lfs_alloc.c revision 1.76 1 /* $NetBSD: lfs_alloc.c,v 1.76 2005/03/08 00:18:19 perseant Exp $ */
2
3 /*-
4 * Copyright (c) 1999, 2000, 2001, 2002, 2003 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Konrad E. Schroder <perseant (at) hhhh.org>.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the NetBSD
21 * Foundation, Inc. and its contributors.
22 * 4. Neither the name of The NetBSD Foundation nor the names of its
23 * contributors may be used to endorse or promote products derived
24 * from this software without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
37 */
38 /*
39 * Copyright (c) 1991, 1993
40 * The Regents of the University of California. All rights reserved.
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)lfs_alloc.c 8.4 (Berkeley) 1/4/94
67 */
68
69 #include <sys/cdefs.h>
70 __KERNEL_RCSID(0, "$NetBSD: lfs_alloc.c,v 1.76 2005/03/08 00:18:19 perseant Exp $");
71
72 #if defined(_KERNEL_OPT)
73 #include "opt_quota.h"
74 #endif
75
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/kernel.h>
79 #include <sys/buf.h>
80 #include <sys/lock.h>
81 #include <sys/vnode.h>
82 #include <sys/syslog.h>
83 #include <sys/mount.h>
84 #include <sys/pool.h>
85 #include <sys/proc.h>
86
87 #include <ufs/ufs/quota.h>
88 #include <ufs/ufs/inode.h>
89 #include <ufs/ufs/ufsmount.h>
90 #include <ufs/ufs/ufs_extern.h>
91
92 #include <ufs/lfs/lfs.h>
93 #include <ufs/lfs/lfs_extern.h>
94
95 extern struct lock ufs_hashlock;
96
97 static int extend_ifile(struct lfs *, struct ucred *);
98 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int,
99 struct vnode **);
100
101 /*
102 * Allocate a particular inode with a particular version number, freeing
103 * any previous versions of this inode that may have gone before.
104 * Used by the roll-forward code.
105 *
106 * XXX this function does not have appropriate locking to be used on a live fs;
107 * XXX but something similar could probably be used for an "undelete" call.
108 *
109 * Called with the Ifile inode locked.
110 */
111 int
112 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p,
113 struct vnode **vpp)
114 {
115 IFILE *ifp;
116 struct buf *bp, *cbp;
117 struct vnode *vp;
118 struct inode *ip;
119 ino_t tino, oldnext;
120 int error;
121 CLEANERINFO *cip;
122
123 /*
124 * First, just try a vget. If the version number is the one we want,
125 * we don't have to do anything else. If the version number is wrong,
126 * take appropriate action.
127 */
128 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp);
129 if (error == 0) {
130 DLOG((DLOG_RF, "lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp));
131
132 *vpp = vp;
133 ip = VTOI(vp);
134 if (ip->i_gen == version)
135 return 0;
136 else if (ip->i_gen < version) {
137 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p);
138 ip->i_gen = ip->i_ffs1_gen = version;
139 LFS_SET_UINO(ip, IN_CHANGE | IN_UPDATE);
140 return 0;
141 } else {
142 DLOG((DLOG_RF, "ino %d: sought version %d, got %d\n",
143 ino, version, ip->i_ffs1_gen));
144 vput(vp);
145 *vpp = NULLVP;
146 return EEXIST;
147 }
148 }
149
150 /*
151 * The inode is not in use. Find it on the free list.
152 */
153 /* If the Ifile is too short to contain this inum, extend it */
154 while (VTOI(fs->lfs_ivnode)->i_size <= (ino /
155 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz)
156 << fs->lfs_bshift) {
157 extend_ifile(fs, NOCRED);
158 }
159
160 LFS_IENTRY(ifp, fs, ino, bp);
161 oldnext = ifp->if_nextfree;
162 ifp->if_version = version;
163 brelse(bp);
164
165 LFS_GET_HEADFREE(fs, cip, cbp, &ino);
166 if (ino) {
167 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext);
168 } else {
169 tino = ino;
170 while (1) {
171 LFS_IENTRY(ifp, fs, tino, bp);
172 if (ifp->if_nextfree == ino ||
173 ifp->if_nextfree == LFS_UNUSED_INUM)
174 break;
175 tino = ifp->if_nextfree;
176 brelse(bp);
177 }
178 if (ifp->if_nextfree == LFS_UNUSED_INUM) {
179 brelse(bp);
180 return ENOENT;
181 }
182 ifp->if_nextfree = oldnext;
183 LFS_BWRITE_LOG(bp);
184 }
185
186 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp);
187 if (error == 0) {
188 /*
189 * Make it VREG so we can put blocks on it. We will change
190 * this later if it turns out to be some other kind of file.
191 */
192 ip = VTOI(vp);
193 ip->i_mode = ip->i_ffs1_mode = IFREG;
194 ip->i_nlink = ip->i_ffs1_nlink = 1;
195 ip->i_ffs_effnlink = 1;
196 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp);
197 ip = VTOI(vp);
198
199 DLOG((DLOG_RF, "lfs_rf_valloc: ino %d vp %p\n", ino, vp));
200
201 /* The dirop-nature of this vnode is past */
202 lfs_unmark_vnode(vp);
203 (void)lfs_vunref(vp);
204 --lfs_dirvcount;
205 vp->v_flag &= ~VDIROP;
206 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
207 }
208 *vpp = vp;
209 return error;
210 }
211
212 /*
213 * Add a new block to the Ifile, to accommodate future file creations.
214 * Called with the segment lock held.
215 */
216 static int
217 extend_ifile(struct lfs *fs, struct ucred *cred)
218 {
219 struct vnode *vp;
220 struct inode *ip;
221 IFILE *ifp;
222 IFILE_V1 *ifp_v1;
223 struct buf *bp, *cbp;
224 int error;
225 daddr_t i, blkno, max;
226 ino_t oldlast;
227 CLEANERINFO *cip;
228
229 vp = fs->lfs_ivnode;
230 ip = VTOI(vp);
231 blkno = lblkno(fs, ip->i_size);
232 if ((error = VOP_BALLOC(vp, ip->i_size, fs->lfs_bsize, cred, 0,
233 &bp)) != 0) {
234 return (error);
235 }
236 ip->i_size += fs->lfs_bsize;
237 ip->i_ffs1_size = ip->i_size;
238 uvm_vnp_setsize(vp, ip->i_size);
239
240 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) *
241 fs->lfs_ifpb;
242 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast);
243 LFS_PUT_HEADFREE(fs, cip, cbp, i);
244 #ifdef DIAGNOSTIC
245 if (fs->lfs_freehd == LFS_UNUSED_INUM)
246 panic("inode 0 allocated [2]");
247 #endif /* DIAGNOSTIC */
248 max = i + fs->lfs_ifpb;
249
250 if (fs->lfs_version == 1) {
251 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) {
252 ifp_v1->if_version = 1;
253 ifp_v1->if_daddr = LFS_UNUSED_DADDR;
254 ifp_v1->if_nextfree = ++i;
255 }
256 ifp_v1--;
257 ifp_v1->if_nextfree = oldlast;
258 } else {
259 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) {
260 ifp->if_version = 1;
261 ifp->if_daddr = LFS_UNUSED_DADDR;
262 ifp->if_nextfree = ++i;
263 }
264 ifp--;
265 ifp->if_nextfree = oldlast;
266 }
267 LFS_PUT_TAILFREE(fs, cip, cbp, max - 1);
268
269 (void) LFS_BWRITE_LOG(bp); /* Ifile */
270
271 return 0;
272 }
273
274 /* Allocate a new inode. */
275 /* ARGSUSED */
276 /* VOP_BWRITE 2i times */
277 int
278 lfs_valloc(void *v)
279 {
280 struct vop_valloc_args /* {
281 struct vnode *a_pvp;
282 int a_mode;
283 struct ucred *a_cred;
284 struct vnode **a_vpp;
285 } */ *ap = v;
286 struct lfs *fs;
287 struct buf *bp, *cbp;
288 struct ifile *ifp;
289 ino_t new_ino;
290 int error;
291 int new_gen;
292 CLEANERINFO *cip;
293
294 fs = VTOI(ap->a_pvp)->i_lfs;
295 if (fs->lfs_ronly)
296 return EROFS;
297 *ap->a_vpp = NULL;
298
299 lfs_seglock(fs, SEGM_PROT);
300
301 /* Get the head of the freelist. */
302 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino);
303
304 #ifdef DIAGNOSTIC
305 if (new_ino == LFS_UNUSED_INUM) {
306 #ifdef DEBUG
307 lfs_dump_super(fs);
308 #endif /* DEBUG */
309 panic("inode 0 allocated [1]");
310 }
311 #endif /* DIAGNOSTIC */
312 DLOG((DLOG_ALLOC, "lfs_valloc: allocate inode %d\n", new_ino));
313
314 /*
315 * Remove the inode from the free list and write the new start
316 * of the free list into the superblock.
317 */
318 LFS_IENTRY(ifp, fs, new_ino, bp);
319 if (ifp->if_daddr != LFS_UNUSED_DADDR)
320 panic("lfs_valloc: inuse inode %d on the free list", new_ino);
321 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree);
322 DLOG((DLOG_ALLOC, "lfs_valloc: headfree %d -> %d\n", new_ino,
323 ifp->if_nextfree));
324
325 new_gen = ifp->if_version; /* version was updated by vfree */
326 brelse(bp);
327
328 /* Extend IFILE so that the next lfs_valloc will succeed. */
329 if (fs->lfs_freehd == LFS_UNUSED_INUM) {
330 if ((error = extend_ifile(fs, ap->a_cred)) != 0) {
331 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino);
332 lfs_segunlock(fs);
333 return error;
334 }
335 }
336 #ifdef DIAGNOSTIC
337 if (fs->lfs_freehd == LFS_UNUSED_INUM)
338 panic("inode 0 allocated [3]");
339 #endif /* DIAGNOSTIC */
340
341 lfs_segunlock(fs);
342
343 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp);
344 }
345
346 /*
347 * Finish allocating a new inode, given an inode and generation number.
348 */
349 static int
350 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen,
351 struct vnode **vpp)
352 {
353 struct inode *ip;
354 struct vnode *vp;
355 IFILE *ifp;
356 struct buf *bp, *cbp;
357 int error;
358 CLEANERINFO *cip;
359
360 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp);
361 DLOG((DLOG_ALLOC, "lfs_ialloc: ino %d vp %p error %d\n", new_ino,
362 vp, error));
363 if (error)
364 goto errout;
365
366 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0);
367 /* Create an inode to associate with the vnode. */
368 lfs_vcreate(pvp->v_mount, new_ino, vp);
369
370 ip = VTOI(vp);
371 LFS_SET_UINO(ip, IN_CHANGE);
372 /* on-disk structure has been zeroed out by lfs_vcreate */
373 ip->i_din.ffs1_din->di_inumber = new_ino;
374
375 /* Set a new generation number for this inode. */
376 if (new_gen) {
377 ip->i_gen = new_gen;
378 ip->i_ffs1_gen = new_gen;
379 }
380
381 /* Insert into the inode hash table. */
382 ufs_ihashins(ip);
383 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
384
385 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp);
386 ip = VTOI(vp);
387
388 memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize));
389
390 uvm_vnp_setsize(vp, 0);
391 *vpp = vp;
392 lfs_mark_vnode(vp);
393 genfs_node_init(vp, &lfs_genfsops);
394 VREF(ip->i_devvp);
395 /* Set superblock modified bit and increment file count. */
396 fs->lfs_fmod = 1;
397 ++fs->lfs_nfiles;
398 return (0);
399
400 errout:
401 /*
402 * Put the new inum back on the free list.
403 */
404 lfs_seglock(fs, SEGM_PROT);
405 LFS_IENTRY(ifp, fs, new_ino, bp);
406 ifp->if_daddr = LFS_UNUSED_DADDR;
407 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree));
408 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino);
409 (void) LFS_BWRITE_LOG(bp); /* Ifile */
410 lfs_segunlock(fs);
411
412 *vpp = NULLVP;
413 return (error);
414 }
415
416 /* Create a new vnode/inode pair and initialize what fields we can. */
417 void
418 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp)
419 {
420 struct inode *ip;
421 struct ufs1_dinode *dp;
422 struct ufsmount *ump;
423 int i;
424
425 /* Get a pointer to the private mount structure. */
426 ump = VFSTOUFS(mp);
427
428 /* Initialize the inode. */
429 ip = pool_get(&lfs_inode_pool, PR_WAITOK);
430 memset(ip, 0, sizeof(*ip));
431 dp = pool_get(&lfs_dinode_pool, PR_WAITOK);
432 memset(dp, 0, sizeof(*dp));
433 ip->inode_ext.lfs = pool_get(&lfs_inoext_pool, PR_WAITOK);
434 memset(ip->inode_ext.lfs, 0, sizeof(*ip->inode_ext.lfs));
435 for (i = 0; i < LFS_BLIST_HASH_WIDTH; i++)
436 LIST_INIT(&(ip->i_lfs_blist[i]));
437 vp->v_data = ip;
438 ip->i_din.ffs1_din = dp;
439 ip->i_ump = ump;
440 ip->i_vnode = vp;
441 ip->i_devvp = ump->um_devvp;
442 ip->i_dev = ump->um_dev;
443 ip->i_number = dp->di_inumber = ino;
444 ip->i_lfs = ump->um_lfs;
445 ip->i_lfs_effnblks = 0;
446 #ifdef QUOTA
447 for (i = 0; i < MAXQUOTAS; i++)
448 ip->i_dquot[i] = NODQUOT;
449 #endif
450 #ifdef DEBUG
451 if (ino == LFS_IFILE_INUM)
452 vp->v_vnlock->lk_wmesg = "inlock";
453 #endif
454 }
455
456 /* Free an inode. */
457 /* ARGUSED */
458 /* VOP_BWRITE 2i times */
459 int
460 lfs_vfree(void *v)
461 {
462 struct vop_vfree_args /* {
463 struct vnode *a_pvp;
464 ino_t a_ino;
465 int a_mode;
466 } */ *ap = v;
467 SEGUSE *sup;
468 CLEANERINFO *cip;
469 struct buf *cbp, *bp;
470 struct ifile *ifp;
471 struct inode *ip;
472 struct vnode *vp;
473 struct lfs *fs;
474 daddr_t old_iaddr;
475 ino_t ino, otail;
476 int s;
477
478 /* Get the inode number and file system. */
479 vp = ap->a_pvp;
480 ip = VTOI(vp);
481 fs = ip->i_lfs;
482 ino = ip->i_number;
483
484 /* Drain of pending writes */
485 s = splbio();
486 if (fs->lfs_version > 1 && WRITEINPROG(vp))
487 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0);
488 splx(s);
489
490 lfs_seglock(fs, SEGM_PROT);
491
492 lfs_unmark_vnode(vp);
493 if (vp->v_flag & VDIROP) {
494 --lfs_dirvcount;
495 vp->v_flag &= ~VDIROP;
496 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
497 wakeup(&lfs_dirvcount);
498 lfs_vunref(vp);
499 }
500
501 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED);
502 ip->i_flag &= ~IN_ALLMOD;
503
504 /*
505 * Set the ifile's inode entry to unused, increment its version number
506 * and link it onto the free chain.
507 */
508 LFS_IENTRY(ifp, fs, ino, bp);
509 old_iaddr = ifp->if_daddr;
510 ifp->if_daddr = LFS_UNUSED_DADDR;
511 ++ifp->if_version;
512 if (fs->lfs_version == 1) {
513 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree));
514 LFS_PUT_HEADFREE(fs, cip, cbp, ino);
515 (void) LFS_BWRITE_LOG(bp); /* Ifile */
516 } else {
517 ifp->if_nextfree = LFS_UNUSED_INUM;
518 /*
519 * XXX Writing the freed node here means that it might not
520 * XXX make it into the free list in the event of a crash
521 * XXX (the ifile could be written before the rest of this
522 * XXX completes).
523 */
524 (void) LFS_BWRITE_LOG(bp); /* Ifile */
525 LFS_GET_TAILFREE(fs, cip, cbp, &otail);
526 LFS_IENTRY(ifp, fs, otail, bp);
527 ifp->if_nextfree = ino;
528 LFS_BWRITE_LOG(bp);
529 LFS_PUT_TAILFREE(fs, cip, cbp, ino);
530 DLOG((DLOG_ALLOC, "lfs_vfree: tailfree %d -> %d\n", otail,
531 ino));
532 }
533 #ifdef DIAGNOSTIC
534 if (ino == LFS_UNUSED_INUM) {
535 panic("inode 0 freed");
536 }
537 #endif /* DIAGNOSTIC */
538 if (old_iaddr != LFS_UNUSED_DADDR) {
539 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp);
540 #ifdef DIAGNOSTIC
541 if (sup->su_nbytes < sizeof (struct ufs1_dinode)) {
542 printf("lfs_vfree: negative byte count"
543 " (segment %" PRIu32 " short by %d)\n",
544 dtosn(fs, old_iaddr),
545 (int)sizeof (struct ufs1_dinode) -
546 sup->su_nbytes);
547 panic("lfs_vfree: negative byte count");
548 sup->su_nbytes = sizeof (struct ufs1_dinode);
549 }
550 #endif
551 sup->su_nbytes -= sizeof (struct ufs1_dinode);
552 LFS_WRITESEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); /* Ifile */
553 }
554
555 /* Set superblock modified bit and decrement file count. */
556 fs->lfs_fmod = 1;
557 --fs->lfs_nfiles;
558
559 lfs_segunlock(fs);
560
561 return (0);
562 }
563