guri: Fix URI scope parsing

The previous parsing code could read off the end of a URI if it had an
incorrect %-escaped character in.

Fix that, and more closely implement parsing for the syntax defined in
RFC 6874, which is the amendment to RFC 3986 which specifies zone ID
syntax.

This requires reworking some network-address tests, which were
previously treating zone IDs incorrectly.

oss-fuzz#23816

Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This commit is contained in:
Philip Withnall 2020-09-30 18:52:18 +01:00
parent c363c3a9a5
commit b43fb9cbfb
3 changed files with 56 additions and 26 deletions

View File

@ -40,10 +40,10 @@ static ParseTest uri_tests[] = {
{ "ftp://[fec0::abcd]/start", "ftp", "fec0::abcd", 8080, -1 },
{ "ftp://[fec0::abcd]:999/start", "ftp", "fec0::abcd", 999, -1 },
{ "ftp://joe%x-@ftp.gnome.org:2020/start", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "http://[fec0::abcd%em1]/start", "http", "fec0::abcd%em1", 8080, -1 },
{ "http://[fec0::abcd%em1]/start", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "http://[fec0::abcd%25em1]/start", "http", "fec0::abcd%em1", 8080, -1 },
{ "http://[fec0::abcd%10]/start", "http", "fec0::abcd%10", 8080, -1 },
{ "http://[fec0::abcd%25em%31]/start", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "http://[fec0::abcd%10]/start", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "http://[fec0::abcd%25em%31]/start", "http", "fec0::abcd%em1", 8080, -1 },
{ "ftp://ftp.gnome.org/start?foo=bar@baz", "ftp", "ftp.gnome.org", 8080, -1 }
};
@ -80,6 +80,7 @@ static ParseTest host_tests[] =
{ "[2001:db8::1]", NULL, "2001:db8::1", 1234, -1 },
{ "[2001:db8::1]:888", NULL, "2001:db8::1", 888, -1 },
{ "[2001:db8::1%em1]", NULL, "2001:db8::1%em1", 1234, -1 },
{ "[2001:db8::1%25em1]", NULL, "2001:db8::1%25em1", 1234, -1 },
{ "[hostname", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "[hostnam]e", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
{ "hostname:", NULL, NULL, 0, G_IO_ERROR_INVALID_ARGUMENT },
@ -337,10 +338,9 @@ test_uri_scope_id (void)
SCOPE_ID_TEST_PORT);
addr = g_network_address_parse_uri (uri, 0, &error);
g_free (uri);
g_assert_no_error (error);
test_scope_id (addr);
g_object_unref (addr);
g_assert_error (error, G_IO_ERROR, G_IO_ERROR_INVALID_ARGUMENT);
g_assert_null (addr);
g_clear_error (&error);
uri = g_strdup_printf ("http://[%s%%25%s]:%d/foo",
SCOPE_ID_TEST_ADDR,

View File

@ -445,15 +445,21 @@ _uri_encoder (GString *out,
/* Parse the IP-literal construction from RFC 6874 (which extends RFC 3986 to
* support IPv6 zone identifiers.
*
* Rules:
* Currently, IP versions beyond 6 (i.e. the IPvFuture rule) are unsupported.
* Theres no point supporting them until (a) they exist and (b) the rest of the
* stack (notably, sockets) supports them.
*
* IP-literal = "[" ( IPv6address / IPvFuture ) "]"
* Rules:
*
* IP-literal = "[" ( IPv6address / IPv6addrz / IPvFuture ) "]"
*
* ZoneID = 1*( unreserved / pct-encoded )
*
* IPv6addrz = IPv6address "%25" ZoneID
*
* If %G_URI_FLAGS_PARSE_RELAXED is specified, this function also accepts:
*
* IPv6addrz = IPv6address "%" ZoneID
*/
static gboolean
parse_ip_literal (const gchar *start,
@ -462,43 +468,67 @@ parse_ip_literal (const gchar *start,
gchar **out,
GError **error)
{
gchar *pct;
gchar *pct, *zone_id = NULL;
gchar *addr = NULL;
gsize addr_length = 0;
gsize zone_id_length = 0;
gchar *decoded_zone_id = NULL;
if (start[length - 1] != ']')
goto bad_ipv6_literal;
/* Drop the square brackets */
addr = g_strndup (start + 1, length - 2);
addr_length = length - 2;
/* If there's an IPv6 scope id, ignore it for the moment. */
/* If there's an IPv6 scope ID, split out the zone. */
pct = strchr (addr, '%');
if (pct)
*pct = '\0';
if (pct != NULL)
{
*pct = '\0';
if (addr_length - (pct - addr) >= 4 &&
*(pct + 1) == '2' && *(pct + 2) == '5')
{
zone_id = pct + 3;
zone_id_length = addr_length - (zone_id - addr);
}
else if (flags & G_URI_FLAGS_PARSE_RELAXED &&
addr_length - (pct - addr) >= 2)
{
zone_id = pct + 1;
zone_id_length = addr_length - (zone_id - addr);
}
else
goto bad_ipv6_literal;
g_assert (zone_id_length >= 1);
}
/* addr must be an IPv6 address */
if (!g_hostname_is_ip_address (addr) || !strchr (addr, ':'))
goto bad_ipv6_literal;
if (pct)
{
*pct = '%';
if (strchr (pct + 1, '%'))
goto bad_ipv6_literal;
/* If the '%' is encoded as '%25' (which it should be), decode it */
if (pct[1] == '2' && pct[2] == '5' && pct[3])
memmove (pct + 1, pct + 3, strlen (pct + 3) + 1);
}
/* Zone ID must be valid. It can contain %-encoded characters. */
if (zone_id != NULL &&
!uri_decode (&decoded_zone_id, NULL, zone_id, zone_id_length, FALSE,
flags, G_URI_ERROR_BAD_HOST, NULL))
goto bad_ipv6_literal;
/* Success */
if (out != NULL)
if (out != NULL && decoded_zone_id != NULL)
*out = g_strconcat (addr, "%", decoded_zone_id, NULL);
else if (out != NULL)
*out = g_steal_pointer (&addr);
g_free (addr);
g_free (decoded_zone_id);
return TRUE;
bad_ipv6_literal:
g_free (addr);
g_free (decoded_zone_id);
g_set_error (error, G_URI_ERROR, G_URI_ERROR_BAD_HOST,
_("Invalid IPv6 address %.*s in URI"),
(gint)length, start);

View File

@ -697,13 +697,13 @@ static const UriAbsoluteTest absolute_tests[] = {
{ "http-ish", NULL, "host", -1, "/path", NULL, NULL } },
/* IPv6 scope ID parsing (both correct and incorrect) */
{ "http://[fe80::dead:beef%em1]/", G_URI_FLAGS_NONE,
{ "http://[fe80::dead:beef%em1]/", G_URI_FLAGS_PARSE_RELAXED,
{ "http", NULL, "fe80::dead:beef%em1", -1, "/", NULL, NULL } },
{ "http://[fe80::dead:beef%25em1]/", G_URI_FLAGS_NONE,
{ "http", NULL, "fe80::dead:beef%em1", -1, "/", NULL, NULL } },
{ "http://[fe80::dead:beef%10]/", G_URI_FLAGS_NONE,
{ "http://[fe80::dead:beef%10]/", G_URI_FLAGS_PARSE_RELAXED,
{ "http", NULL, "fe80::dead:beef%10", -1, "/", NULL, NULL } },
{ "http://[fe80::dead:beef%25]/", G_URI_FLAGS_NONE,
{ "http://[fe80::dead:beef%25]/", G_URI_FLAGS_PARSE_RELAXED,
{ "http", NULL, "fe80::dead:beef%25", -1, "/", NULL, NULL } },
};
static int num_absolute_tests = G_N_ELEMENTS (absolute_tests);