1 /* $NetBSD: mfs_vfsops.c,v 1.65 2005/02/26 22:32:20 perry 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 1.65 2005/02/26 22:32:20 perry 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) __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 NULL, 104 ufs_check_export, 105 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 106 vfs_stdextattrctl, 107 mfs_vnodeopv_descs, 108 }; 109 110 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup") 111 { 112 113 sysctl_createv(clog, 0, NULL, NULL, 114 CTLFLAG_PERMANENT, 115 CTLTYPE_NODE, "vfs", NULL, 116 NULL, 0, NULL, 0, 117 CTL_VFS, CTL_EOL); 118 sysctl_createv(clog, 0, NULL, NULL, 119 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 120 CTLTYPE_NODE, "mfs", 121 SYSCTL_DESCR("Memory based file system"), 122 NULL, 1, NULL, 0, 123 CTL_VFS, 3, CTL_EOL); 124 /* 125 * XXX the "1" and the "3" above could be dynamic, thereby 126 * eliminating one more instance of the "number to vfs" 127 * mapping problem, but they are in order as taken from 128 * sys/mount.h 129 */ 130 } 131 132 /* 133 * Memory based filesystem initialization. 134 */ 135 void 136 mfs_init() 137 { 138 #ifdef _LKM 139 malloc_type_attach(M_MFSNODE); 140 #endif 141 /* 142 * ffs_init() ensures to initialize necessary resources 143 * only once. 144 */ 145 ffs_init(); 146 } 147 148 void 149 mfs_reinit() 150 { 151 ffs_reinit(); 152 } 153 154 void 155 mfs_done() 156 { 157 /* 158 * ffs_done() ensures to free necessary resources 159 * only once, when it's no more needed. 160 */ 161 ffs_done(); 162 #ifdef _LKM 163 malloc_type_detach(M_MFSNODE); 164 #endif 165 } 166 167 /* 168 * Called by main() when mfs is going to be mounted as root. 169 */ 170 171 int 172 mfs_mountroot() 173 { 174 struct fs *fs; 175 struct mount *mp; 176 struct proc *p = curproc; /* 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, BUFQ_FCFS); 196 if ((error = ffs_mountfs(rootvp, mp, p)) != 0) { 197 mp->mnt_op->vfs_refcount--; 198 vfs_unbusy(mp); 199 bufq_free(&mfsp->mfs_buflist); 200 free(mp, M_MOUNT); 201 free(mfsp, M_MFSNODE); 202 return (error); 203 } 204 simple_lock(&mountlist_slock); 205 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 206 simple_unlock(&mountlist_slock); 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, p); 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(base) 222 caddr_t base; 223 { 224 struct fs *fs = (struct fs *)(base + SBLOCK_UFS1); 225 226 /* check for valid super block */ 227 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE || 228 fs->fs_bsize < sizeof(struct fs)) 229 return (0); 230 mountroot = mfs_mountroot; 231 mfs_rootbase = base; 232 mfs_rootsize = fs->fs_fsize * fs->fs_size; 233 rootdev = makedev(255, mfs_minor); 234 mfs_minor++; 235 return (mfs_rootsize); 236 } 237 238 /* 239 * VFS Operations. 240 * 241 * mount system call 242 */ 243 /* ARGSUSED */ 244 int 245 mfs_mount(mp, path, data, ndp, p) 246 struct mount *mp; 247 const char *path; 248 void *data; 249 struct nameidata *ndp; 250 struct proc *p; 251 { 252 struct vnode *devvp; 253 struct mfs_args args; 254 struct ufsmount *ump; 255 struct fs *fs; 256 struct mfsnode *mfsp; 257 int flags, error; 258 259 if (mp->mnt_flag & MNT_GETARGS) { 260 struct vnode *vp; 261 struct mfsnode *mfsp; 262 263 ump = VFSTOUFS(mp); 264 if (ump == NULL) 265 return EIO; 266 267 vp = ump->um_devvp; 268 if (vp == NULL) 269 return EIO; 270 271 mfsp = VTOMFS(vp); 272 if (mfsp == NULL) 273 return EIO; 274 275 args.fspec = NULL; 276 vfs_showexport(mp, &args.export, &ump->um_export); 277 args.base = mfsp->mfs_baseoff; 278 args.size = mfsp->mfs_size; 279 return copyout(&args, data, sizeof(args)); 280 } 281 /* 282 * XXX turn off async to avoid hangs when writing lots of data. 283 * the problem is that MFS needs to allocate pages to clean pages, 284 * so if we wait until the last minute to clean pages then there 285 * may not be any pages available to do the cleaning. 286 * ... and since the default partially-synchronous mode turns out 287 * to not be sufficient under heavy load, make it full synchronous. 288 */ 289 mp->mnt_flag &= ~MNT_ASYNC; 290 mp->mnt_flag |= MNT_SYNCHRONOUS; 291 292 error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args)); 293 if (error) 294 return (error); 295 296 /* 297 * If updating, check whether changing from read-only to 298 * read/write; if there is no device name, that's all we do. 299 */ 300 if (mp->mnt_flag & MNT_UPDATE) { 301 ump = VFSTOUFS(mp); 302 fs = ump->um_fs; 303 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 304 flags = WRITECLOSE; 305 if (mp->mnt_flag & MNT_FORCE) 306 flags |= FORCECLOSE; 307 error = ffs_flushfiles(mp, flags, p); 308 if (error) 309 return (error); 310 } 311 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 312 fs->fs_ronly = 0; 313 if (args.fspec == 0) 314 return (vfs_export(mp, &ump->um_export, &args.export)); 315 return (0); 316 } 317 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp); 318 if (error) 319 return (error); 320 devvp->v_type = VBLK; 321 if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0)) 322 panic("mfs_mount: dup dev"); 323 mfs_minor++; 324 mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 325 devvp->v_data = mfsp; 326 mfsp->mfs_baseoff = args.base; 327 mfsp->mfs_size = args.size; 328 mfsp->mfs_vnode = devvp; 329 mfsp->mfs_proc = p; 330 mfsp->mfs_shutdown = 0; 331 bufq_alloc(&mfsp->mfs_buflist, BUFQ_FCFS); 332 if ((error = ffs_mountfs(devvp, mp, p)) != 0) { 333 mfsp->mfs_shutdown = 1; 334 vrele(devvp); 335 return (error); 336 } 337 ump = VFSTOUFS(mp); 338 fs = ump->um_fs; 339 error = set_statvfs_info(path, UIO_USERSPACE, args.fspec, 340 UIO_USERSPACE, mp, p); 341 if (error) 342 return error; 343 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 344 sizeof(fs->fs_fsmnt)); 345 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 346 /* XXX: cleanup on error */ 347 return 0; 348 } 349 350 int mfs_pri = PWAIT | PCATCH; /* XXX prob. temp */ 351 352 /* 353 * Used to grab the process and keep it in the kernel to service 354 * memory filesystem I/O requests. 355 * 356 * Loop servicing I/O requests. 357 * Copy the requested data into or out of the memory filesystem 358 * address space. 359 */ 360 /* ARGSUSED */ 361 int 362 mfs_start(mp, flags, p) 363 struct mount *mp; 364 int flags; 365 struct proc *p; 366 { 367 struct vnode *vp = VFSTOUFS(mp)->um_devvp; 368 struct mfsnode *mfsp = VTOMFS(vp); 369 struct buf *bp; 370 caddr_t base; 371 int sleepreturn = 0; 372 struct lwp *l; /* XXX NJWLWP */ 373 374 /* XXX NJWLWP the vnode interface again gives us a proc in a 375 * place where we want a execution context. Cheat. 376 */ 377 KASSERT(curproc == p); 378 l = curlwp; 379 base = mfsp->mfs_baseoff; 380 while (mfsp->mfs_shutdown != 1) { 381 while ((bp = BUFQ_GET(&mfsp->mfs_buflist)) != NULL) { 382 mfs_doio(bp, base); 383 wakeup((caddr_t)bp); 384 } 385 /* 386 * If a non-ignored signal is received, try to unmount. 387 * If that fails, or the filesystem is already in the 388 * process of being unmounted, clear the signal (it has been 389 * "processed"), otherwise we will loop here, as tsleep 390 * will always return EINTR/ERESTART. 391 */ 392 if (sleepreturn != 0) { 393 /* 394 * XXX Freeze syncer. Must do this before locking 395 * the mount point. See dounmount() for details. 396 */ 397 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 398 if (vfs_busy(mp, LK_NOWAIT, 0) != 0) 399 lockmgr(&syncer_lock, LK_RELEASE, NULL); 400 else if (dounmount(mp, 0, p) != 0) 401 CLRSIG(p, CURSIG(l)); 402 sleepreturn = 0; 403 continue; 404 } 405 406 sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0); 407 } 408 KASSERT(BUFQ_PEEK(&mfsp->mfs_buflist) == NULL); 409 bufq_free(&mfsp->mfs_buflist); 410 return (sleepreturn); 411 } 412 413 /* 414 * Get file system statistics. 415 */ 416 int 417 mfs_statvfs(mp, sbp, p) 418 struct mount *mp; 419 struct statvfs *sbp; 420 struct proc *p; 421 { 422 int error; 423 424 error = ffs_statvfs(mp, sbp, p); 425 if (error) 426 return error; 427 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 428 sizeof(sbp->f_fstypename)); 429 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 430 return 0; 431 } 432