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 1.40 2007/05/09 00:53:35 dillon 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/linker.h>
  52 #include <sys/fcntl.h>
  53
  54 #include <vm/vm.h>
  55 #include <vm/vm_object.h>
  56 #include <vm/vm_page.h>
  57 #include <vm/vm_pager.h>
  58 #include <vm/vnode_pager.h>
  59
  60 #include <sys/buf2.h>
  61 #include <sys/thread2.h>
  62
  63 #include <vfs/ufs/quota.h>
  64 #include <vfs/ufs/inode.h>
  65 #include <vfs/ufs/ufsmount.h>
  66 #include <vfs/ufs/ufs_extern.h>
  67 #include <vfs/ufs/fs.h>
  68 #include <vfs/ufs/ffs_extern.h>
  69
  70 #include "mfsnode.h"
  71 #include "mfs_extern.h"
  72
  73 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part");
  74
  75 extern struct vop_ops *mfs_vnode_vops_p;
  76
  77 static int      mfs_mount (struct mount *mp,
  78                         char *path, caddr_t data, struct ucred *td);
  79 static int      mfs_start (struct mount *mp, int flags);
  80 static int      mfs_statfs (struct mount *mp, struct statfs *sbp,
  81                         struct ucred *cred);
  82 static int      mfs_init (struct vfsconf *);
  83
  84 d_open_t        mfsopen;
  85 d_close_t       mfsclose;
  86 d_strategy_t    mfsstrategy;
  87
  88 static struct dev_ops mfs_ops = {
  89         { "MFS", -1, D_DISK },
  90         .d_open =       mfsopen,
  91         .d_close =      mfsclose,
  92         .d_read =       physread,
  93         .d_write =      physwrite,
  94         .d_strategy =   mfsstrategy,
  95 };
  96
  97 /*
  98  * mfs vfs operations.
  99  */
 100 static struct vfsops mfs_vfsops = {
 101         .vfs_mount =            mfs_mount,
 102         .vfs_start =            mfs_start,
 103         .vfs_unmount =          ffs_unmount,
 104         .vfs_root =             ufs_root,
 105         .vfs_quotactl =         ufs_quotactl,
 106         .vfs_statfs =           mfs_statfs,
 107         .vfs_sync =             ffs_sync,
 108         .vfs_vget =             ffs_vget,
 109         .vfs_fhtovp =           ffs_fhtovp,
 110         .vfs_checkexp =         ufs_check_export,
 111         .vfs_vptofh =           ffs_vptofh,
 112         .vfs_init =             mfs_init
 113 };
 114
 115 VFS_SET(mfs_vfsops, mfs, 0);
 116
 117 /*
 118  * We allow the underlying MFS block device to be opened and read.
 119  */
 120 int
 121 mfsopen(struct dev_open_args *ap)
 122 {
 123         cdev_t dev = ap->a_head.a_dev;
 124
 125         if (ap->a_oflags & FWRITE)
 126                 return(EROFS);
 127         if (dev->si_drv1)
 128                 return(0);
 129         return(ENXIO);
 130 }
 131
 132 int
 133 mfsclose(struct dev_close_args *ap)
 134 {
 135         return(0);
 136 }
 137
 138 int
 139 mfsstrategy(struct dev_strategy_args *ap)
 140 {
 141         cdev_t dev = ap->a_head.a_dev;
 142         struct bio *bio = ap->a_bio;
 143         struct buf *bp = bio->bio_buf;
 144         off_t boff = bio->bio_offset;
 145         off_t eoff = boff + bp->b_bcount;
 146         struct mfsnode *mfsp;
 147
 148         if ((mfsp = dev->si_drv1) == NULL) {
 149                 bp->b_error = ENXIO;
 150                 goto error;
 151         }
 152         if (boff < 0)
 153                 goto bad;
 154         if (eoff > mfsp->mfs_size) {
 155                 if (boff > mfsp->mfs_size || (bp->b_flags & B_BNOCLIP))
 156                         goto bad;
 157                 /*
 158                  * Return EOF by completing the I/O with 0 bytes transfered.
 159                  * Set B_INVAL to indicate that any data in the buffer is not
 160                  * valid.
 161                  */
 162                 if (boff == mfsp->mfs_size) {
 163                         bp->b_resid = bp->b_bcount;
 164                         bp->b_flags |= B_INVAL;
 165                         goto done;
 166                 }
 167                 bp->b_bcount = mfsp->mfs_size - boff;
 168         }
 169
 170         /*
 171          * Initiate I/O
 172          */
 173         bioq_insert_tail(&mfsp->bio_queue, bio);
 174         wakeup((caddr_t)mfsp);
 175         return(0);
 176
 177         /*
 178          * Failure conditions on bio
 179          */
 180 bad:
 181         bp->b_error = EINVAL;
 182 error:
 183         bp->b_flags |= B_ERROR | B_INVAL;
 184 done:
 185         biodone(bio);
 186         return(0);
 187 }
 188
 189 /*
 190  * mfs_mount
 191  *
 192  * Called when mounting local physical media
 193  *
 194  * PARAMETERS:
 195  *              mountroot
 196  *                      mp      mount point structure
 197  *                      path    NULL (flag for root mount!!!)
 198  *                      data    <unused>
 199  *                      ndp     <unused>
 200  *                      p       process (user credentials check [statfs])
 201  *
 202  *              mount
 203  *                      mp      mount point structure
 204  *                      path    path to mount point
 205  *                      data    pointer to argument struct in user space
 206  *                      ndp     mount point namei() return (used for
 207  *                              credentials on reload), reused to look
 208  *                              up block device.
 209  *                      p       process (user credentials check)
 210  *
 211  * RETURNS:     0       Success
 212  *              !0      error number (errno.h)
 213  *
 214  * LOCK STATE:
 215  *
 216  *              ENTRY
 217  *                      mount point is locked
 218  *              EXIT
 219  *                      mount point is locked
 220  *
 221  * NOTES:
 222  *              A NULL path can be used for a flag since the mount
 223  *              system call will fail with EFAULT in copyinstr in
 224  *              namei() if it is a genuine NULL from the user.
 225  */
 226 /* ARGSUSED */
 227 static int
 228 mfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
 229 {
 230         struct vnode *devvp;
 231         struct mfs_args args;
 232         struct ufsmount *ump;
 233         struct fs *fs;
 234         struct mfsnode *mfsp;
 235         size_t size;
 236         int flags, err;
 237         int minnum;
 238         cdev_t dev;
 239
 240         /*
 241          * Use NULL path to flag a root mount
 242          */
 243         if( path == NULL) {
 244                 /*
 245                  ***
 246                  * Mounting root file system
 247                  ***
 248                  */
 249
 250                 /* you lose */
 251                 panic("mfs_mount: mount MFS as root: not configured!");
 252         }
 253
 254         /*
 255          ***
 256          * Mounting non-root file system or updating a file system
 257          ***
 258          */
 259
 260         /* copy in user arguments*/
 261         if ((err = copyin(data, (caddr_t)&args, sizeof (struct mfs_args))) != 0)
 262                 goto error_1;
 263
 264         /*
 265          * If updating, check whether changing from read-only to
 266          * read/write; if there is no device name, that's all we do.
 267          */
 268         if (mp->mnt_flag & MNT_UPDATE) {
 269                 /*
 270                  ********************
 271                  * UPDATE
 272                  ********************
 273                  */
 274                 ump = VFSTOUFS(mp);
 275                 fs = ump->um_fs;
 276                 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
 277                         flags = WRITECLOSE;
 278                         if (mp->mnt_flag & MNT_FORCE)
 279                                 flags |= FORCECLOSE;
 280                         err = ffs_flushfiles(mp, flags);
 281                         if (err)
 282                                 goto error_1;
 283                 }
 284                 if (fs->fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
 285                         /* XXX reopen the device vnode read-write */
 286                         fs->fs_ronly = 0;
 287                 }
 288                 /* if not updating name...*/
 289                 if (args.fspec == 0) {
 290                         /*
 291                          * Process export requests.  Jumping to "success"
 292                          * will return the vfs_export() error code.
 293                          */
 294                         err = vfs_export(mp, &ump->um_export, &args.export);
 295                         goto success;
 296                 }
 297
 298                 /* XXX MFS does not support name updating*/
 299                 goto success;
 300         }
 301         /*
 302          * Do the MALLOC before the getnewvnode since doing so afterward
 303          * might cause a bogus v_data pointer to get dereferenced
 304          * elsewhere if MALLOC should block.
 305          */
 306         MALLOC(mfsp, struct mfsnode *, sizeof *mfsp, M_MFSNODE, M_WAITOK);
 307
 308         err = getspecialvnode(VT_MFS, NULL, &mfs_vnode_vops_p, &devvp, 0, 0);
 309         if (err) {
 310                 FREE(mfsp, M_MFSNODE);
 311                 goto error_1;
 312         }
 313
 314         minnum = (curproc->p_pid & 0xFF) |
 315                 ((curproc->p_pid & ~0xFF) << 8);
 316
 317         devvp->v_type = VCHR;
 318         dev = make_dev(&mfs_ops, minnum, UID_ROOT, GID_WHEEL, 0600,
 319                         "MFS%d", minnum >> 16);
 320         /* It is not clear that these will get initialized otherwise */
 321         dev->si_bsize_phys = DEV_BSIZE;
 322         dev->si_iosize_max = DFLTPHYS;
 323         dev->si_drv1 = mfsp;
 324         addaliasu(devvp, mfs_ops.head.maj, minnum);
 325         devvp->v_data = mfsp;
 326         mfsp->mfs_baseoff = args.base;
 327         mfsp->mfs_size = args.size;
 328         mfsp->mfs_vnode = devvp;
 329         mfsp->mfs_dev = reference_dev(dev);
 330         mfsp->mfs_td = curthread;
 331         mfsp->mfs_active = 1;
 332         bioq_init(&mfsp->bio_queue);
 333
 334         /*
 335          * Our 'block' device must be backed by a VM object.  Theoretically
 336          * we could use the anonymous memory VM object supplied by userland,
 337          * but it would be somewhat of a complex task to deal with it
 338          * that way since it would result in I/O requests which supply
 339          * the VM pages from our own object.
 340          *
 341          * vnode_pager_alloc() is typically called when a VM object is
 342          * being referenced externally.  We have to undo the refs for
 343          * the self reference between vnode and object.
 344          */
 345         vnode_pager_alloc(devvp, args.size, 0, 0);
 346         vrele(devvp);
 347         --devvp->v_object->ref_count;
 348
 349         /* Save "mounted from" info for mount point (NULL pad)*/
 350         copyinstr(      args.fspec,                     /* device name*/
 351                         mp->mnt_stat.f_mntfromname,     /* save area*/
 352                         MNAMELEN - 1,                   /* max size*/
 353                         &size);                         /* real size*/
 354         bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
 355
 356         vx_unlock(devvp);
 357         if ((err = ffs_mountfs(devvp, mp, M_MFSNODE)) != 0) {
 358                 mfsp->mfs_active = 0;
 359                 goto error_2;
 360         }
 361
 362         /*
 363          * Initialize FS stat information in mount struct; uses
 364          * mp->mnt_stat.f_mntfromname.
 365          *
 366          * This code is common to root and non-root mounts
 367          */
 368         VFS_STATFS(mp, &mp->mnt_stat, cred);
 369
 370         goto success;
 371
 372 error_2:        /* error with devvp held*/
 373
 374         /* release devvp before failing*/
 375         vrele(devvp);
 376
 377 error_1:        /* no state to back out*/
 378
 379 success:
 380         return( err);
 381 }
 382
 383 /*
 384  * Used to grab the process and keep it in the kernel to service
 385  * memory filesystem I/O requests.
 386  *
 387  * Loop servicing I/O requests.
 388  * Copy the requested data into or out of the memory filesystem
 389  * address space.
 390  */
 391 /* ARGSUSED */
 392 static int
 393 mfs_start(struct mount *mp, int flags)
 394 {
 395         struct vnode *vp = VFSTOUFS(mp)->um_devvp;
 396         struct mfsnode *mfsp = VTOMFS(vp);
 397         struct bio *bio;
 398         struct buf *bp;
 399         int gotsig = 0, sig;
 400         thread_t td = curthread;
 401
 402         /*
 403          * We must prevent the system from trying to swap
 404          * out or kill ( when swap space is low, see vm/pageout.c ) the
 405          * process.  A deadlock can occur if the process is swapped out,
 406          * and the system can loop trying to kill the unkillable ( while
 407          * references exist ) MFS process when swap space is low.
 408          */
 409         KKASSERT(curproc);
 410         PHOLD(curproc);
 411
 412         mfsp->mfs_td = td;
 413
 414         while (mfsp->mfs_active) {
 415                 crit_enter();
 416
 417                 while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
 418                         bioq_remove(&mfsp->bio_queue, bio);
 419                         crit_exit();
 420                         bp = bio->bio_buf;
 421                         mfs_doio(bio, mfsp);
 422                         wakeup(bp);
 423                         crit_enter();
 424                 }
 425
 426                 crit_exit();
 427
 428                 /*
 429                  * If a non-ignored signal is received, try to unmount.
 430                  * If that fails, clear the signal (it has been "processed"),
 431                  * otherwise we will loop here, as tsleep will always return
 432                  * EINTR/ERESTART.
 433                  */
 434                 /*
 435                  * Note that dounmount() may fail if work was queued after
 436                  * we slept. We have to jump hoops here to make sure that we
 437                  * process any buffers after the sleep, before we dounmount()
 438                  */
 439                 if (gotsig) {
 440                         gotsig = 0;
 441                         if (dounmount(mp, 0) != 0) {
 442                                 KKASSERT(td->td_proc);
 443                                 sig = CURSIG(td->td_lwp);
 444                                 if (sig)
 445                                         lwp_delsig(td->td_lwp, sig);
 446                         }
 447                 }
 448                 else if (tsleep((caddr_t)mfsp, PCATCH, "mfsidl", 0))
 449                         gotsig++;       /* try to unmount in next pass */
 450         }
 451         PRELE(curproc);
 452         v_release_rdev(vp);     /* hack because we do not implement CLOSE */
 453         /* XXX destroy/release devvp */
 454         return (0);
 455 }
 456
 457 /*
 458  * Get file system statistics.
 459  */
 460 static int
 461 mfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
 462 {
 463         int error;
 464
 465         error = ffs_statfs(mp, sbp, cred);
 466         sbp->f_type = mp->mnt_vfc->vfc_typenum;
 467         return (error);
 468 }
 469
 470 /*
 471  * Memory based filesystem initialization.
 472  */
 473 static int
 474 mfs_init(struct vfsconf *vfsp)
 475 {
 476         dev_ops_add(&mfs_ops, 0, 0);
 477         return (0);
 478 }