1 /* $NetBSD: mfs_vfsops.c,v 1.71 2005/12/11 12:25:28 christos 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 1.71 2005/12/11 12:25:28 christos 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 caddr_t 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_DEFINE(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 mfs_mount, 90 mfs_start, 91 ffs_unmount, 92 ufs_root, 93 ufs_quotactl, 94 mfs_statvfs, 95 ffs_sync, 96 ffs_vget, 97 ffs_fhtovp, 98 ffs_vptofh, 99 mfs_init, 100 mfs_reinit, 101 mfs_done, 102 NULL, 103 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 104 vfs_stdextattrctl, 105 mfs_vnodeopv_descs, 106 }; 107 VFS_ATTACH(mfs_vfsops); 108 109 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup") 110 { 111 112 sysctl_createv(clog, 0, NULL, NULL, 113 CTLFLAG_PERMANENT, 114 CTLTYPE_NODE, "vfs", NULL, 115 NULL, 0, NULL, 0, 116 CTL_VFS, CTL_EOL); 117 sysctl_createv(clog, 0, NULL, NULL, 118 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 119 CTLTYPE_NODE, "mfs", 120 SYSCTL_DESCR("Memory based file system"), 121 NULL, 1, NULL, 0, 122 CTL_VFS, 3, CTL_EOL); 123 /* 124 * XXX the "1" and the "3" above could be dynamic, thereby 125 * eliminating one more instance of the "number to vfs" 126 * mapping problem, but they are in order as taken from 127 * sys/mount.h 128 */ 129 } 130 131 /* 132 * Memory based filesystem initialization. 133 */ 134 void 135 mfs_init(void) 136 { 137 #ifdef _LKM 138 malloc_type_attach(M_MFSNODE); 139 #endif 140 /* 141 * ffs_init() ensures to initialize necessary resources 142 * only once. 143 */ 144 ffs_init(); 145 } 146 147 void 148 mfs_reinit(void) 149 { 150 ffs_reinit(); 151 } 152 153 void 154 mfs_done(void) 155 { 156 /* 157 * ffs_done() ensures to free necessary resources 158 * only once, when it's no more needed. 159 */ 160 ffs_done(); 161 #ifdef _LKM 162 malloc_type_detach(M_MFSNODE); 163 #endif 164 } 165 166 /* 167 * Called by main() when mfs is going to be mounted as root. 168 */ 169 170 int 171 mfs_mountroot(void) 172 { 173 struct fs *fs; 174 struct mount *mp; 175 struct lwp *l = curlwp; /* XXX */ 176 struct ufsmount *ump; 177 struct mfsnode *mfsp; 178 int error = 0; 179 180 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) { 181 vrele(rootvp); 182 return (error); 183 } 184 185 mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 186 rootvp->v_data = mfsp; 187 rootvp->v_op = mfs_vnodeop_p; 188 rootvp->v_tag = VT_MFS; 189 mfsp->mfs_baseoff = mfs_rootbase; 190 mfsp->mfs_size = mfs_rootsize; 191 mfsp->mfs_vnode = rootvp; 192 mfsp->mfs_proc = NULL; /* indicate kernel space */ 193 mfsp->mfs_shutdown = 0; 194 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 195 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) { 196 mp->mnt_op->vfs_refcount--; 197 vfs_unbusy(mp); 198 bufq_free(mfsp->mfs_buflist); 199 free(mp, M_MOUNT); 200 free(mfsp, M_MFSNODE); 201 return (error); 202 } 203 simple_lock(&mountlist_slock); 204 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 205 simple_unlock(&mountlist_slock); 206 mp->mnt_vnodecovered = NULLVP; 207 ump = VFSTOUFS(mp); 208 fs = ump->um_fs; 209 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 210 (void)ffs_statvfs(mp, &mp->mnt_stat, l); 211 vfs_unbusy(mp); 212 return (0); 213 } 214 215 /* 216 * This is called early in boot to set the base address and size 217 * of the mini-root. 218 */ 219 int 220 mfs_initminiroot(caddr_t base) 221 { 222 struct fs *fs = (struct fs *)(base + SBLOCK_UFS1); 223 224 /* check for valid super block */ 225 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE || 226 fs->fs_bsize < sizeof(struct fs)) 227 return (0); 228 mountroot = mfs_mountroot; 229 mfs_rootbase = base; 230 mfs_rootsize = fs->fs_fsize * fs->fs_size; 231 rootdev = makedev(255, mfs_minor); 232 mfs_minor++; 233 return (mfs_rootsize); 234 } 235 236 /* 237 * VFS Operations. 238 * 239 * mount system call 240 */ 241 /* ARGSUSED */ 242 int 243 mfs_mount(struct mount *mp, const char *path, void *data, 244 struct nameidata *ndp, struct lwp *l) 245 { 246 struct vnode *devvp; 247 struct mfs_args args; 248 struct ufsmount *ump; 249 struct fs *fs; 250 struct mfsnode *mfsp; 251 struct proc *p; 252 int flags, error; 253 254 p = l->l_proc; 255 if (mp->mnt_flag & MNT_GETARGS) { 256 struct vnode *vp; 257 258 ump = VFSTOUFS(mp); 259 if (ump == NULL) 260 return EIO; 261 262 vp = ump->um_devvp; 263 if (vp == NULL) 264 return EIO; 265 266 mfsp = VTOMFS(vp); 267 if (mfsp == NULL) 268 return EIO; 269 270 args.fspec = NULL; 271 args.base = mfsp->mfs_baseoff; 272 args.size = mfsp->mfs_size; 273 return copyout(&args, data, sizeof(args)); 274 } 275 /* 276 * XXX turn off async to avoid hangs when writing lots of data. 277 * the problem is that MFS needs to allocate pages to clean pages, 278 * so if we wait until the last minute to clean pages then there 279 * may not be any pages available to do the cleaning. 280 * ... and since the default partially-synchronous mode turns out 281 * to not be sufficient under heavy load, make it full synchronous. 282 */ 283 mp->mnt_flag &= ~MNT_ASYNC; 284 mp->mnt_flag |= MNT_SYNCHRONOUS; 285 286 error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args)); 287 if (error) 288 return (error); 289 290 /* 291 * If updating, check whether changing from read-only to 292 * read/write; if there is no device name, that's all we do. 293 */ 294 if (mp->mnt_flag & MNT_UPDATE) { 295 ump = VFSTOUFS(mp); 296 fs = ump->um_fs; 297 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 298 flags = WRITECLOSE; 299 if (mp->mnt_flag & MNT_FORCE) 300 flags |= FORCECLOSE; 301 error = ffs_flushfiles(mp, flags, l); 302 if (error) 303 return (error); 304 } 305 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 306 fs->fs_ronly = 0; 307 if (args.fspec == NULL) 308 return EINVAL; 309 return (0); 310 } 311 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp); 312 if (error) 313 return (error); 314 devvp->v_type = VBLK; 315 if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0)) 316 panic("mfs_mount: dup dev"); 317 mfs_minor++; 318 mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 319 devvp->v_data = mfsp; 320 mfsp->mfs_baseoff = args.base; 321 mfsp->mfs_size = args.size; 322 mfsp->mfs_vnode = devvp; 323 mfsp->mfs_proc = p; 324 mfsp->mfs_shutdown = 0; 325 bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0); 326 if ((error = ffs_mountfs(devvp, mp, l)) != 0) { 327 mfsp->mfs_shutdown = 1; 328 vrele(devvp); 329 return (error); 330 } 331 ump = VFSTOUFS(mp); 332 fs = ump->um_fs; 333 error = set_statvfs_info(path, UIO_USERSPACE, args.fspec, 334 UIO_USERSPACE, mp, l); 335 if (error) 336 return error; 337 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 338 sizeof(fs->fs_fsmnt)); 339 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 340 /* XXX: cleanup on error */ 341 return 0; 342 } 343 344 int mfs_pri = PWAIT | PCATCH; /* XXX prob. temp */ 345 346 /* 347 * Used to grab the process and keep it in the kernel to service 348 * memory filesystem I/O requests. 349 * 350 * Loop servicing I/O requests. 351 * Copy the requested data into or out of the memory filesystem 352 * address space. 353 */ 354 /* ARGSUSED */ 355 int 356 mfs_start(struct mount *mp, int flags, struct lwp *l) 357 { 358 struct vnode *vp = VFSTOUFS(mp)->um_devvp; 359 struct mfsnode *mfsp = VTOMFS(vp); 360 struct buf *bp; 361 caddr_t base; 362 int sleepreturn = 0; 363 364 base = mfsp->mfs_baseoff; 365 while (mfsp->mfs_shutdown != 1) { 366 while ((bp = BUFQ_GET(mfsp->mfs_buflist)) != NULL) { 367 mfs_doio(bp, base); 368 wakeup((caddr_t)bp); 369 } 370 /* 371 * If a non-ignored signal is received, try to unmount. 372 * If that fails, or the filesystem is already in the 373 * process of being unmounted, clear the signal (it has been 374 * "processed"), otherwise we will loop here, as tsleep 375 * will always return EINTR/ERESTART. 376 */ 377 if (sleepreturn != 0) { 378 /* 379 * XXX Freeze syncer. Must do this before locking 380 * the mount point. See dounmount() for details. 381 */ 382 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 383 if (vfs_busy(mp, LK_NOWAIT, 0) != 0) 384 lockmgr(&syncer_lock, LK_RELEASE, NULL); 385 else if (dounmount(mp, 0, l) != 0) 386 CLRSIG(l->l_proc, CURSIG(l)); 387 sleepreturn = 0; 388 continue; 389 } 390 391 sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0); 392 } 393 KASSERT(BUFQ_PEEK(mfsp->mfs_buflist) == NULL); 394 bufq_free(mfsp->mfs_buflist); 395 return (sleepreturn); 396 } 397 398 /* 399 * Get file system statistics. 400 */ 401 int 402 mfs_statvfs(struct mount *mp, struct statvfs *sbp, struct lwp *l) 403 { 404 int error; 405 406 error = ffs_statvfs(mp, sbp, l); 407 if (error) 408 return error; 409 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 410 sizeof(sbp->f_fstypename)); 411 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 412 return 0; 413 } 414