Commit | Line | Data |
---|---|---|
b688fc87 WJ |
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 | * | |
33 | * @(#)vfs_subr.c 7.60 (Berkeley) 6/21/91 | |
34 | */ | |
35 | ||
36 | /* | |
37 | * External virtual filesystem routines | |
38 | */ | |
39 | ||
40 | #include "param.h" | |
41 | #include "proc.h" | |
42 | #include "mount.h" | |
43 | #include "time.h" | |
44 | #include "vnode.h" | |
45 | #include "specdev.h" | |
46 | #include "namei.h" | |
47 | #include "ucred.h" | |
48 | #include "buf.h" | |
49 | #include "errno.h" | |
50 | #include "malloc.h" | |
51 | ||
52 | /* | |
53 | * Remove a mount point from the list of mounted filesystems. | |
54 | * Unmount of the root is illegal. | |
55 | */ | |
56 | void | |
57 | vfs_remove(mp) | |
58 | register struct mount *mp; | |
59 | { | |
60 | ||
61 | if (mp == rootfs) | |
62 | panic("vfs_remove: unmounting root"); | |
63 | mp->mnt_prev->mnt_next = mp->mnt_next; | |
64 | mp->mnt_next->mnt_prev = mp->mnt_prev; | |
65 | mp->mnt_vnodecovered->v_mountedhere = (struct mount *)0; | |
66 | vfs_unlock(mp); | |
67 | } | |
68 | ||
69 | /* | |
70 | * Lock a filesystem. | |
71 | * Used to prevent access to it while mounting and unmounting. | |
72 | */ | |
73 | vfs_lock(mp) | |
74 | register struct mount *mp; | |
75 | { | |
76 | ||
77 | while(mp->mnt_flag & MNT_MLOCK) { | |
78 | mp->mnt_flag |= MNT_MWAIT; | |
79 | sleep((caddr_t)mp, PVFS); | |
80 | } | |
81 | mp->mnt_flag |= MNT_MLOCK; | |
82 | return (0); | |
83 | } | |
84 | ||
85 | /* | |
86 | * Unlock a locked filesystem. | |
87 | * Panic if filesystem is not locked. | |
88 | */ | |
89 | void | |
90 | vfs_unlock(mp) | |
91 | register struct mount *mp; | |
92 | { | |
93 | ||
94 | if ((mp->mnt_flag & MNT_MLOCK) == 0) | |
95 | panic("vfs_unlock: not locked"); | |
96 | mp->mnt_flag &= ~MNT_MLOCK; | |
97 | if (mp->mnt_flag & MNT_MWAIT) { | |
98 | mp->mnt_flag &= ~MNT_MWAIT; | |
99 | wakeup((caddr_t)mp); | |
100 | } | |
101 | } | |
102 | ||
103 | /* | |
104 | * Mark a mount point as busy. | |
105 | * Used to synchronize access and to delay unmounting. | |
106 | */ | |
107 | vfs_busy(mp) | |
108 | register struct mount *mp; | |
109 | { | |
110 | ||
111 | while(mp->mnt_flag & MNT_MPBUSY) { | |
112 | mp->mnt_flag |= MNT_MPWANT; | |
113 | sleep((caddr_t)&mp->mnt_flag, PVFS); | |
114 | } | |
115 | if (mp->mnt_flag & MNT_UNMOUNT) | |
116 | return (1); | |
117 | mp->mnt_flag |= MNT_MPBUSY; | |
118 | return (0); | |
119 | } | |
120 | ||
121 | /* | |
122 | * Free a busy filesystem. | |
123 | * Panic if filesystem is not busy. | |
124 | */ | |
125 | vfs_unbusy(mp) | |
126 | register struct mount *mp; | |
127 | { | |
128 | ||
129 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
130 | panic("vfs_unbusy: not busy"); | |
131 | mp->mnt_flag &= ~MNT_MPBUSY; | |
132 | if (mp->mnt_flag & MNT_MPWANT) { | |
133 | mp->mnt_flag &= ~MNT_MPWANT; | |
134 | wakeup((caddr_t)&mp->mnt_flag); | |
135 | } | |
136 | } | |
137 | ||
138 | /* | |
139 | * Lookup a mount point by filesystem identifier. | |
140 | */ | |
141 | struct mount * | |
142 | getvfs(fsid) | |
143 | fsid_t *fsid; | |
144 | { | |
145 | register struct mount *mp; | |
146 | ||
147 | mp = rootfs; | |
148 | do { | |
149 | if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && | |
150 | mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) { | |
151 | return (mp); | |
152 | } | |
153 | mp = mp->mnt_next; | |
154 | } while (mp != rootfs); | |
155 | return ((struct mount *)0); | |
156 | } | |
157 | ||
158 | /* | |
159 | * Set vnode attributes to VNOVAL | |
160 | */ | |
161 | void vattr_null(vap) | |
162 | register struct vattr *vap; | |
163 | { | |
164 | ||
165 | vap->va_type = VNON; | |
166 | vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = | |
167 | vap->va_fsid = vap->va_fileid = vap->va_size = | |
168 | vap->va_size_rsv = vap->va_blocksize = vap->va_rdev = | |
169 | vap->va_bytes = vap->va_bytes_rsv = | |
170 | vap->va_atime.tv_sec = vap->va_atime.tv_usec = | |
171 | vap->va_mtime.tv_sec = vap->va_mtime.tv_usec = | |
172 | vap->va_ctime.tv_sec = vap->va_ctime.tv_usec = | |
173 | vap->va_flags = vap->va_gen = VNOVAL; | |
174 | } | |
175 | ||
176 | /* | |
177 | * Routines having to do with the management of the vnode table. | |
178 | */ | |
179 | struct vnode *vfreeh, **vfreet; | |
180 | extern struct vnodeops dead_vnodeops, spec_vnodeops; | |
181 | extern void vclean(); | |
182 | long numvnodes; | |
183 | struct vattr va_null; | |
184 | ||
185 | /* | |
186 | * Initialize the vnode structures and initialize each file system type. | |
187 | */ | |
188 | vfsinit() | |
189 | { | |
190 | struct vfsops **vfsp; | |
191 | ||
192 | /* | |
193 | * Initialize the vnode name cache | |
194 | */ | |
195 | nchinit(); | |
196 | /* | |
197 | * Initialize each file system type. | |
198 | */ | |
199 | vattr_null(&va_null); | |
200 | for (vfsp = &vfssw[0]; vfsp <= &vfssw[MOUNT_MAXTYPE]; vfsp++) { | |
201 | if (*vfsp == NULL) | |
202 | continue; | |
203 | (*(*vfsp)->vfs_init)(); | |
204 | } | |
205 | } | |
206 | ||
207 | /* | |
208 | * Return the next vnode from the free list. | |
209 | */ | |
210 | getnewvnode(tag, mp, vops, vpp) | |
211 | enum vtagtype tag; | |
212 | struct mount *mp; | |
213 | struct vnodeops *vops; | |
214 | struct vnode **vpp; | |
215 | { | |
216 | register struct vnode *vp, *vq; | |
217 | ||
218 | if (numvnodes < desiredvnodes) { | |
219 | vp = (struct vnode *)malloc((u_long)sizeof *vp, | |
220 | M_VNODE, M_WAITOK); | |
221 | bzero((char *)vp, sizeof *vp); | |
222 | numvnodes++; | |
223 | } else { | |
224 | if ((vp = vfreeh) == NULL) { | |
225 | tablefull("vnode"); | |
226 | *vpp = 0; | |
227 | return (ENFILE); | |
228 | } | |
229 | if (vp->v_usecount) | |
230 | panic("free vnode isn't"); | |
231 | if (vq = vp->v_freef) | |
232 | vq->v_freeb = &vfreeh; | |
233 | else | |
234 | vfreet = &vfreeh; | |
235 | vfreeh = vq; | |
236 | vp->v_freef = NULL; | |
237 | vp->v_freeb = NULL; | |
238 | if (vp->v_type != VBAD) | |
239 | vgone(vp); | |
240 | vp->v_flag = 0; | |
241 | vp->v_lastr = 0; | |
242 | vp->v_socket = 0; | |
243 | } | |
244 | vp->v_type = VNON; | |
245 | cache_purge(vp); | |
246 | vp->v_tag = tag; | |
247 | vp->v_op = vops; | |
248 | insmntque(vp, mp); | |
249 | VREF(vp); | |
250 | *vpp = vp; | |
251 | return (0); | |
252 | } | |
253 | ||
254 | /* | |
255 | * Move a vnode from one mount queue to another. | |
256 | */ | |
257 | insmntque(vp, mp) | |
258 | register struct vnode *vp; | |
259 | register struct mount *mp; | |
260 | { | |
261 | register struct vnode *vq; | |
262 | ||
263 | /* | |
264 | * Delete from old mount point vnode list, if on one. | |
265 | */ | |
266 | if (vp->v_mountb) { | |
267 | if (vq = vp->v_mountf) | |
268 | vq->v_mountb = vp->v_mountb; | |
269 | *vp->v_mountb = vq; | |
270 | } | |
271 | /* | |
272 | * Insert into list of vnodes for the new mount point, if available. | |
273 | */ | |
274 | vp->v_mount = mp; | |
275 | if (mp == NULL) { | |
276 | vp->v_mountf = NULL; | |
277 | vp->v_mountb = NULL; | |
278 | return; | |
279 | } | |
280 | if (vq = mp->mnt_mounth) | |
281 | vq->v_mountb = &vp->v_mountf; | |
282 | vp->v_mountf = vq; | |
283 | vp->v_mountb = &mp->mnt_mounth; | |
284 | mp->mnt_mounth = vp; | |
285 | } | |
286 | ||
287 | /* | |
288 | * Make sure all write-behind blocks associated | |
289 | * with mount point are flushed out (from sync). | |
290 | */ | |
291 | mntflushbuf(mountp, flags) | |
292 | struct mount *mountp; | |
293 | int flags; | |
294 | { | |
295 | register struct vnode *vp; | |
296 | ||
297 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
298 | panic("mntflushbuf: not busy"); | |
299 | loop: | |
300 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
301 | if (VOP_ISLOCKED(vp)) | |
302 | continue; | |
303 | if (vget(vp)) | |
304 | goto loop; | |
305 | vflushbuf(vp, flags); | |
306 | vput(vp); | |
307 | if (vp->v_mount != mountp) | |
308 | goto loop; | |
309 | } | |
310 | } | |
311 | ||
312 | /* | |
313 | * Flush all dirty buffers associated with a vnode. | |
314 | */ | |
315 | vflushbuf(vp, flags) | |
316 | register struct vnode *vp; | |
317 | int flags; | |
318 | { | |
319 | register struct buf *bp; | |
320 | struct buf *nbp; | |
321 | int s; | |
322 | ||
323 | loop: | |
324 | s = splbio(); | |
325 | for (bp = vp->v_dirtyblkhd; bp; bp = nbp) { | |
326 | nbp = bp->b_blockf; | |
327 | if ((bp->b_flags & B_BUSY)) | |
328 | continue; | |
329 | if ((bp->b_flags & B_DELWRI) == 0) | |
330 | panic("vflushbuf: not dirty"); | |
331 | bremfree(bp); | |
332 | bp->b_flags |= B_BUSY; | |
333 | splx(s); | |
334 | /* | |
335 | * Wait for I/O associated with indirect blocks to complete, | |
336 | * since there is no way to quickly wait for them below. | |
337 | * NB: This is really specific to ufs, but is done here | |
338 | * as it is easier and quicker. | |
339 | */ | |
340 | if (bp->b_vp == vp || (flags & B_SYNC) == 0) | |
341 | (void) bawrite(bp); | |
342 | else | |
343 | (void) bwrite(bp); | |
344 | goto loop; | |
345 | } | |
346 | splx(s); | |
347 | if ((flags & B_SYNC) == 0) | |
348 | return; | |
349 | s = splbio(); | |
350 | while (vp->v_numoutput) { | |
351 | vp->v_flag |= VBWAIT; | |
352 | sleep((caddr_t)&vp->v_numoutput, PRIBIO + 1); | |
353 | } | |
354 | splx(s); | |
355 | if (vp->v_dirtyblkhd) { | |
356 | vprint("vflushbuf: dirty", vp); | |
357 | goto loop; | |
358 | } | |
359 | } | |
360 | ||
361 | /* | |
362 | * Update outstanding I/O count and do wakeup if requested. | |
363 | */ | |
364 | vwakeup(bp) | |
365 | register struct buf *bp; | |
366 | { | |
367 | register struct vnode *vp; | |
368 | ||
369 | bp->b_dirtyoff = bp->b_dirtyend = 0; | |
370 | if (vp = bp->b_vp) { | |
371 | vp->v_numoutput--; | |
372 | if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { | |
373 | if (vp->v_numoutput < 0) | |
374 | panic("vwakeup: neg numoutput"); | |
375 | vp->v_flag &= ~VBWAIT; | |
376 | wakeup((caddr_t)&vp->v_numoutput); | |
377 | } | |
378 | } | |
379 | } | |
380 | ||
381 | /* | |
382 | * Invalidate in core blocks belonging to closed or umounted filesystem | |
383 | * | |
384 | * Go through the list of vnodes associated with the file system; | |
385 | * for each vnode invalidate any buffers that it holds. Normally | |
386 | * this routine is preceeded by a bflush call, so that on a quiescent | |
387 | * filesystem there will be no dirty buffers when we are done. Binval | |
388 | * returns the count of dirty buffers when it is finished. | |
389 | */ | |
390 | mntinvalbuf(mountp) | |
391 | struct mount *mountp; | |
392 | { | |
393 | register struct vnode *vp; | |
394 | int dirty = 0; | |
395 | ||
396 | if ((mountp->mnt_flag & MNT_MPBUSY) == 0) | |
397 | panic("mntinvalbuf: not busy"); | |
398 | loop: | |
399 | for (vp = mountp->mnt_mounth; vp; vp = vp->v_mountf) { | |
400 | if (vget(vp)) | |
401 | goto loop; | |
402 | dirty += vinvalbuf(vp, 1); | |
403 | vput(vp); | |
404 | if (vp->v_mount != mountp) | |
405 | goto loop; | |
406 | } | |
407 | return (dirty); | |
408 | } | |
409 | ||
410 | /* | |
411 | * Flush out and invalidate all buffers associated with a vnode. | |
412 | * Called with the underlying object locked. | |
413 | */ | |
414 | vinvalbuf(vp, save) | |
415 | register struct vnode *vp; | |
416 | int save; | |
417 | { | |
418 | register struct buf *bp; | |
419 | struct buf *nbp, *blist; | |
420 | int s, dirty = 0; | |
421 | ||
422 | for (;;) { | |
423 | if (blist = vp->v_dirtyblkhd) | |
424 | /* void */; | |
425 | else if (blist = vp->v_cleanblkhd) | |
426 | /* void */; | |
427 | else | |
428 | break; | |
429 | for (bp = blist; bp; bp = nbp) { | |
430 | nbp = bp->b_blockf; | |
431 | s = splbio(); | |
432 | if (bp->b_flags & B_BUSY) { | |
433 | bp->b_flags |= B_WANTED; | |
434 | sleep((caddr_t)bp, PRIBIO + 1); | |
435 | splx(s); | |
436 | break; | |
437 | } | |
438 | bremfree(bp); | |
439 | bp->b_flags |= B_BUSY; | |
440 | splx(s); | |
441 | if (save && (bp->b_flags & B_DELWRI)) { | |
442 | dirty++; | |
443 | (void) bwrite(bp); | |
444 | break; | |
445 | } | |
446 | if (bp->b_vp != vp) | |
447 | reassignbuf(bp, bp->b_vp); | |
448 | else | |
449 | bp->b_flags |= B_INVAL; | |
450 | brelse(bp); | |
451 | } | |
452 | } | |
453 | if (vp->v_dirtyblkhd || vp->v_cleanblkhd) | |
454 | panic("vinvalbuf: flush failed"); | |
455 | return (dirty); | |
456 | } | |
457 | ||
458 | /* | |
459 | * Associate a buffer with a vnode. | |
460 | */ | |
461 | bgetvp(vp, bp) | |
462 | register struct vnode *vp; | |
463 | register struct buf *bp; | |
464 | { | |
465 | register struct vnode *vq; | |
466 | register struct buf *bq; | |
467 | ||
468 | if (bp->b_vp) | |
469 | panic("bgetvp: not free"); | |
470 | VHOLD(vp); | |
471 | bp->b_vp = vp; | |
472 | if (vp->v_type == VBLK || vp->v_type == VCHR) | |
473 | bp->b_dev = vp->v_rdev; | |
474 | else | |
475 | bp->b_dev = NODEV; | |
476 | /* | |
477 | * Insert onto list for new vnode. | |
478 | */ | |
479 | if (bq = vp->v_cleanblkhd) | |
480 | bq->b_blockb = &bp->b_blockf; | |
481 | bp->b_blockf = bq; | |
482 | bp->b_blockb = &vp->v_cleanblkhd; | |
483 | vp->v_cleanblkhd = bp; | |
484 | } | |
485 | ||
486 | /* | |
487 | * Disassociate a buffer from a vnode. | |
488 | */ | |
489 | brelvp(bp) | |
490 | register struct buf *bp; | |
491 | { | |
492 | struct buf *bq; | |
493 | struct vnode *vp; | |
494 | ||
495 | if (bp->b_vp == (struct vnode *) 0) | |
496 | panic("brelvp: NULL"); | |
497 | /* | |
498 | * Delete from old vnode list, if on one. | |
499 | */ | |
500 | if (bp->b_blockb) { | |
501 | if (bq = bp->b_blockf) | |
502 | bq->b_blockb = bp->b_blockb; | |
503 | *bp->b_blockb = bq; | |
504 | bp->b_blockf = NULL; | |
505 | bp->b_blockb = NULL; | |
506 | } | |
507 | vp = bp->b_vp; | |
508 | bp->b_vp = (struct vnode *) 0; | |
509 | HOLDRELE(vp); | |
510 | } | |
511 | ||
512 | /* | |
513 | * Reassign a buffer from one vnode to another. | |
514 | * Used to assign file specific control information | |
515 | * (indirect blocks) to the vnode to which they belong. | |
516 | */ | |
517 | reassignbuf(bp, newvp) | |
518 | register struct buf *bp; | |
519 | register struct vnode *newvp; | |
520 | { | |
521 | register struct buf *bq, **listheadp; | |
522 | ||
523 | if (newvp == NULL) | |
524 | panic("reassignbuf: NULL"); | |
525 | /* | |
526 | * Delete from old vnode list, if on one. | |
527 | */ | |
528 | if (bp->b_blockb) { | |
529 | if (bq = bp->b_blockf) | |
530 | bq->b_blockb = bp->b_blockb; | |
531 | *bp->b_blockb = bq; | |
532 | } | |
533 | /* | |
534 | * If dirty, put on list of dirty buffers; | |
535 | * otherwise insert onto list of clean buffers. | |
536 | */ | |
537 | if (bp->b_flags & B_DELWRI) | |
538 | listheadp = &newvp->v_dirtyblkhd; | |
539 | else | |
540 | listheadp = &newvp->v_cleanblkhd; | |
541 | if (bq = *listheadp) | |
542 | bq->b_blockb = &bp->b_blockf; | |
543 | bp->b_blockf = bq; | |
544 | bp->b_blockb = listheadp; | |
545 | *listheadp = bp; | |
546 | } | |
547 | ||
548 | /* | |
549 | * Create a vnode for a block device. | |
550 | * Used for root filesystem, argdev, and swap areas. | |
551 | * Also used for memory file system special devices. | |
552 | */ | |
553 | bdevvp(dev, vpp) | |
554 | dev_t dev; | |
555 | struct vnode **vpp; | |
556 | { | |
557 | register struct vnode *vp; | |
558 | struct vnode *nvp; | |
559 | int error; | |
560 | ||
561 | if (dev == NODEV) | |
562 | return (0); | |
563 | error = getnewvnode(VT_NON, (struct mount *)0, &spec_vnodeops, &nvp); | |
564 | if (error) { | |
565 | *vpp = 0; | |
566 | return (error); | |
567 | } | |
568 | vp = nvp; | |
569 | vp->v_type = VBLK; | |
570 | if (nvp = checkalias(vp, dev, (struct mount *)0)) { | |
571 | vput(vp); | |
572 | vp = nvp; | |
573 | } | |
574 | *vpp = vp; | |
575 | return (0); | |
576 | } | |
577 | ||
578 | /* | |
579 | * Check to see if the new vnode represents a special device | |
580 | * for which we already have a vnode (either because of | |
581 | * bdevvp() or because of a different vnode representing | |
582 | * the same block device). If such an alias exists, deallocate | |
583 | * the existing contents and return the aliased vnode. The | |
584 | * caller is responsible for filling it with its new contents. | |
585 | */ | |
586 | struct vnode * | |
587 | checkalias(nvp, nvp_rdev, mp) | |
588 | register struct vnode *nvp; | |
589 | dev_t nvp_rdev; | |
590 | struct mount *mp; | |
591 | { | |
592 | register struct vnode *vp; | |
593 | struct vnode **vpp; | |
594 | ||
595 | if (nvp->v_type != VBLK && nvp->v_type != VCHR) | |
596 | return (NULLVP); | |
597 | ||
598 | vpp = &speclisth[SPECHASH(nvp_rdev)]; | |
599 | loop: | |
600 | for (vp = *vpp; vp; vp = vp->v_specnext) { | |
601 | if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) | |
602 | continue; | |
603 | /* | |
604 | * Alias, but not in use, so flush it out. | |
605 | */ | |
606 | if (vp->v_usecount == 0) { | |
607 | vgone(vp); | |
608 | goto loop; | |
609 | } | |
610 | if (vget(vp)) | |
611 | goto loop; | |
612 | break; | |
613 | } | |
614 | if (vp == NULL || vp->v_tag != VT_NON) { | |
615 | MALLOC(nvp->v_specinfo, struct specinfo *, | |
616 | sizeof(struct specinfo), M_VNODE, M_WAITOK); | |
617 | nvp->v_rdev = nvp_rdev; | |
618 | nvp->v_hashchain = vpp; | |
619 | nvp->v_specnext = *vpp; | |
620 | nvp->v_specflags = 0; | |
621 | *vpp = nvp; | |
622 | if (vp != NULL) { | |
623 | nvp->v_flag |= VALIASED; | |
624 | vp->v_flag |= VALIASED; | |
625 | vput(vp); | |
626 | } | |
627 | return (NULLVP); | |
628 | } | |
629 | VOP_UNLOCK(vp); | |
630 | vclean(vp, 0); | |
631 | vp->v_op = nvp->v_op; | |
632 | vp->v_tag = nvp->v_tag; | |
633 | nvp->v_type = VNON; | |
634 | insmntque(vp, mp); | |
635 | return (vp); | |
636 | } | |
637 | ||
638 | /* | |
639 | * Grab a particular vnode from the free list, increment its | |
640 | * reference count and lock it. The vnode lock bit is set the | |
641 | * vnode is being eliminated in vgone. The process is awakened | |
642 | * when the transition is completed, and an error returned to | |
643 | * indicate that the vnode is no longer usable (possibly having | |
644 | * been changed to a new file system type). | |
645 | */ | |
646 | vget(vp) | |
647 | register struct vnode *vp; | |
648 | { | |
649 | register struct vnode *vq; | |
650 | ||
651 | if (vp->v_flag & VXLOCK) { | |
652 | vp->v_flag |= VXWANT; | |
653 | sleep((caddr_t)vp, PINOD); | |
654 | return (1); | |
655 | } | |
656 | if (vp->v_usecount == 0) { | |
657 | if (vq = vp->v_freef) | |
658 | vq->v_freeb = vp->v_freeb; | |
659 | else | |
660 | vfreet = vp->v_freeb; | |
661 | *vp->v_freeb = vq; | |
662 | vp->v_freef = NULL; | |
663 | vp->v_freeb = NULL; | |
664 | } | |
665 | VREF(vp); | |
666 | VOP_LOCK(vp); | |
667 | return (0); | |
668 | } | |
669 | ||
670 | /* | |
671 | * Vnode reference, just increment the count | |
672 | */ | |
673 | void vref(vp) | |
674 | struct vnode *vp; | |
675 | { | |
676 | ||
677 | vp->v_usecount++; | |
678 | } | |
679 | ||
680 | /* | |
681 | * vput(), just unlock and vrele() | |
682 | */ | |
683 | void vput(vp) | |
684 | register struct vnode *vp; | |
685 | { | |
686 | VOP_UNLOCK(vp); | |
687 | vrele(vp); | |
688 | } | |
689 | ||
690 | /* | |
691 | * Vnode release. | |
692 | * If count drops to zero, call inactive routine and return to freelist. | |
693 | */ | |
694 | void vrele(vp) | |
695 | register struct vnode *vp; | |
696 | { | |
697 | struct proc *p = curproc; /* XXX */ | |
698 | ||
699 | #ifdef DIAGNOSTIC | |
700 | if (vp == NULL) | |
701 | panic("vrele: null vp"); | |
702 | #endif | |
703 | vp->v_usecount--; | |
704 | if (vp->v_usecount > 0) | |
705 | return; | |
706 | #ifdef DIAGNOSTIC | |
707 | if (vp->v_usecount != 0 || vp->v_writecount != 0) { | |
708 | vprint("vrele: bad ref count", vp); | |
709 | panic("vrele: ref cnt"); | |
710 | } | |
711 | #endif | |
712 | if (vfreeh == NULLVP) { | |
713 | /* | |
714 | * insert into empty list | |
715 | */ | |
716 | vfreeh = vp; | |
717 | vp->v_freeb = &vfreeh; | |
718 | } else { | |
719 | /* | |
720 | * insert at tail of list | |
721 | */ | |
722 | *vfreet = vp; | |
723 | vp->v_freeb = vfreet; | |
724 | } | |
725 | vp->v_freef = NULL; | |
726 | vfreet = &vp->v_freef; | |
727 | VOP_INACTIVE(vp, p); | |
728 | } | |
729 | ||
730 | /* | |
731 | * Page or buffer structure gets a reference. | |
732 | */ | |
733 | vhold(vp) | |
734 | register struct vnode *vp; | |
735 | { | |
736 | ||
737 | vp->v_holdcnt++; | |
738 | } | |
739 | ||
740 | /* | |
741 | * Page or buffer structure frees a reference. | |
742 | */ | |
743 | holdrele(vp) | |
744 | register struct vnode *vp; | |
745 | { | |
746 | ||
747 | if (vp->v_holdcnt <= 0) | |
748 | panic("holdrele: holdcnt"); | |
749 | vp->v_holdcnt--; | |
750 | } | |
751 | ||
752 | /* | |
753 | * Remove any vnodes in the vnode table belonging to mount point mp. | |
754 | * | |
755 | * If MNT_NOFORCE is specified, there should not be any active ones, | |
756 | * return error if any are found (nb: this is a user error, not a | |
757 | * system error). If MNT_FORCE is specified, detach any active vnodes | |
758 | * that are found. | |
759 | */ | |
760 | int busyprt = 0; /* patch to print out busy vnodes */ | |
761 | ||
762 | vflush(mp, skipvp, flags) | |
763 | struct mount *mp; | |
764 | struct vnode *skipvp; | |
765 | int flags; | |
766 | { | |
767 | register struct vnode *vp, *nvp; | |
768 | int busy = 0; | |
769 | ||
770 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) | |
771 | panic("vflush: not busy"); | |
772 | loop: | |
773 | for (vp = mp->mnt_mounth; vp; vp = nvp) { | |
774 | if (vp->v_mount != mp) | |
775 | goto loop; | |
776 | nvp = vp->v_mountf; | |
777 | /* | |
778 | * Skip over a selected vnode. | |
779 | */ | |
780 | if (vp == skipvp) | |
781 | continue; | |
782 | /* | |
783 | * Skip over a vnodes marked VSYSTEM. | |
784 | */ | |
785 | if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) | |
786 | continue; | |
787 | /* | |
788 | * With v_usecount == 0, all we need to do is clear | |
789 | * out the vnode data structures and we are done. | |
790 | */ | |
791 | if (vp->v_usecount == 0) { | |
792 | vgone(vp); | |
793 | continue; | |
794 | } | |
795 | /* | |
796 | * For block or character devices, revert to an | |
797 | * anonymous device. For all other files, just kill them. | |
798 | */ | |
799 | if (flags & FORCECLOSE) { | |
800 | if (vp->v_type != VBLK && vp->v_type != VCHR) { | |
801 | vgone(vp); | |
802 | } else { | |
803 | vclean(vp, 0); | |
804 | vp->v_op = &spec_vnodeops; | |
805 | insmntque(vp, (struct mount *)0); | |
806 | } | |
807 | continue; | |
808 | } | |
809 | if (busyprt) | |
810 | vprint("vflush: busy vnode", vp); | |
811 | busy++; | |
812 | } | |
813 | if (busy) | |
814 | return (EBUSY); | |
815 | return (0); | |
816 | } | |
817 | ||
818 | /* | |
819 | * Disassociate the underlying file system from a vnode. | |
820 | */ | |
821 | void vclean(vp, flags) | |
822 | register struct vnode *vp; | |
823 | int flags; | |
824 | { | |
825 | struct vnodeops *origops; | |
826 | int active; | |
827 | struct proc *p = curproc; /* XXX */ | |
828 | ||
829 | /* | |
830 | * Check to see if the vnode is in use. | |
831 | * If so we have to reference it before we clean it out | |
832 | * so that its count cannot fall to zero and generate a | |
833 | * race against ourselves to recycle it. | |
834 | */ | |
835 | if (active = vp->v_usecount) | |
836 | VREF(vp); | |
837 | /* | |
838 | * Prevent the vnode from being recycled or | |
839 | * brought into use while we clean it out. | |
840 | */ | |
841 | if (vp->v_flag & VXLOCK) | |
842 | panic("vclean: deadlock"); | |
843 | vp->v_flag |= VXLOCK; | |
844 | /* | |
845 | * Even if the count is zero, the VOP_INACTIVE routine may still | |
846 | * have the object locked while it cleans it out. The VOP_LOCK | |
847 | * ensures that the VOP_INACTIVE routine is done with its work. | |
848 | * For active vnodes, it ensures that no other activity can | |
849 | * occur while the buffer list is being cleaned out. | |
850 | */ | |
851 | VOP_LOCK(vp); | |
852 | if (flags & DOCLOSE) | |
853 | vinvalbuf(vp, 1); | |
854 | /* | |
855 | * Prevent any further operations on the vnode from | |
856 | * being passed through to the old file system. | |
857 | */ | |
858 | origops = vp->v_op; | |
859 | vp->v_op = &dead_vnodeops; | |
860 | vp->v_tag = VT_NON; | |
861 | /* | |
862 | * If purging an active vnode, it must be unlocked, closed, | |
863 | * and deactivated before being reclaimed. | |
864 | */ | |
865 | (*(origops->vop_unlock))(vp); | |
866 | if (active) { | |
867 | if (flags & DOCLOSE) | |
868 | (*(origops->vop_close))(vp, IO_NDELAY, NOCRED, p); | |
869 | (*(origops->vop_inactive))(vp, p); | |
870 | } | |
871 | /* | |
872 | * Reclaim the vnode. | |
873 | */ | |
874 | if ((*(origops->vop_reclaim))(vp)) | |
875 | panic("vclean: cannot reclaim"); | |
876 | if (active) | |
877 | vrele(vp); | |
878 | /* | |
879 | * Done with purge, notify sleepers in vget of the grim news. | |
880 | */ | |
881 | vp->v_flag &= ~VXLOCK; | |
882 | if (vp->v_flag & VXWANT) { | |
883 | vp->v_flag &= ~VXWANT; | |
884 | wakeup((caddr_t)vp); | |
885 | } | |
886 | } | |
887 | ||
888 | /* | |
889 | * Eliminate all activity associated with the requested vnode | |
890 | * and with all vnodes aliased to the requested vnode. | |
891 | */ | |
892 | void vgoneall(vp) | |
893 | register struct vnode *vp; | |
894 | { | |
895 | register struct vnode *vq; | |
896 | ||
897 | if (vp->v_flag & VALIASED) { | |
898 | /* | |
899 | * If a vgone (or vclean) is already in progress, | |
900 | * wait until it is done and return. | |
901 | */ | |
902 | if (vp->v_flag & VXLOCK) { | |
903 | vp->v_flag |= VXWANT; | |
904 | sleep((caddr_t)vp, PINOD); | |
905 | return; | |
906 | } | |
907 | /* | |
908 | * Ensure that vp will not be vgone'd while we | |
909 | * are eliminating its aliases. | |
910 | */ | |
911 | vp->v_flag |= VXLOCK; | |
912 | while (vp->v_flag & VALIASED) { | |
913 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
914 | if (vq->v_rdev != vp->v_rdev || | |
915 | vq->v_type != vp->v_type || vp == vq) | |
916 | continue; | |
917 | vgone(vq); | |
918 | break; | |
919 | } | |
920 | } | |
921 | /* | |
922 | * Remove the lock so that vgone below will | |
923 | * really eliminate the vnode after which time | |
924 | * vgone will awaken any sleepers. | |
925 | */ | |
926 | vp->v_flag &= ~VXLOCK; | |
927 | } | |
928 | vgone(vp); | |
929 | } | |
930 | ||
931 | /* | |
932 | * Eliminate all activity associated with a vnode | |
933 | * in preparation for reuse. | |
934 | */ | |
935 | void vgone(vp) | |
936 | register struct vnode *vp; | |
937 | { | |
938 | register struct vnode *vq; | |
939 | struct vnode *vx; | |
940 | long count; | |
941 | ||
942 | /* | |
943 | * If a vgone (or vclean) is already in progress, | |
944 | * wait until it is done and return. | |
945 | */ | |
946 | if (vp->v_flag & VXLOCK) { | |
947 | vp->v_flag |= VXWANT; | |
948 | sleep((caddr_t)vp, PINOD); | |
949 | return; | |
950 | } | |
951 | /* | |
952 | * Clean out the filesystem specific data. | |
953 | */ | |
954 | vclean(vp, DOCLOSE); | |
955 | /* | |
956 | * Delete from old mount point vnode list, if on one. | |
957 | */ | |
958 | if (vp->v_mountb) { | |
959 | if (vq = vp->v_mountf) | |
960 | vq->v_mountb = vp->v_mountb; | |
961 | *vp->v_mountb = vq; | |
962 | vp->v_mountf = NULL; | |
963 | vp->v_mountb = NULL; | |
964 | } | |
965 | /* | |
966 | * If special device, remove it from special device alias list. | |
967 | */ | |
968 | if (vp->v_type == VBLK || vp->v_type == VCHR) { | |
969 | if (*vp->v_hashchain == vp) { | |
970 | *vp->v_hashchain = vp->v_specnext; | |
971 | } else { | |
972 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
973 | if (vq->v_specnext != vp) | |
974 | continue; | |
975 | vq->v_specnext = vp->v_specnext; | |
976 | break; | |
977 | } | |
978 | if (vq == NULL) | |
979 | panic("missing bdev"); | |
980 | } | |
981 | if (vp->v_flag & VALIASED) { | |
982 | count = 0; | |
983 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
984 | if (vq->v_rdev != vp->v_rdev || | |
985 | vq->v_type != vp->v_type) | |
986 | continue; | |
987 | count++; | |
988 | vx = vq; | |
989 | } | |
990 | if (count == 0) | |
991 | panic("missing alias"); | |
992 | if (count == 1) | |
993 | vx->v_flag &= ~VALIASED; | |
994 | vp->v_flag &= ~VALIASED; | |
995 | } | |
996 | FREE(vp->v_specinfo, M_VNODE); | |
997 | vp->v_specinfo = NULL; | |
998 | } | |
999 | /* | |
1000 | * If it is on the freelist, move it to the head of the list. | |
1001 | */ | |
1002 | if (vp->v_freeb) { | |
1003 | if (vq = vp->v_freef) | |
1004 | vq->v_freeb = vp->v_freeb; | |
1005 | else | |
1006 | vfreet = vp->v_freeb; | |
1007 | *vp->v_freeb = vq; | |
1008 | vp->v_freef = vfreeh; | |
1009 | vp->v_freeb = &vfreeh; | |
1010 | vfreeh->v_freeb = &vp->v_freef; | |
1011 | vfreeh = vp; | |
1012 | } | |
1013 | vp->v_type = VBAD; | |
1014 | } | |
1015 | ||
1016 | /* | |
1017 | * Lookup a vnode by device number. | |
1018 | */ | |
1019 | vfinddev(dev, type, vpp) | |
1020 | dev_t dev; | |
1021 | enum vtype type; | |
1022 | struct vnode **vpp; | |
1023 | { | |
1024 | register struct vnode *vp; | |
1025 | ||
1026 | for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { | |
1027 | if (dev != vp->v_rdev || type != vp->v_type) | |
1028 | continue; | |
1029 | *vpp = vp; | |
1030 | return (0); | |
1031 | } | |
1032 | return (1); | |
1033 | } | |
1034 | ||
1035 | /* | |
1036 | * Calculate the total number of references to a special device. | |
1037 | */ | |
1038 | vcount(vp) | |
1039 | register struct vnode *vp; | |
1040 | { | |
1041 | register struct vnode *vq; | |
1042 | int count; | |
1043 | ||
1044 | if ((vp->v_flag & VALIASED) == 0) | |
1045 | return (vp->v_usecount); | |
1046 | loop: | |
1047 | for (count = 0, vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
1048 | if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) | |
1049 | continue; | |
1050 | /* | |
1051 | * Alias, but not in use, so flush it out. | |
1052 | */ | |
1053 | if (vq->v_usecount == 0) { | |
1054 | vgone(vq); | |
1055 | goto loop; | |
1056 | } | |
1057 | count += vq->v_usecount; | |
1058 | } | |
1059 | return (count); | |
1060 | } | |
1061 | ||
1062 | /* | |
1063 | * Print out a description of a vnode. | |
1064 | */ | |
1065 | static char *typename[] = | |
1066 | { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; | |
1067 | ||
1068 | vprint(label, vp) | |
1069 | char *label; | |
1070 | register struct vnode *vp; | |
1071 | { | |
1072 | char buf[64]; | |
1073 | ||
1074 | if (label != NULL) | |
1075 | printf("%s: ", label); | |
1076 | printf("type %s, usecount %d, writecount %d, refcount %d,", | |
1077 | typename[vp->v_type], vp->v_usecount, vp->v_writecount, | |
1078 | vp->v_holdcnt); | |
1079 | buf[0] = '\0'; | |
1080 | if (vp->v_flag & VROOT) | |
1081 | strcat(buf, "|VROOT"); | |
1082 | if (vp->v_flag & VTEXT) | |
1083 | strcat(buf, "|VTEXT"); | |
1084 | if (vp->v_flag & VSYSTEM) | |
1085 | strcat(buf, "|VSYSTEM"); | |
1086 | if (vp->v_flag & VXLOCK) | |
1087 | strcat(buf, "|VXLOCK"); | |
1088 | if (vp->v_flag & VXWANT) | |
1089 | strcat(buf, "|VXWANT"); | |
1090 | if (vp->v_flag & VBWAIT) | |
1091 | strcat(buf, "|VBWAIT"); | |
1092 | if (vp->v_flag & VALIASED) | |
1093 | strcat(buf, "|VALIASED"); | |
1094 | if (buf[0] != '\0') | |
1095 | printf(" flags (%s)", &buf[1]); | |
1096 | printf("\n\t"); | |
1097 | VOP_PRINT(vp); | |
1098 | } | |
1099 | ||
1100 | #ifdef DEBUG | |
1101 | /* | |
1102 | * List all of the locked vnodes in the system. | |
1103 | * Called when debugging the kernel. | |
1104 | */ | |
1105 | printlockedvnodes() | |
1106 | { | |
1107 | register struct mount *mp; | |
1108 | register struct vnode *vp; | |
1109 | ||
1110 | printf("Locked vnodes\n"); | |
1111 | mp = rootfs; | |
1112 | do { | |
1113 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) | |
1114 | if (VOP_ISLOCKED(vp)) | |
1115 | vprint((char *)0, vp); | |
1116 | mp = mp->mnt_next; | |
1117 | } while (mp != rootfs); | |
1118 | } | |
1119 | #endif | |
1120 | ||
1121 | int kinfo_vdebug = 1; | |
1122 | int kinfo_vgetfailed; | |
1123 | #define KINFO_VNODESLOP 10 | |
1124 | /* | |
1125 | * Dump vnode list (via kinfo). | |
1126 | * Copyout address of vnode followed by vnode. | |
1127 | */ | |
1128 | /* ARGSUSED */ | |
1129 | kinfo_vnode(op, where, acopysize, arg, aneeded) | |
1130 | int op; | |
1131 | char *where; | |
1132 | int *acopysize, arg, *aneeded; | |
1133 | { | |
1134 | register struct mount *mp = rootfs; | |
1135 | struct mount *omp; | |
1136 | struct vnode *vp; | |
1137 | register char *bp = where, *savebp; | |
1138 | char *ewhere = where + *acopysize; | |
1139 | int error; | |
1140 | ||
1141 | #define VPTRSZ sizeof (struct vnode *) | |
1142 | #define VNODESZ sizeof (struct vnode) | |
1143 | if (where == NULL) { | |
1144 | *aneeded = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); | |
1145 | return (0); | |
1146 | } | |
1147 | ||
1148 | do { | |
1149 | if (vfs_busy(mp)) { | |
1150 | mp = mp->mnt_next; | |
1151 | continue; | |
1152 | } | |
1153 | savebp = bp; | |
1154 | again: | |
1155 | for (vp = mp->mnt_mounth; vp; vp = vp->v_mountf) { | |
1156 | /* | |
1157 | * Check that the vp is still associated with | |
1158 | * this filesystem. RACE: could have been | |
1159 | * recycled onto the same filesystem. | |
1160 | */ | |
1161 | if (vp->v_mount != mp) { | |
1162 | if (kinfo_vdebug) | |
1163 | printf("kinfo: vp changed\n"); | |
1164 | bp = savebp; | |
1165 | goto again; | |
1166 | } | |
1167 | if ((bp + VPTRSZ + VNODESZ <= ewhere) && | |
1168 | ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || | |
1169 | (error = copyout((caddr_t)vp, bp + VPTRSZ, | |
1170 | VNODESZ)))) | |
1171 | return (error); | |
1172 | bp += VPTRSZ + VNODESZ; | |
1173 | } | |
1174 | omp = mp; | |
1175 | mp = mp->mnt_next; | |
1176 | vfs_unbusy(omp); | |
1177 | } while (mp != rootfs); | |
1178 | ||
1179 | *aneeded = bp - where; | |
1180 | if (bp > ewhere) | |
1181 | *acopysize = ewhere - where; | |
1182 | else | |
1183 | *acopysize = bp - where; | |
1184 | return (0); | |
1185 | } |