_inode *inode;
663 struct timespec ts;
664
665 ASSERT_VI_LOCKED(vp, __func__);
666
667 node = VTON(vp);
668 inode = &node->nn_inode;
669
670 if ((node->nn_flags & (IN_ACCESS | IN_CHANGE | IN_UPDATE)) == 0)
671 return;
672
673 if (((vp->v_mount->mnt_kern_flag &
674 (MNTK_SUSPENDED | MNTK_SUSPEND)) == 0) ||
675 (node->nn_flags & (IN_CHANGE | IN_UPDATE)))
676 node->nn_flags |= IN_MODIFIED;
677
678 vfs_timestamp(&ts);
679 if (node->nn_flags & IN_UPDATE) {
680 inode->i_mtime = ts.tv_sec;
681 inode->i_mtime_nsec = ts.tv_nsec;
682 }
683 if (node->nn_flags & IN_CHANGE) {
684 inode->i_ctime = ts.tv_sec;
685 inode->i_ctime_nsec = ts.tv_nsec;
686 }
687
688 node->nn_flags &= ~(IN_ACCESS | IN_CHANGE | IN_UPDATE);
689 }
690
691 void
nandfs_itimes(struct vnode * vp)692 nandfs_itimes(struct vnode *vp)
693 {
694
695 VI_LOCK(vp);
696 nandfs_itimes_locked(vp);
697 VI_UNLOCK(vp);
698 }
699
700 static int
nandfs_chmod(struct vnode * vp,int mode,struct ucred * cred,struct thread * td)701 nandfs_chmod(struct vnode *vp, int mode, struct ucred *cred, struct thread *td)
702 {
703 struct nandfs_node *node = VTON(vp);
704 struct nandfs_inode *inode = &node->nn_inode;
705 uint16_t nmode;
706 int error = 0;
707
708 DPRINTF(VNCALL, ("%s: vp %p, mode %x, cred %p, td %p\n", __func__, vp,
709 mode, cred, td));
710 /*
711 * To modify the permissions on a file, must possess VADMIN
712 * for that file.
713 */
714 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
715 return (error);
716
717 /*
718 * Privileged processes may set the sticky bit on non-directories,
719 * as well as set the setgid bit on a file with a group that the
720 * process is not a member of. Both of these are allowed in
721 * jail(8).
722 */
723 if (vp->v_type != VDIR && (mode & S_ISTXT)) {
724 if (priv_check_cred(cred, PRIV_VFS_STICKYFILE, 0))
725 return (EFTYPE);
726 }
727 if (!groupmember(inode->i_gid, cred) && (mode & ISGID)) {
728 error = priv_check_cred(cred, PRIV_VFS_SETGID, 0);
729 if (error)
730 return (error);
731 }
732
733 /*
734 * Deny setting setuid if we are not the file owner.
735 */
736 if ((mode & ISUID) && inode->i_uid != cred->cr_uid) {
737 error = priv_check_cred(cred, PRIV_VFS_ADMIN, 0);
738 if (error)
739 return (error);
740 }
741
742 nmode = inode->i_mode;
743 nmode &= ~ALLPERMS;
744 nmode |= (mode & ALLPERMS);
745 inode->i_mode = nmode;
746 node->nn_flags |= IN_CHANGE;
747
748 DPRINTF(VNCALL, ("%s: to mode %x\n", __func__, nmode));
749
750 return (error);
751 }
752
753 static int
nandfs_chown(struct vnode * vp,uid_t uid,gid_t gid,struct ucred * cred,struct thread * td)754 nandfs_chown(struct vnode *vp, uid_t uid, gid_t gid, struct ucred *cred,
755 struct thread *td)
756 {
757 struct nandfs_node *node = VTON(vp);
758 struct nandfs_inode *inode = &node->nn_inode;
759 uid_t ouid;
760 gid_t ogid;
761 int error = 0;
762
763 if (uid == (uid_t)VNOVAL)
764 uid = inode->i_uid;
765 if (gid == (gid_t)VNOVAL)
766 gid = inode->i_gid;
767 /*
768 * To modify the ownership of a file, must possess VADMIN for that
769 * file.
770 */
771 if ((error = VOP_ACCESSX(vp, VWRITE_OWNER, cred, td)))
772 return (error);
773 /*
774 * To change the owner of a file, or change the group of a file to a
775 * group of which we are not a member, the caller must have
776 * privilege.
777 */
778 if (((uid != inode->i_uid && uid != cred->cr_uid) ||
779 (gid != inode->i_gid && !groupmember(gid, cred))) &&
780 (error = priv_check_cred(cred, PRIV_VFS_CHOWN, 0)))
781 return (error);
782 ogid = inode->i_gid;
783 ouid = inode->i_uid;
784
785 inode->i_gid = gid;
786 inode->i_uid = uid;
787
788 node->nn_flags |= IN_CHANGE;
789 if ((inode->i_mode & (ISUID | ISGID)) &&
790 (ouid != uid || ogid != gid)) {
791 if (priv_check_cred(cred, PRIV_VFS_RETAINSUGID, 0))
792 inode->i_mode &= ~(ISUID | ISGID);
793 }
794 DPRINTF(VNCALL, ("%s: vp %p, cred %p, td %p - ret OK\n", __func__, vp,
795 cred, td));
796 return (0);
797 }
798
799 static int
nandfs_setattr(struct vop_setattr_args * ap)800 nandfs_setattr(struct vop_setattr_args *ap)
801 {
802 struct vnode *vp = ap->a_vp;
803 struct nandfs_node *node = VTON(vp);
804 struct nandfs_inode *inode = &node->nn_inode;
805 struct vattr *vap = ap->a_vap;
806 struct ucred *cred = ap->a_cred;
807 struct thread *td = curthread;
808 uint32_t flags;
809 int error = 0;
810
811 if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) ||
812 (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) ||
813 (vap->va_blocksize != VNOVAL) || (vap->va_rdev != VNOVAL) ||
814 (vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) {
815 DPRINTF(VNCALL, ("%s: unsettable attribute\n", __func__));
816 return (EINVAL);
817 }
818
819 if (vap->va_flags != VNOVAL) {
820 DPRINTF(VNCALL, ("%s: vp:%p td:%p flags:%lx\n", __func__, vp,
821 td, vap->va_flags));
822
823 if (vp->v_mount->mnt_flag & MNT_RDONLY)
824 return (EROFS);
825 /*
826 * Callers may only modify the file flags on objects they
827 * have VADMIN rights for.
828 */
829 if ((error = VOP_ACCESS(vp, VADMIN, cred, td)))
830 return (error);
831 /*
832 * Unprivileged processes are not permitted to unset system
833 * flags, or modify flags if any system flags are set.
834 * Privileged non-jail processes may not modify system flags
835 * if securelevel > 0 and any existing system flags are set.
836 * Privileged jail processes behave like privileged non-jail
837 * processes if the PR_ALLOW_CHFLAGS permission bit is set;
838 * otherwise, they behave like unprivileged processes.
839 */
840
841 flags = inode->i_flags;
842 if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
843 if (flags & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
844 error = securelevel_gt(cred, 0);
845 if (error)
846 return (error);
847 }
848 /* Snapshot flag cannot be set or cleared */
849 if (((vap->va_flags & SF_SNAPSHOT) != 0 &&
850 (flags & SF_SNAPSHOT) == 0) ||
851 ((vap->va_flags & SF_SNAPSHOT) == 0 &&
852 (flags & SF_SNAPSHOT) != 0))
853 return (EPERM);
854
855 inode->i_flags = vap->va_flags;
856 } else {
857 if (flags & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND) ||
858 (vap->va_flags & UF_SETTABLE) != vap->va_flags)
859 return (EPERM);
860
861 flags &= SF_SETTABLE;
862 flags |= (vap->va_flags & UF_SETTABLE);
863 inode->i_flags = flags;
864 }
865 node->nn_flags |= IN_CHANGE;
866 if (vap->va_flags & (IMMUTABLE | APPEND))
867 return (0);
868 }
869 if (inode->i_flags & (IMMUTABLE | APPEND))
870 return (EPERM);
871
872 if (vap->va_size != (u_quad_t)VNOVAL) {
873 DPRINTF(VNCALL, ("%s: vp:%p td:%p size:%jx\n", __func__, vp, td,
874 (uintmax_t)vap->va_size));
875
876 switch (vp->v_type) {
877 case VDIR:
878 return (EISDIR);
879 case VLNK:
880 case VREG:
881 if (vp->v_mount->mnt_flag & MNT_RDONLY)
882 return (EROFS);
883 if ((inode->i_flags & SF_SNAPSHOT) != 0)
884 return (EPERM);
885 break;
886 default:
887 return (0);
888 }
889
890 if (vap->va_size > node->nn_nandfsdev->nd_maxfilesize)
891 return (EFBIG);
892
893 KASSERT((vp->v_type == VREG), ("Set size %d", vp->v_type));
894 nandfs_truncate(vp, vap->va_size);
895 node->nn_flags |= IN_CHANGE;
896
897 return (0);
898 }
899
900 if (vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL) {
901 if (vp->v_mount->mnt_flag & MNT_RDONLY)
902 return (EROFS);
903 DPRINTF(VNCALL, ("%s: vp:%p td:%p uid/gid %x/%x\n", __func__,
904 vp, td, vap->va_uid, vap->va_gid));
905 error = nandfs_chown(vp, vap->va_uid, vap->va_gid, cred, td);
906 if (error)
907 return (error);
908 }
909
910 if (vap->va_mode != (mode_t)VNOVAL) {
911 if (vp->v_mount->mnt_flag & MNT_RDONLY)
912 return (EROFS);
913 DPRINTF(VNCALL, ("%s: vp:%p td:%p mode %x\n", __func__, vp, td,
914 vap->va_mode));
915
916 error = nandfs_chmod(vp, (int)vap->va_mode, cred, td);
917 if (error)
918 return (error);
919 }
920 if (vap->va_atime.tv_sec != VNOVAL ||
921 vap->va_mtime.tv_sec != VNOVAL ||
922 vap->va_birthtime.tv_sec != VNOVAL) {
923 DPRINTF(VNCALL, ("%s: vp:%p td:%p time a/m/b %jx/%jx/%jx\n",
924 __func__, vp, td, (uintmax_t)vap->va_atime.tv_sec,
925 (uintmax_t)vap->va_mtime.tv_sec,
926 (uintmax_t)vap->va_birthtime.tv_sec));
927
928 if (vap->va_atime.tv_sec != VNOVAL)
929 node->nn_flags |= IN_ACCESS;
930 if (vap->va_mtime.tv_sec != VNOVAL)
931 node->nn_flags |= IN_CHANGE | IN_UPDATE;
932 if (vap->va_birthtime.tv_sec != VNOVAL)
933 node->nn_flags |= IN_MODIFIED;
934 nandfs_itimes(vp);
935 return (0);
936 }
937
938 return (0);
939 }
940
941 static int
nandfs_open(struct vop_open_args * ap)942 nandfs_open(struct vop_open_args *ap)
943 {
944 struct nandfs_node *node = VTON(ap->a_vp);
945 uint64_t filesize;
946
947 DPRINTF(VNCALL, ("nandfs_open called ap->a_mode %x\n", ap->a_mode));
948
949 if (ap->a_vp->v_type == VCHR || ap->a_vp->v_type == VBLK)
950 return (EOPNOTSUPP);
951
952 if ((node->nn_inode.i_flags & APPEND) &&
953 (ap->a_mode & (FWRITE | O_APPEND)) == FWRITE)
954 return (EPERM);
955
956 filesize = node->nn_inode.i_size;
957 vnode_create_vobject(ap->a_vp, filesize, ap->a_td);
958
959 return (0);
960 }
961
962 static int
nandfs_close(struct vop_close_args * ap)963 nandfs_close(struct vop_close_args *ap)
964 {
965 struct vnode *vp = ap->a_vp;
966 struct nandfs_node *node = VTON(vp);
967
968 DPRINTF(VNCALL, ("%s: vp %p node %p\n", __func__, vp, node));
969
970 mtx_lock(&vp->v_interlock);
971 if (vp->v_usecount > 1)
972 nandfs_itimes_locked(vp);
973 mtx_unlock(&vp->v_interlock);
974
975 return (0);
976 }
977
978 static int
nandfs_check_possible(struct vnode * vp,struct vattr * vap,mode_t mode)979 nandfs_check_possible(struct vnode *vp, struct vattr *vap, mode_t mode)
980 {
981
982 /* Check if we are allowed to write */
983 switch (vap->va_type) {
984 case VDIR:
985 case VLNK:
986 case VREG:
987 /*
988 * Normal nodes: check if we're on a read-only mounted
989 * filingsystem and bomb out if we're trying to write.
990 */
991 if ((mode & VMODIFY_PERMS) && (vp->v_mount->mnt_flag & MNT_RDONLY))
992 return (EROFS);
993 break;
994 case VBLK:
995 case VCHR:
996 case VSOCK:
997 case VFIFO:
998 /*
999 * Special nodes: even on read-only mounted filingsystems
1000 * these are allowed to be written to if permissions allow.
1001 */
1002 break;
1003 default:
1004 /* No idea what this is */
1005 return (EINVAL);
1006 }
1007
1008 /* No one may write immutable files */
1009 if ((mode & VWRITE) && (VTON(vp)->nn_inode.i_flags & IMMUTABLE))
1010 return (EPERM);
1011
1012 return (0);
1013 }
1014
1015 static int
nandfs_check_permitted(struct vnode * vp,struct vattr * vap,mode_t mode,struct ucred * cred)1016 nandfs_check_permitted(struct vnode *vp, struct vattr *vap, mode_t mode,
1017 struct ucred *cred)
1018 {
1019
1020 return (vaccess(vp->v_type, vap->va_mode, vap->va_uid, vap->va_gid, mode,
1021 cred, NULL));
1022 }
1023
1024 static int
nandfs_advlock(struct vop_advlock_args * ap)1025 nandfs_advlock(struct vop_advlock_args *ap)
1026 {
1027 struct nandfs_node *nvp;
1028 quad_t size;
1029
1030 nvp = VTON(ap->a_vp);
1031 size = nvp->nn_inode.i_size;
1032 return (lf_advlock(ap, &(nvp->nn_lockf), size));
1033 }
1034
1035 static int
nandfs_access(struct vop_access_args * ap)1036 nandfs_access(struct vop_access_args *ap)
1037 {
1038 struct vnode *vp = ap->a_vp;
1039 accmode_t accmode = ap->a_accmode;
1040 struct ucred *cred = ap->a_cred;
1041 struct vattr vap;
1042 int error;
1043
1044 DPRINTF(VNCALL, ("%s: vp:%p mode: %x\n", __func__, vp, accmode));
1045
1046 error = VOP_GETATTR(vp, &vap, NULL);
1047 if (error)
1048 return (error);
1049
1050 error = nandfs_check_possible(vp, &vap, accmode);
1051 if (error)
1052 return (error);
1053
1054 error = nandfs_check_permitted(vp, &vap, accmode, cred);
1055
1056 return (error);
1057 }
1058
1059 static int
nandfs_print(struct vop_print_args * ap)1060 nandfs_print(struct vop_print_args *ap)
1061 {
1062 struct vnode *vp = ap->a_vp;
1063 struct nandfs_node *nvp = VTON(vp);
1064
1065 printf("\tvp=%p, nandfs_node=%p\n", vp, nvp);
1066 printf("nandfs inode %#jx\n", (uintmax_t)nvp->nn_ino);
1067 printf("flags = 0x%b\n", (u_int)nvp->nn_flags, PRINT_NODE_FLAGS);
1068
1069 return (0);
1070 }
1071
1072 static void
nandfs_read_filebuf(struct nandfs_node * node,struct buf * bp)1073 nandfs_read_filebuf(struct nandfs_node *node, struct buf *bp)
1074 {
1075 struct nandfs_device *nandfsdev = node->nn_nandfsdev;
1076 struct buf *nbp;
1077 nandfs_daddr_t vblk, pblk;
1078 nandfs_lbn_t from;
1079 uint32_t blocksize;
1080 int error = 0;
1081 int blk2dev = nandfsdev->nd_blocksize / DEV_BSIZE;
1082
1083 /*
1084 * Translate all the block sectors into a series of buffers to read
1085 * asynchronously from the nandfs device. Note that this lookup may
1086 * induce readin's too.
1087 */
1088
1089 blocksize = nandfsdev->nd_blocksize;
1090 if (bp->b_bcount / blocksize != 1)
1091 panic("invalid b_count in bp %p\n", bp);
1092
1093 from = bp->b_blkno;
1094
1095 DPRINTF(READ, ("\tread in from inode %#jx blkno %#jx"
1096 " count %#lx\n", (uintmax_t)node->nn_ino, from,
1097 bp->b_bcount));
1098
1099 /* Get virtual block numbers for the vnode's buffer span */
1100 error = nandfs_bmap_lookup(node, from, &vblk);
1101 if (error) {
1102 bp->b_error = EINVAL;
1103 bp->b_ioflags |= BIO_ERROR;
1104 bufdone(bp);
1105 return;
1106 }
1107
1108 /* Translate virtual block numbers to physical block numbers */
1109 error = nandfs_vtop(node, vblk, &pblk);
1110 if (error) {
1111 bp->b_error = EINVAL;
1112 bp->b_ioflags |= BIO_ERROR;
1113 bufdone(bp);
1114 return;
1115 }
1116
1117 /* Issue translated blocks */
1118 bp->b_resid = bp->b_bcount;
1119
1120 /* Note virtual block 0 marks not mapped */
1121 if (vblk == 0) {
1122 vfs_bio_clrbuf(bp);
1123 bufdone(bp);
1124 return;
1125 }
1126
1127 nbp = bp;
1128 nbp->b_blkno = pblk * blk2dev;
1129 bp->b_iooffset = dbtob(nbp->b_blkno);
1130 MPASS(bp->b_iooffset >= 0);
1131 BO_STRATEGY(&nandfsdev->nd_devvp->v_bufobj, nbp);
1132 nandfs_vblk_set(bp, vblk);
1133 DPRINTF(READ, ("read_filebuf : ino %#jx blk %#jx -> "
1134 "%#jx -> %#jx [bp %p]\n", (uintmax_t)node->nn_ino,
1135 (uintmax_t)(from), (uintmax_t)vblk,
1136 (uintmax_t)pblk, nbp));
1137 }
1138
1139 static void
nandfs_write_filebuf(struct nandfs_node * node,struct buf * bp)1140 nandfs_write_filebuf(struct nandfs_node *node, struct buf *bp)
1141 {
1142 struct nandfs_device *nandfsdev = node->nn_nandfsdev;
1143
1144 bp->b_iooffset = dbtob(bp->b_blkno);
1145 MPASS(bp->b_iooffset >= 0);
1146 BO_STRATEGY(&nandfsdev->nd_devvp->v_bufobj, bp);
1147 }
1148
1149 static int
nandfs_strategy(struct vop_strategy_args * ap)1150 nandfs_strategy(struct vop_strategy_args *ap)
1151 {
1152 struct vnode *vp = ap->a_vp;
1153 struct buf *bp = ap->a_bp;
1154 struct nandfs_node *node = VTON(vp);
1155
1156
1157 /* check if we ought to be here */
1158 KASSERT((vp->v_type != VBLK && vp->v_type != VCHR),
1159 ("nandfs_strategy on type %d", vp->v_type));
1160
1161 /* Translate if needed and pass on */
1162 if (bp->b_iocmd == BIO_READ) {
1163 nandfs_read_filebuf(node, bp);
1164 return (0);
1165 }
1166
1167 /* Send to segment collector */
1168 nandfs_write_filebuf(node, bp);
1169 return (0);
1170 }
1171
1172 static int
nandfs_readdir(struct vop_readdir_args * ap)1173 nandfs_readdir(struct vop_readdir_args *ap)
1174 {
1175 struct uio *uio = ap->a_uio;
1176 struct vnode *vp = ap->a_vp;
1177 struct nandfs_node *node = VTON(vp);
1178 struct nandfs_dir_entry *ndirent;
1179 struct dirent dirent;
1180 struct buf *bp;
1181 uint64_t file_size, diroffset, transoffset, blkoff;
1182 uint64_t blocknr;
1183 uint32_t blocksize = node->nn_nandfsdev->nd_blocksize;
1184 uint8_t *pos, name_len;
1185 int error;
1186
1187 DPRINTF(READDIR, ("nandfs_readdir called\n"));
1188
1189 if (vp->v_type != VDIR)
1190 return (ENOTDIR);
1191
1192 file_size = node->nn_inode.i_size;
1193 DPRINTF(READDIR, ("nandfs_readdir filesize %jd resid %zd\n",
1194 (uintmax_t)file_size, uio->uio_resid ));
1195
1196 /* We are called just as long as we keep on pushing data in */
1197 error = 0;
1198 if ((uio->uio_offset < file_size) &&
1199 (uio->uio_resid >= sizeof(struct dirent))) {
1200 diroffset = uio->uio_offset;
1201 transoffset = diroffset;
1202
1203 blocknr = diroffset / blocksize;
1204 blkoff = diroffset % blocksize;
1205 error = nandfs_bread(node, blocknr, NOCRED, 0, &bp);
1206 if (error) {
1207 brelse(bp);
1208 return (EIO);
1209 }
1210 while (diroffset < file_size) {
1211 DPRINTF(READDIR, ("readdir : offset = %"PRIu64"\n",
1212 diroffset));
1213 if (blkoff >= blocksize) {
1214 blkoff = 0; blocknr++;
1215 brelse(bp);
1216 error = nandfs_bread(node, blocknr, NOCRED, 0,
1217 &bp);
1218 if (error) {
1219 brelse(bp);
1220 return (EIO);
1221 }
1222 }
1223
1224 /* Read in one dirent */
1225 pos = (uint8_t *)bp->b_data + blkoff;
1226 ndirent = (struct nandfs_dir_entry *)pos;
1227
1228 name_len = ndirent->name_len;
1229 memset(&dirent, 0, sizeof(dirent));
1230 dirent.d_fileno = ndirent->inode;
1231 if (dirent.d_fileno) {
1232 dirent.d_type = ndirent->file_type;
1233 dirent.d_namlen = name_len;
1234 strncpy(dirent.d_name, ndirent->name, name_len);
1235 dirent.d_reclen = GENERIC_DIRSIZ(&dirent);
1236 /* NOTE: d_off is the offset of the *next* entry. */
1237 dirent.d_off = diroffset + ndirent->rec_len;
1238 dirent_terminate(&dirent);
1239 DPRINTF(READDIR, ("copying `%*.*s`\n", name_len,
1240 name_len, dirent.d_name));
1241 }
1242
1243 /*
1244 * If there isn't enough space in the uio to return a
1245 * whole dirent, break off read
1246 */
1247 if (uio->uio_resid < GENERIC_DIRSIZ(&dirent))
1248 break;
1249
1250 /* Transfer */
1251 if (dirent.d_fileno)
1252 uiomove(&dirent, dirent.d_reclen, uio);
1253
1254 /* Advance */
1255 diroffset += ndirent->rec_len;
1256 blkoff += ndirent->rec_len;
1257
1258 /* Remember the last entry we transferred */
1259 transoffset = diroffset;
1260 }
1261 brelse(bp);
1262
1263 /* Pass on last transferred offset */
1264 uio->uio_offset = transoffset;
1265 }
1266
1267 if (ap->a_eofflag)
1268 *ap->a_eofflag = (uio->uio_offset >= file_size);
1269
1270 return (error);
1271 }
1272
1273 static int
nandfs_dirempty(struct vnode * dvp,uint64_t parentino,struct ucred * cred)1274 nandfs_dirempty(struct vnode *dvp, uint64_t parentino, struct ucred *cred)
1275 {
1276 struct nandfs_node *dnode = VTON(dvp);
1277 struct nandfs_dir_entry *dirent;
1278 uint64_t file_size = dnode->nn_inode.i_size;
1279 uint64_t blockcount = dnode->nn_inode.i_blocks;
1280 uint64_t blocknr;
1281 uint32_t blocksize = dnode->nn_nandfsdev->nd_blocksize;
1282 uint32_t limit;
1283 uint32_t off;
1284 uint8_t *pos;
1285 struct buf *bp;
1286 int error;
1287
1288 DPRINTF(LOOKUP, ("%s: dvp %p parentino %#jx cred %p\n", __func__, dvp,
1289 (uintmax_t)parentino, cred));
1290
1291 KASSERT((file_size != 0), ("nandfs_dirempty for NULL dir %p", dvp));
1292
1293 blocknr = 0;
1294 while (blocknr < blockcount) {
1295 error = nandfs_bread(dnode, blocknr, NOCRED, 0, &bp);
1296 if (error) {
1297 brelse(bp);
1298 return (0);
1299 }
1300
1301 pos = (uint8_t *)bp->b_data;
1302 off = 0;
1303
1304 if (blocknr == (blockcount - 1))
1305 limit = file_size % blocksize;
1306 else
1307 limit = blocksize;
1308
1309 while (off < limit) {
1310 dirent = (struct nandfs_dir_entry *)(pos + off);
1311 off += dirent->rec_len;
1312
1313 if (dirent->inode == 0)
1314 continue;
1315
1316 switch (dirent->name_len) {
1317 case 0:
1318 break;
1319 case 1:
1320 if (dirent->name[0] != '.')
1321 goto notempty;
1322
1323 KASSERT(dirent->inode == dnode->nn_ino,
1324 (".'s inode does not match dir"));
1325 break;
1326 case 2:
1327 if (dirent->name[0] != '.' &&
1328 dirent->name[1] != '.')
1329 goto notempty;
1330
1331 KASSERT(dirent->inode == parentino,
1332 ("..'s inode does not match parent"));
1333 break;
1334 default:
1335 goto notempty;
1336 }
1337 }
1338
1339 brelse(bp);
1340 blocknr++;
1341 }
1342
1343 return (1);
1344 notempty:
1345 brelse(bp);
1346 return (0);
1347 }
1348
1349 static int
nandfs_link(struct vop_link_args * ap)1350 nandfs_link(struct vop_link_args *ap)
1351 {
1352 struct vnode *tdvp = ap->a_tdvp;
1353 struct vnode *vp = ap->a_vp;
1354 struct componentname *cnp = ap->a_cnp;
1355 struct nandfs_node *node = VTON(vp);
1356 struct nandfs_inode *inode = &node->nn_inode;
1357 int error;
1358
1359 if (inode->i_links_count >= NANDFS_LINK_MAX)
1360 return (EMLINK);
1361
1362 if (inode->i_flags & (IMMUTABLE | APPEND))
1363 return (EPERM);
1364
1365 /* Update link count */
1366 inode->i_links_count++;
1367
1368 /* Add dir entry */
1369 error = nandfs_add_dirent(tdvp, node->nn_ino, cnp->cn_nameptr,
1370 cnp->cn_namelen, IFTODT(inode->i_mode));
1371 if (error) {
1372 inode->i_links_count--;
1373 }
1374
1375 node->nn_flags |= IN_CHANGE;
1376 nandfs_itimes(vp);
1377 DPRINTF(VNCALL, ("%s: tdvp %p vp %p cnp %p\n",
1378 __func__, tdvp, vp, cnp));
1379
1380 return (0);
1381 }
1382
1383 static int
nandfs_create(struct vop_create_args * ap)1384 nandfs_create(struct vop_create_args *ap)
1385 {
1386 struct vnode *dvp = ap->a_dvp;
1387 struct vnode **vpp = ap->a_vpp;
1388 struct componentname *cnp = ap->a_cnp;
1389 uint16_t mode = MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode);
1390 struct nandfs_node *dir_node = VTON(dvp);
1391 struct nandfsmount *nmp = dir_node->nn_nmp;
1392 struct nandfs_node *node;
1393 int error;
1394
1395 DPRINTF(VNCALL, ("%s: dvp %p\n", __func__, dvp));
1396
1397 if (nandfs_fs_full(dir_node->nn_nandfsdev))
1398 return (ENOSPC);
1399
1400 /* Create new vnode/inode */
1401 error = nandfs_node_create(nmp, &node, mode);
1402 if (error)
1403 return (error);
1404 node->nn_inode.i_gid = dir_node->nn_inode.i_gid;
1405 node->nn_inode.i_uid = cnp->cn_cred->cr_uid;
1406
1407 /* Add new dir entry */
1408 error = nandfs_add_dirent(dvp, node->nn_ino, cnp->cn_nameptr,
1409 cnp->cn_namelen, IFTODT(mode));
1410 if (error) {
1411 if (nandfs_node_destroy(node)) {
1412 nandfs_error("%s: error destroying node %p\n",
1413 __func__, node);
1414 }
1415 return (error);
1416 }
1417 *vpp = NTOV(node);
1418 if ((cnp->cn_flags & MAKEENTRY) != 0)
1419 cache_enter(dvp, *vpp, cnp);
1420
1421 DPRINTF(VNCALL, ("created file vp %p nandnode %p ino %jx\n", *vpp, node,
1422 (uintmax_t)node->nn_ino));
1423 return (0);
1424 }
1425
1426 static int
nandfs_remove(struct vop_remove_args * ap)1427 nandfs_remove(struct vop_remove_args *ap)
1428 {
1429 struct vnode *vp = ap->a_vp;
1430 struct vnode *dvp = ap->a_dvp;
1431 struct nandfs_node *node = VTON(vp);
1432 struct nandfs_node *dnode = VTON(dvp);
1433 struct componentname *cnp = ap->a_cnp;
1434
1435 DPRINTF(VNCALL, ("%s: dvp %p vp %p nandnode %p ino %#jx link %d\n",
1436 __func__, dvp, vp, node, (uintmax_t)node->nn_ino,
1437 node->nn_inode.i_links_count));
1438
1439 if (vp->v_type == VDIR)
1440 return (EISDIR);
1441
1442 /* Files marked as immutable or append-only cannot be deleted. */
1443 if ((node->nn_inode.i_flags & (IMMUTABLE | APPEND | NOUNLINK)) ||
1444 (dnode->nn_inode.i_flags & APPEND))
1445 return (EPERM);
1446
1447 nandfs_remove_dirent(dvp, node, cnp);
1448 node->nn_inode.i_links_count--;
1449 node->nn_flags |= IN_CHANGE;
1450
1451 return (0);
1452 }
1453
1454 /*
1455 * Check if source directory is in the path of the target directory.
1456 * Target is supplied locked, source is unlocked.
1457 * The target is always vput before returning.
1458 */
1459 static int
nandfs_checkpath(struct nandfs_node * src,struct nandfs_node * dest,struct ucred * cred)1460 nandfs_checkpath(struct nandfs_node *src, struct nandfs_node *dest,
1461 struct ucred *cred)
1462 {
1463 struct vnode *vp;
1464 int error, rootino;
1465 struct nandfs_dir_entry dirent;
1466
1467 vp = NTOV(dest);
1468 if (src->nn_ino == dest->nn_ino) {
1469 error = EEXIST;
1470 goto out;
1471 }
1472 rootino = NANDFS_ROOT_INO;
1473 error = 0;
1474 if (dest->nn_ino == rootino)
1475 goto out;
1476
1477 for (;;) {
1478 if (vp->v_type != VDIR) {
1479 error = ENOTDIR;
1480 break;
1481 }
1482
1483 error = vn_rdwr(UIO_READ, vp, (caddr_t)&dirent,
1484 NANDFS_DIR_REC_LEN(2), (off_t)0, UIO_SYSSPACE,
1485 IO_NODELOCKED | IO_NOMACCHECK, cred, NOCRED,
1486 NULL, NULL);
1487 if (error != 0)
1488 break;
1489 if (dirent.name_len != 2 ||
1490 dirent.name[0] != '.' ||
1491 dirent.name[1] != '.') {
1492 error = ENOTDIR;
1493 break;
1494 }
1495 if (dirent.inode == src->nn_ino) {
1496 error = EINVAL;
1497 break;
1498 }
1499 if (dirent.inode == rootino)
1500 break;
1501 vput(vp);
1502 if ((error = VFS_VGET(vp->v_mount, dirent.inode,
1503 LK_EXCLUSIVE, &vp)) != 0) {
1504 vp = NULL;
1505 break;
1506 }
1507 }
1508
1509 out:
1510 if (error == ENOTDIR)
1511 printf("checkpath: .. not a directory\n");
1512 if (vp != NULL)
1513 vput(vp);
1514 return (error);
1515 }
1516
1517 static int
nandfs_rename(struct vop_rename_args * ap)1518 nandfs_rename(struct vop_rename_args *ap)
1519 {
1520 struct vnode *tvp = ap->a_tvp;
1521 struct vnode *tdvp = ap->a_tdvp;
1522 struct vnode *fvp = ap->a_fvp;
1523 struct vnode *fdvp = ap->a_fdvp;
1524 struct componentname *tcnp = ap->a_tcnp;
1525 struct componentname *fcnp = ap->a_fcnp;
1526 int doingdirectory = 0, oldparent = 0, newparent = 0;
1527 int error = 0;
1528
1529 struct nandfs_node *fdnode, *fnode, *fnode1;
1530 struct nandfs_node *tdnode = VTON(tdvp);
1531 struct nandfs_node *tnode;
1532
1533 uint32_t tdflags, fflags, fdflags;
1534 uint16_t mode;
1535
1536 DPRINTF(VNCALL, ("%s: fdvp:%p fvp:%p tdvp:%p tdp:%p\n", __func__, fdvp,
1537 fvp, tdvp, tvp));
1538
1539 /*
1540 * Check for cross-device rename.
1541 */
1542 if ((fvp->v_mount != tdvp->v_mount) ||
1543 (tvp && (fvp->v_mount != tvp->v_mount))) {
1544 error = EXDEV;
1545 abortit:
1546 if (tdvp == tvp)
1547 vrele(tdvp);
1548 else
1549 vput(tdvp);
1550 if (tvp)
1551 vput(tvp);
1552 vrele(fdvp);
1553 vrele(fvp);
1554 return (error);
1555 }
1556
1557 tdflags = tdnode->nn_inode.i_flags;
1558 if (tvp &&
1559 ((VTON(tvp)->nn_inode.i_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
1560 (tdflags & APPEND))) {
1561 error = EPERM;
1562 goto abortit;
1563 }
1564
1565 /*
1566 * Renaming a file to itself has no effect. The upper layers should
1567 * not call us in that case. Temporarily just warn if they do.
1568 */
1569 if (fvp == tvp) {
1570 printf("nandfs_rename: fvp == tvp (can't happen)\n");
1571 error = 0;
1572 goto abortit;
1573 }
1574
1575 if ((error = vn_lock(fvp, LK_EXCLUSIVE)) != 0)
1576 goto abortit;
1577
1578 fdnode = VTON(fdvp);
1579 fnode = VTON(fvp);
1580
1581 if (fnode->nn_inode.i_links_count >= NANDFS_LINK_MAX) {
1582 VOP_UNLOCK(fvp, 0);
1583 error = EMLINK;
1584 goto abortit;
1585 }
1586
1587 fflags = fnode->nn_inode.i_flags;
1588 fdflags = fdnode->nn_inode.i_flags;
1589
1590 if ((fflags & (NOUNLINK | IMMUTABLE | APPEND)) ||
1591 (fdflags & APPEND)) {
1592 VOP_UNLOCK(fvp, 0);
1593 error = EPERM;
1594 goto abortit;
1595 }
1596
1597 mode = fnode->nn_inode.i_mode;
1598 if ((mode & S_IFMT) == S_IFDIR) {
1599 /*
1600 * Avoid ".", "..", and aliases of "." for obvious reasons.
1601 */
1602
1603 if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
1604 (fdvp == fvp) ||
1605 ((fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT) ||
1606 (fnode->nn_flags & IN_RENAME)) {
1607 VOP_UNLOCK(fvp, 0);
1608 error = EINVAL;
1609 goto abortit;
1610 }
1611 fnode->nn_flags |= IN_RENAME;
1612 doingdirectory = 1;
1613 DPRINTF(VNCALL, ("%s: doingdirectory dvp %p\n", __func__,
1614 tdvp));
1615 oldparent = fdnode->nn_ino;
1616 }
1617
1618 vrele(fdvp);
1619
1620 tnode = NULL;
1621 if (tvp)
1622 tnode = VTON(tvp);
1623
1624 /*
1625 * Bump link count on fvp while we are moving stuff around. If we
1626 * crash before completing the work, the link count may be wrong
1627 * but correctable.
1628 */
1629 fnode->nn_inode.i_links_count++;
1630
1631 /* Check for in path moving XXX */
1632 error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, tcnp->cn_thread);
1633 VOP_UNLOCK(fvp, 0);
1634 if (oldparent != tdnode->nn_ino)
1635 newparent = tdnode->nn_ino;
1636 if (doingdirectory && newparent) {
1637 if (error) /* write access check above */
1638 goto bad;
1639 if (tnode != NULL)
1640 vput(tvp);
1641
1642 error = nandfs_checkpath(fnode, tdnode, tcnp->cn_cred);
1643 if (error)
1644 goto out;
1645
1646 VREF(tdvp);
1647 error = relookup(tdvp, &tvp, tcnp);
1648 if (error)
1649 goto out;
1650 vrele(tdvp);
1651 tdnode = VTON(tdvp);
1652 tnode = NULL;
1653 if (tvp)
1654 tnode = VTON(tvp);
1655 }
1656
1657 /*
1658 * If the target doesn't exist, link the target to the source and
1659 * unlink the source. Otherwise, rewrite the target directory to
1660 * reference the source and remove the original entry.
1661 */
1662
1663 if (tvp == NULL) {
1664 /*
1665 * Account for ".." in new directory.
1666 */
1667 if (doingdirectory && fdvp != tdvp)
1668 tdnode->nn_inode.i_links_count++;
1669
1670 DPRINTF(VNCALL, ("%s: new entry in dvp:%p\n", __func__, tdvp));
1671 /*
1672 * Add name in new directory.
1673 */
1674 error = nandfs_add_dirent(tdvp, fnode->nn_ino, tcnp->cn_nameptr,
1675 tcnp->cn_namelen, IFTODT(fnode->nn_inode.i_mode));
1676 if (error) {
1677 if (doingdirectory && fdvp != tdvp)
1678 tdnode->nn_inode.i_links_count--;
1679 goto bad;
1680 }
1681
1682 vput(tdvp);
1683 } else {
1684 /*
1685 * If the parent directory is "sticky", then the user must
1686 * own the parent directory, or the destination of the rename,
1687 * otherwise the destination may not be changed (except by
1688 * root). This implements append-only directories.
1689 */
1690 if ((tdnode->nn_inode.i_mode & S_ISTXT) &&
1691 tcnp->cn_cred->cr_uid != 0 &&
1692 tcnp->cn_cred->cr_uid != tdnode->nn_inode.i_uid &&
1693 tnode->nn_inode.i_uid != tcnp->cn_cred->cr_uid) {
1694 error = EPERM;
1695 goto bad;
1696 }
1697 /*
1698 * Target must be empty if a directory and have no links
1699 * to it. Also, ensure source and target are compatible
1700 * (both directories, or both not directories).
1701 */
1702 mode = tnode->nn_inode.i_mode;
1703 if ((mode & S_IFMT) == S_IFDIR) {
1704 if (!nandfs_dirempty(tvp, tdnode->nn_ino,
1705 tcnp->cn_cred)) {
1706 error = ENOTEMPTY;
1707 goto bad;
1708 }
1709 if (!doingdirectory) {
1710 error = ENOTDIR;
1711 goto bad;
1712 }
1713 /*
1714 * Update name cache since directory is going away.
1715 */
1716 cache_purge(tdvp);
1717 } else if (doingdirectory) {
1718 error = EISDIR;
1719 goto bad;
1720 }
1721
1722 DPRINTF(VNCALL, ("%s: update entry dvp:%p\n", __func__, tdvp));
1723 /*
1724 * Change name tcnp in tdvp to point at fvp.
1725 */
1726 error = nandfs_update_dirent(tdvp, fnode, tnode);
1727 if (error)
1728 goto bad;
1729
1730 if (doingdirectory && !newparent)
1731 tdnode->nn_inode.i_links_count--;
1732
1733 vput(tdvp);
1734
1735 tnode->nn_inode.i_links_count--;
1736 vput(tvp);
1737 tnode = NULL;
1738 }
1739
1740 /*
1741 * Unlink the source.
1742 */
1743 fcnp->cn_flags &= ~MODMASK;
1744 fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
1745 VREF(fdvp);
1746 error = relookup(fdvp, &fvp, fcnp);
1747 if (error == 0)
1748 vrele(fdvp);
1749 if (fvp != NULL) {
1750 fnode1 = VTON(fvp);
1751 fdnode = VTON(fdvp);
1752 } else {
1753 /*
1754 * From name has disappeared.
1755 */
1756 if (doingdirectory)
1757 panic("nandfs_rename: lost dir entry");
1758 vrele(ap->a_fvp);
1759 return (0);
1760 }
1761
1762 DPRINTF(VNCALL, ("%s: unlink source fnode:%p\n", __func__, fnode));
1763
1764 /*
1765 * Ensure that the directory entry still exists and has not
1766 * changed while the new name has been entered. If the source is
1767 * a file then the entry may have been unlinked or renamed. In
1768 * either case there is no further work to be done. If the source
1769 * is a directory then it cannot have been rmdir'ed; its link
1770 * count of three would cause a rmdir to fail with ENOTEMPTY.
1771 * The IN_RENAME flag ensures that it cannot be moved by another
1772 * rename.
1773 */
1774 if (fnode != fnode1) {
1775 if (doingdirectory)
1776 panic("nandfs: lost dir entry");
1777 } else {
1778 /*
1779 * If the source is a directory with a
1780 * new parent, the link count of the old
1781 * parent directory must be decremented
1782 * and ".." set to point to the new parent.
1783 */
1784 if (doingdirectory && newparent) {
1785 DPRINTF(VNCALL, ("%s: new parent %#jx -> %#jx\n",
1786 __func__, (uintmax_t) oldparent,
1787 (uintmax_t) newparent));
1788 error = nandfs_update_parent_dir(fvp, newparent);
1789 if (!error) {
1790 fdnode->nn_inode.i_links_count--;
1791 fdnode->nn_flags |= IN_CHANGE;
1792 }
1793 }
1794 error = nandfs_remove_dirent(fdvp, fnode, fcnp);
1795 if (!error) {
1796 fnode->nn_inode.i_links_count--;
1797 fnode->nn_flags |= IN_CHANGE;
1798 }
1799 fnode->nn_flags &= ~IN_RENAME;
1800 }
1801 if (fdnode)
1802 vput(fdvp);
1803 if (fnode)
1804 vput(fvp);
1805 vrele(ap->a_fvp);
1806 return (error);
1807
1808 bad:
1809 DPRINTF(VNCALL, ("%s: error:%d\n", __func__, error));
1810 if (tnode)
1811 vput(NTOV(tnode));
1812 vput(NTOV(tdnode));
1813 out:
1814 if (doingdirectory)
1815 fnode->nn_flags &= ~IN_RENAME;
1816 if (vn_lock(fvp, LK_EXCLUSIVE) == 0) {
1817 fnode->nn_inode.i_links_count--;
1818 fnode->nn_flags |= IN_CHANGE;
1819 fnode->nn_flags &= ~IN_RENAME;
1820 vput(fvp);
1821 } else
1822 vrele(fvp);
1823 return (error);
1824 }
1825
1826 static int
nandfs_mkdir(struct vop_mkdir_args * ap)1827 nandfs_mkdir(struct vop_mkdir_args *ap)
1828 {
1829 struct vnode *dvp = ap->a_dvp;
1830 struct vnode **vpp = ap->a_vpp;
1831 struct componentname *cnp = ap->a_cnp;
1832 struct nandfs_node *dir_node = VTON(dvp);
1833 struct nandfs_inode *dir_inode = &dir_node->nn_inode;
1834 struct nandfs_node *node;
1835 struct nandfsmount *nmp = dir_node->nn_nmp;
1836 uint16_t mode = MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode);
1837 int error;
1838
1839 DPRINTF(VNCALL, ("%s: dvp %p\n", __func__, dvp));
1840
1841 if (nandfs_fs_full(dir_node->nn_nandfsdev))
1842 return (ENOSPC);
1843
1844 if (dir_inode->i_links_count >= NANDFS_LINK_MAX)
1845 return (EMLINK);
1846
1847 error = nandfs_node_create(nmp, &node, mode);
1848 if (error)
1849 return (error);
1850
1851 node->nn_inode.i_gid = dir_node->nn_inode.i_gid;
1852 node->nn_inode.i_uid = cnp->cn_cred->cr_uid;
1853
1854 *vpp = NTOV(node);
1855
1856 error = nandfs_add_dirent(dvp, node->nn_ino, cnp->cn_nameptr,
1857 cnp->cn_namelen, IFTODT(mode));
1858 if (error) {
1859 vput(*vpp);
1860 return (error);
1861 }
1862
1863 dir_node->nn_inode.i_links_count++;
1864 dir_node->nn_flags |= IN_CHANGE;
1865
1866 error = nandfs_init_dir(NTOV(node), node->nn_ino, dir_node->nn_ino);
1867 if (error) {
1868 vput(NTOV(node));
1869 return (error);
1870 }
1871
1872 DPRINTF(VNCALL, ("created dir vp %p nandnode %p ino %jx\n", *vpp, node,
1873 (uintmax_t)node->nn_ino));
1874 return (0);
1875 }
1876
1877 static int
nandfs_mknod(struct vop_mknod_args * ap)1878 nandfs_mknod(struct vop_mknod_args *ap)
1879 {
1880 struct vnode *dvp = ap->a_dvp;
1881 struct vnode **vpp = ap->a_vpp;
1882 struct vattr *vap = ap->a_vap;
1883 uint16_t mode = MAKEIMODE(vap->va_type, vap->va_mode);
1884 struct componentname *cnp = ap->a_cnp;
1885 struct nandfs_node *dir_node = VTON(dvp);
1886 struct nandfsmount *nmp = dir_node->nn_nmp;
1887 struct nandfs_node *node;
1888 int error;
1889
1890 if (nandfs_fs_full(dir_node->nn_nandfsdev))
1891 return (ENOSPC);
1892
1893 error = nandfs_node_create(nmp, &node, mode);
1894 if (error)
1895 return (error);
1896 node->nn_inode.i_gid = dir_node->nn_inode.i_gid;
1897 node->nn_inode.i_uid = cnp->cn_cred->cr_uid;
1898 if (vap->va_rdev != VNOVAL)
1899 node->nn_inode.i_special = vap->va_rdev;
1900
1901 *vpp = NTOV(node);
1902
1903 if (nandfs_add_dirent(dvp, node->nn_ino, cnp->cn_nameptr,
1904 cnp->cn_namelen, IFTODT(mode))) {
1905 vput(*vpp);
1906 return (ENOTDIR);
1907 }
1908
1909 node->nn_flags |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1910
1911 return (0);
1912 }
1913
1914 static int
nandfs_symlink(struct vop_symlink_args * ap)1915 nandfs_symlink(struct vop_symlink_args *ap)
1916 {
1917 struct vnode **vpp = ap->a_vpp;
1918 struct vnode *dvp = ap->a_dvp;
1919 uint16_t mode = MAKEIMODE(ap->a_vap->va_type, ap->a_vap->va_mode);
1920 struct componentname *cnp = ap->a_cnp;
1921 struct nandfs_node *dir_node = VTON(dvp);
1922 struct nandfsmount *nmp = dir_node->nn_nmp;
1923 struct nandfs_node *node;
1924 int len, error;
1925
1926 if (nandfs_fs_full(dir_node->nn_nandfsdev))
1927 return (ENOSPC);
1928
1929 error = nandfs_node_create(nmp, &node, S_IFLNK | mode);
1930 if (error)
1931 return (error);
1932 node->nn_inode.i_gid = dir_node->nn_inode.i_gid;
1933 node->nn_inode.i_uid = cnp->cn_cred->cr_uid;
1934
1935 *vpp = NTOV(node);
1936
1937 if (nandfs_add_dirent(dvp, node->nn_ino, cnp->cn_nameptr,
1938 cnp->cn_namelen, IFTODT(mode))) {
1939 vput(*vpp);
1940 return (ENOTDIR);
1941 }
1942
1943
1944 len = strlen(ap->a_target);
1945 error = vn_rdwr(UIO_WRITE, *vpp, ap->a_target, len, (off_t)0,
1946 UIO_SYSSPACE, IO_NODELOCKED | IO_NOMACCHECK,
1947 cnp->cn_cred, NOCRED, NULL, NULL);
1948 if (error)
1949 vput(*vpp);
1950
1951 return (error);
1952 }
1953
1954 static int
nandfs_readlink(struct vop_readlink_args * ap)1955 nandfs_readlink(struct vop_readlink_args *ap)
1956 {
1957 struct vnode *vp = ap->a_vp;
1958
1959 return (VOP_READ(vp, ap->a_uio, 0, ap->a_cred));
1960 }
1961
1962 static int
nandfs_rmdir(struct vop_rmdir_args * ap)1963 nandfs_rmdir(struct vop_rmdir_args *ap)
1964 {
1965 struct vnode *vp = ap->a_vp;
1966 struct vnode *dvp = ap->a_dvp;
1967 struct componentname *cnp = ap->a_cnp;
1968 struct nandfs_node *node, *dnode;
1969 uint32_t dflag, flag;
1970 int error = 0;
1971
1972 node = VTON(vp);
1973 dnode = VTON(dvp);
1974
1975 /* Files marked as immutable or append-only cannot be deleted. */
1976 if ((node->nn_inode.i_flags & (IMMUTABLE | APPEND | NOUNLINK)) ||
1977 (dnode->nn_inode.i_flags & APPEND))
1978 return (EPERM);
1979
1980 DPRINTF(VNCALL, ("%s: dvp %p vp %p nandnode %p ino %#jx\n", __func__,
1981 dvp, vp, node, (uintmax_t)node->nn_ino));
1982
1983 if (node->nn_inode.i_links_count < 2)
1984 return (EINVAL);
1985
1986 if (!nandfs_dirempty(vp, dnode->nn_ino, cnp->cn_cred))
1987 return (ENOTEMPTY);
1988
1989 /* Files marked as immutable or append-only cannot be deleted. */
1990 dflag = dnode->nn_inode.i_flags;
1991 flag = node->nn_inode.i_flags;
1992 if ((dflag & APPEND) ||
1993 (flag & (NOUNLINK | IMMUTABLE | APPEND))) {
1994 return (EPERM);
1995 }
1996
1997 if (vp->v_mountedhere != 0)
1998 return (EINVAL);
1999
2000 nandfs_remove_dirent(dvp, node, cnp);
2001 dnode->nn_inode.i_links_count -= 1;
2002 dnode->nn_flags |= IN_CHANGE;
2003
2004 cache_purge(dvp);
2005
2006 error = nandfs_truncate(vp, (uint64_t)0);
2007 if (error)
2008 return (error);
2009
2010 node->nn_inode.i_links_count -= 2;
2011 node->nn_flags |= IN_CHANGE;
2012
2013 cache_purge(vp);
2014
2015 return (error);
2016 }
2017
2018 static int
nandfs_fsync(struct vop_fsync_args * ap)2019 nandfs_fsync(struct vop_fsync_args *ap)
2020 {
2021 struct vnode *vp = ap->a_vp;
2022 struct nandfs_node *node = VTON(vp);
2023 int locked;
2024
2025 DPRINTF(VNCALL, ("%s: vp %p nandnode %p ino %#jx\n", __func__, vp,
2026 node, (uintmax_t)node->nn_ino));
2027
2028 /*
2029 * Start syncing vnode only if inode was modified or
2030 * there are some dirty buffers
2031 */
2032 if (VTON(vp)->nn_flags & IN_MODIFIED ||
2033 vp->v_bufobj.bo_dirty.bv_cnt) {
2034 locked = VOP_ISLOCKED(vp);
2035 VOP_UNLOCK(vp, 0);
2036 nandfs_wakeup_wait_sync(node->nn_nandfsdev, SYNCER_FSYNC);
2037 VOP_LOCK(vp, locked | LK_RETRY);
2038 }
2039
2040 return (0);
2041 }
2042
2043 static int
nandfs_bmap(struct vop_bmap_args * ap)2044 nandfs_bmap(struct vop_bmap_args *ap)
2045 {
2046 struct vnode *vp = ap->a_vp;
2047 struct nandfs_node *nnode = VTON(vp);
2048 struct nandfs_device *nandfsdev = nnode->nn_nandfsdev;
2049 nandfs_daddr_t l2vmap, v2pmap;
2050 int error;
2051 int blk2dev = nandfsdev->nd_blocksize / DEV_BSIZE;
2052
2053 DPRINTF(VNCALL, ("%s: vp %p nandnode %p ino %#jx\n", __func__, vp,
2054 nnode, (uintmax_t)nnode->nn_ino));
2055
2056 if (ap->a_bop != NULL)
2057 *ap->a_bop = &nandfsdev->nd_devvp->v_bufobj;
2058 if (ap->a_bnp == NULL)
2059 return (0);
2060 if (ap->a_runp != NULL)
2061 *ap->a_runp = 0;
2062 if (ap->a_runb != NULL)
2063 *ap->a_runb = 0;
2064
2065 /*
2066 * Translate all the block sectors into a series of buffers to read
2067 * asynchronously from the nandfs device. Note that this lookup may
2068 * induce readin's too.
2069 */
2070
2071 /* Get virtual block numbers for the vnode's buffer span */
2072 error = nandfs_bmap_lookup(nnode, ap->a_bn, &l2vmap);
2073 if (error)
2074 return (-1);
2075
2076 /* Translate virtual block numbers to physical block numbers */
2077 error = nandfs_vtop(nnode, l2vmap, &v2pmap);
2078 if (error)
2079 return (-1);
2080
2081 /* Note virtual block 0 marks not mapped */
2082 if (l2vmap == 0)
2083 *ap->a_bnp = -1;
2084 else
2085 *ap->a_bnp = v2pmap * blk2dev; /* in DEV_BSIZE */
2086
2087 DPRINTF(VNCALL, ("%s: vp %p nandnode %p ino %#jx lblk %jx -> blk %jx\n",
2088 __func__, vp, nnode, (uintmax_t)nnode->nn_ino, (uintmax_t)ap->a_bn,
2089 (uintmax_t)*ap->a_bnp ));
2090
2091 return (0);
2092 }
2093
2094 static void
nandfs_force_syncer(struct nandfsmount * nmp)2095 nandfs_force_syncer(struct nandfsmount *nmp)
2096 {
2097
2098 nmp->nm_flags |= NANDFS_FORCE_SYNCER;
2099 nandfs_wakeup_wait_sync(nmp->nm_nandfsdev, SYNCER_FFORCE);
2100 }
2101
2102 static int
nandfs_ioctl(struct vop_ioctl_args * ap)2103 nandfs_ioctl(struct vop_ioctl_args *ap)
2104 {
2105 struct vnode *vp = ap->a_vp;
2106 u_long command = ap->a_command;
2107 caddr_t data = ap->a_data;
2108 struct nandfs_node *node = VTON(vp);
2109 struct nandfs_device *nandfsdev = node->nn_nandfsdev;
2110 struct nandfsmount *nmp = node->nn_nmp;
2111 uint64_t *tab, *cno;
2112 struct nandfs_seg_stat *nss;
2113 struct nandfs_cpmode *ncpm;
2114 struct nandfs_argv *nargv;
2115 struct nandfs_cpstat *ncp;
2116 int error;
2117
2118 DPRINTF(VNCALL, ("%s: %x\n", __func__, (uint32_t)command));
2119
2120 error = priv_check(ap->a_td, PRIV_VFS_MOUNT);
2121 if (error)
2122 return (error);
2123
2124 if (nmp->nm_ronly) {
2125 switch (command) {
2126 case NANDFS_IOCTL_GET_FSINFO:
2127 case NANDFS_IOCTL_GET_SUSTAT:
2128 case NANDFS_IOCTL_GET_CPINFO:
2129 case NANDFS_IOCTL_GET_CPSTAT:
2130 case NANDFS_IOCTL_GET_SUINFO:
2131 case NANDFS_IOCTL_GET_VINFO:
2132 case NANDFS_IOCTL_GET_BDESCS:
2133 break;
2134 default:
2135 return (EROFS);
2136 }
2137 }
2138
2139 switch (command) {
2140 case NANDFS_IOCTL_GET_FSINFO:
2141 error = nandfs_get_fsinfo(nmp, (struct nandfs_fsinfo *)data);
2142 break;
2143 case NANDFS_IOCTL_GET_SUSTAT:
2144 nss = (struct nandfs_seg_stat *)data;
2145 error = nandfs_get_seg_stat(nandfsdev, nss);
2146 break;
2147 case NANDFS_IOCTL_CHANGE_CPMODE:
2148 ncpm = (struct nandfs_cpmode *)data;
2149 error = nandfs_chng_cpmode(nandfsdev->nd_cp_node, ncpm);
2150 nandfs_force_syncer(nmp);
2151 break;
2152 case NANDFS_IOCTL_GET_CPINFO:
2153 nargv = (struct nandfs_argv *)data;
2154 error = nandfs_get_cpinfo_ioctl(nandfsdev->nd_cp_node, nargv);
2155 break;
2156 case NANDFS_IOCTL_DELETE_CP:
2157 tab = (uint64_t *)data;
2158 error = nandfs_delete_cp(nandfsdev->nd_cp_node, tab[0], tab[1]);
2159 nandfs_force_syncer(nmp);
2160 break;
2161 case NANDFS_IOCTL_GET_CPSTAT:
2162 ncp = (struct nandfs_cpstat *)data;
2163 error = nandfs_get_cpstat(nandfsdev->nd_cp_node, ncp);
2164 break;
2165 case NANDFS_IOCTL_GET_SUINFO:
2166 nargv = (struct nandfs_argv *)data;
2167 error = nandfs_get_segment_info_ioctl(nandfsdev, nargv);
2168 break;
2169 case NANDFS_IOCTL_GET_VINFO:
2170 nargv = (struct nandfs_argv *)data;
2171 error = nandfs_get_dat_vinfo_ioctl(nandfsdev, nargv);
2172 break;
2173 case NANDFS_IOCTL_GET_BDESCS:
2174 nargv = (struct nandfs_argv *)data;
2175 error = nandfs_get_dat_bdescs_ioctl(nandfsdev, nargv);
2176 break;
2177 case NANDFS_IOCTL_SYNC:
2178 cno = (uint64_t *)data;
2179 nandfs_force_syncer(nmp);
2180 *cno = nandfsdev->nd_last_cno;
2181 error = 0;
2182 break;
2183 case NANDFS_IOCTL_MAKE_SNAP:
2184 cno = (uint64_t *)data;
2185 error = nandfs_make_snap(nandfsdev, cno);
2186 nandfs_force_syncer(nmp);
2187 break;
2188 case NANDFS_IOCTL_DELETE_SNAP:
2189 cno = (uint64_t *)data;
2190 error = nandfs_delete_snap(nandfsdev, *cno);
2191 nandfs_force_syncer(nmp);
2192 break;
2193 default:
2194 error = ENOTTY;
2195 break;
2196 }
2197
2198 return (error);
2199 }
2200
2201 /*
2202 * Whiteout vnode call
2203 */
2204 static int
nandfs_whiteout(struct vop_whiteout_args * ap)2205 nandfs_whiteout(struct vop_whiteout_args *ap)
2206 {
2207 struct vnode *dvp = ap->a_dvp;
2208 struct componentname *cnp = ap->a_cnp;
2209 int error = 0;
2210
2211 switch (ap->a_flags) {
2212 case LOOKUP:
2213 return (0);
2214 case CREATE:
2215 /* Create a new directory whiteout */
2216 #ifdef INVARIANTS
2217 if ((cnp->cn_flags & SAVENAME) == 0)
2218 panic("nandfs_whiteout: missing name");
2219 #endif
2220 error = nandfs_add_dirent(dvp, NANDFS_WHT_INO, cnp->cn_nameptr,
2221 cnp->cn_namelen, DT_WHT);
2222 break;
2223
2224 case DELETE:
2225 /* Remove an existing directory whiteout */
2226 cnp->cn_flags &= ~DOWHITEOUT;
2227 error = nandfs_remove_dirent(dvp, NULL, cnp);
2228 break;
2229 default:
2230 panic("nandf_whiteout: unknown op: %d", ap->a_flags);
2231 }
2232
2233 return (error);
2234 }
2235
2236 static int
nandfs_pathconf(struct vop_pathconf_args * ap)2237 nandfs_pathconf(struct vop_pathconf_args *ap)
2238 {
2239 int error;
2240
2241 error = 0;
2242 switch (ap->a_name) {
2243 case _PC_LINK_MAX:
2244 *ap->a_retval = NANDFS_LINK_MAX;
2245 break;
2246 case _PC_NAME_MAX:
2247 *ap->a_retval = NANDFS_NAME_LEN;
2248 break;
2249 case _PC_PIPE_BUF:
2250 if (ap->a_vp->v_type == VDIR || ap->a_vp->v_type == VFIFO)
2251 *ap->a_retval = PIPE_BUF;
2252 else
2253 error = EINVAL;
2254 break;
2255 case _PC_CHOWN_RESTRICTED:
2256 *ap->a_retval = 1;
2257 break;
2258 case _PC_NO_TRUNC:
2259 *ap->a_retval = 1;
2260 break;
2261 case _PC_ALLOC_SIZE_MIN:
2262 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_bsize;
2263 break;
2264 case _PC_FILESIZEBITS:
2265 *ap->a_retval = 64;
2266 break;
2267 case _PC_REC_INCR_XFER_SIZE:
2268 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
2269 break;
2270 case _PC_REC_MAX_XFER_SIZE:
2271 *ap->a_retval = -1; /* means ``unlimited'' */
2272 break;
2273 case _PC_REC_MIN_XFER_SIZE:
2274 *ap->a_retval = ap->a_vp->v_mount->mnt_stat.f_iosize;
2275 break;
2276 default:
2277 error = vop_stdpathconf(ap);
2278 break;
2279 }
2280 return (error);
2281 }
2282
2283 static int
nandfs_vnlock1(struct vop_lock1_args * ap)2284 nandfs_vnlock1(struct vop_lock1_args *ap)
2285 {
2286 struct vnode *vp = ap->a_vp;
2287 struct nandfs_node *node = VTON(vp);
2288 int error, vi_locked;
2289
2290 /*
2291 * XXX can vnode go away while we are sleeping?
2292 */
2293 vi_locked = mtx_owned(&vp->v_interlock);
2294 if (vi_locked)
2295 VI_UNLOCK(vp);
2296 error = NANDFS_WRITELOCKFLAGS(node->nn_nandfsdev,
2297 ap->a_flags & LK_NOWAIT);
2298 if (vi_locked && !error)
2299 VI_LOCK(vp);
2300 if (error)
2301 return (error);
2302
2303 error = vop_stdlock(ap);
2304 if (error) {
2305 NANDFS_WRITEUNLOCK(node->nn_nandfsdev);
2306 return (error);
2307 }
2308
2309 return (0);
2310 }
2311
2312 static int
nandfs_vnunlock(struct vop_unlock_args * ap)2313 nandfs_vnunlock(struct vop_unlock_args *ap)
2314 {
2315 struct vnode *vp = ap->a_vp;
2316 struct nandfs_node *node = VTON(vp);
2317 int error;
2318
2319 error = vop_stdunlock(ap);
2320 if (error)
2321 return (error);
2322
2323 NANDFS_WRITEUNLOCK(node->nn_nandfsdev);
2324
2325 return (0);
2326 }
2327
2328 /*
2329 * Global vfs data structures
2330 */
2331 struct vop_vector nandfs_vnodeops = {
2332 .vop_default = &default_vnodeops,
2333 .vop_access = nandfs_access,
2334 .vop_advlock = nandfs_advlock,
2335 .vop_bmap = nandfs_bmap,
2336 .vop_close = nandfs_close,
2337 .vop_create = nandfs_create,
2338 .vop_fsync = nandfs_fsync,
2339 .vop_getattr = nandfs_getattr,
2340 .vop_inactive = nandfs_inactive,
2341 .vop_cachedlookup = nandfs_lookup,
2342 .vop_ioctl = nandfs_ioctl,
2343 .vop_link = nandfs_link,
2344 .vop_lookup = vfs_cache_lookup,
2345 .vop_mkdir = nandfs_mkdir,
2346 .vop_mknod = nandfs_mknod,
2347 .vop_open = nandfs_open,
2348 .vop_pathconf = nandfs_pathconf,
2349 .vop_print = nandfs_print,
2350 .vop_read = nandfs_read,
2351 .vop_readdir = nandfs_readdir,
2352 .vop_readlink = nandfs_readlink,
2353 .vop_reclaim = nandfs_reclaim,
2354 .vop_remove = nandfs_remove,
2355 .vop_rename = nandfs_rename,
2356 .vop_rmdir = nandfs_rmdir,
2357 .vop_whiteout = nandfs_whiteout,
2358 .vop_write = nandfs_write,
2359 .vop_setattr = nandfs_setattr,
2360 .vop_strategy = nandfs_strategy,
2361 .vop_symlink = nandfs_symlink,
2362 .vop_lock1 = nandfs_vnlock1,
2363 .vop_unlock = nandfs_vnunlock,
2364 };
2365
2366 struct vop_vector nandfs_system_vnodeops = {
2367 .vop_default = &default_vnodeops,
2368 .vop_close = nandfs_close,
2369 .vop_inactive = nandfs_inactive,
2370 .vop_reclaim = nandfs_reclaim,
2371 .vop_strategy = nandfs_strategy,
2372 .vop_fsync = nandfs_fsync,
2373 .vop_bmap = nandfs_bmap,
2374 .vop_access = VOP_PANIC,
2375 .vop_advlock = VOP_PANIC,
2376 .vop_create = VOP_PANIC,
2377 .vop_getattr = VOP_PANIC,
2378 .vop_cachedlookup = VOP_PANIC,
2379 .vop_ioctl = VOP_PANIC,
2380 .vop_link = VOP_PANIC,
2381 .vop_lookup = VOP_PANIC,
2382 .vop_mkdir = VOP_PANIC,
2383 .vop_mknod = VOP_PANIC,
2384 .vop_open = VOP_PANIC,
2385 .vop_pathconf = VOP_PANIC,
2386 .vop_print = VOP_PANIC,
2387 .vop_read = VOP_PANIC,
2388 .vop_readdir = VOP_PANIC,
2389 .vop_readlink = VOP_PANIC,
2390 .vop_remove = VOP_PANIC,
2391 .vop_rename = VOP_PANIC,
2392 .vop_rmdir = VOP_PANIC,
2393 .vop_whiteout = VOP_PANIC,
2394 .vop_write = VOP_PANIC,
2395 .vop_setattr = VOP_PANIC,
2396 .vop_symlink = VOP_PANIC,
2397 };
2398
2399 static int
nandfsfifo_close(struct vop_close_args * ap)2400 nandfsfifo_close(struct vop_close_args *ap)
2401 {
2402 struct vnode *vp = ap->a_vp;
2403 struct nandfs_node *node = VTON(vp);
2404
2405 DPRINTF(VNCALL, ("%s: vp %p node %p\n", __func__, vp, node));
2406
2407 mtx_lock(&vp->v_interlock);
2408 if (vp->v_usecount > 1)
2409 nandfs_itimes_locked(vp);
2410 mtx_unlock(&vp->v_interlock);
2411
2412 return (fifo_specops.vop_close(ap));
2413 }
2414
2415 struct vop_vector nandfs_fifoops = {
2416 .vop_default = &fifo_specops,
2417 .vop_fsync = VOP_PANIC,
2418 .vop_access = nandfs_access,
2419 .vop_close = nandfsfifo_close,
2420 .vop_getattr = nandfs_getattr,
2421 .vop_inactive = nandfs_inactive,
2422 .vop_pathconf = nandfs_pathconf,
2423 .vop_print = nandfs_print,
2424 .vop_read = VOP_PANIC,
2425 .vop_reclaim = nandfs_reclaim,
2426 .vop_setattr = nandfs_setattr,
2427 .vop_write = VOP_PANIC,
2428 .vop_lock1 = nandfs_vnlock1,
2429 .vop_unlock = nandfs_vnunlock,
2430 };
2431
2432 int
nandfs_vinit(struct vnode * vp,uint64_t ino)2433 nandfs_vinit(struct vnode *vp, uint64_t ino)
2434 {
2435 struct nandfs_node *node;
2436
2437 ASSERT_VOP_LOCKED(vp, __func__);
2438
2439 node = VTON(vp);
2440
2441 /* Check if we're fetching the root */
2442 if (ino == NANDFS_ROOT_INO)
2443 vp->v_vflag |= VV_ROOT;
2444
2445 if (ino != NANDFS_GC_INO)
2446 vp->v_type = IFTOVT(node->nn_inode.i_mode);
2447 else
2448 vp->v_type = VREG;
2449
2450 if (vp->v_type == VFIFO)
2451 vp->v_op = &nandfs_fifoops;
2452
2453 return (0);
2454 }
2455