openbsd/src/lib/libc/include/thread_private.h, branch OPENBSD_7_9

replace the flockfile backend with a per FILE recursive mutex.

2025-08-04T01:44:33+00:00

the flockfile implementation in thread/rthread_file.c used an external lock, and associated it with the relevant FILE * as needed. this isn't great for a lot of reasons, complexity being the big one, but the straw that broke the camels back is that it uses a single spinlock to coordinate all of this, which in turn generates a lot of sched_yield syscalls. this avoids all the code complexity and the spinlock by just embedding a small __rctmx in every FILE. tested by and ok tb@ jca@ ok claudio@

This thread-aware profiling support diff comes from Yuichiro NAITO.

2025-07-16T16:22:58+00:00

At runtime, profiling data is stored per-thread. Upon termination, the per-thread profiling data pools are merged into a into a single record, which is then written out (using the new kernel-assisted system). I think the original gmon merging parts may have come from or been inspired by someone at NetBSD. This has been delayed because there were concerns about adding a TLS object which would require other library cranks, so this rides all the major cranks today. ok kettenis tb claudio sthen etc etc etc

add libc specific cmtx and rcmtx locks for internal use.

2025-07-12T23:59:44+00:00

__cmtx provides mutual exclusion using futex(2) and cas on archs that support it, or _spinlocks on the rest. __rcmtx is a recursive mutex built on top of __cmtx, so it inherits the use of futex and cas/spinlock from __cmtx. until now the options we had for locking between threads in libc were spinlocks or pthread mutexes. spinlocks use sched_yield to relax if they have to spin on a contended lock, which we are trying to minimise the use of as much as possible. pthread_mutex is relatively large in memory and offers a lot of complicated features which are unecessary for most of libc. the non cas/futex version of pthread_mutexes currently relies on __thrsleep and __thrwakeup, which we also want to deprecate. having a small futex based lock available everywhere will help us move away from overuse of spinlocks, and deprecate __thrsleep and __thrwakeup. ok kettenis@ jca@ provided his eyes too.

Use struct __sFILE instead of FILE in thread locking callback

2024-08-18T02:25:51+00:00

declarations to reduce pollution. Declare __isthreaded in thread_private.h where it's really needed. ok deraadt@

Fix two issues related to thread private data in asr.

2021-01-06T19:54:17+00:00

- setting up asr in single thread mode and then starting threads using asr would lead to multiple threads sharing the same resolver. - destruction of a thread that has been using asr would leak data. Problem originally reported by Alexey Sokolov and Uli Schlachter. ok kettenis@

New futex(2) based rwlock implementation based on the mutex code.

2019-02-13T13:22:14+00:00

This implementation reduces contention because threads no longer need to spin calling sched_yield(2) before going to sleep. Tested by many, thanks! ok visa@, pirofti@

Make the "not my pool" searching loop a tiny bit smarter, while

2019-01-10T18:45:33+00:00

making the number of pools variable. Do not document the malloc conf settings atm, don't know yet if they will stay. Thanks to all the testers. ok deraadt@

Implement __cxa_thread_atexit to support C++11 thread_local scope. The

2017-12-05T13:45:31+00:00

interface is also made available as __cxa_thread_atexit_impl to satisfy the needs of GNU libstdc++. ok guenther@, millert@

Revert recent changes to unbreak ports/net/samba

2017-11-04T22:53:57+00:00

While it is not clear (to me) why that ports ends up with corrupted shared libs, reverting those changes fixes the issue and should allow us to close p2k17 more smoothly. Discussed with a bunch, ok ajacoutot@ guenther@

Change pthread_cleanup_{push,pop} to macros that store the cleanup info

2017-10-28T21:23:14+00:00

on the stack instead of mallocing the list and move the APIs from libpthread to libc so that they can be used inside libc. Note: the standard was explicitly written to permit/support this "macro with unmatched brace" style and it's what basically everyone else already does. We xor the info with random cookies with a random magic to detect/trip-up overwrites. Major bump to both libc and libpthread due to the API move. ok mpi@