Massive performance regression with openssl 3.0/3.1

[hannob]

Parsing private keys is around 70 times slower with openssl 3.0 and around 10 times slower with openssl 3.1. It does not appear that it has any priority for OpenSSL to fix these issues [1].

This is a regression in the underlying openssl functions. There is unfortunately little we can do here short of implementing a separate key parser, but opening this bug to track the problem. For the time being, it is recommended to run with openssl 1.1 if possible.

[1] openssl/openssl#15199

[hannob]

Update: This got a bit better with 3.2/3.3. Those are around 3 times slower than 1.1. That's still obviously not great, but somewhat more acceptable. For now, avoid using blocklistmaker on openssl 3.0/3.1.

There is ongoing discussion in the upstream bug about potential architectural changes, so hopefully this will get adressed better in a future openssl release.