hash.h

   1 #ifndef HASH_H
   2 #define HASH_H
   3
   4 #if defined(SHA1_APPLE)
   5 #include <CommonCrypto/CommonDigest.h>
   6 #elif defined(SHA1_OPENSSL)
   7 #  include <openssl/sha.h>
   8 #  if defined(OPENSSL_API_LEVEL) && OPENSSL_API_LEVEL >= 3
   9 #    define SHA1_NEEDS_CLONE_HELPER
  10 #    include "sha1/openssl.h"
  11 #  endif
  12 #elif defined(SHA1_DC)
  13 #include "sha1dc_git.h"
  14 #else /* SHA1_BLK */
  15 #include "block-sha1/sha1.h"
  16 #endif
  17
  18 #if defined(SHA256_NETTLE)
  19 #include "sha256/nettle.h"
  20 #elif defined(SHA256_GCRYPT)
  21 #define SHA256_NEEDS_CLONE_HELPER
  22 #include "sha256/gcrypt.h"
  23 #elif defined(SHA256_OPENSSL)
  24 #  include <openssl/sha.h>
  25 #  if defined(OPENSSL_API_LEVEL) && OPENSSL_API_LEVEL >= 3
  26 #    define SHA256_NEEDS_CLONE_HELPER
  27 #    include "sha256/openssl.h"
  28 #  endif
  29 #else
  30 #include "sha256/block/sha256.h"
  31 #endif
  32
  33 #ifndef platform_SHA_CTX
  34 /*
  35  * platform's underlying implementation of SHA-1; could be OpenSSL,
  36  * blk_SHA, Apple CommonCrypto, etc...  Note that the relevant
  37  * SHA-1 header may have already defined platform_SHA_CTX for our
  38  * own implementations like block-sha1, so we list
  39  * the default for OpenSSL compatible SHA-1 implementations here.
  40  */
  41 #define platform_SHA_CTX        SHA_CTX
  42 #define platform_SHA1_Init      SHA1_Init
  43 #define platform_SHA1_Update    SHA1_Update
  44 #define platform_SHA1_Final     SHA1_Final
  45 #endif
  46
  47 #define git_SHA_CTX             platform_SHA_CTX
  48 #define git_SHA1_Init           platform_SHA1_Init
  49 #define git_SHA1_Update         platform_SHA1_Update
  50 #define git_SHA1_Final          platform_SHA1_Final
  51
  52 #ifdef platform_SHA1_Clone
  53 #define git_SHA1_Clone  platform_SHA1_Clone
  54 #endif
  55
  56 #ifndef platform_SHA256_CTX
  57 #define platform_SHA256_CTX     SHA256_CTX
  58 #define platform_SHA256_Init    SHA256_Init
  59 #define platform_SHA256_Update  SHA256_Update
  60 #define platform_SHA256_Final   SHA256_Final
  61 #endif
  62
  63 #define git_SHA256_CTX          platform_SHA256_CTX
  64 #define git_SHA256_Init         platform_SHA256_Init
  65 #define git_SHA256_Update       platform_SHA256_Update
  66 #define git_SHA256_Final        platform_SHA256_Final
  67
  68 #ifdef platform_SHA256_Clone
  69 #define git_SHA256_Clone        platform_SHA256_Clone
  70 #endif
  71
  72 #ifdef SHA1_MAX_BLOCK_SIZE
  73 #include "compat/sha1-chunked.h"
  74 #undef git_SHA1_Update
  75 #define git_SHA1_Update         git_SHA1_Update_Chunked
  76 #endif
  77
  78 #ifndef SHA1_NEEDS_CLONE_HELPER
  79 static inline void git_SHA1_Clone(git_SHA_CTX *dst, const git_SHA_CTX *src)
  80 {
  81         memcpy(dst, src, sizeof(*dst));
  82 }
  83 #endif
  84
  85 #ifndef SHA256_NEEDS_CLONE_HELPER
  86 static inline void git_SHA256_Clone(git_SHA256_CTX *dst, const git_SHA256_CTX *src)
  87 {
  88         memcpy(dst, src, sizeof(*dst));
  89 }
  90 #endif
  91
  92 /*
  93  * Note that these constants are suitable for indexing the hash_algos array and
  94  * comparing against each other, but are otherwise arbitrary, so they should not
  95  * be exposed to the user or serialized to disk.  To know whether a
  96  * git_hash_algo struct points to some usable hash function, test the format_id
  97  * field for being non-zero.  Use the name field for user-visible situations and
  98  * the format_id field for fixed-length fields on disk.
  99  */
 100 /* An unknown hash function. */
 101 #define GIT_HASH_UNKNOWN 0
 102 /* SHA-1 */
 103 #define GIT_HASH_SHA1 1
 104 /* SHA-256  */
 105 #define GIT_HASH_SHA256 2
 106 /* Number of algorithms supported (including unknown). */
 107 #define GIT_HASH_NALGOS (GIT_HASH_SHA256 + 1)
 108
 109 /* "sha1", big-endian */
 110 #define GIT_SHA1_FORMAT_ID 0x73686131
 111
 112 /* The length in bytes and in hex digits of an object name (SHA-1 value). */
 113 #define GIT_SHA1_RAWSZ 20
 114 #define GIT_SHA1_HEXSZ (2 * GIT_SHA1_RAWSZ)
 115 /* The block size of SHA-1. */
 116 #define GIT_SHA1_BLKSZ 64
 117
 118 /* "s256", big-endian */
 119 #define GIT_SHA256_FORMAT_ID 0x73323536
 120
 121 /* The length in bytes and in hex digits of an object name (SHA-256 value). */
 122 #define GIT_SHA256_RAWSZ 32
 123 #define GIT_SHA256_HEXSZ (2 * GIT_SHA256_RAWSZ)
 124 /* The block size of SHA-256. */
 125 #define GIT_SHA256_BLKSZ 64
 126
 127 /* The length in byte and in hex digits of the largest possible hash value. */
 128 #define GIT_MAX_RAWSZ GIT_SHA256_RAWSZ
 129 #define GIT_MAX_HEXSZ GIT_SHA256_HEXSZ
 130 /* The largest possible block size for any supported hash. */
 131 #define GIT_MAX_BLKSZ GIT_SHA256_BLKSZ
 132
 133 struct object_id {
 134         unsigned char hash[GIT_MAX_RAWSZ];
 135         int algo;       /* XXX requires 4-byte alignment */
 136 };
 137
 138 #define GET_OID_QUIETLY           01
 139 #define GET_OID_COMMIT            02
 140 #define GET_OID_COMMITTISH        04
 141 #define GET_OID_TREE             010
 142 #define GET_OID_TREEISH          020
 143 #define GET_OID_BLOB             040
 144 #define GET_OID_FOLLOW_SYMLINKS 0100
 145 #define GET_OID_RECORD_PATH     0200
 146 #define GET_OID_ONLY_TO_DIE    04000
 147 #define GET_OID_REQUIRE_PATH  010000
 148 #define GET_OID_HASH_ANY      020000
 149
 150 #define GET_OID_DISAMBIGUATORS \
 151         (GET_OID_COMMIT | GET_OID_COMMITTISH | \
 152         GET_OID_TREE | GET_OID_TREEISH | \
 153         GET_OID_BLOB)
 154
 155 enum get_oid_result {
 156         FOUND = 0,
 157         MISSING_OBJECT = -1, /* The requested object is missing */
 158         SHORT_NAME_AMBIGUOUS = -2,
 159         /* The following only apply when symlinks are followed */
 160         DANGLING_SYMLINK = -4, /*
 161                                 * The initial symlink is there, but
 162                                 * (transitively) points to a missing
 163                                 * in-tree file
 164                                 */
 165         SYMLINK_LOOP = -5,
 166         NOT_DIR = -6, /*
 167                        * Somewhere along the symlink chain, a path is
 168                        * requested which contains a file as a
 169                        * non-final element.
 170                        */
 171 };
 172
 173 #ifdef USE_THE_REPOSITORY_VARIABLE
 174 # include "repository.h"
 175 # define the_hash_algo the_repository->hash_algo
 176 #endif
 177
 178 /* A suitably aligned type for stack allocations of hash contexts. */
 179 union git_hash_ctx {
 180         git_SHA_CTX sha1;
 181         git_SHA256_CTX sha256;
 182 };
 183 typedef union git_hash_ctx git_hash_ctx;
 184
 185 typedef void (*git_hash_init_fn)(git_hash_ctx *ctx);
 186 typedef void (*git_hash_clone_fn)(git_hash_ctx *dst, const git_hash_ctx *src);
 187 typedef void (*git_hash_update_fn)(git_hash_ctx *ctx, const void *in, size_t len);
 188 typedef void (*git_hash_final_fn)(unsigned char *hash, git_hash_ctx *ctx);
 189 typedef void (*git_hash_final_oid_fn)(struct object_id *oid, git_hash_ctx *ctx);
 190
 191 struct git_hash_algo {
 192         /*
 193          * The name of the algorithm, as appears in the config file and in
 194          * messages.
 195          */
 196         const char *name;
 197
 198         /* A four-byte version identifier, used in pack indices. */
 199         uint32_t format_id;
 200
 201         /* The length of the hash in binary. */
 202         size_t rawsz;
 203
 204         /* The length of the hash in hex characters. */
 205         size_t hexsz;
 206
 207         /* The block size of the hash. */
 208         size_t blksz;
 209
 210         /* The hash initialization function. */
 211         git_hash_init_fn init_fn;
 212
 213         /* The hash context cloning function. */
 214         git_hash_clone_fn clone_fn;
 215
 216         /* The hash update function. */
 217         git_hash_update_fn update_fn;
 218
 219         /* The hash finalization function. */
 220         git_hash_final_fn final_fn;
 221
 222         /* The hash finalization function for object IDs. */
 223         git_hash_final_oid_fn final_oid_fn;
 224
 225         /* The OID of the empty tree. */
 226         const struct object_id *empty_tree;
 227
 228         /* The OID of the empty blob. */
 229         const struct object_id *empty_blob;
 230
 231         /* The all-zeros OID. */
 232         const struct object_id *null_oid;
 233 };
 234 extern const struct git_hash_algo hash_algos[GIT_HASH_NALGOS];
 235
 236 /*
 237  * Return a GIT_HASH_* constant based on the name.  Returns GIT_HASH_UNKNOWN if
 238  * the name doesn't match a known algorithm.
 239  */
 240 int hash_algo_by_name(const char *name);
 241 /* Identical, except based on the format ID. */
 242 int hash_algo_by_id(uint32_t format_id);
 243 /* Identical, except based on the length. */
 244 int hash_algo_by_length(int len);
 245 /* Identical, except for a pointer to struct git_hash_algo. */
 246 static inline int hash_algo_by_ptr(const struct git_hash_algo *p)
 247 {
 248         return p - hash_algos;
 249 }
 250
 251 const struct object_id *null_oid(void);
 252
 253 static inline int hashcmp(const unsigned char *sha1, const unsigned char *sha2, const struct git_hash_algo *algop)
 254 {
 255         /*
 256          * Teach the compiler that there are only two possibilities of hash size
 257          * here, so that it can optimize for this case as much as possible.
 258          */
 259         if (algop->rawsz == GIT_MAX_RAWSZ)
 260                 return memcmp(sha1, sha2, GIT_MAX_RAWSZ);
 261         return memcmp(sha1, sha2, GIT_SHA1_RAWSZ);
 262 }
 263
 264 static inline int hasheq(const unsigned char *sha1, const unsigned char *sha2, const struct git_hash_algo *algop)
 265 {
 266         /*
 267          * We write this here instead of deferring to hashcmp so that the
 268          * compiler can properly inline it and avoid calling memcmp.
 269          */
 270         if (algop->rawsz == GIT_MAX_RAWSZ)
 271                 return !memcmp(sha1, sha2, GIT_MAX_RAWSZ);
 272         return !memcmp(sha1, sha2, GIT_SHA1_RAWSZ);
 273 }
 274
 275 static inline void hashcpy(unsigned char *sha_dst, const unsigned char *sha_src,
 276                            const struct git_hash_algo *algop)
 277 {
 278         memcpy(sha_dst, sha_src, algop->rawsz);
 279 }
 280
 281 static inline void hashclr(unsigned char *hash, const struct git_hash_algo *algop)
 282 {
 283         memset(hash, 0, algop->rawsz);
 284 }
 285
 286 static inline int oidcmp(const struct object_id *oid1, const struct object_id *oid2)
 287 {
 288         return memcmp(oid1->hash, oid2->hash, GIT_MAX_RAWSZ);
 289 }
 290
 291 static inline int oideq(const struct object_id *oid1, const struct object_id *oid2)
 292 {
 293         return !memcmp(oid1->hash, oid2->hash, GIT_MAX_RAWSZ);
 294 }
 295
 296 static inline void oidcpy(struct object_id *dst, const struct object_id *src)
 297 {
 298         memcpy(dst->hash, src->hash, GIT_MAX_RAWSZ);
 299         dst->algo = src->algo;
 300 }
 301
 302 static inline void oidread(struct object_id *oid, const unsigned char *hash,
 303                            const struct git_hash_algo *algop)
 304 {
 305         memcpy(oid->hash, hash, algop->rawsz);
 306         if (algop->rawsz < GIT_MAX_RAWSZ)
 307                 memset(oid->hash + algop->rawsz, 0, GIT_MAX_RAWSZ - algop->rawsz);
 308         oid->algo = hash_algo_by_ptr(algop);
 309 }
 310
 311 static inline void oidclr(struct object_id *oid,
 312                           const struct git_hash_algo *algop)
 313 {
 314         memset(oid->hash, 0, GIT_MAX_RAWSZ);
 315         oid->algo = hash_algo_by_ptr(algop);
 316 }
 317
 318 static inline struct object_id *oiddup(const struct object_id *src)
 319 {
 320         struct object_id *dst = xmalloc(sizeof(struct object_id));
 321         oidcpy(dst, src);
 322         return dst;
 323 }
 324
 325 static inline void oid_set_algo(struct object_id *oid, const struct git_hash_algo *algop)
 326 {
 327         oid->algo = hash_algo_by_ptr(algop);
 328 }
 329
 330 /*
 331  * Converts a cryptographic hash (e.g. SHA-1) into an int-sized hash code
 332  * for use in hash tables. Cryptographic hashes are supposed to have
 333  * uniform distribution, so in contrast to `memhash()`, this just copies
 334  * the first `sizeof(int)` bytes without shuffling any bits. Note that
 335  * the results will be different on big-endian and little-endian
 336  * platforms, so they should not be stored or transferred over the net.
 337  */
 338 static inline unsigned int oidhash(const struct object_id *oid)
 339 {
 340         /*
 341          * Equivalent to 'return *(unsigned int *)oid->hash;', but safe on
 342          * platforms that don't support unaligned reads.
 343          */
 344         unsigned int hash;
 345         memcpy(&hash, oid->hash, sizeof(hash));
 346         return hash;
 347 }
 348
 349 static inline int is_null_oid(const struct object_id *oid)
 350 {
 351         static const unsigned char null_hash[GIT_MAX_RAWSZ];
 352         return !memcmp(oid->hash, null_hash, GIT_MAX_RAWSZ);
 353 }
 354
 355 const char *empty_tree_oid_hex(const struct git_hash_algo *algop);
 356
 357 static inline int is_empty_blob_oid(const struct object_id *oid,
 358                                     const struct git_hash_algo *algop)
 359 {
 360         return oideq(oid, algop->empty_blob);
 361 }
 362
 363 static inline int is_empty_tree_oid(const struct object_id *oid,
 364                                     const struct git_hash_algo *algop)
 365 {
 366         return oideq(oid, algop->empty_tree);
 367 }
 368
 369 #endif