e4defrag.c revision 18a1444b4f1e6a0948fd38fa0de382d86cfe04de
1/* 2 * e4defrag.c - ext4 filesystem defragmenter 3 * 4 * Copyright (C) 2009 NEC Software Tohoku, Ltd. 5 * 6 * Author: Akira Fujita <a-fujita@rs.jp.nec.com> 7 * Takashi Sato <t-sato@yk.jp.nec.com> 8 */ 9 10#ifndef _LARGEFILE_SOURCE 11#define _LARGEFILE_SOURCE 12#endif 13 14#ifndef _LARGEFILE64_SOURCE 15#define _LARGEFILE64_SOURCE 16#endif 17 18#ifndef _GNU_SOURCE 19#define _GNU_SOURCE 20#endif 21 22#include <ctype.h> 23#include <dirent.h> 24#include <endian.h> 25#include <errno.h> 26#include <fcntl.h> 27#include <ftw.h> 28#include <limits.h> 29#include <mntent.h> 30#include <stdio.h> 31#include <stdlib.h> 32#include <string.h> 33#include <unistd.h> 34#include <ext2fs/ext2_types.h> 35#include <ext2fs/ext2fs.h> 36#include <linux/fs.h> 37#include <sys/ioctl.h> 38#include <ext2fs/fiemap.h> 39#include <sys/mman.h> 40#include <sys/stat.h> 41#include <sys/statfs.h> 42#include <sys/syscall.h> 43#include <sys/vfs.h> 44 45/* A relatively new ioctl interface ... */ 46#ifndef EXT4_IOC_MOVE_EXT 47#define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent) 48#endif 49 50/* Macro functions */ 51#define PRINT_ERR_MSG(msg) fprintf(stderr, "%s\n", (msg)) 52#define IN_FTW_PRINT_ERR_MSG(msg) \ 53 fprintf(stderr, "\t%s\t\t[ NG ]\n", (msg)) 54#define PRINT_FILE_NAME(file) fprintf(stderr, " \"%s\"\n", (file)) 55#define PRINT_ERR_MSG_WITH_ERRNO(msg) \ 56 fprintf(stderr, "\t%s:%s\t[ NG ]\n", (msg), strerror(errno)) 57#define STATISTIC_ERR_MSG(msg) \ 58 fprintf(stderr, "\t%s\n", (msg)) 59#define STATISTIC_ERR_MSG_WITH_ERRNO(msg) \ 60 fprintf(stderr, "\t%s:%s\n", (msg), strerror(errno)) 61#define min(x, y) (((x) > (y)) ? (y) : (x)) 62#define CALC_SCORE(ratio) \ 63 ((ratio) > 10 ? (80 + 20 * (ratio) / 100) : (8 * (ratio))) 64/* Wrap up the free function */ 65#define FREE(tmp) \ 66 do { \ 67 if ((tmp) != NULL) \ 68 free(tmp); \ 69 } while (0) \ 70/* Insert list2 after list1 */ 71#define insert(list1, list2) \ 72 do { \ 73 list2->next = list1->next; \ 74 list1->next->prev = list2; \ 75 list2->prev = list1; \ 76 list1->next = list2; \ 77 } while (0) 78 79/* To delete unused warning */ 80#ifdef __GNUC__ 81#define EXT2FS_ATTR(x) __attribute__(x) 82#else 83#define EXT2FS_ATTR(x) 84#endif 85 86/* The mode of defrag */ 87#define DETAIL 0x01 88#define STATISTIC 0x02 89 90#define DEVNAME 0 91#define DIRNAME 1 92#define FILENAME 2 93 94#define FTW_OPEN_FD 2000 95 96#define FS_EXT4 "ext4" 97#define ROOT_UID 0 98 99#define BOUND_SCORE 55 100#define SHOW_FRAG_FILES 5 101 102/* Magic number for ext4 */ 103#define EXT4_SUPER_MAGIC 0xEF53 104 105/* Definition of flex_bg */ 106#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200 107 108/* The following macro is used for ioctl FS_IOC_FIEMAP 109 * EXTENT_MAX_COUNT: the maximum number of extents for exchanging between 110 * kernel-space and user-space per ioctl 111 */ 112#define EXTENT_MAX_COUNT 512 113 114/* The following macros are error message */ 115#define MSG_USAGE \ 116"Usage : e4defrag [-v] file...| directory...| device...\n\ 117 : e4defrag -c file...| directory...| device...\n" 118 119#define NGMSG_EXT4 "Filesystem is not ext4 filesystem" 120#define NGMSG_FILE_EXTENT "Failed to get file extents" 121#define NGMSG_FILE_INFO "Failed to get file information" 122#define NGMSG_FILE_OPEN "Failed to open" 123#define NGMSG_FILE_UNREG "File is not regular file" 124#define NGMSG_LOST_FOUND "Can not process \"lost+found\"" 125 126/* Data type for filesystem-wide blocks number */ 127typedef unsigned long long ext4_fsblk_t; 128 129struct fiemap_extent_data { 130 __u64 len; /* blocks count */ 131 __u64 logical; /* start logical block number */ 132 ext4_fsblk_t physical; /* start physical block number */ 133}; 134 135struct fiemap_extent_list { 136 struct fiemap_extent_list *prev; 137 struct fiemap_extent_list *next; 138 struct fiemap_extent_data data; /* extent belong to file */ 139}; 140 141struct fiemap_extent_group { 142 struct fiemap_extent_group *prev; 143 struct fiemap_extent_group *next; 144 __u64 len; /* length of this continuous region */ 145 struct fiemap_extent_list *start; /* start ext */ 146 struct fiemap_extent_list *end; /* end ext */ 147}; 148 149struct move_extent { 150 __s32 reserved; /* original file descriptor */ 151 __u32 donor_fd; /* donor file descriptor */ 152 __u64 orig_start; /* logical start offset in block for orig */ 153 __u64 donor_start; /* logical start offset in block for donor */ 154 __u64 len; /* block length to be moved */ 155 __u64 moved_len; /* moved block length */ 156}; 157 158struct frag_statistic_ino { 159 int now_count; /* the file's extents count of before defrag */ 160 int best_count; /* the best file's extents count */ 161 __u64 size_per_ext; /* size(KB) per extent */ 162 float ratio; /* the ratio of fragmentation */ 163 char msg_buffer[PATH_MAX + 1]; /* pathname of the file */ 164}; 165 166static char lost_found_dir[PATH_MAX + 1]; 167static int block_size; 168static int extents_before_defrag; 169static int extents_after_defrag; 170static int mode_flag; 171static unsigned int current_uid; 172static unsigned int defraged_file_count; 173static unsigned int frag_files_before_defrag; 174static unsigned int frag_files_after_defrag; 175static unsigned int regular_count; 176static unsigned int succeed_cnt; 177static unsigned int total_count; 178static __u8 log_groups_per_flex; 179static __u32 blocks_per_group; 180static __u32 feature_incompat; 181static ext4_fsblk_t files_block_count; 182static struct frag_statistic_ino frag_rank[SHOW_FRAG_FILES]; 183 184 185/* Local definitions of some syscalls glibc may not yet have */ 186 187#ifndef HAVE_POSIX_FADVISE 188#warning Using locally defined posix_fadvise interface. 189 190#ifndef __NR_fadvise64_64 191#error Your kernel headers dont define __NR_fadvise64_64 192#endif 193 194/* 195 * fadvise() - Give advice about file access. 196 * 197 * @fd: defrag target file's descriptor. 198 * @offset: file offset. 199 * @len: area length. 200 * @advise: process flag. 201 */ 202static int posix_fadvise(int fd, loff_t offset, size_t len, int advise) 203{ 204 return syscall(__NR_fadvise64_64, fd, offset, len, advise); 205} 206#endif /* ! HAVE_FADVISE64_64 */ 207 208#ifndef HAVE_SYNC_FILE_RANGE 209#warning Using locally defined sync_file_range interface. 210 211#ifndef __NR_sync_file_range 212#ifndef __NR_sync_file_range2 /* ppc */ 213#error Your kernel headers dont define __NR_sync_file_range 214#endif 215#endif 216 217/* 218 * sync_file_range() - Sync file region. 219 * 220 * @fd: defrag target file's descriptor. 221 * @offset: file offset. 222 * @length: area length. 223 * @flag: process flag. 224 */ 225int sync_file_range(int fd, loff_t offset, loff_t length, unsigned int flag) 226{ 227#ifdef __NR_sync_file_range 228 return syscall(__NR_sync_file_range, fd, offset, length, flag); 229#else 230 return syscall(__NR_sync_file_range2, fd, flag, offset, length); 231#endif 232} 233#endif /* ! HAVE_SYNC_FILE_RANGE */ 234 235#ifndef HAVE_FALLOCATE64 236#warning Using locally defined fallocate syscall interface. 237 238#ifndef __NR_fallocate 239#error Your kernel headers dont define __NR_fallocate 240#endif 241 242/* 243 * fallocate64() - Manipulate file space. 244 * 245 * @fd: defrag target file's descriptor. 246 * @mode: process flag. 247 * @offset: file offset. 248 * @len: file size. 249 */ 250static int fallocate64(int fd, int mode, loff_t offset, loff_t len) 251{ 252 return syscall(__NR_fallocate, fd, mode, offset, len); 253} 254#endif /* ! HAVE_FALLOCATE */ 255 256/* 257 * get_mount_point() - Get device's mount point. 258 * 259 * @devname: the device's name. 260 * @mount_point: the mount point. 261 * @dir_path_len: the length of directory. 262 */ 263static int get_mount_point(const char *devname, char *mount_point, 264 int dir_path_len) 265{ 266 /* Refer to /etc/mtab */ 267 const char *mtab = MOUNTED; 268 FILE *fp = NULL; 269 struct mntent *mnt = NULL; 270 struct stat64 sb; 271 272 if (stat64(devname, &sb) < 0) { 273 perror(NGMSG_FILE_INFO); 274 PRINT_FILE_NAME(devname); 275 return -1; 276 } 277 278 fp = setmntent(mtab, "r"); 279 if (fp == NULL) { 280 perror("Couldn't access /etc/mtab"); 281 return -1; 282 } 283 284 while ((mnt = getmntent(fp)) != NULL) { 285 struct stat64 ms; 286 287 /* 288 * To handle device symlinks, we see if the 289 * device number matches, not the name 290 */ 291 if (stat64(mnt->mnt_fsname, &ms) < 0) 292 continue; 293 if (sb.st_rdev != ms.st_rdev) 294 continue; 295 296 endmntent(fp); 297 if (strcmp(mnt->mnt_type, FS_EXT4) == 0) { 298 strncpy(mount_point, mnt->mnt_dir, 299 dir_path_len); 300 return 0; 301 } 302 PRINT_ERR_MSG(NGMSG_EXT4); 303 return -1; 304 } 305 endmntent(fp); 306 PRINT_ERR_MSG("Filesystem is not mounted"); 307 return -1; 308} 309 310/* 311 * is_ext4() - Whether on an ext4 filesystem. 312 * 313 * @file: the file's name. 314 */ 315static int is_ext4(const char *file, char *devname) 316{ 317 int maxlen = 0; 318 int len, ret; 319 FILE *fp = NULL; 320 char *mnt_type = NULL; 321 /* Refer to /etc/mtab */ 322 const char *mtab = MOUNTED; 323 char file_path[PATH_MAX + 1]; 324 struct mntent *mnt = NULL; 325 struct statfs64 fsbuf; 326 327 /* Get full path */ 328 if (realpath(file, file_path) == NULL) { 329 perror("Couldn't get full path"); 330 PRINT_FILE_NAME(file); 331 return -1; 332 } 333 334 if (statfs64(file_path, &fsbuf) < 0) { 335 perror("Failed to get filesystem information"); 336 PRINT_FILE_NAME(file); 337 return -1; 338 } 339 340 if (fsbuf.f_type != EXT4_SUPER_MAGIC) { 341 PRINT_ERR_MSG(NGMSG_EXT4); 342 return -1; 343 } 344 345 fp = setmntent(mtab, "r"); 346 if (fp == NULL) { 347 perror("Couldn't access /etc/mtab"); 348 return -1; 349 } 350 351 while ((mnt = getmntent(fp)) != NULL) { 352 if (mnt->mnt_fsname[0] != '/') 353 continue; 354 len = strlen(mnt->mnt_dir); 355 ret = memcmp(file_path, mnt->mnt_dir, len); 356 if (ret != 0) 357 continue; 358 359 if (maxlen >= len) 360 continue; 361 362 maxlen = len; 363 364 mnt_type = realloc(mnt_type, strlen(mnt->mnt_type) + 1); 365 if (mnt_type == NULL) { 366 endmntent(fp); 367 return -1; 368 } 369 memset(mnt_type, 0, strlen(mnt->mnt_type) + 1); 370 strncpy(mnt_type, mnt->mnt_type, strlen(mnt->mnt_type)); 371 strncpy(lost_found_dir, mnt->mnt_dir, PATH_MAX); 372 strncpy(devname, mnt->mnt_fsname, strlen(mnt->mnt_fsname) + 1); 373 } 374 375 endmntent(fp); 376 if (mnt_type && strcmp(mnt_type, FS_EXT4) == 0) { 377 FREE(mnt_type); 378 return 0; 379 } else { 380 FREE(mnt_type); 381 PRINT_ERR_MSG(NGMSG_EXT4); 382 return -1; 383 } 384} 385 386/* 387 * calc_entry_counts() - Calculate file counts. 388 * 389 * @file: file name. 390 * @buf: file info. 391 * @flag: file type. 392 * @ftwbuf: the pointer of a struct FTW. 393 */ 394static int calc_entry_counts(const char *file EXT2FS_ATTR((unused)), 395 const struct stat64 *buf, int flag EXT2FS_ATTR((unused)), 396 struct FTW *ftwbuf EXT2FS_ATTR((unused))) 397{ 398 if (S_ISREG(buf->st_mode)) 399 regular_count++; 400 401 total_count++; 402 403 return 0; 404} 405 406/* 407 * page_in_core() - Get information on whether pages are in core. 408 * 409 * @fd: defrag target file's descriptor. 410 * @defrag_data: data used for defrag. 411 * @vec: page state array. 412 * @page_num: page number. 413 */ 414static int page_in_core(int fd, struct move_extent defrag_data, 415 unsigned char **vec, unsigned int *page_num) 416{ 417 long pagesize; 418 void *page = NULL; 419 loff_t offset, end_offset, length; 420 421 if (vec == NULL || *vec != NULL) 422 return -1; 423 424 pagesize = sysconf(_SC_PAGESIZE); 425 if (pagesize < 0) 426 return -1; 427 /* In mmap, offset should be a multiple of the page size */ 428 offset = (loff_t)defrag_data.orig_start * block_size; 429 length = (loff_t)defrag_data.len * block_size; 430 end_offset = offset + length; 431 /* Round the offset down to the nearest multiple of pagesize */ 432 offset = (offset / pagesize) * pagesize; 433 length = end_offset - offset; 434 435 page = mmap(NULL, length, PROT_READ, MAP_SHARED, fd, offset); 436 if (page == MAP_FAILED) 437 return -1; 438 439 *page_num = 0; 440 *page_num = (length + pagesize - 1) / pagesize; 441 *vec = (unsigned char *)calloc(*page_num, 1); 442 if (*vec == NULL) 443 return -1; 444 445 /* Get information on whether pages are in core */ 446 if (mincore(page, (size_t)length, *vec) == -1 || 447 munmap(page, length) == -1) { 448 FREE(*vec); 449 return -1; 450 } 451 452 return 0; 453} 454 455/* 456 * defrag_fadvise() - Predeclare an access pattern for file data. 457 * 458 * @fd: defrag target file's descriptor. 459 * @defrag_data: data used for defrag. 460 * @vec: page state array. 461 * @page_num: page number. 462 */ 463static int defrag_fadvise(int fd, struct move_extent defrag_data, 464 unsigned char *vec, unsigned int page_num) 465{ 466 int flag = 1; 467 long pagesize = sysconf(_SC_PAGESIZE); 468 int fadvise_flag = POSIX_FADV_DONTNEED; 469 int sync_flag = SYNC_FILE_RANGE_WAIT_BEFORE | 470 SYNC_FILE_RANGE_WRITE | 471 SYNC_FILE_RANGE_WAIT_AFTER; 472 unsigned int i; 473 loff_t offset; 474 475 if (pagesize < 1) 476 return -1; 477 478 offset = (loff_t)defrag_data.orig_start * block_size; 479 offset = (offset / pagesize) * pagesize; 480 481 /* Sync file for fadvise process */ 482 if (sync_file_range(fd, offset, 483 (loff_t)pagesize * page_num, sync_flag) < 0) 484 return -1; 485 486 /* Try to release buffer cache which this process used, 487 * then other process can use the released buffer 488 */ 489 for (i = 0; i < page_num; i++) { 490 if ((vec[i] & 0x1) == 0) { 491 offset += pagesize; 492 continue; 493 } 494 if (posix_fadvise(fd, offset, pagesize, fadvise_flag) < 0) { 495 if ((mode_flag & DETAIL) && flag) { 496 perror("\tFailed to fadvise"); 497 flag = 0; 498 } 499 } 500 offset += pagesize; 501 } 502 503 return 0; 504} 505 506/* 507 * check_free_size() - Check if there's enough disk space. 508 * 509 * @fd: defrag target file's descriptor. 510 * @file: file name. 511 * @blk_count: file blocks. 512 */ 513static int check_free_size(int fd, const char *file, ext4_fsblk_t blk_count) 514{ 515 ext4_fsblk_t free_blk_count; 516 struct statfs64 fsbuf; 517 518 if (fstatfs64(fd, &fsbuf) < 0) { 519 if (mode_flag & DETAIL) { 520 PRINT_FILE_NAME(file); 521 PRINT_ERR_MSG_WITH_ERRNO( 522 "Failed to get filesystem information"); 523 } 524 return -1; 525 } 526 527 /* Compute free space for root and normal user separately */ 528 if (current_uid == ROOT_UID) 529 free_blk_count = fsbuf.f_bfree; 530 else 531 free_blk_count = fsbuf.f_bavail; 532 533 if (free_blk_count >= blk_count) 534 return 0; 535 536 return -ENOSPC; 537} 538 539/* 540 * file_frag_count() - Get file fragment count. 541 * 542 * @fd: defrag target file's descriptor. 543 */ 544static int file_frag_count(int fd) 545{ 546 int ret; 547 struct fiemap fiemap_buf; 548 549 /* When fm_extent_count is 0, 550 * ioctl just get file fragment count. 551 */ 552 memset(&fiemap_buf, 0, sizeof(struct fiemap)); 553 fiemap_buf.fm_start = 0; 554 fiemap_buf.fm_length = FIEMAP_MAX_OFFSET; 555 fiemap_buf.fm_flags |= FIEMAP_FLAG_SYNC; 556 557 ret = ioctl(fd, FS_IOC_FIEMAP, &fiemap_buf); 558 if (ret < 0) 559 return ret; 560 561 return fiemap_buf.fm_mapped_extents; 562} 563 564/* 565 * file_check() - Check file's attributes. 566 * 567 * @fd: defrag target file's descriptor. 568 * @buf: a pointer of the struct stat64. 569 * @file: file name. 570 * @extents: file extents. 571 * @blk_count: file blocks. 572 */ 573static int file_check(int fd, const struct stat64 *buf, const char *file, 574 int extents, ext4_fsblk_t blk_count) 575{ 576 int ret; 577 struct flock lock; 578 579 /* Write-lock check is more reliable */ 580 lock.l_type = F_WRLCK; 581 lock.l_start = 0; 582 lock.l_whence = SEEK_SET; 583 lock.l_len = 0; 584 585 /* Free space */ 586 ret = check_free_size(fd, file, blk_count); 587 if (ret < 0) { 588 if ((mode_flag & DETAIL) && ret == -ENOSPC) { 589 printf("\033[79;0H\033[K[%u/%u] \"%s\"\t\t" 590 " extents: %d -> %d\n", defraged_file_count, 591 total_count, file, extents, extents); 592 IN_FTW_PRINT_ERR_MSG( 593 "Defrag size is larger than filesystem's free space"); 594 } 595 return -1; 596 } 597 598 /* Access authority */ 599 if (current_uid != ROOT_UID && 600 buf->st_uid != current_uid) { 601 if (mode_flag & DETAIL) { 602 printf("\033[79;0H\033[K[%u/%u] \"%s\"\t\t" 603 " extents: %d -> %d\n", defraged_file_count, 604 total_count, file, extents, extents); 605 IN_FTW_PRINT_ERR_MSG( 606 "File is not current user's file" 607 " or current user is not root"); 608 } 609 return -1; 610 } 611 612 /* Lock status */ 613 if (fcntl(fd, F_GETLK, &lock) < 0) { 614 if (mode_flag & DETAIL) { 615 PRINT_FILE_NAME(file); 616 PRINT_ERR_MSG_WITH_ERRNO( 617 "Failed to get lock information"); 618 } 619 return -1; 620 } else if (lock.l_type != F_UNLCK) { 621 if (mode_flag & DETAIL) { 622 PRINT_FILE_NAME(file); 623 IN_FTW_PRINT_ERR_MSG("File has been locked"); 624 } 625 return -1; 626 } 627 628 return 0; 629} 630 631/* 632 * insert_extent_by_logical() - Sequentially insert extent by logical. 633 * 634 * @ext_list_head: the head of logical extent list. 635 * @ext: the extent element which will be inserted. 636 */ 637static int insert_extent_by_logical(struct fiemap_extent_list **ext_list_head, 638 struct fiemap_extent_list *ext) 639{ 640 struct fiemap_extent_list *ext_list_tmp = *ext_list_head; 641 642 if (ext == NULL) 643 goto out; 644 645 /* First element */ 646 if (*ext_list_head == NULL) { 647 (*ext_list_head) = ext; 648 (*ext_list_head)->prev = *ext_list_head; 649 (*ext_list_head)->next = *ext_list_head; 650 return 0; 651 } 652 653 if (ext->data.logical <= ext_list_tmp->data.logical) { 654 /* Insert before head */ 655 if (ext_list_tmp->data.logical < 656 ext->data.logical + ext->data.len) 657 /* Overlap */ 658 goto out; 659 /* Adjust head */ 660 *ext_list_head = ext; 661 } else { 662 /* Insert into the middle or last of the list */ 663 do { 664 if (ext->data.logical < ext_list_tmp->data.logical) 665 break; 666 ext_list_tmp = ext_list_tmp->next; 667 } while (ext_list_tmp != (*ext_list_head)); 668 if (ext->data.logical < 669 ext_list_tmp->prev->data.logical + 670 ext_list_tmp->prev->data.len) 671 /* Overlap */ 672 goto out; 673 674 if (ext_list_tmp != *ext_list_head && 675 ext_list_tmp->data.logical < 676 ext->data.logical + ext->data.len) 677 /* Overlap */ 678 goto out; 679 } 680 ext_list_tmp = ext_list_tmp->prev; 681 /* Insert "ext" after "ext_list_tmp" */ 682 insert(ext_list_tmp, ext); 683 return 0; 684out: 685 errno = EINVAL; 686 return -1; 687} 688 689/* 690 * insert_extent_by_physical() - Sequentially insert extent by physical. 691 * 692 * @ext_list_head: the head of physical extent list. 693 * @ext: the extent element which will be inserted. 694 */ 695static int insert_extent_by_physical(struct fiemap_extent_list **ext_list_head, 696 struct fiemap_extent_list *ext) 697{ 698 struct fiemap_extent_list *ext_list_tmp = *ext_list_head; 699 700 if (ext == NULL) 701 goto out; 702 703 /* First element */ 704 if (*ext_list_head == NULL) { 705 (*ext_list_head) = ext; 706 (*ext_list_head)->prev = *ext_list_head; 707 (*ext_list_head)->next = *ext_list_head; 708 return 0; 709 } 710 711 if (ext->data.physical <= ext_list_tmp->data.physical) { 712 /* Insert before head */ 713 if (ext_list_tmp->data.physical < 714 ext->data.physical + ext->data.len) 715 /* Overlap */ 716 goto out; 717 /* Adjust head */ 718 *ext_list_head = ext; 719 } else { 720 /* Insert into the middle or last of the list */ 721 do { 722 if (ext->data.physical < ext_list_tmp->data.physical) 723 break; 724 ext_list_tmp = ext_list_tmp->next; 725 } while (ext_list_tmp != (*ext_list_head)); 726 if (ext->data.physical < 727 ext_list_tmp->prev->data.physical + 728 ext_list_tmp->prev->data.len) 729 /* Overlap */ 730 goto out; 731 732 if (ext_list_tmp != *ext_list_head && 733 ext_list_tmp->data.physical < 734 ext->data.physical + ext->data.len) 735 /* Overlap */ 736 goto out; 737 } 738 ext_list_tmp = ext_list_tmp->prev; 739 /* Insert "ext" after "ext_list_tmp" */ 740 insert(ext_list_tmp, ext); 741 return 0; 742out: 743 errno = EINVAL; 744 return -1; 745} 746 747/* 748 * insert_exts_group() - Insert a exts_group. 749 * 750 * @ext_group_head: the head of a exts_group list. 751 * @exts_group: the exts_group element which will be inserted. 752 */ 753static int insert_exts_group(struct fiemap_extent_group **ext_group_head, 754 struct fiemap_extent_group *exts_group) 755{ 756 struct fiemap_extent_group *ext_group_tmp = NULL; 757 758 if (exts_group == NULL) { 759 errno = EINVAL; 760 return -1; 761 } 762 763 /* Initialize list */ 764 if (*ext_group_head == NULL) { 765 (*ext_group_head) = exts_group; 766 (*ext_group_head)->prev = *ext_group_head; 767 (*ext_group_head)->next = *ext_group_head; 768 return 0; 769 } 770 771 ext_group_tmp = (*ext_group_head)->prev; 772 insert(ext_group_tmp, exts_group); 773 774 return 0; 775} 776 777/* 778 * join_extents() - Find continuous region(exts_group). 779 * 780 * @ext_list_head: the head of the extent list. 781 * @ext_group_head: the head of the target exts_group list. 782 */ 783static int join_extents(struct fiemap_extent_list *ext_list_head, 784 struct fiemap_extent_group **ext_group_head) 785{ 786 __u64 len = ext_list_head->data.len; 787 struct fiemap_extent_list *ext_list_start = ext_list_head; 788 struct fiemap_extent_list *ext_list_tmp = ext_list_head->next; 789 790 do { 791 struct fiemap_extent_group *ext_group_tmp = NULL; 792 793 /* This extent and previous extent are not continuous, 794 * so, all previous extents are treated as an extent group. 795 */ 796 if ((ext_list_tmp->prev->data.logical + 797 ext_list_tmp->prev->data.len) 798 != ext_list_tmp->data.logical) { 799 ext_group_tmp = 800 malloc(sizeof(struct fiemap_extent_group)); 801 if (ext_group_tmp == NULL) 802 return -1; 803 804 memset(ext_group_tmp, 0, 805 sizeof(struct fiemap_extent_group)); 806 ext_group_tmp->len = len; 807 ext_group_tmp->start = ext_list_start; 808 ext_group_tmp->end = ext_list_tmp->prev; 809 810 if (insert_exts_group(ext_group_head, 811 ext_group_tmp) < 0) { 812 FREE(ext_group_tmp); 813 return -1; 814 } 815 ext_list_start = ext_list_tmp; 816 len = ext_list_tmp->data.len; 817 ext_list_tmp = ext_list_tmp->next; 818 continue; 819 } 820 821 /* This extent and previous extent are continuous, 822 * so, they belong to the same extent group, and we check 823 * if the next extent belongs to the same extent group. 824 */ 825 len += ext_list_tmp->data.len; 826 ext_list_tmp = ext_list_tmp->next; 827 } while (ext_list_tmp != ext_list_head->next); 828 829 return 0; 830} 831 832/* 833 * get_file_extents() - Get file's extent list. 834 * 835 * @fd: defrag target file's descriptor. 836 * @ext_list_head: the head of the extent list. 837 */ 838static int get_file_extents(int fd, struct fiemap_extent_list **ext_list_head) 839{ 840 __u32 i; 841 int ret; 842 int ext_buf_size, fie_buf_size; 843 __u64 pos = 0; 844 struct fiemap *fiemap_buf = NULL; 845 struct fiemap_extent *ext_buf = NULL; 846 struct fiemap_extent_list *ext_list = NULL; 847 848 /* Convert units, in bytes. 849 * Be careful : now, physical block number in extent is 48bit, 850 * and the maximum blocksize for ext4 is 4K(12bit), 851 * so there is no overflow, but in future it may be changed. 852 */ 853 854 /* Alloc space for fiemap */ 855 ext_buf_size = EXTENT_MAX_COUNT * sizeof(struct fiemap_extent); 856 fie_buf_size = sizeof(struct fiemap) + ext_buf_size; 857 858 fiemap_buf = malloc(fie_buf_size); 859 if (fiemap_buf == NULL) 860 return -1; 861 862 ext_buf = fiemap_buf->fm_extents; 863 memset(fiemap_buf, 0, fie_buf_size); 864 fiemap_buf->fm_length = FIEMAP_MAX_OFFSET; 865 fiemap_buf->fm_flags |= FIEMAP_FLAG_SYNC; 866 fiemap_buf->fm_extent_count = EXTENT_MAX_COUNT; 867 868 do { 869 fiemap_buf->fm_start = pos; 870 memset(ext_buf, 0, ext_buf_size); 871 ret = ioctl(fd, FS_IOC_FIEMAP, fiemap_buf); 872 if (ret < 0 || fiemap_buf->fm_mapped_extents == 0) 873 goto out; 874 for (i = 0; i < fiemap_buf->fm_mapped_extents; i++) { 875 ext_list = NULL; 876 ext_list = malloc(sizeof(struct fiemap_extent_list)); 877 if (ext_list == NULL) 878 goto out; 879 880 ext_list->data.physical = ext_buf[i].fe_physical 881 / block_size; 882 ext_list->data.logical = ext_buf[i].fe_logical 883 / block_size; 884 ext_list->data.len = ext_buf[i].fe_length 885 / block_size; 886 887 ret = insert_extent_by_physical( 888 ext_list_head, ext_list); 889 if (ret < 0) { 890 FREE(ext_list); 891 goto out; 892 } 893 } 894 /* Record file's logical offset this time */ 895 pos = ext_buf[EXTENT_MAX_COUNT-1].fe_logical + 896 ext_buf[EXTENT_MAX_COUNT-1].fe_length; 897 /* 898 * If fm_extents array has been filled and 899 * there are extents left, continue to cycle. 900 */ 901 } while (fiemap_buf->fm_mapped_extents 902 == EXTENT_MAX_COUNT && 903 !(ext_buf[EXTENT_MAX_COUNT-1].fe_flags 904 & FIEMAP_EXTENT_LAST)); 905 906 FREE(fiemap_buf); 907 return 0; 908out: 909 FREE(fiemap_buf); 910 return -1; 911} 912 913/* 914 * get_logical_count() - Get the file logical extents count. 915 * 916 * @logical_list_head: the head of the logical extent list. 917 */ 918static int get_logical_count(struct fiemap_extent_list *logical_list_head) 919{ 920 int ret = 0; 921 struct fiemap_extent_list *ext_list_tmp = logical_list_head; 922 923 do { 924 ret++; 925 ext_list_tmp = ext_list_tmp->next; 926 } while (ext_list_tmp != logical_list_head); 927 928 return ret; 929} 930 931/* 932 * get_physical_count() - Get the file physical extents count. 933 * 934 * @physical_list_head: the head of the physical extent list. 935 */ 936static int get_physical_count(struct fiemap_extent_list *physical_list_head) 937{ 938 int ret = 0; 939 struct fiemap_extent_list *ext_list_tmp = physical_list_head; 940 941 do { 942 if ((ext_list_tmp->data.physical + ext_list_tmp->data.len) 943 != ext_list_tmp->next->data.physical) { 944 /* This extent and next extent are not continuous. */ 945 ret++; 946 } 947 948 ext_list_tmp = ext_list_tmp->next; 949 } while (ext_list_tmp != physical_list_head); 950 951 return ret; 952} 953 954/* 955 * change_physical_to_logical() - Change list from physical to logical. 956 * 957 * @physical_list_head: the head of physical extent list. 958 * @logical_list_head: the head of logical extent list. 959 */ 960static int change_physical_to_logical( 961 struct fiemap_extent_list **physical_list_head, 962 struct fiemap_extent_list **logical_list_head) 963{ 964 int ret; 965 struct fiemap_extent_list *ext_list_tmp = *physical_list_head; 966 struct fiemap_extent_list *ext_list_next = ext_list_tmp->next; 967 968 while (1) { 969 if (ext_list_tmp == ext_list_next) { 970 ret = insert_extent_by_logical( 971 logical_list_head, ext_list_tmp); 972 if (ret < 0) 973 return -1; 974 975 *physical_list_head = NULL; 976 break; 977 } 978 979 ext_list_tmp->prev->next = ext_list_tmp->next; 980 ext_list_tmp->next->prev = ext_list_tmp->prev; 981 *physical_list_head = ext_list_next; 982 983 ret = insert_extent_by_logical( 984 logical_list_head, ext_list_tmp); 985 if (ret < 0) { 986 FREE(ext_list_tmp); 987 return -1; 988 } 989 ext_list_tmp = ext_list_next; 990 ext_list_next = ext_list_next->next; 991 } 992 993 return 0; 994} 995 996/* get_file_blocks() - Get total file blocks. 997 * 998 * @ext_list_head: the extent list head of the target file 999 */ 1000static ext4_fsblk_t get_file_blocks(struct fiemap_extent_list *ext_list_head) 1001{ 1002 ext4_fsblk_t blk_count = 0; 1003 struct fiemap_extent_list *ext_list_tmp = ext_list_head; 1004 1005 do { 1006 blk_count += ext_list_tmp->data.len; 1007 ext_list_tmp = ext_list_tmp->next; 1008 } while (ext_list_tmp != ext_list_head); 1009 1010 return blk_count; 1011} 1012 1013/* 1014 * free_ext() - Free the extent list. 1015 * 1016 * @ext_list_head: the extent list head of which will be free. 1017 */ 1018static void free_ext(struct fiemap_extent_list *ext_list_head) 1019{ 1020 struct fiemap_extent_list *ext_list_tmp = NULL; 1021 1022 if (ext_list_head == NULL) 1023 return; 1024 1025 while (ext_list_head->next != ext_list_head) { 1026 ext_list_tmp = ext_list_head; 1027 ext_list_head->prev->next = ext_list_head->next; 1028 ext_list_head->next->prev = ext_list_head->prev; 1029 ext_list_head = ext_list_head->next; 1030 free(ext_list_tmp); 1031 } 1032 free(ext_list_head); 1033} 1034 1035/* 1036 * free_exts_group() - Free the exts_group. 1037 * 1038 * @*ext_group_head: the exts_group list head which will be free. 1039 */ 1040static void free_exts_group(struct fiemap_extent_group *ext_group_head) 1041{ 1042 struct fiemap_extent_group *ext_group_tmp = NULL; 1043 1044 if (ext_group_head == NULL) 1045 return; 1046 1047 while (ext_group_head->next != ext_group_head) { 1048 ext_group_tmp = ext_group_head; 1049 ext_group_head->prev->next = ext_group_head->next; 1050 ext_group_head->next->prev = ext_group_head->prev; 1051 ext_group_head = ext_group_head->next; 1052 free(ext_group_tmp); 1053 } 1054 free(ext_group_head); 1055} 1056 1057/* 1058 * get_best_count() - Get the file best extents count. 1059 * 1060 * @block_count: the file's physical block count. 1061 */ 1062static int get_best_count(ext4_fsblk_t block_count) 1063{ 1064 int ret; 1065 unsigned int flex_bg_num; 1066 1067 /* Calcuate best extents count */ 1068 if (feature_incompat & EXT4_FEATURE_INCOMPAT_FLEX_BG) { 1069 flex_bg_num = 1 << log_groups_per_flex; 1070 ret = ((block_count - 1) / 1071 ((ext4_fsblk_t)blocks_per_group * 1072 flex_bg_num)) + 1; 1073 } else 1074 ret = ((block_count - 1) / blocks_per_group) + 1; 1075 1076 return ret; 1077} 1078 1079 1080/* 1081 * file_statistic() - Get statistic info of the file's fragments. 1082 * 1083 * @file: the file's name. 1084 * @buf: the pointer of the struct stat64. 1085 * @flag: file type. 1086 * @ftwbuf: the pointer of a struct FTW. 1087 */ 1088static int file_statistic(const char *file, const struct stat64 *buf, 1089 int flag EXT2FS_ATTR((unused)), 1090 struct FTW *ftwbuf EXT2FS_ATTR((unused))) 1091{ 1092 int fd; 1093 int ret; 1094 int now_ext_count, best_ext_count = 0, physical_ext_count; 1095 int i, j; 1096 __u64 size_per_ext = 0; 1097 float ratio = 0.0; 1098 ext4_fsblk_t blk_count = 0; 1099 char msg_buffer[PATH_MAX + 24]; 1100 struct fiemap_extent_list *physical_list_head = NULL; 1101 struct fiemap_extent_list *logical_list_head = NULL; 1102 1103 defraged_file_count++; 1104 1105 if (mode_flag & DETAIL) { 1106 if (total_count == 1 && regular_count == 1) 1107 printf("<File>\n"); 1108 else { 1109 printf("[%u/%u]", defraged_file_count, total_count); 1110 fflush(stdout); 1111 } 1112 } 1113 if (lost_found_dir[0] != '\0' && 1114 !memcmp(file, lost_found_dir, strnlen(lost_found_dir, PATH_MAX))) { 1115 if (mode_flag & DETAIL) { 1116 PRINT_FILE_NAME(file); 1117 STATISTIC_ERR_MSG(NGMSG_LOST_FOUND); 1118 } 1119 return 0; 1120 } 1121 1122 if (!S_ISREG(buf->st_mode)) { 1123 if (mode_flag & DETAIL) { 1124 PRINT_FILE_NAME(file); 1125 STATISTIC_ERR_MSG(NGMSG_FILE_UNREG); 1126 } 1127 return 0; 1128 } 1129 1130 /* Access authority */ 1131 if (current_uid != ROOT_UID && 1132 buf->st_uid != current_uid) { 1133 if (mode_flag & DETAIL) { 1134 PRINT_FILE_NAME(file); 1135 STATISTIC_ERR_MSG( 1136 "File is not current user's file" 1137 " or current user is not root"); 1138 } 1139 return 0; 1140 } 1141 1142 /* Empty file */ 1143 if (buf->st_size == 0) { 1144 if (mode_flag & DETAIL) { 1145 PRINT_FILE_NAME(file); 1146 STATISTIC_ERR_MSG("File size is 0"); 1147 } 1148 return 0; 1149 } 1150 1151 /* Has no blocks */ 1152 if (buf->st_blocks == 0) { 1153 if (mode_flag & DETAIL) { 1154 PRINT_FILE_NAME(file); 1155 STATISTIC_ERR_MSG("File has no blocks"); 1156 } 1157 return 0; 1158 } 1159 1160 fd = open64(file, O_RDONLY); 1161 if (fd < 0) { 1162 if (mode_flag & DETAIL) { 1163 PRINT_FILE_NAME(file); 1164 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN); 1165 } 1166 return 0; 1167 } 1168 1169 /* Get file's physical extents */ 1170 ret = get_file_extents(fd, &physical_list_head); 1171 if (ret < 0) { 1172 if (mode_flag & DETAIL) { 1173 PRINT_FILE_NAME(file); 1174 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1175 } 1176 goto out; 1177 } 1178 1179 /* Get the count of file's continuous physical region */ 1180 physical_ext_count = get_physical_count(physical_list_head); 1181 1182 /* Change list from physical to logical */ 1183 ret = change_physical_to_logical(&physical_list_head, 1184 &logical_list_head); 1185 if (ret < 0) { 1186 if (mode_flag & DETAIL) { 1187 PRINT_FILE_NAME(file); 1188 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1189 } 1190 goto out; 1191 } 1192 1193 /* Count file fragments before defrag */ 1194 now_ext_count = get_logical_count(logical_list_head); 1195 1196 if (current_uid == ROOT_UID) { 1197 /* Calculate the size per extent */ 1198 blk_count = get_file_blocks(logical_list_head); 1199 1200 best_ext_count = get_best_count(blk_count); 1201 1202 /* e4defrag rounds size_per_ext up to a block size boundary */ 1203 size_per_ext = blk_count * (buf->st_blksize / 1024) / 1204 now_ext_count; 1205 1206 ratio = (float)(physical_ext_count - best_ext_count) * 100 / 1207 blk_count; 1208 1209 extents_before_defrag += now_ext_count; 1210 extents_after_defrag += best_ext_count; 1211 files_block_count += blk_count; 1212 } 1213 1214 if (total_count == 1 && regular_count == 1) { 1215 /* File only */ 1216 if (mode_flag & DETAIL) { 1217 int count = 0; 1218 struct fiemap_extent_list *ext_list_tmp = 1219 logical_list_head; 1220 1221 /* Print extents info */ 1222 do { 1223 count++; 1224 printf("[ext %d]:\tstart %llu:\tlogical " 1225 "%llu:\tlen %llu\n", count, 1226 ext_list_tmp->data.physical, 1227 ext_list_tmp->data.logical, 1228 ext_list_tmp->data.len); 1229 ext_list_tmp = ext_list_tmp->next; 1230 } while (ext_list_tmp != logical_list_head); 1231 1232 } else { 1233 printf("%-40s%10s/%-10s%9s\n", 1234 "<File>", "now", "best", "size/ext"); 1235 if (current_uid == ROOT_UID) { 1236 if (strlen(file) > 40) 1237 printf("%s\n%50d/%-10d%6llu KB\n", 1238 file, now_ext_count, 1239 best_ext_count, size_per_ext); 1240 else 1241 printf("%-40s%10d/%-10d%6llu KB\n", 1242 file, now_ext_count, 1243 best_ext_count, size_per_ext); 1244 } else { 1245 if (strlen(file) > 40) 1246 printf("%s\n%50d/%-10s%7s\n", 1247 file, now_ext_count, 1248 "-", "-"); 1249 else 1250 printf("%-40s%10d/%-10s%7s\n", 1251 file, now_ext_count, 1252 "-", "-"); 1253 } 1254 } 1255 succeed_cnt++; 1256 goto out; 1257 } 1258 1259 if (mode_flag & DETAIL) { 1260 /* Print statistic info */ 1261 sprintf(msg_buffer, "[%u/%u]%s", 1262 defraged_file_count, total_count, file); 1263 if (current_uid == ROOT_UID) { 1264 if (strlen(msg_buffer) > 40) 1265 printf("\033[79;0H\033[K%s\n" 1266 "%50d/%-10d%6llu KB\n", 1267 msg_buffer, now_ext_count, 1268 best_ext_count, size_per_ext); 1269 else 1270 printf("\033[79;0H\033[K%-40s" 1271 "%10d/%-10d%6llu KB\n", 1272 msg_buffer, now_ext_count, 1273 best_ext_count, size_per_ext); 1274 } else { 1275 if (strlen(msg_buffer) > 40) 1276 printf("\033[79;0H\033[K%s\n%50d/%-10s%7s\n", 1277 msg_buffer, now_ext_count, 1278 "-", "-"); 1279 else 1280 printf("\033[79;0H\033[K%-40s%10d/%-10s%7s\n", 1281 msg_buffer, now_ext_count, 1282 "-", "-"); 1283 } 1284 } 1285 1286 for (i = 0; i < SHOW_FRAG_FILES; i++) { 1287 if (ratio >= frag_rank[i].ratio) { 1288 for (j = SHOW_FRAG_FILES - 1; j > i; j--) { 1289 memset(&frag_rank[j], 0, 1290 sizeof(struct frag_statistic_ino)); 1291 strncpy(frag_rank[j].msg_buffer, 1292 frag_rank[j - 1].msg_buffer, 1293 strnlen(frag_rank[j - 1].msg_buffer, 1294 PATH_MAX)); 1295 frag_rank[j].now_count = 1296 frag_rank[j - 1].now_count; 1297 frag_rank[j].best_count = 1298 frag_rank[j - 1].best_count; 1299 frag_rank[j].size_per_ext = 1300 frag_rank[j - 1].size_per_ext; 1301 frag_rank[j].ratio = 1302 frag_rank[j - 1].ratio; 1303 } 1304 memset(&frag_rank[i], 0, 1305 sizeof(struct frag_statistic_ino)); 1306 strncpy(frag_rank[i].msg_buffer, file, 1307 strnlen(file, PATH_MAX)); 1308 frag_rank[i].now_count = now_ext_count; 1309 frag_rank[i].best_count = best_ext_count; 1310 frag_rank[i].size_per_ext = size_per_ext; 1311 frag_rank[i].ratio = ratio; 1312 break; 1313 } 1314 } 1315 1316 succeed_cnt++; 1317 1318out: 1319 close(fd); 1320 free_ext(physical_list_head); 1321 free_ext(logical_list_head); 1322 return 0; 1323} 1324 1325/* 1326 * print_progress - Print defrag progress 1327 * 1328 * @file: file name. 1329 * @start: logical offset for defrag target file 1330 * @file_size: defrag target filesize 1331 */ 1332static void print_progress(const char *file, loff_t start, loff_t file_size) 1333{ 1334 int percent = (start * 100) / file_size; 1335 printf("\033[79;0H\033[K[%u/%u]%s:\t%3d%%", 1336 defraged_file_count, total_count, file, min(percent, 100)); 1337 fflush(stdout); 1338 1339 return; 1340} 1341 1342/* 1343 * call_defrag() - Execute the defrag program. 1344 * 1345 * @fd: target file descriptor. 1346 * @donor_fd: donor file descriptor. 1347 * @file: target file name. 1348 * @buf: pointer of the struct stat64. 1349 * @ext_list_head: head of the extent list. 1350 */ 1351static int call_defrag(int fd, int donor_fd, const char *file, 1352 const struct stat64 *buf, struct fiemap_extent_list *ext_list_head) 1353{ 1354 loff_t start = 0; 1355 unsigned int page_num; 1356 unsigned char *vec = NULL; 1357 int defraged_ret = 0; 1358 int ret; 1359 struct move_extent move_data; 1360 struct fiemap_extent_list *ext_list_tmp = NULL; 1361 1362 memset(&move_data, 0, sizeof(struct move_extent)); 1363 move_data.donor_fd = donor_fd; 1364 1365 /* Print defrag progress */ 1366 print_progress(file, start, buf->st_size); 1367 1368 ext_list_tmp = ext_list_head; 1369 do { 1370 move_data.orig_start = ext_list_tmp->data.logical; 1371 /* Logical offset of orig and donor should be same */ 1372 move_data.donor_start = move_data.orig_start; 1373 move_data.len = ext_list_tmp->data.len; 1374 move_data.moved_len = 0; 1375 1376 ret = page_in_core(fd, move_data, &vec, &page_num); 1377 if (ret < 0) { 1378 if (mode_flag & DETAIL) { 1379 printf("\n"); 1380 PRINT_ERR_MSG_WITH_ERRNO( 1381 "Failed to get file map"); 1382 } else { 1383 printf("\t[ NG ]\n"); 1384 } 1385 return -1; 1386 } 1387 1388 /* EXT4_IOC_MOVE_EXT */ 1389 defraged_ret = 1390 ioctl(fd, EXT4_IOC_MOVE_EXT, &move_data); 1391 1392 /* Free pages */ 1393 ret = defrag_fadvise(fd, move_data, vec, page_num); 1394 if (vec) { 1395 free(vec); 1396 vec = NULL; 1397 } 1398 if (ret < 0) { 1399 if (mode_flag & DETAIL) { 1400 printf("\n"); 1401 PRINT_ERR_MSG_WITH_ERRNO( 1402 "Failed to free page"); 1403 } else { 1404 printf("\t[ NG ]\n"); 1405 } 1406 return -1; 1407 } 1408 1409 if (defraged_ret < 0) { 1410 if (mode_flag & DETAIL) { 1411 printf("\n"); 1412 PRINT_ERR_MSG_WITH_ERRNO( 1413 "Failed to defrag with " 1414 "EXT4_IOC_MOVE_EXT ioctl"); 1415 if (errno == ENOTTY) 1416 printf("\tAt least 2.6.31-rc1 of " 1417 "vanilla kernel is required\n"); 1418 } else { 1419 printf("\t[ NG ]\n"); 1420 } 1421 return -1; 1422 } 1423 /* Adjust logical offset for next ioctl */ 1424 move_data.orig_start += move_data.moved_len; 1425 move_data.donor_start = move_data.orig_start; 1426 1427 start = move_data.orig_start * buf->st_blksize; 1428 1429 /* Print defrag progress */ 1430 print_progress(file, start, buf->st_size); 1431 1432 /* End of file */ 1433 if (start >= buf->st_size) 1434 break; 1435 1436 ext_list_tmp = ext_list_tmp->next; 1437 } while (ext_list_tmp != ext_list_head); 1438 1439 return 0; 1440} 1441 1442/* 1443 * file_defrag() - Check file attributes and call ioctl to defrag. 1444 * 1445 * @file: the file's name. 1446 * @buf: the pointer of the struct stat64. 1447 * @flag: file type. 1448 * @ftwbuf: the pointer of a struct FTW. 1449 */ 1450static int file_defrag(const char *file, const struct stat64 *buf, 1451 int flag EXT2FS_ATTR((unused)), 1452 struct FTW *ftwbuf EXT2FS_ATTR((unused))) 1453{ 1454 int fd; 1455 int donor_fd = -1; 1456 int ret; 1457 int best; 1458 int file_frags_start, file_frags_end; 1459 int orig_physical_cnt, donor_physical_cnt = 0; 1460 char tmp_inode_name[PATH_MAX + 8]; 1461 ext4_fsblk_t blk_count = 0; 1462 struct fiemap_extent_list *orig_list_physical = NULL; 1463 struct fiemap_extent_list *orig_list_logical = NULL; 1464 struct fiemap_extent_list *donor_list_physical = NULL; 1465 struct fiemap_extent_list *donor_list_logical = NULL; 1466 struct fiemap_extent_group *orig_group_head = NULL; 1467 struct fiemap_extent_group *orig_group_tmp = NULL; 1468 1469 defraged_file_count++; 1470 1471 if (mode_flag & DETAIL) { 1472 printf("[%u/%u]", defraged_file_count, total_count); 1473 fflush(stdout); 1474 } 1475 1476 if (lost_found_dir[0] != '\0' && 1477 !memcmp(file, lost_found_dir, strnlen(lost_found_dir, PATH_MAX))) { 1478 if (mode_flag & DETAIL) { 1479 PRINT_FILE_NAME(file); 1480 IN_FTW_PRINT_ERR_MSG(NGMSG_LOST_FOUND); 1481 } 1482 return 0; 1483 } 1484 1485 if (!S_ISREG(buf->st_mode)) { 1486 if (mode_flag & DETAIL) { 1487 PRINT_FILE_NAME(file); 1488 IN_FTW_PRINT_ERR_MSG(NGMSG_FILE_UNREG); 1489 } 1490 return 0; 1491 } 1492 1493 /* Empty file */ 1494 if (buf->st_size == 0) { 1495 if (mode_flag & DETAIL) { 1496 PRINT_FILE_NAME(file); 1497 IN_FTW_PRINT_ERR_MSG("File size is 0"); 1498 } 1499 return 0; 1500 } 1501 1502 /* Has no blocks */ 1503 if (buf->st_blocks == 0) { 1504 if (mode_flag & DETAIL) { 1505 PRINT_FILE_NAME(file); 1506 STATISTIC_ERR_MSG("File has no blocks"); 1507 } 1508 return 0; 1509 } 1510 1511 fd = open64(file, O_RDWR); 1512 if (fd < 0) { 1513 if (mode_flag & DETAIL) { 1514 PRINT_FILE_NAME(file); 1515 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN); 1516 } 1517 return 0; 1518 } 1519 1520 /* Get file's extents */ 1521 ret = get_file_extents(fd, &orig_list_physical); 1522 if (ret < 0) { 1523 if (mode_flag & DETAIL) { 1524 PRINT_FILE_NAME(file); 1525 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1526 } 1527 goto out; 1528 } 1529 1530 /* Get the count of file's continuous physical region */ 1531 orig_physical_cnt = get_physical_count(orig_list_physical); 1532 1533 /* Change list from physical to logical */ 1534 ret = change_physical_to_logical(&orig_list_physical, 1535 &orig_list_logical); 1536 if (ret < 0) { 1537 if (mode_flag & DETAIL) { 1538 PRINT_FILE_NAME(file); 1539 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1540 } 1541 goto out; 1542 } 1543 1544 /* Count file fragments before defrag */ 1545 file_frags_start = get_logical_count(orig_list_logical); 1546 1547 blk_count = get_file_blocks(orig_list_logical); 1548 if (file_check(fd, buf, file, file_frags_start, blk_count) < 0) 1549 goto out; 1550 1551 if (fsync(fd) < 0) { 1552 if (mode_flag & DETAIL) { 1553 PRINT_FILE_NAME(file); 1554 PRINT_ERR_MSG_WITH_ERRNO("Failed to sync(fsync)"); 1555 } 1556 goto out; 1557 } 1558 1559 if (current_uid == ROOT_UID) 1560 best = get_best_count(blk_count); 1561 else 1562 best = 1; 1563 1564 if (file_frags_start <= best) 1565 goto check_improvement; 1566 1567 /* Combine extents to group */ 1568 ret = join_extents(orig_list_logical, &orig_group_head); 1569 if (ret < 0) { 1570 if (mode_flag & DETAIL) { 1571 PRINT_FILE_NAME(file); 1572 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1573 } 1574 goto out; 1575 } 1576 1577 /* Create donor inode */ 1578 memset(tmp_inode_name, 0, PATH_MAX + 8); 1579 sprintf(tmp_inode_name, "%.*s.defrag", 1580 (int)strnlen(file, PATH_MAX), file); 1581 donor_fd = open64(tmp_inode_name, O_WRONLY | O_CREAT | O_EXCL, S_IRUSR); 1582 if (donor_fd < 0) { 1583 if (mode_flag & DETAIL) { 1584 PRINT_FILE_NAME(file); 1585 if (errno == EEXIST) 1586 PRINT_ERR_MSG_WITH_ERRNO( 1587 "File is being defraged by other program"); 1588 else 1589 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN); 1590 } 1591 goto out; 1592 } 1593 1594 /* Unlink donor inode */ 1595 ret = unlink(tmp_inode_name); 1596 if (ret < 0) { 1597 if (mode_flag & DETAIL) { 1598 PRINT_FILE_NAME(file); 1599 PRINT_ERR_MSG_WITH_ERRNO("Failed to unlink"); 1600 } 1601 goto out; 1602 } 1603 1604 /* Allocate space for donor inode */ 1605 orig_group_tmp = orig_group_head; 1606 do { 1607 ret = fallocate64(donor_fd, 0, 1608 (loff_t)orig_group_tmp->start->data.logical * block_size, 1609 (loff_t)orig_group_tmp->len * block_size); 1610 if (ret < 0) { 1611 if (mode_flag & DETAIL) { 1612 PRINT_FILE_NAME(file); 1613 PRINT_ERR_MSG_WITH_ERRNO("Failed to fallocate"); 1614 } 1615 goto out; 1616 } 1617 1618 orig_group_tmp = orig_group_tmp->next; 1619 } while (orig_group_tmp != orig_group_head); 1620 1621 /* Get donor inode's extents */ 1622 ret = get_file_extents(donor_fd, &donor_list_physical); 1623 if (ret < 0) { 1624 if (mode_flag & DETAIL) { 1625 PRINT_FILE_NAME(file); 1626 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1627 } 1628 goto out; 1629 } 1630 1631 /* Calcuate donor inode's continuous physical region */ 1632 donor_physical_cnt = get_physical_count(donor_list_physical); 1633 1634 /* Change donor extent list from physical to logical */ 1635 ret = change_physical_to_logical(&donor_list_physical, 1636 &donor_list_logical); 1637 if (ret < 0) { 1638 if (mode_flag & DETAIL) { 1639 PRINT_FILE_NAME(file); 1640 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT); 1641 } 1642 goto out; 1643 } 1644 1645check_improvement: 1646 if (mode_flag & DETAIL) { 1647 if (file_frags_start != 1) 1648 frag_files_before_defrag++; 1649 1650 extents_before_defrag += file_frags_start; 1651 } 1652 1653 if (file_frags_start <= best || 1654 orig_physical_cnt <= donor_physical_cnt) { 1655 printf("\033[79;0H\033[K[%u/%u]%s:\t%3d%%", 1656 defraged_file_count, total_count, file, 100); 1657 if (mode_flag & DETAIL) 1658 printf(" extents: %d -> %d", 1659 file_frags_start, file_frags_start); 1660 1661 printf("\t[ OK ]\n"); 1662 succeed_cnt++; 1663 1664 if (file_frags_start != 1) 1665 frag_files_after_defrag++; 1666 1667 extents_after_defrag += file_frags_start; 1668 goto out; 1669 } 1670 1671 /* Defrag the file */ 1672 ret = call_defrag(fd, donor_fd, file, buf, donor_list_logical); 1673 1674 /* Count file fragments after defrag and print extents info */ 1675 if (mode_flag & DETAIL) { 1676 file_frags_end = file_frag_count(fd); 1677 if (file_frags_end < 0) { 1678 printf("\n"); 1679 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_INFO); 1680 goto out; 1681 } 1682 1683 if (file_frags_end != 1) 1684 frag_files_after_defrag++; 1685 1686 extents_after_defrag += file_frags_end; 1687 1688 if (ret < 0) 1689 goto out; 1690 1691 printf(" extents: %d -> %d", 1692 file_frags_start, file_frags_end); 1693 fflush(stdout); 1694 } 1695 1696 if (ret < 0) 1697 goto out; 1698 1699 printf("\t[ OK ]\n"); 1700 fflush(stdout); 1701 succeed_cnt++; 1702 1703out: 1704 close(fd); 1705 if (donor_fd != -1) 1706 close(donor_fd); 1707 free_ext(orig_list_physical); 1708 free_ext(orig_list_logical); 1709 free_ext(donor_list_physical); 1710 free_exts_group(orig_group_head); 1711 return 0; 1712} 1713 1714/* 1715 * main() - Ext4 online defrag. 1716 * 1717 * @argc: the number of parameter. 1718 * @argv[]: the pointer array of parameter. 1719 */ 1720int main(int argc, char *argv[]) 1721{ 1722 int opt; 1723 int i, j, ret = 0; 1724 int flags = FTW_PHYS | FTW_MOUNT; 1725 int arg_type = -1; 1726 int success_flag = 0; 1727 char dir_name[PATH_MAX + 1]; 1728 char dev_name[PATH_MAX + 1]; 1729 struct stat64 buf; 1730 ext2_filsys fs = NULL; 1731 1732 /* Parse arguments */ 1733 if (argc == 1) 1734 goto out; 1735 1736 while ((opt = getopt(argc, argv, "vc")) != EOF) { 1737 switch (opt) { 1738 case 'v': 1739 mode_flag |= DETAIL; 1740 break; 1741 case 'c': 1742 mode_flag |= STATISTIC; 1743 break; 1744 default: 1745 goto out; 1746 } 1747 } 1748 1749 if (argc == optind) 1750 goto out; 1751 1752 current_uid = getuid(); 1753 1754 /* Main process */ 1755 for (i = optind; i < argc; i++) { 1756 succeed_cnt = 0; 1757 regular_count = 0; 1758 total_count = 0; 1759 frag_files_before_defrag = 0; 1760 frag_files_after_defrag = 0; 1761 extents_before_defrag = 0; 1762 extents_after_defrag = 0; 1763 defraged_file_count = 0; 1764 files_block_count = 0; 1765 blocks_per_group = 0; 1766 feature_incompat = 0; 1767 log_groups_per_flex = 0; 1768 1769 memset(dir_name, 0, PATH_MAX + 1); 1770 memset(dev_name, 0, PATH_MAX + 1); 1771 memset(lost_found_dir, 0, PATH_MAX + 1); 1772 memset(frag_rank, 0, 1773 sizeof(struct frag_statistic_ino) * SHOW_FRAG_FILES); 1774 1775 if ((mode_flag & STATISTIC) && i > optind) 1776 printf("\n"); 1777 1778#if BYTE_ORDER != BIG_ENDIAN && BYTE_ORDER != LITTLE_ENDIAN 1779 PRINT_ERR_MSG("Endian's type is not big/little endian"); 1780 PRINT_FILE_NAME(argv[i]); 1781 continue; 1782#endif 1783 1784 if (lstat64(argv[i], &buf) < 0) { 1785 perror(NGMSG_FILE_INFO); 1786 PRINT_FILE_NAME(argv[i]); 1787 continue; 1788 } 1789 1790 /* Handle i.e. lvm device symlinks */ 1791 if (S_ISLNK(buf.st_mode)) { 1792 struct stat64 buf2; 1793 1794 if (stat64(argv[i], &buf2) == 0 && 1795 S_ISBLK(buf2.st_mode)) 1796 buf = buf2; 1797 } 1798 1799 if (S_ISBLK(buf.st_mode)) { 1800 /* Block device */ 1801 strncpy(dev_name, argv[i], strnlen(argv[i], PATH_MAX)); 1802 if (get_mount_point(argv[i], dir_name, PATH_MAX) < 0) 1803 continue; 1804 if (lstat64(dir_name, &buf) < 0) { 1805 perror(NGMSG_FILE_INFO); 1806 PRINT_FILE_NAME(argv[i]); 1807 continue; 1808 } 1809 arg_type = DEVNAME; 1810 if (!(mode_flag & STATISTIC)) 1811 printf("ext4 defragmentation for device(%s)\n", 1812 argv[i]); 1813 } else if (S_ISDIR(buf.st_mode)) { 1814 /* Directory */ 1815 if (access(argv[i], R_OK) < 0) { 1816 perror(argv[i]); 1817 continue; 1818 } 1819 arg_type = DIRNAME; 1820 strncpy(dir_name, argv[i], strnlen(argv[i], PATH_MAX)); 1821 } else if (S_ISREG(buf.st_mode)) { 1822 /* Regular file */ 1823 arg_type = FILENAME; 1824 } else { 1825 /* Irregular file */ 1826 PRINT_ERR_MSG(NGMSG_FILE_UNREG); 1827 PRINT_FILE_NAME(argv[i]); 1828 continue; 1829 } 1830 1831 /* Set blocksize */ 1832 block_size = buf.st_blksize; 1833 1834 /* For device case, 1835 * filesystem type checked in get_mount_point() 1836 */ 1837 if (arg_type == FILENAME || arg_type == DIRNAME) { 1838 if (is_ext4(argv[i], dev_name) < 0) 1839 continue; 1840 if (realpath(argv[i], dir_name) == NULL) { 1841 perror("Couldn't get full path"); 1842 PRINT_FILE_NAME(argv[i]); 1843 continue; 1844 } 1845 } 1846 1847 if (current_uid == ROOT_UID) { 1848 /* Get super block info */ 1849 ret = ext2fs_open(dev_name, 0, 0, block_size, 1850 unix_io_manager, &fs); 1851 if (ret) { 1852 if (mode_flag & DETAIL) { 1853 perror("Can't get super block info"); 1854 PRINT_FILE_NAME(argv[i]); 1855 } 1856 continue; 1857 } 1858 1859 blocks_per_group = fs->super->s_blocks_per_group; 1860 feature_incompat = fs->super->s_feature_incompat; 1861 log_groups_per_flex = fs->super->s_log_groups_per_flex; 1862 1863 ext2fs_close(fs); 1864 } 1865 1866 switch (arg_type) { 1867 int mount_dir_len = 0; 1868 1869 case DIRNAME: 1870 if (!(mode_flag & STATISTIC)) 1871 printf("ext4 defragmentation " 1872 "for directory(%s)\n", argv[i]); 1873 1874 mount_dir_len = strnlen(lost_found_dir, PATH_MAX); 1875 1876 strncat(lost_found_dir, "/lost+found", 1877 PATH_MAX - strnlen(lost_found_dir, PATH_MAX)); 1878 1879 /* Not the case("e4defrag mount_piont_dir") */ 1880 if (dir_name[mount_dir_len] != '\0') { 1881 /* 1882 * "e4defrag mount_piont_dir/lost+found" 1883 * or "e4defrag mount_piont_dir/lost+found/" 1884 */ 1885 if (strncmp(lost_found_dir, dir_name, 1886 strnlen(lost_found_dir, 1887 PATH_MAX)) == 0 && 1888 (dir_name[strnlen(lost_found_dir, 1889 PATH_MAX)] == '\0' || 1890 dir_name[strnlen(lost_found_dir, 1891 PATH_MAX)] == '/')) { 1892 PRINT_ERR_MSG(NGMSG_LOST_FOUND); 1893 PRINT_FILE_NAME(argv[i]); 1894 continue; 1895 } 1896 1897 /* "e4defrag mount_piont_dir/else_dir" */ 1898 memset(lost_found_dir, 0, PATH_MAX + 1); 1899 } 1900 case DEVNAME: 1901 if (arg_type == DEVNAME) { 1902 strncpy(lost_found_dir, dir_name, 1903 strnlen(dir_name, PATH_MAX)); 1904 strncat(lost_found_dir, "/lost+found/", 1905 PATH_MAX - strnlen(lost_found_dir, 1906 PATH_MAX)); 1907 } 1908 1909 nftw64(dir_name, calc_entry_counts, FTW_OPEN_FD, flags); 1910 1911 if (mode_flag & STATISTIC) { 1912 if (mode_flag & DETAIL) 1913 printf("%-40s%10s/%-10s%9s\n", 1914 "<File>", "now", "best", "size/ext"); 1915 1916 if (!(mode_flag & DETAIL) && 1917 current_uid != ROOT_UID) { 1918 printf(" Done.\n"); 1919 success_flag = 1; 1920 continue; 1921 } 1922 1923 nftw64(dir_name, file_statistic, 1924 FTW_OPEN_FD, flags); 1925 1926 if (succeed_cnt != 0 && 1927 current_uid == ROOT_UID) { 1928 if (mode_flag & DETAIL) 1929 printf("\n"); 1930 printf("%-40s%10s/%-10s%9s\n", 1931 "<Fragmented files>", "now", 1932 "best", "size/ext"); 1933 for (j = 0; j < SHOW_FRAG_FILES; j++) { 1934 if (strlen(frag_rank[j]. 1935 msg_buffer) > 37) { 1936 printf("%d. %s\n%50d/" 1937 "%-10d%6llu KB\n", 1938 j + 1, 1939 frag_rank[j].msg_buffer, 1940 frag_rank[j].now_count, 1941 frag_rank[j].best_count, 1942 frag_rank[j]. 1943 size_per_ext); 1944 } else if (strlen(frag_rank[j]. 1945 msg_buffer) > 0) { 1946 printf("%d. %-37s%10d/" 1947 "%-10d%6llu KB\n", 1948 j + 1, 1949 frag_rank[j].msg_buffer, 1950 frag_rank[j].now_count, 1951 frag_rank[j].best_count, 1952 frag_rank[j]. 1953 size_per_ext); 1954 } else 1955 break; 1956 } 1957 } 1958 break; 1959 } 1960 /* File tree walk */ 1961 nftw64(dir_name, file_defrag, FTW_OPEN_FD, flags); 1962 printf("\n\tSuccess:\t\t\t[ %u/%u ]\n", succeed_cnt, 1963 total_count); 1964 printf("\tFailure:\t\t\t[ %u/%u ]\n", 1965 total_count - succeed_cnt, total_count); 1966 if (mode_flag & DETAIL) { 1967 printf("\tTotal extents:\t\t\t%4d->%d\n", 1968 extents_before_defrag, 1969 extents_after_defrag); 1970 printf("\tFragmented percentage:\t\t" 1971 "%3llu%%->%llu%%\n", 1972 !regular_count ? 0 : 1973 ((unsigned long long) 1974 frag_files_before_defrag * 100) / 1975 regular_count, 1976 !regular_count ? 0 : 1977 ((unsigned long long) 1978 frag_files_after_defrag * 100) / 1979 regular_count); 1980 } 1981 break; 1982 case FILENAME: 1983 total_count = 1; 1984 regular_count = 1; 1985 strncat(lost_found_dir, "/lost+found/", 1986 PATH_MAX - strnlen(lost_found_dir, 1987 PATH_MAX)); 1988 if (strncmp(lost_found_dir, dir_name, 1989 strnlen(lost_found_dir, 1990 PATH_MAX)) == 0) { 1991 PRINT_ERR_MSG(NGMSG_LOST_FOUND); 1992 PRINT_FILE_NAME(argv[i]); 1993 continue; 1994 } 1995 1996 if (mode_flag & STATISTIC) { 1997 file_statistic(argv[i], &buf, FTW_F, NULL); 1998 break; 1999 } else 2000 printf("ext4 defragmentation for %s\n", 2001 argv[i]); 2002 /* Defrag single file process */ 2003 file_defrag(argv[i], &buf, FTW_F, NULL); 2004 if (succeed_cnt != 0) 2005 printf(" Success:\t\t\t[1/1]\n"); 2006 else 2007 printf(" Success:\t\t\t[0/1]\n"); 2008 2009 break; 2010 } 2011 2012 if (succeed_cnt != 0) 2013 success_flag = 1; 2014 if (mode_flag & STATISTIC) { 2015 if (current_uid != ROOT_UID) { 2016 printf(" Done.\n"); 2017 continue; 2018 } 2019 2020 if (!succeed_cnt) { 2021 if (mode_flag & DETAIL) 2022 printf("\n"); 2023 2024 if (arg_type == DEVNAME) 2025 printf(" In this device(%s), " 2026 "none can be defragmented.\n", argv[i]); 2027 else if (arg_type == DIRNAME) 2028 printf(" In this directory(%s), " 2029 "none can be defragmented.\n", argv[i]); 2030 else 2031 printf(" This file(%s) " 2032 "can't be defragmented.\n", argv[i]); 2033 } else { 2034 float files_ratio = 0.0; 2035 float score = 0.0; 2036 __u64 size_per_ext = files_block_count * 2037 (buf.st_blksize / 1024) / 2038 extents_before_defrag; 2039 files_ratio = (float)(extents_before_defrag - 2040 extents_after_defrag) * 2041 100 / files_block_count; 2042 score = CALC_SCORE(files_ratio); 2043 printf("\n Total/best extents\t\t\t\t%d/%d\n" 2044 " Average size per extent" 2045 "\t\t\t%llu KB\n" 2046 " Fragmentation score\t\t\t\t%.0f\n", 2047 extents_before_defrag, 2048 extents_after_defrag, 2049 size_per_ext, score); 2050 printf(" [0-30 no problem:" 2051 " 31-55 a little bit fragmented:" 2052 " 56- needs defrag]\n"); 2053 2054 if (arg_type == DEVNAME) 2055 printf(" This device (%s) ", argv[i]); 2056 else if (arg_type == DIRNAME) 2057 printf(" This directory (%s) ", 2058 argv[i]); 2059 else 2060 printf(" This file (%s) ", argv[i]); 2061 2062 if (score > BOUND_SCORE) 2063 printf("needs defragmentation.\n"); 2064 else 2065 printf("does not need " 2066 "defragmentation.\n"); 2067 } 2068 printf(" Done.\n"); 2069 } 2070 2071 } 2072 2073 if (success_flag) 2074 return 0; 2075 2076 exit(1); 2077 2078out: 2079 printf(MSG_USAGE); 2080 exit(1); 2081} 2082 2083