1 /* 2 * ioctl.c - NILFS ioctl operations. 3 * 4 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation. 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License as published by 8 * the Free Software Foundation; either version 2 of the License, or 9 * (at your option) any later version. 10 * 11 * This program is distributed in the hope that it will be useful, 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 * GNU General Public License for more details. 15 * 16 * You should have received a copy of the GNU General Public License 17 * along with this program; if not, write to the Free Software 18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 19 * 20 * Written by Koji Sato <koji@osrg.net>. 21 */ 22 23 #include <linux/fs.h> 24 #include <linux/wait.h> 25 #include <linux/smp_lock.h> /* lock_kernel(), unlock_kernel() */ 26 #include <linux/capability.h> /* capable() */ 27 #include <linux/uaccess.h> /* copy_from_user(), copy_to_user() */ 28 #include <linux/vmalloc.h> 29 #include <linux/nilfs2_fs.h> 30 #include "nilfs.h" 31 #include "segment.h" 32 #include "bmap.h" 33 #include "cpfile.h" 34 #include "sufile.h" 35 #include "dat.h" 36 37 38 static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs, 39 struct nilfs_argv *argv, int dir, 40 ssize_t (*dofunc)(struct the_nilfs *, 41 __u64 *, int, 42 void *, size_t, size_t)) 43 { 44 void *buf; 45 void __user *base = (void __user *)(unsigned long)argv->v_base; 46 size_t maxmembs, total, n; 47 ssize_t nr; 48 int ret, i; 49 __u64 pos, ppos; 50 51 if (argv->v_nmembs == 0) 52 return 0; 53 54 if (argv->v_size > PAGE_SIZE) 55 return -EINVAL; 56 57 buf = (void *)__get_free_pages(GFP_NOFS, 0); 58 if (unlikely(!buf)) 59 return -ENOMEM; 60 maxmembs = PAGE_SIZE / argv->v_size; 61 62 ret = 0; 63 total = 0; 64 pos = argv->v_index; 65 for (i = 0; i < argv->v_nmembs; i += n) { 66 n = (argv->v_nmembs - i < maxmembs) ? 67 argv->v_nmembs - i : maxmembs; 68 if ((dir & _IOC_WRITE) && 69 copy_from_user(buf, base + argv->v_size * i, 70 argv->v_size * n)) { 71 ret = -EFAULT; 72 break; 73 } 74 ppos = pos; 75 nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size, 76 n); 77 if (nr < 0) { 78 ret = nr; 79 break; 80 } 81 if ((dir & _IOC_READ) && 82 copy_to_user(base + argv->v_size * i, buf, 83 argv->v_size * nr)) { 84 ret = -EFAULT; 85 break; 86 } 87 total += nr; 88 if ((size_t)nr < n) 89 break; 90 if (pos == ppos) 91 pos += n; 92 } 93 argv->v_nmembs = total; 94 95 free_pages((unsigned long)buf, 0); 96 return ret; 97 } 98 99 static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp, 100 unsigned int cmd, void __user *argp) 101 { 102 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 103 struct inode *cpfile = nilfs->ns_cpfile; 104 struct nilfs_transaction_info ti; 105 struct nilfs_cpmode cpmode; 106 int ret; 107 108 if (!capable(CAP_SYS_ADMIN)) 109 return -EPERM; 110 if (copy_from_user(&cpmode, argp, sizeof(cpmode))) 111 return -EFAULT; 112 113 mutex_lock(&nilfs->ns_mount_mutex); 114 nilfs_transaction_begin(inode->i_sb, &ti, 0); 115 ret = nilfs_cpfile_change_cpmode( 116 cpfile, cpmode.cm_cno, cpmode.cm_mode); 117 if (unlikely(ret < 0)) { 118 nilfs_transaction_abort(inode->i_sb); 119 mutex_unlock(&nilfs->ns_mount_mutex); 120 return ret; 121 } 122 nilfs_transaction_commit(inode->i_sb); /* never fails */ 123 mutex_unlock(&nilfs->ns_mount_mutex); 124 return ret; 125 } 126 127 static int 128 nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp, 129 unsigned int cmd, void __user *argp) 130 { 131 struct inode *cpfile = NILFS_SB(inode->i_sb)->s_nilfs->ns_cpfile; 132 struct nilfs_transaction_info ti; 133 __u64 cno; 134 int ret; 135 136 if (!capable(CAP_SYS_ADMIN)) 137 return -EPERM; 138 if (copy_from_user(&cno, argp, sizeof(cno))) 139 return -EFAULT; 140 141 nilfs_transaction_begin(inode->i_sb, &ti, 0); 142 ret = nilfs_cpfile_delete_checkpoint(cpfile, cno); 143 if (unlikely(ret < 0)) { 144 nilfs_transaction_abort(inode->i_sb); 145 return ret; 146 } 147 nilfs_transaction_commit(inode->i_sb); /* never fails */ 148 return ret; 149 } 150 151 static ssize_t 152 nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 153 void *buf, size_t size, size_t nmembs) 154 { 155 int ret; 156 157 down_read(&nilfs->ns_segctor_sem); 158 ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf, 159 size, nmembs); 160 up_read(&nilfs->ns_segctor_sem); 161 return ret; 162 } 163 164 static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp, 165 unsigned int cmd, void __user *argp) 166 { 167 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 168 struct nilfs_cpstat cpstat; 169 int ret; 170 171 down_read(&nilfs->ns_segctor_sem); 172 ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat); 173 up_read(&nilfs->ns_segctor_sem); 174 if (ret < 0) 175 return ret; 176 177 if (copy_to_user(argp, &cpstat, sizeof(cpstat))) 178 ret = -EFAULT; 179 return ret; 180 } 181 182 static ssize_t 183 nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 184 void *buf, size_t size, size_t nmembs) 185 { 186 int ret; 187 188 down_read(&nilfs->ns_segctor_sem); 189 ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size, 190 nmembs); 191 up_read(&nilfs->ns_segctor_sem); 192 return ret; 193 } 194 195 static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp, 196 unsigned int cmd, void __user *argp) 197 { 198 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 199 struct nilfs_sustat sustat; 200 int ret; 201 202 down_read(&nilfs->ns_segctor_sem); 203 ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat); 204 up_read(&nilfs->ns_segctor_sem); 205 if (ret < 0) 206 return ret; 207 208 if (copy_to_user(argp, &sustat, sizeof(sustat))) 209 ret = -EFAULT; 210 return ret; 211 } 212 213 static ssize_t 214 nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags, 215 void *buf, size_t size, size_t nmembs) 216 { 217 int ret; 218 219 down_read(&nilfs->ns_segctor_sem); 220 ret = nilfs_dat_get_vinfo(nilfs_dat_inode(nilfs), buf, size, nmembs); 221 up_read(&nilfs->ns_segctor_sem); 222 return ret; 223 } 224 225 static ssize_t 226 nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags, 227 void *buf, size_t size, size_t nmembs) 228 { 229 struct inode *dat = nilfs_dat_inode(nilfs); 230 struct nilfs_bmap *bmap = NILFS_I(dat)->i_bmap; 231 struct nilfs_bdesc *bdescs = buf; 232 int ret, i; 233 234 down_read(&nilfs->ns_segctor_sem); 235 for (i = 0; i < nmembs; i++) { 236 ret = nilfs_bmap_lookup_at_level(bmap, 237 bdescs[i].bd_offset, 238 bdescs[i].bd_level + 1, 239 &bdescs[i].bd_blocknr); 240 if (ret < 0) { 241 if (ret != -ENOENT) { 242 up_read(&nilfs->ns_segctor_sem); 243 return ret; 244 } 245 bdescs[i].bd_blocknr = 0; 246 } 247 } 248 up_read(&nilfs->ns_segctor_sem); 249 return nmembs; 250 } 251 252 static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp, 253 unsigned int cmd, void __user *argp) 254 { 255 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 256 struct nilfs_argv argv; 257 int ret; 258 259 if (copy_from_user(&argv, argp, sizeof(argv))) 260 return -EFAULT; 261 262 if (argv.v_size != sizeof(struct nilfs_bdesc)) 263 return -EINVAL; 264 265 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), 266 nilfs_ioctl_do_get_bdescs); 267 if (ret < 0) 268 return ret; 269 270 if (copy_to_user(argp, &argv, sizeof(argv))) 271 ret = -EFAULT; 272 return ret; 273 } 274 275 static int nilfs_ioctl_move_inode_block(struct inode *inode, 276 struct nilfs_vdesc *vdesc, 277 struct list_head *buffers) 278 { 279 struct buffer_head *bh; 280 int ret; 281 282 if (vdesc->vd_flags == 0) 283 ret = nilfs_gccache_submit_read_data( 284 inode, vdesc->vd_offset, vdesc->vd_blocknr, 285 vdesc->vd_vblocknr, &bh); 286 else 287 ret = nilfs_gccache_submit_read_node( 288 inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh); 289 290 if (unlikely(ret < 0)) { 291 if (ret == -ENOENT) 292 printk(KERN_CRIT 293 "%s: invalid virtual block address (%s): " 294 "ino=%llu, cno=%llu, offset=%llu, " 295 "blocknr=%llu, vblocknr=%llu\n", 296 __func__, vdesc->vd_flags ? "node" : "data", 297 (unsigned long long)vdesc->vd_ino, 298 (unsigned long long)vdesc->vd_cno, 299 (unsigned long long)vdesc->vd_offset, 300 (unsigned long long)vdesc->vd_blocknr, 301 (unsigned long long)vdesc->vd_vblocknr); 302 return ret; 303 } 304 if (unlikely(!list_empty(&bh->b_assoc_buffers))) { 305 printk(KERN_CRIT "%s: conflicting %s buffer: ino=%llu, " 306 "cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu\n", 307 __func__, vdesc->vd_flags ? "node" : "data", 308 (unsigned long long)vdesc->vd_ino, 309 (unsigned long long)vdesc->vd_cno, 310 (unsigned long long)vdesc->vd_offset, 311 (unsigned long long)vdesc->vd_blocknr, 312 (unsigned long long)vdesc->vd_vblocknr); 313 brelse(bh); 314 return -EEXIST; 315 } 316 list_add_tail(&bh->b_assoc_buffers, buffers); 317 return 0; 318 } 319 320 static int nilfs_ioctl_move_blocks(struct the_nilfs *nilfs, 321 struct nilfs_argv *argv, void *buf) 322 { 323 size_t nmembs = argv->v_nmembs; 324 struct inode *inode; 325 struct nilfs_vdesc *vdesc; 326 struct buffer_head *bh, *n; 327 LIST_HEAD(buffers); 328 ino_t ino; 329 __u64 cno; 330 int i, ret; 331 332 for (i = 0, vdesc = buf; i < nmembs; ) { 333 ino = vdesc->vd_ino; 334 cno = vdesc->vd_cno; 335 inode = nilfs_gc_iget(nilfs, ino, cno); 336 if (unlikely(inode == NULL)) { 337 ret = -ENOMEM; 338 goto failed; 339 } 340 do { 341 ret = nilfs_ioctl_move_inode_block(inode, vdesc, 342 &buffers); 343 if (unlikely(ret < 0)) 344 goto failed; 345 vdesc++; 346 } while (++i < nmembs && 347 vdesc->vd_ino == ino && vdesc->vd_cno == cno); 348 } 349 350 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) { 351 ret = nilfs_gccache_wait_and_mark_dirty(bh); 352 if (unlikely(ret < 0)) { 353 WARN_ON(ret == -EEXIST); 354 goto failed; 355 } 356 list_del_init(&bh->b_assoc_buffers); 357 brelse(bh); 358 } 359 return nmembs; 360 361 failed: 362 list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) { 363 list_del_init(&bh->b_assoc_buffers); 364 brelse(bh); 365 } 366 return ret; 367 } 368 369 static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs, 370 struct nilfs_argv *argv, void *buf) 371 { 372 size_t nmembs = argv->v_nmembs; 373 struct inode *cpfile = nilfs->ns_cpfile; 374 struct nilfs_period *periods = buf; 375 int ret, i; 376 377 for (i = 0; i < nmembs; i++) { 378 ret = nilfs_cpfile_delete_checkpoints( 379 cpfile, periods[i].p_start, periods[i].p_end); 380 if (ret < 0) 381 return ret; 382 } 383 return nmembs; 384 } 385 386 static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs, 387 struct nilfs_argv *argv, void *buf) 388 { 389 size_t nmembs = argv->v_nmembs; 390 int ret; 391 392 ret = nilfs_dat_freev(nilfs_dat_inode(nilfs), buf, nmembs); 393 394 return (ret < 0) ? ret : nmembs; 395 } 396 397 static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs, 398 struct nilfs_argv *argv, void *buf) 399 { 400 size_t nmembs = argv->v_nmembs; 401 struct inode *dat = nilfs_dat_inode(nilfs); 402 struct nilfs_bmap *bmap = NILFS_I(dat)->i_bmap; 403 struct nilfs_bdesc *bdescs = buf; 404 int ret, i; 405 406 for (i = 0; i < nmembs; i++) { 407 /* XXX: use macro or inline func to check liveness */ 408 ret = nilfs_bmap_lookup_at_level(bmap, 409 bdescs[i].bd_offset, 410 bdescs[i].bd_level + 1, 411 &bdescs[i].bd_blocknr); 412 if (ret < 0) { 413 if (ret != -ENOENT) 414 return ret; 415 bdescs[i].bd_blocknr = 0; 416 } 417 if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr) 418 /* skip dead block */ 419 continue; 420 if (bdescs[i].bd_level == 0) { 421 ret = nilfs_mdt_mark_block_dirty(dat, 422 bdescs[i].bd_offset); 423 if (ret < 0) { 424 WARN_ON(ret == -ENOENT); 425 return ret; 426 } 427 } else { 428 ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset, 429 bdescs[i].bd_level); 430 if (ret < 0) { 431 WARN_ON(ret == -ENOENT); 432 return ret; 433 } 434 } 435 } 436 return nmembs; 437 } 438 439 int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs, 440 struct nilfs_argv *argv, void **kbufs) 441 { 442 const char *msg; 443 int ret; 444 445 ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]); 446 if (ret < 0) { 447 /* 448 * can safely abort because checkpoints can be removed 449 * independently. 450 */ 451 msg = "cannot delete checkpoints"; 452 goto failed; 453 } 454 ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]); 455 if (ret < 0) { 456 /* 457 * can safely abort because DAT file is updated atomically 458 * using a copy-on-write technique. 459 */ 460 msg = "cannot delete virtual blocks from DAT file"; 461 goto failed; 462 } 463 ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]); 464 if (ret < 0) { 465 /* 466 * can safely abort because the operation is nondestructive. 467 */ 468 msg = "cannot mark copying blocks dirty"; 469 goto failed; 470 } 471 return 0; 472 473 failed: 474 printk(KERN_ERR "NILFS: GC failed during preparation: %s: err=%d\n", 475 msg, ret); 476 return ret; 477 } 478 479 static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp, 480 unsigned int cmd, void __user *argp) 481 { 482 struct nilfs_argv argv[5]; 483 static const size_t argsz[5] = { 484 sizeof(struct nilfs_vdesc), 485 sizeof(struct nilfs_period), 486 sizeof(__u64), 487 sizeof(struct nilfs_bdesc), 488 sizeof(__u64), 489 }; 490 void __user *base; 491 void *kbufs[5]; 492 struct the_nilfs *nilfs; 493 size_t len, nsegs; 494 int n, ret; 495 496 if (!capable(CAP_SYS_ADMIN)) 497 return -EPERM; 498 499 if (copy_from_user(argv, argp, sizeof(argv))) 500 return -EFAULT; 501 502 nsegs = argv[4].v_nmembs; 503 if (argv[4].v_size != argsz[4]) 504 return -EINVAL; 505 /* 506 * argv[4] points to segment numbers this ioctl cleans. We 507 * use kmalloc() for its buffer because memory used for the 508 * segment numbers is enough small. 509 */ 510 kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base, 511 nsegs * sizeof(__u64)); 512 if (IS_ERR(kbufs[4])) 513 return PTR_ERR(kbufs[4]); 514 515 nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 516 517 for (n = 0; n < 4; n++) { 518 ret = -EINVAL; 519 if (argv[n].v_size != argsz[n]) 520 goto out_free; 521 522 if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment) 523 goto out_free; 524 525 len = argv[n].v_size * argv[n].v_nmembs; 526 base = (void __user *)(unsigned long)argv[n].v_base; 527 if (len == 0) { 528 kbufs[n] = NULL; 529 continue; 530 } 531 532 kbufs[n] = vmalloc(len); 533 if (!kbufs[n]) { 534 ret = -ENOMEM; 535 goto out_free; 536 } 537 if (copy_from_user(kbufs[n], base, len)) { 538 ret = -EFAULT; 539 vfree(kbufs[n]); 540 goto out_free; 541 } 542 } 543 544 /* 545 * nilfs_ioctl_move_blocks() will call nilfs_gc_iget(), 546 * which will operates an inode list without blocking. 547 * To protect the list from concurrent operations, 548 * nilfs_ioctl_move_blocks should be atomic operation. 549 */ 550 if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) { 551 ret = -EBUSY; 552 goto out_free; 553 } 554 555 ret = nilfs_ioctl_move_blocks(nilfs, &argv[0], kbufs[0]); 556 if (ret < 0) 557 printk(KERN_ERR "NILFS: GC failed during preparation: " 558 "cannot read source blocks: err=%d\n", ret); 559 else 560 ret = nilfs_clean_segments(inode->i_sb, argv, kbufs); 561 562 if (ret < 0) 563 nilfs_remove_all_gcinode(nilfs); 564 clear_nilfs_gc_running(nilfs); 565 566 out_free: 567 while (--n >= 0) 568 vfree(kbufs[n]); 569 kfree(kbufs[4]); 570 return ret; 571 } 572 573 static int nilfs_ioctl_sync(struct inode *inode, struct file *filp, 574 unsigned int cmd, void __user *argp) 575 { 576 __u64 cno; 577 int ret; 578 579 ret = nilfs_construct_segment(inode->i_sb); 580 if (ret < 0) 581 return ret; 582 583 if (argp != NULL) { 584 cno = NILFS_SB(inode->i_sb)->s_nilfs->ns_cno - 1; 585 if (copy_to_user(argp, &cno, sizeof(cno))) 586 return -EFAULT; 587 } 588 return 0; 589 } 590 591 static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp, 592 unsigned int cmd, void __user *argp, 593 size_t membsz, 594 ssize_t (*dofunc)(struct the_nilfs *, 595 __u64 *, int, 596 void *, size_t, size_t)) 597 598 { 599 struct the_nilfs *nilfs = NILFS_SB(inode->i_sb)->s_nilfs; 600 struct nilfs_argv argv; 601 int ret; 602 603 if (copy_from_user(&argv, argp, sizeof(argv))) 604 return -EFAULT; 605 606 if (argv.v_size < membsz) 607 return -EINVAL; 608 609 ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc); 610 if (ret < 0) 611 return ret; 612 613 if (copy_to_user(argp, &argv, sizeof(argv))) 614 ret = -EFAULT; 615 return ret; 616 } 617 618 long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) 619 { 620 struct inode *inode = filp->f_dentry->d_inode; 621 void __user *argp = (void * __user *)arg; 622 623 switch (cmd) { 624 case NILFS_IOCTL_CHANGE_CPMODE: 625 return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp); 626 case NILFS_IOCTL_DELETE_CHECKPOINT: 627 return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp); 628 case NILFS_IOCTL_GET_CPINFO: 629 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 630 sizeof(struct nilfs_cpinfo), 631 nilfs_ioctl_do_get_cpinfo); 632 case NILFS_IOCTL_GET_CPSTAT: 633 return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp); 634 case NILFS_IOCTL_GET_SUINFO: 635 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 636 sizeof(struct nilfs_suinfo), 637 nilfs_ioctl_do_get_suinfo); 638 case NILFS_IOCTL_GET_SUSTAT: 639 return nilfs_ioctl_get_sustat(inode, filp, cmd, argp); 640 case NILFS_IOCTL_GET_VINFO: 641 return nilfs_ioctl_get_info(inode, filp, cmd, argp, 642 sizeof(struct nilfs_vinfo), 643 nilfs_ioctl_do_get_vinfo); 644 case NILFS_IOCTL_GET_BDESCS: 645 return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp); 646 case NILFS_IOCTL_CLEAN_SEGMENTS: 647 return nilfs_ioctl_clean_segments(inode, filp, cmd, argp); 648 case NILFS_IOCTL_SYNC: 649 return nilfs_ioctl_sync(inode, filp, cmd, argp); 650 default: 651 return -ENOTTY; 652 } 653 } 654