[PATCH] SHA1 naive collision checking
[git/repo.git] / fsck-cache.c
blobbb5cf196af7f06a3aea6fcfc5941565110c94b27
1 #include "cache.h"
3 #include <sys/types.h>
4 #include <dirent.h>
6 struct needs {
7 unsigned char parent[20];
8 unsigned char needs[20];
9 char tag[10];
12 struct seen {
13 unsigned char sha1[20];
14 char tag[10];
15 unsigned needed;
18 static struct needs *needs;
19 static struct seen *seen;
21 static int nr_seen, alloc_seen, nr_needs, alloc_needs;
24 * These two functions build up a graph in memory about
25 * what objects we've referenced, and found, and types..
27 static int compare_seen(const void *s1, const void *s2)
29 return memcmp(s1, s2, 20);
32 static int lookup_seen(unsigned char *sha1, char *tag)
34 int first = 0, last = nr_seen;
36 while (last > first) {
37 int next = (last + first) / 2;
38 struct seen *s = seen + next;
39 int cmp = memcmp(sha1, s->sha1, 20);
41 if (cmp < 0) {
42 last = next;
43 continue;
45 if (cmp > 0) {
46 first = next+1;
47 continue;
49 if (strcmp(tag, s->tag))
50 break;
51 s->needed++;
52 return 1;
54 return 0;
57 static void check_connectivity(void)
59 int i;
61 /* Sort the "seen" tags for quicker lookup */
62 qsort(seen, nr_seen, sizeof(struct seen), compare_seen);
64 /* Look up all the requirements, warn about missing objects.. */
65 for (i = 0; i < nr_needs; i++) {
66 struct needs *n = needs + i;
67 char hex[60];
69 if (lookup_seen(n->needs, n->tag))
70 continue;
71 strcpy(hex, sha1_to_hex(n->parent));
72 printf("missing %s: %s referenced by %s\n", n->tag, sha1_to_hex(n->needs), hex);
75 /* Tell the user about things not referenced.. */
76 for (i = 0; i < nr_seen; i++) {
77 struct seen *s = seen + i;
79 if (s->needed)
80 continue;
81 printf("unreferenced %s: %s\n", s->tag, sha1_to_hex(s->sha1));
85 static void mark_needs_sha1(unsigned char *parent, const char * tag, unsigned char *child)
87 struct needs *n;
89 if (nr_needs == alloc_needs) {
90 alloc_needs = alloc_nr(alloc_needs);
91 needs = realloc(needs, alloc_needs*sizeof(struct needs));
93 n = needs + nr_needs;
94 nr_needs++;
95 memcpy(n->parent, parent, 20);
96 memcpy(n->needs, child, 20);
97 strncpy(n->tag, tag, sizeof(n->tag));
100 static int mark_sha1_seen(unsigned char *sha1, char *tag)
102 struct seen *s;
104 if (nr_seen == alloc_seen) {
105 alloc_seen = alloc_nr(alloc_seen);
106 seen = realloc(seen, alloc_seen*sizeof(struct seen));
108 s = seen + nr_seen;
109 memset(s, 0, sizeof(*s));
110 nr_seen++;
111 memcpy(s->sha1, sha1, 20);
112 strncpy(s->tag, tag, sizeof(s->tag));
114 return 0;
117 static int fsck_tree(unsigned char *sha1, void *data, unsigned long size)
119 int warn_old_tree = 1;
121 while (size) {
122 int len = 1+strlen(data);
123 unsigned char *file_sha1 = data + len;
124 char *path = strchr(data, ' ');
125 unsigned int mode;
126 if (size < len + 20 || !path || sscanf(data, "%o", &mode) != 1)
127 return -1;
129 /* Warn about trees that don't do the recursive thing.. */
130 if (warn_old_tree && strchr(path, '/')) {
131 fprintf(stderr, "warning: fsck-cache: tree %s has full pathnames in it\n", sha1_to_hex(sha1));
132 warn_old_tree = 0;
135 data += len + 20;
136 size -= len + 20;
137 mark_needs_sha1(sha1, S_ISDIR(mode) ? "tree" : "blob", file_sha1);
139 return 0;
142 static int fsck_commit(unsigned char *sha1, void *data, unsigned long size)
144 int parents;
145 unsigned char tree_sha1[20];
146 unsigned char parent_sha1[20];
148 if (memcmp(data, "tree ", 5))
149 return -1;
150 if (get_sha1_hex(data + 5, tree_sha1) < 0)
151 return -1;
152 mark_needs_sha1(sha1, "tree", tree_sha1);
153 data += 5 + 40 + 1; /* "tree " + <hex sha1> + '\n' */
154 parents = 0;
155 while (!memcmp(data, "parent ", 7)) {
156 if (get_sha1_hex(data + 7, parent_sha1) < 0)
157 return -1;
158 mark_needs_sha1(sha1, "commit", parent_sha1);
159 data += 7 + 40 + 1; /* "parent " + <hex sha1> + '\n' */
160 parents++;
162 if (!parents)
163 printf("root: %s\n", sha1_to_hex(sha1));
164 return 0;
167 static int fsck_entry(unsigned char *sha1, char *tag, void *data, unsigned long size)
169 if (!strcmp(tag, "blob")) {
170 /* Nothing to check */;
171 } else if (!strcmp(tag, "tree")) {
172 if (fsck_tree(sha1, data, size) < 0)
173 return -1;
174 } else if (!strcmp(tag, "commit")) {
175 if (fsck_commit(sha1, data, size) < 0)
176 return -1;
177 } else
178 return -1;
179 return mark_sha1_seen(sha1, tag);
182 static int fsck_name(char *hex)
184 unsigned char sha1[20];
185 if (!get_sha1_hex(hex, sha1)) {
186 unsigned long mapsize;
187 void *map = map_sha1_file(sha1, &mapsize);
188 if (map) {
189 char type[100];
190 unsigned long size;
191 void *buffer = NULL;
192 if (!check_sha1_signature(sha1, map, mapsize))
193 buffer = unpack_sha1_file(map, mapsize, type, &size);
194 munmap(map, mapsize);
195 if (buffer && !fsck_entry(sha1, type, buffer, size))
196 return 0;
199 return -1;
202 static int fsck_dir(int i, char *path)
204 DIR *dir = opendir(path);
205 struct dirent *de;
207 if (!dir) {
208 fprintf(stderr, "missing sha1 directory '%s'", path);
209 return -1;
212 while ((de = readdir(dir)) != NULL) {
213 char name[100];
214 int len = strlen(de->d_name);
216 switch (len) {
217 case 2:
218 if (de->d_name[1] != '.')
219 break;
220 case 1:
221 if (de->d_name[0] != '.')
222 break;
223 continue;
224 case 38:
225 sprintf(name, "%02x", i);
226 memcpy(name+2, de->d_name, len+1);
227 if (!fsck_name(name))
228 continue;
230 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
232 closedir(dir);
233 return 0;
236 int main(int argc, char **argv)
238 int i;
239 char *sha1_dir;
241 if (argc != 1)
242 usage("fsck-cache");
243 sha1_dir = getenv(DB_ENVIRONMENT) ? : DEFAULT_DB_ENVIRONMENT;
244 for (i = 0; i < 256; i++) {
245 static char dir[4096];
246 sprintf(dir, "%s/%02x", sha1_dir, i);
247 fsck_dir(i, dir);
249 check_connectivity();
250 return 0;