kernel - Remove the last MP locks from tmpfs (2).
[dragonfly.git] / sys / vfs / tmpfs / tmpfs_vnops.c
CommitLineData
7a2de9a4
MD
1/*-
2 * Copyright (c) 2005, 2006 The NetBSD Foundation, Inc.
3 * All rights reserved.
4 *
5 * This code is derived from software contributed to The NetBSD Foundation
6 * by Julio M. Merino Vidal, developed as part of Google's Summer of Code
7 * 2005 program.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
19 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
22 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28 * POSSIBILITY OF SUCH DAMAGE.
80ae59d7
MD
29 *
30 * $NetBSD: tmpfs_vnops.c,v 1.39 2007/07/23 15:41:01 jmmv Exp $
7a2de9a4
MD
31 */
32
33/*
34 * tmpfs vnode interface.
35 */
7a2de9a4
MD
36
37#include <sys/kernel.h>
38#include <sys/kern_syscall.h>
39#include <sys/param.h>
40#include <sys/fcntl.h>
41#include <sys/lockf.h>
42#include <sys/priv.h>
43#include <sys/proc.h>
44#include <sys/resourcevar.h>
45#include <sys/sched.h>
7a2de9a4
MD
46#include <sys/stat.h>
47#include <sys/systm.h>
48#include <sys/unistd.h>
49#include <sys/vfsops.h>
50#include <sys/vnode.h>
51
7a2de9a4
MD
52#include <vm/vm.h>
53#include <vm/vm_object.h>
54#include <vm/vm_page.h>
55#include <vm/vm_pager.h>
b7545cb3 56#include <vm/swap_pager.h>
7a2de9a4
MD
57
58#include <vfs/fifofs/fifo.h>
59#include <vfs/tmpfs/tmpfs_vnops.h>
60#include <vfs/tmpfs/tmpfs.h>
61
62MALLOC_DECLARE(M_TMPFS);
63
80ae59d7
MD
64static __inline
65void
66tmpfs_knote(struct vnode *vp, int flags)
67{
68 if (flags)
69 KNOTE(&vp->v_pollinfo.vpi_kqinfo.ki_note, flags);
70}
71
72
7a2de9a4
MD
73/* --------------------------------------------------------------------- */
74
75static int
76tmpfs_nresolve(struct vop_nresolve_args *v)
77{
78 struct vnode *dvp = v->a_dvp;
79 struct vnode *vp = NULL;
80 struct namecache *ncp = v->a_nch->ncp;
d89ce96a 81 struct tmpfs_node *tnode;
7a2de9a4
MD
82
83 int error;
84 struct tmpfs_dirent *de;
85 struct tmpfs_node *dnode;
86
87 dnode = VP_TO_TMPFS_DIR(dvp);
88
7a2de9a4
MD
89 de = tmpfs_dir_lookup(dnode, NULL, ncp);
90 if (de == NULL) {
d89ce96a 91 error = ENOENT;
7a2de9a4 92 } else {
d89ce96a
MD
93 /*
94 * Allocate a vnode for the node we found.
95 */
7a2de9a4 96 tnode = de->td_node;
7a2de9a4 97 error = tmpfs_alloc_vp(dvp->v_mount, tnode,
d89ce96a
MD
98 LK_EXCLUSIVE | LK_RETRY, &vp);
99 if (error)
7a2de9a4 100 goto out;
d89ce96a 101 KKASSERT(vp);
7a2de9a4
MD
102 }
103
7a2de9a4 104out:
d89ce96a
MD
105 /*
106 * Store the result of this lookup in the cache. Avoid this if the
7a2de9a4 107 * request was for creation, as it does not improve timings on
d89ce96a
MD
108 * emprical tests.
109 */
7a2de9a4
MD
110 if (vp) {
111 vn_unlock(vp);
112 cache_setvp(v->a_nch, vp);
113 vrele(vp);
d89ce96a 114 } else if (error == ENOENT) {
7a2de9a4
MD
115 cache_setvp(v->a_nch, NULL);
116 }
117 return error;
118}
119
120static int
121tmpfs_nlookupdotdot(struct vop_nlookupdotdot_args *v)
122{
123 struct vnode *dvp = v->a_dvp;
124 struct vnode **vpp = v->a_vpp;
125 struct tmpfs_node *dnode = VP_TO_TMPFS_NODE(dvp);
126 struct ucred *cred = v->a_cred;
127 int error;
128
129 *vpp = NULL;
130 /* Check accessibility of requested node as a first step. */
131 error = VOP_ACCESS(dvp, VEXEC, cred);
132 if (error != 0)
133 return error;
134
135 if (dnode->tn_dir.tn_parent != NULL) {
136 /* Allocate a new vnode on the matching entry. */
137 error = tmpfs_alloc_vp(dvp->v_mount, dnode->tn_dir.tn_parent,
138 LK_EXCLUSIVE | LK_RETRY, vpp);
139
140 if (*vpp)
141 vn_unlock(*vpp);
142 }
143
144 return (*vpp == NULL) ? ENOENT : 0;
145}
146
147/* --------------------------------------------------------------------- */
148
149static int
150tmpfs_ncreate(struct vop_ncreate_args *v)
151{
152 struct vnode *dvp = v->a_dvp;
153 struct vnode **vpp = v->a_vpp;
154 struct namecache *ncp = v->a_nch->ncp;
155 struct vattr *vap = v->a_vap;
156 struct ucred *cred = v->a_cred;
157 int error;
158
159 KKASSERT(vap->va_type == VREG || vap->va_type == VSOCK);
160
7a2de9a4
MD
161 error = tmpfs_alloc_file(dvp, vpp, vap, ncp, cred, NULL);
162 if (error == 0) {
163 cache_setunresolved(v->a_nch);
164 cache_setvp(v->a_nch, *vpp);
80ae59d7 165 tmpfs_knote(dvp, NOTE_WRITE);
7a2de9a4 166 }
7a2de9a4
MD
167
168 return error;
169}
170/* --------------------------------------------------------------------- */
171
172static int
173tmpfs_nmknod(struct vop_nmknod_args *v)
174{
175 struct vnode *dvp = v->a_dvp;
176 struct vnode **vpp = v->a_vpp;
177 struct namecache *ncp = v->a_nch->ncp;
178 struct vattr *vap = v->a_vap;
179 struct ucred *cred = v->a_cred;
180 int error;
181
182 if (vap->va_type != VBLK && vap->va_type != VCHR &&
183 vap->va_type != VFIFO)
184 return EINVAL;
185
7a2de9a4
MD
186 error = tmpfs_alloc_file(dvp, vpp, vap, ncp, cred, NULL);
187 if (error == 0) {
188 cache_setunresolved(v->a_nch);
189 cache_setvp(v->a_nch, *vpp);
80ae59d7 190 tmpfs_knote(dvp, NOTE_WRITE);
7a2de9a4 191 }
7a2de9a4
MD
192
193 return error;
194}
195
196/* --------------------------------------------------------------------- */
197
198static int
199tmpfs_open(struct vop_open_args *v)
200{
201 struct vnode *vp = v->a_vp;
202 int mode = v->a_mode;
203
204 int error;
205 struct tmpfs_node *node;
206
7a2de9a4
MD
207 node = VP_TO_TMPFS_NODE(vp);
208
209 /* The file is still active but all its names have been removed
210 * (e.g. by a "rmdir $(pwd)"). It cannot be opened any more as
211 * it is about to die. */
212 if (node->tn_links < 1)
213 return (ENOENT);
214
215 /* If the file is marked append-only, deny write requests. */
630e3a33
MD
216 if ((node->tn_flags & APPEND) &&
217 (mode & (FWRITE | O_APPEND)) == FWRITE) {
7a2de9a4 218 error = EPERM;
630e3a33 219 } else {
7a2de9a4
MD
220 return (vop_stdopen(v));
221 }
7a2de9a4
MD
222 return error;
223}
224
225/* --------------------------------------------------------------------- */
226
227static int
228tmpfs_close(struct vop_close_args *v)
229{
230 struct vnode *vp = v->a_vp;
231 struct tmpfs_node *node;
232
233 node = VP_TO_TMPFS_NODE(vp);
234
235 if (node->tn_links > 0) {
236 /* Update node times. No need to do it if the node has
237 * been deleted, because it will vanish after we return. */
238 tmpfs_update(vp);
239 }
240
241 return vop_stdclose(v);
242}
243
244/* --------------------------------------------------------------------- */
245
246int
247tmpfs_access(struct vop_access_args *v)
248{
249 struct vnode *vp = v->a_vp;
250 int error;
251 struct tmpfs_node *node;
252
7a2de9a4
MD
253 node = VP_TO_TMPFS_NODE(vp);
254
255 switch (vp->v_type) {
256 case VDIR:
257 /* FALLTHROUGH */
258 case VLNK:
259 /* FALLTHROUGH */
260 case VREG:
5a9e9ac7 261 if ((v->a_mode & VWRITE) && (vp->v_mount->mnt_flag & MNT_RDONLY)) {
7a2de9a4
MD
262 error = EROFS;
263 goto out;
264 }
265 break;
266
267 case VBLK:
268 /* FALLTHROUGH */
269 case VCHR:
270 /* FALLTHROUGH */
271 case VSOCK:
272 /* FALLTHROUGH */
273 case VFIFO:
274 break;
275
276 default:
277 error = EINVAL;
278 goto out;
279 }
280
5a9e9ac7 281 if ((v->a_mode & VWRITE) && (node->tn_flags & IMMUTABLE)) {
7a2de9a4
MD
282 error = EPERM;
283 goto out;
284 }
285
286 error = vop_helper_access(v, node->tn_uid, node->tn_gid, node->tn_mode, 0);
287
288out:
289
290 return error;
291}
292
293/* --------------------------------------------------------------------- */
294
295int
296tmpfs_getattr(struct vop_getattr_args *v)
297{
298 struct vnode *vp = v->a_vp;
299 struct vattr *vap = v->a_vap;
7a2de9a4 300 struct tmpfs_node *node;
7a2de9a4
MD
301
302 node = VP_TO_TMPFS_NODE(vp);
303
e575e508 304 lwkt_gettoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
305 tmpfs_update(vp);
306
307 vap->va_type = vp->v_type;
308 vap->va_mode = node->tn_mode;
309 vap->va_nlink = node->tn_links;
310 vap->va_uid = node->tn_uid;
311 vap->va_gid = node->tn_gid;
312 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
313 vap->va_fileid = node->tn_id;
314 vap->va_size = node->tn_size;
315 vap->va_blocksize = PAGE_SIZE;
316 vap->va_atime.tv_sec = node->tn_atime;
317 vap->va_atime.tv_nsec = node->tn_atimensec;
318 vap->va_mtime.tv_sec = node->tn_mtime;
319 vap->va_mtime.tv_nsec = node->tn_mtimensec;
320 vap->va_ctime.tv_sec = node->tn_ctime;
321 vap->va_ctime.tv_nsec = node->tn_ctimensec;
322 vap->va_gen = node->tn_gen;
323 vap->va_flags = node->tn_flags;
324 if (vp->v_type == VBLK || vp->v_type == VCHR)
325 {
326 vap->va_rmajor = umajor(node->tn_rdev);
327 vap->va_rminor = uminor(node->tn_rdev);
328 }
329 vap->va_bytes = round_page(node->tn_size);
330 vap->va_filerev = 0;
331
e575e508
VS
332 lwkt_reltoken(&vp->v_mount->mnt_token);
333
7a2de9a4
MD
334 return 0;
335}
336
337/* --------------------------------------------------------------------- */
338
339int
340tmpfs_setattr(struct vop_setattr_args *v)
341{
342 struct vnode *vp = v->a_vp;
343 struct vattr *vap = v->a_vap;
344 struct ucred *cred = v->a_cred;
80ae59d7 345 struct tmpfs_node *node = VP_TO_TMPFS_NODE(vp);
7a2de9a4 346 int error = 0;
80ae59d7 347 int kflags = 0;
7a2de9a4 348
80ae59d7 349 if (error == 0 && (vap->va_flags != VNOVAL)) {
7a2de9a4 350 error = tmpfs_chflags(vp, vap->va_flags, cred);
80ae59d7
MD
351 kflags |= NOTE_ATTRIB;
352 }
7a2de9a4 353
80ae59d7
MD
354 if (error == 0 && (vap->va_size != VNOVAL)) {
355 if (vap->va_size > node->tn_size)
356 kflags |= NOTE_WRITE | NOTE_EXTEND;
357 else
358 kflags |= NOTE_WRITE;
7a2de9a4 359 error = tmpfs_chsize(vp, vap->va_size, cred);
80ae59d7 360 }
7a2de9a4 361
d89ce96a
MD
362 if (error == 0 && (vap->va_uid != (uid_t)VNOVAL ||
363 vap->va_gid != (gid_t)VNOVAL)) {
7a2de9a4 364 error = tmpfs_chown(vp, vap->va_uid, vap->va_gid, cred);
80ae59d7 365 kflags |= NOTE_ATTRIB;
d89ce96a 366 }
7a2de9a4 367
80ae59d7 368 if (error == 0 && (vap->va_mode != (mode_t)VNOVAL)) {
7a2de9a4 369 error = tmpfs_chmod(vp, vap->va_mode, cred);
80ae59d7
MD
370 kflags |= NOTE_ATTRIB;
371 }
7a2de9a4
MD
372
373 if (error == 0 && ((vap->va_atime.tv_sec != VNOVAL &&
374 vap->va_atime.tv_nsec != VNOVAL) ||
375 (vap->va_mtime.tv_sec != VNOVAL &&
d89ce96a 376 vap->va_mtime.tv_nsec != VNOVAL) )) {
7a2de9a4 377 error = tmpfs_chtimes(vp, &vap->va_atime, &vap->va_mtime,
d89ce96a 378 vap->va_vaflags, cred);
80ae59d7 379 kflags |= NOTE_ATTRIB;
d89ce96a 380 }
7a2de9a4
MD
381
382 /* Update the node times. We give preference to the error codes
383 * generated by this function rather than the ones that may arise
384 * from tmpfs_update. */
385 tmpfs_update(vp);
80ae59d7 386 tmpfs_knote(vp, kflags);
7a2de9a4 387
7a2de9a4
MD
388 return error;
389}
390
391/* --------------------------------------------------------------------- */
392
9fc94b5f 393/*
630e3a33
MD
394 * fsync is usually a NOP, but we must take action when unmounting or
395 * when recycling.
9fc94b5f 396 */
7a2de9a4
MD
397static int
398tmpfs_fsync(struct vop_fsync_args *v)
399{
9fc94b5f 400 struct tmpfs_mount *tmp;
630e3a33 401 struct tmpfs_node *node;
7a2de9a4
MD
402 struct vnode *vp = v->a_vp;
403
9fc94b5f 404 tmp = VFS_TO_TMPFS(vp->v_mount);
630e3a33
MD
405 node = VP_TO_TMPFS_NODE(vp);
406
407 tmpfs_update(vp);
408 if (vp->v_type == VREG) {
d4623db3 409 if (vp->v_flag & VRECLAIMED) {
630e3a33
MD
410 if (node->tn_links == 0)
411 tmpfs_truncate(vp, 0);
412 else
413 vfsync(v->a_vp, v->a_waitfor, 1, NULL, NULL);
414 }
9fc94b5f 415 }
7a2de9a4
MD
416 return 0;
417}
418
419/* --------------------------------------------------------------------- */
420
421static int
422tmpfs_read (struct vop_read_args *ap)
423{
424 struct buf *bp;
425 struct vnode *vp = ap->a_vp;
426 struct uio *uio = ap->a_uio;
427 struct tmpfs_node *node;
7a2de9a4 428 off_t base_offset;
9fc94b5f 429 size_t offset;
7a2de9a4 430 size_t len;
9fc94b5f 431 int error;
7a2de9a4
MD
432
433 error = 0;
434 if (uio->uio_resid == 0) {
435 return error;
436 }
437
438 node = VP_TO_TMPFS_NODE(vp);
439
440 if (uio->uio_offset < 0)
441 return (EINVAL);
442 if (vp->v_type != VREG)
443 return (EINVAL);
444
7a2de9a4
MD
445 while (uio->uio_resid > 0 && uio->uio_offset < node->tn_size) {
446 /*
447 * Use buffer cache I/O (via tmpfs_strategy)
448 */
9fc94b5f 449 offset = (size_t)uio->uio_offset & BMASK;
7a2de9a4 450 base_offset = (off_t)uio->uio_offset - offset;
72d6a027 451 bp = getcacheblk(vp, base_offset, BSIZE);
7a2de9a4
MD
452 if (bp == NULL)
453 {
b403e861 454 lwkt_gettoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
455 error = bread(vp, base_offset, BSIZE, &bp);
456 if (error) {
457 brelse(bp);
b403e861 458 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
459 kprintf("tmpfs_read bread error %d\n", error);
460 break;
461 }
b403e861 462 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
463 }
464
7a2de9a4
MD
465 /*
466 * Figure out how many bytes we can actually copy this loop.
467 */
468 len = BSIZE - offset;
469 if (len > uio->uio_resid)
470 len = uio->uio_resid;
471 if (len > node->tn_size - uio->uio_offset)
472 len = (size_t)(node->tn_size - uio->uio_offset);
473
474 error = uiomove((char *)bp->b_data + offset, len, uio);
475 bqrelse(bp);
476 if (error) {
477 kprintf("tmpfs_read uiomove error %d\n", error);
478 break;
479 }
480 }
481
7a2de9a4
MD
482 TMPFS_NODE_LOCK(node);
483 node->tn_status |= TMPFS_NODE_ACCESSED;
484 TMPFS_NODE_UNLOCK(node);
485
7a2de9a4
MD
486 return(error);
487}
488
489static int
490tmpfs_write (struct vop_write_args *ap)
491{
492 struct buf *bp;
493 struct vnode *vp = ap->a_vp;
494 struct uio *uio = ap->a_uio;
495 struct thread *td = uio->uio_td;
496 struct tmpfs_node *node;
497 boolean_t extended;
498 off_t oldsize;
499 int error;
7a2de9a4 500 off_t base_offset;
9fc94b5f 501 size_t offset;
7a2de9a4
MD
502 size_t len;
503 struct rlimit limit;
7a2de9a4 504 int trivial = 0;
80ae59d7 505 int kflags = 0;
7a2de9a4
MD
506
507 error = 0;
508 if (uio->uio_resid == 0) {
509 return error;
510 }
511
512 node = VP_TO_TMPFS_NODE(vp);
513
514 if (vp->v_type != VREG)
515 return (EINVAL);
516
1be4932c
VS
517 lwkt_gettoken(&vp->v_mount->mnt_token);
518
7a2de9a4
MD
519 oldsize = node->tn_size;
520 if (ap->a_ioflag & IO_APPEND)
521 uio->uio_offset = node->tn_size;
522
523 /*
524 * Check for illegal write offsets.
525 */
526 if (uio->uio_offset + uio->uio_resid >
1be4932c
VS
527 VFS_TO_TMPFS(vp->v_mount)->tm_maxfilesize) {
528 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4 529 return (EFBIG);
1be4932c 530 }
7a2de9a4
MD
531
532 if (vp->v_type == VREG && td != NULL) {
533 error = kern_getrlimit(RLIMIT_FSIZE, &limit);
1be4932c
VS
534 if (error != 0) {
535 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4 536 return error;
1be4932c 537 }
7a2de9a4
MD
538 if (uio->uio_offset + uio->uio_resid > limit.rlim_cur) {
539 ksignal(td->td_proc, SIGXFSZ);
1be4932c 540 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
541 return (EFBIG);
542 }
543 }
544
545
546 /*
547 * Extend the file's size if necessary
548 */
9fc94b5f 549 extended = ((uio->uio_offset + uio->uio_resid) > node->tn_size);
7a2de9a4 550
7a2de9a4
MD
551 while (uio->uio_resid > 0) {
552 /*
553 * Use buffer cache I/O (via tmpfs_strategy)
554 */
9fc94b5f 555 offset = (size_t)uio->uio_offset & BMASK;
7a2de9a4
MD
556 base_offset = (off_t)uio->uio_offset - offset;
557 len = BSIZE - offset;
558 if (len > uio->uio_resid)
559 len = uio->uio_resid;
560
561 if ((uio->uio_offset + len) > node->tn_size) {
9fc94b5f 562 trivial = (uio->uio_offset <= node->tn_size);
7a2de9a4
MD
563 error = tmpfs_reg_resize(vp, uio->uio_offset + len, trivial);
564 if (error)
565 break;
566 }
567
9fc94b5f
MD
568 /*
569 * Read to fill in any gaps. Theoretically we could
570 * optimize this if the write covers the entire buffer
571 * and is not a UIO_NOCOPY write, however this can lead
572 * to a security violation exposing random kernel memory
573 * (whatever junk was in the backing VM pages before).
574 *
575 * So just use bread() to do the right thing.
576 */
577 error = bread(vp, base_offset, BSIZE, &bp);
7a2de9a4
MD
578 error = uiomove((char *)bp->b_data + offset, len, uio);
579 if (error) {
580 kprintf("tmpfs_write uiomove error %d\n", error);
581 brelse(bp);
582 break;
583 }
584
80ae59d7 585 if (uio->uio_offset > node->tn_size) {
7a2de9a4 586 node->tn_size = uio->uio_offset;
80ae59d7
MD
587 kflags |= NOTE_EXTEND;
588 }
589 kflags |= NOTE_WRITE;
7a2de9a4
MD
590
591 /*
9fc94b5f 592 * The data has been loaded into the buffer, write it out.
7a2de9a4 593 *
9fc94b5f
MD
594 * We want tmpfs to be able to use all available ram, not
595 * just the buffer cache, so if not explicitly paging we
596 * use buwrite() to leave the buffer clean but mark all the
597 * VM pages valid+dirty.
7a2de9a4 598 *
d89ce96a
MD
599 * When the kernel is paging, either via normal pageout
600 * operation or when cleaning the object during a recycle,
601 * the underlying VM pages are going to get thrown away
602 * so we MUST write them to swap.
603 *
604 * XXX unfortunately this catches msync() system calls too
605 * for the moment.
7a2de9a4 606 */
b7545cb3
AH
607 if (vm_swap_size == 0) {
608 /*
609 * if swap isn't configured yet, force a buwrite() to
610 * avoid problems further down the line, due to flushing
611 * to swap.
612 */
9fc94b5f 613 buwrite(bp);
b7545cb3
AH
614 } else {
615 if (ap->a_ioflag & IO_SYNC) {
616 bwrite(bp);
617 } else if ((ap->a_ioflag & IO_ASYNC) ||
618 (uio->uio_segflg == UIO_NOCOPY)) {
619 bawrite(bp);
620 } else {
621 buwrite(bp);
622 }
d89ce96a 623 }
9fc94b5f 624
7a2de9a4 625 if (bp->b_error) {
2cd8c774 626 kprintf("tmpfs_write bwrite error %d\n", bp->b_error);
7a2de9a4
MD
627 break;
628 }
629 }
7a2de9a4 630
7a2de9a4 631 if (error) {
80ae59d7 632 if (extended) {
7a2de9a4 633 (void)tmpfs_reg_resize(vp, oldsize, trivial);
80ae59d7
MD
634 kflags &= ~NOTE_EXTEND;
635 }
636 goto done;
7a2de9a4
MD
637 }
638
639 TMPFS_NODE_LOCK(node);
640 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_MODIFIED |
641 (extended? TMPFS_NODE_CHANGED : 0);
642
643 if (node->tn_mode & (S_ISUID | S_ISGID)) {
644 if (priv_check_cred(ap->a_cred, PRIV_VFS_RETAINSUGID, 0))
645 node->tn_mode &= ~(S_ISUID | S_ISGID);
646 }
647 TMPFS_NODE_UNLOCK(node);
80ae59d7 648done:
7a2de9a4 649
80ae59d7 650 tmpfs_knote(vp, kflags);
1be4932c
VS
651
652
653 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
654 return(error);
655}
656
657static int
658tmpfs_advlock (struct vop_advlock_args *ap)
659{
660 struct tmpfs_node *node;
661 struct vnode *vp = ap->a_vp;
662
663 node = VP_TO_TMPFS_NODE(vp);
664
665 return (lf_advlock(ap, &node->tn_advlock, node->tn_size));
666}
667
7a2de9a4
MD
668static int
669tmpfs_strategy(struct vop_strategy_args *ap)
670{
671 struct bio *bio = ap->a_bio;
9fc94b5f 672 struct buf *bp = bio->bio_buf;
7a2de9a4
MD
673 struct vnode *vp = ap->a_vp;
674 struct tmpfs_node *node;
675 vm_object_t uobj;
676
9fc94b5f
MD
677 if (vp->v_type != VREG) {
678 bp->b_resid = bp->b_bcount;
679 bp->b_flags |= B_ERROR | B_INVAL;
680 bp->b_error = EINVAL;
681 biodone(bio);
682 return(0);
683 }
7a2de9a4 684
d1f61aa2 685 lwkt_gettoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
686 node = VP_TO_TMPFS_NODE(vp);
687
688 uobj = node->tn_reg.tn_aobj;
9fc94b5f 689
7a2de9a4 690 /*
9fc94b5f
MD
691 * Call swap_pager_strategy to read or write between the VM
692 * object and the buffer cache.
7a2de9a4
MD
693 */
694 swap_pager_strategy(uobj, bio);
695
d1f61aa2 696 lwkt_reltoken(&vp->v_mount->mnt_token);
7a2de9a4
MD
697 return 0;
698}
699
700static int
701tmpfs_bmap(struct vop_bmap_args *ap)
702{
703 if (ap->a_doffsetp != NULL)
704 *ap->a_doffsetp = ap->a_loffset;
705 if (ap->a_runp != NULL)
706 *ap->a_runp = 0;
707 if (ap->a_runb != NULL)
708 *ap->a_runb = 0;
709
710 return 0;
711}
9fc94b5f 712
7a2de9a4
MD
713/* --------------------------------------------------------------------- */
714
715static int
716tmpfs_nremove(struct vop_nremove_args *v)
717{
718 struct vnode *dvp = v->a_dvp;
719 struct namecache *ncp = v->a_nch->ncp;
9fc94b5f 720 struct vnode *vp;
7a2de9a4
MD
721 int error;
722 struct tmpfs_dirent *de;
723 struct tmpfs_mount *tmp;
724 struct tmpfs_node *dnode;
725 struct tmpfs_node *node;
726
9fc94b5f
MD
727 /*
728 * We have to acquire the vp from v->a_nch because
729 * we will likely unresolve the namecache entry, and
730 * a vrele is needed to trigger the tmpfs_inactive/tmpfs_reclaim
731 * sequence to recover space from the file.
732 */
733 error = cache_vref(v->a_nch, v->a_cred, &vp);
734 KKASSERT(error == 0);
7a2de9a4
MD
735
736 if (vp->v_type == VDIR) {
737 error = EISDIR;
738 goto out;
739 }
740
741 dnode = VP_TO_TMPFS_DIR(dvp);
742 node = VP_TO_TMPFS_NODE(vp);
743 tmp = VFS_TO_TMPFS(vp->v_mount);
744 de = tmpfs_dir_lookup(dnode, node, ncp);
745 if (de == NULL) {
746 error = ENOENT;
747 goto out;
748 }
749
750 /* Files marked as immutable or append-only cannot be deleted. */
751 if ((node->tn_flags & (IMMUTABLE | APPEND | NOUNLINK)) ||
752 (dnode->tn_flags & APPEND)) {
753 error = EPERM;
754 goto out;
755 }
756
757 /* Remove the entry from the directory; as it is a file, we do not
758 * have to change the number of hard links of the directory. */
22d3b394 759 tmpfs_dir_detach(dnode, de);
7a2de9a4
MD
760
761 /* Free the directory entry we just deleted. Note that the node
762 * referred by it will not be removed until the vnode is really
763 * reclaimed. */
0786baf1 764 tmpfs_free_dirent(tmp, de);
7a2de9a4
MD
765
766 if (node->tn_links > 0) {
767 TMPFS_NODE_LOCK(node);
768 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \
769 TMPFS_NODE_MODIFIED;
770 TMPFS_NODE_UNLOCK(node);
771 }
772
773 cache_setunresolved(v->a_nch);
774 cache_setvp(v->a_nch, NULL);
80ae59d7 775 tmpfs_knote(vp, NOTE_DELETE);
9fc94b5f 776 /*cache_inval_vp(vp, CINV_DESTROY);*/
80ae59d7 777 tmpfs_knote(dvp, NOTE_WRITE);
7a2de9a4
MD
778 error = 0;
779
7a2de9a4 780out:
9fc94b5f 781 vrele(vp);
7a2de9a4
MD
782
783 return error;
784}
785
786/* --------------------------------------------------------------------- */
787
788static int
789tmpfs_nlink(struct vop_nlink_args *v)
790{
791 struct vnode *dvp = v->a_dvp;
792 struct vnode *vp = v->a_vp;
793 struct namecache *ncp = v->a_nch->ncp;
7a2de9a4
MD
794 struct tmpfs_dirent *de;
795 struct tmpfs_node *node;
22d3b394
MD
796 struct tmpfs_node *dnode;
797 int error;
7a2de9a4 798
7a2de9a4
MD
799 KKASSERT(dvp != vp); /* XXX When can this be false? */
800
801 node = VP_TO_TMPFS_NODE(vp);
22d3b394 802 dnode = VP_TO_TMPFS_NODE(dvp);
7a2de9a4
MD
803
804 /* XXX: Why aren't the following two tests done by the caller? */
805
806 /* Hard links of directories are forbidden. */
807 if (vp->v_type == VDIR) {
808 error = EPERM;
809 goto out;
810 }
811
812 /* Cannot create cross-device links. */
813 if (dvp->v_mount != vp->v_mount) {
814 error = EXDEV;
815 goto out;
816 }
817
818 /* Ensure that we do not overflow the maximum number of links imposed
819 * by the system. */
820 KKASSERT(node->tn_links <= LINK_MAX);
821 if (node->tn_links == LINK_MAX) {
822 error = EMLINK;
823 goto out;
824 }
825
826 /* We cannot create links of files marked immutable or append-only. */
827 if (node->tn_flags & (IMMUTABLE | APPEND)) {
828 error = EPERM;
829 goto out;
830 }
831
832 /* Allocate a new directory entry to represent the node. */
833 error = tmpfs_alloc_dirent(VFS_TO_TMPFS(vp->v_mount), node,
834 ncp->nc_name, ncp->nc_nlen, &de);
835 if (error != 0)
836 goto out;
837
838 /* Insert the new directory entry into the appropriate directory. */
22d3b394 839 tmpfs_dir_attach(dnode, de);
7a2de9a4
MD
840
841 /* vp link count has changed, so update node times. */
842
843 TMPFS_NODE_LOCK(node);
844 node->tn_status |= TMPFS_NODE_CHANGED;
845 TMPFS_NODE_UNLOCK(node);
846 tmpfs_update(vp);
847
80ae59d7 848 tmpfs_knote(vp, NOTE_LINK);
7a2de9a4
MD
849 cache_setunresolved(v->a_nch);
850 cache_setvp(v->a_nch, vp);
80ae59d7 851 tmpfs_knote(dvp, NOTE_WRITE);
7a2de9a4
MD
852 error = 0;
853
854out:
7a2de9a4
MD
855 return error;
856}
857
858/* --------------------------------------------------------------------- */
859
860static int
861tmpfs_nrename(struct vop_nrename_args *v)
862{
863 struct vnode *fdvp = v->a_fdvp;
864 struct namecache *fncp = v->a_fnch->ncp;
865 struct vnode *fvp = fncp->nc_vp;
866 struct vnode *tdvp = v->a_tdvp;
867 struct namecache *tncp = v->a_tnch->ncp;
868 struct vnode *tvp = tncp->nc_vp;
7a2de9a4
MD
869 struct tmpfs_dirent *de;
870 struct tmpfs_mount *tmp;
871 struct tmpfs_node *fdnode;
872 struct tmpfs_node *fnode;
873 struct tmpfs_node *tnode;
874 struct tmpfs_node *tdnode;
22d3b394 875 char *newname;
dca262fb 876 char *oldname;
22d3b394 877 int error;
7a2de9a4 878
7a2de9a4
MD
879 tnode = (tvp == NULL) ? NULL : VP_TO_TMPFS_NODE(tvp);
880
881 /* Disallow cross-device renames.
882 * XXX Why isn't this done by the caller? */
883 if (fvp->v_mount != tdvp->v_mount ||
884 (tvp != NULL && fvp->v_mount != tvp->v_mount)) {
885 error = EXDEV;
886 goto out;
887 }
888
889 tmp = VFS_TO_TMPFS(tdvp->v_mount);
890 tdnode = VP_TO_TMPFS_DIR(tdvp);
891
892 /* If source and target are the same file, there is nothing to do. */
893 if (fvp == tvp) {
894 error = 0;
895 goto out;
896 }
897
7a2de9a4
MD
898 fdnode = VP_TO_TMPFS_DIR(fdvp);
899 fnode = VP_TO_TMPFS_NODE(fvp);
900 de = tmpfs_dir_lookup(fdnode, fnode, fncp);
901
902 /* Avoid manipulating '.' and '..' entries. */
903 if (de == NULL) {
904 error = ENOENT;
905 goto out_locked;
906 }
907 KKASSERT(de->td_node == fnode);
908
dca262fb
MD
909 /*
910 * If replacing an entry in the target directory and that entry
911 * is a directory, it must be empty.
912 *
7a2de9a4 913 * Kern_rename gurantees the destination to be a directory
dca262fb
MD
914 * if the source is one (it does?).
915 */
7a2de9a4
MD
916 if (tvp != NULL) {
917 KKASSERT(tnode != NULL);
918
919 if ((tnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
920 (tdnode->tn_flags & (APPEND | IMMUTABLE))) {
921 error = EPERM;
922 goto out_locked;
923 }
924
925 if (fnode->tn_type == VDIR && tnode->tn_type == VDIR) {
926 if (tnode->tn_size > 0) {
927 error = ENOTEMPTY;
928 goto out_locked;
929 }
930 } else if (fnode->tn_type == VDIR && tnode->tn_type != VDIR) {
931 error = ENOTDIR;
932 goto out_locked;
933 } else if (fnode->tn_type != VDIR && tnode->tn_type == VDIR) {
934 error = EISDIR;
935 goto out_locked;
936 } else {
937 KKASSERT(fnode->tn_type != VDIR &&
938 tnode->tn_type != VDIR);
939 }
940 }
941
dca262fb
MD
942 if ((fnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
943 (fdnode->tn_flags & (APPEND | IMMUTABLE))) {
7a2de9a4
MD
944 error = EPERM;
945 goto out_locked;
946 }
947
dca262fb
MD
948 /*
949 * Ensure that we have enough memory to hold the new name, if it
950 * has to be changed.
951 */
7a2de9a4
MD
952 if (fncp->nc_nlen != tncp->nc_nlen ||
953 bcmp(fncp->nc_name, tncp->nc_name, fncp->nc_nlen) != 0) {
d00cd01c 954 newname = kmalloc(tncp->nc_nlen + 1, tmp->tm_name_zone,
42f6f6b1
VS
955 M_WAITOK | M_NULLOK);
956 if (newname == NULL) {
957 error = ENOSPC;
958 goto out_locked;
959 }
dca262fb
MD
960 bcopy(tncp->nc_name, newname, tncp->nc_nlen);
961 newname[tncp->nc_nlen] = '\0';
962 } else {
7a2de9a4 963 newname = NULL;
dca262fb 964 }
7a2de9a4 965
dca262fb
MD
966 /*
967 * Unlink entry from source directory. Note that the kernel has
968 * already checked for illegal recursion cases (renaming a directory
969 * into a subdirectory of itself).
970 */
971 if (fdnode != tdnode)
972 tmpfs_dir_detach(fdnode, de);
973
974 /*
975 * Handle any name change. Swap with newname, we will
976 * deallocate it at the end.
977 */
978 if (newname != NULL) {
979#if 0
980 TMPFS_NODE_LOCK(fnode);
981 fnode->tn_status |= TMPFS_NODE_CHANGED;
982 TMPFS_NODE_UNLOCK(fnode);
983#endif
984 oldname = de->td_name;
985 de->td_name = newname;
986 de->td_namelen = (uint16_t)tncp->nc_nlen;
987 newname = oldname;
988 }
989
990 /*
991 * Link entry to target directory. If the entry
992 * represents a directory move the parent linkage
993 * as well.
994 */
7a2de9a4 995 if (fdnode != tdnode) {
7a2de9a4 996 if (de->td_node->tn_type == VDIR) {
7a2de9a4 997 TMPFS_VALIDATE_DIR(fnode);
7a2de9a4 998
7a2de9a4 999 TMPFS_NODE_LOCK(tdnode);
7a2de9a4 1000 tdnode->tn_links++;
dca262fb
MD
1001 tdnode->tn_status |= TMPFS_NODE_MODIFIED;
1002 TMPFS_NODE_UNLOCK(tdnode);
7a2de9a4 1003
dca262fb
MD
1004 TMPFS_NODE_LOCK(fnode);
1005 fnode->tn_dir.tn_parent = tdnode;
1006 fnode->tn_status |= TMPFS_NODE_CHANGED;
1007 TMPFS_NODE_UNLOCK(fnode);
1008
1009 TMPFS_NODE_LOCK(fdnode);
1010 fdnode->tn_links--;
1011 fdnode->tn_status |= TMPFS_NODE_MODIFIED;
7a2de9a4 1012 TMPFS_NODE_UNLOCK(fdnode);
7a2de9a4 1013 }
22d3b394 1014 tmpfs_dir_attach(tdnode, de);
dca262fb 1015 } else {
7a2de9a4 1016 TMPFS_NODE_LOCK(tdnode);
7a2de9a4 1017 tdnode->tn_status |= TMPFS_NODE_MODIFIED;
7a2de9a4
MD
1018 TMPFS_NODE_UNLOCK(tdnode);
1019 }
1020
dca262fb
MD
1021 /*
1022 * If we are overwriting an entry, we have to remove the old one
1023 * from the target directory.
1024 */
7a2de9a4
MD
1025 if (tvp != NULL) {
1026 /* Remove the old entry from the target directory. */
1027 de = tmpfs_dir_lookup(tdnode, tnode, tncp);
22d3b394 1028 tmpfs_dir_detach(tdnode, de);
80ae59d7 1029 tmpfs_knote(tdnode->tn_vnode, NOTE_DELETE);
7a2de9a4 1030
dca262fb
MD
1031 /*
1032 * Free the directory entry we just deleted. Note that the
7a2de9a4 1033 * node referred by it will not be removed until the vnode is
dca262fb
MD
1034 * really reclaimed.
1035 */
0786baf1 1036 tmpfs_free_dirent(VFS_TO_TMPFS(tvp->v_mount), de);
9fc94b5f 1037 /*cache_inval_vp(tvp, CINV_DESTROY);*/
7a2de9a4
MD
1038 }
1039
dca262fb
MD
1040 /*
1041 * Finish up
1042 */
1043 if (newname) {
d00cd01c 1044 kfree(newname, tmp->tm_name_zone);
dca262fb
MD
1045 newname = NULL;
1046 }
7a2de9a4 1047 cache_rename(v->a_fnch, v->a_tnch);
80ae59d7
MD
1048 tmpfs_knote(v->a_fdvp, NOTE_WRITE);
1049 tmpfs_knote(v->a_tdvp, NOTE_WRITE);
1050 if (fnode->tn_vnode)
1051 tmpfs_knote(fnode->tn_vnode, NOTE_RENAME);
7a2de9a4
MD
1052 error = 0;
1053
1054out_locked:
630e3a33 1055 ;
7a2de9a4
MD
1056
1057out:
1058 /* Release target nodes. */
1059 /* XXX: I don't understand when tdvp can be the same as tvp, but
1060 * other code takes care of this... */
1061 if (tdvp == tvp)
1062 vrele(tdvp);
7a2de9a4
MD
1063
1064 return error;
1065}
1066
1067/* --------------------------------------------------------------------- */
1068
1069static int
1070tmpfs_nmkdir(struct vop_nmkdir_args *v)
1071{
1072 struct vnode *dvp = v->a_dvp;
1073 struct vnode **vpp = v->a_vpp;
1074 struct namecache *ncp = v->a_nch->ncp;
1075 struct vattr *vap = v->a_vap;
1076 struct ucred *cred = v->a_cred;
1077 int error;
1078
1079 KKASSERT(vap->va_type == VDIR);
1080
7a2de9a4
MD
1081 error = tmpfs_alloc_file(dvp, vpp, vap, ncp, cred, NULL);
1082 if (error == 0) {
1083 cache_setunresolved(v->a_nch);
1084 cache_setvp(v->a_nch, *vpp);
80ae59d7 1085 tmpfs_knote(dvp, NOTE_WRITE | NOTE_LINK);
7a2de9a4 1086 }
7a2de9a4
MD
1087
1088 return error;
1089}
1090
1091/* --------------------------------------------------------------------- */
1092
1093static int
1094tmpfs_nrmdir(struct vop_nrmdir_args *v)
1095{
1096 struct vnode *dvp = v->a_dvp;
1097 struct namecache *ncp = v->a_nch->ncp;
9fc94b5f 1098 struct vnode *vp;
7a2de9a4
MD
1099 struct tmpfs_dirent *de;
1100 struct tmpfs_mount *tmp;
1101 struct tmpfs_node *dnode;
1102 struct tmpfs_node *node;
38e5e604
MD
1103 int error;
1104
1105 /*
9fc94b5f
MD
1106 * We have to acquire the vp from v->a_nch because
1107 * we will likely unresolve the namecache entry, and
1108 * a vrele is needed to trigger the tmpfs_inactive/tmpfs_reclaim
1109 * sequence.
1110 */
1111 error = cache_vref(v->a_nch, v->a_cred, &vp);
1112 KKASSERT(error == 0);
7a2de9a4 1113
e527fb6b
MD
1114 /*
1115 * Prevalidate so we don't hit an assertion later
1116 */
1117 if (vp->v_type != VDIR) {
1118 error = ENOTDIR;
1119 goto out;
1120 }
1121
7a2de9a4
MD
1122 tmp = VFS_TO_TMPFS(dvp->v_mount);
1123 dnode = VP_TO_TMPFS_DIR(dvp);
1124 node = VP_TO_TMPFS_DIR(vp);
1125
1126 /* Directories with more than two entries ('.' and '..') cannot be
1127 * removed. */
1128 if (node->tn_size > 0) {
1129 error = ENOTEMPTY;
1130 goto out;
1131 }
1132
1133 if ((dnode->tn_flags & APPEND)
1134 || (node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND))) {
1135 error = EPERM;
1136 goto out;
1137 }
1138
1139 /* This invariant holds only if we are not trying to remove "..".
1140 * We checked for that above so this is safe now. */
1141 KKASSERT(node->tn_dir.tn_parent == dnode);
1142
1143 /* Get the directory entry associated with node (vp). This was
1144 * filled by tmpfs_lookup while looking up the entry. */
1145 de = tmpfs_dir_lookup(dnode, node, ncp);
1146 KKASSERT(TMPFS_DIRENT_MATCHES(de,
1147 ncp->nc_name,
1148 ncp->nc_nlen));
1149
1150 /* Check flags to see if we are allowed to remove the directory. */
b7fe63af
MD
1151 if ((dnode->tn_flags & APPEND) ||
1152 node->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) {
7a2de9a4
MD
1153 error = EPERM;
1154 goto out;
1155 }
1156
1157
1158 /* Detach the directory entry from the directory (dnode). */
22d3b394 1159 tmpfs_dir_detach(dnode, de);
7a2de9a4
MD
1160
1161 /* No vnode should be allocated for this entry from this point */
1162 TMPFS_NODE_LOCK(node);
1163 TMPFS_ASSERT_ELOCKED(node);
1164 TMPFS_NODE_LOCK(dnode);
1165 TMPFS_ASSERT_ELOCKED(dnode);
1166
0786baf1
MD
1167#if 0
1168 /* handled by tmpfs_free_node */
1169 KKASSERT(node->tn_links > 0);
7a2de9a4
MD
1170 node->tn_links--;
1171 node->tn_dir.tn_parent = NULL;
0786baf1 1172#endif
7a2de9a4
MD
1173 node->tn_status |= TMPFS_NODE_ACCESSED | TMPFS_NODE_CHANGED | \
1174 TMPFS_NODE_MODIFIED;
1175
0786baf1
MD
1176#if 0
1177 /* handled by tmpfs_free_node */
1178 KKASSERT(dnode->tn_links > 0);
7a2de9a4 1179 dnode->tn_links--;
0786baf1 1180#endif
7a2de9a4
MD
1181 dnode->tn_status |= TMPFS_NODE_ACCESSED | \
1182 TMPFS_NODE_CHANGED | TMPFS_NODE_MODIFIED;
1183
1184 TMPFS_NODE_UNLOCK(dnode);
1185 TMPFS_NODE_UNLOCK(node);
1186
1187 /* Free the directory entry we just deleted. Note that the node
1188 * referred by it will not be removed until the vnode is really
1189 * reclaimed. */
0786baf1 1190 tmpfs_free_dirent(tmp, de);
7a2de9a4
MD
1191
1192 /* Release the deleted vnode (will destroy the node, notify
1193 * interested parties and clean it from the cache). */
1194
1195 TMPFS_NODE_LOCK(dnode);
1196 dnode->tn_status |= TMPFS_NODE_CHANGED;
1197 TMPFS_NODE_UNLOCK(dnode);
1198 tmpfs_update(dvp);
1199
1200 cache_setunresolved(v->a_nch);
1201 cache_setvp(v->a_nch, NULL);
9fc94b5f 1202 /*cache_inval_vp(vp, CINV_DESTROY);*/
80ae59d7 1203 tmpfs_knote(dvp, NOTE_WRITE | NOTE_LINK);
7a2de9a4
MD
1204 error = 0;
1205
1206out:
9fc94b5f 1207 vrele(vp);
7a2de9a4
MD
1208
1209 return error;
1210}
1211
1212/* --------------------------------------------------------------------- */
1213
1214static int
1215tmpfs_nsymlink(struct vop_nsymlink_args *v)
1216{
1217 struct vnode *dvp = v->a_dvp;
1218 struct vnode **vpp = v->a_vpp;
1219 struct namecache *ncp = v->a_nch->ncp;
1220 struct vattr *vap = v->a_vap;
1221 struct ucred *cred = v->a_cred;
1222 char *target = v->a_target;
1223 int error;
1224
7a2de9a4
MD
1225 vap->va_type = VLNK;
1226 error = tmpfs_alloc_file(dvp, vpp, vap, ncp, cred, target);
1227 if (error == 0) {
80ae59d7 1228 tmpfs_knote(*vpp, NOTE_WRITE);
7a2de9a4
MD
1229 cache_setunresolved(v->a_nch);
1230 cache_setvp(v->a_nch, *vpp);
1231 }
7a2de9a4
MD
1232
1233 return error;
1234}
1235
1236/* --------------------------------------------------------------------- */
1237
1238static int
1239tmpfs_readdir(struct vop_readdir_args *v)
1240{
1241 struct vnode *vp = v->a_vp;
1242 struct uio *uio = v->a_uio;
1243 int *eofflag = v->a_eofflag;
1244 off_t **cookies = v->a_cookies;
1245 int *ncookies = v->a_ncookies;
22d3b394 1246 struct tmpfs_mount *tmp;
7a2de9a4
MD
1247 int error;
1248 off_t startoff;
1249 off_t cnt = 0;
1250 struct tmpfs_node *node;
1251
1252 /* This operation only makes sense on directory nodes. */
1253 if (vp->v_type != VDIR)
1254 return ENOTDIR;
1255
22d3b394 1256 tmp = VFS_TO_TMPFS(vp->v_mount);
7a2de9a4
MD
1257 node = VP_TO_TMPFS_DIR(vp);
1258 startoff = uio->uio_offset;
1259
1260 if (uio->uio_offset == TMPFS_DIRCOOKIE_DOT) {
1261 error = tmpfs_dir_getdotdent(node, uio);
1262 if (error != 0)
1263 goto outok;
1264 cnt++;
1265 }
1266
1267 if (uio->uio_offset == TMPFS_DIRCOOKIE_DOTDOT) {
22d3b394 1268 error = tmpfs_dir_getdotdotdent(tmp, node, uio);
7a2de9a4
MD
1269 if (error != 0)
1270 goto outok;
1271 cnt++;
1272 }
1273
1274 error = tmpfs_dir_getdents(node, uio, &cnt);
1275
1276outok:
1277 KKASSERT(error >= -1);
1278
1279 if (error == -1)
1280 error = 0;
1281
1282 if (eofflag != NULL)
1283 *eofflag =
1284 (error == 0 && uio->uio_offset == TMPFS_DIRCOOKIE_EOF);
1285
1286 /* Update NFS-related variables. */
1287 if (error == 0 && cookies != NULL && ncookies != NULL) {
1288 off_t i;
1289 off_t off = startoff;
1290 struct tmpfs_dirent *de = NULL;
1291
1292 *ncookies = cnt;
1293 *cookies = kmalloc(cnt * sizeof(off_t), M_TEMP, M_WAITOK);
1294
1295 for (i = 0; i < cnt; i++) {
1296 KKASSERT(off != TMPFS_DIRCOOKIE_EOF);
1297 if (off == TMPFS_DIRCOOKIE_DOT) {
1298 off = TMPFS_DIRCOOKIE_DOTDOT;
1299 } else {
1300 if (off == TMPFS_DIRCOOKIE_DOTDOT) {
1301 de = TAILQ_FIRST(&node->tn_dir.tn_dirhead);
1302 } else if (de != NULL) {
1303 de = TAILQ_NEXT(de, td_entries);
1304 } else {
1305 de = tmpfs_dir_lookupbycookie(node,
1306 off);
1307 KKASSERT(de != NULL);
1308 de = TAILQ_NEXT(de, td_entries);
1309 }
1310 if (de == NULL)
1311 off = TMPFS_DIRCOOKIE_EOF;
1312 else
1313 off = tmpfs_dircookie(de);
1314 }
1315
1316 (*cookies)[i] = off;
1317 }
1318 KKASSERT(uio->uio_offset == off);
1319 }
7a2de9a4
MD
1320
1321 return error;
1322}
1323
1324/* --------------------------------------------------------------------- */
1325
1326static int
1327tmpfs_readlink(struct vop_readlink_args *v)
1328{
1329 struct vnode *vp = v->a_vp;
1330 struct uio *uio = v->a_uio;
1331
1332 int error;
1333 struct tmpfs_node *node;
1334
1335 KKASSERT(uio->uio_offset == 0);
1336 KKASSERT(vp->v_type == VLNK);
1337
1338 node = VP_TO_TMPFS_NODE(vp);
1339
1340 error = uiomove(node->tn_link, MIN(node->tn_size, uio->uio_resid),
1341 uio);
1342 TMPFS_NODE_LOCK(node);
1343 node->tn_status |= TMPFS_NODE_ACCESSED;
1344 TMPFS_NODE_UNLOCK(node);
1345
1346 return error;
1347}
1348
1349/* --------------------------------------------------------------------- */
1350
1351static int
1352tmpfs_inactive(struct vop_inactive_args *v)
1353{
1354 struct vnode *vp = v->a_vp;
1355
1356 struct tmpfs_node *node;
1357
7a2de9a4
MD
1358 node = VP_TO_TMPFS_NODE(vp);
1359
9fc94b5f
MD
1360 /*
1361 * Get rid of unreferenced deleted vnodes sooner rather than
1362 * later so the data memory can be recovered immediately.
f96f2f39
MD
1363 *
1364 * We must truncate the vnode to prevent the normal reclamation
1365 * path from flushing the data for the removed file to disk.
9fc94b5f 1366 */
7a2de9a4 1367 TMPFS_NODE_LOCK(node);
b7fe63af
MD
1368 if ((node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0 &&
1369 (node->tn_links == 0 ||
1370 (node->tn_links == 1 && node->tn_type == VDIR &&
1371 node->tn_dir.tn_parent)))
1372 {
9fc94b5f 1373 node->tn_vpstate = TMPFS_VNODE_DOOMED;
7a2de9a4 1374 TMPFS_NODE_UNLOCK(node);
f96f2f39
MD
1375 if (node->tn_type == VREG)
1376 tmpfs_truncate(vp, 0);
7a2de9a4 1377 vrecycle(vp);
9fc94b5f 1378 } else {
7a2de9a4 1379 TMPFS_NODE_UNLOCK(node);
9fc94b5f 1380 }
7a2de9a4
MD
1381
1382 return 0;
1383}
1384
1385/* --------------------------------------------------------------------- */
1386
1387int
1388tmpfs_reclaim(struct vop_reclaim_args *v)
1389{
1390 struct vnode *vp = v->a_vp;
7a2de9a4
MD
1391 struct tmpfs_mount *tmp;
1392 struct tmpfs_node *node;
1393
1394 node = VP_TO_TMPFS_NODE(vp);
1395 tmp = VFS_TO_TMPFS(vp->v_mount);
1396
7a2de9a4
MD
1397 tmpfs_free_vp(vp);
1398
b7fe63af
MD
1399 /*
1400 * If the node referenced by this vnode was deleted by the
1401 * user, we must free its associated data structures now that
1402 * the vnode is being reclaimed.
1403 *
1404 * Directories have an extra link ref.
1405 */
7a2de9a4 1406 TMPFS_NODE_LOCK(node);
b7fe63af
MD
1407 if ((node->tn_vpstate & TMPFS_VNODE_ALLOCATING) == 0 &&
1408 (node->tn_links == 0 ||
1409 (node->tn_links == 1 && node->tn_type == VDIR &&
1410 node->tn_dir.tn_parent)))
1411 {
7a2de9a4 1412 node->tn_vpstate = TMPFS_VNODE_DOOMED;
7a2de9a4 1413 tmpfs_free_node(tmp, node);
0786baf1 1414 /* eats the lock */
9fc94b5f 1415 } else {
7a2de9a4 1416 TMPFS_NODE_UNLOCK(node);
9fc94b5f 1417 }
7a2de9a4
MD
1418
1419 KKASSERT(vp->v_data == NULL);
1420 return 0;
1421}
1422
1423/* --------------------------------------------------------------------- */
1424
1425static int
1426tmpfs_print(struct vop_print_args *v)
1427{
1428 struct vnode *vp = v->a_vp;
1429
1430 struct tmpfs_node *node;
1431
1432 node = VP_TO_TMPFS_NODE(vp);
1433
1434 kprintf("tag VT_TMPFS, tmpfs_node %p, flags 0x%x, links %d\n",
1435 node, node->tn_flags, node->tn_links);
1436 kprintf("\tmode 0%o, owner %d, group %d, size %ju, status 0x%x\n",
1437 node->tn_mode, node->tn_uid, node->tn_gid,
1438 (uintmax_t)node->tn_size, node->tn_status);
1439
1440 if (vp->v_type == VFIFO)
1441 fifo_printinfo(vp);
1442
1443 kprintf("\n");
1444
1445 return 0;
1446}
1447
1448/* --------------------------------------------------------------------- */
1449
1450static int
1451tmpfs_pathconf(struct vop_pathconf_args *v)
1452{
1453 int name = v->a_name;
1454 register_t *retval = v->a_retval;
1455
1456 int error;
1457
1458 error = 0;
1459
1460 switch (name) {
1461 case _PC_LINK_MAX:
1462 *retval = LINK_MAX;
1463 break;
1464
1465 case _PC_NAME_MAX:
1466 *retval = NAME_MAX;
1467 break;
1468
1469 case _PC_PATH_MAX:
1470 *retval = PATH_MAX;
1471 break;
1472
1473 case _PC_PIPE_BUF:
1474 *retval = PIPE_BUF;
1475 break;
1476
1477 case _PC_CHOWN_RESTRICTED:
1478 *retval = 1;
1479 break;
1480
1481 case _PC_NO_TRUNC:
1482 *retval = 1;
1483 break;
1484
1485 case _PC_SYNC_IO:
1486 *retval = 1;
1487 break;
1488
1489 case _PC_FILESIZEBITS:
1490 *retval = 0; /* XXX Don't know which value should I return. */
1491 break;
1492
1493 default:
1494 error = EINVAL;
1495 }
1496
1497 return error;
1498}
1499
80ae59d7
MD
1500/************************************************************************
1501 * KQFILTER OPS *
1502 ************************************************************************/
1503
1504static void filt_tmpfsdetach(struct knote *kn);
1505static int filt_tmpfsread(struct knote *kn, long hint);
1506static int filt_tmpfswrite(struct knote *kn, long hint);
1507static int filt_tmpfsvnode(struct knote *kn, long hint);
1508
1509static struct filterops tmpfsread_filtops =
1510 { FILTEROP_ISFD, NULL, filt_tmpfsdetach, filt_tmpfsread };
1511static struct filterops tmpfswrite_filtops =
1512 { FILTEROP_ISFD, NULL, filt_tmpfsdetach, filt_tmpfswrite };
1513static struct filterops tmpfsvnode_filtops =
1514 { FILTEROP_ISFD, NULL, filt_tmpfsdetach, filt_tmpfsvnode };
1515
1516static int
1517tmpfs_kqfilter (struct vop_kqfilter_args *ap)
1518{
1519 struct vnode *vp = ap->a_vp;
1520 struct knote *kn = ap->a_kn;
1521
1522 switch (kn->kn_filter) {
1523 case EVFILT_READ:
1524 kn->kn_fop = &tmpfsread_filtops;
1525 break;
1526 case EVFILT_WRITE:
1527 kn->kn_fop = &tmpfswrite_filtops;
1528 break;
1529 case EVFILT_VNODE:
1530 kn->kn_fop = &tmpfsvnode_filtops;
1531 break;
1532 default:
1533 return (EOPNOTSUPP);
1534 }
1535
1536 kn->kn_hook = (caddr_t)vp;
1537
1538 knote_insert(&vp->v_pollinfo.vpi_kqinfo.ki_note, kn);
1539
1540 return(0);
1541}
1542
1543static void
1544filt_tmpfsdetach(struct knote *kn)
1545{
1546 struct vnode *vp = (void *)kn->kn_hook;
1547
1548 knote_remove(&vp->v_pollinfo.vpi_kqinfo.ki_note, kn);
1549}
1550
1551static int
1552filt_tmpfsread(struct knote *kn, long hint)
1553{
1554 struct vnode *vp = (void *)kn->kn_hook;
1555 struct tmpfs_node *node = VP_TO_TMPFS_NODE(vp);
1556 off_t off;
1557
1558 if (hint == NOTE_REVOKE) {
1559 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1560 return(1);
1561 }
f79d9cc9
MD
1562
1563 /*
1564 * Interlock against MP races when performing this function.
1565 */
1566 lwkt_gettoken(&vp->v_mount->mnt_token);
80ae59d7
MD
1567 off = node->tn_size - kn->kn_fp->f_offset;
1568 kn->kn_data = (off < INTPTR_MAX) ? off : INTPTR_MAX;
f79d9cc9
MD
1569 if (kn->kn_sfflags & NOTE_OLDAPI) {
1570 lwkt_reltoken(&vp->v_mount->mnt_token);
80ae59d7 1571 return(1);
f79d9cc9 1572 }
80ae59d7 1573
80ae59d7 1574 if (kn->kn_data == 0) {
80ae59d7 1575 kn->kn_data = (off < INTPTR_MAX) ? off : INTPTR_MAX;
80ae59d7 1576 }
f79d9cc9 1577 lwkt_reltoken(&vp->v_mount->mnt_token);
80ae59d7
MD
1578 return (kn->kn_data != 0);
1579}
1580
1581static int
1582filt_tmpfswrite(struct knote *kn, long hint)
1583{
1584 if (hint == NOTE_REVOKE)
1585 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
1586 kn->kn_data = 0;
1587 return (1);
1588}
1589
1590static int
1591filt_tmpfsvnode(struct knote *kn, long hint)
1592{
1593 if (kn->kn_sfflags & hint)
1594 kn->kn_fflags |= hint;
1595 if (hint == NOTE_REVOKE) {
1596 kn->kn_flags |= EV_EOF;
1597 return (1);
1598 }
1599 return (kn->kn_fflags != 0);
1600}
1601
1602
7a2de9a4
MD
1603/* --------------------------------------------------------------------- */
1604
1605/*
1606 * vnode operations vector used for files stored in a tmpfs file system.
1607 */
1608struct vop_ops tmpfs_vnode_vops = {
1609 .vop_default = vop_defaultop,
1610 .vop_getpages = vop_stdgetpages,
1611 .vop_putpages = vop_stdputpages,
1612 .vop_ncreate = tmpfs_ncreate,
1613 .vop_nresolve = tmpfs_nresolve,
1614 .vop_nlookupdotdot = tmpfs_nlookupdotdot,
1615 .vop_nmknod = tmpfs_nmknod,
1616 .vop_open = tmpfs_open,
1617 .vop_close = tmpfs_close,
1618 .vop_access = tmpfs_access,
1619 .vop_getattr = tmpfs_getattr,
1620 .vop_setattr = tmpfs_setattr,
1621 .vop_read = tmpfs_read,
1622 .vop_write = tmpfs_write,
1623 .vop_fsync = tmpfs_fsync,
1624 .vop_nremove = tmpfs_nremove,
1625 .vop_nlink = tmpfs_nlink,
1626 .vop_nrename = tmpfs_nrename,
1627 .vop_nmkdir = tmpfs_nmkdir,
1628 .vop_nrmdir = tmpfs_nrmdir,
1629 .vop_nsymlink = tmpfs_nsymlink,
1630 .vop_readdir = tmpfs_readdir,
1631 .vop_readlink = tmpfs_readlink,
1632 .vop_inactive = tmpfs_inactive,
1633 .vop_reclaim = tmpfs_reclaim,
1634 .vop_print = tmpfs_print,
1635 .vop_pathconf = tmpfs_pathconf,
9fc94b5f 1636 .vop_bmap = tmpfs_bmap,
7a2de9a4
MD
1637 .vop_strategy = tmpfs_strategy,
1638 .vop_advlock = tmpfs_advlock,
80ae59d7 1639 .vop_kqfilter = tmpfs_kqfilter
7a2de9a4 1640};