This incarnation of the pthreads library is ancient and not useful, and
should have been mothballed some time ago...

Copyrights added to each file.

libpthread: Link with -Wl,-z,nodelete.

Can't safely unload libpthread because of the interaction with libc
thread stubs.

PR lib/59784: dlopening and dlclosing libpthread is broken

revert previous; needs more discussion/testing.

Make pthread_setname_np variadic for compatibility with FreeBSD and GLIBC.
Bump.

libpthread: fix lint build

libpthread: Fix expected symbols on alpha and hppa.

These have PTHREAD__ASM_RASOPS and define a different subset of
symbols which get exported. None of these pthread__ras_* and
pthread__lock_* symbols _should_ be exported but we can fix that in a
future revbump.

PR lib/58838: shared libraries in base should all have expsym lists

branches: 1.98.2;
libpthread/Makefile: trailing whitespace

fix rump URLs

Link to https://github.com/rumpkernel/ instead of
a site now taken over by an SEO squatter.

Per discussion on github.com/rumpkernel issues with pooka.

PR misc/57501

branches: 1.96.8;
Update LLVM to 10.0.0git (01f3a59fb3e2542fce74c768718f594d0debd0da)

Add USE_SHLIBDIR=yes to a number of Makefiles for the libraries used
by /sbin/{zfs,zpool,mount_zfs}. The general effect is to move them
from /usr/lib to /lib. Compatibility links are installed in /usr/lib
and nothing that is installed, say in /usr/pkg, appears to break.

With this, it is possible to have a /var and /usr mount using ZFS
legacy mounting early on in the boot process.

Run tested on amd64 and i386 and compile tested on evbarm.

branches: 1.94.2;
Make CLEANFILES actually work. .TARGET is not defined when not in a target
rule.

Thanks xtos for the heads up.

Replace the link command for libpthread.a so that we create a single section
with all the libpthread symbols in it.
This makes -lpthread behave like to -Wl,--whole-archive -lpthread.

This avoids a situation where threaded static binaries use some libc thread
stubs, which are racy.

Fixes PR lib/54001: call_once2_32, call_once2_static test cases failing on
amd64 since gcc7 import.

Suggested by Jonathan Wakely, thanks!

Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

Implement support for "pshared" POSIX semaphores.

Fixes lib/53273 (and Firefox's multi-process tab feature).

branches: 1.90.2;
But set NOSANITIZER

We can't build those three with MKSANITIZER, they are used too early.

branches: 1.88.2;
remove hard-coded -fomit-frame-pointer for pthread stuff,
let these use the same setting as the rest of the tree.
the performance difference is marginal and this allows
dtrace ustack() to work better.

GSoC 2016 Charles Cui: Implement thread priority protection based on work
by Andy Doran. Also document the get/set pshared thread calls as not
implemented, and add a skeleton implementation that is disabled.
XXX: document _sched_protect(2).

Add pthread_getcpuclockid(3)

Allow for arbitrary MI scheduler implementations.

A concrete result is enabling unpatched libpthread to run on the
rumprun stacks (e.g. Xen and bare metal) with a non-NetBSD scheduler.
Those schedulers hook into the existing _lwp_frobnitz() NetBSD syscall
interfaces (well, "syscall" interfaces in that scenario ;)

More specifically about the change itself:

1) instead of calling _lwp_makecontext() followed by _lwp_create()
and passing the entry point in ucontext_t (MD) through the calls, roll
the calls into pthread__makelwp() and allow alternate implementations
for that MI interface.

2) allow compile-time overriding of __lwp_gettcb_fast() or
__lwp_getprivate_fast, which are inline and leak MD scheduler/thread
details into libpthread


Additionally, two small nits:

I) define LIB=pthread before including mk.conf so that it's possible
to test for LIB==pthread in mk.conf

II) make it possible to leave out pthread_cancelstub.c. This is required
by the current implementation of rumprun-posix (i.e. rumprun on
POSIX hosts) due to symbol collisions. It needs to be fixed properly
some day, but for now allows an almost-correct libpthread to run.
I am sure @justin will be happy to explain the details ;)


no change to NetBSD
tested: anita+atf

MKCOMPAT fixes for when compat MACHINE_CPU != normal MACHINE_CPU

Use MACHINE_CPU for m68k. Use similar logic for both tests.

Add support for an optional ARCHSUBDIR/Makefile.inc

Some architectures can't create unwind tables without the frame pointer,
so don't use -fomit-frame-pointer on those.

- Allow libpthread to be dlopened again, by providing libc stubs to libpthread.
- Fail if the dlopened libpthread does pthread_create(). From manu@
- Discussed at length in the mailing lists; approved by core@
- This was chosen as the least intrusive patch that will provide
the necessary functionality.
XXX: pullup to 6

Explain what PTHREAD__COMPAT is for. This comment in
src/libpthread/Makefile is adated from a comment in
src/doc/CHANGES.prev.

Add pthread_condattr_setclock(3) man page link.

branches: 1.77.2;
-Wfoo goes into CWARNFLAGS

Disable new -Wstring-plus-int warning where needed for now.
Adjust various places that add GCC-only options to check for the active
compiler first.

These directories default to WARNS?=5

Move pthread_once implementation into a separate file, it doesn't depend
on the mutex implementation in any way.

branches: 1.73.2;
document pthread_curcpu_np

branches: 1.72.2;
document non-literal format strings

Add support for DF_1_BIND_NOW, DF_1_NODELETE and DF_1_NOOPEN marked
objects, and the RTLD_NODELETE and RTLD_NOLOAD flags to dlopen(3).

Mark libpthread as DF_1_NOOPEN and use it to test the functionality.

Somewhat taken from FreeBSD.

Fixes PR 42029.

OK from christos and joerg.

document pthread_getattr_np

Rationalize by moving

pthread_key_delete(3) to pthread_key_create(3); and
pthread_setspecific(3) to pthread_getspecific(3)

Rationalize by moving

pthread_attr_setname_np(3) to pthread_attr_getname_np(3); and
pthread_setname_np(3) to pthread_getname_np(3).

Group also pthread_attr_getschedparam(3) and pthread_attr_getschedpolicy(3),
as these two pages fit well together.

Merge/group man pages:
- pthread_barrier_{init,destroy,wait} into pthread_barrier(3).
- pthread_cond_* into pthread_cond(3).
- pthread_mutex_* into pthread_mutex(3).
- pthread_rwlock_* into pthread_rwlock(3).
- pthread_spin_* into pthread_spin(3).
Cross-link amongst groups.
OK jruoho@

Finally, split pthread_attr_getschedpolicy() and pthread_attr_setschedpolicy()
to their own page.

Split pthread_attr_getschedparam() and pthread_attr_setschedparam() to a
separate page.

Split pthread_attr_getinheritsched() and pthread_attr_setinheritsched() to a
separate page, and actually document these.

Move pthread_attr_getguardsize() and pthread_attr_setguardsize() to a
separate page, and actually document these.

Move the thread stack attribute functions to a separate page. In addition,
document pthread_attr_getstack() and pthread_attr_setstack(). Try to also
document these a little better (in particular, note some items in the long
list of caveats and questions related to application-controlled thread stack).

Split pthread_attr_getdetachstate(3) and pthread_attr_setdetachstate(3) to a
separate page. Once ready, the main page, pthread_attr(3), will contain a
summary of the various thread attribute functions.

Move pthread_attr_{g,s}etscope() to a separate page and try to document
these better. Should fix PR lib/41831 from Wolfgang Stukenbrock.

XXX: Please proofread.

Document pthread_attr_get_np(). Fixes PR lib/39289 from ahoka@.

Build links for rest of the pthread_attr(3) functions.
Hopefully I got all of them right in the setlist...

Remove unused code that's confusing when using cscope/opengrok.

branches: 1.55.4;
check for ${PTHREAD_MACHINE_ARCH} before ${MACHINE_ARCH} or ${MACHINE_CPU}

Adjust the compat stuff slightly so that the changes are mostly self
contained (Makefile, pthread_compat.c).

make PTHREAD__COMPAT=1

Builds a libpthread that can be dropped into a NetBSD 2/3/4 chroot.

This makes threading work in the chroot when using a NetBSD 5 kernel, no
other modifications required.

- Add affinity(3) manual page, which describes thread affinity,
pthread_setaffinity_np(3) and pthread_getaffinity_np(3) functions,
provides simple code example.
- Add cpuset(3) manual page, which describes API of CPU-sets.

Thanks <wiz> for many improvements!

branches: 1.51.2; 1.51.4;
Add sched(3) manual page, and appropriate links.
Thanks <wiz> for review and fixes!

branches: 1.50.2;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

Implementation of processor-sets, affinity and POSIX real-time extensions.
Add schedctl(8) - a program to control scheduling of processes and threads.

Notes:
- This is supported only by SCHED_M2;
- Migration of LWP mechanism will be revisited;

Proposed on: <tech-kern>. Reviewed by: <ad>.

document following functions.
pthread_attr_getname_np
pthread_attr_setname_np
pthread_getname_np
pthread_setname_np

Remove the debuglog stuff. ktrace is more useful now.

Mutexes:

- Play scrooge again and chop more cycles off acquire/release.
- Spin while the lock holder is running on another CPU (adaptive mutexes).
- Do non-atomic release.

Threadreg:

- Add the necessary hooks to use a thread register.
- Add the code for i386, using %gs.
- Leave i386 code disabled until xen and COMPAT_NETBSD32 have the changes.

SSP doesn't like alloca...

Compile pthread_getspecific / pthread_setspecific with -fomit-frame-pointer
-falign-functions=32, since these two really get hammered on. To make them
faster needs a threadreg or TLS, unless there is a way to tell gcc that a
library-local (pthread__threadmask) variable does not need to be PIC.

- Don't take the mutex's spinlock (ptr_interlock) in pthread_cond_wait().
Instead, make the deferred wakeup list a per-thread array and pass down
the lwpid_t's that way.

- In pthread_cond_wait(), take the mutex before dealing with early wakeup.
In this way there should never be contention on the CV's spinlock if
the app follows POSIX rules (there should only be contention on the
user-provided mutex).

- Add a port of the kernel's rwlocks. The rwlock's spinlock is only taken if
there is contention. This is enabled where atomic ops are available. Right
now that is only i386 and amd64 because I don't have other hardware to
test with. It's trivial to add stubs for other architectures as long as
they have compare-and-swap. When we have proper atomic ops the old rwlock
code can be removed.

- Add a new mutex implementation that's similar to the kernel's mutexes, but
uses compare-and-swap to maintain the waiters list, so no spinlocks are
involved. Same caveats apply as for the rwlocks.

branches: 1.42.2;
Trim fat off libpthread internal spinlock operations. Makes a mesurable
improvement across the board.

Remove PT_FIXEDSTACKSIZE_LG.

branches: 1.40.2;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Sync with kernel changes introduced by merging the newlock2 branch.

set warns to 4

Fix bugs with and improve upon previous.

Conditionalised support for 1:1 threads. Needs associated kernel changes
and more work to be useful.

include libc after our own files to fix ports that have 2 assym.h's. One
in pthread and one in libc.

XXX: Work around libpthread's "intimate" connection with libc.

Create links for pthread_attr_setschedparam.3
and pthread_attr_getschedparam.3 to pthread_attr.3.

From Igor Sobrado in PR/29997.

Don't create links to pthread_barierattr_init.3 and
pthread_barierattr_destroy.3 but to pthread_barrierattr_init.3
and pthread_barrierattr_destroy.3 (note the 2 rs in barrier).

Use TOOL_GENASSYM. Part of PR toolchain/30350

branches: 1.30.2;
Add charles' mlock fixes, protected with PTHREAD_MLOCK_KLUDGE and enabled
by default, until we come up with a real fix.

document sched_yield.

Use CPP macros to cause many libpthread functions used by applications
to be transformed into the do-nothing-when-libpthread-isn't-linked libc
stub names. This will permit library code that uses <pthread.h> and pthread
functions "defensively" to not need to link against libpthread and not need
to be patched to the threadlib.h API.

pthread_spin_lock.3 also describes pthread_spin_trylock(),
so link them together

add pool code for res state.
XXX: could be more efficient.

branches: 1.25.4;
Add dependencies to assym.h on:
pthread.h pthread_int.h pthread_md.h
Should help prevent the problem I raised in [kern/23946], based on
a hint from Christian Limpach.

Add MAKEVERBOSE support.

Set default stack size to the current limit on the stack size as set
with the shell's command to change limits. Make the PTHREAD_STACKSIZE
environment variable override the default stack size. The old fixed
stack size behaviour can be enable with PT_FIXEDSTACKSIZE_LG when building
libpthread.

Document pthread_attr_setcreatesuspend_np, pthread_suspend_np, pthread_resume_np

Use ${HOST_SH} instead of `sh'.

If necessary, pull in <bsd.sys.mk> to get the definition of HOST_SH;
Makefiles that pull in one of (most of) <bsd.*.mk> will get this anyway.

Split out pthread_{set,get}specific() into a separate file and arrange
for that file to not be built with profiling. This makes it reasonable to
use pthread_{set,get}specific() to implement thread-safe profiline call counts.

Rework how dependency generation is performed:

* DPSRCS contains extra dependencies, but is _NOT_ added to CLEANFILES.
This is a change of behaviour. If a Makefile wants the clean semantics
it must specifically append to CLEANFILES.
Resolves PR toolchain/5204.

* To recap: .d (depend) files are generated for all files in SRCS and DPSRCS
that have a suffix of: .c .m .s .S .C .cc .cpp .cxx

* If YHEADER is set, automatically add the .y->.h to DPSRCS & CLEANFILES

* Ensure that ${OBJS} ${POBJS} ${LOBJS} ${SOBJS} *.d depend upon ${DPSRCS}

* Deprecate the (short lived) DEPENDSRCS


Update the various Makefiles to these new semantics; generally either
adding to CLEANFILES (because DPSRCS doesn't do that anymore), or replacing
specific .o dependencies with DPSRCS entries.

Tested with "make -j 8 distribution" and "make distribution".

Add pthread(3), based on one by Hubert Feyrer, improved by yours truly,
reviewed by Nathan J. Williams.

pthread.c was getting a bit unwieldly. Move pthread_attr stuff out
into a new file, and put the shared private structure definition in
pthread_int.h.

Make infrastructure for man pages.

Re-enable nanosleep(); kernel bug is squished (kern_time.c:1.67).

Disable userlevel nanosleep() until I find the timer/alarm bug that it
is tickling.

Implement nanosleep() at userlevel. Threaded applications seem to like
to sleep a lot, and there's no need for each such thread to consume kernel
resources.

(accidentally checked the pthread.c part of this in yesterday; no reason
not to finish the job)

p1003.1b semaphore manual pages have been moved to librt.

add manual pages.

add support for p1003.1b semaphores. From FreeBSD

Always allow includes to be installed. This makes it easier for
people to get straggling ports bootstrapped (pthread.h is required
in order to build libc).

Merge the nathanw_sa branch.

branches: 1.8.2;
This incarnation of the pthreads library is ancient and not useful, and
should have been mothballed some time ago...

The latest pthread library.

Fix for older shells that can't handle then;

Added NOPIC flag.

correct my bad advice

OK one more try at getting it right ...

Makefile needed a preall to make a few links before compiling sources.
CVS doesn't check in symlinks.

branches: 1.1.1;
Initial revision

libpthread_dbg(3) deletion from the base distribution

libpthread_dbg(3) is a remnant library from the M:N thread model
(pre-NetBSD-5.0) API to introspect threads within a process and for use
of debuggers.

Currently in the 1:1 model it's not used in GDB neither in LLDB and it's
not either planned to be used. It's current function to read pthread_t
structures is realizable within a regular debugger capable to
instrospect objects within a tracee (GDB, LLDB...).

Remaining users of this API can still use this library from
pkgsrc/devel/libpthread_dbg.

Sponsored by <The NetBSD Foundation>

branches: 1.6.28; 1.6.32;
Add CAVEATS section:
Due to limitations in the current pthread implementation, makecontext(3)
and sigaltstack(2) should not be used in programs which link against
libpthread (whether threads are used or not).

Wording from drochner@ and ad@

Bump date for man pages.

Add note in libpthread/README so it's not forgotten to update the man
pages when the problem is fixed.

branches: 1.5.2; 1.5.18;
Note that libpthread_dbg needs to be checked after making changes to
libpthread.

This incarnation of the pthreads library is ancient and not useful, and
should have been mothballed some time ago...

The latest pthread library.

Makefile needed a preall to make a few links before compiling sources.
CVS doesn't check in symlinks.

branches: 1.1.1;
Initial revision

Export the guard size of the main thread via vm.guard_size. Add a
complementary writable sysctl for the initial guard size of threads
created via pthread_create. Let the existing attribut accessors do the
right thing. Raise the default guard size for threads to 64KB.

branches: 1.18.4;
libpthread_dbg(3) deletion from the base distribution

libpthread_dbg(3) is a remnant library from the M:N thread model
(pre-NetBSD-5.0) API to introspect threads within a process and for use
of debuggers.

Currently in the 1:1 model it's not used in GDB neither in LLDB and it's
not either planned to be used. It's current function to read pthread_t
structures is realizable within a regular debugger capable to
instrospect objects within a tracee (GDB, LLDB...).

Remaining users of this API can still use this library from
pkgsrc/devel/libpthread_dbg.

Sponsored by <The NetBSD Foundation>

branches: 1.17.22; 1.17.26;
ld.elf_so is reentrant.

branches: 1.16.6;
As per (now closed) PR lib/43581, add a note about unsupported modifications
to the red zone size.

Fix typo.

We've got the priority/class stuff now.

Update.

Mutexes:

- Play scrooge again and chop more cycles off acquire/release.
- Spin while the lock holder is running on another CPU (adaptive mutexes).
- Do non-atomic release.

Threadreg:

- Add the necessary hooks to use a thread register.
- Add the code for i386, using %gs.
- Leave i386 code disabled until xen and COMPAT_NETBSD32 have the changes.

Update.

branches: 1.10.2;
Update.

branches: 1.9.2;
Update.

branches: 1.8.2;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Don't grab the state lock to check for cancellation around condition
variables, as _lwp_wakeup/lwp_park provide the necessary barrier.

fix a race between pthread_exit and pthread_create.
inefficient, but better than crashing.

Add some items for 1:1 threads.

Update list of items.

pthread_kill() was implemented, remove from list
pthread_cond_timedwait() now works even before SA initialization, but some
other routines do not; update the paragraph for this

Merge the nathanw_sa branch.

branches: 1.1.2;
file TODO was initially added on branch nathanw_sa.

branches: 1.1.2;
file Version.map was initially added on branch agc-symver.

libpthread/affinity.3: Fix error reporting in example.

pthread_setaffinity_np does not set errno, so err(...) is wrong.
Instead, it returns the error number, which requires either
errc(..., error, ...) or errno = error; err(...).

slightly expand example

branches: 1.8.52;
Do not xref to itself.

Improvements in secmodel(9). Document secmodel_register(9), _deregister(9)
and _eval(9).

Add secmodel_extensions(9), and indicate the new sysctl(7) to let
ordinary users control the CPU affinity (user_set_cpu_affinity).

branches: 1.6.6;
Sort sections.

Add a compatibility note. Also some minor improvements.

- Make thread-affinity and processor-set interfaces mutually exlusive.
- pset_assign: when CPU is assigned, migrate out all LWPs from it.

branches: 1.3.2; 1.3.4;
Do not recommend to mix affinity(3) with pset(3).

Arguments were swapped for cpuset_set(), cpuset_clr(), cpuset_isset()
functions, but man pages were not updated (hi Christos!).

branches: 1.1.2; 1.1.4;
- Add affinity(3) manual page, which describes thread affinity,
pthread_setaffinity_np(3) and pthread_getaffinity_np(3) functions,
provides simple code example.
- Add cpuset(3) manual page, which describes API of CPU-sets.

Thanks <wiz> for many improvements!

branches: 1.2.2;
Fix some typos, improve wording.

Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

branches: 1.2.2;
Drop error path from C11 call_once

The original implementation of C11 threads(3) contained check for error
paths, but it was stripped in the calls that are documented to return
no status from an operation. Do the same in call_once(3).

Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

libpthread: Say `restrict', not `__restrict', in man pages.

The private macro __restrict exists only to make .h files work with
legacy C/C++ versions without support for the `restrict' keyword.
It's only distracting to use in a man page.

branches: 1.2.2; 1.2.14;
Fix some typos, improve wording.

Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

branches: 1.1.2;
Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

Remove the debuglog stuff. ktrace is more useful now.

branches: 1.7.6;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Set _IOLBF on stdout.

Include <stdlib.h> for exit() prototype.

More printf format size tweaks.

printf format fixes.

Merge the nathanw_sa branch.

branches: 1.1.2;
file debuglog.c was initially added on branch nathanw_sa.

Use TOOL_GENASSYM. Part of PR toolchain/30350

Clean up tmp files on exit. Fixes PR#23723

Merge the nathanw_sa branch.

branches: 1.1.2;
file genassym.sh was initially added on branch nathanw_sa.

libpthread: Say `restrict', not `__restrict', in man pages.

The private macro __restrict exists only to make .h files work with
legacy C/C++ versions without support for the `restrict' keyword.
It's only distracting to use in a man page.

branches: 1.2.2; 1.2.14;
Fix some typos, improve wording.

Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

branches: 1.1.2;
Add a complete C11 threads(3) implementation

C11 Thread support library is a portable threading C API between OSs,
similar to std::threads in the C++ world.

The library is implemented as a thin shim over POSIX interfaces.

NetBSD implements the API as a part of the POSIX threading library
(libpthread(3)).

C11 threads(3) are in the process of making them an integral part
of the POSIX standard. The interface has been implemented in major
OSs and used with stopgap libraries for older versions of them.

C11 threading library is already used (with a stopgap implementation)
in the NetBSD distribution in MESA.

Original implementation by myself from 2016.

ATF tests are new and cover almost all interfaces.

Proposed on tech-userlevel@.

Remove superfluous Tn.

libpthread_dbg(3) deletion from the base distribution

libpthread_dbg(3) is a remnant library from the M:N thread model
(pre-NetBSD-5.0) API to introspect threads within a process and for use
of debuggers.

Currently in the 1:1 model it's not used in GDB neither in LLDB and it's
not either planned to be used. It's current function to read pthread_t
structures is realizable within a regular debugger capable to
instrospect objects within a tracee (GDB, LLDB...).

Remaining users of this API can still use this library from
pkgsrc/devel/libpthread_dbg.

Sponsored by <The NetBSD Foundation>

branches: 1.16.2;
Bump date for previous.

Add reference in SEE ALSO to pthread_dbg(3)

Sponsored by <The NetBSD Foundation>

branches: 1.14.28;
Add the Butenhof's book to SEE ALSO. (It was decent enough when I read it
years ago, but if there are better ones, please feel free to add those.)

Enumerate the functions in a table instead of a big SEE ALSO.
Also some stylistic changes.

pthread(3): mention that we use 1:1 threading model; remove descriptions
of no longer used environment variables.

Add CAVEATS section:
Due to limitations in the current pthread implementation, makecontext(3)
and sigaltstack(2) should not be used in programs which link against
libpthread (whether threads are used or not).

Wording from drochner@ and ad@

Bump date for man pages.

Add note in libpthread/README so it's not forgotten to update the man
pages when the problem is fixed.

Remove trailing whitespace.

branches: 1.9.2; 1.9.4; 1.9.6;
Clarifications for PTHREAD_DIAGASSERT.

Assign to TNF. Make 2-clause instead of 3.

Ok: co-copyright owner hubertf@.

branches: 1.7.6; 1.7.8;
Sync with reality, and note that programs must link against the dynamic
libpthread in order to remain compatible with future releases of NetBSD.

branches: 1.6.6;
Remove scheduler activations references. Bump date.

- Sync with reality.
- Minor cosmetic changes.

Document PTHREAD_CONCURRENCY, ack by wiz@.

branches: 1.3.2; 1.3.4;
Set default stack size to the current limit on the stack size as set
with the shell's command to change limits. Make the PTHREAD_STACKSIZE
environment variable override the default stack size. The old fixed
stack size behaviour can be enable with PT_FIXEDSTACKSIZE_LG when building
libpthread.

netbsd.org->NetBSD.org

Add pthread(3), based on one by Hubert Feyrer, improved by yours truly,
reviewed by Nathan J. Williams.

revert previous; needs more discussion/testing.

Make pthread_setname_np variadic for compatibility with FreeBSD and GLIBC.
Bump.

libc: Expose some more pthread stubs.

Additionally, provide the option for pthread.h to expose
pthread_create so that libc can provide a weak stub for it, if you
define _NETBSD_PTHREAD_CREATE_WEAK.

This allows libraries to use the functions without linking against
libpthread for applications that don't need threads. For example,
with OpenSSL's libcrypto, as long as you don't try to raise
OSSL_set_max_threads above 0, it won't ever try pthread_create.

New pthread.h macros defined:

pthread_attr_destroy __libc_thr_attr_destroy
pthread_attr_init __libc_thr_attr_init
pthread_attr_setdetachstate __libc_thr_attr_setdetachstate
pthread_detach __libc_thr_detach
pthread_join __libc_thr_join

New pthread.h macros defined if _NETBSD_PTHREAD_CREATE_WEAK is
defined first:

pthread_create __libc_thr_create

New symbols defined by libc as weak aliases for stubs that always
fail but can be overridden by libpthread:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that libc already defined __libc_thr_create.

New internal stubs, not for overriding:

__libc_thr_detach_stub
__libc_thr_join_stub

These were formerly called __libc_pthread_detach and
__libc_pthread_join, respectively, and libc has defined weak aliases
pthread_detach and pthread_join for them since 2013. I retained all
four of those symbols, but defined new ones in the __ namespace that
match the pattern for other thread stubs to avoid raising questions
about why the pattern seems to be broken. I left a comment noting
this fact too.

New pthread strong aliases to override the libc symbols:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that pthread already defined a strong __libc_thr_create alias
for the real pthread_create, so there is no need to add a new one.

PR lib/59685: libcrypto should not depend on libpthread

branches: 1.187.2;
pthread_cancel(3): Use _Atomic consistently with C11 atomics.

Should fix clang build:

/home/source/ab/HEAD-llvm/src/lib/libpthread/pthread.c:1264:8: error: address argument to atomic operation must be a pointer to _Atomic type ('volatile unsigned int *' invalid)
(atomic_load_relaxed(&self->pt_cancel) &
^ ~~~~~~~~~~~~~~~~
/home/source/ab/HEAD-llvm/src/lib/libpthread/pthread.c:74:2: note: expanded from macro 'atomic_load_relaxed'
atomic_load_explicit(p, memory_order_relaxed)
^ ~
/home/builds/ab/HEAD-llvm/amd64/202504020240Z-dest/usr/include/clang-13.0/stdatomic.h:121:30: note: expanded from macro 'atomic_load_explicit'
#define atomic_load_explicit __c11_atomic_load

PR lib/59134: POSIX-1.2024: pthread_setcancelstate must be
async-signal-safe

pthread_cancel(3): Rework.

Make pthread_setcancelstate(3) async-signal-safe. (As a side effect,
this also makes pthread_setcanceltype(3) async-signal-safe, although
that is not required.)

PR lib/59134: POSIX-1.2024: pthread_setcancelstate must be
async-signal-safe

branches: 1.185.2;
Fix resource leak in pthread_create(). Pthread field "pt_lwpctl"
must not be accessed after _lwp_exit() as the kernel may free
it, attach it to another thread and change its "lc_curcpu".

The "pthread__deadqueue" will fill up with pthreads with
an invalid "pt_lwpctl" and allocated stack that never get
reused.

Replace the test "lc_curcpu == LWPCTL_CPU_EXITED" with
test "_lwp_kill(newthread->pt_lid, 0) == -1 && errno == ESRCH"
to make sure this thread has finished its _lwp_exit().

PR lib/57831 "Memory leaks in libpthread/libc"

pthread: Don't adjust user-allocated stack addresses by guardsize.

PR lib/57721

XXX pullup-10
XXX pullup-9
XXX pullup-8

raise() has understood threads for a long time, don't reimplment it.

Use snprintf_ss in pthread__assertfunc and update comment in
pthread__errorfunc. snprintf can use locks in some code paths and we
only care about the restricted subset here.

branches: 1.181.2;
libpthread(3): Fix a marvellous interaction with rtld.

Patch from chs@. Comment explaining the story by me. This patch may
not be optimal -- maybe it would be better in pthread__init, or
better for rtld to call _lwp_unpark after _lwp_park in the contened
case -- but we've tested this version and it's annoying to reproduce,
so let's take this version and worry about testing improvements
later.

libpthread: Move namespacing include to top of .c files.

Stuff like libc's namespace.h, or atomic_op_namespace.h, which does
namespacing tricks like `#define atomic_cas_uint _atomic_cas_uint',
has to go at the top of each .c file. If it goes in the middle, it
might be too late to affect the declarations, and result in compile
errors.

I tripped over this by including <sys/atomic.h> in mips
<machine/lock.h>.

(Maybe we should create a new pthread_namespace.h file for the
purpose, but this'll do for now.)

fake-use alloca()'s return value to quieten -Werror=unused-result

s/reseting/resetting/

Don't need to ignore ESRCH from _lwp_park() any more.

Drop self->pt_lock before clearing TSD / malloc TSD.

- Make pthread_condvar and pthread_mutex work on the stack rather than in
pthread_t, so there's less chance of bad things happening if someone calls
(for example) pthread_cond_broadcast() from a signal handler.

- Remove all the deferred waiter handling except for the one case that really
matters which is transferring waiters from condvar -> mutex on wakeup, and
do that by splicing the condvar's waiters onto the mutex.

- Remove the mutex waiters bit as it's another complication that's not
strictly needed.

If _malloc_thread_cleanup is implement, call it from libpthread.
Provide the hook from modern jemalloc to avoid using TSD for the thread
destruction cleanup as it can result in reentrancy crashes if fork is
called from a thread that never called malloc as it will result in a
late malloc from the pre-fork synchronisation handler.

Deal with a couple of problems with threads being awoken early due to
timeouts or cancellation where:

- The restarting thread calls _lwp_exit() before another thread gets around
to waking it with _lwp_unpark(), leading to ESRCH (observed by joerg@).
(I may have removed a similar check mistakenly over the weekend.)

- The restarting thread considers itself gone off the sleep queue but
at the same time another thread is part way through waking it, and hasn't
fully completed that operation yet by setting thread->pt_mutexwait = 0.
I think that could have potentially lead to the list of waiters getting
messed up given the right circumstances.

Pass down errno when calling pthread__errorfunc after a system call.
Allow format arguments for that reason and use (v)snprintf_ss in
pthread_errorfunc to avoid race conditions and the like.

In the interests of reliability simplify waiter handling more and redo
condvars to manage the list of waiters with atomic ops.

- Try to eliminate a hang in "parked" I've been seeing while stress testing.
Centralise wakeup of deferred waiters in pthread__clear_waiters() and use
throughout libpthread. Make fewer assumptions. Be more conservative in
pthread_mutex when dealing with pending waiters.

- Remove the "hint" argument everywhere since the kernel doesn't use it any
more.

Lock/unlock/reinit pthread__deadqueue_lock over fork.

Drop most of the logic associated with pthread__started.

The pthread_cond logic is a questionable optimisation at best and the
post-fork logic is plainly broken.

Revert "Enhance the pthread(3) + malloc(3) init model"

It is reported to hand on aarch64 with gzip.

Set __isthreaded before bootstrapping malloc(3)

jemalloc depends on the __isthreaded dynamic state logic.

Reported by <wiz> for mpv and by <tih> for gzip.

Enhance the pthread(3) + malloc(3) init model

Separate the pthread_atfork(3) call from pthread_tsd_init()
and move it into a distinct function.

Call inside pthread__init() late TSD initialization route, just after
"pthread_atfork(NULL, NULL, pthread__fork_callback);".

Document that malloc(3) initialization is now controlled again and called
during the first pthread_atfork(3) call.

Remove #if 0 code from pthread_mutex.c as we no longer initialize malloc
prematurely.

Change the behavior of pthread_equal()

On error when not aborting, do not return EINVAL as it has a side effect
of being interpreted as matching threads. For invalid threads return
unmatched.

Check pthreads for NULL, before accessing pt_magic field. This avoids
faults on comparision with a NULL pointer.

This behavior is in the scope of UB, but should be easier to deal with
buggy software.

Remove trailing whiteapaces and tab

- pthread_join(): remove temporary hack now kernel returns correct errno.

- kill(getpid(), SIGABRT) -> _lwp_kill(_lwp_self(), SIGABRT)

Chack thread->pt_magic with PT_MAGIC promptly

Rearrange some checks to avoid verifying pthread_t after using it.

Revert previous

Two assignments are correct.

Do not set stackbase2 twice for !__MACHINE_STACK_GROWS_UP

pthread_join(): add a temporary hack to make lib/libpthread/t_detach pass.
The correct fix is to do this in kernel (I have that change, but it's part
of the wider change to index LWPs in a tree).

pthread_detach(), pthread_join(): go back to using _lwp_detach() and
_lwp_wait(), rather than doing it all in userspace. There's less to go
wrong. Doesn't seem to be a performance penalty.

pthread_exit(): it looks there there is at least one path through which
a thread can exit with waiters still hanging off it (cancellation when
waiting on a condvar) so deal with all/any crappy failure like that and
make sure there are never any waiters left before exiting. Maybe of help
for:

PR: bin/50350: rump/rumpkern/t_sp/stress_{long,short} fail on Core 2

- Fix a race between the kernel and libpthread, where a new thread can start
life without its self->pt_lid being filled in.

- Fix an error path in _lwp_create(). If the new LID can't be copied out,
then get rid of the new LWP (i.e. either succeed or fail, not both).

- Mark l_dopreempt and l_nopreempt volatile in struct lwp.

Rip out some very ambitious optimisations around pthread_mutex that are
don't buy much. This stuff is hard enough to get right in the kernel let
alone userspace, and I don't trust that it's right.

branches: 1.153.2;
Transfer all the keys that were created in the libc stub implementation
to the pthread tsd implementation when the main thread is created.
This corrects a problem where a process created keys before libpthread
was loaded (either from the libc constructor or because libpthread
was dlopened later). This fixes a problem with jemalloc which creates
keys in the constructor.

Drop a duplicate instruction line

No functional change intended.

branches: 1.151.2; 1.151.4;

Deal with more lwp_park() timestamp unconsting

Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

Do not look at environmental variables for suid/guid binaries.

Export the guard size of the main thread via vm.guard_size. Add a
complementary writable sysctl for the initial guard size of threads
created via pthread_create. Let the existing attribut accessors do the
right thing. Raise the default guard size for threads to 64KB.

branches: 1.147.8;
Fix previous: Can't use calloc/malloc before we complete initialization
of the thread library, because malloc uses pthread_foo_specific, and it will
end up initializing itself incorrectly.

Make PTHREAD_KEYS_MAX dynamically adjustable

NetBSD's PTHREAD_KEYS_MAX is set to 256, which is low compared to
other systems like Linux (1024) or MacOS X (512). As a result some
setups tested on Linux will exhibit problems on NetBSD because of
pthread_keys usage beyond the limit. This happens for instance on
Apache with various module loaded, and in this case no particular
developper can be blamed for going beyond the limit, since several
modules from different sources contribute to the problem.

This patch makes the limit conigurable through the PTHREAD_KEYS_MAX
environement variable. If undefined, the default remains unchanged
(256). In any case, the value cannot be lowered below POSIX-mandated
_POSIX_THREAD_KEYS_MAX (128).

While there:
- use EXIT_FAILURE instead of 1 when calling err(3) in libpthread.
- Reset _POSIX_THREAD_KEYS_MAX to POSIX mandated 128, instead of 256.

Allow for arbitrary MI scheduler implementations.

A concrete result is enabling unpatched libpthread to run on the
rumprun stacks (e.g. Xen and bare metal) with a non-NetBSD scheduler.
Those schedulers hook into the existing _lwp_frobnitz() NetBSD syscall
interfaces (well, "syscall" interfaces in that scenario ;)

More specifically about the change itself:

1) instead of calling _lwp_makecontext() followed by _lwp_create()
and passing the entry point in ucontext_t (MD) through the calls, roll
the calls into pthread__makelwp() and allow alternate implementations
for that MI interface.

2) allow compile-time overriding of __lwp_gettcb_fast() or
__lwp_getprivate_fast, which are inline and leak MD scheduler/thread
details into libpthread


Additionally, two small nits:

I) define LIB=pthread before including mk.conf so that it's possible
to test for LIB==pthread in mk.conf

II) make it possible to leave out pthread_cancelstub.c. This is required
by the current implementation of rumprun-posix (i.e. rumprun on
POSIX hosts) due to symbol collisions. It needs to be fixed properly
some day, but for now allows an almost-correct libpthread to run.
I am sure @justin will be happy to explain the details ;)


no change to NetBSD
tested: anita+atf

branches: 1.144.4;
remove compatibility code for handling CLOCK_MONOTONIC and handle it in the
syscall directly.

- Allow libpthread to be dlopened again, by providing libc stubs to libpthread.
- Fail if the dlopened libpthread does pthread_create(). From manu@
- Discussed at length in the mailing lists; approved by core@
- This was chosen as the least intrusive patch that will provide
the necessary functionality.
XXX: pullup to 6

Don't rely on sys/lwp.h includeing sys/resource.h for us.

add argsused

Remove duplicate "const" modifier found by clang -Wduplicate-decl-specifier

libpthread: replace the use of obsolete sys/tree.h interface with rbtree(9).

cast return of sysconf

branches: 1.137.2;
return errno if pthread_create hits the system limit, not just -1
(this is not entirely correct because it can return ENOMEM which is
not mentioned in the spec, but there are other places in pthread_create
whete ENOMEM is returned -- it at all, this should be fixed everywhere)

Prefer sysconf to obtain number of CPUs.

Simplify check for TLS definition to not hide code. Drop it in another
place as it is redundant.

Store allocated lwpctl state in the thread actually forked,
which is the only thread lives in the child process.

The problem originally reported here:
https://bugs.ruby-lang.org/issues/6341

don't reuse a dynamically allocated stack if a fixed one is requested

Keep track of the size of the guard area, in case we want to make it
modifiable later. Only reuse the stack if it was allocated by libpthread
and if the expected thread size matches the current stack size.

Further refine stack allocation. If the stack was provided by the user,
don't bother with setting up a guard page. Otherwise, round up the size
to page size. Point stack inside the guarded area, without the guard
page. Fix size when mprotect failed.

Use correct size.

-fix initial stacksize rounding
-minor indentation fix

Actually use the stack thread attributes when creating a new thread.

Fix the stack base pointer for the initial thread on !HPPA.
AT_STACKBASE is pointing to the start of the stack, which is the
upper limit on platforms where the stack grows down.

Separate pthread_t from thread stack. Drop additional alignment
restrictions on the thread stack. Remove remaining parts of stackid.

branches: 1.125.2; 1.125.4;
put back pthread__dbg variable; this is set to no zero by td_open() when
debugging to avoid multiple td_open() mess.

Use __dead

Rework TLS initialisation:
- Update TCB for the initial thread in pthread__initthread, not
pthread__init to get it valid as soon as possible.
- Don't overwrite the pt_tls field in pthread__initthread.
- Don't deallocate pt_tls in pthread__scrubthread. This worked more by
chance than by design.
- Handle freeing the TLS area in pthread_create after removing the
thread instance from the dead queue.

Use __lwp_gettcb_fast if present.

Add TLS support infrastructure. For dynamic binaries, ld.elf_so exports
_rtld_tls_allocate and _rtld_tls_free. libpthread uses this functions to
setup the thread private area of all new threads. ld.elf_so is
responsible for setting up the private area for the initial thread.
Similar functions are called from _libc_init for static binaries, using
dl_iterate_phdr to access the ELF Program Header.

Add test cases to exercise the different TLS storage models. Test cases
are compiled and installed on all platforms, but are skipped on
platforms not marked for TLS support.

This material is based upon work partially supported by
The NetBSD Foundation under a contract with Joerg Sonnenberger.

It is inspired by the TLS support in FreeBSD by Doug Rabson and the
clean ups of the DragonFly port of the original FreeBSD modifications.

use pthread__stacksize since size has not been initialized yet.

only mprotect base if we moved it.

I've had this patch in my tree for a while and since it only improves
the situation, I decided to commit it. There is an inherent problem
with ASLR and the way the pthread library is using the thread stack.

Our pthread library chooses that stack for each thread strategically
so that it can locate the location of the pthread struct for each
thread by masking the stack pointer and looking just below the red
zone it creates. Unfortunately with ASLR you get many random values
for the initial stack, and there are situations where the masked
stack base ends up below the base of the stack. (this happens on
x86 when the stack base happens to be 0x???02000 for example and
your mask is stackmask is 0xffe00000). To fix this, we detect the
pathological cases (this happens only in the main thread), allocate
more stack, and mprotect it appropriately. Then we stash the main
base and the main struct, so that when we look for the pthread
struct in pthread__id, we can special case the main thread.

Another way to work around the problem is unlimiting stacksize,
but the proper way is to use TLS to find the thread structure and
not to play games with the thread stacks.

Don't use internal libc function __findenv().

pthread_create: simplify error path slightly.

fix the pthread pt_lid in the fork callback function that runs in the child instead of a function that may be going away. KNFify

Correctly set pt_lid in the child, after a fork

Don't just look only at the first element in the deadqueue to find lwp's
to reuse, because if we lose the race with the kernel we are never going
to reuse any elements. Look in the whole list instead.
XXX: should be pulled up to 5.x

Restore use of _lwp_makecontext, the AMD64 bug has been fixed.

Partially revert 1.110: Use makecontext again until the problems with
_lwp_makecontext are solved.

- Convert from makecontext() -> _lwp_makecontext().

- Rely on _lwp_makecontext() to set up the thread identity register.
This is not currently done (a bug), nor does libpthread use the
threadreg yet. I'm doing this so it the code can be used by the
person working on TLS to verify that their threadreg code is working.

Fix the comparision function used by the red-black tree global thread list
implementation:
-don't return a difference, this can overflow
-don't try to substract typed pointers which don't belong to the
same object, this gives undefined results

This fixes instabilities of programs which use more than a handful
of threads, eg spuriously failing pthread_join().

Revert the _lwp_ctl which is causing problem. I did test this locally,
I guess not well enough.

- Make the threadreg code use _lwp_setprivate() instead of MD hooks.

XXX This must not be enabled by default because the LWP private mechanism
is reserved for TLS. It is provided only as a test/demo.

XXX Since ucontext_t does not contain the thread private variable, for a
short time after threads are created their thread specific data is unset.
If a signal arrives during that time we are screwed.

- No longer need pthread__osrev.

- Rearrange _lwp_ctl() calls slightly.

branches: 1.106.2; 1.106.4; 1.106.6;
Disable diagnostic assertions by default and just return error codes like
other systems. Allows poorly written applications to appear working. If you
are developing pthread apps please turn it on manually by setting the
environment variable.

Change some type to eliminate some lint warnings.

Minor correction to previous.

Now that we have all the scheduling gunk, make these do something useful:

pthread_attr_get_np
pthread_attr_setschedparam
pthread_attr_getschedparam
pthread_attr_setschedpolicy
pthread_attr_getschedpolicy

pthread_join: explicitly test for cancellation.

PR lib/38741 priority inversion in libpthread breaks apps that use
SCHED_FIFO threads

- Change condvar sync so that we never take the condvar's spinlock without
first holding the caller-provided mutex. Previously, the spinlock was only
taken without the mutex in an error path, but it was enough to trigger the
problem described in the PR.

- Even with this change, applications calling pthread_cond_signal/broadcast
without holding the interlocking mutex are still subject to the problem
described in the PR. POSIX discourages this saying that it leads to
undefined scheduling behaviour, which seems good enough for the time being.

- Elsewhere, use a hash of mutexes instead of per-object spinlocks to
synchronize entry/exit from sleep queues.

- Simplify how sleep queues are maintained.

branches: 1.100.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.99.2;
Back out previous. It seems to expose another bug in libpthread/libc,
potentially errno being used before threading is up and running.

Move pthread__errno() into pthread_specific.c so it gets the "no stack
frame" treatment.

Complain if _lwp_ctl() fails.

branches: 1.96.2;
add missing static decls.

- Use pthread__cancelled() in more places.
- pthread_join(): assert that pthread_cond_wait() returns zero.

- Fix pthread_rwlock_trywrlock() which was broken.

- Add new functions: pthread_mutex_held_np, mutex_owner_np, rwlock_held_np,
rwlock_wrheld_np, rwlock_rdheld_np. These match the kernel's locking
primitives and can be used when porting kernel code to userspace.

- Always create LWPs detached. Do join/exit sync mostly in userland. When
looped on a dual core box this seems ~30% quicker than using lwp_wait().
Reduce number of lock acquire/release ops during thread exit.

Remove test of pthread__osrev that is no longer needed.

remove unnecessary assignments.

- On 64-bit platforms 1/2 the default tsd values were garbage. Fix it.
- The lwpctl block is now needed on uniprocessors, for pthread_curcpu_np().

Remove the debuglog stuff. ktrace is more useful now.

Add pthread_equal() to libc stubs; this makes a lot of sense for
threadsafe libraries implementing own locking functions.
Ride on yesterday's minor version bumps.

Mutexes:

- Play scrooge again and chop more cycles off acquire/release.
- Spin while the lock holder is running on another CPU (adaptive mutexes).
- Do non-atomic release.

Threadreg:

- Add the necessary hooks to use a thread register.
- Add the code for i386, using %gs.
- Leave i386 code disabled until xen and COMPAT_NETBSD32 have the changes.

For PR bin/37347:

- Override __libc_thr_init() instead of using our own constructor.
- Add pthread__getenv() and use instead of getenv(). This is used before
we are up and running and unfortunatley getenv() takes locks.

Other changes:

- Cache the spinlock vectors in pthread__st. Internal spinlock operations
now take 1 function call instead of 3 (i386).
- Use pthread__self() internally, not pthread_self().
- Use __attribute__ ((visibility("hidden"))) in some places.
- Kill PTHREAD_MAIN_DEBUG.

Use _lwp_setname() to pass thread names to the kernel.

... but preserve the linked list, for the debugger only.

Replace the global thread list with a red-black tree. From joerg@.

Put new threads on the tail of pthread__allqueue, for the debugger.

- Get rid of self->pt_mutexhint and use pthread__mutex_owned() instead.
- Update some comments and fix minor bugs. Minor cosmetic changes.
- Replace some spinlocks with mutexes and rwlocks.
- Change the process private semaphores to use mutexes and condition
variables instead of doing the synchronization directly. Spinlocks
are no longer used by the semaphore code.

- Don't take the mutex's spinlock (ptr_interlock) in pthread_cond_wait().
Instead, make the deferred wakeup list a per-thread array and pass down
the lwpid_t's that way.

- In pthread_cond_wait(), take the mutex before dealing with early wakeup.
In this way there should never be contention on the CV's spinlock if
the app follows POSIX rules (there should only be contention on the
user-provided mutex).

- Add a port of the kernel's rwlocks. The rwlock's spinlock is only taken if
there is contention. This is enabled where atomic ops are available. Right
now that is only i386 and amd64 because I don't have other hardware to
test with. It's trivial to add stubs for other architectures as long as
they have compare-and-swap. When we have proper atomic ops the old rwlock
code can be removed.

- Add a new mutex implementation that's similar to the kernel's mutexes, but
uses compare-and-swap to maintain the waiters list, so no spinlocks are
involved. Same caveats apply as for the rwlocks.

branches: 1.80.2;
Acquire the correct lock in pthread_detach(). Spotted by Jan Kryl.

pthread_suspend_np, pthread_resume_np, pthread_detach: return correct code
on error.

Trim fat off libpthread internal spinlock operations. Makes a mesurable
improvement across the board.

- Reinitialize the absolute minimum when recycling user thread state.
Chops another ~10% off create/join in a loop on i386.
- Disable low level debugging as this is stable. Improves benchmarks
across the board by a small percentage. Uncontested mutex acquire
and release in a loop becomes about 8% quicker.
- Minor cleanup.

Remove PT_FIXEDSTACKSIZE_LG.

Cache thread context for creation instead of setting it up every time.
Speeds create/join loop by about 10-15% on i386.

Simplify pthread_join(). Knocks about 5% off pthread_create/join in a loop.

Missed variable rename..

Change the signature of _lwp_park() to accept an lwpid_t and second
hint pointer, but do so in a way that remains compatible with older
pthread libraries. This can be used to wake another thread before the
calling thread goes asleep, saving at least one syscall + involuntary
context switch. This turns out to be a fairly large win on the condvar
benchmarks that I have tried.

branches: 1.71.2;
Make libpthread_dbg build again.

Fix error in previous.

Some significant performance improvements, and a fix for a race with pthread
detach/join.

- Make mutex acquire spin for a short time, as done with spinlocks.
- Make the number of spins controllable with the env var PTHREAD_NSPINS.
- Reduce the amount of time that libpthread internal spinlocks are held.
- Rely more on the barrier effects of park/unpark to avoid taking spinlocks.
- Simplify the locking around pthreads and the global queues.
- Align per-thread sync data on a 128 byte boundary.
- Offset thread stacks by a small amount to try and reduce cache thrash.

branches: 1.68.2;
- Test+branch is usually cheaper than making an indirect function call,
so avoid making them.
- When parking an LWP on a condition variable, point the hint argument at
the mutex's waiters queue. Chances are we will be awoken from that later.

- pthread__park: always check for cancellation.
- pthread__unpark: cosmetic changes.

- Simplify the interface to pthread__park() and friends slightly.
- If sysctl() fails, complain.

Silence a gcc/lint warning.

Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

pthread_detach: acquire the join lock before changing pt_flags.

- Put detached & exiting threads on the tail of the deadqueue, not head.
- Don't forget to set PT_FLAG_DETACHED.

Minor optimisation to previous: only check for continued existence of
the thread if it was detached.

fix a race between pthread_exit and pthread_create.
inefficient, but better than crashing.

Cast the return value of _lwp_unpark_all(), just in case.

use __func__ instead of __FUNCTION__ which is a gcc extension.

pthread_create(): clear newthread->pt_sleeponq before calling _lwp_create().

add a missing volatile.

Fix a race between pthread_exit() and pthread_join().

Fix a race between pthread_create() and pthread_exit() in the 1:1 case.

Pass in the 'hint' argument when parking/unparking LWPs.

Fix bugs with and improve upon previous.

Add another comment.

Add a comment.

Conditionalised support for 1:1 threads. Needs associated kernel changes
and more work to be useful.

branches: 1.48.4; 1.48.6;
There is no point in using the internal sigprocmask() to initialize
the process' signal mask -- this ends up in a no-op.
Use the system call directly instead.
(This might be done in pthread_sig.c, but for now I wanted a simple
patch which is easily tested and pulled up.)

- do PTHREAD_MLOCK_KLUDGE in pthread__stackid_setup, rather than callers,
so that the main thread is not different from others.
as a side effect, fix memory leak in pthread_create on error.
- make pthread__stackid_setup return a error rather than calling err(2).

if mlock() fails in pthread_create(), return EAGAIN instead of
failing an assertion.

starting the pthread library (ie. calling pthread__start()) before
any threads are created turned out to be not such a good idea.
there are stronger requirements on what has to work in a forked child
while a process is still single-threaded. so take all that stuff
back out and fix the problems with single-threaded programs that
are linked with libpthread differently, by checking if the library
has been started and doing completely different stuff if it hasn't been:
- for pthread_rwlock_timedrdlock(), just fail with EDEADLK immediately.
- for sem_wait(), the only thing that can unlock the semaphore is a
signal handler, so use sigsuspend() to wait for a signal.
- for pthread_mutex_lock_slow(), just go into an infinite loop
waiting for signals.

I also noticed that there's a "sem2" test that has never worked in its
single-threaded form. the problem there is that a signal handler tries
to take a sem_t interlock which is already held when the signal is received.
fix this too, by adding a single-threaded case for sig_trywait() that
blocks signals instead of using the userland interlock.

in pthread_kill() and pthread_suspend_np(), return without doing anything
if the target thread is a zombie.

in all the functions that didn't do so already, verify a pthread_t before
dereferencing it (under #ifdef ERRORCHECK, since these checks are not
mandated by the standard).

clean up some debugging stuff.

in pthread_mutex_lock_slow(), pthread_rwlock_timedrdlock() and sem_wait(),
call pthread__start() if it hasn't already been called. this avoids
an internal assertion from the library if these routines are used
before any threads are created and they need to sleep.
fixes PR 20256, PR 24241, PR 25722, PR 26096.

make this compile without PTHREAD_MLOCK_KLUDGE.

branches: 1.41.2;
Keep the kernel updated with signal action signal masks (act.sa_mask) until
threads are started, since before that the traditional signal invocation
method will be used. Fixes regress/lib/libpthread/sigmask2.

Default back pthread__concurrency to 1 until we fix the regression test
related to pthread_kill().

If $PTHREAD_CONCURRENCY is not set, set it to ncpu

Add charles' mlock fixes, protected with PTHREAD_MLOCK_KLUDGE and enabled
by default, until we come up with a real fix.

Add some additional assertions and debugging printf()s.

- fix pthread_detach with an already exit'ed thread.
namely, put the thread to deadqueue rather than just leaking it.
- fix a race between pthread_detach/join and pthread_exit,
which also causes dead thread leaks.

enhance MI pthreads code to support hppa:
- statically initialize all global spin locks. on hppa, 0 means
the lock is held, so leaving them with the default value doesn't work.
- compare functions pointers using a function-pointer type rather than
an integral type. on hppa, function pointers may be indirect,
so we need to trigger gcc to emit calls to the function-pointer
canonicalization routines in the millicode.
- on hppa the stack grows up, so handle that using the STACK_* macros.

const'ify the arguments to pthread__assertfunc and _errorfunc;
needed because the __func__ pseudo-var is strictly "const*" in gcc-3.4

branches: 1.33.2;
add libpthread part of concurrency support for SA on MP systems
- enable concurrency according to environment variable PTHREAD_CONCURRENCY
- add idle VP wakeup if there are additional jobs and idle VPs
- make reidlequeue per VP
- enable spinning for locks
- fix race condition in alarm processing
- fix race condition in mutex locking
- make debugging output line buffered and add VP prefix to debug lines

Handle block/unblock for threads in critical section without
sa_unblockyield.

XXX g/c sa_unblockyield in kernel later

PR/23791: Patrick Latify: Fix memory leak on thread creation failure.

Add:

int pthread_attr_setcreatesuspend_np(pthread_attr_t *);
int pthread_suspend_np(pthread_t);
int pthread_resume_np(pthread_t);

needed for java. Approved and fixed by cl.

Split out pthread_{set,get}specific() into a separate file and arrange
for that file to not be built with profiling. This makes it reasonable to
use pthread_{set,get}specific() to implement thread-safe profiline call counts.

Lock accesses to pt_flags and pt_cancel.

When disabling cancellation, clear the pt_cancel flag if it was set
and note the cancellation request with PT_FLAG_CS_PENDING. This avoids
a problem where a cancellation request entered but not acted upon before
pthread_setcanclstate(PTHREAD_CANCEL_DISABLE) is called would still be
aceted upon before cancellation was re-enabled.

Use _sys_write() instead of write() in the assertion and error
functions, to avoid invoking the cancellation machinery in the middle
of trying to print an assertion or error message.

Move initialization of variable 'nthreads' so that pthread_exit() from
the main thread, with no other threads created, exits cleanly instead
of triggering an assertion in pthread_next().

pthread.c was getting a bit unwieldly. Move pthread_attr stuff out
into a new file, and put the shared private structure definition in
pthread_int.h.

Implement a bunch of pthread_attr_() functions, which genuinely set and examine
pthread_attr_t objects, although most of the properties being set don't really
affect threads yet:

pthread_attr_{get,set}guardsize()
pthread_attr_{get,set}inheritsched()
pthread_attr_{get,set}scope()
pthread_attr_{get,set}stack()
pthread_attr_setstack{size,addr}()

Remove some useless assertions and error checks in the existing pthread_attr()
routines.

Implement pthread_attr_get_np(), to examine the attributes of an existing
thread. Idea and interface from FreeBSD.

Change PTHREAD_ERRORMODE environment variable to PTHREAD_DIAGASSERT, and
make it behave like libc's LIBC_DIAGASSERT. The way to disable error-checking
and aborting is now "PTHREAD_DIAGASSERT=AEL", rather than
"PTHREAD_ERRORMODE=ignore".

Add strong alias for setcancelstate.

Initialize pt_trapuc in pthread__initthread().

Tiny bit of infrastructure for ABI-supported thread-ID storage.

Add pthread_attr_getstack{,size,addr} needed by jdk

pass lint:
1. add new pthread__abort() and change pthread_assert(0) to it.
2. put constcond in the right place (in the macro).
3. no space after pthread__assert macro.

Move call to pthread__alarm_init() from pthread_init() (called at load
time) to pthread__start() (called on the first call of
pthread_create()), so that there's no opportunity for a fork() to
clear the per-process timer.

Problem pointed out by a test program from Mihai CHELARU on
current-users.

Introduce a pthread__error() macro, for detected application errors as
opposed to internal errors. The setting of the PTHREAD_ERRORMODE
environment variable determines the runtime behavior. Valid settings
are "ignore", "abort", and "print". The default is currently "abort".

When threads are started, register an atfork callback to clear
pthread__started in the child process. This lets sleep() work in the
child, as required by POSIX.

Re-enable nanosleep(); kernel bug is squished (kern_time.c:1.67).

add __RCSID()

Disable userlevel nanosleep() until I find the timer/alarm bug that it
is tickling.

In pthread_create(), rearrange name-setting code so that name is set
to NULL when the attribute structure does not have a private field
allocated.

Add support for naming a thread, using an API compatible with Tru64 Unix:
* pthread_attr_getname_np()
* pthread_attr_setname_np()
* pthread_getname_np()
* pthread_setname_np()

In addition to being query'able by the application (for log messages, etc.),
it is intended that these names can show up in the debugger.

Reviewed by nathanw.

In pthread_exit(), clear pt_canceled in addition to setting
PT_FLAG_CS_DISABLED in pt_flags.

Prevents cancellation cleanup handlers from thinking that they've been
canceled and exiting mid-handler. Problem spotted by Matt Thomas.

Define a pthread-specific assert function, pthread__assert(), that
bails out without trying to flush stdio buffers.

Use pthread__sched_sleepers() instead of iterating over sleep queues
ourself.

cosmetics

Make pthread_join() accomodate spurious wakeups from pthread__block().

Add some threadlib stubs that are used by X:
* thr_create() (stub in libc aborts)
* thr_exit() (stub in libc calls exit())
* thr_yield() (stub in libc does nothing)

Also make the libc thr_self() stub not abort.

XXX Should we add thrattr_*() stubs for setting the detach-state?

- constify the get functions
- remove argument variable names since they pollute the namespace
- add mock implementation of the missing stack functions
XXX: we need to do this right [i.e. actually get and set sizes and addresses]
but the current stack setup makes this a bit complicated.

de-lint

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread.c was initially added on branch nathanw_sa.

libpthread: Fix expected symbols on alpha and hppa.

These have PTHREAD__ASM_RASOPS and define a different subset of
symbols which get exported. None of these pthread__ras_* and
pthread__lock_* symbols _should_ be exported but we can fix that in a
future revbump.

PR lib/58838: shared libraries in base should all have expsym lists

libpthread: Add expected symbols list.

PR lib/58838: shared libraries in base should all have expsym lists

revert previous; needs more discussion/testing.

Remove printflike annotation from pthread_setname_np()

This is a temporary measure as the annotation breaks too much existing
code which is not designed with it in mind (has an unnecessary, now, 3rd
arg from when all calls were expected to have three).

This can be reverted once all calls in the tree have been audited, and
corrected as necessary.

Make pthread_setname_np variadic for compatibility with FreeBSD and GLIBC.
Bump.

libc: Expose some more pthread stubs.

Additionally, provide the option for pthread.h to expose
pthread_create so that libc can provide a weak stub for it, if you
define _NETBSD_PTHREAD_CREATE_WEAK.

This allows libraries to use the functions without linking against
libpthread for applications that don't need threads. For example,
with OpenSSL's libcrypto, as long as you don't try to raise
OSSL_set_max_threads above 0, it won't ever try pthread_create.

New pthread.h macros defined:

pthread_attr_destroy __libc_thr_attr_destroy
pthread_attr_init __libc_thr_attr_init
pthread_attr_setdetachstate __libc_thr_attr_setdetachstate
pthread_detach __libc_thr_detach
pthread_join __libc_thr_join

New pthread.h macros defined if _NETBSD_PTHREAD_CREATE_WEAK is
defined first:

pthread_create __libc_thr_create

New symbols defined by libc as weak aliases for stubs that always
fail but can be overridden by libpthread:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that libc already defined __libc_thr_create.

New internal stubs, not for overriding:

__libc_thr_detach_stub
__libc_thr_join_stub

These were formerly called __libc_pthread_detach and
__libc_pthread_join, respectively, and libc has defined weak aliases
pthread_detach and pthread_join for them since 2013. I retained all
four of those symbols, but defined new ones in the __ namespace that
match the pattern for other thread stubs to avoid raising questions
about why the pattern seems to be broken. I left a comment noting
this fact too.

New pthread strong aliases to override the libc symbols:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that pthread already defined a strong __libc_thr_create alias
for the real pthread_create, so there is no need to add a new one.

PR lib/59685: libcrypto should not depend on libpthread

branches: 1.41.18;
Remove namespace restriction from pthread_condattr_{g,s}etclock(3)

These functions were marked as _NETBSD_SOURCE when introduced to the
sources. In fact they are regular POSIX threading functions available
since the 2001 standard. There is an older mention about alignment with
"IEEE Std 1003.1j-2000".

This corrects usage of these functions when a source code is compiled
with a POSIX namespace option.

fix duplicate declaration of pthread_atfork in unistd.h

Needs to be protected since it has a timespec argument. Found by lint(1)

branches: 1.38.6;
POSIX harder the pthread_mutex_timedlock(3) prototype

Add missing __restrict keyword to the first pointer parameter.

It was already used for the second argument, should not be a functional
change and generated code should be the same.

This new form is now aligned with POSIX.

branches: 1.37.2;
GSoC 2016 Charles Cui: Implement thread priority protection based on work
by Andy Doran. Also document the get/set pshared thread calls as not
implemented, and add a skeleton implementation that is disabled.
XXX: document _sched_protect(2).

Add pthread_getcpuclockid(3)

add pthread_condattr_setclock(3)

branches: 1.34.6; 1.34.8; 1.34.12;
Add pthread_getattr_np()

merge christos-time_t

branches: 1.32.6;
add pthread_cond_has_waiters_np()

Back out previous because it breaks the build.

glibc compatibile PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP, requested by
rafal@.

branches: 1.29.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.28.4;
sched_setparam: fix the case when incorrect (according to the class)
in-kernel priority is used. Reported by <drochner>.

Minor fixes for scheduling calls to conform the POSIX:
- If pid is equal to zero, use the calling process;
- In case of permission problem, return EPERM instead of EACESS;
- sched_setscheduler() should return previously used policy;
- pthread_* calls should return the error code or zero;

Should fix the namespace problems (and builds of some packages):
- Move cpuset_t defintion from pset.h to sched.h;
- Remove the #include of pset.h in pthread.h;

s/set/test/

only include <sys/pset.h> if _NETBSD_SOURCE

Implementation of processor-sets, affinity and POSIX real-time extensions.
Add schedctl(8) - a program to control scheduling of processes and threads.

Notes:
- This is supported only by SCHED_M2;
- Migration of LWP mechanism will be revisited;

Proposed on: <tech-kern>. Reviewed by: <ad>.

- Fix pthread_rwlock_trywrlock() which was broken.

- Add new functions: pthread_mutex_held_np, mutex_owner_np, rwlock_held_np,
rwlock_wrheld_np, rwlock_rdheld_np. These match the kernel's locking
primitives and can be used when porting kernel code to userspace.

- Always create LWPs detached. Do join/exit sync mostly in userland. When
looped on a dual core box this seems ~30% quicker than using lwp_wait().
Reduce number of lock acquire/release ops during thread exit.

Add thr_curcpu(), pthread_curcpu_np().

Add pthread_equal() to libc stubs; this makes a lot of sense for
threadsafe libraries implementing own locking functions.
Ride on yesterday's minor version bumps.

branches: 1.21.12;
Add restrict qualifiers to <pthread.h> function arguments.

branches: 1.20.2;
Implement pthread_attr_{set,get}schedpolicy() at the same level as the other
scheduling stuff: only handle SCHED_OTHER. Like the rest of the scheduling
stuff, this is for the benefit of code that can't be bothered to test against
_POSIX_THREAD_PRIORITY_SCHEDULING.

Fix typo in comment.

Add #defines to make use of libc stubs for pthread_cond_wait() and
pthread_cond_timedwait().

XXX as noted in the comments, in the situations where these are
useful, they should never be called in a single-threaded
process. Perhaps they should die rather than return 0.

Addresses xsrc/28630.

Use CPP macros to cause many libpthread functions used by applications
to be transformed into the do-nothing-when-libpthread-isn't-linked libc
stub names. This will permit library code that uses <pthread.h> and pthread
functions "defensively" to not need to link against libpthread and not need
to be patched to the threadlib.h API.

Add pthread_{g,s}etschedparam, that do nothing.

Add:

int pthread_attr_setcreatesuspend_np(pthread_attr_t *);
int pthread_suspend_np(pthread_t);
int pthread_resume_np(pthread_t);

needed for java. Approved and fixed by cl.

Add prototypes and necessary symbols for a bunch of pthread_attr_*() functions,
including (interface borrowed from FreeBSD) pthread_attr_get_np().

Remove a few limit-indicating symbols that don't actually need to be defined
(and if they did need to be defined, it would be in limits.h, not here).

#include <pthread_types.h> instead of "pthread_types.h"

Define PTHREAD_* structure initializers from _PTHREAD_* macros from
pthread_types.h.

pthread_kill() prototype moved to <signal.h>.

Add pthread_attr_getstack{,size,addr} needed by jdk

Add support for naming a thread, using an API compatible with Tru64 Unix:
* pthread_attr_getname_np()
* pthread_attr_setname_np()
* pthread_getname_np()
* pthread_setname_np()

In addition to being query'able by the application (for log messages, etc.),
it is intended that these names can show up in the debugger.

Reviewed by nathanw.

Implement pthread_atfork() (in libc, because the required threadlib
stub behavior is exactly the same as the usual behavior).

Remove the dummy stack attr routines -- best to catch broken applications
and fix them to use the proper feature test instead (as has already been
done with the openldap package).

Move _POSIX_THREADS to the right place and define a few other symbolic
constants for the features we gained with the merge of the nathanw_sa
branch.

Ok'd by thorpej.

Closes my PR 19930.

Remove another variable name.

- constify the get functions
- remove argument variable names since they pollute the namespace
- add mock implementation of the missing stack functions
XXX: we need to do this right [i.e. actually get and set sizes and addresses]
but the current stack setup makes this a bit complicated.

'destructor' is a bit too generic; use 'destruct' instead for
pthread_key_create() prototype
this fixes compilation of Python 2.2 with thread support
XXX perhaps should drop all variable names from prototypes?

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread.h was initially added on branch nathanw_sa.

Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Conditionalised support for 1:1 threads. Needs associated kernel changes
and more work to be useful.

Coverity CID 882: Avoid NULL deref.

gettimeofday();TIMEVAL_TO_TIMESPEC(); is exactly equivalent to
clock_gettime(CLOCK_REALTIME), except the latter may have more preicison
some day. So, use that.

Avoid passing zero-valued timespecs to timer_settime() when we want to set
a timer, as that will clear the timer instead. Pass in a safely in-the-past
value instead.
Addresses PR lib/28700.

(XXX passing in values between 0 and 1000 nanoseconds will still fail, but
that bug needs to be fixed in timer_settime(), not here)

enhance MI pthreads code to support hppa:
- statically initialize all global spin locks. on hppa, 0 means
the lock is held, so leaving them with the default value doesn't work.
- compare functions pointers using a function-pointer type rather than
an integral type. on hppa, function pointers may be indirect,
so we need to trigger gcc to emit calls to the function-pointer
canonicalization routines in the millicode.
- on hppa the stack grows up, so handle that using the STACK_* macros.

branches: 1.9.4;
add libpthread part of concurrency support for SA on MP systems
- enable concurrency according to environment variable PTHREAD_CONCURRENCY
- add idle VP wakeup if there are additional jobs and idle VPs
- make reidlequeue per VP
- enable spinning for locks
- fix race condition in alarm processing
- fix race condition in mutex locking
- make debugging output line buffered and add VP prefix to debug lines

Consistently use pthread__assert() rather than err() to assert that
timer_settime() worked, and don't bother calling err() after a
pthread__assert() call.

Remove debugging code accidentally committed.

* Use a <= comparison rather than a < comparison when searching for
alarms to process.

* Remove a redundant assignment.

* Make a debug message more consistent with the others.

add __RCSID()

Adapt to pthread__assert() and remove local debug toggle.

de-lint

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_alarms.c was initially added on branch nathanw_sa.

Each get function "has" a set counterpart, it doesn't "contain" it.

branches: 1.21.6; 1.21.12;
Update conformance to the 2001 revision.

Add missing word.

Rewrite and include a short summary.

Finally, split pthread_attr_getschedpolicy() and pthread_attr_setschedpolicy()
to their own page.

Split pthread_attr_getschedparam() and pthread_attr_setschedparam() to a
separate page.

Split pthread_attr_getinheritsched() and pthread_attr_setinheritsched() to a
separate page, and actually document these.

Move pthread_attr_getguardsize() and pthread_attr_setguardsize() to a
separate page, and actually document these.

Move the thread stack attribute functions to a separate page. In addition,
document pthread_attr_getstack() and pthread_attr_setstack(). Try to also
document these a little better (in particular, note some items in the long
list of caveats and questions related to application-controlled thread stack).

Split pthread_attr_getdetachstate(3) and pthread_attr_setdetachstate(3) to a
separate page. Once ready, the main page, pthread_attr(3), will contain a
summary of the various thread attribute functions.

Remove pthread_attr_getscope() and pthread_attr_setscope().

pthread_attr(3): sync with reality i.e. uncomment descriptions about
various functions we support. Fix license style.

Avoid duplicate .Pp.

.\" are comments, not ./"

branches: 1.8.4; 1.8.6;
s/explicitely/explicitly/

Move TNF licenses to 2 clause form

branches: 1.6.6;
Use more markup. New sentence, new line.

branches: 1.5.6;
PR lib/34931: PTHREAD_ATTR(3) doesn't document PTHREAD_CREATE_JOINABLE and
PTHREAD_CREATE_DETACHED macros. From Matthew Mondor.

Add restrict qualifiers to <pthread.h> function arguments.

branches: 1.3.6;
Remove trailing whitespace.

$NetBSD$, not $Header$.

Man pages for libpthread, adapted from FreeBSD by Stephen Degler.

libc: Expose some more pthread stubs.

Additionally, provide the option for pthread.h to expose
pthread_create so that libc can provide a weak stub for it, if you
define _NETBSD_PTHREAD_CREATE_WEAK.

This allows libraries to use the functions without linking against
libpthread for applications that don't need threads. For example,
with OpenSSL's libcrypto, as long as you don't try to raise
OSSL_set_max_threads above 0, it won't ever try pthread_create.

New pthread.h macros defined:

pthread_attr_destroy __libc_thr_attr_destroy
pthread_attr_init __libc_thr_attr_init
pthread_attr_setdetachstate __libc_thr_attr_setdetachstate
pthread_detach __libc_thr_detach
pthread_join __libc_thr_join

New pthread.h macros defined if _NETBSD_PTHREAD_CREATE_WEAK is
defined first:

pthread_create __libc_thr_create

New symbols defined by libc as weak aliases for stubs that always
fail but can be overridden by libpthread:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that libc already defined __libc_thr_create.

New internal stubs, not for overriding:

__libc_thr_detach_stub
__libc_thr_join_stub

These were formerly called __libc_pthread_detach and
__libc_pthread_join, respectively, and libc has defined weak aliases
pthread_detach and pthread_join for them since 2013. I retained all
four of those symbols, but defined new ones in the __ namespace that
match the pattern for other thread stubs to avoid raising questions
about why the pattern seems to be broken. I left a comment noting
this fact too.

New pthread strong aliases to override the libc symbols:

__libc_thr_attr_destroy
__libc_thr_attr_init
__libc_thr_attr_setdetachstate
__libc_thr_detach
__libc_thr_join

Note that pthread already defined a strong __libc_thr_create alias
for the real pthread_create, so there is no need to add a new one.

PR lib/59685: libcrypto should not depend on libpthread

branches: 1.21.6;
pthread: Nix trailing whitespace.

libpthread: Move namespacing include to top of .c files.

Stuff like libc's namespace.h, or atomic_op_namespace.h, which does
namespacing tricks like `#define atomic_cas_uint _atomic_cas_uint',
has to go at the top of each .c file. If it goes in the middle, it
might be too late to affect the declarations, and result in compile
errors.

I tripped over this by including <sys/atomic.h> in mips
<machine/lock.h>.

(Maybe we should create a new pthread_namespace.h file for the
purpose, but this'll do for now.)

Use the pta_magic field in pthread attribute

Set PT_ATTR_DEAD on pthread_attr_destroy().
Check pta_magic before using pthread_attr_t in a bunch of other functions.

branches: 1.18.4;
pthread__attr_init_private:
malloc+memset -> calloc. Also initialize all values to the proper
defaults.
This fixes the "rustc panic" discussed on pkgsrc-users.
OK: joerg

Export the guard size of the main thread via vm.guard_size. Add a
complementary writable sysctl for the initial guard size of threads
created via pthread_create. Let the existing attribut accessors do the
right thing. Raise the default guard size for threads to 64KB.

branches: 1.16.24;
Separate pthread_t from thread stack. Drop additional alignment
restrictions on the thread stack. Remove remaining parts of stackid.

branches: 1.15.6;
move namespace protection before inclusion.

Add a weak alias for pthread_attr_get_np. Noted my Matthias Drochner.

Add pthread_getattr_np()

fix -Wsign-compare issues

branches: 1.11.4;
pthread_attr_getschedparam: allow without preceding setparam on the
attr structure.

Now that we have all the scheduling gunk, make these do something useful:

pthread_attr_get_np
pthread_attr_setschedparam
pthread_attr_getschedparam
pthread_attr_setschedpolicy
pthread_attr_getschedpolicy

pthread_attr_setstacksize: EINVAL if the requested size is less than
sysconf(_SC_THREAD_STACK_MIN).

branches: 1.8.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.7.4;
add missing static decls.

branches: 1.6.2;
Remove PT_FIXEDSTACKSIZE_LG.

branches: 1.5.8;
pthread_attr_getschedpolicy() wasn't setting the return buffer at all.
SCHED_OTHER happens to be 0, so this assignment to "int *" succeeds,
and becomes a no-op.

Fix by dereferencing "policy" to do the assignment, thus filling the
return buffer with 0.

branches: 1.4.2;
Implement pthread_attr_{set,get}schedpolicy() at the same level as the other
scheduling stuff: only handle SCHED_OTHER. Like the rest of the scheduling
stuff, this is for the benefit of code that can't be bothered to test against
_POSIX_THREAD_PRIORITY_SCHEDULING.

Add:

int pthread_attr_setcreatesuspend_np(pthread_attr_t *);
int pthread_suspend_np(pthread_t);
int pthread_resume_np(pthread_t);

needed for java. Approved and fixed by cl.

sprinkle ARGSUSED for good cheer.

pthread.c was getting a bit unwieldly. Move pthread_attr stuff out
into a new file, and put the shared private structure definition in
pthread_int.h.

Add pthread_getattr_np to the NAME section

document pthread_getattr_np

Sort sections.

Add missing list end.

Document pthread_attr_get_np(). Fixes PR lib/39289 from ahoka@.

Remove superfluous Tn.

Add pthread_attr_setdetachstate to NAME section

Update conformance to the 2001 revision.

Split pthread_attr_getdetachstate(3) and pthread_attr_setdetachstate(3) to a
separate page. Once ready, the main page, pthread_attr(3), will contain a
summary of the various thread attribute functions.

pthread: Document the setstack vs setguardsize bug.

Suggest the safe, compatible workaround.

branches: 1.5.14;
Add missing function names in the NAME section for rest of the man pages in libpthread

Export the guard size of the main thread via vm.guard_size. Add a
complementary writable sysctl for the initial guard size of threads
created via pthread_create. Let the existing attribut accessors do the
right thing. Raise the default guard size for threads to 64KB.

branches: 1.3.8;
_SC_PAGESIZE is not the page size; it's a symbolic code for retrieving
the page size.

Whitespace nit.

Move pthread_attr_getguardsize() and pthread_attr_setguardsize() to a
separate page, and actually document these.

s/tread/thread/

branches: 1.4.16;
Add missing function names in the NAME section for rest of the man pages in libpthread

s/inherited/inherit/

branches: 1.2.6; 1.2.12;
Sort sections.

Split pthread_attr_getinheritsched() and pthread_attr_setinheritsched() to a
separate page, and actually document these.

Add missing function names in the NAME section for rest of the man pages in libpthread

Add missing prototypes. Noted by wiz@ -- thanks.

Sort errors.

Rationalize by moving

pthread_attr_setname_np(3) to pthread_attr_getname_np(3); and
pthread_setname_np(3) to pthread_getname_np(3).

Sort sections.

Same here; fix LIBRARY, add COMPATIBILITY/ERRORS, etc.

branches: 1.1.2;
document following functions.
pthread_attr_getname_np
pthread_attr_setname_np
pthread_getname_np
pthread_setname_np

Add missing function names in the NAME section for rest of the man pages in libpthread

Group also pthread_attr_getschedparam(3) and pthread_attr_getschedpolicy(3),
as these two pages fit well together.

Split pthread_attr_getschedparam() and pthread_attr_setschedparam() to a
separate page.

Group also pthread_attr_getschedparam(3) and pthread_attr_getschedpolicy(3),
as these two pages fit well together.

Finally, split pthread_attr_getschedpolicy() and pthread_attr_setschedpolicy()
to their own page.

Remove superfluous Tn.

Add missing function names in the NAME section for rest of the man pages in libpthread

Add missing list end.

Move pthread_attr_{g,s}etscope() to a separate page and try to document
these better. Should fix PR lib/41831 from Wolfgang Stukenbrock.

XXX: Please proofread.

pthread: Document the setstack vs setguardsize bug.

Suggest the safe, compatible workaround.

branches: 1.8.14;
Remove superfluous Tn.

Add missing function names in the NAME section for rest of the man pages in libpthread

Fix some lint.

Too much or too little whitespace;
improve macro usage;
add missing .El;
merge error sections for same error code.

Note also alignment via mmap(2), as done in NetBSD.

Use .Ss for non-standard subtitle.

Update conformance to the 2001 revision.

Fix some minor errors/typos and clarify.

Move the thread stack attribute functions to a separate page. In addition,
document pthread_attr_getstack() and pthread_attr_setstack(). Try to also
document these a little better (in particular, note some items in the long
list of caveats and questions related to application-controlled thread stack).

Add the last compatibility note for the NetBSD extensions.

branches: 1.3.8;
Convert TNF licenses to new 2 clause variant

branches: 1.2.32;
Various fixes.

Document pthread_attr_setcreatesuspend_np, pthread_suspend_np, pthread_resume_np

Rationalize by moving

pthread_attr_setname_np(3) to pthread_attr_getname_np(3); and
pthread_setname_np(3) to pthread_getname_np(3).

Sort sections.

Same here; fix LIBRARY, add COMPATIBILITY/ERRORS, etc.

branches: 1.2.6;
Remove trailing whitespace.

branches: 1.1.2; 1.1.6; 1.1.8;
document following functions.
pthread_attr_getname_np
pthread_attr_setname_np
pthread_getname_np
pthread_setname_np

Remove description of pthread_barrierattr_getpshared and pthread_barrierattr_setpshared

These functions are described in pthread_barrierattr(3) man page

Add rest of the pthread_barrier functions in the NAME section

Fix some lint.

Too much or too little whitespace;
improve macro usage;
add missing .El;
merge error sections for same error code.

GSoC 2016 Charles Cui: Implement thread priority protection based on work
by Andy Doran. Also document the get/set pshared thread calls as not
implemented, and add a skeleton implementation that is disabled.
XXX: document _sched_protect(2).

Update ERRORS, tidy SEE ALSO.

Add comma in enumeration, sort errors.

Split long lines, .Dv NULL, note pthread_barrierattr_init(3).

Merge/group man pages:
- pthread_barrier_{init,destroy,wait} into pthread_barrier(3).
- pthread_cond_* into pthread_cond(3).
- pthread_mutex_* into pthread_mutex(3).
- pthread_rwlock_* into pthread_rwlock(3).
- pthread_spin_* into pthread_spin(3).
Cross-link amongst groups.
OK jruoho@

libpthread: Move namespacing include to top of .c files.

Stuff like libc's namespace.h, or atomic_op_namespace.h, which does
namespacing tricks like `#define atomic_cas_uint _atomic_cas_uint',
has to go at the top of each .c file. If it goes in the middle, it
might be too late to affect the declarations, and result in compile
errors.

I tripped over this by including <sys/atomic.h> in mips
<machine/lock.h>.

(Maybe we should create a new pthread_namespace.h file for the
purpose, but this'll do for now.)

- Try to eliminate a hang in "parked" I've been seeing while stress testing.
Centralise wakeup of deferred waiters in pthread__clear_waiters() and use
throughout libpthread. Make fewer assumptions. Be more conservative in
pthread_mutex when dealing with pending waiters.

- Remove the "hint" argument everywhere since the kernel doesn't use it any
more.