1 /* $NetBSD: mfs_vfsops.c,v 1.61 2004/07/05 07:28:46 pk 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 1.61 2004/07/05 07:28:46 pk 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/mount.h> 49 #include <sys/signalvar.h> 50 #include <sys/vnode.h> 51 #include <sys/malloc.h> 52 53 #include <miscfs/syncfs/syncfs.h> 54 55 #include <ufs/ufs/quota.h> 56 #include <ufs/ufs/inode.h> 57 #include <ufs/ufs/ufsmount.h> 58 #include <ufs/ufs/ufs_extern.h> 59 60 #include <ufs/ffs/fs.h> 61 #include <ufs/ffs/ffs_extern.h> 62 63 #include <ufs/mfs/mfsnode.h> 64 #include <ufs/mfs/mfs_extern.h> 65 66 caddr_t mfs_rootbase; /* address of mini-root in kernel virtual memory */ 67 u_long mfs_rootsize; /* size of mini-root in bytes */ 68 69 static int mfs_minor; /* used for building internal dev_t */ 70 71 extern int (**mfs_vnodeop_p) __P((void *)); 72 73 MALLOC_DEFINE(M_MFSNODE, "MFS node", "MFS vnode private part"); 74 75 /* 76 * mfs vfs operations. 77 */ 78 79 extern const struct vnodeopv_desc mfs_vnodeop_opv_desc; 80 81 const struct vnodeopv_desc * const mfs_vnodeopv_descs[] = { 82 &mfs_vnodeop_opv_desc, 83 NULL, 84 }; 85 86 struct vfsops mfs_vfsops = { 87 MOUNT_MFS, 88 mfs_mount, 89 mfs_start, 90 ffs_unmount, 91 ufs_root, 92 ufs_quotactl, 93 mfs_statvfs, 94 ffs_sync, 95 ffs_vget, 96 ffs_fhtovp, 97 ffs_vptofh, 98 mfs_init, 99 mfs_reinit, 100 mfs_done, 101 NULL, 102 NULL, 103 ufs_check_export, 104 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp, 105 mfs_vnodeopv_descs, 106 }; 107 108 SYSCTL_SETUP(sysctl_vfs_mfs_setup, "sysctl vfs.mfs subtree setup") 109 { 110 111 sysctl_createv(clog, 0, NULL, NULL, 112 CTLFLAG_PERMANENT, 113 CTLTYPE_NODE, "vfs", NULL, 114 NULL, 0, NULL, 0, 115 CTL_VFS, CTL_EOL); 116 sysctl_createv(clog, 0, NULL, NULL, 117 CTLFLAG_PERMANENT|CTLFLAG_ALIAS, 118 CTLTYPE_NODE, "mfs", 119 SYSCTL_DESCR("Memory based file system"), 120 NULL, 1, NULL, 0, 121 CTL_VFS, 3, CTL_EOL); 122 /* 123 * XXX the "1" and the "3" above could be dynamic, thereby 124 * eliminating one more instance of the "number to vfs" 125 * mapping problem, but they are in order as taken from 126 * sys/mount.h 127 */ 128 } 129 130 /* 131 * Memory based filesystem initialization. 132 */ 133 void 134 mfs_init() 135 { 136 #ifdef _LKM 137 malloc_type_attach(M_MFSNODE); 138 #endif 139 /* 140 * ffs_init() ensures to initialize necessary resources 141 * only once. 142 */ 143 ffs_init(); 144 } 145 146 void 147 mfs_reinit() 148 { 149 ffs_reinit(); 150 } 151 152 void 153 mfs_done() 154 { 155 /* 156 * ffs_done() ensures to free necessary resources 157 * only once, when it's no more needed. 158 */ 159 ffs_done(); 160 #ifdef _LKM 161 malloc_type_detach(M_MFSNODE); 162 #endif 163 } 164 165 /* 166 * Called by main() when mfs is going to be mounted as root. 167 */ 168 169 int 170 mfs_mountroot() 171 { 172 struct fs *fs; 173 struct mount *mp; 174 struct proc *p = curproc; /* XXX */ 175 struct ufsmount *ump; 176 struct mfsnode *mfsp; 177 int error = 0; 178 179 /* 180 * Get vnodes for rootdev. 181 */ 182 if (bdevvp(rootdev, &rootvp)) { 183 printf("mfs_mountroot: can't setup bdevvp's"); 184 return (error); 185 } 186 187 if ((error = vfs_rootmountalloc(MOUNT_MFS, "mfs_root", &mp))) { 188 vrele(rootvp); 189 return (error); 190 } 191 192 mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 193 rootvp->v_data = mfsp; 194 rootvp->v_op = mfs_vnodeop_p; 195 rootvp->v_tag = VT_MFS; 196 mfsp->mfs_baseoff = mfs_rootbase; 197 mfsp->mfs_size = mfs_rootsize; 198 mfsp->mfs_vnode = rootvp; 199 mfsp->mfs_proc = NULL; /* indicate kernel space */ 200 mfsp->mfs_shutdown = 0; 201 bufq_alloc(&mfsp->mfs_buflist, BUFQ_FCFS); 202 if ((error = ffs_mountfs(rootvp, mp, p)) != 0) { 203 mp->mnt_op->vfs_refcount--; 204 vfs_unbusy(mp); 205 bufq_free(&mfsp->mfs_buflist); 206 free(mp, M_MOUNT); 207 free(mfsp, M_MFSNODE); 208 vrele(rootvp); 209 return (error); 210 } 211 simple_lock(&mountlist_slock); 212 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list); 213 simple_unlock(&mountlist_slock); 214 mp->mnt_vnodecovered = NULLVP; 215 ump = VFSTOUFS(mp); 216 fs = ump->um_fs; 217 (void) copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0); 218 (void)ffs_statvfs(mp, &mp->mnt_stat, p); 219 vfs_unbusy(mp); 220 return (0); 221 } 222 223 /* 224 * This is called early in boot to set the base address and size 225 * of the mini-root. 226 */ 227 int 228 mfs_initminiroot(base) 229 caddr_t base; 230 { 231 struct fs *fs = (struct fs *)(base + SBLOCK_UFS1); 232 233 /* check for valid super block */ 234 if (fs->fs_magic != FS_UFS1_MAGIC || fs->fs_bsize > MAXBSIZE || 235 fs->fs_bsize < sizeof(struct fs)) 236 return (0); 237 mountroot = mfs_mountroot; 238 mfs_rootbase = base; 239 mfs_rootsize = fs->fs_fsize * fs->fs_size; 240 rootdev = makedev(255, mfs_minor); 241 mfs_minor++; 242 return (mfs_rootsize); 243 } 244 245 /* 246 * VFS Operations. 247 * 248 * mount system call 249 */ 250 /* ARGSUSED */ 251 int 252 mfs_mount(mp, path, data, ndp, p) 253 struct mount *mp; 254 const char *path; 255 void *data; 256 struct nameidata *ndp; 257 struct proc *p; 258 { 259 struct vnode *devvp; 260 struct mfs_args args; 261 struct ufsmount *ump; 262 struct fs *fs; 263 struct mfsnode *mfsp; 264 int flags, error; 265 266 if (mp->mnt_flag & MNT_GETARGS) { 267 struct vnode *vp; 268 struct mfsnode *mfsp; 269 270 ump = VFSTOUFS(mp); 271 if (ump == NULL) 272 return EIO; 273 274 vp = ump->um_devvp; 275 if (vp == NULL) 276 return EIO; 277 278 mfsp = VTOMFS(vp); 279 if (mfsp == NULL) 280 return EIO; 281 282 args.fspec = NULL; 283 vfs_showexport(mp, &args.export, &ump->um_export); 284 args.base = mfsp->mfs_baseoff; 285 args.size = mfsp->mfs_size; 286 return copyout(&args, data, sizeof(args)); 287 } 288 /* 289 * XXX turn off async to avoid hangs when writing lots of data. 290 * the problem is that MFS needs to allocate pages to clean pages, 291 * so if we wait until the last minute to clean pages then there 292 * may not be any pages available to do the cleaning. 293 * ... and since the default partially-synchronous mode turns out 294 * to not be sufficient under heavy load, make it full synchronous. 295 */ 296 mp->mnt_flag &= ~MNT_ASYNC; 297 mp->mnt_flag |= MNT_SYNCHRONOUS; 298 299 error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args)); 300 if (error) 301 return (error); 302 303 /* 304 * If updating, check whether changing from read-only to 305 * read/write; if there is no device name, that's all we do. 306 */ 307 if (mp->mnt_flag & MNT_UPDATE) { 308 ump = VFSTOUFS(mp); 309 fs = ump->um_fs; 310 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) { 311 flags = WRITECLOSE; 312 if (mp->mnt_flag & MNT_FORCE) 313 flags |= FORCECLOSE; 314 error = ffs_flushfiles(mp, flags, p); 315 if (error) 316 return (error); 317 } 318 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) 319 fs->fs_ronly = 0; 320 if (args.fspec == 0) 321 return (vfs_export(mp, &ump->um_export, &args.export)); 322 return (0); 323 } 324 error = getnewvnode(VT_MFS, (struct mount *)0, mfs_vnodeop_p, &devvp); 325 if (error) 326 return (error); 327 devvp->v_type = VBLK; 328 if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0)) 329 panic("mfs_mount: dup dev"); 330 mfs_minor++; 331 mfsp = (struct mfsnode *)malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK); 332 devvp->v_data = mfsp; 333 mfsp->mfs_baseoff = args.base; 334 mfsp->mfs_size = args.size; 335 mfsp->mfs_vnode = devvp; 336 mfsp->mfs_proc = p; 337 mfsp->mfs_shutdown = 0; 338 bufq_alloc(&mfsp->mfs_buflist, BUFQ_FCFS); 339 if ((error = ffs_mountfs(devvp, mp, p)) != 0) { 340 mfsp->mfs_shutdown = 1; 341 vrele(devvp); 342 return (error); 343 } 344 ump = VFSTOUFS(mp); 345 fs = ump->um_fs; 346 error = set_statvfs_info(path, UIO_USERSPACE, args.fspec, 347 UIO_USERSPACE, mp, p); 348 if (error) 349 return error; 350 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, 351 sizeof(fs->fs_fsmnt)); 352 fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0'; 353 /* XXX: cleanup on error */ 354 return 0; 355 } 356 357 int mfs_pri = PWAIT | PCATCH; /* XXX prob. temp */ 358 359 /* 360 * Used to grab the process and keep it in the kernel to service 361 * memory filesystem I/O requests. 362 * 363 * Loop servicing I/O requests. 364 * Copy the requested data into or out of the memory filesystem 365 * address space. 366 */ 367 /* ARGSUSED */ 368 int 369 mfs_start(mp, flags, p) 370 struct mount *mp; 371 int flags; 372 struct proc *p; 373 { 374 struct vnode *vp = VFSTOUFS(mp)->um_devvp; 375 struct mfsnode *mfsp = VTOMFS(vp); 376 struct buf *bp; 377 caddr_t base; 378 int sleepreturn = 0; 379 struct lwp *l; /* XXX NJWLWP */ 380 381 /* XXX NJWLWP the vnode interface again gives us a proc in a 382 * place where we want a execution context. Cheat. 383 */ 384 KASSERT(curproc == p); 385 l = curlwp; 386 base = mfsp->mfs_baseoff; 387 while (mfsp->mfs_shutdown != 1) { 388 while ((bp = BUFQ_GET(&mfsp->mfs_buflist)) != NULL) { 389 mfs_doio(bp, base); 390 wakeup((caddr_t)bp); 391 } 392 /* 393 * If a non-ignored signal is received, try to unmount. 394 * If that fails, or the filesystem is already in the 395 * process of being unmounted, clear the signal (it has been 396 * "processed"), otherwise we will loop here, as tsleep 397 * will always return EINTR/ERESTART. 398 */ 399 if (sleepreturn != 0) { 400 /* 401 * XXX Freeze syncer. Must do this before locking 402 * the mount point. See dounmount() for details. 403 */ 404 lockmgr(&syncer_lock, LK_EXCLUSIVE, NULL); 405 if (vfs_busy(mp, LK_NOWAIT, 0) != 0) 406 lockmgr(&syncer_lock, LK_RELEASE, NULL); 407 else if (dounmount(mp, 0, p) != 0) 408 CLRSIG(p, CURSIG(l)); 409 sleepreturn = 0; 410 continue; 411 } 412 413 sleepreturn = tsleep(vp, mfs_pri, "mfsidl", 0); 414 } 415 KASSERT(BUFQ_PEEK(&mfsp->mfs_buflist) == NULL); 416 bufq_free(&mfsp->mfs_buflist); 417 return (sleepreturn); 418 } 419 420 /* 421 * Get file system statistics. 422 */ 423 int 424 mfs_statvfs(mp, sbp, p) 425 struct mount *mp; 426 struct statvfs *sbp; 427 struct proc *p; 428 { 429 int error; 430 431 error = ffs_statvfs(mp, sbp, p); 432 if (error) 433 return error; 434 (void)strncpy(sbp->f_fstypename, mp->mnt_op->vfs_name, 435 sizeof(sbp->f_fstypename)); 436 sbp->f_fstypename[sizeof(sbp->f_fstypename) - 1] = '\0'; 437 return 0; 438 } 439