Fix the uid parser.
[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 /**
35 * Contains info for each user that owns at least one regular file.
36 *
37 * Even users that are not taken into account because of the --uid
38 * option occupy a slot in this hash table. This allows to find out
39 * quicky whether a uid is admissible. And yes, this has to be fast.
40 */
41 static struct user_info *uid_hash_table;
42
43 static inline int ui_used(struct user_info *ui)
44 {
45 return ui->flags & UI_FL_SLOT_USED;
46 }
47
48 static inline int ui_admissible(struct user_info *ui)
49 {
50 return ui->flags & UI_FL_ADMISSIBLE;
51 }
52
53 struct uid_range {
54 uint32_t low;
55 uint32_t high;
56 };
57
58 static struct uid_range *admissible_uids;
59
60 static inline int check_uid_arg(const char *arg, uint32_t *uid)
61 {
62 const uint32_t max = ~0U;
63 /*
64 * we need an 64-bit int for string -> uid conversion because strtoll()
65 * returns a signed value.
66 */
67 int64_t val;
68 int ret = para_atoi64(arg, &val);
69
70 if (ret < 0)
71 return ret;
72 if (val < 0 || val > max)
73 return -ERRNO_TO_ERROR(EINVAL);
74 *uid = val;
75 return 1;
76 }
77
78 static int parse_uid_range(const char *orig_arg, struct uid_range *ur)
79 {
80 int ret;
81 char *arg = para_strdup(orig_arg), *p = strchr(arg, '-');
82
83 if (!p || p == arg) { /* -42 or 42 */
84 ret = check_uid_arg(p? p + 1 : arg, &ur->high);
85 if (ret < 0)
86 goto out;
87 ur->low = p? 0 : ur->high;
88 ret = 1;
89 goto out;
90 }
91 /* 42- or 42-4711 */
92 *p = '\0';
93 p++;
94 ret = check_uid_arg(arg, &ur->low);
95 if (ret < 0)
96 goto out;
97 ur->high = ~0U;
98 if (*p) { /* 42-4711 */
99 ret = check_uid_arg(p, &ur->high);
100 if (ret < 0)
101 goto out;
102 }
103 if (ur->low > ur->high)
104 ret = -ERRNO_TO_ERROR(EINVAL);
105 out:
106 if (ret < 0)
107 ERROR_LOG("bad uid option: %s\n", orig_arg);
108 else
109 INFO_LOG("admissible uid range: %u - %u\n", ur->low,
110 ur->high);
111 free(arg);
112 return ret;
113 }
114
115
116 /** evaluates to 1 if x < y, to -1 if x > y and to 0 if x == y */
117 #define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
118
119 /**
120 * The log function.
121 *
122 * \param ll Loglevel.
123 * \param fml Usual format string.
124 *
125 * All XXX_LOG() macros use this function.
126 */
127 __printf_2_3 void __log(int ll, const char* fmt,...)
128 {
129 va_list argp;
130 FILE *outfd;
131 struct tm *tm;
132 time_t t1;
133 char str[255] = "";
134
135 if (ll < conf.loglevel_arg)
136 return;
137 outfd = stderr;
138 time(&t1);
139 tm = localtime(&t1);
140 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
141 fprintf(outfd, "%s ", str);
142 va_start(argp, fmt);
143 vfprintf(outfd, fmt, argp);
144 va_end(argp);
145 }
146
147 /**
148 * Compare the size of two directories
149 *
150 * \param obj1 Pointer to the first object.
151 * \param obj2 Pointer to the second object.
152 *
153 * This function first compares the size values as usual integers. If they compare as
154 * equal, the address of \a obj1 and \a obj2 are compared. So this compare function
155 * returns zero if and only if \a obj1 and \a obj2 point to the same memory area.
156 */
157 static int size_compare(const struct osl_object *obj1, const struct osl_object *obj2)
158 {
159 uint64_t d1 = *(uint64_t *)obj1->data;
160 uint64_t d2 = *(uint64_t *)obj2->data;
161 int ret = NUM_COMPARE(d2, d1);
162
163 if (ret)
164 return ret;
165 //INFO_LOG("addresses: %p, %p\n", obj1->data, obj2->data);
166 return NUM_COMPARE(obj2->data, obj1->data);
167 }
168
169 /**
170 * Compare two osl objects pointing to unsigned integers of 64 bit size.
171 *
172 * \param obj1 Pointer to the first integer.
173 * \param obj2 Pointer to the second integer.
174 *
175 * \return The values required for an osl compare function.
176 *
177 * \sa osl_compare_func, osl_hash_compare().
178 */
179 static int uint64_compare(const struct osl_object *obj1,
180 const struct osl_object *obj2)
181 {
182 uint64_t d1 = read_u64((const char *)obj1->data);
183 uint64_t d2 = read_u64((const char *)obj2->data);
184
185 if (d1 < d2)
186 return 1;
187 if (d1 > d2)
188 return -1;
189 return 0;
190 }
191
192 /** The columns of the directory table. */
193 enum dir_table_columns {
194 /** The name of the directory. */
195 DT_NAME,
196 /** The dir count number. */
197 DT_NUM,
198 /** The number of the parent directory. */
199 DT_PARENT_NUM,
200 /** The number of bytes of all regular files. */
201 DT_BYTES,
202 /** The number of all regular files. */
203 DT_FILES,
204 /** Number of columns in this table. */
205 NUM_DT_COLUMNS
206 };
207
208 static struct osl_column_description dir_table_cols[] = {
209 [DT_NAME] = {
210 .storage_type = OSL_MAPPED_STORAGE,
211 .storage_flags = 0,
212 .name = "dir",
213 },
214 [DT_NUM] = {
215 .storage_type = OSL_MAPPED_STORAGE,
216 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
217 .name = "num",
218 .compare_function = uint64_compare,
219 .data_size = sizeof(uint64_t)
220 },
221 [DT_PARENT_NUM] = {
222 .storage_type = OSL_MAPPED_STORAGE,
223 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
224 .name = "parent_num",
225 .compare_function = size_compare,
226 .data_size = sizeof(uint64_t)
227 },
228 [DT_BYTES] = {
229 .storage_type = OSL_MAPPED_STORAGE,
230 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
231 .compare_function = size_compare,
232 .name = "num_bytes",
233 .data_size = sizeof(uint64_t)
234 },
235 [DT_FILES] = {
236 .storage_type = OSL_MAPPED_STORAGE,
237 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
238 .compare_function = size_compare,
239 .name = "num_files",
240 .data_size = sizeof(uint64_t)
241 }
242 };
243
244 static struct osl_table_description dir_table_desc = {
245 .name = "dir_table",
246 .num_columns = NUM_DT_COLUMNS,
247 .flags = 0,
248 .column_descriptions = dir_table_cols,
249 };
250
251 /** The columns of the id table. */
252 enum user_table_columns {
253 /** The numer of the directory. */
254 UT_DIR_NUM,
255 /** The number of bytes of all regular files in this dir owned by this id. */
256 UT_BYTES,
257 /** The number of files in this dir owned by this id. */
258 UT_FILES,
259 /** Number of columns in this table. */
260 NUM_UT_COLUMNS
261 };
262
263 static struct osl_column_description user_table_cols[] = {
264 [UT_DIR_NUM] = {
265 .storage_type = OSL_MAPPED_STORAGE,
266 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
267 .name = "dir_num",
268 .compare_function = uint64_compare,
269 .data_size = sizeof(uint64_t)
270 },
271 [UT_BYTES] = {
272 .storage_type = OSL_MAPPED_STORAGE,
273 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
274 .compare_function = size_compare,
275 .name = "num_bytes",
276 .data_size = sizeof(uint64_t)
277 },
278 [UT_FILES] = {
279 .storage_type = OSL_MAPPED_STORAGE,
280 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
281 .compare_function = size_compare,
282 .name = "num_files",
283 .data_size = sizeof(uint64_t)
284 },
285 };
286
287 static struct osl_table *dir_table;
288
289 static int add_directory(char *dirname, uint64_t *dir_num, uint64_t *parent_dir_num,
290 uint64_t *dir_size, uint64_t *dir_files)
291 {
292 struct osl_object dir_objects[NUM_DT_COLUMNS];
293
294 INFO_LOG("adding #%llu: %s\n", (long long unsigned)*dir_num, dirname);
295 dir_objects[DT_NAME].data = dirname;
296 dir_objects[DT_NAME].size = strlen(dirname) + 1;
297 dir_objects[DT_NUM].data = dir_num;
298 dir_objects[DT_NUM].size = sizeof(*dir_num);
299 dir_objects[DT_PARENT_NUM].data = parent_dir_num;
300 dir_objects[DT_PARENT_NUM].size = sizeof(*parent_dir_num);
301 dir_objects[DT_BYTES].data = dir_size;
302 dir_objects[DT_BYTES].size = sizeof(*dir_size);
303 dir_objects[DT_FILES].data = dir_files;
304 dir_objects[DT_FILES].size = sizeof(*dir_files);
305 return osl(osl_add_row(dir_table, dir_objects));
306 }
307
308 static uint32_t num_uids;
309
310 static int open_user_table(struct user_info *ui, int create)
311 {
312 int ret;
313
314 ui->desc = para_malloc(sizeof(*ui->desc));
315 ui->desc->num_columns = NUM_UT_COLUMNS;
316 ui->desc->flags = 0;
317 ui->desc->column_descriptions = user_table_cols;
318 ui->desc->dir = para_strdup(conf.database_dir_arg);
319 ui->desc->name = make_message("%u", (unsigned)ui->uid);
320 INFO_LOG(".............................uid #%u: %u\n",
321 (unsigned)num_uids, (unsigned)ui->uid);
322 if (create) {
323 ret = osl(osl_create_table(ui->desc));
324 if (ret < 0)
325 goto err;
326 num_uids++;
327 }
328 ret = osl(osl_open_table(ui->desc, &ui->table));
329 if (ret < 0)
330 goto err;
331 return 1;
332 err:
333 free((char *)ui->desc->name);
334 free((char *)ui->desc->dir);
335 free(ui->desc);
336 ui->desc->name = NULL;
337 ui->desc->dir = NULL;
338 ui->desc = NULL;
339 ui->table = NULL;
340 ui->flags = 0;
341 return ret;
342 }
343
344 #define uid_hash_bits 8
345 static uint32_t uid_hash_table_size = 1 << uid_hash_bits;
346 #define PRIME1 0x811c9dc5
347 #define PRIME2 0x01000193
348
349 static void create_hash_table(void)
350 {
351 uid_hash_table = para_calloc(uid_hash_table_size
352 * sizeof(struct user_info));
353 }
354
355 static void free_hash_table(void)
356 {
357 free(uid_hash_table);
358 uid_hash_table = NULL;
359 }
360
361 static int create_tables(void)
362 {
363 int ret;
364
365 dir_table_desc.dir = para_strdup(conf.database_dir_arg);
366 ret = osl(osl_create_table(&dir_table_desc));
367 if (ret < 0)
368 return ret;
369 create_hash_table();
370 return 1;
371 }
372
373 /*
374 * We use a hash table of size s=2^uid_hash_bits to map the uids into the
375 * interval [0..s]. Hash collisions are treated by open addressing, i.e.
376 * unused slots in the table are used to store different uids that hash to the
377 * same slot.
378 *
379 * If a hash collision occurs, different slots are successively probed in order
380 * to find an unused slot for the new uid. Probing is implemented via a second
381 * hash function that maps the uid to h=(uid * PRIME2) | 1, which is always an
382 * odd number.
383 *
384 * An odd number is sufficient to make sure each entry of the hash table gets
385 * probed for probe_num between 0 and s-1 because s is a power of two, hence
386 * the second hash value has never a common divisor with the hash table size.
387 * IOW: h is invertible in the ring [0..s].
388 */
389 static uint32_t double_hash(uint32_t uid, uint32_t probe_num)
390 {
391 return (uid * PRIME1 + ((uid * PRIME2) | 1) * probe_num)
392 % uid_hash_table_size;
393 }
394
395 #define FOR_EACH_USER(ui) for (ui = uid_hash_table; ui && ui < uid_hash_table \
396 + uid_hash_table_size; ui++)
397
398 enum search_uid_flags {
399 OPEN_USER_TABLE = 1,
400 CREATE_USER_TABLE = 2,
401 };
402
403 static int uid_is_admissible(uint32_t uid)
404 {
405 int i;
406
407 for (i = 0; i < conf.uid_given; i++) {
408 struct uid_range *ur = admissible_uids + i;
409
410 if (ur->low <= uid && ur->high >= uid)
411 break;
412 }
413 i = !conf.uid_given || i < conf.uid_given;
414 DEBUG_LOG("uid %u is %sadmissible\n", (unsigned)uid,
415 i? "" : "not ");
416 return i;
417 }
418
419 static int search_uid(uint32_t uid, enum search_uid_flags flags,
420 struct user_info **ui_ptr)
421 {
422 uint32_t p;
423
424 for (p = 0; p < uid_hash_table_size; p++) {
425 struct user_info *ui = uid_hash_table + double_hash(uid, p);
426
427 if (!ui_used(ui)) {
428 int ret;
429 if (!flags)
430 return -E_BAD_UID;
431 ui->uid = uid;
432 ui->flags |= UI_FL_SLOT_USED;
433 if (!uid_is_admissible(uid))
434 return 0;
435 ui->flags |= UI_FL_ADMISSIBLE;
436 ret = open_user_table(ui, flags & CREATE_USER_TABLE);
437 if (ret < 0)
438 return ret;
439
440 if (ui_ptr)
441 *ui_ptr = ui;
442 return 1;
443 }
444 if (ui->uid != uid)
445 continue;
446 if (ui_ptr)
447 *ui_ptr = ui;
448 return 0;
449 }
450 return flags? -E_HASH_TABLE_OVERFLOW : -E_BAD_UID;
451 }
452
453 static int update_user_row(struct osl_table *t, uint64_t dir_num,
454 uint64_t *add)
455 {
456 struct osl_row *row;
457 struct osl_object obj = {.data = &dir_num, .size = sizeof(dir_num)};
458
459 int ret = osl(osl_get_row(t, UT_DIR_NUM, &obj, &row));
460
461 if (ret == -E_OSL && osl_errno != E_OSL_RB_KEY_NOT_FOUND)
462 return ret;
463 if (ret < 0) { /* this is the first file we add */
464 struct osl_object objects[NUM_UT_COLUMNS];
465 uint64_t num_files = 1;
466
467 objects[UT_DIR_NUM].data = &dir_num;
468 objects[UT_DIR_NUM].size = sizeof(dir_num);
469 objects[UT_BYTES].data = add;
470 objects[UT_BYTES].size = sizeof(*add);
471 objects[UT_FILES].data = &num_files;
472 objects[UT_FILES].size = sizeof(num_files);
473 INFO_LOG("######################### ret: %d\n", ret);
474 ret = osl(osl_add_row(t, objects));
475 INFO_LOG("######################### ret: %d\n", ret);
476 return ret;
477 } else { /* add size and increment file count */
478 uint64_t num;
479 struct osl_object obj1, obj2 = {.data = &num, .size = sizeof(num)};
480
481 ret = osl(osl_get_object(t, row, UT_BYTES, &obj1));
482 if (ret < 0)
483 return ret;
484 num = *(uint64_t *)obj1.data + *add;
485 ret = osl(osl_update_object(t, row, UT_BYTES, &obj2));
486 if (ret < 0)
487 return ret;
488 ret = osl(osl_get_object(t, row, UT_FILES, &obj1));
489 if (ret < 0)
490 return ret;
491 num = *(uint64_t *)obj1.data + 1;
492 return osl(osl_update_object(t, row, UT_FILES, &obj2));
493 }
494 }
495
496 static uint64_t num_dirs;
497 static uint64_t num_files;
498 static uint64_t num_bytes;
499
500 int scan_dir(char *dirname, uint64_t *parent_dir_num)
501 {
502 DIR *dir;
503 struct dirent *entry;
504 int ret, cwd_fd, ret2;
505 uint64_t dir_size = 0, dir_files = 0;
506 uint64_t this_dir_num = ++num_dirs;
507
508 DEBUG_LOG("----------------- %llu: %s\n", (long long unsigned)num_dirs, dirname);
509 ret = para_opendir(dirname, &dir, &cwd_fd);
510 if (ret < 0) {
511 if (ret != -ERRNO_TO_ERROR(EACCES))
512 return ret;
513 WARNING_LOG("permission denied for %s\n", dirname);
514 return 1;
515 }
516 while ((entry = readdir(dir))) {
517 mode_t m;
518 struct stat s;
519 uint32_t uid;
520 uint64_t size;
521 struct user_info *ui;
522
523 if (!strcmp(entry->d_name, "."))
524 continue;
525 if (!strcmp(entry->d_name, ".."))
526 continue;
527 if (lstat(entry->d_name, &s) == -1) {
528 WARNING_LOG("lstat error for %s/%s\n", dirname,
529 entry->d_name);
530 continue;
531 }
532 m = s.st_mode;
533 if (!S_ISREG(m) && !S_ISDIR(m))
534 continue;
535 if (S_ISDIR(m)) {
536 ret = scan_dir(entry->d_name, &this_dir_num);
537 if (ret < 0)
538 goto out;
539 continue;
540 }
541 /* regular file */
542 size = s.st_size;
543 dir_size += size;
544 num_bytes += size;
545 dir_files++;
546 num_files++;
547 uid = s.st_uid;
548 ret = search_uid(uid, CREATE_USER_TABLE | OPEN_USER_TABLE, &ui);
549 if (ret < 0)
550 goto out;
551 ui->bytes += size;
552 ui->files++;
553 ret = update_user_row(ui->table, this_dir_num, &size);
554 if (ret < 0)
555 goto out;
556 }
557 ret = add_directory(dirname, &this_dir_num, parent_dir_num,
558 &dir_size, &dir_files);
559 out:
560 closedir(dir);
561 ret2 = para_fchdir(cwd_fd);
562 if (ret2 < 0 && ret >= 0)
563 ret = ret2;
564 close(cwd_fd);
565 return ret;
566 }
567
568 static int get_dir_name_by_number(uint64_t *dirnum, char **name)
569 {
570 char *result = NULL, *tmp;
571 struct osl_row *row;
572 uint64_t val = *dirnum;
573 struct osl_object obj = {.data = &val, .size = sizeof(val)};
574 int ret;
575
576 again:
577 ret = osl(osl_get_row(dir_table, DT_NUM, &obj, &row));
578 if (ret < 0)
579 goto out;
580 ret = osl(osl_get_object(dir_table, row, DT_NAME, &obj));
581 if (ret < 0)
582 goto out;
583 if (result) {
584 tmp = make_message("%s/%s", (char *)obj.data, result);
585 free(result);
586 result = tmp;
587 } else
588 result = para_strdup((char *)obj.data);
589 ret = osl(osl_get_object(dir_table, row, DT_PARENT_NUM, &obj));
590 if (ret < 0)
591 goto out;
592 val = *(uint64_t *)obj.data;
593 if (val)
594 goto again;
595 out:
596 if (ret < 0) {
597 free(result);
598 *name = NULL;
599 } else
600 *name = result;
601 return ret;
602 }
603
604 static int get_dir_name_of_row(struct osl_row *dir_table_row, char **name)
605 {
606 struct osl_object obj;
607 int ret;
608 char *this_dir, *prefix = NULL;
609
610 *name = NULL;
611 ret = osl(osl_get_object(dir_table, dir_table_row, DT_NAME, &obj));
612 if (ret < 0)
613 return ret;
614 this_dir = para_strdup((char *)obj.data);
615 ret = osl(osl_get_object(dir_table, dir_table_row, DT_PARENT_NUM, &obj));
616 if (ret < 0)
617 goto out;
618 if (!*(uint64_t *)obj.data) {
619 *name = this_dir;
620 return 1;
621 }
622 ret = get_dir_name_by_number((uint64_t *)obj.data, &prefix);
623 if (ret < 0)
624 goto out;
625 *name = make_message("%s/%s", prefix, this_dir);
626 free(prefix);
627 ret = 1;
628 out:
629 free(this_dir);
630 return ret;
631 }
632
633 const uint64_t size_unit_divisors[] = {
634 [size_unit_arg_b] = 1ULL,
635 [size_unit_arg_k] = 1024ULL,
636 [size_unit_arg_m] = 1024ULL * 1024ULL,
637 [size_unit_arg_g] = 1024ULL * 1024ULL * 1024ULL,
638 [size_unit_arg_t] = 1024ULL * 1024ULL * 1024ULL * 1024ULL,
639 };
640
641 const uint64_t count_unit_divisors[] = {
642
643 [count_unit_arg_n] = 1ULL,
644 [count_unit_arg_k] = 1000ULL,
645 [count_unit_arg_m] = 1000ULL * 1000ULL,
646 [count_unit_arg_g] = 1000ULL * 1000ULL * 1000ULL,
647 [count_unit_arg_t] = 1000ULL * 1000ULL * 1000ULL * 1000ULL,
648 };
649
650 const char size_unit_abbrevs[] = " BKMGT";
651 const char count_unit_abbrevs[] = " KMGT";
652
653 static void format_size_value(enum enum_size_unit unit, uint64_t value, char *result)
654 {
655 if (unit == size_unit_arg_h) /* human readable */
656 for (unit = size_unit_arg_b; unit < size_unit_arg_t && value > size_unit_divisors[unit + 1]; unit++)
657 ; /* nothing */
658 sprintf(result, "%llu%c", (long long unsigned)value / size_unit_divisors[unit], size_unit_abbrevs[unit]);
659 }
660
661 static void format_count_value(enum enum_count_unit unit, uint64_t value, char *result)
662 {
663 if (unit == count_unit_arg_h) /* human readable */
664 for (unit = count_unit_arg_n; unit < count_unit_arg_t && value > count_unit_divisors[unit + 1]; unit++)
665 ; /* nothing */
666 sprintf(result, "%llu%c", (long long unsigned)value / count_unit_divisors[unit], count_unit_abbrevs[unit]);
667 }
668
669 enum global_stats_flags {
670 GSF_PRINT_DIRNAME = 1,
671 GSF_PRINT_BYTES = 2,
672 GSF_PRINT_FILES = 4,
673 GSF_COMPUTE_SUMMARY = 8,
674 };
675
676 struct global_stats_info {
677 uint32_t count;
678 int ret;
679 int osl_errno;
680 enum global_stats_flags flags;
681 };
682
683 static int global_stats_loop_function(struct osl_row *row, void *data)
684 {
685 struct global_stats_info *gsi = data;
686 struct osl_object obj;
687 char *dirname, formated_value[25];
688 int ret, summary = gsi->flags & GSF_COMPUTE_SUMMARY;
689
690 if (!gsi->count && !summary) {
691 ret = -E_LOOP_COMPLETE;
692 goto err;
693 }
694 if (gsi->count && (gsi->flags & GSF_PRINT_DIRNAME)) {
695 ret = get_dir_name_of_row(row, &dirname);
696 if (ret < 0)
697 goto err;
698 printf("%s%s", dirname,
699 (gsi->flags & (GSF_PRINT_FILES | GSF_PRINT_BYTES))?
700 "\t" : "\n"
701 );
702 }
703 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_FILES))) {
704 uint64_t files;
705 ret = osl(osl_get_object(dir_table, row, DT_FILES, &obj));
706 if (ret < 0)
707 goto err;
708 files = *(uint64_t *)obj.data;
709 if (gsi->count && (gsi->flags & GSF_PRINT_FILES)) {
710 format_size_value(conf.size_unit_arg, files,
711 formated_value);
712 printf("%s%s", formated_value,
713 (gsi->flags & GSF_PRINT_BYTES)? "\t" : "\n");
714 }
715 if (summary)
716 num_files += files;
717 }
718 if (summary || (gsi->count && (gsi->flags & GSF_PRINT_BYTES))) {
719 uint64_t bytes;
720 ret = osl(osl_get_object(dir_table, row, DT_BYTES, &obj));
721 if (ret < 0)
722 goto err;
723 bytes = *(uint64_t *)obj.data;
724 if (gsi->count && (gsi->flags & GSF_PRINT_BYTES)) {
725 format_size_value(conf.size_unit_arg, bytes,
726 formated_value);
727 printf("%s\n", formated_value);
728 }
729 if (summary) {
730 num_bytes += bytes;
731 num_dirs++;
732 }
733 }
734 if (gsi->count > 0)
735 gsi->count--;
736 return 1;
737 err:
738 gsi->ret = ret;
739 gsi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
740 return -1;
741 }
742
743 static void print_id_stats(void)
744 {
745 struct user_info *ui;
746
747 printf("--------------------- user summary (uid/dirs/files/bytes):\n");
748 FOR_EACH_USER(ui) {
749 char formated_dir_count[25], formated_file_count[25],
750 formated_bytes[25];
751 if (!ui_used(ui))
752 continue;
753 format_count_value(conf.count_unit_arg, ui->dirs,
754 formated_dir_count);
755 format_count_value(conf.count_unit_arg, ui->files,
756 formated_file_count);
757 format_size_value(conf.size_unit_arg, ui->bytes,
758 formated_bytes);
759 printf("%u\t%s\t%s\t%s\n", (unsigned)ui->uid,
760 formated_dir_count,
761 formated_file_count,
762 formated_bytes
763 );
764 }
765 }
766
767 enum user_stats_flags {
768 USF_PRINT_DIRNAME = 1,
769 USF_PRINT_BYTES = 2,
770 USF_PRINT_FILES = 4,
771 USF_COMPUTE_SUMMARY = 8,
772 };
773
774 struct user_stats_info {
775 uint32_t count;
776 enum user_stats_flags flags;
777 int ret;
778 int osl_errno;
779 struct user_info *ui;
780 };
781
782 static int user_stats_loop_function(struct osl_row *row, void *data)
783 {
784 struct user_stats_info *usi = data;
785 struct osl_object obj;
786 int ret, summary = usi->flags & GSF_COMPUTE_SUMMARY;
787 char formated_value[25];
788
789 if (!usi->count && !summary) {
790 ret = -E_LOOP_COMPLETE;
791 goto err;
792 }
793 if (usi->count && (usi->flags & USF_PRINT_DIRNAME)) {
794 char *dirname;
795 ret = osl(osl_get_object(usi->ui->table, row, UT_DIR_NUM, &obj));
796 if (ret < 0)
797 goto err;
798 ret = get_dir_name_by_number((uint64_t *)obj.data, &dirname);
799 if (ret < 0)
800 goto err;
801 printf("%s%s",
802 dirname,
803 (usi->flags & (USF_PRINT_FILES | USF_PRINT_BYTES))?
804 "\t" : "\n"
805 );
806 }
807 if (summary || (usi->count && (usi->flags & USF_PRINT_FILES))) {
808 uint64_t files;
809 ret = osl(osl_get_object(usi->ui->table, row, UT_FILES, &obj));
810 if (ret < 0)
811 goto err;
812 files = *(uint64_t *)obj.data;
813 if (usi->count && (usi->flags & USF_PRINT_FILES)) {
814 format_size_value(conf.size_unit_arg, files,
815 formated_value);
816 printf("%s%s", formated_value,
817 (usi->flags & USF_PRINT_BYTES)? "\t" : "\n"
818 );
819 }
820 if (summary)
821 usi->ui->files += files;
822 }
823 if (summary || (usi->count && (usi->flags & USF_PRINT_BYTES))) {
824 uint64_t bytes;
825 ret = osl(osl_get_object(usi->ui->table, row, UT_BYTES, &obj));
826 if (ret < 0)
827 goto err;
828 bytes = *(uint64_t *)obj.data;
829 if (usi->count && (usi->flags & USF_PRINT_BYTES)) {
830 format_size_value(conf.size_unit_arg, bytes,
831 formated_value);
832 printf("%s\n", formated_value);
833 }
834 if (summary) {
835 usi->ui->bytes += bytes;
836 usi->ui->dirs++;
837 }
838
839 }
840 if (usi->count > 0)
841 usi->count--;
842 return 1;
843 err:
844 usi->ret = ret;
845 usi->osl_errno = (ret == -E_OSL)? osl_errno : 0;
846 return -1;
847 }
848
849 static int check_loop_return(int ret, int loop_ret, int loop_osl_errno)
850 {
851 if (ret >= 0)
852 return ret;
853 assert(ret == -E_OSL);
854 if (osl_errno != E_OSL_LOOP)
855 /* error not caused by loop function returning negative. */
856 return ret;
857 assert(loop_ret < 0);
858 if (loop_ret == -E_LOOP_COMPLETE) /* no error */
859 return 1;
860 if (loop_ret == -E_OSL) { /* osl error in loop function */
861 assert(loop_osl_errno);
862 osl_errno = loop_osl_errno;
863 }
864 return loop_ret;
865 }
866
867 static int adu_loop_reverse(struct osl_table *t, unsigned col_num, void *private_data,
868 osl_rbtree_loop_func *func, int *loop_ret, int *loop_osl_errno)
869 {
870 int ret = osl(osl_rbtree_loop_reverse(t, col_num, private_data, func));
871 return check_loop_return(ret, *loop_ret, *loop_osl_errno);
872 }
873
874 static int print_user_stats(void)
875 {
876 struct user_info *ui;
877 int ret;
878
879 FOR_EACH_USER(ui) {
880 struct user_stats_info usi = {
881 .count = conf.limit_arg,
882 .ui = ui
883 };
884 if (!ui_used(ui) || !ui_admissible(ui))
885 continue;
886 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_BYTES | USF_COMPUTE_SUMMARY;
887 printf("************************************************ uid %u\n",
888 (unsigned) ui->uid);
889 printf("----------------- Largest dirs -------------------\n");
890 ret = adu_loop_reverse(ui->table, UT_BYTES, &usi, user_stats_loop_function,
891 &usi.ret, &usi.osl_errno);
892 if (ret < 0)
893 return ret;
894 printf("---------- dirs containing most files ------------\n");
895 usi.count = conf.limit_arg,
896 usi.flags = USF_PRINT_DIRNAME | USF_PRINT_FILES;
897 ret = adu_loop_reverse(ui->table, UT_FILES, &usi, user_stats_loop_function,
898 &usi.ret, &usi.osl_errno);
899 if (ret < 0)
900 return ret;
901 }
902 return 1;
903 }
904
905 static int print_statistics(void)
906 {
907 int ret;
908 struct global_stats_info gsi = {
909 .count = conf.limit_arg,
910 .flags = GSF_PRINT_DIRNAME | GSF_PRINT_BYTES | GSF_COMPUTE_SUMMARY
911 };
912
913 printf("----------------- Largest dirs -------------------\n");
914 ret = adu_loop_reverse(dir_table, DT_BYTES, &gsi,
915 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
916 if (ret < 0)
917 return ret;
918 gsi.count = conf.limit_arg;
919
920 gsi.flags = GSF_PRINT_DIRNAME | GSF_PRINT_FILES;
921 printf("---------- dirs containing most files ------------\n");
922 ret = adu_loop_reverse(dir_table, DT_FILES, &gsi,
923 global_stats_loop_function, &gsi.ret, &gsi.osl_errno);
924 if (ret < 0)
925 return ret;
926 printf("------------------ Global summary (dirs/files/bytes)\n"
927 "%llu\t%llu\t%llu\n",
928 (long long unsigned)num_dirs, (long long unsigned)num_files,
929 (long long unsigned)num_bytes);
930 print_user_stats();
931 print_id_stats();
932 return 1;
933 }
934
935 static char *get_uid_list_name(void)
936 {
937 return make_message("%s/uid_list", conf.database_dir_arg);
938 }
939
940 static int write_uid_list(void)
941 {
942 char *buf, *filename;
943 uint32_t count = 0;
944 struct user_info *ui;
945 size_t size = num_uids * sizeof(uint32_t);
946 int ret;
947
948 if (!num_uids)
949 return 0;
950 buf = para_malloc(size);
951 FOR_EACH_USER(ui) {
952 if (!ui_used(ui) || !ui_admissible(ui))
953 continue;
954 DEBUG_LOG("saving uid %u\n", (unsigned) ui->uid);
955 write_u32(buf + count++ * sizeof(uint32_t), ui->uid);
956 }
957 filename = get_uid_list_name();
958 ret = para_write_file(filename, buf, size);
959 free(filename);
960 free(buf);
961 return ret;
962 }
963
964 static int open_dir_table(void)
965 {
966 if (!dir_table_desc.dir) /* we did not create the table */
967 dir_table_desc.dir = para_strdup(conf.database_dir_arg);
968 return osl(osl_open_table(&dir_table_desc, &dir_table));
969 }
970
971 static void close_dir_table(void)
972 {
973 int ret;
974
975 if (!dir_table)
976 return;
977 ret = osl(osl_close_table(dir_table, OSL_MARK_CLEAN));
978 if (ret < 0)
979 ERROR_LOG("failed to close dir table: %s\n", adu_strerror(-ret));
980 free((char *)dir_table_desc.dir);
981 dir_table = NULL;
982 }
983
984 static void close_user_table(struct user_info *ui)
985 {
986 int ret;
987
988 if (!ui || !ui_used(ui) || !ui_admissible(ui))
989 return;
990 ret = osl(osl_close_table(ui->table, OSL_MARK_CLEAN));
991 if (ret < 0)
992 ERROR_LOG("failed to close user table %u: %s\n",
993 (unsigned) ui->uid, adu_strerror(-ret));
994 free((char *)ui->desc->name);
995 ui->desc->name = NULL;
996 free((char *)ui->desc->dir);
997 ui->desc->dir = NULL;
998 free(ui->desc);
999 ui->desc = NULL;
1000 ui->table = NULL;
1001 ui->flags = 0;
1002 }
1003
1004 static void close_user_tables(void)
1005 {
1006 struct user_info *ui;
1007
1008 FOR_EACH_USER(ui)
1009 close_user_table(ui);
1010 }
1011
1012 static void close_all_tables(void)
1013 {
1014 close_dir_table();
1015 close_user_tables();
1016 free_hash_table();
1017 }
1018
1019 static int com_create()
1020 {
1021 uint64_t zero = 0ULL;
1022 int ret = create_tables();
1023
1024 if (ret < 0)
1025 return ret;
1026 ret = open_dir_table();
1027 if (ret < 0)
1028 return ret;
1029 ret = scan_dir(conf.base_dir_arg, &zero);
1030 if (ret < 0)
1031 goto out;
1032 ret = write_uid_list();
1033 out:
1034 close_all_tables();
1035 return ret;
1036 }
1037
1038 static int read_uid_file(void)
1039 {
1040 size_t size;
1041 uint32_t n;
1042 char *filename = get_uid_list_name(), *map;
1043 int ret = mmap_full_file(filename, O_RDONLY, (void **)&map, &size, NULL);
1044
1045 if (ret < 0) {
1046 INFO_LOG("failed to map %s\n", filename);
1047 free(filename);
1048 return ret;
1049 }
1050 num_uids = size / 4;
1051 INFO_LOG("found %u uids in %s\n", (unsigned)num_uids, filename);
1052 free(filename);
1053 /* hash table size should be a power of two and larger than the number of uids */
1054 uid_hash_table_size = 4;
1055 while (uid_hash_table_size < num_uids)
1056 uid_hash_table_size *= 2;
1057 create_hash_table();
1058 for (n = 0; n < num_uids; n++) {
1059 uint32_t uid = read_u32(map + n * sizeof(uid));
1060 ret = search_uid(uid, OPEN_USER_TABLE, NULL);
1061 if (ret < 0)
1062 goto out;
1063 }
1064 out:
1065 para_munmap(map, size);
1066 return ret;
1067 }
1068
1069 static int com_select(void)
1070 {
1071 int ret;
1072
1073 ret = open_dir_table();
1074 if (ret < 0)
1075 return ret;
1076 ret = read_uid_file();
1077 if (ret < 0)
1078 return ret;
1079 ret = print_statistics();
1080 close_all_tables();
1081 return ret;
1082 }
1083
1084 static int check_args(void)
1085 {
1086 int i, ret;
1087
1088 /* remove trailing slashes from base-dir arg */
1089 if (conf.base_dir_given) {
1090 size_t len = strlen(conf.base_dir_arg);
1091 for (;;) {
1092 if (!len) /* empty string */
1093 return -ERRNO_TO_ERROR(EINVAL);
1094 if (!--len) /* length 1 is always OK */
1095 break;
1096 if (conf.base_dir_arg[len] != '/')
1097 break; /* no trailing slash, also OK */
1098 conf.base_dir_arg[len] = '\0';
1099 }
1100 }
1101 if (!conf.uid_given)
1102 return 0;
1103 admissible_uids = para_malloc(conf.uid_given * sizeof(*admissible_uids));
1104 for (i = 0; i < conf.uid_given; i++) {
1105 ret = parse_uid_range(conf.uid_arg[i], admissible_uids + i);
1106 if (ret < 0)
1107 goto err;
1108 }
1109 return 1;
1110 err:
1111 free(admissible_uids);
1112 admissible_uids = NULL;
1113 return ret;
1114 }
1115
1116 int main(int argc, char **argv)
1117 {
1118 int ret;
1119 struct cmdline_parser_params params = {
1120 .override = 0,
1121 .initialize = 1,
1122 .check_required = 0,
1123 .check_ambiguity = 0,
1124 .print_errors = 1
1125 };
1126
1127 cmdline_parser_ext(argc, argv, &conf, &params); /* aborts on errors */
1128 ret = check_args();
1129 if (ret < 0)
1130 goto out;
1131 ret = -E_SYNTAX;
1132 if (conf.select_given)
1133 ret = com_select();
1134 else
1135 ret = com_create();
1136 if (ret < 0)
1137 goto out;
1138 out:
1139 free(admissible_uids);
1140 if (ret < 0) {
1141 ERROR_LOG("%s\n", adu_strerror(-ret));
1142 return -EXIT_FAILURE;
1143 }
1144 return EXIT_SUCCESS;
1145 }