I see where I've not explained myself properly. The existing code I'm referring to is the code that right now handles the TLS sessions in apache/nginx. That's the code I'm suggesting could be run from a forked process instead of in the main process. To need IPC to offload the RSA crypto you'd need to be doing Apache->TLS session code->fork->RSA operations. I'm saying you could do Apache->fork->TLS session code. Just run all your TLS sessions in a different process with the normal single process, no PKCS#11 GnuTLS/OpenSSL code. Is that not feasible?
To do that Apache needs some form of internal IPC to communicate its TLS sessions to the forked process. Maybe that's more complex than forking and doing IPC at the PKCS#11 driver level? Don't know.
> To do that Apache needs some form of internal IPC to communicate its TLS sessions to the forked process. Maybe that's more complex than forking and doing IPC at the PKCS#11 driver level? Don't know.
Yes.
Also, bear in mind that you can't just fork and continue running in modern software.
A process shall be created with a single thread. If a multi-threaded process calls fork(), the new process shall contain a replica of the calling thread and its entire address space, possibly including the states of mutexes and other resources. Consequently, to avoid errors, the child process may only execute async-signal-safe operations until such time as one of the exec functions is called.
You'd be forking at the start of the Apache launch before any connections so that shouldn't be much of an issue.
This construct has a much worse bug. It separates the TLS from the rest of apache so it protects against bugs in other parts of the server (HTTP parsing for example) but it doesn't separate the TLS session code from the crypto primitives, so it wouldn't protect against heartbleed. For that forking at the PKCS#11 boundary would be much safer.
Thinking about it a better OpenSSL patch than the Akamai one of protecting the memory with a different alocator would be to run the actual crypto in a different process with a well-defined IPC between that and the main library. That would give you much of the safety of a software HSM without any changes to Apache/nginx or any other TLS server.
Actually, thinking about it some more forking at PKCS#11 driver will not fix heartbleed completely. It will stop the key being recovered but will still allow you to recover passwords and cookies. To fix it completely you'd need forking at both ends, or just using apache in forking instead of event mode.
To do that Apache needs some form of internal IPC to communicate its TLS sessions to the forked process. Maybe that's more complex than forking and doing IPC at the PKCS#11 driver level? Don't know.