core: fix a crash in g_utf8_validate() (rh #862341)

This commit is contained in:
Jiří Klimeš 2012-11-27 11:51:39 +01:00
parent d73fc1c375
commit 77ef511f26
3 changed files with 207 additions and 1 deletions

View File

@ -4,7 +4,7 @@
Summary: Mobile broadband modem management service
Name: ModemManager
Version: 0.6.0.0
Release: 1%{snapshot}%{?dist}
Release: 2%{snapshot}%{?dist}
#
# Source from git://anongit.freedesktop.org/ModemManager/ModemManager
# tarball built with:
@ -30,6 +30,9 @@ BuildRequires: libxslt
Patch0: buildsys-hates-openpty.patch
Patch1: rh862341-1-UCS2-conversions.patch
Patch2: rh862341-2-parse_operator.patch
%description
The ModemManager service provides a consistent API to operate many different
modems, including mobile broadband (3G) devices.
@ -37,6 +40,8 @@ modems, including mobile broadband (3G) devices.
%prep
%setup -q
%patch0 -p1 -b .pty
%patch1 -p1 -b .UCS2-conversion
%patch2 -p1 -b .parse-operator
%build
@ -90,6 +95,9 @@ gtk-update-icon-cache %{_datadir}/icons/hicolor >&/dev/null || :
%{_datadir}/dbus-1/interfaces/*.xml
%changelog
* Tue Nov 27 2012 Jiří Klimeš <jklimes@redhat.com> - 0.6.0.0-2
- core: fix a crash in g_utf8_validate() (rh #862341)
* Tue Sep 4 2012 Dan Williams <dcbw@redhat.com> - 0.6.0.0-1
- Update to 0.6.0
- core: fix SMS notifications on many Qualcomm devices

View File

@ -0,0 +1,160 @@
commit e07c2162d66ab1a2e938db48744fc00afd29c272
Author: Dan Williams <dcbw@redhat.com>
Date: Tue Sep 11 16:36:16 2012 -0500
core: better handling of non-UCS2 conversions that should be UCS2 (bgo #683817)
Some modems return the +COPS operator name in hex-encoded current
character set (as set with +CSCS). Others return the operator name
in ASCII when set to UCS2, while yet others return the ASCII name
with trash at the end (*cough* Huawei *cough*). Handle that better
by not crashing.
diff --git a/src/mm-charsets.c b/src/mm-charsets.c
index 68bf2d1..65fc33c 100644
--- a/src/mm-charsets.c
+++ b/src/mm-charsets.c
@@ -710,8 +710,7 @@ gsm_pack (const guint8 *src,
* the hex representation of the charset-encoded string, so we need to cope with
* that case. */
gchar *
-mm_charset_take_and_convert_to_utf8 (gchar *str,
- MMModemCharset charset)
+mm_charset_take_and_convert_to_utf8 (gchar *str, MMModemCharset charset)
{
gchar *utf8 = NULL;
@@ -752,6 +751,7 @@ mm_charset_take_and_convert_to_utf8 (gchar *str,
case MM_MODEM_CHARSET_UCS2: {
gsize len;
gboolean possibly_hex = TRUE;
+ gsize bread = 0, bwritten = 0;
/* If the string comes in hex-UCS-2, len needs to be a multiple of 4 */
len = strlen (str);
@@ -765,19 +765,37 @@ mm_charset_take_and_convert_to_utf8 (gchar *str,
possibly_hex = isxdigit (*p++);
}
- /* If we get UCS-2, we expect the HEX representation of the string */
+ /* If hex, then we expect hex-encoded UCS-2 */
if (possibly_hex) {
utf8 = mm_modem_charset_hex_to_utf8 (str, charset);
- if (!utf8) {
- /* If we couldn't convert the string as HEX-UCS-2, try to see if
- * the string is valid UTF-8 itself. */
- utf8 = str;
- } else
+ if (utf8) {
g_free (str);
- } else
- /* If we already know it's not hex, try to use the string as it is */
- utf8 = str;
+ break;
+ }
+ }
+
+ /* If not hex, then it might be raw UCS-2 (very unlikely) or ASCII/UTF-8
+ * (much more likely). Try to convert to UTF-8 and if that fails, use
+ * the partial conversion length to re-convert the part of the string
+ * that is UTF-8, if any.
+ */
+ utf8 = g_convert (str, strlen (str),
+ "UTF-8//TRANSLIT", "UTF-8//TRANSLIT",
+ &bread, &bwritten, NULL);
+ /* Valid conversion, or we didn't get enough valid UTF-8 */
+ if (utf8 || (bwritten <= 2)) {
+ g_free (str);
+ break;
+ }
+
+ /* Last try; chop off the original string at the conversion failure
+ * location and get what we can.
+ */
+ str[bread] = '\0';
+ utf8 = g_convert (str, strlen (str),
+ "UTF-8//TRANSLIT", "UTF-8//TRANSLIT",
+ NULL, NULL, NULL);
break;
}
@@ -791,7 +809,7 @@ mm_charset_take_and_convert_to_utf8 (gchar *str,
/* Validate UTF-8 always before returning. This result will be exposed in DBus
* very likely... */
- if (!g_utf8_validate (utf8, -1, NULL)) {
+ if (utf8 && !g_utf8_validate (utf8, -1, NULL)) {
/* Better return NULL than an invalid UTF-8 string */
g_free (utf8);
utf8 = NULL;
diff --git a/src/mm-charsets.h b/src/mm-charsets.h
index 855387a..96435cc 100644
--- a/src/mm-charsets.h
+++ b/src/mm-charsets.h
@@ -72,7 +72,6 @@ guint8 *gsm_pack (const guint8 *src,
guint8 start_offset, /* in bits */
guint32 *out_packed_len);
-gchar *mm_charset_take_and_convert_to_utf8 (gchar *str,
- MMModemCharset charset);
+gchar *mm_charset_take_and_convert_to_utf8 (gchar *str, MMModemCharset charset);
#endif /* MM_CHARSETS_H */
diff --git a/src/tests/test-charsets.c b/src/tests/test-charsets.c
index f954d93..ced38a6 100644
--- a/src/tests/test-charsets.c
+++ b/src/tests/test-charsets.c
@@ -326,6 +326,41 @@ test_pack_gsm7_7_chars_offset (void *f, gpointer d)
g_free (packed);
}
+static void
+test_take_convert_ucs2_hex_utf8 (void *f, gpointer d)
+{
+ gchar *src, *converted;
+
+ /* Ensure hex-encoded UCS-2 works */
+ src = g_strdup ("0054002d004d006f00620069006c0065");
+ converted = mm_charset_take_and_convert_to_utf8 (src, MM_MODEM_CHARSET_UCS2);
+ g_assert_cmpstr (converted, ==, "T-Mobile");
+ g_free (converted);
+}
+
+static void
+test_take_convert_ucs2_bad_ascii (void *f, gpointer d)
+{
+ gchar *src, *converted;
+
+ /* Test that something mostly ASCII returns most of the original string */
+ src = g_strdup ("Orange\241");
+ converted = mm_charset_take_and_convert_to_utf8 (src, MM_MODEM_CHARSET_UCS2);
+ g_assert_cmpstr (converted, ==, "Orange");
+ g_free (converted);
+}
+
+static void
+test_take_convert_ucs2_bad_ascii2 (void *f, gpointer d)
+{
+ gchar *src, *converted;
+
+ /* Ensure something completely screwed up doesn't crash */
+ src = g_strdup ("\241\255\254\250\244\234");
+ converted = mm_charset_take_and_convert_to_utf8 (src, MM_MODEM_CHARSET_UCS2);
+ g_assert (converted == NULL);
+}
+
#if GLIB_CHECK_VERSION(2,25,12)
typedef GTestFixtureFunc TCFunc;
@@ -360,6 +395,10 @@ int main (int argc, char **argv)
g_test_suite_add (suite, TESTCASE (test_pack_gsm7_7_chars_offset, NULL));
+ g_test_suite_add (suite, TESTCASE (test_take_convert_ucs2_hex_utf8, NULL));
+ g_test_suite_add (suite, TESTCASE (test_take_convert_ucs2_bad_ascii, NULL));
+ g_test_suite_add (suite, TESTCASE (test_take_convert_ucs2_bad_ascii2, NULL));
+
result = g_test_run ();
return result;

View File

@ -0,0 +1,38 @@
commit 219424a6e2d017491a05ecbed661bccde3f991ef
Author: Aleksander Morgado <aleksander@lanedo.com>
Date: Wed Sep 12 07:34:37 2012 +0200
gsm: skip additional UTF-8 validity check when parsing operator
This is a fix related to the previous e07c2162d.
mm_charset_take_and_convert_to_utf8() already does a UTF-8 validity check
internally before returning the string, so it's pointless to do a new one
on the returned string.
Plus, mm_charset_take_and_convert_to_utf8() may really return NULL, which
would end up in segfaulting as g_utf8_validate() expects always a non-NULL
string.
diff --git a/src/mm-generic-gsm.c b/src/mm-generic-gsm.c
index 83e4083..60f2ea1 100644
--- a/src/mm-generic-gsm.c
+++ b/src/mm-generic-gsm.c
@@ -3066,13 +3066,14 @@ parse_operator (const char *reply, MMModemCharset cur_charset)
* string of the bytes of the operator name as encoded by the current
* character set.
*/
- if (cur_charset == MM_MODEM_CHARSET_UCS2)
+ if (cur_charset == MM_MODEM_CHARSET_UCS2) {
+ /* In this case we're already checking UTF-8 validity */
operator = mm_charset_take_and_convert_to_utf8 (operator, MM_MODEM_CHARSET_UCS2);
-
+ }
/* Ensure the operator name is valid UTF-8 so that we can send it
* through D-Bus and such.
*/
- if (!g_utf8_validate (operator, -1, NULL)) {
+ else if (!g_utf8_validate (operator, -1, NULL)) {
g_free (operator);
operator = NULL;
}