sys/vfs/mfs/mfs_vfsops.c

   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 1.41 2008/07/26 22:31:54 mneumann Exp $
  36  */
  37
  38
  39 #include <sys/param.h>
  40 #include <sys/systm.h>
  41 #include <sys/conf.h>
  42 #include <sys/device.h>
  43 #include <sys/kernel.h>
  44 #include <sys/proc.h>
  45 #include <sys/buf.h>
  46 #include <sys/mount.h>
  47 #include <sys/signalvar.h>
  48 #include <sys/signal2.h>
  49 #include <sys/vnode.h>
  50 #include <sys/malloc.h>
  51 #include <sys/sysproto.h>
  52 #include <sys/mman.h>
  53 #include <sys/linker.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/nlookup.h>
  56 #include <sys/devfs.h>
  57
  58 #include <vm/vm.h>
  59 #include <vm/vm_object.h>
  60 #include <vm/vm_page.h>
  61 #include <vm/vm_pager.h>
  62 #include <vm/vnode_pager.h>
  63 #include <vm/vm_extern.h>
  64
  65 #include <sys/buf2.h>
  66 #include <sys/thread2.h>
  67
  68 #include <vfs/ufs/quota.h>
  69 #include <vfs/ufs/inode.h>
  70 #include <vfs/ufs/ufsmount.h>
  71 #include <vfs/ufs/ufs_extern.h>
  72 #include <vfs/ufs/fs.h>
  73 #include <vfs/ufs/ffs_extern.h>
  74
  75 #include "mfsnode.h"
  76 #include "mfs_extern.h"
  77
  78 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
  79
  80 static int      mfs_mount (struct mount *mp,
  81                         char *path, caddr_t data, struct ucred *td);
  82 static int      mfs_start (struct mount *mp, int flags);
  83 static int      mfs_statfs (struct mount *mp, struct statfs *sbp,
  84                         struct ucred *cred);
  85 static int      mfs_init (struct vfsconf *);
  86 static void     mfs_doio(struct bio *bio, struct mfsnode *mfsp);
  87
  88 d_open_t        mfsopen;
  89 d_close_t       mfsclose;
  90 d_strategy_t    mfsstrategy;
  91
  92 static struct dev_ops mfs_ops = {
  93         { "MFS", -1, D_DISK },
  94         .d_open =       mfsopen,
  95         .d_close =      mfsclose,
  96         .d_read =       physread,
  97         .d_write =      physwrite,
  98         .d_strategy =   mfsstrategy,
  99 };
 100
 101 /*
 102  * mfs vfs operations.
 103  */
 104 static struct vfsops mfs_vfsops = {
 105         .vfs_mount =            mfs_mount,
 106         .vfs_start =            mfs_start,
 107         .vfs_unmount =          ffs_unmount,
 108         .vfs_root =             ufs_root,
 109         .vfs_quotactl =         ufs_quotactl,
 110         .vfs_statfs =           mfs_statfs,
 111         .vfs_sync =             ffs_sync,
 112         .vfs_vget =             ffs_vget,
 113         .vfs_fhtovp =           ffs_fhtovp,
 114         .vfs_checkexp =         ufs_check_export,
 115         .vfs_vptofh =           ffs_vptofh,
 116         .vfs_init =             mfs_init
 117 };
 118
 119 VFS_SET(mfs_vfsops, mfs, 0);
 120
 121 /*
 122  * We allow the underlying MFS block device to be opened and read.
 123  */
 124 int
 125 mfsopen(struct dev_open_args *ap)
 126 {
 127         cdev_t dev = ap->a_head.a_dev;
 128
 129 #if 0
 130         if (ap->a_oflags & FWRITE)
 131                 return(EROFS);
 132 #endif
 133         if (dev->si_drv1)
 134                 return(0);
 135         return(ENXIO);
 136 }
 137
 138 int
 139 mfsclose(struct dev_close_args *ap)
 140 {
 141         cdev_t dev = ap->a_head.a_dev;
 142         struct mfsnode *mfsp;
 143
 144         if ((mfsp = dev->si_drv1) == NULL)
 145                 return(0);
 146         mfsp->mfs_active = 0;
 147         wakeup((caddr_t)mfsp);
 148         return(0);
 149 }
 150
 151 int
 152 mfsstrategy(struct dev_strategy_args *ap)
 153 {
 154         cdev_t dev = ap->a_head.a_dev;
 155         struct bio *bio = ap->a_bio;
 156         struct buf *bp = bio->bio_buf;
 157         off_t boff = bio->bio_offset;
 158         off_t eoff = boff + bp->b_bcount;
 159         struct mfsnode *mfsp;
 160
 161         if ((mfsp = dev->si_drv1) == NULL) {
 162                 bp->b_error = ENXIO;
 163                 goto error;
 164         }
 165         if (boff < 0)
 166                 goto bad;
 167         if (eoff > mfsp->mfs_size) {
 168                 if (boff > mfsp->mfs_size || (bp->b_flags & B_BNOCLIP))
 169                         goto bad;
 170                 /*
 171                  * Return EOF by completing the I/O with 0 bytes transfered.
 172                  * Set B_INVAL to indicate that any data in the buffer is not
 173                  * valid.
 174                  */
 175                 if (boff == mfsp->mfs_size) {
 176                         bp->b_resid = bp->b_bcount;
 177                         bp->b_flags |= B_INVAL;
 178                         goto done;
 179                 }
 180                 bp->b_bcount = mfsp->mfs_size - boff;
 181         }
 182
 183         /*
 184          * Initiate I/O
 185          */
 186         if (mfsp->mfs_td == curthread) {
 187                 mfs_doio(bio, mfsp);
 188         } else {
 189                 bioq_insert_tail(&mfsp->bio_queue, bio);
 190                 wakeup((caddr_t)mfsp);
 191         }
 192         return(0);
 193
 194         /*
 195          * Failure conditions on bio
 196          */
 197 bad:
 198         bp->b_error = EINVAL;
 199 error:
 200         bp->b_flags |= B_ERROR | B_INVAL;
 201 done:
 202         biodone(bio);
 203         return(0);
 204 }
 205
 206 /*
 207  * mfs_mount
 208  *
 209  * Called when mounting local physical media
 210  *
 211  * PARAMETERS:
 212  *              mountroot
 213  *                      mp      mount point structure
 214  *                      path    NULL (flag for root mount!!!)
 215  *                      data    <unused>
 216  *                      ndp     <unused>
 217  *                      p       process (user credentials check [statfs])
 218  *
 219  *              mount
 220  *                      mp      mount point structure
 221  *                      path    path to mount point
 222  *                      data    pointer to argument struct in user space
 223  *                      ndp     mount point namei() return (used for
 224  *                              credentials on reload), reused to look
 225  *                              up block device.
 226  *                      p       process (user credentials check)
 227  *
 228  * RETURNS:     0       Success
 229  *              !0      error number (errno.h)
 230  *
 231  * LOCK STATE:
 232  *
 233  *              ENTRY
 234  *                      mount point is locked
 235  *              EXIT
 236  *                      mount point is locked
 237  *
 238  * NOTES:
 239  *              A NULL path can be used for a flag since the mount
 240  *              system call will fail with EFAULT in copyinstr in
 241  *              namei() if it is a genuine NULL from the user.
 242  */
 243 /* ARGSUSED */
 244 static int
 245 mfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
 246 {
 247         struct vnode *devvp;
 248         struct mfs_args args;
 249         struct ufsmount *ump;
 250         struct fs *fs;
 251         struct mfsnode *mfsp;
 252         struct nlookupdata nd;
 253         size_t size;
 254         char devname[16];
 255         int flags;
 256         int minnum;
 257         int error;
 258         cdev_t dev;
 259
 260         /*
 261          * Use NULL path to flag a root mount
 262          */
 263         if (path == NULL) {
 264                 /*
 265                  ***
 266                  * Mounting root file system
 267                  ***
 268                  */
 269
 270                 /* you lose */
 271                 panic("mfs_mount: mount MFS as root: not configured!");
 272         }
 273
 274         mfsp = NULL;
 275
 276         /*
 277          ***
 278          * Mounting non-root file system or updating a file system
 279          ***
 280          */
 281
 282         /* copy in user arguments*/
 283         error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args));
 284         if (error)
 285                 goto error_1;
 286
 287         /*
 288          * If updating, check whether changing from read-only to
 289          * read/write; if there is no device name, that's all we do.
 290          */
 291         if (mp->mnt_flag & MNT_UPDATE) {
 292                 /*
 293                  ********************
 294                  * UPDATE
 295                  ********************
 296                  */
 297                 ump = VFSTOUFS(mp);
 298                 fs = ump->um_fs;
 299                 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
 300                         flags = WRITECLOSE;
 301                         if (mp->mnt_flag & MNT_FORCE)
 302                                 flags |= FORCECLOSE;
 303                         error = ffs_flushfiles(mp, flags);
 304                         if (error)
 305                                 goto error_1;
 306                 }
 307                 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
 308                         /* XXX reopen the device vnode read-write */
 309                         fs->fs_ronly = 0;
 310                 }
 311                 /* if not updating name...*/
 312                 if (args.fspec == 0) {
 313                         /*
 314                          * Process export requests.  Jumping to "success"
 315                          * will return the vfs_export() error code.
 316                          */
 317                         error = vfs_export(mp, &ump->um_export, &args.export);
 318                         goto success;
 319                 }
 320
 321                 /* XXX MFS does not support name updating*/
 322                 goto success;
 323         }
 324
 325         /*
 326          * Do the MALLOC before the make_dev since doing so afterward
 327          * might cause a bogus v_data pointer to get dereferenced
 328          * elsewhere if MALLOC should block.
 329          */
 330         MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE,
 331                M_WAITOK|M_ZERO);
 332
 333         minnum = (int)curproc->p_pid;
 334
 335         dev = make_dev(&mfs_ops, minnum, UID_ROOT, GID_WHEEL, 0600,
 336                        "mfs%d", minnum);
 337         /* It is not clear that these will get initialized otherwise */
 338         dev->si_bsize_phys = DEV_BSIZE;
 339         dev->si_iosize_max = DFLTPHYS;
 340         dev->si_drv1 = mfsp;
 341         mfsp->mfs_baseoff = args.base;
 342         mfsp->mfs_size = args.size;
 343         mfsp->mfs_dev = dev;
 344         mfsp->mfs_td = curthread;
 345         mfsp->mfs_active = 1;
 346         bioq_init(&mfsp->bio_queue);
 347
 348         devfs_config(); /* sync devfs work */
 349         ksnprintf(devname, sizeof(devname), "/dev/mfs%d", minnum);
 350         nlookup_init(&nd, devname, UIO_SYSSPACE, 0);
 351         devvp = NULL;
 352         error = nlookup(&nd);
 353         if (error == 0) {
 354                 devvp = nd.nl_nch.ncp->nc_vp;
 355                 if (devvp == NULL)
 356                         error = ENOENT;
 357                 error = vget(devvp, LK_SHARED);
 358         }
 359         nlookup_done(&nd);
 360
 361         if (error)
 362                 goto error_1;
 363         vn_unlock(devvp);
 364
 365         /*
 366          * Our 'block' device must be backed by a VM object.  Theoretically
 367          * we could use the anonymous memory VM object supplied by userland,
 368          * but it would be somewhat of a complex task to deal with it
 369          * that way since it would result in I/O requests which supply
 370          * the VM pages from our own object.
 371          *
 372          * vnode_pager_alloc() is typically called when a VM object is
 373          * being referenced externally.  We have to undo the refs for
 374          * the self reference between vnode and object.
 375          */
 376         vnode_pager_setsize(devvp, args.size);
 377
 378         /* Save "mounted from" info for mount point (NULL pad)*/
 379         copyinstr(args.fspec,                   /* device name*/
 380                   mp->mnt_stat.f_mntfromname,   /* save area*/
 381                   MNAMELEN - 1,                 /* max size*/
 382                   &size);                       /* real size*/
 383         bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
 384         /* vref is eaten by mount? */
 385
 386         error = ffs_mountfs(devvp, mp, M_MFSNODE);
 387         if (error) {
 388                 mfsp->mfs_active = 0;
 389                 goto error_2;
 390         }
 391
 392         /*
 393          * Initialize FS stat information in mount struct; uses
 394          * mp->mnt_stat.f_mntfromname.
 395          *
 396          * This code is common to root and non-root mounts
 397          */
 398         VFS_STATFS(mp, &mp->mnt_stat, cred);
 399
 400         goto success;
 401
 402 error_2:        /* error with devvp held*/
 403         vrele(devvp);
 404
 405 error_1:        /* no state to back out*/
 406         if (mfsp) {
 407                 if (mfsp->mfs_dev) {
 408                         destroy_dev(mfsp->mfs_dev);
 409                         mfsp->mfs_dev = NULL;
 410                 }
 411                 FREE(mfsp, M_MFSNODE);
 412         }
 413
 414 success:
 415         return(error);
 416 }
 417
 418 /*
 419  * Used to grab the process and keep it in the kernel to service
 420  * memory filesystem I/O requests.
 421  *
 422  * Loop servicing I/O requests.
 423  * Copy the requested data into or out of the memory filesystem
 424  * address space.
 425  */
 426 /* ARGSUSED */
 427 static int
 428 mfs_start(struct mount *mp, int flags)
 429 {
 430         struct vnode *vp = VFSTOUFS(mp)->um_devvp;
 431         struct mfsnode *mfsp = vp->v_rdev->si_drv1;
 432         struct bio *bio;
 433         struct buf *bp;
 434         int gotsig = 0, sig;
 435         thread_t td = curthread;
 436
 437         /*
 438          * We must prevent the system from trying to swap
 439          * out or kill ( when swap space is low, see vm/pageout.c ) the
 440          * process.  A deadlock can occur if the process is swapped out,
 441          * and the system can loop trying to kill the unkillable ( while
 442          * references exist ) MFS process when swap space is low.
 443          */
 444         KKASSERT(curproc);
 445         PHOLD(curproc);
 446
 447         mfsp->mfs_td = td;
 448
 449         while (mfsp->mfs_active) {
 450                 crit_enter();
 451
 452                 while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
 453                         bioq_remove(&mfsp->bio_queue, bio);
 454                         crit_exit();
 455                         bp = bio->bio_buf;
 456                         mfs_doio(bio, mfsp);
 457                         wakeup(bp);
 458                         crit_enter();
 459                 }
 460
 461                 crit_exit();
 462
 463                 /*
 464                  * If a non-ignored signal is received, try to unmount.
 465                  * If that fails, clear the signal (it has been "processed"),
 466                  * otherwise we will loop here, as tsleep will always return
 467                  * EINTR/ERESTART.
 468                  */
 469                 /*
 470                  * Note that dounmount() may fail if work was queued after
 471                  * we slept. We have to jump hoops here to make sure that we
 472                  * process any buffers after the sleep, before we dounmount()
 473                  */
 474                 if (gotsig) {
 475                         gotsig = 0;
 476                         if (dounmount(mp, 0) != 0) {
 477                                 KKASSERT(td->td_proc);
 478                                 sig = CURSIG(td->td_lwp);
 479                                 if (sig)
 480                                         lwp_delsig(td->td_lwp, sig);
 481                         }
 482                 }
 483                 else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
 484                         gotsig++;       /* try to unmount in next pass */
 485         }
 486         PRELE(curproc);
 487         if (mfsp->mfs_dev) {
 488                 destroy_dev(mfsp->mfs_dev);
 489                 mfsp->mfs_dev = NULL;
 490         }
 491         FREE(mfsp, M_MFSNODE);
 492         return (0);
 493 }
 494
 495 /*
 496  * Get file system statistics.
 497  */
 498 static int
 499 mfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
 500 {
 501         int error;
 502
 503         error = ffs_statfs(mp, sbp, cred);
 504         sbp->f_type = mp->mnt_vfc->vfc_typenum;
 505         return (error);
 506 }
 507
 508 /*
 509  * Memory based filesystem initialization.
 510  */
 511 static int
 512 mfs_init(struct vfsconf *vfsp)
 513 {
 514         return (0);
 515 }
 516
 517 /*
 518  * Memory file system I/O.
 519  *
 520  * Trivial on the HP since buffer has already been mapping into KVA space.
 521  *
 522  * Read and Write are handled with a simple copyin and copyout.
 523  *
 524  * We also partially support VOP_FREEBLKS().  We can't implement
 525  * completely -- for example, on fragments or inode metadata, but we can
 526  * implement it for page-aligned requests.
 527  */
 528 static void
 529 mfs_doio(struct bio *bio, struct mfsnode *mfsp)
 530 {
 531         struct buf *bp = bio->bio_buf;
 532         caddr_t base = mfsp->mfs_baseoff + bio->bio_offset;
 533         int bytes;
 534
 535         switch(bp->b_cmd) {
 536         case BUF_CMD_FREEBLKS:
 537                 /*
 538                  * Implement FREEBLKS, which allows the filesystem to tell
 539                  * a block device when blocks are no longer needed (like when
 540                  * a file is deleted).  We use the hook to MADV_FREE the VM.
 541                  * This makes an MFS filesystem work as well or better then
 542                  * a sun-style swap-mounted filesystem.
 543                  */
 544                 bytes = bp->b_bcount;
 545
 546                 if ((vm_offset_t)base & PAGE_MASK) {
 547                         int n = PAGE_SIZE - ((vm_offset_t)base & PAGE_MASK);
 548                         bytes -= n;
 549                         base += n;
 550                 }
 551                 if (bytes > 0) {
 552                         struct madvise_args uap;
 553
 554                         bytes &= ~PAGE_MASK;
 555                         if (bytes != 0) {
 556                                 bzero(&uap, sizeof(uap));
 557                                 uap.addr  = base;
 558                                 uap.len   = bytes;
 559                                 uap.behav = MADV_FREE;
 560                                 sys_madvise(&uap);
 561                         }
 562                 }
 563                 bp->b_error = 0;
 564                 break;
 565         case BUF_CMD_READ:
 566                 /*
 567                  * Read data from our 'memory' disk
 568                  */
 569                 bp->b_error = copyin(base, bp->b_data, bp->b_bcount);
 570                 break;
 571         case BUF_CMD_WRITE:
 572                 /*
 573                  * Write data to our 'memory' disk
 574                  */
 575                 bp->b_error = copyout(bp->b_data, base, bp->b_bcount);
 576                 break;
 577         default:
 578                 panic("mfs: bad b_cmd %d\n", bp->b_cmd);
 579         }
 580         if (bp->b_error)
 581                 bp->b_flags |= B_ERROR;
 582         biodone(bio);
 583 }