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 |
80cc9f07 | 34 | * $Id: vfs_subr.c,v 1.3 1993/10/16 15:25:25 rgrimes Exp $ |
15637ed4 RG |
35 | */ |
36 | ||
37 | /* | |
38 | * External virtual filesystem routines | |
39 | */ | |
40 | ||
41 | #include "param.h" | |
42 | #include "proc.h" | |
43 | #include "mount.h" | |
44 | #include "time.h" | |
45 | #include "vnode.h" | |
46 | #include "specdev.h" | |
47 | #include "namei.h" | |
48 | #include "ucred.h" | |
49 | #include "buf.h" | |
50 | #include "errno.h" | |
51 | #include "malloc.h" | |
52 | ||
53 | /* | |
54 | * Remove a mount point from the list of mounted filesystems. | |
55 | * Unmount of the root is illegal. | |
56 | */ | |
57 | void | |
58 | vfs_remove(mp) | |
59 | register struct mount *mp; | |
60 | { | |
61 | ||
62 | if (mp == rootfs) | |
63 | panic("vfs_remove: unmounting root"); | |
64 | mp->mnt_prev->mnt_next = mp->mnt_next; | |
65 | mp->mnt_next->mnt_prev = mp->mnt_prev; | |
66 | mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0; | |
67 | vfs_unlock(mp); | |
68 | } | |
69 | ||
70 | /* | |
71 | * Lock a filesystem. | |
72 | * Used to prevent access to it while mounting and unmounting. | |
73 | */ | |
74 | vfs_lock(mp) | |
75 | register struct mount *mp; | |
76 | { | |
77 | ||
78 | while(mp->mnt_flag & MNT_MLOCK) { | |
79 | mp->mnt_flag |= MNT_MWAIT; | |
80 | sleep((caddr_t)mp, PVFS); | |
81 | } | |
82 | mp->mnt_flag |= MNT_MLOCK; | |
83 | return (0); | |
84 | } | |
85 | ||
86 | /* | |
87 | * Unlock a locked filesystem. | |
88 | * Panic if filesystem is not locked. | |
89 | */ | |
90 | void | |
91 | vfs_unlock(mp) | |
92 | register struct mount *mp; | |
93 | { | |
94 | ||
95 | if ((mp->mnt_flag & MNT_MLOCK) == 0) | |
96 | panic("vfs_unlock: not locked"); | |
97 | mp->mnt_flag &= ~MNT_MLOCK; | |
98 | if (mp->mnt_flag & MNT_MWAIT) { | |
99 | mp->mnt_flag &= ~MNT_MWAIT; | |
100 | wakeup((caddr_t)mp); | |
101 | } | |
102 | } | |
103 | ||
104 | /* | |
105 | * Mark a mount point as busy. | |
106 | * Used to synchronize access and to delay unmounting. | |
107 | */ | |
108 | vfs_busy(mp) | |
109 | register struct mount *mp; | |
110 | { | |
111 | ||
112 | while(mp->mnt_flag & MNT_MPBUSY) { | |
113 | mp->mnt_flag |= MNT_MPWANT; | |
114 | sleep((caddr_t)&mp->mnt_flag, PVFS); | |
115 | } | |
116 | if (mp->mnt_flag & MNT_UNMOUNT) | |
117 | return (1); | |
118 | mp->mnt_flag |= MNT_MPBUSY; | |
119 | return (0); | |
120 | } | |
121 | ||
122 | /* | |
123 | * Free a busy filesystem. | |
124 | * Panic if filesystem is not busy. | |
125 | */ | |
126 | vfs_unbusy(mp) | |
127 | register struct mount *mp; | |
128 | { | |
129 | ||
130 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
131 | panic("vfs_unbusy: not busy"); | |
132 | mp->mnt_flag &= ~MNT_MPBUSY; | |
133 | if (mp->mnt_flag & MNT_MPWANT) { | |
134 | mp->mnt_flag &= ~MNT_MPWANT; | |
135 | wakeup((caddr_t)&mp->mnt_flag); | |
136 | } | |
137 | } | |
138 | ||
139 | /* | |
140 | * Lookup a mount point by filesystem identifier. | |
141 | */ | |
142 | struct mount * | |
143 | getvfs(fsid) | |
144 | fsid_t *fsid; | |
145 | { | |
146 | register struct mount *mp; | |
147 | ||
148 | mp = rootfs; | |
149 | do { | |
150 | if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && | |
151 | mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { | |
152 | return (mp); | |
153 | } | |
154 | mp = mp->mnt_next; | |
155 | } while (mp != rootfs); | |
156 | return ((struct mount *)0); | |
157 | } | |
158 | ||
159 | /* | |
160 | * Set vnode attributes to VNOVAL | |
161 | */ | |
162 | void vattr_null(vap) | |
163 | register struct vattr *vap; | |
164 | { | |
165 | ||
166 | vap->va_type = VNON; | |
167 | vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = | |
168 | vap->va_fsid = vap->va_fileid = vap->va_size = | |
169 | vap->va_size_rsv = vap->va_blocksize = vap->va_rdev = | |
170 | vap->va_bytes = vap->va_bytes_rsv = | |
171 | vap->va_atime.tv_sec = vap->va_atime.tv_usec = | |
172 | vap->va_mtime.tv_sec = vap->va_mtime.tv_usec = | |
173 | vap->va_ctime.tv_sec = vap->va_ctime.tv_usec = | |
174 | vap->va_flags = vap->va_gen = VNOVAL; | |
175 | } | |
176 | ||
177 | /* | |
178 | * Routines having to do with the management of the vnode table. | |
179 | */ | |
180 | struct vnode *vfreeh, **vfreet; | |
181 | extern struct vnodeops dead_vnodeops, spec_vnodeops; | |
182 | extern void vclean(); | |
183 | long numvnodes; | |
184 | struct vattr va_null; | |
185 | ||
186 | /* | |
187 | * Initialize the vnode structures and initialize each file system type. | |
188 | */ | |
189 | vfsinit() | |
190 | { | |
191 | struct vfsops **vfsp; | |
192 | ||
193 | /* | |
194 | * Initialize the vnode name cache | |
195 | */ | |
196 | nchinit(); | |
197 | /* | |
198 | * Initialize each file system type. | |
199 | */ | |
200 | vattr_null(&va_null); | |
201 | for (vfsp = &vfssw[0]; vfsp <= &vfssw[MOUNT_MAXTYPE]; vfsp++) { | |
202 | if (*vfsp == NULL) | |
203 | continue; | |
204 | (*(*vfsp)->vfs_init)(); | |
205 | } | |
206 | } | |
207 | ||
208 | /* | |
209 | * Return the next vnode from the free list. | |
210 | */ | |
211 | getnewvnode(tag, mp, vops, vpp) | |
212 | enum vtagtype tag; | |
213 | struct mount *mp; | |
214 | struct vnodeops *vops; | |
215 | struct vnode **vpp; | |
216 | { | |
217 | register struct vnode *vp, *vq; | |
218 | ||
219 | if (numvnodes < desiredvnodes) { | |
220 | vp = (struct vnode *)malloc((u_long)sizeof *vp, | |
221 | M_VNODE, M_WAITOK); | |
222 | bzero((char *)vp, sizeof *vp); | |
223 | numvnodes++; | |
224 | } else { | |
225 | if ((vp = vfreeh) == NULL) { | |
226 | tablefull("vnode"); | |
227 | *vpp = 0; | |
228 | return (ENFILE); | |
229 | } | |
230 | if (vp->v_usecount) | |
231 | panic("free vnode isn't"); | |
232 | if (vq = vp->v_freef) | |
233 | vq->v_freeb = &vfreeh; | |
234 | else | |
235 | vfreet = &vfreeh; | |
236 | vfreeh = vq; | |
237 | vp->v_freef = NULL; | |
238 | vp->v_freeb = NULL; | |
239 | if (vp->v_type != VBAD) | |
240 | vgone(vp); | |
241 | vp->v_flag = 0; | |
242 | vp->v_lastr = 0; | |
243 | vp->v_socket = 0; | |
244 | } | |
245 | vp->v_type = VNON; | |
246 | cache_purge(vp); | |
247 | vp->v_tag = tag; | |
248 | vp->v_op = vops; | |
249 | insmntque(vp, mp); | |
250 | VREF(vp); | |
251 | *vpp = vp; | |
252 | return (0); | |
253 | } | |
254 | ||
255 | /* | |
256 | * Move a vnode from one mount queue to another. | |
257 | */ | |
258 | insmntque(vp, mp) | |
259 | register struct vnode *vp; | |
260 | register struct mount *mp; | |
261 | { | |
262 | register struct vnode *vq; | |
263 | ||
264 | /* | |
265 | * Delete from old mount point vnode list, if on one. | |
266 | */ | |
267 | if (vp->v_mountb) { | |
268 | if (vq = vp->v_mountf) | |
269 | vq->v_mountb = vp->v_mountb; | |
270 | *vp->v_mountb = vq; | |
271 | } | |
272 | /* | |
273 | * Insert into list of vnodes for the new mount point, if available. | |
274 | */ | |
275 | vp->v_mount = mp; | |
276 | if (mp == NULL) { | |
277 | vp->v_mountf = NULL; | |
278 | vp->v_mountb = NULL; | |
279 | return; | |
280 | } | |
281 | if (vq = mp->mnt_mounth) | |
282 | vq->v_mountb = &vp->v_mountf; | |
283 | vp->v_mountf = vq; | |
284 | vp->v_mountb = &mp->mnt_mounth; | |
285 | mp->mnt_mounth = vp; | |
286 | } | |
287 | ||
288 | /* | |
289 | * Make sure all write-behind blocks associated | |
290 | * with mount point are flushed out (from sync). | |
291 | */ | |
292 | mntflushbuf(mountp, flags) | |
293 | struct mount *mountp; | |
294 | int flags; | |
295 | { | |
296 | register struct vnode *vp; | |
297 | ||
298 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
299 | panic("mntflushbuf: not busy"); | |
300 | loop: | |
301 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
302 | if (VOP_ISLOCKED(vp)) | |
303 | continue; | |
304 | if (vget(vp)) | |
305 | goto loop; | |
306 | vflushbuf(vp, flags); | |
307 | vput(vp); | |
308 | if (vp->v_mount != mountp) | |
309 | goto loop; | |
310 | } | |
311 | } | |
312 | ||
313 | /* | |
314 | * Flush all dirty buffers associated with a vnode. | |
315 | */ | |
316 | vflushbuf(vp, flags) | |
317 | register struct vnode *vp; | |
318 | int flags; | |
319 | { | |
320 | register struct buf *bp; | |
321 | struct buf *nbp; | |
322 | int s; | |
323 | ||
324 | loop: | |
325 | s = splbio(); | |
326 | for (bp = vp->v_dirtyblkhd; bp; bp = nbp) { | |
327 | nbp = bp->b_blockf; | |
328 | if ((bp->b_flags & B_BUSY)) | |
329 | continue; | |
330 | if ((bp->b_flags & B_DELWRI) == 0) | |
331 | panic("vflushbuf: not dirty"); | |
332 | bremfree(bp); | |
333 | bp->b_flags |= B_BUSY; | |
334 | splx(s); | |
335 | /* | |
336 | * Wait for I/O associated with indirect blocks to complete, | |
337 | * since there is no way to quickly wait for them below. | |
338 | * NB: This is really specific to ufs, but is done here | |
339 | * as it is easier and quicker. | |
340 | */ | |
341 | if (bp->b_vp == vp || (flags & B_SYNC) == 0) | |
342 | (void) bawrite(bp); | |
343 | else | |
344 | (void) bwrite(bp); | |
345 | goto loop; | |
346 | } | |
347 | splx(s); | |
348 | if ((flags & B_SYNC) == 0) | |
349 | return; | |
350 | s = splbio(); | |
351 | while (vp->v_numoutput) { | |
352 | vp->v_flag |= VBWAIT; | |
353 | sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); | |
354 | } | |
355 | splx(s); | |
356 | if (vp->v_dirtyblkhd) { | |
357 | vprint("vflushbuf: dirty", vp); | |
358 | goto loop; | |
359 | } | |
360 | } | |
361 | ||
362 | /* | |
363 | * Update outstanding I/O count and do wakeup if requested. | |
364 | */ | |
365 | vwakeup(bp) | |
366 | register struct buf *bp; | |
367 | { | |
368 | register struct vnode *vp; | |
369 | ||
370 | bp->b_dirtyoff = bp->b_dirtyend = 0; | |
371 | if (vp = bp->b_vp) { | |
372 | vp->v_numoutput--; | |
373 | if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { | |
374 | if (vp->v_numoutput < 0) | |
375 | panic("vwakeup: neg numoutput"); | |
376 | vp->v_flag &= ~VBWAIT; | |
377 | wakeup((caddr_t)&vp->v_numoutput); | |
378 | } | |
379 | } | |
380 | } | |
381 | ||
382 | /* | |
383 | * Invalidate in core blocks belonging to closed or umounted filesystem | |
384 | * | |
385 | * Go through the list of vnodes associated with the file system; | |
386 | * for each vnode invalidate any buffers that it holds. Normally | |
387 | * this routine is preceeded by a bflush call, so that on a quiescent | |
388 | * filesystem there will be no dirty buffers when we are done. Binval | |
389 | * returns the count of dirty buffers when it is finished. | |
390 | */ | |
391 | mntinvalbuf(mountp) | |
392 | struct mount *mountp; | |
393 | { | |
394 | register struct vnode *vp; | |
395 | int dirty = 0; | |
396 | ||
397 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
398 | panic("mntinvalbuf: not busy"); | |
399 | loop: | |
400 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
401 | if (vget(vp)) | |
402 | goto loop; | |
403 | dirty += vinvalbuf(vp, 1); | |
404 | vput(vp); | |
405 | if (vp->v_mount != mountp) | |
406 | goto loop; | |
407 | } | |
408 | return (dirty); | |
409 | } | |
410 | ||
411 | /* | |
412 | * Flush out and invalidate all buffers associated with a vnode. | |
413 | * Called with the underlying object locked. | |
414 | */ | |
415 | vinvalbuf(vp, save) | |
416 | register struct vnode *vp; | |
417 | int save; | |
418 | { | |
419 | register struct buf *bp; | |
420 | struct buf *nbp, *blist; | |
421 | int s, dirty = 0; | |
422 | ||
423 | for (;;) { | |
424 | if (blist = vp->v_dirtyblkhd) | |
425 | /* void */; | |
426 | else if (blist = vp->v_cleanblkhd) | |
427 | /* void */; | |
428 | else | |
429 | break; | |
430 | for (bp = blist; bp; bp = nbp) { | |
431 | nbp = bp->b_blockf; | |
432 | s = splbio(); | |
433 | if (bp->b_flags & B_BUSY) { | |
434 | bp->b_flags |= B_WANTED; | |
435 | sleep((caddr_t)bp, PRIBIO + 1); | |
436 | splx(s); | |
437 | break; | |
438 | } | |
439 | bremfree(bp); | |
440 | bp->b_flags |= B_BUSY; | |
441 | splx(s); | |
442 | if (save && (bp->b_flags & B_DELWRI)) { | |
443 | dirty++; | |
444 | (void) bwrite(bp); | |
445 | break; | |
446 | } | |
447 | if (bp->b_vp != vp) | |
448 | reassignbuf(bp, bp->b_vp); | |
449 | else | |
450 | bp->b_flags |= B_INVAL; | |
451 | brelse(bp); | |
452 | } | |
453 | } | |
454 | if (vp->v_dirtyblkhd || vp->v_cleanblkhd) | |
455 | panic("vinvalbuf: flush failed"); | |
456 | return (dirty); | |
457 | } | |
458 | ||
459 | /* | |
460 | * Associate a buffer with a vnode. | |
461 | */ | |
462 | bgetvp(vp, bp) | |
463 | register struct vnode *vp; | |
464 | register struct buf *bp; | |
465 | { | |
466 | register struct vnode *vq; | |
467 | register struct buf *bq; | |
468 | ||
469 | if (bp->b_vp) | |
470 | panic("bgetvp: not free"); | |
471 | VHOLD(vp); | |
472 | bp->b_vp = vp; | |
473 | if (vp->v_type == VBLK || vp->v_type == VCHR) | |
474 | bp->b_dev = vp->v_rdev; | |
475 | else | |
476 | bp->b_dev = NODEV; | |
477 | /* | |
478 | * Insert onto list for new vnode. | |
479 | */ | |
480 | if (bq = vp->v_cleanblkhd) | |
481 | bq->b_blockb = &bp->b_blockf; | |
482 | bp->b_blockf = bq; | |
483 | bp->b_blockb = &vp->v_cleanblkhd; | |
484 | vp->v_cleanblkhd = bp; | |
485 | } | |
486 | ||
487 | /* | |
488 | * Disassociate a buffer from a vnode. | |
489 | */ | |
490 | brelvp(bp) | |
491 | register struct buf *bp; | |
492 | { | |
493 | struct buf *bq; | |
494 | struct vnode *vp; | |
495 | ||
496 | if (bp->b_vp == (struct vnode *) 0) | |
497 | panic("brelvp: NULL"); | |
498 | /* | |
499 | * Delete from old vnode list, if on one. | |
500 | */ | |
501 | if (bp->b_blockb) { | |
502 | if (bq = bp->b_blockf) | |
503 | bq->b_blockb = bp->b_blockb; | |
504 | *bp->b_blockb = bq; | |
505 | bp->b_blockf = NULL; | |
506 | bp->b_blockb = NULL; | |
507 | } | |
508 | vp = bp->b_vp; | |
509 | bp->b_vp = (struct vnode *) 0; | |
510 | HOLDRELE(vp); | |
511 | } | |
512 | ||
513 | /* | |
514 | * Reassign a buffer from one vnode to another. | |
515 | * Used to assign file specific control information | |
516 | * (indirect blocks) to the vnode to which they belong. | |
517 | */ | |
518 | reassignbuf(bp, newvp) | |
519 | register struct buf *bp; | |
520 | register struct vnode *newvp; | |
521 | { | |
522 | register struct buf *bq, **listheadp; | |
523 | ||
524 | if (newvp == NULL) | |
525 | panic("reassignbuf: NULL"); | |
526 | /* | |
527 | * Delete from old vnode list, if on one. | |
528 | */ | |
529 | if (bp->b_blockb) { | |
530 | if (bq = bp->b_blockf) | |
531 | bq->b_blockb = bp->b_blockb; | |
532 | *bp->b_blockb = bq; | |
533 | } | |
534 | /* | |
535 | * If dirty, put on list of dirty buffers; | |
536 | * otherwise insert onto list of clean buffers. | |
537 | */ | |
538 | if (bp->b_flags & B_DELWRI) | |
539 | listheadp = &newvp->v_dirtyblkhd; | |
540 | else | |
541 | listheadp = &newvp->v_cleanblkhd; | |
542 | if (bq = *listheadp) | |
543 | bq->b_blockb = &bp->b_blockf; | |
544 | bp->b_blockf = bq; | |
545 | bp->b_blockb = listheadp; | |
546 | *listheadp = bp; | |
547 | } | |
548 | ||
549 | /* | |
550 | * Create a vnode for a block device. | |
551 | * Used for root filesystem, argdev, and swap areas. | |
552 | * Also used for memory file system special devices. | |
553 | */ | |
554 | bdevvp(dev, vpp) | |
555 | dev_t dev; | |
556 | struct vnode **vpp; | |
557 | { | |
558 | register struct vnode *vp; | |
559 | struct vnode *nvp; | |
560 | int error; | |
561 | ||
562 | if (dev == NODEV) | |
563 | return (0); | |
564 | error = getnewvnode(VT_NON, (struct mount *)0, &spec_vnodeops, &nvp); | |
565 | if (error) { | |
80cc9f07 | 566 | *vpp = 0; |
15637ed4 RG |
567 | return (error); |
568 | } | |
569 | vp = nvp; | |
80cc9f07 | 570 | vp->v_type = VBLK; |
15637ed4 RG |
571 | if (nvp = checkalias(vp, dev, (struct mount *)0)) { |
572 | vput(vp); | |
573 | vp = nvp; | |
574 | } | |
575 | *vpp = vp; | |
576 | return (0); | |
577 | } | |
578 | ||
579 | /* | |
580 | * Check to see if the new vnode represents a special device | |
581 | * for which we already have a vnode (either because of | |
582 | * bdevvp() or because of a different vnode representing | |
583 | * the same block device). If such an alias exists, deallocate | |
584 | * the existing contents and return the aliased vnode. The | |
585 | * caller is responsible for filling it with its new contents. | |
586 | */ | |
587 | struct vnode * | |
588 | checkalias(nvp, nvp_rdev, mp) | |
589 | register struct vnode *nvp; | |
590 | dev_t nvp_rdev; | |
591 | struct mount *mp; | |
592 | { | |
593 | register struct vnode *vp; | |
594 | struct vnode **vpp; | |
595 | ||
596 | if (nvp->v_type != VBLK && nvp->v_type != VCHR) | |
597 | return (NULLVP); | |
598 | ||
599 | vpp = &speclisth[SPECHASH(nvp_rdev)]; | |
600 | loop: | |
601 | for (vp = *vpp; vp; vp = vp->v_specnext) { | |
602 | if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) | |
603 | continue; | |
604 | /* | |
605 | * Alias, but not in use, so flush it out. | |
606 | */ | |
607 | if (vp->v_usecount == 0) { | |
608 | vgone(vp); | |
609 | goto loop; | |
610 | } | |
611 | if (vget(vp)) | |
612 | goto loop; | |
613 | break; | |
614 | } | |
615 | if (vp == NULL || vp->v_tag != VT_NON) { | |
616 | MALLOC(nvp->v_specinfo, struct specinfo *, | |
617 | sizeof(struct specinfo), M_VNODE, M_WAITOK); | |
618 | nvp->v_rdev = nvp_rdev; | |
619 | nvp->v_hashchain = vpp; | |
620 | nvp->v_specnext = *vpp; | |
621 | nvp->v_specflags = 0; | |
622 | *vpp = nvp; | |
623 | if (vp != NULL) { | |
624 | nvp->v_flag |= VALIASED; | |
625 | vp->v_flag |= VALIASED; | |
626 | vput(vp); | |
627 | } | |
628 | return (NULLVP); | |
629 | } | |
630 | VOP_UNLOCK(vp); | |
631 | vclean(vp, 0); | |
632 | vp->v_op = nvp->v_op; | |
633 | vp->v_tag = nvp->v_tag; | |
634 | nvp->v_type = VNON; | |
635 | insmntque(vp, mp); | |
636 | return (vp); | |
637 | } | |
638 | ||
639 | /* | |
640 | * Grab a particular vnode from the free list, increment its | |
641 | * reference count and lock it. The vnode lock bit is set the | |
642 | * vnode is being eliminated in vgone. The process is awakened | |
643 | * when the transition is completed, and an error returned to | |
644 | * indicate that the vnode is no longer usable (possibly having | |
645 | * been changed to a new file system type). | |
646 | */ | |
647 | vget(vp) | |
648 | register struct vnode *vp; | |
649 | { | |
650 | register struct vnode *vq; | |
651 | ||
652 | if (vp->v_flag & VXLOCK) { | |
653 | vp->v_flag |= VXWANT; | |
654 | sleep((caddr_t)vp, PINOD); | |
655 | return (1); | |
656 | } | |
657 | if (vp->v_usecount == 0) { | |
658 | if (vq = vp->v_freef) | |
659 | vq->v_freeb = vp->v_freeb; | |
660 | else | |
661 | vfreet = vp->v_freeb; | |
662 | *vp->v_freeb = vq; | |
663 | vp->v_freef = NULL; | |
664 | vp->v_freeb = NULL; | |
665 | } | |
666 | VREF(vp); | |
667 | VOP_LOCK(vp); | |
668 | return (0); | |
669 | } | |
670 | ||
671 | /* | |
672 | * Vnode reference, just increment the count | |
673 | */ | |
674 | void vref(vp) | |
675 | struct vnode *vp; | |
676 | { | |
677 | ||
678 | vp->v_usecount++; | |
679 | } | |
680 | ||
681 | /* | |
682 | * vput(), just unlock and vrele() | |
683 | */ | |
684 | void vput(vp) | |
685 | register struct vnode *vp; | |
686 | { | |
687 | VOP_UNLOCK(vp); | |
688 | vrele(vp); | |
689 | } | |
690 | ||
691 | /* | |
692 | * Vnode release. | |
693 | * If count drops to zero, call inactive routine and return to freelist. | |
694 | */ | |
695 | void vrele(vp) | |
696 | register struct vnode *vp; | |
697 | { | |
698 | struct proc *p = curproc; /* XXX */ | |
699 | ||
700 | #ifdef DIAGNOSTIC | |
701 | if (vp == NULL) | |
702 | panic("vrele: null vp"); | |
703 | #endif | |
704 | vp->v_usecount--; | |
705 | if (vp->v_usecount > 0) | |
706 | return; | |
707 | #ifdef DIAGNOSTIC | |
708 | if (vp->v_usecount != 0 || vp->v_writecount != 0) { | |
709 | vprint("vrele: bad ref count", vp); | |
710 | panic("vrele: ref cnt"); | |
711 | } | |
712 | #endif | |
713 | if (vfreeh == NULLVP) { | |
714 | /* | |
715 | * insert into empty list | |
716 | */ | |
717 | vfreeh = vp; | |
718 | vp->v_freeb = &vfreeh; | |
719 | } else { | |
720 | /* | |
721 | * insert at tail of list | |
722 | */ | |
723 | *vfreet = vp; | |
724 | vp->v_freeb = vfreet; | |
725 | } | |
726 | vp->v_freef = NULL; | |
727 | vfreet = &vp->v_freef; | |
728 | VOP_INACTIVE(vp, p); | |
729 | } | |
730 | ||
731 | /* | |
732 | * Page or buffer structure gets a reference. | |
733 | */ | |
734 | vhold(vp) | |
735 | register struct vnode *vp; | |
736 | { | |
737 | ||
738 | vp->v_holdcnt++; | |
739 | } | |
740 | ||
741 | /* | |
742 | * Page or buffer structure frees a reference. | |
743 | */ | |
744 | holdrele(vp) | |
745 | register struct vnode *vp; | |
746 | { | |
747 | ||
748 | if (vp->v_holdcnt <= 0) | |
749 | panic("holdrele: holdcnt"); | |
750 | vp->v_holdcnt--; | |
751 | } | |
752 | ||
753 | /* | |
754 | * Remove any vnodes in the vnode table belonging to mount point mp. | |
755 | * | |
756 | * If MNT_NOFORCE is specified, there should not be any active ones, | |
757 | * return error if any are found (nb: this is a user error, not a | |
758 | * system error). If MNT_FORCE is specified, detach any active vnodes | |
759 | * that are found. | |
760 | */ | |
761 | int busyprt = 0; /* patch to print out busy vnodes */ | |
762 | ||
763 | vflush(mp, skipvp, flags) | |
764 | struct mount *mp; | |
765 | struct vnode *skipvp; | |
766 | int flags; | |
767 | { | |
768 | register struct vnode *vp, *nvp; | |
769 | int busy = 0; | |
770 | ||
771 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
772 | panic("vflush: not busy"); | |
773 | loop: | |
774 | for (vp = mp->mnt_mounth; vp; vp = nvp) { | |
775 | if (vp->v_mount != mp) | |
776 | goto loop; | |
777 | nvp = vp->v_mountf; | |
778 | /* | |
779 | * Skip over a selected vnode. | |
780 | */ | |
781 | if (vp == skipvp) | |
782 | continue; | |
783 | /* | |
784 | * Skip over a vnodes marked VSYSTEM. | |
785 | */ | |
786 | if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) | |
787 | continue; | |
788 | /* | |
789 | * With v_usecount == 0, all we need to do is clear | |
790 | * out the vnode data structures and we are done. | |
791 | */ | |
792 | if (vp->v_usecount == 0) { | |
793 | vgone(vp); | |
794 | continue; | |
795 | } | |
796 | /* | |
797 | * For block or character devices, revert to an | |
798 | * anonymous device. For all other files, just kill them. | |
799 | */ | |
800 | if (flags & FORCECLOSE) { | |
801 | if (vp->v_type != VBLK && vp->v_type != VCHR) { | |
802 | vgone(vp); | |
803 | } else { | |
804 | vclean(vp, 0); | |
805 | vp->v_op = &spec_vnodeops; | |
806 | insmntque(vp, (struct mount *)0); | |
807 | } | |
808 | continue; | |
809 | } | |
810 | if (busyprt) | |
811 | vprint("vflush: busy vnode", vp); | |
812 | busy++; | |
813 | } | |
814 | if (busy) | |
815 | return (EBUSY); | |
816 | return (0); | |
817 | } | |
818 | ||
819 | /* | |
820 | * Disassociate the underlying file system from a vnode. | |
821 | */ | |
822 | void vclean(vp, flags) | |
823 | register struct vnode *vp; | |
824 | int flags; | |
825 | { | |
826 | struct vnodeops *origops; | |
827 | int active; | |
828 | struct proc *p = curproc; /* XXX */ | |
829 | ||
830 | /* | |
831 | * Check to see if the vnode is in use. | |
832 | * If so we have to reference it before we clean it out | |
833 | * so that its count cannot fall to zero and generate a | |
834 | * race against ourselves to recycle it. | |
835 | */ | |
836 | if (active = vp->v_usecount) | |
837 | VREF(vp); | |
838 | /* | |
839 | * Prevent the vnode from being recycled or | |
840 | * brought into use while we clean it out. | |
841 | */ | |
842 | if (vp->v_flag & VXLOCK) | |
843 | panic("vclean: deadlock"); | |
844 | vp->v_flag |= VXLOCK; | |
845 | /* | |
846 | * Even if the count is zero, the VOP_INACTIVE routine may still | |
847 | * have the object locked while it cleans it out. The VOP_LOCK | |
848 | * ensures that the VOP_INACTIVE routine is done with its work. | |
849 | * For active vnodes, it ensures that no other activity can | |
850 | * occur while the buffer list is being cleaned out. | |
851 | */ | |
852 | VOP_LOCK(vp); | |
853 | if (flags & DOCLOSE) | |
854 | vinvalbuf(vp, 1); | |
855 | /* | |
856 | * Prevent any further operations on the vnode from | |
857 | * being passed through to the old file system. | |
858 | */ | |
859 | origops = vp->v_op; | |
860 | vp->v_op = &dead_vnodeops; | |
861 | vp->v_tag = VT_NON; | |
862 | /* | |
863 | * If purging an active vnode, it must be unlocked, closed, | |
864 | * and deactivated before being reclaimed. | |
865 | */ | |
866 | (*(origops->vop_unlock))(vp); | |
867 | if (active) { | |
868 | if (flags & DOCLOSE) | |
869 | (*(origops->vop_close))(vp, IO_NDELAY, NOCRED, p); | |
870 | (*(origops->vop_inactive))(vp, p); | |
871 | } | |
872 | /* | |
873 | * Reclaim the vnode. | |
874 | */ | |
875 | if ((*(origops->vop_reclaim))(vp)) | |
876 | panic("vclean: cannot reclaim"); | |
877 | if (active) | |
878 | vrele(vp); | |
879 | /* | |
880 | * Done with purge, notify sleepers in vget of the grim news. | |
881 | */ | |
882 | vp->v_flag &= ~VXLOCK; | |
883 | if (vp->v_flag & VXWANT) { | |
884 | vp->v_flag &= ~VXWANT; | |
885 | wakeup((caddr_t)vp); | |
886 | } | |
887 | } | |
888 | ||
889 | /* | |
890 | * Eliminate all activity associated with the requested vnode | |
891 | * and with all vnodes aliased to the requested vnode. | |
892 | */ | |
893 | void vgoneall(vp) | |
894 | register struct vnode *vp; | |
895 | { | |
896 | register struct vnode *vq; | |
897 | ||
898 | if (vp->v_flag & VALIASED) { | |
899 | /* | |
900 | * If a vgone (or vclean) is already in progress, | |
901 | * wait until it is done and return. | |
902 | */ | |
903 | if (vp->v_flag & VXLOCK) { | |
904 | vp->v_flag |= VXWANT; | |
905 | sleep((caddr_t)vp, PINOD); | |
906 | return; | |
907 | } | |
908 | /* | |
909 | * Ensure that vp will not be vgone'd while we | |
910 | * are eliminating its aliases. | |
911 | */ | |
912 | vp->v_flag |= VXLOCK; | |
913 | while (vp->v_flag & VALIASED) { | |
914 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
915 | if (vq->v_rdev != vp->v_rdev || | |
916 | vq->v_type != vp->v_type || vp == vq) | |
917 | continue; | |
918 | vgone(vq); | |
919 | break; | |
920 | } | |
921 | } | |
922 | /* | |
923 | * Remove the lock so that vgone below will | |
924 | * really eliminate the vnode after which time | |
925 | * vgone will awaken any sleepers. | |
926 | */ | |
927 | vp->v_flag &= ~VXLOCK; | |
928 | } | |
929 | vgone(vp); | |
930 | } | |
931 | ||
932 | /* | |
933 | * Eliminate all activity associated with a vnode | |
934 | * in preparation for reuse. | |
935 | */ | |
936 | void vgone(vp) | |
937 | register struct vnode *vp; | |
938 | { | |
939 | register struct vnode *vq; | |
940 | struct vnode *vx; | |
941 | long count; | |
942 | ||
943 | /* | |
944 | * If a vgone (or vclean) is already in progress, | |
945 | * wait until it is done and return. | |
946 | */ | |
947 | if (vp->v_flag & VXLOCK) { | |
948 | vp->v_flag |= VXWANT; | |
949 | sleep((caddr_t)vp, PINOD); | |
950 | return; | |
951 | } | |
952 | /* | |
953 | * Clean out the filesystem specific data. | |
954 | */ | |
955 | vclean(vp, DOCLOSE); | |
956 | /* | |
957 | * Delete from old mount point vnode list, if on one. | |
958 | */ | |
959 | if (vp->v_mountb) { | |
960 | if (vq = vp->v_mountf) | |
961 | vq->v_mountb = vp->v_mountb; | |
962 | *vp->v_mountb = vq; | |
963 | vp->v_mountf = NULL; | |
964 | vp->v_mountb = NULL; | |
965 | } | |
966 | /* | |
967 | * If special device, remove it from special device alias list. | |
968 | */ | |
969 | if (vp->v_type == VBLK || vp->v_type == VCHR) { | |
970 | if (*vp->v_hashchain == vp) { | |
971 | *vp->v_hashchain = vp->v_specnext; | |
972 | } else { | |
973 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
974 | if (vq->v_specnext != vp) | |
975 | continue; | |
976 | vq->v_specnext = vp->v_specnext; | |
977 | break; | |
978 | } | |
979 | if (vq == NULL) | |
980 | panic("missing bdev"); | |
981 | } | |
982 | if (vp->v_flag & VALIASED) { | |
983 | count = 0; | |
984 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
985 | if (vq->v_rdev != vp->v_rdev || | |
986 | vq->v_type != vp->v_type) | |
987 | continue; | |
988 | count++; | |
989 | vx = vq; | |
990 | } | |
991 | if (count == 0) | |
992 | panic("missing alias"); | |
993 | if (count == 1) | |
994 | vx->v_flag &= ~VALIASED; | |
995 | vp->v_flag &= ~VALIASED; | |
996 | } | |
997 | FREE(vp->v_specinfo, M_VNODE); | |
998 | vp->v_specinfo = NULL; | |
999 | } | |
1000 | /* | |
1001 | * If it is on the freelist, move it to the head of the list. | |
1002 | */ | |
1003 | if (vp->v_freeb) { | |
1004 | if (vq = vp->v_freef) | |
1005 | vq->v_freeb = vp->v_freeb; | |
1006 | else | |
1007 | vfreet = vp->v_freeb; | |
1008 | *vp->v_freeb = vq; | |
1009 | vp->v_freef = vfreeh; | |
1010 | vp->v_freeb = &vfreeh; | |
1011 | vfreeh->v_freeb = &vp->v_freef; | |
1012 | vfreeh = vp; | |
1013 | } | |
1014 | vp->v_type = VBAD; | |
1015 | } | |
1016 | ||
1017 | /* | |
1018 | * Lookup a vnode by device number. | |
1019 | */ | |
1020 | vfinddev(dev, type, vpp) | |
1021 | dev_t dev; | |
1022 | enum vtype type; | |
1023 | struct vnode **vpp; | |
1024 | { | |
1025 | register struct vnode *vp; | |
1026 | ||
1027 | for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { | |
1028 | if (dev != vp->v_rdev || type != vp->v_type) | |
1029 | continue; | |
1030 | *vpp = vp; | |
1031 | return (0); | |
1032 | } | |
1033 | return (1); | |
1034 | } | |
1035 | ||
1036 | /* | |
1037 | * Calculate the total number of references to a special device. | |
1038 | */ | |
1039 | vcount(vp) | |
1040 | register struct vnode *vp; | |
1041 | { | |
1042 | register struct vnode *vq; | |
1043 | int count; | |
1044 | ||
1045 | if ((vp->v_flag & VALIASED) == 0) | |
1046 | return (vp->v_usecount); | |
1047 | loop: | |
1048 | for (count = 0, vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
1049 | if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) | |
1050 | continue; | |
1051 | /* | |
1052 | * Alias, but not in use, so flush it out. | |
1053 | */ | |
1054 | if (vq->v_usecount == 0) { | |
1055 | vgone(vq); | |
1056 | goto loop; | |
1057 | } | |
1058 | count += vq->v_usecount; | |
1059 | } | |
1060 | return (count); | |
1061 | } | |
1062 | ||
1063 | /* | |
1064 | * Print out a description of a vnode. | |
1065 | */ | |
1066 | static char *typename[] = | |
1067 | { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; | |
1068 | ||
1069 | vprint(label, vp) | |
1070 | char *label; | |
1071 | register struct vnode *vp; | |
1072 | { | |
1073 | char buf[64]; | |
1074 | ||
1075 | if (label != NULL) | |
1076 | printf("%s: ", label); | |
1077 | printf("type %s, usecount %d, writecount %d, refcount %d,", | |
1078 | typename[vp->v_type], vp->v_usecount, vp->v_writecount, | |
1079 | vp->v_holdcnt); | |
1080 | buf[0] = '\0'; | |
1081 | if (vp->v_flag & VROOT) | |
1082 | strcat(buf, "|VROOT"); | |
1083 | if (vp->v_flag & VTEXT) | |
1084 | strcat(buf, "|VTEXT"); | |
1085 | if (vp->v_flag & VSYSTEM) | |
1086 | strcat(buf, "|VSYSTEM"); | |
1087 | if (vp->v_flag & VXLOCK) | |
1088 | strcat(buf, "|VXLOCK"); | |
1089 | if (vp->v_flag & VXWANT) | |
1090 | strcat(buf, "|VXWANT"); | |
1091 | if (vp->v_flag & VBWAIT) | |
1092 | strcat(buf, "|VBWAIT"); | |
1093 | if (vp->v_flag & VALIASED) | |
1094 | strcat(buf, "|VALIASED"); | |
1095 | if (buf[0] != '\0') | |
1096 | printf(" flags (%s)", &buf[1]); | |
1097 | printf("\n\t"); | |
1098 | VOP_PRINT(vp); | |
1099 | } | |
1100 | ||
1101 | #ifdef DEBUG | |
1102 | /* | |
1103 | * List all of the locked vnodes in the system. | |
1104 | * Called when debugging the kernel. | |
1105 | */ | |
1106 | printlockedvnodes() | |
1107 | { | |
1108 | register struct mount *mp; | |
1109 | register struct vnode *vp; | |
1110 | ||
1111 | printf("Locked vnodes\n"); | |
1112 | mp = rootfs; | |
1113 | do { | |
1114 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) | |
1115 | if (VOP_ISLOCKED(vp)) | |
1116 | vprint((char *)0, vp); | |
1117 | mp = mp->mnt_next; | |
1118 | } while (mp != rootfs); | |
1119 | } | |
1120 | #endif | |
1121 | ||
1122 | int kinfo_vdebug = 1; | |
1123 | int kinfo_vgetfailed; | |
1124 | #define KINFO_VNODESLOP 10 | |
1125 | /* | |
1126 | * Dump vnode list (via kinfo). | |
1127 | * Copyout address of vnode followed by vnode. | |
1128 | */ | |
1129 | /* ARGSUSED */ | |
1130 | kinfo_vnode(op, where, acopysize, arg, aneeded) | |
1131 | int op; | |
1132 | char *where; | |
1133 | int *acopysize, arg, *aneeded; | |
1134 | { | |
1135 | register struct mount *mp = rootfs; | |
1136 | struct mount *omp; | |
1137 | struct vnode *vp; | |
1138 | register char *bp = where, *savebp; | |
1139 | char *ewhere = where + *acopysize; | |
1140 | int error; | |
1141 | ||
1142 | #define VPTRSZ sizeof (struct vnode *) | |
1143 | #define VNODESZ sizeof (struct vnode) | |
1144 | if (where == NULL) { | |
1145 | *aneeded = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); | |
1146 | return (0); | |
1147 | } | |
1148 | ||
1149 | do { | |
1150 | if (vfs_busy(mp)) { | |
1151 | mp = mp->mnt_next; | |
1152 | continue; | |
1153 | } | |
1154 | savebp = bp; | |
1155 | again: | |
1156 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) { | |
1157 | /* | |
1158 | * Check that the vp is still associated with | |
1159 | * this filesystem. RACE: could have been | |
1160 | * recycled onto the same filesystem. | |
1161 | */ | |
1162 | if (vp->v_mount != mp) { | |
1163 | if (kinfo_vdebug) | |
1164 | printf("kinfo: vp changed\n"); | |
1165 | bp = savebp; | |
1166 | goto again; | |
1167 | } | |
1168 | if ((bp + VPTRSZ + VNODESZ <= ewhere) && | |
1169 | ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || | |
1170 | (error = copyout((caddr_t)vp, bp + VPTRSZ, | |
1171 | VNODESZ)))) | |
1172 | return (error); | |
1173 | bp += VPTRSZ + VNODESZ; | |
1174 | } | |
1175 | omp = mp; | |
1176 | mp = mp->mnt_next; | |
1177 | vfs_unbusy(omp); | |
1178 | } while (mp != rootfs); | |
1179 | ||
1180 | *aneeded = bp - where; | |
1181 | if (bp > ewhere) | |
1182 | *acopysize = ewhere - where; | |
1183 | else | |
1184 | *acopysize = bp - where; | |
1185 | return (0); | |
1186 | } |