libpthread: New pthread__smt_wait to put CPU in low power for spin.

This is now distinct from pthread__smt_pause, which is for spin lock
backoff with no paired wakeup.

On Arm, there is a single-bit event register per CPU, and there are two
instructions to manage it:

- wfe, wait for event -- if event register is clear, enter low power
mode and wait until event register is set; then exit low power mode
and clear event register

- sev, signal event -- sets event register on all CPUs (other
circumstances like interrupts also set the event register and cause
wfe to wake)

These can be used to reduce the power consumption of spinning for a
lock, but only if they are actually paired -- if there's no sev, wfe
might hang indefinitely. Currently only pthread_spin(3) actually
pairs them; the other lock primitives (internal lock, mutex, rwlock)
do not -- they have spin lock backoff loops, but no corresponding
wakeup to cancel a wfe.

It may be worthwhile to teach the other lock primitives to pair
wfe/sev, but that requires some performance measurement to verify
it's actually worthwhile. So for now, we just make sure not to use
wfe when there's no sev, and keep everything else the same -- this
should fix severe performance degredation in libpthread on Arm
without hurting anything else.

No change in the generated code on amd64 and i386. No change in the
generated code for pthread_spin.c on arm and aarch64 -- changes only
the generated code for pthread_lock.c, pthread_mutex.c, and
pthread_rwlock.c, as intended.

PR port-arm/57437

XXX pullup-10

branches: 1.1.4; 1.1.18; 1.1.28; 1.1.36;
Preliminary files for AARCH64 (64-bit ARM) support.
Enough for a distribution build.

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

branches: 1.4.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.3.6;
Provide PTHREAD__ASM_RASOPS for alpha.

The gcc generated lock try RAS is broken as the store needs to be the last
instruction.

branches: 1.2.24;
Merge the nathanw_sa branch.

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

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

branches: 1.7.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Merge the nathanw_sa branch.

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

Restore PTHREAD__ASM_RASOPS for alpha (from _context_u.S), that was
lost in previous cleanup.

ok by ad@.

branches: 1.1.4;
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

make pthread__sp unsigned long.

branches: 1.6.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.5.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.4.6;
Provide PTHREAD__ASM_RASOPS for alpha.

The gcc generated lock try RAS is broken as the store needs to be the last
instruction.

branches: 1.3.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

Remove possible race condition in upcall recycling.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Two fixes:
* In switch-away cases, write PT_SWITCHTO last (after PT_SWITCHTOUC), so
that pthread__resolve_locks() doesn't see an empty SWITCHTOUC value. This
also permits pthread__resolve_locks() to use the presence of PT_SWITCHTO
as a sign that the thread has done all of its necessary chain work.

* Make the return-point of pthread__switch global and visible, so that its
address can be compared to the PC of a thread, again as a sign that its
chain-work is done.

(other architectures in progress, after they get the *previous* asm fix...)

Merge the nathanw_sa branch.

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

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

Use RET macro for returning.

NetBSD/acorn26 has used APCS-32 for years, so unifdef -U__APCS_26__.

Merge the nathanw_sa branch.

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

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

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

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Include <sys/types> before <sys/lock.h>, shuffle <ucontext.h>.

Merge the nathanw_sa branch.

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

libpthread: New pthread__smt_wait to put CPU in low power for spin.

This is now distinct from pthread__smt_pause, which is for spin lock
backoff with no paired wakeup.

On Arm, there is a single-bit event register per CPU, and there are two
instructions to manage it:

- wfe, wait for event -- if event register is clear, enter low power
mode and wait until event register is set; then exit low power mode
and clear event register

- sev, signal event -- sets event register on all CPUs (other
circumstances like interrupts also set the event register and cause
wfe to wake)

These can be used to reduce the power consumption of spinning for a
lock, but only if they are actually paired -- if there's no sev, wfe
might hang indefinitely. Currently only pthread_spin(3) actually
pairs them; the other lock primitives (internal lock, mutex, rwlock)
do not -- they have spin lock backoff loops, but no corresponding
wakeup to cancel a wfe.

It may be worthwhile to teach the other lock primitives to pair
wfe/sev, but that requires some performance measurement to verify
it's actually worthwhile. So for now, we just make sure not to use
wfe when there's no sev, and keep everything else the same -- this
should fix severe performance degredation in libpthread on Arm
without hurting anything else.

No change in the generated code on amd64 and i386. No change in the
generated code for pthread_spin.c on arm and aarch64 -- changes only
the generated code for pthread_lock.c, pthread_mutex.c, and
pthread_rwlock.c, as intended.

PR port-arm/57437

XXX pullup-10

libpthread: Use __nothing, not /* nothing */, for empty macros.

No functional change intended -- just safer to do it this way in case
the macros are used in if branches or comma expressions.

PR port-arm/57437 (pthread__smt_pause/wake issue)

XXX pullup-10

branches: 1.11.2; 1.11.10;
G/C __APCS_26__ support

branches: 1.10.2; 1.10.4; 1.10.6;
Typo in comment

branches: 1.9.18;
Use the thumb1 versions of sev/wfe for thumb && armv6+.
if using armv5t don't do anything for thumb.

Use .inst instead of wfe/sev to shut up gas.

branches: 1.7.2;
Add a pthread__smt_wake and add support for it on arm along with
pthread__smt_pause. These are implemented using the ARM instructions
SEV (wake) and WFE (pause). These are treated as NOPs on ARM CPUs that
don't support them.

branches: 1.6.4;
make pthread__sp unsigned long.

branches: 1.5.2;
Remove unused code that's confusing when using cscope/opengrok.

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

add missing backslash

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

Use RET macro for returning.

Remove possible race condition in upcall recycling.

Typo in comment.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Two fixes:
* In switch-away cases, write PT_SWITCHTO last (after PT_SWITCHTOUC), so
that pthread__resolve_locks() doesn't see an empty SWITCHTOUC value. This
also permits pthread__resolve_locks() to use the presence of PT_SWITCHTO
as a sign that the thread has done all of its necessary chain work.

* Make the return-point of pthread__switch global and visible, so that its
address can be compared to the PC of a thread, again as a sign that its
chain-work is done.

(other architectures in progress, after they get the *previous* asm fix...)

NetBSD/acorn26 has used APCS-32 for years, so unifdef -U__APCS_26__.

Merge the nathanw_sa branch.

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

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

branches: 1.4.6;
Remove unnecessary include.

branches: 1.3.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.6;
Resurrect the function pointers for lock operations and allow each
architecture to provide asm versions of the RAS operations.

We do this because relying on the compiler to get the RAS right is not
sensible. (It gets alpha wrong and hppa is suboptimal)

Provide asm RAS ops for hppa.

(A slightly different version) reviewed by Andrew Doran.

branches: 1.1.18;
add hppa MD libpthread bits. translated from MIPS.

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

branches: 1.3.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

add hppa MD libpthread bits. translated from MIPS.

Use RAS_{START,END}_ASM_HIDDEN

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

branches: 1.1.4;
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

make pthread__sp unsigned long.

branches: 1.7.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.6.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

Resurrect the function pointers for lock operations and allow each
architecture to provide asm versions of the RAS operations.

We do this because relying on the compiler to get the RAS right is not
sensible. (It gets alpha wrong and hppa is suboptimal)

Provide asm RAS ops for hppa.

(A slightly different version) reviewed by Andrew Doran.

branches: 1.3.12;
__asm__ -> __asm

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

add hppa MD libpthread bits. translated from MIPS.

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

add hppa MD libpthread bits. translated from MIPS.

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

Delete special rules for syscall.S; they are not needed, and one of them
is wrong anyway.

branches: 1.1.1;
Initial revision

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

branches: 1.9.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.8.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

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.

- 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.

Add: pthread__atomic_cas_ptr, pthread__atomic_swap_ptr, pthread__membar_full
This is a stopgap until the thorpej-atomic branch is complete.

branches: 1.2.20; 1.2.24;
Merge the nathanw_sa branch.

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

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

branches: 1.8.6;
Selector registers are 16-bit and binutils 2.18 insists that only 16-bit
accesses are permitted on them. Therefore, change movl to movw. No change to
machine code generated.

branches: 1.7.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.12;
fix warning about indirect call without *

Punt to setcontext() system call if the PSL_T bit (single-step trap)
is set, so that the single-step trap happens in the thread's context
and not in the middle of _setcontext_u.

XXX might be able to do something here with iret, too, but it needs
more testing.

save&restore %fs and %gs registers for USER_LDT applications.
PR#26900

branches: 1.3.2;
use explicit "l" suffixes. (eg. lea -> leal)

Merge the nathanw_sa branch.

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

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

branches: 1.7.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Merge the nathanw_sa branch.

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

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

More wrapper functions, and some hacks for machine independent sleep
mechanisms.

Copyrights added to each file.

branches: 1.1.1;
Initial revision

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

Copyrights added to each file.

OK one more try at getting it right ...

branches: 1.1.1;
Initial revision

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

- 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: 1.5.8; 1.5.10;
Remove clause 3 and 4 from TNF licenses

branches: 1.4.6;
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.

branches: 1.3.24;
add __RCSID()

Merge the nathanw_sa branch.

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

libpthread: New pthread__smt_wait to put CPU in low power for spin.

This is now distinct from pthread__smt_pause, which is for spin lock
backoff with no paired wakeup.

On Arm, there is a single-bit event register per CPU, and there are two
instructions to manage it:

- wfe, wait for event -- if event register is clear, enter low power
mode and wait until event register is set; then exit low power mode
and clear event register

- sev, signal event -- sets event register on all CPUs (other
circumstances like interrupts also set the event register and cause
wfe to wake)

These can be used to reduce the power consumption of spinning for a
lock, but only if they are actually paired -- if there's no sev, wfe
might hang indefinitely. Currently only pthread_spin(3) actually
pairs them; the other lock primitives (internal lock, mutex, rwlock)
do not -- they have spin lock backoff loops, but no corresponding
wakeup to cancel a wfe.

It may be worthwhile to teach the other lock primitives to pair
wfe/sev, but that requires some performance measurement to verify
it's actually worthwhile. So for now, we just make sure not to use
wfe when there's no sev, and keep everything else the same -- this
should fix severe performance degredation in libpthread on Arm
without hurting anything else.

No change in the generated code on amd64 and i386. No change in the
generated code for pthread_spin.c on arm and aarch64 -- changes only
the generated code for pthread_lock.c, pthread_mutex.c, and
pthread_rwlock.c, as intended.

PR port-arm/57437

XXX pullup-10

branches: 1.20.24; 1.20.34; 1.20.42;
Avoid getcontext() as it triggers clobbering warnings. Use inline
assembler to get the fields directly. Saves a system call as side
effect.

branches: 1.19.4;
Allow storing and receiving the LWP private pointer via ucontext_t
on all platforms except VAX and IA64. Add fast access via register for
AMD64, i386 and SH3 ports. Use this fast access in libpthread to replace
the stack based pthread_self(). Implement skeleton support for Alpha,
HPPA, PowerPC, SPARC and SPARC64, but leave it disabled.

Ports that support this feature provide __HAVE____LWP_GETPRIVATE_FAST in
machine/types.h and a corresponding __lwp_getprivate_fast in
machine/mcontext.h.

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

make pthread__sp unsigned long.

branches: 1.17.2;
Remove unused code that's confusing when using cscope/opengrok.

- 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: 1.15.6; 1.15.8;
pthread__threadreg_get: mark it const.

branches: 1.14.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.13.2;
Cheat and add inlines for _atomic_cas_ptr() to work around gcc emitting
unneeded PIC stuff in mutex_lock() and mutex_unlock(), when a thread
register is used.

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

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.

Resurrect the function pointers for lock operations and allow each
architecture to provide asm versions of the RAS operations.

We do this because relying on the compiler to get the RAS right is not
sensible. (It gets alpha wrong and hppa is suboptimal)

Provide asm RAS ops for hppa.

(A slightly different version) reviewed by Andrew Doran.

- 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.

Add: pthread__atomic_cas_ptr, pthread__atomic_swap_ptr, pthread__membar_full
This is a stopgap until the thorpej-atomic branch is complete.

branches: 1.7.8; 1.7.12;
Instead of using hard-coded values for various registers, get them from the
current context. Valid values can change depending on how the kernel is
setup. i386 and amd64 happen to be setup differently.

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

branches: 1.5.4; 1.5.6;
Add ucontext conversion macros for an "extra" register set.

delint

de-lint

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

A couple of fixes to make libpthread really shared, i.e. not have text re-
locations:

- Don't declare pthread__switch_away global
- Do the PIC dance for pthread__switch_return_point and
pthread__locked_switch. Ideally these (and other) symbols would
be hidden.

Thanks to uwe@, dyoung@ and elad@ for help.

XXX sh3 is still to be done.
XXX vax does strange things.

branches: 1.8.2;
s/the the/the/ (only in sources that aren't regularly imported from
elsewhere).

use explicit "l" suffixes. (eg. lea -> leal)

Remove possible race condition in upcall recycling.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Two fixes:
* In switch-away cases, write PT_SWITCHTO last (after PT_SWITCHTOUC), so
that pthread__resolve_locks() doesn't see an empty SWITCHTOUC value. This
also permits pthread__resolve_locks() to use the presence of PT_SWITCHTO
as a sign that the thread has done all of its necessary chain work.

* Make the return-point of pthread__switch global and visible, so that its
address can be compared to the PC of a thread, again as a sign that its
chain-work is done.

(other architectures in progress, after they get the *previous* asm fix...)

Continue at the plain switch return point in pthread__switch, not the
locked one, in the !PIC case. From Tor Egge via Havard Eidnes.

Merge the nathanw_sa branch.

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

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

More wrapper functions, and some hacks for machine independent sleep
mechanisms.

Stylistic change.

branches: 1.1.1;
Initial revision

Dummy ia64 support

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

branches: 1.5.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.4.32;
fix typo in previous.

Don't use FP instructions when building for m68010. Fixes broken sun2 build.
G/C some unused code.

Merge the nathanw_sa branch.

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

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

branches: 1.7.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Merge the nathanw_sa branch.

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

This is not needed anymore.

SA_SIGINFO support for m68k (libpthread)

need <string.h>

add __RCSID()

Merge the nathanw_sa branch.

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

make pthread__sp unsigned long.

branches: 1.7.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.6.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

branches: 1.4.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

m68k does not sigcontext<->mcontext anymore.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

s/the the/the/ (only in sources that aren't regularly imported from
elsewhere).

fix pt_trapuc handling errors:
- movl to address register doesn't set flags (add explicit test)
- clr only clears a word (use clrl)

Remove possible race condition in upcall recycling.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Apply Nathan's switch-away fix and previous save-PT_UC from new stack fix.
Compile-tested only.

Merge the nathanw_sa branch.

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

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

branches: 1.3.48;
Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Merge the nathanw_sa branch.

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

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

branches: 1.3.8; 1.3.14;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.34;
Merge the nathanw_sa branch.

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

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

branches: 1.7.8; 1.7.14;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.14;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Merge the nathanw_sa branch.

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

make pthread__sp unsigned long.

branches: 1.8.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.7.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.22;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Catch up with changes to __fpregset_t.
In __longjmp14(), copy the FP CSR from the correct array and array slot.

Completes fix for PR port-mips/25942.

branches: 1.4.2;
Hide the register number constants behind an _R_ prefix, and also
rename FPBASE to _FPBASE, so that we avoid polluting the user's
name space when e.g. <sys/ptrace.h> is included. Previously, the
PC symbol in mips/regnum.h would conflict with the declaration of
the external variable by the same name in termcap.h, as discovered
by the ``okheaders'' regression test.

STACKSPACE of 16 is too small, given the use of CALLFRAME_SIZ;
increase to match.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

Remove possible race condition in upcall recycling.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Oops, that's switch_return_point, not locked_return_point. The perils
of cut-and-paste.

Apply switch-away fix and previous save-PT_UC from new stack fix.

Merge the nathanw_sa branch.

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

New files for OR1K support

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

When _SOFT_FLOAT is defined, don't save or restore user FP context.
(note: still needs some mk rules tweaking for MKSOFTFLOAT=yes to pass
-msoft-float to asm builds).

Remove a comment made obsolete by the previous commit.

branches: 1.3.4;
Don't need to set the MSR field here; _INITCONTEXT_U_MD() takes care
of that where necessary.

Merge the nathanw_sa branch.

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

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

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

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Add _REG_foo to ppc mcontext and use them instead of constants.

Merge the nathanw_sa branch.

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

Revert previous:
http://mail-index.netbsd.org/source-changes/2020/02/20/msg114173.html

Comment turned out to be wrong, and KASSERT fires for oea.

XXX
Need to revisit shortly...

libpthread sets initial value of MSR for lwp's. However, appropriate
value differs b/w oea/booke/ibm4xx, and there's no way to obtain it
from userland. Therefore, this initial value should be corrected by
cpu_setmcontext().

- Comment this in libpthread
- Add KASSERT in cpu_mcontext_validate()

branches: 1.7.44; 1.7.46;
make pthread__sp unsigned long.

branches: 1.6.2;
Remove unused code that's confusing when using cscope/opengrok.

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Add ucontext conversion macros for an "extra" register set.

Add _REG_foo to ppc mcontext and use them instead of constants.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

remove unreferenced temporary label

Fix a comment to say "Edit" instead of "Exit".

Remove possible race condition in upcall recycling.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Oops, reordering the PT_STATE/PT_SWITCHTOUC/PT_SWITCHTO assembly
killed a live value too soon. Fix by using a different register for
the PT_STATE immediate.

Update the switch_return point name in the static case.

Two fixes:
* In switch-away cases, write PT_SWITCHTO last (after PT_SWITCHTOUC), so
that pthread__resolve_locks() doesn't see an empty SWITCHTOUC value. This
also permits pthread__resolve_locks() to use the presence of PT_SWITCHTO
as a sign that the thread has done all of its necessary chain work.

* Make the return-point of pthread__switch global and visible, so that its
address can be compared to the PC of a thread, again as a sign that its
chain-work is done.

(other architectures in progress, after they get the *previous* asm fix...)

Fix typo in comment.

Merge the nathanw_sa branch.

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

RISC-V support that works on QEMU with a single hart.

Thanks for Simon Burge for plic(4).

We have _REG_SP so use it.

New files for Userland support of UCB RISC-V (both 32-bit and 64-bit)

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

branches: 1.6.6;
Recycle unused _REG_EXPEVT slot in mcontext for _REG_GBR.

As the size of mcontext is not changed, we avoid the hassle of
versioning all the calls that use it.

_REG_EXPEVT was never used by any code in the tree. Reporting EXPEVT
makes sense only for signals and in that case we pass it to userland
in ksi_trap already which is official MI way to get this (MD) information.

Old binaries running on new kernels will now have their GBR set from
new mcontext, but that's ok too, as GBR was not properly supported by
old kernels (not saved in trapframe), so old binaries couldn't have
possibly used it anyway.

branches: 1.5.20; 1.5.22;
In PIC code call setcontext(2) via PLT to avoid text reloc in the
shared library.

branches: 1.4.6;
Cause SIGTRAP if NOTREACHED code is reached.

Not that the branch in SETC doens't have a delay slot, remove the
.empty comment as it's no longer pertinent.

PIC patch from Valeriy E. Ushakov applied.
Also, removed bogus delay slot flag from branch instruction.

First bits of SH3 support. Only _context_u.S is implemented (passess
cu[1-6] tests), the pthread_switch.S is stubbed out for now.

Code posted by Christian Groessler (cpg at aladdin dot de) to port-sh3.

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

branches: 1.8.6;
Recycle unused _REG_EXPEVT slot in mcontext for _REG_GBR.

As the size of mcontext is not changed, we avoid the hassle of
versioning all the calls that use it.

_REG_EXPEVT was never used by any code in the tree. Reporting EXPEVT
makes sense only for signals and in that case we pass it to userland
in ksi_trap already which is official MI way to get this (MD) information.

Old binaries running on new kernels will now have their GBR set from
new mcontext, but that's ok too, as GBR was not properly supported by
old kernels (not saved in trapframe), so old binaries couldn't have
possibly used it anyway.

branches: 1.7.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Add necessary symbols.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

The field "pt_sleepuc" doesn't exist in struct pthread_st.

First bits of SH3 support. Only _context_u.S is implemented (passess
cu[1-6] tests), the pthread_switch.S is stubbed out for now.

Code posted by Christian Groessler (cpg at aladdin dot de) to port-sh3.

make pthread__sp unsigned long.

branches: 1.7.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.6.4;
Fix typo in comment.

Pull down revision 1.3.6.1 by skrll@ (adapted to include reg::r_gbr).

struct mcontext != struct reg on sh3.

Fix PTHREAD_UCONTEXT_TO_REG / PTHREAD_REG_TO_UCONTEXT to deal with this.

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

branches: 1.3.4; 1.3.6; 1.3.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Add _INITCONTEXT_U_MD, we need to clean _REG_SR.

First bits of SH3 support. Only _context_u.S is implemented (passess
cu[1-6] tests), the pthread_switch.S is stubbed out for now.

Code posted by Christian Groessler (cpg at aladdin dot de) to port-sh3.

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

Use PLT for PIC calls to avoid text relocs in the shared library.

Adapt to new PIC macros that are now in <machine/asm.h>. Same binary
code is generated (still with text relocs, but eliminating them is the
next step).

branches: 1.5.2;
Fix typo in comment.

Do not move incoming paramters to callee-save registers when not
necessary. Saves about a dozen of instructions.

In pthread__locked_switch set self->pt_uc only when the context is inited.
In STACK_SWITCH subtract STACKSPACE, not add it (it's zero for now anyway).

First cut at pthreads MD code for sh3. Based on m68k version.
Regression tests still failing: sem, sigalarm.

First bits of SH3 support. Only _context_u.S is implemented (passess
cu[1-6] tests), the pthread_switch.S is stubbed out for now.

Code posted by Christian Groessler (cpg at aladdin dot de) to port-sh3.

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

branches: 1.3.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.32;
Merge the nathanw_sa branch.

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

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

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

Remove possible race condition in upcall recycling.

Remove PT_SLEEPUC, which somehow wasn't.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Include <sys/types.h>
Remove v9 register defs.

Merge the nathanw_sa branch.

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

make pthread__sp unsigned long.

branches: 1.7.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.6.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

branches: 1.4.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

All the ucontext <-> {fp}reg macros were slightly simplistic and untested
before (old gdb support on sparc did not use this).
Pointed out by Christian Limpach.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

A couple of fixes to make libpthread really shared, i.e. not have text re-
locations:

- Don't declare pthread__switch_away global
- Do the PIC dance for pthread__switch_return_point and
pthread__locked_switch. Ideally these (and other) symbols would
be hidden.

Thanks to uwe@, dyoung@ and elad@ for help.

XXX sh3 is still to be done.
XXX vax does strange things.

branches: 1.7.6;
Remove possible race condition in upcall recycling.

Use delay slot in STACK_SWITCH (catching up with sparc64 changes).

Use correct context to switch to.
From sparc64 fix by Andrey Petrov <petrov@netbsd>.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Catch up with Nathan's changes to other archs.

Merge the nathanw_sa branch.

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

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

branches: 1.3.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.32;
Merge the nathanw_sa branch.

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

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

branches: 1.8.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.7.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

Add missing include and copyright.

Merge the nathanw_sa branch.

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

make pthread__sp unsigned long.

branches: 1.6.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.5.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.4.20;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Fix ucontext conversion macros. From Christian Limpach.

Merge the nathanw_sa branch.

branches: 1.1.2;
file pthread_md.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.

A couple of fixes to make libpthread really shared, i.e. not have text re-
locations:

- Don't declare pthread__switch_away global
- Do the PIC dance for pthread__switch_return_point and
pthread__locked_switch. Ideally these (and other) symbols would
be hidden.

Thanks to uwe@, dyoung@ and elad@ for help.

XXX sh3 is still to be done.
XXX vax does strange things.

branches: 1.8.6;
Remove possible race condition in upcall recycling.

typo in previous commit, annul isn't good there,
noticed by Eduardo Horvath.

Use branch-slot in STACK_SWITCH, suggested by Martin Husemann.

Use correct context to switch to.

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Catch up with changes Nathan did to other archs.

Merge the nathanw_sa branch.

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

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

branches: 1.4.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.3.20;
Rework and cleanup. Don't use REI, fake a call frame instead.

Set _UC_USER in ucontext. Call setcontext if _UC_USER is not in
_setcontext_u

VAX versions of _getcontext_u/_setcontext_u/_swapcontext_u

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

branches: 1.8.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.7.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Add STACKSPACE

pt_sleepuc was removed a long time ago.

Remove possible race condition in upcall recycling.

Add PT_TRAPUC and pthread__switch_return_point.
Not tested (as most of the pthread stuff on vax) but at least
allow the system to compile.

Adapt to structure name changes.

Add of pthread support for VAX.

make pthread__sp unsigned long.

branches: 1.7.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.6.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

branches: 1.4.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Define FPREG macros even though we don't have any.

Need to define STACKSIZE (even though we don't need any).

Add of pthread support for VAX.

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

Merge updates to algorithms from i386 switch code.

Remove possible race condition in upcall recycling.

Add PT_TRAPUC and pthread__switch_return_point.
Not tested (as most of the pthread stuff on vax) but at least
allow the system to compile.

Update to new world order. **not tested** But will allow builds to
finish.

Add myself to the authors list.

Add of pthread support for VAX.

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

branches: 1.11.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.10.4;
- Remove libpthread's atomic ops.
- Remove the old spinlock-based mutex and rwlock implementations.
- Use the atomic ops from libc.

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.

Fix a dodgy bit of assembly.

- 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.

Add: pthread__atomic_cas_ptr, pthread__atomic_swap_ptr, pthread__membar_full
This is a stopgap until the thorpej-atomic branch is complete.

branches: 1.4.14; 1.4.18;
Fix thread context switching to take the stack ABI into account.
From Wolfgang Solfrank.

branches: 1.3.2;
Restore %rax correctly during a full context restore (oops).

netbsd.org->NetBSD.org

libpthread support for x86_64.

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

branches: 1.6.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.12;
Remove the PTHREAD_SA option. If M:N threads is reimplemented it's
better off done with a seperate library.

Remove possible race condition in upcall recycling.

Adapt to structure name changes.

Remove PT_SLEEPUC and add PT_TRAPUC.

libpthread support for x86_64.

libpthread: New pthread__smt_wait to put CPU in low power for spin.

This is now distinct from pthread__smt_pause, which is for spin lock
backoff with no paired wakeup.

On Arm, there is a single-bit event register per CPU, and there are two
instructions to manage it:

- wfe, wait for event -- if event register is clear, enter low power
mode and wait until event register is set; then exit low power mode
and clear event register

- sev, signal event -- sets event register on all CPUs (other
circumstances like interrupts also set the event register and cause
wfe to wake)

These can be used to reduce the power consumption of spinning for a
lock, but only if they are actually paired -- if there's no sev, wfe
might hang indefinitely. Currently only pthread_spin(3) actually
pairs them; the other lock primitives (internal lock, mutex, rwlock)
do not -- they have spin lock backoff loops, but no corresponding
wakeup to cancel a wfe.

It may be worthwhile to teach the other lock primitives to pair
wfe/sev, but that requires some performance measurement to verify
it's actually worthwhile. So for now, we just make sure not to use
wfe when there's no sev, and keep everything else the same -- this
should fix severe performance degredation in libpthread on Arm
without hurting anything else.

No change in the generated code on amd64 and i386. No change in the
generated code for pthread_spin.c on arm and aarch64 -- changes only
the generated code for pthread_lock.c, pthread_mutex.c, and
pthread_rwlock.c, as intended.

PR port-arm/57437

XXX pullup-10

branches: 1.12.36; 1.12.46; 1.12.54;
make pthread__sp unsigned long.

branches: 1.11.2;
Remove unused code that's confusing when using cscope/opengrok.

branches: 1.10.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.9.2;
Cheat and add inlines for _atomic_cas_ptr() to work around gcc emitting
unneeded PIC stuff in mutex_lock() and mutex_unlock(), when a thread
register is used.

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

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.

Resurrect the function pointers for lock operations and allow each
architecture to provide asm versions of the RAS operations.

We do this because relying on the compiler to get the RAS right is not
sensible. (It gets alpha wrong and hppa is suboptimal)

Provide asm RAS ops for hppa.

(A slightly different version) reviewed by Andrew Doran.

Add: pthread__atomic_cas_ptr, pthread__atomic_swap_ptr, pthread__membar_full
This is a stopgap until the thorpej-atomic branch is complete.

branches: 1.4.8; 1.4.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

netbsd.org->NetBSD.org

Catch up with changed struct reg (same as gregs in mcontext now).

libpthread support for x86_64.

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

I was a bit over zealous with my last change so revert the
locked_return_point change.

Loading the instruction at locked_return_point as a return value didn't
work so well.

Thanks to uwe, cube, and dsl.

A couple of fixes to make libpthread really shared, i.e. not have text re-
locations:

- Don't declare pthread__switch_away global
- Do the PIC dance for pthread__switch_return_point and
pthread__locked_switch. Ideally these (and other) symbols would
be hidden.

Thanks to uwe@, dyoung@ and elad@ for help.

XXX sh3 is still to be done.
XXX vax does strange things.

branches: 1.10.2;
s/the the/the/ (only in sources that aren't regularly imported from
elsewhere).

Make register usage more consistent, also in comparison with the i386
version, for easier maintenance.

Correct and simplify computing the return point; just use PC-relative
addressing.

Make sure the stack stays aligned.

Remove possible race condition in upcall recycling.

netbsd.org->NetBSD.org

Adapt to pt_trapuc: change STACK_SWITCH to check for a value in pt_trapuc
and use it preferentially to a value in pt_uc, clearing it once on the new
stack. Move stores into pt_uc back to before the stack switch; storing
after the stack switch opened a one-instruction race condition where an upcall
that had just started a chain could be preempted again, and would bomb when
restarted due to its pt_uc not yet having been updated. Now that pt_trapuc
is what the upcall code writes to, it is safe to store to pt_uc before
switching stacks.

Remove obsolete pt_sleepuc code.

Apply Nathan's switch-away fix.

Continue at the plain switch return point in pthread__switch, not the
locked one, in the !PIC case. From Tor Egge via Havard Eidnes.

libpthread support for x86_64.