mfs_vfsops.c revision 1.97
1 /* $NetBSD: mfs_vfsops.c,v 1.97 2008/05/10 02:26:11 rumble 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 1.97 2008/05/10 02:26:11 rumble 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/kmem.h> 53 #include <sys/module.h> 54 55 #include <miscfs/genfs/genfs.h> 56 #include <miscfs/specfs/specdev.h> 57 58 #include <ufs/ufs/quota.h> 59 #include <ufs/ufs/inode.h> 60 #include <ufs/ufs/ufsmount.h> 61 #include <ufs/ufs/ufs_extern.h> 62 63 #include <ufs/ffs/fs.h> 64 #include <ufs/ffs/ffs_extern.h> 65 66 #include <ufs/mfs/mfsnode.h> 67 #include <ufs/mfs/mfs_extern.h> 68 69 MODULE(MODULE_CLASS_VFS, mfs, NULL); 70 71 void * mfs_rootbase; /* address of mini-root in kernel virtual memory */ 72 u_long mfs_rootsize; /* size of mini-root in bytes */ 73 kmutex_t mfs_lock; /* global lock */ 74 75 static int mfs_minor; /* used for building internal dev_t */ 76 static int mfs_initcnt; 77 78 extern int (**mfs_vnodeop_p)(void *); 79 80 /* 81 * mfs vfs operations. 82 */ 83 84 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc; 85 86 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = { 87 &mfs_vnodeop_opv_desc, 88 NULL, 89 }; 90 91 struct vfsops mfs_vfsops = { 92 MOUNT_MFS, 93 sizeof (struct mfs_args), 94 mfs_mount, 95 mfs_start, 96 ffs_unmount, 97 ufs_root, 98 ufs_quotactl, 99 mfs_statvfs, 100 ffs_sync, 101 ffs_vget, 102 ffs_fhtovp, 103 ffs_vptofh, 104 mfs_init, 105 mfs_reinit, 106 mfs_done, 107 NULL, 108 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 109 vfs_stdextattrctl, 110 (void *)eopnotsupp, /* vfs_suspendctl */ 111 genfs_renamelock_enter, 112 genfs_renamelock_exit, 113 (void *)eopnotsupp, 114 mfs_vnodeopv_descs, 115 0, 116 { NULL, NULL }, 117 }; 118 119 static int 120 mfs_modcmd(modcmd_t cmd, void *arg) 121 { 122 123 switch (cmd) { 124 case MODULE_CMD_INIT: 125 return vfs_attach(&mfs_vfsops); 126 case MODULE_CMD_FINI: 127 return vfs_detach(&mfs_vfsops); 128 default: 129 return ENOTTY; 130 } 131 } 132 133 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup") 134 { 135 136 sysctl_createv(clog, 0, NULL, NULL, 137 CTLFLAG_PERMANENT, 138 CTLTYPE_NODE, "vfs", NULL, 139 NULL, 0, NULL, 0, 140 CTL_VFS, CTL_EOL); 141 sysctl_createv(clog, 0, NULL, NULL, 142 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 143 CTLTYPE_NODE, "mfs", 144 SYSCTL_DESCR("Memory based file system"), 145 NULL, 1, NULL, 0, 146 CTL_VFS, 3, CTL_EOL); 147 /* 148 * XXX the "1" and the "3" above could be dynamic, thereby 149 * eliminating one more instance of the "number to vfs" 150 * mapping problem, but they are in order as taken from 151 * sys/mount.h 152 */ 153 } 154 155 /* 156 * Memory based filesystem initialization. 157 */ 158 void 159 mfs_init(void) 160 { 161 162 if (mfs_initcnt++ == 0) { 163 mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE); 164 ffs_init(); 165 } 166 } 167 168 void 169 mfs_reinit(void) 170 { 171 172 ffs_reinit(); 173 } 174 175 void 176 mfs_done(void) 177 { 178 179 if (--mfs_initcnt == 0) { 180 ffs_done(); 181 mutex_destroy(&mfs_lock); 182 } 183 } 184 185 /* 186 * Called by main() when mfs is going to be mounted as root. 187 */ 188 189 int 190 mfs_mountroot(void) 191 { 192 struct fs *fs; 193 struct mount *mp; 194 struct lwp *l = curlwp; /* XXX */ 195 struct ufsmount *ump; 196 struct mfsnode *mfsp; 197 int error = 0; 198 199 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) { 200 vrele(rootvp); 201 return (error); 202 } 203 204 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP); 205 rootvp->v_data = mfsp; 206 rootvp->v_op = mfs_vnodeop_p; 207 rootvp->v_tag = VT_MFS; 208 mfsp->mfs_baseoff = mfs_rootbase; 209 mfsp->mfs_size = mfs_rootsize; 210 mfsp->mfs_vnode = rootvp; 211 mfsp->mfs_proc = NULL; /* indicate kernel space */ 212 mfsp->mfs_shutdown = 0; 213 cv_init(&mfsp->mfs_cv, "mfs"); 214 mfsp->mfs_refcnt = 1; 215 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 216 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 217 vfs_unbusy(mp, false, NULL); 218 bufq_free(mfsp->mfs_buflist); 219 vfs_destroy(mp); 220 kmem_free(mfsp, sizeof(*mfsp)); 221 return (error); 222 } 223 mutex_enter(&mountlist_lock); 224 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 225 mutex_exit(&mountlist_lock); 226 mp->mnt_vnodecovered = NULLVP; 227 ump = VFSTOUFS(mp); 228 fs = ump->um_fs; 229 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 230 (void)ffs_statvfs(mp, &mp->mnt_stat); 231 vfs_unbusy(mp, false, NULL); 232 return (0); 233 } 234 235 /* 236 * This is called early in boot to set the base address and size 237 * of the mini-root. 238 */ 239 int 240 mfs_initminiroot(void *base) 241 { 242 struct fs *fs = (struct fs *)((char *)base + SBLOCK_UFS1); 243 244 /* check for valid super block */ 245 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE || 246 fs->fs_bsize < sizeof(struct fs)) 247 return (0); 248 mountroot = mfs_mountroot; 249 mfs_rootbase = base; 250 mfs_rootsize = fs->fs_fsize * fs->fs_size; 251 rootdev = makedev(255, mfs_minor); 252 mfs_minor++; 253 return (mfs_rootsize); 254 } 255 256 /* 257 * VFS Operations. 258 * 259 * mount system call 260 */ 261 /* ARGSUSED */ 262 int 263 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 264 { 265 struct lwp *l = curlwp; 266 struct vnode *devvp; 267 struct mfs_args *args = data; 268 struct ufsmount *ump; 269 struct fs *fs; 270 struct mfsnode *mfsp; 271 struct proc *p; 272 int flags, error = 0; 273 274 if (*data_len < sizeof *args) 275 return EINVAL; 276 277 p = l->l_proc; 278 if (mp->mnt_flag & MNT_GETARGS) { 279 struct vnode *vp; 280 281 ump = VFSTOUFS(mp); 282 if (ump == NULL) 283 return EIO; 284 285 vp = ump->um_devvp; 286 if (vp == NULL) 287 return EIO; 288 289 mfsp = VTOMFS(vp); 290 if (mfsp == NULL) 291 return EIO; 292 293 args->fspec = NULL; 294 args->base = mfsp->mfs_baseoff; 295 args->size = mfsp->mfs_size; 296 *data_len = sizeof *args; 297 return 0; 298 } 299 /* 300 * XXX turn off async to avoid hangs when writing lots of data. 301 * the problem is that MFS needs to allocate pages to clean pages, 302 * so if we wait until the last minute to clean pages then there 303 * may not be any pages available to do the cleaning. 304 * ... and since the default partially-synchronous mode turns out 305 * to not be sufficient under heavy load, make it full synchronous. 306 */ 307 mp->mnt_flag &= ~MNT_ASYNC; 308 mp->mnt_flag |= MNT_SYNCHRONOUS; 309 310 /* 311 * If updating, check whether changing from read-only to 312 * read/write; if there is no device name, that's all we do. 313 */ 314 if (mp->mnt_flag & MNT_UPDATE) { 315 ump = VFSTOUFS(mp); 316 fs = ump->um_fs; 317 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 318 flags = WRITECLOSE; 319 if (mp->mnt_flag & MNT_FORCE) 320 flags |= FORCECLOSE; 321 error = ffs_flushfiles(mp, flags, l); 322 if (error) 323 return (error); 324 } 325 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 326 fs->fs_ronly = 0; 327 if (args->fspec == NULL) 328 return EINVAL; 329 return (0); 330 } 331 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp); 332 if (error) 333 return (error); 334 devvp->v_vflag |= VV_MPSAFE; 335 devvp->v_type = VBLK; 336 spec_node_init(devvp, makedev(255, mfs_minor)); 337 mfs_minor++; 338 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP); 339 devvp->v_data = mfsp; 340 mfsp->mfs_baseoff = args->base; 341 mfsp->mfs_size = args->size; 342 mfsp->mfs_vnode = devvp; 343 mfsp->mfs_proc = p; 344 mfsp->mfs_shutdown = 0; 345 cv_init(&mfsp->mfs_cv, "mfsidl"); 346 mfsp->mfs_refcnt = 1; 347 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 348 if ((error = ffs_mountfs(devvp, mp, l)) != 0) { 349 mfsp->mfs_shutdown = 1; 350 vrele(devvp); 351 return (error); 352 } 353 ump = VFSTOUFS(mp); 354 fs = ump->um_fs; 355 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 356 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 357 if (error) 358 return error; 359 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 360 sizeof(fs->fs_fsmnt)); 361 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 362 /* XXX: cleanup on error */ 363 return 0; 364 } 365 366 /* 367 * Used to grab the process and keep it in the kernel to service 368 * memory filesystem I/O requests. 369 * 370 * Loop servicing I/O requests. 371 * Copy the requested data into or out of the memory filesystem 372 * address space. 373 */ 374 /* ARGSUSED */ 375 int 376 mfs_start(struct mount *mp, int flags) 377 { 378 struct vnode *vp; 379 struct mfsnode *mfsp; 380 struct proc *p; 381 struct buf *bp; 382 void *base; 383 int sleepreturn = 0, refcnt, error; 384 ksiginfoq_t kq; 385 386 /* 387 * Ensure that file system is still mounted when getting mfsnode. 388 * Add a reference to the mfsnode to prevent it disappearing in 389 * this routine. 390 */ 391 if ((error = vfs_busy(mp, NULL)) != 0) 392 return error; 393 vp = VFSTOUFS(mp)->um_devvp; 394 mfsp = VTOMFS(vp); 395 mutex_enter(&mfs_lock); 396 mfsp->mfs_refcnt++; 397 mutex_exit(&mfs_lock); 398 vfs_unbusy(mp, false, NULL); 399 400 base = mfsp->mfs_baseoff; 401 mutex_enter(&mfs_lock); 402 while (mfsp->mfs_shutdown != 1) { 403 while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) { 404 mutex_exit(&mfs_lock); 405 mfs_doio(bp, base); 406 mutex_enter(&mfs_lock); 407 } 408 /* 409 * If a non-ignored signal is received, try to unmount. 410 * If that fails, or the filesystem is already in the 411 * process of being unmounted, clear the signal (it has been 412 * "processed"), otherwise we will loop here, as tsleep 413 * will always return EINTR/ERESTART. 414 */ 415 if (sleepreturn != 0) { 416 mutex_exit(&mfs_lock); 417 if (dounmount(mp, 0, curlwp) != 0) { 418 p = curproc; 419 ksiginfo_queue_init(&kq); 420 mutex_enter(p->p_lock); 421 sigclearall(p, NULL, &kq); 422 mutex_exit(p->p_lock); 423 ksiginfo_queue_drain(&kq); 424 } 425 sleepreturn = 0; 426 mutex_enter(&mfs_lock); 427 continue; 428 } 429 430 sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock); 431 } 432 KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL); 433 refcnt = --mfsp->mfs_refcnt; 434 mutex_exit(&mfs_lock); 435 if (refcnt == 0) { 436 bufq_free(mfsp->mfs_buflist); 437 cv_destroy(&mfsp->mfs_cv); 438 kmem_free(mfsp, sizeof(*mfsp)); 439 } 440 return (sleepreturn); 441 } 442 443 /* 444 * Get file system statistics. 445 */ 446 int 447 mfs_statvfs(struct mount *mp, struct statvfs *sbp) 448 { 449 int error; 450 451 error = ffs_statvfs(mp, sbp); 452 if (error) 453 return error; 454 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 455 sizeof(sbp->f_fstypename)); 456 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 457 return 0; 458 } 459
Indexes created Mon Sep 29 21:09:56 GMT 2025