lfs_syscalls.c revision 1.10
1 /* $NetBSD: lfs_syscalls.c,v 1.10 1996/02/09 22:28:56 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1991, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)lfs_syscalls.c 8.6 (Berkeley) 6/16/94 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/proc.h> 41 #include <sys/buf.h> 42 #include <sys/mount.h> 43 #include <sys/vnode.h> 44 #include <sys/malloc.h> 45 #include <sys/kernel.h> 46 47 #include <sys/syscallargs.h> 48 49 #include <ufs/ufs/quota.h> 50 #include <ufs/ufs/inode.h> 51 #include <ufs/ufs/ufsmount.h> 52 #include <ufs/ufs/ufs_extern.h> 53 54 #include <ufs/lfs/lfs.h> 55 #include <ufs/lfs/lfs_extern.h> 56 57 #define BUMP_FIP(SP) \ 58 (SP)->fip = (FINFO *) (&(SP)->fip->fi_blocks[(SP)->fip->fi_nblocks]) 59 60 #define INC_FINFO(SP) ++((SEGSUM *)((SP)->segsum))->ss_nfinfo 61 #define DEC_FINFO(SP) --((SEGSUM *)((SP)->segsum))->ss_nfinfo 62 63 /* 64 * Before committing to add something to a segment summary, make sure there 65 * is enough room. S is the bytes added to the summary. 66 */ 67 #define CHECK_SEG(s) \ 68 if (sp->sum_bytes_left < (s)) { \ 69 (void) lfs_writeseg(fs, sp); \ 70 } 71 struct buf *lfs_fakebuf __P((struct vnode *, int, size_t, caddr_t)); 72 73 /* 74 * lfs_markv: 75 * 76 * This will mark inodes and blocks dirty, so they are written into the log. 77 * It will block until all the blocks have been written. The segment create 78 * time passed in the block_info and inode_info structures is used to decide 79 * if the data is valid for each block (in case some process dirtied a block 80 * or inode that is being cleaned between the determination that a block is 81 * live and the lfs_markv call). 82 * 83 * 0 on success 84 * -1/errno is return on error. 85 */ 86 int 87 lfs_markv(p, v, retval) 88 struct proc *p; 89 void *v; 90 register_t *retval; 91 { 92 struct lfs_markv_args /* { 93 syscallarg(fsid_t *) fsidp; 94 syscallarg(struct block_info *) blkiov; 95 syscallarg(int) blkcnt; 96 } */ *uap = v; 97 struct segment *sp; 98 BLOCK_INFO *blkp; 99 IFILE *ifp; 100 struct buf *bp, **bpp; 101 struct inode *ip = NULL; 102 struct lfs *fs; 103 struct mount *mntp; 104 struct vnode *vp; 105 fsid_t fsid; 106 void *start; 107 ino_t lastino; 108 daddr_t b_daddr, v_daddr; 109 u_long bsize; 110 int cnt, error; 111 112 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 113 return (error); 114 115 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 116 return (error); 117 if ((mntp = getvfs(&fsid)) == NULL) 118 return (EINVAL); 119 120 cnt = SCARG(uap, blkcnt); 121 start = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK); 122 error = copyin(SCARG(uap, blkiov), start, cnt * sizeof(BLOCK_INFO)); 123 if (error) 124 goto err1; 125 126 /* Mark blocks/inodes dirty. */ 127 fs = VFSTOUFS(mntp)->um_lfs; 128 bsize = fs->lfs_bsize; 129 error = 0; 130 131 lfs_seglock(fs, SEGM_SYNC | SEGM_CLEAN); 132 sp = fs->lfs_sp; 133 for (v_daddr = LFS_UNUSED_DADDR, lastino = LFS_UNUSED_INUM, 134 blkp = start; cnt--; ++blkp) { 135 /* 136 * Get the IFILE entry (only once) and see if the file still 137 * exists. 138 */ 139 if (lastino != blkp->bi_inode) { 140 if (lastino != LFS_UNUSED_INUM) { 141 /* Finish up last file */ 142 if (sp->fip->fi_nblocks == 0) { 143 DEC_FINFO(sp); 144 sp->sum_bytes_left += 145 sizeof(FINFO) - sizeof(daddr_t); 146 } else { 147 lfs_updatemeta(sp); 148 BUMP_FIP(sp); 149 } 150 151 lfs_writeinode(fs, sp, ip); 152 lfs_vunref(vp); 153 } 154 155 /* Start a new file */ 156 CHECK_SEG(sizeof(FINFO)); 157 sp->sum_bytes_left -= sizeof(FINFO) - sizeof(daddr_t); 158 INC_FINFO(sp); 159 sp->start_lbp = &sp->fip->fi_blocks[0]; 160 sp->vp = NULL; 161 sp->fip->fi_version = blkp->bi_version; 162 sp->fip->fi_nblocks = 0; 163 sp->fip->fi_ino = blkp->bi_inode; 164 lastino = blkp->bi_inode; 165 if (blkp->bi_inode == LFS_IFILE_INUM) 166 v_daddr = fs->lfs_idaddr; 167 else { 168 LFS_IENTRY(ifp, fs, blkp->bi_inode, bp); 169 v_daddr = ifp->if_daddr; 170 brelse(bp); 171 } 172 if (v_daddr == LFS_UNUSED_DADDR) 173 continue; 174 175 /* Get the vnode/inode. */ 176 if (lfs_fastvget(mntp, blkp->bi_inode, v_daddr, &vp, 177 blkp->bi_lbn == LFS_UNUSED_LBN ? 178 blkp->bi_bp : NULL)) { 179 #ifdef DIAGNOSTIC 180 printf("lfs_markv: VFS_VGET failed (%d)\n", 181 blkp->bi_inode); 182 #endif 183 lastino = LFS_UNUSED_INUM; 184 v_daddr = LFS_UNUSED_DADDR; 185 continue; 186 } 187 sp->vp = vp; 188 ip = VTOI(vp); 189 } else if (v_daddr == LFS_UNUSED_DADDR) 190 continue; 191 192 /* If this BLOCK_INFO didn't contain a block, keep going. */ 193 if (blkp->bi_lbn == LFS_UNUSED_LBN) 194 continue; 195 if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &b_daddr, NULL) || 196 b_daddr != blkp->bi_daddr) 197 continue; 198 /* 199 * If we got to here, then we are keeping the block. If it 200 * is an indirect block, we want to actually put it in the 201 * buffer cache so that it can be updated in the finish_meta 202 * section. If it's not, we need to allocate a fake buffer 203 * so that writeseg can perform the copyin and write the buffer. 204 */ 205 if (blkp->bi_lbn >= 0) /* Data Block */ 206 bp = lfs_fakebuf(vp, blkp->bi_lbn, bsize, 207 blkp->bi_bp); 208 else { 209 bp = getblk(vp, blkp->bi_lbn, bsize, 0, 0); 210 if (!(bp->b_flags & (B_DELWRI | B_DONE | B_CACHE)) && 211 (error = copyin(blkp->bi_bp, bp->b_data, 212 bsize))) 213 goto err2; 214 if ((error = VOP_BWRITE(bp)) != 0); 215 goto err2; 216 } 217 while (lfs_gatherblock(sp, bp, NULL)); 218 } 219 if (sp->vp) { 220 if (sp->fip->fi_nblocks == 0) { 221 DEC_FINFO(sp); 222 sp->sum_bytes_left += 223 sizeof(FINFO) - sizeof(daddr_t); 224 } else 225 lfs_updatemeta(sp); 226 227 lfs_writeinode(fs, sp, ip); 228 lfs_vunref(vp); 229 } 230 (void) lfs_writeseg(fs, sp); 231 lfs_segunlock(fs); 232 free(start, M_SEGMENT); 233 return (error); 234 235 /* 236 * XXX 237 * If we come in to error 2, we might have indirect blocks that were 238 * updated and now have bad block pointers. I don't know what to do 239 * about this. 240 */ 241 242 err2: lfs_vunref(vp); 243 /* Free up fakebuffers */ 244 for (bpp = --sp->cbpp; bpp >= sp->bpp; --bpp) 245 if ((*bpp)->b_flags & B_CALL) { 246 brelvp(*bpp); 247 free(*bpp, M_SEGMENT); 248 } else 249 brelse(*bpp); 250 lfs_segunlock(fs); 251 err1: 252 free(start, M_SEGMENT); 253 return (error); 254 } 255 256 /* 257 * lfs_bmapv: 258 * 259 * This will fill in the current disk address for arrays of blocks. 260 * 261 * 0 on success 262 * -1/errno is return on error. 263 */ 264 int 265 lfs_bmapv(p, v, retval) 266 struct proc *p; 267 void *v; 268 register_t *retval; 269 { 270 struct lfs_bmapv_args /* { 271 syscallarg(fsid_t *) fsidp; 272 syscallarg(struct block_info *) blkiov; 273 syscallarg(int) blkcnt; 274 } */ *uap = v; 275 BLOCK_INFO *blkp; 276 struct mount *mntp; 277 struct vnode *vp; 278 fsid_t fsid; 279 void *start; 280 daddr_t daddr; 281 int cnt, error, step; 282 283 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 284 return (error); 285 286 error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t)); 287 if (error) 288 return (error); 289 if ((mntp = getvfs(&fsid)) == NULL) 290 return (EINVAL); 291 292 cnt = SCARG(uap, blkcnt); 293 start = blkp = malloc(cnt * sizeof(BLOCK_INFO), M_SEGMENT, M_WAITOK); 294 error = copyin(SCARG(uap, blkiov), blkp, cnt * sizeof(BLOCK_INFO)); 295 if (error) { 296 free(blkp, M_SEGMENT); 297 return (error); 298 } 299 300 for (step = cnt; step--; ++blkp) { 301 if (blkp->bi_lbn == LFS_UNUSED_LBN) 302 continue; 303 /* Could be a deadlock ? */ 304 if (VFS_VGET(mntp, blkp->bi_inode, &vp)) 305 daddr = LFS_UNUSED_DADDR; 306 else { 307 if (VOP_BMAP(vp, blkp->bi_lbn, NULL, &daddr, NULL)) 308 daddr = LFS_UNUSED_DADDR; 309 vput(vp); 310 } 311 blkp->bi_daddr = daddr; 312 } 313 copyout(start, SCARG(uap, blkiov), cnt * sizeof(BLOCK_INFO)); 314 free(start, M_SEGMENT); 315 return (0); 316 } 317 318 /* 319 * lfs_segclean: 320 * 321 * Mark the segment clean. 322 * 323 * 0 on success 324 * -1/errno is return on error. 325 */ 326 int 327 lfs_segclean(p, v, retval) 328 struct proc *p; 329 void *v; 330 register_t *retval; 331 { 332 struct lfs_segclean_args /* { 333 syscallarg(fsid_t *) fsidp; 334 syscallarg(u_long) segment; 335 } */ *uap = v; 336 CLEANERINFO *cip; 337 SEGUSE *sup; 338 struct buf *bp; 339 struct mount *mntp; 340 struct lfs *fs; 341 fsid_t fsid; 342 int error; 343 344 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 345 return (error); 346 347 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 348 return (error); 349 if ((mntp = getvfs(&fsid)) == NULL) 350 return (EINVAL); 351 352 fs = VFSTOUFS(mntp)->um_lfs; 353 354 if (datosn(fs, fs->lfs_curseg) == SCARG(uap, segment)) 355 return (EBUSY); 356 357 LFS_SEGENTRY(sup, fs, SCARG(uap, segment), bp); 358 if (sup->su_flags & SEGUSE_ACTIVE) { 359 brelse(bp); 360 return (EBUSY); 361 } 362 fs->lfs_avail += fsbtodb(fs, fs->lfs_ssize) - 1; 363 fs->lfs_bfree += (sup->su_nsums * LFS_SUMMARY_SIZE / DEV_BSIZE) + 364 sup->su_ninos * btodb(fs->lfs_bsize); 365 sup->su_flags &= ~SEGUSE_DIRTY; 366 (void) VOP_BWRITE(bp); 367 368 LFS_CLEANERINFO(cip, fs, bp); 369 ++cip->clean; 370 --cip->dirty; 371 (void) VOP_BWRITE(bp); 372 wakeup(&fs->lfs_avail); 373 return (0); 374 } 375 376 /* 377 * lfs_segwait: 378 * 379 * This will block until a segment in file system fsid is written. A timeout 380 * in milliseconds may be specified which will awake the cleaner automatically. 381 * An fsid of -1 means any file system, and a timeout of 0 means forever. 382 * 383 * 0 on success 384 * 1 on timeout 385 * -1/errno is return on error. 386 */ 387 int 388 lfs_segwait(p, v, retval) 389 struct proc *p; 390 void *v; 391 register_t *retval; 392 { 393 struct lfs_segwait_args /* { 394 syscallarg(fsid_t *) fsidp; 395 syscallarg(struct timeval *) tv; 396 } */ *uap = v; 397 extern int lfs_allclean_wakeup; 398 struct mount *mntp; 399 struct timeval atv; 400 fsid_t fsid; 401 void *addr; 402 u_long timeout; 403 int error, s; 404 405 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) { 406 return (error); 407 } 408 #ifdef WHEN_QUADS_WORK 409 if (error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) 410 return (error); 411 if (fsid == (fsid_t)-1) 412 addr = &lfs_allclean_wakeup; 413 else { 414 if ((mntp = getvfs(&fsid)) == NULL) 415 return (EINVAL); 416 addr = &VFSTOUFS(mntp)->um_lfs->lfs_nextseg; 417 } 418 #else 419 if ((error = copyin(SCARG(uap, fsidp), &fsid, sizeof(fsid_t))) != 0) 420 return (error); 421 if ((mntp = getvfs(&fsid)) == NULL) 422 addr = &lfs_allclean_wakeup; 423 else 424 addr = &VFSTOUFS(mntp)->um_lfs->lfs_nextseg; 425 #endif 426 427 if (SCARG(uap, tv)) { 428 error = copyin(SCARG(uap, tv), &atv, sizeof(struct timeval)); 429 if (error) 430 return (error); 431 if (itimerfix(&atv)) 432 return (EINVAL); 433 s = splclock(); 434 timeradd(&atv, &time, &atv); 435 timeout = hzto(&atv); 436 splx(s); 437 } else 438 timeout = 0; 439 440 error = tsleep(addr, PCATCH | PUSER, "segment", timeout); 441 return (error == ERESTART ? EINTR : 0); 442 } 443 444 /* 445 * VFS_VGET call specialized for the cleaner. The cleaner already knows the 446 * daddr from the ifile, so don't look it up again. If the cleaner is 447 * processing IINFO structures, it may have the ondisk inode already, so 448 * don't go retrieving it again. 449 */ 450 int 451 lfs_fastvget(mp, ino, daddr, vpp, dinp) 452 struct mount *mp; 453 ino_t ino; 454 daddr_t daddr; 455 struct vnode **vpp; 456 struct dinode *dinp; 457 { 458 register struct inode *ip; 459 struct vnode *vp; 460 struct ufsmount *ump; 461 struct buf *bp; 462 dev_t dev; 463 int error; 464 465 ump = VFSTOUFS(mp); 466 dev = ump->um_dev; 467 /* 468 * This is playing fast and loose. Someone may have the inode 469 * locked, in which case they are going to be distinctly unhappy 470 * if we trash something. 471 */ 472 if ((*vpp = ufs_ihashlookup(dev, ino)) != NULL) { 473 lfs_vref(*vpp); 474 if ((*vpp)->v_flag & VXLOCK) 475 printf ("Cleaned vnode VXLOCKED\n"); 476 ip = VTOI(*vpp); 477 if (ip->i_flag & IN_LOCKED) 478 printf("cleaned vnode locked\n"); 479 if (!(ip->i_flag & IN_MODIFIED)) { 480 ++ump->um_lfs->lfs_uinodes; 481 ip->i_flag |= IN_MODIFIED; 482 } 483 ip->i_flag |= IN_MODIFIED; 484 return (0); 485 } 486 487 /* Allocate new vnode/inode. */ 488 if ((error = lfs_vcreate(mp, ino, &vp)) != 0) { 489 *vpp = NULL; 490 return (error); 491 } 492 493 /* 494 * Put it onto its hash chain and lock it so that other requests for 495 * this inode will block if they arrive while we are sleeping waiting 496 * for old data structures to be purged or for the contents of the 497 * disk portion of this inode to be read. 498 */ 499 ip = VTOI(vp); 500 ufs_ihashins(ip); 501 502 /* 503 * XXX 504 * This may not need to be here, logically it should go down with 505 * the i_devvp initialization. 506 * Ask Kirk. 507 */ 508 ip->i_lfs = ump->um_lfs; 509 510 /* Read in the disk contents for the inode, copy into the inode. */ 511 if (dinp) { 512 error = copyin(dinp, &ip->i_din, sizeof(struct dinode)); 513 if (error) 514 return (error); 515 } 516 else { 517 error = bread(ump->um_devvp, daddr, 518 (int)ump->um_lfs->lfs_bsize, NOCRED, &bp); 519 if (error) { 520 /* 521 * The inode does not contain anything useful, so it 522 * would be misleading to leave it on its hash chain. 523 * Iput() will return it to the free list. 524 */ 525 ufs_ihashrem(ip); 526 527 /* Unlock and discard unneeded inode. */ 528 lfs_vunref(vp); 529 brelse(bp); 530 *vpp = NULL; 531 return (error); 532 } 533 ip->i_din = 534 *lfs_ifind(ump->um_lfs, ino, (struct dinode *)bp->b_data); 535 brelse(bp); 536 } 537 538 /* Inode was just read from user space or disk, make sure it's locked */ 539 ip->i_flag |= IN_LOCKED; 540 541 /* 542 * Initialize the vnode from the inode, check for aliases. In all 543 * cases re-init ip, the underlying vnode/inode may have changed. 544 */ 545 error = ufs_vinit(mp, lfs_specop_p, LFS_FIFOOPS, &vp); 546 if (error) { 547 lfs_vunref(vp); 548 *vpp = NULL; 549 return (error); 550 } 551 /* 552 * Finish inode initialization now that aliasing has been resolved. 553 */ 554 ip->i_devvp = ump->um_devvp; 555 ip->i_flag |= IN_MODIFIED; 556 ++ump->um_lfs->lfs_uinodes; 557 VREF(ip->i_devvp); 558 *vpp = vp; 559 return (0); 560 } 561 struct buf * 562 lfs_fakebuf(vp, lbn, size, uaddr) 563 struct vnode *vp; 564 int lbn; 565 size_t size; 566 caddr_t uaddr; 567 { 568 struct buf *bp; 569 570 bp = lfs_newbuf(vp, lbn, 0); 571 bp->b_saveaddr = uaddr; 572 bp->b_bufsize = size; 573 bp->b_bcount = size; 574 bp->b_flags |= B_INVAL; 575 return (bp); 576 } 577
Indexes created Tue Sep 30 11:09:46 GMT 2025