Commit | Line | Data |
---|---|---|
da7c5cc6 | 1 | /* |
7188ac27 KM |
2 | * Copyright (c) 1982, 1986, 1989 Regents of the University of California. |
3 | * All rights reserved. | |
da7c5cc6 | 4 | * |
7188ac27 KM |
5 | * Redistribution and use in source and binary forms are permitted |
6 | * provided that the above copyright notice and this paragraph are | |
7 | * duplicated in all such forms and that any documentation, | |
8 | * advertising materials, and other materials related to such | |
9 | * distribution and use acknowledge that the software was developed | |
10 | * by the University of California, Berkeley. The name of the | |
11 | * University may not be used to endorse or promote products derived | |
12 | * from this software without specific prior written permission. | |
13 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR | |
14 | * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED | |
15 | * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE. | |
16 | * | |
0d73cd0c | 17 | * @(#)lfs_inode.c 7.13 (Berkeley) %G% |
da7c5cc6 | 18 | */ |
5d5124a1 | 19 | |
94368568 JB |
20 | #include "param.h" |
21 | #include "systm.h" | |
22 | #include "mount.h" | |
94368568 | 23 | #include "user.h" |
7188ac27 | 24 | #include "file.h" |
94368568 | 25 | #include "buf.h" |
0b355a6e | 26 | #include "cmap.h" |
7188ac27 KM |
27 | #include "vnode.h" |
28 | #include "../ufs/inode.h" | |
29 | #include "../ufs/fs.h" | |
30 | #include "../ufs/ufsmount.h" | |
b4567e9c | 31 | #ifdef QUOTA |
7188ac27 | 32 | #include "../ufs/quota.h" |
4147b3f6 | 33 | #endif |
94368568 | 34 | #include "kernel.h" |
b30358ab | 35 | #include "malloc.h" |
5d5124a1 | 36 | |
c22c66ff | 37 | #define INOHSZ 512 |
3ebac878 RE |
38 | #if ((INOHSZ&(INOHSZ-1)) == 0) |
39 | #define INOHASH(dev,ino) (((dev)+(ino))&(INOHSZ-1)) | |
40 | #else | |
a3a9487d | 41 | #define INOHASH(dev,ino) (((unsigned)((dev)+(ino)))%INOHSZ) |
3ebac878 RE |
42 | #endif |
43 | ||
7188ac27 KM |
44 | #define INSFREE(ip) {\ |
45 | if (ifreeh) { \ | |
46 | *ifreet = (ip); \ | |
47 | (ip)->i_freeb = ifreet; \ | |
48 | } else { \ | |
49 | ifreeh = (ip); \ | |
50 | (ip)->i_freeb = &ifreeh; \ | |
51 | } \ | |
52 | (ip)->i_freef = NULL; \ | |
53 | ifreet = &(ip)->i_freef; \ | |
54 | } | |
55 | ||
3ebac878 RE |
56 | union ihead { /* inode LRU cache, Chris Maltby */ |
57 | union ihead *ih_head[2]; | |
58 | struct inode *ih_chain[2]; | |
59 | } ihead[INOHSZ]; | |
60 | ||
7188ac27 | 61 | struct inode *ifreeh, **ifreet, *bdevlisth; |
5d5124a1 BJ |
62 | |
63 | /* | |
64 | * Initialize hash links for inodes | |
65 | * and build inode free list. | |
66 | */ | |
67 | ihinit() | |
68 | { | |
69 | register int i; | |
75105cf0 | 70 | register struct inode *ip = inode; |
3ebac878 | 71 | register union ihead *ih = ihead; |
5d5124a1 | 72 | |
3ebac878 RE |
73 | for (i = INOHSZ; --i >= 0; ih++) { |
74 | ih->ih_head[0] = ih; | |
75 | ih->ih_head[1] = ih; | |
76 | } | |
77 | ifreeh = ip; | |
78 | ifreet = &ip->i_freef; | |
79 | ip->i_freeb = &ifreeh; | |
80 | ip->i_forw = ip; | |
81 | ip->i_back = ip; | |
7188ac27 | 82 | ITOV(ip)->v_data = (qaddr_t)ip; |
3ebac878 RE |
83 | for (i = ninode; --i > 0; ) { |
84 | ++ip; | |
85 | ip->i_forw = ip; | |
86 | ip->i_back = ip; | |
7188ac27 | 87 | ITOV(ip)->v_data = (qaddr_t)ip; |
3ebac878 RE |
88 | *ifreet = ip; |
89 | ip->i_freeb = ifreet; | |
90 | ifreet = &ip->i_freef; | |
91 | } | |
92 | ip->i_freef = NULL; | |
5d5124a1 BJ |
93 | } |
94 | ||
3ebac878 | 95 | /* |
7188ac27 | 96 | * Look up an vnode/inode by device,inumber. |
5d5124a1 BJ |
97 | * If it is in core (in the inode structure), |
98 | * honor the locking protocol. | |
99 | * If it is not in core, read it in from the | |
100 | * specified device. | |
7188ac27 | 101 | * Callers must check for mount points!! |
5d5124a1 BJ |
102 | * In all cases, a pointer to a locked |
103 | * inode structure is returned. | |
5d5124a1 | 104 | */ |
7188ac27 KM |
105 | iget(xp, ino, ipp) |
106 | struct inode *xp; | |
7494ef16 | 107 | ino_t ino; |
7188ac27 | 108 | struct inode **ipp; |
5d5124a1 | 109 | { |
7188ac27 KM |
110 | dev_t dev = xp->i_dev; |
111 | struct mount *mntp = ITOV(xp)->v_mount; | |
112 | register struct fs *fs = VFSTOUFS(mntp)->um_fs; | |
113 | register struct inode *ip, *iq; | |
114 | register struct vnode *vp; | |
115 | struct inode *nip; | |
116 | struct buf *bp; | |
117 | struct dinode tdip, *dp; | |
118 | union ihead *ih; | |
119 | int error; | |
2e64ab65 | 120 | |
5d5124a1 | 121 | loop: |
3ebac878 RE |
122 | ih = &ihead[INOHASH(dev, ino)]; |
123 | for (ip = ih->ih_chain[0]; ip != (struct inode *)ih; ip = ip->i_forw) | |
7494ef16 | 124 | if (ino == ip->i_number && dev == ip->i_dev) { |
8ac1234a SL |
125 | /* |
126 | * Following is essentially an inline expanded | |
127 | * copy of igrab(), expanded inline for speed, | |
128 | * and so that the test for a mounted on inode | |
129 | * can be deferred until after we are sure that | |
130 | * the inode isn't busy. | |
131 | */ | |
5c2ba954 | 132 | if ((ip->i_flag&ILOCKED) != 0) { |
5d5124a1 BJ |
133 | ip->i_flag |= IWANT; |
134 | sleep((caddr_t)ip, PINOD); | |
135 | goto loop; | |
136 | } | |
7188ac27 KM |
137 | vp = ITOV(ip); |
138 | if (vp->v_count == 0) { /* ino on free list */ | |
3ebac878 RE |
139 | if (iq = ip->i_freef) |
140 | iq->i_freeb = ip->i_freeb; | |
141 | else | |
142 | ifreet = ip->i_freeb; | |
143 | *ip->i_freeb = iq; | |
144 | ip->i_freef = NULL; | |
145 | ip->i_freeb = NULL; | |
146 | } | |
aed86454 | 147 | ILOCK(ip); |
8fe1c702 | 148 | VREF(vp); |
7188ac27 KM |
149 | *ipp = ip; |
150 | return(0); | |
151 | } | |
152 | if (error = getnewino(dev, ino, &nip)) { | |
153 | *ipp = 0; | |
154 | return (error); | |
155 | } | |
156 | ip = nip; | |
157 | /* | |
158 | * Read in the disk contents for the inode. | |
159 | */ | |
160 | if (error = bread(VFSTOUFS(mntp)->um_devvp, fsbtodb(fs, itod(fs, ino)), | |
a937f856 | 161 | (int)fs->fs_bsize, NOCRED, &bp)) { |
7188ac27 KM |
162 | /* |
163 | * The inode doesn't contain anything useful, so it would | |
164 | * be misleading to leave it on its hash chain. Iput() will | |
165 | * take care of putting it back on the free list. We also | |
166 | * lose its inumber, just in case. | |
167 | */ | |
168 | remque(ip); | |
169 | ip->i_forw = ip; | |
170 | ip->i_back = ip; | |
171 | ip->i_number = 0; | |
9b651315 | 172 | ITOV(ip)->v_type = VNON; |
7188ac27 | 173 | INSFREE(ip); |
aed86454 | 174 | iunlock(ip); |
7188ac27 KM |
175 | ip->i_flag = 0; |
176 | brelse(bp); | |
177 | *ipp = 0; | |
178 | return(error); | |
179 | } | |
180 | /* | |
181 | * Check to see if the new inode represents a block device | |
182 | * for which we already have an inode (either because of | |
183 | * bdevvp() or because of a different inode representing | |
184 | * the same block device). If such an alias exists, put the | |
185 | * just allocated inode back on the free list, and replace | |
186 | * the contents of the existing inode with the contents of | |
187 | * the new inode. | |
188 | */ | |
189 | dp = bp->b_un.b_dino; | |
190 | dp += itoo(fs, ino); | |
191 | if ((dp->di_mode & IFMT) != IFBLK) { | |
192 | ip->i_ic = dp->di_ic; | |
193 | brelse(bp); | |
194 | } else { | |
195 | again: | |
196 | for (iq = bdevlisth; iq; iq = iq->i_devlst) { | |
9b651315 KM |
197 | vp = ITOV(iq); |
198 | if (dp->di_rdev != vp->v_rdev) | |
7188ac27 KM |
199 | continue; |
200 | igrab(iq); | |
9b651315 | 201 | if (dp->di_rdev != vp->v_rdev) { |
7188ac27 KM |
202 | iput(iq); |
203 | goto again; | |
204 | } | |
205 | /* | |
206 | * Discard unneeded inode. | |
207 | */ | |
208 | remque(ip); | |
209 | ip->i_forw = ip; | |
210 | ip->i_back = ip; | |
211 | ip->i_number = 0; | |
9b651315 | 212 | ITOV(ip)->v_type = VNON; |
7188ac27 | 213 | INSFREE(ip); |
aed86454 | 214 | iunlock(ip); |
7188ac27 KM |
215 | ip->i_flag = 0; |
216 | /* | |
217 | * Reinitialize aliased inode. | |
218 | * We must release the buffer that we just read | |
219 | * before doing the iupdat() to avoid a possible | |
220 | * deadlock with updating an inode in the same | |
221 | * disk block. | |
222 | */ | |
223 | ip = iq; | |
7188ac27 KM |
224 | tdip.di_ic = dp->di_ic; |
225 | brelse(bp); | |
226 | error = iupdat(ip, &time, &time, 1); | |
227 | ip->i_ic = tdip.di_ic; | |
228 | remque(ip); | |
229 | insque(ip, ih); | |
230 | ip->i_dev = dev; | |
231 | ip->i_number = ino; | |
232 | if (ip->i_devvp) { | |
233 | vrele(ip->i_devvp); | |
234 | ip->i_devvp = 0; | |
235 | } | |
236 | cache_purge(vp); | |
237 | break; | |
238 | } | |
239 | if (iq == 0) { | |
240 | ip->i_ic = dp->di_ic; | |
241 | brelse(bp); | |
242 | ip->i_devlst = bdevlisth; | |
243 | bdevlisth = ip; | |
5d5124a1 | 244 | } |
7188ac27 KM |
245 | } |
246 | /* | |
247 | * Finish inode initialization. | |
248 | */ | |
249 | ip->i_fs = fs; | |
250 | ip->i_devvp = VFSTOUFS(mntp)->um_devvp; | |
8fe1c702 | 251 | VREF(ip->i_devvp); |
7188ac27 KM |
252 | /* |
253 | * Initialize the associated vnode | |
254 | */ | |
255 | vp = ITOV(ip); | |
256 | vinit(vp, mntp, IFTOVT(ip->i_mode), &ufs_vnodeops); | |
257 | if (vp->v_type == VCHR || vp->v_type == VBLK) { | |
258 | vp->v_rdev = ip->i_rdev; | |
259 | vp->v_op = &blk_vnodeops; | |
260 | } | |
261 | if (ino == ROOTINO) | |
262 | vp->v_flag |= VROOT; | |
263 | #ifdef QUOTA | |
264 | if (ip->i_mode != 0) | |
265 | ip->i_dquot = inoquota(ip); | |
266 | #endif | |
afd7e202 KM |
267 | /* |
268 | * Set up a generation number for this inode if it does not | |
269 | * already have one. This should only happen on old filesystems. | |
270 | */ | |
271 | if (ip->i_gen == 0) { | |
272 | if (++nextgennumber < (u_long)time.tv_sec) | |
273 | nextgennumber = time.tv_sec; | |
274 | ip->i_gen = nextgennumber; | |
275 | if ((vp->v_mount->m_flag & M_RDONLY) == 0) | |
276 | ip->i_flag |= IMOD; | |
277 | } | |
7188ac27 KM |
278 | *ipp = ip; |
279 | return (0); | |
280 | } | |
3ebac878 | 281 | |
7188ac27 KM |
282 | /* |
283 | * Allocate a new inode. | |
284 | * | |
285 | * Put it onto its hash chain and lock it so that other requests for | |
286 | * this inode will block if they arrive while we are sleeping waiting | |
287 | * for old data structures to be purged or for the contents of the disk | |
288 | * portion of this inode to be read. | |
289 | */ | |
290 | getnewino(dev, ino, ipp) | |
291 | dev_t dev; | |
292 | ino_t ino; | |
293 | struct inode **ipp; | |
294 | { | |
295 | union ihead *ih; | |
296 | register struct inode *ip, *iq; | |
297 | register struct vnode *vp; | |
298 | ||
299 | /* | |
300 | * Remove the next inode from the free list. | |
301 | */ | |
3ebac878 | 302 | if ((ip = ifreeh) == NULL) { |
945fbb1b | 303 | tablefull("inode"); |
7188ac27 KM |
304 | *ipp = 0; |
305 | return(ENFILE); | |
5d5124a1 | 306 | } |
7188ac27 KM |
307 | vp = ITOV(ip); |
308 | if (vp->v_count) | |
bed1bb6e | 309 | panic("free inode isn't"); |
3ebac878 RE |
310 | if (iq = ip->i_freef) |
311 | iq->i_freeb = &ifreeh; | |
312 | ifreeh = iq; | |
313 | ip->i_freef = NULL; | |
314 | ip->i_freeb = NULL; | |
315 | /* | |
316 | * Now to take inode off the hash chain it was on | |
317 | * (initially, or after an iflush, it is on a "hash chain" | |
7188ac27 KM |
318 | * consisting entirely of itself, and pointed to by no-one) |
319 | * and put it on the chain for its new (ino, dev) pair. | |
3ebac878 | 320 | */ |
32dc2b7e | 321 | remque(ip); |
5d5124a1 BJ |
322 | ip->i_dev = dev; |
323 | ip->i_number = ino; | |
7188ac27 KM |
324 | if (dev != NODEV) { |
325 | ih = &ihead[INOHASH(dev, ino)]; | |
326 | insque(ip, ih); | |
327 | } | |
aed86454 KM |
328 | ip->i_flag = 0; |
329 | ILOCK(ip); | |
6459ebe0 | 330 | ip->i_lastr = 0; |
ec67a3ce | 331 | #endif SECSIZE |
5d5124a1 | 332 | /* |
7188ac27 | 333 | * Purge old data structures associated with the inode. |
5d5124a1 | 334 | */ |
7188ac27 KM |
335 | cache_purge(vp); |
336 | if (ip->i_devvp) { | |
337 | vrele(ip->i_devvp); | |
338 | ip->i_devvp = 0; | |
5d5124a1 | 339 | } |
b4567e9c | 340 | #ifdef QUOTA |
7188ac27 KM |
341 | dqrele(ip->i_dquot); |
342 | ip->i_dquot = NODQUOT; | |
89045c38 | 343 | #endif |
7188ac27 KM |
344 | if (vp->v_type == VBLK) { |
345 | if (bdevlisth == ip) { | |
346 | bdevlisth = ip->i_devlst; | |
347 | } else { | |
348 | for (iq = bdevlisth; iq; iq = iq->i_devlst) { | |
349 | if (iq->i_devlst != ip) | |
350 | continue; | |
351 | iq->i_devlst = ip->i_devlst; | |
352 | break; | |
353 | } | |
354 | if (iq == NULL) | |
355 | panic("missing bdev"); | |
356 | } | |
357 | } | |
358 | *ipp = ip; | |
359 | return (0); | |
5d5124a1 BJ |
360 | } |
361 | ||
8ac1234a SL |
362 | /* |
363 | * Convert a pointer to an inode into a reference to an inode. | |
364 | * | |
365 | * This is basically the internal piece of iget (after the | |
366 | * inode pointer is located) but without the test for mounted | |
367 | * filesystems. It is caller's responsibility to check that | |
368 | * the inode pointer is valid. | |
369 | */ | |
370 | igrab(ip) | |
371 | register struct inode *ip; | |
372 | { | |
7188ac27 KM |
373 | register struct vnode *vp = ITOV(ip); |
374 | ||
8ac1234a SL |
375 | while ((ip->i_flag&ILOCKED) != 0) { |
376 | ip->i_flag |= IWANT; | |
377 | sleep((caddr_t)ip, PINOD); | |
378 | } | |
7188ac27 | 379 | if (vp->v_count == 0) { /* ino on free list */ |
8ac1234a SL |
380 | register struct inode *iq; |
381 | ||
382 | if (iq = ip->i_freef) | |
383 | iq->i_freeb = ip->i_freeb; | |
384 | else | |
385 | ifreet = ip->i_freeb; | |
386 | *ip->i_freeb = iq; | |
387 | ip->i_freef = NULL; | |
388 | ip->i_freeb = NULL; | |
389 | } | |
8fe1c702 | 390 | VREF(vp); |
aed86454 | 391 | ILOCK(ip); |
8ac1234a SL |
392 | } |
393 | ||
7188ac27 KM |
394 | /* |
395 | * Create a vnode for a block device. | |
396 | * Used for root filesystem, argdev, and swap areas. | |
397 | */ | |
398 | bdevvp(dev, vpp) | |
399 | dev_t dev; | |
400 | struct vnode **vpp; | |
401 | { | |
402 | register struct inode *ip; | |
403 | register struct vnode *vp; | |
404 | struct inode *nip; | |
405 | int error; | |
406 | ||
407 | /* | |
408 | * Check for the existence of an existing vnode. | |
409 | */ | |
410 | again: | |
411 | for (ip = bdevlisth; ip; ip = ip->i_devlst) { | |
412 | vp = ITOV(ip); | |
413 | if (dev != vp->v_rdev) | |
414 | continue; | |
415 | igrab(ip); | |
416 | if (dev != vp->v_rdev) { | |
417 | iput(ip); | |
418 | goto again; | |
419 | } | |
420 | IUNLOCK(ip); | |
421 | *vpp = vp; | |
422 | return (0); | |
423 | } | |
424 | if (error = getnewino(NODEV, (ino_t)0, &nip)) { | |
425 | *vpp = 0; | |
426 | return (error); | |
427 | } | |
428 | ip = nip; | |
429 | ip->i_fs = 0; | |
430 | ip->i_devlst = bdevlisth; | |
431 | bdevlisth = ip; | |
432 | vp = ITOV(ip); | |
433 | vinit(vp, 0, VBLK, &blk_vnodeops); | |
434 | vp->v_rdev = dev; | |
435 | IUNLOCK(ip); | |
436 | *vpp = vp; | |
437 | return (0); | |
438 | } | |
439 | ||
5d5124a1 BJ |
440 | /* |
441 | * Decrement reference count of | |
442 | * an inode structure. | |
443 | * On the last reference, | |
444 | * write the inode out and if necessary, | |
445 | * truncate and deallocate the file. | |
446 | */ | |
447 | iput(ip) | |
7494ef16 | 448 | register struct inode *ip; |
5d5124a1 | 449 | { |
ff56f48a | 450 | |
5c2ba954 | 451 | if ((ip->i_flag & ILOCKED) == 0) |
ff56f48a | 452 | panic("iput"); |
a388503d | 453 | IUNLOCK(ip); |
7188ac27 | 454 | vrele(ITOV(ip)); |
ff56f48a KM |
455 | } |
456 | ||
7188ac27 KM |
457 | |
458 | ufs_inactive(vp) | |
459 | struct vnode *vp; | |
ff56f48a | 460 | { |
7188ac27 KM |
461 | register struct inode *ip = VTOI(vp); |
462 | int mode, error; | |
463 | ||
464 | if (ITOV(ip)->v_count != 0) | |
465 | panic("ufs_inactive: not inactive"); | |
b5ea418e KM |
466 | /* |
467 | * Get rid of inodes related to stale file handles. | |
468 | */ | |
469 | if (ip->i_mode == 0) | |
470 | goto freeit; | |
aed86454 | 471 | ILOCK(ip); |
7188ac27 KM |
472 | if (ip->i_nlink <= 0 && (ITOV(ip)->v_mount->m_flag&M_RDONLY) == 0) { |
473 | error = itrunc(ip, (u_long)0); | |
474 | mode = ip->i_mode; | |
475 | ip->i_mode = 0; | |
476 | ip->i_rdev = 0; | |
477 | ip->i_flag |= IUPD|ICHG; | |
478 | ifree(ip, ip->i_number, mode); | |
b4567e9c | 479 | #ifdef QUOTA |
7188ac27 KM |
480 | (void) chkiq(ip->i_dev, ip, ip->i_uid, 0); |
481 | dqrele(ip->i_dquot); | |
482 | ip->i_dquot = NODQUOT; | |
89045c38 | 483 | #endif |
7188ac27 KM |
484 | } |
485 | IUPDAT(ip, &time, &time, 0); | |
486 | IUNLOCK(ip); | |
b5ea418e | 487 | freeit: |
7188ac27 KM |
488 | ip->i_flag = 0; |
489 | /* | |
490 | * Put the inode on the end of the free list. | |
491 | * Possibly in some cases it would be better to | |
492 | * put the inode at the head of the free list, | |
493 | * (eg: where i_mode == 0 || i_number == 0). | |
494 | */ | |
495 | INSFREE(ip); | |
496 | return (error); | |
5d5124a1 BJ |
497 | } |
498 | ||
499 | /* | |
500 | * Check accessed and update flags on | |
501 | * an inode structure. | |
502 | * If any is on, update the inode | |
503 | * with the current time. | |
c0bb1685 BJ |
504 | * If waitfor is given, then must insure |
505 | * i/o order so wait for write to complete. | |
5d5124a1 | 506 | */ |
c0bb1685 | 507 | iupdat(ip, ta, tm, waitfor) |
7494ef16 | 508 | register struct inode *ip; |
b32450f4 | 509 | struct timeval *ta, *tm; |
7494ef16 | 510 | int waitfor; |
5d5124a1 | 511 | { |
7188ac27 KM |
512 | struct buf *bp; |
513 | struct vnode *vp = ITOV(ip); | |
5d5124a1 | 514 | struct dinode *dp; |
ec67a3ce | 515 | register struct fs *fs; |
5d5124a1 | 516 | |
ec67a3ce | 517 | fs = ip->i_fs; |
7188ac27 KM |
518 | if ((ip->i_flag & (IUPD|IACC|ICHG|IMOD)) == 0) |
519 | return (0); | |
520 | if (vp->v_mount->m_flag & M_RDONLY) | |
521 | return (0); | |
522 | error = bread(ip->i_devvp, fsbtodb(fs, itod(fs, ip->i_number)), | |
a937f856 | 523 | (int)fs->fs_bsize, NOCRED, &bp); |
7188ac27 KM |
524 | if (error) { |
525 | brelse(bp); | |
526 | return (error); | |
527 | } | |
528 | if (ip->i_flag&IACC) | |
529 | ip->i_atime = ta->tv_sec; | |
530 | if (ip->i_flag&IUPD) | |
531 | ip->i_mtime = tm->tv_sec; | |
532 | if (ip->i_flag&ICHG) | |
533 | ip->i_ctime = time.tv_sec; | |
534 | ip->i_flag &= ~(IUPD|IACC|ICHG|IMOD); | |
535 | dp = bp->b_un.b_dino + itoo(fs, ip->i_number); | |
536 | dp->di_ic = ip->i_ic; | |
537 | if (waitfor) { | |
538 | return (bwrite(bp)); | |
539 | } else { | |
540 | bdwrite(bp); | |
541 | return (0); | |
5d5124a1 BJ |
542 | } |
543 | } | |
544 | ||
9c03b2c0 SL |
545 | #define SINGLE 0 /* index of single indirect block */ |
546 | #define DOUBLE 1 /* index of double indirect block */ | |
547 | #define TRIPLE 2 /* index of triple indirect block */ | |
5d5124a1 | 548 | /* |
528f664c SL |
549 | * Truncate the inode ip to at most |
550 | * length size. Free affected disk | |
551 | * blocks -- the blocks of the file | |
552 | * are removed in reverse order. | |
9c03b2c0 SL |
553 | * |
554 | * NB: triple indirect blocks are untested. | |
5d5124a1 | 555 | */ |
9c03b2c0 | 556 | itrunc(oip, length) |
28821bc5 | 557 | register struct inode *oip; |
4f083fd7 | 558 | u_long length; |
5d5124a1 | 559 | { |
4f083fd7 | 560 | register daddr_t lastblock; |
a5e62f37 | 561 | daddr_t bn, lbn, lastiblock[NIADDR]; |
6459ebe0 | 562 | register struct fs *fs; |
9c03b2c0 | 563 | register struct inode *ip; |
28821bc5 | 564 | struct buf *bp; |
7188ac27 KM |
565 | int offset, osize, size, level; |
566 | long count, nblocks, blocksreleased = 0; | |
28821bc5 | 567 | register int i; |
7188ac27 | 568 | int error, allerror = 0; |
9c03b2c0 | 569 | struct inode tip; |
4f083fd7 | 570 | |
7b2e4f05 SL |
571 | if (oip->i_size <= length) { |
572 | oip->i_flag |= ICHG|IUPD; | |
7188ac27 KM |
573 | error = iupdat(oip, &time, &time, 1); |
574 | return (error); | |
7b2e4f05 | 575 | } |
c0bb1685 | 576 | /* |
9c03b2c0 SL |
577 | * Calculate index into inode's block list of |
578 | * last direct and indirect blocks (if any) | |
579 | * which we want to keep. Lastblock is -1 when | |
580 | * the file is truncated to 0. | |
c0bb1685 | 581 | */ |
9c03b2c0 | 582 | fs = oip->i_fs; |
4f083fd7 | 583 | lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1; |
9c03b2c0 SL |
584 | lastiblock[SINGLE] = lastblock - NDADDR; |
585 | lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs); | |
586 | lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs); | |
08d9a8ec | 587 | nblocks = btodb(fs->fs_bsize); |
6459ebe0 | 588 | /* |
28821bc5 KM |
589 | * Update the size of the file. If the file is not being |
590 | * truncated to a block boundry, the contents of the | |
591 | * partial block following the end of the file must be | |
592 | * zero'ed in case it ever become accessable again because | |
593 | * of subsequent file growth. | |
594 | */ | |
595 | osize = oip->i_size; | |
596 | offset = blkoff(fs, length); | |
597 | if (offset == 0) { | |
598 | oip->i_size = length; | |
599 | } else { | |
600 | lbn = lblkno(fs, length); | |
7188ac27 KM |
601 | error = balloc(oip, lbn, offset, &bn, B_CLRBUF); |
602 | if (error) | |
603 | return (error); | |
604 | if ((long)bn < 0) | |
605 | panic("itrunc: hole"); | |
28821bc5 KM |
606 | oip->i_size = length; |
607 | size = blksize(fs, oip, lbn); | |
ec67a3ce | 608 | count = howmany(size, CLBYTES); |
7188ac27 | 609 | munhash(oip->i_devvp, bn + i * CLBYTES / DEV_BSIZE); |
a937f856 | 610 | error = bread(oip->i_devvp, bn, size, NOCRED, &bp); |
7188ac27 | 611 | if (error) { |
28821bc5 KM |
612 | oip->i_size = osize; |
613 | brelse(bp); | |
7188ac27 | 614 | return (error); |
28821bc5 | 615 | } |
a5e62f37 | 616 | bzero(bp->b_un.b_addr + offset, (unsigned)(size - offset)); |
28821bc5 KM |
617 | bdwrite(bp); |
618 | } | |
619 | /* | |
620 | * Update file and block pointers | |
9c03b2c0 SL |
621 | * on disk before we start freeing blocks. |
622 | * If we crash before free'ing blocks below, | |
623 | * the blocks will be returned to the free list. | |
624 | * lastiblock values are also normalized to -1 | |
625 | * for calls to indirtrunc below. | |
6459ebe0 | 626 | */ |
9c03b2c0 | 627 | tip = *oip; |
28821bc5 | 628 | tip.i_size = osize; |
9c03b2c0 SL |
629 | for (level = TRIPLE; level >= SINGLE; level--) |
630 | if (lastiblock[level] < 0) { | |
631 | oip->i_ib[level] = 0; | |
632 | lastiblock[level] = -1; | |
4f083fd7 | 633 | } |
9c03b2c0 SL |
634 | for (i = NDADDR - 1; i > lastblock; i--) |
635 | oip->i_db[i] = 0; | |
9c03b2c0 | 636 | oip->i_flag |= ICHG|IUPD; |
7188ac27 | 637 | allerror = syncip(oip); |
9c03b2c0 | 638 | |
6459ebe0 | 639 | /* |
9c03b2c0 | 640 | * Indirect blocks first. |
6459ebe0 | 641 | */ |
28821bc5 | 642 | ip = &tip; |
9c03b2c0 SL |
643 | for (level = TRIPLE; level >= SINGLE; level--) { |
644 | bn = ip->i_ib[level]; | |
4f083fd7 | 645 | if (bn != 0) { |
7188ac27 KM |
646 | error = indirtrunc(ip, bn, lastiblock[level], level, |
647 | &count); | |
648 | if (error) | |
649 | allerror = error; | |
650 | blocksreleased += count; | |
9c03b2c0 SL |
651 | if (lastiblock[level] < 0) { |
652 | ip->i_ib[level] = 0; | |
ced3a252 | 653 | blkfree(ip, bn, (off_t)fs->fs_bsize); |
9c03b2c0 | 654 | blocksreleased += nblocks; |
9c03b2c0 SL |
655 | } |
656 | } | |
657 | if (lastiblock[level] >= 0) | |
658 | goto done; | |
4f083fd7 | 659 | } |
9c03b2c0 | 660 | |
6459ebe0 | 661 | /* |
9c03b2c0 | 662 | * All whole direct blocks or frags. |
6459ebe0 | 663 | */ |
4f083fd7 | 664 | for (i = NDADDR - 1; i > lastblock; i--) { |
8011f5df | 665 | register off_t bsize; |
4f083fd7 | 666 | |
6459ebe0 | 667 | bn = ip->i_db[i]; |
4f083fd7 | 668 | if (bn == 0) |
5d5124a1 | 669 | continue; |
4f083fd7 | 670 | ip->i_db[i] = 0; |
0b355a6e | 671 | bsize = (off_t)blksize(fs, ip, i); |
ced3a252 | 672 | blkfree(ip, bn, bsize); |
0b355a6e | 673 | blocksreleased += btodb(bsize); |
4f083fd7 | 674 | } |
9c03b2c0 SL |
675 | if (lastblock < 0) |
676 | goto done; | |
677 | ||
4f083fd7 SL |
678 | /* |
679 | * Finally, look for a change in size of the | |
680 | * last direct block; release any frags. | |
681 | */ | |
9c03b2c0 SL |
682 | bn = ip->i_db[lastblock]; |
683 | if (bn != 0) { | |
8011f5df | 684 | off_t oldspace, newspace; |
9c03b2c0 | 685 | |
4f083fd7 SL |
686 | /* |
687 | * Calculate amount of space we're giving | |
688 | * back as old block size minus new block size. | |
689 | */ | |
9c03b2c0 | 690 | oldspace = blksize(fs, ip, lastblock); |
4f083fd7 | 691 | ip->i_size = length; |
9c03b2c0 SL |
692 | newspace = blksize(fs, ip, lastblock); |
693 | if (newspace == 0) | |
694 | panic("itrunc: newspace"); | |
695 | if (oldspace - newspace > 0) { | |
4f083fd7 SL |
696 | /* |
697 | * Block number of space to be free'd is | |
698 | * the old block # plus the number of frags | |
699 | * required for the storage we're keeping. | |
700 | */ | |
9c03b2c0 | 701 | bn += numfrags(fs, newspace); |
ced3a252 | 702 | blkfree(ip, bn, oldspace - newspace); |
08d9a8ec | 703 | blocksreleased += btodb(oldspace - newspace); |
4f083fd7 | 704 | } |
5d5124a1 | 705 | } |
4f083fd7 | 706 | done: |
9c03b2c0 SL |
707 | /* BEGIN PARANOIA */ |
708 | for (level = SINGLE; level <= TRIPLE; level++) | |
709 | if (ip->i_ib[level] != oip->i_ib[level]) | |
710 | panic("itrunc1"); | |
711 | for (i = 0; i < NDADDR; i++) | |
712 | if (ip->i_db[i] != oip->i_db[i]) | |
713 | panic("itrunc2"); | |
714 | /* END PARANOIA */ | |
08d9a8ec SL |
715 | oip->i_blocks -= blocksreleased; |
716 | if (oip->i_blocks < 0) /* sanity */ | |
717 | oip->i_blocks = 0; | |
718 | oip->i_flag |= ICHG; | |
b4567e9c | 719 | #ifdef QUOTA |
08d9a8ec | 720 | (void) chkdq(oip, -blocksreleased, 0); |
89045c38 | 721 | #endif |
7188ac27 | 722 | return (allerror); |
5d5124a1 BJ |
723 | } |
724 | ||
4f083fd7 SL |
725 | /* |
726 | * Release blocks associated with the inode ip and | |
727 | * stored in the indirect block bn. Blocks are free'd | |
728 | * in LIFO order up to (but not including) lastbn. If | |
9c03b2c0 SL |
729 | * level is greater than SINGLE, the block is an indirect |
730 | * block and recursive calls to indirtrunc must be used to | |
731 | * cleanse other indirect blocks. | |
732 | * | |
733 | * NB: triple indirect blocks are untested. | |
4f083fd7 | 734 | */ |
7188ac27 | 735 | indirtrunc(ip, bn, lastbn, level, countp) |
6459ebe0 | 736 | register struct inode *ip; |
4f083fd7 | 737 | daddr_t bn, lastbn; |
9c03b2c0 | 738 | int level; |
7188ac27 | 739 | long *countp; |
5d5124a1 | 740 | { |
4f083fd7 | 741 | register int i; |
b30358ab | 742 | struct buf *bp; |
9c03b2c0 | 743 | register struct fs *fs = ip->i_fs; |
b30358ab KM |
744 | register daddr_t *bap; |
745 | daddr_t *copy, nb, last; | |
7188ac27 KM |
746 | long blkcount, factor; |
747 | int nblocks, blocksreleased = 0; | |
748 | int error, allerror = 0; | |
5d5124a1 | 749 | |
9c03b2c0 SL |
750 | /* |
751 | * Calculate index in current block of last | |
752 | * block to be kept. -1 indicates the entire | |
753 | * block so we need not calculate the index. | |
754 | */ | |
755 | factor = 1; | |
756 | for (i = SINGLE; i < level; i++) | |
757 | factor *= NINDIR(fs); | |
4f083fd7 | 758 | last = lastbn; |
9c03b2c0 SL |
759 | if (lastbn > 0) |
760 | last /= factor; | |
08d9a8ec | 761 | nblocks = btodb(fs->fs_bsize); |
9c03b2c0 SL |
762 | /* |
763 | * Get buffer of block pointers, zero those | |
764 | * entries corresponding to blocks to be free'd, | |
765 | * and update on disk copy first. | |
766 | */ | |
ec67a3ce MK |
767 | #ifdef SECSIZE |
768 | bp = bread(ip->i_dev, fsbtodb(fs, bn), (int)fs->fs_bsize, | |
769 | fs->fs_dbsize); | |
770 | #else SECSIZE | |
a937f856 KM |
771 | error = bread(ip->i_devvp, fsbtodb(fs, bn), (int)fs->fs_bsize, |
772 | NOCRED, &bp); | |
7188ac27 | 773 | if (error) { |
9c03b2c0 | 774 | brelse(bp); |
7188ac27 KM |
775 | *countp = 0; |
776 | return (error); | |
9c03b2c0 SL |
777 | } |
778 | bap = bp->b_un.b_daddr; | |
b30358ab KM |
779 | MALLOC(copy, daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK); |
780 | bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize); | |
9c03b2c0 SL |
781 | bzero((caddr_t)&bap[last + 1], |
782 | (u_int)(NINDIR(fs) - (last + 1)) * sizeof (daddr_t)); | |
7188ac27 KM |
783 | error = bwrite(bp); |
784 | if (error) | |
785 | allerror = error; | |
b30358ab | 786 | bap = copy; |
4f083fd7 | 787 | |
9c03b2c0 SL |
788 | /* |
789 | * Recursively free totally unused blocks. | |
790 | */ | |
791 | for (i = NINDIR(fs) - 1; i > last; i--) { | |
5d5124a1 | 792 | nb = bap[i]; |
4f083fd7 | 793 | if (nb == 0) |
5d5124a1 | 794 | continue; |
7188ac27 KM |
795 | if (level > SINGLE) { |
796 | error = indirtrunc(ip, nb, (daddr_t)-1, level - 1, | |
797 | &blkcount); | |
798 | if (error) | |
799 | allerror = error; | |
800 | blocksreleased += blkcount; | |
801 | } | |
ced3a252 | 802 | blkfree(ip, nb, (off_t)fs->fs_bsize); |
4f083fd7 | 803 | blocksreleased += nblocks; |
4f083fd7 | 804 | } |
9c03b2c0 SL |
805 | |
806 | /* | |
807 | * Recursively free last partial block. | |
808 | */ | |
809 | if (level > SINGLE && lastbn >= 0) { | |
810 | last = lastbn % factor; | |
4f083fd7 | 811 | nb = bap[i]; |
7188ac27 KM |
812 | if (nb != 0) { |
813 | error = indirtrunc(ip, nb, last, level - 1, &blkcount); | |
814 | if (error) | |
815 | allerror = error; | |
816 | blocksreleased += blkcount; | |
817 | } | |
5d5124a1 | 818 | } |
b30358ab | 819 | FREE(copy, M_TEMP); |
7188ac27 KM |
820 | *countp = blocksreleased; |
821 | return (allerror); | |
5d5124a1 BJ |
822 | } |
823 | ||
3ebac878 | 824 | /* |
ec67a3ce | 825 | * Remove any inodes in the inode cache belonging to dev. |
3ebac878 RE |
826 | * |
827 | * There should not be any active ones, return error if any are found | |
ec67a3ce | 828 | * (nb: this is a user error, not a system err). |
3ebac878 | 829 | */ |
0d73cd0c KM |
830 | int busyprt = 0; /* patch to print out busy inodes */ |
831 | ||
b4567e9c | 832 | #ifdef QUOTA |
4147b3f6 | 833 | iflush(dev, iq) |
89045c38 | 834 | dev_t dev; |
4147b3f6 | 835 | struct inode *iq; |
89045c38 | 836 | #else |
3ebac878 RE |
837 | iflush(dev) |
838 | dev_t dev; | |
89045c38 | 839 | #endif |
3ebac878 | 840 | { |
32dc2b7e | 841 | register struct inode *ip; |
0d73cd0c | 842 | int busy = 0; |
3ebac878 RE |
843 | |
844 | for (ip = inode; ip < inodeNINODE; ip++) { | |
b4567e9c | 845 | #ifdef QUOTA |
0d73cd0c KM |
846 | if (ip == iq || ip->i_dev != dev) |
847 | continue; | |
89045c38 | 848 | #else |
0d73cd0c KM |
849 | if (ip->i_dev != dev) |
850 | continue; | |
89045c38 | 851 | #endif |
0d73cd0c KM |
852 | if (ITOV(ip)->v_count) { |
853 | busy++; | |
854 | if (!busyprt) | |
855 | continue; | |
856 | printf("%s %d on dev 0x%x count %d type %d\n", | |
857 | "iflush: busy inode ", ip->i_number, ip->i_dev, | |
858 | ITOV(ip)->v_count, ITOV(ip)->v_type); | |
859 | continue; | |
860 | } | |
861 | remque(ip); | |
862 | ip->i_forw = ip; | |
863 | ip->i_back = ip; | |
864 | /* | |
865 | * As v_count == 0, the inode was on the free list already, | |
866 | * just leave it there, it will fall off the bottom eventually. | |
867 | * We could perhaps move it to the head of the free list, | |
868 | * but as umounts are done so infrequently, we would gain | |
869 | * very little, while making the code bigger. | |
870 | */ | |
b4567e9c | 871 | #ifdef QUOTA |
0d73cd0c KM |
872 | dqrele(ip->i_dquot); |
873 | ip->i_dquot = NODQUOT; | |
89045c38 | 874 | #endif |
0d73cd0c KM |
875 | if (ip->i_devvp) { |
876 | vrele(ip->i_devvp); | |
877 | ip->i_devvp = 0; | |
878 | } | |
3ebac878 | 879 | } |
0d73cd0c KM |
880 | if (busy) |
881 | return (EBUSY); | |
ec67a3ce | 882 | return (0); |
3ebac878 RE |
883 | } |
884 | ||
d6a210b8 | 885 | /* |
7494ef16 | 886 | * Lock an inode. If its already locked, set the WANT bit and sleep. |
d6a210b8 | 887 | */ |
7494ef16 BJ |
888 | ilock(ip) |
889 | register struct inode *ip; | |
d6a210b8 BJ |
890 | { |
891 | ||
7188ac27 KM |
892 | while (ip->i_flag & ILOCKED) { |
893 | ip->i_flag |= IWANT; | |
894 | (void) sleep((caddr_t)ip, PINOD); | |
895 | } | |
896 | ip->i_flag |= ILOCKED; | |
d6a210b8 BJ |
897 | } |
898 | ||
899 | /* | |
7494ef16 | 900 | * Unlock an inode. If WANT bit is on, wakeup. |
d6a210b8 | 901 | */ |
ff56f48a | 902 | iunlock(ip) |
7494ef16 | 903 | register struct inode *ip; |
d6a210b8 BJ |
904 | { |
905 | ||
7188ac27 KM |
906 | if ((ip->i_flag & ILOCKED) == 0) |
907 | printf("unlocking unlocked inode %d on dev 0x%x\n", | |
908 | ip->i_number, ip->i_dev); | |
909 | ip->i_flag &= ~ILOCKED; | |
910 | if (ip->i_flag&IWANT) { | |
911 | ip->i_flag &= ~IWANT; | |
912 | wakeup((caddr_t)ip); | |
913 | } | |
914 | } | |
915 | ||
916 | /* | |
917 | * Check mode permission on inode pointer. Mode is READ, WRITE or EXEC. | |
918 | * The mode is shifted to select the owner/group/other fields. The | |
919 | * super user is granted all permissions. | |
920 | * | |
921 | * NB: Called from vnode op table. It seems this could all be done | |
922 | * using vattr's but... | |
923 | */ | |
924 | iaccess(ip, mode, cred) | |
925 | register struct inode *ip; | |
926 | register int mode; | |
927 | struct ucred *cred; | |
928 | { | |
929 | register gid_t *gp; | |
930 | register struct vnode *vp = ITOV(ip); | |
931 | int i; | |
932 | ||
933 | /* | |
934 | * If you're the super-user, | |
935 | * you always get access. | |
936 | */ | |
937 | if (cred->cr_uid == 0) | |
938 | return (0); | |
939 | /* | |
940 | * Access check is based on only one of owner, group, public. | |
941 | * If not owner, then check group. If not a member of the | |
942 | * group, then check public access. | |
943 | */ | |
944 | if (cred->cr_uid != ip->i_uid) { | |
945 | mode >>= 3; | |
946 | gp = cred->cr_groups; | |
947 | for (i = 0; i < cred->cr_ngroups; i++, gp++) | |
948 | if (ip->i_gid == *gp) | |
949 | goto found; | |
950 | mode >>= 3; | |
951 | found: | |
952 | ; | |
953 | } | |
954 | if ((ip->i_mode & mode) != 0) | |
955 | return (0); | |
956 | return (EACCES); | |
d6a210b8 | 957 | } |