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