sys/vfs/dirfs/dirfs_vfsops.c

   1 /*
   2  * Copyright (c) 2013 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Antonio Huete Jimenez <tuxillo@quantumachine.net>
   6  * by Matthew Dillon <dillon@dragonflybsd.org>
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  *
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in
  16  *    the documentation and/or other materials provided with the
  17  *    distribution.
  18  * 3. Neither the name of The DragonFly Project nor the names of its
  19  *    contributors may be used to endorse or promote products derived
  20  *    from this software without specific, prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  23  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  27  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  30  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  31  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  32  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  33  * SUCH DAMAGE.
  34  *
  35  */
  36
  37 #include <sys/vfsops.h>
  38 #include <sys/mount.h>
  39 #include <sys/types.h>
  40 #include <sys/systm.h>
  41 #include <sys/param.h>
  42 #include <sys/module.h>
  43 #include <sys/kernel.h>
  44 #include <sys/malloc.h>
  45 #include <sys/sysctl.h>
  46 #include <sys/queue.h>
  47 #include <sys/spinlock2.h>
  48 #include <sys/sysref2.h>
  49 #include <sys/ktr.h>
  50
  51 #include <string.h>
  52
  53 #include "dirfs.h"
  54
  55 MALLOC_DEFINE(M_DIRFS, "dirfs", "dirfs mount allocation");
  56 MALLOC_DEFINE(M_DIRFS_NODE, "dirfs nodes", "dirfs nodes memory allocation");
  57 MALLOC_DEFINE(M_DIRFS_MISC, "dirfs misc", "dirfs miscellaneous allocation");
  58
  59 /*
  60  * Kernel tracing facilities
  61  */
  62 KTR_INFO_MASTER(dirfs);
  63
  64 KTR_INFO(KTR_DIRFS, dirfs, root, 20,
  65     "DIRFS(root dnp=%p vnode=%p hostdir=%s fd=%d error=%d)",
  66     dirfs_node_t dnp, struct vnode *vp, char *hostdir, int fd, int error);
  67
  68 KTR_INFO(KTR_DIRFS, dirfs, mount, 21,
  69     "DIRFS(mount path=%s dmp=%p mp=%p error=%d)",
  70     char *path, dirfs_mount_t dmp, struct mount *mp, int error);
  71
  72 KTR_INFO(KTR_DIRFS, dirfs, unmount, 22,
  73     "DIRFS(unmount dmp=%p mp=%p error=%d)",
  74     dirfs_mount_t dmp, struct mount *mp, int error);
  75
  76 /* System wide sysctl stuff */
  77 int debuglvl = 0;
  78 int dirfs_fd_limit = 100;
  79 int dirfs_fd_used = 0;
  80 long passive_fd_list_miss = 0;
  81 long passive_fd_list_hits = 0;
  82
  83 SYSCTL_NODE(_vfs, OID_AUTO, dirfs, CTLFLAG_RW, 0,
  84     "dirfs filesystem for vkernels");
  85 SYSCTL_INT(_vfs_dirfs, OID_AUTO, debug, CTLFLAG_RW,
  86     &debuglvl, 0, "dirfs debug level");
  87 SYSCTL_INT(_vfs_dirfs, OID_AUTO, fd_limit, CTLFLAG_RW,
  88     &dirfs_fd_limit, 0, "Maximum number of passive nodes to cache");
  89 SYSCTL_INT(_vfs_dirfs, OID_AUTO, fd_used, CTLFLAG_RD,
  90     &dirfs_fd_used, 0, "Current number of passive nodes cached");
  91 SYSCTL_LONG(_vfs_dirfs, OID_AUTO, passive_fd_list_miss, CTLFLAG_RD,
  92     &passive_fd_list_miss, 0, "Passive fd list cache misses");
  93 SYSCTL_LONG(_vfs_dirfs, OID_AUTO, passive_fd_list_hits, CTLFLAG_RD,
  94     &passive_fd_list_hits, 0, "Passive fd list cache misses");
  95
  96 static int dirfs_statfs(struct mount *, struct statfs *, struct ucred *);
  97
  98 static int
  99 dirfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
 100 {
 101         dirfs_mount_t dmp;
 102         struct stat st;
 103         size_t done, nlen;
 104         int error;
 105
 106         dbg(1, "called\n");
 107
 108         if (mp->mnt_flag & MNT_UPDATE) {
 109                 dmp = VFS_TO_DIRFS(mp);
 110                 if (dmp->dm_rdonly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
 111                         /* XXX We should make sure all writes are synced */
 112                         dmp->dm_rdonly = 1;
 113                         debug(2, "dirfs read-write -> read-only\n");
 114                 }
 115
 116                 if (dmp->dm_rdonly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
 117                         debug(2, "dirfs read-only -> read-write\n");
 118                         dmp->dm_rdonly = 0;
 119                 }
 120                 return 0;
 121         }
 122
 123         dmp = kmalloc(sizeof(*dmp), M_DIRFS, M_WAITOK | M_ZERO);
 124         mp->mnt_data = (qaddr_t)dmp;
 125         dmp->dm_mount = mp;
 126
 127         error = copyinstr(data, &dmp->dm_path, MAXPATHLEN, &done);
 128         if (error) {
 129                 /* Attempt to copy from kernel address */
 130                 error = copystr(data, &dmp->dm_path, MAXPATHLEN, &done);
 131                 if (error) {
 132                         kfree(dmp, M_DIRFS);
 133                         goto failure;
 134                 }
 135         }
 136
 137         /* Strip / character at the end to avoid problems */
 138         nlen = strnlen(dmp->dm_path, MAXPATHLEN);
 139         if (dmp->dm_path[nlen-1] == '/')
 140                 dmp->dm_path[nlen-1] = 0;
 141
 142         /* Make sure host directory exists and it is indeed a directory. */
 143         if ((stat(dmp->dm_path, &st)) == 0) {
 144                 if (!S_ISDIR(st.st_mode)) {
 145                         kfree(dmp, M_DIRFS);
 146                         error = EINVAL;
 147                         goto failure;
 148                 }
 149         } else {
 150                 error = errno;
 151                 goto failure;
 152         }
 153
 154         lockinit(&dmp->dm_lock, "dfsmnt", 0, LK_CANRECURSE);
 155
 156         vfs_add_vnodeops(mp, &dirfs_vnode_vops, &mp->mnt_vn_norm_ops);
 157         vfs_getnewfsid(mp);
 158
 159         /* Who is running the vkernel */
 160         dmp->dm_uid = getuid();
 161         dmp->dm_gid = getgid();
 162
 163         TAILQ_INIT(&dmp->dm_fdlist);
 164         RB_INIT(&dmp->dm_inotree);
 165
 166         kmalloc_raise_limit(M_DIRFS_NODE, 0);
 167
 168         dirfs_statfs(mp, &mp->mnt_stat, cred);
 169
 170 failure:
 171         KTR_LOG(dirfs_mount, (dmp->dm_path) ? dmp->dm_path : "NULL",
 172             dmp, mp, error);
 173
 174         return error;
 175 }
 176
 177 static int
 178 dirfs_unmount(struct mount *mp, int mntflags)
 179 {
 180         dirfs_mount_t dmp;
 181         dirfs_node_t dnp;
 182         int cnt;
 183         int error;
 184
 185         dbg(1, "called\n");
 186         cnt = 0;
 187         dmp = VFS_TO_DIRFS(mp);
 188
 189         error = vflush(mp, 0, 0);
 190         if (error)
 191                 goto failure;
 192
 193         /*
 194          * Clean up dm_fdlist.  There should be no vnodes left so the
 195          * only ref should be from the fdlist.
 196          */
 197         while ((dnp = TAILQ_FIRST(&dmp->dm_fdlist)) != NULL) {
 198                 dirfs_node_setpassive(dmp, dnp, 0);
 199         }
 200
 201         /*
 202          * Cleanup root node. In the case the filesystem is mounted
 203          * but no operation is done on it, there will be no call to
 204          * VFS_ROOT() so better check dnp is not NULL before attempting
 205          * to release it.
 206          */
 207         dnp = dmp->dm_root;
 208         if (dnp != NULL) {
 209                 dirfs_close_helper(dnp);
 210                 debug_node2(dnp);
 211                 dirfs_node_drop(dmp, dnp); /* last ref should free structure */
 212         }
 213         kfree(dmp, M_DIRFS);
 214         mp->mnt_data = (qaddr_t) 0;
 215
 216 failure:
 217         KTR_LOG(dirfs_unmount, dmp, mp, error);
 218
 219         return error;
 220 }
 221
 222 static int
 223 dirfs_root(struct mount *mp, struct vnode **vpp)
 224 {
 225         dirfs_mount_t dmp;
 226         dirfs_node_t dnp;
 227         int fd;
 228         int error;
 229
 230         dbg(1, "called\n");
 231
 232         dmp = VFS_TO_DIRFS(mp);
 233         KKASSERT(dmp != NULL);
 234
 235         if (dmp->dm_root == NULL) {
 236                 /*
 237                  * dm_root holds the root dirfs node. Allocate a new one since
 238                  * there is none. Also attempt to lstat(2) it, in order to set
 239                  * data for VOP_ACCESS()
 240                  */
 241                 dnp = dirfs_node_alloc(mp);
 242                 error = dirfs_node_stat(DIRFS_NOFD, dmp->dm_path, dnp);
 243                 if (error != 0) {
 244                         dirfs_node_free(dmp, dnp);
 245                         return error;
 246                 }
 247                 dirfs_node_ref(dnp);    /* leave inact for life of mount */
 248
 249                 /* Root inode's parent is NULL, used for verification */
 250                 dnp->dn_parent = NULL;
 251                 dmp->dm_root = dnp;
 252                 dirfs_node_setflags(dnp, DIRFS_ROOT);
 253
 254                 /*
 255                  * Maintain an open descriptor on the root dnp.  The
 256                  * normal open/close/cache does not apply for the root
 257                  * so the descriptor is ALWAYS available.
 258                  */
 259                 fd = open(dmp->dm_path, O_DIRECTORY);
 260                 if (fd == -1) {
 261                         dbg(9, "failed to open ROOT node\n");
 262                         dirfs_free_vp(dmp, dnp);
 263                         dirfs_node_free(dmp, dnp);
 264                         return errno;
 265                 }
 266                 dnp->dn_fd = fd;
 267                 dnp->dn_type = VDIR;
 268         } else {
 269                 dnp = dmp->dm_root;
 270         }
 271
 272         /*
 273          * Acquire the root vnode (dn_type already set above).  This
 274          * call will handle any races and return a locked vnode.
 275          */
 276         dirfs_alloc_vp(mp, vpp, LK_CANRECURSE, dnp);
 277         KTR_LOG(dirfs_root, dnp, *vpp, dmp->dm_path, dnp->dn_fd, error);
 278
 279         return 0;
 280 }
 281
 282 static int
 283 dirfs_fhtovp(struct mount *mp, struct vnode *rootvp, struct fid *fhp, struct vnode **vpp)
 284 {
 285         dbg(1, "called\n");
 286
 287         return EOPNOTSUPP;
 288 }
 289
 290 static int
 291 dirfs_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
 292 {
 293         dirfs_mount_t dmp = VFS_TO_DIRFS(mp);
 294         struct statfs st;
 295
 296         dbg(1, "called\n");
 297
 298         if((statfs(dmp->dm_path, &st)) == -1)
 299                 return errno;
 300
 301         ksnprintf(st.f_mntfromname, MNAMELEN - 1, "dirfs@%s", dmp->dm_path);
 302         bcopy(&st, sbp, sizeof(st));
 303         strlcpy(sbp->f_fstypename, mp->mnt_vfc->vfc_name, MFSNAMELEN);
 304         dbg(5, "iosize = %zd\n", sbp->f_iosize);
 305
 306         return 0;
 307 }
 308
 309 static int
 310 dirfs_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
 311 {
 312         dirfs_mount_t dmp = VFS_TO_DIRFS(mp);
 313         struct statvfs st;
 314
 315         dbg(1, "called\n");
 316
 317         if ((statvfs(dmp->dm_path, &st)) == -1)
 318                 return errno;
 319
 320         bcopy(&st, sbp, sizeof(st));
 321
 322         return 0;
 323 }
 324
 325 static int
 326 dirfs_vptofh(struct vnode *vp, struct fid *fhp)
 327 {
 328         dirfs_node_t dnp;
 329
 330         dnp = VP_TO_NODE(vp);
 331         debug_node2(dnp);
 332         dbg(1, "called\n");
 333
 334         return EOPNOTSUPP;
 335 }
 336
 337 static int
 338 dirfs_checkexp(struct mount *mp, struct sockaddr *nam, int *exflagsp,
 339                struct ucred **credanonp)
 340 {
 341         dbg(1, "called\n");
 342
 343         return EOPNOTSUPP;
 344 }
 345
 346 static struct vfsops dirfs_vfsops = {
 347         .vfs_mount =                    dirfs_mount,
 348         .vfs_unmount =                  dirfs_unmount,
 349         .vfs_root =                     dirfs_root,
 350         .vfs_vget =                     vfs_stdvget,
 351         .vfs_statfs =                   dirfs_statfs,
 352         .vfs_statvfs =                  dirfs_statvfs,
 353         .vfs_fhtovp =                   dirfs_fhtovp,
 354         .vfs_vptofh =                   dirfs_vptofh,
 355         .vfs_checkexp =                 dirfs_checkexp
 356 };
 357
 358 VFS_SET(dirfs_vfsops, dirfs, 0);
 359 MODULE_VERSION(dirfs, 1);