ulfs_readwrite.c revision 1.3
1 /* $NetBSD: ulfs_readwrite.c,v 1.3 2013/06/08 22:05:15 dholland Exp $ */ 2 /* from NetBSD: ufs_readwrite.c,v 1.105 2013/01/22 09:39:18 dholland Exp */ 3 4 /*- 5 * Copyright (c) 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)ufs_readwrite.c 8.11 (Berkeley) 5/8/95 33 */ 34 35 #include <sys/cdefs.h> 36 __KERNEL_RCSID(1, "$NetBSD: ulfs_readwrite.c,v 1.3 2013/06/08 22:05:15 dholland Exp $"); 37 38 #ifdef LFS_READWRITE 39 #define FS struct lfs 40 #define I_FS i_lfs 41 #define READ lfs_read 42 #define READ_S "lfs_read" 43 #define WRITE lfs_write 44 #define WRITE_S "lfs_write" 45 #define fs_bsize lfs_bsize 46 #define fs_bmask lfs_bmask 47 #else 48 #define FS struct fs 49 #define I_FS i_fs 50 #define READ ffs_read 51 #define READ_S "ffs_read" 52 #define WRITE ffs_write 53 #define WRITE_S "ffs_write" 54 #endif 55 56 /* 57 * Vnode op for reading. 58 */ 59 /* ARGSUSED */ 60 int 61 READ(void *v) 62 { 63 struct vop_read_args /* { 64 struct vnode *a_vp; 65 struct uio *a_uio; 66 int a_ioflag; 67 kauth_cred_t a_cred; 68 } */ *ap = v; 69 struct vnode *vp; 70 struct inode *ip; 71 struct uio *uio; 72 struct ulfsmount *ump; 73 struct buf *bp; 74 FS *fs; 75 vsize_t bytelen; 76 daddr_t lbn, nextlbn; 77 off_t bytesinfile; 78 long size, xfersize, blkoffset; 79 int error, ioflag; 80 bool usepc = false; 81 82 vp = ap->a_vp; 83 ip = VTOI(vp); 84 ump = ip->i_ump; 85 uio = ap->a_uio; 86 ioflag = ap->a_ioflag; 87 error = 0; 88 89 #ifdef DIAGNOSTIC 90 if (uio->uio_rw != UIO_READ) 91 panic("%s: mode", READ_S); 92 93 if (vp->v_type == VLNK) { 94 if (ip->i_size < ump->um_maxsymlinklen || 95 (ump->um_maxsymlinklen == 0 && DIP(ip, blocks) == 0)) 96 panic("%s: short symlink", READ_S); 97 } else if (vp->v_type != VREG && vp->v_type != VDIR) 98 panic("%s: type %d", READ_S, vp->v_type); 99 #endif 100 fs = ip->I_FS; 101 if ((u_int64_t)uio->uio_offset > ump->um_maxfilesize) 102 return (EFBIG); 103 if (uio->uio_resid == 0) 104 return (0); 105 106 #ifndef LFS_READWRITE 107 if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT) 108 return ffs_snapshot_read(vp, uio, ioflag); 109 #endif /* !LFS_READWRITE */ 110 111 fstrans_start(vp->v_mount, FSTRANS_SHARED); 112 113 if (uio->uio_offset >= ip->i_size) 114 goto out; 115 116 #ifdef LFS_READWRITE 117 usepc = (vp->v_type == VREG && ip->i_number != LFS_IFILE_INUM); 118 #else /* !LFS_READWRITE */ 119 usepc = vp->v_type == VREG; 120 #endif /* !LFS_READWRITE */ 121 if (usepc) { 122 const int advice = IO_ADV_DECODE(ap->a_ioflag); 123 124 while (uio->uio_resid > 0) { 125 if (ioflag & IO_DIRECT) { 126 genfs_directio(vp, uio, ioflag); 127 } 128 bytelen = MIN(ip->i_size - uio->uio_offset, 129 uio->uio_resid); 130 if (bytelen == 0) 131 break; 132 error = ubc_uiomove(&vp->v_uobj, uio, bytelen, advice, 133 UBC_READ | UBC_PARTIALOK | UBC_UNMAP_FLAG(vp)); 134 if (error) 135 break; 136 } 137 goto out; 138 } 139 140 for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) { 141 bytesinfile = ip->i_size - uio->uio_offset; 142 if (bytesinfile <= 0) 143 break; 144 lbn = lblkno(fs, uio->uio_offset); 145 nextlbn = lbn + 1; 146 size = blksize(fs, ip, lbn); 147 blkoffset = blkoff(fs, uio->uio_offset); 148 xfersize = MIN(MIN(fs->fs_bsize - blkoffset, uio->uio_resid), 149 bytesinfile); 150 151 if (lblktosize(fs, nextlbn) >= ip->i_size) 152 error = bread(vp, lbn, size, NOCRED, 0, &bp); 153 else { 154 int nextsize = blksize(fs, ip, nextlbn); 155 error = breadn(vp, lbn, 156 size, &nextlbn, &nextsize, 1, NOCRED, 0, &bp); 157 } 158 if (error) 159 break; 160 161 /* 162 * We should only get non-zero b_resid when an I/O error 163 * has occurred, which should cause us to break above. 164 * However, if the short read did not cause an error, 165 * then we want to ensure that we do not uiomove bad 166 * or uninitialized data. 167 */ 168 size -= bp->b_resid; 169 if (size < xfersize) { 170 if (size == 0) 171 break; 172 xfersize = size; 173 } 174 error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio); 175 if (error) 176 break; 177 brelse(bp, 0); 178 } 179 if (bp != NULL) 180 brelse(bp, 0); 181 182 out: 183 if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) { 184 ip->i_flag |= IN_ACCESS; 185 if ((ap->a_ioflag & IO_SYNC) == IO_SYNC) { 186 error = ULFS_UPDATE(vp, NULL, NULL, UPDATE_WAIT); 187 } 188 } 189 190 fstrans_done(vp->v_mount); 191 return (error); 192 } 193 194 /* 195 * Vnode op for writing. 196 */ 197 int 198 WRITE(void *v) 199 { 200 struct vop_write_args /* { 201 struct vnode *a_vp; 202 struct uio *a_uio; 203 int a_ioflag; 204 kauth_cred_t a_cred; 205 } */ *ap = v; 206 struct vnode *vp; 207 struct uio *uio; 208 struct inode *ip; 209 FS *fs; 210 struct buf *bp; 211 kauth_cred_t cred; 212 daddr_t lbn; 213 off_t osize, origoff, oldoff, preallocoff, endallocoff, nsize; 214 int blkoffset, error, flags, ioflag, resid, size, xfersize; 215 int aflag; 216 int extended=0; 217 vsize_t bytelen; 218 bool async; 219 bool usepc = false; 220 #ifdef LFS_READWRITE 221 bool need_unreserve = false; 222 #endif 223 struct ulfsmount *ump; 224 225 cred = ap->a_cred; 226 ioflag = ap->a_ioflag; 227 uio = ap->a_uio; 228 vp = ap->a_vp; 229 ip = VTOI(vp); 230 ump = ip->i_ump; 231 232 KASSERT(vp->v_size == ip->i_size); 233 #ifdef DIAGNOSTIC 234 if (uio->uio_rw != UIO_WRITE) 235 panic("%s: mode", WRITE_S); 236 #endif 237 238 switch (vp->v_type) { 239 case VREG: 240 if (ioflag & IO_APPEND) 241 uio->uio_offset = ip->i_size; 242 if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size) 243 return (EPERM); 244 /* FALLTHROUGH */ 245 case VLNK: 246 break; 247 case VDIR: 248 if ((ioflag & IO_SYNC) == 0) 249 panic("%s: nonsync dir write", WRITE_S); 250 break; 251 default: 252 panic("%s: type", WRITE_S); 253 } 254 255 fs = ip->I_FS; 256 if (uio->uio_offset < 0 || 257 (u_int64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize) 258 return (EFBIG); 259 #ifdef LFS_READWRITE 260 /* Disallow writes to the Ifile, even if noschg flag is removed */ 261 /* XXX can this go away when the Ifile is no longer in the namespace? */ 262 if (vp == fs->lfs_ivnode) 263 return (EPERM); 264 #endif 265 if (uio->uio_resid == 0) 266 return (0); 267 268 fstrans_start(vp->v_mount, FSTRANS_SHARED); 269 270 flags = ioflag & IO_SYNC ? B_SYNC : 0; 271 async = vp->v_mount->mnt_flag & MNT_ASYNC; 272 origoff = uio->uio_offset; 273 resid = uio->uio_resid; 274 osize = ip->i_size; 275 error = 0; 276 277 usepc = vp->v_type == VREG; 278 279 #ifdef LFS_READWRITE 280 async = true; 281 lfs_availwait(fs, btofsb(fs, uio->uio_resid)); 282 lfs_check(vp, LFS_UNUSED_LBN, 0); 283 #endif /* !LFS_READWRITE */ 284 if (!usepc) 285 goto bcache; 286 287 preallocoff = round_page(blkroundup(fs, MAX(osize, uio->uio_offset))); 288 aflag = ioflag & IO_SYNC ? B_SYNC : 0; 289 nsize = MAX(osize, uio->uio_offset + uio->uio_resid); 290 endallocoff = nsize - blkoff(fs, nsize); 291 292 /* 293 * if we're increasing the file size, deal with expanding 294 * the fragment if there is one. 295 */ 296 297 if (nsize > osize && lblkno(fs, osize) < ULFS_NDADDR && 298 lblkno(fs, osize) != lblkno(fs, nsize) && 299 blkroundup(fs, osize) != osize) { 300 off_t eob; 301 302 eob = blkroundup(fs, osize); 303 uvm_vnp_setwritesize(vp, eob); 304 error = ulfs_balloc_range(vp, osize, eob - osize, cred, aflag); 305 if (error) 306 goto out; 307 if (flags & B_SYNC) { 308 mutex_enter(vp->v_interlock); 309 VOP_PUTPAGES(vp, trunc_page(osize & fs->fs_bmask), 310 round_page(eob), 311 PGO_CLEANIT | PGO_SYNCIO | PGO_JOURNALLOCKED); 312 } 313 } 314 315 while (uio->uio_resid > 0) { 316 int ubc_flags = UBC_WRITE; 317 bool overwrite; /* if we're overwrite a whole block */ 318 off_t newoff; 319 320 if (ioflag & IO_DIRECT) { 321 genfs_directio(vp, uio, ioflag | IO_JOURNALLOCKED); 322 } 323 324 oldoff = uio->uio_offset; 325 blkoffset = blkoff(fs, uio->uio_offset); 326 bytelen = MIN(fs->fs_bsize - blkoffset, uio->uio_resid); 327 if (bytelen == 0) { 328 break; 329 } 330 331 /* 332 * if we're filling in a hole, allocate the blocks now and 333 * initialize the pages first. if we're extending the file, 334 * we can safely allocate blocks without initializing pages 335 * since the new blocks will be inaccessible until the write 336 * is complete. 337 */ 338 overwrite = uio->uio_offset >= preallocoff && 339 uio->uio_offset < endallocoff; 340 if (!overwrite && (vp->v_vflag & VV_MAPPED) == 0 && 341 blkoff(fs, uio->uio_offset) == 0 && 342 (uio->uio_offset & PAGE_MASK) == 0) { 343 vsize_t len; 344 345 len = trunc_page(bytelen); 346 len -= blkoff(fs, len); 347 if (len > 0) { 348 overwrite = true; 349 bytelen = len; 350 } 351 } 352 353 newoff = oldoff + bytelen; 354 if (vp->v_size < newoff) { 355 uvm_vnp_setwritesize(vp, newoff); 356 } 357 358 if (!overwrite) { 359 error = ulfs_balloc_range(vp, uio->uio_offset, bytelen, 360 cred, aflag); 361 if (error) 362 break; 363 } else { 364 genfs_node_wrlock(vp); 365 error = GOP_ALLOC(vp, uio->uio_offset, bytelen, 366 aflag, cred); 367 genfs_node_unlock(vp); 368 if (error) 369 break; 370 ubc_flags |= UBC_FAULTBUSY; 371 } 372 373 /* 374 * copy the data. 375 */ 376 377 error = ubc_uiomove(&vp->v_uobj, uio, bytelen, 378 IO_ADV_DECODE(ioflag), ubc_flags | UBC_UNMAP_FLAG(vp)); 379 380 /* 381 * update UVM's notion of the size now that we've 382 * copied the data into the vnode's pages. 383 * 384 * we should update the size even when uiomove failed. 385 */ 386 387 if (vp->v_size < newoff) { 388 uvm_vnp_setsize(vp, newoff); 389 extended = 1; 390 } 391 392 if (error) 393 break; 394 395 /* 396 * flush what we just wrote if necessary. 397 * XXXUBC simplistic async flushing. 398 */ 399 400 #ifndef LFS_READWRITE 401 if (!async && oldoff >> 16 != uio->uio_offset >> 16) { 402 mutex_enter(vp->v_interlock); 403 error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16, 404 (uio->uio_offset >> 16) << 16, 405 PGO_CLEANIT | PGO_JOURNALLOCKED | PGO_LAZY); 406 if (error) 407 break; 408 } 409 #endif 410 } 411 if (error == 0 && ioflag & IO_SYNC) { 412 mutex_enter(vp->v_interlock); 413 error = VOP_PUTPAGES(vp, trunc_page(origoff & fs->fs_bmask), 414 round_page(blkroundup(fs, uio->uio_offset)), 415 PGO_CLEANIT | PGO_SYNCIO | PGO_JOURNALLOCKED); 416 } 417 goto out; 418 419 bcache: 420 mutex_enter(vp->v_interlock); 421 VOP_PUTPAGES(vp, trunc_page(origoff), round_page(origoff + resid), 422 PGO_CLEANIT | PGO_FREE | PGO_SYNCIO | PGO_JOURNALLOCKED); 423 while (uio->uio_resid > 0) { 424 lbn = lblkno(fs, uio->uio_offset); 425 blkoffset = blkoff(fs, uio->uio_offset); 426 xfersize = MIN(fs->fs_bsize - blkoffset, uio->uio_resid); 427 if (fs->fs_bsize > xfersize) 428 flags |= B_CLRBUF; 429 else 430 flags &= ~B_CLRBUF; 431 432 #ifdef LFS_READWRITE 433 error = lfs_reserve(fs, vp, NULL, 434 btofsb(fs, (ULFS_NIADDR + 1) << fs->lfs_bshift)); 435 if (error) 436 break; 437 need_unreserve = true; 438 #endif 439 error = ULFS_BALLOC(vp, uio->uio_offset, xfersize, 440 ap->a_cred, flags, &bp); 441 442 if (error) 443 break; 444 if (uio->uio_offset + xfersize > ip->i_size) { 445 ip->i_size = uio->uio_offset + xfersize; 446 DIP_ASSIGN(ip, size, ip->i_size); 447 uvm_vnp_setsize(vp, ip->i_size); 448 extended = 1; 449 } 450 size = blksize(fs, ip, lbn) - bp->b_resid; 451 if (xfersize > size) 452 xfersize = size; 453 454 error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio); 455 456 /* 457 * if we didn't clear the block and the uiomove failed, 458 * the buf will now contain part of some other file, 459 * so we need to invalidate it. 460 */ 461 if (error && (flags & B_CLRBUF) == 0) { 462 brelse(bp, BC_INVAL); 463 break; 464 } 465 #ifdef LFS_READWRITE 466 (void)VOP_BWRITE(bp->b_vp, bp); 467 lfs_reserve(fs, vp, NULL, 468 -btofsb(fs, (ULFS_NIADDR + 1) << fs->lfs_bshift)); 469 need_unreserve = false; 470 #else 471 if (ioflag & IO_SYNC) 472 (void)bwrite(bp); 473 else if (xfersize + blkoffset == fs->fs_bsize) 474 bawrite(bp); 475 else 476 bdwrite(bp); 477 #endif 478 if (error || xfersize == 0) 479 break; 480 } 481 #ifdef LFS_READWRITE 482 if (need_unreserve) { 483 lfs_reserve(fs, vp, NULL, 484 -btofsb(fs, (ULFS_NIADDR + 1) << fs->lfs_bshift)); 485 } 486 #endif 487 488 /* 489 * If we successfully wrote any data, and we are not the superuser 490 * we clear the setuid and setgid bits as a precaution against 491 * tampering. 492 */ 493 out: 494 ip->i_flag |= IN_CHANGE | IN_UPDATE; 495 if (vp->v_mount->mnt_flag & MNT_RELATIME) 496 ip->i_flag |= IN_ACCESS; 497 if (resid > uio->uio_resid && ap->a_cred) { 498 if (ip->i_mode & ISUID) { 499 if (kauth_authorize_vnode(ap->a_cred, 500 KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) { 501 ip->i_mode &= ~ISUID; 502 DIP_ASSIGN(ip, mode, ip->i_mode); 503 } 504 } 505 506 if (ip->i_mode & ISGID) { 507 if (kauth_authorize_vnode(ap->a_cred, 508 KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) { 509 ip->i_mode &= ~ISGID; 510 DIP_ASSIGN(ip, mode, ip->i_mode); 511 } 512 } 513 } 514 if (resid > uio->uio_resid) 515 VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0)); 516 if (error) { 517 (void) ULFS_TRUNCATE(vp, osize, ioflag & IO_SYNC, ap->a_cred); 518 uio->uio_offset -= resid - uio->uio_resid; 519 uio->uio_resid = resid; 520 } else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC) { 521 error = ULFS_UPDATE(vp, NULL, NULL, UPDATE_WAIT); 522 } else { 523 /* nothing */ 524 } 525 KASSERT(vp->v_size == ip->i_size); 526 fstrans_done(vp->v_mount); 527 528 return (error); 529 } 530
Indexes created Tue Sep 30 20:09:53 GMT 2025