Solved! 🥳

This was a pretty "interesting" bug. Remember when I invented a way to implement #async / #await in #C, for jobs running on a threadpool. Back then I said it only works when completion of the task resumes execution on the same pool thread.

Trying to improve overall performance, I found the complex logic to identify the thread job to put on a pool thread a real deal-breaker. Just having one single MPMC queue with a single semaphore for all pool threads to wait on is a lot more efficient. But then, a job continued after an awaited task will resume on a "random" thread.

It theoretically works by making sure to restore the CORRECT context (the original one of the pool thread) every time after executing a job, whether partially (up to the next await) or completely.

Only it didn't, at least here on #FreeBSD, and I finally understood the reason for this was that I was using #TLS (thread-local storage) to find the context to restore.

Well, most architectures store a pointer to the current thread metadata in a register. #POSIX user #context #switching saves and restores registers. I found a source claiming that the #Linux ( #glibc) implementation explicitly does NOT include the register holding a thread pointer. Obviously, #FreeBSD's implementation DOES include it. POSIX doesn't have to say anything about that.

In short, avoiding TLS accesses when running with a custom context solved the crash. 🤯

Today, I implemented the #async / #await pattern (as known from #csharp and meanwhile quite some other languages) ...

... in good old #C! 😎

Well, at least sort of.

* It requires some standard library support, namely #POSIX user context switching with #getcontext and friends, which was deprecated in POSIX-1.2008. But it's still available on many systems, including #FreeBSD, #NetBSD, #Linux (with #glibc). It's NOT available e.g. on #OpenBSD, or Linux with some alternative libc.

* I can't do anything about the basic language syntax, so some boilerplate comes with using it.

* It has some overhead (room for extra stacks, even extra syscalls as getcontext unfortunately also always saves/restores the signal mask)

But then ... async/await in C! 🥳

Here are the docs:
https://zirias.github.io/poser/api/latest/class_p_s_c___async_task.html

#C #coding

I finally eliminated the need for a dedicated #thread controlling the pam helper #process in #swad. 🥳

The building block that was still missing from #poser was a way to await some async I/O task performed on the main thread from a worker thread. So I added a class to allow exactly that. The naive implementation just signals the main thread to carry out the requested task and then waits on a #semaphore for completion, which of course blocks the worker thread.

Turns out we can actually do better, reaching similar functionality like e.g. #async / #await in C#: Release the worker thread to do other jobs while waiting. The key to this is user context switching support like offered by #POSIX-1.2001 #getcontext and friends. Unfortunately it was deprecated in POSIX-1.2008 without an obvious replacement (the docs basically say "use threads", which doesn't work for my scenario), but still lots of systems provide it, e.g. #FreeBSD, #NetBSD, #Linux (with #glibc) ...

The posercore lib now offers both implementations, prefering to use user context switching if available. It comes at a price: Every thread job now needs its private stack space (I allocated 64kiB there for now), and of course the switching takes some time as well, but that's very likely better than leaving a task idle waiting. And there's a restriction, resuming must still happen on the same thread that called the "await", so if this thread is currently busy, we have to wait a little bit longer. I still think it's a very nice solution. 😎

In any case, the code for the PAM credential checker module looks much cleaner now (the await "magic" happens on line 174):
https://github.com/Zirias/swad/blob/57eefe93cdad0df55ebede4bd877d22e7be1a7f8/src/bin/swad/cred/pamchecker.c

#C #coding