Commit | Line | Data |
---|---|---|
3c4390e8 | 1 | /* |
ec54f0cc KB |
2 | * Copyright (c) 1989, 1993 |
3 | * The Regents of the University of California. All rights reserved. | |
3c4390e8 | 4 | * |
dbf0c423 | 5 | * %sccs.include.redist.c% |
3c4390e8 | 6 | * |
8981e258 | 7 | * @(#)vfs_subr.c 8.7 (Berkeley) %G% |
3c4390e8 KM |
8 | */ |
9 | ||
10 | /* | |
11 | * External virtual filesystem routines | |
12 | */ | |
13 | ||
cb796a23 | 14 | #include <sys/param.h> |
917dc539 | 15 | #include <sys/systm.h> |
cb796a23 KB |
16 | #include <sys/proc.h> |
17 | #include <sys/mount.h> | |
18 | #include <sys/time.h> | |
19 | #include <sys/vnode.h> | |
807cc430 | 20 | #include <sys/stat.h> |
cb796a23 KB |
21 | #include <sys/namei.h> |
22 | #include <sys/ucred.h> | |
23 | #include <sys/buf.h> | |
24 | #include <sys/errno.h> | |
25 | #include <sys/malloc.h> | |
8981e258 MH |
26 | #include <sys/domain.h> |
27 | #include <sys/mbuf.h> | |
3c4390e8 | 28 | |
bb4964fd KM |
29 | #include <vm/vm.h> |
30 | #include <sys/sysctl.h> | |
31 | ||
021de758 JSP |
32 | #include <miscfs/specfs/specdev.h> |
33 | ||
807cc430 KM |
34 | enum vtype iftovt_tab[16] = { |
35 | VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON, | |
36 | VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VBAD, | |
37 | }; | |
38 | int vttoif_tab[9] = { | |
39 | 0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, | |
40 | S_IFSOCK, S_IFIFO, S_IFMT, | |
41 | }; | |
42 | ||
e3249ec0 KM |
43 | /* |
44 | * Insq/Remq for the vnode usage lists. | |
45 | */ | |
3fc2ac18 KM |
46 | #define bufinsvn(bp, dp) LIST_INSERT_HEAD(dp, bp, b_vnbufs) |
47 | #define bufremvn(bp) { \ | |
48 | LIST_REMOVE(bp, b_vnbufs); \ | |
49 | (bp)->b_vnbufs.le_next = NOLIST; \ | |
50 | } | |
51 | ||
52 | TAILQ_HEAD(freelst, vnode) vnode_free_list; /* vnode free list */ | |
53 | struct mntlist mountlist; /* mounted filesystem list */ | |
e3249ec0 | 54 | |
3c4390e8 | 55 | /* |
3fc2ac18 | 56 | * Initialize the vnode management data structures. |
3c4390e8 | 57 | */ |
3fc2ac18 | 58 | vntblinit() |
3c4390e8 KM |
59 | { |
60 | ||
3fc2ac18 KM |
61 | TAILQ_INIT(&vnode_free_list); |
62 | TAILQ_INIT(&mountlist); | |
3c4390e8 KM |
63 | } |
64 | ||
65 | /* | |
66 | * Lock a filesystem. | |
67 | * Used to prevent access to it while mounting and unmounting. | |
68 | */ | |
69 | vfs_lock(mp) | |
70 | register struct mount *mp; | |
71 | { | |
72 | ||
54fb9dc2 KM |
73 | while(mp->mnt_flag & MNT_MLOCK) { |
74 | mp->mnt_flag |= MNT_MWAIT; | |
594501df KM |
75 | sleep((caddr_t)mp, PVFS); |
76 | } | |
54fb9dc2 | 77 | mp->mnt_flag |= MNT_MLOCK; |
3c4390e8 KM |
78 | return (0); |
79 | } | |
80 | ||
81 | /* | |
82 | * Unlock a locked filesystem. | |
83 | * Panic if filesystem is not locked. | |
84 | */ | |
85 | void | |
86 | vfs_unlock(mp) | |
87 | register struct mount *mp; | |
88 | { | |
89 | ||
54fb9dc2 | 90 | if ((mp->mnt_flag & MNT_MLOCK) == 0) |
36ef03ec | 91 | panic("vfs_unlock: not locked"); |
54fb9dc2 KM |
92 | mp->mnt_flag &= ~MNT_MLOCK; |
93 | if (mp->mnt_flag & MNT_MWAIT) { | |
94 | mp->mnt_flag &= ~MNT_MWAIT; | |
3c4390e8 KM |
95 | wakeup((caddr_t)mp); |
96 | } | |
97 | } | |
98 | ||
36ef03ec KM |
99 | /* |
100 | * Mark a mount point as busy. | |
101 | * Used to synchronize access and to delay unmounting. | |
102 | */ | |
103 | vfs_busy(mp) | |
104 | register struct mount *mp; | |
105 | { | |
106 | ||
54fb9dc2 KM |
107 | while(mp->mnt_flag & MNT_MPBUSY) { |
108 | mp->mnt_flag |= MNT_MPWANT; | |
109 | sleep((caddr_t)&mp->mnt_flag, PVFS); | |
36ef03ec | 110 | } |
d8b63609 KM |
111 | if (mp->mnt_flag & MNT_UNMOUNT) |
112 | return (1); | |
54fb9dc2 | 113 | mp->mnt_flag |= MNT_MPBUSY; |
36ef03ec KM |
114 | return (0); |
115 | } | |
116 | ||
117 | /* | |
118 | * Free a busy filesystem. | |
119 | * Panic if filesystem is not busy. | |
120 | */ | |
36ef03ec KM |
121 | vfs_unbusy(mp) |
122 | register struct mount *mp; | |
123 | { | |
124 | ||
54fb9dc2 | 125 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) |
36ef03ec | 126 | panic("vfs_unbusy: not busy"); |
54fb9dc2 KM |
127 | mp->mnt_flag &= ~MNT_MPBUSY; |
128 | if (mp->mnt_flag & MNT_MPWANT) { | |
129 | mp->mnt_flag &= ~MNT_MPWANT; | |
130 | wakeup((caddr_t)&mp->mnt_flag); | |
36ef03ec KM |
131 | } |
132 | } | |
133 | ||
3c4390e8 KM |
134 | /* |
135 | * Lookup a mount point by filesystem identifier. | |
136 | */ | |
137 | struct mount * | |
138 | getvfs(fsid) | |
139 | fsid_t *fsid; | |
140 | { | |
141 | register struct mount *mp; | |
142 | ||
3fc2ac18 | 143 | for (mp = mountlist.tqh_first; mp != NULL; mp = mp->mnt_list.tqe_next) { |
54fb9dc2 | 144 | if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] && |
3fc2ac18 | 145 | mp->mnt_stat.f_fsid.val[1] == fsid->val[1]) |
d713f801 | 146 | return (mp); |
3fc2ac18 | 147 | } |
d713f801 | 148 | return ((struct mount *)0); |
3c4390e8 KM |
149 | } |
150 | ||
917dc539 JSP |
151 | /* |
152 | * Get a new unique fsid | |
153 | */ | |
154 | void | |
155 | getnewfsid(mp, mtype) | |
156 | struct mount *mp; | |
157 | int mtype; | |
158 | { | |
159 | static u_short xxxfs_mntid; | |
160 | ||
161 | fsid_t tfsid; | |
162 | ||
1209b9a4 | 163 | mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0); |
917dc539 JSP |
164 | mp->mnt_stat.f_fsid.val[1] = mtype; |
165 | if (xxxfs_mntid == 0) | |
166 | ++xxxfs_mntid; | |
1209b9a4 | 167 | tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid); |
917dc539 | 168 | tfsid.val[1] = mtype; |
3fc2ac18 | 169 | if (mountlist.tqh_first != NULL) { |
17fd1cc7 JSP |
170 | while (getvfs(&tfsid)) { |
171 | tfsid.val[0]++; | |
172 | xxxfs_mntid++; | |
173 | } | |
917dc539 JSP |
174 | } |
175 | mp->mnt_stat.f_fsid.val[0] = tfsid.val[0]; | |
176 | } | |
177 | ||
3c4390e8 KM |
178 | /* |
179 | * Set vnode attributes to VNOVAL | |
180 | */ | |
181 | void vattr_null(vap) | |
182 | register struct vattr *vap; | |
183 | { | |
184 | ||
185 | vap->va_type = VNON; | |
83504fd5 | 186 | vap->va_size = vap->va_bytes = VNOVAL; |
3c4390e8 | 187 | vap->va_mode = vap->va_nlink = vap->va_uid = vap->va_gid = |
83504fd5 KM |
188 | vap->va_fsid = vap->va_fileid = |
189 | vap->va_blocksize = vap->va_rdev = | |
ecf75a7d KM |
190 | vap->va_atime.ts_sec = vap->va_atime.ts_nsec = |
191 | vap->va_mtime.ts_sec = vap->va_mtime.ts_nsec = | |
192 | vap->va_ctime.ts_sec = vap->va_ctime.ts_nsec = | |
8cf4d4fb | 193 | vap->va_flags = vap->va_gen = VNOVAL; |
fcba749b | 194 | vap->va_vaflags = 0; |
3c4390e8 | 195 | } |
c60798ca | 196 | |
36d09cb1 KM |
197 | /* |
198 | * Routines having to do with the management of the vnode table. | |
199 | */ | |
9342689a | 200 | extern int (**dead_vnodeop_p)(); |
32339c94 | 201 | extern void vclean(); |
1a80f56e | 202 | long numvnodes; |
e781da98 | 203 | extern struct vattr va_null; |
3e787e54 KM |
204 | int newnodes = 0; |
205 | int printcnt = 0; | |
36d09cb1 KM |
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; | |
cf74dd57 | 213 | int (**vops)(); |
36d09cb1 KM |
214 | struct vnode **vpp; |
215 | { | |
c768e50f | 216 | register struct vnode *vp; |
1f9d2249 | 217 | int s; |
36d09cb1 | 218 | |
3e787e54 | 219 | newnodes++; |
3fc2ac18 KM |
220 | if ((vnode_free_list.tqh_first == NULL && |
221 | numvnodes < 2 * desiredvnodes) || | |
ecf75a7d | 222 | numvnodes < desiredvnodes) { |
aacc1bff KM |
223 | vp = (struct vnode *)malloc((u_long)sizeof *vp, |
224 | M_VNODE, M_WAITOK); | |
1a80f56e | 225 | bzero((char *)vp, sizeof *vp); |
3e787e54 KM |
226 | vp->v_freelist.tqe_next = (struct vnode *)0xdeadf; |
227 | vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb; | |
228 | vp->v_mntvnodes.le_next = (struct vnode *)0xdeadf; | |
229 | vp->v_mntvnodes.le_prev = (struct vnode **)0xdeadb; | |
1a80f56e | 230 | numvnodes++; |
3e787e54 | 231 | vp->v_spare[0] = numvnodes; |
1a80f56e | 232 | } else { |
3fc2ac18 | 233 | if ((vp = vnode_free_list.tqh_first) == NULL) { |
1a80f56e KM |
234 | tablefull("vnode"); |
235 | *vpp = 0; | |
236 | return (ENFILE); | |
237 | } | |
238 | if (vp->v_usecount) | |
239 | panic("free vnode isn't"); | |
3e787e54 KM |
240 | if (vp->v_freelist.tqe_next == (struct vnode *)0xdeadf || |
241 | vp->v_freelist.tqe_prev == (struct vnode **)0xdeadb) | |
242 | panic("getnewvnode: not on queue"); | |
3fc2ac18 | 243 | TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); |
3e787e54 | 244 | vp->v_freelist.tqe_next = (struct vnode *)0xdeadf; |
0bf9bb76 KM |
245 | /* see comment on why 0xdeadb is set at end of vgone (below) */ |
246 | vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb; | |
39b99eb6 | 247 | vp->v_lease = NULL; |
1a80f56e KM |
248 | if (vp->v_type != VBAD) |
249 | vgone(vp); | |
1f9d2249 | 250 | #ifdef DIAGNOSTIC |
2345b093 KM |
251 | if (vp->v_data) |
252 | panic("cleaned vnode isn't"); | |
1f9d2249 MS |
253 | s = splbio(); |
254 | if (vp->v_numoutput) | |
255 | panic("Clean vnode has pending I/O's"); | |
256 | splx(s); | |
257 | #endif | |
1a80f56e | 258 | vp->v_flag = 0; |
1a80f56e | 259 | vp->v_lastr = 0; |
1f9d2249 MS |
260 | vp->v_lastw = 0; |
261 | vp->v_lasta = 0; | |
262 | vp->v_cstart = 0; | |
263 | vp->v_clen = 0; | |
1a80f56e | 264 | vp->v_socket = 0; |
36d09cb1 | 265 | } |
1f9d2249 | 266 | vp->v_ralen = 1; |
b027498b | 267 | vp->v_type = VNON; |
36d09cb1 KM |
268 | cache_purge(vp); |
269 | vp->v_tag = tag; | |
ef24f6dd | 270 | vp->v_op = vops; |
36d09cb1 | 271 | insmntque(vp, mp); |
36d09cb1 | 272 | *vpp = vp; |
0bf9bb76 | 273 | vp->v_usecount = 1; |
3fc2ac18 | 274 | vp->v_data = 0; |
3e787e54 | 275 | if (printcnt-- > 0) vprint("getnewvnode got", vp); |
36d09cb1 KM |
276 | return (0); |
277 | } | |
8981e258 | 278 | |
36d09cb1 KM |
279 | /* |
280 | * Move a vnode from one mount queue to another. | |
281 | */ | |
282 | insmntque(vp, mp) | |
283 | register struct vnode *vp; | |
284 | register struct mount *mp; | |
285 | { | |
36d09cb1 KM |
286 | |
287 | /* | |
288 | * Delete from old mount point vnode list, if on one. | |
289 | */ | |
3e787e54 KM |
290 | if (vp->v_mount != NULL) { |
291 | if (vp->v_mntvnodes.le_next == (struct vnode *)0xdeadf || | |
292 | vp->v_mntvnodes.le_prev == (struct vnode **)0xdeadb) | |
293 | panic("insmntque: not on queue"); | |
3fc2ac18 | 294 | LIST_REMOVE(vp, v_mntvnodes); |
3e787e54 KM |
295 | vp->v_mntvnodes.le_next = (struct vnode *)0xdeadf; |
296 | vp->v_mntvnodes.le_prev = (struct vnode **)0xdeadb; | |
297 | } | |
36d09cb1 KM |
298 | /* |
299 | * Insert into list of vnodes for the new mount point, if available. | |
300 | */ | |
3fc2ac18 | 301 | if ((vp->v_mount = mp) == NULL) |
36d09cb1 | 302 | return; |
3e787e54 KM |
303 | if (vp->v_mntvnodes.le_next != (struct vnode *)0xdeadf || |
304 | vp->v_mntvnodes.le_prev != (struct vnode **)0xdeadb) | |
305 | panic("insmntque: already on queue"); | |
3fc2ac18 | 306 | LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes); |
36d09cb1 KM |
307 | } |
308 | ||
76429560 KM |
309 | /* |
310 | * Update outstanding I/O count and do wakeup if requested. | |
311 | */ | |
312 | vwakeup(bp) | |
313 | register struct buf *bp; | |
314 | { | |
315 | register struct vnode *vp; | |
316 | ||
a9338fad | 317 | bp->b_flags &= ~B_WRITEINPROG; |
76429560 KM |
318 | if (vp = bp->b_vp) { |
319 | vp->v_numoutput--; | |
1f9d2249 MS |
320 | if (vp->v_numoutput < 0) |
321 | panic("vwakeup: neg numoutput"); | |
76429560 KM |
322 | if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) { |
323 | if (vp->v_numoutput < 0) | |
324 | panic("vwakeup: neg numoutput"); | |
325 | vp->v_flag &= ~VBWAIT; | |
326 | wakeup((caddr_t)&vp->v_numoutput); | |
327 | } | |
328 | } | |
329 | } | |
330 | ||
76429560 KM |
331 | /* |
332 | * Flush out and invalidate all buffers associated with a vnode. | |
333 | * Called with the underlying object locked. | |
334 | */ | |
d024c2ce | 335 | int |
c33e9e8b | 336 | vinvalbuf(vp, flags, cred, p, slpflag, slptimeo) |
76429560 | 337 | register struct vnode *vp; |
12079a9d | 338 | int flags; |
d024c2ce KM |
339 | struct ucred *cred; |
340 | struct proc *p; | |
c33e9e8b | 341 | int slpflag, slptimeo; |
76429560 KM |
342 | { |
343 | register struct buf *bp; | |
344 | struct buf *nbp, *blist; | |
d024c2ce | 345 | int s, error; |
76429560 | 346 | |
12079a9d | 347 | if (flags & V_SAVE) { |
d024c2ce KM |
348 | if (error = VOP_FSYNC(vp, cred, MNT_WAIT, p)) |
349 | return (error); | |
3fc2ac18 | 350 | if (vp->v_dirtyblkhd.lh_first != NULL) |
d024c2ce KM |
351 | panic("vinvalbuf: dirty bufs"); |
352 | } | |
76429560 | 353 | for (;;) { |
3fc2ac18 | 354 | if ((blist = vp->v_cleanblkhd.lh_first) && flags & V_SAVEMETA) |
12079a9d | 355 | while (blist && blist->b_lblkno < 0) |
3fc2ac18 KM |
356 | blist = blist->b_vnbufs.le_next; |
357 | if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && | |
e3249ec0 | 358 | (flags & V_SAVEMETA)) |
12079a9d | 359 | while (blist && blist->b_lblkno < 0) |
3fc2ac18 | 360 | blist = blist->b_vnbufs.le_next; |
12079a9d | 361 | if (!blist) |
76429560 | 362 | break; |
12079a9d | 363 | |
76429560 | 364 | for (bp = blist; bp; bp = nbp) { |
3fc2ac18 | 365 | nbp = bp->b_vnbufs.le_next; |
12079a9d MS |
366 | if (flags & V_SAVEMETA && bp->b_lblkno < 0) |
367 | continue; | |
76429560 KM |
368 | s = splbio(); |
369 | if (bp->b_flags & B_BUSY) { | |
370 | bp->b_flags |= B_WANTED; | |
c33e9e8b KM |
371 | error = tsleep((caddr_t)bp, |
372 | slpflag | (PRIBIO + 1), "vinvalbuf", | |
373 | slptimeo); | |
76429560 | 374 | splx(s); |
c33e9e8b KM |
375 | if (error) |
376 | return (error); | |
76429560 KM |
377 | break; |
378 | } | |
379 | bremfree(bp); | |
380 | bp->b_flags |= B_BUSY; | |
381 | splx(s); | |
c33e9e8b KM |
382 | /* |
383 | * XXX Since there are no node locks for NFS, I believe | |
384 | * there is a slight chance that a delayed write will | |
385 | * occur while sleeping just above, so check for it. | |
386 | */ | |
387 | if ((bp->b_flags & B_DELWRI) && (flags & V_SAVE)) { | |
388 | (void) VOP_BWRITE(bp); | |
389 | break; | |
390 | } | |
12079a9d | 391 | bp->b_flags |= B_INVAL; |
76429560 KM |
392 | brelse(bp); |
393 | } | |
394 | } | |
e3249ec0 | 395 | if (!(flags & V_SAVEMETA) && |
3fc2ac18 | 396 | (vp->v_dirtyblkhd.lh_first || vp->v_cleanblkhd.lh_first)) |
76429560 | 397 | panic("vinvalbuf: flush failed"); |
d024c2ce | 398 | return (0); |
76429560 KM |
399 | } |
400 | ||
401 | /* | |
402 | * Associate a buffer with a vnode. | |
403 | */ | |
404 | bgetvp(vp, bp) | |
405 | register struct vnode *vp; | |
406 | register struct buf *bp; | |
407 | { | |
408 | ||
409 | if (bp->b_vp) | |
410 | panic("bgetvp: not free"); | |
411 | VHOLD(vp); | |
412 | bp->b_vp = vp; | |
413 | if (vp->v_type == VBLK || vp->v_type == VCHR) | |
414 | bp->b_dev = vp->v_rdev; | |
415 | else | |
416 | bp->b_dev = NODEV; | |
417 | /* | |
418 | * Insert onto list for new vnode. | |
419 | */ | |
e3249ec0 | 420 | bufinsvn(bp, &vp->v_cleanblkhd); |
76429560 KM |
421 | } |
422 | ||
423 | /* | |
424 | * Disassociate a buffer from a vnode. | |
425 | */ | |
426 | brelvp(bp) | |
427 | register struct buf *bp; | |
428 | { | |
76429560 KM |
429 | struct vnode *vp; |
430 | ||
431 | if (bp->b_vp == (struct vnode *) 0) | |
432 | panic("brelvp: NULL"); | |
433 | /* | |
434 | * Delete from old vnode list, if on one. | |
435 | */ | |
3fc2ac18 | 436 | if (bp->b_vnbufs.le_next != NOLIST) |
e3249ec0 | 437 | bufremvn(bp); |
76429560 KM |
438 | vp = bp->b_vp; |
439 | bp->b_vp = (struct vnode *) 0; | |
440 | HOLDRELE(vp); | |
441 | } | |
442 | ||
443 | /* | |
444 | * Reassign a buffer from one vnode to another. | |
445 | * Used to assign file specific control information | |
446 | * (indirect blocks) to the vnode to which they belong. | |
447 | */ | |
448 | reassignbuf(bp, newvp) | |
449 | register struct buf *bp; | |
450 | register struct vnode *newvp; | |
451 | { | |
3fc2ac18 | 452 | register struct buflists *listheadp; |
76429560 | 453 | |
e5c3f16e KM |
454 | if (newvp == NULL) { |
455 | printf("reassignbuf: NULL"); | |
456 | return; | |
457 | } | |
76429560 KM |
458 | /* |
459 | * Delete from old vnode list, if on one. | |
460 | */ | |
3fc2ac18 | 461 | if (bp->b_vnbufs.le_next != NOLIST) |
e3249ec0 | 462 | bufremvn(bp); |
76429560 KM |
463 | /* |
464 | * If dirty, put on list of dirty buffers; | |
465 | * otherwise insert onto list of clean buffers. | |
466 | */ | |
467 | if (bp->b_flags & B_DELWRI) | |
468 | listheadp = &newvp->v_dirtyblkhd; | |
469 | else | |
470 | listheadp = &newvp->v_cleanblkhd; | |
e3249ec0 | 471 | bufinsvn(bp, listheadp); |
76429560 KM |
472 | } |
473 | ||
36d09cb1 | 474 | /* |
ef24f6dd KM |
475 | * Create a vnode for a block device. |
476 | * Used for root filesystem, argdev, and swap areas. | |
477 | * Also used for memory file system special devices. | |
478 | */ | |
479 | bdevvp(dev, vpp) | |
480 | dev_t dev; | |
481 | struct vnode **vpp; | |
482 | { | |
ef24f6dd KM |
483 | register struct vnode *vp; |
484 | struct vnode *nvp; | |
485 | int error; | |
486 | ||
1c89915d KM |
487 | if (dev == NODEV) |
488 | return (0); | |
9342689a | 489 | error = getnewvnode(VT_NON, (struct mount *)0, spec_vnodeop_p, &nvp); |
ef24f6dd KM |
490 | if (error) { |
491 | *vpp = 0; | |
492 | return (error); | |
493 | } | |
494 | vp = nvp; | |
495 | vp->v_type = VBLK; | |
c0de8792 | 496 | if (nvp = checkalias(vp, dev, (struct mount *)0)) { |
ef24f6dd KM |
497 | vput(vp); |
498 | vp = nvp; | |
499 | } | |
500 | *vpp = vp; | |
501 | return (0); | |
502 | } | |
503 | ||
504 | /* | |
505 | * Check to see if the new vnode represents a special device | |
506 | * for which we already have a vnode (either because of | |
507 | * bdevvp() or because of a different vnode representing | |
508 | * the same block device). If such an alias exists, deallocate | |
f0556f86 | 509 | * the existing contents and return the aliased vnode. The |
ef24f6dd KM |
510 | * caller is responsible for filling it with its new contents. |
511 | */ | |
512 | struct vnode * | |
c0de8792 | 513 | checkalias(nvp, nvp_rdev, mp) |
ef24f6dd | 514 | register struct vnode *nvp; |
c0de8792 | 515 | dev_t nvp_rdev; |
ef24f6dd KM |
516 | struct mount *mp; |
517 | { | |
518 | register struct vnode *vp; | |
c0de8792 | 519 | struct vnode **vpp; |
ef24f6dd KM |
520 | |
521 | if (nvp->v_type != VBLK && nvp->v_type != VCHR) | |
54fb9dc2 | 522 | return (NULLVP); |
c0de8792 KM |
523 | |
524 | vpp = &speclisth[SPECHASH(nvp_rdev)]; | |
ef24f6dd | 525 | loop: |
c0de8792 KM |
526 | for (vp = *vpp; vp; vp = vp->v_specnext) { |
527 | if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type) | |
ef24f6dd | 528 | continue; |
c0de8792 KM |
529 | /* |
530 | * Alias, but not in use, so flush it out. | |
531 | */ | |
7f7b7d89 | 532 | if (vp->v_usecount == 0) { |
c0de8792 KM |
533 | vgone(vp); |
534 | goto loop; | |
535 | } | |
3fc2ac18 | 536 | if (vget(vp, 1)) |
ef62830d | 537 | goto loop; |
ef24f6dd KM |
538 | break; |
539 | } | |
c0de8792 | 540 | if (vp == NULL || vp->v_tag != VT_NON) { |
c0de8792 KM |
541 | MALLOC(nvp->v_specinfo, struct specinfo *, |
542 | sizeof(struct specinfo), M_VNODE, M_WAITOK); | |
543 | nvp->v_rdev = nvp_rdev; | |
7f7b7d89 | 544 | nvp->v_hashchain = vpp; |
c0de8792 | 545 | nvp->v_specnext = *vpp; |
2c957a90 | 546 | nvp->v_specflags = 0; |
c0de8792 | 547 | *vpp = nvp; |
40452d5e KM |
548 | if (vp != NULL) { |
549 | nvp->v_flag |= VALIASED; | |
550 | vp->v_flag |= VALIASED; | |
551 | vput(vp); | |
552 | } | |
54fb9dc2 | 553 | return (NULLVP); |
ef24f6dd | 554 | } |
2bae1875 KM |
555 | VOP_UNLOCK(vp); |
556 | vclean(vp, 0); | |
ef24f6dd KM |
557 | vp->v_op = nvp->v_op; |
558 | vp->v_tag = nvp->v_tag; | |
559 | nvp->v_type = VNON; | |
560 | insmntque(vp, mp); | |
561 | return (vp); | |
562 | } | |
563 | ||
564 | /* | |
565 | * Grab a particular vnode from the free list, increment its | |
566 | * reference count and lock it. The vnode lock bit is set the | |
567 | * vnode is being eliminated in vgone. The process is awakened | |
568 | * when the transition is completed, and an error returned to | |
569 | * indicate that the vnode is no longer usable (possibly having | |
570 | * been changed to a new file system type). | |
36d09cb1 | 571 | */ |
3fc2ac18 | 572 | vget(vp, lockflag) |
36d09cb1 | 573 | register struct vnode *vp; |
3fc2ac18 | 574 | int lockflag; |
36d09cb1 | 575 | { |
36d09cb1 | 576 | |
ef24f6dd KM |
577 | if (vp->v_flag & VXLOCK) { |
578 | vp->v_flag |= VXWANT; | |
579 | sleep((caddr_t)vp, PINOD); | |
580 | return (1); | |
581 | } | |
3e787e54 KM |
582 | if (vp->v_usecount == 0) { |
583 | if (vp->v_freelist.tqe_next == (struct vnode *)0xdeadf || | |
584 | vp->v_freelist.tqe_prev == (struct vnode **)0xdeadb) | |
585 | panic("vget: not on queue"); | |
3fc2ac18 | 586 | TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); |
3e787e54 KM |
587 | vp->v_freelist.tqe_next = (struct vnode *)0xdeadf; |
588 | vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb; | |
589 | } | |
ec04fc59 | 590 | vp->v_usecount++; |
3fc2ac18 KM |
591 | if (lockflag) |
592 | VOP_LOCK(vp); | |
3e787e54 | 593 | if (printcnt-- > 0) vprint("vget got", vp); |
ef24f6dd | 594 | return (0); |
36d09cb1 KM |
595 | } |
596 | ||
d32390ea KM |
597 | int bug_refs = 0; |
598 | ||
36d09cb1 KM |
599 | /* |
600 | * Vnode reference, just increment the count | |
601 | */ | |
602 | void vref(vp) | |
603 | struct vnode *vp; | |
604 | { | |
605 | ||
ec04fc59 KM |
606 | if (vp->v_usecount <= 0) |
607 | panic("vref used where vget required"); | |
3e787e54 KM |
608 | if (vp->v_freelist.tqe_next != (struct vnode *)0xdeadf || |
609 | vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb) | |
610 | panic("vref: not free"); | |
7f7b7d89 | 611 | vp->v_usecount++; |
3e787e54 | 612 | if (printcnt-- > 0) vprint("vref get", vp); |
d32390ea KM |
613 | if (vp->v_type != VBLK && curproc) |
614 | curproc->p_spare[0]++; | |
615 | if (bug_refs) | |
616 | vprint("vref: "); | |
36d09cb1 KM |
617 | } |
618 | ||
619 | /* | |
620 | * vput(), just unlock and vrele() | |
621 | */ | |
622 | void vput(vp) | |
623 | register struct vnode *vp; | |
624 | { | |
4d1ee2eb | 625 | |
36d09cb1 KM |
626 | VOP_UNLOCK(vp); |
627 | vrele(vp); | |
628 | } | |
629 | ||
630 | /* | |
631 | * Vnode release. | |
632 | * If count drops to zero, call inactive routine and return to freelist. | |
633 | */ | |
634 | void vrele(vp) | |
635 | register struct vnode *vp; | |
636 | { | |
637 | ||
65c3b3a8 | 638 | #ifdef DIAGNOSTIC |
36d09cb1 | 639 | if (vp == NULL) |
ef24f6dd | 640 | panic("vrele: null vp"); |
65c3b3a8 | 641 | #endif |
7f7b7d89 | 642 | vp->v_usecount--; |
3e787e54 | 643 | if (printcnt-- > 0) vprint("vrele put", vp); |
d32390ea KM |
644 | if (vp->v_type != VBLK && curproc) |
645 | curproc->p_spare[0]--; | |
646 | if (bug_refs) | |
647 | vprint("vref: "); | |
7f7b7d89 | 648 | if (vp->v_usecount > 0) |
36d09cb1 | 649 | return; |
65c3b3a8 KM |
650 | #ifdef DIAGNOSTIC |
651 | if (vp->v_usecount != 0 || vp->v_writecount != 0) { | |
652 | vprint("vrele: bad ref count", vp); | |
653 | panic("vrele: ref cnt"); | |
654 | } | |
655 | #endif | |
dc998e72 KM |
656 | /* |
657 | * insert at tail of LRU list | |
658 | */ | |
3e787e54 KM |
659 | if (vp->v_freelist.tqe_next != (struct vnode *)0xdeadf || |
660 | vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb) | |
661 | panic("vrele: not free"); | |
3fc2ac18 | 662 | TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist); |
d024c2ce | 663 | VOP_INACTIVE(vp); |
ef24f6dd KM |
664 | } |
665 | ||
7f7b7d89 KM |
666 | /* |
667 | * Page or buffer structure gets a reference. | |
668 | */ | |
451df175 | 669 | void vhold(vp) |
7f7b7d89 KM |
670 | register struct vnode *vp; |
671 | { | |
672 | ||
673 | vp->v_holdcnt++; | |
674 | } | |
675 | ||
676 | /* | |
677 | * Page or buffer structure frees a reference. | |
678 | */ | |
451df175 | 679 | void holdrele(vp) |
7f7b7d89 KM |
680 | register struct vnode *vp; |
681 | { | |
682 | ||
683 | if (vp->v_holdcnt <= 0) | |
684 | panic("holdrele: holdcnt"); | |
685 | vp->v_holdcnt--; | |
686 | } | |
687 | ||
f0556f86 KM |
688 | /* |
689 | * Remove any vnodes in the vnode table belonging to mount point mp. | |
690 | * | |
691 | * If MNT_NOFORCE is specified, there should not be any active ones, | |
692 | * return error if any are found (nb: this is a user error, not a | |
693 | * system error). If MNT_FORCE is specified, detach any active vnodes | |
694 | * that are found. | |
695 | */ | |
8981e258 | 696 | #ifdef DIAGNOSTIC |
bb4964fd KM |
697 | int busyprt = 0; /* print out busy vnodes */ |
698 | struct ctldebug debug1 = { "busyprt", &busyprt }; | |
8981e258 | 699 | #endif |
f0556f86 KM |
700 | |
701 | vflush(mp, skipvp, flags) | |
702 | struct mount *mp; | |
703 | struct vnode *skipvp; | |
704 | int flags; | |
705 | { | |
706 | register struct vnode *vp, *nvp; | |
707 | int busy = 0; | |
708 | ||
54fb9dc2 | 709 | if ((mp->mnt_flag & MNT_MPBUSY) == 0) |
36ef03ec | 710 | panic("vflush: not busy"); |
4597dd33 | 711 | loop: |
3fc2ac18 | 712 | for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { |
4597dd33 KM |
713 | if (vp->v_mount != mp) |
714 | goto loop; | |
3fc2ac18 | 715 | nvp = vp->v_mntvnodes.le_next; |
f0556f86 KM |
716 | /* |
717 | * Skip over a selected vnode. | |
f0556f86 KM |
718 | */ |
719 | if (vp == skipvp) | |
720 | continue; | |
36ef03ec KM |
721 | /* |
722 | * Skip over a vnodes marked VSYSTEM. | |
723 | */ | |
724 | if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) | |
725 | continue; | |
da374605 KM |
726 | /* |
727 | * If WRITECLOSE is set, only flush out regular file | |
728 | * vnodes open for writing. | |
729 | */ | |
730 | if ((flags & WRITECLOSE) && | |
731 | (vp->v_writecount == 0 || vp->v_type != VREG)) | |
732 | continue; | |
f0556f86 | 733 | /* |
7f7b7d89 | 734 | * With v_usecount == 0, all we need to do is clear |
f0556f86 KM |
735 | * out the vnode data structures and we are done. |
736 | */ | |
7f7b7d89 | 737 | if (vp->v_usecount == 0) { |
f0556f86 KM |
738 | vgone(vp); |
739 | continue; | |
740 | } | |
741 | /* | |
da374605 | 742 | * If FORCECLOSE is set, forcibly close the vnode. |
f0556f86 KM |
743 | * For block or character devices, revert to an |
744 | * anonymous device. For all other files, just kill them. | |
745 | */ | |
36ef03ec | 746 | if (flags & FORCECLOSE) { |
f0556f86 KM |
747 | if (vp->v_type != VBLK && vp->v_type != VCHR) { |
748 | vgone(vp); | |
749 | } else { | |
750 | vclean(vp, 0); | |
9342689a | 751 | vp->v_op = spec_vnodeop_p; |
f0556f86 KM |
752 | insmntque(vp, (struct mount *)0); |
753 | } | |
754 | continue; | |
755 | } | |
8981e258 | 756 | #ifdef DIAGNOSTIC |
f0556f86 | 757 | if (busyprt) |
0bf84b18 | 758 | vprint("vflush: busy vnode", vp); |
8981e258 | 759 | #endif |
f0556f86 KM |
760 | busy++; |
761 | } | |
762 | if (busy) | |
763 | return (EBUSY); | |
764 | return (0); | |
765 | } | |
766 | ||
ef24f6dd KM |
767 | /* |
768 | * Disassociate the underlying file system from a vnode. | |
ef24f6dd | 769 | */ |
ecf75a7d KM |
770 | void |
771 | vclean(vp, flags) | |
ef24f6dd | 772 | register struct vnode *vp; |
aacc1bff | 773 | int flags; |
ef24f6dd | 774 | { |
2bae1875 | 775 | int active; |
ef24f6dd | 776 | |
2bae1875 KM |
777 | /* |
778 | * Check to see if the vnode is in use. | |
0bf84b18 KM |
779 | * If so we have to reference it before we clean it out |
780 | * so that its count cannot fall to zero and generate a | |
781 | * race against ourselves to recycle it. | |
2bae1875 | 782 | */ |
7f7b7d89 | 783 | if (active = vp->v_usecount) |
2bae1875 | 784 | VREF(vp); |
669df1aa KM |
785 | /* |
786 | * Even if the count is zero, the VOP_INACTIVE routine may still | |
787 | * have the object locked while it cleans it out. The VOP_LOCK | |
788 | * ensures that the VOP_INACTIVE routine is done with its work. | |
789 | * For active vnodes, it ensures that no other activity can | |
790 | * occur while the underlying object is being cleaned out. | |
791 | */ | |
792 | VOP_LOCK(vp); | |
2bae1875 KM |
793 | /* |
794 | * Prevent the vnode from being recycled or | |
795 | * brought into use while we clean it out. | |
796 | */ | |
0bf84b18 KM |
797 | if (vp->v_flag & VXLOCK) |
798 | panic("vclean: deadlock"); | |
ef24f6dd | 799 | vp->v_flag |= VXLOCK; |
0bf84b18 | 800 | /* |
669df1aa | 801 | * Clean out any buffers associated with the vnode. |
0bf84b18 | 802 | */ |
36ef03ec | 803 | if (flags & DOCLOSE) |
c33e9e8b | 804 | vinvalbuf(vp, V_SAVE, NOCRED, NULL, 0, 0); |
ef24f6dd | 805 | /* |
669df1aa KM |
806 | * Any other processes trying to obtain this lock must first |
807 | * wait for VXLOCK to clear, then call the new lock operation. | |
ef24f6dd | 808 | */ |
669df1aa | 809 | VOP_UNLOCK(vp); |
ef24f6dd | 810 | /* |
669df1aa KM |
811 | * If purging an active vnode, it must be closed and |
812 | * deactivated before being reclaimed. | |
ef24f6dd | 813 | */ |
2bae1875 | 814 | if (active) { |
669df1aa KM |
815 | if (flags & DOCLOSE) |
816 | VOP_CLOSE(vp, IO_NDELAY, NOCRED, NULL); | |
817 | VOP_INACTIVE(vp); | |
ef24f6dd KM |
818 | } |
819 | /* | |
820 | * Reclaim the vnode. | |
821 | */ | |
669df1aa | 822 | if (VOP_RECLAIM(vp)) |
ef24f6dd | 823 | panic("vclean: cannot reclaim"); |
2bae1875 KM |
824 | if (active) |
825 | vrele(vp); | |
38c46eee | 826 | |
ef24f6dd | 827 | /* |
669df1aa | 828 | * Done with purge, notify sleepers of the grim news. |
ef24f6dd | 829 | */ |
669df1aa KM |
830 | vp->v_op = dead_vnodeop_p; |
831 | vp->v_tag = VT_NON; | |
ef24f6dd KM |
832 | vp->v_flag &= ~VXLOCK; |
833 | if (vp->v_flag & VXWANT) { | |
834 | vp->v_flag &= ~VXWANT; | |
835 | wakeup((caddr_t)vp); | |
836 | } | |
837 | } | |
838 | ||
ef62830d KM |
839 | /* |
840 | * Eliminate all activity associated with the requested vnode | |
841 | * and with all vnodes aliased to the requested vnode. | |
842 | */ | |
843 | void vgoneall(vp) | |
844 | register struct vnode *vp; | |
845 | { | |
7f7b7d89 | 846 | register struct vnode *vq; |
ef62830d | 847 | |
7a7b3a95 KM |
848 | if (vp->v_flag & VALIASED) { |
849 | /* | |
850 | * If a vgone (or vclean) is already in progress, | |
851 | * wait until it is done and return. | |
852 | */ | |
853 | if (vp->v_flag & VXLOCK) { | |
854 | vp->v_flag |= VXWANT; | |
855 | sleep((caddr_t)vp, PINOD); | |
856 | return; | |
857 | } | |
858 | /* | |
859 | * Ensure that vp will not be vgone'd while we | |
860 | * are eliminating its aliases. | |
861 | */ | |
862 | vp->v_flag |= VXLOCK; | |
863 | while (vp->v_flag & VALIASED) { | |
864 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
865 | if (vq->v_rdev != vp->v_rdev || | |
866 | vq->v_type != vp->v_type || vp == vq) | |
867 | continue; | |
868 | vgone(vq); | |
869 | break; | |
870 | } | |
ef62830d | 871 | } |
7a7b3a95 KM |
872 | /* |
873 | * Remove the lock so that vgone below will | |
874 | * really eliminate the vnode after which time | |
875 | * vgone will awaken any sleepers. | |
876 | */ | |
877 | vp->v_flag &= ~VXLOCK; | |
ef62830d KM |
878 | } |
879 | vgone(vp); | |
880 | } | |
881 | ||
ef24f6dd KM |
882 | /* |
883 | * Eliminate all activity associated with a vnode | |
884 | * in preparation for reuse. | |
885 | */ | |
886 | void vgone(vp) | |
887 | register struct vnode *vp; | |
888 | { | |
7f7b7d89 | 889 | register struct vnode *vq; |
c0de8792 | 890 | struct vnode *vx; |
ef24f6dd | 891 | |
4f55e3ec KM |
892 | /* |
893 | * If a vgone (or vclean) is already in progress, | |
894 | * wait until it is done and return. | |
895 | */ | |
896 | if (vp->v_flag & VXLOCK) { | |
897 | vp->v_flag |= VXWANT; | |
898 | sleep((caddr_t)vp, PINOD); | |
899 | return; | |
900 | } | |
ef24f6dd KM |
901 | /* |
902 | * Clean out the filesystem specific data. | |
903 | */ | |
36ef03ec | 904 | vclean(vp, DOCLOSE); |
ef24f6dd KM |
905 | /* |
906 | * Delete from old mount point vnode list, if on one. | |
907 | */ | |
3fc2ac18 | 908 | if (vp->v_mount != NULL) { |
3e787e54 KM |
909 | if (vp->v_mntvnodes.le_next == (struct vnode *)0xdeadf || |
910 | vp->v_mntvnodes.le_prev == (struct vnode **)0xdeadb) | |
911 | panic("vgone: not on queue"); | |
3fc2ac18 | 912 | LIST_REMOVE(vp, v_mntvnodes); |
3e787e54 KM |
913 | vp->v_mntvnodes.le_next = (struct vnode *)0xdeadf; |
914 | vp->v_mntvnodes.le_prev = (struct vnode **)0xdeadb; | |
d10e9258 | 915 | vp->v_mount = NULL; |
ef24f6dd KM |
916 | } |
917 | /* | |
918 | * If special device, remove it from special device alias list. | |
919 | */ | |
920 | if (vp->v_type == VBLK || vp->v_type == VCHR) { | |
7f7b7d89 KM |
921 | if (*vp->v_hashchain == vp) { |
922 | *vp->v_hashchain = vp->v_specnext; | |
ef24f6dd | 923 | } else { |
7f7b7d89 | 924 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { |
c0de8792 | 925 | if (vq->v_specnext != vp) |
ef24f6dd | 926 | continue; |
c0de8792 | 927 | vq->v_specnext = vp->v_specnext; |
ef24f6dd KM |
928 | break; |
929 | } | |
c0de8792 | 930 | if (vq == NULL) |
ef24f6dd KM |
931 | panic("missing bdev"); |
932 | } | |
c0de8792 | 933 | if (vp->v_flag & VALIASED) { |
4d1ee2eb | 934 | vx = NULL; |
7f7b7d89 | 935 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { |
de81e10c KM |
936 | if (vq->v_rdev != vp->v_rdev || |
937 | vq->v_type != vp->v_type) | |
c0de8792 | 938 | continue; |
4d1ee2eb CT |
939 | if (vx) |
940 | break; | |
c0de8792 KM |
941 | vx = vq; |
942 | } | |
4d1ee2eb | 943 | if (vx == NULL) |
c0de8792 | 944 | panic("missing alias"); |
4d1ee2eb | 945 | if (vq == NULL) |
c0de8792 KM |
946 | vx->v_flag &= ~VALIASED; |
947 | vp->v_flag &= ~VALIASED; | |
948 | } | |
949 | FREE(vp->v_specinfo, M_VNODE); | |
950 | vp->v_specinfo = NULL; | |
ef24f6dd KM |
951 | } |
952 | /* | |
3387ef89 | 953 | * If it is on the freelist and not already at the head, |
0bf9bb76 KM |
954 | * move it to the head of the list. The test of the back |
955 | * pointer and the reference count of zero is because | |
956 | * it will be removed from the free list by getnewvnode, | |
957 | * but will not have its reference count incremented until | |
958 | * after calling vgone. If the reference count were | |
959 | * incremented first, vgone would (incorrectly) try to | |
960 | * close the previous instance of the underlying object. | |
961 | * So, the back pointer is explicitly set to `0xdeadb' in | |
962 | * getnewvnode after removing it from the freelist to ensure | |
963 | * that we do not try to move it here. | |
ef24f6dd | 964 | */ |
0bf9bb76 KM |
965 | if (vp->v_usecount == 0 && |
966 | vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb && | |
967 | vnode_free_list.tqh_first != vp) { | |
3e787e54 KM |
968 | if (vp->v_freelist.tqe_next == (struct vnode *)0xdeadf) |
969 | panic("vgone: use 0, not free"); | |
3fc2ac18 KM |
970 | TAILQ_REMOVE(&vnode_free_list, vp, v_freelist); |
971 | TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist); | |
ef24f6dd | 972 | } |
2bae1875 | 973 | vp->v_type = VBAD; |
36d09cb1 | 974 | } |
ef62830d | 975 | |
2bcd6066 KM |
976 | /* |
977 | * Lookup a vnode by device number. | |
978 | */ | |
979 | vfinddev(dev, type, vpp) | |
980 | dev_t dev; | |
981 | enum vtype type; | |
982 | struct vnode **vpp; | |
983 | { | |
984 | register struct vnode *vp; | |
985 | ||
986 | for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) { | |
987 | if (dev != vp->v_rdev || type != vp->v_type) | |
988 | continue; | |
989 | *vpp = vp; | |
05378ee4 | 990 | return (1); |
2bcd6066 | 991 | } |
05378ee4 | 992 | return (0); |
2bcd6066 KM |
993 | } |
994 | ||
ef62830d KM |
995 | /* |
996 | * Calculate the total number of references to a special device. | |
997 | */ | |
998 | vcount(vp) | |
999 | register struct vnode *vp; | |
1000 | { | |
7f7b7d89 | 1001 | register struct vnode *vq; |
ef62830d KM |
1002 | int count; |
1003 | ||
1004 | if ((vp->v_flag & VALIASED) == 0) | |
7f7b7d89 | 1005 | return (vp->v_usecount); |
ef62830d | 1006 | loop: |
7f7b7d89 | 1007 | for (count = 0, vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { |
de81e10c | 1008 | if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type) |
ef62830d KM |
1009 | continue; |
1010 | /* | |
1011 | * Alias, but not in use, so flush it out. | |
1012 | */ | |
7f7b7d89 | 1013 | if (vq->v_usecount == 0) { |
ef62830d KM |
1014 | vgone(vq); |
1015 | goto loop; | |
1016 | } | |
7f7b7d89 | 1017 | count += vq->v_usecount; |
ef62830d KM |
1018 | } |
1019 | return (count); | |
1020 | } | |
0bf84b18 KM |
1021 | |
1022 | /* | |
1023 | * Print out a description of a vnode. | |
1024 | */ | |
1025 | static char *typename[] = | |
61f846a8 | 1026 | { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" }; |
0bf84b18 KM |
1027 | |
1028 | vprint(label, vp) | |
1029 | char *label; | |
1030 | register struct vnode *vp; | |
1031 | { | |
f2f730c6 | 1032 | char buf[64]; |
0bf84b18 KM |
1033 | |
1034 | if (label != NULL) | |
1035 | printf("%s: ", label); | |
3e787e54 | 1036 | printf("num %d ", vp->v_spare[0]); |
65c3b3a8 KM |
1037 | printf("type %s, usecount %d, writecount %d, refcount %d,", |
1038 | typename[vp->v_type], vp->v_usecount, vp->v_writecount, | |
1039 | vp->v_holdcnt); | |
f2f730c6 KM |
1040 | buf[0] = '\0'; |
1041 | if (vp->v_flag & VROOT) | |
1042 | strcat(buf, "|VROOT"); | |
1043 | if (vp->v_flag & VTEXT) | |
1044 | strcat(buf, "|VTEXT"); | |
36ef03ec KM |
1045 | if (vp->v_flag & VSYSTEM) |
1046 | strcat(buf, "|VSYSTEM"); | |
36ef03ec KM |
1047 | if (vp->v_flag & VXLOCK) |
1048 | strcat(buf, "|VXLOCK"); | |
1049 | if (vp->v_flag & VXWANT) | |
1050 | strcat(buf, "|VXWANT"); | |
f2f730c6 KM |
1051 | if (vp->v_flag & VBWAIT) |
1052 | strcat(buf, "|VBWAIT"); | |
36ef03ec KM |
1053 | if (vp->v_flag & VALIASED) |
1054 | strcat(buf, "|VALIASED"); | |
f2f730c6 KM |
1055 | if (buf[0] != '\0') |
1056 | printf(" flags (%s)", &buf[1]); | |
3fc2ac18 KM |
1057 | if (vp->v_data == NULL) { |
1058 | printf("\n"); | |
1059 | } else { | |
1060 | printf("\n\t"); | |
1061 | VOP_PRINT(vp); | |
1062 | } | |
0bf84b18 | 1063 | } |
985cbdd5 | 1064 | |
34c62e18 KM |
1065 | #ifdef DEBUG |
1066 | /* | |
1067 | * List all of the locked vnodes in the system. | |
1068 | * Called when debugging the kernel. | |
1069 | */ | |
1070 | printlockedvnodes() | |
1071 | { | |
1072 | register struct mount *mp; | |
1073 | register struct vnode *vp; | |
1074 | ||
1075 | printf("Locked vnodes\n"); | |
3fc2ac18 KM |
1076 | for (mp = mountlist.tqh_first; mp != NULL; mp = mp->mnt_list.tqe_next) { |
1077 | for (vp = mp->mnt_vnodelist.lh_first; | |
1078 | vp != NULL; | |
1079 | vp = vp->v_mntvnodes.le_next) | |
34c62e18 KM |
1080 | if (VOP_ISLOCKED(vp)) |
1081 | vprint((char *)0, vp); | |
3fc2ac18 | 1082 | } |
34c62e18 KM |
1083 | } |
1084 | #endif | |
1085 | ||
985cbdd5 MT |
1086 | int kinfo_vdebug = 1; |
1087 | int kinfo_vgetfailed; | |
1088 | #define KINFO_VNODESLOP 10 | |
1089 | /* | |
786fb484 | 1090 | * Dump vnode list (via sysctl). |
985cbdd5 MT |
1091 | * Copyout address of vnode followed by vnode. |
1092 | */ | |
aacc1bff | 1093 | /* ARGSUSED */ |
786fb484 | 1094 | sysctl_vnode(where, sizep) |
985cbdd5 | 1095 | char *where; |
c1909da4 | 1096 | size_t *sizep; |
985cbdd5 | 1097 | { |
3fc2ac18 | 1098 | register struct mount *mp, *nmp; |
985cbdd5 | 1099 | struct vnode *vp; |
985cbdd5 | 1100 | register char *bp = where, *savebp; |
5bf57294 | 1101 | char *ewhere; |
985cbdd5 MT |
1102 | int error; |
1103 | ||
1104 | #define VPTRSZ sizeof (struct vnode *) | |
1105 | #define VNODESZ sizeof (struct vnode) | |
1106 | if (where == NULL) { | |
786fb484 | 1107 | *sizep = (numvnodes + KINFO_VNODESLOP) * (VPTRSZ + VNODESZ); |
985cbdd5 MT |
1108 | return (0); |
1109 | } | |
786fb484 | 1110 | ewhere = where + *sizep; |
985cbdd5 | 1111 | |
3fc2ac18 KM |
1112 | for (mp = mountlist.tqh_first; mp != NULL; mp = nmp) { |
1113 | nmp = mp->mnt_list.tqe_next; | |
1114 | if (vfs_busy(mp)) | |
36ef03ec | 1115 | continue; |
985cbdd5 MT |
1116 | savebp = bp; |
1117 | again: | |
3fc2ac18 KM |
1118 | for (vp = mp->mnt_vnodelist.lh_first; |
1119 | vp != NULL; | |
1120 | vp = vp->v_mntvnodes.le_next) { | |
41185b3b KM |
1121 | /* |
1122 | * Check that the vp is still associated with | |
1123 | * this filesystem. RACE: could have been | |
1124 | * recycled onto the same filesystem. | |
1125 | */ | |
4597dd33 KM |
1126 | if (vp->v_mount != mp) { |
1127 | if (kinfo_vdebug) | |
1128 | printf("kinfo: vp changed\n"); | |
1129 | bp = savebp; | |
1130 | goto again; | |
1131 | } | |
786fb484 KM |
1132 | if (bp + VPTRSZ + VNODESZ > ewhere) { |
1133 | *sizep = bp - where; | |
1134 | return (ENOMEM); | |
1135 | } | |
1136 | if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) || | |
1137 | (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ))) | |
985cbdd5 | 1138 | return (error); |
985cbdd5 | 1139 | bp += VPTRSZ + VNODESZ; |
985cbdd5 | 1140 | } |
3fc2ac18 KM |
1141 | vfs_unbusy(mp); |
1142 | } | |
985cbdd5 | 1143 | |
786fb484 | 1144 | *sizep = bp - where; |
985cbdd5 MT |
1145 | return (0); |
1146 | } | |
8981e258 MH |
1147 | |
1148 | /* | |
1149 | * Check to see if a filesystem is mounted on a block device. | |
1150 | */ | |
1151 | int | |
1152 | vfs_mountedon(vp) | |
1153 | register struct vnode *vp; | |
1154 | { | |
1155 | register struct vnode *vq; | |
1156 | ||
1157 | if (vp->v_specflags & SI_MOUNTEDON) | |
1158 | return (EBUSY); | |
1159 | if (vp->v_flag & VALIASED) { | |
1160 | for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) { | |
1161 | if (vq->v_rdev != vp->v_rdev || | |
1162 | vq->v_type != vp->v_type) | |
1163 | continue; | |
1164 | if (vq->v_specflags & SI_MOUNTEDON) | |
1165 | return (EBUSY); | |
1166 | } | |
1167 | } | |
1168 | return (0); | |
1169 | } | |
1170 | ||
1171 | /* | |
1172 | * Build hash lists of net addresses and hang them off the mount point. | |
1173 | * Called by ufs_mount() to set up the lists of export addresses. | |
1174 | */ | |
1175 | static int | |
1176 | vfs_hang_addrlist(mp, nep, argp) | |
1177 | struct mount *mp; | |
1178 | struct netexport *nep; | |
1179 | struct export_args *argp; | |
1180 | { | |
1181 | register struct netcred *np; | |
1182 | register struct radix_node_head *rnh; | |
1183 | register int i; | |
1184 | struct radix_node *rn; | |
1185 | struct sockaddr *saddr, *smask = 0; | |
1186 | struct domain *dom; | |
1187 | int error; | |
1188 | ||
1189 | if (argp->ex_addrlen == 0) { | |
1190 | if (mp->mnt_flag & MNT_DEFEXPORTED) | |
1191 | return (EPERM); | |
1192 | np = &nep->ne_defexported; | |
1193 | np->netc_exflags = argp->ex_flags; | |
1194 | np->netc_anon = argp->ex_anon; | |
1195 | np->netc_anon.cr_ref = 1; | |
1196 | mp->mnt_flag |= MNT_DEFEXPORTED; | |
1197 | return (0); | |
1198 | } | |
1199 | i = sizeof(struct netcred) + argp->ex_addrlen + argp->ex_masklen; | |
1200 | np = (struct netcred *)malloc(i, M_NETADDR, M_WAITOK); | |
1201 | bzero((caddr_t)np, i); | |
1202 | saddr = (struct sockaddr *)(np + 1); | |
1203 | if (error = copyin(argp->ex_addr, (caddr_t)saddr, argp->ex_addrlen)) | |
1204 | goto out; | |
1205 | if (saddr->sa_len > argp->ex_addrlen) | |
1206 | saddr->sa_len = argp->ex_addrlen; | |
1207 | if (argp->ex_masklen) { | |
1208 | smask = (struct sockaddr *)((caddr_t)saddr + argp->ex_addrlen); | |
1209 | error = copyin(argp->ex_addr, (caddr_t)smask, argp->ex_masklen); | |
1210 | if (error) | |
1211 | goto out; | |
1212 | if (smask->sa_len > argp->ex_masklen) | |
1213 | smask->sa_len = argp->ex_masklen; | |
1214 | } | |
1215 | i = saddr->sa_family; | |
1216 | if ((rnh = nep->ne_rtable[i]) == 0) { | |
1217 | /* | |
1218 | * Seems silly to initialize every AF when most are not | |
1219 | * used, do so on demand here | |
1220 | */ | |
1221 | for (dom = domains; dom; dom = dom->dom_next) | |
1222 | if (dom->dom_family == i && dom->dom_rtattach) { | |
1223 | dom->dom_rtattach((void **)&nep->ne_rtable[i], | |
1224 | dom->dom_rtoffset); | |
1225 | break; | |
1226 | } | |
1227 | if ((rnh = nep->ne_rtable[i]) == 0) { | |
1228 | error = ENOBUFS; | |
1229 | goto out; | |
1230 | } | |
1231 | } | |
1232 | rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, | |
1233 | np->netc_rnodes); | |
1234 | if (rn == 0 || np != (struct netcred *)rn) { /* already exists */ | |
1235 | error = EPERM; | |
1236 | goto out; | |
1237 | } | |
1238 | np->netc_exflags = argp->ex_flags; | |
1239 | np->netc_anon = argp->ex_anon; | |
1240 | np->netc_anon.cr_ref = 1; | |
1241 | return (0); | |
1242 | out: | |
1243 | free(np, M_NETADDR); | |
1244 | return (error); | |
1245 | } | |
1246 | ||
1247 | /* ARGSUSED */ | |
1248 | static int | |
1249 | vfs_free_netcred(rn, w) | |
1250 | struct radix_node *rn; | |
1251 | caddr_t w; | |
1252 | { | |
1253 | register struct radix_node_head *rnh = (struct radix_node_head *)w; | |
1254 | ||
1255 | (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh); | |
1256 | free((caddr_t)rn, M_NETADDR); | |
1257 | return (0); | |
1258 | } | |
1259 | ||
1260 | /* | |
1261 | * Free the net address hash lists that are hanging off the mount points. | |
1262 | */ | |
1263 | static void | |
1264 | vfs_free_addrlist(nep) | |
1265 | struct netexport *nep; | |
1266 | { | |
1267 | register int i; | |
1268 | register struct radix_node_head *rnh; | |
1269 | ||
1270 | for (i = 0; i <= AF_MAX; i++) | |
1271 | if (rnh = nep->ne_rtable[i]) { | |
1272 | (*rnh->rnh_walktree)(rnh, vfs_free_netcred, | |
1273 | (caddr_t)rnh); | |
1274 | free((caddr_t)rnh, M_RTABLE); | |
1275 | nep->ne_rtable[i] = 0; | |
1276 | } | |
1277 | } | |
1278 | ||
1279 | int | |
1280 | vfs_export(mp, nep, argp) | |
1281 | struct mount *mp; | |
1282 | struct netexport *nep; | |
1283 | struct export_args *argp; | |
1284 | { | |
1285 | int error; | |
1286 | ||
1287 | if (argp->ex_flags & MNT_DELEXPORT) { | |
1288 | vfs_free_addrlist(nep); | |
1289 | mp->mnt_flag &= ~(MNT_EXPORTED | MNT_DEFEXPORTED); | |
1290 | } | |
1291 | if (argp->ex_flags & MNT_EXPORTED) { | |
1292 | if (error = vfs_hang_addrlist(mp, nep, argp)) | |
1293 | return (error); | |
1294 | mp->mnt_flag |= MNT_EXPORTED; | |
1295 | } | |
1296 | return (0); | |
1297 | } | |
1298 | ||
1299 | struct netcred * | |
1300 | vfs_export_lookup(mp, nep, nam) | |
1301 | register struct mount *mp; | |
1302 | struct netexport *nep; | |
1303 | struct mbuf *nam; | |
1304 | { | |
1305 | register struct netcred *np; | |
1306 | register struct radix_node_head *rnh; | |
1307 | struct sockaddr *saddr; | |
1308 | ||
1309 | np = NULL; | |
1310 | if (mp->mnt_flag & MNT_EXPORTED) { | |
1311 | /* | |
1312 | * Lookup in the export list first. | |
1313 | */ | |
1314 | if (nam != NULL) { | |
1315 | saddr = mtod(nam, struct sockaddr *); | |
1316 | rnh = nep->ne_rtable[saddr->sa_family]; | |
1317 | if (rnh != NULL) { | |
1318 | np = (struct netcred *) | |
1319 | (*rnh->rnh_matchaddr)((caddr_t)saddr, | |
1320 | rnh); | |
1321 | if (np && np->netc_rnodes->rn_flags & RNF_ROOT) | |
1322 | np = NULL; | |
1323 | } | |
1324 | } | |
1325 | /* | |
1326 | * If no address match, use the default if it exists. | |
1327 | */ | |
1328 | if (np == NULL && mp->mnt_flag & MNT_DEFEXPORTED) | |
1329 | np = &nep->ne_defexported; | |
1330 | } | |
1331 | return (np); | |
1332 | } |