X-Git-Url: https://git.exim.org/exim.git/blobdiff_plain/3634fc257bd0667daef14d72005cd87c735bbb24..f3ebb786e451da973560f1c9d8cdb151d25108b5:/src/src/parse.c diff --git a/src/src/parse.c b/src/src/parse.c index 94e42c182..e64cb94a1 100644 --- a/src/src/parse.c +++ b/src/src/parse.c @@ -2,7 +2,7 @@ * Exim - an Internet mail transport agent * *************************************************/ -/* Copyright (c) University of Cambridge 1995 - 2009 */ +/* Copyright (c) University of Cambridge 1995 - 2018 */ /* See the file NOTICE for conditions of use and distribution. */ /* Functions for parsing addresses */ @@ -23,7 +23,7 @@ redundant apparatus. */ address_item *deliver_make_addr(uschar *address, BOOL copy) { -address_item *addr = store_get(sizeof(address_item)); +address_item *addr = store_get(sizeof(address_item), FALSE); addr->next = NULL; addr->parent = NULL; addr->address = address; @@ -187,7 +187,7 @@ The start of the last potential comment position is remembered to make it possible to ignore comments at the end of compound items. Argument: current character pointer -Regurns: new character pointer +Returns: new character pointer */ static uschar * @@ -421,10 +421,10 @@ for (;;) if (*s == '\"') { *t++ = '\"'; - while ((c = *(++s)) != 0 && c != '\"') + while ((c = *++s) && c != '\"') { *t++ = c; - if (c == '\\' && s[1] != 0) *t++ = *(++s); + if (c == '\\' && s[1]) *t++ = *++s; } if (c == '\"') { @@ -443,7 +443,7 @@ for (;;) else while (!mac_iscntrl_or_special(*s) || *s == '\\') { c = *t++ = *s++; - if (c == '\\' && *s != 0) *t++ = *s++; + if (c == '\\' && *s) *t++ = *s++; } /* Terminate the word and skip subsequent comment */ @@ -550,9 +550,7 @@ read_addr_spec(uschar *s, uschar *t, int term, uschar **errorptr, { s = read_local_part(s, t, errorptr, FALSE); if (*errorptr == NULL) - { if (*s != term) - { if (*s != '@') *errorptr = string_sprintf("\"@\" or \".\" expected after \"%s\"", t); else @@ -562,8 +560,6 @@ if (*errorptr == NULL) *domainptr = t; s = read_domain(s, t, errorptr); } - } - } return s; } @@ -622,10 +618,10 @@ uschar * parse_extract_address(uschar *mailbox, uschar **errorptr, int *start, int *end, int *domain, BOOL allow_null) { -uschar *yield = store_get(Ustrlen(mailbox) + 1); +uschar *yield = store_get(Ustrlen(mailbox) + 1, is_tainted(mailbox)); uschar *startptr, *endptr; -uschar *s = (uschar *)mailbox; -uschar *t = (uschar *)yield; +uschar *s = US mailbox; +uschar *t = US yield; *domain = 0; @@ -642,7 +638,7 @@ RESTART: /* Come back here after passing a group name */ s = skip_comment(s); startptr = s; /* In case addr-spec */ s = read_local_part(s, t, errorptr, TRUE); /* Dot separated words */ -if (*errorptr != NULL) goto PARSE_FAILED; +if (*errorptr) goto PARSE_FAILED; /* If the terminator is neither < nor @ then the format of the address must either be a bare local-part (we are now at the end), or a phrase @@ -662,10 +658,10 @@ if (*s != '@' && *s != '<') end of string will produce a null local_part and therefore fail. We don't need to keep updating t, as the phrase isn't to be kept. */ - while (*s != '<' && (!parse_allow_group || *s != ':')) + while (*s != '<' && (!f.parse_allow_group || *s != ':')) { s = read_local_part(s, t, errorptr, FALSE); - if (*errorptr != NULL) + if (*errorptr) { *errorptr = string_sprintf("%s (expected word or \"<\")", *errorptr); goto PARSE_FAILED; @@ -674,8 +670,8 @@ if (*s != '@' && *s != '<') if (*s == ':') { - parse_found_group = TRUE; - parse_allow_group = FALSE; + f.parse_found_group = TRUE; + f.parse_allow_group = FALSE; s++; goto RESTART; } @@ -690,8 +686,8 @@ processing it. Note that this is "if" rather than "else if" because it's also used after reading a preceding phrase. There are a lot of broken sendmails out there that put additional pairs of <> -round s. If strip_excess_angle_brackets is set, allow any number of -them, as long as they match. */ +round s. If strip_excess_angle_brackets is set, allow a limited +number of them, as long as they match. */ if (*s == '<') { @@ -700,8 +696,11 @@ if (*s == '<') int bracket_count = 1; s++; - if (strip_excess_angle_brackets) - while (*s == '<') { bracket_count++; s++; } + if (strip_excess_angle_brackets) while (*s == '<') + { + if(bracket_count++ > 5) FAILED(US"angle-brackets nested too deep"); + s++; + } t = yield; startptr = s; @@ -715,7 +714,7 @@ if (*s == '<') if (*s == '@') { s = read_route(s, t, errorptr); - if (*errorptr != NULL) goto PARSE_FAILED; + if (*errorptr) goto PARSE_FAILED; *t = 0; /* Ensure route is ignored - probably overkill */ source_routed = TRUE; } @@ -733,7 +732,7 @@ if (*s == '<') else { s = read_addr_spec(s, t, '>', errorptr, &domainptr); - if (*errorptr != NULL) goto PARSE_FAILED; + if (*errorptr) goto PARSE_FAILED; *domain = domainptr - yield; if (source_routed && *domain == 0) FAILED(US"domain missing in source-routed address"); @@ -743,9 +742,10 @@ if (*s == '<') if (*errorptr != NULL) goto PARSE_FAILED; while (bracket_count-- > 0) if (*s++ != '>') { - *errorptr = (s[-1] == 0)? US"'>' missing at end of address" : - string_sprintf("malformed address: %.32s may not follow %.*s", - s-1, s - (uschar *)mailbox - 1, mailbox); + *errorptr = s[-1] == 0 + ? US"'>' missing at end of address" + : string_sprintf("malformed address: %.32s may not follow %.*s", + s-1, (int)(s - US mailbox - 1), mailbox); goto PARSE_FAILED; } @@ -790,21 +790,21 @@ move it back past white space if necessary. */ PARSE_SUCCEEDED: if (*s != 0) { - if (parse_found_group && *s == ';') + if (f.parse_found_group && *s == ';') { - parse_found_group = FALSE; - parse_allow_group = TRUE; + f.parse_found_group = FALSE; + f.parse_allow_group = TRUE; } else { *errorptr = string_sprintf("malformed address: %.32s may not follow %.*s", - s, s - (uschar *)mailbox, mailbox); + s, (int)(s - US mailbox), mailbox); goto PARSE_FAILED; } } -*start = startptr - (uschar *)mailbox; /* Return offsets */ +*start = startptr - US mailbox; /* Return offsets */ while (isspace(endptr[-1])) endptr--; -*end = endptr - (uschar *)mailbox; +*end = endptr - US mailbox; /* Although this code has no limitation on the length of address extracted, other parts of Exim may have limits, and in any case, RFC 2821 limits local @@ -817,17 +817,17 @@ if (*end - *start > ADDRESS_MAXLENGTH) return NULL; } -return (uschar *)yield; +return yield; /* Use goto (via the macro FAILED) to get to here from a variety of places. We might have an empty address in a group - the caller can choose to ignore this. We must, however, keep the flags correct. */ PARSE_FAILED: -if (parse_found_group && *s == ';') +if (f.parse_found_group && *s == ';') { - parse_found_group = FALSE; - parse_allow_group = TRUE; + f.parse_found_group = FALSE; + f.parse_allow_group = TRUE; } return NULL; } @@ -866,17 +866,17 @@ Returns: pointer to the original string, if no quoting needed, or the introduction */ -uschar * -parse_quote_2047(uschar *string, int len, uschar *charset, uschar *buffer, +const uschar * +parse_quote_2047(const uschar *string, int len, uschar *charset, uschar *buffer, int buffer_size, BOOL fold) { -uschar *s = string; +const uschar *s = string; uschar *p, *t; int hlen; BOOL coded = FALSE; BOOL first_byte = FALSE; -if (charset == NULL) charset = US"iso-8859-1"; +if (!charset) charset = US"iso-8859-1"; /* We don't expect this to fail! */ @@ -910,11 +910,10 @@ for (; len > 0; len--) { *t++ = '_'; first_byte = FALSE; - } + } else { - sprintf(CS t, "=%02X", ch); - while (*t != 0) t++; + t += sprintf(CS t, "=%02X", ch); coded = TRUE; first_byte = !first_byte; } @@ -926,7 +925,7 @@ for (; len > 0; len--) *t++ = '='; *t = 0; -return coded? buffer : string; +return coded ? buffer : string; } @@ -985,12 +984,13 @@ Arguments: Returns: the fixed RFC822 phrase */ -uschar * -parse_fix_phrase(uschar *phrase, int len, uschar *buffer, int buffer_size) +const uschar * +parse_fix_phrase(const uschar *phrase, int len, uschar *buffer, int buffer_size) { int ch, i; BOOL quoted = FALSE; -uschar *s, *t, *end, *yield; +const uschar *s, *end; +uschar *t, *yield; while (len > 0 && isspace(*phrase)) { phrase++; len--; } if (len > buffer_size/4) return US"Name too long"; @@ -1119,7 +1119,7 @@ while (s < end) else if (ch == '(') { - uschar *ss = s; /* uschar after '(' */ + const uschar *ss = s; /* uschar after '(' */ int level = 1; while(ss < end) { @@ -1245,7 +1245,7 @@ Returns: FF_DELIVERED addresses extracted int parse_forward_list(uschar *s, int options, address_item **anchor, - uschar **error, uschar *incoming_domain, uschar *directory, + uschar **error, const uschar *incoming_domain, uschar *directory, error_block **syntax_errors) { int count = 0; @@ -1396,7 +1396,7 @@ for (;;) if (flen <= 0) { - *error = string_sprintf("file name missing after :include:"); + *error = US"file name missing after :include:"; return FF_ERROR; } @@ -1428,7 +1428,7 @@ for (;;) /* Check file name if required */ - if (directory != NULL) + if (directory) { int len = Ustrlen(directory); uschar *p = filename + len; @@ -1440,16 +1440,53 @@ for (;;) return FF_ERROR; } +#ifdef EXIM_HAVE_OPENAT + /* It is necessary to check that every component inside the directory + is NOT a symbolic link, in order to keep the file inside the directory. + This is mighty tedious. We open the directory and openat every component, + with a flag that fails symlinks. */ + + { + int fd = open(CS directory, O_RDONLY); + if (fd < 0) + { + *error = string_sprintf("failed to open directory %s", directory); + return FF_ERROR; + } + while (*p) + { + uschar temp; + int fd2; + uschar * q = p; + + while (*++p && *p != '/') ; + temp = *p; + *p = '\0'; + + fd2 = openat(fd, CS q, O_RDONLY|O_NOFOLLOW); + close(fd); + *p = temp; + if (fd2 < 0) + { + *error = string_sprintf("failed to open %s (component of included " + "file); could be symbolic link", filename); + return FF_ERROR; + } + fd = fd2; + } + f = fdopen(fd, "rb"); + } +#else /* It is necessary to check that every component inside the directory is NOT a symbolic link, in order to keep the file inside the directory. This is mighty tedious. It is also not totally foolproof in that it leaves the possibility of a race attack, but I don't know how to do any better. */ - while (*p != 0) + while (*p) { int temp; - while (*(++p) != 0 && *p != '/'); + while (*++p && *p != '/'); temp = *p; *p = 0; if (Ulstat(filename, &statbuf) != 0) @@ -1469,11 +1506,16 @@ for (;;) return FF_ERROR; } } +#endif } - /* Open and stat the file */ +#ifdef EXIM_HAVE_OPENAT + else +#endif + /* Open and stat the file */ + f = Ufopen(filename, "rb"); - if ((f = Ufopen(filename, "rb")) == NULL) + if (!f) { *error = string_open_failed(errno, "included file %s", filename); return FF_INCLUDEFAIL; @@ -1489,7 +1531,7 @@ for (;;) /* If directory was checked, double check that we opened a regular file */ - if (directory != NULL && (statbuf.st_mode & S_IFMT) != S_IFREG) + if (directory && (statbuf.st_mode & S_IFMT) != S_IFREG) { *error = string_sprintf("included file %s is not a regular file in " "the %s directory", filename, directory); @@ -1505,7 +1547,7 @@ for (;;) return FF_ERROR; } - filebuf = store_get(statbuf.st_size + 1); + filebuf = store_get(statbuf.st_size + 1, is_tainted(filename)); if (fread(filebuf, 1, statbuf.st_size, f) != statbuf.st_size) { *error = string_sprintf("error while reading included file %s: %s", @@ -1521,10 +1563,9 @@ for (;;) error, incoming_domain, directory, syntax_errors); if (frc != FF_DELIVERED && frc != FF_NOTDELIVERED) return frc; - if (addr != NULL) + if (addr) { - last = addr; - while (last->next != NULL) { count++; last = last->next; } + for (last = addr; last->next; last = last->next) count++; last->next = *anchor; *anchor = addr; count++; @@ -1582,7 +1623,7 @@ for (;;) if ((*s == '|' || *s == '/') && (recipient == NULL || domain == 0)) { - uschar *t = store_get(Ustrlen(s) + 1); + uschar *t = store_get(Ustrlen(s) + 1, is_tainted(s)); uschar *p = t; uschar *q = s; while (*q != 0) @@ -1621,7 +1662,7 @@ for (;;) if (syntax_errors != NULL) { - error_block *e = store_get(sizeof(error_block)); + error_block *e = store_get(sizeof(error_block), FALSE); error_block *last = *syntax_errors; if (last == NULL) *syntax_errors = e; else { @@ -1689,6 +1730,7 @@ parse_message_id(uschar *str, uschar **yield, uschar **error) { uschar *domain = NULL; uschar *id; +rmark reset_point; str = skip_comment(str); if (*str != '<') @@ -1701,27 +1743,28 @@ if (*str != '<') for the answer, but it may also be very long if we are processing a header line. Therefore, take care to release unwanted store afterwards. */ -id = *yield = store_get(Ustrlen(str) + 1); +reset_point = store_mark(); +id = *yield = store_get(Ustrlen(str) + 1, is_tainted(str)); *id++ = *str++; str = read_addr_spec(str, id, '>', error, &domain); -if (*error == NULL) +if (!*error) { if (*str != '>') *error = US"Missing '>' after message-id"; else if (domain == NULL) *error = US"domain missing in message-id"; } -if (*error != NULL) +if (*error) { - store_reset(*yield); + store_reset(reset_point); return NULL; } -while (*id != 0) id++; +while (*id) id++; *id++ = *str++; *id++ = 0; -store_reset(id); +store_release_above(id); str = skip_comment(str); return str; @@ -2108,7 +2151,7 @@ allow_utf8_domains = FALSE; printf("Testing parse_extract_address with group syntax\n"); -parse_allow_group = TRUE; +f.parse_allow_group = TRUE; while (Ufgets(buffer, sizeof(buffer), stdin) != NULL) { uschar *out;