On ia64, where there's no libc_r at all, libkse is now the default
thread library by virtue of these links.
The reasons for this change are:
1. libkse is slated to become the default thread library anyway,
2. active development and maintenance is only present for libkse,
3. GNOME and KDE, both in the process of being supported on ia64,
work better with KSE; even on ia64.
can clear the pointer to mutex, not the thread doing mutex
handoff. Because _mutex_lock_backout does not hold scheduler
lock while testing THR_FLAGS_IN_SYNCQ and then reading mutex
pointer, it is possible mutex owner begin to unlock and
handoff the mutex to the current thread, and mutex pointer
will be cleared to NULL before current thread reading it, so
current thread will end up with deferencing a NULL pointer,
Fix the race by making mutex waiters to clear their mutex pointers.
While I am here, also save inherited priority in mutex for
PTHREAD_PRIO_INERIT mutex in mutex_trylock_common just like what
we did in mutex_lock_common.
for interrupted field.
Also in _thr_sig_handler, retrieve current signal mask from kernel not
from ucp, the later is pre-unioned mask, not current signal mask.
pthread_md.h. This commit only moves the definition; it does not
change it for any of the platforms. This more easily allows 64-bit
architectures (in particular) to pick a slightly larger stack size.
THR_SETCONTEXT as PANIC(). The THR_SETCONTEXT macro is currently not
used, which means that the definition we had could be wrong, overly
pessimistic or unknowingly right. I don't like the odds...
The new _ia64_break_setcontext() and corresponding kernel fixes make
KSE mostly usable. There's still a case where we don't properly
restore a context and end up with a NaT consumption fault (typically
an indication for not handling NaT collection points correctly),
but at least now mutex_d works...
state for amd64 was twice as large as necessary. Peter
recently fixed this, so the comment no longer applies.
Also, since the size of struct mcontext changed, adjust
the threads library version of get&set context to match.
FYI, any change layout/size change to any arch's struct
mcontext will likely need some minor changes in libpthread.
to avoid potential memory leak, also fix a bug in pthread_create, contention
scope should be inherited when PTHREAD_INHERIT_SCHED is set, and also check
right field for PTHREAD_INHERIT_SCHED, scheduling inherit flag is in sched_inherit.
2. Execute hooks registered by atexit() on thread stack but not on scheduler
stack.
3. Simplify some code in _kse_single_thread by calling xxx_destroy functions.
Reviewed by: deischen
should be a value past to pthread_attr_setguardsize, not a rounded up value.
Also fix a stack size matching bug in thr_stack.c, now stack matching code
uses number of pages but not bytes length to match stack size, so for example,
size 512 bytes and size 513 bytes should both match 1 page stack size.
Reviewed by: deischen
a shared library or any other dyanmic allocated data block, once
pthread_once_t is initialized, a mutex is allocated, if we unload the
shared library or free those data block, then there is no way to deallocate
the mutex, result is memory leak.
To fix this problem, we don't use mutex field in pthread_once_t, instead,
we use its state field and an internal mutex and conditional variable in
libkse to do any synchronization, we introduce a third state IN_PROGRESS to
wait if another thread is already in invoking init_routine().
Also while I am here, make pthread_once() conformed to pthread cancellation
point specification.
Reviewed by: deischen
instead of long types for low-level locks.
Add prototypes for some internal libc functions that are
wrapped by the library as cancellation points.
Add memory barriers to alpha atomic swap functions (submitted
by davidxu).
Requested by: bde
critical region, we wrap some syscalls for thread cancellation point, and
when syscalls returns, we call _thr_leave_cancellation_point, at the time
if a signal comes in, it would be buffered, and when the thread leaves
_thr_leave_cancellation_point, buffered signals will be processed, to avoid
messing up normal syscall errno, we should save and restore errno around
signal handling code.
yet, so we can protect some locking code from being interrupted by signal
handling. When KSE mode is turned on, reset the thread flag to scope process
except we are running in 1:1 mode which we needn't turn it off.
Also remove some unused member variables in structure kse.
Tested by: deischen
have execute permissions. Run "perl verify" instead. Replace all
occurences of the hardcoding of ./verify with $(VERIFY) to allow
it to be overridden as well.
otherwise masks all signals until fork() returns, in child process,
we reset library state before restoring signal masks until we reach
a safe to point.
Reviewed by: deischen
happens, the context of the interrupted thread is exported to
userland. Unlike most contexts, it will be an async context and
we cannot easily use our existing functions to set such a
context.
To avoid a lot of complexity that may possibly interfere with
the common case, we simply let the kernel deal with it. However,
we don't use the EPC based syscall path to invoke setcontext(2).
No, we use the break-based syscall path. That way the trapframe
will be compatible with the context we're trying to restore and
we save the kernel a lot of trouble. The kind of trouble we did
not want to go though ourselves...
However, we also need to set the threads mailbox and there's no
syscall to help us out. To avoid creating a new syscall, we use
the context itself to pass the information to the kernel so that
the kernel can update the mailbox. This involves setting a flag
(_MC_FLAGS_KSE_SET_MBOX) and setting ifa (the address) and isr
(the value).
TCB. We know that the thread pointer points to &tcb->tcb_tp, so all
we have to do is subtract offsetof(struct tcb, tcb_tp) from the
thread pointer to get to the TCB. Any reasonably smart compiler will
translate accesses to fields in the TCB as negative offsets from TP.
In _tcb_set() make sure the fake TCB gets a pointer to the current
KCB, just like any other TCB. This fixes a NULL-pointer dereference
in _thr_ref_add() when it tried to get the current KSE.
makecontext(). We only supply 3, not 4. This is mostly harmless,
except that on ia64 the garbage can include NaT bits, resulting
in NaT consumption faults.
that the TLS is 16-byte aligned, as well as guarantee that the thread
pointer is 16-byte aligned as it points to struct ia64_tp. Likewise,
struct tcb and struct ksd are also guaranteed to be 16-byte aligned
(if they weren't already).
archs that can (or are required to) have per-thread registers.
Tested on i386, amd64; marcel is testing on ia64 and will
have some follow-up commits.
Reviewed by: davidxu
context functions. We don't need to enter the kernel anymore. The
contexts are compatible (ie a context created by getcontext() can
be restored by _ia64_restore_context()).
While here, make the use of THR_ALIGNBYTES and THR_ALIGN a no-op.
They are going to be removed anyway.
We write 1 for r8 in the context so that _ia64_restore_context()
will return with a non-zero value. _ia64_save_context() always
return 0.
o In _ia64_restore_context(), don't restore the thread pointer. It
is not normally part of the context. Also, restore the return
registers. We get called for contexts created by getcontext(),
which means we have to restore all the syscall return values.
the userland version of [gs]etcontext to switch between a thread
and the UTS scheduler (and back again). This also fixes a bug
in i386 _thr_setcontext() which wasn't properly restoring the
context.
Reviewed by: davidxu
This eliminates ping-ponging of locks, where the idle KSE wakes
up only to find the lock it needs is being held. This gives
little or no gain to M:N mode but greatly speeds up 1:1 mode.
Reviewed & Tested by: davidxu
handed-off/signaled to a higher priority thread. Note that when
there are idle KSEs that could run the higher priority thread,
we still add the preemption point because it seems to take the
kernel a while to schedule an idle KSE. The drawbacks are that
threads will be swapped more often between CPUs (KSEs) and
that there will be an extra userland context switch (the idle
KSE is still woken and will probably resume the preempted
thread). We'll revisit this if and when idle CPU/KSE wakeup
times improve.
Inspired by: Petri Helenius <pete@he.iki.fi>
Reviewed by: davidxu
is system bound thread and when it is blocked, no upcall is generated.
o Add ability to libkse to allow it run in pure 1:1 threading mode,
defining SYSTEM_SCOPE_ONLY in Makefile can turn on this option.
o Eliminate code for installing dummy signal handler for sigwait call.
o Add hash table to find thread.
Reviewed by: deischen
its waitset, but if the signal is not masked by the thread, the signal
can interrupt the thread and signal action can be invoked by the thread,
sigwait should return with errno set to EINTR.
Also save and restore thread internal state(timeout and interrupted)
around signal handler invoking.
signals were changed in kernel, it will retrieve the pending set and
try to find a thread to dispatch the signal. The dispatching process
can be rolled back if the signal is no longer in kernel.
o Create two functions _thr_signal_init() and _thr_signal_deinit(),
all signal action settings are retrieved from kernel when threading
mode is turned on, after a fork(), child process will reset them to
user settings by calling _thr_signal_deinit(). when threading mode
is not turned on, all signal operations are direct past to kernel.
o When a thread generated a synchoronous signals and its context returned
from completed list, UTS will retrieve the signal from its mailbox and try
to deliver the signal to thread.
o Context signal mask is now only used when delivering signals, thread's
current signal mask is always the one in pthread structure.
o Remove have_signals field in pthread structure, replace it with
psf_valid in pthread_signal_frame. when psf_valid is true, in context
switch time, thread will backout itself from some mutex/condition
internal queues, then begin to process signals. when a thread is not
at blocked state and running, check_pending indicates there are signals
for the thread, after preempted and then resumed time, UTS will try to
deliver signals to the thread.
o At signal delivering time, not only pending signals in thread will be
scanned, process's pending signals will be scanned too.
o Change sigwait code a bit, remove field sigwait in pthread_wait_data,
replace it with oldsigmask in pthread structure, when a thread calls
sigwait(), its current signal mask is backuped to oldsigmask, and waitset
is copied to its signal mask and when the thread gets a signal in the
waitset range, its current signal mask is restored from oldsigmask,
these are done in atomic fashion.
o Two additional POSIX APIs are implemented, sigwaitinfo() and sigtimedwait().
o Signal code locking is better than previous, there is fewer race conditions.
o Temporary disable most of code in _kse_single_thread as it is not safe
after fork().
functions are derived from the swapctx() and restorectx() (resp)
from sys/ia64/ia64/context.s. The code is expected to be 99%
correct, but has not yet been tested.
Note that with these functions operating on mcontext_t, we also
created the foundation upon which we can implement getcontext(2)
and setcontext(2) replacements. It's not guaranteed that the use
of these syscalls and _ia64_{save|restore}_context() on the same
uicontext_t is actually going to work. Replacing the syscalls is
now trivially achieved.
This commit completes the ia64 port of libpthread itself (modulo
testing and bugfixes).
the register stack and memory stack and call the function given to it.
While here, provide empty, non-working, stubs for the context functions
(_ia64_save_context() and _ia64_restore_context()) so that anyone can at
least compile libkse from CVS sources. Real implementations will follow
soon.
minimize the amount and complexity of assembly code that needs to be
written. This way the core functionality is spread over 3 elementary
functions that don't have to do anything that can more easily and
more safely be done in C. As such, assembly code will only have to
know about the definition of mcontext_t.
The runtime cost of not having these functions being inlined is less
important than the cleanliness and maintainability of the code at
this stage of the implementation.
platforms the compiler warns about incompatible integer/pointer casts
and on ia64 this generally is bad news. We know that what we're doing
here is valid/correct, so suppress the warning. No functional change.
Sleeps better: marcel
by moving the definition of struct ksd to pthread_md.h and removing
the inclusion of ksd.h from thr_private.h (which has the definition
of struct kse and kse_critical_t). This allows ksd.h to have inline
functions that use struct kse and kse_critical_t and generally
yields a cleaner implementation at the cost of not having all ksd
related types/definitions in one header.
Implement the ksd functionality on ia64 by using inline functions
and permanently remove ksd.c from the ia64 specific makefile.
This change does not clean up the i386 specific version of ksd.h.
NOTE: The ksd code on ia64 abuses the tp register in the same way
as it is abused in libthr in that it is incompatible with the
runtime specification. This will be address when support for TLS
hits the tree.
_ksd_readandclear_tmbx to be function-like. That way we
can define them as inline functions or create prototypes
for them.
This change allows the ksd interface on ia64 to be fully
inlined.
the chance of getting the same thread id when allocating a
new thread is reduced. This won't work if the application
creates a new thread for every time a thread exits, but
we're still within the allowances of POSIX.
debugging is enabled so the symbol needs to be resolved before rtld
locking is enabled. I may not really know what I'm talking about,
but it works.
Submitted by: kan
path, making them suitable for direct use by the dynamic loader.
Register libpthread-specific locking API with rtld on startup.
This still has some rough edges with signals which should be
addresses later.
Approved by: re (scottl)
is called and the application is not threaded. This works around
a problem when an application that hasn't yet become threaded
tries to jump out of a signal handler.
Reported by: mbr
Approved by: re@ (rwatson)
low-level lock used by the libpthread implementation. In the
future, we'll eliminate spinlocks from libc but that will wait
until after 5.1-release.
Don't call an application signal handler if the handler is
the same as the library-installed handler. This seems to
be possible after a fork and is the cause of konsole hangs.
Approved by: re@ (jhb)
a lock is being waitied on.
Fix a races in join and cancellation.
When trying to wait on a CV and the library is not yet
threaded, make it threaded so that waiting actually works.
When trying to nanosleep() and we're not threaded, just
call the system call nanosleep instead of adding the thread
to the wait queue.
Clean up adding/removing new threads to the "all threads queue",
assigning them unique ids, and tracking how many active threads
there are. Do it all when the thread is added to the scheduling
queue instead of making pthread_create() know how to do it.
Fix a race where a thread could be marked for signal delivery
but it could be exited before we actually add the signal to it.
Other minor cleanups and bug fixes.
Submitted by: davidxu
Approved by: re@ (blanket for libpthread)
be external (initialize()!).
Remove cancellation points from _pthread_cond_wait and
_pthread_cond_timedwait (single underscore versions are
libc private functions). Point the weak reference(!) for
these functions to the versions with cancellation points.
Approved by: re@(blanket till 5/19)
Pointed out by: kan (cancellation point bug)
lock held (_thr_sched_switch_unlocked()) and use this to avoid
dropping the scheduler lock and having the scheduler retake the
same lock again.
Add a better way of detecting if a low-level lock is in use.
When switching out a thread due to blocking in the UTS, don't
switch to the KSE's scheduler stack only to switch back to
another thread. If possible switch to the new thread directly
from the old thread and avoid the overhead of the extra
context switch.
Check for pending signals on a thread when entering the scheduler
and add them to the threads signal frame. This includes some
other minor signal fixes.
Most of this was a joint effor between davidxu and myself.
Reviewed by: davidxu
Approved by: re@ (blanket for libpthread)
a thread lock).
Better protect access to thread state while searching for
threads to handle a signal.
Better protect access to process pending signals while processing
a thread in sigwait().
Submitted by: davidxu
KSEs when it's thread exits; allow the GC handler to do that.
o Make spinlock/spinlock critical regions.
The following were submitted by davidxu
o Alow thr_switch() to take a null mailbox argument.
o Better protect cancellation checks.
o Don't set KSE specific data when creating new KSEs; rely on the
first upcall of the KSE to set it.
o Add the ability to set the maximum concurrency level and do this
automatically. We should have a way to enable/disable this with
some sort of tunable because some applications may not want this
to be the default.
o Hold the scheduling lock across thread switch calls.
o If scheduling of a thread fails, make sure to remove it from the list
of active threads.
o Better protect accesses to a joining threads when the target thread is
exited and detached.
o Remove some macro definitions that are now provided by <sys/kse.h>.
o Don't leave the library in threaded mode if creation of the initial
KSE fails.
o Wakeup idle KSEs when there are threads ready to run.
o Maintain the number of threads active in the priority queue.
While I'm here, use the TAILQ_FOREACH macro instead of a more
manual method which was inherited from libc_r (so we could
remove elements from the list which isn't needed for libpthread).
Submitted by: Kazuaki Oda <kaakun@highway.ne.jp>
provided by Sergey A. Osokin <osa@freebsd.org.ru>.
In order to test this on a single CPU machine, you need to:
sysctl kern.threads.debug=1
sysctl kern.threads.virtual_cpu=2
