1 /* $NetBSD: mfs_vfsops.c,v 1.84 2007/10/10 20:42:40 ad Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1990, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)mfs_vfsops.c 8.11 (Berkeley) 6/19/95 32 */ 33 34 #include <sys/cdefs.h> 35 __KERNEL_RCSID(0, "$NetBSD: mfs_vfsops.c,v 1.84 2007/10/10 20:42:40 ad Exp $"); 36 37 #if defined(_KERNEL_OPT) 38 #include "opt_compat_netbsd.h" 39 #endif 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/sysctl.h> 44 #include <sys/time.h> 45 #include <sys/kernel.h> 46 #include <sys/proc.h> 47 #include <sys/buf.h> 48 #include <sys/bufq.h> 49 #include <sys/mount.h> 50 #include <sys/signalvar.h> 51 #include <sys/vnode.h> 52 #include <sys/malloc.h> 53 54 #include <miscfs/syncfs/syncfs.h> 55 56 #include <ufs/ufs/quota.h> 57 #include <ufs/ufs/inode.h> 58 #include <ufs/ufs/ufsmount.h> 59 #include <ufs/ufs/ufs_extern.h> 60 61 #include <ufs/ffs/fs.h> 62 #include <ufs/ffs/ffs_extern.h> 63 64 #include <ufs/mfs/mfsnode.h> 65 #include <ufs/mfs/mfs_extern.h> 66 67 void * mfs_rootbase; /* address of mini-root in kernel virtual memory */ 68 u_long mfs_rootsize; /* size of mini-root in bytes */ 69 70 static int mfs_minor; /* used for building internal dev_t */ 71 72 extern int (**mfs_vnodeop_p)(void *); 73 74 MALLOC_JUSTDEFINE(M_MFSNODE, "MFS node", "MFS vnode private part"); 75 76 /* 77 * mfs vfs operations. 78 */ 79 80 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc; 81 82 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = { 83 &mfs_vnodeop_opv_desc, 84 NULL, 85 }; 86 87 struct vfsops mfs_vfsops = { 88 MOUNT_MFS, 89 sizeof (struct mfs_args), 90 mfs_mount, 91 mfs_start, 92 ffs_unmount, 93 ufs_root, 94 ufs_quotactl, 95 mfs_statvfs, 96 ffs_sync, 97 ffs_vget, 98 ffs_fhtovp, 99 ffs_vptofh, 100 mfs_init, 101 mfs_reinit, 102 mfs_done, 103 NULL, 104 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 105 vfs_stdextattrctl, 106 (void *)eopnotsupp, /* vfs_suspendctl */ 107 mfs_vnodeopv_descs, 108 0, 109 { NULL, NULL }, 110 }; 111 VFS_ATTACH(mfs_vfsops); 112 113 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup") 114 { 115 116 sysctl_createv(clog, 0, NULL, NULL, 117 CTLFLAG_PERMANENT, 118 CTLTYPE_NODE, "vfs", NULL, 119 NULL, 0, NULL, 0, 120 CTL_VFS, CTL_EOL); 121 sysctl_createv(clog, 0, NULL, NULL, 122 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 123 CTLTYPE_NODE, "mfs", 124 SYSCTL_DESCR("Memory based file system"), 125 NULL, 1, NULL, 0, 126 CTL_VFS, 3, CTL_EOL); 127 /* 128 * XXX the "1" and the "3" above could be dynamic, thereby 129 * eliminating one more instance of the "number to vfs" 130 * mapping problem, but they are in order as taken from 131 * sys/mount.h 132 */ 133 } 134 135 /* 136 * Memory based filesystem initialization. 137 */ 138 void 139 mfs_init(void) 140 { 141 142 malloc_type_attach(M_MFSNODE); 143 /* 144 * ffs_init() ensures to initialize necessary resources 145 * only once. 146 */ 147 ffs_init(); 148 } 149 150 void 151 mfs_reinit(void) 152 { 153 ffs_reinit(); 154 } 155 156 void 157 mfs_done(void) 158 { 159 /* 160 * ffs_done() ensures to free necessary resources 161 * only once, when it's no more needed. 162 */ 163 ffs_done(); 164 malloc_type_detach(M_MFSNODE); 165 } 166 167 /* 168 * Called by main() when mfs is going to be mounted as root. 169 */ 170 171 int 172 mfs_mountroot(void) 173 { 174 struct fs *fs; 175 struct mount *mp; 176 struct lwp *l = curlwp; /* XXX */ 177 struct ufsmount *ump; 178 struct mfsnode *mfsp; 179 int error = 0; 180 181 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) { 182 vrele(rootvp); 183 return (error); 184 } 185 186 mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 187 rootvp->v_data = mfsp; 188 rootvp->v_op = mfs_vnodeop_p; 189 rootvp->v_tag = VT_MFS; 190 mfsp->mfs_baseoff = mfs_rootbase; 191 mfsp->mfs_size = mfs_rootsize; 192 mfsp->mfs_vnode = rootvp; 193 mfsp->mfs_proc = NULL; /* indicate kernel space */ 194 mfsp->mfs_shutdown = 0; 195 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 196 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 197 mp->mnt_op->vfs_refcount--; 198 vfs_unbusy(mp); 199 bufq_free(mfsp->mfs_buflist); 200 vfs_destroy(mp); 201 free(mfsp, M_MFSNODE); 202 return (error); 203 } 204 mutex_enter(&mountlist_lock); 205 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 206 mutex_exit(&mountlist_lock); 207 mp->mnt_vnodecovered = NULLVP; 208 ump = VFSTOUFS(mp); 209 fs = ump->um_fs; 210 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 211 (void)ffs_statvfs(mp, &mp->mnt_stat, l); 212 vfs_unbusy(mp); 213 return (0); 214 } 215 216 /* 217 * This is called early in boot to set the base address and size 218 * of the mini-root. 219 */ 220 int 221 mfs_initminiroot(void *base) 222 { 223 struct fs *fs = (struct fs *)((char *)base + SBLOCK_UFS1); 224 225 /* check for valid super block */ 226 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE || 227 fs->fs_bsize < sizeof(struct fs)) 228 return (0); 229 mountroot = mfs_mountroot; 230 mfs_rootbase = base; 231 mfs_rootsize = fs->fs_fsize * fs->fs_size; 232 rootdev = makedev(255, mfs_minor); 233 mfs_minor++; 234 return (mfs_rootsize); 235 } 236 237 /* 238 * VFS Operations. 239 * 240 * mount system call 241 */ 242 /* ARGSUSED */ 243 int 244 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len, 245 struct lwp *l) 246 { 247 struct vnode *devvp; 248 struct mfs_args *args = data; 249 struct ufsmount *ump; 250 struct fs *fs; 251 struct mfsnode *mfsp; 252 struct proc *p; 253 int flags, error = 0; 254 255 if (*data_len < sizeof *args) 256 return EINVAL; 257 258 p = l->l_proc; 259 if (mp->mnt_flag & MNT_GETARGS) { 260 struct vnode *vp; 261 262 ump = VFSTOUFS(mp); 263 if (ump == NULL) 264 return EIO; 265 266 vp = ump->um_devvp; 267 if (vp == NULL) 268 return EIO; 269 270 mfsp = VTOMFS(vp); 271 if (mfsp == NULL) 272 return EIO; 273 274 args->fspec = NULL; 275 args->base = mfsp->mfs_baseoff; 276 args->size = mfsp->mfs_size; 277 *data_len = sizeof *args; 278 return 0; 279 } 280 /* 281 * XXX turn off async to avoid hangs when writing lots of data. 282 * the problem is that MFS needs to allocate pages to clean pages, 283 * so if we wait until the last minute to clean pages then there 284 * may not be any pages available to do the cleaning. 285 * ... and since the default partially-synchronous mode turns out 286 * to not be sufficient under heavy load, make it full synchronous. 287 */ 288 mp->mnt_flag &= ~MNT_ASYNC; 289 mp->mnt_flag |= MNT_SYNCHRONOUS; 290 291 /* 292 * If updating, check whether changing from read-only to 293 * read/write; if there is no device name, that's all we do. 294 */ 295 if (mp->mnt_flag & MNT_UPDATE) { 296 ump = VFSTOUFS(mp); 297 fs = ump->um_fs; 298 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 299 flags = WRITECLOSE; 300 if (mp->mnt_flag & MNT_FORCE) 301 flags |= FORCECLOSE; 302 error = ffs_flushfiles(mp, flags, l); 303 if (error) 304 return (error); 305 } 306 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 307 fs->fs_ronly = 0; 308 if (args->fspec == NULL) 309 return EINVAL; 310 return (0); 311 } 312 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp); 313 if (error) 314 return (error); 315 devvp->v_type = VBLK; 316 if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0)) 317 panic("mfs_mount: dup dev"); 318 mfs_minor++; 319 mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 320 devvp->v_data = mfsp; 321 mfsp->mfs_baseoff = args->base; 322 mfsp->mfs_size = args->size; 323 mfsp->mfs_vnode = devvp; 324 mfsp->mfs_proc = p; 325 mfsp->mfs_shutdown = 0; 326 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 327 if ((error = ffs_mountfs(devvp, mp, l)) != 0) { 328 mfsp->mfs_shutdown = 1; 329 vrele(devvp); 330 return (error); 331 } 332 ump = VFSTOUFS(mp); 333 fs = ump->um_fs; 334 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 335 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 336 if (error) 337 return error; 338 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 339 sizeof(fs->fs_fsmnt)); 340 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 341 /* XXX: cleanup on error */ 342 return 0; 343 } 344 345 int mfs_pri = PWAIT | PCATCH; /* XXX prob. temp */ 346 347 /* 348 * Used to grab the process and keep it in the kernel to service 349 * memory filesystem I/O requests. 350 * 351 * Loop servicing I/O requests. 352 * Copy the requested data into or out of the memory filesystem 353 * address space. 354 */ 355 /* ARGSUSED */ 356 int 357 mfs_start(struct mount *mp, int flags, struct lwp *l) 358 { 359 struct vnode *vp = VFSTOUFS(mp)->um_devvp; 360 struct mfsnode *mfsp = VTOMFS(vp); 361 struct proc *p; 362 struct buf *bp; 363 void *base; 364 int sleepreturn = 0; 365 ksiginfoq_t kq; 366 367 base = mfsp->mfs_baseoff; 368 while (mfsp->mfs_shutdown != 1) { 369 while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) { 370 mfs_doio(bp, base); 371 wakeup((void *)bp); 372 } 373 /* 374 * If a non-ignored signal is received, try to unmount. 375 * If that fails, or the filesystem is already in the 376 * process of being unmounted, clear the signal (it has been 377 * "processed"), otherwise we will loop here, as tsleep 378 * will always return EINTR/ERESTART. 379 */ 380 if (sleepreturn != 0) { 381 /* 382 * XXX Freeze syncer. Must do this before locking 383 * the mount point. See dounmount() for details. 384 */ 385 mutex_enter(&syncer_mutex); 386 if (vfs_busy(mp, LK_NOWAIT, 0) != 0) 387 mutex_exit(&syncer_mutex); 388 else if (dounmount(mp, 0, l) != 0) { 389 p = l->l_proc; 390 ksiginfo_queue_init(&kq); 391 mutex_enter(&p->p_smutex); 392 sigclearall(p, NULL, &kq); 393 mutex_exit(&p->p_smutex); 394 ksiginfo_queue_drain(&kq); 395 } 396 sleepreturn = 0; 397 continue; 398 } 399 400 sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0); 401 } 402 KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL); 403 bufq_free(mfsp->mfs_buflist); 404 return (sleepreturn); 405 } 406 407 /* 408 * Get file system statistics. 409 */ 410 int 411 mfs_statvfs(struct mount *mp, struct statvfs *sbp, struct lwp *l) 412 { 413 int error; 414 415 error = ffs_statvfs(mp, sbp, l); 416 if (error) 417 return error; 418 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 419 sizeof(sbp->f_fstypename)); 420 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 421 return 0; 422 } 423