Implement --one-file-system (-x).
[adu.git] / adu.c
1 #include "adu.h"
2 #include <dirent.h> /* readdir() */
3
4 #include "gcc-compat.h"
5 #include "cmdline.h"
6 #include "fd.h"
7 #include "string.h"
8 #include "error.h"
9 #include "portable_io.h"
10
11 DEFINE_ERRLIST;
12 int osl_errno;
13
14 /** Command line and config file options. */
15 static struct gengetopt_args_info conf;
16
17 enum uid_info_flags {
18 /** whether this slot of the hash table is used. */
19 UI_FL_SLOT_USED = 1,
20 /** whether this uid should be taken into account. */
21 UI_FL_ADMISSIBLE = 2,
22 };
23
24 struct user_info {
25 uint32_t uid;
26 uint32_t flags;
27 struct osl_table *table;
28 uint64_t files;
29 uint64_t bytes;
30 uint64_t dirs;
31 struct osl_table_description *desc;
32 };
33
34 /** The decimal representation of an uint64_t never exceeds that size. */
35 #define FORMATED_VALUE_SIZE 25
36
37 #define FOR_EACH_USER(ui) for (ui = uid_hash_table; ui && ui < uid_hash_table \
38 + uid_hash_table_size; ui++)
39
40
41 /**
42 * Contains info for each user that owns at least one regular file.
43 *
44 * Even users that are not taken into account because of the --uid
45 * option occupy a slot in this hash table. This allows to find out
46 * quicky whether a uid is admissible. And yes, this has to be fast.
47 */
48 static struct user_info *uid_hash_table;
49
50 /* these get filled in by the select command. */
51 static char count_unit_buf[4] = "( )", size_unit_buf[4] = "( )";
52
53 static inline int ui_used(struct user_info *ui)
54 {
55 return ui->flags & UI_FL_SLOT_USED;
56 }
57
58 static inline int ui_admissible(struct user_info *ui)
59 {
60 return ui->flags & UI_FL_ADMISSIBLE;
61 }
62
63 struct uid_range {
64 uint32_t low;
65 uint32_t high;
66 };
67
68 static struct uid_range *admissible_uids;
69
70 static inline int check_uid_arg(const char *arg, uint32_t *uid)
71 {
72 const uint32_t max = ~0U;
73 /*
74 * we need an 64-bit int for string -> uid conversion because strtoll()
75 * returns a signed value.
76 */
77 int64_t val;
78 int ret = atoi64(arg, &val);
79
80 if (ret < 0)
81 return ret;
82 if (val < 0 || val > max)
83 return -ERRNO_TO_ERROR(EINVAL);
84 *uid = val;
85 return 1;
86 }
87
88 static int parse_uid_range(const char *orig_arg, struct uid_range *ur)
89 {
90 int ret;
91 char *arg = adu_strdup(orig_arg), *p = strchr(arg, '-');
92
93 if (!p || p == arg) { /* -42 or 42 */
94 ret = check_uid_arg(p? p + 1 : arg, &ur->high);
95 if (ret < 0)
96 goto out;
97 ur->low = p? 0 : ur->high;
98 ret = 1;
99 goto out;
100 }
101 /* 42- or 42-4711 */
102 *p = '\0';
103 p++;
104 ret = check_uid_arg(arg, &ur->low);
105 if (ret < 0)
106 goto out;
107 ur->high = ~0U;
108 if (*p) { /* 42-4711 */
109 ret = check_uid_arg(p, &ur->high);
110 if (ret < 0)
111 goto out;
112 }
113 if (ur->low > ur->high)
114 ret = -ERRNO_TO_ERROR(EINVAL);
115 out:
116 if (ret < 0)
117 ERROR_LOG("bad uid option: %s\n", orig_arg);
118 else
119 INFO_LOG("admissible uid range: %u - %u\n", ur->low,
120 ur->high);
121 free(arg);
122 return ret;
123 }
124
125 /** evaluates to 1 if x < y, to -1 if x > y and to 0 if x == y */
126 #define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
127
128 /**
129 * The log function.
130 *
131 * \param ll Loglevel.
132 * \param fml Usual format string.
133 *
134 * All XXX_LOG() macros use this function.
135 */
136 __printf_2_3 void __log(int ll, const char* fmt,...)
137 {
138 va_list argp;
139 FILE *outfd;
140 struct tm *tm;
141 time_t t1;
142 char str[255] = "";
143
144 if (ll < conf.loglevel_arg)
145 return;
146 outfd = stderr;
147 time(&t1);
148 tm = localtime(&t1);
149 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
150 fprintf(outfd, "%s ", str);
151 va_start(argp, fmt);
152 vfprintf(outfd, fmt, argp);
153 va_end(argp);
154 }
155
156 /**
157 * Compare the size of two directories
158 *
159 * \param obj1 Pointer to the first object.
160 * \param obj2 Pointer to the second object.
161 *
162 * This function first compares the size values as usual integers. If they compare as
163 * equal, the address of \a obj1 and \a obj2 are compared. So this compare function
164 * returns zero if and only if \a obj1 and \a obj2 point to the same memory area.
165 */
166 static int size_compare(const struct osl_object *obj1, const struct osl_object *obj2)
167 {
168 uint64_t d1 = *(uint64_t *)obj1->data;
169 uint64_t d2 = *(uint64_t *)obj2->data;
170 int ret = NUM_COMPARE(d2, d1);
171
172 if (ret)
173 return ret;
174 //INFO_LOG("addresses: %p, %p\n", obj1->data, obj2->data);
175 return NUM_COMPARE(obj2->data, obj1->data);
176 }
177
178 /**
179 * Compare two osl objects pointing to unsigned integers of 64 bit size.
180 *
181 * \param obj1 Pointer to the first integer.
182 * \param obj2 Pointer to the second integer.
183 *
184 * \return The values required for an osl compare function.
185 *
186 * \sa osl_compare_func, osl_hash_compare().
187 */
188 static int uint64_compare(const struct osl_object *obj1,
189 const struct osl_object *obj2)
190 {
191 uint64_t d1 = read_u64((const char *)obj1->data);
192 uint64_t d2 = read_u64((const char *)obj2->data);
193
194 if (d1 < d2)
195 return 1;
196 if (d1 > d2)
197 return -1;
198 return 0;
199 }
200
201 /** The columns of the directory table. */
202 enum dir_table_columns {
203 /** The name of the directory. */
204 DT_NAME,
205 /** The dir count number. */
206 DT_NUM,
207 /** The number of the parent directory. */
208 DT_PARENT_NUM,
209 /** The number of bytes of all regular files. */
210 DT_BYTES,
211 /** The number of all regular files. */
212 DT_FILES,
213 /** Number of columns in this table. */
214 NUM_DT_COLUMNS
215 };
216
217 static struct osl_column_description dir_table_cols[] = {
218 [DT_NAME] = {
219 .storage_type = OSL_MAPPED_STORAGE,
220 .storage_flags = 0,
221 .name = "dir",
222 },
223 [DT_NUM] = {
224 .storage_type = OSL_MAPPED_STORAGE,
225 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
226 .name = "num",
227 .compare_function = uint64_compare,
228 .data_size = sizeof(uint64_t)
229 },
230 [DT_PARENT_NUM] = {
231 .storage_type = OSL_MAPPED_STORAGE,
232 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
233 .name = "parent_num",
234 .compare_function = size_compare,
235 .data_size = sizeof(uint64_t)
236 },
237 [DT_BYTES] = {
238 .storage_type = OSL_MAPPED_STORAGE,
239 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
240 .compare_function = size_compare,
241 .name = "num_bytes",
242 .data_size = sizeof(uint64_t)
243 },
244 [DT_FILES] = {
245 .storage_type = OSL_MAPPED_STORAGE,
246 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
247 .compare_function = size_compare,
248 .name = "num_files",
249 .data_size = sizeof(uint64_t)
250 }
251 };
252
253 static struct osl_table_description dir_table_desc = {
254 .name = "dir_table",
255 .num_columns = NUM_DT_COLUMNS,
256 .flags = 0,
257 .column_descriptions = dir_table_cols,
258 };
259
260 /** The columns of the id table. */
261 enum user_table_columns {
262 /** The numer of the directory. */
263 UT_DIR_NUM,
264 /** The number of bytes of all regular files in this dir owned by this id. */
265 UT_BYTES,
266 /** The number of files in this dir owned by this id. */
267 UT_FILES,
268 /** Number of columns in this table. */
269 NUM_UT_COLUMNS
270 };
271
272 static struct osl_column_description user_table_cols[] = {
273 [UT_DIR_NUM] = {
274 .storage_type = OSL_MAPPED_STORAGE,
275 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
276 .name = "dir_num",
277 .compare_function = uint64_compare,
278 .data_size = sizeof(uint64_t)
279 },
280 [UT_BYTES] = {
281 .storage_type = OSL_MAPPED_STORAGE,
282 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
283 .compare_function = size_compare,
284 .name = "num_bytes",
285 .data_size = sizeof(uint64_t)
286 },
287 [UT_FILES] = {
288 .storage_type = OSL_MAPPED_STORAGE,
289 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
290 .compare_function = size_compare,
291 .name = "num_files",
292 .data_size = sizeof(uint64_t)
293 },
294 };
295
296 static struct osl_table *dir_table;
297
298 static int add_directory(char *dirname, uint64_t *dir_num, uint64_t *parent_dir_num,
299 uint64_t *dir_size, uint64_t *dir_files)
300 {
301 struct osl_object dir_objects[NUM_DT_COLUMNS];
302
303 INFO_LOG("adding #%llu: %s\n", (long long unsigned)*dir_num, dirname);
304 dir_objects[DT_NAME].data = dirname;
305 dir_objects[DT_NAME].size = strlen(dirname) + 1;
306 dir_objects[DT_NUM].data = dir_num;
307 dir_objects[DT_NUM].size = sizeof(*dir_num);
308 dir_objects[DT_PARENT_NUM].data = parent_dir_num;
309 dir_objects[DT_PARENT_NUM].size = sizeof(*parent_dir_num);
310 dir_objects[DT_BYTES].data = dir_size;
311 dir_objects[DT_BYTES].size = sizeof(*dir_size);
312 dir_objects[DT_FILES].data = dir_files;
313 dir_objects[DT_FILES].size = sizeof(*dir_files);
314 return osl(osl_add_row(dir_table, dir_objects));
315 }
316
317 static uint32_t num_uids;
318
319 static int open_user_table(struct user_info *ui, int create)
320 {
321 int ret;
322
323 ui->desc = adu_malloc(sizeof(*ui->desc));
324 ui->desc->num_columns = NUM_UT_COLUMNS;
325 ui->desc->flags = 0;
326 ui->desc->column_descriptions = user_table_cols;
327 ui->desc->dir = adu_strdup(conf.database_dir_arg);
328 ui->desc->name = make_message("%u", (unsigned)ui->uid);
329 INFO_LOG(".............................uid #%u: %u\n",
330 (unsigned)num_uids, (unsigned)ui->uid);
331 if (create) {
332 ret = osl(osl_create_table(ui->desc));
333 if (ret < 0)
334 goto err;
335 num_uids++;
336 }
337 ret = osl(osl_open_table(ui->desc, &ui->table));
338 if (ret < 0)
339 goto err;
340 return 1;
341 err:
342 free((char *)ui->desc->name);
343 free((char *)ui->desc->dir);
344 free(ui->desc);
345 ui->desc->name = NULL;
346 ui->desc->dir = NULL;
347 ui->desc = NULL;
348 ui->table = NULL;
349 ui->flags = 0;
350 return ret;
351 }
352
353 #define uid_hash_bits 8
354 static uint32_t uid_hash_table_size = 1 << uid_hash_bits;
355 #define PRIME1 0x811c9dc5
356 #define PRIME2 0x01000193
357
358 static void create_hash_table(void)
359 {
360 uid_hash_table = adu_calloc(uid_hash_table_size
361 * sizeof(struct user_info));
362 }
363
364 static void free_hash_table(void)
365 {
366 free(uid_hash_table);
367 uid_hash_table = NULL;
368 }
369
370 static int create_tables(void)
371 {
372 int ret;
373
374 dir_table_desc.dir = adu_strdup(conf.database_dir_arg);
375 ret = osl(osl_create_table(&dir_table_desc));
376 if (ret < 0)
377 return ret;
378 create_hash_table();
379 return 1;
380 }
381
382 static void close_dir_table(void)
383 {
384 int ret;
385
386 if (!dir_table)
387 return;
388 ret = osl(osl_close_table(dir_table, OSL_MARK_CLEAN));
389 if (ret < 0)
390 ERROR_LOG("failed to close dir table: %s\n", adu_strerror(-ret));
391 free((char *)dir_table_desc.dir);
392 dir_table = NULL;
393 }
394
395 static void close_user_table(struct user_info *ui)
396 {
397 int ret;
398
399 if (!ui || !ui_used(ui) || !ui_admissible(ui))
400 return;
401 ret = osl(osl_close_table(ui->table, OSL_MARK_CLEAN));
402 if (ret < 0)
403 ERROR_LOG("failed to close user table %u: %s\n",
404 (unsigned) ui->uid, adu_strerror(-ret));
405 free((char *)ui->desc->name);
406 ui->desc->name = NULL;
407 free((char *)ui->desc->dir);
408 ui->desc->dir = NULL;
409 free(ui->desc);
410 ui->desc = NULL;
411 ui->table = NULL;
412 ui->flags = 0;
413 }
414
415 static void close_user_tables(void)
416 {
417 struct user_info *ui;
418
419 FOR_EACH_USER(ui)
420 close_user_table(ui);
421 }
422
423 static void close_all_tables(void)
424 {
425 close_dir_table();
426 close_user_tables();
427 free_hash_table();
428 }
429
430 static int signum;
431
432 static void signal_handler(int s)
433 {
434 signum = s;
435 }
436
437 static void check_signals(void)
438 {
439 if (likely(!signum))
440 return;
441 EMERG_LOG("caught signal %d\n", signum);
442 close_all_tables();
443 exit(EXIT_FAILURE);
444 }
445
446 static int init_signals(void)
447 {
448 if (signal(SIGINT, &signal_handler) == SIG_ERR)
449 return -E_SIGNAL_SIG_ERR;
450 if (signal(SIGTERM, &signal_handler) == SIG_ERR)
451 return -E_SIGNAL_SIG_ERR;
452 return 1;
453 }
454
455 /*
456 * We use a hash table of size s=2^uid_hash_bits to map the uids into the
457 * interval [0..s]. Hash collisions are treated by open addressing, i.e.
458 * unused slots in the table are used to store different uids that hash to the
459 * same slot.
460 *
461 * If a hash collision occurs, different slots are successively probed in order
462 * to find an unused slot for the new uid. Probing is implemented via a second
463 * hash function that maps the uid to h=(uid * PRIME2) | 1, which is always an
464 * odd number.
465 *
466 * An odd number is sufficient to make sure each entry of the hash table gets
467 * probed for probe_num between 0 and s-1 because s is a power of two, hence
468 * the second hash value has never a common divisor with the hash table size.
469 * IOW: h is invertible in the ring [0..s].
470 */
471 static uint32_t double_hash(uint32_t uid, uint32_t probe_num)
472 {
473 return (uid * PRIME1 + ((uid * PRIME2) | 1) * probe_num)
474 % uid_hash_table_size;
475 }
476
477 enum search_uid_flags {
478 OPEN_USER_TABLE = 1,
479 CREATE_USER_TABLE = 2,
480 };
481
482 static int uid_is_admissible(uint32_t uid)
483 {
484 int i;
485
486 for (i = 0; i < conf.uid_given; i++) {
487 struct uid_range *ur = admissible_uids + i;
488
489 if (ur->low <= uid && ur->high >= uid)
490 break;
491 }
492 i = !conf.uid_given || i < conf.uid_given;
493 DEBUG_LOG("uid %u is %sadmissible\n", (unsigned)uid,
494 i? "" : "not ");
495 return i;
496 }
497
498 static int search_uid(uint32_t uid, enum search_uid_flags flags,
499 struct user_info **ui_ptr)
500 {
501 uint32_t p;
502
503 for (p = 0; p < uid_hash_table_size; p++) {
504 struct user_info *ui = uid_hash_table + double_hash(uid, p);
505
506 if (!ui_used(ui)) {
507 int ret;
508 if (!flags)
509 return -E_BAD_UID;
510 ui->uid = uid;
511 ui->flags |= UI_FL_SLOT_USED;
512 if (!uid_is_admissible(uid))
513 return 0;
514 ui->flags |= UI_FL_ADMISSIBLE;
515 ret = open_user_table(ui, flags & CREATE_USER_TABLE);
516 if (ret < 0)
517 return ret;
518
519 if (ui_ptr)
520 *ui_ptr = ui;
521 return 1;
522 }
523 if (ui->uid != uid)
524 continue;
525 if (ui_ptr)
526 *ui_ptr = ui;
527 return 0;
528 }
529 return flags? -E_HASH_TABLE_OVERFLOW : -E_BAD_UID;
530 }
531
532 static int update_user_row(struct osl_table *t, uint64_t dir_num,
533 uint64_t *add)
534 {
535 struct osl_row *row;
536 struct osl_object obj = {.data = &dir_num, .size = sizeof(dir_num)};
537
538 int ret = osl(osl_get_row(t, UT_DIR_NUM, &obj, &row));
539
540 if (ret == -E_OSL && osl_errno != E_OSL_RB_KEY_NOT_FOUND)
541 return ret;
542 if (ret < 0) { /* this is the first file we add */
543 struct osl_object objects[NUM_UT_COLUMNS];
544 uint64_t num_files = 1;
545
546 objects[UT_DIR_NUM].data = &dir_num;
547 objects[UT_DIR_NUM].size = sizeof(dir_num);
548 objects[UT_BYTES].data = add;
549 objects[UT_BYTES].size = sizeof(*add);
550 objects[UT_FILES].data = &num_files;
551 objects[UT_FILES].size = sizeof(num_files);
552 INFO_LOG("######################### ret: %d\n", ret);
553 ret = osl(osl_add_row(t, objects));
554 INFO_LOG("######################### ret: %d\n", ret);
555 return ret;
556 } else { /* add size and increment file count */
557 uint64_t num;
558 struct osl_object obj1, obj2 = {.data = &num, .size = sizeof(num)};
559
560 ret = osl(osl_get_object(t, row, UT_BYTES, &obj1));
561 if (ret < 0)
562 return ret;
563 num = *(uint64_t *)obj1.data + *add;
564 ret = osl(osl_update_object(t, row, UT_BYTES, &obj2));
565 if (ret < 0)
566 return ret;
567 ret = osl(osl_get_object(t, row, UT_FILES, &obj1));
568 if (ret < 0)
569 return ret;
570 num = *(uint64_t *)obj1.data + 1;
571 return osl(osl_update_object(t, row, UT_FILES, &obj2));
572 }
573 }
574
575 static uint64_t num_dirs;
576 static uint64_t num_files;
577 static uint64_t num_bytes;
578
579 /* id of the device containing the base dir. */
580 static dev_t device_id;
581
582 static int scan_dir(char *dirname, uint64_t *parent_dir_num)
583 {
584 DIR *dir;
585 struct dirent *entry;
586 int ret, cwd_fd, ret2;
587 uint64_t dir_size = 0, dir_files = 0;
588 uint64_t this_dir_num = ++num_dirs;
589
590 check_signals();
591 DEBUG_LOG("----------------- %llu: %s\n", (long long unsigned)num_dirs, dirname);
592 ret = adu_opendir(dirname, &dir, &cwd_fd);
593 if (ret < 0) {
594 if (ret != -ERRNO_TO_ERROR(EACCES))
595 return ret;
596 WARNING_LOG("permission denied for %s\n", dirname);
597 return 1;
598 }
599 while ((entry = readdir(dir))) {
600 mode_t m;
601 struct stat s;
602 uint32_t uid;
603 uint64_t size;
604 struct user_info *ui;
605
606 if (!strcmp(entry->d_name, "."))
607 continue;
608 if (!strcmp(entry->d_name, ".."))
609 continue;
610 if (lstat(entry->d_name, &s) == -1) {
611 WARNING_LOG("lstat error for %s/%s\n", dirname,
612 entry->d_name);
613 continue;
614 }
615 m = s.st_mode;
616 if (!S_ISREG(m) && !S_ISDIR(m))
617 continue;
618 if (S_ISDIR(m)) {
619 if (conf.one_file_system_given && s.st_dev != device_id)
620 continue;
621 ret = scan_dir(entry->d_name, &this_dir_num);
622 if (ret < 0)
623 goto out;
624 continue;
625 }
626 /* regular file */
627 size = s.st_size;
628 dir_size += size;
629 num_bytes += size;
630 dir_files++;
631 num_files++;
632 uid = s.st_uid;
633 ret = search_uid(uid, CREATE_USER_TABLE | OPEN_USER_TABLE, &ui);
634 if (ret < 0)
635 goto out;
636 ui->bytes += size;
637 ui->files++;
638 ret = update_user_row(ui->table, this_dir_num, &size);
639 if (ret < 0)
640 goto out;
641 }
642 ret = add_directory(dirname, &this_dir_num, parent_dir_num,
643 &dir_size, &dir_files);
644 out:
645 closedir(dir);
646 ret2 = adu_fchdir(cwd_fd);
647 if (ret2 < 0 && ret >= 0)
648 ret = ret2;
649 close(cwd_fd);
650 return ret;
651 }
652
653 static int get_dir_name_by_number(uint64_t *dirnum, char **name)
654 {
655 char *result = NULL, *tmp;
656 struct osl_row *row;
657 uint64_t val = *dirnum;
658 struct osl_object obj = {.data = &val, .size = sizeof(val)};
659 int ret;
660
661 again:
662 ret = osl(osl_get_row(dir_table, DT_NUM, &obj, &row));
663 if (ret < 0)
664 goto out;
665 ret = osl(osl_get_object(dir_table, row, DT_NAME, &obj));
666 if (ret < 0)
667 goto out;
668 if (result) {
669 tmp = make_message("%s/%s", (char *)obj.data, result);
670 free(result);
671 result = tmp;
672 } else
673 result = adu_strdup((char *)obj.data);
674 ret = osl(osl_get_object(dir_table, row, DT_PARENT_NUM, &obj));
675 if (ret < 0)
676 goto out;
677 val = *(uint64_t *)obj.data;
678 if (val)
679 goto again;
680 out:
681 if (ret < 0) {
682 free(result);
683 *name = NULL;
684 } else
685 *name = result;
686 return ret;
687 }
688
689 static int get_dir_name_of_row(struct osl_row *dir_table_row, char **name)
690 {
691 struct osl_object obj;
692 int ret;
693 char *this_dir, *prefix = NULL;
694
695 *name = NULL;
696 ret = osl(osl_get_object(dir_table, dir_table_row, DT_NAME, &obj));
697 if (ret < 0)
698 return ret;
699 this_dir = adu_strdup((char *)obj.data);
700 ret = osl(osl_get_object(dir_table, dir_table_row, DT_PARENT_NUM, &obj));
701 if (ret < 0)
702 goto out;
703 if (!*(uint64_t *)obj.data) {
704 *name = this_dir;
705 return 1;
706 }
707 ret = get_dir_name_by_number((uint64_t *)obj.data, &prefix);
708 if (ret < 0)
709 goto out;
710 *name = make_message("%s/%s", prefix, this_dir);
711 free(prefix);
712 ret = 1;
713 out:
714 free(this_dir);
715 return ret;
716 }
717
718 const uint64_t size_unit_divisors[] = {
719 [size_unit_arg_b] = 1ULL,
720 [size_unit_arg_k] = 1024ULL,
721 [size_unit_arg_m] = 1024ULL * 1024ULL,
722 [size_unit_arg_g] = 1024ULL * 1024ULL * 1024ULL,
723 [size_unit_arg_t] = 1024ULL * 1024ULL * 1024ULL * 1024ULL,
724 };
725
726 const uint64_t count_unit_divisors[] = {
727
728 [count_unit_arg_n] = 1ULL,
729 [count_unit_arg_k] = 1000ULL,
730 [count_unit_arg_m] = 1000ULL * 1000ULL,
731 [count_unit_arg_g] = 1000ULL * 1000ULL * 1000ULL,
732 [count_unit_arg_t] = 1000ULL * 1000ULL * 1000ULL * 1000ULL,
733 };
734
735 const char size_unit_abbrevs[] = " BKMGT";
736 const char count_unit_abbrevs[] = " kmgt";
737
738 static enum enum_size_unit format_size_value(enum enum_size_unit unit,
739 uint64_t value, int print_unit, char *result)
740 {
741 enum enum_size_unit u = unit;
742 char unit_buf[2] = "\0\0";
743
744 if (unit == size_unit_arg_h) /* human readable */
745 for (u = size_unit_arg_b; u < size_unit_arg_t &&
746 value > size_unit_divisors[u + 1]; u++)
747 ; /* nothing */
748 if (print_unit)
749 unit_buf[0] = size_unit_abbrevs[u];
750 sprintf(result, "%llu%s",
751 (long long unsigned)value / size_unit_divisors[u], unit_buf);
752 return u;
753 }
754
755 static enum enum_count_unit format_count_value(enum enum_count_unit unit,
756 uint64_t value, int print_unit, char *result)
757 {
758 enum enum_count_unit u = unit;
759 char unit_buf[2] = "\0\0";
760
761 if (unit == count_unit_arg_h) /* human readable */
762 for (u = count_unit_arg_n; u < count_unit_arg_t &&
763 value > count_unit_divisors[u + 1]; u++)
764 ; /* nothing */
765 if (print_unit)
766 unit_buf[0] = count_unit_abbrevs[u];
767 sprintf(result, "%llu%s",
768 (long long unsigned)value / count_unit_divisors[u], unit_buf);
769 return u;
770 }
771
772 enum global_stats_flags {
773 GSF_PRINT_DIRNAME = 1,
774 GSF_PRINT_BYTES = 2,
775 GSF_PRINT_FILES = 4,
776 GSF_COMPUTE_SUMMARY = 8,
777 };
778
779 struct global_stats_info {
780 uint32_t count;
781 int ret;
782 int osl_errno;
783 enum global_stats_flags flags;
784 };
785
786 static int global_stats_loop_function(struct osl_row *row, void *data)
787 {
788 struct global_stats_info *gsi = data;
789 struct osl_object obj;
790 char *dirname, formated_value[FORMATED_VALUE_SIZE];
791 int ret, summary = gsi->flags & GSF_COMPUTE_SUMMARY;
792
793 check_signals();
794 if (!gsi->count && !summary) {
795 ret = -E_LOOP_COMPLETE;
796 goto err;
797 }
798 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_FILES))) {
799 uint64_t files;
800 ret = osl(osl_get_object(dir_table, row, DT_FILES, &obj));
801 if (ret < 0)
802 goto err;
803 files = *(uint64_t *)obj.data;
804 if (gsi->count && (gsi->flags & GSF_PRINT_FILES)) {
805 format_count_value(conf.count_unit_arg, files,
806 conf.count_unit_arg == count_unit_arg_h,
807 formated_value);
808 printf("\t%s%s", formated_value,
809 (gsi->flags & (GSF_PRINT_BYTES | GSF_PRINT_DIRNAME))?
810 "\t" : "\n");
811 }
812 if (summary)
813 num_files += files;
814 }
815 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_BYTES))) {
816 uint64_t bytes;
817 ret = osl(osl_get_object(dir_table, row, DT_BYTES, &obj));
818 if (ret < 0)
819 goto err;
820 bytes = *(uint64_t *)obj.data;
821 if (gsi->count && (gsi->flags & GSF_PRINT_BYTES)) {
822 format_size_value(conf.size_unit_arg, bytes,
823 conf.size_unit_arg == size_unit_arg_h,
824 formated_value);
825 printf("%s%s%s",
826 (gsi->flags & GSF_PRINT_FILES)? "" : "\t",
827 formated_value,
828 (gsi->flags & GSF_PRINT_DIRNAME)? "\t" : "\n"
829 );
830 }
831 if (summary) {
832 num_bytes += bytes;
833 num_dirs++;
834 }
835 }
836 if (gsi->count && (gsi->flags & GSF_PRINT_DIRNAME)) {
837 ret = get_dir_name_of_row(row, &dirname);
838 if (ret < 0)
839 goto err;
840 printf("%s%s\n",
841 (gsi->flags & (GSF_PRINT_BYTES | GSF_PRINT_FILES))? "" : "\t",
842 dirname);
843 free(dirname);
844 }
845 if (gsi->count > 0)
846 gsi->count--;
847 return 1;
848 err:
849 gsi->ret = ret;
850 gsi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
851 return -1;
852 }
853
854 static void print_id_stats(void)
855 {
856 struct user_info *ui;
857
858 printf("User summary "
859 "(uid/dirs%s/files%s/size%s):\n",
860 count_unit_buf, count_unit_buf, size_unit_buf);
861 FOR_EACH_USER(ui) {
862 char formated_dir_count[FORMATED_VALUE_SIZE],
863 formated_file_count[FORMATED_VALUE_SIZE],
864 formated_bytes[FORMATED_VALUE_SIZE ];
865 if (!ui_used(ui) || !ui_admissible(ui))
866 continue;
867 format_count_value(conf.count_unit_arg, ui->dirs,
868 conf.count_unit_arg == count_unit_arg_h,
869 formated_dir_count);
870 format_count_value(conf.count_unit_arg, ui->files,
871 conf.count_unit_arg == count_unit_arg_h,
872 formated_file_count);
873 format_size_value(conf.size_unit_arg, ui->bytes,
874 conf.size_unit_arg == size_unit_arg_h,
875 formated_bytes);
876 printf("\t%u\t%s\t%s\t%s\n", (unsigned)ui->uid,
877 formated_dir_count,
878 formated_file_count,
879 formated_bytes
880 );
881 }
882 }
883
884 enum user_stats_flags {
885 USF_PRINT_DIRNAME = 1,
886 USF_PRINT_BYTES = 2,
887 USF_PRINT_FILES = 4,
888 USF_COMPUTE_SUMMARY = 8,
889 };
890
891 struct user_stats_info {
892 uint32_t count;
893 enum user_stats_flags flags;
894 int ret;
895 int osl_errno;
896 struct user_info *ui;
897 };
898
899 static int user_stats_loop_function(struct osl_row *row, void *data)
900 {
901 struct user_stats_info *usi = data;
902 struct osl_object obj;
903 int ret, summary = usi->flags & GSF_COMPUTE_SUMMARY;
904 char formated_value[FORMATED_VALUE_SIZE];
905
906 check_signals();
907 if (!usi->count && !summary) {
908 ret = -E_LOOP_COMPLETE;
909 goto err;
910 }
911 if (summary || (usi->count && (usi->flags & USF_PRINT_FILES))) {
912 uint64_t files;
913 ret = osl(osl_get_object(usi->ui->table, row, UT_FILES, &obj));
914 if (ret < 0)
915 goto err;
916 files = *(uint64_t *)obj.data;
917 if (usi->count && (usi->flags & USF_PRINT_FILES)) {
918 format_count_value(conf.count_unit_arg, files,
919 conf.count_unit_arg == count_unit_arg_h,
920 formated_value);
921 printf("\t%s%s", formated_value,
922 (usi->flags & (USF_PRINT_BYTES | USF_PRINT_DIRNAME))?
923 "\t" : "\n"
924 );
925 }
926 if (summary)
927 usi->ui->files += files;
928 }
929 if (summary || (usi->count && (usi->flags & USF_PRINT_BYTES))) {
930 uint64_t bytes;
931 ret = osl(osl_get_object(usi->ui->table, row, UT_BYTES, &obj));
932 if (ret < 0)
933 goto err;
934 bytes = *(uint64_t *)obj.data;
935 if (usi->count && (usi->flags & USF_PRINT_BYTES)) {
936 format_size_value(conf.size_unit_arg, bytes,
937 conf.size_unit_arg == size_unit_arg_h,
938 formated_value);
939 printf("%s%s%s",
940 (usi->flags & USF_PRINT_FILES)? "" : "\t",
941 formated_value,
942 usi->flags & USF_PRINT_DIRNAME? "\t" : "\n"
943 );
944 }
945 if (summary) {
946 usi->ui->bytes += bytes;
947 usi->ui->dirs++;
948 }
949
950 }
951 if (usi->count && (usi->flags & USF_PRINT_DIRNAME)) {
952 char *dirname;
953 ret = osl(osl_get_object(usi->ui->table, row, UT_DIR_NUM, &obj));
954 if (ret < 0)
955 goto err;
956 ret = get_dir_name_by_number((uint64_t *)obj.data, &dirname);
957 if (ret < 0)
958 goto err;
959 printf("%s%s\n",
960 (usi->flags & (USF_PRINT_BYTES | USF_PRINT_FILES))? "" : "\t",
961 dirname);
962 free(dirname);
963 }
964 if (usi->count > 0)
965 usi->count--;
966 return 1;
967 err:
968 usi->ret = ret;
969 usi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
970 return -1;
971 }
972
973 static int check_loop_return(int ret, int loop_ret, int loop_osl_errno)
974 {
975 if (ret >= 0)
976 return ret;
977 assert(ret == -E_OSL);
978 if (osl_errno != E_OSL_LOOP)
979 /* error not caused by loop function returning negative. */
980 return ret;
981 assert(loop_ret < 0);
982 if (loop_ret == -E_LOOP_COMPLETE) /* no error */
983 return 1;
984 if (loop_ret == -E_OSL) { /* osl error in loop function */
985 assert(loop_osl_errno);
986 osl_errno = loop_osl_errno;
987 }
988 return loop_ret;
989 }
990
991 static int adu_loop_reverse(struct osl_table *t, unsigned col_num, void *private_data,
992 osl_rbtree_loop_func *func, int *loop_ret, int *loop_osl_errno)
993 {
994 int ret = osl(osl_rbtree_loop_reverse(t, col_num, private_data, func));
995 return check_loop_return(ret, *loop_ret, *loop_osl_errno);
996 }
997
998 static int print_user_stats(void)
999 {
1000 struct user_info *ui;
1001 int ret;
1002
1003 FOR_EACH_USER(ui) {
1004 struct user_stats_info usi = {
1005 .count = conf.limit_arg,
1006 .ui = ui
1007 };
1008 if (!ui_used(ui) || !ui_admissible(ui))
1009 continue;
1010 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_BYTES | USF_COMPUTE_SUMMARY;
1011 printf("uid %u, by size%s:\n",
1012 (unsigned) ui->uid, size_unit_buf);
1013 ret = adu_loop_reverse(ui->table, UT_BYTES, &usi, user_stats_loop_function,
1014 &usi.ret, &usi.osl_errno);
1015 if (ret < 0)
1016 return ret;
1017 printf("\nuid %u, by file count%s:\n",
1018 (unsigned) ui->uid, count_unit_buf);
1019 usi.count = conf.limit_arg,
1020 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_FILES;
1021 ret = adu_loop_reverse(ui->table, UT_FILES, &usi, user_stats_loop_function,
1022 &usi.ret, &usi.osl_errno);
1023 if (ret < 0)
1024 return ret;
1025 printf("\n");
1026 }
1027 return 1;
1028 }
1029
1030 static void print_global_summary(void)
1031 {
1032 char d[FORMATED_VALUE_SIZE], f[FORMATED_VALUE_SIZE],
1033 s[FORMATED_VALUE_SIZE];
1034 enum enum_count_unit ud, uf;
1035 enum enum_size_unit us;
1036
1037 ud = format_count_value(conf.count_unit_arg, num_dirs, 0, d);
1038 uf = format_count_value(conf.count_unit_arg, num_files, 0, f);
1039 us = format_size_value(conf.size_unit_arg, num_bytes, 0, s);
1040
1041 printf("Global summary "
1042 "(dirs(%c)/files(%c)/size(%c))\n"
1043 "\t%s\t%s\t%s\n\n",
1044 count_unit_abbrevs[ud],
1045 count_unit_abbrevs[uf],
1046 size_unit_abbrevs[us],
1047 d, f, s
1048 );
1049
1050 }
1051
1052 static int print_statistics(void)
1053 {
1054 int ret;
1055 struct global_stats_info gsi = {
1056 .count = conf.limit_arg,
1057 .flags = GSF_PRINT_DIRNAME | GSF_PRINT_BYTES | GSF_COMPUTE_SUMMARY
1058 };
1059
1060 printf("By size%s:\n",
1061 size_unit_buf);
1062 ret = adu_loop_reverse(dir_table, DT_BYTES, &gsi,
1063 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
1064 if (ret < 0)
1065 return ret;
1066 printf("\n");
1067
1068 gsi.count = conf.limit_arg;
1069 gsi.flags = GSF_PRINT_DIRNAME | GSF_PRINT_FILES;
1070 printf("By file count%s:\n",
1071 count_unit_buf);
1072 ret = adu_loop_reverse(dir_table, DT_FILES, &gsi,
1073 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
1074 if (ret < 0)
1075 return ret;
1076 printf("\n");
1077 print_global_summary();
1078 print_user_stats();
1079 print_id_stats();
1080 return 1;
1081 }
1082
1083 static char *get_uid_list_name(void)
1084 {
1085 return make_message("%s/uid_list", conf.database_dir_arg);
1086 }
1087
1088 static int write_uid_list(void)
1089 {
1090 char *buf, *filename;
1091 uint32_t count = 0;
1092 struct user_info *ui;
1093 size_t size = num_uids * sizeof(uint32_t);
1094 int ret;
1095
1096 if (!num_uids)
1097 return 0;
1098 buf = adu_malloc(size);
1099 FOR_EACH_USER(ui) {
1100 if (!ui_used(ui) || !ui_admissible(ui))
1101 continue;
1102 DEBUG_LOG("saving uid %u\n", (unsigned) ui->uid);
1103 write_u32(buf + count++ * sizeof(uint32_t), ui->uid);
1104 }
1105 filename = get_uid_list_name();
1106 ret = adu_write_file(filename, buf, size);
1107 free(filename);
1108 free(buf);
1109 return ret;
1110 }
1111
1112 static int open_dir_table(void)
1113 {
1114 if (!dir_table_desc.dir) /* we did not create the table */
1115 dir_table_desc.dir = adu_strdup(conf.database_dir_arg);
1116 return osl(osl_open_table(&dir_table_desc, &dir_table));
1117 }
1118 static int com_create()
1119 {
1120 uint64_t zero = 0ULL;
1121 int ret;
1122 struct stat statbuf;
1123
1124 if (lstat(conf.base_dir_arg, &statbuf) == -1)
1125 return -ERRNO_TO_ERROR(errno);
1126 if (!S_ISDIR(statbuf.st_mode))
1127 return -ERRNO_TO_ERROR(ENOTDIR);
1128 device_id = statbuf.st_dev;
1129 ret = create_tables();
1130 if (ret < 0)
1131 return ret;
1132 check_signals();
1133 ret = open_dir_table();
1134 if (ret < 0)
1135 return ret;
1136 check_signals();
1137 ret = scan_dir(conf.base_dir_arg, &zero);
1138 if (ret < 0)
1139 goto out;
1140 ret = write_uid_list();
1141 out:
1142 close_all_tables();
1143 return ret;
1144 }
1145
1146 static int read_uid_file(void)
1147 {
1148 size_t size;
1149 uint32_t n;
1150 char *filename = get_uid_list_name(), *map;
1151 int ret = mmap_full_file(filename, O_RDONLY, (void **)&map, &size, NULL);
1152
1153 if (ret < 0) {
1154 INFO_LOG("failed to map %s\n", filename);
1155 free(filename);
1156 return ret;
1157 }
1158 num_uids = size / 4;
1159 INFO_LOG("found %u uids in %s\n", (unsigned)num_uids, filename);
1160 free(filename);
1161 /* hash table size should be a power of two and larger than the number of uids */
1162 uid_hash_table_size = 4;
1163 while (uid_hash_table_size < num_uids)
1164 uid_hash_table_size *= 2;
1165 create_hash_table();
1166 for (n = 0; n < num_uids; n++) {
1167 uint32_t uid = read_u32(map + n * sizeof(uid));
1168 ret = search_uid(uid, OPEN_USER_TABLE, NULL);
1169 if (ret < 0)
1170 goto out;
1171 }
1172 out:
1173 adu_munmap(map, size);
1174 return ret;
1175 }
1176
1177 static int com_select(void)
1178 {
1179 int ret;
1180
1181 if (conf.count_unit_arg != count_unit_arg_h)
1182 count_unit_buf[1] = count_unit_abbrevs[conf.count_unit_arg];
1183 else
1184 count_unit_buf[0] = '\0';
1185 if (conf.size_unit_arg != size_unit_arg_h)
1186 size_unit_buf[1] = size_unit_abbrevs[conf.size_unit_arg];
1187 else
1188 size_unit_buf[0] = '\0';
1189
1190 ret = open_dir_table();
1191 if (ret < 0)
1192 return ret;
1193 check_signals();
1194 ret = read_uid_file();
1195 if (ret < 0)
1196 return ret;
1197 check_signals();
1198 ret = print_statistics();
1199 close_all_tables();
1200 return ret;
1201 }
1202
1203 static int check_args(void)
1204 {
1205 int i, ret;
1206
1207 /* remove trailing slashes from base-dir arg */
1208 if (conf.base_dir_given) {
1209 size_t len = strlen(conf.base_dir_arg);
1210 for (;;) {
1211 if (!len) /* empty string */
1212 return -ERRNO_TO_ERROR(EINVAL);
1213 if (!--len) /* length 1 is always OK */
1214 break;
1215 if (conf.base_dir_arg[len] != '/')
1216 break; /* no trailing slash, also OK */
1217 conf.base_dir_arg[len] = '\0';
1218 }
1219 }
1220 if (!conf.uid_given)
1221 return 0;
1222 admissible_uids = adu_malloc(conf.uid_given * sizeof(*admissible_uids));
1223 for (i = 0; i < conf.uid_given; i++) {
1224 ret = parse_uid_range(conf.uid_arg[i], admissible_uids + i);
1225 if (ret < 0)
1226 goto err;
1227 }
1228 return 1;
1229 err:
1230 free(admissible_uids);
1231 admissible_uids = NULL;
1232 return ret;
1233 }
1234
1235 int main(int argc, char **argv)
1236 {
1237 int ret;
1238 struct cmdline_parser_params params = {
1239 .override = 0,
1240 .initialize = 1,
1241 .check_required = 0,
1242 .check_ambiguity = 0,
1243 .print_errors = 1
1244 };
1245
1246 cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
1247 ret = init_signals();
1248 if (ret < 0)
1249 goto out;
1250 ret = check_args();
1251 if (ret < 0)
1252 goto out;
1253 ret = -E_SYNTAX;
1254 if (conf.select_given)
1255 ret = com_select();
1256 else
1257 ret = com_create();
1258 if (ret < 0)
1259 goto out;
1260 out:
1261 free(admissible_uids);
1262 if (ret < 0) {
1263 ERROR_LOG("%s\n", adu_strerror(-ret));
1264 return -EXIT_FAILURE;
1265 }
1266 return EXIT_SUCCESS;
1267 }