[Openvpn-devel,1/4] re-implement argv_printf_*()

Message ID 20181019155638.9498-2-davids@openvpn.net
State New
Headers show
Series
  • struct argv overhaul - round 2+x
Related show

Commit Message

David Sommerseth Oct. 19, 2018, 3:56 p.m.
From: Heiko Hund <heiko.hund@sophos.com>

The previous implementation had the problem that it was not fully
compatible with printf() and could only detect % format directives
following a space character (0x20).

It modifies the format string and inserts marks to separate groups
before passing it to the regular printf in libc. The marks are
later used to separate the output string into individual command
line arguments.

The choice of 0x1D as the argument delimiter is based on the
assumption that no "regular" string passed to argv_printf_*() will
ever have to contain that byte (and the fact that it actually is
the ASCII "group separator" control character, which fits its
purpose).

This commit has been updated by David Sommerseth based on his feedback
on the mailing list discussions earlier on.

Signed-off-by: Heiko Hund <heiko.hund@sophos.com>
Signed-off-by: David Sommerseth <davids@openvpn.net>
---
 src/openvpn/argv.c                   | 266 ++++++++++++---------------
 src/openvpn/argv.h                   |   4 +-
 src/openvpn/route.c                  |   8 +-
 src/openvpn/tun.c                    |  24 +--
 tests/unit_tests/openvpn/test_argv.c |  58 +++++-
 5 files changed, 192 insertions(+), 168 deletions(-)

Comments

David Sommerseth Oct. 24, 2018, 10:11 a.m. | #1
On 19/10/18 17:56, David Sommerseth wrote:
> From: Heiko Hund <heiko.hund@sophos.com>
> 
> The previous implementation had the problem that it was not fully
> compatible with printf() and could only detect % format directives
> following a space character (0x20).
> 
> It modifies the format string and inserts marks to separate groups
> before passing it to the regular printf in libc. The marks are
> later used to separate the output string into individual command
> line arguments.
> 
> The choice of 0x1D as the argument delimiter is based on the
> assumption that no "regular" string passed to argv_printf_*() will
> ever have to contain that byte (and the fact that it actually is
> the ASCII "group separator" control character, which fits its
> purpose).
> 
> This commit has been updated by David Sommerseth based on his feedback
> on the mailing list discussions earlier on.
> 
> Signed-off-by: Heiko Hund <heiko.hund@sophos.com>
> Signed-off-by: David Sommerseth <davids@openvpn.net>
> ---
>  src/openvpn/argv.c                   | 266 ++++++++++++---------------
>  src/openvpn/argv.h                   |   4 +-
>  src/openvpn/route.c                  |   8 +-
>  src/openvpn/tun.c                    |  24 +--
>  tests/unit_tests/openvpn/test_argv.c |  58 +++++-
>  5 files changed, 192 insertions(+), 168 deletions(-)

This is a rebase of [0,1] on top of git master 2b15c11716e0d.  I've added my
review suggestions directly to this commit, as can be found here [2].  The
doxygen suggestion is coming as the last patch in this patch-set.

[0] <https://patchwork.openvpn.net/patch/71/>
[1]
<https://www.mail-archive.com/openvpn-devel@lists.sourceforge.net/msg15834.html>
[2]
<https://www.mail-archive.com/openvpn-devel@lists.sourceforge.net/msg15868.html>

Patch

diff --git a/src/openvpn/argv.c b/src/openvpn/argv.c
index 9100a196..9606f382 100644
--- a/src/openvpn/argv.c
+++ b/src/openvpn/argv.c
@@ -131,64 +131,6 @@  argv_insert_head(const struct argv *a, const char *head)
     return r;
 }
 
-static char *
-argv_term(const char **f)
-{
-    const char *p = *f;
-    const char *term = NULL;
-    size_t termlen = 0;
-
-    if (*p == '\0')
-    {
-        return NULL;
-    }
-
-    while (true)
-    {
-        const int c = *p;
-        if (c == '\0')
-        {
-            break;
-        }
-        if (term)
-        {
-            if (!isspace(c))
-            {
-                ++termlen;
-            }
-            else
-            {
-                break;
-            }
-        }
-        else
-        {
-            if (!isspace(c))
-            {
-                term = p;
-                termlen = 1;
-            }
-        }
-        ++p;
-    }
-    *f = p;
-
-    if (term)
-    {
-        char *ret;
-        ASSERT(termlen > 0);
-        ret = malloc(termlen + 1);
-        check_malloc_return(ret);
-        memcpy(ret, term, termlen);
-        ret[termlen] = '\0';
-        return ret;
-    }
-    else
-    {
-        return NULL;
-    }
-}
-
 const char *
 argv_str(const struct argv *a, struct gc_arena *gc, const unsigned int flags)
 {
@@ -218,119 +160,151 @@  argv_msg_prefix(const int msglev, const struct argv *a, const char *prefix)
     gc_free(&gc);
 }
 
-static void
+
+/*
+ * argv_prep_format - prepare argv format string for further processing
+ *
+ * Individual argument must be separated by space. Ignores leading and trailing spaces.
+ * Consecutive spaces count as one. Returns prepared format string, with space replaced
+ * by delim and adds the number of arguments to the count parameter.
+ */
+static char *
+argv_prep_format(const char *format, const char delim, size_t *count, struct gc_arena *gc)
+{
+    int i, j;
+    char *f;
+    bool in_token = false;
+
+    if (format == NULL)
+    {
+        return NULL;
+    }
+
+    f = gc_malloc(strlen(format) + 1, true, gc);
+    for (i = 0, j = 0; i < strlen(format); i++)
+    {
+        if (format[i] == ' ')
+        {
+            in_token = false;
+            continue;
+        }
+
+        if (!in_token)
+        {
+            ++*count;
+
+            /*
+             * We don't add any delimiter to the resulting
+             * format string before something has been added there.
+             * The resulting format string will never start with a delimiter.
+             */
+            if (j > 0)
+            {
+                f[j++] = delim;
+            }
+        }
+
+        f[j++] = format[i];
+        in_token = true;
+    }
+
+    return f;
+}
+
+/*
+ * argv_printf_arglist - create a struct argv from a format string
+ *
+ * Instead of parsing the format string ourselves place delimiters via argv_prep_format()
+ * before we let libc's printf() do the parsing. Then split the resulting string at the
+ * injected delimiters.
+ */
+static bool
 argv_printf_arglist(struct argv *a, const char *format, va_list arglist)
 {
-    char *term;
-    const char *f = format;
+    struct gc_arena gc = gc_new();
+    const char delim = 0x1D;  /* ASCII Group Separator (GS) */
+    char *f, *buf, *end;
+    size_t argc, size;
+    bool res = false;
+    va_list tmplist;
+    int len;
 
     argv_extend(a, 1); /* ensure trailing NULL */
 
-    while ((term = argv_term(&f)) != NULL)
+    argc = a->argc;
+    f = argv_prep_format(format, delim, &argc, &gc);
+    if (f == NULL)
     {
-        if (term[0] == '%')
-        {
-            if (!strcmp(term, "%s"))
-            {
-                char *s = va_arg(arglist, char *);
-                if (!s)
-                {
-                    s = "";
-                }
-                argv_append(a, string_alloc(s, NULL));
-            }
-            else if (!strcmp(term, "%d"))
-            {
-                char numstr[64];
-                openvpn_snprintf(numstr, sizeof(numstr), "%d", va_arg(arglist, int));
-                argv_append(a, string_alloc(numstr, NULL));
-            }
-            else if (!strcmp(term, "%u"))
-            {
-                char numstr[64];
-                openvpn_snprintf(numstr, sizeof(numstr), "%u", va_arg(arglist, unsigned int));
-                argv_append(a, string_alloc(numstr, NULL));
-            }
-            else if (!strcmp(term, "%lu"))
-            {
-                char numstr[64];
-                openvpn_snprintf(numstr, sizeof(numstr), "%lu",
-                                 va_arg(arglist, unsigned long));
-                argv_append(a, string_alloc(numstr, NULL));
-            }
-            else if (!strcmp(term, "%s/%d"))
-            {
-                char numstr[64];
-                char *s = va_arg(arglist, char *);
-
-                if (!s)
-                {
-                    s = "";
-                }
+        goto out;
+    }
 
-                openvpn_snprintf(numstr, sizeof(numstr), "%d", va_arg(arglist, int));
+    /* determine minimum buffer size */
+    va_copy(tmplist, arglist);
+    len = vsnprintf(NULL, 0, f, tmplist);
+    va_end(tmplist);
+    if (len < 0)
+    {
+        goto out;
+    }
 
-                {
-                    const size_t len = strlen(s) + strlen(numstr) + 2;
-                    char *combined = (char *) malloc(len);
-                    check_malloc_return(combined);
+    size = adjust_power_of_2(len + 1);
+    buf = gc_malloc(size, false, &gc);
+    len = vsnprintf(buf, size, f, arglist);
+    if (len < 0 || len >= size)
+    {
+        goto out;
+    }
 
-                    strcpy(combined, s);
-                    strcat(combined, "/");
-                    strcat(combined, numstr);
-                    argv_append(a, combined);
-                }
-            }
-            else if (!strcmp(term, "%s%sc"))
-            {
-                char *s1 = va_arg(arglist, char *);
-                char *s2 = va_arg(arglist, char *);
-                char *combined;
+    /* split the string at the delimiters */
+    end = strchr(buf, delim);
+    while (end)
+    {
+        *end = '\0';
+        argv_append(a, string_alloc(buf, NULL));
+        buf = end + 1;
+        end = strchr(buf, delim);
+    }
+    argv_append(a, string_alloc(buf, NULL));
 
-                if (!s1)
-                {
-                    s1 = "";
-                }
-                if (!s2)
-                {
-                    s2 = "";
-                }
-                combined = (char *) malloc(strlen(s1) + strlen(s2) + 1);
-                check_malloc_return(combined);
-                strcpy(combined, s1);
-                strcat(combined, s2);
-                argv_append(a, combined);
-            }
-            else
-            {
-                ASSERT(0);
-            }
-            free(term);
-        }
-        else
-        {
-            argv_append(a, term);
-        }
+    if (a->argc != argc)
+    {
+        /* Someone snuck in a GS (0x1D), fail gracefully */
+        argv_reset(a);
+        argv_extend(a, 1); /* ensure trailing NULL */
+        goto out;
     }
+
+    res = true;
+
+out:
+    gc_free(&gc);
+    return res;
 }
 
-void
+
+
+bool
 argv_printf(struct argv *a, const char *format, ...)
 {
+    bool res;
     va_list arglist;
-    argv_reset(a);
+
     va_start(arglist, format);
-    argv_printf_arglist(a, format, arglist);
+    argv_reset(a);
+    res = argv_printf_arglist(a, format, arglist);
     va_end(arglist);
+    return res;
 }
 
-void
+bool
 argv_printf_cat(struct argv *a, const char *format, ...)
 {
+    bool res;
     va_list arglist;
     va_start(arglist, format);
-    argv_printf_arglist(a, format, arglist);
+    res = argv_printf_arglist(a, format, arglist);
     va_end(arglist);
+    return res;
 }
 
 void
diff --git a/src/openvpn/argv.h b/src/openvpn/argv.h
index 9d9f3873..b9105a43 100644
--- a/src/openvpn/argv.h
+++ b/src/openvpn/argv.h
@@ -52,7 +52,7 @@  void argv_msg_prefix(const int msglev, const struct argv *a, const char *prefix)
 
 void argv_parse_cmd(struct argv *a, const char *s);
 
-void argv_printf(struct argv *a, const char *format, ...)
+bool argv_printf(struct argv *a, const char *format, ...)
 #ifdef __GNUC__
 #if __USE_MINGW_ANSI_STDIO
 __attribute__ ((format(gnu_printf, 2, 3)))
@@ -62,7 +62,7 @@  __attribute__ ((format(__printf__, 2, 3)))
 #endif
 ;
 
-void argv_printf_cat(struct argv *a, const char *format, ...)
+bool argv_printf_cat(struct argv *a, const char *format, ...)
 #ifdef __GNUC__
 #if __USE_MINGW_ANSI_STDIO
 __attribute__ ((format(gnu_printf, 2, 3)))
diff --git a/src/openvpn/route.c b/src/openvpn/route.c
index d97e8dba..9649d197 100644
--- a/src/openvpn/route.c
+++ b/src/openvpn/route.c
@@ -1614,7 +1614,7 @@  add_route(struct route_ipv4 *r,
 #elif defined (_WIN32)
     {
         DWORD ai = TUN_ADAPTER_INDEX_INVALID;
-        argv_printf(&argv, "%s%sc ADD %s MASK %s %s",
+        argv_printf(&argv, "%s%s ADD %s MASK %s %s",
                     get_win_sys_path(),
                     WIN_ROUTE_PATH_SUFFIX,
                     network,
@@ -1979,7 +1979,7 @@  add_route_ipv6(struct route_ipv6 *r6, const struct tuntap *tt, unsigned int flag
         device = buf_bptr(&out);
 
         /* netsh interface ipv6 add route 2001:db8::/32 MyTunDevice */
-        argv_printf(&argv, "%s%sc interface ipv6 add route %s/%d %s",
+        argv_printf(&argv, "%s%s interface ipv6 add route %s/%d %s",
                     get_win_sys_path(),
                     NETSH_PATH_SUFFIX,
                     network,
@@ -2186,7 +2186,7 @@  delete_route(struct route_ipv4 *r,
 
 #elif defined (_WIN32)
 
-    argv_printf(&argv, "%s%sc DELETE %s MASK %s %s",
+    argv_printf(&argv, "%s%s DELETE %s MASK %s %s",
                 get_win_sys_path(),
                 WIN_ROUTE_PATH_SUFFIX,
                 network,
@@ -2426,7 +2426,7 @@  delete_route_ipv6(const struct route_ipv6 *r6, const struct tuntap *tt, unsigned
         device = buf_bptr(&out);
 
         /* netsh interface ipv6 delete route 2001:db8::/32 MyTunDevice */
-        argv_printf(&argv, "%s%sc interface ipv6 delete route %s/%d %s",
+        argv_printf(&argv, "%s%s interface ipv6 delete route %s/%d %s",
                     get_win_sys_path(),
                     NETSH_PATH_SUFFIX,
                     network,
diff --git a/src/openvpn/tun.c b/src/openvpn/tun.c
index 948fd17d..f80aedf8 100644
--- a/src/openvpn/tun.c
+++ b/src/openvpn/tun.c
@@ -994,7 +994,7 @@  do_ifconfig_ipv6(struct tuntap *tt, const char *ifname, int tun_mtu,
 
         openvpn_snprintf(iface, sizeof(iface), "interface=%lu",
                          tt->adapter_index);
-        argv_printf(&argv, "%s%sc interface ipv6 set address %s %s store=active",
+        argv_printf(&argv, "%s%s interface ipv6 set address %s %s store=active",
                     get_win_sys_path(), NETSH_PATH_SUFFIX, iface,
                     ifconfig_ipv6_local);
         netsh_command(&argv, 4, M_FATAL);
@@ -4873,14 +4873,14 @@  ipconfig_register_dns(const struct env_set *es)
     msg(D_TUNTAP_INFO, "Start ipconfig commands for register-dns...");
     netcmd_semaphore_lock();
 
-    argv_printf(&argv, "%s%sc /flushdns",
+    argv_printf(&argv, "%s%s /flushdns",
                 get_win_sys_path(),
                 WIN_IPCONFIG_PATH_SUFFIX);
     argv_msg(D_TUNTAP_INFO, &argv);
     status = openvpn_execve_check(&argv, es, 0, err);
     argv_reset(&argv);
 
-    argv_printf(&argv, "%s%sc /registerdns",
+    argv_printf(&argv, "%s%s /registerdns",
                 get_win_sys_path(),
                 WIN_IPCONFIG_PATH_SUFFIX);
     argv_msg(D_TUNTAP_INFO, &argv);
@@ -4994,8 +4994,8 @@  netsh_set_dns6_servers(const struct in6_addr *addr_list,
     for (int i = 0; i < addr_len; ++i)
     {
         const char *fmt = (i == 0) ?
-                          "%s%sc interface ipv6 set dns %s static %s"
-                          : "%s%sc interface ipv6 add dns %s %s";
+                          "%s%s interface ipv6 set dns %s static %s"
+                          : "%s%s interface ipv6 add dns %s %s";
         argv_printf(&argv, fmt, get_win_sys_path(),
                     NETSH_PATH_SUFFIX, flex_name,
                     print_in6_addr(addr_list[i], 0, &gc));
@@ -5042,7 +5042,7 @@  netsh_ifconfig_options(const char *type,
     /* delete existing DNS/WINS settings from TAP interface */
     if (delete_first)
     {
-        argv_printf(&argv, "%s%sc interface ip delete %s %s all",
+        argv_printf(&argv, "%s%s interface ip delete %s %s all",
                     get_win_sys_path(),
                     NETSH_PATH_SUFFIX,
                     type,
@@ -5059,8 +5059,8 @@  netsh_ifconfig_options(const char *type,
             if (delete_first || !test_first || !ip_addr_member_of(addr_list[i], current))
             {
                 const char *fmt = count ?
-                                  "%s%sc interface ip add %s %s %s"
-                                  : "%s%sc interface ip set %s %s static %s";
+                                  "%s%s interface ip add %s %s %s"
+                                  : "%s%s interface ip set %s %s static %s";
 
                 argv_printf(&argv, fmt,
                             get_win_sys_path(),
@@ -5136,7 +5136,7 @@  netsh_ifconfig(const struct tuntap_options *to,
         else
         {
             /* example: netsh interface ip set address my-tap static 10.3.0.1 255.255.255.0 */
-            argv_printf(&argv, "%s%sc interface ip set address %s static %s %s",
+            argv_printf(&argv, "%s%s interface ip set address %s static %s %s",
                         get_win_sys_path(),
                         NETSH_PATH_SUFFIX,
                         flex_name,
@@ -5185,7 +5185,7 @@  netsh_enable_dhcp(const struct tuntap_options *to,
 
     /* example: netsh interface ip set address my-tap dhcp */
     argv_printf(&argv,
-                "%s%sc interface ip set address %s dhcp",
+                "%s%s interface ip set address %s dhcp",
                 get_win_sys_path(),
                 NETSH_PATH_SUFFIX,
                 actual_name);
@@ -6100,7 +6100,7 @@  close_tun(struct tuntap *tt)
             /* netsh interface ipv6 delete address \"%s\" %s */
             ifconfig_ipv6_local = print_in6_addr(tt->local_ipv6, 0,  &gc);
             argv_printf(&argv,
-                        "%s%sc interface ipv6 delete address %s %s store=active",
+                        "%s%s interface ipv6 delete address %s %s store=active",
                         get_win_sys_path(),
                         NETSH_PATH_SUFFIX,
                         tt->actual_name,
@@ -6112,7 +6112,7 @@  close_tun(struct tuntap *tt)
             if (tt->options.dns6_len > 0)
             {
                 argv_printf(&argv,
-                            "%s%sc interface ipv6 delete dns %s all",
+                            "%s%s interface ipv6 delete dns %s all",
                             get_win_sys_path(),
                             NETSH_PATH_SUFFIX,
                             tt->actual_name);
diff --git a/tests/unit_tests/openvpn/test_argv.c b/tests/unit_tests/openvpn/test_argv.c
index 0fdd3f0a..9b67fade 100644
--- a/tests/unit_tests/openvpn/test_argv.c
+++ b/tests/unit_tests/openvpn/test_argv.c
@@ -9,6 +9,7 @@ 
 #include <setjmp.h>
 #include <cmocka.h>
 #include <assert.h>
+#include <stdbool.h>
 
 #include "argv.h"
 #include "buffer.h"
@@ -53,23 +54,69 @@  argv_printf_cat__multiple_spaces_in_format__parsed_as_one(void **state)
     argv_reset(&a);
 }
 
+static void
+argv_printf__embedded_format_directive__replaced_in_output(void **state)
+{
+    struct argv a = argv_new();
+
+    argv_printf(&a, "<p1:%s>", PATH1);
+    assert_int_equal(a.argc, 1);
+    assert_string_equal(a.argv[0], "<p1:" PATH1 ">");
+
+    argv_reset(&a);
+}
+
+static void
+argv_printf__group_sep_in_arg__fail_no_ouput(void **state)
+{
+    struct argv a = argv_new();
+
+    assert_false(argv_printf(&a, "tool --do %s", "this\035--harmful"));
+    assert_int_equal(a.argc, 0);
+
+    argv_reset(&a);
+}
+
 static void
 argv_printf__combined_path_with_spaces__argc_correct(void **state)
 {
     struct argv a = argv_new();
 
-    argv_printf(&a, "%s%sc", PATH1, PATH2);
+    argv_printf(&a, "%s%s", PATH1, PATH2);
     assert_int_equal(a.argc, 1);
 
-    argv_printf(&a, "%s%sc %d", PATH1, PATH2, 42);
+    argv_printf(&a, "%s%s %d", PATH1, PATH2, 42);
     assert_int_equal(a.argc, 2);
 
-    argv_printf(&a, "foo %s%sc %s x y", PATH2, PATH1, "foo");
+    argv_printf(&a, "foo %s%s %s x y", PATH2, PATH1, "foo");
     assert_int_equal(a.argc, 5);
 
     argv_reset(&a);
 }
 
+static void
+argv_printf__empty_parameter__argc_correct(void **state)
+{
+    struct argv a = argv_new();
+
+    argv_printf(&a, "%s", "");
+    assert_int_equal(a.argc, 1);
+
+    argv_printf(&a, "%s %s", PATH1, "");
+    assert_int_equal(a.argc, 2);
+
+    argv_printf(&a, "%s %s %s", PATH1, "", PARAM1);
+    assert_int_equal(a.argc, 3);
+
+    argv_printf(&a, "%s %s %s %s", PATH1, "", "", PARAM1);
+    assert_int_equal(a.argc, 4);
+
+    argv_printf(&a, "%s %s", "", PARAM1);
+    assert_int_equal(a.argc, 2);
+
+    argv_free(&a);
+}
+
 static void
 argv_parse_cmd__command_string__argc_correct(void **state)
 {
@@ -113,7 +160,7 @@  argv_str__multiple_argv__correct_output(void **state)
     struct gc_arena gc = gc_new();
     const char *output;
 
-    argv_printf(&a, "%s%sc", PATH1, PATH2);
+    argv_printf(&a, "%s%s", PATH1, PATH2);
     argv_printf_cat(&a, "%s", PARAM1);
     argv_printf_cat(&a, "%s", PARAM2);
     argv_printf_cat(&a, "%d", -1);
@@ -172,7 +219,10 @@  main(void)
     const struct CMUnitTest tests[] = {
         cmocka_unit_test(argv_printf__multiple_spaces_in_format__parsed_as_one),
         cmocka_unit_test(argv_printf_cat__multiple_spaces_in_format__parsed_as_one),
+        cmocka_unit_test(argv_printf__embedded_format_directive__replaced_in_output),
+        cmocka_unit_test(argv_printf__group_sep_in_arg__fail_no_ouput),
         cmocka_unit_test(argv_printf__combined_path_with_spaces__argc_correct),
+        cmocka_unit_test(argv_printf__empty_parameter__argc_correct),
         cmocka_unit_test(argv_parse_cmd__command_string__argc_correct),
         cmocka_unit_test(argv_parse_cmd__command_and_extra_options__argc_correct),
         cmocka_unit_test(argv_printf_cat__used_twice__argc_correct),