| 1 | /* $NetBSD: ufs_readwrite.c,v 1.120 2015/04/12 22:48:38 riastradh Exp $ */ |
| 2 | |
| 3 | /*- |
| 4 | * Copyright (c) 1993 |
| 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. Neither the name of the University nor the names of its contributors |
| 16 | * may be used to endorse or promote products derived from this software |
| 17 | * without specific prior written permission. |
| 18 | * |
| 19 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 20 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 21 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 22 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 23 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 24 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 25 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 26 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 27 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 28 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 29 | * SUCH DAMAGE. |
| 30 | * |
| 31 | * @(#)ufs_readwrite.c 8.11 (Berkeley) 5/8/95 |
| 32 | */ |
| 33 | |
| 34 | #include <sys/cdefs.h> |
| 35 | __KERNEL_RCSID(1, "$NetBSD: ufs_readwrite.c,v 1.120 2015/04/12 22:48:38 riastradh Exp $" ); |
| 36 | |
| 37 | #ifdef LFS_READWRITE |
| 38 | #define FS struct lfs |
| 39 | #define I_FS i_lfs |
| 40 | #define READ lfs_read |
| 41 | #define READ_S "lfs_read" |
| 42 | #define WRITE lfs_write |
| 43 | #define WRITE_S "lfs_write" |
| 44 | #define BUFRD lfs_bufrd |
| 45 | #define BUFWR lfs_bufwr |
| 46 | #define fs_bsize lfs_bsize |
| 47 | #define fs_bmask lfs_bmask |
| 48 | #define UFS_WAPBL_BEGIN(mp) 0 |
| 49 | #define UFS_WAPBL_END(mp) do { } while (0) |
| 50 | #define UFS_WAPBL_UPDATE(vp, access, modify, flags) do { } while (0) |
| 51 | #define ufs_blkoff lfs_blkoff |
| 52 | #define ufs_blksize lfs_blksize |
| 53 | #define ufs_lblkno lfs_lblkno |
| 54 | #define ufs_lblktosize lfs_lblktosize |
| 55 | #define ufs_blkroundup lfs_blkroundup |
| 56 | #else |
| 57 | #define FS struct fs |
| 58 | #define I_FS i_fs |
| 59 | #define READ ffs_read |
| 60 | #define READ_S "ffs_read" |
| 61 | #define WRITE ffs_write |
| 62 | #define WRITE_S "ffs_write" |
| 63 | #define BUFRD ffs_bufrd |
| 64 | #define BUFWR ffs_bufwr |
| 65 | #define ufs_blkoff ffs_blkoff |
| 66 | #define ufs_blksize ffs_blksize |
| 67 | #define ufs_lblkno ffs_lblkno |
| 68 | #define ufs_lblktosize ffs_lblktosize |
| 69 | #define ufs_blkroundup ffs_blkroundup |
| 70 | #endif |
| 71 | |
| 72 | static int ufs_post_read_update(struct vnode *, int, int); |
| 73 | static int ufs_post_write_update(struct vnode *, struct uio *, int, |
| 74 | kauth_cred_t, off_t, int, int, int); |
| 75 | |
| 76 | /* |
| 77 | * Vnode op for reading. |
| 78 | */ |
| 79 | /* ARGSUSED */ |
| 80 | int |
| 81 | READ(void *v) |
| 82 | { |
| 83 | struct vop_read_args /* { |
| 84 | struct vnode *a_vp; |
| 85 | struct uio *a_uio; |
| 86 | int a_ioflag; |
| 87 | kauth_cred_t a_cred; |
| 88 | } */ *ap = v; |
| 89 | struct vnode *vp; |
| 90 | struct inode *ip; |
| 91 | struct uio *uio; |
| 92 | struct ufsmount *ump; |
| 93 | vsize_t bytelen; |
| 94 | int error, ioflag, advice; |
| 95 | |
| 96 | vp = ap->a_vp; |
| 97 | ip = VTOI(vp); |
| 98 | ump = ip->i_ump; |
| 99 | uio = ap->a_uio; |
| 100 | ioflag = ap->a_ioflag; |
| 101 | error = 0; |
| 102 | |
| 103 | KASSERT(uio->uio_rw == UIO_READ); |
| 104 | KASSERT(vp->v_type == VREG || vp->v_type == VDIR); |
| 105 | |
| 106 | /* XXX Eliminate me by refusing directory reads from userland. */ |
| 107 | if (vp->v_type == VDIR) |
| 108 | return BUFRD(vp, uio, ioflag, ap->a_cred); |
| 109 | #ifdef LFS_READWRITE |
| 110 | /* XXX Eliminate me by using ufs_bufio in lfs. */ |
| 111 | if (vp->v_type == VREG && ip->i_number == LFS_IFILE_INUM) |
| 112 | return BUFRD(vp, uio, ioflag, ap->a_cred); |
| 113 | #endif |
| 114 | if ((u_int64_t)uio->uio_offset > ump->um_maxfilesize) |
| 115 | return (EFBIG); |
| 116 | if (uio->uio_resid == 0) |
| 117 | return (0); |
| 118 | |
| 119 | #ifndef LFS_READWRITE |
| 120 | if ((ip->i_flags & (SF_SNAPSHOT | SF_SNAPINVAL)) == SF_SNAPSHOT) |
| 121 | return ffs_snapshot_read(vp, uio, ioflag); |
| 122 | #endif /* !LFS_READWRITE */ |
| 123 | |
| 124 | fstrans_start(vp->v_mount, FSTRANS_SHARED); |
| 125 | |
| 126 | if (uio->uio_offset >= ip->i_size) |
| 127 | goto out; |
| 128 | |
| 129 | KASSERT(vp->v_type == VREG); |
| 130 | advice = IO_ADV_DECODE(ap->a_ioflag); |
| 131 | while (uio->uio_resid > 0) { |
| 132 | if (ioflag & IO_DIRECT) { |
| 133 | genfs_directio(vp, uio, ioflag); |
| 134 | } |
| 135 | bytelen = MIN(ip->i_size - uio->uio_offset, uio->uio_resid); |
| 136 | if (bytelen == 0) |
| 137 | break; |
| 138 | error = ubc_uiomove(&vp->v_uobj, uio, bytelen, advice, |
| 139 | UBC_READ | UBC_PARTIALOK | UBC_UNMAP_FLAG(vp)); |
| 140 | if (error) |
| 141 | break; |
| 142 | } |
| 143 | |
| 144 | out: |
| 145 | error = ufs_post_read_update(vp, ap->a_ioflag, error); |
| 146 | fstrans_done(vp->v_mount); |
| 147 | return (error); |
| 148 | } |
| 149 | |
| 150 | /* |
| 151 | * UFS op for reading via the buffer cache |
| 152 | */ |
| 153 | int |
| 154 | BUFRD(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t cred) |
| 155 | { |
| 156 | struct inode *ip; |
| 157 | struct ufsmount *ump; |
| 158 | FS *fs; |
| 159 | struct buf *bp; |
| 160 | daddr_t lbn, nextlbn; |
| 161 | off_t bytesinfile; |
| 162 | long size, xfersize, blkoffset; |
| 163 | int error; |
| 164 | |
| 165 | KASSERT(VOP_ISLOCKED(vp)); |
| 166 | KASSERT(vp->v_type == VDIR || vp->v_type == VLNK); |
| 167 | KASSERT(uio->uio_rw == UIO_READ); |
| 168 | |
| 169 | ip = VTOI(vp); |
| 170 | ump = ip->i_ump; |
| 171 | fs = ip->I_FS; |
| 172 | error = 0; |
| 173 | |
| 174 | KASSERT(vp->v_type != VLNK || ip->i_size >= ump->um_maxsymlinklen); |
| 175 | KASSERT(vp->v_type != VLNK || ump->um_maxsymlinklen != 0 || |
| 176 | DIP(ip, blocks) == 0); |
| 177 | |
| 178 | if (uio->uio_offset > ump->um_maxfilesize) |
| 179 | return EFBIG; |
| 180 | if (uio->uio_resid == 0) |
| 181 | return 0; |
| 182 | |
| 183 | #ifndef LFS_READWRITE |
| 184 | KASSERT(!ISSET(ip->i_flags, (SF_SNAPSHOT | SF_SNAPINVAL))); |
| 185 | #endif |
| 186 | |
| 187 | fstrans_start(vp->v_mount, FSTRANS_SHARED); |
| 188 | |
| 189 | if (uio->uio_offset >= ip->i_size) |
| 190 | goto out; |
| 191 | |
| 192 | for (error = 0, bp = NULL; uio->uio_resid > 0; bp = NULL) { |
| 193 | bytesinfile = ip->i_size - uio->uio_offset; |
| 194 | if (bytesinfile <= 0) |
| 195 | break; |
| 196 | lbn = ufs_lblkno(fs, uio->uio_offset); |
| 197 | nextlbn = lbn + 1; |
| 198 | size = ufs_blksize(fs, ip, lbn); |
| 199 | blkoffset = ufs_blkoff(fs, uio->uio_offset); |
| 200 | xfersize = MIN(MIN(fs->fs_bsize - blkoffset, uio->uio_resid), |
| 201 | bytesinfile); |
| 202 | |
| 203 | if (ufs_lblktosize(fs, nextlbn) >= ip->i_size) |
| 204 | error = bread(vp, lbn, size, 0, &bp); |
| 205 | else { |
| 206 | int nextsize = ufs_blksize(fs, ip, nextlbn); |
| 207 | error = breadn(vp, lbn, |
| 208 | size, &nextlbn, &nextsize, 1, 0, &bp); |
| 209 | } |
| 210 | if (error) |
| 211 | break; |
| 212 | |
| 213 | /* |
| 214 | * We should only get non-zero b_resid when an I/O error |
| 215 | * has occurred, which should cause us to break above. |
| 216 | * However, if the short read did not cause an error, |
| 217 | * then we want to ensure that we do not uiomove bad |
| 218 | * or uninitialized data. |
| 219 | */ |
| 220 | size -= bp->b_resid; |
| 221 | if (size < xfersize) { |
| 222 | if (size == 0) |
| 223 | break; |
| 224 | xfersize = size; |
| 225 | } |
| 226 | error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio); |
| 227 | if (error) |
| 228 | break; |
| 229 | brelse(bp, 0); |
| 230 | } |
| 231 | if (bp != NULL) |
| 232 | brelse(bp, 0); |
| 233 | |
| 234 | out: |
| 235 | error = ufs_post_read_update(vp, ioflag, error); |
| 236 | fstrans_done(vp->v_mount); |
| 237 | return (error); |
| 238 | } |
| 239 | |
| 240 | static int |
| 241 | ufs_post_read_update(struct vnode *vp, int ioflag, int oerror) |
| 242 | { |
| 243 | struct inode *ip = VTOI(vp); |
| 244 | int error = oerror; |
| 245 | |
| 246 | if (!(vp->v_mount->mnt_flag & MNT_NOATIME)) { |
| 247 | ip->i_flag |= IN_ACCESS; |
| 248 | if ((ioflag & IO_SYNC) == IO_SYNC) { |
| 249 | error = UFS_WAPBL_BEGIN(vp->v_mount); |
| 250 | if (error) |
| 251 | goto out; |
| 252 | error = UFS_UPDATE(vp, NULL, NULL, UPDATE_WAIT); |
| 253 | UFS_WAPBL_END(vp->v_mount); |
| 254 | } |
| 255 | } |
| 256 | |
| 257 | out: |
| 258 | /* Read error overrides any inode update error. */ |
| 259 | if (oerror) |
| 260 | error = oerror; |
| 261 | return error; |
| 262 | } |
| 263 | |
| 264 | /* |
| 265 | * Vnode op for writing. |
| 266 | */ |
| 267 | int |
| 268 | WRITE(void *v) |
| 269 | { |
| 270 | struct vop_write_args /* { |
| 271 | struct vnode *a_vp; |
| 272 | struct uio *a_uio; |
| 273 | int a_ioflag; |
| 274 | kauth_cred_t a_cred; |
| 275 | } */ *ap = v; |
| 276 | struct vnode *vp; |
| 277 | struct uio *uio; |
| 278 | struct inode *ip; |
| 279 | FS *fs; |
| 280 | kauth_cred_t cred; |
| 281 | off_t osize, origoff, oldoff, preallocoff, endallocoff, nsize; |
| 282 | int blkoffset, error, flags, ioflag, resid; |
| 283 | int aflag; |
| 284 | int extended=0; |
| 285 | vsize_t bytelen; |
| 286 | bool async; |
| 287 | struct ufsmount *ump; |
| 288 | |
| 289 | cred = ap->a_cred; |
| 290 | ioflag = ap->a_ioflag; |
| 291 | uio = ap->a_uio; |
| 292 | vp = ap->a_vp; |
| 293 | ip = VTOI(vp); |
| 294 | ump = ip->i_ump; |
| 295 | |
| 296 | KASSERT(vp->v_size == ip->i_size); |
| 297 | KASSERT(uio->uio_rw == UIO_WRITE); |
| 298 | KASSERT(vp->v_type == VREG); |
| 299 | KASSERT(!ISSET(ioflag, IO_JOURNALLOCKED)); |
| 300 | UFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount); |
| 301 | |
| 302 | if (ioflag & IO_APPEND) |
| 303 | uio->uio_offset = ip->i_size; |
| 304 | if ((ip->i_flags & APPEND) && uio->uio_offset != ip->i_size) |
| 305 | return (EPERM); |
| 306 | |
| 307 | fs = ip->I_FS; |
| 308 | if (uio->uio_offset < 0 || |
| 309 | (u_int64_t)uio->uio_offset + uio->uio_resid > ump->um_maxfilesize) |
| 310 | return (EFBIG); |
| 311 | #ifdef LFS_READWRITE |
| 312 | /* Disallow writes to the Ifile, even if noschg flag is removed */ |
| 313 | /* XXX can this go away when the Ifile is no longer in the namespace? */ |
| 314 | if (vp == fs->lfs_ivnode) |
| 315 | return (EPERM); |
| 316 | #endif |
| 317 | if (uio->uio_resid == 0) |
| 318 | return (0); |
| 319 | |
| 320 | fstrans_start(vp->v_mount, FSTRANS_SHARED); |
| 321 | |
| 322 | flags = ioflag & IO_SYNC ? B_SYNC : 0; |
| 323 | async = vp->v_mount->mnt_flag & MNT_ASYNC; |
| 324 | origoff = uio->uio_offset; |
| 325 | resid = uio->uio_resid; |
| 326 | osize = ip->i_size; |
| 327 | error = 0; |
| 328 | |
| 329 | KASSERT(vp->v_type == VREG); |
| 330 | |
| 331 | /* |
| 332 | * XXX The entire write operation must occur in a single WAPBL |
| 333 | * transaction because it may allocate disk blocks, if |
| 334 | * appending or filling holes, which is allowed to happen only |
| 335 | * if the write fully succeeds. |
| 336 | * |
| 337 | * If ubc_uiomove fails in the middle with EFAULT, we can clean |
| 338 | * up at the end with UFS_TRUNCATE. But if the power fails in |
| 339 | * the middle, there would be nobody to deallocate the blocks, |
| 340 | * without an fsck to globally analyze the file system. |
| 341 | * |
| 342 | * If the increasingly inaccurately named WAPBL were augmented |
| 343 | * with rollback records for block allocations, then we could |
| 344 | * split this into multiple transactions and commit the |
| 345 | * allocations in the last one. |
| 346 | * |
| 347 | * But WAPBL doesn't have that notion now, so we'll have to |
| 348 | * live with gigantic transactions and WAPBL tentacles in |
| 349 | * genfs_getpages/putpages to cope with the possibility that |
| 350 | * the transaction may or may not be locked on entry to the |
| 351 | * page cache. |
| 352 | * |
| 353 | * And even if we added that notion to WAPBL, it wouldn't help |
| 354 | * us get rid of the tentacles in genfs_getpages/putpages |
| 355 | * because we'd have to interoperate with old implementations |
| 356 | * that assume they can replay the log without fsck. |
| 357 | */ |
| 358 | error = UFS_WAPBL_BEGIN(vp->v_mount); |
| 359 | if (error) { |
| 360 | fstrans_done(vp->v_mount); |
| 361 | return error; |
| 362 | } |
| 363 | |
| 364 | #ifdef LFS_READWRITE |
| 365 | async = true; |
| 366 | lfs_availwait(fs, btofsb(fs, uio->uio_resid)); |
| 367 | lfs_check(vp, LFS_UNUSED_LBN, 0); |
| 368 | #endif /* !LFS_READWRITE */ |
| 369 | |
| 370 | preallocoff = round_page(ufs_blkroundup(fs, MAX(osize, uio->uio_offset))); |
| 371 | aflag = ioflag & IO_SYNC ? B_SYNC : 0; |
| 372 | nsize = MAX(osize, uio->uio_offset + uio->uio_resid); |
| 373 | endallocoff = nsize - ufs_blkoff(fs, nsize); |
| 374 | |
| 375 | /* |
| 376 | * if we're increasing the file size, deal with expanding |
| 377 | * the fragment if there is one. |
| 378 | */ |
| 379 | |
| 380 | if (nsize > osize && ufs_lblkno(fs, osize) < UFS_NDADDR && |
| 381 | ufs_lblkno(fs, osize) != ufs_lblkno(fs, nsize) && |
| 382 | ufs_blkroundup(fs, osize) != osize) { |
| 383 | off_t eob; |
| 384 | |
| 385 | eob = ufs_blkroundup(fs, osize); |
| 386 | uvm_vnp_setwritesize(vp, eob); |
| 387 | error = ufs_balloc_range(vp, osize, eob - osize, cred, aflag); |
| 388 | if (error) |
| 389 | goto out; |
| 390 | if (flags & B_SYNC) { |
| 391 | mutex_enter(vp->v_interlock); |
| 392 | VOP_PUTPAGES(vp, trunc_page(osize & fs->fs_bmask), |
| 393 | round_page(eob), |
| 394 | PGO_CLEANIT | PGO_SYNCIO | PGO_JOURNALLOCKED); |
| 395 | } |
| 396 | } |
| 397 | |
| 398 | while (uio->uio_resid > 0) { |
| 399 | int ubc_flags = UBC_WRITE; |
| 400 | bool overwrite; /* if we're overwrite a whole block */ |
| 401 | off_t newoff; |
| 402 | |
| 403 | if (ioflag & IO_DIRECT) { |
| 404 | genfs_directio(vp, uio, ioflag | IO_JOURNALLOCKED); |
| 405 | } |
| 406 | |
| 407 | oldoff = uio->uio_offset; |
| 408 | blkoffset = ufs_blkoff(fs, uio->uio_offset); |
| 409 | bytelen = MIN(fs->fs_bsize - blkoffset, uio->uio_resid); |
| 410 | if (bytelen == 0) { |
| 411 | break; |
| 412 | } |
| 413 | |
| 414 | /* |
| 415 | * if we're filling in a hole, allocate the blocks now and |
| 416 | * initialize the pages first. if we're extending the file, |
| 417 | * we can safely allocate blocks without initializing pages |
| 418 | * since the new blocks will be inaccessible until the write |
| 419 | * is complete. |
| 420 | */ |
| 421 | overwrite = uio->uio_offset >= preallocoff && |
| 422 | uio->uio_offset < endallocoff; |
| 423 | if (!overwrite && (vp->v_vflag & VV_MAPPED) == 0 && |
| 424 | ufs_blkoff(fs, uio->uio_offset) == 0 && |
| 425 | (uio->uio_offset & PAGE_MASK) == 0) { |
| 426 | vsize_t len; |
| 427 | |
| 428 | len = trunc_page(bytelen); |
| 429 | len -= ufs_blkoff(fs, len); |
| 430 | if (len > 0) { |
| 431 | overwrite = true; |
| 432 | bytelen = len; |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | newoff = oldoff + bytelen; |
| 437 | if (vp->v_size < newoff) { |
| 438 | uvm_vnp_setwritesize(vp, newoff); |
| 439 | } |
| 440 | |
| 441 | if (!overwrite) { |
| 442 | error = ufs_balloc_range(vp, uio->uio_offset, bytelen, |
| 443 | cred, aflag); |
| 444 | if (error) |
| 445 | break; |
| 446 | } else { |
| 447 | genfs_node_wrlock(vp); |
| 448 | error = GOP_ALLOC(vp, uio->uio_offset, bytelen, |
| 449 | aflag, cred); |
| 450 | genfs_node_unlock(vp); |
| 451 | if (error) |
| 452 | break; |
| 453 | ubc_flags |= UBC_FAULTBUSY; |
| 454 | } |
| 455 | |
| 456 | /* |
| 457 | * copy the data. |
| 458 | */ |
| 459 | |
| 460 | error = ubc_uiomove(&vp->v_uobj, uio, bytelen, |
| 461 | IO_ADV_DECODE(ioflag), ubc_flags | UBC_UNMAP_FLAG(vp)); |
| 462 | |
| 463 | /* |
| 464 | * update UVM's notion of the size now that we've |
| 465 | * copied the data into the vnode's pages. |
| 466 | * |
| 467 | * we should update the size even when uiomove failed. |
| 468 | */ |
| 469 | |
| 470 | if (vp->v_size < newoff) { |
| 471 | uvm_vnp_setsize(vp, newoff); |
| 472 | extended = 1; |
| 473 | } |
| 474 | |
| 475 | if (error) |
| 476 | break; |
| 477 | |
| 478 | /* |
| 479 | * flush what we just wrote if necessary. |
| 480 | * XXXUBC simplistic async flushing. |
| 481 | */ |
| 482 | |
| 483 | #ifndef LFS_READWRITE |
| 484 | if (!async && oldoff >> 16 != uio->uio_offset >> 16) { |
| 485 | mutex_enter(vp->v_interlock); |
| 486 | error = VOP_PUTPAGES(vp, (oldoff >> 16) << 16, |
| 487 | (uio->uio_offset >> 16) << 16, |
| 488 | PGO_CLEANIT | PGO_JOURNALLOCKED | PGO_LAZY); |
| 489 | if (error) |
| 490 | break; |
| 491 | } |
| 492 | #endif |
| 493 | } |
| 494 | if (error == 0 && ioflag & IO_SYNC) { |
| 495 | mutex_enter(vp->v_interlock); |
| 496 | error = VOP_PUTPAGES(vp, trunc_page(origoff & fs->fs_bmask), |
| 497 | round_page(ufs_blkroundup(fs, uio->uio_offset)), |
| 498 | PGO_CLEANIT | PGO_SYNCIO | PGO_JOURNALLOCKED); |
| 499 | } |
| 500 | |
| 501 | out: |
| 502 | error = ufs_post_write_update(vp, uio, ioflag, cred, osize, resid, |
| 503 | extended, error); |
| 504 | UFS_WAPBL_END(vp->v_mount); |
| 505 | fstrans_done(vp->v_mount); |
| 506 | |
| 507 | return (error); |
| 508 | } |
| 509 | |
| 510 | /* |
| 511 | * UFS op for writing via the buffer cache |
| 512 | */ |
| 513 | int |
| 514 | BUFWR(struct vnode *vp, struct uio *uio, int ioflag, kauth_cred_t cred) |
| 515 | { |
| 516 | struct inode *ip; |
| 517 | struct ufsmount *ump; |
| 518 | FS *fs; |
| 519 | int flags; |
| 520 | struct buf *bp; |
| 521 | off_t osize; |
| 522 | int resid, xfersize, size, blkoffset; |
| 523 | daddr_t lbn; |
| 524 | int extended=0; |
| 525 | int error; |
| 526 | #ifdef LFS_READWRITE |
| 527 | bool need_unreserve = false; |
| 528 | #endif |
| 529 | |
| 530 | KASSERT(ISSET(ioflag, IO_NODELOCKED)); |
| 531 | KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE); |
| 532 | KASSERT(vp->v_type == VDIR || vp->v_type == VLNK); |
| 533 | KASSERT(vp->v_type != VDIR || ISSET(ioflag, IO_SYNC)); |
| 534 | KASSERT(uio->uio_rw == UIO_WRITE); |
| 535 | KASSERT(ISSET(ioflag, IO_JOURNALLOCKED)); |
| 536 | UFS_WAPBL_JLOCK_ASSERT(vp->v_mount); |
| 537 | |
| 538 | ip = VTOI(vp); |
| 539 | ump = ip->i_ump; |
| 540 | fs = ip->I_FS; |
| 541 | |
| 542 | KASSERT(vp->v_size == ip->i_size); |
| 543 | |
| 544 | if (uio->uio_offset < 0 || |
| 545 | uio->uio_resid > ump->um_maxfilesize || |
| 546 | uio->uio_offset > (ump->um_maxfilesize - uio->uio_resid)) |
| 547 | return EFBIG; |
| 548 | #ifdef LFS_READWRITE |
| 549 | KASSERT(vp != fs->lfs_ivnode); |
| 550 | #endif |
| 551 | if (uio->uio_resid == 0) |
| 552 | return 0; |
| 553 | |
| 554 | fstrans_start(vp->v_mount, FSTRANS_SHARED); |
| 555 | |
| 556 | flags = ioflag & IO_SYNC ? B_SYNC : 0; |
| 557 | resid = uio->uio_resid; |
| 558 | osize = ip->i_size; |
| 559 | error = 0; |
| 560 | |
| 561 | KASSERT(vp->v_type != VREG); |
| 562 | |
| 563 | #ifdef LFS_READWRITE |
| 564 | lfs_availwait(fs, btofsb(fs, uio->uio_resid)); |
| 565 | lfs_check(vp, LFS_UNUSED_LBN, 0); |
| 566 | #endif /* !LFS_READWRITE */ |
| 567 | |
| 568 | /* XXX Should never have pages cached here. */ |
| 569 | KASSERT(vp->v_uobj.uo_npages == 0); |
| 570 | while (uio->uio_resid > 0) { |
| 571 | lbn = ufs_lblkno(fs, uio->uio_offset); |
| 572 | blkoffset = ufs_blkoff(fs, uio->uio_offset); |
| 573 | xfersize = MIN(fs->fs_bsize - blkoffset, uio->uio_resid); |
| 574 | if (fs->fs_bsize > xfersize) |
| 575 | flags |= B_CLRBUF; |
| 576 | else |
| 577 | flags &= ~B_CLRBUF; |
| 578 | |
| 579 | #ifdef LFS_READWRITE |
| 580 | error = lfs_reserve(fs, vp, NULL, |
| 581 | btofsb(fs, (UFS_NIADDR + 1) << fs->lfs_bshift)); |
| 582 | if (error) |
| 583 | break; |
| 584 | need_unreserve = true; |
| 585 | #endif |
| 586 | error = UFS_BALLOC(vp, uio->uio_offset, xfersize, cred, flags, |
| 587 | &bp); |
| 588 | |
| 589 | if (error) |
| 590 | break; |
| 591 | if (uio->uio_offset + xfersize > ip->i_size) { |
| 592 | ip->i_size = uio->uio_offset + xfersize; |
| 593 | DIP_ASSIGN(ip, size, ip->i_size); |
| 594 | uvm_vnp_setsize(vp, ip->i_size); |
| 595 | extended = 1; |
| 596 | } |
| 597 | size = ufs_blksize(fs, ip, lbn) - bp->b_resid; |
| 598 | if (xfersize > size) |
| 599 | xfersize = size; |
| 600 | |
| 601 | error = uiomove((char *)bp->b_data + blkoffset, xfersize, uio); |
| 602 | |
| 603 | /* |
| 604 | * if we didn't clear the block and the uiomove failed, |
| 605 | * the buf will now contain part of some other file, |
| 606 | * so we need to invalidate it. |
| 607 | */ |
| 608 | if (error && (flags & B_CLRBUF) == 0) { |
| 609 | brelse(bp, BC_INVAL); |
| 610 | break; |
| 611 | } |
| 612 | #ifdef LFS_READWRITE |
| 613 | (void)VOP_BWRITE(bp->b_vp, bp); |
| 614 | lfs_reserve(fs, vp, NULL, |
| 615 | -btofsb(fs, (UFS_NIADDR + 1) << fs->lfs_bshift)); |
| 616 | need_unreserve = false; |
| 617 | #else |
| 618 | if (ioflag & IO_SYNC) |
| 619 | (void)bwrite(bp); |
| 620 | else if (xfersize + blkoffset == fs->fs_bsize) |
| 621 | bawrite(bp); |
| 622 | else |
| 623 | bdwrite(bp); |
| 624 | #endif |
| 625 | if (error || xfersize == 0) |
| 626 | break; |
| 627 | } |
| 628 | #ifdef LFS_READWRITE |
| 629 | if (need_unreserve) { |
| 630 | lfs_reserve(fs, vp, NULL, |
| 631 | -btofsb(fs, (UFS_NIADDR + 1) << fs->lfs_bshift)); |
| 632 | } |
| 633 | #endif |
| 634 | |
| 635 | error = ufs_post_write_update(vp, uio, ioflag, cred, osize, resid, |
| 636 | extended, error); |
| 637 | fstrans_done(vp->v_mount); |
| 638 | |
| 639 | return (error); |
| 640 | } |
| 641 | |
| 642 | static int |
| 643 | ufs_post_write_update(struct vnode *vp, struct uio *uio, int ioflag, |
| 644 | kauth_cred_t cred, off_t osize, int resid, int extended, int oerror) |
| 645 | { |
| 646 | struct inode *ip = VTOI(vp); |
| 647 | int error = oerror; |
| 648 | |
| 649 | /* Trigger ctime and mtime updates, and atime if MNT_RELATIME. */ |
| 650 | ip->i_flag |= IN_CHANGE | IN_UPDATE; |
| 651 | if (vp->v_mount->mnt_flag & MNT_RELATIME) |
| 652 | ip->i_flag |= IN_ACCESS; |
| 653 | |
| 654 | /* |
| 655 | * If we successfully wrote any data and we are not the superuser, |
| 656 | * we clear the setuid and setgid bits as a precaution against |
| 657 | * tampering. |
| 658 | */ |
| 659 | if (resid > uio->uio_resid && cred) { |
| 660 | if (ip->i_mode & ISUID) { |
| 661 | if (kauth_authorize_vnode(cred, |
| 662 | KAUTH_VNODE_RETAIN_SUID, vp, NULL, EPERM) != 0) { |
| 663 | ip->i_mode &= ~ISUID; |
| 664 | DIP_ASSIGN(ip, mode, ip->i_mode); |
| 665 | } |
| 666 | } |
| 667 | |
| 668 | if (ip->i_mode & ISGID) { |
| 669 | if (kauth_authorize_vnode(cred, |
| 670 | KAUTH_VNODE_RETAIN_SGID, vp, NULL, EPERM) != 0) { |
| 671 | ip->i_mode &= ~ISGID; |
| 672 | DIP_ASSIGN(ip, mode, ip->i_mode); |
| 673 | } |
| 674 | } |
| 675 | } |
| 676 | |
| 677 | /* If we successfully wrote anything, notify kevent listeners. */ |
| 678 | if (resid > uio->uio_resid) |
| 679 | VN_KNOTE(vp, NOTE_WRITE | (extended ? NOTE_EXTEND : 0)); |
| 680 | |
| 681 | /* |
| 682 | * Update the size on disk: truncate back to original size on |
| 683 | * error, or reflect the new size on success. |
| 684 | */ |
| 685 | if (error) { |
| 686 | (void) UFS_TRUNCATE(vp, osize, ioflag & IO_SYNC, cred); |
| 687 | uio->uio_offset -= resid - uio->uio_resid; |
| 688 | uio->uio_resid = resid; |
| 689 | } else if (resid > uio->uio_resid && (ioflag & IO_SYNC) == IO_SYNC) |
| 690 | error = UFS_UPDATE(vp, NULL, NULL, UPDATE_WAIT); |
| 691 | else |
| 692 | UFS_WAPBL_UPDATE(vp, NULL, NULL, 0); |
| 693 | |
| 694 | /* Make sure the vnode uvm size matches the inode file size. */ |
| 695 | KASSERT(vp->v_size == ip->i_size); |
| 696 | |
| 697 | /* Write error overrides any inode update error. */ |
| 698 | if (oerror) |
| 699 | error = oerror; |
| 700 | return error; |
| 701 | } |
| 702 | |