Michał Trojnara comments

Results 96 comments of


                                            Michał Trojnara

Memory leak when PKCS11 is enabled in openssl.cnf

@vgottardi2 Do you use libp11 on an operating system that does not automatically release memory at application exit?

Provide Windows binaries

OpenSSL does not provide official Windows installers: https://wiki.openssl.org/index.php/Binaries Would you like to sponsor building Windows windows installers for libp11 releases?

Provide Windows binaries

Provided in https://github.com/OpenSC/libp11/releases/tag/libp11-0.4.11

Provide Windows binaries

I'm working on automatic Windows builds with GitHub Actions. Please be patient.

Provide Windows binaries

Please take a look at Artifacts at the bottom of https://github.com/OpenSC/libp11/actions/runs/2856550857#artifacts. What do you think?

[BROKEN] Attempt to refcount the engine with EVP_PKEYs we give out

With all due respect, this PR is insane. Does any other engine do it this way?

[BROKEN] Attempt to refcount the engine with EVP_PKEYs we give out

> Allegedly they all do I guess I should have been more precise... I checked OpenSSL 1.1.0 engines for `RSA_new_method()` invocations. They are used by `e_capi.c` and `e_chil.c`. None of...

[BROKEN] Attempt to refcount the engine with EVP_PKEYs we give out

... or even better to obsolete them (for compatibility) and add new `PKCS11_engine_get_private_key()` and `PKCS11_engine_get_public_key()`.

[BROKEN] Attempt to refcount the engine with EVP_PKEYs we give out

I can see quite a few problems here. Which of them is the fundamental one for you? I guess we can implement the fall through to the OpenSSL default methods...

[BROKEN] Attempt to refcount the engine with EVP_PKEYs we give out

Existing engines (`e_capi.c`, `e_chil.c`, `e_dasync.c`) **do** call `ENGINE_set_RSA()`. If this is a problem then it is bigger than just our engine.