]> git.ipfire.org Git - thirdparty/e2fsprogs.git/blob - misc/e4defrag.c
b6e2e319524320b94ea5642713478bbf993d73bf
[thirdparty/e2fsprogs.git] / misc / e4defrag.c
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 "config.h"
23 #include <ctype.h>
24 #include <dirent.h>
25 #include <endian.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #include <ftw.h>
29 #include <limits.h>
30 #include <mntent.h>
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <unistd.h>
35 #include <ext2fs/ext2_types.h>
36 #include <ext2fs/ext2fs.h>
37 #include <linux/fs.h>
38 #include <sys/ioctl.h>
39 #include <ext2fs/fiemap.h>
40 #include <sys/mman.h>
41 #include <sys/stat.h>
42 #include <sys/statfs.h>
43 #include <sys/syscall.h>
44 #include <sys/vfs.h>
45
46 /* A relatively new ioctl interface ... */
47 #ifndef EXT4_IOC_MOVE_EXT
48 #define EXT4_IOC_MOVE_EXT _IOWR('f', 15, struct move_extent)
49 #endif
50
51 /* Macro functions */
52 #define PRINT_ERR_MSG(msg) fprintf(stderr, "%s\n", (msg))
53 #define IN_FTW_PRINT_ERR_MSG(msg) \
54 fprintf(stderr, "\t%s\t\t[ NG ]\n", (msg))
55 #define PRINT_FILE_NAME(file) fprintf(stderr, " \"%s\"\n", (file))
56 #define PRINT_ERR_MSG_WITH_ERRNO(msg) \
57 fprintf(stderr, "\t%s:%s\t[ NG ]\n", (msg), strerror(errno))
58 #define STATISTIC_ERR_MSG(msg) \
59 fprintf(stderr, "\t%s\n", (msg))
60 #define STATISTIC_ERR_MSG_WITH_ERRNO(msg) \
61 fprintf(stderr, "\t%s:%s\n", (msg), strerror(errno))
62 #define min(x, y) (((x) > (y)) ? (y) : (x))
63 #define CALC_SCORE(ratio) \
64 ((ratio) > 10 ? (80 + 20 * (ratio) / 100) : (8 * (ratio)))
65 /* Wrap up the free function */
66 #define FREE(tmp) \
67 do { \
68 if ((tmp) != NULL) \
69 free(tmp); \
70 } while (0) \
71 /* Insert list2 after list1 */
72 #define insert(list1, list2) \
73 do { \
74 list2->next = list1->next; \
75 list1->next->prev = list2; \
76 list2->prev = list1; \
77 list1->next = list2; \
78 } while (0)
79
80 /* To delete unused warning */
81 #ifdef __GNUC__
82 #define EXT2FS_ATTR(x) __attribute__(x)
83 #else
84 #define EXT2FS_ATTR(x)
85 #endif
86
87 /* The mode of defrag */
88 #define DETAIL 0x01
89 #define STATISTIC 0x02
90
91 #define DEVNAME 0
92 #define DIRNAME 1
93 #define FILENAME 2
94
95 #define FTW_OPEN_FD 2000
96
97 #define FS_EXT4 "ext4"
98 #define ROOT_UID 0
99
100 #define BOUND_SCORE 55
101 #define SHOW_FRAG_FILES 5
102
103 /* Magic number for ext4 */
104 #define EXT4_SUPER_MAGIC 0xEF53
105
106 /* Definition of flex_bg */
107 #define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
108
109 /* The following macro is used for ioctl FS_IOC_FIEMAP
110 * EXTENT_MAX_COUNT: the maximum number of extents for exchanging between
111 * kernel-space and user-space per ioctl
112 */
113 #define EXTENT_MAX_COUNT 512
114
115 /* The following macros are error message */
116 #define MSG_USAGE \
117 "Usage : e4defrag [-v] file...| directory...| device...\n\
118 : e4defrag -c file...| directory...| device...\n"
119
120 #define NGMSG_EXT4 "Filesystem is not ext4 filesystem"
121 #define NGMSG_FILE_EXTENT "Failed to get file extents"
122 #define NGMSG_FILE_INFO "Failed to get file information"
123 #define NGMSG_FILE_OPEN "Failed to open"
124 #define NGMSG_FILE_UNREG "File is not regular file"
125 #define NGMSG_LOST_FOUND "Can not process \"lost+found\""
126
127 /* Data type for filesystem-wide blocks number */
128 typedef unsigned long long ext4_fsblk_t;
129
130 struct fiemap_extent_data {
131 __u64 len; /* blocks count */
132 __u64 logical; /* start logical block number */
133 ext4_fsblk_t physical; /* start physical block number */
134 };
135
136 struct fiemap_extent_list {
137 struct fiemap_extent_list *prev;
138 struct fiemap_extent_list *next;
139 struct fiemap_extent_data data; /* extent belong to file */
140 };
141
142 struct fiemap_extent_group {
143 struct fiemap_extent_group *prev;
144 struct fiemap_extent_group *next;
145 __u64 len; /* length of this continuous region */
146 struct fiemap_extent_list *start; /* start ext */
147 struct fiemap_extent_list *end; /* end ext */
148 };
149
150 struct move_extent {
151 __s32 reserved; /* original file descriptor */
152 __u32 donor_fd; /* donor file descriptor */
153 __u64 orig_start; /* logical start offset in block for orig */
154 __u64 donor_start; /* logical start offset in block for donor */
155 __u64 len; /* block length to be moved */
156 __u64 moved_len; /* moved block length */
157 };
158
159 struct frag_statistic_ino {
160 int now_count; /* the file's extents count of before defrag */
161 int best_count; /* the best file's extents count */
162 __u64 size_per_ext; /* size(KB) per extent */
163 float ratio; /* the ratio of fragmentation */
164 char msg_buffer[PATH_MAX + 1]; /* pathname of the file */
165 };
166
167 char lost_found_dir[PATH_MAX + 1];
168 int block_size;
169 int extents_before_defrag;
170 int extents_after_defrag;
171 int mode_flag;
172 unsigned int current_uid;
173 unsigned int defraged_file_count;
174 unsigned int frag_files_before_defrag;
175 unsigned int frag_files_after_defrag;
176 unsigned int regular_count;
177 unsigned int succeed_cnt;
178 unsigned int total_count;
179 __u8 log_groups_per_flex;
180 __u32 blocks_per_group;
181 __u32 feature_incompat;
182 ext4_fsblk_t files_block_count;
183 struct frag_statistic_ino frag_rank[SHOW_FRAG_FILES];
184
185
186 /* Local definitions of some syscalls glibc may not yet have */
187
188 #ifndef HAVE_POSIX_FADVISE
189 #warning Using locally defined posix_fadvise interface.
190
191 #ifndef __NR_fadvise64_64
192 #error Your kernel headers dont define __NR_fadvise64_64
193 #endif
194
195 /*
196 * fadvise() - Give advice about file access.
197 *
198 * @fd: defrag target file's descriptor.
199 * @offset: file offset.
200 * @len: area length.
201 * @advise: process flag.
202 */
203 static int posix_fadvise(int fd, loff_t offset, size_t len, int advise)
204 {
205 return syscall(__NR_fadvise64_64, fd, offset, len, advise);
206 }
207 #endif /* ! HAVE_FADVISE64_64 */
208
209 #ifndef HAVE_SYNC_FILE_RANGE
210 #warning Using locally defined sync_file_range interface.
211
212 #ifndef __NR_sync_file_range
213 #ifndef __NR_sync_file_range2 /* ppc */
214 #error Your kernel headers dont define __NR_sync_file_range
215 #endif
216 #endif
217
218 /*
219 * sync_file_range() - Sync file region.
220 *
221 * @fd: defrag target file's descriptor.
222 * @offset: file offset.
223 * @length: area length.
224 * @flag: process flag.
225 */
226 int sync_file_range(int fd, loff_t offset, loff_t length, unsigned int flag)
227 {
228 #ifdef __NR_sync_file_range
229 return syscall(__NR_sync_file_range, fd, offset, length, flag);
230 #else
231 return syscall(__NR_sync_file_range2, fd, flag, offset, length);
232 #endif
233 }
234 #endif /* ! HAVE_SYNC_FILE_RANGE */
235
236 #ifndef HAVE_FALLOCATE64
237 #warning Using locally defined fallocate syscall interface.
238
239 #ifndef __NR_fallocate
240 #error Your kernel headers dont define __NR_fallocate
241 #endif
242
243 /*
244 * fallocate64() - Manipulate file space.
245 *
246 * @fd: defrag target file's descriptor.
247 * @mode: process flag.
248 * @offset: file offset.
249 * @len: file size.
250 */
251 static int fallocate64(int fd, int mode, loff_t offset, loff_t len)
252 {
253 return syscall(__NR_fallocate, fd, mode, offset, len);
254 }
255 #endif /* ! HAVE_FALLOCATE */
256
257 /*
258 * get_mount_point() - Get device's mount point.
259 *
260 * @devname: the device's name.
261 * @mount_point: the mount point.
262 * @dir_path_len: the length of directory.
263 */
264 static int get_mount_point(const char *devname, char *mount_point,
265 int dir_path_len)
266 {
267 /* Refer to /etc/mtab */
268 const char *mtab = MOUNTED;
269 FILE *fp = NULL;
270 struct mntent *mnt = NULL;
271 struct stat64 sb;
272
273 if (stat64(devname, &sb) < 0) {
274 perror(NGMSG_FILE_INFO);
275 PRINT_FILE_NAME(devname);
276 return -1;
277 }
278
279 fp = setmntent(mtab, "r");
280 if (fp == NULL) {
281 perror("Couldn't access /etc/mtab");
282 return -1;
283 }
284
285 while ((mnt = getmntent(fp)) != NULL) {
286 struct stat64 ms;
287
288 /*
289 * To handle device symlinks, we see if the
290 * device number matches, not the name
291 */
292 if (stat64(mnt->mnt_fsname, &ms) < 0)
293 continue;
294 if (sb.st_rdev != ms.st_rdev)
295 continue;
296
297 endmntent(fp);
298 if (strcmp(mnt->mnt_type, FS_EXT4) == 0) {
299 strncpy(mount_point, mnt->mnt_dir,
300 dir_path_len);
301 return 0;
302 }
303 PRINT_ERR_MSG(NGMSG_EXT4);
304 return -1;
305 }
306 endmntent(fp);
307 PRINT_ERR_MSG("Filesystem is not mounted");
308 return -1;
309 }
310
311 /*
312 * is_ext4() - Whether on an ext4 filesystem.
313 *
314 * @file: the file's name.
315 */
316 static int is_ext4(const char *file, char *devname)
317 {
318 int maxlen = 0;
319 int len, ret;
320 FILE *fp = NULL;
321 char *mnt_type = NULL;
322 /* Refer to /etc/mtab */
323 const char *mtab = MOUNTED;
324 char file_path[PATH_MAX + 1];
325 struct mntent *mnt = NULL;
326 struct statfs64 fsbuf;
327
328 /* Get full path */
329 if (realpath(file, file_path) == NULL) {
330 perror("Couldn't get full path");
331 PRINT_FILE_NAME(file);
332 return -1;
333 }
334
335 if (statfs64(file_path, &fsbuf) < 0) {
336 perror("Failed to get filesystem information");
337 PRINT_FILE_NAME(file);
338 return -1;
339 }
340
341 if (fsbuf.f_type != EXT4_SUPER_MAGIC) {
342 PRINT_ERR_MSG(NGMSG_EXT4);
343 return -1;
344 }
345
346 fp = setmntent(mtab, "r");
347 if (fp == NULL) {
348 perror("Couldn't access /etc/mtab");
349 return -1;
350 }
351
352 while ((mnt = getmntent(fp)) != NULL) {
353 if (mnt->mnt_fsname[0] != '/')
354 continue;
355 len = strlen(mnt->mnt_dir);
356 ret = memcmp(file_path, mnt->mnt_dir, len);
357 if (ret != 0)
358 continue;
359
360 if (maxlen >= len)
361 continue;
362
363 maxlen = len;
364
365 mnt_type = realloc(mnt_type, strlen(mnt->mnt_type) + 1);
366 if (mnt_type == NULL) {
367 endmntent(fp);
368 return -1;
369 }
370 memset(mnt_type, 0, strlen(mnt->mnt_type) + 1);
371 strncpy(mnt_type, mnt->mnt_type, strlen(mnt->mnt_type));
372 strncpy(lost_found_dir, mnt->mnt_dir, PATH_MAX);
373 strncpy(devname, mnt->mnt_fsname, strlen(mnt->mnt_fsname) + 1);
374 }
375
376 endmntent(fp);
377 if (mnt_type && strcmp(mnt_type, FS_EXT4) == 0) {
378 FREE(mnt_type);
379 return 0;
380 } else {
381 FREE(mnt_type);
382 PRINT_ERR_MSG(NGMSG_EXT4);
383 return -1;
384 }
385 }
386
387 /*
388 * calc_entry_counts() - Calculate file counts.
389 *
390 * @file: file name.
391 * @buf: file info.
392 * @flag: file type.
393 * @ftwbuf: the pointer of a struct FTW.
394 */
395 static int calc_entry_counts(const char *file EXT2FS_ATTR((unused)),
396 const struct stat64 *buf, int flag EXT2FS_ATTR((unused)),
397 struct FTW *ftwbuf EXT2FS_ATTR((unused)))
398 {
399 if (S_ISREG(buf->st_mode))
400 regular_count++;
401
402 total_count++;
403
404 return 0;
405 }
406
407 /*
408 * page_in_core() - Get information on whether pages are in core.
409 *
410 * @fd: defrag target file's descriptor.
411 * @defrag_data: data used for defrag.
412 * @vec: page state array.
413 * @page_num: page number.
414 */
415 static int page_in_core(int fd, struct move_extent defrag_data,
416 unsigned char **vec, unsigned int *page_num)
417 {
418 long pagesize;
419 void *page = NULL;
420 loff_t offset, end_offset, length;
421
422 if (vec == NULL || *vec != NULL)
423 return -1;
424
425 pagesize = sysconf(_SC_PAGESIZE);
426 if (pagesize < 0)
427 return -1;
428 /* In mmap, offset should be a multiple of the page size */
429 offset = (loff_t)defrag_data.orig_start * block_size;
430 length = (loff_t)defrag_data.len * block_size;
431 end_offset = offset + length;
432 /* Round the offset down to the nearest multiple of pagesize */
433 offset = (offset / pagesize) * pagesize;
434 length = end_offset - offset;
435
436 page = mmap(NULL, length, PROT_READ, MAP_SHARED, fd, offset);
437 if (page == MAP_FAILED)
438 return -1;
439
440 *page_num = 0;
441 *page_num = (length + pagesize - 1) / pagesize;
442 *vec = (unsigned char *)calloc(*page_num, 1);
443 if (*vec == NULL)
444 return -1;
445
446 /* Get information on whether pages are in core */
447 if (mincore(page, (size_t)length, *vec) == -1 ||
448 munmap(page, length) == -1) {
449 FREE(*vec);
450 return -1;
451 }
452
453 return 0;
454 }
455
456 /*
457 * defrag_fadvise() - Predeclare an access pattern for file data.
458 *
459 * @fd: defrag target file's descriptor.
460 * @defrag_data: data used for defrag.
461 * @vec: page state array.
462 * @page_num: page number.
463 */
464 static int defrag_fadvise(int fd, struct move_extent defrag_data,
465 unsigned char *vec, unsigned int page_num)
466 {
467 int flag = 1;
468 long pagesize = sysconf(_SC_PAGESIZE);
469 int fadvise_flag = POSIX_FADV_DONTNEED;
470 int sync_flag = SYNC_FILE_RANGE_WAIT_BEFORE |
471 SYNC_FILE_RANGE_WRITE |
472 SYNC_FILE_RANGE_WAIT_AFTER;
473 unsigned int i;
474 loff_t offset;
475
476 offset = (loff_t)defrag_data.orig_start * block_size;
477 offset = (offset / pagesize) * pagesize;
478
479 /* Sync file for fadvise process */
480 if (sync_file_range(fd, offset,
481 (loff_t)pagesize * page_num, sync_flag) < 0)
482 return -1;
483
484 /* Try to release buffer cache which this process used,
485 * then other process can use the released buffer
486 */
487 for (i = 0; i < page_num; i++) {
488 if ((vec[i] & 0x1) == 0) {
489 offset += pagesize;
490 continue;
491 }
492 if (posix_fadvise(fd, offset, pagesize, fadvise_flag) < 0) {
493 if ((mode_flag & DETAIL) && flag) {
494 perror("\tFailed to fadvise");
495 flag = 0;
496 }
497 }
498 offset += pagesize;
499 }
500
501 return 0;
502 }
503
504 /*
505 * check_free_size() - Check if there's enough disk space.
506 *
507 * @fd: defrag target file's descriptor.
508 * @file: file name.
509 * @blk_count: file blocks.
510 */
511 static int check_free_size(int fd, const char *file, ext4_fsblk_t blk_count)
512 {
513 ext4_fsblk_t free_blk_count;
514 struct statfs64 fsbuf;
515
516 if (fstatfs64(fd, &fsbuf) < 0) {
517 if (mode_flag & DETAIL) {
518 PRINT_FILE_NAME(file);
519 PRINT_ERR_MSG_WITH_ERRNO(
520 "Failed to get filesystem information");
521 }
522 return -1;
523 }
524
525 /* Compute free space for root and normal user separately */
526 if (current_uid == ROOT_UID)
527 free_blk_count = fsbuf.f_bfree;
528 else
529 free_blk_count = fsbuf.f_bavail;
530
531 if (free_blk_count >= blk_count)
532 return 0;
533
534 return -ENOSPC;
535 }
536
537 /*
538 * file_frag_count() - Get file fragment count.
539 *
540 * @fd: defrag target file's descriptor.
541 */
542 static int file_frag_count(int fd)
543 {
544 int ret;
545 struct fiemap fiemap_buf;
546
547 /* When fm_extent_count is 0,
548 * ioctl just get file fragment count.
549 */
550 memset(&fiemap_buf, 0, sizeof(struct fiemap));
551 fiemap_buf.fm_start = 0;
552 fiemap_buf.fm_length = FIEMAP_MAX_OFFSET;
553 fiemap_buf.fm_flags |= FIEMAP_FLAG_SYNC;
554
555 ret = ioctl(fd, FS_IOC_FIEMAP, &fiemap_buf);
556 if (ret < 0)
557 return ret;
558
559 return fiemap_buf.fm_mapped_extents;
560 }
561
562 /*
563 * file_check() - Check file's attributes.
564 *
565 * @fd: defrag target file's descriptor.
566 * @buf: a pointer of the struct stat64.
567 * @file: file name.
568 * @extents: file extents.
569 * @blk_count: file blocks.
570 */
571 static int file_check(int fd, const struct stat64 *buf, const char *file,
572 int extents, ext4_fsblk_t blk_count)
573 {
574 int ret;
575 struct flock lock;
576
577 /* Write-lock check is more reliable */
578 lock.l_type = F_WRLCK;
579 lock.l_start = 0;
580 lock.l_whence = SEEK_SET;
581 lock.l_len = 0;
582
583 /* Free space */
584 ret = check_free_size(fd, file, blk_count);
585 if (ret < 0) {
586 if ((mode_flag & DETAIL) && ret == -ENOSPC) {
587 printf("\033[79;0H\033[K[%u/%u] \"%s\"\t\t"
588 " extents: %d -> %d\n", defraged_file_count,
589 total_count, file, extents, extents);
590 IN_FTW_PRINT_ERR_MSG(
591 "Defrag size is larger than filesystem's free space");
592 }
593 return -1;
594 }
595
596 /* Access authority */
597 if (current_uid != ROOT_UID &&
598 buf->st_uid != current_uid) {
599 if (mode_flag & DETAIL) {
600 printf("\033[79;0H\033[K[%u/%u] \"%s\"\t\t"
601 " extents: %d -> %d\n", defraged_file_count,
602 total_count, file, extents, extents);
603 IN_FTW_PRINT_ERR_MSG(
604 "File is not current user's file"
605 " or current user is not root");
606 }
607 return -1;
608 }
609
610 /* Lock status */
611 if (fcntl(fd, F_GETLK, &lock) < 0) {
612 if (mode_flag & DETAIL) {
613 PRINT_FILE_NAME(file);
614 PRINT_ERR_MSG_WITH_ERRNO(
615 "Failed to get lock information");
616 }
617 return -1;
618 } else if (lock.l_type != F_UNLCK) {
619 if (mode_flag & DETAIL) {
620 PRINT_FILE_NAME(file);
621 IN_FTW_PRINT_ERR_MSG("File has been locked");
622 }
623 return -1;
624 }
625
626 return 0;
627 }
628
629 /*
630 * insert_extent_by_logical() - Sequentially insert extent by logical.
631 *
632 * @ext_list_head: the head of logical extent list.
633 * @ext: the extent element which will be inserted.
634 */
635 static int insert_extent_by_logical(struct fiemap_extent_list **ext_list_head,
636 struct fiemap_extent_list *ext)
637 {
638 struct fiemap_extent_list *ext_list_tmp = *ext_list_head;
639
640 if (ext == NULL)
641 goto out;
642
643 /* First element */
644 if (*ext_list_head == NULL) {
645 (*ext_list_head) = ext;
646 (*ext_list_head)->prev = *ext_list_head;
647 (*ext_list_head)->next = *ext_list_head;
648 return 0;
649 }
650
651 if (ext->data.logical <= ext_list_tmp->data.logical) {
652 /* Insert before head */
653 if (ext_list_tmp->data.logical <
654 ext->data.logical + ext->data.len)
655 /* Overlap */
656 goto out;
657 /* Adjust head */
658 *ext_list_head = ext;
659 } else {
660 /* Insert into the middle or last of the list */
661 do {
662 if (ext->data.logical < ext_list_tmp->data.logical)
663 break;
664 ext_list_tmp = ext_list_tmp->next;
665 } while (ext_list_tmp != (*ext_list_head));
666 if (ext->data.logical <
667 ext_list_tmp->prev->data.logical +
668 ext_list_tmp->prev->data.len)
669 /* Overlap */
670 goto out;
671
672 if (ext_list_tmp != *ext_list_head &&
673 ext_list_tmp->data.logical <
674 ext->data.logical + ext->data.len)
675 /* Overlap */
676 goto out;
677 }
678 ext_list_tmp = ext_list_tmp->prev;
679 /* Insert "ext" after "ext_list_tmp" */
680 insert(ext_list_tmp, ext);
681 return 0;
682 out:
683 errno = EINVAL;
684 return -1;
685 }
686
687 /*
688 * insert_extent_by_physical() - Sequentially insert extent by physical.
689 *
690 * @ext_list_head: the head of physical extent list.
691 * @ext: the extent element which will be inserted.
692 */
693 static int insert_extent_by_physical(struct fiemap_extent_list **ext_list_head,
694 struct fiemap_extent_list *ext)
695 {
696 struct fiemap_extent_list *ext_list_tmp = *ext_list_head;
697
698 if (ext == NULL)
699 goto out;
700
701 /* First element */
702 if (*ext_list_head == NULL) {
703 (*ext_list_head) = ext;
704 (*ext_list_head)->prev = *ext_list_head;
705 (*ext_list_head)->next = *ext_list_head;
706 return 0;
707 }
708
709 if (ext->data.physical <= ext_list_tmp->data.physical) {
710 /* Insert before head */
711 if (ext_list_tmp->data.physical <
712 ext->data.physical + ext->data.len)
713 /* Overlap */
714 goto out;
715 /* Adjust head */
716 *ext_list_head = ext;
717 } else {
718 /* Insert into the middle or last of the list */
719 do {
720 if (ext->data.physical < ext_list_tmp->data.physical)
721 break;
722 ext_list_tmp = ext_list_tmp->next;
723 } while (ext_list_tmp != (*ext_list_head));
724 if (ext->data.physical <
725 ext_list_tmp->prev->data.physical +
726 ext_list_tmp->prev->data.len)
727 /* Overlap */
728 goto out;
729
730 if (ext_list_tmp != *ext_list_head &&
731 ext_list_tmp->data.physical <
732 ext->data.physical + ext->data.len)
733 /* Overlap */
734 goto out;
735 }
736 ext_list_tmp = ext_list_tmp->prev;
737 /* Insert "ext" after "ext_list_tmp" */
738 insert(ext_list_tmp, ext);
739 return 0;
740 out:
741 errno = EINVAL;
742 return -1;
743 }
744
745 /*
746 * insert_exts_group() - Insert a exts_group.
747 *
748 * @ext_group_head: the head of a exts_group list.
749 * @exts_group: the exts_group element which will be inserted.
750 */
751 static int insert_exts_group(struct fiemap_extent_group **ext_group_head,
752 struct fiemap_extent_group *exts_group)
753 {
754 struct fiemap_extent_group *ext_group_tmp = NULL;
755
756 if (exts_group == NULL) {
757 errno = EINVAL;
758 return -1;
759 }
760
761 /* Initialize list */
762 if (*ext_group_head == NULL) {
763 (*ext_group_head) = exts_group;
764 (*ext_group_head)->prev = *ext_group_head;
765 (*ext_group_head)->next = *ext_group_head;
766 return 0;
767 }
768
769 ext_group_tmp = (*ext_group_head)->prev;
770 insert(ext_group_tmp, exts_group);
771
772 return 0;
773 }
774
775 /*
776 * join_extents() - Find continuous region(exts_group).
777 *
778 * @ext_list_head: the head of the extent list.
779 * @ext_group_head: the head of the target exts_group list.
780 */
781 static int join_extents(struct fiemap_extent_list *ext_list_head,
782 struct fiemap_extent_group **ext_group_head)
783 {
784 __u64 len = ext_list_head->data.len;
785 struct fiemap_extent_list *ext_list_start = ext_list_head;
786 struct fiemap_extent_list *ext_list_tmp = ext_list_head->next;
787
788 do {
789 struct fiemap_extent_group *ext_group_tmp = NULL;
790
791 /* This extent and previous extent are not continuous,
792 * so, all previous extents are treated as an extent group.
793 */
794 if ((ext_list_tmp->prev->data.logical +
795 ext_list_tmp->prev->data.len)
796 != ext_list_tmp->data.logical) {
797 ext_group_tmp =
798 malloc(sizeof(struct fiemap_extent_group));
799 if (ext_group_tmp == NULL)
800 return -1;
801
802 memset(ext_group_tmp, 0,
803 sizeof(struct fiemap_extent_group));
804 ext_group_tmp->len = len;
805 ext_group_tmp->start = ext_list_start;
806 ext_group_tmp->end = ext_list_tmp->prev;
807
808 if (insert_exts_group(ext_group_head,
809 ext_group_tmp) < 0) {
810 FREE(ext_group_tmp);
811 return -1;
812 }
813 ext_list_start = ext_list_tmp;
814 len = ext_list_tmp->data.len;
815 ext_list_tmp = ext_list_tmp->next;
816 continue;
817 }
818
819 /* This extent and previous extent are continuous,
820 * so, they belong to the same extent group, and we check
821 * if the next extent belongs to the same extent group.
822 */
823 len += ext_list_tmp->data.len;
824 ext_list_tmp = ext_list_tmp->next;
825 } while (ext_list_tmp != ext_list_head->next);
826
827 return 0;
828 }
829
830 /*
831 * get_file_extents() - Get file's extent list.
832 *
833 * @fd: defrag target file's descriptor.
834 * @ext_list_head: the head of the extent list.
835 */
836 static int get_file_extents(int fd, struct fiemap_extent_list **ext_list_head)
837 {
838 __u32 i;
839 int ret;
840 int ext_buf_size, fie_buf_size;
841 __u64 pos = 0;
842 struct fiemap *fiemap_buf = NULL;
843 struct fiemap_extent *ext_buf = NULL;
844 struct fiemap_extent_list *ext_list = NULL;
845
846 /* Convert units, in bytes.
847 * Be careful : now, physical block number in extent is 48bit,
848 * and the maximum blocksize for ext4 is 4K(12bit),
849 * so there is no overflow, but in future it may be changed.
850 */
851
852 /* Alloc space for fiemap */
853 ext_buf_size = EXTENT_MAX_COUNT * sizeof(struct fiemap_extent);
854 fie_buf_size = sizeof(struct fiemap) + ext_buf_size;
855
856 fiemap_buf = malloc(fie_buf_size);
857 if (fiemap_buf == NULL)
858 return -1;
859
860 ext_buf = fiemap_buf->fm_extents;
861 memset(fiemap_buf, 0, fie_buf_size);
862 fiemap_buf->fm_length = FIEMAP_MAX_OFFSET;
863 fiemap_buf->fm_flags |= FIEMAP_FLAG_SYNC;
864 fiemap_buf->fm_extent_count = EXTENT_MAX_COUNT;
865
866 do {
867 fiemap_buf->fm_start = pos;
868 memset(ext_buf, 0, ext_buf_size);
869 ret = ioctl(fd, FS_IOC_FIEMAP, fiemap_buf);
870 if (ret < 0 || fiemap_buf->fm_mapped_extents == 0)
871 goto out;
872 for (i = 0; i < fiemap_buf->fm_mapped_extents; i++) {
873 ext_list = NULL;
874 ext_list = malloc(sizeof(struct fiemap_extent_list));
875 if (ext_list == NULL)
876 goto out;
877
878 ext_list->data.physical = ext_buf[i].fe_physical
879 / block_size;
880 ext_list->data.logical = ext_buf[i].fe_logical
881 / block_size;
882 ext_list->data.len = ext_buf[i].fe_length
883 / block_size;
884
885 ret = insert_extent_by_physical(
886 ext_list_head, ext_list);
887 if (ret < 0) {
888 FREE(ext_list);
889 goto out;
890 }
891 }
892 /* Record file's logical offset this time */
893 pos = ext_buf[EXTENT_MAX_COUNT-1].fe_logical +
894 ext_buf[EXTENT_MAX_COUNT-1].fe_length;
895 /*
896 * If fm_extents array has been filled and
897 * there are extents left, continue to cycle.
898 */
899 } while (fiemap_buf->fm_mapped_extents
900 == EXTENT_MAX_COUNT &&
901 !(ext_buf[EXTENT_MAX_COUNT-1].fe_flags
902 & FIEMAP_EXTENT_LAST));
903
904 FREE(fiemap_buf);
905 return 0;
906 out:
907 FREE(fiemap_buf);
908 return -1;
909 }
910
911 /*
912 * get_logical_count() - Get the file logical extents count.
913 *
914 * @logical_list_head: the head of the logical extent list.
915 */
916 static int get_logical_count(struct fiemap_extent_list *logical_list_head)
917 {
918 int ret = 0;
919 struct fiemap_extent_list *ext_list_tmp = logical_list_head;
920
921 do {
922 ret++;
923 ext_list_tmp = ext_list_tmp->next;
924 } while (ext_list_tmp != logical_list_head);
925
926 return ret;
927 }
928
929 /*
930 * get_physical_count() - Get the file physical extents count.
931 *
932 * @physical_list_head: the head of the physical extent list.
933 */
934 static int get_physical_count(struct fiemap_extent_list *physical_list_head)
935 {
936 int ret = 0;
937 struct fiemap_extent_list *ext_list_tmp = physical_list_head;
938
939 do {
940 if ((ext_list_tmp->data.physical + ext_list_tmp->data.len)
941 != ext_list_tmp->next->data.physical) {
942 /* This extent and next extent are not continuous. */
943 ret++;
944 }
945
946 ext_list_tmp = ext_list_tmp->next;
947 } while (ext_list_tmp != physical_list_head);
948
949 return ret;
950 }
951
952 /*
953 * change_physical_to_logical() - Change list from physical to logical.
954 *
955 * @physical_list_head: the head of physical extent list.
956 * @logical_list_head: the head of logical extent list.
957 */
958 static int change_physical_to_logical(
959 struct fiemap_extent_list **physical_list_head,
960 struct fiemap_extent_list **logical_list_head)
961 {
962 int ret;
963 struct fiemap_extent_list *ext_list_tmp = *physical_list_head;
964 struct fiemap_extent_list *ext_list_next = ext_list_tmp->next;
965
966 while (1) {
967 if (ext_list_tmp == ext_list_next) {
968 ret = insert_extent_by_logical(
969 logical_list_head, ext_list_tmp);
970 if (ret < 0)
971 return -1;
972
973 *physical_list_head = NULL;
974 break;
975 }
976
977 ext_list_tmp->prev->next = ext_list_tmp->next;
978 ext_list_tmp->next->prev = ext_list_tmp->prev;
979 *physical_list_head = ext_list_next;
980
981 ret = insert_extent_by_logical(
982 logical_list_head, ext_list_tmp);
983 if (ret < 0) {
984 FREE(ext_list_tmp);
985 return -1;
986 }
987 ext_list_tmp = ext_list_next;
988 ext_list_next = ext_list_next->next;
989 }
990
991 return 0;
992 }
993
994 /* get_file_blocks() - Get total file blocks.
995 *
996 * @ext_list_head: the extent list head of the target file
997 */
998 static ext4_fsblk_t get_file_blocks(struct fiemap_extent_list *ext_list_head)
999 {
1000 ext4_fsblk_t blk_count = 0;
1001 struct fiemap_extent_list *ext_list_tmp = ext_list_head;
1002
1003 do {
1004 blk_count += ext_list_tmp->data.len;
1005 ext_list_tmp = ext_list_tmp->next;
1006 } while (ext_list_tmp != ext_list_head);
1007
1008 return blk_count;
1009 }
1010
1011 /*
1012 * free_ext() - Free the extent list.
1013 *
1014 * @ext_list_head: the extent list head of which will be free.
1015 */
1016 static void free_ext(struct fiemap_extent_list *ext_list_head)
1017 {
1018 struct fiemap_extent_list *ext_list_tmp = NULL;
1019
1020 if (ext_list_head == NULL)
1021 return;
1022
1023 while (ext_list_head->next != ext_list_head) {
1024 ext_list_tmp = ext_list_head;
1025 ext_list_head->prev->next = ext_list_head->next;
1026 ext_list_head->next->prev = ext_list_head->prev;
1027 ext_list_head = ext_list_head->next;
1028 free(ext_list_tmp);
1029 }
1030 free(ext_list_head);
1031 }
1032
1033 /*
1034 * free_exts_group() - Free the exts_group.
1035 *
1036 * @*ext_group_head: the exts_group list head which will be free.
1037 */
1038 static void free_exts_group(struct fiemap_extent_group *ext_group_head)
1039 {
1040 struct fiemap_extent_group *ext_group_tmp = NULL;
1041
1042 if (ext_group_head == NULL)
1043 return;
1044
1045 while (ext_group_head->next != ext_group_head) {
1046 ext_group_tmp = ext_group_head;
1047 ext_group_head->prev->next = ext_group_head->next;
1048 ext_group_head->next->prev = ext_group_head->prev;
1049 ext_group_head = ext_group_head->next;
1050 free(ext_group_tmp);
1051 }
1052 free(ext_group_head);
1053 }
1054
1055 /*
1056 * get_best_count() - Get the file best extents count.
1057 *
1058 * @block_count: the file's physical block count.
1059 */
1060 static int get_best_count(ext4_fsblk_t block_count)
1061 {
1062 int ret;
1063 unsigned int flex_bg_num;
1064
1065 /* Calcuate best extents count */
1066 if (feature_incompat & EXT4_FEATURE_INCOMPAT_FLEX_BG) {
1067 flex_bg_num = 1 << log_groups_per_flex;
1068 ret = ((block_count - 1) /
1069 ((ext4_fsblk_t)blocks_per_group *
1070 flex_bg_num)) + 1;
1071 } else
1072 ret = ((block_count - 1) / blocks_per_group) + 1;
1073
1074 return ret;
1075 }
1076
1077
1078 /*
1079 * file_statistic() - Get statistic info of the file's fragments.
1080 *
1081 * @file: the file's name.
1082 * @buf: the pointer of the struct stat64.
1083 * @flag: file type.
1084 * @ftwbuf: the pointer of a struct FTW.
1085 */
1086 static int file_statistic(const char *file, const struct stat64 *buf,
1087 int flag EXT2FS_ATTR((unused)),
1088 struct FTW *ftwbuf EXT2FS_ATTR((unused)))
1089 {
1090 int fd;
1091 int ret;
1092 int now_ext_count, best_ext_count = 0, physical_ext_count;
1093 int i, j;
1094 __u64 size_per_ext = 0;
1095 float ratio = 0.0;
1096 ext4_fsblk_t blk_count = 0;
1097 char msg_buffer[PATH_MAX + 24];
1098 struct fiemap_extent_list *physical_list_head = NULL;
1099 struct fiemap_extent_list *logical_list_head = NULL;
1100
1101 defraged_file_count++;
1102
1103 if (mode_flag & DETAIL) {
1104 if (total_count == 1 && regular_count == 1)
1105 printf("<File>\n");
1106 else {
1107 printf("[%u/%u]", defraged_file_count, total_count);
1108 fflush(stdout);
1109 }
1110 }
1111 if (lost_found_dir[0] != '\0' &&
1112 !memcmp(file, lost_found_dir, strnlen(lost_found_dir, PATH_MAX))) {
1113 if (mode_flag & DETAIL) {
1114 PRINT_FILE_NAME(file);
1115 STATISTIC_ERR_MSG(NGMSG_LOST_FOUND);
1116 }
1117 return 0;
1118 }
1119
1120 if (!S_ISREG(buf->st_mode)) {
1121 if (mode_flag & DETAIL) {
1122 PRINT_FILE_NAME(file);
1123 STATISTIC_ERR_MSG(NGMSG_FILE_UNREG);
1124 }
1125 return 0;
1126 }
1127
1128 /* Access authority */
1129 if (current_uid != ROOT_UID &&
1130 buf->st_uid != current_uid) {
1131 if (mode_flag & DETAIL) {
1132 PRINT_FILE_NAME(file);
1133 STATISTIC_ERR_MSG(
1134 "File is not current user's file"
1135 " or current user is not root");
1136 }
1137 return 0;
1138 }
1139
1140 /* Empty file */
1141 if (buf->st_size == 0) {
1142 if (mode_flag & DETAIL) {
1143 PRINT_FILE_NAME(file);
1144 STATISTIC_ERR_MSG("File size is 0");
1145 }
1146 return 0;
1147 }
1148
1149 /* Has no blocks */
1150 if (buf->st_blocks == 0) {
1151 if (mode_flag & DETAIL) {
1152 PRINT_FILE_NAME(file);
1153 STATISTIC_ERR_MSG("File has no blocks");
1154 }
1155 return 0;
1156 }
1157
1158 fd = open64(file, O_RDONLY);
1159 if (fd < 0) {
1160 if (mode_flag & DETAIL) {
1161 PRINT_FILE_NAME(file);
1162 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN);
1163 }
1164 return 0;
1165 }
1166
1167 /* Get file's physical extents */
1168 ret = get_file_extents(fd, &physical_list_head);
1169 if (ret < 0) {
1170 if (mode_flag & DETAIL) {
1171 PRINT_FILE_NAME(file);
1172 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1173 }
1174 goto out;
1175 }
1176
1177 /* Get the count of file's continuous physical region */
1178 physical_ext_count = get_physical_count(physical_list_head);
1179
1180 /* Change list from physical to logical */
1181 ret = change_physical_to_logical(&physical_list_head,
1182 &logical_list_head);
1183 if (ret < 0) {
1184 if (mode_flag & DETAIL) {
1185 PRINT_FILE_NAME(file);
1186 STATISTIC_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1187 }
1188 goto out;
1189 }
1190
1191 /* Count file fragments before defrag */
1192 now_ext_count = get_logical_count(logical_list_head);
1193
1194 if (current_uid == ROOT_UID) {
1195 /* Calculate the size per extent */
1196 blk_count = get_file_blocks(logical_list_head);
1197
1198 best_ext_count = get_best_count(blk_count);
1199
1200 /* e4defrag rounds size_per_ext up to a block size boundary */
1201 size_per_ext = blk_count * (buf->st_blksize / 1024) /
1202 now_ext_count;
1203
1204 ratio = (float)(physical_ext_count - best_ext_count) * 100 /
1205 blk_count;
1206
1207 extents_before_defrag += now_ext_count;
1208 extents_after_defrag += best_ext_count;
1209 files_block_count += blk_count;
1210 }
1211
1212 if (total_count == 1 && regular_count == 1) {
1213 /* File only */
1214 if (mode_flag & DETAIL) {
1215 int count = 0;
1216 struct fiemap_extent_list *ext_list_tmp =
1217 logical_list_head;
1218
1219 /* Print extents info */
1220 do {
1221 count++;
1222 printf("[ext %d]:\tstart %llu:\tlogical "
1223 "%llu:\tlen %llu\n", count,
1224 ext_list_tmp->data.physical,
1225 ext_list_tmp->data.logical,
1226 ext_list_tmp->data.len);
1227 ext_list_tmp = ext_list_tmp->next;
1228 } while (ext_list_tmp != logical_list_head);
1229
1230 } else {
1231 printf("%-40s%10s/%-10s%9s\n",
1232 "<File>", "now", "best", "size/ext");
1233 if (current_uid == ROOT_UID) {
1234 if (strlen(file) > 40)
1235 printf("%s\n%50d/%-10d%6llu KB\n",
1236 file, now_ext_count,
1237 best_ext_count, size_per_ext);
1238 else
1239 printf("%-40s%10d/%-10d%6llu KB\n",
1240 file, now_ext_count,
1241 best_ext_count, size_per_ext);
1242 } else {
1243 if (strlen(file) > 40)
1244 printf("%s\n%50d/%-10s%7s\n",
1245 file, now_ext_count,
1246 "-", "-");
1247 else
1248 printf("%-40s%10d/%-10s%7s\n",
1249 file, now_ext_count,
1250 "-", "-");
1251 }
1252 }
1253 succeed_cnt++;
1254 goto out;
1255 }
1256
1257 if (mode_flag & DETAIL) {
1258 /* Print statistic info */
1259 sprintf(msg_buffer, "[%u/%u]%s",
1260 defraged_file_count, total_count, file);
1261 if (current_uid == ROOT_UID) {
1262 if (strlen(msg_buffer) > 40)
1263 printf("\033[79;0H\033[K%s\n"
1264 "%50d/%-10d%6llu KB\n",
1265 msg_buffer, now_ext_count,
1266 best_ext_count, size_per_ext);
1267 else
1268 printf("\033[79;0H\033[K%-40s"
1269 "%10d/%-10d%6llu KB\n",
1270 msg_buffer, now_ext_count,
1271 best_ext_count, size_per_ext);
1272 } else {
1273 if (strlen(msg_buffer) > 40)
1274 printf("\033[79;0H\033[K%s\n%50d/%-10s%7s\n",
1275 msg_buffer, now_ext_count,
1276 "-", "-");
1277 else
1278 printf("\033[79;0H\033[K%-40s%10d/%-10s%7s\n",
1279 msg_buffer, now_ext_count,
1280 "-", "-");
1281 }
1282 }
1283
1284 for (i = 0; i < SHOW_FRAG_FILES; i++) {
1285 if (ratio >= frag_rank[i].ratio) {
1286 for (j = SHOW_FRAG_FILES - 1; j > i; j--) {
1287 memset(&frag_rank[j], 0,
1288 sizeof(struct frag_statistic_ino));
1289 strncpy(frag_rank[j].msg_buffer,
1290 frag_rank[j - 1].msg_buffer,
1291 strnlen(frag_rank[j - 1].msg_buffer,
1292 PATH_MAX));
1293 frag_rank[j].now_count =
1294 frag_rank[j - 1].now_count;
1295 frag_rank[j].best_count =
1296 frag_rank[j - 1].best_count;
1297 frag_rank[j].size_per_ext =
1298 frag_rank[j - 1].size_per_ext;
1299 frag_rank[j].ratio =
1300 frag_rank[j - 1].ratio;
1301 }
1302 memset(&frag_rank[i], 0,
1303 sizeof(struct frag_statistic_ino));
1304 strncpy(frag_rank[i].msg_buffer, file,
1305 strnlen(file, PATH_MAX));
1306 frag_rank[i].now_count = now_ext_count;
1307 frag_rank[i].best_count = best_ext_count;
1308 frag_rank[i].size_per_ext = size_per_ext;
1309 frag_rank[i].ratio = ratio;
1310 break;
1311 }
1312 }
1313
1314 succeed_cnt++;
1315
1316 out:
1317 close(fd);
1318 free_ext(physical_list_head);
1319 free_ext(logical_list_head);
1320 return 0;
1321 }
1322
1323 /*
1324 * print_progress - Print defrag progress
1325 *
1326 * @file: file name.
1327 * @start: logical offset for defrag target file
1328 * @file_size: defrag target filesize
1329 */
1330 static void print_progress(const char *file, loff_t start, loff_t file_size)
1331 {
1332 int percent = (start * 100) / file_size;
1333 printf("\033[79;0H\033[K[%u/%u]%s:\t%3d%%",
1334 defraged_file_count, total_count, file, min(percent, 100));
1335 fflush(stdout);
1336
1337 return;
1338 }
1339
1340 /*
1341 * call_defrag() - Execute the defrag program.
1342 *
1343 * @fd: target file descriptor.
1344 * @donor_fd: donor file descriptor.
1345 * @file: target file name.
1346 * @buf: pointer of the struct stat64.
1347 * @ext_list_head: head of the extent list.
1348 */
1349 static int call_defrag(int fd, int donor_fd, const char *file,
1350 const struct stat64 *buf, struct fiemap_extent_list *ext_list_head)
1351 {
1352 loff_t start = 0;
1353 unsigned int page_num;
1354 unsigned char *vec = NULL;
1355 int defraged_ret = 0;
1356 int ret;
1357 struct move_extent move_data;
1358 struct fiemap_extent_list *ext_list_tmp = NULL;
1359
1360 memset(&move_data, 0, sizeof(struct move_extent));
1361 move_data.donor_fd = donor_fd;
1362
1363 /* Print defrag progress */
1364 print_progress(file, start, buf->st_size);
1365
1366 ext_list_tmp = ext_list_head;
1367 do {
1368 move_data.orig_start = ext_list_tmp->data.logical;
1369 /* Logical offset of orig and donor should be same */
1370 move_data.donor_start = move_data.orig_start;
1371 move_data.len = ext_list_tmp->data.len;
1372 move_data.moved_len = 0;
1373
1374 ret = page_in_core(fd, move_data, &vec, &page_num);
1375 if (ret < 0) {
1376 if (mode_flag & DETAIL) {
1377 printf("\n");
1378 PRINT_ERR_MSG_WITH_ERRNO(
1379 "Failed to get file map");
1380 } else {
1381 printf("\t[ NG ]\n");
1382 }
1383 return -1;
1384 }
1385
1386 /* EXT4_IOC_MOVE_EXT */
1387 defraged_ret =
1388 ioctl(fd, EXT4_IOC_MOVE_EXT, &move_data);
1389
1390 /* Free pages */
1391 ret = defrag_fadvise(fd, move_data, vec, page_num);
1392 if (vec) {
1393 free(vec);
1394 vec = NULL;
1395 }
1396 if (ret < 0) {
1397 if (mode_flag & DETAIL) {
1398 printf("\n");
1399 PRINT_ERR_MSG_WITH_ERRNO(
1400 "Failed to free page");
1401 } else {
1402 printf("\t[ NG ]\n");
1403 }
1404 return -1;
1405 }
1406
1407 if (defraged_ret < 0) {
1408 if (mode_flag & DETAIL) {
1409 printf("\n");
1410 PRINT_ERR_MSG_WITH_ERRNO(
1411 "Failed to defrag with "
1412 "EXT4_IOC_MOVE_EXT ioctl");
1413 if (errno == ENOTTY)
1414 printf("\tAt least 2.6.31-rc1 of "
1415 "vanilla kernel is required\n");
1416 } else {
1417 printf("\t[ NG ]\n");
1418 }
1419 return -1;
1420 }
1421 /* Adjust logical offset for next ioctl */
1422 move_data.orig_start += move_data.moved_len;
1423 move_data.donor_start = move_data.orig_start;
1424
1425 start = move_data.orig_start * buf->st_blksize;
1426
1427 /* Print defrag progress */
1428 print_progress(file, start, buf->st_size);
1429
1430 /* End of file */
1431 if (start >= buf->st_size)
1432 break;
1433
1434 ext_list_tmp = ext_list_tmp->next;
1435 } while (ext_list_tmp != ext_list_head);
1436
1437 return 0;
1438 }
1439
1440 /*
1441 * file_defrag() - Check file attributes and call ioctl to defrag.
1442 *
1443 * @file: the file's name.
1444 * @buf: the pointer of the struct stat64.
1445 * @flag: file type.
1446 * @ftwbuf: the pointer of a struct FTW.
1447 */
1448 static int file_defrag(const char *file, const struct stat64 *buf,
1449 int flag EXT2FS_ATTR((unused)),
1450 struct FTW *ftwbuf EXT2FS_ATTR((unused)))
1451 {
1452 int fd;
1453 int donor_fd = -1;
1454 int ret;
1455 int best;
1456 int file_frags_start, file_frags_end;
1457 int orig_physical_cnt, donor_physical_cnt = 0;
1458 char tmp_inode_name[PATH_MAX + 8];
1459 ext4_fsblk_t blk_count = 0;
1460 struct fiemap_extent_list *orig_list_physical = NULL;
1461 struct fiemap_extent_list *orig_list_logical = NULL;
1462 struct fiemap_extent_list *donor_list_physical = NULL;
1463 struct fiemap_extent_list *donor_list_logical = NULL;
1464 struct fiemap_extent_group *orig_group_head = NULL;
1465 struct fiemap_extent_group *orig_group_tmp = NULL;
1466
1467 defraged_file_count++;
1468
1469 if (mode_flag & DETAIL) {
1470 printf("[%u/%u]", defraged_file_count, total_count);
1471 fflush(stdout);
1472 }
1473
1474 if (lost_found_dir[0] != '\0' &&
1475 !memcmp(file, lost_found_dir, strnlen(lost_found_dir, PATH_MAX))) {
1476 if (mode_flag & DETAIL) {
1477 PRINT_FILE_NAME(file);
1478 IN_FTW_PRINT_ERR_MSG(NGMSG_LOST_FOUND);
1479 }
1480 return 0;
1481 }
1482
1483 if (!S_ISREG(buf->st_mode)) {
1484 if (mode_flag & DETAIL) {
1485 PRINT_FILE_NAME(file);
1486 IN_FTW_PRINT_ERR_MSG(NGMSG_FILE_UNREG);
1487 }
1488 return 0;
1489 }
1490
1491 /* Empty file */
1492 if (buf->st_size == 0) {
1493 if (mode_flag & DETAIL) {
1494 PRINT_FILE_NAME(file);
1495 IN_FTW_PRINT_ERR_MSG("File size is 0");
1496 }
1497 return 0;
1498 }
1499
1500 /* Has no blocks */
1501 if (buf->st_blocks == 0) {
1502 if (mode_flag & DETAIL) {
1503 PRINT_FILE_NAME(file);
1504 STATISTIC_ERR_MSG("File has no blocks");
1505 }
1506 return 0;
1507 }
1508
1509 fd = open64(file, O_RDWR);
1510 if (fd < 0) {
1511 if (mode_flag & DETAIL) {
1512 PRINT_FILE_NAME(file);
1513 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN);
1514 }
1515 return 0;
1516 }
1517
1518 /* Get file's extents */
1519 ret = get_file_extents(fd, &orig_list_physical);
1520 if (ret < 0) {
1521 if (mode_flag & DETAIL) {
1522 PRINT_FILE_NAME(file);
1523 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1524 }
1525 goto out;
1526 }
1527
1528 /* Get the count of file's continuous physical region */
1529 orig_physical_cnt = get_physical_count(orig_list_physical);
1530
1531 /* Change list from physical to logical */
1532 ret = change_physical_to_logical(&orig_list_physical,
1533 &orig_list_logical);
1534 if (ret < 0) {
1535 if (mode_flag & DETAIL) {
1536 PRINT_FILE_NAME(file);
1537 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1538 }
1539 goto out;
1540 }
1541
1542 /* Count file fragments before defrag */
1543 file_frags_start = get_logical_count(orig_list_logical);
1544
1545 blk_count = get_file_blocks(orig_list_logical);
1546 if (file_check(fd, buf, file, file_frags_start, blk_count) < 0)
1547 goto out;
1548
1549 if (fsync(fd) < 0) {
1550 if (mode_flag & DETAIL) {
1551 PRINT_FILE_NAME(file);
1552 PRINT_ERR_MSG_WITH_ERRNO("Failed to sync(fsync)");
1553 }
1554 goto out;
1555 }
1556
1557 if (current_uid == ROOT_UID)
1558 best = get_best_count(blk_count);
1559 else
1560 best = 1;
1561
1562 if (file_frags_start <= best)
1563 goto check_improvement;
1564
1565 /* Combine extents to group */
1566 ret = join_extents(orig_list_logical, &orig_group_head);
1567 if (ret < 0) {
1568 if (mode_flag & DETAIL) {
1569 PRINT_FILE_NAME(file);
1570 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1571 }
1572 goto out;
1573 }
1574
1575 /* Create donor inode */
1576 memset(tmp_inode_name, 0, PATH_MAX + 8);
1577 sprintf(tmp_inode_name, "%.*s.defrag",
1578 (int)strnlen(file, PATH_MAX), file);
1579 donor_fd = open64(tmp_inode_name, O_WRONLY | O_CREAT | O_EXCL, S_IRUSR);
1580 if (donor_fd < 0) {
1581 if (mode_flag & DETAIL) {
1582 PRINT_FILE_NAME(file);
1583 if (errno == EEXIST)
1584 PRINT_ERR_MSG_WITH_ERRNO(
1585 "File is being defraged by other program");
1586 else
1587 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_OPEN);
1588 }
1589 goto out;
1590 }
1591
1592 /* Unlink donor inode */
1593 ret = unlink(tmp_inode_name);
1594 if (ret < 0) {
1595 if (mode_flag & DETAIL) {
1596 PRINT_FILE_NAME(file);
1597 PRINT_ERR_MSG_WITH_ERRNO("Failed to unlink");
1598 }
1599 goto out;
1600 }
1601
1602 /* Allocate space for donor inode */
1603 orig_group_tmp = orig_group_head;
1604 do {
1605 ret = fallocate64(donor_fd, 0,
1606 (loff_t)orig_group_tmp->start->data.logical * block_size,
1607 (loff_t)orig_group_tmp->len * block_size);
1608 if (ret < 0) {
1609 if (mode_flag & DETAIL) {
1610 PRINT_FILE_NAME(file);
1611 PRINT_ERR_MSG_WITH_ERRNO("Failed to fallocate");
1612 }
1613 goto out;
1614 }
1615
1616 orig_group_tmp = orig_group_tmp->next;
1617 } while (orig_group_tmp != orig_group_head);
1618
1619 /* Get donor inode's extents */
1620 ret = get_file_extents(donor_fd, &donor_list_physical);
1621 if (ret < 0) {
1622 if (mode_flag & DETAIL) {
1623 PRINT_FILE_NAME(file);
1624 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1625 }
1626 goto out;
1627 }
1628
1629 /* Calcuate donor inode's continuous physical region */
1630 donor_physical_cnt = get_physical_count(donor_list_physical);
1631
1632 /* Change donor extent list from physical to logical */
1633 ret = change_physical_to_logical(&donor_list_physical,
1634 &donor_list_logical);
1635 if (ret < 0) {
1636 if (mode_flag & DETAIL) {
1637 PRINT_FILE_NAME(file);
1638 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_EXTENT);
1639 }
1640 goto out;
1641 }
1642
1643 check_improvement:
1644 if (mode_flag & DETAIL) {
1645 if (file_frags_start != 1)
1646 frag_files_before_defrag++;
1647
1648 extents_before_defrag += file_frags_start;
1649 }
1650
1651 if (file_frags_start <= best ||
1652 orig_physical_cnt <= donor_physical_cnt) {
1653 printf("\033[79;0H\033[K[%u/%u]%s:\t%3d%%",
1654 defraged_file_count, total_count, file, 100);
1655 if (mode_flag & DETAIL)
1656 printf(" extents: %d -> %d",
1657 file_frags_start, file_frags_start);
1658
1659 printf("\t[ OK ]\n");
1660 succeed_cnt++;
1661
1662 if (file_frags_start != 1)
1663 frag_files_after_defrag++;
1664
1665 extents_after_defrag += file_frags_start;
1666 goto out;
1667 }
1668
1669 /* Defrag the file */
1670 ret = call_defrag(fd, donor_fd, file, buf, donor_list_logical);
1671
1672 /* Count file fragments after defrag and print extents info */
1673 if (mode_flag & DETAIL) {
1674 file_frags_end = file_frag_count(fd);
1675 if (file_frags_end < 0) {
1676 printf("\n");
1677 PRINT_ERR_MSG_WITH_ERRNO(NGMSG_FILE_INFO);
1678 goto out;
1679 }
1680
1681 if (file_frags_end != 1)
1682 frag_files_after_defrag++;
1683
1684 extents_after_defrag += file_frags_end;
1685
1686 if (ret < 0)
1687 goto out;
1688
1689 printf(" extents: %d -> %d",
1690 file_frags_start, file_frags_end);
1691 fflush(stdout);
1692 }
1693
1694 if (ret < 0)
1695 goto out;
1696
1697 printf("\t[ OK ]\n");
1698 fflush(stdout);
1699 succeed_cnt++;
1700
1701 out:
1702 close(fd);
1703 if (donor_fd != -1)
1704 close(donor_fd);
1705 free_ext(orig_list_physical);
1706 free_ext(orig_list_logical);
1707 free_ext(donor_list_physical);
1708 free_exts_group(orig_group_head);
1709 return 0;
1710 }
1711
1712 /*
1713 * main() - Ext4 online defrag.
1714 *
1715 * @argc: the number of parameter.
1716 * @argv[]: the pointer array of parameter.
1717 */
1718 int main(int argc, char *argv[])
1719 {
1720 int opt;
1721 int i, j, ret = 0;
1722 int flags = FTW_PHYS | FTW_MOUNT;
1723 int arg_type = -1;
1724 int success_flag = 0;
1725 char dir_name[PATH_MAX + 1];
1726 char dev_name[PATH_MAX + 1];
1727 struct stat64 buf;
1728 ext2_filsys fs = NULL;
1729
1730 /* Parse arguments */
1731 if (argc == 1)
1732 goto out;
1733
1734 while ((opt = getopt(argc, argv, "vc")) != EOF) {
1735 switch (opt) {
1736 case 'v':
1737 mode_flag |= DETAIL;
1738 break;
1739 case 'c':
1740 mode_flag |= STATISTIC;
1741 break;
1742 default:
1743 goto out;
1744 }
1745 }
1746
1747 if (argc == optind)
1748 goto out;
1749
1750 current_uid = getuid();
1751
1752 /* Main process */
1753 for (i = optind; i < argc; i++) {
1754 succeed_cnt = 0;
1755 regular_count = 0;
1756 total_count = 0;
1757 frag_files_before_defrag = 0;
1758 frag_files_after_defrag = 0;
1759 extents_before_defrag = 0;
1760 extents_after_defrag = 0;
1761 defraged_file_count = 0;
1762 files_block_count = 0;
1763 blocks_per_group = 0;
1764 feature_incompat = 0;
1765 log_groups_per_flex = 0;
1766
1767 memset(dir_name, 0, PATH_MAX + 1);
1768 memset(dev_name, 0, PATH_MAX + 1);
1769 memset(lost_found_dir, 0, PATH_MAX + 1);
1770 memset(frag_rank, 0,
1771 sizeof(struct frag_statistic_ino) * SHOW_FRAG_FILES);
1772
1773 if ((mode_flag & STATISTIC) && i > optind)
1774 printf("\n");
1775
1776 #if BYTE_ORDER != BIG_ENDIAN && BYTE_ORDER != LITTLE_ENDIAN
1777 PRINT_ERR_MSG("Endian's type is not big/little endian");
1778 PRINT_FILE_NAME(argv[i]);
1779 continue;
1780 #endif
1781
1782 if (lstat64(argv[i], &buf) < 0) {
1783 perror(NGMSG_FILE_INFO);
1784 PRINT_FILE_NAME(argv[i]);
1785 continue;
1786 }
1787
1788 /* Handle i.e. lvm device symlinks */
1789 if (S_ISLNK(buf.st_mode)) {
1790 struct stat64 buf2;
1791
1792 if (stat64(argv[i], &buf2) == 0 &&
1793 S_ISBLK(buf2.st_mode))
1794 buf = buf2;
1795 }
1796
1797 if (S_ISBLK(buf.st_mode)) {
1798 /* Block device */
1799 strncpy(dev_name, argv[i], strnlen(argv[i], PATH_MAX));
1800 if (get_mount_point(argv[i], dir_name, PATH_MAX) < 0)
1801 continue;
1802 if (lstat64(dir_name, &buf) < 0) {
1803 perror(NGMSG_FILE_INFO);
1804 PRINT_FILE_NAME(argv[i]);
1805 continue;
1806 }
1807 arg_type = DEVNAME;
1808 if (!(mode_flag & STATISTIC))
1809 printf("ext4 defragmentation for device(%s)\n",
1810 argv[i]);
1811 } else if (S_ISDIR(buf.st_mode)) {
1812 /* Directory */
1813 if (access(argv[i], R_OK) < 0) {
1814 perror(argv[i]);
1815 continue;
1816 }
1817 arg_type = DIRNAME;
1818 strncpy(dir_name, argv[i], strnlen(argv[i], PATH_MAX));
1819 } else if (S_ISREG(buf.st_mode)) {
1820 /* Regular file */
1821 arg_type = FILENAME;
1822 } else {
1823 /* Irregular file */
1824 PRINT_ERR_MSG(NGMSG_FILE_UNREG);
1825 PRINT_FILE_NAME(argv[i]);
1826 continue;
1827 }
1828
1829 /* Set blocksize */
1830 block_size = buf.st_blksize;
1831
1832 /* For device case,
1833 * filesystem type checked in get_mount_point()
1834 */
1835 if (arg_type == FILENAME || arg_type == DIRNAME) {
1836 if (is_ext4(argv[i], dev_name) < 0)
1837 continue;
1838 if (realpath(argv[i], dir_name) == NULL) {
1839 perror("Couldn't get full path");
1840 PRINT_FILE_NAME(argv[i]);
1841 continue;
1842 }
1843 }
1844
1845 if (current_uid == ROOT_UID) {
1846 /* Get super block info */
1847 ret = ext2fs_open(dev_name, 0, 0, block_size,
1848 unix_io_manager, &fs);
1849 if (ret) {
1850 if (mode_flag & DETAIL) {
1851 perror("Can't get super block info");
1852 PRINT_FILE_NAME(argv[i]);
1853 }
1854 continue;
1855 }
1856
1857 blocks_per_group = fs->super->s_blocks_per_group;
1858 feature_incompat = fs->super->s_feature_incompat;
1859 log_groups_per_flex = fs->super->s_log_groups_per_flex;
1860
1861 ext2fs_close(fs);
1862 }
1863
1864 switch (arg_type) {
1865 case DIRNAME:
1866 if (!(mode_flag & STATISTIC))
1867 printf("ext4 defragmentation "
1868 "for directory(%s)\n", argv[i]);
1869
1870 int mount_dir_len = 0;
1871 mount_dir_len = strnlen(lost_found_dir, PATH_MAX);
1872
1873 strncat(lost_found_dir, "/lost+found",
1874 PATH_MAX - strnlen(lost_found_dir, PATH_MAX));
1875
1876 /* Not the case("e4defrag mount_piont_dir") */
1877 if (dir_name[mount_dir_len] != '\0') {
1878 /*
1879 * "e4defrag mount_piont_dir/lost+found"
1880 * or "e4defrag mount_piont_dir/lost+found/"
1881 */
1882 if (strncmp(lost_found_dir, dir_name,
1883 strnlen(lost_found_dir,
1884 PATH_MAX)) == 0 &&
1885 (dir_name[strnlen(lost_found_dir,
1886 PATH_MAX)] == '\0' ||
1887 dir_name[strnlen(lost_found_dir,
1888 PATH_MAX)] == '/')) {
1889 PRINT_ERR_MSG(NGMSG_LOST_FOUND);
1890 PRINT_FILE_NAME(argv[i]);
1891 continue;
1892 }
1893
1894 /* "e4defrag mount_piont_dir/else_dir" */
1895 memset(lost_found_dir, 0, PATH_MAX + 1);
1896 }
1897 case DEVNAME:
1898 if (arg_type == DEVNAME) {
1899 strncpy(lost_found_dir, dir_name,
1900 strnlen(dir_name, PATH_MAX));
1901 strncat(lost_found_dir, "/lost+found/",
1902 PATH_MAX - strnlen(lost_found_dir,
1903 PATH_MAX));
1904 }
1905
1906 nftw64(dir_name, calc_entry_counts, FTW_OPEN_FD, flags);
1907
1908 if (mode_flag & STATISTIC) {
1909 if (mode_flag & DETAIL)
1910 printf("%-40s%10s/%-10s%9s\n",
1911 "<File>", "now", "best", "size/ext");
1912
1913 if (!(mode_flag & DETAIL) &&
1914 current_uid != ROOT_UID) {
1915 printf(" Done.\n");
1916 success_flag = 1;
1917 continue;
1918 }
1919
1920 nftw64(dir_name, file_statistic,
1921 FTW_OPEN_FD, flags);
1922
1923 if (succeed_cnt != 0 &&
1924 current_uid == ROOT_UID) {
1925 if (mode_flag & DETAIL)
1926 printf("\n");
1927 printf("%-40s%10s/%-10s%9s\n",
1928 "<Fragmented files>", "now",
1929 "best", "size/ext");
1930 for (j = 0; j < SHOW_FRAG_FILES; j++) {
1931 if (strlen(frag_rank[j].
1932 msg_buffer) > 37) {
1933 printf("%d. %s\n%50d/"
1934 "%-10d%6llu KB\n",
1935 j + 1,
1936 frag_rank[j].msg_buffer,
1937 frag_rank[j].now_count,
1938 frag_rank[j].best_count,
1939 frag_rank[j].
1940 size_per_ext);
1941 } else if (strlen(frag_rank[j].
1942 msg_buffer) > 0) {
1943 printf("%d. %-37s%10d/"
1944 "%-10d%6llu KB\n",
1945 j + 1,
1946 frag_rank[j].msg_buffer,
1947 frag_rank[j].now_count,
1948 frag_rank[j].best_count,
1949 frag_rank[j].
1950 size_per_ext);
1951 } else
1952 break;
1953 }
1954 }
1955 break;
1956 }
1957 /* File tree walk */
1958 nftw64(dir_name, file_defrag, FTW_OPEN_FD, flags);
1959 printf("\n\tSuccess:\t\t\t[ %u/%u ]\n", succeed_cnt,
1960 total_count);
1961 printf("\tFailure:\t\t\t[ %u/%u ]\n",
1962 total_count - succeed_cnt, total_count);
1963 if (mode_flag & DETAIL) {
1964 printf("\tTotal extents:\t\t\t%4d->%d\n",
1965 extents_before_defrag,
1966 extents_after_defrag);
1967 printf("\tFragmented percentage:\t\t"
1968 "%3llu%%->%llu%%\n",
1969 !regular_count ? 0 :
1970 ((unsigned long long)
1971 frag_files_before_defrag * 100) /
1972 regular_count,
1973 !regular_count ? 0 :
1974 ((unsigned long long)
1975 frag_files_after_defrag * 100) /
1976 regular_count);
1977 }
1978 break;
1979 case FILENAME:
1980 total_count = 1;
1981 regular_count = 1;
1982 strncat(lost_found_dir, "/lost+found/",
1983 PATH_MAX - strnlen(lost_found_dir,
1984 PATH_MAX));
1985 if (strncmp(lost_found_dir, dir_name,
1986 strnlen(lost_found_dir,
1987 PATH_MAX)) == 0) {
1988 PRINT_ERR_MSG(NGMSG_LOST_FOUND);
1989 PRINT_FILE_NAME(argv[i]);
1990 continue;
1991 }
1992
1993 if (mode_flag & STATISTIC) {
1994 file_statistic(argv[i], &buf, FTW_F, NULL);
1995 break;
1996 } else
1997 printf("ext4 defragmentation for %s\n",
1998 argv[i]);
1999 /* Defrag single file process */
2000 file_defrag(argv[i], &buf, FTW_F, NULL);
2001 if (succeed_cnt != 0)
2002 printf(" Success:\t\t\t[1/1]\n");
2003 else
2004 printf(" Success:\t\t\t[0/1]\n");
2005
2006 break;
2007 }
2008
2009 if (succeed_cnt != 0)
2010 success_flag = 1;
2011 if (mode_flag & STATISTIC) {
2012 if (current_uid != ROOT_UID) {
2013 printf(" Done.\n");
2014 continue;
2015 }
2016
2017 if (!succeed_cnt) {
2018 if (mode_flag & DETAIL)
2019 printf("\n");
2020
2021 if (arg_type == DEVNAME)
2022 printf(" In this device(%s), "
2023 "none can be defragmented.\n", argv[i]);
2024 else if (arg_type == DIRNAME)
2025 printf(" In this directory(%s), "
2026 "none can be defragmented.\n", argv[i]);
2027 else
2028 printf(" This file(%s) "
2029 "can't be defragmented.\n", argv[i]);
2030 } else {
2031 float files_ratio = 0.0;
2032 float score = 0.0;
2033 __u64 size_per_ext = files_block_count *
2034 (buf.st_blksize / 1024) /
2035 extents_before_defrag;
2036 files_ratio = (float)(extents_before_defrag -
2037 extents_after_defrag) *
2038 100 / files_block_count;
2039 score = CALC_SCORE(files_ratio);
2040 printf("\n Total/best extents\t\t\t\t%d/%d\n"
2041 " Average size per extent"
2042 "\t\t\t%llu KB\n"
2043 " Fragmentation score\t\t\t\t%.0f\n",
2044 extents_before_defrag,
2045 extents_after_defrag,
2046 size_per_ext, score);
2047 printf(" [0-30 no problem:"
2048 " 31-55 a little bit fragmented:"
2049 " 56- needs defrag]\n");
2050
2051 if (arg_type == DEVNAME)
2052 printf(" This device (%s) ", argv[i]);
2053 else if (arg_type == DIRNAME)
2054 printf(" This directory (%s) ",
2055 argv[i]);
2056 else
2057 printf(" This file (%s) ", argv[i]);
2058
2059 if (score > BOUND_SCORE)
2060 printf("needs defragmentation.\n");
2061 else
2062 printf("does not need "
2063 "defragmentation.\n");
2064 }
2065 printf(" Done.\n");
2066 }
2067
2068 }
2069
2070 if (success_flag)
2071 return 0;
2072
2073 exit(1);
2074
2075 out:
2076 printf(MSG_USAGE);
2077 exit(1);
2078 }
2079