[php-src] master: Fix infinite loop in mb_encode_mimeheader
| From: | Alex Dowad via Ben Ramsey | Date: | Wed, 10 Apr 2024 04:55:43 +0000 |
| Subject: | [php-src] master: Fix infinite loop in mb_encode_mimeheader | ||
| Groups: | php.cvs | ||
| Request: | Send a blank email to php-cvs+get-128030@lists.php.net to get a copy of this message | ||
Author: Alex Dowad (alexdowad)
Committer: Ben Ramsey (ramsey)
Date: 2024-04-09T23:52:11-05:00
Commit: https://github.com/php/php-src/commit/3394efc63e52a017995f92d8da4ef28224247bb3
Raw diff: https://github.com/php/php-src/commit/3394efc63e52a017995f92d8da4ef28224247bb3.diff
Fix infinite loop in mb_encode_mimeheader
Changed paths:
M ext/mbstring/mbstring.c
M ext/mbstring/tests/mb_encode_mimeheader_basic4.phpt
Diff:
diff --git a/ext/mbstring/mbstring.c b/ext/mbstring/mbstring.c
index cf4c1c32c5915..618fff55362e1 100644
--- a/ext/mbstring/mbstring.c
+++ b/ext/mbstring/mbstring.c
@@ -5858,6 +5858,9 @@ static zend_string* mb_mime_header_encode(zend_string *input, const
mbfl_encodin
unsigned char *in = (unsigned char*)ZSTR_VAL(input);
size_t in_len = ZSTR_LEN(input);
+ ZEND_ASSERT(outcode->mime_name != NULL);
+ ZEND_ASSERT(outcode->mime_name[0] != '\0');
+
if (!in_len) {
return zend_empty_string;
}
@@ -5880,7 +5883,8 @@ static zend_string* mb_mime_header_encode(zend_string *input, const
mbfl_encodin
unsigned int state = 0;
/* wchar_buf should be big enough that when it is full, we definitely have enough
* wchars to fill an entire line of output */
- uint32_t wchar_buf[80];
+ const size_t wchar_buf_len = 90;
+ uint32_t wchar_buf[wchar_buf_len];
uint32_t *p, *e;
/* What part of wchar_buf is filled with still-unprocessed data which should not
* be overwritten? */
@@ -5891,7 +5895,7 @@ static zend_string* mb_mime_header_encode(zend_string *input, const
mbfl_encodin
* spaces), just pass it through unchanged */
bool checking_leading_spaces = true;
while (in_len) {
- size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf, 80, &state);
+ size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf, wchar_buf_len, &state);
p = wchar_buf;
e = wchar_buf + out_len;
@@ -5925,9 +5929,9 @@ no_passthrough: ;
* do so all the way to the end of the string */
while (in_len) {
/* Decode part of the input string, refill wchar_buf */
- ZEND_ASSERT(offset < 80);
- size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf + offset, 80 - offset,
&state);
- ZEND_ASSERT(out_len <= 80 - offset);
+ ZEND_ASSERT(offset + MBSTRING_MIN_WCHAR_BUFSIZE <= wchar_buf_len);
+ size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf + offset, wchar_buf_len -
offset, &state);
+ ZEND_ASSERT(out_len <= wchar_buf_len - offset);
p = wchar_buf;
e = wchar_buf + offset + out_len;
/* ASCII output is broken into space-delimited 'words'
@@ -5948,6 +5952,7 @@ no_passthrough: ;
* If we are already too far along on a line to include Base64/QPrint encoded data
* on the same line (without overrunning max line length), then add a line feed
* right now */
+feed_and_mime_encode:
if (mb_convert_buf_len(&buf) - line_start + indent + strlen(outcode->mime_name) > 55)
{
MB_CONVERT_BUF_ENSURE(&buf, buf.out, buf.limit, (e - word_start) + linefeed_len + 1);
buf.out = mb_convert_buf_appendn(buf.out, linefeed, linefeed_len);
@@ -5985,7 +5990,13 @@ no_passthrough: ;
if (in_len) {
/* Copy chars which are part of an incomplete 'word' to the beginning
- * of wchar_buf and reprocess them on the next iteration */
+ * of wchar_buf and reprocess them on the next iteration.
+ * But first make sure that the incomplete 'word' isn't so big that
+ * there will be no space to add any more decoded wchars in the buffer
+ * (which could lead to an infinite loop) */
+ if ((word_start - wchar_buf) < MBSTRING_MIN_WCHAR_BUFSIZE) {
+ goto feed_and_mime_encode;
+ }
offset = e - word_start;
if (offset) {
memmove(wchar_buf, word_start, offset * sizeof(uint32_t));
@@ -6027,17 +6038,17 @@ mime_encoding_needed: ;
/* Do we need to refill wchar_buf to make sure we don't run out of wchars
* in the middle of a line? */
- if (p == wchar_buf) {
+ offset = e - p;
+ if (wchar_buf_len - offset < MBSTRING_MIN_WCHAR_BUFSIZE) {
goto start_new_line;
}
- offset = e - p;
memmove(wchar_buf, p, offset * sizeof(uint32_t));
while(true) {
refill_wchar_buf: ;
- ZEND_ASSERT(offset < 80);
- size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf + offset, 80 - offset,
&state);
- ZEND_ASSERT(out_len <= 80 - offset);
+ ZEND_ASSERT(offset + MBSTRING_MIN_WCHAR_BUFSIZE <= wchar_buf_len);
+ size_t out_len = incode->to_wchar(&in, &in_len, wchar_buf + offset, wchar_buf_len -
offset, &state);
+ ZEND_ASSERT(out_len <= wchar_buf_len - offset);
p = wchar_buf;
e = wchar_buf + offset + out_len;
@@ -6112,22 +6123,18 @@ start_new_line: ;
indent = 0; /* Indent argument must only affect the first line */
- if (in_len) {
- /* We still have more of input string remaining to decode */
+ if (in_len || p < e) {
+ /* We still have more input to process */
buf.out = mb_convert_buf_appendn(buf.out, linefeed, linefeed_len);
buf.out = mb_convert_buf_add(buf.out, ' ');
line_start = mb_convert_buf_len(&buf);
- /* Copy remaining wchars to beginning of buffer so they will be
- * processed on the next iteration of outer 'do' loop */
offset = e - p;
- memmove(wchar_buf, p, offset * sizeof(uint32_t));
- goto refill_wchar_buf;
- } else if (p < e) {
- /* Input string is finished, but we still have trailing wchars
- * remaining to be processed in wchar_buf */
- buf.out = mb_convert_buf_appendn(buf.out, linefeed, linefeed_len);
- buf.out = mb_convert_buf_add(buf.out, ' ');
- line_start = mb_convert_buf_len(&buf);
+ if (in_len && (wchar_buf_len - offset >= MBSTRING_MIN_WCHAR_BUFSIZE)) {
+ /* Copy any remaining wchars to beginning of buffer and refill
+ * the rest of the buffer */
+ memmove(wchar_buf, p, offset * sizeof(uint32_t));
+ goto refill_wchar_buf;
+ }
goto start_new_line;
} else {
/* We are done! */
@@ -6165,7 +6172,7 @@ PHP_FUNCTION(mb_encode_mimeheader)
charset = php_mb_get_encoding(charset_name, 2);
if (!charset) {
RETURN_THROWS();
- } else if (charset->mime_name == NULL || charset->mime_name[0] == '\0') {
+ } else if (charset->mime_name == NULL || charset->mime_name[0] == '\0' || charset
== &mbfl_encoding_qprint) {
zend_argument_value_error(2, "\"%s\" cannot be used for MIME header
encoding", ZSTR_VAL(charset_name));
RETURN_THROWS();
}
diff --git a/ext/mbstring/tests/mb_encode_mimeheader_basic4.phpt
b/ext/mbstring/tests/mb_encode_mimeheader_basic4.phpt
index 7bf05b43ae36b..d9a3407bf3f9f 100644
--- a/ext/mbstring/tests/mb_encode_mimeheader_basic4.phpt
+++ b/ext/mbstring/tests/mb_encode_mimeheader_basic4.phpt
@@ -115,11 +115,29 @@
var_dump(mb_encode_mimeheader("\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\
// In the general case, matching the old implementation's decision to transfer-encode or not
// perfectly would require allocating potentially unbounded scratch memory (up to the size of
// the input string), but we aim to only use a constant amount of temporarily allocated memory
-var_dump(mb_encode_mimeheader("2\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20!3",
"GB18030", "Q", ""));
+var_dump(mb_encode_mimeheader("2\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20\x20!3",
"GB18030", "Q", ""));
+
+// Regression test for infinite loop which was unintentionally caused when refactoring
+var_dump(mb_encode_mimeheader(",9868949,9868978,9869015,9689100,9869121,9869615,9870690,9867116,98558119861183.
", "utf-8", "B"));
+var_dump(mb_encode_mimeheader('xx ' . str_repeat("A", 81) . " ",
"utf-8", "B"));
+
+// Regression test for problem where MIME encoding loop would not leave enough space in wchar
+// buffer for the next iteration, causing an assertion failure
+mb_internal_encoding('MacJapanese');
+var_dump(mb_encode_mimeheader("ne\xf6\xff\xff\xffs\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff1\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff1",
'CP50220', 'B', "A", 44));
+
+// Regression test for failing assertion caused by the fact that QPrint deliberately generates no
+// wchars for CR (0x0D) bytes
+try {
+ mb_internal_encoding('Quoted-Printable');
+ var_dump(mb_encode_mimeheader("=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=00=00=00=00=00=00=00=01=00=00=00=00=00=00=00850r=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=0D=00=00=00=0050r=08=0DCP850r850r0r",
"Quoted-Printable", "B", "", 184));
+} catch (\ValueError $e) {
+ echo $e->getMessage() . \PHP_EOL;
+}
echo "Done";
?>
---EXPECT--
+--EXPECTF--
string(0) ""
string(21) "=?UTF-8?Q?abc=00abc?="
string(16) "=?UTF-8?B?Pw==?="
@@ -156,5 +174,14 @@ string(75) "
111111111111111111111111111111111111111111111111111111111111111111
string(33) "=?HZ-GB-2312?Q?=7E=7Bs=5B=7E=7D?="
string(77) "2
!3"
string(282) "=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20!=33=20?="
-string(296) "2 =?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20!=33?="
+string(344) "2 =?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20?=
=?GB18030?Q?=20=20=20=20=20=20=20=20=20!=33?="
+string(135) "=?UTF-8?B?LDk4Njg5NDksOTg2ODk3OCw5ODY5MDE1LDk2ODkxMDAsOTg2OTEyMSw5ODY5?=
+ =?UTF-8?B?NjE1LDk4NzA2OTAsOTg2NzExNiw5ODU1ODExOTg2MTE4My4g?="
+string(142) "xx =?UTF-8?B?QUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFB?=
+ =?UTF-8?B?QUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBQUFBIA==?="
+string(690) "=?ISO-2022-JP?B?bmU/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/cxskQiFEGyhCPw==?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/MRskQiFEGyhCPxskQiFEGyhCPw==?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/GyRCIUQbKEI/?=A
=?ISO-2022-JP?B?GyRCIUQbKEI/GyRCIUQbKEI/MQ==?="
+
+
+Deprecated: mb_encode_mimeheader(): Handling QPrint via mbstring is deprecated; use
quoted_printable_encode/quoted_printable_decode instead in %s on line %d
+mb_encode_mimeheader(): Argument #2 ($charset) "Quoted-Printable" cannot be used for MIME
header encoding
Done