1 /* du -- summarize disk usage
2 Copyright (C) 1988, 1989, 1990, 1991, 1995 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
18 /* Differences from the Unix du:
19 * Doesn't simply ignore the names of regular files given as arguments
22 -l Count the size of all files, even if they have appeared
23 already in another hard link.
24 -x Do not cross file-system boundaries during the recursion.
25 -c Write a grand total of all of the arguments after all
26 arguments have been processed. This can be used to find
27 out the disk usage of a directory, with some files excluded.
28 -h Print sizes in human readable format (1k 234M 2G, etc).
29 -k Print sizes in kilobytes instead of 512 byte blocks
30 (the default required by POSIX).
31 -m Print sizes in megabytes instead of 512 byte blocks
32 -b Print sizes in bytes.
33 -S Count the size of each directory separately, not including
34 the sizes of subdirectories.
35 -D Dereference only symbolic links given on the command line.
36 -L Dereference all symbolic links.
38 By tege@sics.se, Torbjorn Granlund,
39 and djm@ai.mit.edu, David MacKenzie.
40 Variable blocks added by lm@sgi.com.
50 #include <sys/types.h>
59 #define convert_blocks(b, size) (size == size_kilobytes ? ((b) + 1) / 2 : \
60 size == size_megabytes ? ((b) + 1024) / 2048 : (b))
62 /* Initial number of entries in each hash table entry's table of inodes. */
63 #define INITIAL_HASH_MODULE 100
65 /* Initial number of entries in the inode hash table. */
66 #define INITIAL_ENTRY_TAB_SIZE 70
68 /* Initial size to allocate for `path'. */
69 #define INITIAL_PATH_SIZE 100
71 /* The maximum length of a human-readable string. Be pessimistic
72 and assume `int' is 64-bits wide. Converting 2^63 - 1 gives the
73 11-character string, 8589934592G. */
74 #define LONGEST_HUMAN_READABLE 11
76 /* Hash structure for inode and device numbers. The separate entry
77 structure makes it easier to rehash "in place". */
83 struct entry
*coll_link
;
86 /* Structure for a hash table for inode numbers. */
90 unsigned modulus
; /* Size of the `hash' pointer vector. */
91 struct entry
*entry_tab
; /* Pointer to dynamically growing vector. */
92 unsigned entry_tab_size
; /* Size of current `entry_tab' allocation. */
93 unsigned first_free_entry
; /* Index in `entry_tab'. */
94 struct entry
*hash
[1]; /* Vector of pointers in `entry_tab'. */
98 /* Structure for dynamically resizable strings. */
102 unsigned alloc
; /* Size of allocation for the text. */
103 unsigned length
; /* Length of the text currently. */
104 char *text
; /* Pointer to the text. */
105 } *string
, stringstruct
;
114 static int hash_insert
__P ((ino_t ino
, dev_t dev
));
115 static int hash_insert2
__P ((struct htab
*htab
, ino_t ino
, dev_t dev
));
116 static long count_entry
__P ((char *ent
, int top
, dev_t last_dev
));
117 static void du_files
__P ((char **files
));
118 static void hash_init
__P ((unsigned int modulus
,
119 unsigned int entry_tab_size
));
120 static void hash_reset
__P ((void));
121 static void str_concatc
__P ((string s1
, char *cstr
));
122 static void str_copyc
__P ((string s1
, char *cstr
));
123 static void str_init
__P ((string
*s1
, unsigned int size
));
124 static void str_trunc
__P ((string s1
, unsigned int length
));
126 /* Name under which this program was invoked. */
129 /* If nonzero, display only a total for each argument. */
130 static int opt_summarize_only
= 0;
132 /* If nonzero, display counts for all files, not just directories. */
133 static int opt_all
= 0;
135 /* If nonzero, count each hard link of files with multiple links. */
136 static int opt_count_all
= 0;
138 /* If nonzero, do not cross file-system boundaries. */
139 static int opt_one_file_system
= 0;
141 /* If nonzero, print a grand total at the end. */
142 static int opt_combined_arguments
= 0;
144 /* If nonzero, do not add sizes of subdirectories. */
145 static int opt_separate_dirs
= 0;
147 /* If nonzero, dereference symlinks that are command line arguments. */
148 static int opt_dereference_arguments
= 0;
152 size_blocks
, /* 512-byte blocks. */
153 size_kilobytes
, /* 1K blocks. */
154 size_megabytes
, /* 1024K blocks. */
155 size_bytes
/* 1-byte blocks. */
158 /* human style output */
159 static int opt_human_readable
;
161 /* The units to count in. */
162 static enum output_size output_size
;
164 /* Accumulated path for file or directory being processed. */
167 /* Pointer to hash structure, used by the hash routines. */
168 static struct htab
*htab
;
170 /* Globally used stat buffer. */
171 static struct stat stat_buf
;
173 /* A pointer to either lstat or stat, depending on whether
174 dereferencing of all symbolic links is to be done. */
175 static int (*xstat
) ();
177 /* The exit status to use if we don't get any fatal errors. */
178 static int exit_status
;
180 /* If nonzero, display usage information and exit. */
181 static int show_help
;
183 /* If nonzero, print the version on standard output and exit. */
184 static int show_version
;
186 /* Grand total size of all args. */
187 static long tot_size
= 0L;
189 static struct option
const long_options
[] =
191 {"all", no_argument
, &opt_all
, 1},
192 {"bytes", no_argument
, NULL
, 'b'},
193 {"count-links", no_argument
, &opt_count_all
, 1},
194 {"dereference", no_argument
, NULL
, 'L'},
195 {"dereference-args", no_argument
, &opt_dereference_arguments
, 1},
196 {"human-readable", no_argument
, NULL
, 'h'},
197 {"kilobytes", no_argument
, NULL
, 'k'},
198 {"megabytes", no_argument
, NULL
, 'm'},
199 {"one-file-system", no_argument
, &opt_one_file_system
, 1},
200 {"separate-dirs", no_argument
, &opt_separate_dirs
, 1},
201 {"summarize", no_argument
, &opt_summarize_only
, 1},
202 {"total", no_argument
, &opt_combined_arguments
, 1},
203 {"help", no_argument
, &show_help
, 1},
204 {"version", no_argument
, &show_version
, 1},
209 usage (int status
, char *reason
)
212 fprintf (status
== 0 ? stdout
: stderr
, "%s: %s\n",
213 program_name
, reason
);
216 fprintf (stderr
, _("Try `%s --help' for more information.\n"),
220 printf (_("Usage: %s [OPTION]... [FILE]...\n"), program_name
);
222 Summarize disk usage of each FILE, recursively for directories.\n\
224 -a, --all write counts for all files, not just directories\n\
225 -b, --bytes print size in bytes\n\
226 -c, --total produce a grand total\n\
227 -h, --human print sizes in human readable format (e.g. 1K 234M 2G)\n\
228 -k, --kilobytes use 1024-byte blocks, not 512 despite POSIXLY_CORRECT\n\
229 -l, --count-links count sizes many times if hard linked\n\
230 -m, --megabytes use 1024K-byte blocks, not 512 despite POSIXLY_CORRECT\n\
231 -s, --summarize display only a total for each argument\n\
232 -x, --one-file-system skip directories on different filesystems\n\
233 -D, --dereference-args dereference PATHs when symbolic link\n\
234 -L, --dereference dereference all symbolic links\n\
235 -S, --separate-dirs do not include size of subdirectories\n\
236 --help display this help and exit\n\
237 --version output version information and exit\n"));
243 main (int argc
, char **argv
)
252 program_name
= argv
[0];
255 if (getenv ("POSIXLY_CORRECT"))
256 output_size
= size_blocks
;
257 else if ((bs
= getenv ("BLOCKSIZE"))
258 && strncmp (bs
, "HUMAN", sizeof ("HUMAN") - 1) == 0)
260 opt_human_readable
= 1;
261 output_size
= size_bytes
;
264 output_size
= size_kilobytes
;
266 while ((c
= getopt_long (argc
, argv
, "abchklmsxDLS", long_options
,
272 case 0: /* Long option. */
280 output_size
= size_bytes
;
281 opt_human_readable
= 0;
285 opt_combined_arguments
= 1;
289 output_size
= size_bytes
;
290 opt_human_readable
= 1;
294 output_size
= size_kilobytes
;
295 opt_human_readable
= 0;
299 output_size
= size_megabytes
;
300 opt_human_readable
= 0;
308 opt_summarize_only
= 1;
312 opt_one_file_system
= 1;
316 opt_dereference_arguments
= 1;
324 opt_separate_dirs
= 1;
328 usage (2, (char *) 0);
334 printf ("du - %s\n", version_string
);
341 if (opt_all
&& opt_summarize_only
)
342 usage (2, _("cannot both summarize and show all entries"));
344 /* Initialize the hash structure for inode numbers. */
345 hash_init (INITIAL_HASH_MODULE
, INITIAL_ENTRY_TAB_SIZE
);
347 str_init (&path
, INITIAL_PATH_SIZE
);
349 du_files (optind
== argc
? cwd_only
: argv
+ optind
);
354 /* Convert N_BYTES to a more readable string than %d would.
355 Most people visually process strings of 3-4 digits effectively,
356 but longer strings of digits are more prone to misinterpretation.
357 Hence, converting to an abbreviated form usually improves readability.
358 Use a suffix indicating multiples of 1024 (K), 1024*1024 (M), and
359 1024*1024*1024 (G). For example, 8500 would be converted to 8.3K,
360 133456345 to 127M, 56990456345 to 53G, and so on. Numbers smaller
361 than 1024 aren't modified. */
364 human_readable (int n_bytes
, char *buf
, int buf_len
)
370 assert (buf_len
> LONGEST_HUMAN_READABLE
);
375 if (amt
>= 1024 * 1024 * 1024)
377 amt
/= (1024 * 1024 * 1024);
380 else if (amt
>= 1024 * 1024)
382 amt
/= (1024 * 1024);
385 else if (amt
>= 1024)
397 sprintf (p
, "%.0f%s", amt
, suffix
);
405 sprintf (p
, "%.1f%s", amt
, suffix
);
410 /* Recursively print the sizes of the directories (and, if selected, files)
411 named in FILES, the last entry of which is NULL. */
414 du_files (char **files
)
416 struct saved_cwd cwd
;
417 ino_t initial_ino
; /* Initial directory's inode. */
418 dev_t initial_dev
; /* Initial directory's device. */
419 int i
; /* Index in FILES. */
424 /* Remember the inode and device number of the current directory. */
425 if (stat (".", &stat_buf
))
426 error (1, errno
, _("current directory"));
427 initial_ino
= stat_buf
.st_ino
;
428 initial_dev
= stat_buf
.st_dev
;
430 for (i
= 0; files
[i
]; i
++)
437 /* Delete final slash in the argument, unless the slash is alone. */
438 s
= strlen (arg
) - 1;
444 str_copyc (path
, arg
);
446 else if (arg
[0] == '/')
447 str_trunc (path
, 0); /* Null path for root directory. */
449 str_copyc (path
, arg
);
451 if (!opt_combined_arguments
)
454 count_entry (arg
, 1, 0);
456 /* chdir if `count_entry' has changed the working directory. */
457 if (stat (".", &stat_buf
))
458 error (1, errno
, ".");
459 if (stat_buf
.st_ino
!= initial_ino
|| stat_buf
.st_dev
!= initial_dev
)
461 if (restore_cwd (&cwd
, _("starting directory"), NULL
))
466 if (opt_combined_arguments
)
468 if (opt_human_readable
)
470 char buf
[LONGEST_HUMAN_READABLE
+ 1];
471 printf("%s\ttotal\n", human_readable (tot_size
, buf
,
472 LONGEST_HUMAN_READABLE
+ 1));
476 printf (_("%ld\ttotal\n"), output_size
== size_bytes
? tot_size
477 : convert_blocks (tot_size
, output_size
== size_kilobytes
));
485 /* Print (if appropriate) and return the size
486 (in units determined by `output_size') of file or directory ENT.
487 TOP is one for external calls, zero for recursive calls.
488 LAST_DEV is the device that the parent directory of ENT is on. */
491 count_entry (char *ent
, int top
, dev_t last_dev
)
495 if (((top
&& opt_dereference_arguments
)
496 ? stat (ent
, &stat_buf
)
497 : (*xstat
) (ent
, &stat_buf
)) < 0)
499 error (0, errno
, "%s", path
->text
);
505 && stat_buf
.st_nlink
> 1
506 && hash_insert (stat_buf
.st_ino
, stat_buf
.st_dev
))
507 return 0; /* Have counted this already. */
509 if (output_size
== size_bytes
)
510 size
= stat_buf
.st_size
;
512 size
= ST_NBLOCKS (stat_buf
);
516 if (S_ISDIR (stat_buf
.st_mode
))
522 struct saved_cwd cwd
;
526 dir_dev
= stat_buf
.st_dev
;
528 if (opt_one_file_system
&& !top
&& last_dev
!= dir_dev
)
529 return 0; /* Don't enter a new file system. */
532 # define S_ISDIR(s) 0
534 /* If we're dereferencing symlinks and we're about to chdir through
535 a symlink, remember the current directory so we can return to it
536 later. In other cases, chdir ("..") works fine. */
537 through_symlink
= (xstat
== stat
538 && lstat (ent
, &e_buf
) == 0
539 && S_ISLNK (e_buf
.st_mode
));
546 error (0, errno
, _("cannot change to directory %s"), path
->text
);
552 name_space
= savedir (".", stat_buf
.st_size
);
553 if (name_space
== NULL
)
557 error (0, errno
, "%s", path
->text
);
560 if (restore_cwd (&cwd
, "..", path
->text
))
564 else if (chdir ("..") < 0)
565 error (1, errno
, _("cannot change to `..' from directory %s"),
571 error (1, 0, _("virtual memory exhausted"));
574 /* Remember the current path. */
576 str_concatc (path
, "/");
577 pathlen
= path
->length
;
582 str_concatc (path
, namep
);
584 size
+= count_entry (namep
, 0, dir_dev
);
586 str_trunc (path
, pathlen
);
587 namep
+= strlen (namep
) + 1;
592 restore_cwd (&cwd
, "..", path
->text
);
595 else if (chdir ("..") < 0)
597 _("cannot change to `..' from directory %s"), path
->text
);
599 str_trunc (path
, pathlen
- 1); /* Remove the "/" we added. */
600 if (!opt_summarize_only
|| top
)
602 if (opt_human_readable
)
604 char buf
[LONGEST_HUMAN_READABLE
+ 1];
606 human_readable (size
, buf
, LONGEST_HUMAN_READABLE
+ 1),
607 path
->length
> 0 ? path
->text
: "/");
611 printf ("%ld\t%s\n", (output_size
== size_bytes
613 : convert_blocks (size
, output_size
)),
614 path
->length
> 0 ? path
->text
: "/");
618 return opt_separate_dirs
? 0 : size
;
620 else if (opt_all
|| top
)
622 /* FIXME: make this an option. */
623 int print_only_dir_size
= 0;
624 if (!print_only_dir_size
)
626 if (opt_human_readable
)
628 char buf
[LONGEST_HUMAN_READABLE
+ 1];
630 human_readable (size
, buf
, LONGEST_HUMAN_READABLE
+ 1),
631 path
->length
> 0 ? path
->text
: "/");
635 printf ("%ld\t%s\n", output_size
== size_bytes
? size
636 : convert_blocks (size
, output_size
== size_kilobytes
),
646 /* Allocate space for the hash structures, and set the global
647 variable `htab' to point to it. The initial hash module is specified in
648 MODULUS, and the number of entries are specified in ENTRY_TAB_SIZE. (The
649 hash structure will be rebuilt when ENTRY_TAB_SIZE entries have been
650 inserted, and MODULUS and ENTRY_TAB_SIZE in the global `htab' will be
654 hash_init (unsigned int modulus
, unsigned int entry_tab_size
)
658 htab_r
= (struct htab
*)
659 xmalloc (sizeof (struct htab
) + sizeof (struct entry
*) * modulus
);
661 htab_r
->entry_tab
= (struct entry
*)
662 xmalloc (sizeof (struct entry
) * entry_tab_size
);
664 htab_r
->modulus
= modulus
;
665 htab_r
->entry_tab_size
= entry_tab_size
;
671 /* Reset the hash structure in the global variable `htab' to
672 contain no entries. */
680 htab
->first_free_entry
= 0;
683 for (i
= htab
->modulus
; i
> 0; i
--)
687 /* Insert an item (inode INO and device DEV) in the hash
688 structure in the global variable `htab', if an entry with the same data
689 was not found already. Return zero if the item was inserted and nonzero
693 hash_insert (ino_t ino
, dev_t dev
)
695 struct htab
*htab_r
= htab
; /* Initially a copy of the global `htab'. */
697 if (htab_r
->first_free_entry
>= htab_r
->entry_tab_size
)
702 unsigned entry_tab_size
;
704 /* Increase the number of hash entries, and re-hash the data.
705 The method of shrimping and increasing is made to compactify
706 the heap. If twice as much data would be allocated
707 straightforwardly, we would never re-use a byte of memory. */
709 /* Let `htab' shrimp. Keep only the header, not the pointer vector. */
711 htab_r
= (struct htab
*)
712 xrealloc ((char *) htab_r
, sizeof (struct htab
));
714 modulus
= 2 * htab_r
->modulus
;
715 entry_tab_size
= 2 * htab_r
->entry_tab_size
;
717 /* Increase the number of possible entries. */
719 htab_r
->entry_tab
= (struct entry
*)
720 xrealloc ((char *) htab_r
->entry_tab
,
721 sizeof (struct entry
) * entry_tab_size
);
723 /* Increase the size of htab again. */
725 htab_r
= (struct htab
*)
726 xrealloc ((char *) htab_r
,
727 sizeof (struct htab
) + sizeof (struct entry
*) * modulus
);
729 htab_r
->modulus
= modulus
;
730 htab_r
->entry_tab_size
= entry_tab_size
;
733 i
= htab_r
->first_free_entry
;
735 /* Make the increased hash table empty. The entries are still
736 available in htab->entry_tab. */
740 /* Go through the entries and install them in the pointer vector
741 htab->hash. The items are actually inserted in htab->entry_tab at
742 the position where they already are. The htab->coll_link need
743 however be updated. Could be made a little more efficient. */
745 for (ep
= htab_r
->entry_tab
; i
> 0; i
--)
747 hash_insert2 (htab_r
, ep
->ino
, ep
->dev
);
752 return hash_insert2 (htab_r
, ino
, dev
);
755 /* Insert INO and DEV in the hash structure HTAB, if not
756 already present. Return zero if inserted and nonzero if it
760 hash_insert2 (struct htab
*htab
, ino_t ino
, dev_t dev
)
762 struct entry
**hp
, *ep2
, *ep
;
763 hp
= &htab
->hash
[ino
% htab
->modulus
];
772 /* Search for an entry with the same data. */
776 if (ep
->ino
== ino
&& ep
->dev
== dev
)
777 return 1; /* Found an entry with the same data. */
782 /* Did not find it. */
786 ep
= *hp
= &htab
->entry_tab
[htab
->first_free_entry
++];
789 ep
->coll_link
= ep2
; /* `ep2' is NULL if no collision. */
794 /* Initialize the struct string S1 for holding SIZE characters. */
797 str_init (string
*s1
, unsigned int size
)
801 s
= (string
) xmalloc (sizeof (stringstruct
));
802 s
->text
= xmalloc (size
+ 1);
809 ensure_space (string s
, unsigned int size
)
813 s
->text
= xrealloc (s
->text
, size
+ 1);
818 /* Assign the null-terminated C-string CSTR to S1. */
821 str_copyc (string s1
, char *cstr
)
823 unsigned l
= strlen (cstr
);
824 ensure_space (s1
, l
);
825 strcpy (s1
->text
, cstr
);
830 str_concatc (string s1
, char *cstr
)
832 unsigned l1
= s1
->length
;
833 unsigned l2
= strlen (cstr
);
834 unsigned l
= l1
+ l2
;
836 ensure_space (s1
, l
);
837 strcpy (s1
->text
+ l1
, cstr
);
841 /* Truncate the string S1 to have length LENGTH. */
844 str_trunc (string s1
, unsigned int length
)
846 if (s1
->length
> length
)
848 s1
->text
[length
] = 0;