diff options
author | Mark Dickinson <dickinsm@gmail.com> | 2022-09-04 16:21:18 (GMT) |
---|---|---|
committer | GitHub <noreply@github.com> | 2022-09-04 16:21:18 (GMT) |
commit | b126196838bbaf5f4d35120e0e6bcde435b0b480 (patch) | |
tree | bcae84c9fed45c63d62485a2b66c27f826b3342e /Lib | |
parent | 6adb89f50a0b032e0264cb3cd400a71c0fe6e0f8 (diff) | |
download | cpython-b126196838bbaf5f4d35120e0e6bcde435b0b480.zip cpython-b126196838bbaf5f4d35120e0e6bcde435b0b480.tar.gz cpython-b126196838bbaf5f4d35120e0e6bcde435b0b480.tar.bz2 |
gh-95778: Correctly pre-check for int-to-str conversion (#96537)
Converting a large enough `int` to a decimal string raises `ValueError` as expected. However, the raise comes _after_ the quadratic-time base-conversion algorithm has run to completion. For effective DOS prevention, we need some kind of check before entering the quadratic-time loop. Oops! =)
The quick fix: essentially we catch _most_ values that exceed the threshold up front. Those that slip through will still be on the small side (read: sufficiently fast), and will get caught by the existing check so that the limit remains exact.
The justification for the current check. The C code check is:
```c
max_str_digits / (3 * PyLong_SHIFT) <= (size_a - 11) / 10
```
In GitHub markdown math-speak, writing $M$ for `max_str_digits`, $L$ for `PyLong_SHIFT` and $s$ for `size_a`, that check is:
$$\left\lfloor\frac{M}{3L}\right\rfloor \le \left\lfloor\frac{s - 11}{10}\right\rfloor$$
From this it follows that
$$\frac{M}{3L} < \frac{s-1}{10}$$
hence that
$$\frac{L(s-1)}{M} > \frac{10}{3} > \log_2(10).$$
So
$$2^{L(s-1)} > 10^M.$$
But our input integer $a$ satisfies $|a| \ge 2^{L(s-1)}$, so $|a|$ is larger than $10^M$. This shows that we don't accidentally capture anything _below_ the intended limit in the check.
<!-- gh-issue-number: gh-95778 -->
* Issue: gh-95778
<!-- /gh-issue-number -->
Co-authored-by: Gregory P. Smith [Google LLC] <greg@krypto.org>
Diffstat (limited to 'Lib')
-rw-r--r-- | Lib/test/test_int.py | 82 |
1 files changed, 82 insertions, 0 deletions
diff --git a/Lib/test/test_int.py b/Lib/test/test_int.py index e9561b0..800c0b0 100644 --- a/Lib/test/test_int.py +++ b/Lib/test/test_int.py @@ -1,4 +1,5 @@ import sys +import time import unittest from test import support @@ -632,6 +633,87 @@ class IntStrDigitLimitsTests(unittest.TestCase): with self.assertRaises(ValueError): str(i) + def test_denial_of_service_prevented_int_to_str(self): + """Regression test: ensure we fail before performing O(N**2) work.""" + maxdigits = sys.get_int_max_str_digits() + assert maxdigits < 50_000, maxdigits # A test prerequisite. + get_time = time.process_time + if get_time() <= 0: # some platforms like WASM lack process_time() + get_time = time.monotonic + + huge_int = int(f'0x{"c"*65_000}', base=16) # 78268 decimal digits. + digits = 78_268 + with support.adjust_int_max_str_digits(digits): + start = get_time() + huge_decimal = str(huge_int) + seconds_to_convert = get_time() - start + self.assertEqual(len(huge_decimal), digits) + # Ensuring that we chose a slow enough conversion to measure. + # It takes 0.1 seconds on a Zen based cloud VM in an opt build. + if seconds_to_convert < 0.005: + raise unittest.SkipTest('"slow" conversion took only ' + f'{seconds_to_convert} seconds.') + + # We test with the limit almost at the size needed to check performance. + # The performant limit check is slightly fuzzy, give it a some room. + with support.adjust_int_max_str_digits(int(.995 * digits)): + with self.assertRaises(ValueError) as err: + start = get_time() + str(huge_int) + seconds_to_fail_huge = get_time() - start + self.assertIn('conversion', str(err.exception)) + self.assertLess(seconds_to_fail_huge, seconds_to_convert/8) + + # Now we test that a conversion that would take 30x as long also fails + # in a similarly fast fashion. + extra_huge_int = int(f'0x{"c"*500_000}', base=16) # 602060 digits. + with self.assertRaises(ValueError) as err: + start = get_time() + # If not limited, 8 seconds said Zen based cloud VM. + str(extra_huge_int) + seconds_to_fail_extra_huge = get_time() - start + self.assertIn('conversion', str(err.exception)) + self.assertLess(seconds_to_fail_extra_huge, seconds_to_convert/8) + + def test_denial_of_service_prevented_str_to_int(self): + """Regression test: ensure we fail before performing O(N**2) work.""" + maxdigits = sys.get_int_max_str_digits() + assert maxdigits < 100_000, maxdigits # A test prerequisite. + get_time = time.process_time + if get_time() <= 0: # some platforms like WASM lack process_time() + get_time = time.monotonic + + digits = 133700 + huge = '8'*digits + with support.adjust_int_max_str_digits(digits): + start = get_time() + int(huge) + seconds_to_convert = get_time() - start + # Ensuring that we chose a slow enough conversion to measure. + # It takes 0.1 seconds on a Zen based cloud VM in an opt build. + if seconds_to_convert < 0.005: + raise unittest.SkipTest('"slow" conversion took only ' + f'{seconds_to_convert} seconds.') + + with support.adjust_int_max_str_digits(digits - 1): + with self.assertRaises(ValueError) as err: + start = get_time() + int(huge) + seconds_to_fail_huge = get_time() - start + self.assertIn('conversion', str(err.exception)) + self.assertLess(seconds_to_fail_huge, seconds_to_convert/8) + + # Now we test that a conversion that would take 30x as long also fails + # in a similarly fast fashion. + extra_huge = '7'*1_200_000 + with self.assertRaises(ValueError) as err: + start = get_time() + # If not limited, 8 seconds in the Zen based cloud VM. + int(extra_huge) + seconds_to_fail_extra_huge = get_time() - start + self.assertIn('conversion', str(err.exception)) + self.assertLess(seconds_to_fail_extra_huge, seconds_to_convert/8) + def test_power_of_two_bases_unlimited(self): """The limit does not apply to power of 2 bases.""" maxdigits = sys.get_int_max_str_digits() |