`bigIntArrayToUnsignedInt` used the signed `<<` operator, so when the most significant byte of a 4-byte input had its top bit set (e.g. `[0x80, 0x00, 0x00, 0x01]`) the result was a negative Int32 instead of the intended unsigned 32-bit value. This caused any RSA `publicExponent` exactly 4 bytes long with the top bit set to be parsed incorrectly. Coerce the final value with `>>> 0` and add a unit test. Assisted-by: Claude <noreply@anthropic.com> Co-Authored-By: Claude <noreply@anthropic.com> Co-Authored-By: DeepView Autofix <276251120+deepview-autofix@users.noreply.github.com> Co-Authored-By: Nikita Skovoroda <chalkerx@gmail.com> Signed-off-by: Nikita Skovoroda <chalkerx@gmail.com> PR-URL: https://github.com/nodejs/node/pull/62839 Reviewed-By: Filip Skokan <panva.ip@gmail.com>
80 lines
2.2 KiB
JavaScript
80 lines
2.2 KiB
JavaScript
// Flags: --expose-internals
|
|
'use strict';
|
|
|
|
const common = require('../common');
|
|
if (!common.hasCrypto)
|
|
common.skip('missing crypto');
|
|
|
|
const assert = require('assert');
|
|
|
|
const {
|
|
bigIntArrayToUnsignedInt,
|
|
normalizeAlgorithm,
|
|
validateKeyOps,
|
|
} = require('internal/crypto/util');
|
|
|
|
// bigIntArrayToUnsignedInt must return an unsigned 32-bit value even when
|
|
// the most significant byte has its top bit set. Otherwise the signed `<<`
|
|
// operator yields a negative Int32 for inputs like [0x80, 0x00, 0x00, 0x01].
|
|
{
|
|
assert.strictEqual(
|
|
bigIntArrayToUnsignedInt(new Uint8Array([0x80, 0x00, 0x00, 0x01])),
|
|
0x80000001);
|
|
assert.strictEqual(
|
|
bigIntArrayToUnsignedInt(new Uint8Array([0xff, 0xff, 0xff, 0xff])),
|
|
0xffffffff);
|
|
assert.strictEqual(
|
|
bigIntArrayToUnsignedInt(new Uint8Array([1, 0, 1])),
|
|
65537);
|
|
assert.strictEqual(
|
|
bigIntArrayToUnsignedInt(new Uint8Array([1, 0, 0, 0, 0])),
|
|
undefined);
|
|
}
|
|
|
|
{
|
|
// Check that normalizeAlgorithm does not mutate object inputs.
|
|
const algorithm = { name: 'ECDSA', hash: 'SHA-256' };
|
|
assert.strictEqual(normalizeAlgorithm(algorithm, 'sign') !== algorithm, true);
|
|
assert.deepStrictEqual(algorithm, { name: 'ECDSA', hash: 'SHA-256' });
|
|
}
|
|
|
|
// The algorithm name getter should only be invoked once during
|
|
// normalizeAlgorithm, including for algorithms with a non-null desiredType
|
|
// where step 6 runs the specialized dictionary converter.
|
|
// Refs: https://github.com/web-platform-tests/wpt/pull/57614#pullrequestreview-3808145365
|
|
{
|
|
let nameReadCount = 0;
|
|
const algorithm = {
|
|
get name() {
|
|
nameReadCount++;
|
|
return 'AES-GCM';
|
|
},
|
|
iv: new Uint8Array(12),
|
|
};
|
|
const normalized = normalizeAlgorithm(algorithm, 'encrypt');
|
|
assert.strictEqual(normalized.name, 'AES-GCM');
|
|
assert.strictEqual(nameReadCount, 1);
|
|
}
|
|
|
|
{
|
|
let nameReadCount = 0;
|
|
const algorithm = {
|
|
get name() {
|
|
nameReadCount++;
|
|
return 'ECDSA';
|
|
},
|
|
hash: 'SHA-256',
|
|
};
|
|
const normalized = normalizeAlgorithm(algorithm, 'sign');
|
|
assert.strictEqual(normalized.name, 'ECDSA');
|
|
assert.strictEqual(nameReadCount, 1);
|
|
}
|
|
|
|
{
|
|
for (const ops of [
|
|
['sign', 'toString', 'constructor'],
|
|
['sign', '__proto__', 'constructor'],
|
|
]) {
|
|
validateKeyOps(ops);
|
|
}
|
|
}
|