| 1 | /* |
| 2 | * Copyright (c) 2004-2006 The DragonFly Project. All rights reserved. |
| 3 | * |
| 4 | * This code is derived from software contributed to The DragonFly Project |
| 5 | * by Matthew Dillon <dillon@backplane.com> |
| 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 | * |
| 11 | * 1. Redistributions of source code must retain the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer. |
| 13 | * 2. Redistributions in binary form must reproduce the above copyright |
| 14 | * notice, this list of conditions and the following disclaimer in |
| 15 | * the documentation and/or other materials provided with the |
| 16 | * distribution. |
| 17 | * 3. Neither the name of The DragonFly Project nor the names of its |
| 18 | * contributors may be used to endorse or promote products derived |
| 19 | * from this software without specific, prior written permission. |
| 20 | * |
| 21 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 22 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 23 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
| 24 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
| 25 | * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 26 | * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, |
| 27 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
| 28 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED |
| 29 | * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 30 | * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT |
| 31 | * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 32 | * SUCH DAMAGE. |
| 33 | * |
| 34 | * $DragonFly: src/sys/kern/vfs_jops.c,v 1.31 2006/09/30 21:10:19 swildner Exp $ |
| 35 | */ |
| 36 | /* |
| 37 | * Each mount point may have zero or more independantly configured journals |
| 38 | * attached to it. Each journal is represented by a memory FIFO and worker |
| 39 | * thread. Journal events are streamed through the FIFO to the thread, |
| 40 | * batched up (typically on one-second intervals), and written out by the |
| 41 | * thread. |
| 42 | * |
| 43 | * Journal vnode ops are executed instead of mnt_vn_norm_ops when one or |
| 44 | * more journals have been installed on a mount point. It becomes the |
| 45 | * responsibility of the journal op to call the underlying normal op as |
| 46 | * appropriate. |
| 47 | */ |
| 48 | #include <sys/param.h> |
| 49 | #include <sys/systm.h> |
| 50 | #include <sys/buf.h> |
| 51 | #include <sys/conf.h> |
| 52 | #include <sys/kernel.h> |
| 53 | #include <sys/queue.h> |
| 54 | #include <sys/lock.h> |
| 55 | #include <sys/malloc.h> |
| 56 | #include <sys/mount.h> |
| 57 | #include <sys/unistd.h> |
| 58 | #include <sys/vnode.h> |
| 59 | #include <sys/poll.h> |
| 60 | #include <sys/mountctl.h> |
| 61 | #include <sys/journal.h> |
| 62 | #include <sys/file.h> |
| 63 | #include <sys/proc.h> |
| 64 | #include <sys/msfbuf.h> |
| 65 | #include <sys/socket.h> |
| 66 | #include <sys/socketvar.h> |
| 67 | |
| 68 | #include <machine/limits.h> |
| 69 | |
| 70 | #include <vm/vm.h> |
| 71 | #include <vm/vm_object.h> |
| 72 | #include <vm/vm_page.h> |
| 73 | #include <vm/vm_pager.h> |
| 74 | #include <vm/vnode_pager.h> |
| 75 | |
| 76 | #include <sys/file2.h> |
| 77 | #include <sys/thread2.h> |
| 78 | |
| 79 | static int journal_attach(struct mount *mp); |
| 80 | static void journal_detach(struct mount *mp); |
| 81 | static int journal_install_vfs_journal(struct mount *mp, struct file *fp, |
| 82 | const struct mountctl_install_journal *info); |
| 83 | static int journal_restart_vfs_journal(struct mount *mp, struct file *fp, |
| 84 | const struct mountctl_restart_journal *info); |
| 85 | static int journal_remove_vfs_journal(struct mount *mp, |
| 86 | const struct mountctl_remove_journal *info); |
| 87 | static int journal_restart(struct mount *mp, struct file *fp, |
| 88 | struct journal *jo, int flags); |
| 89 | static int journal_destroy(struct mount *mp, struct journal *jo, int flags); |
| 90 | static int journal_resync_vfs_journal(struct mount *mp, const void *ctl); |
| 91 | static int journal_status_vfs_journal(struct mount *mp, |
| 92 | const struct mountctl_status_journal *info, |
| 93 | struct mountctl_journal_ret_status *rstat, |
| 94 | int buflen, int *res); |
| 95 | |
| 96 | static void jrecord_undo_file(struct jrecord *jrec, struct vnode *vp, |
| 97 | int jrflags, off_t off, off_t bytes); |
| 98 | |
| 99 | static int journal_setattr(struct vop_setattr_args *ap); |
| 100 | static int journal_write(struct vop_write_args *ap); |
| 101 | static int journal_fsync(struct vop_fsync_args *ap); |
| 102 | static int journal_putpages(struct vop_putpages_args *ap); |
| 103 | static int journal_setacl(struct vop_setacl_args *ap); |
| 104 | static int journal_setextattr(struct vop_setextattr_args *ap); |
| 105 | static int journal_ncreate(struct vop_ncreate_args *ap); |
| 106 | static int journal_nmknod(struct vop_nmknod_args *ap); |
| 107 | static int journal_nlink(struct vop_nlink_args *ap); |
| 108 | static int journal_nsymlink(struct vop_nsymlink_args *ap); |
| 109 | static int journal_nwhiteout(struct vop_nwhiteout_args *ap); |
| 110 | static int journal_nremove(struct vop_nremove_args *ap); |
| 111 | static int journal_nmkdir(struct vop_nmkdir_args *ap); |
| 112 | static int journal_nrmdir(struct vop_nrmdir_args *ap); |
| 113 | static int journal_nrename(struct vop_nrename_args *ap); |
| 114 | |
| 115 | #define JRUNDO_SIZE 0x00000001 |
| 116 | #define JRUNDO_UID 0x00000002 |
| 117 | #define JRUNDO_GID 0x00000004 |
| 118 | #define JRUNDO_FSID 0x00000008 |
| 119 | #define JRUNDO_MODES 0x00000010 |
| 120 | #define JRUNDO_INUM 0x00000020 |
| 121 | #define JRUNDO_ATIME 0x00000040 |
| 122 | #define JRUNDO_MTIME 0x00000080 |
| 123 | #define JRUNDO_CTIME 0x00000100 |
| 124 | #define JRUNDO_GEN 0x00000200 |
| 125 | #define JRUNDO_FLAGS 0x00000400 |
| 126 | #define JRUNDO_UDEV 0x00000800 |
| 127 | #define JRUNDO_NLINK 0x00001000 |
| 128 | #define JRUNDO_FILEDATA 0x00010000 |
| 129 | #define JRUNDO_GETVP 0x00020000 |
| 130 | #define JRUNDO_CONDLINK 0x00040000 /* write file data if link count 1 */ |
| 131 | #define JRUNDO_VATTR (JRUNDO_SIZE|JRUNDO_UID|JRUNDO_GID|JRUNDO_FSID|\ |
| 132 | JRUNDO_MODES|JRUNDO_INUM|JRUNDO_ATIME|JRUNDO_MTIME|\ |
| 133 | JRUNDO_CTIME|JRUNDO_GEN|JRUNDO_FLAGS|JRUNDO_UDEV|\ |
| 134 | JRUNDO_NLINK) |
| 135 | #define JRUNDO_ALL (JRUNDO_VATTR|JRUNDO_FILEDATA) |
| 136 | |
| 137 | static struct vop_ops journal_vnode_vops = { |
| 138 | .vop_default = vop_journal_operate_ap, |
| 139 | .vop_mountctl = journal_mountctl, |
| 140 | .vop_setattr = journal_setattr, |
| 141 | .vop_write = journal_write, |
| 142 | .vop_fsync = journal_fsync, |
| 143 | .vop_putpages = journal_putpages, |
| 144 | .vop_setacl = journal_setacl, |
| 145 | .vop_setextattr = journal_setextattr, |
| 146 | .vop_ncreate = journal_ncreate, |
| 147 | .vop_nmknod = journal_nmknod, |
| 148 | .vop_nlink = journal_nlink, |
| 149 | .vop_nsymlink = journal_nsymlink, |
| 150 | .vop_nwhiteout = journal_nwhiteout, |
| 151 | .vop_nremove = journal_nremove, |
| 152 | .vop_nmkdir = journal_nmkdir, |
| 153 | .vop_nrmdir = journal_nrmdir, |
| 154 | .vop_nrename = journal_nrename |
| 155 | }; |
| 156 | |
| 157 | int |
| 158 | journal_mountctl(struct vop_mountctl_args *ap) |
| 159 | { |
| 160 | struct mount *mp; |
| 161 | int error = 0; |
| 162 | |
| 163 | mp = ap->a_head.a_ops->head.vv_mount; |
| 164 | KKASSERT(mp); |
| 165 | |
| 166 | if (mp->mnt_vn_journal_ops == NULL) { |
| 167 | switch(ap->a_op) { |
| 168 | case MOUNTCTL_INSTALL_VFS_JOURNAL: |
| 169 | error = journal_attach(mp); |
| 170 | if (error == 0 && ap->a_ctllen != sizeof(struct mountctl_install_journal)) |
| 171 | error = EINVAL; |
| 172 | if (error == 0 && ap->a_fp == NULL) |
| 173 | error = EBADF; |
| 174 | if (error == 0) |
| 175 | error = journal_install_vfs_journal(mp, ap->a_fp, ap->a_ctl); |
| 176 | if (TAILQ_EMPTY(&mp->mnt_jlist)) |
| 177 | journal_detach(mp); |
| 178 | break; |
| 179 | case MOUNTCTL_RESTART_VFS_JOURNAL: |
| 180 | case MOUNTCTL_REMOVE_VFS_JOURNAL: |
| 181 | case MOUNTCTL_RESYNC_VFS_JOURNAL: |
| 182 | case MOUNTCTL_STATUS_VFS_JOURNAL: |
| 183 | error = ENOENT; |
| 184 | break; |
| 185 | default: |
| 186 | error = EOPNOTSUPP; |
| 187 | break; |
| 188 | } |
| 189 | } else { |
| 190 | switch(ap->a_op) { |
| 191 | case MOUNTCTL_INSTALL_VFS_JOURNAL: |
| 192 | if (ap->a_ctllen != sizeof(struct mountctl_install_journal)) |
| 193 | error = EINVAL; |
| 194 | if (error == 0 && ap->a_fp == NULL) |
| 195 | error = EBADF; |
| 196 | if (error == 0) |
| 197 | error = journal_install_vfs_journal(mp, ap->a_fp, ap->a_ctl); |
| 198 | break; |
| 199 | case MOUNTCTL_RESTART_VFS_JOURNAL: |
| 200 | if (ap->a_ctllen != sizeof(struct mountctl_restart_journal)) |
| 201 | error = EINVAL; |
| 202 | if (error == 0 && ap->a_fp == NULL) |
| 203 | error = EBADF; |
| 204 | if (error == 0) |
| 205 | error = journal_restart_vfs_journal(mp, ap->a_fp, ap->a_ctl); |
| 206 | break; |
| 207 | case MOUNTCTL_REMOVE_VFS_JOURNAL: |
| 208 | if (ap->a_ctllen != sizeof(struct mountctl_remove_journal)) |
| 209 | error = EINVAL; |
| 210 | if (error == 0) |
| 211 | error = journal_remove_vfs_journal(mp, ap->a_ctl); |
| 212 | if (TAILQ_EMPTY(&mp->mnt_jlist)) |
| 213 | journal_detach(mp); |
| 214 | break; |
| 215 | case MOUNTCTL_RESYNC_VFS_JOURNAL: |
| 216 | if (ap->a_ctllen != 0) |
| 217 | error = EINVAL; |
| 218 | error = journal_resync_vfs_journal(mp, ap->a_ctl); |
| 219 | break; |
| 220 | case MOUNTCTL_STATUS_VFS_JOURNAL: |
| 221 | if (ap->a_ctllen != sizeof(struct mountctl_status_journal)) |
| 222 | error = EINVAL; |
| 223 | if (error == 0) { |
| 224 | error = journal_status_vfs_journal(mp, ap->a_ctl, |
| 225 | ap->a_buf, ap->a_buflen, ap->a_res); |
| 226 | } |
| 227 | break; |
| 228 | default: |
| 229 | error = EOPNOTSUPP; |
| 230 | break; |
| 231 | } |
| 232 | } |
| 233 | return (error); |
| 234 | } |
| 235 | |
| 236 | /* |
| 237 | * High level mount point setup. When a |
| 238 | */ |
| 239 | static int |
| 240 | journal_attach(struct mount *mp) |
| 241 | { |
| 242 | KKASSERT(mp->mnt_jbitmap == NULL); |
| 243 | vfs_add_vnodeops(mp, &journal_vnode_vops, &mp->mnt_vn_journal_ops); |
| 244 | mp->mnt_jbitmap = kmalloc(JREC_STREAMID_JMAX/8, M_JOURNAL, M_WAITOK|M_ZERO); |
| 245 | mp->mnt_streamid = JREC_STREAMID_JMIN; |
| 246 | return(0); |
| 247 | } |
| 248 | |
| 249 | static void |
| 250 | journal_detach(struct mount *mp) |
| 251 | { |
| 252 | KKASSERT(mp->mnt_jbitmap != NULL); |
| 253 | if (mp->mnt_vn_journal_ops) |
| 254 | vfs_rm_vnodeops(mp, &journal_vnode_vops, &mp->mnt_vn_journal_ops); |
| 255 | kfree(mp->mnt_jbitmap, M_JOURNAL); |
| 256 | mp->mnt_jbitmap = NULL; |
| 257 | } |
| 258 | |
| 259 | /* |
| 260 | * Install a journal on a mount point. Each journal has an associated worker |
| 261 | * thread which is responsible for buffering and spooling the data to the |
| 262 | * target. A mount point may have multiple journals attached to it. An |
| 263 | * initial start record is generated when the journal is associated. |
| 264 | */ |
| 265 | static int |
| 266 | journal_install_vfs_journal(struct mount *mp, struct file *fp, |
| 267 | const struct mountctl_install_journal *info) |
| 268 | { |
| 269 | struct journal *jo; |
| 270 | struct jrecord jrec; |
| 271 | int error = 0; |
| 272 | int size; |
| 273 | |
| 274 | jo = kmalloc(sizeof(struct journal), M_JOURNAL, M_WAITOK|M_ZERO); |
| 275 | bcopy(info->id, jo->id, sizeof(jo->id)); |
| 276 | jo->flags = info->flags & ~(MC_JOURNAL_WACTIVE | MC_JOURNAL_RACTIVE | |
| 277 | MC_JOURNAL_STOP_REQ); |
| 278 | |
| 279 | /* |
| 280 | * Memory FIFO size, round to nearest power of 2 |
| 281 | */ |
| 282 | if (info->membufsize) { |
| 283 | if (info->membufsize < 65536) |
| 284 | size = 65536; |
| 285 | else if (info->membufsize > 128 * 1024 * 1024) |
| 286 | size = 128 * 1024 * 1024; |
| 287 | else |
| 288 | size = (int)info->membufsize; |
| 289 | } else { |
| 290 | size = 1024 * 1024; |
| 291 | } |
| 292 | jo->fifo.size = 1; |
| 293 | while (jo->fifo.size < size) |
| 294 | jo->fifo.size <<= 1; |
| 295 | |
| 296 | /* |
| 297 | * Other parameters. If not specified the starting transaction id |
| 298 | * will be the current date. |
| 299 | */ |
| 300 | if (info->transid) { |
| 301 | jo->transid = info->transid; |
| 302 | } else { |
| 303 | struct timespec ts; |
| 304 | getnanotime(&ts); |
| 305 | jo->transid = ((int64_t)ts.tv_sec << 30) | ts.tv_nsec; |
| 306 | } |
| 307 | |
| 308 | jo->fp = fp; |
| 309 | |
| 310 | /* |
| 311 | * Allocate the memory FIFO |
| 312 | */ |
| 313 | jo->fifo.mask = jo->fifo.size - 1; |
| 314 | jo->fifo.membase = kmalloc(jo->fifo.size, M_JFIFO, M_WAITOK|M_ZERO|M_NULLOK); |
| 315 | if (jo->fifo.membase == NULL) |
| 316 | error = ENOMEM; |
| 317 | |
| 318 | /* |
| 319 | * Create the worker threads and generate the association record. |
| 320 | */ |
| 321 | if (error) { |
| 322 | kfree(jo, M_JOURNAL); |
| 323 | } else { |
| 324 | fhold(fp); |
| 325 | journal_create_threads(jo); |
| 326 | jrecord_init(jo, &jrec, JREC_STREAMID_DISCONT); |
| 327 | jrecord_write(&jrec, JTYPE_ASSOCIATE, 0); |
| 328 | jrecord_done(&jrec, 0); |
| 329 | TAILQ_INSERT_TAIL(&mp->mnt_jlist, jo, jentry); |
| 330 | } |
| 331 | return(error); |
| 332 | } |
| 333 | |
| 334 | /* |
| 335 | * Restart a journal with a new descriptor. The existing reader and writer |
| 336 | * threads are terminated and a new descriptor is associated with the |
| 337 | * journal. The FIFO rindex is reset to xindex and the threads are then |
| 338 | * restarted. |
| 339 | */ |
| 340 | static int |
| 341 | journal_restart_vfs_journal(struct mount *mp, struct file *fp, |
| 342 | const struct mountctl_restart_journal *info) |
| 343 | { |
| 344 | struct journal *jo; |
| 345 | int error; |
| 346 | |
| 347 | TAILQ_FOREACH(jo, &mp->mnt_jlist, jentry) { |
| 348 | if (bcmp(jo->id, info->id, sizeof(jo->id)) == 0) |
| 349 | break; |
| 350 | } |
| 351 | if (jo) |
| 352 | error = journal_restart(mp, fp, jo, info->flags); |
| 353 | else |
| 354 | error = EINVAL; |
| 355 | return (error); |
| 356 | } |
| 357 | |
| 358 | static int |
| 359 | journal_restart(struct mount *mp, struct file *fp, |
| 360 | struct journal *jo, int flags) |
| 361 | { |
| 362 | /* |
| 363 | * XXX lock the jo |
| 364 | */ |
| 365 | |
| 366 | #if 0 |
| 367 | /* |
| 368 | * Record the fact that we are doing a restart in the journal. |
| 369 | * XXX it isn't safe to do this if the journal is being restarted |
| 370 | * because it was locked up and the writer thread has already exited. |
| 371 | */ |
| 372 | jrecord_init(jo, &jrec, JREC_STREAMID_RESTART); |
| 373 | jrecord_write(&jrec, JTYPE_DISASSOCIATE, 0); |
| 374 | jrecord_done(&jrec, 0); |
| 375 | #endif |
| 376 | |
| 377 | /* |
| 378 | * Stop the reader and writer threads and clean up the current |
| 379 | * descriptor. |
| 380 | */ |
| 381 | printf("RESTART WITH FP %p KILLING %p\n", fp, jo->fp); |
| 382 | journal_destroy_threads(jo, flags); |
| 383 | |
| 384 | if (jo->fp) |
| 385 | fdrop(jo->fp); |
| 386 | |
| 387 | /* |
| 388 | * Associate the new descriptor, reset the FIFO index, and recreate |
| 389 | * the threads. |
| 390 | */ |
| 391 | fhold(fp); |
| 392 | jo->fp = fp; |
| 393 | jo->fifo.rindex = jo->fifo.xindex; |
| 394 | journal_create_threads(jo); |
| 395 | |
| 396 | return(0); |
| 397 | } |
| 398 | |
| 399 | /* |
| 400 | * Disassociate a journal from a mount point and terminate its worker thread. |
| 401 | * A final termination record is written out before the file pointer is |
| 402 | * dropped. |
| 403 | */ |
| 404 | static int |
| 405 | journal_remove_vfs_journal(struct mount *mp, |
| 406 | const struct mountctl_remove_journal *info) |
| 407 | { |
| 408 | struct journal *jo; |
| 409 | int error; |
| 410 | |
| 411 | TAILQ_FOREACH(jo, &mp->mnt_jlist, jentry) { |
| 412 | if (bcmp(jo->id, info->id, sizeof(jo->id)) == 0) |
| 413 | break; |
| 414 | } |
| 415 | if (jo) |
| 416 | error = journal_destroy(mp, jo, info->flags); |
| 417 | else |
| 418 | error = EINVAL; |
| 419 | return (error); |
| 420 | } |
| 421 | |
| 422 | /* |
| 423 | * Remove all journals associated with a mount point. Usually called |
| 424 | * by the umount code. |
| 425 | */ |
| 426 | void |
| 427 | journal_remove_all_journals(struct mount *mp, int flags) |
| 428 | { |
| 429 | struct journal *jo; |
| 430 | |
| 431 | while ((jo = TAILQ_FIRST(&mp->mnt_jlist)) != NULL) { |
| 432 | journal_destroy(mp, jo, flags); |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | static int |
| 437 | journal_destroy(struct mount *mp, struct journal *jo, int flags) |
| 438 | { |
| 439 | struct jrecord jrec; |
| 440 | |
| 441 | TAILQ_REMOVE(&mp->mnt_jlist, jo, jentry); |
| 442 | |
| 443 | jrecord_init(jo, &jrec, JREC_STREAMID_DISCONT); |
| 444 | jrecord_write(&jrec, JTYPE_DISASSOCIATE, 0); |
| 445 | jrecord_done(&jrec, 0); |
| 446 | |
| 447 | journal_destroy_threads(jo, flags); |
| 448 | |
| 449 | if (jo->fp) |
| 450 | fdrop(jo->fp); |
| 451 | if (jo->fifo.membase) |
| 452 | kfree(jo->fifo.membase, M_JFIFO); |
| 453 | kfree(jo, M_JOURNAL); |
| 454 | |
| 455 | return(0); |
| 456 | } |
| 457 | |
| 458 | static int |
| 459 | journal_resync_vfs_journal(struct mount *mp, const void *ctl) |
| 460 | { |
| 461 | return(EINVAL); |
| 462 | } |
| 463 | |
| 464 | static int |
| 465 | journal_status_vfs_journal(struct mount *mp, |
| 466 | const struct mountctl_status_journal *info, |
| 467 | struct mountctl_journal_ret_status *rstat, |
| 468 | int buflen, int *res) |
| 469 | { |
| 470 | struct journal *jo; |
| 471 | int error = 0; |
| 472 | int index; |
| 473 | |
| 474 | index = 0; |
| 475 | *res = 0; |
| 476 | TAILQ_FOREACH(jo, &mp->mnt_jlist, jentry) { |
| 477 | if (info->index == MC_JOURNAL_INDEX_ID) { |
| 478 | if (bcmp(jo->id, info->id, sizeof(jo->id)) != 0) |
| 479 | continue; |
| 480 | } else if (info->index >= 0) { |
| 481 | if (info->index < index) |
| 482 | continue; |
| 483 | } else if (info->index != MC_JOURNAL_INDEX_ALL) { |
| 484 | continue; |
| 485 | } |
| 486 | if (buflen < sizeof(*rstat)) { |
| 487 | if (*res) |
| 488 | rstat[-1].flags |= MC_JOURNAL_STATUS_MORETOCOME; |
| 489 | else |
| 490 | error = EINVAL; |
| 491 | break; |
| 492 | } |
| 493 | bzero(rstat, sizeof(*rstat)); |
| 494 | rstat->recsize = sizeof(*rstat); |
| 495 | bcopy(jo->id, rstat->id, sizeof(jo->id)); |
| 496 | rstat->index = index; |
| 497 | rstat->membufsize = jo->fifo.size; |
| 498 | rstat->membufused = jo->fifo.windex - jo->fifo.xindex; |
| 499 | rstat->membufunacked = jo->fifo.rindex - jo->fifo.xindex; |
| 500 | rstat->bytessent = jo->total_acked; |
| 501 | rstat->fifostalls = jo->fifostalls; |
| 502 | ++rstat; |
| 503 | ++index; |
| 504 | *res += sizeof(*rstat); |
| 505 | buflen -= sizeof(*rstat); |
| 506 | } |
| 507 | return(error); |
| 508 | } |
| 509 | |
| 510 | /************************************************************************ |
| 511 | * PARALLEL TRANSACTION SUPPORT ROUTINES * |
| 512 | ************************************************************************ |
| 513 | * |
| 514 | * JRECLIST_*() - routines which create and iterate over jrecord structures, |
| 515 | * because a mount point may have multiple attached journals. |
| 516 | */ |
| 517 | |
| 518 | /* |
| 519 | * Initialize the passed jrecord_list and create a jrecord for each |
| 520 | * journal we need to write to. Unnecessary mallocs are avoided by |
| 521 | * using the passed jrecord structure as the first jrecord in the list. |
| 522 | * A starting transaction is pushed for each jrecord. |
| 523 | * |
| 524 | * Returns non-zero if any of the journals require undo records. |
| 525 | */ |
| 526 | static |
| 527 | int |
| 528 | jreclist_init(struct mount *mp, struct jrecord_list *jreclist, |
| 529 | struct jrecord *jreccache, int16_t rectype) |
| 530 | { |
| 531 | struct journal *jo; |
| 532 | struct jrecord *jrec; |
| 533 | int wantrev; |
| 534 | int count; |
| 535 | int16_t streamid; |
| 536 | |
| 537 | TAILQ_INIT(&jreclist->list); |
| 538 | |
| 539 | /* |
| 540 | * Select the stream ID to use for the transaction. We must select |
| 541 | * a stream ID that is not currently in use by some other parallel |
| 542 | * transaction. |
| 543 | * |
| 544 | * Don't bother calculating the next streamid when reassigning |
| 545 | * mnt_streamid, since parallel transactions are fairly rare. This |
| 546 | * also allows someone observing the raw records to clearly see |
| 547 | * when parallel transactions occur. |
| 548 | */ |
| 549 | streamid = mp->mnt_streamid; |
| 550 | count = 0; |
| 551 | while (mp->mnt_jbitmap[streamid >> 3] & (1 << (streamid & 7))) { |
| 552 | if (++streamid == JREC_STREAMID_JMAX) |
| 553 | streamid = JREC_STREAMID_JMIN; |
| 554 | if (++count == JREC_STREAMID_JMAX - JREC_STREAMID_JMIN) { |
| 555 | printf("jreclist_init: all streamid's in use! sleeping\n"); |
| 556 | tsleep(jreclist, 0, "jsidfl", hz * 10); |
| 557 | count = 0; |
| 558 | } |
| 559 | } |
| 560 | mp->mnt_jbitmap[streamid >> 3] |= 1 << (streamid & 7); |
| 561 | mp->mnt_streamid = streamid; |
| 562 | jreclist->streamid = streamid; |
| 563 | |
| 564 | /* |
| 565 | * Now initialize a stream on each journal. |
| 566 | */ |
| 567 | count = 0; |
| 568 | wantrev = 0; |
| 569 | TAILQ_FOREACH(jo, &mp->mnt_jlist, jentry) { |
| 570 | if (count == 0) |
| 571 | jrec = jreccache; |
| 572 | else |
| 573 | jrec = kmalloc(sizeof(*jrec), M_JOURNAL, M_WAITOK); |
| 574 | jrecord_init(jo, jrec, streamid); |
| 575 | jrec->user_save = jrecord_push(jrec, rectype); |
| 576 | TAILQ_INSERT_TAIL(&jreclist->list, jrec, user_entry); |
| 577 | if (jo->flags & MC_JOURNAL_WANT_REVERSABLE) |
| 578 | wantrev = 1; |
| 579 | ++count; |
| 580 | } |
| 581 | return(wantrev); |
| 582 | } |
| 583 | |
| 584 | /* |
| 585 | * Terminate the journaled transactions started by jreclist_init(). If |
| 586 | * an error occured, the transaction records will be aborted. |
| 587 | */ |
| 588 | static |
| 589 | void |
| 590 | jreclist_done(struct mount *mp, struct jrecord_list *jreclist, int error) |
| 591 | { |
| 592 | struct jrecord *jrec; |
| 593 | int count; |
| 594 | |
| 595 | /* |
| 596 | * Cleanup the jrecord state on each journal. |
| 597 | */ |
| 598 | TAILQ_FOREACH(jrec, &jreclist->list, user_entry) { |
| 599 | jrecord_pop(jrec, jrec->user_save); |
| 600 | jrecord_done(jrec, error); |
| 601 | } |
| 602 | |
| 603 | /* |
| 604 | * Free allocated jrec's (the first is always supplied) |
| 605 | */ |
| 606 | count = 0; |
| 607 | while ((jrec = TAILQ_FIRST(&jreclist->list)) != NULL) { |
| 608 | TAILQ_REMOVE(&jreclist->list, jrec, user_entry); |
| 609 | if (count) |
| 610 | kfree(jrec, M_JOURNAL); |
| 611 | ++count; |
| 612 | } |
| 613 | |
| 614 | /* |
| 615 | * Clear the streamid so it can be reused. |
| 616 | */ |
| 617 | mp->mnt_jbitmap[jreclist->streamid >> 3] &= ~(1 << (jreclist->streamid & 7)); |
| 618 | } |
| 619 | |
| 620 | /* |
| 621 | * This procedure writes out UNDO records for available reversable |
| 622 | * journals. |
| 623 | * |
| 624 | * XXX could use improvement. There is no need to re-read the file |
| 625 | * for each journal. |
| 626 | */ |
| 627 | static |
| 628 | void |
| 629 | jreclist_undo_file(struct jrecord_list *jreclist, struct vnode *vp, |
| 630 | int jrflags, off_t off, off_t bytes) |
| 631 | { |
| 632 | struct jrecord *jrec; |
| 633 | int error; |
| 634 | |
| 635 | error = 0; |
| 636 | if (jrflags & JRUNDO_GETVP) |
| 637 | error = vget(vp, LK_SHARED); |
| 638 | if (error == 0) { |
| 639 | TAILQ_FOREACH(jrec, &jreclist->list, user_entry) { |
| 640 | if (jrec->jo->flags & MC_JOURNAL_WANT_REVERSABLE) { |
| 641 | jrecord_undo_file(jrec, vp, jrflags, off, bytes); |
| 642 | } |
| 643 | } |
| 644 | } |
| 645 | if (error == 0 && jrflags & JRUNDO_GETVP) |
| 646 | vput(vp); |
| 647 | } |
| 648 | |
| 649 | /************************************************************************ |
| 650 | * LOW LEVEL UNDO SUPPORT ROUTINE * |
| 651 | ************************************************************************ |
| 652 | * |
| 653 | * This function is used to support UNDO records. It will generate an |
| 654 | * appropriate record with the requested portion of the file data. Note |
| 655 | * that file data is only recorded if JRUNDO_FILEDATA is passed. If bytes |
| 656 | * is -1, it will be set to the size of the file. |
| 657 | */ |
| 658 | static void |
| 659 | jrecord_undo_file(struct jrecord *jrec, struct vnode *vp, int jrflags, |
| 660 | off_t off, off_t bytes) |
| 661 | { |
| 662 | struct vattr attr; |
| 663 | void *save1; /* warning, save pointers do not always remain valid */ |
| 664 | void *save2; |
| 665 | int error; |
| 666 | |
| 667 | /* |
| 668 | * Setup. Start the UNDO record, obtain a shared lock on the vnode, |
| 669 | * and retrieve attribute info. |
| 670 | */ |
| 671 | save1 = jrecord_push(jrec, JTYPE_UNDO); |
| 672 | error = VOP_GETATTR(vp, &attr); |
| 673 | if (error) |
| 674 | goto done; |
| 675 | |
| 676 | /* |
| 677 | * Generate UNDO records as requested. |
| 678 | */ |
| 679 | if (jrflags & JRUNDO_VATTR) { |
| 680 | save2 = jrecord_push(jrec, JTYPE_VATTR); |
| 681 | jrecord_leaf(jrec, JLEAF_VTYPE, &attr.va_type, sizeof(attr.va_type)); |
| 682 | if ((jrflags & JRUNDO_NLINK) && attr.va_nlink != VNOVAL) |
| 683 | jrecord_leaf(jrec, JLEAF_NLINK, &attr.va_nlink, sizeof(attr.va_nlink)); |
| 684 | if ((jrflags & JRUNDO_SIZE) && attr.va_size != VNOVAL) |
| 685 | jrecord_leaf(jrec, JLEAF_SIZE, &attr.va_size, sizeof(attr.va_size)); |
| 686 | if ((jrflags & JRUNDO_UID) && attr.va_uid != VNOVAL) |
| 687 | jrecord_leaf(jrec, JLEAF_UID, &attr.va_uid, sizeof(attr.va_uid)); |
| 688 | if ((jrflags & JRUNDO_GID) && attr.va_gid != VNOVAL) |
| 689 | jrecord_leaf(jrec, JLEAF_GID, &attr.va_gid, sizeof(attr.va_gid)); |
| 690 | if ((jrflags & JRUNDO_FSID) && attr.va_fsid != VNOVAL) |
| 691 | jrecord_leaf(jrec, JLEAF_FSID, &attr.va_fsid, sizeof(attr.va_fsid)); |
| 692 | if ((jrflags & JRUNDO_MODES) && attr.va_mode != (mode_t)VNOVAL) |
| 693 | jrecord_leaf(jrec, JLEAF_MODES, &attr.va_mode, sizeof(attr.va_mode)); |
| 694 | if ((jrflags & JRUNDO_INUM) && attr.va_fileid != VNOVAL) |
| 695 | jrecord_leaf(jrec, JLEAF_INUM, &attr.va_fileid, sizeof(attr.va_fileid)); |
| 696 | if ((jrflags & JRUNDO_ATIME) && attr.va_atime.tv_sec != VNOVAL) |
| 697 | jrecord_leaf(jrec, JLEAF_ATIME, &attr.va_atime, sizeof(attr.va_atime)); |
| 698 | if ((jrflags & JRUNDO_MTIME) && attr.va_mtime.tv_sec != VNOVAL) |
| 699 | jrecord_leaf(jrec, JLEAF_MTIME, &attr.va_mtime, sizeof(attr.va_mtime)); |
| 700 | if ((jrflags & JRUNDO_CTIME) && attr.va_ctime.tv_sec != VNOVAL) |
| 701 | jrecord_leaf(jrec, JLEAF_CTIME, &attr.va_ctime, sizeof(attr.va_ctime)); |
| 702 | if ((jrflags & JRUNDO_GEN) && attr.va_gen != VNOVAL) |
| 703 | jrecord_leaf(jrec, JLEAF_GEN, &attr.va_gen, sizeof(attr.va_gen)); |
| 704 | if ((jrflags & JRUNDO_FLAGS) && attr.va_flags != VNOVAL) |
| 705 | jrecord_leaf(jrec, JLEAF_FLAGS, &attr.va_flags, sizeof(attr.va_flags)); |
| 706 | if ((jrflags & JRUNDO_UDEV) && attr.va_rdev != VNOVAL) |
| 707 | jrecord_leaf(jrec, JLEAF_UDEV, &attr.va_rdev, sizeof(attr.va_rdev)); |
| 708 | jrecord_pop(jrec, save2); |
| 709 | } |
| 710 | |
| 711 | /* |
| 712 | * Output the file data being overwritten by reading the file and |
| 713 | * writing it out to the journal prior to the write operation. We |
| 714 | * do not need to write out data past the current file EOF. |
| 715 | * |
| 716 | * XXX support JRUNDO_CONDLINK - do not write out file data for files |
| 717 | * with a link count > 1. The undo code needs to locate the inode and |
| 718 | * regenerate the hardlink. |
| 719 | */ |
| 720 | if ((jrflags & JRUNDO_FILEDATA) && attr.va_type == VREG) { |
| 721 | if (attr.va_size != VNOVAL) { |
| 722 | if (bytes == -1) |
| 723 | bytes = attr.va_size - off; |
| 724 | if (off + bytes > attr.va_size) |
| 725 | bytes = attr.va_size - off; |
| 726 | if (bytes > 0) |
| 727 | jrecord_file_data(jrec, vp, off, bytes); |
| 728 | } else { |
| 729 | error = EINVAL; |
| 730 | } |
| 731 | } |
| 732 | if ((jrflags & JRUNDO_FILEDATA) && attr.va_type == VLNK) { |
| 733 | struct iovec aiov; |
| 734 | struct uio auio; |
| 735 | char *buf; |
| 736 | |
| 737 | buf = kmalloc(PATH_MAX, M_JOURNAL, M_WAITOK); |
| 738 | aiov.iov_base = buf; |
| 739 | aiov.iov_len = PATH_MAX; |
| 740 | auio.uio_iov = &aiov; |
| 741 | auio.uio_iovcnt = 1; |
| 742 | auio.uio_offset = 0; |
| 743 | auio.uio_rw = UIO_READ; |
| 744 | auio.uio_segflg = UIO_SYSSPACE; |
| 745 | auio.uio_td = curthread; |
| 746 | auio.uio_resid = PATH_MAX; |
| 747 | error = VOP_READLINK(vp, &auio, proc0.p_ucred); |
| 748 | if (error == 0) { |
| 749 | jrecord_leaf(jrec, JLEAF_SYMLINKDATA, buf, |
| 750 | PATH_MAX - auio.uio_resid); |
| 751 | } |
| 752 | kfree(buf, M_JOURNAL); |
| 753 | } |
| 754 | done: |
| 755 | if (error) |
| 756 | jrecord_leaf(jrec, JLEAF_ERROR, &error, sizeof(error)); |
| 757 | jrecord_pop(jrec, save1); |
| 758 | } |
| 759 | |
| 760 | /************************************************************************ |
| 761 | * JOURNAL VNOPS * |
| 762 | ************************************************************************ |
| 763 | * |
| 764 | * These are function shims replacing the normal filesystem ops. We become |
| 765 | * responsible for calling the underlying filesystem ops. We have the choice |
| 766 | * of executing the underlying op first and then generating the journal entry, |
| 767 | * or starting the journal entry, executing the underlying op, and then |
| 768 | * either completing or aborting it. |
| 769 | * |
| 770 | * The journal is supposed to be a high-level entity, which generally means |
| 771 | * identifying files by name rather then by inode. Supplying both allows |
| 772 | * the journal to be used both for inode-number-compatible 'mirrors' and |
| 773 | * for simple filesystem replication. |
| 774 | * |
| 775 | * Writes are particularly difficult to deal with because a single write may |
| 776 | * represent a hundred megabyte buffer or more, and both writes and truncations |
| 777 | * require the 'old' data to be written out as well as the new data if the |
| 778 | * log is reversable. Other issues: |
| 779 | * |
| 780 | * - How to deal with operations on unlinked files (no path available), |
| 781 | * but which may still be filesystem visible due to hard links. |
| 782 | * |
| 783 | * - How to deal with modifications made via a memory map. |
| 784 | * |
| 785 | * - Future cache coherency support will require cache coherency API calls |
| 786 | * both prior to and after the call to the underlying VFS. |
| 787 | * |
| 788 | * ALSO NOTE: We do not have to shim compatibility VOPs like MKDIR which have |
| 789 | * new VFS equivalents (NMKDIR). |
| 790 | */ |
| 791 | |
| 792 | /* |
| 793 | * Journal vop_settattr { a_vp, a_vap, a_cred, a_td } |
| 794 | */ |
| 795 | static |
| 796 | int |
| 797 | journal_setattr(struct vop_setattr_args *ap) |
| 798 | { |
| 799 | struct jrecord_list jreclist; |
| 800 | struct jrecord jreccache; |
| 801 | struct jrecord *jrec; |
| 802 | struct mount *mp; |
| 803 | void *save; |
| 804 | int error; |
| 805 | |
| 806 | mp = ap->a_head.a_ops->head.vv_mount; |
| 807 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_SETATTR)) { |
| 808 | jreclist_undo_file(&jreclist, ap->a_vp, JRUNDO_VATTR, 0, 0); |
| 809 | } |
| 810 | error = vop_journal_operate_ap(&ap->a_head); |
| 811 | if (error == 0) { |
| 812 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 813 | jrecord_write_cred(jrec, curthread, ap->a_cred); |
| 814 | jrecord_write_vnode_ref(jrec, ap->a_vp); |
| 815 | save = jrecord_push(jrec, JTYPE_REDO); |
| 816 | jrecord_write_vattr(jrec, ap->a_vap); |
| 817 | jrecord_pop(jrec, save); |
| 818 | } |
| 819 | } |
| 820 | jreclist_done(mp, &jreclist, error); |
| 821 | return (error); |
| 822 | } |
| 823 | |
| 824 | /* |
| 825 | * Journal vop_write { a_vp, a_uio, a_ioflag, a_cred } |
| 826 | */ |
| 827 | static |
| 828 | int |
| 829 | journal_write(struct vop_write_args *ap) |
| 830 | { |
| 831 | struct jrecord_list jreclist; |
| 832 | struct jrecord jreccache; |
| 833 | struct jrecord *jrec; |
| 834 | struct mount *mp; |
| 835 | struct uio uio_copy; |
| 836 | struct iovec uio_one_iovec; |
| 837 | void *save; |
| 838 | int error; |
| 839 | |
| 840 | /* |
| 841 | * This is really nasty. UIO's don't retain sufficient information to |
| 842 | * be reusable once they've gone through the VOP chain. The iovecs get |
| 843 | * cleared, so we have to copy the UIO. |
| 844 | * |
| 845 | * XXX fix the UIO code to not destroy iov's during a scan so we can |
| 846 | * reuse the uio over and over again. |
| 847 | * |
| 848 | * XXX UNDO code needs to journal the old data prior to the write. |
| 849 | */ |
| 850 | uio_copy = *ap->a_uio; |
| 851 | if (uio_copy.uio_iovcnt == 1) { |
| 852 | uio_one_iovec = ap->a_uio->uio_iov[0]; |
| 853 | uio_copy.uio_iov = &uio_one_iovec; |
| 854 | } else { |
| 855 | uio_copy.uio_iov = kmalloc(uio_copy.uio_iovcnt * sizeof(struct iovec), |
| 856 | M_JOURNAL, M_WAITOK); |
| 857 | bcopy(ap->a_uio->uio_iov, uio_copy.uio_iov, |
| 858 | uio_copy.uio_iovcnt * sizeof(struct iovec)); |
| 859 | } |
| 860 | |
| 861 | /* |
| 862 | * Write out undo data. Note that uio_offset is incorrect if |
| 863 | * IO_APPEND is set, but fortunately we have no undo file data to |
| 864 | * write out in that case. |
| 865 | */ |
| 866 | mp = ap->a_head.a_ops->head.vv_mount; |
| 867 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_WRITE)) { |
| 868 | if (ap->a_ioflag & IO_APPEND) { |
| 869 | jreclist_undo_file(&jreclist, ap->a_vp, JRUNDO_SIZE|JRUNDO_MTIME, 0, 0); |
| 870 | } else { |
| 871 | jreclist_undo_file(&jreclist, ap->a_vp, |
| 872 | JRUNDO_FILEDATA|JRUNDO_SIZE|JRUNDO_MTIME, |
| 873 | uio_copy.uio_offset, uio_copy.uio_resid); |
| 874 | } |
| 875 | } |
| 876 | error = vop_journal_operate_ap(&ap->a_head); |
| 877 | |
| 878 | /* |
| 879 | * XXX bad hack to figure out the offset for O_APPEND writes (note: |
| 880 | * uio field state after the VFS operation). |
| 881 | */ |
| 882 | uio_copy.uio_offset = ap->a_uio->uio_offset - |
| 883 | (uio_copy.uio_resid - ap->a_uio->uio_resid); |
| 884 | |
| 885 | /* |
| 886 | * Output the write data to the journal. |
| 887 | */ |
| 888 | if (error == 0) { |
| 889 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 890 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 891 | jrecord_write_vnode_ref(jrec, ap->a_vp); |
| 892 | save = jrecord_push(jrec, JTYPE_REDO); |
| 893 | jrecord_write_uio(jrec, JLEAF_FILEDATA, &uio_copy); |
| 894 | jrecord_pop(jrec, save); |
| 895 | } |
| 896 | } |
| 897 | jreclist_done(mp, &jreclist, error); |
| 898 | |
| 899 | if (uio_copy.uio_iov != &uio_one_iovec) |
| 900 | kfree(uio_copy.uio_iov, M_JOURNAL); |
| 901 | return (error); |
| 902 | } |
| 903 | |
| 904 | /* |
| 905 | * Journal vop_fsync { a_vp, a_waitfor, a_td } |
| 906 | */ |
| 907 | static |
| 908 | int |
| 909 | journal_fsync(struct vop_fsync_args *ap) |
| 910 | { |
| 911 | #if 0 |
| 912 | struct mount *mp; |
| 913 | struct journal *jo; |
| 914 | #endif |
| 915 | int error; |
| 916 | |
| 917 | error = vop_journal_operate_ap(&ap->a_head); |
| 918 | #if 0 |
| 919 | mp = ap->a_head.a_ops->head.vv_mount; |
| 920 | if (error == 0) { |
| 921 | TAILQ_FOREACH(jo, &mp->mnt_jlist, jentry) { |
| 922 | /* XXX synchronize pending journal records */ |
| 923 | } |
| 924 | } |
| 925 | #endif |
| 926 | return (error); |
| 927 | } |
| 928 | |
| 929 | /* |
| 930 | * Journal vop_putpages { a_vp, a_m, a_count, a_sync, a_rtvals, a_offset } |
| 931 | * |
| 932 | * note: a_count is in bytes. |
| 933 | */ |
| 934 | static |
| 935 | int |
| 936 | journal_putpages(struct vop_putpages_args *ap) |
| 937 | { |
| 938 | struct jrecord_list jreclist; |
| 939 | struct jrecord jreccache; |
| 940 | struct jrecord *jrec; |
| 941 | struct mount *mp; |
| 942 | void *save; |
| 943 | int error; |
| 944 | |
| 945 | mp = ap->a_head.a_ops->head.vv_mount; |
| 946 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_PUTPAGES) && |
| 947 | ap->a_count > 0 |
| 948 | ) { |
| 949 | jreclist_undo_file(&jreclist, ap->a_vp, |
| 950 | JRUNDO_FILEDATA|JRUNDO_SIZE|JRUNDO_MTIME, |
| 951 | ap->a_offset, btoc(ap->a_count)); |
| 952 | } |
| 953 | error = vop_journal_operate_ap(&ap->a_head); |
| 954 | if (error == 0 && ap->a_count > 0) { |
| 955 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 956 | jrecord_write_vnode_ref(jrec, ap->a_vp); |
| 957 | save = jrecord_push(jrec, JTYPE_REDO); |
| 958 | jrecord_write_pagelist(jrec, JLEAF_FILEDATA, ap->a_m, ap->a_rtvals, |
| 959 | btoc(ap->a_count), ap->a_offset); |
| 960 | jrecord_pop(jrec, save); |
| 961 | } |
| 962 | } |
| 963 | jreclist_done(mp, &jreclist, error); |
| 964 | return (error); |
| 965 | } |
| 966 | |
| 967 | /* |
| 968 | * Journal vop_setacl { a_vp, a_type, a_aclp, a_cred, a_td } |
| 969 | */ |
| 970 | static |
| 971 | int |
| 972 | journal_setacl(struct vop_setacl_args *ap) |
| 973 | { |
| 974 | struct jrecord_list jreclist; |
| 975 | struct jrecord jreccache; |
| 976 | struct jrecord *jrec; |
| 977 | struct mount *mp; |
| 978 | int error; |
| 979 | |
| 980 | mp = ap->a_head.a_ops->head.vv_mount; |
| 981 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_SETACL); |
| 982 | error = vop_journal_operate_ap(&ap->a_head); |
| 983 | if (error == 0) { |
| 984 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 985 | #if 0 |
| 986 | if ((jo->flags & MC_JOURNAL_WANT_REVERSABLE)) |
| 987 | jrecord_undo_file(jrec, ap->a_vp, JRUNDO_XXX, 0, 0); |
| 988 | #endif |
| 989 | jrecord_write_cred(jrec, curthread, ap->a_cred); |
| 990 | jrecord_write_vnode_ref(jrec, ap->a_vp); |
| 991 | #if 0 |
| 992 | save = jrecord_push(jrec, JTYPE_REDO); |
| 993 | /* XXX type, aclp */ |
| 994 | jrecord_pop(jrec, save); |
| 995 | #endif |
| 996 | } |
| 997 | } |
| 998 | jreclist_done(mp, &jreclist, error); |
| 999 | return (error); |
| 1000 | } |
| 1001 | |
| 1002 | /* |
| 1003 | * Journal vop_setextattr { a_vp, a_name, a_uio, a_cred, a_td } |
| 1004 | */ |
| 1005 | static |
| 1006 | int |
| 1007 | journal_setextattr(struct vop_setextattr_args *ap) |
| 1008 | { |
| 1009 | struct jrecord_list jreclist; |
| 1010 | struct jrecord jreccache; |
| 1011 | struct jrecord *jrec; |
| 1012 | struct mount *mp; |
| 1013 | void *save; |
| 1014 | int error; |
| 1015 | |
| 1016 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1017 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_SETEXTATTR); |
| 1018 | error = vop_journal_operate_ap(&ap->a_head); |
| 1019 | if (error == 0) { |
| 1020 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1021 | #if 0 |
| 1022 | if ((jo->flags & MC_JOURNAL_WANT_REVERSABLE)) |
| 1023 | jrecord_undo_file(jrec, ap->a_vp, JRUNDO_XXX, 0, 0); |
| 1024 | #endif |
| 1025 | jrecord_write_cred(jrec, curthread, ap->a_cred); |
| 1026 | jrecord_write_vnode_ref(jrec, ap->a_vp); |
| 1027 | jrecord_leaf(jrec, JLEAF_ATTRNAME, ap->a_name, strlen(ap->a_name)); |
| 1028 | save = jrecord_push(jrec, JTYPE_REDO); |
| 1029 | jrecord_write_uio(jrec, JLEAF_FILEDATA, ap->a_uio); |
| 1030 | jrecord_pop(jrec, save); |
| 1031 | } |
| 1032 | } |
| 1033 | jreclist_done(mp, &jreclist, error); |
| 1034 | return (error); |
| 1035 | } |
| 1036 | |
| 1037 | /* |
| 1038 | * Journal vop_ncreate { a_ncp, a_vpp, a_cred, a_vap } |
| 1039 | */ |
| 1040 | static |
| 1041 | int |
| 1042 | journal_ncreate(struct vop_ncreate_args *ap) |
| 1043 | { |
| 1044 | struct jrecord_list jreclist; |
| 1045 | struct jrecord jreccache; |
| 1046 | struct jrecord *jrec; |
| 1047 | struct mount *mp; |
| 1048 | void *save; |
| 1049 | int error; |
| 1050 | |
| 1051 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1052 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_CREATE); |
| 1053 | error = vop_journal_operate_ap(&ap->a_head); |
| 1054 | if (error == 0) { |
| 1055 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1056 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1057 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1058 | if (*ap->a_vpp) |
| 1059 | jrecord_write_vnode_ref(jrec, *ap->a_vpp); |
| 1060 | save = jrecord_push(jrec, JTYPE_REDO); |
| 1061 | jrecord_write_vattr(jrec, ap->a_vap); |
| 1062 | jrecord_pop(jrec, save); |
| 1063 | } |
| 1064 | } |
| 1065 | jreclist_done(mp, &jreclist, error); |
| 1066 | return (error); |
| 1067 | } |
| 1068 | |
| 1069 | /* |
| 1070 | * Journal vop_nmknod { a_ncp, a_vpp, a_cred, a_vap } |
| 1071 | */ |
| 1072 | static |
| 1073 | int |
| 1074 | journal_nmknod(struct vop_nmknod_args *ap) |
| 1075 | { |
| 1076 | struct jrecord_list jreclist; |
| 1077 | struct jrecord jreccache; |
| 1078 | struct jrecord *jrec; |
| 1079 | struct mount *mp; |
| 1080 | void *save; |
| 1081 | int error; |
| 1082 | |
| 1083 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1084 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_MKNOD); |
| 1085 | error = vop_journal_operate_ap(&ap->a_head); |
| 1086 | if (error == 0) { |
| 1087 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1088 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1089 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1090 | save = jrecord_push(jrec, JTYPE_REDO); |
| 1091 | jrecord_write_vattr(jrec, ap->a_vap); |
| 1092 | jrecord_pop(jrec, save); |
| 1093 | if (*ap->a_vpp) |
| 1094 | jrecord_write_vnode_ref(jrec, *ap->a_vpp); |
| 1095 | } |
| 1096 | } |
| 1097 | jreclist_done(mp, &jreclist, error); |
| 1098 | return (error); |
| 1099 | } |
| 1100 | |
| 1101 | /* |
| 1102 | * Journal vop_nlink { a_ncp, a_vp, a_cred } |
| 1103 | */ |
| 1104 | static |
| 1105 | int |
| 1106 | journal_nlink(struct vop_nlink_args *ap) |
| 1107 | { |
| 1108 | struct jrecord_list jreclist; |
| 1109 | struct jrecord jreccache; |
| 1110 | struct jrecord *jrec; |
| 1111 | struct mount *mp; |
| 1112 | void *save; |
| 1113 | int error; |
| 1114 | |
| 1115 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1116 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_LINK); |
| 1117 | error = vop_journal_operate_ap(&ap->a_head); |
| 1118 | if (error == 0) { |
| 1119 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1120 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1121 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1122 | /* XXX PATH to VP and inode number */ |
| 1123 | /* XXX this call may not record the correct path when |
| 1124 | * multiple paths are available */ |
| 1125 | save = jrecord_push(jrec, JTYPE_REDO); |
| 1126 | jrecord_write_vnode_link(jrec, ap->a_vp, ap->a_ncp); |
| 1127 | jrecord_pop(jrec, save); |
| 1128 | } |
| 1129 | } |
| 1130 | jreclist_done(mp, &jreclist, error); |
| 1131 | return (error); |
| 1132 | } |
| 1133 | |
| 1134 | /* |
| 1135 | * Journal vop_symlink { a_ncp, a_vpp, a_cred, a_vap, a_target } |
| 1136 | */ |
| 1137 | static |
| 1138 | int |
| 1139 | journal_nsymlink(struct vop_nsymlink_args *ap) |
| 1140 | { |
| 1141 | struct jrecord_list jreclist; |
| 1142 | struct jrecord jreccache; |
| 1143 | struct jrecord *jrec; |
| 1144 | struct mount *mp; |
| 1145 | void *save; |
| 1146 | int error; |
| 1147 | |
| 1148 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1149 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_SYMLINK); |
| 1150 | error = vop_journal_operate_ap(&ap->a_head); |
| 1151 | if (error == 0) { |
| 1152 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1153 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1154 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1155 | save = jrecord_push(jrec, JTYPE_REDO); |
| 1156 | jrecord_leaf(jrec, JLEAF_SYMLINKDATA, |
| 1157 | ap->a_target, strlen(ap->a_target)); |
| 1158 | jrecord_pop(jrec, save); |
| 1159 | if (*ap->a_vpp) |
| 1160 | jrecord_write_vnode_ref(jrec, *ap->a_vpp); |
| 1161 | } |
| 1162 | } |
| 1163 | jreclist_done(mp, &jreclist, error); |
| 1164 | return (error); |
| 1165 | } |
| 1166 | |
| 1167 | /* |
| 1168 | * Journal vop_nwhiteout { a_ncp, a_cred, a_flags } |
| 1169 | */ |
| 1170 | static |
| 1171 | int |
| 1172 | journal_nwhiteout(struct vop_nwhiteout_args *ap) |
| 1173 | { |
| 1174 | struct jrecord_list jreclist; |
| 1175 | struct jrecord jreccache; |
| 1176 | struct jrecord *jrec; |
| 1177 | struct mount *mp; |
| 1178 | int error; |
| 1179 | |
| 1180 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1181 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_WHITEOUT); |
| 1182 | error = vop_journal_operate_ap(&ap->a_head); |
| 1183 | if (error == 0) { |
| 1184 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1185 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1186 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1187 | } |
| 1188 | } |
| 1189 | jreclist_done(mp, &jreclist, error); |
| 1190 | return (error); |
| 1191 | } |
| 1192 | |
| 1193 | /* |
| 1194 | * Journal vop_nremove { a_ncp, a_cred } |
| 1195 | */ |
| 1196 | static |
| 1197 | int |
| 1198 | journal_nremove(struct vop_nremove_args *ap) |
| 1199 | { |
| 1200 | struct jrecord_list jreclist; |
| 1201 | struct jrecord jreccache; |
| 1202 | struct jrecord *jrec; |
| 1203 | struct mount *mp; |
| 1204 | int error; |
| 1205 | |
| 1206 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1207 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_REMOVE) && |
| 1208 | ap->a_ncp->nc_vp |
| 1209 | ) { |
| 1210 | jreclist_undo_file(&jreclist, ap->a_ncp->nc_vp, |
| 1211 | JRUNDO_ALL|JRUNDO_GETVP|JRUNDO_CONDLINK, 0, -1); |
| 1212 | } |
| 1213 | error = vop_journal_operate_ap(&ap->a_head); |
| 1214 | if (error == 0) { |
| 1215 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1216 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1217 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1218 | } |
| 1219 | } |
| 1220 | jreclist_done(mp, &jreclist, error); |
| 1221 | return (error); |
| 1222 | } |
| 1223 | |
| 1224 | /* |
| 1225 | * Journal vop_nmkdir { a_ncp, a_vpp, a_cred, a_vap } |
| 1226 | */ |
| 1227 | static |
| 1228 | int |
| 1229 | journal_nmkdir(struct vop_nmkdir_args *ap) |
| 1230 | { |
| 1231 | struct jrecord_list jreclist; |
| 1232 | struct jrecord jreccache; |
| 1233 | struct jrecord *jrec; |
| 1234 | struct mount *mp; |
| 1235 | int error; |
| 1236 | |
| 1237 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1238 | jreclist_init(mp, &jreclist, &jreccache, JTYPE_MKDIR); |
| 1239 | error = vop_journal_operate_ap(&ap->a_head); |
| 1240 | if (error == 0) { |
| 1241 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1242 | #if 0 |
| 1243 | if (jo->flags & MC_JOURNAL_WANT_AUDIT) { |
| 1244 | jrecord_write_audit(jrec); |
| 1245 | } |
| 1246 | #endif |
| 1247 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1248 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1249 | jrecord_write_vattr(jrec, ap->a_vap); |
| 1250 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1251 | if (*ap->a_vpp) |
| 1252 | jrecord_write_vnode_ref(jrec, *ap->a_vpp); |
| 1253 | } |
| 1254 | } |
| 1255 | jreclist_done(mp, &jreclist, error); |
| 1256 | return (error); |
| 1257 | } |
| 1258 | |
| 1259 | /* |
| 1260 | * Journal vop_nrmdir { a_ncp, a_cred } |
| 1261 | */ |
| 1262 | static |
| 1263 | int |
| 1264 | journal_nrmdir(struct vop_nrmdir_args *ap) |
| 1265 | { |
| 1266 | struct jrecord_list jreclist; |
| 1267 | struct jrecord jreccache; |
| 1268 | struct jrecord *jrec; |
| 1269 | struct mount *mp; |
| 1270 | int error; |
| 1271 | |
| 1272 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1273 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_RMDIR)) { |
| 1274 | jreclist_undo_file(&jreclist, ap->a_ncp->nc_vp, |
| 1275 | JRUNDO_VATTR|JRUNDO_GETVP, 0, 0); |
| 1276 | } |
| 1277 | error = vop_journal_operate_ap(&ap->a_head); |
| 1278 | if (error == 0) { |
| 1279 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1280 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1281 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_ncp); |
| 1282 | } |
| 1283 | } |
| 1284 | jreclist_done(mp, &jreclist, error); |
| 1285 | return (error); |
| 1286 | } |
| 1287 | |
| 1288 | /* |
| 1289 | * Journal vop_nrename { a_fncp, a_tncp, a_cred } |
| 1290 | */ |
| 1291 | static |
| 1292 | int |
| 1293 | journal_nrename(struct vop_nrename_args *ap) |
| 1294 | { |
| 1295 | struct jrecord_list jreclist; |
| 1296 | struct jrecord jreccache; |
| 1297 | struct jrecord *jrec; |
| 1298 | struct mount *mp; |
| 1299 | int error; |
| 1300 | |
| 1301 | mp = ap->a_head.a_ops->head.vv_mount; |
| 1302 | if (jreclist_init(mp, &jreclist, &jreccache, JTYPE_RENAME) && |
| 1303 | ap->a_tncp->nc_vp |
| 1304 | ) { |
| 1305 | jreclist_undo_file(&jreclist, ap->a_tncp->nc_vp, |
| 1306 | JRUNDO_ALL|JRUNDO_GETVP|JRUNDO_CONDLINK, 0, -1); |
| 1307 | } |
| 1308 | error = vop_journal_operate_ap(&ap->a_head); |
| 1309 | if (error == 0) { |
| 1310 | TAILQ_FOREACH(jrec, &jreclist.list, user_entry) { |
| 1311 | jrecord_write_cred(jrec, NULL, ap->a_cred); |
| 1312 | jrecord_write_path(jrec, JLEAF_PATH1, ap->a_fncp); |
| 1313 | jrecord_write_path(jrec, JLEAF_PATH2, ap->a_tncp); |
| 1314 | } |
| 1315 | } |
| 1316 | jreclist_done(mp, &jreclist, error); |
| 1317 | return (error); |
| 1318 | } |
| 1319 | |