diff options
-rw-r--r-- | include/afl-fuzz.h | 46 | ||||
-rw-r--r-- | src/afl-fuzz-src/afl-fuzz.c | 615 | ||||
-rw-r--r-- | src/afl-fuzz-src/extras.c | 484 | ||||
-rw-r--r-- | src/afl-fuzz-src/misc.c | 152 |
4 files changed, 667 insertions, 630 deletions
diff --git a/include/afl-fuzz.h b/include/afl-fuzz.h index 22a78373..c50a21a7 100644 --- a/include/afl-fuzz.h +++ b/include/afl-fuzz.h @@ -468,31 +468,47 @@ void trim_py(char**, size_t*); /* Queue */ -void mark_as_det_done(struct queue_entry* q); -void mark_as_variable(struct queue_entry* q); -void mark_as_redundant(struct queue_entry* q, u8 state); -void add_to_queue(u8* fname, u32 len, u8 passed_det); +void mark_as_det_done(struct queue_entry*); +void mark_as_variable(struct queue_entry*); +void mark_as_redundant(struct queue_entry*, u8); +void add_to_queue(u8*, u32, u8); void destroy_queue(void); -void update_bitmap_score(struct queue_entry* q); +void update_bitmap_score(struct queue_entry*); void cull_queue(void); /* Bitmap */ void write_bitmap(void); -void read_bitmap(u8* fname); -u8 has_new_bits(u8* virgin_map); -u32 count_bits(u8* mem); -u32 count_bytes(u8* mem); -u32 count_non_255_bytes(u8* mem); +void read_bitmap(u8*); +u8 has_new_bits(u8*); +u32 count_bits(u8*); +u32 count_bytes(u8*); +u32 count_non_255_bytes(u8*); #ifdef __x86_64__ -void simplify_trace(u64* mem); -void classify_counts(u64* mem); +void simplify_trace(u64*); +void classify_counts(u64*); #else -void simplify_trace(u32* mem); -void classify_counts(u32* mem); +void simplify_trace(u32*); +void classify_counts(u32*); #endif void init_count_class16(void); -void minimize_bits(u8* dst, u8* src); +void minimize_bits(u8*, u8*); + +/* Misc */ + +u8* DI(u64); +u8* DF(double); +u8* DMS(u64); +u8* DTD(u64, u64); + +/* Extras */ + +void load_extras_file(u8*, u32*, u32*, u32); +void load_extras(u8*); +void maybe_add_auto(u8*, u32); +void save_auto(void); +void load_auto(void); +void destroy_extras(void); /**** Inline routines ****/ diff --git a/src/afl-fuzz-src/afl-fuzz.c b/src/afl-fuzz-src/afl-fuzz.c index dcb97387..7cc05a39 100644 --- a/src/afl-fuzz-src/afl-fuzz.c +++ b/src/afl-fuzz-src/afl-fuzz.c @@ -211,159 +211,6 @@ static void locate_diffs(u8* ptr1, u8* ptr2, u32 len, s32* first, s32* last) { #endif /* !IGNORE_FINDS */ -/* Describe integer. Uses 12 cyclic static buffers for return values. The value - returned should be five characters or less for all the integers we reasonably - expect to see. */ - -static u8* DI(u64 val) { - - static u8 tmp[12][16]; - static u8 cur; - - cur = (cur + 1) % 12; - -#define CHK_FORMAT(_divisor, _limit_mult, _fmt, _cast) do { \ - if (val < (_divisor) * (_limit_mult)) { \ - sprintf(tmp[cur], _fmt, ((_cast)val) / (_divisor)); \ - return tmp[cur]; \ - } \ - } while (0) - - /* 0-9999 */ - CHK_FORMAT(1, 10000, "%llu", u64); - - /* 10.0k - 99.9k */ - CHK_FORMAT(1000, 99.95, "%0.01fk", double); - - /* 100k - 999k */ - CHK_FORMAT(1000, 1000, "%lluk", u64); - - /* 1.00M - 9.99M */ - CHK_FORMAT(1000 * 1000, 9.995, "%0.02fM", double); - - /* 10.0M - 99.9M */ - CHK_FORMAT(1000 * 1000, 99.95, "%0.01fM", double); - - /* 100M - 999M */ - CHK_FORMAT(1000 * 1000, 1000, "%lluM", u64); - - /* 1.00G - 9.99G */ - CHK_FORMAT(1000LL * 1000 * 1000, 9.995, "%0.02fG", double); - - /* 10.0G - 99.9G */ - CHK_FORMAT(1000LL * 1000 * 1000, 99.95, "%0.01fG", double); - - /* 100G - 999G */ - CHK_FORMAT(1000LL * 1000 * 1000, 1000, "%lluG", u64); - - /* 1.00T - 9.99G */ - CHK_FORMAT(1000LL * 1000 * 1000 * 1000, 9.995, "%0.02fT", double); - - /* 10.0T - 99.9T */ - CHK_FORMAT(1000LL * 1000 * 1000 * 1000, 99.95, "%0.01fT", double); - - /* 100T+ */ - strcpy(tmp[cur], "infty"); - return tmp[cur]; - -} - - -/* Describe float. Similar to the above, except with a single - static buffer. */ - -static u8* DF(double val) { - - static u8 tmp[16]; - - if (val < 99.995) { - sprintf(tmp, "%0.02f", val); - return tmp; - } - - if (val < 999.95) { - sprintf(tmp, "%0.01f", val); - return tmp; - } - - return DI((u64)val); - -} - - -/* Describe integer as memory size. */ - -static u8* DMS(u64 val) { - - static u8 tmp[12][16]; - static u8 cur; - - cur = (cur + 1) % 12; - - /* 0-9999 */ - CHK_FORMAT(1, 10000, "%llu B", u64); - - /* 10.0k - 99.9k */ - CHK_FORMAT(1024, 99.95, "%0.01f kB", double); - - /* 100k - 999k */ - CHK_FORMAT(1024, 1000, "%llu kB", u64); - - /* 1.00M - 9.99M */ - CHK_FORMAT(1024 * 1024, 9.995, "%0.02f MB", double); - - /* 10.0M - 99.9M */ - CHK_FORMAT(1024 * 1024, 99.95, "%0.01f MB", double); - - /* 100M - 999M */ - CHK_FORMAT(1024 * 1024, 1000, "%llu MB", u64); - - /* 1.00G - 9.99G */ - CHK_FORMAT(1024LL * 1024 * 1024, 9.995, "%0.02f GB", double); - - /* 10.0G - 99.9G */ - CHK_FORMAT(1024LL * 1024 * 1024, 99.95, "%0.01f GB", double); - - /* 100G - 999G */ - CHK_FORMAT(1024LL * 1024 * 1024, 1000, "%llu GB", u64); - - /* 1.00T - 9.99G */ - CHK_FORMAT(1024LL * 1024 * 1024 * 1024, 9.995, "%0.02f TB", double); - - /* 10.0T - 99.9T */ - CHK_FORMAT(1024LL * 1024 * 1024 * 1024, 99.95, "%0.01f TB", double); - -#undef CHK_FORMAT - - /* 100T+ */ - strcpy(tmp[cur], "infty"); - return tmp[cur]; - -} - - -/* Describe time delta. Returns one static buffer, 34 chars of less. */ - -static u8* DTD(u64 cur_ms, u64 event_ms) { - - static u8 tmp[64]; - u64 delta; - s32 t_d, t_h, t_m, t_s; - - if (!event_ms) return "none seen yet"; - - delta = cur_ms - event_ms; - - t_d = delta / 1000 / 60 / 60 / 24; - t_h = (delta / 1000 / 60 / 60) % 24; - t_m = (delta / 1000 / 60) % 60; - t_s = (delta / 1000) % 60; - - sprintf(tmp, "%s days, %d hrs, %d min, %d sec", DI(t_d), t_h, t_m, t_s); - return tmp; - -} - /* Load postprocessor, if available. */ @@ -516,468 +363,6 @@ static void read_testcases(void) { } -/* Helper function for load_extras. */ - -static int compare_extras_len(const void* p1, const void* p2) { - struct extra_data *e1 = (struct extra_data*)p1, - *e2 = (struct extra_data*)p2; - - return e1->len - e2->len; -} - -static int compare_extras_use_d(const void* p1, const void* p2) { - struct extra_data *e1 = (struct extra_data*)p1, - *e2 = (struct extra_data*)p2; - - return e2->hit_cnt - e1->hit_cnt; -} - - -/* Read extras from a file, sort by size. */ - -static void load_extras_file(u8* fname, u32* min_len, u32* max_len, - u32 dict_level) { - - FILE* f; - u8 buf[MAX_LINE]; - u8 *lptr; - u32 cur_line = 0; - - f = fopen(fname, "r"); - - if (!f) PFATAL("Unable to open '%s'", fname); - - while ((lptr = fgets(buf, MAX_LINE, f))) { - - u8 *rptr, *wptr; - u32 klen = 0; - - ++cur_line; - - /* Trim on left and right. */ - - while (isspace(*lptr)) ++lptr; - - rptr = lptr + strlen(lptr) - 1; - while (rptr >= lptr && isspace(*rptr)) --rptr; - ++rptr; - *rptr = 0; - - /* Skip empty lines and comments. */ - - if (!*lptr || *lptr == '#') continue; - - /* All other lines must end with '"', which we can consume. */ - - --rptr; - - if (rptr < lptr || *rptr != '"') - FATAL("Malformed name=\"value\" pair in line %u.", cur_line); - - *rptr = 0; - - /* Skip alphanumerics and dashes (label). */ - - while (isalnum(*lptr) || *lptr == '_') ++lptr; - - /* If @number follows, parse that. */ - - if (*lptr == '@') { - - ++lptr; - if (atoi(lptr) > dict_level) continue; - while (isdigit(*lptr)) ++lptr; - - } - - /* Skip whitespace and = signs. */ - - while (isspace(*lptr) || *lptr == '=') ++lptr; - - /* Consume opening '"'. */ - - if (*lptr != '"') - FATAL("Malformed name=\"keyword\" pair in line %u.", cur_line); - - ++lptr; - - if (!*lptr) FATAL("Empty keyword in line %u.", cur_line); - - /* Okay, let's allocate memory and copy data between "...", handling - \xNN escaping, \\, and \". */ - - extras = ck_realloc_block(extras, (extras_cnt + 1) * - sizeof(struct extra_data)); - - wptr = extras[extras_cnt].data = ck_alloc(rptr - lptr); - - while (*lptr) { - - char* hexdigits = "0123456789abcdef"; - - switch (*lptr) { - - case 1 ... 31: - case 128 ... 255: - FATAL("Non-printable characters in line %u.", cur_line); - - case '\\': - - ++lptr; - - if (*lptr == '\\' || *lptr == '"') { - *(wptr++) = *(lptr++); - klen++; - break; - } - - if (*lptr != 'x' || !isxdigit(lptr[1]) || !isxdigit(lptr[2])) - FATAL("Invalid escaping (not \\xNN) in line %u.", cur_line); - - *(wptr++) = - ((strchr(hexdigits, tolower(lptr[1])) - hexdigits) << 4) | - (strchr(hexdigits, tolower(lptr[2])) - hexdigits); - - lptr += 3; - ++klen; - - break; - - default: - - *(wptr++) = *(lptr++); - ++klen; - - } - - } - - extras[extras_cnt].len = klen; - - if (extras[extras_cnt].len > MAX_DICT_FILE) - FATAL("Keyword too big in line %u (%s, limit is %s)", cur_line, - DMS(klen), DMS(MAX_DICT_FILE)); - - if (*min_len > klen) *min_len = klen; - if (*max_len < klen) *max_len = klen; - - ++extras_cnt; - - } - - fclose(f); - -} - - -/* Read extras from the extras directory and sort them by size. */ - -static void load_extras(u8* dir) { - - DIR* d; - struct dirent* de; - u32 min_len = MAX_DICT_FILE, max_len = 0, dict_level = 0; - u8* x; - - /* If the name ends with @, extract level and continue. */ - - if ((x = strchr(dir, '@'))) { - - *x = 0; - dict_level = atoi(x + 1); - - } - - ACTF("Loading extra dictionary from '%s' (level %u)...", dir, dict_level); - - d = opendir(dir); - - if (!d) { - - if (errno == ENOTDIR) { - load_extras_file(dir, &min_len, &max_len, dict_level); - goto check_and_sort; - } - - PFATAL("Unable to open '%s'", dir); - - } - - if (x) FATAL("Dictionary levels not supported for directories."); - - while ((de = readdir(d))) { - - struct stat st; - u8* fn = alloc_printf("%s/%s", dir, de->d_name); - s32 fd; - - if (lstat(fn, &st) || access(fn, R_OK)) - PFATAL("Unable to access '%s'", fn); - - /* This also takes care of . and .. */ - if (!S_ISREG(st.st_mode) || !st.st_size) { - - ck_free(fn); - continue; - - } - - if (st.st_size > MAX_DICT_FILE) - FATAL("Extra '%s' is too big (%s, limit is %s)", fn, - DMS(st.st_size), DMS(MAX_DICT_FILE)); - - if (min_len > st.st_size) min_len = st.st_size; - if (max_len < st.st_size) max_len = st.st_size; - - extras = ck_realloc_block(extras, (extras_cnt + 1) * - sizeof(struct extra_data)); - - extras[extras_cnt].data = ck_alloc(st.st_size); - extras[extras_cnt].len = st.st_size; - - fd = open(fn, O_RDONLY); - - if (fd < 0) PFATAL("Unable to open '%s'", fn); - - ck_read(fd, extras[extras_cnt].data, st.st_size, fn); - - close(fd); - ck_free(fn); - - ++extras_cnt; - - } - - closedir(d); - -check_and_sort: - - if (!extras_cnt) FATAL("No usable files in '%s'", dir); - - qsort(extras, extras_cnt, sizeof(struct extra_data), compare_extras_len); - - OKF("Loaded %u extra tokens, size range %s to %s.", extras_cnt, - DMS(min_len), DMS(max_len)); - - if (max_len > 32) - WARNF("Some tokens are relatively large (%s) - consider trimming.", - DMS(max_len)); - - if (extras_cnt > MAX_DET_EXTRAS) - WARNF("More than %d tokens - will use them probabilistically.", - MAX_DET_EXTRAS); - -} - - - - -/* Helper function for maybe_add_auto() */ - -static inline u8 memcmp_nocase(u8* m1, u8* m2, u32 len) { - - while (len--) if (tolower(*(m1++)) ^ tolower(*(m2++))) return 1; - return 0; - -} - - -/* Maybe add automatic extra. */ - -static void maybe_add_auto(u8* mem, u32 len) { - - u32 i; - - /* Allow users to specify that they don't want auto dictionaries. */ - - if (!MAX_AUTO_EXTRAS || !USE_AUTO_EXTRAS) return; - - /* Skip runs of identical bytes. */ - - for (i = 1; i < len; ++i) - if (mem[0] ^ mem[i]) break; - - if (i == len) return; - - /* Reject builtin interesting values. */ - - if (len == 2) { - - i = sizeof(interesting_16) >> 1; - - while (i--) - if (*((u16*)mem) == interesting_16[i] || - *((u16*)mem) == SWAP16(interesting_16[i])) return; - - } - - if (len == 4) { - - i = sizeof(interesting_32) >> 2; - - while (i--) - if (*((u32*)mem) == interesting_32[i] || - *((u32*)mem) == SWAP32(interesting_32[i])) return; - - } - - /* Reject anything that matches existing extras. Do a case-insensitive - match. We optimize by exploiting the fact that extras[] are sorted - by size. */ - - for (i = 0; i < extras_cnt; ++i) - if (extras[i].len >= len) break; - - for (; i < extras_cnt && extras[i].len == len; ++i) - if (!memcmp_nocase(extras[i].data, mem, len)) return; - - /* Last but not least, check a_extras[] for matches. There are no - guarantees of a particular sort order. */ - - auto_changed = 1; - - for (i = 0; i < a_extras_cnt; ++i) { - - if (a_extras[i].len == len && !memcmp_nocase(a_extras[i].data, mem, len)) { - - a_extras[i].hit_cnt++; - goto sort_a_extras; - - } - - } - - /* At this point, looks like we're dealing with a new entry. So, let's - append it if we have room. Otherwise, let's randomly evict some other - entry from the bottom half of the list. */ - - if (a_extras_cnt < MAX_AUTO_EXTRAS) { - - a_extras = ck_realloc_block(a_extras, (a_extras_cnt + 1) * - sizeof(struct extra_data)); - - a_extras[a_extras_cnt].data = ck_memdup(mem, len); - a_extras[a_extras_cnt].len = len; - ++a_extras_cnt; - - } else { - - i = MAX_AUTO_EXTRAS / 2 + - UR((MAX_AUTO_EXTRAS + 1) / 2); - - ck_free(a_extras[i].data); - - a_extras[i].data = ck_memdup(mem, len); - a_extras[i].len = len; - a_extras[i].hit_cnt = 0; - - } - -sort_a_extras: - - /* First, sort all auto extras by use count, descending order. */ - - qsort(a_extras, a_extras_cnt, sizeof(struct extra_data), - compare_extras_use_d); - - /* Then, sort the top USE_AUTO_EXTRAS entries by size. */ - - qsort(a_extras, MIN(USE_AUTO_EXTRAS, a_extras_cnt), - sizeof(struct extra_data), compare_extras_len); - -} - - -/* Save automatically generated extras. */ - -static void save_auto(void) { - - u32 i; - - if (!auto_changed) return; - auto_changed = 0; - - for (i = 0; i < MIN(USE_AUTO_EXTRAS, a_extras_cnt); ++i) { - - u8* fn = alloc_printf("%s/queue/.state/auto_extras/auto_%06u", out_dir, i); - s32 fd; - - fd = open(fn, O_WRONLY | O_CREAT | O_TRUNC, 0600); - - if (fd < 0) PFATAL("Unable to create '%s'", fn); - - ck_write(fd, a_extras[i].data, a_extras[i].len, fn); - - close(fd); - ck_free(fn); - - } - -} - - -/* Load automatically generated extras. */ - -static void load_auto(void) { - - u32 i; - - for (i = 0; i < USE_AUTO_EXTRAS; ++i) { - - u8 tmp[MAX_AUTO_EXTRA + 1]; - u8* fn = alloc_printf("%s/.state/auto_extras/auto_%06u", in_dir, i); - s32 fd, len; - - fd = open(fn, O_RDONLY, 0600); - - if (fd < 0) { - - if (errno != ENOENT) PFATAL("Unable to open '%s'", fn); - ck_free(fn); - break; - - } - - /* We read one byte more to cheaply detect tokens that are too - long (and skip them). */ - - len = read(fd, tmp, MAX_AUTO_EXTRA + 1); - - if (len < 0) PFATAL("Unable to read from '%s'", fn); - - if (len >= MIN_AUTO_EXTRA && len <= MAX_AUTO_EXTRA) - maybe_add_auto(tmp, len); - - close(fd); - ck_free(fn); - - } - - if (i) OKF("Loaded %u auto-discovered dictionary tokens.", i); - else OKF("No auto-generated dictionary tokens to reuse."); - -} - - -/* Destroy extras. */ - -static void destroy_extras(void) { - - u32 i; - - for (i = 0; i < extras_cnt; ++i) - ck_free(extras[i].data); - - ck_free(extras); - - for (i = 0; i < a_extras_cnt; ++i) - ck_free(a_extras[i].data); - - ck_free(a_extras); - -} - - /* Execute target application, monitoring for timeouts. Return status information. The called program will update trace_bits[]. */ diff --git a/src/afl-fuzz-src/extras.c b/src/afl-fuzz-src/extras.c new file mode 100644 index 00000000..1f52181d --- /dev/null +++ b/src/afl-fuzz-src/extras.c @@ -0,0 +1,484 @@ +/* + american fuzzy lop - fuzzer code + -------------------------------- + + Written and maintained by Michal Zalewski <lcamtuf@google.com> + + Forkserver design by Jann Horn <jannhorn@googlemail.com> + + Copyright 2013, 2014, 2015, 2016, 2017 Google Inc. All rights reserved. + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at: + + http://www.apache.org/licenses/LICENSE-2.0 + + This is the real deal: the program takes an instrumented binary and + attempts a variety of basic fuzzing tricks, paying close attention to + how they affect the execution path. + + */ + +#include "afl-fuzz.h" + + +/* Helper function for load_extras. */ + +static int compare_extras_len(const void* p1, const void* p2) { + struct extra_data *e1 = (struct extra_data*)p1, + *e2 = (struct extra_data*)p2; + + return e1->len - e2->len; +} + +static int compare_extras_use_d(const void* p1, const void* p2) { + struct extra_data *e1 = (struct extra_data*)p1, + *e2 = (struct extra_data*)p2; + + return e2->hit_cnt - e1->hit_cnt; +} + + +/* Read extras from a file, sort by size. */ + +void load_extras_file(u8* fname, u32* min_len, u32* max_len, u32 dict_level) { + + FILE* f; + u8 buf[MAX_LINE]; + u8 *lptr; + u32 cur_line = 0; + + f = fopen(fname, "r"); + + if (!f) PFATAL("Unable to open '%s'", fname); + + while ((lptr = fgets(buf, MAX_LINE, f))) { + + u8 *rptr, *wptr; + u32 klen = 0; + + ++cur_line; + + /* Trim on left and right. */ + + while (isspace(*lptr)) ++lptr; + + rptr = lptr + strlen(lptr) - 1; + while (rptr >= lptr && isspace(*rptr)) --rptr; + ++rptr; + *rptr = 0; + + /* Skip empty lines and comments. */ + + if (!*lptr || *lptr == '#') continue; + + /* All other lines must end with '"', which we can consume. */ + + --rptr; + + if (rptr < lptr || *rptr != '"') + FATAL("Malformed name=\"value\" pair in line %u.", cur_line); + + *rptr = 0; + + /* Skip alphanumerics and dashes (label). */ + + while (isalnum(*lptr) || *lptr == '_') ++lptr; + + /* If @number follows, parse that. */ + + if (*lptr == '@') { + + ++lptr; + if (atoi(lptr) > dict_level) continue; + while (isdigit(*lptr)) ++lptr; + + } + + /* Skip whitespace and = signs. */ + + while (isspace(*lptr) || *lptr == '=') ++lptr; + + /* Consume opening '"'. */ + + if (*lptr != '"') + FATAL("Malformed name=\"keyword\" pair in line %u.", cur_line); + + ++lptr; + + if (!*lptr) FATAL("Empty keyword in line %u.", cur_line); + + /* Okay, let's allocate memory and copy data between "...", handling + \xNN escaping, \\, and \". */ + + extras = ck_realloc_block(extras, (extras_cnt + 1) * + sizeof(struct extra_data)); + + wptr = extras[extras_cnt].data = ck_alloc(rptr - lptr); + + while (*lptr) { + + char* hexdigits = "0123456789abcdef"; + + switch (*lptr) { + + case 1 ... 31: + case 128 ... 255: + FATAL("Non-printable characters in line %u.", cur_line); + + case '\\': + + ++lptr; + + if (*lptr == '\\' || *lptr == '"') { + *(wptr++) = *(lptr++); + klen++; + break; + } + + if (*lptr != 'x' || !isxdigit(lptr[1]) || !isxdigit(lptr[2])) + FATAL("Invalid escaping (not \\xNN) in line %u.", cur_line); + + *(wptr++) = + ((strchr(hexdigits, tolower(lptr[1])) - hexdigits) << 4) | + (strchr(hexdigits, tolower(lptr[2])) - hexdigits); + + lptr += 3; + ++klen; + + break; + + default: + + *(wptr++) = *(lptr++); + ++klen; + + } + + } + + extras[extras_cnt].len = klen; + + if (extras[extras_cnt].len > MAX_DICT_FILE) + FATAL("Keyword too big in line %u (%s, limit is %s)", cur_line, + DMS(klen), DMS(MAX_DICT_FILE)); + + if (*min_len > klen) *min_len = klen; + if (*max_len < klen) *max_len = klen; + + ++extras_cnt; + + } + + fclose(f); + +} + + +/* Read extras from the extras directory and sort them by size. */ + +void load_extras(u8* dir) { + + DIR* d; + struct dirent* de; + u32 min_len = MAX_DICT_FILE, max_len = 0, dict_level = 0; + u8* x; + + /* If the name ends with @, extract level and continue. */ + + if ((x = strchr(dir, '@'))) { + + *x = 0; + dict_level = atoi(x + 1); + + } + + ACTF("Loading extra dictionary from '%s' (level %u)...", dir, dict_level); + + d = opendir(dir); + + if (!d) { + + if (errno == ENOTDIR) { + load_extras_file(dir, &min_len, &max_len, dict_level); + goto check_and_sort; + } + + PFATAL("Unable to open '%s'", dir); + + } + + if (x) FATAL("Dictionary levels not supported for directories."); + + while ((de = readdir(d))) { + + struct stat st; + u8* fn = alloc_printf("%s/%s", dir, de->d_name); + s32 fd; + + if (lstat(fn, &st) || access(fn, R_OK)) + PFATAL("Unable to access '%s'", fn); + + /* This also takes care of . and .. */ + if (!S_ISREG(st.st_mode) || !st.st_size) { + + ck_free(fn); + continue; + + } + + if (st.st_size > MAX_DICT_FILE) + FATAL("Extra '%s' is too big (%s, limit is %s)", fn, + DMS(st.st_size), DMS(MAX_DICT_FILE)); + + if (min_len > st.st_size) min_len = st.st_size; + if (max_len < st.st_size) max_len = st.st_size; + + extras = ck_realloc_block(extras, (extras_cnt + 1) * + sizeof(struct extra_data)); + + extras[extras_cnt].data = ck_alloc(st.st_size); + extras[extras_cnt].len = st.st_size; + + fd = open(fn, O_RDONLY); + + if (fd < 0) PFATAL("Unable to open '%s'", fn); + + ck_read(fd, extras[extras_cnt].data, st.st_size, fn); + + close(fd); + ck_free(fn); + + ++extras_cnt; + + } + + closedir(d); + +check_and_sort: + + if (!extras_cnt) FATAL("No usable files in '%s'", dir); + + qsort(extras, extras_cnt, sizeof(struct extra_data), compare_extras_len); + + OKF("Loaded %u extra tokens, size range %s to %s.", extras_cnt, + DMS(min_len), DMS(max_len)); + + if (max_len > 32) + WARNF("Some tokens are relatively large (%s) - consider trimming.", + DMS(max_len)); + + if (extras_cnt > MAX_DET_EXTRAS) + WARNF("More than %d tokens - will use them probabilistically.", + MAX_DET_EXTRAS); + +} + + + +/* Helper function for maybe_add_auto() */ + +static inline u8 memcmp_nocase(u8* m1, u8* m2, u32 len) { + + while (len--) if (tolower(*(m1++)) ^ tolower(*(m2++))) return 1; + return 0; + +} + + +/* Maybe add automatic extra. */ + +void maybe_add_auto(u8* mem, u32 len) { + + u32 i; + + /* Allow users to specify that they don't want auto dictionaries. */ + + if (!MAX_AUTO_EXTRAS || !USE_AUTO_EXTRAS) return; + + /* Skip runs of identical bytes. */ + + for (i = 1; i < len; ++i) + if (mem[0] ^ mem[i]) break; + + if (i == len) return; + + /* Reject builtin interesting values. */ + + if (len == 2) { + + i = sizeof(interesting_16) >> 1; + + while (i--) + if (*((u16*)mem) == interesting_16[i] || + *((u16*)mem) == SWAP16(interesting_16[i])) return; + + } + + if (len == 4) { + + i = sizeof(interesting_32) >> 2; + + while (i--) + if (*((u32*)mem) == interesting_32[i] || + *((u32*)mem) == SWAP32(interesting_32[i])) return; + + } + + /* Reject anything that matches existing extras. Do a case-insensitive + match. We optimize by exploiting the fact that extras[] are sorted + by size. */ + + for (i = 0; i < extras_cnt; ++i) + if (extras[i].len >= len) break; + + for (; i < extras_cnt && extras[i].len == len; ++i) + if (!memcmp_nocase(extras[i].data, mem, len)) return; + + /* Last but not least, check a_extras[] for matches. There are no + guarantees of a particular sort order. */ + + auto_changed = 1; + + for (i = 0; i < a_extras_cnt; ++i) { + + if (a_extras[i].len == len && !memcmp_nocase(a_extras[i].data, mem, len)) { + + a_extras[i].hit_cnt++; + goto sort_a_extras; + + } + + } + + /* At this point, looks like we're dealing with a new entry. So, let's + append it if we have room. Otherwise, let's randomly evict some other + entry from the bottom half of the list. */ + + if (a_extras_cnt < MAX_AUTO_EXTRAS) { + + a_extras = ck_realloc_block(a_extras, (a_extras_cnt + 1) * + sizeof(struct extra_data)); + + a_extras[a_extras_cnt].data = ck_memdup(mem, len); + a_extras[a_extras_cnt].len = len; + ++a_extras_cnt; + + } else { + + i = MAX_AUTO_EXTRAS / 2 + + UR((MAX_AUTO_EXTRAS + 1) / 2); + + ck_free(a_extras[i].data); + + a_extras[i].data = ck_memdup(mem, len); + a_extras[i].len = len; + a_extras[i].hit_cnt = 0; + + } + +sort_a_extras: + + /* First, sort all auto extras by use count, descending order. */ + + qsort(a_extras, a_extras_cnt, sizeof(struct extra_data), + compare_extras_use_d); + + /* Then, sort the top USE_AUTO_EXTRAS entries by size. */ + + qsort(a_extras, MIN(USE_AUTO_EXTRAS, a_extras_cnt), + sizeof(struct extra_data), compare_extras_len); + +} + + +/* Save automatically generated extras. */ + +void save_auto(void) { + + u32 i; + + if (!auto_changed) return; + auto_changed = 0; + + for (i = 0; i < MIN(USE_AUTO_EXTRAS, a_extras_cnt); ++i) { + + u8* fn = alloc_printf("%s/queue/.state/auto_extras/auto_%06u", out_dir, i); + s32 fd; + + fd = open(fn, O_WRONLY | O_CREAT | O_TRUNC, 0600); + + if (fd < 0) PFATAL("Unable to create '%s'", fn); + + ck_write(fd, a_extras[i].data, a_extras[i].len, fn); + + close(fd); + ck_free(fn); + + } + +} + + +/* Load automatically generated extras. */ + +void load_auto(void) { + + u32 i; + + for (i = 0; i < USE_AUTO_EXTRAS; ++i) { + + u8 tmp[MAX_AUTO_EXTRA + 1]; + u8* fn = alloc_printf("%s/.state/auto_extras/auto_%06u", in_dir, i); + s32 fd, len; + + fd = open(fn, O_RDONLY, 0600); + + if (fd < 0) { + + if (errno != ENOENT) PFATAL("Unable to open '%s'", fn); + ck_free(fn); + break; + + } + + /* We read one byte more to cheaply detect tokens that are too + long (and skip them). */ + + len = read(fd, tmp, MAX_AUTO_EXTRA + 1); + + if (len < 0) PFATAL("Unable to read from '%s'", fn); + + if (len >= MIN_AUTO_EXTRA && len <= MAX_AUTO_EXTRA) + maybe_add_auto(tmp, len); + + close(fd); + ck_free(fn); + + } + + if (i) OKF("Loaded %u auto-discovered dictionary tokens.", i); + else OKF("No auto-generated dictionary tokens to reuse."); + +} + + +/* Destroy extras. */ + +void destroy_extras(void) { + + u32 i; + + for (i = 0; i < extras_cnt; ++i) + ck_free(extras[i].data); + + ck_free(extras); + + for (i = 0; i < a_extras_cnt; ++i) + ck_free(a_extras[i].data); + + ck_free(a_extras); + +} + diff --git a/src/afl-fuzz-src/misc.c b/src/afl-fuzz-src/misc.c index 58e57c8f..69ff2f6b 100644 --- a/src/afl-fuzz-src/misc.c +++ b/src/afl-fuzz-src/misc.c @@ -22,3 +22,155 @@ #include "afl-fuzz.h" +/* Describe integer. Uses 12 cyclic static buffers for return values. The value + returned should be five characters or less for all the integers we reasonably + expect to see. */ + +u8* DI(u64 val) { + + static u8 tmp[12][16]; + static u8 cur; + + cur = (cur + 1) % 12; + +#define CHK_FORMAT(_divisor, _limit_mult, _fmt, _cast) do { \ + if (val < (_divisor) * (_limit_mult)) { \ + sprintf(tmp[cur], _fmt, ((_cast)val) / (_divisor)); \ + return tmp[cur]; \ + } \ + } while (0) + + /* 0-9999 */ + CHK_FORMAT(1, 10000, "%llu", u64); + + /* 10.0k - 99.9k */ + CHK_FORMAT(1000, 99.95, "%0.01fk", double); + + /* 100k - 999k */ + CHK_FORMAT(1000, 1000, "%lluk", u64); + + /* 1.00M - 9.99M */ + CHK_FORMAT(1000 * 1000, 9.995, "%0.02fM", double); + + /* 10.0M - 99.9M */ + CHK_FORMAT(1000 * 1000, 99.95, "%0.01fM", double); + + /* 100M - 999M */ + CHK_FORMAT(1000 * 1000, 1000, "%lluM", u64); + + /* 1.00G - 9.99G */ + CHK_FORMAT(1000LL * 1000 * 1000, 9.995, "%0.02fG", double); + + /* 10.0G - 99.9G */ + CHK_FORMAT(1000LL * 1000 * 1000, 99.95, "%0.01fG", double); + + /* 100G - 999G */ + CHK_FORMAT(1000LL * 1000 * 1000, 1000, "%lluG", u64); + + /* 1.00T - 9.99G */ + CHK_FORMAT(1000LL * 1000 * 1000 * 1000, 9.995, "%0.02fT", double); + + /* 10.0T - 99.9T */ + CHK_FORMAT(1000LL * 1000 * 1000 * 1000, 99.95, "%0.01fT", double); + + /* 100T+ */ + strcpy(tmp[cur], "infty"); + return tmp[cur]; + +} + + +/* Describe float. Similar to the above, except with a single + static buffer. */ + +u8* DF(double val) { + + static u8 tmp[16]; + + if (val < 99.995) { + sprintf(tmp, "%0.02f", val); + return tmp; + } + + if (val < 999.95) { + sprintf(tmp, "%0.01f", val); + return tmp; + } + + return DI((u64)val); + +} + + +/* Describe integer as memory size. */ + +u8* DMS(u64 val) { + + static u8 tmp[12][16]; + static u8 cur; + + cur = (cur + 1) % 12; + + /* 0-9999 */ + CHK_FORMAT(1, 10000, "%llu B", u64); + + /* 10.0k - 99.9k */ + CHK_FORMAT(1024, 99.95, "%0.01f kB", double); + + /* 100k - 999k */ + CHK_FORMAT(1024, 1000, "%llu kB", u64); + + /* 1.00M - 9.99M */ + CHK_FORMAT(1024 * 1024, 9.995, "%0.02f MB", double); + + /* 10.0M - 99.9M */ + CHK_FORMAT(1024 * 1024, 99.95, "%0.01f MB", double); + + /* 100M - 999M */ + CHK_FORMAT(1024 * 1024, 1000, "%llu MB", u64); + + /* 1.00G - 9.99G */ + CHK_FORMAT(1024LL * 1024 * 1024, 9.995, "%0.02f GB", double); + + /* 10.0G - 99.9G */ + CHK_FORMAT(1024LL * 1024 * 1024, 99.95, "%0.01f GB", double); + + /* 100G - 999G */ + CHK_FORMAT(1024LL * 1024 * 1024, 1000, "%llu GB", u64); + + /* 1.00T - 9.99G */ + CHK_FORMAT(1024LL * 1024 * 1024 * 1024, 9.995, "%0.02f TB", double); + + /* 10.0T - 99.9T */ + CHK_FORMAT(1024LL * 1024 * 1024 * 1024, 99.95, "%0.01f TB", double); + +#undef CHK_FORMAT + + /* 100T+ */ + strcpy(tmp[cur], "infty"); + return tmp[cur]; + +} + + +/* Describe time delta. Returns one static buffer, 34 chars of less. */ + +u8* DTD(u64 cur_ms, u64 event_ms) { + + static u8 tmp[64]; + u64 delta; + s32 t_d, t_h, t_m, t_s; + + if (!event_ms) return "none seen yet"; + + delta = cur_ms - event_ms; + + t_d = delta / 1000 / 60 / 60 / 24; + t_h = (delta / 1000 / 60 / 60) % 24; + t_m = (delta / 1000 / 60) % 60; + t_s = (delta / 1000) % 60; + + sprintf(tmp, "%s days, %d hrs, %d min, %d sec", DI(t_d), t_h, t_m, t_s); + return tmp; + +} |