system/icu: patch CVE-2020-10531

Also remove obsolete CVE-2017-7867-7868.patch - this was merged since at
least 59.1 and was left over from when icu was originally pulled into the
system/ tree in 2018.
https://github.com/unicode-org/icu/commit/35a07bf89d64809b2e9af3cc90b53e3261677c53

3 files changed, 125 insertions, 157 deletions

diff --git a/system/icu/APKBUILD b/system/icu/APKBUILD
index 4165246a3..046698316 100644
--- a/system/icu/APKBUILD
+++ b/system/icu/APKBUILD
@@ -6,7 +6,7 @@ pkgver=65.1
 # convert x.y.z to x_y_z
 _ver=$(printf '%s' "$pkgver" | tr . _)
 
-pkgrel=0
+pkgrel=1
 pkgdesc="International Components for Unicode"
 url="http://site.icu-project.org/"
 arch="all"
@@ -17,6 +17,7 @@ makedepends=""
 subpackages="$pkgname-static $pkgname-dev $pkgname-doc $pkgname-libs"
 source="https://github.com/unicode-org/icu/releases/download/release-$(printf '%s' "$pkgver" | tr . -)/${pkgname}4c-$_ver-src.tgz
 	icu-60.2-always-use-utf8.patch
+	CVE-2020-10531.patch
 	"
 
 # secfixes:
@@ -27,6 +28,9 @@ source="https://github.com/unicode-org/icu/releases/download/release-$(printf '%
 #   58.2-r2:
 #     - CVE-2017-7867
 #     - CVE-2017-7868
+#   65.1-r1:
+#     - CVE-2020-10531
+
 builddir="$srcdir"/icu/source
 
 prepare() {
@@ -73,4 +77,5 @@ static() {
 }
 
 sha512sums="8f1ef33e1f4abc9a8ee870331c59f01b473d6da1251a19ce403f822f3e3871096f0791855d39c8f20c612fc49cda2c62c06864aa32ddab2dbd186d2b21ce9139  icu4c-65_1-src.tgz
-f86c62422f38f6485c58d4766e629bab69e4b0e00fa910854e40e7db1ace299152eaefa99ae2fbab7465e65d3156cbea7124612defa60680db58ab5c34d6262f  icu-60.2-always-use-utf8.patch"
+f86c62422f38f6485c58d4766e629bab69e4b0e00fa910854e40e7db1ace299152eaefa99ae2fbab7465e65d3156cbea7124612defa60680db58ab5c34d6262f  icu-60.2-always-use-utf8.patch
+2901235a24b1973516e495473188b077eb0486714ac47f702f226a3406176e913672159978b67548a1316284ecc716e06158b05c190ddfa83ad576a7a3539d72  CVE-2020-10531.patch"
diff --git a/system/icu/CVE-2017-7867-7868.patch b/system/icu/CVE-2017-7867-7868.patch
deleted file mode 100644
index df18283c2..000000000
--- a/system/icu/CVE-2017-7867-7868.patch
+++ /dev/null
@@ -1,155 +0,0 @@
-Index: source/common/utext.cpp

-===================================================================

---- source/common/utext.cpp	(revision 39670)

-+++ source/common/utext.cpp	(revision 39671)

-@@ -848,7 +848,13 @@

- 

- // Chunk size.

--//     Must be less than 85, because of byte mapping from UChar indexes to native indexes.

--//     Worst case is three native bytes to one UChar.  (Supplemenaries are 4 native bytes

--//     to two UChars.)

-+//     Must be less than 42  (256/6), because of byte mapping from UChar indexes to native indexes.

-+//     Worst case there are six UTF-8 bytes per UChar.

-+//         obsolete 6 byte form fd + 5 trails maps to fffd

-+//         obsolete 5 byte form fc + 4 trails maps to fffd

-+//         non-shortest 4 byte forms maps to fffd

-+//         normal supplementaries map to a pair of utf-16, two utf8 bytes per utf-16 unit

-+//     mapToUChars array size must allow for the worst case, 6.

-+//     This could be brought down to 4, by treating fd and fc as pure illegal,

-+//     rather than obsolete lead bytes. But that is not compatible with the utf-8 access macros.

- //

- enum { UTF8_TEXT_CHUNK_SIZE=32 };

-@@ -890,5 +896,5 @@

-                                                      //    one for a supplementary starting in the last normal position,

-                                                      //    and one for an entry for the buffer limit position.

--    uint8_t   mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to

-+    uint8_t   mapToUChars[UTF8_TEXT_CHUNK_SIZE*6+6]; // Map native offset from bufNativeStart to

-                                                      //   correspoding offset in filled part of buf.

-     int32_t   align;

-@@ -1033,4 +1039,5 @@

-             u8b = (UTF8Buf *)ut->p;   // the current buffer

-             mapIndex = ix - u8b->toUCharsMapStart;

-+            U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));

-             ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;

-             return TRUE;

-@@ -1299,4 +1306,8 @@

-         //   If index is at the end, there is no character there to look at.

-         if (ix != ut->b) {

-+            // Note: this function will only move the index back if it is on a trail byte

-+            //       and there is a preceding lead byte and the sequence from the lead 

-+            //       through this trail could be part of a valid UTF-8 sequence

-+            //       Otherwise the index remains unchanged.

-             U8_SET_CP_START(s8, 0, ix);

-         }

-@@ -1312,5 +1323,8 @@

-         uint8_t *mapToNative = u8b->mapToNative;

-         uint8_t *mapToUChars = u8b->mapToUChars;

--        int32_t  toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);

-+        int32_t  toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;

-+        // Note that toUCharsMapStart can be negative. Happens when the remaining

-+        // text from current position to the beginning is less than the buffer size.

-+        // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.

-         int32_t  destIx = UTF8_TEXT_CHUNK_SIZE+2;   // Start in the overflow region

-                                                     //   at end of buffer to leave room

-@@ -1339,4 +1353,5 @@

-                 // Special case ASCII range for speed.

-                 buf[destIx] = (UChar)c;

-+                U_ASSERT(toUCharsMapStart <= srcIx);

-                 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;

-                 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);

-@@ -1368,4 +1383,5 @@

-                     mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;

-                 } while (sIx >= srcIx);

-+                U_ASSERT(toUCharsMapStart <= (srcIx+1));

- 

-                 // Set native indexing limit to be the current position.

-@@ -1542,4 +1558,5 @@

-     U_ASSERT(index<=ut->chunkNativeLimit);

-     int32_t mapIndex = index - u8b->toUCharsMapStart;

-+    U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));

-     int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;

-     U_ASSERT(offset>=0 && offset<=ut->chunkLength);

-Index: source/test/intltest/utxttest.cpp

-===================================================================

---- source/test/intltest/utxttest.cpp	(revision 39670)

-+++ source/test/intltest/utxttest.cpp	(revision 39671)

-@@ -68,4 +68,6 @@

-         case 7: name = "Ticket12130";

-             if (exec) Ticket12130(); break;

-+        case 8: name = "Ticket12888";

-+            if (exec) Ticket12888(); break;

-         default: name = "";          break;

-     }

-@@ -1584,2 +1586,62 @@

-     utext_close(&ut);

- }

-+

-+// Ticket 12888: bad handling of illegal utf-8 containing many instances of the archaic, now illegal,

-+//               six byte utf-8 forms. Original implementation had an assumption that

-+//               there would be at most three utf-8 bytes per UTF-16 code unit.

-+//               The five and six byte sequences map to a single replacement character.

-+

-+void UTextTest::Ticket12888() {

-+    const char *badString = 

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80"

-+            "\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80\xfd\x80\x80\x80\x80\x80";

-+

-+    UErrorCode status = U_ZERO_ERROR;

-+    LocalUTextPointer ut(utext_openUTF8(NULL, badString, -1, &status));

-+    TEST_SUCCESS(status);

-+    for (;;) {

-+        UChar32 c = utext_next32(ut.getAlias());

-+        if (c == U_SENTINEL) {

-+            break;

-+        }

-+    }

-+    int32_t endIdx = utext_getNativeIndex(ut.getAlias());

-+    if (endIdx != (int32_t)strlen(badString)) {

-+        errln("%s:%d expected=%d, actual=%d", __FILE__, __LINE__, strlen(badString), endIdx);

-+        return;

-+    }

-+

-+    for (int32_t prevIndex = endIdx; prevIndex>0;) {

-+        UChar32 c = utext_previous32(ut.getAlias());

-+        int32_t currentIndex = utext_getNativeIndex(ut.getAlias());

-+        if (c != 0xfffd) {

-+            errln("%s:%d (expected, actual, index) = (%d, %d, %d)\n",

-+                    __FILE__, __LINE__, 0xfffd, c, currentIndex);

-+            break;

-+        }

-+        if (currentIndex != prevIndex - 6) {

-+            errln("%s:%d: wrong index. Expected, actual = %d, %d",

-+                    __FILE__, __LINE__, prevIndex - 6, currentIndex);

-+            break;

-+        }

-+        prevIndex = currentIndex;

-+    }

-+}

-Index: source/test/intltest/utxttest.h

-===================================================================

---- source/test/intltest/utxttest.h	(revision 39670)

-+++ source/test/intltest/utxttest.h	(revision 39671)

-@@ -39,4 +39,5 @@

-     void Ticket10983();

-     void Ticket12130();

-+    void Ticket12888();

- 

- private:

diff --git a/system/icu/CVE-2020-10531.patch b/system/icu/CVE-2020-10531.patch
new file mode 100644
index 000000000..f456b06ec
--- /dev/null
+++ b/system/icu/CVE-2020-10531.patch
@@ -0,0 +1,118 @@
+From b7d08bc04a4296982fcef8b6b8a354a9e4e7afca Mon Sep 17 00:00:00 2001
+From: Frank Tang <ftang@chromium.org>
+Date: Sat, 1 Feb 2020 02:39:04 +0000
+Subject: [PATCH] ICU-20958 Prevent SEGV_MAPERR in append
+
+See #971
+---
+ source/common/unistr.cpp          |  6 ++-
+ source/test/intltest/ustrtest.cpp | 62 +++++++++++++++++++++++++
+ source/test/intltest/ustrtest.h   |  1 +
+ 3 files changed, 68 insertions(+), 1 deletion(-)
+
+diff --git source/common/unistr.cpp source/common/unistr.cpp
+index 901bb3358ba..077b4d6ef20 100644
+--- source/common/unistr.cpp
++++ source/common/unistr.cpp
+@@ -1563,7 +1563,11 @@ UnicodeString::doAppend(const UChar *srcChars, int32_t srcStart, int32_t srcLeng
+   }
+ 
+   int32_t oldLength = length();
+-  int32_t newLength = oldLength + srcLength;
++  int32_t newLength;
++  if (uprv_add32_overflow(oldLength, srcLength, &newLength)) {
++    setToBogus();
++    return *this;
++  }
+ 
+   // Check for append onto ourself
+   const UChar* oldArray = getArrayStart();
+diff --git source/test/intltest/ustrtest.cpp source/test/intltest/ustrtest.cpp
+index b6515ea813c..ad38bdf53a3 100644
+--- source/test/intltest/ustrtest.cpp
++++ source/test/intltest/ustrtest.cpp
+@@ -67,6 +67,7 @@ void UnicodeStringTest::runIndexedTest( int32_t index, UBool exec, const char* &
+     TESTCASE_AUTO(TestWCharPointers);
+     TESTCASE_AUTO(TestNullPointers);
+     TESTCASE_AUTO(TestUnicodeStringInsertAppendToSelf);
++    TESTCASE_AUTO(TestLargeAppend);
+     TESTCASE_AUTO_END;
+ }
+ 
+@@ -2310,3 +2311,64 @@ void UnicodeStringTest::TestUnicodeStringInsertAppendToSelf() {
+     str.insert(2, sub);
+     assertEquals("", u"abbcdcde", str);
+ }
++
++void UnicodeStringTest::TestLargeAppend() {
++    if(quick) return;
++
++    IcuTestErrorCode status(*this, "TestLargeAppend");
++    // Make a large UnicodeString
++    int32_t len = 0xAFFFFFF;
++    UnicodeString str;
++    char16_t *buf = str.getBuffer(len);
++    // A fast way to set buffer to valid Unicode.
++    // 4E4E is a valid unicode character
++    uprv_memset(buf, 0x4e, len * 2);
++    str.releaseBuffer(len);
++    UnicodeString dest;
++    // Append it 16 times
++    // 0xAFFFFFF times 16 is 0xA4FFFFF1,
++    // which is greater than INT32_MAX, which is 0x7FFFFFFF.
++    int64_t total = 0;
++    for (int32_t i = 0; i < 16; i++) {
++        dest.append(str);
++        total += len;
++        if (total <= INT32_MAX) {
++            assertFalse("dest is not bogus", dest.isBogus());
++        } else {
++            assertTrue("dest should be bogus", dest.isBogus());
++        }
++    }
++    dest.remove();
++    total = 0;
++    for (int32_t i = 0; i < 16; i++) {
++        dest.append(str);
++        total += len;
++        if (total + len <= INT32_MAX) {
++            assertFalse("dest is not bogus", dest.isBogus());
++        } else if (total <= INT32_MAX) {
++            // Check that a string of exactly the maximum size works
++            UnicodeString str2;
++            int32_t remain = INT32_MAX - total;
++            char16_t *buf2 = str2.getBuffer(remain);
++            if (buf2 == nullptr) {
++                // if somehow memory allocation fail, return the test
++                return;
++            }
++            uprv_memset(buf2, 0x4e, remain * 2);
++            str2.releaseBuffer(remain);
++            dest.append(str2);
++            total += remain;
++            assertEquals("When a string of exactly the maximum size works", (int64_t)INT32_MAX, total);
++            assertEquals("When a string of exactly the maximum size works", INT32_MAX, dest.length());
++            assertFalse("dest is not bogus", dest.isBogus());
++
++            // Check that a string size+1 goes bogus
++            str2.truncate(1);
++            dest.append(str2);
++            total++;
++            assertTrue("dest should be bogus", dest.isBogus());
++        } else {
++            assertTrue("dest should be bogus", dest.isBogus());
++        }
++    }
++}
+diff --git source/test/intltest/ustrtest.h source/test/intltest/ustrtest.h
+index 218befdcc68..4a356a92c7a 100644
+--- source/test/intltest/ustrtest.h
++++ source/test/intltest/ustrtest.h
+@@ -97,6 +97,7 @@ class UnicodeStringTest: public IntlTest {
+     void TestWCharPointers();
+     void TestNullPointers();
+     void TestUnicodeStringInsertAppendToSelf();
++    void TestLargeAppend();
+ };
+ 
+ #endif