lfs_alloc.c revision 1.70 1 /* $NetBSD: lfs_alloc.c,v 1.70 2003/07/12 16:17:06 yamt 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. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
57 *
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * SUCH DAMAGE.
69 *
70 * @(#)lfs_alloc.c 8.4 (Berkeley) 1/4/94
71 */
72
73 #include <sys/cdefs.h>
74 __KERNEL_RCSID(0, "$NetBSD: lfs_alloc.c,v 1.70 2003/07/12 16:17:06 yamt Exp $");
75
76 #if defined(_KERNEL_OPT)
77 #include "opt_quota.h"
78 #endif
79
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
83 #include <sys/buf.h>
84 #include <sys/lock.h>
85 #include <sys/vnode.h>
86 #include <sys/syslog.h>
87 #include <sys/mount.h>
88 #include <sys/pool.h>
89 #include <sys/proc.h>
90
91 #include <ufs/ufs/quota.h>
92 #include <ufs/ufs/inode.h>
93 #include <ufs/ufs/ufsmount.h>
94 #include <ufs/ufs/ufs_extern.h>
95
96 #include <ufs/lfs/lfs.h>
97 #include <ufs/lfs/lfs_extern.h>
98
99 extern struct lock ufs_hashlock;
100
101 static int extend_ifile(struct lfs *, struct ucred *);
102 static int lfs_ialloc(struct lfs *, struct vnode *, ino_t, int,
103 struct vnode **);
104
105 /*
106 * Allocate a particular inode with a particular version number, freeing
107 * any previous versions of this inode that may have gone before.
108 * Used by the roll-forward code.
109 *
110 * XXX this function does not have appropriate locking to be used on a live fs;
111 * XXX but something similar could probably be used for an "undelete" call.
112 *
113 * Called with the Ifile inode locked.
114 */
115 int
116 lfs_rf_valloc(struct lfs *fs, ino_t ino, int version, struct proc *p,
117 struct vnode **vpp)
118 {
119 IFILE *ifp;
120 struct buf *bp, *cbp;
121 struct vnode *vp;
122 struct inode *ip;
123 ino_t tino, oldnext;
124 int error;
125 CLEANERINFO *cip;
126
127 /*
128 * First, just try a vget. If the version number is the one we want,
129 * we don't have to do anything else. If the version number is wrong,
130 * take appropriate action.
131 */
132 error = VFS_VGET(fs->lfs_ivnode->v_mount, ino, &vp);
133 if (error == 0) {
134 /* printf("lfs_rf_valloc[1]: ino %d vp %p\n", ino, vp); */
135
136 *vpp = vp;
137 ip = VTOI(vp);
138 if (ip->i_gen == version)
139 return 0;
140 else if (ip->i_gen < version) {
141 VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, p);
142 ip->i_gen = ip->i_ffs1_gen = version;
143 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED | IN_UPDATE);
144 return 0;
145 } else {
146 /* printf("ino %d: asked for version %d but got %d\n",
147 ino, version, ip->i_ffs1_gen); */
148 vput(vp);
149 *vpp = NULLVP;
150 return EEXIST;
151 }
152 }
153
154 /*
155 * The inode is not in use. Find it on the free list.
156 */
157 /* If the Ifile is too short to contain this inum, extend it */
158 while (VTOI(fs->lfs_ivnode)->i_size <= (ino /
159 fs->lfs_ifpb + fs->lfs_cleansz + fs->lfs_segtabsz)
160 << fs->lfs_bshift) {
161 extend_ifile(fs, NOCRED);
162 }
163
164 LFS_IENTRY(ifp, fs, ino, bp);
165 oldnext = ifp->if_nextfree;
166 ifp->if_version = version;
167 brelse(bp);
168
169 LFS_GET_HEADFREE(fs, cip, cbp, &ino);
170 if (ino) {
171 LFS_PUT_HEADFREE(fs, cip, cbp, oldnext);
172 } else {
173 tino = ino;
174 while (1) {
175 LFS_IENTRY(ifp, fs, tino, bp);
176 if (ifp->if_nextfree == ino ||
177 ifp->if_nextfree == LFS_UNUSED_INUM)
178 break;
179 tino = ifp->if_nextfree;
180 brelse(bp);
181 }
182 if (ifp->if_nextfree == LFS_UNUSED_INUM) {
183 brelse(bp);
184 return ENOENT;
185 }
186 ifp->if_nextfree = oldnext;
187 LFS_BWRITE_LOG(bp);
188 }
189
190 error = lfs_ialloc(fs, fs->lfs_ivnode, ino, version, &vp);
191 if (error == 0) {
192 /*
193 * Make it VREG so we can put blocks on it. We will change
194 * this later if it turns out to be some other kind of file.
195 */
196 ip = VTOI(vp);
197 ip->i_mode = ip->i_ffs1_mode = IFREG;
198 ip->i_nlink = ip->i_ffs1_nlink = 1;
199 ip->i_ffs_effnlink = 1;
200 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp);
201 ip = VTOI(vp);
202
203 /* printf("lfs_rf_valloc: ino %d vp %p\n", ino, vp); */
204
205 /* The dirop-nature of this vnode is past */
206 (void)lfs_vunref(vp);
207 --lfs_dirvcount;
208 vp->v_flag &= ~VDIROP;
209 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
210 --fs->lfs_nadirop;
211 ip->i_flag &= ~IN_ADIROP;
212 }
213 *vpp = vp;
214 return error;
215 }
216
217 /*
218 * Add a new block to the Ifile, to accommodate future file creations.
219 * Called with the segment lock held.
220 */
221 static int
222 extend_ifile(struct lfs *fs, struct ucred *cred)
223 {
224 struct vnode *vp;
225 struct inode *ip;
226 IFILE *ifp;
227 IFILE_V1 *ifp_v1;
228 struct buf *bp, *cbp;
229 int error;
230 daddr_t i, blkno, max;
231 ino_t oldlast;
232 CLEANERINFO *cip;
233
234 vp = fs->lfs_ivnode;
235 ip = VTOI(vp);
236 blkno = lblkno(fs, ip->i_size);
237 if ((error = VOP_BALLOC(vp, ip->i_size, fs->lfs_bsize, cred, 0,
238 &bp)) != 0) {
239 return (error);
240 }
241 ip->i_size += fs->lfs_bsize;
242 ip->i_ffs1_size = ip->i_size;
243 uvm_vnp_setsize(vp, ip->i_size);
244
245 i = (blkno - fs->lfs_segtabsz - fs->lfs_cleansz) *
246 fs->lfs_ifpb;
247 LFS_GET_HEADFREE(fs, cip, cbp, &oldlast);
248 LFS_PUT_HEADFREE(fs, cip, cbp, i);
249 #ifdef DIAGNOSTIC
250 if (fs->lfs_freehd == LFS_UNUSED_INUM)
251 panic("inode 0 allocated [2]");
252 #endif /* DIAGNOSTIC */
253 max = i + fs->lfs_ifpb;
254 /* printf("extend_ifile: new block ino %d--%d\n", i, max - 1); */
255
256 if (fs->lfs_version == 1) {
257 for (ifp_v1 = (IFILE_V1 *)bp->b_data; i < max; ++ifp_v1) {
258 ifp_v1->if_version = 1;
259 ifp_v1->if_daddr = LFS_UNUSED_DADDR;
260 ifp_v1->if_nextfree = ++i;
261 }
262 ifp_v1--;
263 ifp_v1->if_nextfree = oldlast;
264 } else {
265 for (ifp = (IFILE *)bp->b_data; i < max; ++ifp) {
266 ifp->if_version = 1;
267 ifp->if_daddr = LFS_UNUSED_DADDR;
268 ifp->if_nextfree = ++i;
269 }
270 ifp--;
271 ifp->if_nextfree = oldlast;
272 }
273 LFS_PUT_TAILFREE(fs, cip, cbp, max - 1);
274
275 (void) LFS_BWRITE_LOG(bp); /* Ifile */
276
277 return 0;
278 }
279
280 /* Allocate a new inode. */
281 /* ARGSUSED */
282 /* VOP_BWRITE 2i times */
283 int
284 lfs_valloc(void *v)
285 {
286 struct vop_valloc_args /* {
287 struct vnode *a_pvp;
288 int a_mode;
289 struct ucred *a_cred;
290 struct vnode **a_vpp;
291 } */ *ap = v;
292 struct lfs *fs;
293 struct buf *bp, *cbp;
294 struct ifile *ifp;
295 ino_t new_ino;
296 int error;
297 int new_gen;
298 CLEANERINFO *cip;
299
300 fs = VTOI(ap->a_pvp)->i_lfs;
301 if (fs->lfs_ronly)
302 return EROFS;
303 *ap->a_vpp = NULL;
304
305 lfs_seglock(fs, SEGM_PROT);
306
307 /* Get the head of the freelist. */
308 LFS_GET_HEADFREE(fs, cip, cbp, &new_ino);
309
310 #ifdef DIAGNOSTIC
311 if (new_ino == LFS_UNUSED_INUM) {
312 #ifdef DEBUG
313 lfs_dump_super(fs);
314 #endif /* DEBUG */
315 panic("inode 0 allocated [1]");
316 }
317 #endif /* DIAGNOSTIC */
318 #ifdef ALLOCPRINT
319 printf("lfs_valloc: allocate inode %d\n", new_ino);
320 #endif
321
322 /*
323 * Remove the inode from the free list and write the new start
324 * of the free list into the superblock.
325 */
326 LFS_IENTRY(ifp, fs, new_ino, bp);
327 if (ifp->if_daddr != LFS_UNUSED_DADDR)
328 panic("lfs_valloc: inuse inode %d on the free list", new_ino);
329 LFS_PUT_HEADFREE(fs, cip, cbp, ifp->if_nextfree);
330 /* printf("lfs_valloc: headfree %d -> %d\n", new_ino, ifp->if_nextfree); */
331
332 new_gen = ifp->if_version; /* version was updated by vfree */
333 brelse(bp);
334
335 /* Extend IFILE so that the next lfs_valloc will succeed. */
336 if (fs->lfs_freehd == LFS_UNUSED_INUM) {
337 if ((error = extend_ifile(fs, ap->a_cred)) != 0) {
338 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino);
339 lfs_segunlock(fs);
340 return error;
341 }
342 }
343 #ifdef DIAGNOSTIC
344 if (fs->lfs_freehd == LFS_UNUSED_INUM)
345 panic("inode 0 allocated [3]");
346 #endif /* DIAGNOSTIC */
347
348 lfs_segunlock(fs);
349
350 return lfs_ialloc(fs, ap->a_pvp, new_ino, new_gen, ap->a_vpp);
351 }
352
353 /*
354 * Finish allocating a new inode, given an inode and generation number.
355 */
356 static int
357 lfs_ialloc(struct lfs *fs, struct vnode *pvp, ino_t new_ino, int new_gen,
358 struct vnode **vpp)
359 {
360 struct inode *ip;
361 struct vnode *vp;
362 IFILE *ifp;
363 struct buf *bp, *cbp;
364 int error;
365 CLEANERINFO *cip;
366
367 error = getnewvnode(VT_LFS, pvp->v_mount, lfs_vnodeop_p, &vp);
368 /* printf("lfs_ialloc: ino %d vp %p error %d\n", new_ino, vp, error);*/
369 if (error)
370 goto errout;
371
372 lockmgr(&ufs_hashlock, LK_EXCLUSIVE, 0);
373 /* Create an inode to associate with the vnode. */
374 lfs_vcreate(pvp->v_mount, new_ino, vp);
375
376 ip = VTOI(vp);
377 LFS_SET_UINO(ip, IN_CHANGE | IN_MODIFIED);
378 /* on-disk structure has been zeroed out by lfs_vcreate */
379 ip->i_din.ffs1_din->di_inumber = new_ino;
380
381 /* Set a new generation number for this inode. */
382 if (new_gen) {
383 ip->i_gen = new_gen;
384 ip->i_ffs1_gen = new_gen;
385 }
386
387 /* Insert into the inode hash table. */
388 ufs_ihashins(ip);
389 lockmgr(&ufs_hashlock, LK_RELEASE, 0);
390
391 ufs_vinit(vp->v_mount, lfs_specop_p, lfs_fifoop_p, &vp);
392 ip = VTOI(vp);
393 /* printf("lfs_ialloc[2]: ino %d vp %p\n", new_ino, vp);*/
394
395 memset(ip->i_lfs_fragsize, 0, NDADDR * sizeof(*ip->i_lfs_fragsize));
396
397 uvm_vnp_setsize(vp, 0);
398 *vpp = vp;
399 if (!(vp->v_flag & VDIROP)) {
400 (void)lfs_vref(vp);
401 ++lfs_dirvcount;
402 TAILQ_INSERT_TAIL(&fs->lfs_dchainhd, ip, i_lfs_dchain);
403 }
404 vp->v_flag |= VDIROP;
405
406 if (!(ip->i_flag & IN_ADIROP))
407 ++fs->lfs_nadirop;
408 ip->i_flag |= IN_ADIROP;
409 genfs_node_init(vp, &lfs_genfsops);
410 VREF(ip->i_devvp);
411 /* Set superblock modified bit and increment file count. */
412 fs->lfs_fmod = 1;
413 ++fs->lfs_nfiles;
414 return (0);
415
416 errout:
417 /*
418 * Put the new inum back on the free list.
419 */
420 lfs_seglock(fs, SEGM_PROT);
421 LFS_IENTRY(ifp, fs, new_ino, bp);
422 ifp->if_daddr = LFS_UNUSED_DADDR;
423 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree));
424 LFS_PUT_HEADFREE(fs, cip, cbp, new_ino);
425 (void) LFS_BWRITE_LOG(bp); /* Ifile */
426 lfs_segunlock(fs);
427
428 *vpp = NULLVP;
429 return (error);
430 }
431
432 /* Create a new vnode/inode pair and initialize what fields we can. */
433 void
434 lfs_vcreate(struct mount *mp, ino_t ino, struct vnode *vp)
435 {
436 struct inode *ip;
437 struct ufs1_dinode *dp;
438 struct ufsmount *ump;
439 #ifdef QUOTA
440 int i;
441 #endif
442
443 /* Get a pointer to the private mount structure. */
444 ump = VFSTOUFS(mp);
445
446 /* Initialize the inode. */
447 ip = pool_get(&lfs_inode_pool, PR_WAITOK);
448 memset(ip, 0, sizeof(*ip));
449 dp = pool_get(&lfs_dinode_pool, PR_WAITOK);
450 memset(dp, 0, sizeof(*dp));
451 ip->inode_ext.lfs = pool_get(&lfs_inoext_pool, PR_WAITOK);
452 vp->v_data = ip;
453 ip->i_din.ffs1_din = dp;
454 ip->i_ump = ump;
455 ip->i_vnode = vp;
456 ip->i_devvp = ump->um_devvp;
457 ip->i_dev = ump->um_dev;
458 ip->i_number = dp->di_inumber = ino;
459 ip->i_lfs = ump->um_lfs;
460 ip->i_lfs_effnblks = 0;
461 #ifdef QUOTA
462 for (i = 0; i < MAXQUOTAS; i++)
463 ip->i_dquot[i] = NODQUOT;
464 #endif
465 #ifdef DEBUG_LFS_VNLOCK
466 if (ino == LFS_IFILE_INUM)
467 vp->v_vnlock->lk_wmesg = "inlock";
468 #endif
469 }
470
471 /* Free an inode. */
472 /* ARGUSED */
473 /* VOP_BWRITE 2i times */
474 int
475 lfs_vfree(void *v)
476 {
477 struct vop_vfree_args /* {
478 struct vnode *a_pvp;
479 ino_t a_ino;
480 int a_mode;
481 } */ *ap = v;
482 SEGUSE *sup;
483 CLEANERINFO *cip;
484 struct buf *cbp, *bp;
485 struct ifile *ifp;
486 struct inode *ip;
487 struct vnode *vp;
488 struct lfs *fs;
489 daddr_t old_iaddr;
490 ino_t ino, otail;
491 int s;
492
493 /* Get the inode number and file system. */
494 vp = ap->a_pvp;
495 ip = VTOI(vp);
496 fs = ip->i_lfs;
497 ino = ip->i_number;
498
499 /* Drain of pending writes */
500 s = splbio();
501 if (fs->lfs_version > 1 && WRITEINPROG(vp))
502 tsleep(vp, (PRIBIO+1), "lfs_vfree", 0);
503 splx(s);
504
505 lfs_seglock(fs, SEGM_PROT);
506
507 if (vp->v_flag & VDIROP) {
508 --lfs_dirvcount;
509 vp->v_flag &= ~VDIROP;
510 TAILQ_REMOVE(&fs->lfs_dchainhd, ip, i_lfs_dchain);
511 wakeup(&lfs_dirvcount);
512 lfs_vunref(vp);
513 }
514 lfs_unmark_vnode(vp);
515
516 LFS_CLR_UINO(ip, IN_ACCESSED|IN_CLEANING|IN_MODIFIED);
517 ip->i_flag &= ~IN_ALLMOD;
518
519 /*
520 * Set the ifile's inode entry to unused, increment its version number
521 * and link it onto the free chain.
522 */
523 LFS_IENTRY(ifp, fs, ino, bp);
524 old_iaddr = ifp->if_daddr;
525 ifp->if_daddr = LFS_UNUSED_DADDR;
526 ++ifp->if_version;
527 if (fs->lfs_version == 1) {
528 LFS_GET_HEADFREE(fs, cip, cbp, &(ifp->if_nextfree));
529 LFS_PUT_HEADFREE(fs, cip, cbp, ino);
530 (void) LFS_BWRITE_LOG(bp); /* Ifile */
531 } else {
532 ifp->if_nextfree = LFS_UNUSED_INUM;
533 /*
534 * XXX Writing the freed node here means that it might not
535 * XXX make it into the free list in the event of a crash
536 * XXX (the ifile could be written before the rest of this
537 * XXX completes).
538 */
539 (void) LFS_BWRITE_LOG(bp); /* Ifile */
540 LFS_GET_TAILFREE(fs, cip, cbp, &otail);
541 LFS_IENTRY(ifp, fs, otail, bp);
542 ifp->if_nextfree = ino;
543 LFS_BWRITE_LOG(bp);
544 LFS_PUT_TAILFREE(fs, cip, cbp, ino);
545 /* printf("lfs_vfree: tailfree %d -> %d\n", otail, ino); */
546 }
547 #ifdef DIAGNOSTIC
548 if (ino == LFS_UNUSED_INUM) {
549 panic("inode 0 freed");
550 }
551 #endif /* DIAGNOSTIC */
552 if (old_iaddr != LFS_UNUSED_DADDR) {
553 LFS_SEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp);
554 #ifdef DIAGNOSTIC
555 if (sup->su_nbytes < sizeof (struct ufs1_dinode)) {
556 printf("lfs_vfree: negative byte count"
557 " (segment %" PRIu32 " short by %d)\n",
558 dtosn(fs, old_iaddr),
559 (int)sizeof (struct ufs1_dinode) -
560 sup->su_nbytes);
561 panic("lfs_vfree: negative byte count");
562 sup->su_nbytes = sizeof (struct ufs1_dinode);
563 }
564 #endif
565 sup->su_nbytes -= sizeof (struct ufs1_dinode);
566 LFS_WRITESEGENTRY(sup, fs, dtosn(fs, old_iaddr), bp); /* Ifile */
567 }
568
569 /* Set superblock modified bit and decrement file count. */
570 fs->lfs_fmod = 1;
571 --fs->lfs_nfiles;
572
573 lfs_segunlock(fs);
574
575 return (0);
576 }
577