Write the list of all uids to a file.
[adu.git] / adu.c
1 #include "adu.h"
2 #include <dirent.h> /* readdir() */
3
4 #include "gcc-compat.h"
5 #include "osl.h"
6 #include "fd.h"
7 #include "hash.h"
8 #include "string.h"
9 #include "error.h"
10
11 DEFINE_ERRLIST;
12
13 #define DATABASE_DIR "/tmp/adu"
14
15 /** evaluates to 1 if x < y, to -1 if x > y and to 0 if x == y */
16 #define NUM_COMPARE(x, y) ((int)((x) < (y)) - (int)((x) > (y)))
17
18
19 /**
20 * The log function.
21 *
22 * \param ll Loglevel.
23 * \param fml Usual format string.
24 *
25 * All XXX_LOG() macros use this function.
26 */
27 __printf_2_3 void __log(int ll, const char* fmt,...)
28 {
29 va_list argp;
30 FILE *outfd;
31 struct tm *tm;
32 time_t t1;
33 char str[255] = "";
34
35 if (ll < 4)
36 return;
37 outfd = stderr;
38 time(&t1);
39 tm = localtime(&t1);
40 strftime(str, sizeof(str), "%b %d %H:%M:%S", tm);
41 fprintf(outfd, "%s ", str);
42 va_start(argp, fmt);
43 vfprintf(outfd, fmt, argp);
44 va_end(argp);
45 }
46
47 /**
48 * Compare the size of two directories
49 *
50 * \param obj1 Pointer to the first object.
51 * \param obj2 Pointer to the second object.
52 *
53 * This function first compares the size values as usual integers. If they compare as
54 * equal, the address of \a obj1 and \a obj2 are compared. So this compare function
55 * returns zero if and only if \a obj1 and \a obj2 point to the same memory area.
56 */
57 static int size_compare(const struct osl_object *obj1, const struct osl_object *obj2)
58 {
59 uint64_t d1 = *(uint64_t *)obj1->data;
60 uint64_t d2 = *(uint64_t *)obj2->data;
61 int ret = NUM_COMPARE(d2, d1);
62
63 if (ret)
64 return ret;
65 //INFO_LOG("addresses: %p, %p\n", obj1->data, obj2->data);
66 return NUM_COMPARE(obj2->data, obj1->data);
67 }
68
69 /**
70 * Compare two osl objects of string type.
71 *
72 * \param obj1 Pointer to the first object.
73 * \param obj2 Pointer to the second object.
74 *
75 * In any case, only \p MIN(obj1->size, obj2->size) characters of each string
76 * are taken into account.
77 *
78 * \return It returns an integer less than, equal to, or greater than zero if
79 * \a obj1 is found, respectively, to be less than, to match, or be greater than
80 * obj2.
81 *
82 * \sa strcmp(3), strncmp(3), osl_compare_func.
83 */
84 int string_compare(const struct osl_object *obj1, const struct osl_object *obj2)
85 {
86 const char *str1 = (const char *)obj1->data;
87 const char *str2 = (const char *)obj2->data;
88 return strncmp(str1, str2, MIN(obj1->size, obj2->size));
89 }
90
91 /** The columns of the directory table. */
92 enum dir_table_columns {
93 /** The name of the directory. */
94 DT_NAME,
95 /** The dir count number. */
96 DT_NUM,
97 /** The number of bytes of all regular files. */
98 DT_BYTES,
99 /** The number of all regular files. */
100 DT_FILES,
101 /** Number of columns in this table. */
102 NUM_DT_COLUMNS
103 };
104
105 static struct osl_column_description dir_table_cols[] = {
106 [DT_NAME] = {
107 .storage_type = OSL_MAPPED_STORAGE,
108 .storage_flags = OSL_RBTREE | OSL_UNIQUE,
109 .name = "dir",
110 .compare_function = string_compare,
111 },
112 [DT_NUM] = {
113 .storage_type = OSL_MAPPED_STORAGE,
114 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
115 .name = "num",
116 .compare_function = uint32_compare,
117 .data_size = sizeof(uint32_t)
118 },
119 [DT_BYTES] = {
120 .storage_type = OSL_MAPPED_STORAGE,
121 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
122 .compare_function = size_compare,
123 .name = "num_bytes",
124 .data_size = sizeof(uint64_t)
125 },
126 [DT_FILES] = {
127 .storage_type = OSL_MAPPED_STORAGE,
128 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
129 .compare_function = size_compare,
130 .name = "num_files",
131 .data_size = sizeof(uint64_t)
132 }
133 };
134
135 static struct osl_table_description dir_table_desc = {
136 .name = "dir_table",
137 .num_columns = NUM_DT_COLUMNS,
138 .flags = 0,
139 .column_descriptions = dir_table_cols,
140 .dir = DATABASE_DIR
141 };
142
143 /** The columns of the id table. */
144 enum user_table_columns {
145 /** The numer of the directory. */
146 UT_DIR_NUM,
147 /** The number of bytes of all regular files in this dir owned by this id. */
148 UT_BYTES,
149 /** The number of files in this dir owned by this id. */
150 UT_FILES,
151 /** Number of columns in this table. */
152 NUM_UT_COLUMNS
153 };
154
155 static struct osl_column_description user_table_cols[] = {
156 [UT_DIR_NUM] = {
157 .storage_type = OSL_MAPPED_STORAGE,
158 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
159 .name = "dir_num",
160 .compare_function = uint32_compare,
161 .data_size = sizeof(uint32_t)
162 },
163 [UT_BYTES] = {
164 .storage_type = OSL_MAPPED_STORAGE,
165 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
166 .compare_function = size_compare,
167 .name = "num_bytes",
168 .data_size = sizeof(uint64_t)
169 },
170 [UT_FILES] = {
171 .storage_type = OSL_MAPPED_STORAGE,
172 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE,
173 .compare_function = size_compare,
174 .name = "num_files",
175 .data_size = sizeof(uint64_t)
176 },
177 };
178
179 static struct osl_table *dir_table;
180
181 int add_directory(char *dirname, uint32_t dir_num, uint64_t *dir_size,
182 uint64_t *dir_files)
183 {
184 struct osl_object dir_objects[NUM_DT_COLUMNS];
185
186 INFO_LOG("adding #%u: %s\n", dir_num, dirname);
187 dir_objects[DT_NAME].data = dirname;
188 dir_objects[DT_NAME].size = strlen(dirname) + 1;
189 dir_objects[DT_NUM].data = &dir_num;
190 dir_objects[DT_NUM].size = sizeof(dir_num);
191 dir_objects[DT_BYTES].data = dir_size;
192 dir_objects[DT_BYTES].size = sizeof(*dir_size);
193 dir_objects[DT_FILES].data = dir_files;
194 dir_objects[DT_FILES].size = sizeof(*dir_files);
195
196 return osl_add_row(dir_table, dir_objects);
197 }
198
199 static uint32_t num_uids;
200
201 int create_and_open_user_table(uint32_t uid, struct osl_table **t)
202 {
203 int ret;
204 struct osl_table_description *desc = para_malloc(sizeof(*desc));
205
206 desc->num_columns = NUM_UT_COLUMNS;
207 desc->flags = 0;
208 desc->column_descriptions = user_table_cols;
209 desc->dir = para_strdup(DATABASE_DIR);
210 desc->name = make_message("%u", uid);
211 num_uids++;
212 INFO_LOG(".............................uid #%u: %u\n",
213 (unsigned)num_uids, (unsigned)uid);
214 // user_table_desc.name = make_message("%u", uid);
215 ret = osl_create_table(desc);
216 if (ret < 0)
217 return ret;
218 return osl_open_table(desc, t);
219 }
220
221
222 #define uid_hash_bits 8
223 static uint32_t uid_hash_table_size = 1 << uid_hash_bits;
224 #define PRIME1 0x811c9dc5
225 #define PRIME2 0x01000193
226
227 struct user_info {
228 uint32_t uid;
229 struct osl_table *table;
230 uint64_t files;
231 uint64_t bytes;
232 };
233
234 static struct user_info *uid_hash_table;
235
236 static void create_hash_table(void)
237 {
238 uid_hash_table = para_calloc(uid_hash_table_size
239 * sizeof(struct user_info));
240 }
241
242 static int create_tables(void)
243 {
244 create_hash_table();
245 return osl_create_table(&dir_table_desc);
246 }
247
248
249 static uint32_t double_hash(uint32_t uid, uint32_t probe_num)
250 {
251 return (uid * PRIME1 + ((uid * PRIME2) | 1) * probe_num) % uid_hash_table_size;
252 }
253
254 #define FOR_EACH_USER(ui) for (ui = uid_hash_table; ui < uid_hash_table \
255 + uid_hash_table_size; ui++)
256
257 static int search_uid(uint32_t uid, int insert, struct user_info **ui)
258 {
259 uint32_t p;
260
261 for (p = 0; p < uid_hash_table_size; p++) {
262 struct user_info *i = uid_hash_table + double_hash(uid, p);
263 if (!i->table) {
264 if (!insert)
265 return -E_BAD_UID;
266 int ret = create_and_open_user_table(uid, &i->table);
267 if (ret < 0)
268 return ret;
269 i->uid = uid;
270 *ui = i;
271 return 1;
272 }
273 if (i->uid != uid)
274 continue;
275 *ui = i;
276 return 0;
277 }
278 return insert? -E_HASH_TABLE_OVERFLOW : -E_BAD_UID;
279 }
280
281 static int update_user_row(struct osl_table *t, uint32_t dir_num,
282 uint64_t *add)
283 {
284 struct osl_row *row;
285 struct osl_object obj = {.data = &dir_num, .size = sizeof(dir_num)};
286
287 int ret = osl_get_row(t, UT_DIR_NUM, &obj, &row);
288
289 if (ret < 0 && ret != -E_RB_KEY_NOT_FOUND)
290 return ret;
291 if (ret < 0) { /* this is the first file we add */
292 struct osl_object objects[NUM_UT_COLUMNS];
293 uint64_t num_files = 1;
294
295 objects[UT_DIR_NUM].data = &dir_num;
296 objects[UT_DIR_NUM].size = sizeof(dir_num);
297 objects[UT_BYTES].data = add;
298 objects[UT_BYTES].size = sizeof(*add);
299 objects[UT_FILES].data = &num_files;
300 objects[UT_FILES].size = sizeof(num_files);
301 INFO_LOG("######################### ret: %d\n", ret);
302 ret = osl_add_row(t, objects);
303 INFO_LOG("######################### ret: %d\n", ret);
304 return ret;
305 } else { /* add size and increment file count */
306 uint64_t num;
307 struct osl_object obj1, obj2 = {.data = &num, .size = sizeof(num)};
308
309 ret = osl_get_object(t, row, UT_BYTES, &obj1);
310 if (ret < 0)
311 return ret;
312 num = *(uint64_t *)obj1.data + *add;
313 ret = osl_update_object(t, row, UT_BYTES, &obj2);
314 if (ret < 0)
315 return ret;
316 ret = osl_get_object(t, row, UT_FILES, &obj1);
317 if (ret < 0)
318 return ret;
319 num = *(uint64_t *)obj1.data + 1;
320 return osl_update_object(t, row, UT_FILES, &obj2);
321 }
322 }
323
324 static uint32_t num_dirs;
325 static uint32_t num_files;
326 static uint64_t num_bytes;
327
328 int scan_dir(char *dirname)
329 {
330 DIR *dir;
331 struct dirent *entry;
332 int ret, cwd_fd, ret2;
333 uint64_t dir_size = 0, dir_files = 0;
334 uint32_t this_dir_num = num_dirs++;
335
336 DEBUG_LOG("----------------- %u: %s\n", num_dirs, dirname);
337 ret = para_opendir(dirname, &dir, &cwd_fd);
338 if (ret < 0) {
339 if (ret != -ERRNO_TO_ERROR(EACCES))
340 return ret;
341 WARNING_LOG("permission denied for %s\n", dirname);
342 return 1;
343 }
344 while ((entry = readdir(dir))) {
345 mode_t m;
346 char *tmp;
347 struct stat s;
348 uint32_t uid;
349 uint64_t size;
350 struct user_info *ui;
351
352 if (!strcmp(entry->d_name, "."))
353 continue;
354 if (!strcmp(entry->d_name, ".."))
355 continue;
356 if (lstat(entry->d_name, &s) == -1) {
357 WARNING_LOG("lstat error for %s/%s\n", dirname,
358 entry->d_name);
359 continue;
360 }
361 m = s.st_mode;
362 if (!S_ISREG(m) && !S_ISDIR(m))
363 continue;
364 if (S_ISDIR(m)) {
365 tmp = make_message("%s/%s", dirname, entry->d_name);
366 ret = scan_dir(tmp);
367 free(tmp);
368 if (ret < 0)
369 goto out;
370 continue;
371 }
372 /* regular file */
373 size = s.st_size;
374 dir_size += size;
375 num_bytes += size;
376 dir_files++;
377 num_files++;
378 uid = s.st_uid;
379 ret = search_uid(uid, 1, &ui);
380 if (ret < 0)
381 goto out;
382 ui->bytes += size;
383 ui->files++;
384 ret = update_user_row(ui->table, this_dir_num, &size);
385 if (ret < 0)
386 goto out;
387 }
388 ret = add_directory(dirname, this_dir_num, &dir_size, &dir_files);
389 out:
390 closedir(dir);
391 ret2 = para_fchdir(cwd_fd);
392 if (ret2 < 0 && ret >= 0)
393 ret = ret2;
394 close(cwd_fd);
395 return ret;
396 }
397
398 static int get_dir_name(struct osl_row *row, char **name)
399 {
400 struct osl_object obj;
401 int ret = osl_get_object(dir_table, row, DT_NAME, &obj);
402
403 if (ret < 0)
404 return ret;
405 *name = obj.data;
406 return 1;
407 }
408
409 static int print_dirname_and_size(struct osl_row *row, void *data)
410 {
411 unsigned *count = data;
412 struct osl_object obj;
413 char *name;
414 int ret;
415
416 if ((*count)++ > 100)
417 return -E_LOOP_COMPLETE;
418 ret = get_dir_name(row, &name);
419 if (ret < 0)
420 return ret;
421 ret = osl_get_object(dir_table, row, DT_BYTES, &obj);
422 if (ret < 0)
423 return ret;
424 printf("%s\t%llu\n", name, *(long long unsigned *)obj.data);
425 return 1;
426 }
427
428 static int print_dirname_and_file_count(struct osl_row *row, void *data)
429 {
430 unsigned *count = data;
431 struct osl_object obj;
432 char *name;
433 int ret;
434
435 if ((*count)++ > 100)
436 return -E_LOOP_COMPLETE;
437 ret = get_dir_name(row, &name);
438 if (ret < 0)
439 return ret;
440 ret = osl_get_object(dir_table, row, DT_FILES, &obj);
441 if (ret < 0)
442 return ret;
443 printf("%s\t%llu\n", name, *(long long unsigned *)obj.data);
444 return 1;
445 }
446
447 static void print_id_stats(void)
448 {
449 struct user_info *ui;
450
451 FOR_EACH_USER(ui) {
452 if (!ui->table)
453 continue;
454 printf("%u\t%llu\t%llu\n", (unsigned)ui->uid, (long long unsigned)ui->files,
455 (long long unsigned)ui->bytes);
456 }
457 }
458
459 struct big_dir_info {
460 unsigned count;
461 struct osl_table *user_table;
462 };
463
464 static int print_big_dir(struct osl_row *row, void *data)
465 {
466 struct big_dir_info *bdi = data;
467 int ret;
468 struct osl_row *dir_row;
469 char *dirname;
470 uint64_t bytes;
471 struct osl_object obj;
472
473 if (bdi->count++ > 10)
474 return -E_LOOP_COMPLETE;
475 ret = osl_get_object(bdi->user_table, row, UT_BYTES, &obj);
476 if (ret < 0)
477 return ret;
478 bytes = *(uint64_t *)obj.data;
479 ret = osl_get_object(bdi->user_table, row, UT_DIR_NUM, &obj);
480 if (ret < 0)
481 return ret;
482 ret = osl_get_row(dir_table, DT_NUM, &obj, &dir_row);
483 if (ret < 0)
484 return ret;
485 ret = osl_get_object(dir_table, dir_row, DT_NAME, &obj);
486 if (ret < 0)
487 return ret;
488 dirname = obj.data;
489 printf("%s: %llu\n", dirname, (long long unsigned)bytes);
490 return 1;
491 }
492
493 static void print_id_dir_stats(void)
494 {
495 struct user_info *ui;
496
497 FOR_EACH_USER(ui) {
498 struct big_dir_info bdi = {.count = 0};
499 if (!ui->table)
500 continue;
501 bdi.user_table = ui->table;
502 printf("************************* Big dirs owned by uid %u\n", (unsigned) ui->uid);
503 osl_rbtree_loop_reverse(ui->table, UT_BYTES, &bdi, print_big_dir);
504 }
505 }
506
507 static int print_statistics(void)
508 {
509 unsigned count = 0;
510 int ret;
511
512 printf("Summary: %u dirs, %u files, %llu bytes\n", (unsigned)num_dirs,
513 (unsigned)num_files, (long long unsigned)num_bytes);
514 printf("************************* Biggest dirs\n");
515 ret = osl_rbtree_loop_reverse(dir_table, DT_BYTES, &count, print_dirname_and_size);
516 if (ret < 0 && ret != -E_LOOP_COMPLETE)
517 return ret;
518 count = 0;
519 printf("************************* dirs containing many files\n");
520 ret = osl_rbtree_loop_reverse(dir_table, DT_FILES, &count, print_dirname_and_file_count);
521 if (ret < 0 && ret != -E_LOOP_COMPLETE)
522 return ret;
523
524 printf("************************* dirs stats by owner\n");
525 print_id_stats();
526 print_id_dir_stats();
527 return 1;
528 }
529
530 static int write_uid_list(void)
531 {
532 char *buf, *filename = DATABASE_DIR "/" "uid_list";
533 uint32_t count = 0;
534 struct user_info *ui;
535 size_t size = num_uids * sizeof(uint32_t);
536 int ret;
537
538 if (!num_uids)
539 return 0;
540 buf = para_malloc(size);
541 FOR_EACH_USER(ui) {
542 if (!ui->table)
543 continue;
544 write_u32(buf + count++ * sizeof(uint32_t), ui->uid);
545 }
546 ret = para_write_file(filename, buf, size);
547 free(buf);
548 return ret;
549 }
550
551 static int com_create(char *dirname)
552 {
553 int ret = scan_dir(dirname);
554
555 if (ret < 0)
556 return ret;
557 return write_uid_list();
558 }
559
560 int main(int argc, char **argv)
561 {
562 int ret = create_tables();
563 if (ret < 0)
564 goto out;
565 ret = osl_open_table(&dir_table_desc, &dir_table);
566 if (ret < 0)
567 goto out;
568 ret = -E_SYNTAX;
569 if (argc != 2)
570 goto out;
571 ret = com_create(argv[1]);
572 if (ret < 0)
573 goto out;
574 print_statistics();
575 out:
576 if (ret < 0) {
577 ERROR_LOG("%s\n", error_txt(-ret));
578 return -EXIT_FAILURE;
579 }
580 return EXIT_SUCCESS;
581 }
582