| 1 | /* |
| 2 | * Copyright (c) 1989, 1993 |
| 3 | * The Regents of the University of California. All rights reserved. |
| 4 | * |
| 5 | * Redistribution and use in source and binary forms, with or without |
| 6 | * modification, are permitted provided that the following conditions |
| 7 | * are met: |
| 8 | * 1. Redistributions of source code must retain the above copyright |
| 9 | * notice, this list of conditions and the following disclaimer. |
| 10 | * 2. Redistributions in binary form must reproduce the above copyright |
| 11 | * notice, this list of conditions and the following disclaimer in the |
| 12 | * documentation and/or other materials provided with the distribution. |
| 13 | * 3. All advertising materials mentioning features or use of this software |
| 14 | * must display the following acknowledgement: |
| 15 | * This product includes software developed by the University of |
| 16 | * California, Berkeley and its contributors. |
| 17 | * 4. Neither the name of the University nor the names of its contributors |
| 18 | * may be used to endorse or promote products derived from this software |
| 19 | * without specific prior written permission. |
| 20 | * |
| 21 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 22 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 23 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 24 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 25 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 26 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 27 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 28 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 29 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 30 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 31 | * SUCH DAMAGE. |
| 32 | * |
| 33 | * @(#)mfs_vnops.c 8.11 (Berkeley) 5/22/95 |
| 34 | * $FreeBSD: src/sys/ufs/mfs/mfs_vnops.c,v 1.47.2.1 2001/05/22 02:06:43 bp Exp $ |
| 35 | * $DragonFly: src/sys/vfs/mfs/mfs_vnops.c,v 1.6 2003/06/26 05:55:21 dillon Exp $ |
| 36 | */ |
| 37 | |
| 38 | #include <sys/param.h> |
| 39 | #include <sys/systm.h> |
| 40 | #include <sys/kernel.h> |
| 41 | #include <sys/proc.h> |
| 42 | #include <sys/buf.h> |
| 43 | #include <sys/vnode.h> |
| 44 | #include <sys/malloc.h> |
| 45 | #include <sys/sysproto.h> |
| 46 | #include <sys/mman.h> |
| 47 | #include <sys/conf.h> |
| 48 | |
| 49 | #include <sys/buf2.h> |
| 50 | |
| 51 | #include <ufs/mfs/mfsnode.h> |
| 52 | #include <ufs/mfs/mfs_extern.h> |
| 53 | |
| 54 | static int mfs_badop __P((struct vop_generic_args *)); |
| 55 | static int mfs_bmap __P((struct vop_bmap_args *)); |
| 56 | static int mfs_close __P((struct vop_close_args *)); |
| 57 | static int mfs_fsync __P((struct vop_fsync_args *)); |
| 58 | static int mfs_freeblks __P((struct vop_freeblks_args *)); |
| 59 | static int mfs_inactive __P((struct vop_inactive_args *)); /* XXX */ |
| 60 | static int mfs_open __P((struct vop_open_args *)); |
| 61 | static int mfs_reclaim __P((struct vop_reclaim_args *)); /* XXX */ |
| 62 | static int mfs_print __P((struct vop_print_args *)); /* XXX */ |
| 63 | static int mfs_strategy __P((struct vop_strategy_args *)); /* XXX */ |
| 64 | static int mfs_getpages __P((struct vop_getpages_args *)); /* XXX */ |
| 65 | /* |
| 66 | * mfs vnode operations. |
| 67 | */ |
| 68 | vop_t **mfs_vnodeop_p; |
| 69 | static struct vnodeopv_entry_desc mfs_vnodeop_entries[] = { |
| 70 | { &vop_default_desc, (vop_t *) mfs_badop }, |
| 71 | { &vop_bmap_desc, (vop_t *) mfs_bmap }, |
| 72 | { &vop_bwrite_desc, (vop_t *) vop_defaultop }, |
| 73 | { &vop_close_desc, (vop_t *) mfs_close }, |
| 74 | { &vop_createvobject_desc, (vop_t *) vop_stdcreatevobject }, |
| 75 | { &vop_destroyvobject_desc, (vop_t *) vop_stddestroyvobject }, |
| 76 | { &vop_freeblks_desc, (vop_t *) mfs_freeblks }, |
| 77 | { &vop_fsync_desc, (vop_t *) mfs_fsync }, |
| 78 | { &vop_getpages_desc, (vop_t *) mfs_getpages }, |
| 79 | { &vop_getvobject_desc, (vop_t *) vop_stdgetvobject }, |
| 80 | { &vop_inactive_desc, (vop_t *) mfs_inactive }, |
| 81 | { &vop_ioctl_desc, (vop_t *) vop_enotty }, |
| 82 | { &vop_islocked_desc, (vop_t *) vop_defaultop }, |
| 83 | { &vop_lock_desc, (vop_t *) vop_defaultop }, |
| 84 | { &vop_open_desc, (vop_t *) mfs_open }, |
| 85 | { &vop_print_desc, (vop_t *) mfs_print }, |
| 86 | { &vop_reclaim_desc, (vop_t *) mfs_reclaim }, |
| 87 | { &vop_strategy_desc, (vop_t *) mfs_strategy }, |
| 88 | { &vop_unlock_desc, (vop_t *) vop_defaultop }, |
| 89 | { NULL, NULL } |
| 90 | }; |
| 91 | static struct vnodeopv_desc mfs_vnodeop_opv_desc = |
| 92 | { &mfs_vnodeop_p, mfs_vnodeop_entries }; |
| 93 | |
| 94 | VNODEOP_SET(mfs_vnodeop_opv_desc); |
| 95 | |
| 96 | /* |
| 97 | * Vnode Operations. |
| 98 | * |
| 99 | * Open called to allow memory filesystem to initialize and |
| 100 | * validate before actual IO. Record our process identifier |
| 101 | * so we can tell when we are doing I/O to ourself. |
| 102 | */ |
| 103 | /* ARGSUSED */ |
| 104 | static int |
| 105 | mfs_open(ap) |
| 106 | struct vop_open_args /* { |
| 107 | struct vnode *a_vp; |
| 108 | int a_mode; |
| 109 | struct ucred *a_cred; |
| 110 | struct thread *a_td; |
| 111 | } */ *ap; |
| 112 | { |
| 113 | |
| 114 | if (ap->a_vp->v_type != VBLK) { |
| 115 | panic("mfs_open not VBLK"); |
| 116 | /* NOTREACHED */ |
| 117 | } |
| 118 | return (0); |
| 119 | } |
| 120 | |
| 121 | static int |
| 122 | mfs_fsync(ap) |
| 123 | struct vop_fsync_args *ap; |
| 124 | { |
| 125 | |
| 126 | return (VOCALL(spec_vnodeop_p, VOFFSET(vop_fsync), ap)); |
| 127 | } |
| 128 | |
| 129 | /* |
| 130 | * mfs_freeblks() - hook to allow us to free physical memory. |
| 131 | * |
| 132 | * We implement the B_FREEBUF strategy. We can't just madvise() |
| 133 | * here because we have to do it in the correct order vs other bio |
| 134 | * requests, so we queue it. |
| 135 | * |
| 136 | * Note: geteblk() sets B_INVAL. We leave it set to guarentee buffer |
| 137 | * throw-away on brelse()? XXX |
| 138 | */ |
| 139 | |
| 140 | static int |
| 141 | mfs_freeblks(ap) |
| 142 | struct vop_freeblks_args /* { |
| 143 | struct vnode *a_vp; |
| 144 | daddr_t a_addr; |
| 145 | daddr_t a_length; |
| 146 | } */ *ap; |
| 147 | { |
| 148 | struct buf *bp; |
| 149 | struct vnode *vp; |
| 150 | |
| 151 | if (!vfinddev(ap->a_vp->v_rdev, VBLK, &vp) || vp->v_usecount == 0) |
| 152 | panic("mfs_freeblks: bad dev"); |
| 153 | |
| 154 | bp = geteblk(ap->a_length); |
| 155 | bp->b_flags |= B_FREEBUF | B_ASYNC; |
| 156 | bp->b_dev = ap->a_vp->v_rdev; |
| 157 | bp->b_blkno = ap->a_addr; |
| 158 | bp->b_offset = dbtob(ap->a_addr); |
| 159 | bp->b_bcount = ap->a_length; |
| 160 | BUF_KERNPROC(bp); |
| 161 | VOP_STRATEGY(vp, bp); |
| 162 | return(0); |
| 163 | } |
| 164 | |
| 165 | /* |
| 166 | * Pass I/O requests to the memory filesystem process. |
| 167 | */ |
| 168 | static int |
| 169 | mfs_strategy(ap) |
| 170 | struct vop_strategy_args /* { |
| 171 | struct vnode *a_vp; |
| 172 | struct buf *a_bp; |
| 173 | } */ *ap; |
| 174 | { |
| 175 | register struct buf *bp = ap->a_bp; |
| 176 | register struct mfsnode *mfsp; |
| 177 | struct vnode *vp; |
| 178 | struct thread *td = curthread; /* XXX */ |
| 179 | int s; |
| 180 | |
| 181 | if (!vfinddev(bp->b_dev, VBLK, &vp) || vp->v_usecount == 0) |
| 182 | panic("mfs_strategy: bad dev"); |
| 183 | mfsp = VTOMFS(vp); |
| 184 | |
| 185 | /* |
| 186 | * splbio required for queueing/dequeueing, in case of forwarded |
| 187 | * BPs from bio interrupts (?). It may not be necessary. |
| 188 | */ |
| 189 | |
| 190 | s = splbio(); |
| 191 | |
| 192 | if (mfsp->mfs_td == NULL) { |
| 193 | /* |
| 194 | * mini-root. Note: B_FREEBUF not supported at the moment, |
| 195 | * I'm not sure what kind of dataspace b_data is in. |
| 196 | */ |
| 197 | caddr_t base; |
| 198 | |
| 199 | base = mfsp->mfs_baseoff + (bp->b_blkno << DEV_BSHIFT); |
| 200 | if (bp->b_flags & B_FREEBUF) |
| 201 | ; |
| 202 | if (bp->b_flags & B_READ) |
| 203 | bcopy(base, bp->b_data, bp->b_bcount); |
| 204 | else |
| 205 | bcopy(bp->b_data, base, bp->b_bcount); |
| 206 | biodone(bp); |
| 207 | } else if (mfsp->mfs_td == td) { |
| 208 | /* |
| 209 | * VOP to self |
| 210 | */ |
| 211 | splx(s); |
| 212 | mfs_doio(bp, mfsp); |
| 213 | s = splbio(); |
| 214 | } else { |
| 215 | /* |
| 216 | * VOP from some other process, queue to MFS process and |
| 217 | * wake it up. |
| 218 | */ |
| 219 | bufq_insert_tail(&mfsp->buf_queue, bp); |
| 220 | wakeup((caddr_t)vp); |
| 221 | } |
| 222 | splx(s); |
| 223 | return (0); |
| 224 | } |
| 225 | |
| 226 | /* |
| 227 | * Memory file system I/O. |
| 228 | * |
| 229 | * Trivial on the HP since buffer has already been mapping into KVA space. |
| 230 | * |
| 231 | * Read and Write are handled with a simple copyin and copyout. |
| 232 | * |
| 233 | * We also partially support VOP_FREEBLKS() via B_FREEBUF. We can't implement |
| 234 | * completely -- for example, on fragments or inode metadata, but we can |
| 235 | * implement it for page-aligned requests. |
| 236 | */ |
| 237 | void |
| 238 | mfs_doio(bp, mfsp) |
| 239 | register struct buf *bp; |
| 240 | struct mfsnode *mfsp; |
| 241 | { |
| 242 | caddr_t base = mfsp->mfs_baseoff + (bp->b_blkno << DEV_BSHIFT); |
| 243 | |
| 244 | if (bp->b_flags & B_FREEBUF) { |
| 245 | /* |
| 246 | * Implement B_FREEBUF, which allows the filesystem to tell |
| 247 | * a block device when blocks are no longer needed (like when |
| 248 | * a file is deleted). We use the hook to MADV_FREE the VM. |
| 249 | * This makes an MFS filesystem work as well or better then |
| 250 | * a sun-style swap-mounted filesystem. |
| 251 | */ |
| 252 | int bytes = bp->b_bcount; |
| 253 | |
| 254 | if ((vm_offset_t)base & PAGE_MASK) { |
| 255 | int n = PAGE_SIZE - ((vm_offset_t)base & PAGE_MASK); |
| 256 | bytes -= n; |
| 257 | base += n; |
| 258 | } |
| 259 | if (bytes > 0) { |
| 260 | struct madvise_args uap; |
| 261 | |
| 262 | bytes &= ~PAGE_MASK; |
| 263 | if (bytes != 0) { |
| 264 | bzero(&uap, sizeof(uap)); |
| 265 | uap.addr = base; |
| 266 | uap.len = bytes; |
| 267 | uap.behav = MADV_FREE; |
| 268 | madvise(&uap); |
| 269 | } |
| 270 | } |
| 271 | bp->b_error = 0; |
| 272 | } else if (bp->b_flags & B_READ) { |
| 273 | /* |
| 274 | * Read data from our 'memory' disk |
| 275 | */ |
| 276 | bp->b_error = copyin(base, bp->b_data, bp->b_bcount); |
| 277 | } else { |
| 278 | /* |
| 279 | * Write data to our 'memory' disk |
| 280 | */ |
| 281 | bp->b_error = copyout(bp->b_data, base, bp->b_bcount); |
| 282 | } |
| 283 | if (bp->b_error) |
| 284 | bp->b_flags |= B_ERROR; |
| 285 | biodone(bp); |
| 286 | } |
| 287 | |
| 288 | /* |
| 289 | * This is a noop, simply returning what one has been given. |
| 290 | */ |
| 291 | static int |
| 292 | mfs_bmap(ap) |
| 293 | struct vop_bmap_args /* { |
| 294 | struct vnode *a_vp; |
| 295 | ufs_daddr_t a_bn; |
| 296 | struct vnode **a_vpp; |
| 297 | ufs_daddr_t *a_bnp; |
| 298 | int *a_runp; |
| 299 | } */ *ap; |
| 300 | { |
| 301 | |
| 302 | if (ap->a_vpp != NULL) |
| 303 | *ap->a_vpp = ap->a_vp; |
| 304 | if (ap->a_bnp != NULL) |
| 305 | *ap->a_bnp = ap->a_bn; |
| 306 | if (ap->a_runp != NULL) |
| 307 | *ap->a_runp = 0; |
| 308 | return (0); |
| 309 | } |
| 310 | |
| 311 | /* |
| 312 | * Memory filesystem close routine |
| 313 | */ |
| 314 | /* ARGSUSED */ |
| 315 | static int |
| 316 | mfs_close(ap) |
| 317 | struct vop_close_args /* { |
| 318 | struct vnode *a_vp; |
| 319 | int a_fflag; |
| 320 | struct ucred *a_cred; |
| 321 | struct thread *a_td; |
| 322 | } */ *ap; |
| 323 | { |
| 324 | register struct vnode *vp = ap->a_vp; |
| 325 | register struct mfsnode *mfsp = VTOMFS(vp); |
| 326 | register struct buf *bp; |
| 327 | int error; |
| 328 | |
| 329 | /* |
| 330 | * Finish any pending I/O requests. |
| 331 | */ |
| 332 | while ((bp = bufq_first(&mfsp->buf_queue)) != NULL) { |
| 333 | bufq_remove(&mfsp->buf_queue, bp); |
| 334 | mfs_doio(bp, mfsp); |
| 335 | wakeup((caddr_t)bp); |
| 336 | } |
| 337 | /* |
| 338 | * On last close of a memory filesystem |
| 339 | * we must invalidate any in core blocks, so that |
| 340 | * we can, free up its vnode. |
| 341 | */ |
| 342 | if ((error = vinvalbuf(vp, 1, ap->a_td, 0, 0)) != 0) |
| 343 | return (error); |
| 344 | /* |
| 345 | * There should be no way to have any more uses of this |
| 346 | * vnode, so if we find any other uses, it is a panic. |
| 347 | */ |
| 348 | if (vp->v_usecount > 1) |
| 349 | printf("mfs_close: ref count %d > 1\n", vp->v_usecount); |
| 350 | if (vp->v_usecount > 1 || (bufq_first(&mfsp->buf_queue) != NULL)) |
| 351 | panic("mfs_close"); |
| 352 | /* |
| 353 | * Send a request to the filesystem server to exit. |
| 354 | */ |
| 355 | mfsp->mfs_active = 0; |
| 356 | wakeup((caddr_t)vp); |
| 357 | return (0); |
| 358 | } |
| 359 | |
| 360 | /* |
| 361 | * Memory filesystem inactive routine |
| 362 | */ |
| 363 | /* ARGSUSED */ |
| 364 | static int |
| 365 | mfs_inactive(ap) |
| 366 | struct vop_inactive_args /* { |
| 367 | struct vnode *a_vp; |
| 368 | struct thread *a_td; |
| 369 | } */ *ap; |
| 370 | { |
| 371 | struct vnode *vp = ap->a_vp; |
| 372 | struct mfsnode *mfsp = VTOMFS(vp); |
| 373 | |
| 374 | if (bufq_first(&mfsp->buf_queue) != NULL) |
| 375 | panic("mfs_inactive: not inactive (next buffer %p)", |
| 376 | bufq_first(&mfsp->buf_queue)); |
| 377 | VOP_UNLOCK(vp, 0, ap->a_td); |
| 378 | return (0); |
| 379 | } |
| 380 | |
| 381 | /* |
| 382 | * Reclaim a memory filesystem devvp so that it can be reused. |
| 383 | */ |
| 384 | static int |
| 385 | mfs_reclaim(ap) |
| 386 | struct vop_reclaim_args /* { |
| 387 | struct vnode *a_vp; |
| 388 | } */ *ap; |
| 389 | { |
| 390 | register struct vnode *vp = ap->a_vp; |
| 391 | |
| 392 | FREE(vp->v_data, M_MFSNODE); |
| 393 | vp->v_data = NULL; |
| 394 | return (0); |
| 395 | } |
| 396 | |
| 397 | /* |
| 398 | * Print out the contents of an mfsnode. |
| 399 | */ |
| 400 | static int |
| 401 | mfs_print(ap) |
| 402 | struct vop_print_args /* { |
| 403 | struct vnode *a_vp; |
| 404 | } */ *ap; |
| 405 | { |
| 406 | register struct mfsnode *mfsp = VTOMFS(ap->a_vp); |
| 407 | |
| 408 | printf("tag VT_MFS, td %p, base %p, size %ld\n", |
| 409 | mfsp->mfs_td, (void *)mfsp->mfs_baseoff, mfsp->mfs_size); |
| 410 | return (0); |
| 411 | } |
| 412 | |
| 413 | /* |
| 414 | * Block device bad operation |
| 415 | */ |
| 416 | static int |
| 417 | mfs_badop(struct vop_generic_args *ap) |
| 418 | { |
| 419 | int i; |
| 420 | |
| 421 | printf("mfs_badop[%s]\n", ap->a_desc->vdesc_name); |
| 422 | i = vop_defaultop(ap); |
| 423 | printf("mfs_badop[%s] = %d\n", ap->a_desc->vdesc_name,i); |
| 424 | return (i); |
| 425 | } |
| 426 | |
| 427 | |
| 428 | static int |
| 429 | mfs_getpages(ap) |
| 430 | struct vop_getpages_args *ap; |
| 431 | { |
| 432 | |
| 433 | return (VOCALL(spec_vnodeop_p, VOFFSET(vop_getpages), ap)); |
| 434 | } |