/* * Copyright (C) 2016 The Android Open Source Project * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation * files (the "Software"), to deal in the Software without * restriction, including without limitation the rights to use, copy, * modify, merge, publish, distribute, sublicense, and/or sell copies * of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ #include "avb_util.h" #include <stdarg.h> uint32_t avb_be32toh(uint32_t in) { uint8_t* d = (uint8_t*)∈ uint32_t ret; ret = ((uint32_t)d[0]) << 24; ret |= ((uint32_t)d[1]) << 16; ret |= ((uint32_t)d[2]) << 8; ret |= ((uint32_t)d[3]); return ret; } uint64_t avb_be64toh(uint64_t in) { uint8_t* d = (uint8_t*)∈ uint64_t ret; ret = ((uint64_t)d[0]) << 56; ret |= ((uint64_t)d[1]) << 48; ret |= ((uint64_t)d[2]) << 40; ret |= ((uint64_t)d[3]) << 32; ret |= ((uint64_t)d[4]) << 24; ret |= ((uint64_t)d[5]) << 16; ret |= ((uint64_t)d[6]) << 8; ret |= ((uint64_t)d[7]); return ret; } /* Converts a 32-bit unsigned integer from host to big-endian byte order. */ uint32_t avb_htobe32(uint32_t in) { union { uint32_t word; uint8_t bytes[4]; } ret; ret.bytes[0] = (in >> 24) & 0xff; ret.bytes[1] = (in >> 16) & 0xff; ret.bytes[2] = (in >> 8) & 0xff; ret.bytes[3] = in & 0xff; return ret.word; } /* Converts a 64-bit unsigned integer from host to big-endian byte order. */ uint64_t avb_htobe64(uint64_t in) { union { uint64_t word; uint8_t bytes[8]; } ret; ret.bytes[0] = (in >> 56) & 0xff; ret.bytes[1] = (in >> 48) & 0xff; ret.bytes[2] = (in >> 40) & 0xff; ret.bytes[3] = (in >> 32) & 0xff; ret.bytes[4] = (in >> 24) & 0xff; ret.bytes[5] = (in >> 16) & 0xff; ret.bytes[6] = (in >> 8) & 0xff; ret.bytes[7] = in & 0xff; return ret.word; } int avb_safe_memcmp(const void* s1, const void* s2, size_t n) { const unsigned char* us1 = s1; const unsigned char* us2 = s2; int result = 0; if (0 == n) { return 0; } /* * Code snippet without data-dependent branch due to Nate Lawson * (nate@root.org) of Root Labs. */ while (n--) { result |= *us1++ ^ *us2++; } return result != 0; } bool avb_safe_add_to(uint64_t* value, uint64_t value_to_add) { uint64_t original_value; avb_assert(value != NULL); original_value = *value; *value += value_to_add; if (*value < original_value) { avb_error("Overflow when adding values.\n"); return false; } return true; } bool avb_safe_add(uint64_t* out_result, uint64_t a, uint64_t b) { uint64_t dummy; if (out_result == NULL) { out_result = &dummy; } *out_result = a; return avb_safe_add_to(out_result, b); } bool avb_validate_utf8(const uint8_t* data, size_t num_bytes) { size_t n; unsigned int num_cc; for (n = 0, num_cc = 0; n < num_bytes; n++) { uint8_t c = data[n]; if (num_cc > 0) { if ((c & (0x80 | 0x40)) == 0x80) { /* 10xx xxxx */ } else { goto fail; } num_cc--; } else { if (c < 0x80) { num_cc = 0; } else if ((c & (0x80 | 0x40 | 0x20)) == (0x80 | 0x40)) { /* 110x xxxx */ num_cc = 1; } else if ((c & (0x80 | 0x40 | 0x20 | 0x10)) == (0x80 | 0x40 | 0x20)) { /* 1110 xxxx */ num_cc = 2; } else if ((c & (0x80 | 0x40 | 0x20 | 0x10 | 0x08)) == (0x80 | 0x40 | 0x20 | 0x10)) { /* 1111 0xxx */ num_cc = 3; } else { goto fail; } } } if (num_cc != 0) { goto fail; } return true; fail: return false; } bool avb_str_concat(char* buf, size_t buf_size, const char* str1, size_t str1_len, const char* str2, size_t str2_len) { uint64_t combined_len; if (!avb_safe_add(&combined_len, str1_len, str2_len)) { avb_error("Overflow when adding string sizes.\n"); return false; } if (combined_len > buf_size - 1) { avb_error("Insufficient buffer space.\n"); return false; } avb_memcpy(buf, str1, str1_len); avb_memcpy(buf + str1_len, str2, str2_len); buf[combined_len] = '\0'; return true; } void* avb_malloc(size_t size) { void* ret = avb_malloc_(size); if (ret == NULL) { avb_error("Failed to allocate memory.\n"); return NULL; } return ret; } void* avb_calloc(size_t size) { void* ret = avb_malloc(size); if (ret == NULL) { return NULL; } avb_memset(ret, '\0', size); return ret; } char* avb_strdup(const char* str) { size_t len = avb_strlen(str); char* ret = avb_malloc(len + 1); if (ret == NULL) { return NULL; } avb_memcpy(ret, str, len); ret[len] = '\0'; return ret; } const char* avb_strstr(const char* haystack, const char* needle) { size_t n, m; /* Look through |haystack| and check if the first character of * |needle| matches. If so, check the rest of |needle|. */ for (n = 0; haystack[n] != '\0'; n++) { if (haystack[n] != needle[0]) { continue; } for (m = 1;; m++) { if (needle[m] == '\0') { return haystack + n; } if (haystack[n + m] != needle[m]) { break; } } } return NULL; } const char* avb_strv_find_str(const char* const* strings, const char* str, size_t str_size) { size_t n; for (n = 0; strings[n] != NULL; n++) { if (avb_strlen(strings[n]) == str_size && avb_memcmp(strings[n], str, str_size) == 0) { return strings[n]; } } return NULL; } char* avb_replace(const char* str, const char* search, const char* replace) { char* ret = NULL; size_t ret_len = 0; size_t search_len, replace_len; const char* str_after_last_replace; search_len = avb_strlen(search); replace_len = avb_strlen(replace); str_after_last_replace = str; while (*str != '\0') { const char* s; size_t num_before; size_t num_new; s = avb_strstr(str, search); if (s == NULL) { break; } num_before = s - str; if (ret == NULL) { num_new = num_before + replace_len + 1; ret = avb_malloc(num_new); if (ret == NULL) { goto out; } avb_memcpy(ret, str, num_before); avb_memcpy(ret + num_before, replace, replace_len); ret[num_new - 1] = '\0'; ret_len = num_new - 1; } else { char* new_str; num_new = ret_len + num_before + replace_len + 1; new_str = avb_malloc(num_new); if (new_str == NULL) { goto out; } avb_memcpy(new_str, ret, ret_len); avb_memcpy(new_str + ret_len, str, num_before); avb_memcpy(new_str + ret_len + num_before, replace, replace_len); new_str[num_new - 1] = '\0'; avb_free(ret); ret = new_str; ret_len = num_new - 1; } str = s + search_len; str_after_last_replace = str; } if (ret == NULL) { ret = avb_strdup(str_after_last_replace); if (ret == NULL) { goto out; } } else { size_t num_remaining = avb_strlen(str_after_last_replace); size_t num_new = ret_len + num_remaining + 1; char* new_str = avb_malloc(num_new); if (new_str == NULL) { goto out; } avb_memcpy(new_str, ret, ret_len); avb_memcpy(new_str + ret_len, str_after_last_replace, num_remaining); new_str[num_new - 1] = '\0'; avb_free(ret); ret = new_str; ret_len = num_new - 1; } out: return ret; } /* We only support a limited amount of strings in avb_strdupv(). */ #define AVB_STRDUPV_MAX_NUM_STRINGS 32 char* avb_strdupv(const char* str, ...) { va_list ap; const char* strings[AVB_STRDUPV_MAX_NUM_STRINGS]; size_t lengths[AVB_STRDUPV_MAX_NUM_STRINGS]; size_t num_strings, n; uint64_t total_length; char *ret = NULL, *dest; num_strings = 0; total_length = 0; va_start(ap, str); do { size_t str_len = avb_strlen(str); strings[num_strings] = str; lengths[num_strings] = str_len; if (!avb_safe_add_to(&total_length, str_len)) { avb_fatal("Overflow while determining total length.\n"); break; } num_strings++; if (num_strings == AVB_STRDUPV_MAX_NUM_STRINGS) { avb_fatal("Too many strings passed.\n"); break; } str = va_arg(ap, const char*); } while (str != NULL); va_end(ap); ret = avb_malloc(total_length + 1); if (ret == NULL) { goto out; } dest = ret; for (n = 0; n < num_strings; n++) { avb_memcpy(dest, strings[n], lengths[n]); dest += lengths[n]; } *dest = '\0'; avb_assert(dest == ret + total_length); out: return ret; } const char* avb_basename(const char* str) { int64_t n; size_t len; len = avb_strlen(str); if (len >= 2) { for (n = len - 2; n >= 0; n--) { if (str[n] == '/') { return str + n + 1; } } } return str; } void avb_uppercase(char* str) { size_t i; for (i = 0; str[i] != '\0'; ++i) { if (str[i] <= 0x7A && str[i] >= 0x61) { str[i] -= 0x20; } } } char* avb_bin2hex(const uint8_t* data, size_t data_len) { const char hex_digits[17] = "0123456789abcdef"; char* hex_data; size_t n; hex_data = avb_malloc(data_len * 2 + 1); if (hex_data == NULL) { return NULL; } for (n = 0; n < data_len; n++) { hex_data[n * 2] = hex_digits[data[n] >> 4]; hex_data[n * 2 + 1] = hex_digits[data[n] & 0x0f]; } hex_data[n * 2] = '\0'; return hex_data; }