X-Git-Url: https://git.exim.org/users/jgh/exim.git/blobdiff_plain/1688f43b3071b3b4d7d3a88a6ccf28c1bc3272e0..79bc02a3499931de53f5e9ea74795d691b3a9569:/src/src/string.c diff --git a/src/src/string.c b/src/src/string.c index 2fe57b303..fbdc0246d 100644 --- a/src/src/string.c +++ b/src/src/string.c @@ -1,10 +1,8 @@ -/* $Cambridge: exim/src/src/string.c,v 1.9 2006/02/13 11:13:37 ph10 Exp $ */ - /************************************************* * Exim - an Internet mail transport agent * *************************************************/ -/* Copyright (c) University of Cambridge 1995 - 2006 */ +/* Copyright (c) University of Cambridge 1995 - 2018 */ /* See the file NOTICE for conditions of use and distribution. */ /* Miscellaneous string-handling functions. Some are not required for @@ -12,7 +10,9 @@ utilities and tests, and are cut out by the COMPILE_UTILITY macro. */ #include "exim.h" +#include +static void gstring_rebuffer(gstring * g); #ifndef COMPILE_UTILITY /************************************************* @@ -36,17 +36,16 @@ Returns: 0 if the string is not a textual representation of an IP address */ int -string_is_ip_address(uschar *s, int *maskptr) +string_is_ip_address(const uschar *s, int *maskptr) { -int i; int yield = 4; /* If an optional mask is permitted, check for it. If found, pass back the offset. */ -if (maskptr != NULL) +if (maskptr) { - uschar *ss = s + Ustrlen(s); + const uschar *ss = s + Ustrlen(s); *maskptr = 0; if (s != ss && isdigit(*(--ss))) { @@ -61,7 +60,6 @@ if (Ustrchr(s, ':') != NULL) { BOOL had_double_colon = FALSE; BOOL v4end = FALSE; - int count = 0; yield = 6; @@ -74,14 +72,14 @@ if (Ustrchr(s, ':') != NULL) may be one and only one appearance of double colon, which implies any number of binary zero bits. The number of preceding components is held in count. */ - for (count = 0; count < 8; count++) + for (int count = 0; count < 8; count++) { /* If the end of the string is reached before reading 8 components, the address is valid provided a double colon has been read. This also applies if we hit the / that introduces a mask or the % that introduces the interface specifier (scope id) of a link-local address. */ - if (*s == 0 || *s == '%' || *s == '/') return had_double_colon? yield : 0; + if (*s == 0 || *s == '%' || *s == '/') return had_double_colon ? yield : 0; /* If a component starts with an additional colon, we have hit a double colon. This is permitted to appear once only, and counts as at least @@ -135,15 +133,18 @@ if (Ustrchr(s, ':') != NULL) /* Test for IPv4 address, which may be the tail-end of an IPv6 address. */ -for (i = 0; i < 4; i++) +for (int i = 0; i < 4; i++) { + long n; + uschar * end; + if (i != 0 && *s++ != '.') return 0; - if (!isdigit(*s++)) return 0; - if (isdigit(*s) && isdigit(*(++s))) s++; + n = strtol(CCS s, CSS &end, 10); + if (n > 255 || n < 0 || end <= s || end > s+3) return 0; + s = end; } -return (*s == 0 || (*s == '/' && maskptr != NULL && *maskptr != 0))? - yield : 0; +return !*s || (*s == '/' && maskptr && *maskptr != 0) ? yield : 0; } #endif /* COMPILE_UTILITY */ @@ -167,7 +168,7 @@ Returns: pointer to the buffer uschar * string_format_size(int size, uschar *buffer) { -if (size == 0) Ustrcpy(CS buffer, " "); +if (size == 0) Ustrcpy(buffer, US" "); else if (size < 1024) sprintf(CS buffer, "%5d", size); else if (size < 10*1024) sprintf(CS buffer, "%4.1fK", (double)size / 1024.0); @@ -212,7 +213,6 @@ return yield; -#ifndef COMPILE_UTILITY /************************************************* * Interpret escape sequence * *************************************************/ @@ -223,15 +223,21 @@ interpreted in strings. Arguments: pp points a pointer to the initiating "\" in the string; the pointer gets updated to point to the final character + If the backslash is the last character in the string, it + is not interpreted. Returns: the value of the character escape */ int -string_interpret_escape(uschar **pp) +string_interpret_escape(const uschar **pp) { +#ifdef COMPILE_UTILITY +const uschar *hex_digits= CUS"0123456789abcdef"; +#endif int ch; -uschar *p = *pp; +const uschar *p = *pp; ch = *(++p); +if (ch == '\0') return **pp; if (isdigit(ch) && ch != '8' && ch != '9') { ch -= '0'; @@ -244,9 +250,12 @@ if (isdigit(ch) && ch != '8' && ch != '9') } else switch(ch) { + case 'b': ch = '\b'; break; + case 'f': ch = '\f'; break; case 'n': ch = '\n'; break; case 'r': ch = '\r'; break; case 't': ch = '\t'; break; + case 'v': ch = '\v'; break; case 'x': ch = 0; if (isxdigit(p[1])) @@ -261,7 +270,6 @@ else switch(ch) *pp = p; return ch; } -#endif /* COMPILE_UTILITY */ @@ -282,12 +290,12 @@ Arguments: Returns: string with non-printers encoded as printing sequences */ -uschar * -string_printing2(uschar *s, BOOL allow_tab) +const uschar * +string_printing2(const uschar *s, BOOL allow_tab) { int nonprintcount = 0; int length = 0; -uschar *t = s; +const uschar *t = s; uschar *ss, *tt; while (*t != 0) @@ -302,9 +310,9 @@ if (nonprintcount == 0) return s; /* Get a new block of store guaranteed big enough to hold the expanded string. */ -ss = store_get(length + nonprintcount * 4 + 1); +ss = store_get(length + nonprintcount * 3 + 1, is_tainted(s)); -/* Copy everying, escaping non printers. */ +/* Copy everything, escaping non printers. */ t = s; tt = ss; @@ -333,67 +341,103 @@ return ss; } #endif /* COMPILE_UTILITY */ - - - /************************************************* -* Copy and save string * +* Undo printing escapes in string * *************************************************/ -/* This function assumes that memcpy() is faster than strcpy(). +/* This function is the reverse of string_printing2. It searches for +backslash characters and if any are found, it makes a new copy of the +string with escape sequences parsed. Otherwise it returns the original +string. -Argument: string to copy -Returns: copy of string in new store +Arguments: + s the input string + +Returns: string with printing escapes parsed back */ uschar * -string_copy(uschar *s) +string_unprinting(uschar *s) { -int len = Ustrlen(s) + 1; -uschar *ss = store_get(len); -memcpy(ss, s, len); -return ss; -} +uschar *p, *q, *r, *ss; +int len, off; +p = Ustrchr(s, '\\'); +if (!p) return s; +len = Ustrlen(s) + 1; +ss = store_get(len, is_tainted(s)); -/************************************************* -* Copy and save string in malloc'd store * -*************************************************/ - -/* This function assumes that memcpy() is faster than strcpy(). +q = ss; +off = p - s; +if (off) + { + memcpy(q, s, off); + q += off; + } -Argument: string to copy -Returns: copy of string in new store -*/ +while (*p) + { + if (*p == '\\') + { + *q++ = string_interpret_escape((const uschar **)&p); + p++; + } + else + { + r = Ustrchr(p, '\\'); + if (!r) + { + off = Ustrlen(p); + memcpy(q, p, off); + p += off; + q += off; + break; + } + else + { + off = r - p; + memcpy(q, p, off); + q += off; + p = r; + } + } + } +*q = '\0'; -uschar * -string_copy_malloc(uschar *s) -{ -int len = Ustrlen(s) + 1; -uschar *ss = store_malloc(len); -memcpy(ss, s, len); return ss; } + +#if (defined(HAVE_LOCAL_SCAN) || defined(EXPAND_DLFUNC)) \ + && !defined(MACRO_PREDEF) && !defined(COMPILE_UTILITY) /************************************************* -* Copy, lowercase and save string * +* Copy and save string * *************************************************/ /* Argument: string to copy -Returns: copy of string in new store, with letters lowercased +Returns: copy of string in new store with the same taint status */ uschar * -string_copylc(uschar *s) +string_copy_function(const uschar *s) { -uschar *ss = store_get(Ustrlen(s) + 1); -uschar *p = ss; -while (*s != 0) *p++ = tolower(*s++); -*p = 0; +return string_copy_taint(s, is_tainted(s)); +} + +/* This function assumes that memcpy() is faster than strcpy(). +As above, but explicitly specifying the result taint status +*/ + +uschar * +string_copy_taint_function(const uschar * s, BOOL tainted) +{ +int len = Ustrlen(s) + 1; +uschar *ss = store_get(len, tainted); +memcpy(ss, s, len); return ss; } @@ -414,41 +458,98 @@ Returns: copy of string in new store */ uschar * -string_copyn(uschar *s, int n) +string_copyn_function(const uschar *s, int n) { -uschar *ss = store_get(n + 1); +uschar *ss = store_get(n + 1, is_tainted(s)); Ustrncpy(ss, s, n); ss[n] = 0; return ss; } +#endif /************************************************* -* Copy, lowercase, and save string, given length * +* Copy and save string in malloc'd store * *************************************************/ -/* It is assumed the data contains no zeros. A zero is added -onto the end. - -Arguments: - s string to copy - n number of characters +/* This function assumes that memcpy() is faster than strcpy(). -Returns: copy of string in new store, with letters lowercased +Argument: string to copy +Returns: copy of string in new store */ uschar * -string_copynlc(uschar *s, int n) +string_copy_malloc(const uschar *s) { -uschar *ss = store_get(n + 1); -uschar *p = ss; -while (n-- > 0) *p++ = tolower(*s++); -*p = 0; +int len = Ustrlen(s) + 1; +uschar *ss = store_malloc(len); +memcpy(ss, s, len); return ss; } +/************************************************* +* Copy string if long, inserting newlines * +*************************************************/ + +/* If the given string is longer than 75 characters, it is copied, and within +the copy, certain space characters are converted into newlines. + +Argument: pointer to the string +Returns: pointer to the possibly altered string +*/ + +uschar * +string_split_message(uschar *msg) +{ +uschar *s, *ss; + +if (msg == NULL || Ustrlen(msg) <= 75) return msg; +s = ss = msg = string_copy(msg); + +for (;;) + { + int i = 0; + while (i < 75 && *ss != 0 && *ss != '\n') ss++, i++; + if (*ss == 0) break; + if (*ss == '\n') + s = ++ss; + else + { + uschar *t = ss + 1; + uschar *tt = NULL; + while (--t > s + 35) + { + if (*t == ' ') + { + if (t[-1] == ':') { tt = t; break; } + if (tt == NULL) tt = t; + } + } + + if (tt == NULL) /* Can't split behind - try ahead */ + { + t = ss + 1; + while (*t != 0) + { + if (*t == ' ' || *t == '\n') + { tt = t; break; } + t++; + } + } + + if (tt == NULL) break; /* Can't find anywhere to split */ + *tt = '\n'; + s = ss = tt+1; + } + } + +return msg; +} + + + /************************************************* * Copy returned DNS domain name, de-escaping * *************************************************/ @@ -469,7 +570,7 @@ uschar * string_copy_dnsdomain(uschar *s) { uschar *yield; -uschar *ss = yield = store_get(Ustrlen(s) + 1); +uschar *ss = yield = store_get(Ustrlen(s) + 1, is_tainted(s)); while (*s != 0) { @@ -509,49 +610,44 @@ Returns: the new string */ uschar * -string_dequote(uschar **sptr) +string_dequote(const uschar **sptr) { -uschar *s = *sptr; +const uschar *s = *sptr; uschar *t, *yield; /* First find the end of the string */ if (*s != '\"') - { while (*s != 0 && !isspace(*s)) s++; - } else { s++; - while (*s != 0 && *s != '\"') + while (*s && *s != '\"') { if (*s == '\\') (void)string_interpret_escape(&s); s++; } - if (*s != 0) s++; + if (*s) s++; } /* Get enough store to copy into */ -t = yield = store_get(s - *sptr + 1); +t = yield = store_get(s - *sptr + 1, is_tainted(*sptr)); s = *sptr; /* Do the copy */ if (*s != '\"') - { while (*s != 0 && !isspace(*s)) *t++ = *s++; - } else { s++; while (*s != 0 && *s != '\"') { - if (*s == '\\') *t++ = string_interpret_escape(&s); - else *t++ = *s; + *t++ = *s == '\\' ? string_interpret_escape(&s) : *s; s++; } - if (*s != 0) s++; + if (*s) s++; } /* Update the pointer and return the terminated copy */ @@ -568,8 +664,8 @@ return yield; * Format a string and save it * *************************************************/ -/* The formatting is done by string_format, which checks the length of -everything. +/* The formatting is done by string_vformat, which checks the length of +everything. Taint is taken from the worst of the arguments. Arguments: format a printf() format - deliberately char * rather than uschar * @@ -580,16 +676,36 @@ Returns: pointer to fresh piece of store containing sprintf'ed string */ uschar * -string_sprintf(char *format, ...) +string_sprintf_trc(const char *format, const uschar * func, unsigned line, ...) { -va_list ap; +#ifdef COMPILE_UTILITY uschar buffer[STRING_SPRINTF_BUFFER_SIZE]; -va_start(ap, format); -if (!string_vformat(buffer, sizeof(buffer), format, ap)) - log_write(0, LOG_MAIN|LOG_PANIC_DIE, - "string_sprintf expansion was longer than %d", sizeof(buffer)); +gstring gs = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer }; +gstring * g = &gs; +unsigned flags = 0; +#else +gstring * g = NULL; +unsigned flags = SVFMT_REBUFFER|SVFMT_EXTEND; +#endif + +va_list ap; +va_start(ap, line); +g = string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE, + flags, format, ap); va_end(ap); -return string_copy(buffer); + +if (!g) + log_write(0, LOG_MAIN|LOG_PANIC_DIE, + "string_sprintf expansion was longer than %d; format string was (%s)\n" + " called from %s %d\n", + STRING_SPRINTF_BUFFER_SIZE, format, func, line); + +#ifdef COMPILE_UTILITY +return string_copyn(g->s, g->ptr); +#else +gstring_release_unused(g); +return string_from_gstring(g); +#endif } @@ -608,7 +724,7 @@ Returns: < 0, = 0, or > 0, according to the comparison */ int -strncmpic(uschar *s, uschar *t, int n) +strncmpic(const uschar *s, const uschar *t, int n) { while (n--) { @@ -632,7 +748,7 @@ Returns: < 0, = 0, or > 0, according to the comparison */ int -strcmpic(uschar *s, uschar *t) +strcmpic(const uschar *s, const uschar *t) { while (*s != 0) { @@ -695,6 +811,17 @@ return NULL; +#ifdef COMPILE_UTILITY +/* Dummy version for this function; it should never be called */ +static void +gstring_grow(gstring * g, int count) +{ +assert(FALSE); +} +#endif + + + #ifndef COMPILE_UTILITY /************************************************* * Get next string from separated list * @@ -703,19 +830,26 @@ return NULL; /* Leading and trailing space is removed from each item. The separator in the list is controlled by the int pointed to by the separator argument as follows: - If its value is > 0 it is used as the delimiter. - (If its value is actually > UCHAR_MAX there is only one item in the list. + If the value is > 0 it is used as the separator. This is typically used for + sublists such as slash-separated options. The value is always a printing + character. + + (If the value is actually > UCHAR_MAX there is only one item in the list. This is used for some cases when called via functions that sometimes plough through lists, and sometimes are given single items.) - If its value is <= 0, the string is inspected for a leading 0 && isspace(buffer[p-1])) p--; - buffer[p] = 0; + buffer[p] = '\0'; } /* Handle the case when a buffer is not provided. */ else { + gstring * g = NULL; + /* We know that *s != 0 at this point. However, it might be pointing to a - separator, which could indicate an empty string, or could be doubled to - indicate a separator character as data at the start of a string. */ + separator, which could indicate an empty string, or (if an ispunct() + character) could be doubled to indicate a separator character as data at the + start of a string. Avoid getting working memory for an empty item. */ if (*s == sep) { s++; - if (*s != sep) buffer = string_copy(US""); + if (*s != sep || sep_is_special) + { + *listptr = s; + return string_copy(US""); + } } - if (buffer == NULL) + /* Not an empty string; the first character is guaranteed to be a data + character. */ + + for (;;) { - int size = 0; - int ptr = 0; - uschar *ss; - for (;;) - { - for (ss = s + 1; *ss != 0 && *ss != sep; ss++); - buffer = string_cat(buffer, &size, &ptr, s, ss-s); - s = ss; - if (*s == 0 || *(++s) != sep) break; - } - while (ptr > 0 && isspace(buffer[ptr-1])) ptr--; - buffer[ptr] = 0; + const uschar * ss; + for (ss = s + 1; *ss && *ss != sep; ) ss++; + g = string_catn(g, s, ss-s); + s = ss; + if (!*s || *++s != sep || sep_is_special) break; } + while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; + buffer = string_from_gstring(g); + gstring_release_unused(g); } /* Update the current pointer and return the new string */ @@ -805,15 +958,154 @@ else *listptr = s; return buffer; } -#endif /* COMPILE_UTILITY */ + + +static const uschar * +Ustrnchr(const uschar * s, int c, unsigned * len) +{ +unsigned siz = *len; +while (siz) + { + if (!*s) return NULL; + if (*s == c) + { + *len = siz; + return s; + } + s++; + siz--; + } +return NULL; +} + + +/************************************************ +* Add element to separated list * +************************************************/ +/* This function is used to build a list, returning an allocated null-terminated +growable string. The given element has any embedded separator characters +doubled. + +Despite having the same growable-string interface as string_cat() the list is +always returned null-terminated. + +Arguments: + list expanding-string for the list that is being built, or NULL + if this is a new list that has no contents yet + sep list separator character + ele new element to be appended to the list + +Returns: pointer to the start of the list, changed if copied for expansion. +*/ + +gstring * +string_append_listele(gstring * list, uschar sep, const uschar * ele) +{ +uschar * sp; + +if (list && list->ptr) + list = string_catn(list, &sep, 1); + +while((sp = Ustrchr(ele, sep))) + { + list = string_catn(list, ele, sp-ele+1); + list = string_catn(list, &sep, 1); + ele = sp+1; + } +list = string_cat(list, ele); +(void) string_from_gstring(list); +return list; +} + + +gstring * +string_append_listele_n(gstring * list, uschar sep, const uschar * ele, + unsigned len) +{ +const uschar * sp; + +if (list && list->ptr) + list = string_catn(list, &sep, 1); + +while((sp = Ustrnchr(ele, sep, &len))) + { + list = string_catn(list, ele, sp-ele+1); + list = string_catn(list, &sep, 1); + ele = sp+1; + len--; + } +list = string_catn(list, ele, len); +(void) string_from_gstring(list); +return list; +} + + + +/* A slightly-bogus listmaker utility; the separator is a string so +can be multiple chars - there is no checking for the element content +containing any of the separator. */ + +gstring * +string_append2_listele_n(gstring * list, const uschar * sepstr, + const uschar * ele, unsigned len) +{ +if (list && list->ptr) + list = string_cat(list, sepstr); + +list = string_catn(list, ele, len); +(void) string_from_gstring(list); +return list; +} + + + +/************************************************/ +/* Add more space to a growable-string. The caller should check +first if growth is required. The gstring struct is modified on +return; specifically, the string-base-pointer may have been changed. + +Arguments: + g the growable-string + count amount needed for g->ptr to increase by +*/ + +static void +gstring_grow(gstring * g, int count) +{ +int p = g->ptr; +int oldsize = g->size; +BOOL tainted = is_tainted(g->s); + +/* Mostly, string_cat() is used to build small strings of a few hundred +characters at most. There are times, however, when the strings are very much +longer (for example, a lookup that returns a vast number of alias addresses). +To try to keep things reasonable, we use increments whose size depends on the +existing length of the string. */ + +unsigned inc = oldsize < 4096 ? 127 : 1023; + +if (count <= 0) return; +g->size = (p + count + inc + 1) & ~inc; /* one for a NUL */ + +/* Try to extend an existing allocation. If the result of calling +store_extend() is false, either there isn't room in the current memory block, +or this string is not the top item on the dynamic store stack. We then have +to get a new chunk of store and copy the old string. When building large +strings, it is helpful to call store_release() on the old string, to release +memory blocks that have become empty. (The block will be freed if the string +is at its start.) However, we can do this only if we know that the old string +was the last item on the dynamic memory stack. This is the case if it matches +store_last_get. */ + +if (!store_extend(g->s, tainted, oldsize, g->size)) + g->s = store_newblock(g->s, tainted, g->size, p); +} -#ifndef COMPILE_UTILITY /************************************************* * Add chars to string * *************************************************/ - /* This function is used when building up strings of unknown length. Room is always left for a terminating zero to be added to the string that is being built. This function does not require the string that is being added to be NUL @@ -823,77 +1115,55 @@ sometimes called to extract parts of other strings. Arguments: string points to the start of the string that is being built, or NULL if this is a new string that has no contents yet - size points to a variable that holds the current capacity of the memory - block (updated if changed) - ptr points to a variable that holds the offset at which to add - characters, updated to the new offset s points to characters to add count count of characters to add; must not exceed the length of s, if s - is a C string - -If string is given as NULL, *size and *ptr should both be zero. + is a C string. Returns: pointer to the start of the string, changed if copied for expansion. Note that a NUL is not added, though space is left for one. This is because string_cat() is often called multiple times to build up a string - there's no point adding the NUL till the end. + */ +/* coverity[+alloc] */ -uschar * -string_cat(uschar *string, int *size, int *ptr, const uschar *s, int count) +gstring * +string_catn(gstring * g, const uschar *s, int count) { -int p = *ptr; +int p; +BOOL srctaint = is_tainted(s); -if (p + count >= *size) +if (!g) { - int oldsize = *size; - - /* Mostly, string_cat() is used to build small strings of a few hundred - characters at most. There are times, however, when the strings are very much - longer (for example, a lookup that returns a vast number of alias addresses). - To try to keep things reasonable, we use increments whose size depends on the - existing length of the string. */ - - int inc = (oldsize < 4096)? 100 : 1024; - while (*size <= p + count) *size += inc; - - /* New string */ - - if (string == NULL) string = store_get(*size); - - /* Try to extend an existing allocation. If the result of calling - store_extend() is false, either there isn't room in the current memory block, - or this string is not the top item on the dynamic store stack. We then have - to get a new chunk of store and copy the old string. When building large - strings, it is helpful to call store_release() on the old string, to release - memory blocks that have become empty. (The block will be freed if the string - is at its start.) However, we can do this only if we know that the old string - was the last item on the dynamic memory stack. This is the case if it matches - store_last_get. */ - - else if (!store_extend(string, oldsize, *size)) - { - BOOL release_ok = store_last_get[store_pool] == string; - uschar *newstring = store_get(*size); - memcpy(newstring, string, p); - if (release_ok) store_release(string); - string = newstring; - } + unsigned inc = count < 4096 ? 127 : 1023; + unsigned size = ((count + inc) & ~inc) + 1; + g = string_get_tainted(size, srctaint); } +else if (srctaint && !is_tainted(g->s)) + gstring_rebuffer(g); + +p = g->ptr; +if (p + count >= g->size) + gstring_grow(g, count); /* Because we always specify the exact number of characters to copy, we can use memcpy(), which is likely to be more efficient than strncopy() because the latter has to check for zero bytes. */ -memcpy(string + p, s, count); -*ptr = p + count; -return string; +memcpy(g->s + p, s, count); +g->ptr = p + count; +return g; +} + + +gstring * +string_cat(gstring *string, const uschar *s) +{ +return string_catn(string, s, Ustrlen(s)); } -#endif /* COMPILE_UTILITY */ -#ifndef COMPILE_UTILITY /************************************************* * Append strings to another string * *************************************************/ @@ -902,12 +1172,8 @@ return string; It calls string_cat() to do the dirty work. Arguments: - string points to the start of the string that is being built, or NULL + string expanding-string that is being built, or NULL if this is a new string that has no contents yet - size points to a variable that holds the current capacity of the memory - block (updated if changed) - ptr points to a variable that holds the offset at which to add - characters, updated to the new offset count the number of strings to append ... "count" uschar* arguments, which must be valid zero-terminated C strings @@ -916,17 +1182,16 @@ Returns: pointer to the start of the string, changed if copied for expansion. The string is not zero-terminated - see string_cat() above. */ -uschar * -string_append(uschar *string, int *size, int *ptr, int count, ...) +__inline__ gstring * +string_append(gstring *string, int count, ...) { va_list ap; -int i; va_start(ap, count); -for (i = 0; i < count; i++) +while (count-- > 0) { uschar *t = va_arg(ap, uschar *); - string = string_cat(string, size, ptr, t, Ustrlen(t)); + string = string_cat(string, t); } va_end(ap); @@ -948,10 +1213,10 @@ on whether the variable length list of data arguments are given explicitly or as a va_list item. The formats are the usual printf() ones, with some omissions (never used) and -two additions for strings: %S forces lower case, and %#s or %#S prints nothing -for a NULL string. Without the # "NULL" is printed (useful in debugging). There -is also the addition of %D, which inserts the date in the form used for -datestamped log files. +three additions for strings: %S forces lower case, %T forces upper case, and +%#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed +(useful in debugging). There is also the addition of %D and %M, which insert +the date in the form used for datestamped log files. Arguments: buffer a buffer in which to put the formatted string @@ -963,47 +1228,107 @@ Returns: TRUE if the result fitted in the buffer */ BOOL -string_format(uschar *buffer, int buflen, char *format, ...) +string_format_trc(uschar * buffer, int buflen, + const uschar * func, unsigned line, const char * format, ...) { -BOOL yield; +gstring g = { .size = buflen, .ptr = 0, .s = buffer }, *gp; va_list ap; va_start(ap, format); -yield = string_vformat(buffer, buflen, format, ap); +gp = string_vformat_trc(&g, func, line, STRING_SPRINTF_BUFFER_SIZE, + 0, format, ap); va_end(ap); -return yield; +g.s[g.ptr] = '\0'; +return !!gp; } -BOOL -string_vformat(uschar *buffer, int buflen, char *format, va_list ap) + +/* Copy the content of a string to tainted memory */ +static void +gstring_rebuffer(gstring * g) { -enum { L_NORMAL, L_SHORT, L_LONG, L_LONGLONG, L_LONGDOUBLE }; +uschar * s = store_get(g->size, TRUE); +memcpy(s, g->s, g->ptr); +g->s = s; +} + + + +/* Build or append to a growing-string, sprintf-style. + +If the "extend" flag is true, the string passed in can be NULL, +empty, or non-empty. Growing is subject to an overall limit given +by the size_limit argument. -BOOL yield = TRUE; -int width, precision; -char *fp = format; /* Deliberately not unsigned */ -uschar *p = buffer; -uschar *last = buffer + buflen - 1; +If the "extend" flag is false, the string passed in may not be NULL, +will not be grown, and is usable in the original place after return. +The return value can be NULL to signify overflow. + +Returns the possibly-new (if copy for growth or taint-handling was needed) +string, not nul-terminated. +*/ + +gstring * +string_vformat_trc(gstring * g, const uschar * func, unsigned line, + unsigned size_limit, unsigned flags, const char *format, va_list ap) +{ +enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 }; + +int width, precision, off, lim, need; +const char * fp = format; /* Deliberately not unsigned */ +BOOL dest_tainted = FALSE; + +string_datestamp_offset = -1; /* Datestamp not inserted */ +string_datestamp_length = 0; /* Datestamp not inserted */ +string_datestamp_type = 0; /* Datestamp not inserted */ + +#ifdef COMPILE_UTILITY +assert(!(flags & SVFMT_EXTEND)); +assert(g); +#else + +/* Ensure we have a string, to save on checking later */ +if (!g) g = string_get(16); +else if (!(flags & SVFMT_TAINT_NOCHK)) dest_tainted = is_tainted(g->s); + +if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(format)) + { +#ifndef MACRO_PREDEF + if (!(flags & SVFMT_REBUFFER)) + die_tainted(US"string_vformat", func, line); +#endif + gstring_rebuffer(g); + dest_tainted = TRUE; + } +#endif /*!COMPILE_UTILITY*/ -string_datestamp_offset = -1; /* Datestamp not inserted */ +lim = g->size - 1; /* leave one for a nul */ +off = g->ptr; /* remember initial offset in gstring */ /* Scan the format and handle the insertions */ -while (*fp != 0) +while (*fp) { int length = L_NORMAL; int *nptr; int slen; - char *null = "NULL"; /* ) These variables */ - char *item_start, *s; /* ) are deliberately */ - char newformat[16]; /* ) not unsigned */ + const char *null = "NULL"; /* ) These variables */ + const char *item_start, *s; /* ) are deliberately */ + char newformat[16]; /* ) not unsigned */ + char * gp = CS g->s + g->ptr; /* ) */ /* Non-% characters just get copied verbatim */ if (*fp != '%') { - if (p >= last) { yield = FALSE; break; } - *p++ = (uschar)*fp++; + /* Avoid string_copyn() due to COMPILE_UTILITY */ + if ((need = g->ptr + 1) > lim) + { + if (!(flags & SVFMT_EXTEND) || need > size_limit) return NULL; + gstring_grow(g, 1); + lim = g->size - 1; + } + g->s[g->ptr++] = (uschar) *fp++; continue; } @@ -1031,78 +1356,92 @@ while (*fp != 0) } if (*fp == '.') - { if (*(++fp) == '*') { precision = va_arg(ap, int); fp++; } else - { - precision = 0; - while (isdigit((uschar)*fp)) - precision = precision*10 + *fp++ - '0'; - } - } + for (precision = 0; isdigit((uschar)*fp); fp++) + precision = precision*10 + *fp - '0'; - /* Skip over 'h', 'L', 'l', and 'll', remembering the item length */ + /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */ if (*fp == 'h') { fp++; length = L_SHORT; } else if (*fp == 'L') { fp++; length = L_LONGDOUBLE; } else if (*fp == 'l') - { if (fp[1] == 'l') - { - fp += 2; - length = L_LONGLONG; - } + { fp += 2; length = L_LONGLONG; } else - { - fp++; - length = L_LONG; - } - } + { fp++; length = L_LONG; } + else if (*fp == 'z') + { fp++; length = L_SIZE; } /* Handle each specific format type. */ switch (*fp++) { case 'n': - nptr = va_arg(ap, int *); - *nptr = p - buffer; - break; + nptr = va_arg(ap, int *); + *nptr = g->ptr - off; + break; case 'd': case 'o': case 'u': case 'x': case 'X': - if (p >= last - ((length > L_LONG)? 24 : 12)) - { yield = FALSE; goto END_FORMAT; } - strncpy(newformat, item_start, fp - item_start); - newformat[fp - item_start] = 0; - - /* Short int is promoted to int when passing through ..., so we must use - int for va_arg(). */ + width = length > L_LONG ? 24 : 12; + if ((need = g->ptr + width) > lim) + { + if (!(flags & SVFMT_EXTEND) || need >= size_limit) return NULL; + gstring_grow(g, width); + lim = g->size - 1; + gp = CS g->s + g->ptr; + } + strncpy(newformat, item_start, fp - item_start); + newformat[fp - item_start] = 0; + + /* Short int is promoted to int when passing through ..., so we must use + int for va_arg(). */ + + switch(length) + { + case L_SHORT: + case L_NORMAL: + g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break; + case L_LONG: + g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break; + case L_LONGLONG: + g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break; + case L_SIZE: + g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break; + } + break; - switch(length) + case 'p': { - case L_SHORT: - case L_NORMAL: sprintf(CS p, newformat, va_arg(ap, int)); break; - case L_LONG: sprintf(CS p, newformat, va_arg(ap, long int)); break; - case L_LONGLONG: sprintf(CS p, newformat, va_arg(ap, LONGLONG_T)); break; + void * ptr; + if ((need = g->ptr + 24) > lim) + { + if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; + gstring_grow(g, 24); + lim = g->size - 1; + gp = CS g->s + g->ptr; + } + /* sprintf() saying "(nil)" for a null pointer seems unreliable. + Handle it explicitly. */ + if ((ptr = va_arg(ap, void *))) + { + strncpy(newformat, item_start, fp - item_start); + newformat[fp - item_start] = 0; + g->ptr += sprintf(gp, newformat, ptr); + } + else + g->ptr += sprintf(gp, "(nil)"); } - while (*p) p++; - break; - - case 'p': - if (p >= last - 24) { yield = FALSE; goto END_FORMAT; } - strncpy(newformat, item_start, fp - item_start); - newformat[fp - item_start] = 0; - sprintf(CS p, newformat, va_arg(ap, void *)); - while (*p) p++; break; /* %f format is inherently insecure if the numbers that it may be @@ -1117,97 +1456,148 @@ while (*fp != 0) case 'E': case 'g': case 'G': - if (precision < 0) precision = 6; - if (p >= last - precision - 8) { yield = FALSE; goto END_FORMAT; } - strncpy(newformat, item_start, fp - item_start); - newformat[fp-item_start] = 0; - if (length == L_LONGDOUBLE) - sprintf(CS p, newformat, va_arg(ap, long double)); - else - sprintf(CS p, newformat, va_arg(ap, double)); - while (*p) p++; - break; + if (precision < 0) precision = 6; + if ((need = g->ptr + precision + 8) > lim) + { + if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; + gstring_grow(g, precision+8); + lim = g->size - 1; + gp = CS g->s + g->ptr; + } + strncpy(newformat, item_start, fp - item_start); + newformat[fp-item_start] = 0; + if (length == L_LONGDOUBLE) + g->ptr += sprintf(gp, newformat, va_arg(ap, long double)); + else + g->ptr += sprintf(gp, newformat, va_arg(ap, double)); + break; /* String types */ case '%': - if (p >= last) { yield = FALSE; goto END_FORMAT; } - *p++ = '%'; - break; + if ((need = g->ptr + 1) > lim) + { + if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; + gstring_grow(g, 1); + lim = g->size - 1; + } + g->s[g->ptr++] = (uschar) '%'; + break; case 'c': - if (p >= last) { yield = FALSE; goto END_FORMAT; } - *p++ = va_arg(ap, int); - break; + if ((need = g->ptr + 1) > lim) + { + if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; + gstring_grow(g, 1); + lim = g->size - 1; + } + g->s[g->ptr++] = (uschar) va_arg(ap, int); + break; - case 'D': /* Insert datestamp for log file names */ - s = CS tod_stamp(tod_log_datestamp); - string_datestamp_offset = p - buffer; /* Passed back via global */ - goto INSERT_STRING; + case 'D': /* Insert daily datestamp for log file names */ + s = CS tod_stamp(tod_log_datestamp_daily); + string_datestamp_offset = g->ptr; /* Passed back via global */ + string_datestamp_length = Ustrlen(s); /* Passed back via global */ + string_datestamp_type = tod_log_datestamp_daily; + slen = string_datestamp_length; + goto INSERT_STRING; + + case 'M': /* Insert monthly datestamp for log file names */ + s = CS tod_stamp(tod_log_datestamp_monthly); + string_datestamp_offset = g->ptr; /* Passed back via global */ + string_datestamp_length = Ustrlen(s); /* Passed back via global */ + string_datestamp_type = tod_log_datestamp_monthly; + slen = string_datestamp_length; + goto INSERT_STRING; case 's': case 'S': /* Forces *lower* case */ - s = va_arg(ap, char *); - - INSERT_STRING: /* Come to from %D above */ - if (s == NULL) s = null; - slen = Ustrlen(s); + case 'T': /* Forces *upper* case */ + s = va_arg(ap, char *); + + if (!s) s = null; + slen = Ustrlen(s); + + if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(s)) + if (flags & SVFMT_REBUFFER) + { + gstring_rebuffer(g); + gp = CS g->s + g->ptr; + dest_tainted = TRUE; + } +#ifndef MACRO_PREDEF + else + die_tainted(US"string_vformat", func, line); +#endif - /* If the width is specified, check that there is a precision - set; if not, set it to the width to prevent overruns of long - strings. */ + INSERT_STRING: /* Come to from %D or %M above */ - if (width >= 0) { - if (precision < 0) precision = width; - } + BOOL truncated = FALSE; - /* If a width is not specified and the precision is specified, set - the width to the precision, or the string length if shorted. */ + /* If the width is specified, check that there is a precision + set; if not, set it to the width to prevent overruns of long + strings. */ - else if (precision >= 0) - { - width = (precision < slen)? precision : slen; - } + if (width >= 0) + { + if (precision < 0) precision = width; + } - /* If neither are specified, set them both to the string length. */ + /* If a width is not specified and the precision is specified, set + the width to the precision, or the string length if shorted. */ - else width = precision = slen; + else if (precision >= 0) + width = precision < slen ? precision : slen; - /* Check string space, and add the string to the buffer if ok. If - not OK, add part of the string (debugging uses this to show as - much as possible). */ + /* If neither are specified, set them both to the string length. */ - if (p >= last - width) - { - yield = FALSE; - width = precision = last - p - 1; + else + width = precision = slen; + + if ((need = g->ptr + width) >= size_limit || !(flags & SVFMT_EXTEND)) + { + if (g->ptr == lim) return NULL; + if (need > lim) + { + truncated = TRUE; + width = precision = lim - g->ptr - 1; + if (width < 0) width = 0; + if (precision < 0) precision = 0; + } + } + else if (need > lim) + { + gstring_grow(g, width); + lim = g->size - 1; + gp = CS g->s + g->ptr; + } + + g->ptr += sprintf(gp, "%*.*s", width, precision, s); + if (fp[-1] == 'S') + while (*gp) { *gp = tolower(*gp); gp++; } + else if (fp[-1] == 'T') + while (*gp) { *gp = toupper(*gp); gp++; } + + if (truncated) return NULL; + break; } - sprintf(CS p, "%*.*s", width, precision, s); - if (fp[-1] == 'S') - while (*p) { *p = tolower(*p); p++; } - else - while (*p) p++; - if (!yield) goto END_FORMAT; - break; /* Some things are never used in Exim; also catches junk. */ default: - strncpy(newformat, item_start, fp - item_start); - newformat[fp-item_start] = 0; - log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type " - "in \"%s\" in \"%s\"", newformat, format); - break; + strncpy(newformat, item_start, fp - item_start); + newformat[fp-item_start] = 0; + log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type " + "in \"%s\" in \"%s\"", newformat, format); + break; } } -/* Ensure string is complete; return TRUE if got to the end of the format */ - -END_FORMAT: - -*p = 0; -return yield; +if (g->ptr > g->size) + log_write(0, LOG_MAIN|LOG_PANIC_DIE, + "string_format internal error: caller %s %d", func, line); +return g; } @@ -1230,182 +1620,48 @@ Returns: a message, in dynamic store */ uschar * -string_open_failed(int eno, char *format, ...) +string_open_failed_trc(int eno, const uschar * func, unsigned line, + const char *format, ...) { va_list ap; -uschar buffer[1024]; +gstring * g = string_get(1024); -Ustrcpy(buffer, "failed to open "); -va_start(ap, format); +g = string_catn(g, US"failed to open ", 15); /* Use the checked formatting routine to ensure that the buffer does not overflow. It should not, since this is called only for internally specified messages. If it does, the message just gets truncated, and there doesn't seem much we can do about that. */ -(void)string_vformat(buffer+15, sizeof(buffer) - 15, format, ap); +va_start(ap, format); +(void) string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE, + 0, format, ap); +string_from_gstring(g); +gstring_release_unused(g); +va_end(ap); -return (eno == EACCES)? - string_sprintf("%s: %s (euid=%ld egid=%ld)", buffer, strerror(eno), - (long int)geteuid(), (long int)getegid()) : - string_sprintf("%s: %s", buffer, strerror(eno)); +return eno == EACCES + ? string_sprintf("%s: %s (euid=%ld egid=%ld)", g->s, strerror(eno), + (long int)geteuid(), (long int)getegid()) + : string_sprintf("%s: %s", g->s, strerror(eno)); } #endif /* COMPILE_UTILITY */ -#ifndef COMPILE_UTILITY -/************************************************* -* Generate local prt for logging * -*************************************************/ - -/* This function is a subroutine for use in string_log_address() below. - -Arguments: - addr the address being logged - yield the current dynamic buffer pointer - sizeptr points to current size - ptrptr points to current insert pointer - -Returns: the new value of the buffer pointer -*/ - -static uschar * -string_get_localpart(address_item *addr, uschar *yield, int *sizeptr, - int *ptrptr) -{ -if (testflag(addr, af_include_affixes) && addr->prefix != NULL) - yield = string_cat(yield, sizeptr, ptrptr, addr->prefix, - Ustrlen(addr->prefix)); -yield = string_cat(yield, sizeptr, ptrptr, addr->local_part, - Ustrlen(addr->local_part)); -if (testflag(addr, af_include_affixes) && addr->suffix != NULL) - yield = string_cat(yield, sizeptr, ptrptr, addr->suffix, - Ustrlen(addr->suffix)); -return yield; -} -/************************************************* -* Generate log address list * -*************************************************/ - -/* This function generates a list consisting of an address and its parents, for -use in logging lines. For saved onetime aliased addresses, the onetime parent -field is used. If the address was delivered by a transport with rcpt_include_ -affixes set, the af_include_affixes bit will be set in the address. In that -case, we include the affixes here too. - -Arguments: - addr bottom (ultimate) address - all_parents if TRUE, include all parents - success TRUE for successful delivery - -Returns: a string in dynamic store -*/ +#ifndef COMPILE_UTILITY +/* qsort(3), currently used to sort the environment variables +for -bP environment output, needs a function to compare two pointers to string +pointers. Here it is. */ -uschar * -string_log_address(address_item *addr, BOOL all_parents, BOOL success) +int +string_compare_by_pointer(const void *a, const void *b) { -int size = 64; -int ptr = 0; -BOOL add_topaddr = TRUE; -uschar *yield = store_get(size); -address_item *topaddr; - -/* Find the ultimate parent */ - -for (topaddr = addr; topaddr->parent != NULL; topaddr = topaddr->parent); - -/* We start with just the local part for pipe, file, and reply deliveries, and -for successful local deliveries from routers that have the log_as_local flag -set. File deliveries from filters can be specified as non-absolute paths in -cases where the transport is goin to complete the path. If there is an error -before this happens (expansion failure) the local part will not be updated, and -so won't necessarily look like a path. Add extra text for this case. */ - -if (testflag(addr, af_pfr) || - (success && - addr->router != NULL && addr->router->log_as_local && - addr->transport != NULL && addr->transport->info->local)) - { - if (testflag(addr, af_file) && addr->local_part[0] != '/') - yield = string_cat(yield, &size, &ptr, CUS"save ", 5); - yield = string_get_localpart(addr, yield, &size, &ptr); - } - -/* Other deliveries start with the full address. It we have split it into local -part and domain, use those fields. Some early failures can happen before the -splitting is done; in those cases use the original field. */ - -else - { - if (addr->local_part != NULL) - { - yield = string_get_localpart(addr, yield, &size, &ptr); - yield = string_cat(yield, &size, &ptr, US"@", 1); - yield = string_cat(yield, &size, &ptr, addr->domain, - Ustrlen(addr->domain) ); - } - else - { - yield = string_cat(yield, &size, &ptr, addr->address, Ustrlen(addr->address)); - } - yield[ptr] = 0; - - /* If the address we are going to print is the same as the top address, - and all parents are not being included, don't add on the top address. First - of all, do a caseless comparison; if this succeeds, do a caseful comparison - on the local parts. */ - - if (strcmpic(yield, topaddr->address) == 0 && - Ustrncmp(yield, topaddr->address, Ustrchr(yield, '@') - yield) == 0 && - addr->onetime_parent == NULL && - (!all_parents || addr->parent == NULL || addr->parent == topaddr)) - add_topaddr = FALSE; - } - -/* If all parents are requested, or this is a local pipe/file/reply, and -there is at least one intermediate parent, show it in brackets, and continue -with all of them if all are wanted. */ - -if ((all_parents || testflag(addr, af_pfr)) && - addr->parent != NULL && - addr->parent != topaddr) - { - uschar *s = US" ("; - address_item *addr2; - for (addr2 = addr->parent; addr2 != topaddr; addr2 = addr2->parent) - { - yield = string_cat(yield, &size, &ptr, s, 2); - yield = string_cat(yield, &size, &ptr, addr2->address, Ustrlen(addr2->address)); - if (!all_parents) break; - s = US", "; - } - yield = string_cat(yield, &size, &ptr, US")", 1); - } - -/* Add the top address if it is required */ - -if (add_topaddr) - { - yield = string_cat(yield, &size, &ptr, US" <", 2); - - if (addr->onetime_parent == NULL) - yield = string_cat(yield, &size, &ptr, topaddr->address, - Ustrlen(topaddr->address)); - else - yield = string_cat(yield, &size, &ptr, addr->onetime_parent, - Ustrlen(addr->onetime_parent)); - - yield = string_cat(yield, &size, &ptr, US">", 1); - } - -yield[ptr] = 0; /* string_cat() leaves space */ -return yield; +return Ustrcmp(* CUSS a, * CUSS b); } -#endif /* COMPILE_UTILITY */ - +#endif /* COMPILE_UTILITY */ @@ -1472,8 +1728,10 @@ printf("Testing string_format\n"); while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) { void *args[3]; + long long llargs[3]; double dargs[3]; int dflag = 0; + int llflag = 0; int n = 0; int count; int countset = 0; @@ -1504,6 +1762,11 @@ while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) dflag = 1; dargs[n++] = Ustrtod(outbuf, NULL); } + else if (Ustrstr(outbuf, "ll") != NULL) + { + llflag = 1; + llargs[n++] = strtoull(CS outbuf, NULL, 10); + } else { args[n++] = (void *)Uatoi(outbuf); @@ -1526,11 +1789,16 @@ while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) if (*s == ',') s++; } - if (!dflag) printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, - args[0], args[1], args[2])? "True" : "False"); + if (!dflag && !llflag) + printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, + args[0], args[1], args[2])? "True" : "False"); + + else if (dflag) + printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, + dargs[0], dargs[1], dargs[2])? "True" : "False"); else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, - dargs[0], dargs[1], dargs[2])? "True" : "False"); + llargs[0], llargs[1], llargs[2])? "True" : "False"); printf("%s\n", CS outbuf); if (countset) printf("count=%d\n", count);