1 /* $NetBSD: mfs_vfsops.c,v 1.117 2025/02/16 18:38:59 joe 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 1.117 2025/02/16 18:38:59 joe 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, "ffs"); 70 71 kmutex_t mfs_lock; /* global lock */ 72 73 /* used for building internal dev_t, minor == 0 reserved for miniroot */ 74 static devminor_t mfs_minor = 1; 75 static int mfs_initcnt; 76 77 extern int (**mfs_vnodeop_p)(void *); 78 79 /* 80 * mfs vfs operations. 81 */ 82 83 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc; 84 85 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = { 86 &mfs_vnodeop_opv_desc, 87 NULL, 88 }; 89 90 struct vfsops mfs_vfsops = { 91 .vfs_name = MOUNT_MFS, 92 .vfs_min_mount_data = sizeof (struct mfs_args), 93 .vfs_mount = mfs_mount, 94 .vfs_start = mfs_start, 95 .vfs_unmount = ffs_unmount, 96 .vfs_root = ufs_root, 97 .vfs_quotactl = ufs_quotactl, 98 .vfs_statvfs = mfs_statvfs, 99 .vfs_sync = ffs_sync, 100 .vfs_vget = ufs_vget, 101 .vfs_loadvnode = ffs_loadvnode, 102 .vfs_newvnode = ffs_newvnode, 103 .vfs_fhtovp = ffs_fhtovp, 104 .vfs_vptofh = ffs_vptofh, 105 .vfs_init = mfs_init, 106 .vfs_reinit = mfs_reinit, 107 .vfs_done = mfs_done, 108 .vfs_snapshot = (void *)eopnotsupp, 109 .vfs_extattrctl = vfs_stdextattrctl, 110 .vfs_suspendctl = genfs_suspendctl, 111 .vfs_renamelock_enter = genfs_renamelock_enter, 112 .vfs_renamelock_exit = genfs_renamelock_exit, 113 .vfs_fsync = (void *)eopnotsupp, 114 .vfs_opv_descs = mfs_vnodeopv_descs 115 }; 116 117 SYSCTL_SETUP(mfs_sysctl_setup, "mfs sysctl") 118 { 119 120 sysctl_createv(clog, 0, NULL, NULL, 121 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 122 CTLTYPE_NODE, "mfs", 123 SYSCTL_DESCR("Memory based file system"), 124 NULL, 1, NULL, 0, 125 CTL_VFS, 3, CTL_EOL); 126 /* 127 * XXX the "1" and the "3" above could be dynamic, thereby 128 * eliminating one more instance of the "number to vfs" 129 * mapping problem, but they are in order as taken from 130 * sys/mount.h 131 */ 132 } 133 134 static int 135 mfs_modcmd(modcmd_t cmd, void *arg) 136 { 137 int error; 138 139 switch (cmd) { 140 case MODULE_CMD_INIT: 141 error = vfs_attach(&mfs_vfsops); 142 break; 143 case MODULE_CMD_FINI: 144 error = vfs_detach(&mfs_vfsops); 145 break; 146 default: 147 error = ENOTTY; 148 break; 149 } 150 151 return (error); 152 } 153 154 /* 155 * Memory based filesystem initialization. 156 */ 157 void 158 mfs_init(void) 159 { 160 161 if (mfs_initcnt++ == 0) { 162 mutex_init(&mfs_lock, MUTEX_DEFAULT, IPL_NONE); 163 ffs_init(); 164 } 165 } 166 167 void 168 mfs_reinit(void) 169 { 170 171 ffs_reinit(); 172 } 173 174 void 175 mfs_done(void) 176 { 177 178 if (--mfs_initcnt == 0) { 179 ffs_done(); 180 mutex_destroy(&mfs_lock); 181 } 182 } 183 184 /* 185 * Called by main() when mfs is going to be mounted as root. 186 */ 187 188 int 189 mfs_mountroot(void) 190 { 191 struct fs *fs; 192 struct mount *mp; 193 struct lwp *l = curlwp; /* XXX */ 194 struct ufsmount *ump; 195 struct mfsnode *mfsp; 196 int error = 0; 197 198 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) { 199 vrele(rootvp); 200 return (error); 201 } 202 203 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP); 204 rootvp->v_data = mfsp; 205 rootvp->v_op = mfs_vnodeop_p; 206 rootvp->v_tag = VT_MFS; 207 mfsp->mfs_baseoff = mfs_rootbase; 208 mfsp->mfs_size = mfs_rootsize; 209 mfsp->mfs_vnode = rootvp; 210 mfsp->mfs_proc = NULL; /* indicate kernel space */ 211 mfsp->mfs_shutdown = 0; 212 cv_init(&mfsp->mfs_cv, "mfs"); 213 mfsp->mfs_refcnt = 1; 214 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 215 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 216 vfs_unbusy(mp); 217 bufq_free(mfsp->mfs_buflist); 218 vfs_rele(mp); 219 kmem_free(mfsp, sizeof(*mfsp)); 220 return (error); 221 } 222 mountlist_append(mp); 223 mp->mnt_vnodecovered = NULLVP; 224 ump = VFSTOUFS(mp); 225 fs = ump->um_fs; 226 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 227 (void)ffs_statvfs(mp, &mp->mnt_stat); 228 vfs_unbusy(mp); 229 return (0); 230 } 231 232 /* 233 * VFS Operations. 234 * 235 * mount system call 236 */ 237 /* ARGSUSED */ 238 int 239 mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len) 240 { 241 struct lwp *l = curlwp; 242 struct vnode *devvp; 243 struct mfs_args *args = data; 244 struct ufsmount *ump; 245 struct fs *fs; 246 struct mfsnode *mfsp; 247 struct proc *p; 248 devminor_t minor; 249 int flags, error = 0; 250 251 if (args == NULL) 252 return EINVAL; 253 if (*data_len < sizeof *args) 254 return EINVAL; 255 256 p = l->l_proc; 257 if (mp->mnt_flag & MNT_GETARGS) { 258 struct vnode *vp; 259 260 ump = VFSTOUFS(mp); 261 if (ump == NULL) 262 return EIO; 263 264 vp = ump->um_devvp; 265 if (vp == NULL) 266 return EIO; 267 268 mfsp = VTOMFS(vp); 269 if (mfsp == NULL) 270 return EIO; 271 272 args->fspec = NULL; 273 args->base = mfsp->mfs_baseoff; 274 args->size = mfsp->mfs_size; 275 *data_len = sizeof *args; 276 return 0; 277 } 278 /* 279 * XXX turn off async to avoid hangs when writing lots of data. 280 * the problem is that MFS needs to allocate pages to clean pages, 281 * so if we wait until the last minute to clean pages then there 282 * may not be any pages available to do the cleaning. 283 * ... and since the default partially-synchronous mode turns out 284 * to not be sufficient under heavy load, make it full synchronous. 285 */ 286 mp->mnt_flag &= ~MNT_ASYNC; 287 mp->mnt_flag |= MNT_SYNCHRONOUS; 288 289 /* 290 * If updating, check whether changing from read-only to 291 * read/write; if there is no device name, that's all we do. 292 */ 293 if (mp->mnt_flag & MNT_UPDATE) { 294 ump = VFSTOUFS(mp); 295 fs = ump->um_fs; 296 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 297 flags = WRITECLOSE; 298 if (mp->mnt_flag & MNT_FORCE) 299 flags |= FORCECLOSE; 300 error = ffs_flushfiles(mp, flags, l); 301 if (error) 302 return (error); 303 } 304 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 305 fs->fs_ronly = 0; 306 if (args->fspec == NULL) 307 return EINVAL; 308 return (0); 309 } 310 mutex_enter(&mfs_lock); 311 minor = mfs_minor++; 312 mutex_exit(&mfs_lock); 313 error = bdevvp(makedev(255, minor), &devvp); 314 if (error) 315 return (error); 316 mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP); 317 /* 318 * Changing v_op and v_data here is safe as we are 319 * the exclusive owner of this device node. 320 */ 321 KASSERT(devvp->v_op == spec_vnodeop_p); 322 KASSERT(devvp->v_data == NULL); 323 devvp->v_op = mfs_vnodeop_p; 324 devvp->v_data = mfsp; 325 mfsp->mfs_baseoff = args->base; 326 mfsp->mfs_size = args->size; 327 mfsp->mfs_vnode = devvp; 328 mfsp->mfs_proc = p; 329 mfsp->mfs_shutdown = 0; 330 cv_init(&mfsp->mfs_cv, "mfsidl"); 331 mfsp->mfs_refcnt = 1; 332 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 333 if ((error = ffs_mountfs(devvp, mp, l)) != 0) { 334 mfsp->mfs_shutdown = 1; 335 vrele(devvp); 336 return (error); 337 } 338 ump = VFSTOUFS(mp); 339 fs = ump->um_fs; 340 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, 341 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l); 342 if (error) 343 return error; 344 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 345 sizeof(fs->fs_fsmnt)); 346 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 347 /* XXX: cleanup on error */ 348 return 0; 349 } 350 351 /* 352 * Used to grab the process and keep it in the kernel to service 353 * memory filesystem I/O requests. 354 * 355 * Loop servicing I/O requests. 356 * Copy the requested data into or out of the memory filesystem 357 * address space. 358 */ 359 /* ARGSUSED */ 360 int 361 mfs_start(struct mount *mp, int flags) 362 { 363 struct vnode *vp; 364 struct mfsnode *mfsp; 365 struct proc *p; 366 struct buf *bp; 367 void *base; 368 int sleepreturn = 0, refcnt, error; 369 ksiginfoq_t kq; 370 371 /* 372 * Ensure that file system is still mounted when getting mfsnode. 373 * Add a reference to the mfsnode to prevent it disappearing in 374 * this routine. 375 */ 376 if ((error = vfs_busy(mp)) != 0) 377 return error; 378 vp = VFSTOUFS(mp)->um_devvp; 379 mfsp = VTOMFS(vp); 380 mutex_enter(&mfs_lock); 381 mfsp->mfs_refcnt++; 382 mutex_exit(&mfs_lock); 383 vfs_unbusy(mp); 384 385 base = mfsp->mfs_baseoff; 386 mutex_enter(&mfs_lock); 387 while (mfsp->mfs_shutdown != 1) { 388 while ((bp = bufq_get(mfsp->mfs_buflist)) != NULL) { 389 mutex_exit(&mfs_lock); 390 mfs_doio(bp, base); 391 mutex_enter(&mfs_lock); 392 } 393 /* 394 * If a non-ignored signal is received, try to unmount. 395 * If that fails, or the filesystem is already in the 396 * process of being unmounted, clear the signal (it has been 397 * "processed"), otherwise we will loop here, as tsleep 398 * will always return EINTR/ERESTART. 399 */ 400 if (sleepreturn != 0) { 401 mutex_exit(&mfs_lock); 402 if (dounmount(mp, 0, curlwp) != 0) { 403 p = curproc; 404 ksiginfo_queue_init(&kq); 405 mutex_enter(p->p_lock); 406 sigclearall(p, NULL, &kq); 407 mutex_exit(p->p_lock); 408 ksiginfo_queue_drain(&kq); 409 } 410 sleepreturn = 0; 411 mutex_enter(&mfs_lock); 412 continue; 413 } 414 415 sleepreturn = cv_wait_sig(&mfsp->mfs_cv, &mfs_lock); 416 } 417 KASSERT(bufq_peek(mfsp->mfs_buflist) == NULL); 418 refcnt = --mfsp->mfs_refcnt; 419 mutex_exit(&mfs_lock); 420 if (refcnt == 0) { 421 bufq_free(mfsp->mfs_buflist); 422 cv_destroy(&mfsp->mfs_cv); 423 kmem_free(mfsp, sizeof(*mfsp)); 424 } 425 return (sleepreturn); 426 } 427 428 /* 429 * Get file system statistics. 430 */ 431 int 432 mfs_statvfs(struct mount *mp, struct statvfs *sbp) 433 { 434 int error; 435 436 error = ffs_statvfs(mp, sbp); 437 if (error) 438 return error; 439 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 440 sizeof(sbp->f_fstypename)); 441 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 442 return 0; 443 } 444