kernel/fs/ctfs/ctfs_root.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
  23  */
  24
  25 #include <sys/modctl.h>
  26 #include <sys/types.h>
  27 #include <sys/param.h>
  28 #include <sys/time.h>
  29 #include <sys/cred.h>
  30 #include <sys/vfs.h>
  31 #include <sys/gfs.h>
  32 #include <sys/vnode.h>
  33 #include <sys/systm.h>
  34 #include <sys/cmn_err.h>
  35 #include <sys/errno.h>
  36 #include <sys/sysmacros.h>
  37 #include <sys/policy.h>
  38 #include <sys/mount.h>
  39 #include <sys/pathname.h>
  40 #include <sys/dirent.h>
  41 #include <sys/fs_subr.h>
  42 #include <sys/contract.h>
  43 #include <sys/contract_impl.h>
  44 #include <sys/ctfs.h>
  45 #include <sys/ctfs_impl.h>
  46 #include <sys/uio.h>
  47 #include <sys/file.h>
  48 #include <sys/atomic.h>
  49 #include <sys/sunddi.h>
  50
  51 /*
  52  * ctfs, the contract filesystem.
  53  *
  54  * Exposes the construct subsystem to userland.  The structure of the
  55  * filesystem is a public interface, but the behavior of the files is
  56  * private and unstable.  Contract consumers are expected to use
  57  * libcontract(3lib) to operate on ctfs file descriptors.
  58  *
  59  * We're trying something a little different here.  Rather than make
  60  * each vnode op itself call into a vector of file type operations, we
  61  * actually use different vnode types (gasp!), the implementations of
  62  * which may call into routines providing common functionality.  This
  63  * design should hopefully make it easier to factor and maintain the
  64  * code.  For the most part, there is a separate file for each vnode
  65  * type's implementation.  The exceptions to this are the ctl/stat
  66  * nodes, which are very similar, and the three event endpoint types.
  67  *
  68  * This file contains common routines used by some or all of the vnode
  69  * types, the filesystem's module linkage and VFS operations, and the
  70  * implementation of the root vnode.
  71  */
  72
  73 static const struct vfsops ctfs_vfsops;
  74
  75 static int ctfs_init(int, char *);
  76
  77 static ino64_t ctfs_root_do_inode(vnode_t *, int);
  78
  79
  80 /*
  81  * File system module linkage
  82  */
  83 static mntopts_t ctfs_mntopts = {
  84         0,
  85         NULL
  86 };
  87
  88 static vfsdef_t vfw = {
  89         VFSDEF_VERSION,
  90         "ctfs",
  91         ctfs_init,
  92         VSW_HASPROTO|VSW_ZMOUNT,
  93         &ctfs_mntopts,
  94 };
  95
  96 extern struct mod_ops mod_fsops;
  97
  98 static struct modlfs modlfs = {
  99         &mod_fsops, "contract filesystem", &vfw
 100 };
 101
 102 static struct modlinkage modlinkage = {
 103         MODREV_1, (void *)&modlfs, NULL
 104 };
 105
 106 int
 107 _init(void)
 108 {
 109         return (mod_install(&modlinkage));
 110 }
 111
 112 int
 113 _info(struct modinfo *modinfop)
 114 {
 115         return (mod_info(&modlinkage, modinfop));
 116 }
 117
 118 int
 119 _fini(void)
 120 {
 121         /*
 122          * As unloading filesystem modules isn't completely safe, we
 123          * don't allow it.
 124          */
 125         return (EBUSY);
 126 }
 127
 128 static int ctfs_fstype;
 129 static major_t ctfs_major;
 130 static minor_t ctfs_minor = 0;
 131
 132 /*
 133  * ctfs_init - the vfsdef_t init entry point
 134  *
 135  * Sets the VFS ops, builds all the vnode ops, and allocates a device
 136  * number.
 137  */
 138 /* ARGSUSED */
 139 static int
 140 ctfs_init(int fstype, char *name)
 141 {
 142         int error;
 143
 144         ctfs_fstype = fstype;
 145         if (error = vfs_setfsops(fstype, &ctfs_vfsops)) {
 146                 cmn_err(CE_WARN, "ctfs_init: bad fstype");
 147                 return (error);
 148         }
 149
 150         if ((ctfs_major = getudev()) == (major_t)-1) {
 151                 cmn_err(CE_WARN, "ctfs_init: can't get unique device number");
 152                 ctfs_major = 0;
 153         }
 154
 155         return (0);
 156 }
 157
 158 /*
 159  * ctfs_mount - the VFS_MOUNT entry point
 160  */
 161 static int
 162 ctfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
 163 {
 164         ctfs_vfs_t *data;
 165         dev_t dev;
 166         gfs_dirent_t *dirent;
 167         int i;
 168
 169         if (secpolicy_fs_mount(cr, mvp, vfsp) != 0)
 170                 return (EPERM);
 171
 172         if (mvp->v_type != VDIR)
 173                 return (ENOTDIR);
 174
 175         if ((uap->flags & MS_OVERLAY) == 0 &&
 176             (mvp->v_count > 1 || (mvp->v_flag & VROOT)))
 177                 return (EBUSY);
 178
 179         data = kmem_alloc(sizeof (ctfs_vfs_t), KM_SLEEP);
 180
 181         /*
 182          * Initialize vfs fields not initialized by VFS_INIT/domount
 183          */
 184         vfsp->vfs_bsize = DEV_BSIZE;
 185         vfsp->vfs_fstype = ctfs_fstype;
 186         do {
 187                 dev = makedevice(ctfs_major,
 188                     atomic_inc_32_nv(&ctfs_minor) & L_MAXMIN32);
 189         } while (vfs_devismounted(dev));
 190         vfs_make_fsid(&vfsp->vfs_fsid, dev, ctfs_fstype);
 191         vfsp->vfs_data = data;
 192         vfsp->vfs_dev = dev;
 193
 194         /*
 195          * Dynamically create gfs_dirent_t array for the root directory.
 196          */
 197         dirent = kmem_zalloc((ct_ntypes + 2) * sizeof (gfs_dirent_t), KM_SLEEP);
 198         for (i = 0; i < ct_ntypes; i++) {
 199                 dirent[i].gfse_name = (char *)ct_types[i]->ct_type_name;
 200                 dirent[i].gfse_ctor = ctfs_create_tdirnode;
 201                 dirent[i].gfse_flags = GFS_CACHE_VNODE;
 202         }
 203         dirent[i].gfse_name = "all";
 204         dirent[i].gfse_ctor = ctfs_create_adirnode;
 205         dirent[i].gfse_flags = GFS_CACHE_VNODE;
 206         dirent[i+1].gfse_name = NULL;
 207
 208         /*
 209          * Create root vnode
 210          */
 211         data->ctvfs_root = gfs_root_create(sizeof (ctfs_rootnode_t),
 212             vfsp, &ctfs_ops_root, CTFS_INO_ROOT, dirent, ctfs_root_do_inode,
 213             CTFS_NAME_MAX, NULL, NULL);
 214
 215         kmem_free(dirent, (ct_ntypes + 2) * sizeof (gfs_dirent_t));
 216
 217         return (0);
 218 }
 219
 220 /*
 221  * ctfs_unmount - the VFS_UNMOUNT entry point
 222  */
 223 static int
 224 ctfs_unmount(vfs_t *vfsp, int flag, struct cred *cr)
 225 {
 226         ctfs_vfs_t *data;
 227
 228         if (secpolicy_fs_unmount(cr, vfsp) != 0)
 229                 return (EPERM);
 230
 231         /*
 232          * Supporting forced unmounts would be nice to do at some
 233          * point.
 234          */
 235         if (flag & MS_FORCE)
 236                 return (ENOTSUP);
 237
 238         /*
 239          * We should never have a reference count less than 2: one for
 240          * the caller, one for the root vnode.
 241          */
 242         ASSERT(vfsp->vfs_count >= 2);
 243
 244         /*
 245          * If we have any active vnodes, they will (transitively) have
 246          * holds on the root vnode.
 247          */
 248         data = vfsp->vfs_data;
 249         if (data->ctvfs_root->v_count > 1)
 250                 return (EBUSY);
 251
 252         /*
 253          * Release the last hold on the root vnode.  It will, in turn,
 254          * release its hold on us.
 255          */
 256         VN_RELE(data->ctvfs_root);
 257
 258         /*
 259          * Disappear.
 260          */
 261         kmem_free(data, sizeof (ctfs_vfs_t));
 262
 263         return (0);
 264 }
 265
 266 /*
 267  * ctfs_root - the VFS_ROOT entry point
 268  */
 269 static int
 270 ctfs_root(vfs_t *vfsp, vnode_t **vpp)
 271 {
 272         vnode_t *vp;
 273
 274         vp = ((ctfs_vfs_t *)vfsp->vfs_data)->ctvfs_root;
 275         VN_HOLD(vp);
 276         *vpp = vp;
 277
 278         return (0);
 279 }
 280
 281 /*
 282  * ctfs_statvfs - the VFS_STATVFS entry point
 283  */
 284 static int
 285 ctfs_statvfs(vfs_t *vfsp, statvfs64_t *sp)
 286 {
 287         dev32_t d32;
 288         int     total, i;
 289
 290         bzero(sp, sizeof (*sp));
 291         sp->f_bsize = DEV_BSIZE;
 292         sp->f_frsize = DEV_BSIZE;
 293         for (i = 0, total = 0; i < ct_ntypes; i++)
 294                 total += contract_type_count(ct_types[i]);
 295         sp->f_files = total;
 296         sp->f_favail = sp->f_ffree = INT_MAX - total;
 297         (void) cmpldev(&d32, vfsp->vfs_dev);
 298         sp->f_fsid = d32;
 299         (void) strlcpy(sp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name,
 300             sizeof (sp->f_basetype));
 301         sp->f_flag = vf_to_stf(vfsp->vfs_flag);
 302         sp->f_namemax = CTFS_NAME_MAX;
 303         (void) strlcpy(sp->f_fstr, "contract", sizeof (sp->f_fstr));
 304
 305         return (0);
 306 }
 307
 308 static const struct vfsops ctfs_vfsops = {
 309         .vfs_mount = ctfs_mount,
 310         .vfs_unmount = ctfs_unmount,
 311         .vfs_root = ctfs_root,
 312         .vfs_statvfs = ctfs_statvfs,
 313 };
 314
 315 /*
 316  * ctfs_common_getattr
 317  *
 318  * Implements functionality common to all ctfs fop_getattr entry
 319  * points.  It assumes vap->va_size is set.
 320  */
 321 void
 322 ctfs_common_getattr(vnode_t *vp, vattr_t *vap)
 323 {
 324         vap->va_uid = 0;
 325         vap->va_gid = 0;
 326         vap->va_rdev = 0;
 327         vap->va_blksize = DEV_BSIZE;
 328         vap->va_nblocks = howmany(vap->va_size, vap->va_blksize);
 329         vap->va_seq = 0;
 330         vap->va_fsid = vp->v_vfsp->vfs_dev;
 331         vap->va_nodeid = gfs_file_inode(vp);
 332 }
 333
 334 /*
 335  * ctfs_open - common fop_open entry point
 336  *
 337  * Used by all ctfs directories; just verifies we are using large-file
 338  * aware interfaces and we aren't trying to open the directories
 339  * writable.
 340  */
 341 /* ARGSUSED */
 342 int
 343 ctfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
 344 {
 345         if ((flag & (FOFFMAX | FWRITE)) != FOFFMAX)
 346                 return (EINVAL);
 347
 348         return (0);
 349 }
 350
 351 /*
 352  * ctfs_close - common fop_close entry point
 353  *
 354  * For all ctfs vnode types which have no close-time clean-up to do.
 355  */
 356 /* ARGSUSED */
 357 int
 358 ctfs_close(
 359         vnode_t *vp,
 360         int flag,
 361         int count,
 362         offset_t offset,
 363         cred_t *cr,
 364         caller_context_t *ct)
 365 {
 366         return (0);
 367 }
 368
 369 /*
 370  * ctfs_access_dir - common fop_access entry point for directories
 371  */
 372 /* ARGSUSED */
 373 int
 374 ctfs_access_dir(
 375         vnode_t *vp,
 376         int mode,
 377         int flags,
 378         cred_t *cr,
 379         caller_context_t *ct)
 380 {
 381         if (mode & VWRITE)
 382                 return (EACCES);
 383
 384         return (0);
 385 }
 386
 387 /*
 388  * ctfs_access_dir - common fop_access entry point for read-only files
 389  */
 390 /* ARGSUSED */
 391 int
 392 ctfs_access_readonly(
 393         vnode_t *vp,
 394         int mode,
 395         int flags,
 396         cred_t *cr,
 397         caller_context_t *ct)
 398 {
 399         if (mode & (VWRITE | VEXEC))
 400                 return (EACCES);
 401
 402         return (0);
 403 }
 404
 405 /*
 406  * ctfs_access_dir - common fop_access entry point for read-write files
 407  */
 408 /* ARGSUSED */
 409 int
 410 ctfs_access_readwrite(
 411         vnode_t *vp,
 412         int mode,
 413         int flags,
 414         cred_t *cr,
 415         caller_context_t *ct)
 416 {
 417         if (mode & VEXEC)
 418                 return (EACCES);
 419
 420         return (0);
 421 }
 422
 423 /*
 424  * ctfs_root_getattr - fop_getattr entry point
 425  */
 426 /* ARGSUSED */
 427 static int
 428 ctfs_root_getattr(
 429         vnode_t *vp,
 430         vattr_t *vap,
 431         int flags,
 432         cred_t *cr,
 433         caller_context_t *ct)
 434 {
 435         vap->va_type = VDIR;
 436         vap->va_mode = 0555;
 437         vap->va_nlink = 2 + ct_ntypes + 1;
 438         vap->va_size = vap->va_nlink;
 439         vap->va_atime.tv_sec = vp->v_vfsp->vfs_mtime;
 440         vap->va_atime.tv_nsec = 0;
 441         vap->va_mtime = vap->va_ctime = vap->va_atime;
 442         ctfs_common_getattr(vp, vap);
 443
 444         return (0);
 445 }
 446
 447 /* ARGSUSED */
 448 static ino64_t
 449 ctfs_root_do_inode(vnode_t *vp, int index)
 450 {
 451         return (CTFS_INO_TYPE_DIR(index));
 452 }
 453
 454 const struct vnodeops ctfs_ops_root = {
 455         .vnop_name = "ctfs root directory",
 456         .vop_open = ctfs_open,
 457         .vop_close = ctfs_close,
 458         .vop_ioctl = fs_inval,
 459         .vop_getattr = ctfs_root_getattr,
 460         .vop_access = ctfs_access_dir,
 461         .vop_readdir = gfs_vop_readdir,
 462         .vop_lookup = gfs_vop_lookup,
 463         .vop_seek = fs_seek,
 464         .vop_inactive = gfs_vop_inactive,
 465 };