symlinks.con commit fast-import: Improve robustness when D->F changes provided in wrong order (253fb5f)
   1#include "cache.h"
   2
   3/*
   4 * Returns the length (on a path component basis) of the longest
   5 * common prefix match of 'name_a' and 'name_b'.
   6 */
   7static int longest_path_match(const char *name_a, int len_a,
   8                              const char *name_b, int len_b,
   9                              int *previous_slash)
  10{
  11        int max_len, match_len = 0, match_len_prev = 0, i = 0;
  12
  13        max_len = len_a < len_b ? len_a : len_b;
  14        while (i < max_len && name_a[i] == name_b[i]) {
  15                if (name_a[i] == '/') {
  16                        match_len_prev = match_len;
  17                        match_len = i;
  18                }
  19                i++;
  20        }
  21        /*
  22         * Is 'name_b' a substring of 'name_a', the other way around,
  23         * or is 'name_a' and 'name_b' the exact same string?
  24         */
  25        if (i >= max_len && ((len_a > len_b && name_a[len_b] == '/') ||
  26                             (len_a < len_b && name_b[len_a] == '/') ||
  27                             (len_a == len_b))) {
  28                match_len_prev = match_len;
  29                match_len = i;
  30        }
  31        *previous_slash = match_len_prev;
  32        return match_len;
  33}
  34
  35static struct cache_def default_cache;
  36
  37static inline void reset_lstat_cache(struct cache_def *cache)
  38{
  39        cache->path[0] = '\0';
  40        cache->len = 0;
  41        cache->flags = 0;
  42        /*
  43         * The track_flags and prefix_len_stat_func members is only
  44         * set by the safeguard rule inside lstat_cache()
  45         */
  46}
  47
  48#define FL_DIR      (1 << 0)
  49#define FL_NOENT    (1 << 1)
  50#define FL_SYMLINK  (1 << 2)
  51#define FL_LSTATERR (1 << 3)
  52#define FL_ERR      (1 << 4)
  53#define FL_FULLPATH (1 << 5)
  54
  55/*
  56 * Check if name 'name' of length 'len' has a symlink leading
  57 * component, or if the directory exists and is real, or not.
  58 *
  59 * To speed up the check, some information is allowed to be cached.
  60 * This can be indicated by the 'track_flags' argument, which also can
  61 * be used to indicate that we should check the full path.
  62 *
  63 * The 'prefix_len_stat_func' parameter can be used to set the length
  64 * of the prefix, where the cache should use the stat() function
  65 * instead of the lstat() function to test each path component.
  66 */
  67static int lstat_cache(struct cache_def *cache, const char *name, int len,
  68                       int track_flags, int prefix_len_stat_func)
  69{
  70        int match_len, last_slash, last_slash_dir, previous_slash;
  71        int match_flags, ret_flags, save_flags, max_len, ret;
  72        struct stat st;
  73
  74        if (cache->track_flags != track_flags ||
  75            cache->prefix_len_stat_func != prefix_len_stat_func) {
  76                /*
  77                 * As a safeguard rule we clear the cache if the
  78                 * values of track_flags and/or prefix_len_stat_func
  79                 * does not match with the last supplied values.
  80                 */
  81                reset_lstat_cache(cache);
  82                cache->track_flags = track_flags;
  83                cache->prefix_len_stat_func = prefix_len_stat_func;
  84                match_len = last_slash = 0;
  85        } else {
  86                /*
  87                 * Check to see if we have a match from the cache for
  88                 * the 2 "excluding" path types.
  89                 */
  90                match_len = last_slash =
  91                        longest_path_match(name, len, cache->path, cache->len,
  92                                           &previous_slash);
  93                match_flags = cache->flags & track_flags & (FL_NOENT|FL_SYMLINK);
  94
  95                if (!(track_flags & FL_FULLPATH) && match_len == len)
  96                        match_len = last_slash = previous_slash;
  97
  98                if (match_flags && match_len == cache->len)
  99                        return match_flags;
 100                /*
 101                 * If we now have match_len > 0, we would know that
 102                 * the matched part will always be a directory.
 103                 *
 104                 * Also, if we are tracking directories and 'name' is
 105                 * a substring of the cache on a path component basis,
 106                 * we can return immediately.
 107                 */
 108                match_flags = track_flags & FL_DIR;
 109                if (match_flags && len == match_len)
 110                        return match_flags;
 111        }
 112
 113        /*
 114         * Okay, no match from the cache so far, so now we have to
 115         * check the rest of the path components.
 116         */
 117        ret_flags = FL_DIR;
 118        last_slash_dir = last_slash;
 119        max_len = len < PATH_MAX ? len : PATH_MAX;
 120        while (match_len < max_len) {
 121                do {
 122                        cache->path[match_len] = name[match_len];
 123                        match_len++;
 124                } while (match_len < max_len && name[match_len] != '/');
 125                if (match_len >= max_len && !(track_flags & FL_FULLPATH))
 126                        break;
 127                last_slash = match_len;
 128                cache->path[last_slash] = '\0';
 129
 130                if (last_slash <= prefix_len_stat_func)
 131                        ret = stat(cache->path, &st);
 132                else
 133                        ret = lstat(cache->path, &st);
 134
 135                if (ret) {
 136                        ret_flags = FL_LSTATERR;
 137                        if (errno == ENOENT)
 138                                ret_flags |= FL_NOENT;
 139                } else if (S_ISDIR(st.st_mode)) {
 140                        last_slash_dir = last_slash;
 141                        continue;
 142                } else if (S_ISLNK(st.st_mode)) {
 143                        ret_flags = FL_SYMLINK;
 144                } else {
 145                        ret_flags = FL_ERR;
 146                }
 147                break;
 148        }
 149
 150        /*
 151         * At the end update the cache.  Note that max 3 different
 152         * path types, FL_NOENT, FL_SYMLINK and FL_DIR, can be cached
 153         * for the moment!
 154         */
 155        save_flags = ret_flags & track_flags & (FL_NOENT|FL_SYMLINK);
 156        if (save_flags && last_slash > 0 && last_slash <= PATH_MAX) {
 157                cache->path[last_slash] = '\0';
 158                cache->len = last_slash;
 159                cache->flags = save_flags;
 160        } else if ((track_flags & FL_DIR) &&
 161                   last_slash_dir > 0 && last_slash_dir <= PATH_MAX) {
 162                /*
 163                 * We have a separate test for the directory case,
 164                 * since it could be that we have found a symlink or a
 165                 * non-existing directory and the track_flags says
 166                 * that we cannot cache this fact, so the cache would
 167                 * then have been left empty in this case.
 168                 *
 169                 * But if we are allowed to track real directories, we
 170                 * can still cache the path components before the last
 171                 * one (the found symlink or non-existing component).
 172                 */
 173                cache->path[last_slash_dir] = '\0';
 174                cache->len = last_slash_dir;
 175                cache->flags = FL_DIR;
 176        } else {
 177                reset_lstat_cache(cache);
 178        }
 179        return ret_flags;
 180}
 181
 182#define USE_ONLY_LSTAT  0
 183
 184/*
 185 * Return non-zero if path 'name' has a leading symlink component
 186 */
 187int threaded_has_symlink_leading_path(struct cache_def *cache, const char *name, int len)
 188{
 189        return lstat_cache(cache, name, len, FL_SYMLINK|FL_DIR, USE_ONLY_LSTAT) & FL_SYMLINK;
 190}
 191
 192/*
 193 * Return non-zero if path 'name' has a leading symlink component
 194 */
 195int has_symlink_leading_path(const char *name, int len)
 196{
 197        return threaded_has_symlink_leading_path(&default_cache, name, len);
 198}
 199
 200/*
 201 * Return non-zero if path 'name' has a leading symlink component or
 202 * if some leading path component does not exists.
 203 */
 204int has_symlink_or_noent_leading_path(const char *name, int len)
 205{
 206        struct cache_def *cache = &default_cache;       /* FIXME */
 207        return lstat_cache(cache, name, len,
 208                           FL_SYMLINK|FL_NOENT|FL_DIR, USE_ONLY_LSTAT) &
 209                (FL_SYMLINK|FL_NOENT);
 210}
 211
 212/*
 213 * Return non-zero if all path components of 'name' exists as a
 214 * directory.  If prefix_len > 0, we will test with the stat()
 215 * function instead of the lstat() function for a prefix length of
 216 * 'prefix_len', thus we then allow for symlinks in the prefix part as
 217 * long as those points to real existing directories.
 218 */
 219int has_dirs_only_path(const char *name, int len, int prefix_len)
 220{
 221        struct cache_def *cache = &default_cache;       /* FIXME */
 222        return lstat_cache(cache, name, len,
 223                           FL_DIR|FL_FULLPATH, prefix_len) &
 224                FL_DIR;
 225}
 226
 227static struct removal_def {
 228        char path[PATH_MAX];
 229        int len;
 230} removal;
 231
 232static void do_remove_scheduled_dirs(int new_len)
 233{
 234        while (removal.len > new_len) {
 235                removal.path[removal.len] = '\0';
 236                if (rmdir(removal.path))
 237                        break;
 238                do {
 239                        removal.len--;
 240                } while (removal.len > new_len &&
 241                         removal.path[removal.len] != '/');
 242        }
 243        removal.len = new_len;
 244}
 245
 246void schedule_dir_for_removal(const char *name, int len)
 247{
 248        int match_len, last_slash, i, previous_slash;
 249
 250        match_len = last_slash = i =
 251                longest_path_match(name, len, removal.path, removal.len,
 252                                   &previous_slash);
 253        /* Find last slash inside 'name' */
 254        while (i < len) {
 255                if (name[i] == '/')
 256                        last_slash = i;
 257                i++;
 258        }
 259
 260        /*
 261         * If we are about to go down the directory tree, we check if
 262         * we must first go upwards the tree, such that we then can
 263         * remove possible empty directories as we go upwards.
 264         */
 265        if (match_len < last_slash && match_len < removal.len)
 266                do_remove_scheduled_dirs(match_len);
 267        /*
 268         * If we go deeper down the directory tree, we only need to
 269         * save the new path components as we go down.
 270         */
 271        if (match_len < last_slash) {
 272                memcpy(&removal.path[match_len], &name[match_len],
 273                       last_slash - match_len);
 274                removal.len = last_slash;
 275        }
 276}
 277
 278void remove_scheduled_dirs(void)
 279{
 280        do_remove_scheduled_dirs(0);
 281}