5.3.0 regression: SSL_set_cert_cb(..., NULL, ...) stores a null function pointer instead of clearing, crashing the next handshake
Los mantenedores suelen responder en 1 día
Evaluación
- Dificultad
- 3/5
- Tiempo estimado
- Medio día
- Aptitud para principiantes
- 62/100
Línea de trabajo
Start in ssl/ssl_cert.cc at ssl_cert_set_cert_cb (both overloads), where a null callback is stored as an OldCallback instead of resetting to std::monostate, and check CertCb::operator bool() in ssl/internal.h. Then trace do_cert_callback in ssl/handshake_server.cc. A regression test should call SSL_set_cert_cb(ssl, NULL, NULL) and complete a server handshake without invoking a callback. Done when clearing with NULL leaves the callback absent and that test passes, with any boring-sys/patches/ entry kept consistent with the upstream fix.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
Summary
In boring / boring-sys 5.3.0 (BoringSSL revision 427ec40c), clearing a certificate callback by passing NULL no longer resets it to the empty state. The null pointer is stored as an OldCallback alternative inside CertCb::cb, so CertCb::operator bool() reports the callback as present, and do_cert_callback then invokes a null function pointer. On our machines this crashes the server process during the resumed TLS handshake and currently keeps us from upgrading to 5.3.0.
Root cause
Introduced by google/boringssl commit 902518a787 ("ssl: Allow certificate selection callback to pick fatal alert"), which turned CERT::cert_cb into struct CertCb { std::variant<std::monostate, OldCallback, NewCallback> }:
// ssl/ssl_cert.cc, ssl_cert_set_cert_cb (both overloads)
- cert->cert_cb = cb;
+ cert->cert_cb.cb = cb;
Assigning a null function pointer selects the OldCallback alternative by exact type match rather than monostate, so the "cleared" representation is no longer reachable through this setter. CertCb::operator bool() (ssl/internal.h) only excludes monostate, and do_cert_callback (ssl/handshake_server.cc) calls the callback whenever that conversion is true.
Reproduction (real world)
Pingora's resumable TLS accept uses exactly this pattern: it arms a blocking cert_cb (suspend_when_need_ssl_cert), pauses the handshake, installs the certificate, then clears the callback with SSL_set_cert_cb(ssl, NULL, NULL) before resuming (unblock_ssl_cert in pingora-boringssl). With 5.3.0 every resumed handshake crashes:
bssl::CertCb::operator()(ssl_st*, void*, unsigned char*) const
bssl::do_cert_callback(bssl::SSL_HANDSHAKE*)
bssl::ssl_server_handshake(bssl::SSL_HANDSHAKE*)
...
SSL_do_handshake
SSL_accept
(excerpt from the macOS crash report: EXC_BAD_ACCESS, invalid address 0x0)
Bisecting boring/boring-sys back to 5.2.0 makes all of this disappear: 33 failing integration tests go back to green. SSL_CTX_set_cert_cb(ctx, NULL, arg) alone should be enough to hit the same path on the next server handshake.
Suggested fix
Normalize a null callback to the empty state in ssl_cert_set_cert_cb, restoring the previous "NULL clears" behavior:
if (cb == nullptr) {
cert->cert_cb.cb = std::monostate();
} else {
cert->cert_cb.cb = cb;
}
(same for the _ex overload.)
Notes
- The same code is present on google/boringssl
mainas of 2026-10-15. - Reproduced on macOS 27.0.1 (26A434) aarch64, rustc 1.99.0; the defect in the C++ itself is platform-independent.
- Happy to send a PR that carries the change as a
boring-sys/patches/entry with a regression test, if that is the preferred route — or to help get it into BoringSSL if you would rather pick it up with the next revision bump.
- Lenguaje dominante
- Rust
- Estrellas
- 526
- Forks
- 178
- Merge medio
- 15 h 34 min
- PR fusionados (30 d)
- 9
Preparar el entorno
- Sin Dockerfile ni archivo de Docker Compose
- Sin plantilla de pull request
- Leer la guía de contribución
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de cloudflare/boring
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 82/100
cloudflare/boring#527 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
cloudflare/boring#526 ·
Los mantenedores suelen responder en 1 día
-
SLH-DSA supportPosiblemente ocupada Un pull request vinculado a esta issue está abierto o ya se fusionó. Abierto
Dificultad 4/5 3-5 días Aptitud para principiantes 45/100
cloudflare/boring#522 ·
Los mantenedores suelen responder en 1 día
-
Build failure on 32 bit freebsdAbierto
Dificultad 4/5 3-5 días Aptitud para principiantes 45/100
cloudflare/boring#517 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 4/5 3-5 días Aptitud para principiantes 45/100
cloudflare/boring#500 · 2 comentarios ·
Los mantenedores suelen responder en 1 día
Todos los issues de cloudflare/boring
Issues similares
-
bug
Dificultad 2/5 1-3 horas Aptitud para principiantes 72/100
Los mantenedores suelen responder en 1 día
-
needs: triage type: bug
Dificultad 2/5 1-3 horas Aptitud para principiantes 70/100
Los mantenedores suelen responder en 1 día
-
area:docs documentation P3
Dificultad 1/5 Menos de una hora Aptitud para principiantes 88/100
Raaif-Yousuf/Wingman#837 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 80/100
Los mantenedores suelen responder en 1 día
-
priority: medium
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
storytold/photocraft#6759 ·
Los mantenedores suelen responder en 1 día