A bug in the SmallVec::insert_many method caused it to allocate a buffer that was smaller than needed. It then wrote past the end of the buffer, causing a buffer overflow and memory corruption on the heap.

This bug was only triggered if the iterator passed to insert_many yielded more items than the lower bound returned from its size_hint method.

The flaw was corrected in smallvec 0.6.14 and 1.6.1, by ensuring that additional space is always reserved for each item inserted. The fix also simplified the implementation of insert_many to use less unsafe code, so it is easier to verify its correctness.

Thank you to Yechan Bae (@Qwaz) and the Rust group at Georgia Tech’s SSLab for finding and reporting this bug.

On Windows, configuring a named pipe server with pipe_mode will force ServerOptions::reject_remote_clients as false.

This drops any intended explicit configuration for the reject_remote_clients that may have been set as true previously.

The default setting of reject_remote_clients is normally true meaning the default is also overridden as false.

Workarounds

Ensure that pipe_mode is set first after initializing a ServerOptions. For example:

let mut opts = ServerOptions::new();
opts.pipe_mode(PipeMode::Message);
opts.reject_remote_clients(true);

When this crate is given a pathological certificate chain to validate, it will spend CPU time exponential with the number of candidate certificates at each step of path building.

Both TLS clients and TLS servers that accept client certificate are affected.

This was previously reported in https://github.com/briansmith/webpki/issues/69 and re-reported recently by Luke Malinowski.

webpki 0.22.1 included a partial fix and webpki 0.22.2 added further fixes.

An attacker can send a flood of CONTINUATION frames, causing h2 to process them indefinitely. This results in an increase in CPU usage.

Tokio task budget helps prevent this from a complete denial-of-service, as the server can still respond to legitimate requests, albeit with increased latency.

More details at "https://seanmonstar.com/blog/hyper-http2-continuation-flood/.

Patches available for 0.4.x and 0.3.x versions.

If a close_notify alert is received during a handshake, complete_io does not terminate.

Callers which do not call complete_io are not affected.

rustls-tokio and rustls-ffi do not call complete_io and are not affected.

rustls::Stream and rustls::StreamOwned types use complete_io and are affected.

idna 0.5.0 and earlier accepts Punycode labels that do not produce any non-ASCII output, which means that either ASCII labels or the empty root label can be masked such that they appear unequal without IDNA processing or when processed with a different implementation and equal when processed with idna 0.5.0 or earlier.

Concretely, example.org and xn--example-.org become equal after processing by idna 0.5.0 or earlier. Also, example.org.xn-- and example.org. become equal after processing by idna 0.5.0 or earlier.

In applications using idna (but not in idna itself) this may be able to lead to privilege escalation when host name comparison is part of a privilege check and the behavior is combined with a client that resolves domains with such labels instead of treating them as errors that preclude DNS resolution / URL fetching and with the attacker managing to introduce a DNS entry (and TLS certificate) for an xn---masked name that turns into the name of the target when processed by idna 0.5.0 or earlier.

Remedy

Upgrade to idna 1.0.3 or later, if depending on idna directly, or to url 2.5.4 or later, if depending on idna via url. (This issue was fixed in idna 1.0.0, but versions earlier than 1.0.3 are not recommended for other reasons.)

When upgrading, please take a moment to read about alternative Unicode back ends for idna.

If you are using Rust earlier than 1.81 in combination with SQLx 0.8.2 or earlier, please also read an issue about combining them with url 2.5.4 and idna 1.0.3.

Additional information

This issue resulted from idna 0.5.0 and earlier implementing the UTS 46 specification literally on this point and the specification having this bug. The specification bug has been fixed in revision 33 of UTS 46.

Acknowledgements

Thanks to kageshiron for recognizing the security implications of this behavior.

In openssl versions before 0.10.70, ssl::select_next_proto can return a slice pointing into the server argument's buffer but with a lifetime bound to the client argument. In situations where the server buffer's lifetime is shorter than the client buffer's, this can cause a use after free. This could cause the server to crash or to return arbitrary memory contents to the client.

openssl 0.10.70 fixes the signature of ssl::select_next_proto to properly constrain the output buffer's lifetime to that of both input buffers.

In standard usage of ssl::select_next_proto in the callback passed to SslContextBuilder::set_alpn_select_callback, code is only affected if the server buffer is constructed within the callback. For example:

Not vulnerable - the server buffer has a 'static lifetime:

builder.set_alpn_select_callback(|_, client_protos| {
    ssl::select_next_proto(b"\x02h2", client_protos).ok_or_else(AlpnError::NOACK)
});

Not vulnerable - the server buffer outlives the handshake:

let server_protos = b"\x02h2".to_vec();
builder.set_alpn_select_callback(|_, client_protos| {
    ssl::select_next_proto(&server_protos, client_protos).ok_or_else(AlpnError::NOACK)
});

Vulnerable - the server buffer is freed when the callback returns:

builder.set_alpn_select_callback(|_, client_protos| {
    let server_protos = b"\x02h2".to_vec();
    ssl::select_next_proto(&server_protos, client_protos).ok_or_else(AlpnError::NOACK)
});

ring::aead::quic::HeaderProtectionKey::new_mask() may panic when overflow checking is enabled. In the QUIC protocol, an attacker can induce this panic by sending a specially-crafted packet. Even unintentionally it is likely to occur in 1 out of every 2**32 packets sent and/or received.

On 64-bit targets operations using ring::aead::{AES_128_GCM, AES_256_GCM} may panic when overflow checking is enabled, when encrypting/decrypting approximately 68,719,476,700 bytes (about 64 gigabytes) of data in a single chunk. Protocols like TLS and SSH are not affected by this because those protocols break large amounts of data into small chunks. Similarly, most applications will not attempt to encrypt/decrypt 64GB of data in one chunk.

Overflow checking is not enabled in release mode by default, but RUSTFLAGS="-C overflow-checks" or overflow-checks = true in the Cargo.toml profile can override this. Overflow checking is usually enabled by default in debug mode.