1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2000-2003 Tor Egge 5 * 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 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 #include <sys/systm.h> 34 #include <sys/fcntl.h> 35 #include <sys/file.h> 36 #include <sys/stat.h> 37 #include <sys/proc.h> 38 #include <sys/limits.h> 39 #include <sys/mount.h> 40 #include <sys/namei.h> 41 #include <sys/vnode.h> 42 #include <sys/conf.h> 43 #include <sys/filio.h> 44 #include <sys/ttycom.h> 45 #include <sys/bio.h> 46 #include <sys/buf.h> 47 #include <sys/rwlock.h> 48 #include <ufs/ufs/extattr.h> 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 #include <ufs/ffs/fs.h> 54 #include <ufs/ffs/ffs_extern.h> 55 56 #include <vm/vm.h> 57 #include <vm/vm_extern.h> 58 #include <vm/vm_object.h> 59 #include <sys/kernel.h> 60 #include <sys/sysctl.h> 61 62 static int ffs_rawread_readahead(struct vnode *vp, 63 caddr_t udata, 64 off_t offset, 65 size_t len, 66 struct thread *td, 67 struct buf *bp); 68 static int ffs_rawread_main(struct vnode *vp, 69 struct uio *uio); 70 71 static int ffs_rawread_sync(struct vnode *vp); 72 73 int ffs_rawread(struct vnode *vp, struct uio *uio, int *workdone); 74 75 SYSCTL_DECL(_vfs_ffs); 76 77 static uma_zone_t ffsraw_pbuf_zone; 78 79 static int allowrawread = 1; 80 SYSCTL_INT(_vfs_ffs, OID_AUTO, allowrawread, CTLFLAG_RW, &allowrawread, 0, 81 "Flag to enable raw reads"); 82 83 static int rawreadahead = 1; 84 SYSCTL_INT(_vfs_ffs, OID_AUTO, rawreadahead, CTLFLAG_RW, &rawreadahead, 0, 85 "Flag to enable readahead for long raw reads"); 86 87 static void 88 ffs_rawread_setup(void *arg __unused) 89 { 90 91 ffsraw_pbuf_zone = pbuf_zsecond_create("ffsrawpbuf", 92 (nswbuf > 100 ) ? (nswbuf - (nswbuf >> 4)) : nswbuf - 8); 93 } 94 SYSINIT(ffs_raw, SI_SUB_VM_CONF, SI_ORDER_ANY, ffs_rawread_setup, NULL); 95 96 static int 97 ffs_rawread_sync(struct vnode *vp) 98 { 99 int error; 100 int upgraded; 101 struct bufobj *bo; 102 struct mount *mp; 103 vm_object_t obj; 104 105 /* Check for dirty mmap, pending writes and dirty buffers */ 106 bo = &vp->v_bufobj; 107 BO_LOCK(bo); 108 VI_LOCK(vp); 109 if (bo->bo_numoutput > 0 || 110 bo->bo_dirty.bv_cnt > 0 || 111 ((obj = vp->v_object) != NULL && 112 vm_object_mightbedirty(obj))) { 113 VI_UNLOCK(vp); 114 BO_UNLOCK(bo); 115 116 if (vn_start_write(vp, &mp, V_NOWAIT) != 0) { 117 if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) 118 upgraded = 1; 119 else 120 upgraded = 0; 121 VOP_UNLOCK(vp); 122 (void) vn_start_write(vp, &mp, V_WAIT); 123 VOP_LOCK(vp, LK_EXCLUSIVE); 124 } else if (VOP_ISLOCKED(vp) != LK_EXCLUSIVE) { 125 upgraded = 1; 126 /* Upgrade to exclusive lock, this might block */ 127 VOP_LOCK(vp, LK_UPGRADE); 128 } else 129 upgraded = 0; 130 131 132 VI_LOCK(vp); 133 /* Check if vnode was reclaimed while unlocked. */ 134 if (VN_IS_DOOMED(vp)) { 135 VI_UNLOCK(vp); 136 if (upgraded != 0) 137 VOP_LOCK(vp, LK_DOWNGRADE); 138 vn_finished_write(mp); 139 return (EIO); 140 } 141 /* Attempt to msync mmap() regions to clean dirty mmap */ 142 if ((obj = vp->v_object) != NULL && 143 vm_object_mightbedirty(obj)) { 144 VI_UNLOCK(vp); 145 VM_OBJECT_WLOCK(obj); 146 vm_object_page_clean(obj, 0, 0, OBJPC_SYNC); 147 VM_OBJECT_WUNLOCK(obj); 148 } else 149 VI_UNLOCK(vp); 150 151 /* Wait for pending writes to complete */ 152 BO_LOCK(bo); 153 error = bufobj_wwait(&vp->v_bufobj, 0, 0); 154 if (error != 0) { 155 /* XXX: can't happen with a zero timeout ??? */ 156 BO_UNLOCK(bo); 157 if (upgraded != 0) 158 VOP_LOCK(vp, LK_DOWNGRADE); 159 vn_finished_write(mp); 160 return (error); 161 } 162 /* Flush dirty buffers */ 163 if (bo->bo_dirty.bv_cnt > 0) { 164 BO_UNLOCK(bo); 165 if ((error = ffs_syncvnode(vp, MNT_WAIT, 0)) != 0) { 166 if (upgraded != 0) 167 VOP_LOCK(vp, LK_DOWNGRADE); 168 vn_finished_write(mp); 169 return (error); 170 } 171 BO_LOCK(bo); 172 if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0) 173 panic("ffs_rawread_sync: dirty bufs"); 174 } 175 BO_UNLOCK(bo); 176 if (upgraded != 0) 177 VOP_LOCK(vp, LK_DOWNGRADE); 178 vn_finished_write(mp); 179 } else { 180 VI_UNLOCK(vp); 181 BO_UNLOCK(bo); 182 } 183 return 0; 184 } 185 186 static int 187 ffs_rawread_readahead(struct vnode *vp, 188 caddr_t udata, 189 off_t offset, 190 size_t len, 191 struct thread *td, 192 struct buf *bp) 193 { 194 int error; 195 u_int iolen; 196 off_t blockno; 197 int blockoff; 198 int bsize; 199 struct vnode *dp; 200 int bforwards; 201 struct inode *ip; 202 ufs2_daddr_t blkno; 203 204 bsize = vp->v_mount->mnt_stat.f_iosize; 205 206 ip = VTOI(vp); 207 dp = ITODEVVP(ip); 208 209 iolen = ((vm_offset_t) udata) & PAGE_MASK; 210 bp->b_bcount = len; 211 if (bp->b_bcount + iolen > bp->b_kvasize) { 212 bp->b_bcount = bp->b_kvasize; 213 if (iolen != 0) 214 bp->b_bcount -= PAGE_SIZE; 215 } 216 bp->b_flags = 0; /* XXX necessary ? */ 217 bp->b_iocmd = BIO_READ; 218 bp->b_iodone = bdone; 219 bp->b_data = udata; 220 blockno = offset / bsize; 221 blockoff = (offset % bsize) / DEV_BSIZE; 222 if ((daddr_t) blockno != blockno) { 223 return EINVAL; /* blockno overflow */ 224 } 225 226 bp->b_lblkno = bp->b_blkno = blockno; 227 228 error = ufs_bmaparray(vp, bp->b_lblkno, &blkno, NULL, &bforwards, NULL); 229 if (error != 0) 230 return error; 231 if (blkno == -1) { 232 /* Fill holes with NULs to preserve semantics */ 233 234 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize) 235 bp->b_bcount = bsize - blockoff * DEV_BSIZE; 236 bp->b_bufsize = bp->b_bcount; 237 238 if (vmapbuf(bp, 1) < 0) 239 return EFAULT; 240 241 maybe_yield(); 242 bzero(bp->b_data, bp->b_bufsize); 243 244 /* Mark operation completed (similar to bufdone()) */ 245 246 bp->b_resid = 0; 247 bp->b_flags |= B_DONE; 248 return 0; 249 } 250 bp->b_blkno = blkno + blockoff; 251 bp->b_offset = bp->b_iooffset = (blkno + blockoff) * DEV_BSIZE; 252 253 if (bp->b_bcount + blockoff * DEV_BSIZE > bsize * (1 + bforwards)) 254 bp->b_bcount = bsize * (1 + bforwards) - blockoff * DEV_BSIZE; 255 bp->b_bufsize = bp->b_bcount; 256 257 if (vmapbuf(bp, 1) < 0) 258 return EFAULT; 259 260 BO_STRATEGY(&dp->v_bufobj, bp); 261 return 0; 262 } 263 264 static int 265 ffs_rawread_main(struct vnode *vp, 266 struct uio *uio) 267 { 268 int error, nerror; 269 struct buf *bp, *nbp, *tbp; 270 u_int iolen; 271 caddr_t udata; 272 long resid; 273 off_t offset; 274 struct thread *td; 275 276 td = uio->uio_td ? uio->uio_td : curthread; 277 udata = uio->uio_iov->iov_base; 278 resid = uio->uio_resid; 279 offset = uio->uio_offset; 280 281 /* 282 * keep the process from being swapped 283 */ 284 PHOLD(td->td_proc); 285 286 error = 0; 287 nerror = 0; 288 289 bp = NULL; 290 nbp = NULL; 291 292 while (resid > 0) { 293 294 if (bp == NULL) { /* Setup first read */ 295 bp = uma_zalloc(ffsraw_pbuf_zone, M_WAITOK); 296 pbgetvp(vp, bp); 297 error = ffs_rawread_readahead(vp, udata, offset, 298 resid, td, bp); 299 if (error != 0) 300 break; 301 302 if (resid > bp->b_bufsize) { /* Setup fist readahead */ 303 if (rawreadahead != 0) 304 nbp = uma_zalloc(ffsraw_pbuf_zone, 305 M_NOWAIT); 306 else 307 nbp = NULL; 308 if (nbp != NULL) { 309 pbgetvp(vp, nbp); 310 311 nerror = ffs_rawread_readahead(vp, 312 udata + 313 bp->b_bufsize, 314 offset + 315 bp->b_bufsize, 316 resid - 317 bp->b_bufsize, 318 td, 319 nbp); 320 if (nerror) { 321 pbrelvp(nbp); 322 uma_zfree(ffsraw_pbuf_zone, 323 nbp); 324 nbp = NULL; 325 } 326 } 327 } 328 } 329 330 bwait(bp, PRIBIO, "rawrd"); 331 vunmapbuf(bp); 332 333 iolen = bp->b_bcount - bp->b_resid; 334 if (iolen == 0 && (bp->b_ioflags & BIO_ERROR) == 0) { 335 nerror = 0; /* Ignore possible beyond EOF error */ 336 break; /* EOF */ 337 } 338 339 if ((bp->b_ioflags & BIO_ERROR) != 0) { 340 error = bp->b_error; 341 break; 342 } 343 resid -= iolen; 344 udata += iolen; 345 offset += iolen; 346 if (iolen < bp->b_bufsize) { 347 /* Incomplete read. Try to read remaining part */ 348 error = ffs_rawread_readahead(vp, 349 udata, 350 offset, 351 bp->b_bufsize - iolen, 352 td, 353 bp); 354 if (error != 0) 355 break; 356 } else if (nbp != NULL) { /* Complete read with readahead */ 357 358 tbp = bp; 359 bp = nbp; 360 nbp = tbp; 361 362 if (resid <= bp->b_bufsize) { /* No more readaheads */ 363 pbrelvp(nbp); 364 uma_zfree(ffsraw_pbuf_zone, nbp); 365 nbp = NULL; 366 } else { /* Setup next readahead */ 367 nerror = ffs_rawread_readahead(vp, 368 udata + 369 bp->b_bufsize, 370 offset + 371 bp->b_bufsize, 372 resid - 373 bp->b_bufsize, 374 td, 375 nbp); 376 if (nerror != 0) { 377 pbrelvp(nbp); 378 uma_zfree(ffsraw_pbuf_zone, nbp); 379 nbp = NULL; 380 } 381 } 382 } else if (nerror != 0) {/* Deferred Readahead error */ 383 break; 384 } else if (resid > 0) { /* More to read, no readahead */ 385 error = ffs_rawread_readahead(vp, udata, offset, 386 resid, td, bp); 387 if (error != 0) 388 break; 389 } 390 } 391 392 if (bp != NULL) { 393 pbrelvp(bp); 394 uma_zfree(ffsraw_pbuf_zone, bp); 395 } 396 if (nbp != NULL) { /* Run down readahead buffer */ 397 bwait(nbp, PRIBIO, "rawrd"); 398 vunmapbuf(nbp); 399 pbrelvp(nbp); 400 uma_zfree(ffsraw_pbuf_zone, nbp); 401 } 402 403 if (error == 0) 404 error = nerror; 405 PRELE(td->td_proc); 406 uio->uio_iov->iov_base = udata; 407 uio->uio_resid = resid; 408 uio->uio_offset = offset; 409 return error; 410 } 411 412 int 413 ffs_rawread(struct vnode *vp, 414 struct uio *uio, 415 int *workdone) 416 { 417 if (allowrawread != 0 && 418 uio->uio_iovcnt == 1 && 419 uio->uio_segflg == UIO_USERSPACE && 420 uio->uio_resid == uio->uio_iov->iov_len && 421 (((uio->uio_td != NULL) ? uio->uio_td : curthread)->td_pflags & 422 TDP_DEADLKTREAT) == 0) { 423 int secsize; /* Media sector size */ 424 off_t filebytes; /* Bytes left of file */ 425 int blockbytes; /* Bytes left of file in full blocks */ 426 int partialbytes; /* Bytes in last partial block */ 427 int skipbytes; /* Bytes not to read in ffs_rawread */ 428 struct inode *ip; 429 int error; 430 431 432 /* Only handle sector aligned reads */ 433 ip = VTOI(vp); 434 secsize = ITODEVVP(ip)->v_bufobj.bo_bsize; 435 if ((uio->uio_offset & (secsize - 1)) == 0 && 436 (uio->uio_resid & (secsize - 1)) == 0) { 437 438 /* Sync dirty pages and buffers if needed */ 439 error = ffs_rawread_sync(vp); 440 if (error != 0) 441 return error; 442 443 /* Check for end of file */ 444 if (ip->i_size > uio->uio_offset) { 445 filebytes = ip->i_size - uio->uio_offset; 446 447 /* No special eof handling needed ? */ 448 if (uio->uio_resid <= filebytes) { 449 *workdone = 1; 450 return ffs_rawread_main(vp, uio); 451 } 452 453 partialbytes = ((unsigned int) ip->i_size) % 454 ITOFS(ip)->fs_bsize; 455 blockbytes = (int) filebytes - partialbytes; 456 if (blockbytes > 0) { 457 skipbytes = uio->uio_resid - 458 blockbytes; 459 uio->uio_resid = blockbytes; 460 error = ffs_rawread_main(vp, uio); 461 uio->uio_resid += skipbytes; 462 if (error != 0) 463 return error; 464 /* Read remaining part using buffer */ 465 } 466 } 467 } 468 } 469 *workdone = 0; 470 return 0; 471 } 472