diff --git a/fuzz/c/fuzzlib/fuzzlib.c b/fuzz/c/fuzzlib/fuzzlib.c index 29f9a4a2..d4c72efb 100644 --- a/fuzz/c/fuzzlib/fuzzlib.c +++ b/fuzz/c/fuzzlib/fuzzlib.c @@ -27,14 +27,12 @@ intentional_segfault() { *ptr = 0; } -const char* // -fuzz(wuffs_base__io_buffer* src, uint32_t hash); - -static const char* // -llvmFuzzerTestOneInput(const uint8_t* data, size_t size) { - // Hash input as per https://en.wikipedia.org/wiki/Jenkins_hash_function - size_t i = 0; +// jenkins_hash_u32 implements +// https://en.wikipedia.org/wiki/Jenkins_hash_function +static uint32_t // +jenkins_hash_u32(const uint8_t* data, size_t size) { uint32_t hash = 0; + size_t i = 0; while (i != size) { hash += data[i++]; hash += hash << 10; @@ -43,6 +41,19 @@ llvmFuzzerTestOneInput(const uint8_t* data, size_t size) { hash += hash << 3; hash ^= hash >> 11; hash += hash << 15; + return hash; +} + +const char* // +fuzz(wuffs_base__io_buffer* src, uint64_t hash); + +static const char* // +llvmFuzzerTestOneInput(const uint8_t* data, size_t size) { + // Make a 64-bit hash out of two 32-bit hashes, each on half of the data. + size_t s2 = size / 2; + uint32_t hash0 = jenkins_hash_u32(data, s2); + uint32_t hash1 = jenkins_hash_u32(data + s2, size - s2); + uint64_t hash = (((uint64_t)hash0) << 32) | ((uint64_t)hash1); wuffs_base__io_buffer src = ((wuffs_base__io_buffer){ .data = ((wuffs_base__slice_u8){ diff --git a/fuzz/c/std/cbor_fuzzer.c b/fuzz/c/std/cbor_fuzzer.c index fe02c202..a2c5020c 100644 --- a/fuzz/c/std/cbor_fuzzer.c +++ b/fuzz/c/std/cbor_fuzzer.c @@ -218,7 +218,7 @@ fuzz_one_token(wuffs_base__token t, } uint64_t // -buffer_limit(uint32_t hash_6_bits, uint64_t min, uint64_t max) { +buffer_limit(uint64_t hash_6_bits, uint64_t min, uint64_t max) { uint64_t n; if (hash_6_bits < 0x20) { n = min + hash_6_bits; @@ -234,14 +234,14 @@ buffer_limit(uint32_t hash_6_bits, uint64_t min, uint64_t max) { } const char* // -fuzz_complex(wuffs_base__io_buffer* full_src, uint32_t hash_24_bits) { +fuzz_complex(wuffs_base__io_buffer* full_src, uint64_t hash_56_bits) { uint64_t tok_limit = buffer_limit( - hash_24_bits & 0x3F, WUFFS_CBOR__DECODER_DST_TOKEN_BUFFER_LENGTH_MIN_INCL, + hash_56_bits & 0x3F, WUFFS_CBOR__DECODER_DST_TOKEN_BUFFER_LENGTH_MIN_INCL, TOK_BUFFER_ARRAY_SIZE); - uint32_t hash_18_bits = hash_24_bits >> 6; + uint64_t hash_50_bits = hash_56_bits >> 6; uint64_t src_limit = - buffer_limit(hash_18_bits & 0x3F, + buffer_limit(hash_50_bits & 0x3F, WUFFS_CBOR__DECODER_SRC_IO_BUFFER_LENGTH_MIN_INCL, 4096); // ---- @@ -391,7 +391,7 @@ fuzz_simple(wuffs_base__io_buffer* full_src) { } const char* // -fuzz(wuffs_base__io_buffer* full_src, uint32_t hash) { +fuzz(wuffs_base__io_buffer* full_src, uint64_t hash) { // Send 99.6% of inputs to fuzz_complex and the remainder to fuzz_simple. The // 0xA5 constant is arbitrary but non-zero. If the hash function maps the // empty input to 0, this still sends the empty input to fuzz_complex. diff --git a/fuzz/c/std/gif_fuzzer.c b/fuzz/c/std/gif_fuzzer.c index d2fd1e33..5c483f8d 100644 --- a/fuzz/c/std/gif_fuzzer.c +++ b/fuzz/c/std/gif_fuzzer.c @@ -61,7 +61,7 @@ It should print "PASS", amongst other information, and exit(0). #include "../fuzzlib/fuzzlib.c" const char* // -fuzz(wuffs_base__io_buffer* src, uint32_t hash) { +fuzz(wuffs_base__io_buffer* src, uint64_t hash) { const char* ret = NULL; wuffs_base__slice_u8 pixbuf = ((wuffs_base__slice_u8){}); wuffs_base__slice_u8 workbuf = ((wuffs_base__slice_u8){}); diff --git a/fuzz/c/std/json_fuzzer.c b/fuzz/c/std/json_fuzzer.c index 6cb02700..e47bc36e 100644 --- a/fuzz/c/std/json_fuzzer.c +++ b/fuzz/c/std/json_fuzzer.c @@ -209,7 +209,7 @@ fuzz_one_token(wuffs_base__token t, } uint64_t // -buffer_limit(uint32_t hash_6_bits, uint64_t min, uint64_t max) { +buffer_limit(uint64_t hash_6_bits, uint64_t min, uint64_t max) { uint64_t n; if (hash_6_bits < 0x20) { n = min + hash_6_bits; @@ -224,7 +224,7 @@ buffer_limit(uint32_t hash_6_bits, uint64_t min, uint64_t max) { return n; } -void set_quirks(wuffs_json__decoder* dec, uint32_t hash_12_bits) { +void set_quirks(wuffs_json__decoder* dec, uint64_t hash_44_bits) { uint32_t quirks[] = { WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_A, WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_CAPITAL_U, @@ -232,8 +232,8 @@ void set_quirks(wuffs_json__decoder* dec, uint32_t hash_12_bits) { WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_QUESTION_MARK, WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_SINGLE_QUOTE, WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_V, - WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_X_AS_BYTES, - WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_X_AS_CODE_POINTS, + (hash_44_bits & 1) ? WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_X_AS_BYTES + : WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_X_AS_CODE_POINTS, WUFFS_JSON__QUIRK_ALLOW_BACKSLASH_ZERO, WUFFS_JSON__QUIRK_ALLOW_COMMENT_BLOCK, WUFFS_JSON__QUIRK_ALLOW_COMMENT_LINE, @@ -246,26 +246,29 @@ void set_quirks(wuffs_json__decoder* dec, uint32_t hash_12_bits) { 0, }; + // Shift off a bit for the ETC_BACKSLASH_X_ETC choice. + uint64_t hash_43_bits = hash_44_bits >> 1; + uint32_t i; for (i = 0; quirks[i]; i++) { - uint32_t bit = 1 << (i % 12); - if (hash_12_bits & bit) { + uint64_t bit = 1 << (i % 43); + if (hash_43_bits & bit) { wuffs_json__decoder__set_quirk_enabled(dec, quirks[i], true); } } } const char* // -fuzz_complex(wuffs_base__io_buffer* full_src, uint32_t hash_24_bits) { +fuzz_complex(wuffs_base__io_buffer* full_src, uint64_t hash_56_bits) { uint64_t tok_limit = buffer_limit( - hash_24_bits & 0x3F, WUFFS_JSON__DECODER_DST_TOKEN_BUFFER_LENGTH_MIN_INCL, + hash_56_bits & 0x3F, WUFFS_JSON__DECODER_DST_TOKEN_BUFFER_LENGTH_MIN_INCL, TOK_BUFFER_ARRAY_SIZE); - uint32_t hash_18_bits = hash_24_bits >> 6; + uint64_t hash_50_bits = hash_56_bits >> 6; uint64_t src_limit = - buffer_limit(hash_18_bits & 0x3F, + buffer_limit(hash_50_bits & 0x3F, WUFFS_JSON__DECODER_SRC_IO_BUFFER_LENGTH_MIN_INCL, 4096); - uint32_t hash_12_bits = hash_18_bits >> 6; + uint64_t hash_44_bits = hash_50_bits >> 6; // ---- @@ -276,7 +279,7 @@ fuzz_complex(wuffs_base__io_buffer* full_src, uint32_t hash_24_bits) { if (!wuffs_base__status__is_ok(&status)) { return wuffs_base__status__message(&status); } - set_quirks(&dec, hash_12_bits); + set_quirks(&dec, hash_44_bits); wuffs_base__token tok_array[TOK_BUFFER_ARRAY_SIZE]; wuffs_base__token_buffer tok = ((wuffs_base__token_buffer){ @@ -415,7 +418,7 @@ fuzz_simple(wuffs_base__io_buffer* full_src) { } const char* // -fuzz(wuffs_base__io_buffer* full_src, uint32_t hash) { +fuzz(wuffs_base__io_buffer* full_src, uint64_t hash) { // Send 99.6% of inputs to fuzz_complex and the remainder to fuzz_simple. The // 0xA5 constant is arbitrary but non-zero. If the hash function maps the // empty input to 0, this still sends the empty input to fuzz_complex. diff --git a/fuzz/c/std/zlib_fuzzer.c b/fuzz/c/std/zlib_fuzzer.c index 2571b65a..4f104d48 100644 --- a/fuzz/c/std/zlib_fuzzer.c +++ b/fuzz/c/std/zlib_fuzzer.c @@ -75,7 +75,7 @@ uint8_t g_work_buffer_array[1]; #endif const char* // -fuzz(wuffs_base__io_buffer* src, uint32_t hash) { +fuzz(wuffs_base__io_buffer* src, uint64_t hash) { wuffs_zlib__decoder dec; wuffs_base__status status = wuffs_zlib__decoder__initialize( &dec, sizeof dec, WUFFS_VERSION,