David Chisnall

I think the Docker comparison is missing the most important part. The OCI container ecosystem (of which Docker is the most well-known part) is huge. It defines container file formats...

Thanks @josejachuf. That seems to have a custom container format and so on. Unfortunately, there's a huge amount of investment in OCI-derived infrastructure across cloud providers (and, as a result,...

LKL uses a small tweak on the generic Linux syscall numbers. x86-64 uses a large tweak to those numbers. Linux does not standardise syscall numbers across architectures. Musl has code...

Once the relayering is done (specifically the pthread work that @vtikoo is doing), we will have two different kinds of %fs regions. For threads created in userspace, libc’s pthread implementation...

I don’t understand what you mean by ‘lthreads without usermode threading’. Lthreads switch fs if the new thread has an itls flag set. Since none of the kernel threads use...

Hmm, that's a good point. That means that @vtikoo will have to make sure not to replace the stack canary value in ethread TLS value (or, rather, to copy that...

We run signals here: https://github.com/lsds/lkl/blob/86819ac5446e1fa31ca67db68aa165bbf07b90a9/arch/lkl/kernel/syscalls.c#L234 This does a loop here, popping signals off the pending list and delivering them: https://github.com/lsds/lkl/blob/86819ac5446e1fa31ca67db68aa165bbf07b90a9/arch/lkl/kernel/signal.c#L68 There are a few things wrong here: - Signals are...

> Note that for synchronous signals, we switch to a host task explicitly This is the correct thing to do for sync signals because they are delivered to an lthread,...

No, #403 makes this worse in the short term.

No, it's not a layering violation, it just makes merging changes from upstream musl slightly harder.