diff options
| -rw-r--r-- | src/common/log.c | 2 | ||||
| -rw-r--r-- | src/common/torlog.h | 4 | ||||
| -rw-r--r-- | src/or/consdiff.c | 1301 | ||||
| -rw-r--r-- | src/or/consdiff.h | 93 | ||||
| -rw-r--r-- | src/or/include.am | 2 | ||||
| -rw-r--r-- | src/test/Makefile.nmake | 2 | ||||
| -rw-r--r-- | src/test/bench.c | 22 | ||||
| -rw-r--r-- | src/test/fuzz/fuzz_diff.c | 67 | ||||
| -rw-r--r-- | src/test/fuzz/fuzz_diff_apply.c | 65 | ||||
| -rw-r--r-- | src/test/fuzz/include.am | 47 | ||||
| -rw-r--r-- | src/test/include.am | 1 | ||||
| -rw-r--r-- | src/test/test.c | 1 | ||||
| -rw-r--r-- | src/test/test.h | 1 | ||||
| -rw-r--r-- | src/test/test_consdiff.c | 1097 |
14 files changed, 2701 insertions, 4 deletions
diff --git a/src/common/log.c b/src/common/log.c index 2878ab5349..6a5819064a 100644 --- a/src/common/log.c +++ b/src/common/log.c @@ -1177,7 +1177,7 @@ static const char *domain_list[] = { "GENERAL", "CRYPTO", "NET", "CONFIG", "FS", "PROTOCOL", "MM", "HTTP", "APP", "CONTROL", "CIRC", "REND", "BUG", "DIR", "DIRSERV", "OR", "EDGE", "ACCT", "HIST", "HANDSHAKE", "HEARTBEAT", "CHANNEL", - "SCHED", "GUARD", NULL + "SCHED", "GUARD", "CONSDIFF", NULL }; /** Return a bitmask for the log domain for which <b>domain</b> is the name, diff --git a/src/common/torlog.h b/src/common/torlog.h index ce7aff5301..6e374b1c11 100644 --- a/src/common/torlog.h +++ b/src/common/torlog.h @@ -101,8 +101,10 @@ #define LD_SCHED (1u<<22) /** Guard nodes */ #define LD_GUARD (1u<<23) +/** Generation and application of consensus diffs. */ +#define LD_CONSDIFF (1u<<24) /** Number of logging domains in the code. */ -#define N_LOGGING_DOMAINS 24 +#define N_LOGGING_DOMAINS 25 /** This log message is not safe to send to a callback-based logger * immediately. Used as a flag, not a log domain. */ diff --git a/src/or/consdiff.c b/src/or/consdiff.c new file mode 100644 index 0000000000..13b267f4a0 --- /dev/null +++ b/src/or/consdiff.c @@ -0,0 +1,1301 @@ +/* Copyright (c) 2014, Daniel Martà + * Copyright (c) 2014, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file consdiff.c + * \brief Consensus diff implementation, including both the generation and the + * application of diffs in a minimal ed format. + * + * The consensus diff application is done in consdiff_apply_diff, which relies + * on apply_ed_diff for the main ed diff part and on some digest helper + * functions to check the digest hashes found in the consensus diff header. + * + * The consensus diff generation is more complex. consdiff_gen_diff generates + * it, relying on gen_ed_diff to generate the ed diff and some digest helper + * functions to generate the digest hashes. + * + * gen_ed_diff is the tricky bit. In it simplest form, it will take quadratic + * time and linear space to generate an ed diff given two smartlists. As shown + * in its comment section, calling calc_changes on the entire two consensuses + * will calculate what is to be added and what is to be deleted in the diff. + * Its comment section briefly explains how it works. + * + * In our case specific to consensuses, we take advantage of the fact that + * consensuses list routers sorted by their identities. We use that + * information to avoid running calc_changes on the whole smartlists. + * gen_ed_diff will navigate through the two consensuses identity by identity + * and will send small couples of slices to calc_changes, keeping the running + * time near-linear. This is explained in more detail in the gen_ed_diff + * comments. + * + * The allocation strategy tries to save time and memory by avoiding needless + * copies. Instead of actually splitting the inputs into separate strings, we + * allocate cdline_t objects, each of which represents a line in the original + * object or in the output. We use memarea_t allocators to manage the + * temporary memory we use when generating or applying diffs. + **/ + +#define CONSDIFF_PRIVATE + +#include "or.h" +#include "consdiff.h" +#include "memarea.h" +#include "routerparse.h" + +static const char* ns_diff_version = "network-status-diff-version 1"; +static const char* hash_token = "hash"; + +static char *consensus_join_lines(const smartlist_t *inp); + +/** Return true iff a and b have the same contents. */ +STATIC int +lines_eq(const cdline_t *a, const cdline_t *b) +{ + return a->len == b->len && fast_memeq(a->s, b->s, a->len); +} + +/** Return true iff a has the same contents as the nul-terminated string b. */ +STATIC int +line_str_eq(const cdline_t *a, const char *b) +{ + const size_t len = strlen(b); + tor_assert(len <= UINT32_MAX); + cdline_t bline = { b, (uint32_t)len }; + return lines_eq(a, &bline); +} + +/** Add a cdline_t to <b>lst</b> holding as its contents the nul-terminated + * string s. Use the provided memory area for storage. */ +STATIC void +smartlist_add_linecpy(smartlist_t *lst, memarea_t *area, const char *s) +{ + size_t len = strlen(s); + const char *ss = memarea_memdup(area, s, len); + cdline_t *line = memarea_alloc(area, sizeof(cdline_t)); + line->s = ss; + line->len = (uint32_t)len; + smartlist_add(lst, line); +} + +/** Compute the digest of <b>cons</b>, and store the result in + * <b>digest_out</b>. Return 0 on success, -1 on failure. */ +/* This is a separate, mockable function so that we can override it when + * fuzzing. */ +MOCK_IMPL(STATIC int, +consensus_compute_digest,(const char *cons, + consensus_digest_t *digest_out)) +{ + int r = crypto_digest256((char*)digest_out->sha3_256, + cons, strlen(cons), DIGEST_SHA3_256); + return r; +} + +/** Return true iff <b>d1</b> and <b>d2</b> contain the same digest */ +/* This is a separate, mockable function so that we can override it when + * fuzzing. */ +MOCK_IMPL(STATIC int, +consensus_digest_eq,(const uint8_t *d1, + const uint8_t *d2)) +{ + return fast_memeq(d1, d2, DIGEST256_LEN); +} + +/** Create (allocate) a new slice from a smartlist. Assumes that the start + * and the end indexes are within the bounds of the initial smartlist. The end + * element is not part of the resulting slice. If end is -1, the slice is to + * reach the end of the smartlist. + */ +STATIC smartlist_slice_t * +smartlist_slice(const smartlist_t *list, int start, int end) +{ + int list_len = smartlist_len(list); + tor_assert(start >= 0); + tor_assert(start <= list_len); + if (end == -1) { + end = list_len; + } + tor_assert(start <= end); + + smartlist_slice_t *slice = tor_malloc(sizeof(smartlist_slice_t)); + slice->list = list; + slice->offset = start; + slice->len = end - start; + return slice; +} + +/** Helper: Compute the longest common subsequence lengths for the two slices. + * Used as part of the diff generation to find the column at which to split + * slice2 while still having the optimal solution. + * If direction is -1, the navigation is reversed. Otherwise it must be 1. + * The length of the resulting integer array is that of the second slice plus + * one. + */ +STATIC int * +lcs_lengths(const smartlist_slice_t *slice1, const smartlist_slice_t *slice2, + int direction) +{ + size_t a_size = sizeof(int) * (slice2->len+1); + + /* Resulting lcs lengths. */ + int *result = tor_malloc_zero(a_size); + /* Copy of the lcs lengths from the last iteration. */ + int *prev = tor_malloc(a_size); + + tor_assert(direction == 1 || direction == -1); + + int si = slice1->offset; + if (direction == -1) { + si += (slice1->len-1); + } + + for (int i = 0; i < slice1->len; ++i, si+=direction) { + + const cdline_t *line1 = smartlist_get(slice1->list, si); + /* Store the last results. */ + memcpy(prev, result, a_size); + + int sj = slice2->offset; + if (direction == -1) { + sj += (slice2->len-1); + } + + for (int j = 0; j < slice2->len; ++j, sj+=direction) { + + const cdline_t *line2 = smartlist_get(slice2->list, sj); + if (lines_eq(line1, line2)) { + /* If the lines are equal, the lcs is one line longer. */ + result[j + 1] = prev[j] + 1; + } else { + /* If not, see what lcs parent path is longer. */ + result[j + 1] = MAX(result[j], prev[j + 1]); + } + } + } + tor_free(prev); + return result; +} + +/** Helper: Trim any number of lines that are equally at the start or the end + * of both slices. + */ +STATIC void +trim_slices(smartlist_slice_t *slice1, smartlist_slice_t *slice2) +{ + while (slice1->len>0 && slice2->len>0) { + const cdline_t *line1 = smartlist_get(slice1->list, slice1->offset); + const cdline_t *line2 = smartlist_get(slice2->list, slice2->offset); + if (!lines_eq(line1, line2)) { + break; + } + slice1->offset++; slice1->len--; + slice2->offset++; slice2->len--; + } + + int i1 = (slice1->offset+slice1->len)-1; + int i2 = (slice2->offset+slice2->len)-1; + + while (slice1->len>0 && slice2->len>0) { + const cdline_t *line1 = smartlist_get(slice1->list, i1); + const cdline_t *line2 = smartlist_get(slice2->list, i2); + if (!lines_eq(line1, line2)) { + break; + } + i1--; + slice1->len--; + i2--; + slice2->len--; + } +} + +/** Like smartlist_string_pos, but uses a cdline_t, and is restricted to the + * bounds of the slice. + */ +STATIC int +smartlist_slice_string_pos(const smartlist_slice_t *slice, + const cdline_t *string) +{ + int end = slice->offset + slice->len; + for (int i = slice->offset; i < end; ++i) { + const cdline_t *el = smartlist_get(slice->list, i); + if (lines_eq(el, string)) { + return i; + } + } + return -1; +} + +/** Helper: Set all the appropriate changed booleans to true. The first slice + * must be of length 0 or 1. All the lines of slice1 and slice2 which are not + * present in the other slice will be set to changed in their bool array. + * The two changed bool arrays are passed in the same order as the slices. + */ +STATIC void +set_changed(bitarray_t *changed1, bitarray_t *changed2, + const smartlist_slice_t *slice1, const smartlist_slice_t *slice2) +{ + int toskip = -1; + tor_assert(slice1->len == 0 || slice1->len == 1); + + if (slice1->len == 1) { + const cdline_t *line_common = smartlist_get(slice1->list, slice1->offset); + toskip = smartlist_slice_string_pos(slice2, line_common); + if (toskip == -1) { + bitarray_set(changed1, slice1->offset); + } + } + int end = slice2->offset + slice2->len; + for (int i = slice2->offset; i < end; ++i) { + if (i != toskip) { + bitarray_set(changed2, i); + } + } +} + +/* + * Helper: Given that slice1 has been split by half into top and bot, we want + * to fetch the column at which to split slice2 so that we are still on track + * to the optimal diff solution, i.e. the shortest one. We use lcs_lengths + * since the shortest diff is just another way to say the longest common + * subsequence. + */ +static int +optimal_column_to_split(const smartlist_slice_t *top, + const smartlist_slice_t *bot, + const smartlist_slice_t *slice2) +{ + int *lens_top = lcs_lengths(top, slice2, 1); + int *lens_bot = lcs_lengths(bot, slice2, -1); + int column=0, max_sum=-1; + + for (int i = 0; i < slice2->len+1; ++i) { + int sum = lens_top[i] + lens_bot[slice2->len-i]; + if (sum > max_sum) { + column = i; + max_sum = sum; + } + } + tor_free(lens_top); + tor_free(lens_bot); + + return column; +} + +/** + * Helper: Figure out what elements are new or gone on the second smartlist + * relative to the first smartlist, and store the booleans in the bitarrays. + * True on the first bitarray means the element is gone, true on the second + * bitarray means it's new. + * + * In its base case, either of the smartlists is of length <= 1 and we can + * quickly see what elements are new or are gone. In the other case, we will + * split one smartlist by half and we'll use optimal_column_to_split to find + * the optimal column at which to split the second smartlist so that we are + * finding the smallest diff possible. + */ +STATIC void +calc_changes(smartlist_slice_t *slice1, + smartlist_slice_t *slice2, + bitarray_t *changed1, bitarray_t *changed2) +{ + trim_slices(slice1, slice2); + + if (slice1->len <= 1) { + set_changed(changed1, changed2, slice1, slice2); + + } else if (slice2->len <= 1) { + set_changed(changed2, changed1, slice2, slice1); + + /* Keep on splitting the slices in two. */ + } else { + smartlist_slice_t *top, *bot, *left, *right; + + /* Split the first slice in half. */ + int mid = slice1->len/2; + top = smartlist_slice(slice1->list, slice1->offset, slice1->offset+mid); + bot = smartlist_slice(slice1->list, slice1->offset+mid, + slice1->offset+slice1->len); + + /* Split the second slice by the optimal column. */ + int mid2 = optimal_column_to_split(top, bot, slice2); + left = smartlist_slice(slice2->list, slice2->offset, slice2->offset+mid2); + right = smartlist_slice(slice2->list, slice2->offset+mid2, + slice2->offset+slice2->len); + + calc_changes(top, left, changed1, changed2); + calc_changes(bot, right, changed1, changed2); + tor_free(top); + tor_free(bot); + tor_free(left); + tor_free(right); + } +} + +/* This table is from crypto.c. The SP and PAD defines are different. */ +#define NOT_VALID_BASE64 255 +#define X NOT_VALID_BASE64 +#define SP NOT_VALID_BASE64 +#define PAD NOT_VALID_BASE64 +static const uint8_t base64_compare_table[256] = { + X, X, X, X, X, X, X, X, X, SP, SP, SP, X, SP, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + SP, X, X, X, X, X, X, X, X, X, X, 62, X, X, X, 63, + 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, X, X, X, PAD, X, X, + X, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, + 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, X, X, X, X, X, + X, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, + 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, + X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, X, +}; + +/** Helper: Get the identity hash from a router line, assuming that the line + * at least appears to be a router line and thus starts with "r ". + * + * If an identity hash is found, store it (without decoding it) in + * <b>hash_out</b>, and return 0. On failure, return -1. + */ +STATIC int +get_id_hash(const cdline_t *line, cdline_t *hash_out) +{ + if (line->len < 2) + return -1; + + /* Skip the router name. */ + const char *hash = memchr(line->s + 2, ' ', line->len - 2); + if (!hash) { + return -1; + } + + hash++; + const char *hash_end = hash; + /* Stop when the first non-base64 character is found. Use unsigned chars to + * avoid negative indexes causing crashes. + */ + while (base64_compare_table[*((unsigned char*)hash_end)] + != NOT_VALID_BASE64 && + hash_end < line->s + line->len) { + hash_end++; + } + + /* Empty hash. */ + if (hash_end == hash) { + return -1; + } + + hash_out->s = hash; + /* Always true because lines are limited to this length */ + tor_assert(hash_end - hash <= UINT32_MAX); + hash_out->len = (uint32_t)(hash_end - hash); + + return 0; +} + +/** Helper: Check that a line is a valid router entry. We must at least be + * able to fetch a proper identity hash from it for it to be valid. + */ +STATIC int +is_valid_router_entry(const cdline_t *line) +{ + if (line->len < 2 || fast_memneq(line->s, "r ", 2)) + return 0; + cdline_t tmp; + return (get_id_hash(line, &tmp) == 0); +} + +/** Helper: Find the next router line starting at the current position. + * Assumes that cur is lower than the length of the smartlist, i.e. it is a + * line within the bounds of the consensus. The only exception is when we + * don't want to skip the first line, in which case cur will be -1. + */ +STATIC int +next_router(const smartlist_t *cons, int cur) +{ + int len = smartlist_len(cons); + tor_assert(cur >= -1 && cur < len); + + if (++cur >= len) { + return len; + } + + const cdline_t *line = smartlist_get(cons, cur); + while (!is_valid_router_entry(line)) { + if (++cur >= len) { + return len; + } + line = smartlist_get(cons, cur); + } + return cur; +} + +/** Helper: compare two base64-encoded identity hashes, which may be of + * different lengths. Comparison ends when the first non-base64 char is found. + */ +STATIC int +base64cmp(const cdline_t *hash1, const cdline_t *hash2) +{ + /* NULL is always lower, useful for last_hash which starts at NULL. */ + if (!hash1->s && !hash2->s) { + return 0; + } + if (!hash1->s) { + return -1; + } + if (!hash2->s) { + return 1; + } + + /* Don't index with a char; char may be signed. */ + const unsigned char *a = (unsigned char*)hash1->s; + const unsigned char *b = (unsigned char*)hash2->s; + const unsigned char *a_end = a + hash1->len; + const unsigned char *b_end = b + hash2->len; + while (1) { + uint8_t av = base64_compare_table[*a]; + uint8_t bv = base64_compare_table[*b]; + if (av == NOT_VALID_BASE64) { + if (bv == NOT_VALID_BASE64) { + /* Both ended with exactly the same characters. */ + return 0; + } else { + /* hash2 goes on longer than hash1 and thus hash1 is lower. */ + return -1; + } + } else if (bv == NOT_VALID_BASE64) { + /* hash1 goes on longer than hash2 and thus hash1 is greater. */ + return 1; + } else if (av < bv) { + /* The first difference shows that hash1 is lower. */ + return -1; + } else if (av > bv) { + /* The first difference shows that hash1 is greater. */ + return 1; + } else { + a++; + b++; + if (a == a_end) { + if (b == b_end) { + return 0; + } else { + return -1; + } + } else if (b == b_end) { + return 1; + } + } + } +} + +/** Structure used to remember the previous and current identity hash of + * the "r " lines in a consensus, to enforce well-ordering. */ +typedef struct router_id_iterator_t { + cdline_t last_hash; + cdline_t hash; +} router_id_iterator_t; + +/** + * Initializer for a router_id_iterator_t. + */ +#define ROUTER_ID_ITERATOR_INIT { { NULL, 0 }, { NULL, 0 } } + +/** Given an index *<b>idxp</b> into the consensus at <b>cons</b>, advance + * the index to the next router line ("r ...") in the consensus, or to + * an index one after the end of the list if there is no such line. + * + * Use <b>iter</b> to record the hash of the found router line, if any, + * and to enforce ordering on the hashes. If the hashes are mis-ordered, + * return -1. Else, return 0. + **/ +static int +find_next_router_line(const smartlist_t *cons, + const char *consname, + int *idxp, + router_id_iterator_t *iter) +{ + *idxp = next_router(cons, *idxp); + if (*idxp < smartlist_len(cons)) { + memcpy(&iter->last_hash, &iter->hash, sizeof(cdline_t)); + if (get_id_hash(smartlist_get(cons, *idxp), &iter->hash) < 0 || + base64cmp(&iter->hash, &iter->last_hash) <= 0) { + log_warn(LD_CONSDIFF, "Refusing to generate consensus diff because " + "the %s consensus doesn't have its router entries sorted " + "properly.", consname); + return -1; + } + } + return 0; +} + +/** Generate an ed diff as a smartlist from two consensuses, also given as + * smartlists. Will return NULL if the diff could not be generated, which can + * happen if any lines the script had to add matched "." or if the routers + * were not properly ordered. + * + * All cdline_t objects in the resulting object are either references to lines + * in one of the inputs, or are newly allocated lines in the provided memarea. + * + * This implementation is consensus-specific. To generate an ed diff for any + * given input in quadratic time, you can replace all the code until the + * navigation in reverse order with the following: + * + * int len1 = smartlist_len(cons1); + * int len2 = smartlist_len(cons2); + * bitarray_t *changed1 = bitarray_init_zero(len1); + * bitarray_t *changed2 = bitarray_init_zero(len2); + * cons1_sl = smartlist_slice(cons1, 0, -1); + * cons2_sl = smartlist_slice(cons2, 0, -1); + * calc_changes(cons1_sl, cons2_sl, changed1, changed2); + */ +STATIC smartlist_t * +gen_ed_diff(const smartlist_t *cons1, const smartlist_t *cons2, + memarea_t *area) +{ + int len1 = smartlist_len(cons1); + int len2 = smartlist_len(cons2); + smartlist_t *result = smartlist_new(); + + /* Initialize the changed bitarrays to zero, so that calc_changes only needs + * to set the ones that matter and leave the rest untouched. + */ + bitarray_t *changed1 = bitarray_init_zero(len1); + bitarray_t *changed2 = bitarray_init_zero(len2); + int i1=-1, i2=-1; + int start1=0, start2=0; + + /* To check that hashes are ordered properly */ + router_id_iterator_t iter1 = ROUTER_ID_ITERATOR_INIT; + router_id_iterator_t iter2 = ROUTER_ID_ITERATOR_INIT; + + /* i1 and i2 are initialized at the first line of each consensus. They never + * reach past len1 and len2 respectively, since next_router doesn't let that + * happen. i1 and i2 are advanced by at least one line at each iteration as + * long as they have not yet reached len1 and len2, so the loop is + * guaranteed to end, and each pair of (i1,i2) will be inspected at most + * once. + */ + while (i1 < len1 || i2 < len2) { + + /* Advance each of the two navigation positions by one router entry if not + * yet at the end. + */ + if (i1 < len1) { + if (find_next_router_line(cons1, "base", &i1, &iter1) < 0) { + goto error_cleanup; + } + } + + if (i2 < len2) { + if (find_next_router_line(cons2, "target", &i2, &iter2) < 0) { + goto error_cleanup; + } + } + + /* If we have reached the end of both consensuses, there is no need to + * compare hashes anymore, since this is the last iteration. + */ + if (i1 < len1 || i2 < len2) { + + /* Keep on advancing the lower (by identity hash sorting) position until + * we have two matching positions. The only other possible outcome is + * that a lower position reaches the end of the consensus before it can + * reach a hash that is no longer the lower one. Since there will always + * be a lower hash for as long as the loop runs, one of the two indexes + * will always be incremented, thus assuring that the loop must end + * after a finite number of iterations. If that cannot be because said + * consensus has already reached the end, both are extended to their + * respecting ends since we are done. + */ + int cmp = base64cmp(&iter1.hash, &iter2.hash); + while (cmp != 0) { + if (i1 < len1 && cmp < 0) { + if (find_next_router_line(cons1, "base", &i1, &iter1) < 0) { + goto error_cleanup; + } + if (i1 == len1) { + /* We finished the first consensus, so grab all the remaining + * lines of the second consensus and finish up. + */ + i2 = len2; + break; + } + } else if (i2 < len2 && cmp > 0) { + if (find_next_router_line(cons2, "target", &i2, &iter2) < 0) { + goto error_cleanup; + } + if (i2 == len2) { + /* We finished the second consensus, so grab all the remaining + * lines of the first consensus and finish up. + */ + i1 = len1; + break; + } + } else { + i1 = len1; + i2 = len2; + break; + } + cmp = base64cmp(&iter1.hash, &iter2.hash); + } + } + + /* Make slices out of these chunks (up to the common router entry) and + * calculate the changes for them. + * Error if any of the two slices are longer than 10K lines. That should + * never happen with any pair of real consensuses. Feeding more than 10K + * lines to calc_changes would be very slow anyway. + */ +#define MAX_LINE_COUNT (10000) + if (i1-start1 > MAX_LINE_COUNT || i2-start2 > MAX_LINE_COUNT) { + log_warn(LD_CONSDIFF, "Refusing to generate consensus diff because " + "we found too few common router ids."); + goto error_cleanup; + } + + smartlist_slice_t *cons1_sl = smartlist_slice(cons1, start1, i1); + smartlist_slice_t *cons2_sl = smartlist_slice(cons2, start2, i2); + calc_changes(cons1_sl, cons2_sl, changed1, changed2); + tor_free(cons1_sl); + tor_free(cons2_sl); + start1 = i1, start2 = i2; + } + + /* Navigate the changes in reverse order and generate one ed command for + * each chunk of changes. + */ + i1=len1-1, i2=len2-1; + char buf[128]; + while (i1 > 0 || i2 > 0) { + + int start1x, start2x, end1, end2, added, deleted; + + /* We are at a point were no changed bools are true, so just keep going. */ + if (!(i1 >= 0 && bitarray_is_set(changed1, i1)) && + !(i2 >= 0 && bitarray_is_set(changed2, i2))) { + if (i1 >= 0) { + i1--; + } + if (i2 >= 0) { + i2--; + } + continue; + } + + end1 = i1, end2 = i2; + + /* Grab all contiguous changed lines */ + while (i1 >= 0 && bitarray_is_set(changed1, i1)) { + i1--; + } + while (i2 >= 0 && bitarray_is_set(changed2, i2)) { + i2--; + } + + start1x = i1+1, start2x = i2+1; + added = end2-i2, deleted = end1-i1; + + if (added == 0) { + if (deleted == 1) { + tor_snprintf(buf, sizeof(buf), "%id", start1x+1); + smartlist_add_linecpy(result, area, buf); + } else { + tor_snprintf(buf, sizeof(buf), "%i,%id", start1x+1, start1x+deleted); + smartlist_add_linecpy(result, area, buf); + } + } else { + int i; + if (deleted == 0) { + tor_snprintf(buf, sizeof(buf), "%ia", start1x); + smartlist_add_linecpy(result, area, buf); + } else if (deleted == 1) { + tor_snprintf(buf, sizeof(buf), "%ic", start1x+1); + smartlist_add_linecpy(result, area, buf); + } else { + tor_snprintf(buf, sizeof(buf), "%i,%ic", start1x+1, start1x+deleted); + smartlist_add_linecpy(result, area, buf); + } + + for (i = start2x; i <= end2; ++i) { + cdline_t *line = smartlist_get(cons2, i); + if (line_str_eq(line, ".")) { + log_warn(LD_CONSDIFF, "Cannot generate consensus diff because " + "one of the lines to be added is \".\"."); + goto error_cleanup; + } + smartlist_add(result, line); + } + smartlist_add_linecpy(result, area, "."); + } + } + + bitarray_free(changed1); + bitarray_free(changed2); + + return result; + + error_cleanup: + + bitarray_free(changed1); + bitarray_free(changed2); + + smartlist_free(result); + + return NULL; +} + +/* Helper: Read a base-10 number between 0 and INT32_MAX from <b>s</b> and + * store it in <b>num_out</b>. Advance <b>s</b> to the characer immediately + * after the number. Return 0 on success, -1 on failure. */ +static int +get_linenum(const char **s, int *num_out) +{ + int ok; + char *next; + *num_out = (int) tor_parse_long(*s, 10, 0, INT32_MAX, &ok, &next); + if (ok && next) { + *s = next; + return 0; + } else { + return -1; + } +} + +/** Apply the ed diff, starting at <b>diff_starting_line</b>, to the consensus + * and return a new consensus, also as a line-based smartlist. Will return + * NULL if the ed diff is not properly formatted. + * + * All cdline_t objects in the resulting object are references to lines + * in one of the inputs; nothing is copied. + */ +STATIC smartlist_t * +apply_ed_diff(const smartlist_t *cons1, const smartlist_t *diff, + int diff_starting_line) +{ + int diff_len = smartlist_len(diff); + int j = smartlist_len(cons1); + smartlist_t *cons2 = smartlist_new(); + + for (int i=diff_starting_line; i<diff_len; ++i) { + const cdline_t *diff_cdline = smartlist_get(diff, i); + char diff_line[128]; + + if (diff_cdline->len > sizeof(diff_line) - 1) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an ed command was far too long"); + goto error_cleanup; + } + /* Copy the line to make it nul-terminated. */ + memcpy(diff_line, diff_cdline->s, diff_cdline->len); + diff_line[diff_cdline->len] = 0; + const char *ptr = diff_line; + int start = 0, end = 0; + if (get_linenum(&ptr, &start) < 0) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an ed command was missing a line number."); + goto error_cleanup; + } + if (*ptr == ',') { + /* Two-item range */ + ++ptr; + if (get_linenum(&ptr, &end) < 0) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an ed command was missing a range end line number."); + goto error_cleanup; + } + /* Incoherent range. */ + if (end <= start) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an invalid range was found in an ed command."); + goto error_cleanup; + } + } else { + /* We'll take <n1> as <n1>,<n1> for simplicity. */ + end = start; + } + + if (end > j) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "its commands are not properly sorted in reverse order."); + goto error_cleanup; + } + + if (*ptr == '\0') { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "a line with no ed command was found"); + goto error_cleanup; + } + + if (*(ptr+1) != '\0') { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an ed command longer than one char was found."); + goto error_cleanup; + } + + char action = *ptr; + + switch (action) { + case 'a': + case 'c': + case 'd': + break; + default: + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "an unrecognised ed command was found."); + goto error_cleanup; + } + + /* Add unchanged lines. */ + for (; j && j > end; --j) { + cdline_t *cons_line = smartlist_get(cons1, j-1); + smartlist_add(cons2, cons_line); + } + + /* Ignore removed lines. */ + if (action == 'c' || action == 'd') { + while (--j >= start) { + /* Skip line */ + } + } + + /* Add new lines in reverse order, since it will all be reversed at the + * end. + */ + if (action == 'a' || action == 'c') { + int added_end = i; + + i++; /* Skip the line with the range and command. */ + while (i < diff_len) { + if (line_str_eq(smartlist_get(diff, i), ".")) { + break; + } + if (++i == diff_len) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "it has lines to be inserted that don't end with a \".\"."); + goto error_cleanup; + } + } + + int added_i = i-1; + + /* It would make no sense to add zero new lines. */ + if (added_i == added_end) { + log_warn(LD_CONSDIFF, "Could not apply consensus diff because " + "it has an ed command that tries to insert zero lines."); + goto error_cleanup; + } + + while (added_i > added_end) { + cdline_t *added_line = smartlist_get(diff, added_i--); + smartlist_add(cons2, added_line); + } + } + } + + /* Add remaining unchanged lines. */ + for (; j > 0; --j) { + cdline_t *cons_line = smartlist_get(cons1, j-1); + smartlist_add(cons2, cons_line); + } + + /* Reverse the whole thing since we did it from the end. */ + smartlist_reverse(cons2); + return cons2; + + error_cleanup: + + smartlist_free(cons2); + + return NULL; +} + +/** Generate a consensus diff as a smartlist from two given consensuses, also + * as smartlists. Will return NULL if the consensus diff could not be + * generated. Neither of the two consensuses are modified in any way, so it's + * up to the caller to free their resources. + */ +smartlist_t * +consdiff_gen_diff(const smartlist_t *cons1, + const smartlist_t *cons2, + const consensus_digest_t *digests1, + const consensus_digest_t *digests2, + memarea_t *area) +{ + smartlist_t *ed_diff = gen_ed_diff(cons1, cons2, area); + /* ed diff could not be generated - reason already logged by gen_ed_diff. */ + if (!ed_diff) { + goto error_cleanup; + } + + /* See that the script actually produces what we want. */ + smartlist_t *ed_cons2 = apply_ed_diff(cons1, ed_diff, 0); + if (!ed_cons2) { + /* LCOV_EXCL_START -- impossible if diff generation is correct */ + log_warn(LD_BUG|LD_CONSDIFF, "Refusing to generate consensus diff because " + "the generated ed diff could not be tested to successfully generate " + "the target consensus."); + goto error_cleanup; + /* LCOV_EXCL_STOP */ + } + + int cons2_eq = 1; + if (smartlist_len(cons2) == smartlist_len(ed_cons2)) { + SMARTLIST_FOREACH_BEGIN(cons2, const cdline_t *, line1) { + const cdline_t *line2 = smartlist_get(ed_cons2, line1_sl_idx); + if (! lines_eq(line1, line2) ) { + cons2_eq = 0; + break; + } + } SMARTLIST_FOREACH_END(line1); + } else { + cons2_eq = 0; + } + smartlist_free(ed_cons2); + if (!cons2_eq) { + /* LCOV_EXCL_START -- impossible if diff generation is correct. */ + log_warn(LD_BUG|LD_CONSDIFF, "Refusing to generate consensus diff because " + "the generated ed diff did not generate the target consensus " + "successfully when tested."); + goto error_cleanup; + /* LCOV_EXCL_STOP */ + } + + char cons1_hash_hex[HEX_DIGEST256_LEN+1]; + char cons2_hash_hex[HEX_DIGEST256_LEN+1]; + base16_encode(cons1_hash_hex, HEX_DIGEST256_LEN+1, + (const char*)digests1->sha3_256, DIGEST256_LEN); + base16_encode(cons2_hash_hex, HEX_DIGEST256_LEN+1, + (const char*)digests2->sha3_256, DIGEST256_LEN); + + /* Create the resulting consensus diff. */ + char buf[160]; + smartlist_t *result = smartlist_new(); + tor_snprintf(buf, sizeof(buf), "%s", ns_diff_version); + smartlist_add_linecpy(result, area, buf); + tor_snprintf(buf, sizeof(buf), "%s %s %s", hash_token, + cons1_hash_hex, cons2_hash_hex); + smartlist_add_linecpy(result, area, buf); + smartlist_add_all(result, ed_diff); + smartlist_free(ed_diff); + return result; + + error_cleanup: + + if (ed_diff) { + /* LCOV_EXCL_START -- ed_diff is NULL except in unreachable cases above */ + smartlist_free(ed_diff); + /* LCOV_EXCL_STOP */ + } + + return NULL; +} + +/** Fetch the digest of the base consensus in the consensus diff, encoded in + * base16 as found in the diff itself. digest1_out and digest2_out must be of + * length DIGEST256_LEN or larger if not NULL. + */ +int +consdiff_get_digests(const smartlist_t *diff, + char *digest1_out, + char *digest2_out) +{ + smartlist_t *hash_words = NULL; + const cdline_t *format; + char cons1_hash[DIGEST256_LEN], cons2_hash[DIGEST256_LEN]; + char *cons1_hash_hex, *cons2_hash_hex; + if (smartlist_len(diff) < 2) { + log_info(LD_CONSDIFF, "The provided consensus diff is too short."); + goto error_cleanup; + } + + /* Check that it's the format and version we know. */ + format = smartlist_get(diff, 0); + if (!line_str_eq(format, ns_diff_version)) { + log_warn(LD_CONSDIFF, "The provided consensus diff format is not known."); + goto error_cleanup; + } + + /* Grab the base16 digests. */ + hash_words = smartlist_new(); + { + const cdline_t *line2 = smartlist_get(diff, 1); + char *h = tor_memdup_nulterm(line2->s, line2->len); + smartlist_split_string(hash_words, h, " ", 0, 0); + tor_free(h); + } + + /* There have to be three words, the first of which must be hash_token. */ + if (smartlist_len(hash_words) != 3 || + strcmp(smartlist_get(hash_words, 0), hash_token)) { + log_info(LD_CONSDIFF, "The provided consensus diff does not include " + "the necessary digests."); + goto error_cleanup; + } + + /* Expected hashes as found in the consensus diff header. They must be of + * length HEX_DIGEST256_LEN, normally 64 hexadecimal characters. + * If any of the decodings fail, error to make sure that the hashes are + * proper base16-encoded digests. + */ + cons1_hash_hex = smartlist_get(hash_words, 1); + cons2_hash_hex = smartlist_get(hash_words, 2); + if (strlen(cons1_hash_hex) != HEX_DIGEST256_LEN || + strlen(cons2_hash_hex) != HEX_DIGEST256_LEN) { + log_info(LD_CONSDIFF, "The provided consensus diff includes " + "base16-encoded digests of incorrect size."); + goto error_cleanup; + } + + if (base16_decode(cons1_hash, DIGEST256_LEN, + cons1_hash_hex, HEX_DIGEST256_LEN) != DIGEST256_LEN || + base16_decode(cons2_hash, DIGEST256_LEN, + cons2_hash_hex, HEX_DIGEST256_LEN) != DIGEST256_LEN) { + log_info(LD_CONSDIFF, "The provided consensus diff includes " + "malformed digests."); + goto error_cleanup; + } + + if (digest1_out) { + memcpy(digest1_out, cons1_hash, DIGEST256_LEN); + } + if (digest2_out) { + memcpy(digest2_out, cons2_hash, DIGEST256_LEN); + } + + SMARTLIST_FOREACH(hash_words, char *, cp, tor_free(cp)); + smartlist_free(hash_words); + return 0; + + error_cleanup: + + if (hash_words) { + SMARTLIST_FOREACH(hash_words, char *, cp, tor_free(cp)); + smartlist_free(hash_words); + } + return 1; +} + +/** Apply the consensus diff to the given consensus and return a new + * consensus, also as a line-based smartlist. Will return NULL if the diff + * could not be applied. Neither the consensus nor the diff are modified in + * any way, so it's up to the caller to free their resources. + */ +char * +consdiff_apply_diff(const smartlist_t *cons1, + const smartlist_t *diff, + const consensus_digest_t *digests1) +{ + smartlist_t *cons2 = NULL; + char *cons2_str = NULL; + char e_cons1_hash[DIGEST256_LEN]; + char e_cons2_hash[DIGEST256_LEN]; + + if (consdiff_get_digests(diff, e_cons1_hash, e_cons2_hash) != 0) { + goto error_cleanup; + } + + /* See that the consensus that was given to us matches its hash. */ + if (!consensus_digest_eq(digests1->sha3_256, + (const uint8_t*)e_cons1_hash)) { + char hex_digest1[HEX_DIGEST256_LEN+1]; + char e_hex_digest1[HEX_DIGEST256_LEN+1]; + log_warn(LD_CONSDIFF, "Refusing to apply consensus diff because " + "the base consensus doesn't match the digest as found in " + "the consensus diff header."); + base16_encode(hex_digest1, HEX_DIGEST256_LEN+1, + (const char *)digests1->sha3_256, DIGEST256_LEN); + base16_encode(e_hex_digest1, HEX_DIGEST256_LEN+1, + e_cons1_hash, DIGEST256_LEN); + log_warn(LD_CONSDIFF, "Expected: %s; found: %s", + hex_digest1, e_hex_digest1); + goto error_cleanup; + } + + /* Grab the ed diff and calculate the resulting consensus. */ + /* Skip the first two lines. */ + cons2 = apply_ed_diff(cons1, diff, 2); + + /* ed diff could not be applied - reason already logged by apply_ed_diff. */ + if (!cons2) { + goto error_cleanup; + } + + cons2_str = consensus_join_lines(cons2); + + consensus_digest_t cons2_digests; + if (consensus_compute_digest(cons2_str, &cons2_digests) < 0) { + /* LCOV_EXCL_START -- digest can't fail */ + log_warn(LD_CONSDIFF, "Could not compute digests of the consensus " + "resulting from applying a consensus diff."); + goto error_cleanup; + /* LCOV_EXCL_STOP */ + } + + /* See that the resulting consensus matches its hash. */ + if (!consensus_digest_eq(cons2_digests.sha3_256, + (const uint8_t*)e_cons2_hash)) { + log_warn(LD_CONSDIFF, "Refusing to apply consensus diff because " + "the resulting consensus doesn't match the digest as found in " + "the consensus diff header."); + char hex_digest2[HEX_DIGEST256_LEN+1]; + char e_hex_digest2[HEX_DIGEST256_LEN+1]; + base16_encode(hex_digest2, HEX_DIGEST256_LEN+1, + (const char *)cons2_digests.sha3_256, DIGEST256_LEN); + base16_encode(e_hex_digest2, HEX_DIGEST256_LEN+1, + e_cons2_hash, DIGEST256_LEN); + log_warn(LD_CONSDIFF, "Expected: %s; found: %s", + hex_digest2, e_hex_digest2); + goto error_cleanup; + } + + goto done; + + error_cleanup: + tor_free(cons2_str); /* Sets it to NULL */ + + done: + if (cons2) { + smartlist_free(cons2); + } + + return cons2_str; +} + +/** Any consensus line longer than this means that the input is invalid. */ +#define CONSENSUS_LINE_MAX_LEN (1<<20) + +/** + * Helper: For every NL-terminated line in <b>s</b>, add a cdline referring to + * that line (without trailing newline) to <b>out</b>. Return -1 if there are + * any non-NL terminated lines; 0 otherwise. + * + * Unlike tor_split_lines, this function avoids ambiguity on its + * handling of a final line that isn't NL-terminated. + * + * All cdline_t objects are allocated in the provided memarea. Strings + * are not copied: if <b>s</b> changes or becomes invalid, then all + * generated cdlines will become invalid. + */ +STATIC int +consensus_split_lines(smartlist_t *out, const char *s, memarea_t *area) +{ + while (*s) { + const char *eol = strchr(s, '\n'); + if (!eol) { + /* File doesn't end with newline. */ + return -1; + } + if (eol - s > CONSENSUS_LINE_MAX_LEN) { + /* Line is far too long. */ + return -1; + } + cdline_t *line = memarea_alloc(area, sizeof(cdline_t)); + line->s = s; + line->len = (uint32_t)(eol - s); + smartlist_add(out, line); + s = eol+1; + } + return 0; +} + +/** Given a list of cdline_t, return a newly allocated string containing + * all of the lines, terminated with NL, concatenated. + * + * Unlike smartlist_join_strings(), avoids lossy operations on empty + * lists. */ +static char * +consensus_join_lines(const smartlist_t *inp) +{ + size_t n = 0; + SMARTLIST_FOREACH(inp, const cdline_t *, cdline, n += cdline->len + 1); + n += 1; + char *result = tor_malloc(n); + char *out = result; + SMARTLIST_FOREACH_BEGIN(inp, const cdline_t *, cdline) { + memcpy(out, cdline->s, cdline->len); + out += cdline->len; + *out++ = '\n'; + } SMARTLIST_FOREACH_END(cdline); + *out++ = '\0'; + tor_assert(out == result+n); + return result; +} + +/** Given two consensus documents, try to compute a diff between them. On + * success, retun a newly allocated string containing that diff. On failure, + * return NULL. */ +char * +consensus_diff_generate(const char *cons1, + const char *cons2) +{ + consensus_digest_t d1, d2; + smartlist_t *lines1 = NULL, *lines2 = NULL, *result_lines = NULL; + int r1, r2; + char *result = NULL; + + r1 = consensus_compute_digest(cons1, &d1); + r2 = consensus_compute_digest(cons2, &d2); + if (BUG(r1 < 0 || r2 < 0)) + return NULL; // LCOV_EXCL_LINE + + memarea_t *area = memarea_new(); + lines1 = smartlist_new(); + lines2 = smartlist_new(); + if (consensus_split_lines(lines1, cons1, area) < 0) + goto done; + if (consensus_split_lines(lines2, cons2, area) < 0) + goto done; + + result_lines = consdiff_gen_diff(lines1, lines2, &d1, &d2, area); + + done: + if (result_lines) { + result = consensus_join_lines(result_lines); + smartlist_free(result_lines); + } + + memarea_drop_all(area); + smartlist_free(lines1); + smartlist_free(lines2); + + return result; +} + +/** Given a consensus document and a diff, try to apply the diff to the + * consensus. On success return a newly allocated string containing the new + * consensus. On failure, return NULL. */ +char * +consensus_diff_apply(const char *consensus, + const char *diff) +{ + consensus_digest_t d1; + smartlist_t *lines1 = NULL, *lines2 = NULL; + int r1; + char *result = NULL; + memarea_t *area = memarea_new(); + + r1 = consensus_compute_digest(consensus, &d1); + if (BUG(r1 < 0)) + return NULL; // LCOV_EXCL_LINE + + lines1 = smartlist_new(); + lines2 = smartlist_new(); + if (consensus_split_lines(lines1, consensus, area) < 0) + goto done; + if (consensus_split_lines(lines2, diff, area) < 0) + goto done; + + result = consdiff_apply_diff(lines1, lines2, &d1); + + done: + smartlist_free(lines1); + smartlist_free(lines2); + memarea_drop_all(area); + + return result; +} + diff --git a/src/or/consdiff.h b/src/or/consdiff.h new file mode 100644 index 0000000000..e9d175136e --- /dev/null +++ b/src/or/consdiff.h @@ -0,0 +1,93 @@ +/* Copyright (c) 2014, Daniel Martà + * Copyright (c) 2014, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#ifndef TOR_CONSDIFF_H +#define TOR_CONSDIFF_H + +#include "or.h" + +char *consensus_diff_generate(const char *cons1, + const char *cons2); +char *consensus_diff_apply(const char *consensus, + const char *diff); + +#ifdef CONSDIFF_PRIVATE +struct memarea_t; + +/** Line type used for constructing consensus diffs. Each of these lines + * refers to a chunk of memory allocated elsewhere, and is not necessarily + * NUL-terminated: this helps us avoid copies and save memory. */ +typedef struct cdline_t { + const char *s; + uint32_t len; +} cdline_t; + +typedef struct consensus_digest_t { + uint8_t sha3_256[DIGEST256_LEN]; +} consensus_digest_t; + +STATIC smartlist_t *consdiff_gen_diff(const smartlist_t *cons1, + const smartlist_t *cons2, + const consensus_digest_t *digests1, + const consensus_digest_t *digests2, + struct memarea_t *area); +STATIC char *consdiff_apply_diff(const smartlist_t *cons1, + const smartlist_t *diff, + const consensus_digest_t *digests1); +STATIC int consdiff_get_digests(const smartlist_t *diff, + char *digest1_out, + char *digest2_out); + +/** Data structure to define a slice of a smarltist. */ +typedef struct smartlist_slice_t { + /** + * Smartlist that this slice is made from. + * References the whole original smartlist that the slice was made out of. + * */ + const smartlist_t *list; + /** Starting position of the slice in the smartlist. */ + int offset; + /** Length of the slice, i.e. the number of elements it holds. */ + int len; +} smartlist_slice_t; +STATIC smartlist_t *gen_ed_diff(const smartlist_t *cons1, + const smartlist_t *cons2, + struct memarea_t *area); +STATIC smartlist_t *apply_ed_diff(const smartlist_t *cons1, + const smartlist_t *diff, + int start_line); +STATIC void calc_changes(smartlist_slice_t *slice1, smartlist_slice_t *slice2, + bitarray_t *changed1, bitarray_t *changed2); +STATIC smartlist_slice_t *smartlist_slice(const smartlist_t *list, + int start, int end); +STATIC int next_router(const smartlist_t *cons, int cur); +STATIC int *lcs_lengths(const smartlist_slice_t *slice1, + const smartlist_slice_t *slice2, + int direction); +STATIC void trim_slices(smartlist_slice_t *slice1, smartlist_slice_t *slice2); +STATIC int base64cmp(const cdline_t *hash1, const cdline_t *hash2); +STATIC int get_id_hash(const cdline_t *line, cdline_t *hash_out); +STATIC int is_valid_router_entry(const cdline_t *line); +STATIC int smartlist_slice_string_pos(const smartlist_slice_t *slice, + const cdline_t *string); +STATIC void set_changed(bitarray_t *changed1, bitarray_t *changed2, + const smartlist_slice_t *slice1, + const smartlist_slice_t *slice2); +STATIC int consensus_split_lines(smartlist_t *out, const char *s, + struct memarea_t *area); +STATIC void smartlist_add_linecpy(smartlist_t *lst, struct memarea_t *area, + const char *s); +STATIC int lines_eq(const cdline_t *a, const cdline_t *b); +STATIC int line_str_eq(const cdline_t *a, const char *b); + +MOCK_DECL(STATIC int, + consensus_compute_digest,(const char *cons, + consensus_digest_t *digest_out)); +MOCK_DECL(STATIC int, + consensus_digest_eq,(const uint8_t *d1, + const uint8_t *d2)); +#endif + +#endif + diff --git a/src/or/include.am b/src/or/include.am index 4e54deca55..d65d0d3657 100644 --- a/src/or/include.am +++ b/src/or/include.am @@ -36,6 +36,7 @@ LIBTOR_A_SOURCES = \ src/or/connection.c \ src/or/connection_edge.c \ src/or/connection_or.c \ + src/or/consdiff.c \ src/or/control.c \ src/or/cpuworker.c \ src/or/dircollate.c \ @@ -151,6 +152,7 @@ ORHEADERS = \ src/or/connection.h \ src/or/connection_edge.h \ src/or/connection_or.h \ + src/or/consdiff.h \ src/or/control.h \ src/or/cpuworker.h \ src/or/dircollate.h \ diff --git a/src/test/Makefile.nmake b/src/test/Makefile.nmake index 0ba56d7036..575198388b 100644 --- a/src/test/Makefile.nmake +++ b/src/test/Makefile.nmake @@ -12,7 +12,7 @@ LIBS = ..\..\..\build-alpha\lib\libevent.lib \ crypt32.lib gdi32.lib user32.lib TEST_OBJECTS = test.obj test_addr.obj test_channel.obj test_channeltls.obj \ - test_containers.obj \ + test_consdiff.obj test_containers.obj \ test_controller_events.obj test_crypto.obj test_data.obj test_dir.obj \ test_checkdir.obj test_microdesc.obj test_pt.obj test_util.obj \ test_config.obj test_connection.obj \ diff --git a/src/test/bench.c b/src/test/bench.c index c3a9d1c4ba..5da9168cdc 100644 --- a/src/test/bench.c +++ b/src/test/bench.c @@ -28,6 +28,7 @@ const char tor_git_revision[] = ""; #include "crypto_curve25519.h" #include "onion_ntor.h" #include "crypto_ed25519.h" +#include "consdiff.h" #if defined(HAVE_CLOCK_GETTIME) && defined(CLOCK_PROCESS_CPUTIME_ID) static uint64_t nanostart; @@ -674,6 +675,27 @@ main(int argc, const char **argv) tor_threads_init(); + if (argc == 4 && !strcmp(argv[1], "diff")) { + init_logging(1); + const int N = 200; + char *f1 = read_file_to_str(argv[2], RFTS_BIN, NULL); + char *f2 = read_file_to_str(argv[3], RFTS_BIN, NULL); + if (! f1 || ! f2) { + perror("X"); + return 1; + } + for (i = 0; i < N; ++i) { + char *diff = consensus_diff_generate(f1, f2); + tor_free(diff); + } + char *diff = consensus_diff_generate(f1, f2); + printf("%s", diff); + tor_free(f1); + tor_free(f2); + tor_free(diff); + return 0; + } + for (i = 1; i < argc; ++i) { if (!strcmp(argv[i], "--list")) { list = 1; diff --git a/src/test/fuzz/fuzz_diff.c b/src/test/fuzz/fuzz_diff.c new file mode 100644 index 0000000000..c241f63dc4 --- /dev/null +++ b/src/test/fuzz/fuzz_diff.c @@ -0,0 +1,67 @@ +/* Copyright (c) 2016, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#define CONSDIFF_PRIVATE + +#include "orconfig.h" +#include "or.h" +#include "consdiff.h" + +#include "fuzzing.h" + +static int +mock_consensus_compute_digest_(const char *c, consensus_digest_t *d) +{ + (void)c; + memset(d->sha3_256, 3, sizeof(d->sha3_256)); + return 0; +} + +int +fuzz_init(void) +{ + MOCK(consensus_compute_digest, mock_consensus_compute_digest_); + return 0; +} + +int +fuzz_cleanup(void) +{ + UNMOCK(consensus_compute_digest); + return 0; +} + +int +fuzz_main(const uint8_t *stdin_buf, size_t data_size) +{ +#define SEP "=====\n" +#define SEPLEN strlen(SEP) + const uint8_t *separator = tor_memmem(stdin_buf, data_size, SEP, SEPLEN); + if (! separator) + return 0; + size_t c1_len = separator - stdin_buf; + char *c1 = tor_memdup_nulterm(stdin_buf, c1_len); + size_t c2_len = data_size - c1_len - SEPLEN; + char *c2 = tor_memdup_nulterm(separator + SEPLEN, c2_len); + + char *c3 = consensus_diff_generate(c1, c2); + + if (c3) { + char *c4 = consensus_diff_apply(c1, c3); + tor_assert(c4); + if (strcmp(c2, c4)) { + printf("%s\n", escaped(c1)); + printf("%s\n", escaped(c2)); + printf("%s\n", escaped(c3)); + printf("%s\n", escaped(c4)); + } + tor_assert(! strcmp(c2, c4)); + tor_free(c3); + tor_free(c4); + } + tor_free(c1); + tor_free(c2); + + return 0; +} + diff --git a/src/test/fuzz/fuzz_diff_apply.c b/src/test/fuzz/fuzz_diff_apply.c new file mode 100644 index 0000000000..8d7bf751bf --- /dev/null +++ b/src/test/fuzz/fuzz_diff_apply.c @@ -0,0 +1,65 @@ +/* Copyright (c) 2016, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#define CONSDIFF_PRIVATE + +#include "orconfig.h" +#include "or.h" +#include "consdiff.h" + +#include "fuzzing.h" + +static int +mock_consensus_compute_digest_(const char *c, consensus_digest_t *d) +{ + (void)c; + memset(d->sha3_256, 3, sizeof(d->sha3_256)); + return 0; +} + +static int +mock_consensus_digest_eq_(const uint8_t *a, const uint8_t *b) +{ + (void)a; + (void)b; + return 1; +} + +int +fuzz_init(void) +{ + MOCK(consensus_compute_digest, mock_consensus_compute_digest_); + MOCK(consensus_digest_eq, mock_consensus_digest_eq_); + return 0; +} + +int +fuzz_cleanup(void) +{ + UNMOCK(consensus_compute_digest); + UNMOCK(consensus_digest_eq); + return 0; +} + +int +fuzz_main(const uint8_t *stdin_buf, size_t data_size) +{ +#define SEP "=====\n" +#define SEPLEN strlen(SEP) + const uint8_t *separator = tor_memmem(stdin_buf, data_size, SEP, SEPLEN); + if (! separator) + return 0; + size_t c1_len = separator - stdin_buf; + char *c1 = tor_memdup_nulterm(stdin_buf, c1_len); + size_t c2_len = data_size - c1_len - SEPLEN; + char *c2 = tor_memdup_nulterm(separator + SEPLEN, c2_len); + + char *c3 = consensus_diff_apply(c1, c2); + + tor_free(c1); + tor_free(c2); + tor_free(c3); + + return 0; +} + diff --git a/src/test/fuzz/include.am b/src/test/fuzz/include.am index 806710879b..97afd18215 100644 --- a/src/test/fuzz/include.am +++ b/src/test/fuzz/include.am @@ -48,6 +48,22 @@ src_test_fuzz_fuzz_descriptor_CFLAGS = $(FUZZING_CFLAGS) src_test_fuzz_fuzz_descriptor_LDFLAGS = $(FUZZING_LDFLAG) src_test_fuzz_fuzz_descriptor_LDADD = $(FUZZING_LIBS) +src_test_fuzz_fuzz_diff_SOURCES = \ + src/test/fuzz/fuzzing_common.c \ + src/test/fuzz/fuzz_diff.c +src_test_fuzz_fuzz_diff_CPPFLAGS = $(FUZZING_CPPFLAGS) +src_test_fuzz_fuzz_diff_CFLAGS = $(FUZZING_CFLAGS) +src_test_fuzz_fuzz_diff_LDFLAGS = $(FUZZING_LDFLAG) +src_test_fuzz_fuzz_diff_LDADD = $(FUZZING_LIBS) + +src_test_fuzz_fuzz_diff_apply_SOURCES = \ + src/test/fuzz/fuzzing_common.c \ + src/test/fuzz/fuzz_diff_apply.c +src_test_fuzz_fuzz_diff_apply_CPPFLAGS = $(FUZZING_CPPFLAGS) +src_test_fuzz_fuzz_diff_apply_CFLAGS = $(FUZZING_CFLAGS) +src_test_fuzz_fuzz_diff_apply_LDFLAGS = $(FUZZING_LDFLAG) +src_test_fuzz_fuzz_diff_apply_LDADD = $(FUZZING_LIBS) + src_test_fuzz_fuzz_http_SOURCES = \ src/test/fuzz/fuzzing_common.c \ src/test/fuzz/fuzz_http.c @@ -99,6 +115,8 @@ src_test_fuzz_fuzz_vrs_LDADD = $(FUZZING_LIBS) FUZZERS = \ src/test/fuzz/fuzz-consensus \ src/test/fuzz/fuzz-descriptor \ + src/test/fuzz/fuzz-diff \ + src/test/fuzz/fuzz-diff-apply \ src/test/fuzz/fuzz-extrainfo \ src/test/fuzz/fuzz-http \ src/test/fuzz/fuzz-hsdescv2 \ @@ -106,7 +124,6 @@ FUZZERS = \ src/test/fuzz/fuzz-microdesc \ src/test/fuzz/fuzz-vrs - LIBFUZZER = /home/nickm/build/libfuzz/libFuzzer.a LIBFUZZER_CPPFLAGS = $(FUZZING_CPPFLAGS) -DLLVM_FUZZ LIBFUZZER_CFLAGS = $(FUZZING_CFLAGS) @@ -128,6 +145,20 @@ src_test_fuzz_lf_fuzz_descriptor_CFLAGS = $(LIBFUZZER_CFLAGS) src_test_fuzz_lf_fuzz_descriptor_LDFLAGS = $(LIBFUZZER_LDFLAG) src_test_fuzz_lf_fuzz_descriptor_LDADD = $(LIBFUZZER_LIBS) +src_test_fuzz_lf_fuzz_diff_SOURCES = \ + $(src_test_fuzz_fuzz_diff_SOURCES) +src_test_fuzz_lf_fuzz_diff_CPPFLAGS = $(LIBFUZZER_CPPFLAGS) +src_test_fuzz_lf_fuzz_diff_CFLAGS = $(LIBFUZZER_CFLAGS) +src_test_fuzz_lf_fuzz_diff_LDFLAGS = $(LIBFUZZER_LDFLAG) +src_test_fuzz_lf_fuzz_diff_LDADD = $(LIBFUZZER_LIBS) + +src_test_fuzz_lf_fuzz_diff_apply_SOURCES = \ + $(src_test_fuzz_fuzz_diff_apply_SOURCES) +src_test_fuzz_lf_fuzz_diff_apply_CPPFLAGS = $(LIBFUZZER_CPPFLAGS) +src_test_fuzz_lf_fuzz_diff_apply_CFLAGS = $(LIBFUZZER_CFLAGS) +src_test_fuzz_lf_fuzz_diff_apply_LDFLAGS = $(LIBFUZZER_LDFLAG) +src_test_fuzz_lf_fuzz_diff_apply_LDADD = $(LIBFUZZER_LIBS) + src_test_fuzz_lf_fuzz_extrainfo_SOURCES = \ $(src_test_fuzz_fuzz_extrainfo_SOURCES) src_test_fuzz_lf_fuzz_extrainfo_CPPFLAGS = $(LIBFUZZER_CPPFLAGS) @@ -172,6 +203,8 @@ src_test_fuzz_lf_fuzz_vrs_LDADD = $(LIBFUZZER_LIBS) LIBFUZZER_FUZZERS = \ src/test/fuzz/lf-fuzz-consensus \ + src/test/fuzz/lf-fuzz-diff \ + src/test/fuzz/lf-fuzz-diff-apply \ src/test/fuzz/lf-fuzz-descriptor \ src/test/fuzz/lf-fuzz-extrainfo \ src/test/fuzz/lf-fuzz-http \ @@ -198,6 +231,16 @@ src_test_fuzz_liboss_fuzz_descriptor_a_SOURCES = \ src_test_fuzz_liboss_fuzz_descriptor_a_CPPFLAGS = $(LIBOSS_FUZZ_CPPFLAGS) src_test_fuzz_liboss_fuzz_descriptor_a_CFLAGS = $(LIBOSS_FUZZ_CFLAGS) +src_test_fuzz_liboss_fuzz_diff_a_SOURCES = \ + $(src_test_fuzz_fuzz_diff_SOURCES) +src_test_fuzz_liboss_fuzz_diff_a_CPPFLAGS = $(LIBOSS_FUZZ_CPPFLAGS) +src_test_fuzz_liboss_fuzz_diff_a_CFLAGS = $(LIBOSS_FUZZ_CFLAGS) + +src_test_fuzz_liboss_fuzz_diff_apply_a_SOURCES = \ + $(src_test_fuzz_fuzz_diff_apply_SOURCES) +src_test_fuzz_liboss_fuzz_diff_apply_a_CPPFLAGS = $(LIBOSS_FUZZ_CPPFLAGS) +src_test_fuzz_liboss_fuzz_diff_apply_a_CFLAGS = $(LIBOSS_FUZZ_CFLAGS) + src_test_fuzz_liboss_fuzz_extrainfo_a_SOURCES = \ $(src_test_fuzz_fuzz_extrainfo_SOURCES) src_test_fuzz_liboss_fuzz_extrainfo_a_CPPFLAGS = $(LIBOSS_FUZZ_CPPFLAGS) @@ -231,6 +274,8 @@ src_test_fuzz_liboss_fuzz_vrs_a_CFLAGS = $(LIBOSS_FUZZ_CFLAGS) OSS_FUZZ_FUZZERS = \ src/test/fuzz/liboss-fuzz-consensus.a \ src/test/fuzz/liboss-fuzz-descriptor.a \ + src/test/fuzz/liboss-fuzz-diff.a \ + src/test/fuzz/liboss-fuzz-diff-apply.a \ src/test/fuzz/liboss-fuzz-extrainfo.a \ src/test/fuzz/liboss-fuzz-http.a \ src/test/fuzz/liboss-fuzz-hsdescv2.a \ diff --git a/src/test/include.am b/src/test/include.am index 1c0726fd3a..23be3e8334 100644 --- a/src/test/include.am +++ b/src/test/include.am @@ -86,6 +86,7 @@ src_test_test_SOURCES = \ src/test/test_compat_libevent.c \ src/test/test_config.c \ src/test/test_connection.c \ + src/test/test_consdiff.c \ src/test/test_containers.c \ src/test/test_controller.c \ src/test/test_controller_events.c \ diff --git a/src/test/test.c b/src/test/test.c index 1f91405a61..0039818cb8 100644 --- a/src/test/test.c +++ b/src/test/test.c @@ -1194,6 +1194,7 @@ struct testgroup_t testgroups[] = { { "compat/libevent/", compat_libevent_tests }, { "config/", config_tests }, { "connection/", connection_tests }, + { "consdiff/", consdiff_tests }, { "container/", container_tests }, { "control/", controller_tests }, { "control/event/", controller_event_tests }, diff --git a/src/test/test.h b/src/test/test.h index c78e272178..0cf56ab74a 100644 --- a/src/test/test.h +++ b/src/test/test.h @@ -189,6 +189,7 @@ extern struct testcase_t circuituse_tests[]; extern struct testcase_t compat_libevent_tests[]; extern struct testcase_t config_tests[]; extern struct testcase_t connection_tests[]; +extern struct testcase_t consdiff_tests[]; extern struct testcase_t container_tests[]; extern struct testcase_t controller_tests[]; extern struct testcase_t controller_event_tests[]; diff --git a/src/test/test_consdiff.c b/src/test/test_consdiff.c new file mode 100644 index 0000000000..1b4e2ad3c1 --- /dev/null +++ b/src/test/test_consdiff.c @@ -0,0 +1,1097 @@ +/* Copyright (c) 2014, Daniel Martà + * Copyright (c) 2014, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#define CONSDIFF_PRIVATE + +#include "or.h" +#include "test.h" + +#include "consdiff.h" +#include "memarea.h" +#include "log_test_helpers.h" + +#define tt_str_eq_line(a,b) \ + tt_assert(line_str_eq((b),(a))) + +static void +test_consdiff_smartlist_slice(void *arg) +{ + smartlist_t *sl = smartlist_new(); + smartlist_slice_t *sls; + + /* Create a regular smartlist. */ + (void)arg; + smartlist_add(sl, (void*)1); + smartlist_add(sl, (void*)2); + smartlist_add(sl, (void*)3); + smartlist_add(sl, (void*)4); + smartlist_add(sl, (void*)5); + + /* See if the slice was done correctly. */ + sls = smartlist_slice(sl, 2, 5); + tt_ptr_op(sl, OP_EQ, sls->list); + tt_ptr_op((void*)3, OP_EQ, smartlist_get(sls->list, sls->offset)); + tt_ptr_op((void*)5, OP_EQ, + smartlist_get(sls->list, sls->offset + (sls->len-1))); + tor_free(sls); + + /* See that using -1 as the end does get to the last element. */ + sls = smartlist_slice(sl, 2, -1); + tt_ptr_op(sl, OP_EQ, sls->list); + tt_ptr_op((void*)3, OP_EQ, smartlist_get(sls->list, sls->offset)); + tt_ptr_op((void*)5, OP_EQ, + smartlist_get(sls->list, sls->offset + (sls->len-1))); + + done: + tor_free(sls); + smartlist_free(sl); +} + +static void +test_consdiff_smartlist_slice_string_pos(void *arg) +{ + smartlist_t *sl = smartlist_new(); + smartlist_slice_t *sls; + memarea_t *area = memarea_new(); + + /* Create a regular smartlist. */ + (void)arg; + consensus_split_lines(sl, "a\nd\nc\na\nb\n", area); + + /* See that smartlist_slice_string_pos respects the bounds of the slice. */ + sls = smartlist_slice(sl, 2, 5); + cdline_t a_line = { "a", 1 }; + tt_int_op(3, OP_EQ, smartlist_slice_string_pos(sls, &a_line)); + cdline_t d_line = { "d", 1 }; + tt_int_op(-1, OP_EQ, smartlist_slice_string_pos(sls, &d_line)); + + done: + tor_free(sls); + smartlist_free(sl); + memarea_drop_all(area); +} + +static void +test_consdiff_lcs_lengths(void *arg) +{ + smartlist_t *sl1 = smartlist_new(); + smartlist_t *sl2 = smartlist_new(); + smartlist_slice_t *sls1, *sls2; + int *lengths1, *lengths2; + memarea_t *area = memarea_new(); + + /* Expected lcs lengths in regular and reverse order. */ + int e_lengths1[] = { 0, 1, 2, 3, 3, 4 }; + int e_lengths2[] = { 0, 1, 1, 2, 3, 4 }; + + (void)arg; + consensus_split_lines(sl1, "a\nb\nc\nd\ne\n", area); + consensus_split_lines(sl2, "a\nc\nd\ni\ne\n", area); + + sls1 = smartlist_slice(sl1, 0, -1); + sls2 = smartlist_slice(sl2, 0, -1); + + lengths1 = lcs_lengths(sls1, sls2, 1); + lengths2 = lcs_lengths(sls1, sls2, -1); + tt_mem_op(e_lengths1, OP_EQ, lengths1, sizeof(int) * 6); + tt_mem_op(e_lengths2, OP_EQ, lengths2, sizeof(int) * 6); + + done: + tor_free(lengths1); + tor_free(lengths2); + tor_free(sls1); + tor_free(sls2); + smartlist_free(sl1); + smartlist_free(sl2); + memarea_drop_all(area); +} + +static void +test_consdiff_trim_slices(void *arg) +{ + smartlist_t *sl1 = smartlist_new(); + smartlist_t *sl2 = smartlist_new(); + smartlist_t *sl3 = smartlist_new(); + smartlist_t *sl4 = smartlist_new(); + smartlist_slice_t *sls1, *sls2, *sls3, *sls4; + memarea_t *area = memarea_new(); + + (void)arg; + consensus_split_lines(sl1, "a\nb\nb\nb\nd\n", area); + consensus_split_lines(sl2, "a\nc\nc\nc\nd\n", area); + consensus_split_lines(sl3, "a\nb\nb\nb\na\n", area); + consensus_split_lines(sl4, "c\nb\nb\nb\nc\n", area); + sls1 = smartlist_slice(sl1, 0, -1); + sls2 = smartlist_slice(sl2, 0, -1); + sls3 = smartlist_slice(sl3, 0, -1); + sls4 = smartlist_slice(sl4, 0, -1); + + /* They should be trimmed by one line at each end. */ + tt_int_op(5, OP_EQ, sls1->len); + tt_int_op(5, OP_EQ, sls2->len); + trim_slices(sls1, sls2); + tt_int_op(3, OP_EQ, sls1->len); + tt_int_op(3, OP_EQ, sls2->len); + + /* They should not be trimmed at all. */ + tt_int_op(5, OP_EQ, sls3->len); + tt_int_op(5, OP_EQ, sls4->len); + trim_slices(sls3, sls4); + tt_int_op(5, OP_EQ, sls3->len); + tt_int_op(5, OP_EQ, sls4->len); + + done: + tor_free(sls1); + tor_free(sls2); + tor_free(sls3); + tor_free(sls4); + smartlist_free(sl1); + smartlist_free(sl2); + smartlist_free(sl3); + smartlist_free(sl4); + memarea_drop_all(area); +} + +static void +test_consdiff_set_changed(void *arg) +{ + smartlist_t *sl1 = smartlist_new(); + smartlist_t *sl2 = smartlist_new(); + bitarray_t *changed1 = bitarray_init_zero(4); + bitarray_t *changed2 = bitarray_init_zero(4); + smartlist_slice_t *sls1, *sls2; + memarea_t *area = memarea_new(); + + (void)arg; + consensus_split_lines(sl1, "a\nb\na\na\n", area); + consensus_split_lines(sl2, "a\na\na\na\n", area); + + /* Length of sls1 is 0. */ + sls1 = smartlist_slice(sl1, 0, 0); + sls2 = smartlist_slice(sl2, 1, 3); + set_changed(changed1, changed2, sls1, sls2); + + /* The former is not changed, the latter changes all of its elements. */ + tt_assert(!bitarray_is_set(changed1, 0)); + tt_assert(!bitarray_is_set(changed1, 1)); + tt_assert(!bitarray_is_set(changed1, 2)); + tt_assert(!bitarray_is_set(changed1, 3)); + + tt_assert(!bitarray_is_set(changed2, 0)); + tt_assert(bitarray_is_set(changed2, 1)); + tt_assert(bitarray_is_set(changed2, 2)); + tt_assert(!bitarray_is_set(changed2, 3)); + bitarray_clear(changed2, 1); + bitarray_clear(changed2, 2); + + /* Length of sls1 is 1 and its element is in sls2. */ + tor_free(sls1); + sls1 = smartlist_slice(sl1, 0, 1); + set_changed(changed1, changed2, sls1, sls2); + + /* The latter changes all elements but the (first) common one. */ + tt_assert(!bitarray_is_set(changed1, 0)); + tt_assert(!bitarray_is_set(changed1, 1)); + tt_assert(!bitarray_is_set(changed1, 2)); + tt_assert(!bitarray_is_set(changed1, 3)); + + tt_assert(!bitarray_is_set(changed2, 0)); + tt_assert(!bitarray_is_set(changed2, 1)); + tt_assert(bitarray_is_set(changed2, 2)); + tt_assert(!bitarray_is_set(changed2, 3)); + bitarray_clear(changed2, 2); + + /* Length of sls1 is 1 and its element is not in sls2. */ + tor_free(sls1); + sls1 = smartlist_slice(sl1, 1, 2); + set_changed(changed1, changed2, sls1, sls2); + + /* The former changes its element, the latter changes all elements. */ + tt_assert(!bitarray_is_set(changed1, 0)); + tt_assert(bitarray_is_set(changed1, 1)); + tt_assert(!bitarray_is_set(changed1, 2)); + tt_assert(!bitarray_is_set(changed1, 3)); + + tt_assert(!bitarray_is_set(changed2, 0)); + tt_assert(bitarray_is_set(changed2, 1)); + tt_assert(bitarray_is_set(changed2, 2)); + tt_assert(!bitarray_is_set(changed2, 3)); + + done: + bitarray_free(changed1); + bitarray_free(changed2); + smartlist_free(sl1); + smartlist_free(sl2); + tor_free(sls1); + tor_free(sls2); + memarea_drop_all(area); +} + +static void +test_consdiff_calc_changes(void *arg) +{ + smartlist_t *sl1 = smartlist_new(); + smartlist_t *sl2 = smartlist_new(); + smartlist_slice_t *sls1, *sls2; + bitarray_t *changed1 = bitarray_init_zero(4); + bitarray_t *changed2 = bitarray_init_zero(4); + memarea_t *area = memarea_new(); + + (void)arg; + consensus_split_lines(sl1, "a\na\na\na\n", area); + consensus_split_lines(sl2, "a\na\na\na\n", area); + + sls1 = smartlist_slice(sl1, 0, -1); + sls2 = smartlist_slice(sl2, 0, -1); + calc_changes(sls1, sls2, changed1, changed2); + + /* Nothing should be set to changed. */ + tt_assert(!bitarray_is_set(changed1, 0)); + tt_assert(!bitarray_is_set(changed1, 1)); + tt_assert(!bitarray_is_set(changed1, 2)); + tt_assert(!bitarray_is_set(changed1, 3)); + + tt_assert(!bitarray_is_set(changed2, 0)); + tt_assert(!bitarray_is_set(changed2, 1)); + tt_assert(!bitarray_is_set(changed2, 2)); + tt_assert(!bitarray_is_set(changed2, 3)); + + smartlist_clear(sl2); + consensus_split_lines(sl2, "a\nb\na\nb\n", area); + tor_free(sls1); + tor_free(sls2); + sls1 = smartlist_slice(sl1, 0, -1); + sls2 = smartlist_slice(sl2, 0, -1); + calc_changes(sls1, sls2, changed1, changed2); + + /* Two elements are changed. */ + tt_assert(!bitarray_is_set(changed1, 0)); + tt_assert(bitarray_is_set(changed1, 1)); + tt_assert(bitarray_is_set(changed1, 2)); + tt_assert(!bitarray_is_set(changed1, 3)); + bitarray_clear(changed1, 1); + bitarray_clear(changed1, 2); + + tt_assert(!bitarray_is_set(changed2, 0)); + tt_assert(bitarray_is_set(changed2, 1)); + tt_assert(!bitarray_is_set(changed2, 2)); + tt_assert(bitarray_is_set(changed2, 3)); + bitarray_clear(changed1, 1); + bitarray_clear(changed1, 3); + + smartlist_clear(sl2); + consensus_split_lines(sl2, "b\nb\nb\nb\n", area); + tor_free(sls1); + tor_free(sls2); + sls1 = smartlist_slice(sl1, 0, -1); + sls2 = smartlist_slice(sl2, 0, -1); + calc_changes(sls1, sls2, changed1, changed2); + + /* All elements are changed. */ + tt_assert(bitarray_is_set(changed1, 0)); + tt_assert(bitarray_is_set(changed1, 1)); + tt_assert(bitarray_is_set(changed1, 2)); + tt_assert(bitarray_is_set(changed1, 3)); + + tt_assert(bitarray_is_set(changed2, 0)); + tt_assert(bitarray_is_set(changed2, 1)); + tt_assert(bitarray_is_set(changed2, 2)); + tt_assert(bitarray_is_set(changed2, 3)); + + done: + bitarray_free(changed1); + bitarray_free(changed2); + smartlist_free(sl1); + smartlist_free(sl2); + tor_free(sls1); + tor_free(sls2); + memarea_drop_all(area); +} + +static void +test_consdiff_get_id_hash(void *arg) +{ + (void)arg; + + cdline_t line1 = { "r name", 6 }; + cdline_t line2 = { "r name _hash_isnt_base64 etc", 28 }; + cdline_t line3 = { "r name hash+valid+base64 etc", 28 }; + cdline_t tmp; + + /* No hash. */ + tt_int_op(-1, OP_EQ, get_id_hash(&line1, &tmp)); + /* The hash contains characters that are not base64. */ + tt_int_op(-1, OP_EQ, get_id_hash(&line2, &tmp)); + + /* valid hash. */ + tt_int_op(0, OP_EQ, get_id_hash(&line3, &tmp)); + tt_ptr_op(tmp.s, OP_EQ, line3.s + 7); + tt_uint_op(tmp.len, OP_EQ, line3.len - 11); + + done: + ; +} + +static void +test_consdiff_is_valid_router_entry(void *arg) +{ + /* Doesn't start with "r ". */ + (void)arg; + cdline_t line0 = { "foo", 3 }; + tt_int_op(0, OP_EQ, is_valid_router_entry(&line0)); + + /* These are already tested with get_id_hash, but make sure it's run + * properly. */ + + cdline_t line1 = { "r name", 6 }; + cdline_t line2 = { "r name _hash_isnt_base64 etc", 28 }; + cdline_t line3 = { "r name hash+valid+base64 etc", 28 }; + tt_int_op(0, OP_EQ, is_valid_router_entry(&line1)); + tt_int_op(0, OP_EQ, is_valid_router_entry(&line2)); + tt_int_op(1, OP_EQ, is_valid_router_entry(&line3)); + + done: + ; +} + +static void +test_consdiff_next_router(void *arg) +{ + smartlist_t *sl = smartlist_new(); + memarea_t *area = memarea_new(); + (void)arg; + smartlist_add_linecpy(sl, area, "foo"); + smartlist_add_linecpy(sl, area, + "r name hash+longer+than+27+chars+and+valid+base64 etc"); + smartlist_add_linecpy(sl, area, "foo"); + smartlist_add_linecpy(sl, area, "foo"); + smartlist_add_linecpy(sl, area, + "r name hash+longer+than+27+chars+and+valid+base64 etc"); + smartlist_add_linecpy(sl, area, "foo"); + + /* Not currently on a router entry line, finding the next one. */ + tt_int_op(1, OP_EQ, next_router(sl, 0)); + tt_int_op(4, OP_EQ, next_router(sl, 2)); + + /* Already at the beginning of a router entry line, ignore it. */ + tt_int_op(4, OP_EQ, next_router(sl, 1)); + + /* There are no more router entries, so return the line after the last. */ + tt_int_op(6, OP_EQ, next_router(sl, 4)); + tt_int_op(6, OP_EQ, next_router(sl, 5)); + + done: + smartlist_free(sl); + memarea_drop_all(area); +} + +static int +base64cmp_wrapper(const char *a, const char *b) +{ + cdline_t aa = { a, a ? (uint32_t) strlen(a) : 0 }; + cdline_t bb = { b, b ? (uint32_t) strlen(b) : 0 }; + return base64cmp(&aa, &bb); +} + +static void +test_consdiff_base64cmp(void *arg) +{ + /* NULL arguments. */ + (void)arg; + tt_int_op(0, OP_EQ, base64cmp_wrapper(NULL, NULL)); + tt_int_op(-1, OP_EQ, base64cmp_wrapper(NULL, "foo")); + tt_int_op(1, OP_EQ, base64cmp_wrapper("bar", NULL)); + + /* Nil base64 values. */ + tt_int_op(0, OP_EQ, base64cmp_wrapper("", "")); + tt_int_op(0, OP_EQ, base64cmp_wrapper("_", "&")); + + /* Exact same valid strings. */ + tt_int_op(0, OP_EQ, base64cmp_wrapper("abcABC/+", "abcABC/+")); + /* Both end with an invalid base64 char other than '\0'. */ + tt_int_op(0, OP_EQ, base64cmp_wrapper("abcABC/+ ", "abcABC/+ ")); + /* Only one ends with an invalid base64 char other than '\0'. */ + tt_int_op(-1, OP_EQ, base64cmp_wrapper("abcABC/+ ", "abcABC/+a")); + + /* Comparisons that would return differently with strcmp(). */ + tt_int_op(-1, OP_EQ, strcmp("/foo", "Afoo")); + tt_int_op(1, OP_EQ, base64cmp_wrapper("/foo", "Afoo")); + tt_int_op(1, OP_EQ, strcmp("Afoo", "0foo")); + tt_int_op(-1, OP_EQ, base64cmp_wrapper("Afoo", "0foo")); + + /* Comparisons that would return the same as with strcmp(). */ + tt_int_op(1, OP_EQ, strcmp("afoo", "Afoo")); + tt_int_op(1, OP_EQ, base64cmp_wrapper("afoo", "Afoo")); + + /* Different lengths */ + tt_int_op(-1, OP_EQ, base64cmp_wrapper("afoo", "afooo")); + tt_int_op(1, OP_EQ, base64cmp_wrapper("afooo", "afoo")); + + done: + ; +} + +static void +test_consdiff_gen_ed_diff(void *arg) +{ + smartlist_t *cons1=NULL, *cons2=NULL, *diff=NULL; + int i; + memarea_t *area = memarea_new(); + setup_capture_of_logs(LOG_WARN); + + (void)arg; + cons1 = smartlist_new(); + cons2 = smartlist_new(); + + /* Identity hashes are not sorted properly, return NULL. */ + smartlist_add_linecpy(cons1, area, "r name bbbbbbbbbbbbbbbbbbbbbbbbbbb etc"); + smartlist_add_linecpy(cons1, area, "foo"); + smartlist_add_linecpy(cons1, area, "r name aaaaaaaaaaaaaaaaaaaaaaaaaaa etc"); + smartlist_add_linecpy(cons1, area, "bar"); + + smartlist_add_linecpy(cons2, area, "r name aaaaaaaaaaaaaaaaaaaaaaaaaaa etc"); + smartlist_add_linecpy(cons2, area, "foo"); + smartlist_add_linecpy(cons2, area, "r name ccccccccccccccccccccccccccc etc"); + smartlist_add_linecpy(cons2, area, "bar"); + + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because the base consensus doesn't have its router entries sorted " + "properly."); + + /* Same, but now with the second consensus. */ + mock_clean_saved_logs(); + diff = gen_ed_diff(cons2, cons1, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because the target consensus doesn't have its router entries sorted " + "properly."); + + /* Same as the two above, but with the reversed thing immediately after a + match. (The code handles this differently) */ + smartlist_del(cons1, 0); + smartlist_add_linecpy(cons1, area, "r name aaaaaaaaaaaaaaaaaaaaaaaaaaa etc"); + + mock_clean_saved_logs(); + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because the base consensus doesn't have its router entries sorted " + "properly."); + + mock_clean_saved_logs(); + diff = gen_ed_diff(cons2, cons1, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because the target consensus doesn't have its router entries sorted " + "properly."); + + /* Identity hashes are repeated, return NULL. */ + smartlist_clear(cons1); + + smartlist_add_linecpy(cons1, area, "r name bbbbbbbbbbbbbbbbbbbbbbbbbbb etc"); + smartlist_add_linecpy(cons1, area, "foo"); + smartlist_add_linecpy(cons1, area, "r name bbbbbbbbbbbbbbbbbbbbbbbbbbb etc"); + smartlist_add_linecpy(cons1, area, "bar"); + + mock_clean_saved_logs(); + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because the base consensus doesn't have its router entries sorted " + "properly."); + + /* We have to add a line that is just a dot, return NULL. */ + smartlist_clear(cons1); + smartlist_clear(cons2); + + smartlist_add_linecpy(cons1, area, "foo1"); + smartlist_add_linecpy(cons1, area, "foo2"); + + smartlist_add_linecpy(cons2, area, "foo1"); + smartlist_add_linecpy(cons2, area, "."); + smartlist_add_linecpy(cons2, area, "foo2"); + + mock_clean_saved_logs(); + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Cannot generate consensus diff " + "because one of the lines to be added is \".\"."); + +#define MAX_LINE_COUNT (10000) + /* Too many lines to be fed to the quadratic-time function. */ + smartlist_clear(cons1); + smartlist_clear(cons2); + + for (i=0; i < MAX_LINE_COUNT; ++i) smartlist_add_linecpy(cons1, area, "a"); + for (i=0; i < MAX_LINE_COUNT; ++i) smartlist_add_linecpy(cons1, area, "b"); + + mock_clean_saved_logs(); + diff = gen_ed_diff(cons1, cons2, area); + + tt_ptr_op(NULL, OP_EQ, diff); + expect_single_log_msg_containing("Refusing to generate consensus diff " + "because we found too few common router ids."); + + /* We have dot lines, but they don't interfere with the script format. */ + smartlist_clear(cons1); + smartlist_clear(cons2); + + smartlist_add_linecpy(cons1, area, "foo1"); + smartlist_add_linecpy(cons1, area, "."); + smartlist_add_linecpy(cons1, area, "."); + smartlist_add_linecpy(cons1, area, "foo2"); + + smartlist_add_linecpy(cons2, area, "foo1"); + smartlist_add_linecpy(cons2, area, "."); + smartlist_add_linecpy(cons2, area, "foo2"); + + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + smartlist_free(diff); + + /* Empty diff tests. */ + smartlist_clear(cons1); + smartlist_clear(cons2); + + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(0, OP_EQ, smartlist_len(diff)); + smartlist_free(diff); + + smartlist_add_linecpy(cons1, area, "foo"); + smartlist_add_linecpy(cons1, area, "bar"); + + smartlist_add_linecpy(cons2, area, "foo"); + smartlist_add_linecpy(cons2, area, "bar"); + + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(0, OP_EQ, smartlist_len(diff)); + smartlist_free(diff); + + /* Everything is deleted. */ + smartlist_clear(cons2); + + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(1, OP_EQ, smartlist_len(diff)); + tt_str_eq_line("1,2d", smartlist_get(diff, 0)); + + smartlist_free(diff); + + /* Everything is added. */ + diff = gen_ed_diff(cons2, cons1, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(4, OP_EQ, smartlist_len(diff)); + tt_str_eq_line("0a", smartlist_get(diff, 0)); + tt_str_eq_line("foo", smartlist_get(diff, 1)); + tt_str_eq_line("bar", smartlist_get(diff, 2)); + tt_str_eq_line(".", smartlist_get(diff, 3)); + + smartlist_free(diff); + + /* Everything is changed. */ + smartlist_add_linecpy(cons2, area, "foo2"); + smartlist_add_linecpy(cons2, area, "bar2"); + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(4, OP_EQ, smartlist_len(diff)); + tt_str_eq_line("1,2c", smartlist_get(diff, 0)); + tt_str_eq_line("foo2", smartlist_get(diff, 1)); + tt_str_eq_line("bar2", smartlist_get(diff, 2)); + tt_str_eq_line(".", smartlist_get(diff, 3)); + + smartlist_free(diff); + + /* Test 'a', 'c' and 'd' together. See that it is done in reverse order. */ + smartlist_clear(cons1); + smartlist_clear(cons2); + consensus_split_lines(cons1, "A\nB\nC\nD\nE\n", area); + consensus_split_lines(cons2, "A\nC\nO\nE\nU\n", area); + diff = gen_ed_diff(cons1, cons2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(7, OP_EQ, smartlist_len(diff)); + tt_str_eq_line("5a", smartlist_get(diff, 0)); + tt_str_eq_line("U", smartlist_get(diff, 1)); + tt_str_eq_line(".", smartlist_get(diff, 2)); + tt_str_eq_line("4c", smartlist_get(diff, 3)); + tt_str_eq_line("O", smartlist_get(diff, 4)); + tt_str_eq_line(".", smartlist_get(diff, 5)); + tt_str_eq_line("2d", smartlist_get(diff, 6)); + + /* TODO: small real use-cases, i.e. consensuses. */ + + done: + teardown_capture_of_logs(); + smartlist_free(cons1); + smartlist_free(cons2); + smartlist_free(diff); + memarea_drop_all(area); +} + +static void +test_consdiff_apply_ed_diff(void *arg) +{ + smartlist_t *cons1=NULL, *cons2=NULL, *diff=NULL; + memarea_t *area = memarea_new(); + (void)arg; + cons1 = smartlist_new(); + diff = smartlist_new(); + setup_capture_of_logs(LOG_WARN); + + consensus_split_lines(cons1, "A\nB\nC\nD\nE\n", area); + + /* Command without range. */ + smartlist_add_linecpy(diff, area, "a"); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + smartlist_clear(diff); + expect_single_log_msg_containing("an ed command was missing a line number"); + + /* Range without command. */ + smartlist_add_linecpy(diff, area, "1"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("a line with no ed command was found"); + + smartlist_clear(diff); + + /* Range without end. */ + smartlist_add_linecpy(diff, area, "1,"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("an ed command was missing a range " + "end line number."); + + smartlist_clear(diff); + + /* Incoherent ranges. */ + smartlist_add_linecpy(diff, area, "1,1"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("an invalid range was found"); + + smartlist_clear(diff); + + smartlist_add_linecpy(diff, area, "3,2"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("an invalid range was found"); + + smartlist_clear(diff); + + /* Script is not in reverse order. */ + smartlist_add_linecpy(diff, area, "1d"); + smartlist_add_linecpy(diff, area, "3d"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("its commands are not properly sorted"); + + smartlist_clear(diff); + + /* Script contains unrecognised commands longer than one char. */ + smartlist_add_linecpy(diff, area, "1foo"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("an ed command longer than one char was " + "found"); + + smartlist_clear(diff); + + /* Script contains unrecognised commands. */ + smartlist_add_linecpy(diff, area, "1e"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("an unrecognised ed command was found"); + + smartlist_clear(diff); + + /* Command that should be followed by at least one line and a ".", but + * isn't. */ + smartlist_add_linecpy(diff, area, "0a"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("it has an ed command that tries to " + "insert zero lines."); + + /* Now it is followed by a ".", but it inserts zero lines. */ + smartlist_add_linecpy(diff, area, "."); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("it has an ed command that tries to " + "insert zero lines."); + + smartlist_clear(diff); + + /* Now it it inserts something, but has no terminator. */ + smartlist_add_linecpy(diff, area, "0a"); + smartlist_add_linecpy(diff, area, "hello"); + mock_clean_saved_logs(); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("lines to be inserted that don't end with " + "a \".\"."); + + smartlist_clear(diff); + + /* Test appending text, 'a'. */ + consensus_split_lines(diff, "3a\nU\nO\n.\n0a\nV\n.\n", area); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_NE, cons2); + tt_int_op(8, OP_EQ, smartlist_len(cons2)); + tt_str_eq_line("V", smartlist_get(cons2, 0)); + tt_str_eq_line("A", smartlist_get(cons2, 1)); + tt_str_eq_line("B", smartlist_get(cons2, 2)); + tt_str_eq_line("C", smartlist_get(cons2, 3)); + tt_str_eq_line("U", smartlist_get(cons2, 4)); + tt_str_eq_line("O", smartlist_get(cons2, 5)); + tt_str_eq_line("D", smartlist_get(cons2, 6)); + tt_str_eq_line("E", smartlist_get(cons2, 7)); + + smartlist_clear(diff); + smartlist_free(cons2); + + /* Test deleting text, 'd'. */ + consensus_split_lines(diff, "4d\n1,2d\n", area); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_NE, cons2); + tt_int_op(2, OP_EQ, smartlist_len(cons2)); + tt_str_eq_line("C", smartlist_get(cons2, 0)); + tt_str_eq_line("E", smartlist_get(cons2, 1)); + + smartlist_clear(diff); + smartlist_free(cons2); + + /* Test changing text, 'c'. */ + consensus_split_lines(diff, "4c\nT\nX\n.\n1, 2c\nM\n.\n", area); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_NE, cons2); + tt_int_op(5, OP_EQ, smartlist_len(cons2)); + tt_str_eq_line("M", smartlist_get(cons2, 0)); + tt_str_eq_line("C", smartlist_get(cons2, 1)); + tt_str_eq_line("T", smartlist_get(cons2, 2)); + tt_str_eq_line("X", smartlist_get(cons2, 3)); + tt_str_eq_line("E", smartlist_get(cons2, 4)); + + smartlist_clear(diff); + smartlist_free(cons2); + + /* Test 'a', 'd' and 'c' together. */ + consensus_split_lines(diff, "4c\nT\nX\n.\n2d\n0a\nM\n.\n", area); + cons2 = apply_ed_diff(cons1, diff, 0); + tt_ptr_op(NULL, OP_NE, cons2); + tt_int_op(6, OP_EQ, smartlist_len(cons2)); + tt_str_eq_line("M", smartlist_get(cons2, 0)); + tt_str_eq_line("A", smartlist_get(cons2, 1)); + tt_str_eq_line("C", smartlist_get(cons2, 2)); + tt_str_eq_line("T", smartlist_get(cons2, 3)); + tt_str_eq_line("X", smartlist_get(cons2, 4)); + tt_str_eq_line("E", smartlist_get(cons2, 5)); + + done: + teardown_capture_of_logs(); + smartlist_free(cons1); + smartlist_free(cons2); + smartlist_free(diff); + memarea_drop_all(area); +} + +static void +test_consdiff_gen_diff(void *arg) +{ + char *cons1_str=NULL, *cons2_str=NULL; + smartlist_t *cons1=NULL, *cons2=NULL, *diff=NULL; + consensus_digest_t digests1, digests2; + memarea_t *area = memarea_new(); + (void)arg; + cons1 = smartlist_new(); + cons2 = smartlist_new(); + + /* Identity hashes are not sorted properly, return NULL. + * Already tested in gen_ed_diff, but see that a NULL ed diff also makes + * gen_diff return NULL. */ + cons1_str = tor_strdup( + "network-status-version foo\n" + "r name bbbbbbbbbbbbbbbbb etc\nfoo\n" + "r name aaaaaaaaaaaaaaaaa etc\nbar\n" + "directory-signature foo bar\nbar\n" + ); + cons2_str = tor_strdup( + "network-status-version foo\n" + "r name aaaaaaaaaaaaaaaaa etc\nfoo\n" + "r name ccccccccccccccccc etc\nbar\n" + "directory-signature foo bar\nbar\n" + ); + + tt_int_op(0, OP_EQ, + consensus_compute_digest(cons1_str, &digests1)); + tt_int_op(0, OP_EQ, + consensus_compute_digest(cons2_str, &digests2)); + + consensus_split_lines(cons1, cons1_str, area); + consensus_split_lines(cons2, cons2_str, area); + + diff = consdiff_gen_diff(cons1, cons2, &digests1, &digests2, area); + tt_ptr_op(NULL, OP_EQ, diff); + + /* Check that the headers are done properly. */ + tor_free(cons1_str); + cons1_str = tor_strdup( + "network-status-version foo\n" + "r name ccccccccccccccccc etc\nfoo\n" + "r name eeeeeeeeeeeeeeeee etc\nbar\n" + "directory-signature foo bar\nbar\n" + ); + tt_int_op(0, OP_EQ, + consensus_compute_digest(cons1_str, &digests1)); + smartlist_clear(cons1); + consensus_split_lines(cons1, cons1_str, area); + diff = consdiff_gen_diff(cons1, cons2, &digests1, &digests2, area); + tt_ptr_op(NULL, OP_NE, diff); + tt_int_op(7, OP_EQ, smartlist_len(diff)); + tt_assert(line_str_eq(smartlist_get(diff, 0), + "network-status-diff-version 1")); + tt_assert(line_str_eq(smartlist_get(diff, 1), "hash " + "06646D6CF563A41869D3B02E73254372AE3140046C5E7D83C9F71E54976AF9B4 " + "7AFECEFA4599BA33D603653E3D2368F648DF4AC4723929B0F7CF39281596B0C1")); + tt_assert(line_str_eq(smartlist_get(diff, 2), "3,4d")); + tt_assert(line_str_eq(smartlist_get(diff, 3), "1a")); + tt_assert(line_str_eq(smartlist_get(diff, 4), + "r name aaaaaaaaaaaaaaaaa etc")); + tt_assert(line_str_eq(smartlist_get(diff, 5), "foo")); + tt_assert(line_str_eq(smartlist_get(diff, 6), ".")); + + /* TODO: small real use-cases, i.e. consensuses. */ + + done: + tor_free(cons1_str); + tor_free(cons2_str); + smartlist_free(cons1); + smartlist_free(cons2); + smartlist_free(diff); + memarea_drop_all(area); +} + +static void +test_consdiff_apply_diff(void *arg) +{ + smartlist_t *cons1=NULL, *diff=NULL; + char *cons1_str=NULL, *cons2 = NULL; + consensus_digest_t digests1; + (void)arg; + memarea_t *area = memarea_new(); + cons1 = smartlist_new(); + diff = smartlist_new(); + setup_capture_of_logs(LOG_INFO); + + cons1_str = tor_strdup( + "network-status-version foo\n" + "r name ccccccccccccccccc etc\nfoo\n" + "r name eeeeeeeeeeeeeeeee etc\nbar\n" + "directory-signature foo bar\nbar\n" + ); + tt_int_op(0, OP_EQ, + consensus_compute_digest(cons1_str, &digests1)); + consensus_split_lines(cons1, cons1_str, area); + + /* diff doesn't have enough lines. */ + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("too short") + + /* first line doesn't match format-version string. */ + smartlist_add_linecpy(diff, area, "foo-bar"); + smartlist_add_linecpy(diff, area, "header-line"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("format is not known") + + /* The first word of the second header line is not "hash". */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "word a b"); + smartlist_add_linecpy(diff, area, "x"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("does not include the necessary digests") + + /* Wrong number of words after "hash". */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash a b c"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("does not include the necessary digests") + + /* base16 digests do not have the expected length. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash aaa bbb"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("includes base16-encoded digests of " + "incorrect size") + + /* base16 digests contain non-base16 characters. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + " ????????????????????????????????????????????????????????????????" + " ----------------------------------------------------------------"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("includes malformed digests") + + /* Invalid ed diff. + * As tested in apply_ed_diff, but check that apply_diff does return NULL if + * the ed diff can't be applied. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* sha3 of cons1. */ + " 06646D6CF563A41869D3B02E73254372AE3140046C5E7D83C9F71E54976AF9B4" + /* sha256 of cons2. */ + " 635D34593020C08E5ECD865F9986E29D50028EFA62843766A8197AD228A7F6AA"); + smartlist_add_linecpy(diff, area, "foobar"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_single_log_msg_containing("because an ed command was missing a line " + "number") + + /* Base consensus doesn't match its digest as found in the diff. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* bogus sha256. */ + " 3333333333333333333333333333333333333333333333333333333333333333" + /* sha256 of cons2. */ + " 635D34593020C08E5ECD865F9986E29D50028EFA62843766A8197AD228A7F6AA"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_log_msg_containing("base consensus doesn't match the digest " + "as found"); + + /* Resulting consensus doesn't match its digest as found in the diff. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* sha3 of cons1. */ + " 06646D6CF563A41869D3B02E73254372AE3140046C5E7D83C9F71E54976AF9B4" + /* bogus sha3. */ + " 3333333333333333333333333333333333333333333333333333333333333333"); + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_log_msg_containing("resulting consensus doesn't match the " + "digest as found"); + +#if 0 + /* XXXX No longer possible, since we aren't using the other algorithm. */ + /* Resulting consensus digest cannot be computed */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* sha3 of cons1. */ + " 06646D6CF563A41869D3B02E73254372AE3140046C5E7D83C9F71E54976AF9B4" + /* bogus sha3. */ + " 3333333333333333333333333333333333333333333333333333333333333333"); + smartlist_add_linecpy(diff, area, "1,2d"); // remove starting line + mock_clean_saved_logs(); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_EQ, cons2); + expect_log_msg_containing("Could not compute digests of the consensus " + "resulting from applying a consensus diff."); +#endif + + /* Very simple test, only to see that nothing errors. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* sha3 of cons1. */ + " 06646D6CF563A41869D3B02E73254372AE3140046C5E7D83C9F71E54976AF9B4" + /* sha3 of cons2. */ + " 90A418881B2FCAB3D9E60EE02E4D666D56CFA38F8A3B7AA3E0ADBA530DDA9353"); + smartlist_add_linecpy(diff, area, "3c"); + smartlist_add_linecpy(diff, area, "sample"); + smartlist_add_linecpy(diff, area, "."); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_NE, cons2); + tt_str_op( + "network-status-version foo\n" + "r name ccccccccccccccccc etc\nsample\n" + "r name eeeeeeeeeeeeeeeee etc\nbar\n" + "directory-signature foo bar\nbar\n", OP_EQ, + cons2); + tor_free(cons2); + + /* Check that lowercase letters in base16-encoded digests work too. */ + smartlist_clear(diff); + smartlist_add_linecpy(diff, area, "network-status-diff-version 1"); + smartlist_add_linecpy(diff, area, "hash" + /* sha3 of cons1. */ + " 06646d6cf563a41869d3b02e73254372ae3140046c5e7d83c9f71e54976af9b4" + /* sha3 of cons2. */ + " 90a418881b2fcab3d9e60ee02e4d666d56cfa38f8a3b7aa3e0adba530dda9353"); + smartlist_add_linecpy(diff, area, "3c"); + smartlist_add_linecpy(diff, area, "sample"); + smartlist_add_linecpy(diff, area, "."); + cons2 = consdiff_apply_diff(cons1, diff, &digests1); + tt_ptr_op(NULL, OP_NE, cons2); + tt_str_op( + "network-status-version foo\n" + "r name ccccccccccccccccc etc\nsample\n" + "r name eeeeeeeeeeeeeeeee etc\nbar\n" + "directory-signature foo bar\nbar\n", OP_EQ, + cons2); + tor_free(cons2); + + smartlist_clear(diff); + + done: + teardown_capture_of_logs(); + tor_free(cons1_str); + smartlist_free(cons1); + smartlist_free(diff); + memarea_drop_all(area); +} + +#define CONSDIFF_LEGACY(name) \ + { #name, test_consdiff_ ## name , 0, NULL, NULL } + +struct testcase_t consdiff_tests[] = { + CONSDIFF_LEGACY(smartlist_slice), + CONSDIFF_LEGACY(smartlist_slice_string_pos), + CONSDIFF_LEGACY(lcs_lengths), + CONSDIFF_LEGACY(trim_slices), + CONSDIFF_LEGACY(set_changed), + CONSDIFF_LEGACY(calc_changes), + CONSDIFF_LEGACY(get_id_hash), + CONSDIFF_LEGACY(is_valid_router_entry), + CONSDIFF_LEGACY(next_router), + CONSDIFF_LEGACY(base64cmp), + CONSDIFF_LEGACY(gen_ed_diff), + CONSDIFF_LEGACY(apply_ed_diff), + CONSDIFF_LEGACY(gen_diff), + CONSDIFF_LEGACY(apply_diff), + END_OF_TESTCASES +}; + |