Commit | Line | Data |
---|---|---|
15637ed4 RG |
1 | /* |
2 | * Copyright (c) 1989 The Regents of the University of California. | |
3 | * All rights reserved. | |
4 | * | |
5 | * Redistribution and use in source and binary forms, with or without | |
6 | * modification, are permitted provided that the following conditions | |
7 | * are met: | |
8 | * 1. Redistributions of source code must retain the above copyright | |
9 | * notice, this list of conditions and the following disclaimer. | |
10 | * 2. Redistributions in binary form must reproduce the above copyright | |
11 | * notice, this list of conditions and the following disclaimer in the | |
12 | * documentation and/or other materials provided with the distribution. | |
13 | * 3. All advertising materials mentioning features or use of this software | |
14 | * must display the following acknowledgement: | |
15 | * This product includes software developed by the University of | |
16 | * California, Berkeley and its contributors. | |
17 | * 4. Neither the name of the University nor the names of its contributors | |
18 | * may be used to endorse or promote products derived from this software | |
19 | * without specific prior written permission. | |
20 | * | |
21 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
22 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
23 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
24 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
25 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
26 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
27 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
28 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
29 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
30 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
31 | * SUCH DAMAGE. | |
32 | * | |
600f7f07 | 33 | * from: @(#)vfs_subr.c 7.60 (Berkeley) 6/21/91 |
4c45483e | 34 | * $Id: vfs_subr.c,v 1.6 1993/11/07 21:44:50 wollman Exp $ |
15637ed4 RG |
35 | */ |
36 | ||
37 | /* | |
38 | * External virtual filesystem routines | |
39 | */ | |
40 | ||
41 | #include "param.h" | |
4c45483e | 42 | #include "systm.h" |
15637ed4 RG |
43 | #include "proc.h" |
44 | #include "mount.h" | |
45 | #include "time.h" | |
46 | #include "vnode.h" | |
47 | #include "specdev.h" | |
48 | #include "namei.h" | |
49 | #include "ucred.h" | |
50 | #include "buf.h" | |
51 | #include "errno.h" | |
52 | #include "malloc.h" | |
53 | ||
4c45483e GW |
54 | static void insmntque(struct vnode *, struct mount *); |
55 | ||
bbc3f849 GW |
56 | struct vnode *speclisth[SPECHSZ]; /* device special file vnode hash table */ |
57 | ||
15637ed4 RG |
58 | /* |
59 | * Remove a mount point from the list of mounted filesystems. | |
60 | * Unmount of the root is illegal. | |
61 | */ | |
62 | void | |
63 | vfs_remove(mp) | |
64 | register struct mount *mp; | |
65 | { | |
66 | ||
67 | if (mp == rootfs) | |
68 | panic("vfs_remove: unmounting root"); | |
69 | mp->mnt_prev->mnt_next = mp->mnt_next; | |
70 | mp->mnt_next->mnt_prev = mp->mnt_prev; | |
71 | mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0; | |
72 | vfs_unlock(mp); | |
73 | } | |
74 | ||
75 | /* | |
76 | * Lock a filesystem. | |
77 | * Used to prevent access to it while mounting and unmounting. | |
78 | */ | |
4c45483e | 79 | int |
15637ed4 RG |
80 | vfs_lock(mp) |
81 | register struct mount *mp; | |
82 | { | |
83 | ||
84 | while(mp->mnt_flag & MNT_MLOCK) { | |
85 | mp->mnt_flag |= MNT_MWAIT; | |
73419b27 | 86 | tsleep((caddr_t)mp, PVFS, "vfslock", 0); |
15637ed4 RG |
87 | } |
88 | mp->mnt_flag |= MNT_MLOCK; | |
89 | return (0); | |
90 | } | |
91 | ||
92 | /* | |
93 | * Unlock a locked filesystem. | |
94 | * Panic if filesystem is not locked. | |
95 | */ | |
96 | void | |
97 | vfs_unlock(mp) | |
98 | register struct mount *mp; | |
99 | { | |
100 | ||
101 | if ((mp->mnt_flag & MNT_MLOCK) == 0) | |
102 | panic("vfs_unlock: not locked"); | |
103 | mp->mnt_flag &= ~MNT_MLOCK; | |
104 | if (mp->mnt_flag & MNT_MWAIT) { | |
105 | mp->mnt_flag &= ~MNT_MWAIT; | |
106 | wakeup((caddr_t)mp); | |
107 | } | |
108 | } | |
109 | ||
110 | /* | |
111 | * Mark a mount point as busy. | |
112 | * Used to synchronize access and to delay unmounting. | |
113 | */ | |
4c45483e | 114 | int |
15637ed4 RG |
115 | vfs_busy(mp) |
116 | register struct mount *mp; | |
117 | { | |
118 | ||
119 | while(mp->mnt_flag & MNT_MPBUSY) { | |
120 | mp->mnt_flag |= MNT_MPWANT; | |
73419b27 | 121 | tsleep((caddr_t)&mp->mnt_flag, PVFS, "vfsbusy", 0); |
15637ed4 RG |
122 | } |
123 | if (mp->mnt_flag & MNT_UNMOUNT) | |
124 | return (1); | |
125 | mp->mnt_flag |= MNT_MPBUSY; | |
126 | return (0); | |
127 | } | |
128 | ||
129 | /* | |
130 | * Free a busy filesystem. | |
131 | * Panic if filesystem is not busy. | |
132 | */ | |
4c45483e | 133 | void |
15637ed4 RG |
134 | vfs_unbusy(mp) |
135 | register struct mount *mp; | |
136 | { | |
137 | ||
138 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
139 | panic("vfs_unbusy: not busy"); | |
140 | mp->mnt_flag &= ~MNT_MPBUSY; | |
141 | if (mp->mnt_flag & MNT_MPWANT) { | |
142 | mp->mnt_flag &= ~MNT_MPWANT; | |
143 | wakeup((caddr_t)&mp->mnt_flag); | |
144 | } | |
145 | } | |
146 | ||
147 | /* | |
148 | * Lookup a mount point by filesystem identifier. | |
149 | */ | |
150 | struct mount * | |
151 | getvfs(fsid) | |
152 | fsid_t *fsid; | |
153 | { | |
154 | register struct mount *mp; | |
155 | ||
156 | mp = rootfs; | |
157 | do { | |
158 | if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && | |
159 | mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { | |
160 | return (mp); | |
161 | } | |
162 | mp = mp->mnt_next; | |
163 | } while (mp != rootfs); | |
164 | return ((struct mount *)0); | |
165 | } | |
166 | ||
167 | /* | |
168 | * Set vnode attributes to VNOVAL | |
169 | */ | |
170 | void vattr_null(vap) | |
171 | register struct vattr *vap; | |
172 | { | |
173 | ||
174 | vap->va_type = VNON; | |
175 | vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = | |
176 | vap->va_fsid = vap->va_fileid = vap->va_size = | |
177 | vap->va_size_rsv = vap->va_blocksize = vap->va_rdev = | |
178 | vap->va_bytes = vap->va_bytes_rsv = | |
179 | vap->va_atime.tv_sec = vap->va_atime.tv_usec = | |
180 | vap->va_mtime.tv_sec = vap->va_mtime.tv_usec = | |
181 | vap->va_ctime.tv_sec = vap->va_ctime.tv_usec = | |
182 | vap->va_flags = vap->va_gen = VNOVAL; | |
183 | } | |
184 | ||
185 | /* | |
186 | * Routines having to do with the management of the vnode table. | |
187 | */ | |
188 | struct vnode *vfreeh, **vfreet; | |
189 | extern struct vnodeops dead_vnodeops, spec_vnodeops; | |
15637ed4 RG |
190 | long numvnodes; |
191 | struct vattr va_null; | |
192 | ||
193 | /* | |
194 | * Initialize the vnode structures and initialize each file system type. | |
195 | */ | |
4c45483e | 196 | void |
15637ed4 RG |
197 | vfsinit() |
198 | { | |
199 | struct vfsops **vfsp; | |
200 | ||
201 | /* | |
202 | * Initialize the vnode name cache | |
203 | */ | |
204 | nchinit(); | |
205 | /* | |
206 | * Initialize each file system type. | |
207 | */ | |
208 | vattr_null(&va_null); | |
209 | for (vfsp = &vfssw[0]; vfsp <= &vfssw[MOUNT_MAXTYPE]; vfsp++) { | |
210 | if (*vfsp == NULL) | |
211 | continue; | |
212 | (*(*vfsp)->vfs_init)(); | |
213 | } | |
214 | } | |
215 | ||
216 | /* | |
217 | * Return the next vnode from the free list. | |
218 | */ | |
4c45483e | 219 | int |
15637ed4 RG |
220 | getnewvnode(tag, mp, vops, vpp) |
221 | enum vtagtype tag; | |
222 | struct mount *mp; | |
223 | struct vnodeops *vops; | |
224 | struct vnode **vpp; | |
225 | { | |
226 | register struct vnode *vp, *vq; | |
227 | ||
228 | if (numvnodes < desiredvnodes) { | |
229 | vp = (struct vnode *)malloc((u_long)sizeof *vp, | |
230 | M_VNODE, M_WAITOK); | |
231 | bzero((char *)vp, sizeof *vp); | |
232 | numvnodes++; | |
233 | } else { | |
234 | if ((vp = vfreeh) == NULL) { | |
235 | tablefull("vnode"); | |
236 | *vpp = 0; | |
237 | return (ENFILE); | |
238 | } | |
239 | if (vp->v_usecount) | |
240 | panic("free vnode isn't"); | |
241 | if (vq = vp->v_freef) | |
242 | vq->v_freeb = &vfreeh; | |
243 | else | |
244 | vfreet = &vfreeh; | |
245 | vfreeh = vq; | |
246 | vp->v_freef = NULL; | |
247 | vp->v_freeb = NULL; | |
248 | if (vp->v_type != VBAD) | |
249 | vgone(vp); | |
250 | vp->v_flag = 0; | |
251 | vp->v_lastr = 0; | |
252 | vp->v_socket = 0; | |
253 | } | |
254 | vp->v_type = VNON; | |
255 | cache_purge(vp); | |
256 | vp->v_tag = tag; | |
257 | vp->v_op = vops; | |
258 | insmntque(vp, mp); | |
259 | VREF(vp); | |
260 | *vpp = vp; | |
261 | return (0); | |
262 | } | |
263 | ||
264 | /* | |
265 | * Move a vnode from one mount queue to another. | |
266 | */ | |
4c45483e | 267 | static void |
15637ed4 RG |
268 | insmntque(vp, mp) |
269 | register struct vnode *vp; | |
270 | register struct mount *mp; | |
271 | { | |
272 | register struct vnode *vq; | |
273 | ||
274 | /* | |
275 | * Delete from old mount point vnode list, if on one. | |
276 | */ | |
277 | if (vp->v_mountb) { | |
278 | if (vq = vp->v_mountf) | |
279 | vq->v_mountb = vp->v_mountb; | |
280 | *vp->v_mountb = vq; | |
281 | } | |
282 | /* | |
283 | * Insert into list of vnodes for the new mount point, if available. | |
284 | */ | |
285 | vp->v_mount = mp; | |
286 | if (mp == NULL) { | |
287 | vp->v_mountf = NULL; | |
288 | vp->v_mountb = NULL; | |
289 | return; | |
290 | } | |
291 | if (vq = mp->mnt_mounth) | |
292 | vq->v_mountb = &vp->v_mountf; | |
293 | vp->v_mountf = vq; | |
294 | vp->v_mountb = &mp->mnt_mounth; | |
295 | mp->mnt_mounth = vp; | |
296 | } | |
297 | ||
298 | /* | |
299 | * Make sure all write-behind blocks associated | |
300 | * with mount point are flushed out (from sync). | |
301 | */ | |
4c45483e | 302 | void |
15637ed4 RG |
303 | mntflushbuf(mountp, flags) |
304 | struct mount *mountp; | |
305 | int flags; | |
306 | { | |
307 | register struct vnode *vp; | |
308 | ||
309 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
310 | panic("mntflushbuf: not busy"); | |
311 | loop: | |
312 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
313 | if (VOP_ISLOCKED(vp)) | |
314 | continue; | |
315 | if (vget(vp)) | |
316 | goto loop; | |
317 | vflushbuf(vp, flags); | |
318 | vput(vp); | |
319 | if (vp->v_mount != mountp) | |
320 | goto loop; | |
321 | } | |
322 | } | |
323 | ||
324 | /* | |
325 | * Flush all dirty buffers associated with a vnode. | |
326 | */ | |
4c45483e | 327 | void |
15637ed4 RG |
328 | vflushbuf(vp, flags) |
329 | register struct vnode *vp; | |
330 | int flags; | |
331 | { | |
332 | register struct buf *bp; | |
333 | struct buf *nbp; | |
334 | int s; | |
335 | ||
336 | loop: | |
337 | s = splbio(); | |
338 | for (bp = vp->v_dirtyblkhd; bp; bp = nbp) { | |
339 | nbp = bp->b_blockf; | |
340 | if ((bp->b_flags & B_BUSY)) | |
341 | continue; | |
342 | if ((bp->b_flags & B_DELWRI) == 0) | |
343 | panic("vflushbuf: not dirty"); | |
344 | bremfree(bp); | |
345 | bp->b_flags |= B_BUSY; | |
346 | splx(s); | |
347 | /* | |
348 | * Wait for I/O associated with indirect blocks to complete, | |
349 | * since there is no way to quickly wait for them below. | |
350 | * NB: This is really specific to ufs, but is done here | |
351 | * as it is easier and quicker. | |
352 | */ | |
353 | if (bp->b_vp == vp || (flags & B_SYNC) == 0) | |
354 | (void) bawrite(bp); | |
355 | else | |
356 | (void) bwrite(bp); | |
357 | goto loop; | |
358 | } | |
359 | splx(s); | |
360 | if ((flags & B_SYNC) == 0) | |
361 | return; | |
362 | s = splbio(); | |
363 | while (vp->v_numoutput) { | |
364 | vp->v_flag |= VBWAIT; | |
73419b27 | 365 | tsleep((caddr_t)&vp->v_numoutput, PRIBIO + 1, "vflushbf", 0); |
15637ed4 RG |
366 | } |
367 | splx(s); | |
368 | if (vp->v_dirtyblkhd) { | |
369 | vprint("vflushbuf: dirty", vp); | |
370 | goto loop; | |
371 | } | |
372 | } | |
373 | ||
374 | /* | |
375 | * Update outstanding I/O count and do wakeup if requested. | |
376 | */ | |
4c45483e | 377 | void |
15637ed4 RG |
378 | vwakeup(bp) |
379 | register struct buf *bp; | |
380 | { | |
381 | register struct vnode *vp; | |
382 | ||
383 | bp->b_dirtyoff = bp->b_dirtyend = 0; | |
384 | if (vp = bp->b_vp) { | |
385 | vp->v_numoutput--; | |
386 | if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { | |
387 | if (vp->v_numoutput < 0) | |
388 | panic("vwakeup: neg numoutput"); | |
389 | vp->v_flag &= ~VBWAIT; | |
390 | wakeup((caddr_t)&vp->v_numoutput); | |
391 | } | |
392 | } | |
393 | } | |
394 | ||
395 | /* | |
396 | * Invalidate in core blocks belonging to closed or umounted filesystem | |
397 | * | |
398 | * Go through the list of vnodes associated with the file system; | |
399 | * for each vnode invalidate any buffers that it holds. Normally | |
400 | * this routine is preceeded by a bflush call, so that on a quiescent | |
401 | * filesystem there will be no dirty buffers when we are done. Binval | |
402 | * returns the count of dirty buffers when it is finished. | |
403 | */ | |
4c45483e | 404 | int |
15637ed4 RG |
405 | mntinvalbuf(mountp) |
406 | struct mount *mountp; | |
407 | { | |
408 | register struct vnode *vp; | |
409 | int dirty = 0; | |
410 | ||
411 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
412 | panic("mntinvalbuf: not busy"); | |
413 | loop: | |
414 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
415 | if (vget(vp)) | |
416 | goto loop; | |
417 | dirty += vinvalbuf(vp, 1); | |
418 | vput(vp); | |
419 | if (vp->v_mount != mountp) | |
420 | goto loop; | |
421 | } | |
422 | return (dirty); | |
423 | } | |
424 | ||
425 | /* | |
426 | * Flush out and invalidate all buffers associated with a vnode. | |
427 | * Called with the underlying object locked. | |
428 | */ | |
4c45483e | 429 | int |
15637ed4 RG |
430 | vinvalbuf(vp, save) |
431 | register struct vnode *vp; | |
432 | int save; | |
433 | { | |
434 | register struct buf *bp; | |
435 | struct buf *nbp, *blist; | |
436 | int s, dirty = 0; | |
437 | ||
438 | for (;;) { | |
439 | if (blist = vp->v_dirtyblkhd) | |
440 | /* void */; | |
441 | else if (blist = vp->v_cleanblkhd) | |
442 | /* void */; | |
443 | else | |
444 | break; | |
445 | for (bp = blist; bp; bp = nbp) { | |
446 | nbp = bp->b_blockf; | |
447 | s = splbio(); | |
448 | if (bp->b_flags & B_BUSY) { | |
449 | bp->b_flags |= B_WANTED; | |
73419b27 | 450 | tsleep((caddr_t)bp, PRIBIO + 1, "vinvalbf", 0); |
15637ed4 RG |
451 | splx(s); |
452 | break; | |
453 | } | |
454 | bremfree(bp); | |
455 | bp->b_flags |= B_BUSY; | |
456 | splx(s); | |
457 | if (save && (bp->b_flags & B_DELWRI)) { | |
458 | dirty++; | |
459 | (void) bwrite(bp); | |
460 | break; | |
461 | } | |
462 | if (bp->b_vp != vp) | |
463 | reassignbuf(bp, bp->b_vp); | |
464 | else | |
465 | bp->b_flags |= B_INVAL; | |
466 | brelse(bp); | |
467 | } | |
468 | } | |
469 | if (vp->v_dirtyblkhd || vp->v_cleanblkhd) | |
470 | panic("vinvalbuf: flush failed"); | |
471 | return (dirty); | |
472 | } | |
473 | ||
474 | /* | |
475 | * Associate a buffer with a vnode. | |
476 | */ | |
4c45483e | 477 | void |
15637ed4 RG |
478 | bgetvp(vp, bp) |
479 | register struct vnode *vp; | |
480 | register struct buf *bp; | |
481 | { | |
482 | register struct vnode *vq; | |
483 | register struct buf *bq; | |
484 | ||
485 | if (bp->b_vp) | |
486 | panic("bgetvp: not free"); | |
487 | VHOLD(vp); | |
488 | bp->b_vp = vp; | |
489 | if (vp->v_type == VBLK || vp->v_type == VCHR) | |
490 | bp->b_dev = vp->v_rdev; | |
491 | else | |
492 | bp->b_dev = NODEV; | |
493 | /* | |
494 | * Insert onto list for new vnode. | |
495 | */ | |
496 | if (bq = vp->v_cleanblkhd) | |
497 | bq->b_blockb = &bp->b_blockf; | |
498 | bp->b_blockf = bq; | |
499 | bp->b_blockb = &vp->v_cleanblkhd; | |
500 | vp->v_cleanblkhd = bp; | |
501 | } | |
502 | ||
503 | /* | |
504 | * Disassociate a buffer from a vnode. | |
505 | */ | |
4c45483e | 506 | void |
15637ed4 RG |
507 | brelvp(bp) |
508 | register struct buf *bp; | |
509 | { | |
510 | struct buf *bq; | |
511 | struct vnode *vp; | |
512 | ||
513 | if (bp->b_vp == (struct vnode *) 0) | |
514 | panic("brelvp: NULL"); | |
515 | /* | |
516 | * Delete from old vnode list, if on one. | |
517 | */ | |
518 | if (bp->b_blockb) { | |
519 | if (bq = bp->b_blockf) | |
520 | bq->b_blockb = bp->b_blockb; | |
521 | *bp->b_blockb = bq; | |
522 | bp->b_blockf = NULL; | |
523 | bp->b_blockb = NULL; | |
524 | } | |
525 | vp = bp->b_vp; | |
526 | bp->b_vp = (struct vnode *) 0; | |
527 | HOLDRELE(vp); | |
528 | } | |
529 | ||
530 | /* | |
531 | * Reassign a buffer from one vnode to another. | |
532 | * Used to assign file specific control information | |
533 | * (indirect blocks) to the vnode to which they belong. | |
534 | */ | |
4c45483e | 535 | void |
15637ed4 RG |
536 | reassignbuf(bp, newvp) |
537 | register struct buf *bp; | |
538 | register struct vnode *newvp; | |
539 | { | |
540 | register struct buf *bq, **listheadp; | |
541 | ||
542 | if (newvp == NULL) | |
543 | panic("reassignbuf: NULL"); | |
544 | /* | |
545 | * Delete from old vnode list, if on one. | |
546 | */ | |
547 | if (bp->b_blockb) { | |
548 | if (bq = bp->b_blockf) | |
549 | bq->b_blockb = bp->b_blockb; | |
550 | *bp->b_blockb = bq; | |
551 | } | |
552 | /* | |
553 | * If dirty, put on list of dirty buffers; | |
554 | * otherwise insert onto list of clean buffers. | |
555 | */ | |
556 | if (bp->b_flags & B_DELWRI) | |
557 | listheadp = &newvp->v_dirtyblkhd; | |
558 | else | |
559 | listheadp = &newvp->v_cleanblkhd; | |
560 | if (bq = *listheadp) | |
561 | bq->b_blockb = &bp->b_blockf; | |
562 | bp->b_blockf = bq; | |
563 | bp->b_blockb = listheadp; | |
564 | *listheadp = bp; | |
565 | } | |
566 | ||
567 | /* | |
568 | * Create a vnode for a block device. | |
569 | * Used for root filesystem, argdev, and swap areas. | |
570 | * Also used for memory file system special devices. | |
571 | */ | |
4c45483e | 572 | int |
15637ed4 RG |
573 | bdevvp(dev, vpp) |
574 | dev_t dev; | |
575 | struct vnode **vpp; | |
576 | { | |
577 | register struct vnode *vp; | |
578 | struct vnode *nvp; | |
579 | int error; | |
580 | ||
581 | if (dev == NODEV) | |
582 | return (0); | |
583 | error = getnewvnode(VT_NON, (struct mount *)0, &spec_vnodeops, &nvp); | |
584 | if (error) { | |
80cc9f07 | 585 | *vpp = 0; |
15637ed4 RG |
586 | return (error); |
587 | } | |
588 | vp = nvp; | |
80cc9f07 | 589 | vp->v_type = VBLK; |
15637ed4 RG |
590 | if (nvp = checkalias(vp, dev, (struct mount *)0)) { |
591 | vput(vp); | |
592 | vp = nvp; | |
593 | } | |
594 | *vpp = vp; | |
595 | return (0); | |
596 | } | |
597 | ||
598 | /* | |
599 | * Check to see if the new vnode represents a special device | |
600 | * for which we already have a vnode (either because of | |
601 | * bdevvp() or because of a different vnode representing | |
602 | * the same block device). If such an alias exists, deallocate | |
603 | * the existing contents and return the aliased vnode. The | |
604 | * caller is responsible for filling it with its new contents. | |
605 | */ | |
606 | struct vnode * | |
607 | checkalias(nvp, nvp_rdev, mp) | |
608 | register struct vnode *nvp; | |
609 | dev_t nvp_rdev; | |
610 | struct mount *mp; | |
611 | { | |
612 | register struct vnode *vp; | |
613 | struct vnode **vpp; | |
614 | ||
615 | if (nvp->v_type != VBLK && nvp->v_type != VCHR) | |
616 | return (NULLVP); | |
617 | ||
618 | vpp = &speclisth[SPECHASH(nvp_rdev)]; | |
619 | loop: | |
620 | for (vp = *vpp; vp; vp = vp->v_specnext) { | |
621 | if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) | |
622 | continue; | |
623 | /* | |
624 | * Alias, but not in use, so flush it out. | |
625 | */ | |
626 | if (vp->v_usecount == 0) { | |
627 | vgone(vp); | |
628 | goto loop; | |
629 | } | |
630 | if (vget(vp)) | |
631 | goto loop; | |
632 | break; | |
633 | } | |
634 | if (vp == NULL || vp->v_tag != VT_NON) { | |
635 | MALLOC(nvp->v_specinfo, struct specinfo *, | |
636 | sizeof(struct specinfo), M_VNODE, M_WAITOK); | |
637 | nvp->v_rdev = nvp_rdev; | |
638 | nvp->v_hashchain = vpp; | |
639 | nvp->v_specnext = *vpp; | |
640 | nvp->v_specflags = 0; | |
641 | *vpp = nvp; | |
642 | if (vp != NULL) { | |
643 | nvp->v_flag |= VALIASED; | |
644 | vp->v_flag |= VALIASED; | |
645 | vput(vp); | |
646 | } | |
647 | return (NULLVP); | |
648 | } | |
649 | VOP_UNLOCK(vp); | |
650 | vclean(vp, 0); | |
651 | vp->v_op = nvp->v_op; | |
652 | vp->v_tag = nvp->v_tag; | |
653 | nvp->v_type = VNON; | |
654 | insmntque(vp, mp); | |
655 | return (vp); | |
656 | } | |
657 | ||
658 | /* | |
659 | * Grab a particular vnode from the free list, increment its | |
660 | * reference count and lock it. The vnode lock bit is set the | |
661 | * vnode is being eliminated in vgone. The process is awakened | |
662 | * when the transition is completed, and an error returned to | |
663 | * indicate that the vnode is no longer usable (possibly having | |
664 | * been changed to a new file system type). | |
665 | */ | |
4c45483e | 666 | int |
15637ed4 RG |
667 | vget(vp) |
668 | register struct vnode *vp; | |
669 | { | |
670 | register struct vnode *vq; | |
671 | ||
672 | if (vp->v_flag & VXLOCK) { | |
673 | vp->v_flag |= VXWANT; | |
73419b27 | 674 | tsleep((caddr_t)vp, PINOD, "vget", 0); |
15637ed4 RG |
675 | return (1); |
676 | } | |
677 | if (vp->v_usecount == 0) { | |
678 | if (vq = vp->v_freef) | |
679 | vq->v_freeb = vp->v_freeb; | |
680 | else | |
681 | vfreet = vp->v_freeb; | |
682 | *vp->v_freeb = vq; | |
683 | vp->v_freef = NULL; | |
684 | vp->v_freeb = NULL; | |
685 | } | |
686 | VREF(vp); | |
687 | VOP_LOCK(vp); | |
688 | return (0); | |
689 | } | |
690 | ||
691 | /* | |
692 | * Vnode reference, just increment the count | |
693 | */ | |
694 | void vref(vp) | |
695 | struct vnode *vp; | |
696 | { | |
697 | ||
698 | vp->v_usecount++; | |
699 | } | |
700 | ||
701 | /* | |
702 | * vput(), just unlock and vrele() | |
703 | */ | |
704 | void vput(vp) | |
705 | register struct vnode *vp; | |
706 | { | |
707 | VOP_UNLOCK(vp); | |
708 | vrele(vp); | |
709 | } | |
710 | ||
711 | /* | |
712 | * Vnode release. | |
713 | * If count drops to zero, call inactive routine and return to freelist. | |
714 | */ | |
715 | void vrele(vp) | |
716 | register struct vnode *vp; | |
717 | { | |
718 | struct proc *p = curproc; /* XXX */ | |
719 | ||
720 | #ifdef DIAGNOSTIC | |
721 | if (vp == NULL) | |
722 | panic("vrele: null vp"); | |
723 | #endif | |
724 | vp->v_usecount--; | |
725 | if (vp->v_usecount > 0) | |
726 | return; | |
727 | #ifdef DIAGNOSTIC | |
728 | if (vp->v_usecount != 0 || vp->v_writecount != 0) { | |
729 | vprint("vrele: bad ref count", vp); | |
730 | panic("vrele: ref cnt"); | |
731 | } | |
732 | #endif | |
733 | if (vfreeh == NULLVP) { | |
734 | /* | |
735 | * insert into empty list | |
736 | */ | |
737 | vfreeh = vp; | |
738 | vp->v_freeb = &vfreeh; | |
739 | } else { | |
740 | /* | |
741 | * insert at tail of list | |
742 | */ | |
743 | *vfreet = vp; | |
744 | vp->v_freeb = vfreet; | |
745 | } | |
746 | vp->v_freef = NULL; | |
747 | vfreet = &vp->v_freef; | |
748 | VOP_INACTIVE(vp, p); | |
749 | } | |
750 | ||
751 | /* | |
752 | * Page or buffer structure gets a reference. | |
753 | */ | |
4c45483e | 754 | void |
15637ed4 RG |
755 | vhold(vp) |
756 | register struct vnode *vp; | |
757 | { | |
758 | ||
759 | vp->v_holdcnt++; | |
760 | } | |
761 | ||
762 | /* | |
763 | * Page or buffer structure frees a reference. | |
764 | */ | |
4c45483e | 765 | void |
15637ed4 RG |
766 | holdrele(vp) |
767 | register struct vnode *vp; | |
768 | { | |
769 | ||
770 | if (vp->v_holdcnt <= 0) | |
771 | panic("holdrele: holdcnt"); | |
772 | vp->v_holdcnt--; | |
773 | } | |
774 | ||
775 | /* | |
776 | * Remove any vnodes in the vnode table belonging to mount point mp. | |
777 | * | |
778 | * If MNT_NOFORCE is specified, there should not be any active ones, | |
779 | * return error if any are found (nb: this is a user error, not a | |
780 | * system error). If MNT_FORCE is specified, detach any active vnodes | |
781 | * that are found. | |
782 | */ | |
783 | int busyprt = 0; /* patch to print out busy vnodes */ | |
784 | ||
4c45483e | 785 | int |
15637ed4 RG |
786 | vflush(mp, skipvp, flags) |
787 | struct mount *mp; | |
788 | struct vnode *skipvp; | |
789 | int flags; | |
790 | { | |
791 | register struct vnode *vp, *nvp; | |
792 | int busy = 0; | |
793 | ||
794 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
795 | panic("vflush: not busy"); | |
796 | loop: | |
797 | for (vp = mp->mnt_mounth; vp; vp = nvp) { | |
798 | if (vp->v_mount != mp) | |
799 | goto loop; | |
800 | nvp = vp->v_mountf; | |
801 | /* | |
802 | * Skip over a selected vnode. | |
803 | */ | |
804 | if (vp == skipvp) | |
805 | continue; | |
806 | /* | |
807 | * Skip over a vnodes marked VSYSTEM. | |
808 | */ | |
809 | if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) | |
810 | continue; | |
811 | /* | |
812 | * With v_usecount == 0, all we need to do is clear | |
813 | * out the vnode data structures and we are done. | |
814 | */ | |
815 | if (vp->v_usecount == 0) { | |
816 | vgone(vp); | |
817 | continue; | |
818 | } | |
819 | /* | |
820 | * For block or character devices, revert to an | |
821 | * anonymous device. For all other files, just kill them. | |
822 | */ | |
823 | if (flags & FORCECLOSE) { | |
824 | if (vp->v_type != VBLK && vp->v_type != VCHR) { | |
825 | vgone(vp); | |
826 | } else { | |
827 | vclean(vp, 0); | |
828 | vp->v_op = &spec_vnodeops; | |
829 | insmntque(vp, (struct mount *)0); | |
830 | } | |
831 | continue; | |
832 | } | |
833 | if (busyprt) | |
834 | vprint("vflush: busy vnode", vp); | |
835 | busy++; | |
836 | } | |
837 | if (busy) | |
838 | return (EBUSY); | |
839 | return (0); | |
840 | } | |
841 | ||
842 | /* | |
843 | * Disassociate the underlying file system from a vnode. | |
844 | */ | |
845 | void vclean(vp, flags) | |
846 | register struct vnode *vp; | |
847 | int flags; | |
848 | { | |
849 | struct vnodeops *origops; | |
850 | int active; | |
851 | struct proc *p = curproc; /* XXX */ | |
852 | ||
853 | /* | |
854 | * Check to see if the vnode is in use. | |
855 | * If so we have to reference it before we clean it out | |
856 | * so that its count cannot fall to zero and generate a | |
857 | * race against ourselves to recycle it. | |
858 | */ | |
859 | if (active = vp->v_usecount) | |
860 | VREF(vp); | |
861 | /* | |
862 | * Prevent the vnode from being recycled or | |
863 | * brought into use while we clean it out. | |
864 | */ | |
865 | if (vp->v_flag & VXLOCK) | |
866 | panic("vclean: deadlock"); | |
867 | vp->v_flag |= VXLOCK; | |
868 | /* | |
869 | * Even if the count is zero, the VOP_INACTIVE routine may still | |
870 | * have the object locked while it cleans it out. The VOP_LOCK | |
871 | * ensures that the VOP_INACTIVE routine is done with its work. | |
872 | * For active vnodes, it ensures that no other activity can | |
873 | * occur while the buffer list is being cleaned out. | |
874 | */ | |
875 | VOP_LOCK(vp); | |
876 | if (flags & DOCLOSE) | |
877 | vinvalbuf(vp, 1); | |
878 | /* | |
879 | * Prevent any further operations on the vnode from | |
880 | * being passed through to the old file system. | |
881 | */ | |
882 | origops = vp->v_op; | |
883 | vp->v_op = &dead_vnodeops; | |
884 | vp->v_tag = VT_NON; | |
885 | /* | |
886 | * If purging an active vnode, it must be unlocked, closed, | |
887 | * and deactivated before being reclaimed. | |
888 | */ | |
889 | (*(origops->vop_unlock))(vp); | |
890 | if (active) { | |
891 | if (flags & DOCLOSE) | |
892 | (*(origops->vop_close))(vp, IO_NDELAY, NOCRED, p); | |
893 | (*(origops->vop_inactive))(vp, p); | |
894 | } | |
895 | /* | |
896 | * Reclaim the vnode. | |
897 | */ | |
898 | if ((*(origops->vop_reclaim))(vp)) | |
899 | panic("vclean: cannot reclaim"); | |
900 | if (active) | |
901 | vrele(vp); | |
902 | /* | |
903 | * Done with purge, notify sleepers in vget of the grim news. | |
904 | */ | |
905 | vp->v_flag &= ~VXLOCK; | |
906 | if (vp->v_flag & VXWANT) { | |
907 | vp->v_flag &= ~VXWANT; | |
908 | wakeup((caddr_t)vp); | |
909 | } | |
910 | } | |
911 | ||
912 | /* | |
913 | * Eliminate all activity associated with the requested vnode | |
914 | * and with all vnodes aliased to the requested vnode. | |
915 | */ | |
916 | void vgoneall(vp) | |
917 | register struct vnode *vp; | |
918 | { | |
919 | register struct vnode *vq; | |
920 | ||
921 | if (vp->v_flag & VALIASED) { | |
922 | /* | |
923 | * If a vgone (or vclean) is already in progress, | |
924 | * wait until it is done and return. | |
925 | */ | |
926 | if (vp->v_flag & VXLOCK) { | |
927 | vp->v_flag |= VXWANT; | |
73419b27 | 928 | tsleep((caddr_t)vp, PINOD, "vgoneall", 0); |
15637ed4 RG |
929 | return; |
930 | } | |
931 | /* | |
932 | * Ensure that vp will not be vgone'd while we | |
933 | * are eliminating its aliases. | |
934 | */ | |
935 | vp->v_flag |= VXLOCK; | |
936 | while (vp->v_flag & VALIASED) { | |
937 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
938 | if (vq->v_rdev != vp->v_rdev || | |
939 | vq->v_type != vp->v_type || vp == vq) | |
940 | continue; | |
941 | vgone(vq); | |
942 | break; | |
943 | } | |
944 | } | |
945 | /* | |
946 | * Remove the lock so that vgone below will | |
947 | * really eliminate the vnode after which time | |
948 | * vgone will awaken any sleepers. | |
949 | */ | |
950 | vp->v_flag &= ~VXLOCK; | |
951 | } | |
952 | vgone(vp); | |
953 | } | |
954 | ||
955 | /* | |
956 | * Eliminate all activity associated with a vnode | |
957 | * in preparation for reuse. | |
958 | */ | |
4c45483e GW |
959 | void |
960 | vgone(vp) | |
15637ed4 RG |
961 | register struct vnode *vp; |
962 | { | |
963 | register struct vnode *vq; | |
4c45483e | 964 | struct vnode *vx = 0; |
15637ed4 RG |
965 | long count; |
966 | ||
967 | /* | |
968 | * If a vgone (or vclean) is already in progress, | |
969 | * wait until it is done and return. | |
970 | */ | |
971 | if (vp->v_flag & VXLOCK) { | |
972 | vp->v_flag |= VXWANT; | |
73419b27 | 973 | tsleep((caddr_t)vp, PINOD, "vgone", 0); |
15637ed4 RG |
974 | return; |
975 | } | |
976 | /* | |
977 | * Clean out the filesystem specific data. | |
978 | */ | |
979 | vclean(vp, DOCLOSE); | |
980 | /* | |
981 | * Delete from old mount point vnode list, if on one. | |
982 | */ | |
983 | if (vp->v_mountb) { | |
984 | if (vq = vp->v_mountf) | |
985 | vq->v_mountb = vp->v_mountb; | |
986 | *vp->v_mountb = vq; | |
987 | vp->v_mountf = NULL; | |
988 | vp->v_mountb = NULL; | |
989 | } | |
990 | /* | |
991 | * If special device, remove it from special device alias list. | |
992 | */ | |
993 | if (vp->v_type == VBLK || vp->v_type == VCHR) { | |
994 | if (*vp->v_hashchain == vp) { | |
995 | *vp->v_hashchain = vp->v_specnext; | |
996 | } else { | |
997 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
998 | if (vq->v_specnext != vp) | |
999 | continue; | |
1000 | vq->v_specnext = vp->v_specnext; | |
1001 | break; | |
1002 | } | |
1003 | if (vq == NULL) | |
1004 | panic("missing bdev"); | |
1005 | } | |
1006 | if (vp->v_flag & VALIASED) { | |
1007 | count = 0; | |
1008 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
1009 | if (vq->v_rdev != vp->v_rdev || | |
1010 | vq->v_type != vp->v_type) | |
1011 | continue; | |
1012 | count++; | |
1013 | vx = vq; | |
1014 | } | |
1015 | if (count == 0) | |
1016 | panic("missing alias"); | |
1017 | if (count == 1) | |
1018 | vx->v_flag &= ~VALIASED; | |
1019 | vp->v_flag &= ~VALIASED; | |
1020 | } | |
1021 | FREE(vp->v_specinfo, M_VNODE); | |
1022 | vp->v_specinfo = NULL; | |
1023 | } | |
1024 | /* | |
1025 | * If it is on the freelist, move it to the head of the list. | |
1026 | */ | |
1027 | if (vp->v_freeb) { | |
1028 | if (vq = vp->v_freef) | |
1029 | vq->v_freeb = vp->v_freeb; | |
1030 | else | |
1031 | vfreet = vp->v_freeb; | |
1032 | *vp->v_freeb = vq; | |
1033 | vp->v_freef = vfreeh; | |
1034 | vp->v_freeb = &vfreeh; | |
1035 | vfreeh->v_freeb = &vp->v_freef; | |
1036 | vfreeh = vp; | |
1037 | } | |
1038 | vp->v_type = VBAD; | |
1039 | } | |
1040 | ||
1041 | /* | |
1042 | * Lookup a vnode by device number. | |
1043 | */ | |
4c45483e | 1044 | int |
15637ed4 RG |
1045 | vfinddev(dev, type, vpp) |
1046 | dev_t dev; | |
1047 | enum vtype type; | |
1048 | struct vnode **vpp; | |
1049 | { | |
1050 | register struct vnode *vp; | |
1051 | ||
1052 | for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { | |
1053 | if (dev != vp->v_rdev || type != vp->v_type) | |
1054 | continue; | |
1055 | *vpp = vp; | |
1056 | return (0); | |
1057 | } | |
1058 | return (1); | |
1059 | } | |
1060 | ||
1061 | /* | |
1062 | * Calculate the total number of references to a special device. | |
1063 | */ | |
4c45483e | 1064 | int |
15637ed4 RG |
1065 | vcount(vp) |
1066 | register struct vnode *vp; | |
1067 | { | |
1068 | register struct vnode *vq; | |
1069 | int count; | |
1070 | ||
1071 | if ((vp->v_flag & VALIASED) == 0) | |
1072 | return (vp->v_usecount); | |
1073 | loop: | |
1074 | for (count = 0, vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
1075 | if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) | |
1076 | continue; | |
1077 | /* | |
1078 | * Alias, but not in use, so flush it out. | |
1079 | */ | |
1080 | if (vq->v_usecount == 0) { | |
1081 | vgone(vq); | |
1082 | goto loop; | |
1083 | } | |
1084 | count += vq->v_usecount; | |
1085 | } | |
1086 | return (count); | |
1087 | } | |
1088 | ||
1089 | /* | |
1090 | * Print out a description of a vnode. | |
1091 | */ | |
1092 | static char *typename[] = | |
1093 | { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; | |
1094 | ||
4c45483e | 1095 | void |
15637ed4 | 1096 | vprint(label, vp) |
4c45483e | 1097 | const char *label; |
15637ed4 RG |
1098 | register struct vnode *vp; |
1099 | { | |
1100 | char buf[64]; | |
1101 | ||
1102 | if (label != NULL) | |
1103 | printf("%s: ", label); | |
1104 | printf("type %s, usecount %d, writecount %d, refcount %d,", | |
1105 | typename[vp->v_type], vp->v_usecount, vp->v_writecount, | |
1106 | vp->v_holdcnt); | |
1107 | buf[0] = '\0'; | |
1108 | if (vp->v_flag & VROOT) | |
1109 | strcat(buf, "|VROOT"); | |
1110 | if (vp->v_flag & VTEXT) | |
1111 | strcat(buf, "|VTEXT"); | |
1112 | if (vp->v_flag & VSYSTEM) | |
1113 | strcat(buf, "|VSYSTEM"); | |
1114 | if (vp->v_flag & VXLOCK) | |
1115 | strcat(buf, "|VXLOCK"); | |
1116 | if (vp->v_flag & VXWANT) | |
1117 | strcat(buf, "|VXWANT"); | |
1118 | if (vp->v_flag & VBWAIT) | |
1119 | strcat(buf, "|VBWAIT"); | |
1120 | if (vp->v_flag & VALIASED) | |
1121 | strcat(buf, "|VALIASED"); | |
1122 | if (buf[0] != '\0') | |
1123 | printf(" flags (%s)", &buf[1]); | |
1124 | printf("\n\t"); | |
1125 | VOP_PRINT(vp); | |
1126 | } | |
1127 | ||
1128 | #ifdef DEBUG | |
1129 | /* | |
1130 | * List all of the locked vnodes in the system. | |
1131 | * Called when debugging the kernel. | |
1132 | */ | |
1133 | printlockedvnodes() | |
1134 | { | |
1135 | register struct mount *mp; | |
1136 | register struct vnode *vp; | |
1137 | ||
1138 | printf("Locked vnodes\n"); | |
1139 | mp = rootfs; | |
1140 | do { | |
1141 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) | |
1142 | if (VOP_ISLOCKED(vp)) | |
1143 | vprint((char *)0, vp); | |
1144 | mp = mp->mnt_next; | |
1145 | } while (mp != rootfs); | |
1146 | } | |
1147 | #endif | |
1148 | ||
1149 | int kinfo_vdebug = 1; | |
1150 | int kinfo_vgetfailed; | |
1151 | #define KINFO_VNODESLOP 10 | |
1152 | /* | |
1153 | * Dump vnode list (via kinfo). | |
1154 | * Copyout address of vnode followed by vnode. | |
1155 | */ | |
1156 | /* ARGSUSED */ | |
4c45483e | 1157 | int |
15637ed4 RG |
1158 | kinfo_vnode(op, where, acopysize, arg, aneeded) |
1159 | int op; | |
1160 | char *where; | |
1161 | int *acopysize, arg, *aneeded; | |
1162 | { | |
1163 | register struct mount *mp = rootfs; | |
1164 | struct mount *omp; | |
1165 | struct vnode *vp; | |
1166 | register char *bp = where, *savebp; | |
1167 | char *ewhere = where + *acopysize; | |
1168 | int error; | |
1169 | ||
1170 | #define VPTRSZ sizeof (struct vnode *) | |
1171 | #define VNODESZ sizeof (struct vnode) | |
1172 | if (where == NULL) { | |
1173 | *aneeded = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); | |
1174 | return (0); | |
1175 | } | |
1176 | ||
1177 | do { | |
1178 | if (vfs_busy(mp)) { | |
1179 | mp = mp->mnt_next; | |
1180 | continue; | |
1181 | } | |
1182 | savebp = bp; | |
1183 | again: | |
1184 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) { | |
1185 | /* | |
1186 | * Check that the vp is still associated with | |
1187 | * this filesystem. RACE: could have been | |
1188 | * recycled onto the same filesystem. | |
1189 | */ | |
1190 | if (vp->v_mount != mp) { | |
1191 | if (kinfo_vdebug) | |
1192 | printf("kinfo: vp changed\n"); | |
1193 | bp = savebp; | |
1194 | goto again; | |
1195 | } | |
1196 | if ((bp + VPTRSZ + VNODESZ <= ewhere) && | |
1197 | ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || | |
1198 | (error = copyout((caddr_t)vp, bp + VPTRSZ, | |
1199 | VNODESZ)))) | |
1200 | return (error); | |
1201 | bp += VPTRSZ + VNODESZ; | |
1202 | } | |
1203 | omp = mp; | |
1204 | mp = mp->mnt_next; | |
1205 | vfs_unbusy(omp); | |
1206 | } while (mp != rootfs); | |
1207 | ||
1208 | *aneeded = bp - where; | |
1209 | if (bp > ewhere) | |
1210 | *acopysize = ewhere - where; | |
1211 | else | |
1212 | *acopysize = bp - where; | |
1213 | return (0); | |
1214 | } |