gitweb: safely output binary files for 'blob_plain' action
[git] / fsck-objects.c
1 #include <sys/types.h>
2 #include <dirent.h>
3
4 #include "cache.h"
5 #include "commit.h"
6 #include "tree.h"
7 #include "blob.h"
8 #include "tag.h"
9 #include "refs.h"
10 #include "pack.h"
11 #include "cache-tree.h"
12 #include "tree-walk.h"
13
14 #define REACHABLE 0x0001
15 #define SEEN      0x0002
16
17 static int show_root = 0;
18 static int show_tags = 0;
19 static int show_unreachable = 0;
20 static int check_full = 0;
21 static int check_strict = 0;
22 static int keep_cache_objects = 0;
23 static unsigned char head_sha1[20];
24
25 #ifdef NO_D_INO_IN_DIRENT
26 #define SORT_DIRENT 0
27 #define DIRENT_SORT_HINT(de) 0
28 #else
29 #define SORT_DIRENT 1
30 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
31 #endif
32
33 static void objreport(struct object *obj, const char *severity,
34                       const char *err, va_list params)
35 {
36         fprintf(stderr, "%s in %s %s: ",
37                 severity, typename(obj->type), sha1_to_hex(obj->sha1));
38         vfprintf(stderr, err, params);
39         fputs("\n", stderr);
40 }
41
42 static int objerror(struct object *obj, const char *err, ...)
43 {
44         va_list params;
45         va_start(params, err);
46         objreport(obj, "error", err, params);
47         va_end(params);
48         return -1;
49 }
50
51 static int objwarning(struct object *obj, const char *err, ...)
52 {
53         va_list params;
54         va_start(params, err);
55         objreport(obj, "warning", err, params);
56         va_end(params);
57         return -1;
58 }
59
60
61 static void check_connectivity(void)
62 {
63         int i;
64
65         /* Look up all the requirements, warn about missing objects.. */
66         for (i = 0; i < obj_allocs; i++) {
67                 const struct object_refs *refs;
68                 struct object *obj = objs[i];
69
70                 if (!obj)
71                         continue;
72
73                 if (!obj->parsed) {
74                         if (has_sha1_file(obj->sha1))
75                                 ; /* it is in pack */
76                         else
77                                 printf("missing %s %s\n",
78                                        typename(obj->type), sha1_to_hex(obj->sha1));
79                         continue;
80                 }
81
82                 refs = lookup_object_refs(obj);
83                 if (refs) {
84                         unsigned j;
85                         for (j = 0; j < refs->count; j++) {
86                                 struct object *ref = refs->ref[j];
87                                 if (ref->parsed ||
88                                     (has_sha1_file(ref->sha1)))
89                                         continue;
90                                 printf("broken link from %7s %s\n",
91                                        typename(obj->type), sha1_to_hex(obj->sha1));
92                                 printf("              to %7s %s\n",
93                                        typename(ref->type), sha1_to_hex(ref->sha1));
94                         }
95                 }
96
97                 if (show_unreachable && !(obj->flags & REACHABLE)) {
98                         printf("unreachable %s %s\n",
99                                typename(obj->type), sha1_to_hex(obj->sha1));
100                         continue;
101                 }
102
103                 if (!obj->used) {
104                         printf("dangling %s %s\n", typename(obj->type),
105                                sha1_to_hex(obj->sha1));
106                 }
107         }
108 }
109
110 /*
111  * The entries in a tree are ordered in the _path_ order,
112  * which means that a directory entry is ordered by adding
113  * a slash to the end of it.
114  *
115  * So a directory called "a" is ordered _after_ a file
116  * called "a.c", because "a/" sorts after "a.c".
117  */
118 #define TREE_UNORDERED (-1)
119 #define TREE_HAS_DUPS  (-2)
120
121 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
122 {
123         int len1 = strlen(name1);
124         int len2 = strlen(name2);
125         int len = len1 < len2 ? len1 : len2;
126         unsigned char c1, c2;
127         int cmp;
128
129         cmp = memcmp(name1, name2, len);
130         if (cmp < 0)
131                 return 0;
132         if (cmp > 0)
133                 return TREE_UNORDERED;
134
135         /*
136          * Ok, the first <len> characters are the same.
137          * Now we need to order the next one, but turn
138          * a '\0' into a '/' for a directory entry.
139          */
140         c1 = name1[len];
141         c2 = name2[len];
142         if (!c1 && !c2)
143                 /*
144                  * git-write-tree used to write out a nonsense tree that has
145                  * entries with the same name, one blob and one tree.  Make
146                  * sure we do not have duplicate entries.
147                  */
148                 return TREE_HAS_DUPS;
149         if (!c1 && S_ISDIR(mode1))
150                 c1 = '/';
151         if (!c2 && S_ISDIR(mode2))
152                 c2 = '/';
153         return c1 < c2 ? 0 : TREE_UNORDERED;
154 }
155
156 static int fsck_tree(struct tree *item)
157 {
158         int retval;
159         int has_full_path = 0;
160         int has_zero_pad = 0;
161         int has_bad_modes = 0;
162         int has_dup_entries = 0;
163         int not_properly_sorted = 0;
164         struct tree_desc desc;
165         unsigned o_mode;
166         const char *o_name;
167         const unsigned char *o_sha1;
168
169         desc.buf = item->buffer;
170         desc.size = item->size;
171
172         o_mode = 0;
173         o_name = NULL;
174         o_sha1 = NULL;
175         while (desc.size) {
176                 unsigned mode;
177                 const char *name;
178                 const unsigned char *sha1;
179
180                 sha1 = tree_entry_extract(&desc, &name, &mode);
181
182                 if (strchr(name, '/'))
183                         has_full_path = 1;
184                 has_zero_pad |= *(char *)desc.buf == '0';
185                 update_tree_entry(&desc);
186
187                 switch (mode) {
188                 /*
189                  * Standard modes..
190                  */
191                 case S_IFREG | 0755:
192                 case S_IFREG | 0644:
193                 case S_IFLNK:
194                 case S_IFDIR:
195                         break;
196                 /*
197                  * This is nonstandard, but we had a few of these
198                  * early on when we honored the full set of mode
199                  * bits..
200                  */
201                 case S_IFREG | 0664:
202                         if (!check_strict)
203                                 break;
204                 default:
205                         has_bad_modes = 1;
206                 }
207
208                 if (o_name) {
209                         switch (verify_ordered(o_mode, o_name, mode, name)) {
210                         case TREE_UNORDERED:
211                                 not_properly_sorted = 1;
212                                 break;
213                         case TREE_HAS_DUPS:
214                                 has_dup_entries = 1;
215                                 break;
216                         default:
217                                 break;
218                         }
219                 }
220
221                 o_mode = mode;
222                 o_name = name;
223                 o_sha1 = sha1;
224         }
225         free(item->buffer);
226         item->buffer = NULL;
227
228         retval = 0;
229         if (has_full_path) {
230                 objwarning(&item->object, "contains full pathnames");
231         }
232         if (has_zero_pad) {
233                 objwarning(&item->object, "contains zero-padded file modes");
234         }
235         if (has_bad_modes) {
236                 objwarning(&item->object, "contains bad file modes");
237         }
238         if (has_dup_entries) {
239                 retval = objerror(&item->object, "contains duplicate file entries");
240         }
241         if (not_properly_sorted) {
242                 retval = objerror(&item->object, "not properly sorted");
243         }
244         return retval;
245 }
246
247 static int fsck_commit(struct commit *commit)
248 {
249         char *buffer = commit->buffer;
250         unsigned char tree_sha1[20], sha1[20];
251
252         if (memcmp(buffer, "tree ", 5))
253                 return objerror(&commit->object, "invalid format - expected 'tree' line");
254         if (get_sha1_hex(buffer+5, tree_sha1) || buffer[45] != '\n')
255                 return objerror(&commit->object, "invalid 'tree' line format - bad sha1");
256         buffer += 46;
257         while (!memcmp(buffer, "parent ", 7)) {
258                 if (get_sha1_hex(buffer+7, sha1) || buffer[47] != '\n')
259                         return objerror(&commit->object, "invalid 'parent' line format - bad sha1");
260                 buffer += 48;
261         }
262         if (memcmp(buffer, "author ", 7))
263                 return objerror(&commit->object, "invalid format - expected 'author' line");
264         free(commit->buffer);
265         commit->buffer = NULL;
266         if (!commit->tree)
267                 return objerror(&commit->object, "could not load commit's tree %s", tree_sha1);
268         if (!commit->parents && show_root)
269                 printf("root %s\n", sha1_to_hex(commit->object.sha1));
270         if (!commit->date)
271                 printf("bad commit date in %s\n", 
272                        sha1_to_hex(commit->object.sha1));
273         return 0;
274 }
275
276 static int fsck_tag(struct tag *tag)
277 {
278         struct object *tagged = tag->tagged;
279
280         if (!tagged) {
281                 return objerror(&tag->object, "could not load tagged object");
282         }
283         if (!show_tags)
284                 return 0;
285
286         printf("tagged %s %s", typename(tagged->type), sha1_to_hex(tagged->sha1));
287         printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
288         return 0;
289 }
290
291 static int fsck_sha1(unsigned char *sha1)
292 {
293         struct object *obj = parse_object(sha1);
294         if (!obj)
295                 return error("%s: object not found", sha1_to_hex(sha1));
296         if (obj->flags & SEEN)
297                 return 0;
298         obj->flags |= SEEN;
299         if (obj->type == TYPE_BLOB)
300                 return 0;
301         if (obj->type == TYPE_TREE)
302                 return fsck_tree((struct tree *) obj);
303         if (obj->type == TYPE_COMMIT)
304                 return fsck_commit((struct commit *) obj);
305         if (obj->type == TYPE_TAG)
306                 return fsck_tag((struct tag *) obj);
307         /* By now, parse_object() would've returned NULL instead. */
308         return objerror(obj, "unknown type '%d' (internal fsck error)", obj->type);
309 }
310
311 /*
312  * This is the sorting chunk size: make it reasonably
313  * big so that we can sort well..
314  */
315 #define MAX_SHA1_ENTRIES (1024)
316
317 struct sha1_entry {
318         unsigned long ino;
319         unsigned char sha1[20];
320 };
321
322 static struct {
323         unsigned long nr;
324         struct sha1_entry *entry[MAX_SHA1_ENTRIES];
325 } sha1_list;
326
327 static int ino_compare(const void *_a, const void *_b)
328 {
329         const struct sha1_entry *a = _a, *b = _b;
330         unsigned long ino1 = a->ino, ino2 = b->ino;
331         return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
332 }
333
334 static void fsck_sha1_list(void)
335 {
336         int i, nr = sha1_list.nr;
337
338         if (SORT_DIRENT)
339                 qsort(sha1_list.entry, nr,
340                       sizeof(struct sha1_entry *), ino_compare);
341         for (i = 0; i < nr; i++) {
342                 struct sha1_entry *entry = sha1_list.entry[i];
343                 unsigned char *sha1 = entry->sha1;
344
345                 sha1_list.entry[i] = NULL;
346                 fsck_sha1(sha1);
347                 free(entry);
348         }
349         sha1_list.nr = 0;
350 }
351
352 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
353 {
354         struct sha1_entry *entry = xmalloc(sizeof(*entry));
355         int nr;
356
357         entry->ino = ino;
358         memcpy(entry->sha1, sha1, 20);
359         nr = sha1_list.nr;
360         if (nr == MAX_SHA1_ENTRIES) {
361                 fsck_sha1_list();
362                 nr = 0;
363         }
364         sha1_list.entry[nr] = entry;
365         sha1_list.nr = ++nr;
366 }
367
368 static int fsck_dir(int i, char *path)
369 {
370         DIR *dir = opendir(path);
371         struct dirent *de;
372
373         if (!dir)
374                 return 0;
375
376         while ((de = readdir(dir)) != NULL) {
377                 char name[100];
378                 unsigned char sha1[20];
379                 int len = strlen(de->d_name);
380
381                 switch (len) {
382                 case 2:
383                         if (de->d_name[1] != '.')
384                                 break;
385                 case 1:
386                         if (de->d_name[0] != '.')
387                                 break;
388                         continue;
389                 case 38:
390                         sprintf(name, "%02x", i);
391                         memcpy(name+2, de->d_name, len+1);
392                         if (get_sha1_hex(name, sha1) < 0)
393                                 break;
394                         add_sha1_list(sha1, DIRENT_SORT_HINT(de));
395                         continue;
396                 }
397                 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
398         }
399         closedir(dir);
400         return 0;
401 }
402
403 static int default_refs = 0;
404
405 static int fsck_handle_ref(const char *refname, const unsigned char *sha1)
406 {
407         struct object *obj;
408
409         obj = lookup_object(sha1);
410         if (!obj) {
411                 if (has_sha1_file(sha1)) {
412                         default_refs++;
413                         return 0; /* it is in a pack */
414                 }
415                 error("%s: invalid sha1 pointer %s", refname, sha1_to_hex(sha1));
416                 /* We'll continue with the rest despite the error.. */
417                 return 0;
418         }
419         default_refs++;
420         obj->used = 1;
421         mark_reachable(obj, REACHABLE);
422         return 0;
423 }
424
425 static void get_default_heads(void)
426 {
427         for_each_ref(fsck_handle_ref);
428         if (!default_refs)
429                 die("No default references");
430 }
431
432 static void fsck_object_dir(const char *path)
433 {
434         int i;
435         for (i = 0; i < 256; i++) {
436                 static char dir[4096];
437                 sprintf(dir, "%s/%02x", path, i);
438                 fsck_dir(i, dir);
439         }
440         fsck_sha1_list();
441 }
442
443 static int fsck_head_link(void)
444 {
445         unsigned char sha1[20];
446         const char *git_HEAD = strdup(git_path("HEAD"));
447         const char *git_refs_heads_master = resolve_ref(git_HEAD, sha1, 1);
448         int pfxlen = strlen(git_HEAD) - 4; /* strip .../.git/ part */
449
450         if (!git_refs_heads_master)
451                 return error("HEAD is not a symbolic ref");
452         if (strncmp(git_refs_heads_master + pfxlen, "refs/heads/", 11))
453                 return error("HEAD points to something strange (%s)",
454                              git_refs_heads_master + pfxlen);
455         if (!memcmp(null_sha1, sha1, 20))
456                 return error("HEAD: not a valid git pointer");
457         return 0;
458 }
459
460 static int fsck_cache_tree(struct cache_tree *it)
461 {
462         int i;
463         int err = 0;
464
465         if (0 <= it->entry_count) {
466                 struct object *obj = parse_object(it->sha1);
467                 if (!obj) {
468                         error("%s: invalid sha1 pointer in cache-tree",
469                               sha1_to_hex(it->sha1));
470                         return 1;
471                 }
472                 mark_reachable(obj, REACHABLE);
473                 obj->used = 1;
474                 if (obj->type != TYPE_TREE)
475                         err |= objerror(obj, "non-tree in cache-tree");
476         }
477         for (i = 0; i < it->subtree_nr; i++)
478                 err |= fsck_cache_tree(it->down[i]->cache_tree);
479         return err;
480 }
481
482 int main(int argc, char **argv)
483 {
484         int i, heads;
485
486         track_object_refs = 1;
487         setup_git_directory();
488
489         for (i = 1; i < argc; i++) {
490                 const char *arg = argv[i];
491
492                 if (!strcmp(arg, "--unreachable")) {
493                         show_unreachable = 1;
494                         continue;
495                 }
496                 if (!strcmp(arg, "--tags")) {
497                         show_tags = 1;
498                         continue;
499                 }
500                 if (!strcmp(arg, "--root")) {
501                         show_root = 1;
502                         continue;
503                 }
504                 if (!strcmp(arg, "--cache")) {
505                         keep_cache_objects = 1;
506                         continue;
507                 }
508                 if (!strcmp(arg, "--full")) {
509                         check_full = 1;
510                         continue;
511                 }
512                 if (!strcmp(arg, "--strict")) {
513                         check_strict = 1;
514                         continue;
515                 }
516                 if (*arg == '-')
517                         usage("git-fsck-objects [--tags] [--root] [[--unreachable] [--cache] [--full] [--strict] <head-sha1>*]");
518         }
519
520         fsck_head_link();
521         fsck_object_dir(get_object_directory());
522         if (check_full) {
523                 struct alternate_object_database *alt;
524                 struct packed_git *p;
525                 prepare_alt_odb();
526                 for (alt = alt_odb_list; alt; alt = alt->next) {
527                         char namebuf[PATH_MAX];
528                         int namelen = alt->name - alt->base;
529                         memcpy(namebuf, alt->base, namelen);
530                         namebuf[namelen - 1] = 0;
531                         fsck_object_dir(namebuf);
532                 }
533                 prepare_packed_git();
534                 for (p = packed_git; p; p = p->next)
535                         /* verify gives error messages itself */
536                         verify_pack(p, 0);
537
538                 for (p = packed_git; p; p = p->next) {
539                         int num = num_packed_objects(p);
540                         for (i = 0; i < num; i++) {
541                                 unsigned char sha1[20];
542                                 nth_packed_object_sha1(p, i, sha1);
543                                 fsck_sha1(sha1);
544                         }
545                 }
546         }
547
548         heads = 0;
549         for (i = 1; i < argc; i++) {
550                 const char *arg = argv[i]; 
551
552                 if (*arg == '-')
553                         continue;
554
555                 if (!get_sha1(arg, head_sha1)) {
556                         struct object *obj = lookup_object(head_sha1);
557
558                         /* Error is printed by lookup_object(). */
559                         if (!obj)
560                                 continue;
561
562                         obj->used = 1;
563                         mark_reachable(obj, REACHABLE);
564                         heads++;
565                         continue;
566                 }
567                 error("invalid parameter: expected sha1, got '%s'", arg);
568         }
569
570         /*
571          * If we've not been given any explicit head information, do the
572          * default ones from .git/refs. We also consider the index file
573          * in this case (ie this implies --cache).
574          */
575         if (!heads) {
576                 get_default_heads();
577                 keep_cache_objects = 1;
578         }
579
580         if (keep_cache_objects) {
581                 int i;
582                 read_cache();
583                 for (i = 0; i < active_nr; i++) {
584                         struct blob *blob = lookup_blob(active_cache[i]->sha1);
585                         struct object *obj;
586                         if (!blob)
587                                 continue;
588                         obj = &blob->object;
589                         obj->used = 1;
590                         mark_reachable(obj, REACHABLE);
591                 }
592                 if (active_cache_tree)
593                         fsck_cache_tree(active_cache_tree);
594         }
595
596         check_connectivity();
597         return 0;
598 }