shared: json - support escaping utf16 surrogate pairs

We originally only supported escaping ucs2 encoded characters (as \uxxxx). This
only covers the BMP. Support escaping also utf16 surrogate pairs (on the form
\uxxxx\uyyyy) to cover all of unicode.

2 files changed, 62 insertions, 22 deletions

diff --git a/src/shared/json.c b/src/shared/json.c
index 47f801c858..bb3d26f0e5 100644
--- a/src/shared/json.c
+++ b/src/shared/json.c
@@ -53,6 +53,42 @@ static void inc_lines(unsigned *line, const char *s, size_t n) {
         }
 }
 
+static int unhex_ucs2(const char *c, uint16_t *ret) {
+        int aa, bb, cc, dd;
+        uint16_t x;
+
+        assert(c);
+        assert(ret);
+
+        aa = unhexchar(c[0]);
+        if (aa < 0)
+                return -EINVAL;
+
+        bb = unhexchar(c[1]);
+        if (bb < 0)
+                return -EINVAL;
+
+        cc = unhexchar(c[2]);
+        if (cc < 0)
+                return -EINVAL;
+
+        dd = unhexchar(c[3]);
+        if (dd < 0)
+                return -EINVAL;
+
+        x =     ((uint16_t) aa << 12) |
+                ((uint16_t) bb << 8) |
+                ((uint16_t) cc << 4) |
+                ((uint16_t) dd);
+
+        if (x <= 0)
+                return -EINVAL;
+
+        *ret = x;
+
+        return 0;
+}
+
 static int json_parse_string(const char **p, char **ret) {
         _cleanup_free_ char *s = NULL;
         size_t n = 0, allocated = 0;
@@ -119,39 +155,40 @@ static int json_parse_string(const char **p, char **ret) {
                         else if (*c == 't')
                                 ch = '\t';
                         else if (*c == 'u') {
-                                int aa, bb, cc, dd;
                                 uint16_t x;
+                                int r;
 
-                                aa = unhexchar(c[1]);
-                                if (aa < 0)
-                                        return -EINVAL;
+                                r = unhex_ucs2(c + 1, &x);
+                                if (r < 0)
+                                        return r;
 
-                                bb = unhexchar(c[2]);
-                                if (bb < 0)
-                                        return -EINVAL;
+                                c += 5;
 
-                                cc = unhexchar(c[3]);
-                                if (cc < 0)
-                                        return -EINVAL;
+                                if (!GREEDY_REALLOC(s, allocated, n + 4))
+                                        return -ENOMEM;
 
-                                dd = unhexchar(c[4]);
-                                if (dd < 0)
+                                if (!utf16_is_surrogate(x))
+                                        n += utf8_encode_unichar(s + n, x);
+                                else if (utf16_is_trailing_surrogate(x))
                                         return -EINVAL;
+                                else {
+                                        uint16_t y;
 
+                                        if (c[0] != '\\' || c[1] != 'u')
+                                                return -EINVAL;
 
-                                x =     ((uint16_t) aa << 12) |
-                                        ((uint16_t) bb << 8) |
-                                        ((uint16_t) cc << 4) |
-                                        ((uint16_t) dd);
+                                        r = unhex_ucs2(c + 2, &y);
+                                        if (r < 0)
+                                                return r;
 
-                                if (x <= 0)
-                                        return -EINVAL;
+                                        c += 6;
 
-                                if (!GREEDY_REALLOC(s, allocated, n + 4))
-                                        return -ENOMEM;
+                                        if (!utf16_is_trailing_surrogate(y))
+                                                return -EINVAL;
+
+                                        n += utf8_encode_unichar(s + n, utf16_surrogate_pair_to_unichar(x, y));
+                                }
 
-                                n += utf8_encode_unichar(s + n, x);
-                                c += 5;
                                 continue;
                         } else
                                 return -EINVAL;
diff --git a/src/test/test-json.c b/src/test/test-json.c
index e53e8ed50f..b09131891c 100644
--- a/src/test/test-json.c
+++ b/src/test/test-json.c
@@ -99,6 +99,9 @@ int main(int argc, char *argv[]) {
         test_one("\"\xef\xbf\xbd\"", JSON_STRING, "\xef\xbf\xbd", JSON_END);
         test_one("\"\\ufffd\"", JSON_STRING, "\xef\xbf\xbd", JSON_END);
         test_one("\"\\uf\"", -EINVAL);
+        test_one("\"\\ud800a\"", -EINVAL);
+        test_one("\"\\udc00\\udc00\"", -EINVAL);
+        test_one("\"\\ud801\\udc37\"", JSON_STRING, "\xf0\x90\x90\xb7", JSON_END);
 
         return 0;
 }