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>
58 #define convert_blocks(b, size) (size == size_kilobytes ? ((b) + 1) / 2 : \
59 size == size_megabytes ? ((b) + 1024) / 2048 : (b))
61 /* Initial number of entries in each hash table entry's table of inodes. */
62 #define INITIAL_HASH_MODULE 100
64 /* Initial number of entries in the inode hash table. */
65 #define INITIAL_ENTRY_TAB_SIZE 70
67 /* Initial size to allocate for `path'. */
68 #define INITIAL_PATH_SIZE 100
70 /* The maximum length of a human-readable string. Be pessimistic
71 and assume `int' is 64-bits wide. Converting 2^63 - 1 gives the
72 11-character string, 8589934592G. */
73 #define LONGEST_HUMAN_READABLE 11
75 /* Hash structure for inode and device numbers. The separate entry
76 structure makes it easier to rehash "in place". */
82 struct entry
*coll_link
;
85 /* Structure for a hash table for inode numbers. */
89 unsigned modulus
; /* Size of the `hash' pointer vector. */
90 struct entry
*entry_tab
; /* Pointer to dynamically growing vector. */
91 unsigned entry_tab_size
; /* Size of current `entry_tab' allocation. */
92 unsigned first_free_entry
; /* Index in `entry_tab'. */
93 struct entry
*hash
[1]; /* Vector of pointers in `entry_tab'. */
97 /* Structure for dynamically resizable strings. */
101 unsigned alloc
; /* Size of allocation for the text. */
102 unsigned length
; /* Length of the text currently. */
103 char *text
; /* Pointer to the text. */
104 } *string
, stringstruct
;
113 static int hash_insert
__P ((ino_t ino
, dev_t dev
));
114 static int hash_insert2
__P ((struct htab
*htab
, ino_t ino
, dev_t dev
));
115 static long count_entry
__P ((char *ent
, int top
, dev_t last_dev
));
116 static void du_files
__P ((char **files
));
117 static void hash_init
__P ((unsigned int modulus
,
118 unsigned int entry_tab_size
));
119 static void hash_reset
__P ((void));
120 static void str_concatc
__P ((string s1
, char *cstr
));
121 static void str_copyc
__P ((string s1
, char *cstr
));
122 static void str_init
__P ((string
*s1
, unsigned int size
));
123 static void str_trunc
__P ((string s1
, unsigned int length
));
125 /* Name under which this program was invoked. */
128 /* If nonzero, display only a total for each argument. */
129 static int opt_summarize_only
= 0;
131 /* If nonzero, display counts for all files, not just directories. */
132 static int opt_all
= 0;
134 /* If nonzero, count each hard link of files with multiple links. */
135 static int opt_count_all
= 0;
137 /* If nonzero, do not cross file-system boundaries. */
138 static int opt_one_file_system
= 0;
140 /* If nonzero, print a grand total at the end. */
141 static int opt_combined_arguments
= 0;
143 /* If nonzero, do not add sizes of subdirectories. */
144 static int opt_separate_dirs
= 0;
146 /* If nonzero, dereference symlinks that are command line arguments. */
147 static int opt_dereference_arguments
= 0;
151 size_blocks
, /* 512-byte blocks. */
152 size_kilobytes
, /* 1K blocks. */
153 size_megabytes
, /* 1024K blocks. */
154 size_bytes
/* 1-byte blocks. */
157 /* human style output */
158 static int opt_human_readable
;
160 /* The units to count in. */
161 static enum output_size output_size
;
163 /* Accumulated path for file or directory being processed. */
166 /* Pointer to hash structure, used by the hash routines. */
167 static struct htab
*htab
;
169 /* Globally used stat buffer. */
170 static struct stat stat_buf
;
172 /* A pointer to either lstat or stat, depending on whether
173 dereferencing of all symbolic links is to be done. */
174 static int (*xstat
) ();
176 /* The exit status to use if we don't get any fatal errors. */
177 static int exit_status
;
179 /* If nonzero, display usage information and exit. */
180 static int show_help
;
182 /* If nonzero, print the version on standard output and exit. */
183 static int show_version
;
185 /* Grand total size of all args. */
186 static long tot_size
= 0L;
188 static struct option
const long_options
[] =
190 {"all", no_argument
, &opt_all
, 1},
191 {"bytes", no_argument
, NULL
, 'b'},
192 {"count-links", no_argument
, &opt_count_all
, 1},
193 {"dereference", no_argument
, NULL
, 'L'},
194 {"dereference-args", no_argument
, &opt_dereference_arguments
, 1},
195 {"human-readable", no_argument
, NULL
, 'h'},
196 {"kilobytes", no_argument
, NULL
, 'k'},
197 {"megabytes", no_argument
, NULL
, 'm'},
198 {"one-file-system", no_argument
, &opt_one_file_system
, 1},
199 {"separate-dirs", no_argument
, &opt_separate_dirs
, 1},
200 {"summarize", no_argument
, &opt_summarize_only
, 1},
201 {"total", no_argument
, &opt_combined_arguments
, 1},
202 {"help", no_argument
, &show_help
, 1},
203 {"version", no_argument
, &show_version
, 1},
208 usage (int status
, char *reason
)
211 fprintf (status
== 0 ? stdout
: stderr
, "%s: %s\n",
212 program_name
, reason
);
215 fprintf (stderr
, _("Try `%s --help' for more information.\n"),
219 printf (_("Usage: %s [OPTION]... [FILE]...\n"), program_name
);
221 Summarize disk usage of each FILE, recursively for directories.\n\
223 -a, --all write counts for all files, not just directories\n\
224 -b, --bytes print size in bytes\n\
225 -c, --total produce a grand total\n\
226 -h, --human print sizes in human readable format (e.g. 1K 234M 2G)\n\
227 -k, --kilobytes use 1024-byte blocks, not 512 despite POSIXLY_CORRECT\n\
228 -l, --count-links count sizes many times if hard linked\n\
229 -m, --megabytes use 1024K-byte blocks, not 512 despite POSIXLY_CORRECT\n\
230 -s, --summarize display only a total for each argument\n\
231 -x, --one-file-system skip directories on different filesystems\n\
232 -D, --dereference-args dereference PATHs when symbolic link\n\
233 -L, --dereference dereference all symbolic links\n\
234 -S, --separate-dirs do not include size of subdirectories\n\
235 --help display this help and exit\n\
236 --version output version information and exit\n"));
242 main (int argc
, char **argv
)
251 program_name
= argv
[0];
252 setlocale (LC_ALL
, "");
253 bindtextdomain (PACKAGE
, LOCALEDIR
);
254 textdomain (PACKAGE
);
258 if (getenv ("POSIXLY_CORRECT"))
259 output_size
= size_blocks
;
260 else if ((bs
= getenv ("BLOCKSIZE"))
261 && strncmp (bs
, "HUMAN", sizeof ("HUMAN") - 1) == 0)
263 opt_human_readable
= 1;
264 output_size
= size_bytes
;
267 output_size
= size_kilobytes
;
269 while ((c
= getopt_long (argc
, argv
, "abchklmsxDLS", long_options
,
275 case 0: /* Long option. */
283 output_size
= size_bytes
;
284 opt_human_readable
= 0;
288 opt_combined_arguments
= 1;
292 output_size
= size_bytes
;
293 opt_human_readable
= 1;
297 output_size
= size_kilobytes
;
298 opt_human_readable
= 0;
302 output_size
= size_megabytes
;
303 opt_human_readable
= 0;
311 opt_summarize_only
= 1;
315 opt_one_file_system
= 1;
319 opt_dereference_arguments
= 1;
327 opt_separate_dirs
= 1;
331 usage (2, (char *) 0);
337 printf ("du - %s\n", PACKAGE_VERSION
);
344 if (opt_all
&& opt_summarize_only
)
345 usage (2, _("cannot both summarize and show all entries"));
347 /* Initialize the hash structure for inode numbers. */
348 hash_init (INITIAL_HASH_MODULE
, INITIAL_ENTRY_TAB_SIZE
);
350 str_init (&path
, INITIAL_PATH_SIZE
);
352 du_files (optind
== argc
? cwd_only
: argv
+ optind
);
357 /* Convert N_BYTES to a more readable string than %d would.
358 Most people visually process strings of 3-4 digits effectively,
359 but longer strings of digits are more prone to misinterpretation.
360 Hence, converting to an abbreviated form usually improves readability.
361 Use a suffix indicating multiples of 1024 (K), 1024*1024 (M), and
362 1024*1024*1024 (G). For example, 8500 would be converted to 8.3K,
363 133456345 to 127M, 56990456345 to 53G, and so on. Numbers smaller
364 than 1024 aren't modified. */
367 human_readable (int n_bytes
, char *buf
, int buf_len
)
373 assert (buf_len
> LONGEST_HUMAN_READABLE
);
378 if (amt
>= 1024 * 1024 * 1024)
380 amt
/= (1024 * 1024 * 1024);
383 else if (amt
>= 1024 * 1024)
385 amt
/= (1024 * 1024);
388 else if (amt
>= 1024)
400 sprintf (p
, "%.0f%s", amt
, suffix
);
408 sprintf (p
, "%.1f%s", amt
, suffix
);
413 /* Recursively print the sizes of the directories (and, if selected, files)
414 named in FILES, the last entry of which is NULL. */
417 du_files (char **files
)
419 struct saved_cwd cwd
;
420 ino_t initial_ino
; /* Initial directory's inode. */
421 dev_t initial_dev
; /* Initial directory's device. */
422 int i
; /* Index in FILES. */
427 /* Remember the inode and device number of the current directory. */
428 if (stat (".", &stat_buf
))
429 error (1, errno
, _("current directory"));
430 initial_ino
= stat_buf
.st_ino
;
431 initial_dev
= stat_buf
.st_dev
;
433 for (i
= 0; files
[i
]; i
++)
440 /* Delete final slash in the argument, unless the slash is alone. */
441 s
= strlen (arg
) - 1;
447 str_copyc (path
, arg
);
449 else if (arg
[0] == '/')
450 str_trunc (path
, 0); /* Null path for root directory. */
452 str_copyc (path
, arg
);
454 if (!opt_combined_arguments
)
457 count_entry (arg
, 1, 0);
459 /* chdir if `count_entry' has changed the working directory. */
460 if (stat (".", &stat_buf
))
461 error (1, errno
, ".");
462 if (stat_buf
.st_ino
!= initial_ino
|| stat_buf
.st_dev
!= initial_dev
)
464 if (restore_cwd (&cwd
, _("starting directory"), NULL
))
469 if (opt_combined_arguments
)
471 if (opt_human_readable
)
473 char buf
[LONGEST_HUMAN_READABLE
+ 1];
474 printf("%s\ttotal\n", human_readable (tot_size
, buf
,
475 LONGEST_HUMAN_READABLE
+ 1));
479 printf (_("%ld\ttotal\n"), output_size
== size_bytes
? tot_size
480 : convert_blocks (tot_size
, output_size
== size_kilobytes
));
488 /* Print (if appropriate) and return the size
489 (in units determined by `output_size') of file or directory ENT.
490 TOP is one for external calls, zero for recursive calls.
491 LAST_DEV is the device that the parent directory of ENT is on. */
494 count_entry (char *ent
, int top
, dev_t last_dev
)
498 if (((top
&& opt_dereference_arguments
)
499 ? stat (ent
, &stat_buf
)
500 : (*xstat
) (ent
, &stat_buf
)) < 0)
502 error (0, errno
, "%s", path
->text
);
508 && stat_buf
.st_nlink
> 1
509 && hash_insert (stat_buf
.st_ino
, stat_buf
.st_dev
))
510 return 0; /* Have counted this already. */
512 if (output_size
== size_bytes
)
513 size
= stat_buf
.st_size
;
515 size
= ST_NBLOCKS (stat_buf
);
519 if (S_ISDIR (stat_buf
.st_mode
))
525 struct saved_cwd cwd
;
529 dir_dev
= stat_buf
.st_dev
;
531 if (opt_one_file_system
&& !top
&& last_dev
!= dir_dev
)
532 return 0; /* Don't enter a new file system. */
535 # define S_ISDIR(s) 0
537 /* If we're dereferencing symlinks and we're about to chdir through
538 a symlink, remember the current directory so we can return to it
539 later. In other cases, chdir ("..") works fine. */
540 through_symlink
= (xstat
== stat
541 && lstat (ent
, &e_buf
) == 0
542 && S_ISLNK (e_buf
.st_mode
));
549 error (0, errno
, _("cannot change to directory %s"), path
->text
);
555 name_space
= savedir (".", stat_buf
.st_size
);
556 if (name_space
== NULL
)
560 error (0, errno
, "%s", path
->text
);
563 if (restore_cwd (&cwd
, "..", path
->text
))
567 else if (chdir ("..") < 0)
568 error (1, errno
, _("cannot change to `..' from directory %s"),
574 error (1, 0, _("virtual memory exhausted"));
577 /* Remember the current path. */
579 str_concatc (path
, "/");
580 pathlen
= path
->length
;
585 str_concatc (path
, namep
);
587 size
+= count_entry (namep
, 0, dir_dev
);
589 str_trunc (path
, pathlen
);
590 namep
+= strlen (namep
) + 1;
595 restore_cwd (&cwd
, "..", path
->text
);
598 else if (chdir ("..") < 0)
600 _("cannot change to `..' from directory %s"), path
->text
);
602 str_trunc (path
, pathlen
- 1); /* Remove the "/" we added. */
603 if (!opt_summarize_only
|| top
)
605 if (opt_human_readable
)
607 char buf
[LONGEST_HUMAN_READABLE
+ 1];
609 human_readable (size
, buf
, LONGEST_HUMAN_READABLE
+ 1),
610 path
->length
> 0 ? path
->text
: "/");
614 printf ("%ld\t%s\n", (output_size
== size_bytes
616 : convert_blocks (size
, output_size
)),
617 path
->length
> 0 ? path
->text
: "/");
621 return opt_separate_dirs
? 0 : size
;
623 else if (opt_all
|| top
)
625 /* FIXME: make this an option. */
626 int print_only_dir_size
= 0;
627 if (!print_only_dir_size
)
629 if (opt_human_readable
)
631 char buf
[LONGEST_HUMAN_READABLE
+ 1];
633 human_readable (size
, buf
, LONGEST_HUMAN_READABLE
+ 1),
634 path
->length
> 0 ? path
->text
: "/");
638 printf ("%ld\t%s\n", output_size
== size_bytes
? size
639 : convert_blocks (size
, output_size
== size_kilobytes
),
649 /* Allocate space for the hash structures, and set the global
650 variable `htab' to point to it. The initial hash module is specified in
651 MODULUS, and the number of entries are specified in ENTRY_TAB_SIZE. (The
652 hash structure will be rebuilt when ENTRY_TAB_SIZE entries have been
653 inserted, and MODULUS and ENTRY_TAB_SIZE in the global `htab' will be
657 hash_init (unsigned int modulus
, unsigned int entry_tab_size
)
661 htab_r
= (struct htab
*)
662 xmalloc (sizeof (struct htab
) + sizeof (struct entry
*) * modulus
);
664 htab_r
->entry_tab
= (struct entry
*)
665 xmalloc (sizeof (struct entry
) * entry_tab_size
);
667 htab_r
->modulus
= modulus
;
668 htab_r
->entry_tab_size
= entry_tab_size
;
674 /* Reset the hash structure in the global variable `htab' to
675 contain no entries. */
683 htab
->first_free_entry
= 0;
686 for (i
= htab
->modulus
; i
> 0; i
--)
690 /* Insert an item (inode INO and device DEV) in the hash
691 structure in the global variable `htab', if an entry with the same data
692 was not found already. Return zero if the item was inserted and nonzero
696 hash_insert (ino_t ino
, dev_t dev
)
698 struct htab
*htab_r
= htab
; /* Initially a copy of the global `htab'. */
700 if (htab_r
->first_free_entry
>= htab_r
->entry_tab_size
)
705 unsigned entry_tab_size
;
707 /* Increase the number of hash entries, and re-hash the data.
708 The method of shrimping and increasing is made to compactify
709 the heap. If twice as much data would be allocated
710 straightforwardly, we would never re-use a byte of memory. */
712 /* Let `htab' shrimp. Keep only the header, not the pointer vector. */
714 htab_r
= (struct htab
*)
715 xrealloc ((char *) htab_r
, sizeof (struct htab
));
717 modulus
= 2 * htab_r
->modulus
;
718 entry_tab_size
= 2 * htab_r
->entry_tab_size
;
720 /* Increase the number of possible entries. */
722 htab_r
->entry_tab
= (struct entry
*)
723 xrealloc ((char *) htab_r
->entry_tab
,
724 sizeof (struct entry
) * entry_tab_size
);
726 /* Increase the size of htab again. */
728 htab_r
= (struct htab
*)
729 xrealloc ((char *) htab_r
,
730 sizeof (struct htab
) + sizeof (struct entry
*) * modulus
);
732 htab_r
->modulus
= modulus
;
733 htab_r
->entry_tab_size
= entry_tab_size
;
736 i
= htab_r
->first_free_entry
;
738 /* Make the increased hash table empty. The entries are still
739 available in htab->entry_tab. */
743 /* Go through the entries and install them in the pointer vector
744 htab->hash. The items are actually inserted in htab->entry_tab at
745 the position where they already are. The htab->coll_link need
746 however be updated. Could be made a little more efficient. */
748 for (ep
= htab_r
->entry_tab
; i
> 0; i
--)
750 hash_insert2 (htab_r
, ep
->ino
, ep
->dev
);
755 return hash_insert2 (htab_r
, ino
, dev
);
758 /* Insert INO and DEV in the hash structure HTAB, if not
759 already present. Return zero if inserted and nonzero if it
763 hash_insert2 (struct htab
*htab
, ino_t ino
, dev_t dev
)
765 struct entry
**hp
, *ep2
, *ep
;
766 hp
= &htab
->hash
[ino
% htab
->modulus
];
775 /* Search for an entry with the same data. */
779 if (ep
->ino
== ino
&& ep
->dev
== dev
)
780 return 1; /* Found an entry with the same data. */
785 /* Did not find it. */
789 ep
= *hp
= &htab
->entry_tab
[htab
->first_free_entry
++];
792 ep
->coll_link
= ep2
; /* `ep2' is NULL if no collision. */
797 /* Initialize the struct string S1 for holding SIZE characters. */
800 str_init (string
*s1
, unsigned int size
)
804 s
= (string
) xmalloc (sizeof (stringstruct
));
805 s
->text
= xmalloc (size
+ 1);
812 ensure_space (string s
, unsigned int size
)
816 s
->text
= xrealloc (s
->text
, size
+ 1);
821 /* Assign the null-terminated C-string CSTR to S1. */
824 str_copyc (string s1
, char *cstr
)
826 unsigned l
= strlen (cstr
);
827 ensure_space (s1
, l
);
828 strcpy (s1
->text
, cstr
);
833 str_concatc (string s1
, char *cstr
)
835 unsigned l1
= s1
->length
;
836 unsigned l2
= strlen (cstr
);
837 unsigned l
= l1
+ l2
;
839 ensure_space (s1
, l
);
840 strcpy (s1
->text
+ l1
, cstr
);
844 /* Truncate the string S1 to have length LENGTH. */
847 str_trunc (string s1
, unsigned int length
)
849 if (s1
->length
> length
)
851 s1
->text
[length
] = 0;