sys/vfs/mfs/mfs_vnops.c

   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 1.37 2007/08/13 17:31:56 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 <vm/vm.h>
  50 #include <vm/vm_object.h>
  51 #include <vm/vm_page.h>
  52 #include <vm/vm_pager.h>
  53 #include <vm/vnode_pager.h>
  54
  55 #include <sys/buf2.h>
  56 #include <sys/thread2.h>
  57
  58 #include "mfsnode.h"
  59 #include "mfs_extern.h"
  60
  61 static int      mfs_badop (struct vop_generic_args *);
  62 static int      mfs_bmap (struct vop_bmap_args *);
  63 static int      mfs_close (struct vop_close_args *);
  64 static int      mfs_fsync (struct vop_fsync_args *);
  65 static int      mfs_freeblks (struct vop_freeblks_args *);
  66 static int      mfs_inactive (struct vop_inactive_args *); /* XXX */
  67 static int      mfs_open (struct vop_open_args *);
  68 static int      mfs_reclaim (struct vop_reclaim_args *); /* XXX */
  69 static int      mfs_print (struct vop_print_args *); /* XXX */
  70 static int      mfs_strategy (struct vop_strategy_args *); /* XXX */
  71 static int      mfs_getpages (struct vop_getpages_args *); /* XXX */
  72 /*
  73  * mfs vnode operations.  Note: the vops here are used for the MFS block
  74  * device, not for operations on files (MFS calls the ffs mount code for that)
  75  */
  76 static struct vop_ops mfs_vnode_vops = {
  77         .vop_default =          mfs_badop,
  78         .vop_bmap =             mfs_bmap,
  79         .vop_close =            mfs_close,
  80         .vop_freeblks =         mfs_freeblks,
  81         .vop_fsync =            mfs_fsync,
  82         .vop_getpages =         mfs_getpages,
  83         .vop_inactive =         mfs_inactive,
  84         .vop_ioctl =            (void *)vop_enotty,
  85         .vop_open =             mfs_open,
  86         .vop_print =            mfs_print,
  87         .vop_reclaim =          mfs_reclaim,
  88         .vop_strategy =         mfs_strategy,
  89 };
  90
  91 struct vop_ops *mfs_vnode_vops_p = &mfs_vnode_vops;
  92
  93 VNODEOP_SET(mfs_vnode_vops);
  94
  95 /*
  96  * Vnode Operations.
  97  *
  98  * Open called to allow memory filesystem to initialize and
  99  * validate before actual IO. Record our process identifier
 100  * so we can tell when we are doing I/O to ourself.
 101  *
 102  * NOTE: new device sequencing.  mounts check the device reference count
 103  * before calling open, so we must associate the device in open and
 104  * disassociate it in close rather then faking it when we created the vnode.
 105  *
 106  * mfs_open(struct vnode *a_vp, int a_mode, struct ucred *a_cred,
 107  *          struct file *a_fp)
 108  */
 109 /* ARGSUSED */
 110 static int
 111 mfs_open(struct vop_open_args *ap)
 112 {
 113         struct vnode *vp = ap->a_vp;
 114
 115         if (vp->v_type != VCHR)
 116                 panic("mfs_open not VCHR");
 117         v_associate_rdev(vp, get_dev(vp->v_umajor, vp->v_uminor));
 118         return (vop_stdopen(ap));
 119 }
 120
 121 static int
 122 mfs_fsync(struct vop_fsync_args *ap)
 123 {
 124         return (VOCALL(&spec_vnode_vops, &ap->a_head));
 125 }
 126
 127 /*
 128  * mfs_freeblks() - hook to allow us to free physical memory.
 129  *
 130  *      We implement the BUF_CMD_FREEBLKS strategy.  We can't just madvise()
 131  *      here because we have to do it in the correct order vs other bio
 132  *      requests, so we queue it.
 133  *
 134  *      Note: geteblk() sets B_INVAL.  We leave it set to guarentee buffer
 135  *      throw-away on brelse()? XXX
 136  *
 137  * mfs_freeblks(struct vnode *a_vp, daddr_t a_addr, daddr_t a_length)
 138  */
 139 static int
 140 mfs_freeblks(struct vop_freeblks_args *ap)
 141 {
 142         struct buf *bp;
 143         struct vnode *vp = ap->a_vp;
 144
 145         bp = geteblk(ap->a_length);
 146         bp->b_flags |= B_ASYNC;
 147         bp->b_cmd = BUF_CMD_FREEBLKS;
 148         bp->b_bio1.bio_offset = ap->a_offset;
 149         bp->b_bcount = ap->a_length;
 150         BUF_KERNPROC(bp);
 151         vn_strategy(vp, &bp->b_bio1);
 152         return(0);
 153 }
 154
 155 /*
 156  * Pass I/O requests to the memory filesystem process.
 157  *
 158  * mfs_strategy(struct vnode *a_vp, struct bio *a_bio)
 159  */
 160 static int
 161 mfs_strategy(struct vop_strategy_args *ap)
 162 {
 163         struct bio *bio = ap->a_bio;
 164         struct buf *bp = bio->bio_buf;
 165         struct mfsnode *mfsp;
 166         struct thread *td = curthread;          /* XXX */
 167
 168         mfsp = ap->a_vp->v_rdev->si_drv1;
 169         if (mfsp == NULL) {
 170                 bp->b_error = ENXIO;
 171                 bp->b_flags |= B_ERROR;
 172                 biodone(bio);
 173                 return(0);
 174         }
 175
 176         /*
 177          * splbio required for queueing/dequeueing, in case of forwarded
 178          * BPs from bio interrupts (?).  It may not be necessary.
 179          */
 180
 181         crit_enter();
 182
 183         if (mfsp->mfs_td == NULL) {
 184                 /*
 185                  * mini-root.  Note: BUF_CMD_FREEBLKS not supported at the
 186                  * moment, since we do not know what kind of dataspace
 187                  * b_data is in.
 188                  */
 189                 caddr_t base;
 190
 191                 base = mfsp->mfs_baseoff + bio->bio_offset;
 192                 switch(bp->b_cmd) {
 193                 case BUF_CMD_FREEBLKS:
 194                         break;
 195                 case BUF_CMD_READ:
 196                         bcopy(base, bp->b_data, bp->b_bcount);
 197                         break;
 198                 case BUF_CMD_WRITE:
 199                         bcopy(bp->b_data, base, bp->b_bcount);
 200                         break;
 201                 default:
 202                         panic("mfs: bad b_cmd %d\n", bp->b_cmd);
 203                 }
 204                 biodone(bio);
 205         } else if (mfsp->mfs_td == td) {
 206                 /*
 207                  * VOP to self
 208                  */
 209                 crit_exit();
 210                 mfs_doio(bio, mfsp);
 211                 crit_enter();
 212         } else {
 213                 /*
 214                  * VOP from some other process, queue to MFS process and
 215                  * wake it up.
 216                  */
 217                 bioq_insert_tail(&mfsp->bio_queue, bio);
 218                 wakeup((caddr_t)mfsp);
 219         }
 220         crit_exit();
 221         return (0);
 222 }
 223
 224 /*
 225  * Memory file system I/O.
 226  *
 227  * Trivial on the HP since buffer has already been mapping into KVA space.
 228  *
 229  * Read and Write are handled with a simple copyin and copyout.
 230  *
 231  * We also partially support VOP_FREEBLKS().  We can't implement
 232  * completely -- for example, on fragments or inode metadata, but we can
 233  * implement it for page-aligned requests.
 234  */
 235 void
 236 mfs_doio(struct bio *bio, struct mfsnode *mfsp)
 237 {
 238         struct buf *bp = bio->bio_buf;
 239         caddr_t base = mfsp->mfs_baseoff + bio->bio_offset;
 240         int bytes;
 241
 242         switch(bp->b_cmd) {
 243         case BUF_CMD_FREEBLKS:
 244                 /*
 245                  * Implement FREEBLKS, which allows the filesystem to tell
 246                  * a block device when blocks are no longer needed (like when
 247                  * a file is deleted).  We use the hook to MADV_FREE the VM.
 248                  * This makes an MFS filesystem work as well or better then
 249                  * a sun-style swap-mounted filesystem.
 250                  */
 251                 bytes = bp->b_bcount;
 252
 253                 if ((vm_offset_t)base & PAGE_MASK) {
 254                         int n = PAGE_SIZE - ((vm_offset_t)base & PAGE_MASK);
 255                         bytes -= n;
 256                         base += n;
 257                 }
 258                 if (bytes > 0) {
 259                         struct madvise_args uap;
 260
 261                         bytes &= ~PAGE_MASK;
 262                         if (bytes != 0) {
 263                                 bzero(&uap, sizeof(uap));
 264                                 uap.addr  = base;
 265                                 uap.len   = bytes;
 266                                 uap.behav = MADV_FREE;
 267                                 sys_madvise(&uap);
 268                         }
 269                 }
 270                 bp->b_error = 0;
 271                 break;
 272         case BUF_CMD_READ:
 273                 /*
 274                  * Read data from our 'memory' disk
 275                  */
 276                 bp->b_error = copyin(base, bp->b_data, bp->b_bcount);
 277                 break;
 278         case BUF_CMD_WRITE:
 279                 /*
 280                  * Write data to our 'memory' disk
 281                  */
 282                 bp->b_error = copyout(bp->b_data, base, bp->b_bcount);
 283                 break;
 284         default:
 285                 panic("mfs: bad b_cmd %d\n", bp->b_cmd);
 286         }
 287         if (bp->b_error)
 288                 bp->b_flags |= B_ERROR;
 289         biodone(bio);
 290 }
 291
 292 /*
 293  * This is a noop, simply returning what one has been given.
 294  *
 295  * mfs_bmap(struct vnode *a_vp, off_t a_loffset,
 296  *          off_t *a_doffsetp, int *a_runp, int *a_runb)
 297  */
 298 static int
 299 mfs_bmap(struct vop_bmap_args *ap)
 300 {
 301         if (ap->a_doffsetp != NULL)
 302                 *ap->a_doffsetp = ap->a_loffset;
 303         if (ap->a_runp != NULL)
 304                 *ap->a_runp = 0;
 305         if (ap->a_runb != NULL)
 306                 *ap->a_runb = 0;
 307         return (0);
 308 }
 309
 310 /*
 311  * Memory filesystem close routine
 312  *
 313  * mfs_close(struct vnode *a_vp, int a_fflag)
 314  */
 315 /* ARGSUSED */
 316 static int
 317 mfs_close(struct vop_close_args *ap)
 318 {
 319         struct vnode *vp = ap->a_vp;
 320         struct mfsnode *mfsp = VTOMFS(vp);
 321         struct bio *bio;
 322         int error = 0;
 323
 324         /*
 325          * Finish any pending I/O requests.
 326          */
 327         while ((bio = bioq_first(&mfsp->bio_queue)) != NULL) {
 328                 bioq_remove(&mfsp->bio_queue, bio);
 329                 mfs_doio(bio, mfsp);
 330                 wakeup((caddr_t)bio->bio_buf);
 331         }
 332
 333         /*
 334          * We really only care about the last close
 335          */
 336         if (vp->v_opencount > 1)
 337                 goto done;
 338
 339         /*
 340          * Synchronize any remaining buffers and then destroy them.
 341          */
 342         if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0)
 343                 goto done;
 344
 345         /*
 346          * Get rid of the pseudo-backing object.  Since the object is
 347          * not directly memory mapped, we don't have to worry about
 348          * synchronizing it.
 349          */
 350         if (vp->v_object)
 351                 vm_pager_deallocate(vp->v_object);
 352
 353         /*
 354          * There should be no way to have any more uses of this
 355          * vnode, so if we find any other uses, it is a panic.
 356          */
 357         if (vp->v_sysref.refcnt > 1)
 358                 kprintf("mfs_close: ref count %d > 1\n", vp->v_sysref.refcnt);
 359         if (vp->v_sysref.refcnt > 1 || (bioq_first(&mfsp->bio_queue) != NULL))
 360                 panic("mfs_close");
 361         /*
 362          * Send a request to the filesystem server to exit.
 363          */
 364         mfsp->mfs_active = 0;
 365         v_release_rdev(vp);
 366         if (mfsp->mfs_dev) {
 367                 destroy_dev(mfsp->mfs_dev);
 368                 mfsp->mfs_dev = NULL;
 369         }
 370         wakeup((caddr_t)mfsp);
 371 done:
 372         vop_stdclose(ap);
 373         return (error);
 374 }
 375
 376 /*
 377  * Memory filesystem inactive routine
 378  *
 379  * mfs_inactive(struct vnode *a_vp)
 380  */
 381 /* ARGSUSED */
 382 static int
 383 mfs_inactive(struct vop_inactive_args *ap)
 384 {
 385         struct vnode *vp = ap->a_vp;
 386         struct mfsnode *mfsp = VTOMFS(vp);
 387
 388         if (bioq_first(&mfsp->bio_queue) != NULL)
 389                 panic("mfs_inactive: not inactive (next buffer %p)",
 390                         bioq_first(&mfsp->bio_queue));
 391         return (0);
 392 }
 393
 394 /*
 395  * Reclaim a memory filesystem devvp so that it can be reused.
 396  *
 397  * mfs_reclaim(struct vnode *a_vp)
 398  */
 399 static int
 400 mfs_reclaim(struct vop_reclaim_args *ap)
 401 {
 402         struct vnode *vp = ap->a_vp;
 403
 404         FREE(vp->v_data, M_MFSNODE);
 405         vp->v_data = NULL;
 406         return (0);
 407 }
 408
 409 /*
 410  * Print out the contents of an mfsnode.
 411  *
 412  * mfs_print(struct vnode *a_vp)
 413  */
 414 static int
 415 mfs_print(struct vop_print_args *ap)
 416 {
 417         struct mfsnode *mfsp = VTOMFS(ap->a_vp);
 418
 419         kprintf("tag VT_MFS, td %p, base %p, size %ld\n",
 420             mfsp->mfs_td, (void *)mfsp->mfs_baseoff, mfsp->mfs_size);
 421         return (0);
 422 }
 423
 424 /*
 425  * Block device bad operation
 426  */
 427 static int
 428 mfs_badop(struct vop_generic_args *ap)
 429 {
 430         int i;
 431
 432         kprintf("mfs_badop[%s]\n", ap->a_desc->sd_name);
 433         i = vop_defaultop(ap);
 434         kprintf("mfs_badop[%s] = %d\n", ap->a_desc->sd_name, i);
 435         return (i);
 436 }
 437
 438 static int
 439 mfs_getpages(struct vop_getpages_args *ap)
 440 {
 441         return (VOCALL(&spec_vnode_vops, &ap->a_head));
 442 }