1 Fix CVE-2017-7867 and CVE-2017-7868:
3 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-7867
4 https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-7868
6 Patch copied from upstream source repository:
8 http://bugs.icu-project.org/trac/changeset/39671
10 Index: icu/source/common/utext.cpp
11 ===================================================================
12 --- icu/source/common/utext.cpp (revision 39670)
13 +++ icu/source/common/utext.cpp (revision 39671)
17 -// Must be less than 85, because of byte mapping from UChar indexes to native indexes.
18 -// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
20 +// Must be less than 42 (256/6), because of byte mapping from UChar indexes to native indexes.
21 +// Worst case there are six UTF-8 bytes per UChar.
22 +// obsolete 6 byte form fd + 5 trails maps to fffd
23 +// obsolete 5 byte form fc + 4 trails maps to fffd
24 +// non-shortest 4 byte forms maps to fffd
25 +// normal supplementaries map to a pair of utf-16, two utf8 bytes per utf-16 unit
26 +// mapToUChars array size must allow for the worst case, 6.
27 +// This could be brought down to 4, by treating fd and fc as pure illegal,
28 +// rather than obsolete lead bytes. But that is not compatible with the utf-8 access macros.
30 enum { UTF8_TEXT_CHUNK_SIZE=32 };
32 // one for a supplementary starting in the last normal position,
33 // and one for an entry for the buffer limit position.
34 - uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
35 + uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*6+6]; // Map native offset from bufNativeStart to
36 // correspoding offset in filled part of buf.
39 u8b = (UTF8Buf *)ut->p; // the current buffer
40 mapIndex = ix - u8b->toUCharsMapStart;
41 + U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
42 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
45 // If index is at the end, there is no character there to look at.
47 + // Note: this function will only move the index back if it is on a trail byte
48 + // and there is a preceding lead byte and the sequence from the lead
49 + // through this trail could be part of a valid UTF-8 sequence
50 + // Otherwise the index remains unchanged.
51 U8_SET_CP_START(s8, 0, ix);
54 uint8_t *mapToNative = u8b->mapToNative;
55 uint8_t *mapToUChars = u8b->mapToUChars;
56 - int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
57 + int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
58 + // Note that toUCharsMapStart can be negative. Happens when the remaining
59 + // text from current position to the beginning is less than the buffer size.
60 + // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
61 int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
62 // at end of buffer to leave room
64 // Special case ASCII range for speed.
65 buf[destIx] = (UChar)c;
66 + U_ASSERT(toUCharsMapStart <= srcIx);
67 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
68 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
70 mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
71 } while (sIx >= srcIx);
72 + U_ASSERT(toUCharsMapStart <= (srcIx+1));
74 // Set native indexing limit to be the current position.
76 U_ASSERT(index<=ut->chunkNativeLimit);
77 int32_t mapIndex = index - u8b->toUCharsMapStart;
78 + U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
79 int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
80 U_ASSERT(offset>=0 && offset<=ut->chunkLength);
81 Index: icu/source/test/intltest/utxttest.cpp
82 ===================================================================
83 --- icu/source/test/intltest/utxttest.cpp (revision 39670)
84 +++ icu/source/test/intltest/utxttest.cpp (revision 39671)
86 case 7: name = "Ticket12130";
87 if (exec) Ticket12130(); break;
88 + case 8: name = "Ticket12888";
89 + if (exec) Ticket12888(); break;
90 default: name = ""; break;
92 @@ -1584,2 +1586,62 @@
96 +// Ticket 12888: bad handling of illegal utf-8 containing many instances of the archaic, now illegal,
97 +// six byte utf-8 forms. Original implementation had an assumption that
98 +// there would be at most three utf-8 bytes per UTF-16 code unit.
99 +// The five and six byte sequences map to a single replacement character.
101 +void UTextTest::Ticket12888() {
102 + const char *badString =
103 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
104 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
105 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
106 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
107 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
108 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
109 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
110 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
111 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
112 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
113 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
114 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
115 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
116 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
117 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
118 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
119 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
120 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
121 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"
122 + "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80";
124 + UErrorCode status = U_ZERO_ERROR;
125 + LocalUTextPointer ut(utext_openUTF8(NULL, badString, -1, &status));
126 + TEST_SUCCESS(status);
128 + UChar32 c = utext_next32(ut.getAlias());
129 + if (c == U_SENTINEL) {
133 + int32_t endIdx = utext_getNativeIndex(ut.getAlias());
134 + if (endIdx != (int32_t)strlen(badString)) {
135 + errln("%s:%d expected=%d, actual=%d", __FILE__, __LINE__, strlen(badString), endIdx);
139 + for (int32_t prevIndex = endIdx; prevIndex>0;) {
140 + UChar32 c = utext_previous32(ut.getAlias());
141 + int32_t currentIndex = utext_getNativeIndex(ut.getAlias());
143 + errln("%s:%d (expected, actual, index) = (%d, %d, %d)\n",
144 + __FILE__, __LINE__, 0xfffd, c, currentIndex);
147 + if (currentIndex != prevIndex - 6) {
148 + errln("%s:%d: wrong index. Expected, actual = %d, %d",
149 + __FILE__, __LINE__, prevIndex - 6, currentIndex);
152 + prevIndex = currentIndex;
155 Index: icu/source/test/intltest/utxttest.h
156 ===================================================================
157 --- icu/source/test/intltest/utxttest.h (revision 39670)
158 +++ icu/source/test/intltest/utxttest.h (revision 39671)
162 + void Ticket12888();