v8/test/mjsunit/regress/regress-782145.js
jgruber 55a9807682 [string] Fix regexp fast path in MaybeCallFunctionAtSymbol
The regexp fast path in MaybeCallFunctionAtSymbol had an issue in which
we'd call ToString after checking that the given {object} was a fast
regexp and deciding to take the fast path. This is invalid since
ToString() can call into user-controlled JS and may mutate {object}.

There's no way to place the ToString call correctly in this instance:
1 before BranchIfFastRegExp, it's a spec violation if we end up on the
  slow regexp path;
2 the problem with the current location is already described above;
3 and we can't place it into the fast-path regexp builtin (e.g.
  RegExpReplace) either due to the same reasons as 1.

The solution in this CL is to restrict the fast path to string
arguments only, i.e. cases where ToString would be a nop and can safely
be skipped.

Bug: chromium:782145
Change-Id: Ifd35b3a9a6cf2e77c96cb860a8ec98eaec35aa85
Reviewed-on: https://chromium-review.googlesource.com/758257
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49213}
2017-11-08 09:49:33 +00:00

22 lines
715 B
JavaScript

// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
function newFastRegExp() { return new RegExp('.'); }
function toSlowRegExp(re) { re.exec = 42; }
let re = newFastRegExp();
const evil_nonstring = { [Symbol.toPrimitive]: () => toSlowRegExp(re) };
const empty_string = "";
String.prototype.replace.call(evil_nonstring, re, empty_string);
re = newFastRegExp();
String.prototype.match.call(evil_nonstring, re, empty_string);
re = newFastRegExp();
String.prototype.search.call(evil_nonstring, re, empty_string);
re = newFastRegExp();
String.prototype.split.call(evil_nonstring, re, empty_string);