Posts: 148
Joined: Tue Jan 30, 2018 12:18 pm

What compiler options are used for building official binaries?

Thu Jul 11, 2019 10:47 pm

We have a 3B+ on stretch. It is shipped with cryptsetup 1.7.3
Look at the benchmarks (I've ensured that while benchmarking there is no thermal throttling from 1,4 to 1,2 GHz or even below, so they are comparable)
First Output is from official 1.7.3, then the own build 2.1.0. The encryption speed is not really different because it is from the kernel. BTW: Does anybody know how to load other encryption modules? Kernel config says they're build as modules.
But then look at the iterations. They are a lot slower then the prebuild cryptsetup.
1.jpg (169.15 KiB) Viewed 182 times
Then I've installed the cryptsetuppackage from buster. It is also slower than cryptsetup 1.7.3 BTW: Thats also very bad because cryptsetup chooses iterations based on speed. The slower it is on your device the more easily it can be cracked on GPUs. argon2 is made resistant against ASIC/GPU attack, thats why I've updated.
Cryptsetup 2.1.0 is still a lot slower than 1.7.3. Big question: Because of compile options or did the code change that much? But the prebuild cryptsetup is still significantly faster than my own build version.
2.jpg (107.23 KiB) Viewed 182 times

That leads to the questions: Which options are used to compile the official binaries to get these better speed results?

Return to “Advanced users”