Kill a bunch of unused cruft.
[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 = para_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 = para_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 = para_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 = para_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 = para_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 = para_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 int scan_dir(char *dirname, uint64_t *parent_dir_num)
580 {
581 DIR *dir;
582 struct dirent *entry;
583 int ret, cwd_fd, ret2;
584 uint64_t dir_size = 0, dir_files = 0;
585 uint64_t this_dir_num = ++num_dirs;
586
587 check_signals();
588 DEBUG_LOG("----------------- %llu: %s\n", (long long unsigned)num_dirs, dirname);
589 ret = para_opendir(dirname, &dir, &cwd_fd);
590 if (ret < 0) {
591 if (ret != -ERRNO_TO_ERROR(EACCES))
592 return ret;
593 WARNING_LOG("permission denied for %s\n", dirname);
594 return 1;
595 }
596 while ((entry = readdir(dir))) {
597 mode_t m;
598 struct stat s;
599 uint32_t uid;
600 uint64_t size;
601 struct user_info *ui;
602
603 if (!strcmp(entry->d_name, "."))
604 continue;
605 if (!strcmp(entry->d_name, ".."))
606 continue;
607 if (lstat(entry->d_name, &s) == -1) {
608 WARNING_LOG("lstat error for %s/%s\n", dirname,
609 entry->d_name);
610 continue;
611 }
612 m = s.st_mode;
613 if (!S_ISREG(m) && !S_ISDIR(m))
614 continue;
615 if (S_ISDIR(m)) {
616 ret = scan_dir(entry->d_name, &this_dir_num);
617 if (ret < 0)
618 goto out;
619 continue;
620 }
621 /* regular file */
622 size = s.st_size;
623 dir_size += size;
624 num_bytes += size;
625 dir_files++;
626 num_files++;
627 uid = s.st_uid;
628 ret = search_uid(uid, CREATE_USER_TABLE | OPEN_USER_TABLE, &ui);
629 if (ret < 0)
630 goto out;
631 ui->bytes += size;
632 ui->files++;
633 ret = update_user_row(ui->table, this_dir_num, &size);
634 if (ret < 0)
635 goto out;
636 }
637 ret = add_directory(dirname, &this_dir_num, parent_dir_num,
638 &dir_size, &dir_files);
639 out:
640 closedir(dir);
641 ret2 = para_fchdir(cwd_fd);
642 if (ret2 < 0 && ret >= 0)
643 ret = ret2;
644 close(cwd_fd);
645 return ret;
646 }
647
648 static int get_dir_name_by_number(uint64_t *dirnum, char **name)
649 {
650 char *result = NULL, *tmp;
651 struct osl_row *row;
652 uint64_t val = *dirnum;
653 struct osl_object obj = {.data = &val, .size = sizeof(val)};
654 int ret;
655
656 again:
657 ret = osl(osl_get_row(dir_table, DT_NUM, &obj, &row));
658 if (ret < 0)
659 goto out;
660 ret = osl(osl_get_object(dir_table, row, DT_NAME, &obj));
661 if (ret < 0)
662 goto out;
663 if (result) {
664 tmp = make_message("%s/%s", (char *)obj.data, result);
665 free(result);
666 result = tmp;
667 } else
668 result = para_strdup((char *)obj.data);
669 ret = osl(osl_get_object(dir_table, row, DT_PARENT_NUM, &obj));
670 if (ret < 0)
671 goto out;
672 val = *(uint64_t *)obj.data;
673 if (val)
674 goto again;
675 out:
676 if (ret < 0) {
677 free(result);
678 *name = NULL;
679 } else
680 *name = result;
681 return ret;
682 }
683
684 static int get_dir_name_of_row(struct osl_row *dir_table_row, char **name)
685 {
686 struct osl_object obj;
687 int ret;
688 char *this_dir, *prefix = NULL;
689
690 *name = NULL;
691 ret = osl(osl_get_object(dir_table, dir_table_row, DT_NAME, &obj));
692 if (ret < 0)
693 return ret;
694 this_dir = para_strdup((char *)obj.data);
695 ret = osl(osl_get_object(dir_table, dir_table_row, DT_PARENT_NUM, &obj));
696 if (ret < 0)
697 goto out;
698 if (!*(uint64_t *)obj.data) {
699 *name = this_dir;
700 return 1;
701 }
702 ret = get_dir_name_by_number((uint64_t *)obj.data, &prefix);
703 if (ret < 0)
704 goto out;
705 *name = make_message("%s/%s", prefix, this_dir);
706 free(prefix);
707 ret = 1;
708 out:
709 free(this_dir);
710 return ret;
711 }
712
713 const uint64_t size_unit_divisors[] = {
714 [size_unit_arg_b] = 1ULL,
715 [size_unit_arg_k] = 1024ULL,
716 [size_unit_arg_m] = 1024ULL * 1024ULL,
717 [size_unit_arg_g] = 1024ULL * 1024ULL * 1024ULL,
718 [size_unit_arg_t] = 1024ULL * 1024ULL * 1024ULL * 1024ULL,
719 };
720
721 const uint64_t count_unit_divisors[] = {
722
723 [count_unit_arg_n] = 1ULL,
724 [count_unit_arg_k] = 1000ULL,
725 [count_unit_arg_m] = 1000ULL * 1000ULL,
726 [count_unit_arg_g] = 1000ULL * 1000ULL * 1000ULL,
727 [count_unit_arg_t] = 1000ULL * 1000ULL * 1000ULL * 1000ULL,
728 };
729
730 const char size_unit_abbrevs[] = " BKMGT";
731 const char count_unit_abbrevs[] = " kmgt";
732
733 static enum enum_size_unit format_size_value(enum enum_size_unit unit,
734 uint64_t value, int print_unit, char *result)
735 {
736 enum enum_size_unit u = unit;
737 char unit_buf[2] = "\0\0";
738
739 if (unit == size_unit_arg_h) /* human readable */
740 for (u = size_unit_arg_b; u < size_unit_arg_t &&
741 value > size_unit_divisors[u + 1]; u++)
742 ; /* nothing */
743 if (print_unit)
744 unit_buf[0] = size_unit_abbrevs[u];
745 sprintf(result, "%llu%s",
746 (long long unsigned)value / size_unit_divisors[u], unit_buf);
747 return u;
748 }
749
750 static enum enum_count_unit format_count_value(enum enum_count_unit unit,
751 uint64_t value, int print_unit, char *result)
752 {
753 enum enum_count_unit u = unit;
754 char unit_buf[2] = "\0\0";
755
756 if (unit == count_unit_arg_h) /* human readable */
757 for (u = count_unit_arg_n; u < count_unit_arg_t &&
758 value > count_unit_divisors[u + 1]; u++)
759 ; /* nothing */
760 if (print_unit)
761 unit_buf[0] = count_unit_abbrevs[u];
762 sprintf(result, "%llu%s",
763 (long long unsigned)value / count_unit_divisors[u], unit_buf);
764 return u;
765 }
766
767 enum global_stats_flags {
768 GSF_PRINT_DIRNAME = 1,
769 GSF_PRINT_BYTES = 2,
770 GSF_PRINT_FILES = 4,
771 GSF_COMPUTE_SUMMARY = 8,
772 };
773
774 struct global_stats_info {
775 uint32_t count;
776 int ret;
777 int osl_errno;
778 enum global_stats_flags flags;
779 };
780
781 static int global_stats_loop_function(struct osl_row *row, void *data)
782 {
783 struct global_stats_info *gsi = data;
784 struct osl_object obj;
785 char *dirname, formated_value[FORMATED_VALUE_SIZE];
786 int ret, summary = gsi->flags & GSF_COMPUTE_SUMMARY;
787
788 check_signals();
789 if (!gsi->count && !summary) {
790 ret = -E_LOOP_COMPLETE;
791 goto err;
792 }
793 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_FILES))) {
794 uint64_t files;
795 ret = osl(osl_get_object(dir_table, row, DT_FILES, &obj));
796 if (ret < 0)
797 goto err;
798 files = *(uint64_t *)obj.data;
799 if (gsi->count && (gsi->flags & GSF_PRINT_FILES)) {
800 format_count_value(conf.count_unit_arg, files,
801 conf.count_unit_arg == count_unit_arg_h,
802 formated_value);
803 printf("\t%s%s", formated_value,
804 (gsi->flags & (GSF_PRINT_BYTES | GSF_PRINT_DIRNAME))?
805 "\t" : "\n");
806 }
807 if (summary)
808 num_files += files;
809 }
810 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_BYTES))) {
811 uint64_t bytes;
812 ret = osl(osl_get_object(dir_table, row, DT_BYTES, &obj));
813 if (ret < 0)
814 goto err;
815 bytes = *(uint64_t *)obj.data;
816 if (gsi->count && (gsi->flags & GSF_PRINT_BYTES)) {
817 format_size_value(conf.size_unit_arg, bytes,
818 conf.size_unit_arg == size_unit_arg_h,
819 formated_value);
820 printf("%s%s%s",
821 (gsi->flags & GSF_PRINT_FILES)? "" : "\t",
822 formated_value,
823 (gsi->flags & GSF_PRINT_DIRNAME)? "\t" : "\n"
824 );
825 }
826 if (summary) {
827 num_bytes += bytes;
828 num_dirs++;
829 }
830 }
831 if (gsi->count && (gsi->flags & GSF_PRINT_DIRNAME)) {
832 ret = get_dir_name_of_row(row, &dirname);
833 if (ret < 0)
834 goto err;
835 printf("%s%s\n",
836 (gsi->flags & (GSF_PRINT_BYTES | GSF_PRINT_FILES))? "" : "\t",
837 dirname);
838 free(dirname);
839 }
840 if (gsi->count > 0)
841 gsi->count--;
842 return 1;
843 err:
844 gsi->ret = ret;
845 gsi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
846 return -1;
847 }
848
849 static void print_id_stats(void)
850 {
851 struct user_info *ui;
852
853 printf("User summary "
854 "(uid/dirs%s/files%s/size%s):\n",
855 count_unit_buf, count_unit_buf, size_unit_buf);
856 FOR_EACH_USER(ui) {
857 char formated_dir_count[FORMATED_VALUE_SIZE],
858 formated_file_count[FORMATED_VALUE_SIZE],
859 formated_bytes[FORMATED_VALUE_SIZE ];
860 if (!ui_used(ui) || !ui_admissible(ui))
861 continue;
862 format_count_value(conf.count_unit_arg, ui->dirs,
863 conf.count_unit_arg == count_unit_arg_h,
864 formated_dir_count);
865 format_count_value(conf.count_unit_arg, ui->files,
866 conf.count_unit_arg == count_unit_arg_h,
867 formated_file_count);
868 format_size_value(conf.size_unit_arg, ui->bytes,
869 conf.size_unit_arg == size_unit_arg_h,
870 formated_bytes);
871 printf("\t%u\t%s\t%s\t%s\n", (unsigned)ui->uid,
872 formated_dir_count,
873 formated_file_count,
874 formated_bytes
875 );
876 }
877 }
878
879 enum user_stats_flags {
880 USF_PRINT_DIRNAME = 1,
881 USF_PRINT_BYTES = 2,
882 USF_PRINT_FILES = 4,
883 USF_COMPUTE_SUMMARY = 8,
884 };
885
886 struct user_stats_info {
887 uint32_t count;
888 enum user_stats_flags flags;
889 int ret;
890 int osl_errno;
891 struct user_info *ui;
892 };
893
894 static int user_stats_loop_function(struct osl_row *row, void *data)
895 {
896 struct user_stats_info *usi = data;
897 struct osl_object obj;
898 int ret, summary = usi->flags & GSF_COMPUTE_SUMMARY;
899 char formated_value[FORMATED_VALUE_SIZE];
900
901 check_signals();
902 if (!usi->count && !summary) {
903 ret = -E_LOOP_COMPLETE;
904 goto err;
905 }
906 if (summary || (usi->count && (usi->flags & USF_PRINT_FILES))) {
907 uint64_t files;
908 ret = osl(osl_get_object(usi->ui->table, row, UT_FILES, &obj));
909 if (ret < 0)
910 goto err;
911 files = *(uint64_t *)obj.data;
912 if (usi->count && (usi->flags & USF_PRINT_FILES)) {
913 format_count_value(conf.count_unit_arg, files,
914 conf.count_unit_arg == count_unit_arg_h,
915 formated_value);
916 printf("\t%s%s", formated_value,
917 (usi->flags & (USF_PRINT_BYTES | USF_PRINT_DIRNAME))?
918 "\t" : "\n"
919 );
920 }
921 if (summary)
922 usi->ui->files += files;
923 }
924 if (summary || (usi->count && (usi->flags & USF_PRINT_BYTES))) {
925 uint64_t bytes;
926 ret = osl(osl_get_object(usi->ui->table, row, UT_BYTES, &obj));
927 if (ret < 0)
928 goto err;
929 bytes = *(uint64_t *)obj.data;
930 if (usi->count && (usi->flags & USF_PRINT_BYTES)) {
931 format_size_value(conf.size_unit_arg, bytes,
932 conf.size_unit_arg == size_unit_arg_h,
933 formated_value);
934 printf("%s%s%s",
935 (usi->flags & USF_PRINT_FILES)? "" : "\t",
936 formated_value,
937 usi->flags & USF_PRINT_DIRNAME? "\t" : "\n"
938 );
939 }
940 if (summary) {
941 usi->ui->bytes += bytes;
942 usi->ui->dirs++;
943 }
944
945 }
946 if (usi->count && (usi->flags & USF_PRINT_DIRNAME)) {
947 char *dirname;
948 ret = osl(osl_get_object(usi->ui->table, row, UT_DIR_NUM, &obj));
949 if (ret < 0)
950 goto err;
951 ret = get_dir_name_by_number((uint64_t *)obj.data, &dirname);
952 if (ret < 0)
953 goto err;
954 printf("%s%s\n",
955 (usi->flags & (USF_PRINT_BYTES | USF_PRINT_FILES))? "" : "\t",
956 dirname);
957 free(dirname);
958 }
959 if (usi->count > 0)
960 usi->count--;
961 return 1;
962 err:
963 usi->ret = ret;
964 usi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
965 return -1;
966 }
967
968 static int check_loop_return(int ret, int loop_ret, int loop_osl_errno)
969 {
970 if (ret >= 0)
971 return ret;
972 assert(ret == -E_OSL);
973 if (osl_errno != E_OSL_LOOP)
974 /* error not caused by loop function returning negative. */
975 return ret;
976 assert(loop_ret < 0);
977 if (loop_ret == -E_LOOP_COMPLETE) /* no error */
978 return 1;
979 if (loop_ret == -E_OSL) { /* osl error in loop function */
980 assert(loop_osl_errno);
981 osl_errno = loop_osl_errno;
982 }
983 return loop_ret;
984 }
985
986 static int adu_loop_reverse(struct osl_table *t, unsigned col_num, void *private_data,
987 osl_rbtree_loop_func *func, int *loop_ret, int *loop_osl_errno)
988 {
989 int ret = osl(osl_rbtree_loop_reverse(t, col_num, private_data, func));
990 return check_loop_return(ret, *loop_ret, *loop_osl_errno);
991 }
992
993 static int print_user_stats(void)
994 {
995 struct user_info *ui;
996 int ret;
997
998 FOR_EACH_USER(ui) {
999 struct user_stats_info usi = {
1000 .count = conf.limit_arg,
1001 .ui = ui
1002 };
1003 if (!ui_used(ui) || !ui_admissible(ui))
1004 continue;
1005 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_BYTES | USF_COMPUTE_SUMMARY;
1006 printf("uid %u, by size%s:\n",
1007 (unsigned) ui->uid, size_unit_buf);
1008 ret = adu_loop_reverse(ui->table, UT_BYTES, &usi, user_stats_loop_function,
1009 &usi.ret, &usi.osl_errno);
1010 if (ret < 0)
1011 return ret;
1012 printf("\nuid %u, by file count%s:\n",
1013 (unsigned) ui->uid, count_unit_buf);
1014 usi.count = conf.limit_arg,
1015 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_FILES;
1016 ret = adu_loop_reverse(ui->table, UT_FILES, &usi, user_stats_loop_function,
1017 &usi.ret, &usi.osl_errno);
1018 if (ret < 0)
1019 return ret;
1020 printf("\n");
1021 }
1022 return 1;
1023 }
1024
1025 static void print_global_summary(void)
1026 {
1027 char d[FORMATED_VALUE_SIZE], f[FORMATED_VALUE_SIZE],
1028 s[FORMATED_VALUE_SIZE];
1029 enum enum_count_unit ud, uf;
1030 enum enum_size_unit us;
1031
1032 ud = format_count_value(conf.count_unit_arg, num_dirs, 0, d);
1033 uf = format_count_value(conf.count_unit_arg, num_files, 0, f);
1034 us = format_size_value(conf.size_unit_arg, num_bytes, 0, s);
1035
1036 printf("Global summary "
1037 "(dirs(%c)/files(%c)/size(%c))\n"
1038 "\t%s\t%s\t%s\n\n",
1039 count_unit_abbrevs[ud],
1040 count_unit_abbrevs[uf],
1041 size_unit_abbrevs[us],
1042 d, f, s
1043 );
1044
1045 }
1046
1047 static int print_statistics(void)
1048 {
1049 int ret;
1050 struct global_stats_info gsi = {
1051 .count = conf.limit_arg,
1052 .flags = GSF_PRINT_DIRNAME | GSF_PRINT_BYTES | GSF_COMPUTE_SUMMARY
1053 };
1054
1055 printf("By size%s:\n",
1056 size_unit_buf);
1057 ret = adu_loop_reverse(dir_table, DT_BYTES, &gsi,
1058 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
1059 if (ret < 0)
1060 return ret;
1061 printf("\n");
1062
1063 gsi.count = conf.limit_arg;
1064 gsi.flags = GSF_PRINT_DIRNAME | GSF_PRINT_FILES;
1065 printf("By file count%s:\n",
1066 count_unit_buf);
1067 ret = adu_loop_reverse(dir_table, DT_FILES, &gsi,
1068 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
1069 if (ret < 0)
1070 return ret;
1071 printf("\n");
1072 print_global_summary();
1073 print_user_stats();
1074 print_id_stats();
1075 return 1;
1076 }
1077
1078 static char *get_uid_list_name(void)
1079 {
1080 return make_message("%s/uid_list", conf.database_dir_arg);
1081 }
1082
1083 static int write_uid_list(void)
1084 {
1085 char *buf, *filename;
1086 uint32_t count = 0;
1087 struct user_info *ui;
1088 size_t size = num_uids * sizeof(uint32_t);
1089 int ret;
1090
1091 if (!num_uids)
1092 return 0;
1093 buf = para_malloc(size);
1094 FOR_EACH_USER(ui) {
1095 if (!ui_used(ui) || !ui_admissible(ui))
1096 continue;
1097 DEBUG_LOG("saving uid %u\n", (unsigned) ui->uid);
1098 write_u32(buf + count++ * sizeof(uint32_t), ui->uid);
1099 }
1100 filename = get_uid_list_name();
1101 ret = para_write_file(filename, buf, size);
1102 free(filename);
1103 free(buf);
1104 return ret;
1105 }
1106
1107 static int open_dir_table(void)
1108 {
1109 if (!dir_table_desc.dir) /* we did not create the table */
1110 dir_table_desc.dir = para_strdup(conf.database_dir_arg);
1111 return osl(osl_open_table(&dir_table_desc, &dir_table));
1112 }
1113 static int com_create()
1114 {
1115 uint64_t zero = 0ULL;
1116 int ret = create_tables();
1117
1118 if (ret < 0)
1119 return ret;
1120 check_signals();
1121 ret = open_dir_table();
1122 if (ret < 0)
1123 return ret;
1124 check_signals();
1125 ret = scan_dir(conf.base_dir_arg, &zero);
1126 if (ret < 0)
1127 goto out;
1128 ret = write_uid_list();
1129 out:
1130 close_all_tables();
1131 return ret;
1132 }
1133
1134 static int read_uid_file(void)
1135 {
1136 size_t size;
1137 uint32_t n;
1138 char *filename = get_uid_list_name(), *map;
1139 int ret = mmap_full_file(filename, O_RDONLY, (void **)&map, &size, NULL);
1140
1141 if (ret < 0) {
1142 INFO_LOG("failed to map %s\n", filename);
1143 free(filename);
1144 return ret;
1145 }
1146 num_uids = size / 4;
1147 INFO_LOG("found %u uids in %s\n", (unsigned)num_uids, filename);
1148 free(filename);
1149 /* hash table size should be a power of two and larger than the number of uids */
1150 uid_hash_table_size = 4;
1151 while (uid_hash_table_size < num_uids)
1152 uid_hash_table_size *= 2;
1153 create_hash_table();
1154 for (n = 0; n < num_uids; n++) {
1155 uint32_t uid = read_u32(map + n * sizeof(uid));
1156 ret = search_uid(uid, OPEN_USER_TABLE, NULL);
1157 if (ret < 0)
1158 goto out;
1159 }
1160 out:
1161 para_munmap(map, size);
1162 return ret;
1163 }
1164
1165 static int com_select(void)
1166 {
1167 int ret;
1168
1169 if (conf.count_unit_arg != count_unit_arg_h)
1170 count_unit_buf[1] = count_unit_abbrevs[conf.count_unit_arg];
1171 else
1172 count_unit_buf[0] = '\0';
1173 if (conf.size_unit_arg != size_unit_arg_h)
1174 size_unit_buf[1] = size_unit_abbrevs[conf.size_unit_arg];
1175 else
1176 size_unit_buf[0] = '\0';
1177
1178 ret = open_dir_table();
1179 if (ret < 0)
1180 return ret;
1181 check_signals();
1182 ret = read_uid_file();
1183 if (ret < 0)
1184 return ret;
1185 check_signals();
1186 ret = print_statistics();
1187 close_all_tables();
1188 return ret;
1189 }
1190
1191 static int check_args(void)
1192 {
1193 int i, ret;
1194
1195 /* remove trailing slashes from base-dir arg */
1196 if (conf.base_dir_given) {
1197 size_t len = strlen(conf.base_dir_arg);
1198 for (;;) {
1199 if (!len) /* empty string */
1200 return -ERRNO_TO_ERROR(EINVAL);
1201 if (!--len) /* length 1 is always OK */
1202 break;
1203 if (conf.base_dir_arg[len] != '/')
1204 break; /* no trailing slash, also OK */
1205 conf.base_dir_arg[len] = '\0';
1206 }
1207 }
1208 if (!conf.uid_given)
1209 return 0;
1210 admissible_uids = para_malloc(conf.uid_given * sizeof(*admissible_uids));
1211 for (i = 0; i < conf.uid_given; i++) {
1212 ret = parse_uid_range(conf.uid_arg[i], admissible_uids + i);
1213 if (ret < 0)
1214 goto err;
1215 }
1216 return 1;
1217 err:
1218 free(admissible_uids);
1219 admissible_uids = NULL;
1220 return ret;
1221 }
1222
1223 int main(int argc, char **argv)
1224 {
1225 int ret;
1226 struct cmdline_parser_params params = {
1227 .override = 0,
1228 .initialize = 1,
1229 .check_required = 0,
1230 .check_ambiguity = 0,
1231 .print_errors = 1
1232 };
1233
1234 cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
1235 ret = init_signals();
1236 if (ret < 0)
1237 goto out;
1238 ret = check_args();
1239 if (ret < 0)
1240 goto out;
1241 ret = -E_SYNTAX;
1242 if (conf.select_given)
1243 ret = com_select();
1244 else
1245 ret = com_create();
1246 if (ret < 0)
1247 goto out;
1248 out:
1249 free(admissible_uids);
1250 if (ret < 0) {
1251 ERROR_LOG("%s\n", adu_strerror(-ret));
1252 return -EXIT_FAILURE;
1253 }
1254 return EXIT_SUCCESS;
1255 }