Merge branch 'jc/fsck-retire-require-eoh'
[git/git.git] / fsck.c
1 #include "cache.h"
2 #include "object.h"
3 #include "blob.h"
4 #include "tree.h"
5 #include "tree-walk.h"
6 #include "commit.h"
7 #include "tag.h"
8 #include "fsck.h"
9 #include "refs.h"
10 #include "utf8.h"
11
12 static int fsck_walk_tree(struct tree *tree, fsck_walk_func walk, void *data)
13 {
14 struct tree_desc desc;
15 struct name_entry entry;
16 int res = 0;
17
18 if (parse_tree(tree))
19 return -1;
20
21 init_tree_desc(&desc, tree->buffer, tree->size);
22 while (tree_entry(&desc, &entry)) {
23 int result;
24
25 if (S_ISGITLINK(entry.mode))
26 continue;
27 if (S_ISDIR(entry.mode))
28 result = walk(&lookup_tree(entry.sha1)->object, OBJ_TREE, data);
29 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode))
30 result = walk(&lookup_blob(entry.sha1)->object, OBJ_BLOB, data);
31 else {
32 result = error("in tree %s: entry %s has bad mode %.6o",
33 sha1_to_hex(tree->object.sha1), entry.path, entry.mode);
34 }
35 if (result < 0)
36 return result;
37 if (!res)
38 res = result;
39 }
40 return res;
41 }
42
43 static int fsck_walk_commit(struct commit *commit, fsck_walk_func walk, void *data)
44 {
45 struct commit_list *parents;
46 int res;
47 int result;
48
49 if (parse_commit(commit))
50 return -1;
51
52 result = walk((struct object *)commit->tree, OBJ_TREE, data);
53 if (result < 0)
54 return result;
55 res = result;
56
57 parents = commit->parents;
58 while (parents) {
59 result = walk((struct object *)parents->item, OBJ_COMMIT, data);
60 if (result < 0)
61 return result;
62 if (!res)
63 res = result;
64 parents = parents->next;
65 }
66 return res;
67 }
68
69 static int fsck_walk_tag(struct tag *tag, fsck_walk_func walk, void *data)
70 {
71 if (parse_tag(tag))
72 return -1;
73 return walk(tag->tagged, OBJ_ANY, data);
74 }
75
76 int fsck_walk(struct object *obj, fsck_walk_func walk, void *data)
77 {
78 if (!obj)
79 return -1;
80 switch (obj->type) {
81 case OBJ_BLOB:
82 return 0;
83 case OBJ_TREE:
84 return fsck_walk_tree((struct tree *)obj, walk, data);
85 case OBJ_COMMIT:
86 return fsck_walk_commit((struct commit *)obj, walk, data);
87 case OBJ_TAG:
88 return fsck_walk_tag((struct tag *)obj, walk, data);
89 default:
90 error("Unknown object type for %s", sha1_to_hex(obj->sha1));
91 return -1;
92 }
93 }
94
95 /*
96 * The entries in a tree are ordered in the _path_ order,
97 * which means that a directory entry is ordered by adding
98 * a slash to the end of it.
99 *
100 * So a directory called "a" is ordered _after_ a file
101 * called "a.c", because "a/" sorts after "a.c".
102 */
103 #define TREE_UNORDERED (-1)
104 #define TREE_HAS_DUPS (-2)
105
106 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
107 {
108 int len1 = strlen(name1);
109 int len2 = strlen(name2);
110 int len = len1 < len2 ? len1 : len2;
111 unsigned char c1, c2;
112 int cmp;
113
114 cmp = memcmp(name1, name2, len);
115 if (cmp < 0)
116 return 0;
117 if (cmp > 0)
118 return TREE_UNORDERED;
119
120 /*
121 * Ok, the first <len> characters are the same.
122 * Now we need to order the next one, but turn
123 * a '\0' into a '/' for a directory entry.
124 */
125 c1 = name1[len];
126 c2 = name2[len];
127 if (!c1 && !c2)
128 /*
129 * git-write-tree used to write out a nonsense tree that has
130 * entries with the same name, one blob and one tree. Make
131 * sure we do not have duplicate entries.
132 */
133 return TREE_HAS_DUPS;
134 if (!c1 && S_ISDIR(mode1))
135 c1 = '/';
136 if (!c2 && S_ISDIR(mode2))
137 c2 = '/';
138 return c1 < c2 ? 0 : TREE_UNORDERED;
139 }
140
141 static int fsck_tree(struct tree *item, int strict, fsck_error error_func)
142 {
143 int retval;
144 int has_null_sha1 = 0;
145 int has_full_path = 0;
146 int has_empty_name = 0;
147 int has_dot = 0;
148 int has_dotdot = 0;
149 int has_dotgit = 0;
150 int has_zero_pad = 0;
151 int has_bad_modes = 0;
152 int has_dup_entries = 0;
153 int not_properly_sorted = 0;
154 struct tree_desc desc;
155 unsigned o_mode;
156 const char *o_name;
157
158 init_tree_desc(&desc, item->buffer, item->size);
159
160 o_mode = 0;
161 o_name = NULL;
162
163 while (desc.size) {
164 unsigned mode;
165 const char *name;
166 const unsigned char *sha1;
167
168 sha1 = tree_entry_extract(&desc, &name, &mode);
169
170 has_null_sha1 |= is_null_sha1(sha1);
171 has_full_path |= !!strchr(name, '/');
172 has_empty_name |= !*name;
173 has_dot |= !strcmp(name, ".");
174 has_dotdot |= !strcmp(name, "..");
175 has_dotgit |= (!strcmp(name, ".git") ||
176 is_hfs_dotgit(name) ||
177 is_ntfs_dotgit(name));
178 has_zero_pad |= *(char *)desc.buffer == '0';
179 update_tree_entry(&desc);
180
181 switch (mode) {
182 /*
183 * Standard modes..
184 */
185 case S_IFREG | 0755:
186 case S_IFREG | 0644:
187 case S_IFLNK:
188 case S_IFDIR:
189 case S_IFGITLINK:
190 break;
191 /*
192 * This is nonstandard, but we had a few of these
193 * early on when we honored the full set of mode
194 * bits..
195 */
196 case S_IFREG | 0664:
197 if (!strict)
198 break;
199 default:
200 has_bad_modes = 1;
201 }
202
203 if (o_name) {
204 switch (verify_ordered(o_mode, o_name, mode, name)) {
205 case TREE_UNORDERED:
206 not_properly_sorted = 1;
207 break;
208 case TREE_HAS_DUPS:
209 has_dup_entries = 1;
210 break;
211 default:
212 break;
213 }
214 }
215
216 o_mode = mode;
217 o_name = name;
218 }
219
220 retval = 0;
221 if (has_null_sha1)
222 retval += error_func(&item->object, FSCK_WARN, "contains entries pointing to null sha1");
223 if (has_full_path)
224 retval += error_func(&item->object, FSCK_WARN, "contains full pathnames");
225 if (has_empty_name)
226 retval += error_func(&item->object, FSCK_WARN, "contains empty pathname");
227 if (has_dot)
228 retval += error_func(&item->object, FSCK_WARN, "contains '.'");
229 if (has_dotdot)
230 retval += error_func(&item->object, FSCK_WARN, "contains '..'");
231 if (has_dotgit)
232 retval += error_func(&item->object, FSCK_WARN, "contains '.git'");
233 if (has_zero_pad)
234 retval += error_func(&item->object, FSCK_WARN, "contains zero-padded file modes");
235 if (has_bad_modes)
236 retval += error_func(&item->object, FSCK_WARN, "contains bad file modes");
237 if (has_dup_entries)
238 retval += error_func(&item->object, FSCK_ERROR, "contains duplicate file entries");
239 if (not_properly_sorted)
240 retval += error_func(&item->object, FSCK_ERROR, "not properly sorted");
241 return retval;
242 }
243
244 static int verify_headers(const void *data, unsigned long size,
245 struct object *obj, fsck_error error_func)
246 {
247 const char *buffer = (const char *)data;
248 unsigned long i;
249
250 for (i = 0; i < size; i++) {
251 switch (buffer[i]) {
252 case '\0':
253 return error_func(obj, FSCK_ERROR,
254 "unterminated header: NUL at offset %d", i);
255 case '\n':
256 if (i + 1 < size && buffer[i + 1] == '\n')
257 return 0;
258 }
259 }
260
261 /*
262 * We did not find double-LF that separates the header
263 * and the body. Not having a body is not a crime but
264 * we do want to see the terminating LF for the last header
265 * line.
266 */
267 if (size && buffer[size - 1] == '\n')
268 return 0;
269
270 return error_func(obj, FSCK_ERROR, "unterminated header");
271 }
272
273 static int fsck_ident(const char **ident, struct object *obj, fsck_error error_func)
274 {
275 char *end;
276
277 if (**ident == '<')
278 return error_func(obj, FSCK_ERROR, "invalid author/committer line - missing space before email");
279 *ident += strcspn(*ident, "<>\n");
280 if (**ident == '>')
281 return error_func(obj, FSCK_ERROR, "invalid author/committer line - bad name");
282 if (**ident != '<')
283 return error_func(obj, FSCK_ERROR, "invalid author/committer line - missing email");
284 if ((*ident)[-1] != ' ')
285 return error_func(obj, FSCK_ERROR, "invalid author/committer line - missing space before email");
286 (*ident)++;
287 *ident += strcspn(*ident, "<>\n");
288 if (**ident != '>')
289 return error_func(obj, FSCK_ERROR, "invalid author/committer line - bad email");
290 (*ident)++;
291 if (**ident != ' ')
292 return error_func(obj, FSCK_ERROR, "invalid author/committer line - missing space before date");
293 (*ident)++;
294 if (**ident == '0' && (*ident)[1] != ' ')
295 return error_func(obj, FSCK_ERROR, "invalid author/committer line - zero-padded date");
296 if (date_overflows(strtoul(*ident, &end, 10)))
297 return error_func(obj, FSCK_ERROR, "invalid author/committer line - date causes integer overflow");
298 if (end == *ident || *end != ' ')
299 return error_func(obj, FSCK_ERROR, "invalid author/committer line - bad date");
300 *ident = end + 1;
301 if ((**ident != '+' && **ident != '-') ||
302 !isdigit((*ident)[1]) ||
303 !isdigit((*ident)[2]) ||
304 !isdigit((*ident)[3]) ||
305 !isdigit((*ident)[4]) ||
306 ((*ident)[5] != '\n'))
307 return error_func(obj, FSCK_ERROR, "invalid author/committer line - bad time zone");
308 (*ident) += 6;
309 return 0;
310 }
311
312 static int fsck_commit_buffer(struct commit *commit, const char *buffer,
313 unsigned long size, fsck_error error_func)
314 {
315 unsigned char tree_sha1[20], sha1[20];
316 struct commit_graft *graft;
317 unsigned parent_count, parent_line_count = 0;
318 int err;
319
320 if (verify_headers(buffer, size, &commit->object, error_func))
321 return -1;
322
323 if (!skip_prefix(buffer, "tree ", &buffer))
324 return error_func(&commit->object, FSCK_ERROR, "invalid format - expected 'tree' line");
325 if (get_sha1_hex(buffer, tree_sha1) || buffer[40] != '\n')
326 return error_func(&commit->object, FSCK_ERROR, "invalid 'tree' line format - bad sha1");
327 buffer += 41;
328 while (skip_prefix(buffer, "parent ", &buffer)) {
329 if (get_sha1_hex(buffer, sha1) || buffer[40] != '\n')
330 return error_func(&commit->object, FSCK_ERROR, "invalid 'parent' line format - bad sha1");
331 buffer += 41;
332 parent_line_count++;
333 }
334 graft = lookup_commit_graft(commit->object.sha1);
335 parent_count = commit_list_count(commit->parents);
336 if (graft) {
337 if (graft->nr_parent == -1 && !parent_count)
338 ; /* shallow commit */
339 else if (graft->nr_parent != parent_count)
340 return error_func(&commit->object, FSCK_ERROR, "graft objects missing");
341 } else {
342 if (parent_count != parent_line_count)
343 return error_func(&commit->object, FSCK_ERROR, "parent objects missing");
344 }
345 if (!skip_prefix(buffer, "author ", &buffer))
346 return error_func(&commit->object, FSCK_ERROR, "invalid format - expected 'author' line");
347 err = fsck_ident(&buffer, &commit->object, error_func);
348 if (err)
349 return err;
350 if (!skip_prefix(buffer, "committer ", &buffer))
351 return error_func(&commit->object, FSCK_ERROR, "invalid format - expected 'committer' line");
352 err = fsck_ident(&buffer, &commit->object, error_func);
353 if (err)
354 return err;
355 if (!commit->tree)
356 return error_func(&commit->object, FSCK_ERROR, "could not load commit's tree %s", sha1_to_hex(tree_sha1));
357
358 return 0;
359 }
360
361 static int fsck_commit(struct commit *commit, const char *data,
362 unsigned long size, fsck_error error_func)
363 {
364 const char *buffer = data ? data : get_commit_buffer(commit, &size);
365 int ret = fsck_commit_buffer(commit, buffer, size, error_func);
366 if (!data)
367 unuse_commit_buffer(commit, buffer);
368 return ret;
369 }
370
371 static int fsck_tag_buffer(struct tag *tag, const char *data,
372 unsigned long size, fsck_error error_func)
373 {
374 unsigned char sha1[20];
375 int ret = 0;
376 const char *buffer;
377 char *to_free = NULL, *eol;
378 struct strbuf sb = STRBUF_INIT;
379
380 if (data)
381 buffer = data;
382 else {
383 enum object_type type;
384
385 buffer = to_free =
386 read_sha1_file(tag->object.sha1, &type, &size);
387 if (!buffer)
388 return error_func(&tag->object, FSCK_ERROR,
389 "cannot read tag object");
390
391 if (type != OBJ_TAG) {
392 ret = error_func(&tag->object, FSCK_ERROR,
393 "expected tag got %s",
394 typename(type));
395 goto done;
396 }
397 }
398
399 if (verify_headers(buffer, size, &tag->object, error_func))
400 goto done;
401
402 if (!skip_prefix(buffer, "object ", &buffer)) {
403 ret = error_func(&tag->object, FSCK_ERROR, "invalid format - expected 'object' line");
404 goto done;
405 }
406 if (get_sha1_hex(buffer, sha1) || buffer[40] != '\n') {
407 ret = error_func(&tag->object, FSCK_ERROR, "invalid 'object' line format - bad sha1");
408 goto done;
409 }
410 buffer += 41;
411
412 if (!skip_prefix(buffer, "type ", &buffer)) {
413 ret = error_func(&tag->object, FSCK_ERROR, "invalid format - expected 'type' line");
414 goto done;
415 }
416 eol = strchr(buffer, '\n');
417 if (!eol) {
418 ret = error_func(&tag->object, FSCK_ERROR, "invalid format - unexpected end after 'type' line");
419 goto done;
420 }
421 if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
422 ret = error_func(&tag->object, FSCK_ERROR, "invalid 'type' value");
423 if (ret)
424 goto done;
425 buffer = eol + 1;
426
427 if (!skip_prefix(buffer, "tag ", &buffer)) {
428 ret = error_func(&tag->object, FSCK_ERROR, "invalid format - expected 'tag' line");
429 goto done;
430 }
431 eol = strchr(buffer, '\n');
432 if (!eol) {
433 ret = error_func(&tag->object, FSCK_ERROR, "invalid format - unexpected end after 'type' line");
434 goto done;
435 }
436 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
437 if (check_refname_format(sb.buf, 0))
438 error_func(&tag->object, FSCK_WARN, "invalid 'tag' name: %.*s",
439 (int)(eol - buffer), buffer);
440 buffer = eol + 1;
441
442 if (!skip_prefix(buffer, "tagger ", &buffer))
443 /* early tags do not contain 'tagger' lines; warn only */
444 error_func(&tag->object, FSCK_WARN, "invalid format - expected 'tagger' line");
445 else
446 ret = fsck_ident(&buffer, &tag->object, error_func);
447
448 done:
449 strbuf_release(&sb);
450 free(to_free);
451 return ret;
452 }
453
454 static int fsck_tag(struct tag *tag, const char *data,
455 unsigned long size, fsck_error error_func)
456 {
457 struct object *tagged = tag->tagged;
458
459 if (!tagged)
460 return error_func(&tag->object, FSCK_ERROR, "could not load tagged object");
461
462 return fsck_tag_buffer(tag, data, size, error_func);
463 }
464
465 int fsck_object(struct object *obj, void *data, unsigned long size,
466 int strict, fsck_error error_func)
467 {
468 if (!obj)
469 return error_func(obj, FSCK_ERROR, "no valid object to fsck");
470
471 if (obj->type == OBJ_BLOB)
472 return 0;
473 if (obj->type == OBJ_TREE)
474 return fsck_tree((struct tree *) obj, strict, error_func);
475 if (obj->type == OBJ_COMMIT)
476 return fsck_commit((struct commit *) obj, (const char *) data,
477 size, error_func);
478 if (obj->type == OBJ_TAG)
479 return fsck_tag((struct tag *) obj, (const char *) data,
480 size, error_func);
481
482 return error_func(obj, FSCK_ERROR, "unknown type '%d' (internal fsck error)",
483 obj->type);
484 }
485
486 int fsck_error_function(struct object *obj, int type, const char *fmt, ...)
487 {
488 va_list ap;
489 struct strbuf sb = STRBUF_INIT;
490
491 strbuf_addf(&sb, "object %s:", sha1_to_hex(obj->sha1));
492
493 va_start(ap, fmt);
494 strbuf_vaddf(&sb, fmt, ap);
495 va_end(ap);
496
497 error("%s", sb.buf);
498 strbuf_release(&sb);
499 return 1;
500 }