commit e8f285b522a907603501329e5b4212755f525fdf
Author: Tomas Korbar <tkorbar@redhat.com>
Date:   Thu Mar 3 12:04:09 2022 +0100

    CVE-2022-25235

diff --git a/lib/xmltok.c b/lib/xmltok.c
index 6b415d8..b55732a 100644
--- a/lib/xmltok.c
+++ b/lib/xmltok.c
@@ -103,13 +103,6 @@
                       + ((((byte)[2]) >> 5) & 1)] \
          & (1u << (((byte)[2]) & 0x1F)))
 
-#define UTF8_GET_NAMING(pages, p, n) \
-  ((n) == 2 \
-  ? UTF8_GET_NAMING2(pages, (const unsigned char *)(p)) \
-  : ((n) == 3 \
-     ? UTF8_GET_NAMING3(pages, (const unsigned char *)(p)) \
-     : 0))
-
 /* Detection of invalid UTF-8 sequences is based on Table 3.1B
    of Unicode 3.2: http://www.unicode.org/unicode/reports/tr28/
    with the additional restriction of not allowing the Unicode
diff --git a/lib/xmltok_impl.c b/lib/xmltok_impl.c
index 0403dd3..56d7a40 100644
--- a/lib/xmltok_impl.c
+++ b/lib/xmltok_impl.c
@@ -61,7 +61,7 @@
    case BT_LEAD ## n: \
      if (end - ptr < n) \
        return XML_TOK_PARTIAL_CHAR; \
-     if (!IS_NAME_CHAR(enc, ptr, n)) { \
+     if (IS_INVALID_CHAR(enc, ptr, n) || ! IS_NAME_CHAR(enc, ptr, n)) { \
        *nextTokPtr = ptr; \
        return XML_TOK_INVALID; \
      } \
@@ -89,7 +89,7 @@
    case BT_LEAD ## n: \
      if (end - ptr < n) \
        return XML_TOK_PARTIAL_CHAR; \
-     if (!IS_NMSTRT_CHAR(enc, ptr, n)) { \
+     if (IS_INVALID_CHAR(enc, ptr, n) || ! IS_NMSTRT_CHAR(enc, ptr, n)) { \
        *nextTokPtr = ptr; \
        return XML_TOK_INVALID; \
      } \
@@ -1117,6 +1117,10 @@ PREFIX(prologTok)(const ENCODING *enc, const char *ptr, const char *end,
   case BT_LEAD ## n: \
     if (end - ptr < n) \
       return XML_TOK_PARTIAL_CHAR; \
+    if (IS_INVALID_CHAR(enc, ptr, n)) {                                        \
+      *nextTokPtr = ptr;                                                       \
+      return XML_TOK_INVALID;                                                  \
+    }                                                                          \
     if (IS_NMSTRT_CHAR(enc, ptr, n)) { \
       ptr += n; \
       tok = XML_TOK_NAME; \
diff --git a/tests/runtests.c b/tests/runtests.c
index 278bfa1..0f3afde 100644
--- a/tests/runtests.c
+++ b/tests/runtests.c
@@ -6540,6 +6540,106 @@ START_TEST(test_utf8_in_cdata_section_2)
 }
 END_TEST
 
+START_TEST(test_utf8_in_start_tags) {
+  struct test_case {
+    bool goodName;
+    bool goodNameStart;
+    const char *tagName;
+  };
+
+  // The idea with the tests below is this:
+  // We want to cover 1-, 2- and 3-byte sequences, 4-byte sequences
+  // go to isNever and are hence not a concern.
+  //
+  // We start with a character that is a valid name character
+  // (or even name-start character, see XML 1.0r4 spec) and then we flip
+  // single bits at places where (1) the result leaves the UTF-8 encoding space
+  // and (2) we stay in the same n-byte sequence family.
+  //
+  // The flipped bits are highlighted in angle brackets in comments,
+  // e.g. "[<1>011 1001]" means we had [0011 1001] but we now flipped
+  // the most significant bit to 1 to leave UTF-8 encoding space.
+  struct test_case cases[] = {
+      // 1-byte UTF-8: [0xxx xxxx]
+      {true, true, "\x3A"},   // [0011 1010] = ASCII colon ':'
+      {false, false, "\xBA"}, // [<1>011 1010]
+      {true, false, "\x39"},  // [0011 1001] = ASCII nine '9'
+      {false, false, "\xB9"}, // [<1>011 1001]
+
+      // 2-byte UTF-8: [110x xxxx] [10xx xxxx]
+      {true, true, "\xDB\xA5"},   // [1101 1011] [1010 0101] =
+                                  // Arabic small waw U+06E5
+      {false, false, "\x9B\xA5"}, // [1<0>01 1011] [1010 0101]
+      {false, false, "\xDB\x25"}, // [1101 1011] [<0>010 0101]
+      {false, false, "\xDB\xE5"}, // [1101 1011] [1<1>10 0101]
+      {true, false, "\xCC\x81"},  // [1100 1100] [1000 0001] =
+                                  // combining char U+0301
+      {false, false, "\x8C\x81"}, // [1<0>00 1100] [1000 0001]
+      {false, false, "\xCC\x01"}, // [1100 1100] [<0>000 0001]
+      {false, false, "\xCC\xC1"}, // [1100 1100] [1<1>00 0001]
+
+      // 3-byte UTF-8: [1110 xxxx] [10xx xxxx] [10xxxxxx]
+      {true, true, "\xE0\xA4\x85"},   // [1110 0000] [1010 0100] [1000 0101] =
+                                      // Devanagari Letter A U+0905
+      {false, false, "\xA0\xA4\x85"}, // [1<0>10 0000] [1010 0100] [1000 0101]
+      {false, false, "\xE0\x24\x85"}, // [1110 0000] [<0>010 0100] [1000 0101]
+      {false, false, "\xE0\xE4\x85"}, // [1110 0000] [1<1>10 0100] [1000 0101]
+      {false, false, "\xE0\xA4\x05"}, // [1110 0000] [1010 0100] [<0>000 0101]
+      {false, false, "\xE0\xA4\xC5"}, // [1110 0000] [1010 0100] [1<1>00 0101]
+      {true, false, "\xE0\xA4\x81"},  // [1110 0000] [1010 0100] [1000 0001] =
+                                      // combining char U+0901
+      {false, false, "\xA0\xA4\x81"}, // [1<0>10 0000] [1010 0100] [1000 0001]
+      {false, false, "\xE0\x24\x81"}, // [1110 0000] [<0>010 0100] [1000 0001]
+      {false, false, "\xE0\xE4\x81"}, // [1110 0000] [1<1>10 0100] [1000 0001]
+      {false, false, "\xE0\xA4\x01"}, // [1110 0000] [1010 0100] [<0>000 0001]
+      {false, false, "\xE0\xA4\xC1"}, // [1110 0000] [1010 0100] [1<1>00 0001]
+  };
+  const bool atNameStart[] = {true, false};
+
+  size_t i = 0;
+  char doc[1024];
+  size_t failCount = 0;
+
+  for (; i < sizeof(cases) / sizeof(cases[0]); i++) {
+    size_t j = 0;
+    for (; j < sizeof(atNameStart) / sizeof(atNameStart[0]); j++) {
+      const bool expectedSuccess
+          = atNameStart[j] ? cases[i].goodNameStart : cases[i].goodName;
+      sprintf(doc, "<%s%s><!--", atNameStart[j] ? "" : "a", cases[i].tagName);
+      XML_Parser parser = XML_ParserCreate(NULL);
+
+      const enum XML_Status status
+          = XML_Parse(parser, doc, (int)strlen(doc), /*isFinal=*/XML_FALSE);
+
+      bool success = true;
+      if ((status == XML_STATUS_OK) != expectedSuccess) {
+        success = false;
+      }
+      if ((status == XML_STATUS_ERROR)
+          && (XML_GetErrorCode(parser) != XML_ERROR_INVALID_TOKEN)) {
+        success = false;
+      }
+
+      if (! success) {
+        fprintf(
+            stderr,
+            "FAIL case %2u (%sat name start, %u-byte sequence, error code %d)\n",
+            (unsigned)i + 1u, atNameStart[j] ? "    " : "not ",
+            (unsigned)strlen(cases[i].tagName), XML_GetErrorCode(parser));
+        failCount++;
+      }
+
+      XML_ParserFree(parser);
+    }
+  }
+
+  if (failCount > 0) {
+    fail("UTF-8 regression detected");
+  }
+}
+END_TEST
+
+
 /* Test trailing spaces in elements are accepted */
 static void XMLCALL
 record_element_end_handler(void *userData,
@@ -6734,6 +6834,15 @@ START_TEST(test_bad_doctype)
 }
 END_TEST
 
+START_TEST(test_bad_doctype_utf8) {
+  const char *text = "<!DOCTYPE \xDB\x25"
+                     "doc><doc/>"; // [1101 1011] [<0>010 0101]
+  expect_failure(text, XML_ERROR_INVALID_TOKEN,
+                 "Invalid UTF-8 in DOCTYPE not faulted");
+}
+END_TEST
+
+
 START_TEST(test_bad_doctype_utf16)
 {
     const char text[] =
@@ -12256,6 +12365,7 @@ make_suite(void)
     tcase_add_test(tc_basic, test_ext_entity_utf8_non_bom);
     tcase_add_test(tc_basic, test_utf8_in_cdata_section);
     tcase_add_test(tc_basic, test_utf8_in_cdata_section_2);
+    tcase_add_test(tc_basic, test_utf8_in_start_tags);
     tcase_add_test(tc_basic, test_trailing_spaces_in_elements);
     tcase_add_test(tc_basic, test_utf16_attribute);
     tcase_add_test(tc_basic, test_utf16_second_attr);
@@ -12264,6 +12374,7 @@ make_suite(void)
     tcase_add_test(tc_basic, test_bad_attr_desc_keyword);
     tcase_add_test(tc_basic, test_bad_attr_desc_keyword_utf16);
     tcase_add_test(tc_basic, test_bad_doctype);
+    tcase_add_test(tc_basic, test_bad_doctype_utf8);
     tcase_add_test(tc_basic, test_bad_doctype_utf16);
     tcase_add_test(tc_basic, test_bad_doctype_plus);
     tcase_add_test(tc_basic, test_bad_doctype_star);