read-cache: avoid misaligned reads in index v4
The process for reading the index into memory from disk is to first read its contents into a single memory-mapped file buffer (type 'char *'), then sequentially convert each on-disk index entry into a corresponding incore 'cache_entry'. To access the contents of the on-disk entry for processing, a moving pointer within the memory-mapped file is cast to type 'struct ondisk_cache_entry *'. In index v4, the entries in the on-disk index file are written *without* aligning their first byte to a 4-byte boundary; entries are a variable length (depending on the entry name and whether or not extended flags are used). As a result, casting the 'char *' buffer pointer to 'struct ondisk_cache_entry *' then accessing its contents in a 'SANITIZE=undefined' build can trigger the following error: read-cache.c:1886:46: runtime error: member access within misaligned address <address> for type 'struct ondisk_cache_entry', which requires 4 byte alignment Avoid this error by reading fields directly from the 'char *' buffer, using the 'offsetof' individual fields in 'struct ondisk_cache_entry'. Additionally, add documentation describing why the new approach avoids the misaligned address error, as well as advice on how to improve the implementation in the future. Reported-by: Jeff King <peff@peff.net> Signed-off-by: Victoria Dye <vdye@github.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Victoria Dye committed
Sep 28, 2022 at 17:19 UTC
4a6ed30f96c0313dc55fc052707b33d4d3518912
1 file changed
+42
-19
read-cache.c
+42
-19
@@ -1873,9 +1873,20 @@ static int read_index_extension(struct index_state *istate,
1873
return 0;
1874
}
1875
1876
+/*
1877
+ * Parses the contents of the cache entry contained within the 'ondisk' buffer
1878
+ * into a new incore 'cache_entry'.
1879
+ *
1880
+ * Note that 'char *ondisk' may not be aligned to a 4-byte address interval in
1881
+ * index v4, so we cannot cast it to 'struct ondisk_cache_entry *' and access
1882
+ * its members. Instead, we use the byte offsets of members within the struct to
1883
+ * identify where 'get_be16()', 'get_be32()', and 'oidread()' (which can all
1884
+ * read from an unaligned memory buffer) should read from the 'ondisk' buffer
1885
+ * into the corresponding incore 'cache_entry' members.
1886
+ */
1887
static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
1888
unsigned int version,
1878
- struct ondisk_cache_entry *ondisk,
1889
+ const char *ondisk,
1890
unsigned long *ent_size,
1891
const struct cache_entry *previous_ce)
1892
{
@@ -1883,7 +1894,7 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
1894
size_t len;
1895
const char *name;
1896
const unsigned hashsz = the_hash_algo->rawsz;
1886
- const uint16_t *flagsp = (const uint16_t *)(ondisk->data + hashsz);
1897
+ const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz;
1898
unsigned int flags;
1899
size_t copy_len = 0;
1900
/*
@@ -1901,15 +1912,15 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
1912
1913
if (flags & CE_EXTENDED) {
1914
int extended_flags;
1904
- extended_flags = get_be16(flagsp + 1) << 16;
1915
+ extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16;
1916
/* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
1917
if (extended_flags & ~CE_EXTENDED_FLAGS)
1918
die(_("unknown index entry format 0x%08x"), extended_flags);
1919
flags |= extended_flags;
1909
- name = (const char *)(flagsp + 2);
1920
+ name = (const char *)(flagsp + 2 * sizeof(uint16_t));
1921
}
1922
else
1912
- name = (const char *)(flagsp + 1);
1923
+ name = (const char *)(flagsp + sizeof(uint16_t));
1924
1925
if (expand_name_field) {
1926
const unsigned char *cp = (const unsigned char *)name;
@@ -1935,20 +1946,32 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
1946
1947
ce = mem_pool__ce_alloc(ce_mem_pool, len);
1948
1938
- ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec);
1939
- ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec);
1940
- ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec);
1941
- ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec);
1942
- ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev);
1943
- ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino);
1944
- ce->ce_mode = get_be32(&ondisk->mode);
1945
- ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid);
1946
- ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid);
1947
- ce->ce_stat_data.sd_size = get_be32(&ondisk->size);
1949
+ /*
1950
+ * NEEDSWORK: using 'offsetof()' is cumbersome and should be replaced
1951
+ * with something more akin to 'load_bitmap_entries_v1()'s use of
1952
+ * 'read_be16'/'read_be32'. For consistency with the corresponding
1953
+ * ondisk entry write function ('copy_cache_entry_to_ondisk()'), this
1954
+ * should be done at the same time as removing references to
1955
+ * 'ondisk_cache_entry' there.
1956
+ */
1957
+ ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime)
1958
+ + offsetof(struct cache_time, sec));
1959
+ ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime)
1960
+ + offsetof(struct cache_time, sec));
1961
+ ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime)
1962
+ + offsetof(struct cache_time, nsec));
1963
+ ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime)
1964
+ + offsetof(struct cache_time, nsec));
1965
+ ce->ce_stat_data.sd_dev = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev));
1966
+ ce->ce_stat_data.sd_ino = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino));
1967
+ ce->ce_mode = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode));
1968
+ ce->ce_stat_data.sd_uid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid));
1969
+ ce->ce_stat_data.sd_gid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid));
1970
+ ce->ce_stat_data.sd_size = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size));
1971
ce->ce_flags = flags & ~CE_NAMEMASK;
1972
ce->ce_namelen = len;
1973
ce->index = 0;
1951
- oidread(&ce->oid, ondisk->data);
1974
+ oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data));
1975
1976
if (expand_name_field) {
1977
if (copy_len)
@@ -2117,12 +2140,12 @@ static unsigned long load_cache_entry_block(struct index_state *istate,
2140
unsigned long src_offset = start_offset;
2141
2142
for (i = offset; i < offset + nr; i++) {
2120
- struct ondisk_cache_entry *disk_ce;
2143
struct cache_entry *ce;
2144
unsigned long consumed;
2145
2124
- disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset);
2125
- ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce);
2146
+ ce = create_from_disk(ce_mem_pool, istate->version,
2147
+ mmap + src_offset,
2148
+ &consumed, previous_ce);
2149
set_index_entry(istate, i, ce);
2150
2151
src_offset += consumed;