weecrypto_totp_validate compared the generated and client-supplied OTPs
with strcmp and broke out of the time-window loop on the first match.
Both choices leaked information via response timing: strcmp leaked the
expected OTP digit-by-digit (shrinking the brute-force search from
~10^digits to a handful of guesses within the 30-second window), and
the early break leaked which window offset matched.
Compare in constant time with string_memcmp_constant_time and always
iterate the full window, OR-ing the result into otp_ok without an
early exit.
This affects both relay protocols (which call totp_validate via the
public info hook) and any other caller of the info hook.
The IRC relay protocol's PASS handler compared the server password with
the client-supplied value using strcmp, leaking the password byte-by-byte
via response timing. This is the same class of bug fixed for the api and
weechat protocols, on a separate code path that did not go through
relay_auth_check_password_plain.
Extract the HMAC-then-constant-time-compare logic from
relay_auth_check_password_plain into relay_auth_password_equals, then
use it in both the plain-auth wrapper and the IRC PASS handler.
The relay authentication used non-constant-time comparisons (strcasecmp,
strcmp) to verify password hashes and plaintext passwords, allowing an
attacker to derive the expected hash byte-by-byte from response timing
and then authenticate without knowing the password.
- SHA/PBKDF2 hex hash comparisons: normalize the client-supplied hash to
uppercase and compare in constant time over the fixed expected length.
- Plaintext password comparison: HMAC-SHA256 both passwords with a fresh
per-call random key and compare the fixed-size MACs in constant time,
hiding both per-byte timing and the password length.
Add string_memcmp_constant_time helper in core, exposed via the plugin
API. Bump WEECHAT_PLUGIN_API_VERSION accordingly.
An authenticated relay client using the permessage-deflate websocket
extension could send a small compressed frame that decompresses to an
unbounded amount of data, exhausting all memory and crashing WeeChat.
The output buffer in relay_websocket_inflate is now capped to
WEBSOCKET_INFLATE_MAX_SIZE: frames decompressing beyond this limit are
rejected and the connection is closed.
In commit 9a9a262ea1 we moved from
pkg-config to find_package() to work around a deficiency in the pkgsrc
package manager, which does not ship pkg-config files as intended by
CPython.
Modern CMake discourages use of "FOO_LIBRARIES" in all cases, when
imported "interface" libraries can and should be used instead. The meson
equivalent is `dependency()` versus `cc.find_library()`, so this is
certainly a general trend among modern build systems.
An imported interface target, such as the previous PkgConfig::PYTHON,
carries with it the various internal properties such as DEFINITIONS,
INCLUDE_DIRS, or LIBRARIES, and batch applies them. It also avoids
leaking across cmake 2.x style whole-directory scopes.
Use the documented cmake imported interface target for embedding Python
and avoid `add_definitions(${Python_DEFINITIONS})` and similar. As a
bonus, it's also shorter and more concise.
Fixes: 9a9a262ea1
Fixes: https://github.com/weechat/weechat/pull/2251
Signed-off-by: Eli Schwartz <eschwartz@gentoo.org>
In commit 9a9a262ea1 we moved from
pkg-config to find_package() to work around a deficiency in the pkgsrc
package manager, which does not ship pkg-config files as intended by
CPython. In the process, Gentoo and other platforms that, unlike pkgsrc,
publicly support multiple versions of python installed in parallel, had
python version selection broken. Consequently, weechat linked to the
wrong python, which happened to be installed in build chroots but was
not the versioned python package that the weechat package listed as a
dependency. Attempting to install weechat then broke on some systems
(which installed one version of python as a dependency but actually
linked to a totally different one).
This happens due to a design bug in upstream CMake. It is never
conceptually reasonable to use
```
find_package(Python COMPONENTS ...)
```
and omit the "Interpreter" component; if you do, CMake will ignore its
own documentation on how to control the build to use a specific python,
and choose one randomly (== "latest version available"). If, and only
if, the Interpreter component is checked, the development headers /
libraries for python will be guaranteed consistent with the documented
lookup variables from FindPython.cmake's documentation.
Bug: https://bugs.gentoo.org/968814
Fixes: 9a9a262ea1
Fixes: https://github.com/weechat/weechat/pull/2251
Signed-off-by: Eli Schwartz <eschwartz@gentoo.org>
Sébastien Helleu