[tbb-commits] [tor-browser/tor-browser-60.2.0esr-8.5-1] Bug 1489744 - Fix a bounds violation crash in the prefs parser. r=glandium, a=RyanVM
gk at torproject.org
gk at torproject.org
Thu Sep 20 08:52:53 UTC 2018
commit e4f063eb30d140dcf0c04074dfe4ab82307fd7c3
Author: Nicholas Nethercote <nnethercote at mozilla.com>
Date: Tue Sep 11 09:36:07 2018 +1000
Bug 1489744 - Fix a bounds violation crash in the prefs parser. r=glandium, a=RyanVM
Currently, if a get_char() call returns EOF, the index moves beyond the
buffer's bounds and get_char() cannot be called again without triggering a
panic. As a result, everywhere that encounters an EOF and then does subsequent
parsing ungets the EOF... except there was one place that failed to do that:
the match case for CharKind::Slash in get_token(). This meant that a single '/'
at the end of the input could trigger a bounds violation (but only if it is the
start of a new token).
This EOF-unget requirement is subtle and easy to get wrong, so this patch
eliminates it. get_char() now can be called repeatedly after an EOF, and will
return EOF on each subsequent call. This means that some of the existing
unget_char() calls can be removed. Some others are still necessary to get line
numbers correct in error messages, but the outcome of mishandled cases is now
much less drastic -- an incorrect line number in an error message instead of a
panic.
The patch also clarifies a couple of related comments.
--HG--
extra : source : 5097db630d1fb51f7f3fcd5523a924fa823e77ce
extra : intermediate-source : 2676e53112584b2c08b9a1bf02e15cd9bb0d0ee5
---
modules/libpref/parser/src/lib.rs | 70 +++++++++++++++++++----------------
modules/libpref/test/gtest/Parser.cpp | 9 +++--
2 files changed, 45 insertions(+), 34 deletions(-)
diff --git a/modules/libpref/parser/src/lib.rs b/modules/libpref/parser/src/lib.rs
index 8d8b9a6125e4..5ea7650252f5 100644
--- a/modules/libpref/parser/src/lib.rs
+++ b/modules/libpref/parser/src/lib.rs
@@ -54,16 +54,10 @@
// This parser uses several important optimizations.
//
-// - Because "'\0' means EOF" is part of the grammar (see above) we can match
-// EOF as a normal char/token, which means we can avoid a typical "do we
-// still have chars remaining?" test in get_char(), which gives a speedup
-// because get_char() is a very hot function. (Actually, Rust would
-// bounds-check this function anyway, so we have get_char_unchecked() which
-// is used for the two hottest call sites.)
-//
-// This also means EOF is representable by a u8. If EOF was represented by an
-// out-of-band value such as -1 or 256, we'd have to return a larger type
-// such as u16 or i16 from get_char().
+// - Because "'\0' means EOF" is part of the grammar (see above), EOF is
+// representable by a u8. If EOF was represented by an out-of-band value such
+// as -1 or 256, we'd have to return a larger type such as u16 or i16 from
+// get_char().
//
// - When starting a new token, it uses a lookup table with the first char,
// which quickly identifies what kind of token it will be. Furthermore, if
@@ -192,7 +186,7 @@ enum Token {
#[derive(Clone, Copy, PartialEq)]
enum CharKind {
// These are ordered by frequency. See the comment in GetToken().
- SingleChar, // Unambiguous single-char tokens: [()+,-]
+ SingleChar, // Unambiguous single-char tokens: [()+,-] or EOF
SpaceNL, // [\t\v\f \n]
Keyword, // [A-Za-z_]
Quote, // ["']
@@ -535,13 +529,21 @@ impl<'t> Parser<'t> {
#[inline(always)]
fn get_char(&mut self) -> u8 {
- let c = self.buf[self.i];
- self.i += 1;
- c
+ // We do the bounds check ourselves so we can return EOF on failure.
+ // (Although the buffer is guaranteed to end in an EOF char, we might
+ // go one char past that, whereupon we must return EOF again.)
+ if self.i < self.buf.len() {
+ let c = unsafe { *self.buf.get_unchecked(self.i) };
+ self.i += 1;
+ c
+ } else {
+ debug_assert!(self.i == self.buf.len());
+ EOF
+ }
}
- // This function skips the bounds check in non-optimized builds. Using it
- // at the hottest two call sites gives a ~15% parsing speed boost.
+ // This function skips the bounds check in optimized builds. Using it at
+ // the hottest two call sites gives a ~15% parsing speed boost.
#[inline(always)]
unsafe fn get_char_unchecked(&mut self) -> u8 {
debug_assert!(self.i < self.buf.len());
@@ -568,9 +570,11 @@ impl<'t> Parser<'t> {
#[inline(always)]
fn match_single_line_comment(&mut self) {
loop {
- // To reach here, the previous char must have been '/', and
- // assertions elsewhere ensure that there must be at least one
- // subsequent char (the '\0' for EOF).
+ // To reach here, the previous char must have been '/' (if this is
+ // the first loop iteration) or non-special (if this is the second
+ // or subsequent iteration), and assertions elsewhere ensure that
+ // there must be at least one subsequent char after those chars
+ // (the '\0' for EOF).
let c = unsafe { self.get_char_unchecked() };
// All the special chars have value <= b'\r'.
@@ -588,8 +592,6 @@ impl<'t> Parser<'t> {
break;
}
EOF => {
- // Unget EOF so subsequent calls to get_char() are safe.
- self.unget_char();
break;
}
_ => continue
@@ -615,8 +617,6 @@ impl<'t> Parser<'t> {
self.match_char(b'\n');
}
EOF => {
- // Unget EOF so subsequent calls to get_char() are safe.
- self.unget_char();
return false
}
_ => continue
@@ -630,11 +630,10 @@ impl<'t> Parser<'t> {
for _ in 0..ndigits {
value = value << 4;
match self.get_char() {
- c @ b'0'... b'9' => value += (c - b'0') as u16,
+ c @ b'0'...b'9' => value += (c - b'0') as u16,
c @ b'A'...b'F' => value += (c - b'A') as u16 + 10,
c @ b'a'...b'f' => value += (c - b'a') as u16 + 10,
_ => {
- // Unget in case the char was a closing quote or EOF.
self.unget_char();
return None;
}
@@ -716,7 +715,15 @@ impl<'t> Parser<'t> {
return Token::Error("unterminated /* comment");
}
}
- _ => return Token::Error("expected '/' or '*' after '/'")
+ c @ _ => {
+ if c == b'\n' || c == b'\r' {
+ // Unget the newline char; the outer loop will
+ // reget it and adjust self.line_num
+ // appropriately.
+ self.unget_char();
+ }
+ return Token::Error("expected '/' or '*' after '/'");
+ }
}
continue;
}
@@ -871,9 +878,12 @@ impl<'t> Parser<'t> {
}
continue; // We don't want to str_buf.push(c2) below.
}
- _ => {
- // Unget in case the char is an EOF.
- self.unget_char();
+ c @ _ => {
+ if c == b'\n' || c == b'\r' {
+ // Unget the newline char; the outer loop will
+ // reget it and adjust self.line_num appropriately.
+ self.unget_char();
+ }
self.string_error_token(
&mut token, "unexpected escape sequence character after '\\'");
continue;
@@ -894,8 +904,6 @@ impl<'t> Parser<'t> {
}
} else if c == EOF {
- // Unget EOF so subsequent calls to get_char() are safe.
- self.unget_char();
self.string_error_token(&mut token, "unterminated string literal");
break;
diff --git a/modules/libpref/test/gtest/Parser.cpp b/modules/libpref/test/gtest/Parser.cpp
index d5469dd38007..fe9dea87ee60 100644
--- a/modules/libpref/test/gtest/Parser.cpp
+++ b/modules/libpref/test/gtest/Parser.cpp
@@ -293,11 +293,14 @@ pref("string.bad-keyword", TRUE);
"test:3: prefs parse error: unterminated /* comment\n"
);
- // Malformed comment.
+ // Malformed comments (single slashes), followed by whitespace, newline, EOF.
DEFAULT(R"(
-/ comment
- )",
+/ comment;
+/
+; /)",
"test:2: prefs parse error: expected '/' or '*' after '/'\n"
+ "test:3: prefs parse error: expected '/' or '*' after '/'\n"
+ "test:4: prefs parse error: expected '/' or '*' after '/'\n"
);
// C++-style comment ending in EOF (1).
More information about the tbb-commits
mailing list