* Exim - an Internet mail transport agent *
*************************************************/
-/* Copyright (c) University of Cambridge 1995 - 2016 */
+/* Copyright (c) University of Cambridge 1995 - 2018 */
+/* Copyright (c) The Exim Maintainers 2020 - 2021 */
/* See the file NOTICE for conditions of use and distribution. */
/* Miscellaneous string-handling functions. Some are not required for
#include "exim.h"
+#include <assert.h>
#ifndef COMPILE_UTILITY
int
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
{
BOOL had_double_colon = FALSE;
BOOL v4end = FALSE;
- int count = 0;
yield = 6;
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
/* 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;
uschar *
string_format_size(int size, uschar *buffer)
{
-if (size == 0) Ustrcpy(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);
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 ch;
const uschar *p = *pp;
ch = *(++p);
+if (ch == '\0') return **pp;
if (isdigit(ch) && ch != '8' && ch != '9')
{
ch -= '0';
/* This function is called for critical strings. It checks for any
non-printing characters, and if any are found, it makes a new copy
of the string with suitable escape sequences. It is most often called by the
-macro string_printing(), which sets allow_tab TRUE.
+macro string_printing(), which sets flags to 0.
Arguments:
s the input string
- allow_tab TRUE to allow tab as a printing character
+ flags Bit 0: convert tabs. Bit 1: convert spaces.
Returns: string with non-printers encoded as printing sequences
*/
const uschar *
-string_printing2(const uschar *s, BOOL allow_tab)
+string_printing2(const uschar *s, int flags)
{
int nonprintcount = 0;
int length = 0;
while (*t != 0)
{
int c = *t++;
- if (!mac_isprint(c) || (!allow_tab && c == '\t')) nonprintcount++;
+ if ( !mac_isprint(c)
+ || flags & SP_TAB && c == '\t'
+ || flags & SP_SPACE && c == ' '
+ ) nonprintcount++;
length++;
}
/* Get a new block of store guaranteed big enough to hold the
expanded string. */
-ss = store_get(length + nonprintcount * 3 + 1);
+tt = ss = store_get(length + nonprintcount * 3 + 1, is_tainted(s));
/* Copy everything, escaping non printers. */
-t = s;
-tt = ss;
-
-while (*t != 0)
+for (t = s; *t; )
{
int c = *t;
- if (mac_isprint(c) && (allow_tab || c != '\t')) *tt++ = *t++; else
+ if ( mac_isprint(c)
+ && (!(flags & SP_TAB) || c != '\t')
+ && (!(flags & SP_SPACE) || c != ' ')
+ )
+ *tt++ = *t++;
+ else
{
*tt++ = '\\';
switch (*t)
if (!p) return s;
len = Ustrlen(s) + 1;
-ss = store_get(len);
+ss = store_get(len, is_tainted(s));
q = ss;
off = p - s;
+#if (defined(HAVE_LOCAL_SCAN) || defined(EXPAND_DLFUNC)) \
+ && !defined(MACRO_PREDEF) && !defined(COMPILE_UTILITY)
/*************************************************
* Copy and save string *
*************************************************/
-/* This function assumes that memcpy() is faster than strcpy().
-
+/*
Argument: string to copy
-Returns: copy of string in new store
+Returns: copy of string in new store with the same taint status
*/
uschar *
-string_copy(const uschar *s)
+string_copy_function(const uschar *s)
{
-int len = Ustrlen(s) + 1;
-uschar *ss = store_get(len);
-memcpy(ss, s, len);
-return ss;
+return string_copy_taint(s, is_tainted(s));
}
-
-
-/*************************************************
-* Copy and save string in malloc'd store *
-*************************************************/
-
/* This function assumes that memcpy() is faster than strcpy().
-
-Argument: string to copy
-Returns: copy of string in new store
+As above, but explicitly specifying the result taint status
*/
uschar *
-string_copy_malloc(const uschar *s)
+string_copy_taint_function(const uschar * s, BOOL tainted)
{
int len = Ustrlen(s) + 1;
-uschar *ss = store_malloc(len);
+uschar *ss = store_get(len, tainted);
memcpy(ss, s, len);
return ss;
}
-/*************************************************
-* Copy, lowercase and save string *
-*************************************************/
-
-/*
-Argument: string to copy
-Returns: copy of string in new store, with letters lowercased
-*/
-
-uschar *
-string_copylc(const uschar *s)
-{
-uschar *ss = store_get(Ustrlen(s) + 1);
-uschar *p = ss;
-while (*s != 0) *p++ = tolower(*s++);
-*p = 0;
-return ss;
-}
-
-
-
/*************************************************
* Copy and save string, given length *
*************************************************/
*/
uschar *
-string_copyn(const 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;
}
string_copy_dnsdomain(uschar *s)
{
uschar *yield;
-uschar *ss = yield = store_get(Ustrlen(s) + 1);
+uschar *ss = yield = store_get(Ustrlen(s) + 1, TRUE); /* always treat as tainted */
while (*s != 0)
{
if (*s != '\\')
- {
*ss++ = *s++;
- }
else if (isdigit(s[1]))
{
*ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0';
s += 4;
}
else if (*(++s) != 0)
- {
*ss++ = *s++;
- }
}
*ss = 0;
/* 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 */
* 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 *
*/
uschar *
-string_sprintf(const 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 " SIZE_T_FMT
- "; format string was (%s)\nexpansion started '%.32s'",
- sizeof(buffer), format, 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
}
+#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 *
separator a pointer to the separator character in an int (see above)
buffer where to put a copy of the next string in the list; or
NULL if the next string is returned in new memory
+ Note that if the list is tainted then a provided buffer must be
+ also (else we trap, with a message referencing the callsite).
+ If we do the allocation, taint is handled there.
buflen when buffer is not NULL, the size of buffer; otherwise ignored
Returns: pointer to buffer, containing the next substring,
*/
uschar *
-string_nextinlist(const uschar **listptr, int *separator, uschar *buffer, int buflen)
+string_nextinlist_trc(const uschar **listptr, int *separator, uschar *buffer, int buflen,
+ const uschar * func, int line)
{
int sep = *separator;
const uschar *s = *listptr;
BOOL sep_is_special;
-if (s == NULL) return NULL;
+if (!s) return NULL;
/* This allows for a fixed specified separator to be an iscntrl() character,
but at the time of implementation, this is never the case. However, it's best
if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1])))
{
sep = s[1];
- s += 2;
+ if (*++s) ++s;
while (isspace(*s) && *s != sep) s++;
}
else
- {
- sep = (sep == 0)? ':' : -sep;
- }
+ sep = sep ? -sep : ':';
*separator = sep;
}
/* An empty string has no list elements */
-if (*s == 0) return NULL;
+if (!*s) return NULL;
/* Note whether whether or not the separator is an iscntrl() character. */
/* Handle the case when a buffer is provided. */
-if (buffer != NULL)
+if (buffer)
{
int p = 0;
- for (; *s != 0; s++)
+ if (is_tainted(s) && !is_tainted(buffer))
+ die_tainted(US"string_nextinlist", func, line);
+ for (; *s; s++)
{
if (*s == sep && (*(++s) != sep || sep_is_special)) break;
if (p < buflen - 1) buffer[p++] = *s;
}
while (p > 0 && isspace(buffer[p-1])) p--;
- buffer[p] = 0;
+ buffer[p] = '\0';
}
/* Handle the case when a buffer is not provided. */
else
{
- int size = 0;
- int ptr = 0;
- const uschar *ss;
+ 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 (if an ispunct()
start of a string. Avoid getting working memory for an empty item. */
if (*s == sep)
- {
- s++;
- if (*s != sep || sep_is_special)
+ if (*++s != sep || sep_is_special)
{
*listptr = s;
return string_copy(US"");
}
- }
/* Not an empty string; the first character is guaranteed to be a data
character. */
for (;;)
{
- for (ss = s + 1; *ss != 0 && *ss != sep; ss++);
- buffer = string_catn(buffer, &size, &ptr, s, ss-s);
+ const uschar * ss;
+ for (ss = s + 1; *ss && *ss != sep; ) ss++;
+ g = string_catn(g, s, ss-s);
s = ss;
- if (*s == 0 || *(++s) != sep || sep_is_special) break;
+ if (!*s || *++s != sep || sep_is_special) break;
}
- while (ptr > 0 && isspace(buffer[ptr-1])) ptr--;
- buffer[ptr] = 0;
+
+ /* Trim trailing spaces from the returned string */
+
+ /* while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; */
+ while ( g->ptr > 0 && isspace(g->s[g->ptr-1])
+ && (g->ptr == 1 || g->s[g->ptr-2] != '\\') )
+ g->ptr--;
+ buffer = string_from_gstring(g);
+ gstring_release_unused_trc(g, CCS func, line);
}
/* Update the current pointer and return the new string */
always returned null-terminated.
Arguments:
- list points to the start of the list that is being built, or NULL
+ list expanding-string for the list that is being built, or NULL
if this is a new list that has no contents yet
- sz (ptr to) amount of memory allocated for list; zero for a new list
- off (ptr to) current list length in chars (insert point for next addition),
- zero for a new list
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.
*/
-uschar *
-string_append_listele(uschar * list, int * sz, int * off,
- uschar sep, const uschar * ele)
+gstring *
+string_append_listele(gstring * list, uschar sep, const uschar * ele)
{
uschar * sp;
-if (list)
- list = string_catn(list, sz, off, &sep, 1);
+if (list && list->ptr)
+ list = string_catn(list, &sep, 1);
while((sp = Ustrchr(ele, sep)))
{
- list = string_catn(list, sz, off, ele, sp-ele+1);
- list = string_catn(list, sz, off, &sep, 1);
+ list = string_catn(list, ele, sp-ele+1);
+ list = string_catn(list, &sep, 1);
ele = sp+1;
}
-list = string_cat(list, sz, off, ele);
-list[*off] = '\0';
+list = string_cat(list, ele);
+(void) string_from_gstring(list);
return list;
}
-uschar *
-string_append_listele_n(uschar * list, int * sz, int * off,
- uschar sep, const uschar * ele, unsigned len)
+gstring *
+string_append_listele_n(gstring * list, uschar sep, const uschar * ele,
+ unsigned len)
{
const uschar * sp;
-if (list)
- list = string_catn(list, sz, off, &sep, 1);
+if (list && list->ptr)
+ list = string_catn(list, &sep, 1);
while((sp = Ustrnchr(ele, sep, &len)))
{
- list = string_catn(list, sz, off, ele, sp-ele+1);
- list = string_catn(list, sz, off, &sep, 1);
+ list = string_catn(list, ele, sp-ele+1);
+ list = string_catn(list, &sep, 1);
ele = sp+1;
len--;
}
-list = string_catn(list, sz, off, ele, len);
-list[*off] = '\0';
+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 (g->ptr < 0 || g->ptr > g->size || g->size >= INT_MAX/2)
+ log_write(0, LOG_MAIN|LOG_PANIC_DIE,
+ "internal error in gstring_grow (ptr %d size %d)", g->ptr, g->size);
+
+if (count <= 0) return;
+
+if (count >= INT_MAX/2 - g->ptr)
+ log_write(0, LOG_MAIN|LOG_PANIC_DIE,
+ "internal error in gstring_grow (ptr %d count %d)", g->ptr, count);
+
+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);
+}
+
+
+
/*************************************************
* 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
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.
-
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
*/
/* coverity[+alloc] */
-uschar *
-string_catn(uschar *string, int *size, int *ptr, const uschar *s, int count)
+gstring *
+string_catn(gstring * g, const uschar *s, int count)
{
-int p = *ptr;
-
-if (p + count >= *size)
- {
- 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 p;
+BOOL srctaint = is_tainted(s);
- int inc = (oldsize < 4096)? 100 : 1024;
- while (*size <= p + count) *size += inc;
+if (count < 0)
+ log_write(0, LOG_MAIN|LOG_PANIC_DIE,
+ "internal error in string_catn (count %d)", count);
- /* New string */
+if (!g)
+ {
+ unsigned inc = count < 4096 ? 127 : 1023;
+ unsigned size = ((count + inc) & ~inc) + 1; /* round up requested count */
+ g = string_get_tainted(size, srctaint);
+ }
+else if (srctaint && !is_tainted(g->s))
+ gstring_rebuffer(g);
- if (string == NULL) string = store_get(*size);
+if (g->ptr < 0 || g->ptr > g->size)
+ log_write(0, LOG_MAIN|LOG_PANIC_DIE,
+ "internal error in string_catn (ptr %d size %d)", g->ptr, g->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;
- }
- }
+p = g->ptr;
+if (count >= g->size - p)
+ 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.
+latter has to check for zero bytes. */
-The Coverity annotation deals with the lack of correlated variable tracking;
-common use is a null string and zero size and pointer, on first use for a
-string being built. The "if" above then allocates, but Coverity assume that
-the "if" might not happen and whines for a null-deref done by the memcpy(). */
-
-/* coverity[deref_parm_field_in_call] : FALSE */
-memcpy(string + p, s, count);
-*ptr = p + count;
-return string;
+memcpy(g->s + p, s, count);
+g->ptr = p + count;
+return g;
}
-uschar *
-string_cat(uschar *string, int *size, int *ptr, const uschar *s)
+gstring *
+string_cat(gstring *string, const uschar *s)
{
-return string_catn(string, size, ptr, s, Ustrlen(s));
+return string_catn(string, s, Ustrlen(s));
}
-#endif /* COMPILE_UTILITY */
-#ifndef COMPILE_UTILITY
/*************************************************
* Append strings to another 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
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);
+ string = string_cat(string, t);
}
va_end(ap);
The formats are the usual printf() ones, with some omissions (never used) and
three additions for strings: %S forces lower case, %T forces upper case, and
-%#s or %#S prints nothing for a NULL string. Without thr # "NULL" is printed
+%#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.
*/
BOOL
-string_format(uschar *buffer, int buflen, const 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, const char *format, va_list ap)
+
+
+/* Build or append to a growing-string, sprintf-style.
+
+Arguments:
+ g a growable-string
+ func called-from function name, for debug
+ line called-from file line number, for debug
+ limit maximum string size
+ flags see below
+ format printf-like format string
+ ap variable-args pointer
+
+Flags:
+ SVFMT_EXTEND buffer can be created or exteded as needed
+ SVFMT_REBUFFER buffer can be recopied to tainted mem as needed
+ SVFMT_TAINT_NOCHK do not check inputs for taint
+
+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 limit argument.
+
+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)
{
-/* We assume numbered ascending order, C does not guarantee that */
-enum { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 };
+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;
-BOOL yield = TRUE;
-int width, precision;
-const char *fp = format; /* Deliberately not unsigned */
-uschar *p = buffer;
-uschar *last = buffer + buflen - 1;
+string_datestamp_offset = -1; /* Datestamp not inserted */
+string_datestamp_length = 0; /* Datestamp not inserted */
+string_datestamp_type = 0; /* Datestamp not inserted */
-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*/
+
+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;
- const char *null = "NULL"; /* ) These variables */
- const 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;
}
}
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', 'll' and 'z', remembering the item length */
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; }
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: p += sprintf(CS p, newformat, va_arg(ap, int)); break;
- case L_LONG: p += sprintf(CS p, newformat, va_arg(ap, long int)); break;
- case L_LONGLONG: p += sprintf(CS p, newformat, va_arg(ap, LONGLONG_T)); break;
- case L_SIZE: p += sprintf(CS p, newformat, va_arg(ap, size_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)");
}
break;
- case 'p':
- if (p >= last - 24) { yield = FALSE; goto END_FORMAT; }
- strncpy(newformat, item_start, fp - item_start);
- newformat[fp - item_start] = 0;
- p += sprintf(CS p, newformat, va_arg(ap, void *));
- break;
-
/* %f format is inherently insecure if the numbers that it may be
handed are unknown (e.g. 1e300). However, in Exim, %f is used for
printing load averages, and these are actually stored as integers
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)
- p += sprintf(CS p, newformat, va_arg(ap, long double));
- else
- p += sprintf(CS p, newformat, va_arg(ap, double));
- 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 daily datestamp for log file names */
- s = CS tod_stamp(tod_log_datestamp_daily);
- string_datestamp_offset = p - buffer; /* 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;
+ 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 = p - buffer; /* 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;
+ 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 */
case 'T': /* Forces *upper* case */
- s = va_arg(ap, char *);
-
- if (s == NULL) s = null;
- slen = Ustrlen(s);
+ 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
INSERT_STRING: /* Come to from %D or %M above */
- /* 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. */
-
- 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)
- {
- yield = FALSE;
- goto END_FORMAT;
- }
- if (p >= last - width)
- {
- yield = FALSE;
- width = precision = last - p - 1;
- if (width < 0) width = 0;
- if (precision < 0) precision = 0;
+ 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 if (fp[-1] == 'T')
- while (*p) { *p = toupper(*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;
}
"Permission denied", reads and includes the euid and egid.
Arguments:
- eno the value of errno after the failure
format a text format string - deliberately not uschar *
... arguments for the format string
*/
uschar *
-string_open_failed(int eno, const char *format, ...)
+string_open_failed_trc(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,
+ SVFMT_REBUFFER, format, ap);
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));
+g = string_catn(g, US": ", 2);
+g = string_cat(g, US strerror(errno));
+
+if (errno == EACCES)
+ {
+ int save_errno = errno;
+ g = string_fmt_append(g, " (euid=%ld egid=%ld)",
+ (long int)geteuid(), (long int)getegid());
+ errno = save_errno;
+ }
+gstring_release_unused(g);
+return string_from_gstring(g);
}
-#endif /* COMPILE_UTILITY */
-#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. */
+
/*************************************************
**************************************************
* Stand-alone test program *
uschar buffer[256];
printf("Testing is_ip_address\n");
+store_init();
while (fgets(CS buffer, sizeof(buffer), stdin) != NULL)
{
git://git.exim.org / exim.git / blobdiff