1 /* 2 * Copyright (c) 1989, 1990, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)mfs_vfsops.c 8.11 (Berkeley) 6/19/95 34 * $FreeBSD: src/sys/ufs/mfs/mfs_vfsops.c,v 1.81.2.3 2001/07/04 17:35:21 tegge Exp $ 35 * $DragonFly: src/sys/vfs/mfs/mfs_vfsops.c,v 1.32 2006/05/11 08:23:20 swildner Exp $ 36 */ 37 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/conf.h> 42 #include <sys/kernel.h> 43 #include <sys/proc.h> 44 #include <sys/buf.h> 45 #include <sys/mount.h> 46 #include <sys/signalvar.h> 47 #include <sys/vnode.h> 48 #include <sys/malloc.h> 49 #include <sys/linker.h> 50 #include <sys/fcntl.h> 51 52 #include <vm/vm.h> 53 #include <vm/vm_object.h> 54 #include <vm/vm_page.h> 55 #include <vm/vm_pager.h> 56 #include <vm/vnode_pager.h> 57 58 #include <sys/buf2.h> 59 #include <sys/thread2.h> 60 61 #include <vfs/ufs/quota.h> 62 #include <vfs/ufs/inode.h> 63 #include <vfs/ufs/ufsmount.h> 64 #include <vfs/ufs/ufs_extern.h> 65 #include <vfs/ufs/fs.h> 66 #include <vfs/ufs/ffs_extern.h> 67 68 #include "mfsnode.h" 69 #include "mfs_extern.h" 70 71 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part"); 72 73 74 extern struct vop_ops *mfs_vnode_vops; 75 76 static int mfs_mount (struct mount *mp, 77 char *path, caddr_t data, struct ucred *td); 78 static int mfs_start (struct mount *mp, int flags); 79 static int mfs_statfs (struct mount *mp, struct statfs *sbp, 80 struct ucred *cred); 81 static int mfs_init (struct vfsconf *); 82 83 d_open_t mfsopen; 84 d_close_t mfsclose; 85 d_strategy_t mfsstrategy; 86 87 #define MFS_CDEV_MAJOR 253 88 89 static struct cdevsw mfs_cdevsw = { 90 /* name */ "MFS", 91 /* maj */ MFS_CDEV_MAJOR, 92 /* flags */ D_DISK, 93 /* port */ NULL, 94 /* clone */ NULL, 95 96 /* open */ mfsopen, 97 /* close */ mfsclose, 98 /* read */ physread, 99 /* write */ physwrite, 100 /* ioctl */ noioctl, 101 /* poll */ nopoll, 102 /* mmap */ nommap, 103 /* strategy */ mfsstrategy, 104 /* dump */ nodump, 105 /* psize */ nopsize 106 }; 107 108 /* 109 * mfs vfs operations. 110 */ 111 static struct vfsops mfs_vfsops = { 112 .vfs_mount = mfs_mount, 113 .vfs_start = mfs_start, 114 .vfs_unmount = ffs_unmount, 115 .vfs_root = ufs_root, 116 .vfs_quotactl = ufs_quotactl, 117 .vfs_statfs = mfs_statfs, 118 .vfs_sync = ffs_sync, 119 .vfs_vget = ffs_vget, 120 .vfs_fhtovp = ffs_fhtovp, 121 .vfs_checkexp = ufs_check_export, 122 .vfs_vptofh = ffs_vptofh, 123 .vfs_init = mfs_init 124 }; 125 126 VFS_SET(mfs_vfsops, mfs, 0); 127 128 /* 129 * We allow the underlying MFS block device to be opened and read. 130 */ 131 int 132 mfsopen(dev_t dev, int flags, int mode, struct thread *td) 133 { 134 if (flags & FWRITE) 135 return(EROFS); 136 if (dev->si_drv1) 137 return(0); 138 return(ENXIO); 139 } 140 141 int 142 mfsclose(dev_t dev, int flags, int mode, struct thread *td) 143 { 144 return(0); 145 } 146 147 void 148 mfsstrategy(dev_t dev, struct bio *bio) 149 { 150 struct buf *bp = bio->bio_buf; 151 off_t boff = bio->bio_offset; 152 off_t eoff = boff + bp->b_bcount; 153 struct mfsnode *mfsp; 154 155 if ((mfsp = dev->si_drv1) == NULL) { 156 bp->b_error = ENXIO; 157 goto error; 158 } 159 if (boff < 0) 160 goto bad; 161 if (eoff > mfsp->mfs_size) { 162 if (boff > mfsp->mfs_size || (bp->b_flags & B_BNOCLIP)) 163 goto bad; 164 /* 165 * Return EOF by completing the I/O with 0 bytes transfered. 166 * Set B_INVAL to indicate that any data in the buffer is not 167 * valid. 168 */ 169 if (boff == mfsp->mfs_size) { 170 bp->b_resid = bp->b_bcount; 171 bp->b_flags |= B_INVAL; 172 goto done; 173 } 174 bp->b_bcount = mfsp->mfs_size - boff; 175 } 176 177 /* 178 * Initiate I/O 179 */ 180 bioq_insert_tail(&mfsp->bio_queue, bio); 181 wakeup((caddr_t)mfsp); 182 return; 183 184 /* 185 * Failure conditions on bio 186 */ 187 bad: 188 bp->b_error = EINVAL; 189 error: 190 bp->b_flags |= B_ERROR | B_INVAL; 191 done: 192 biodone(bio); 193 } 194 195 /* 196 * mfs_mount 197 * 198 * Called when mounting local physical media 199 * 200 * PARAMETERS: 201 * mountroot 202 * mp mount point structure 203 * path NULL (flag for root mount!!!) 204 * data <unused> 205 * ndp <unused> 206 * p process (user credentials check [statfs]) 207 * 208 * mount 209 * mp mount point structure 210 * path path to mount point 211 * data pointer to argument struct in user space 212 * ndp mount point namei() return (used for 213 * credentials on reload), reused to look 214 * up block device. 215 * p process (user credentials check) 216 * 217 * RETURNS: 0 Success 218 * !0 error number (errno.h) 219 * 220 * LOCK STATE: 221 * 222 * ENTRY 223 * mount point is locked 224 * EXIT 225 * mount point is locked 226 * 227 * NOTES: 228 * A NULL path can be used for a flag since the mount 229 * system call will fail with EFAULT in copyinstr in 230 * namei() if it is a genuine NULL from the user. 231 */ 232 /* ARGSUSED */ 233 static int 234 mfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred) 235 { 236 struct vnode *devvp; 237 struct mfs_args args; 238 struct ufsmount *ump; 239 struct fs *fs; 240 struct mfsnode *mfsp; 241 size_t size; 242 int flags, err; 243 int minnum; 244 dev_t dev; 245 246 /* 247 * Use NULL path to flag a root mount 248 */ 249 if( path == NULL) { 250 /* 251 *** 252 * Mounting root file system 253 *** 254 */ 255 256 /* you lose */ 257 panic("mfs_mount: mount MFS as root: not configured!"); 258 } 259 260 /* 261 *** 262 * Mounting non-root file system or updating a file system 263 *** 264 */ 265 266 /* copy in user arguments*/ 267 if ((err = copyin(data, (caddr_t)&args, sizeof (struct mfs_args))) != 0) 268 goto error_1; 269 270 /* 271 * If updating, check whether changing from read-only to 272 * read/write; if there is no device name, that's all we do. 273 */ 274 if (mp->mnt_flag & MNT_UPDATE) { 275 /* 276 ******************** 277 * UPDATE 278 ******************** 279 */ 280 ump = VFSTOUFS(mp); 281 fs = ump->um_fs; 282 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 283 flags = WRITECLOSE; 284 if (mp->mnt_flag & MNT_FORCE) 285 flags |= FORCECLOSE; 286 err = ffs_flushfiles(mp, flags); 287 if (err) 288 goto error_1; 289 } 290 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) { 291 /* XXX reopen the device vnode read-write */ 292 fs->fs_ronly = 0; 293 } 294 /* if not updating name...*/ 295 if (args.fspec == 0) { 296 /* 297 * Process export requests. Jumping to "success" 298 * will return the vfs_export() error code. 299 */ 300 err = vfs_export(mp, &ump->um_export, &args.export); 301 goto success; 302 } 303 304 /* XXX MFS does not support name updating*/ 305 goto success; 306 } 307 /* 308 * Do the MALLOC before the getnewvnode since doing so afterward 309 * might cause a bogus v_data pointer to get dereferenced 310 * elsewhere if MALLOC should block. 311 */ 312 MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE, M_WAITOK); 313 314 err = getspecialvnode(VT_MFS, NULL, &mfs_vnode_vops, &devvp, 0, 0); 315 if (err) { 316 FREE(mfsp, M_MFSNODE); 317 goto error_1; 318 } 319 320 minnum = (curproc->p_pid & 0xFF) | 321 ((curproc->p_pid & ~0xFF) << 8); 322 323 devvp->v_type = VCHR; 324 dev = make_dev(&mfs_cdevsw, minnum, UID_ROOT, GID_WHEEL, 0600, 325 "MFS%d", minnum >> 16); 326 /* It is not clear that these will get initialized otherwise */ 327 dev->si_bsize_phys = DEV_BSIZE; 328 dev->si_iosize_max = DFLTPHYS; 329 dev->si_drv1 = mfsp; 330 addaliasu(devvp, makeudev(MFS_CDEV_MAJOR, minnum)); 331 devvp->v_data = mfsp; 332 mfsp->mfs_baseoff = args.base; 333 mfsp->mfs_size = args.size; 334 mfsp->mfs_vnode = devvp; 335 mfsp->mfs_dev = reference_dev(dev); 336 mfsp->mfs_td = curthread; 337 mfsp->mfs_active = 1; 338 bioq_init(&mfsp->bio_queue); 339 340 /* 341 * Our 'block' device must be backed by a VM object. Theoretically 342 * we could use the anonymous memory VM object supplied by userland, 343 * but it would be somewhat of a complex task to deal with it 344 * that way since it would result in I/O requests which supply 345 * the VM pages from our own object. 346 * 347 * vnode_pager_alloc() is typically called when a VM object is 348 * being referenced externally. We have to undo the refs for 349 * the self reference between vnode and object. 350 */ 351 vnode_pager_alloc(devvp, args.size, 0, 0); 352 --devvp->v_usecount; 353 --devvp->v_object->ref_count; 354 355 /* Save "mounted from" info for mount point (NULL pad)*/ 356 copyinstr( args.fspec, /* device name*/ 357 mp->mnt_stat.f_mntfromname, /* save area*/ 358 MNAMELEN - 1, /* max size*/ 359 &size); /* real size*/ 360 bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size); 361 362 vx_unlock(devvp); 363 if ((err = ffs_mountfs(devvp, mp, M_MFSNODE)) != 0) { 364 mfsp->mfs_active = 0; 365 goto error_2; 366 } 367 368 /* 369 * Initialize FS stat information in mount struct; uses 370 * mp->mnt_stat.f_mntfromname. 371 * 372 * This code is common to root and non-root mounts 373 */ 374 VFS_STATFS(mp, &mp->mnt_stat, cred); 375 376 goto success; 377 378 error_2: /* error with devvp held*/ 379 380 /* release devvp before failing*/ 381 vrele(devvp); 382 383 error_1: /* no state to back out*/ 384 385 success: 386 return( err); 387 } 388 389 /* 390 * Used to grab the process and keep it in the kernel to service 391 * memory filesystem I/O requests. 392 * 393 * Loop servicing I/O requests. 394 * Copy the requested data into or out of the memory filesystem 395 * address space. 396 */ 397 /* ARGSUSED */ 398 static int 399 mfs_start(struct mount *mp, int flags) 400 { 401 struct vnode *vp = VFSTOUFS(mp)->um_devvp; 402 struct mfsnode *mfsp = VTOMFS(vp); 403 struct bio *bio; 404 struct buf *bp; 405 int gotsig = 0, sig; 406 thread_t td = curthread; 407 408 /* 409 * We must prevent the system from trying to swap 410 * out or kill ( when swap space is low, see vm/pageout.c ) the 411 * process. A deadlock can occur if the process is swapped out, 412 * and the system can loop trying to kill the unkillable ( while 413 * references exist ) MFS process when swap space is low. 414 */ 415 KKASSERT(curproc); 416 PHOLD(curproc); 417 418 mfsp->mfs_td = td; 419 420 while (mfsp->mfs_active) { 421 crit_enter(); 422 423 while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) { 424 bioq_remove(&mfsp->bio_queue, bio); 425 crit_exit(); 426 bp = bio->bio_buf; 427 mfs_doio(bio, mfsp); 428 wakeup(bp); 429 crit_enter(); 430 } 431 432 crit_exit(); 433 434 /* 435 * If a non-ignored signal is received, try to unmount. 436 * If that fails, clear the signal (it has been "processed"), 437 * otherwise we will loop here, as tsleep will always return 438 * EINTR/ERESTART. 439 */ 440 /* 441 * Note that dounmount() may fail if work was queued after 442 * we slept. We have to jump hoops here to make sure that we 443 * process any buffers after the sleep, before we dounmount() 444 */ 445 if (gotsig) { 446 gotsig = 0; 447 if (dounmount(mp, 0) != 0) { 448 KKASSERT(td->td_proc); 449 sig = CURSIG(td->td_proc); 450 if (sig) 451 SIGDELSET(td->td_proc->p_siglist, sig); 452 } 453 } 454 else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0)) 455 gotsig++; /* try to unmount in next pass */ 456 } 457 PRELE(curproc); 458 v_release_rdev(vp); /* hack because we do not implement CLOSE */ 459 /* XXX destroy/release devvp */ 460 return (0); 461 } 462 463 /* 464 * Get file system statistics. 465 */ 466 static int 467 mfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred) 468 { 469 int error; 470 471 error = ffs_statfs(mp, sbp, cred); 472 sbp->f_type = mp->mnt_vfc->vfc_typenum; 473 return (error); 474 } 475 476 /* 477 * Memory based filesystem initialization. 478 */ 479 static int 480 mfs_init(struct vfsconf *vfsp) 481 { 482 cdevsw_add(&mfs_cdevsw, 0, 0); 483 return (0); 484 } 485