So far pthread::postcreate() only sets the thread priority at all, only
if the inherit-scheduler attribute is PTHREAD_EXPLICIT_SCHED. This
completely ignores the PTHREAD_INHERIT_SCHED case, since in contrast
to POSIX, a thread does not inherit its priority from the creating
thread, but always starts with THREAD_PRIORITY_NORMAL.
pthread_getschedparam() only returns what's stored in the thread attributes,
not the actual thread priority.
This patch fixes both problems.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
So far the scheduler priority handling is not POSIX compatible.
The priorities use a range of -14 up to +15, which means it's not clear
if the POSIX-required return value of -1 in case of an error is *really*
an error or just the valid priority value -1. Even more confusing, -14
is the *max* value and 15 is the *min* value. Last but not least this
range doesn't match the POSIX requirement of at least 32 priority values.
This patch cleans up scheduler priority handling and moves the valid
priority range to 1 (min) - 32 (max). It also adds a function
sched_get_thread_priority() which will help to make thread priority
more POSIX-like.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
For all pthread init functions, POSIX says
Results are undefined if pthread_FOO_init() is called specifying an
already initialized pthread_FOO object.
So far our pthread init functions tested the incoming object if it's
already an initialized object and, if so, returned EBUSY. That's ok
*iff* the object was already initialized. However, as the example in
https://cygwin.com/ml/cygwin/2016-04/msg00473.html shows, an uninitialized
pthread object could also accidentally look like an initialized object
and then returning EBUSY is not ok.
Consequentially, all those tests are dangerous. Per POSIX, an application
has to know what its doing when calling any of the pthread init functions
anyway, and re-initializing the object is just as well as undefined
behaviour as is returning EBUSY on already initialized objects.
* thread.cc (pthread_attr_init): Drop check for already initialized
object.
(pthread_condattr_init): Ditto.
(pthread_rwlockattr_init): Ditto.
(pthread_mutexattr_init): Ditto.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Hi.
I have noticed that Cygwin's spinlock goes into heavy sleeping code
for each spin. It seems it would be a good idea to actually try to
spin a bit first. There is this 'pause' instruction which let's the
CPU make such busy loops be less busy. Here is a patch to do this.
--
VH
* signal.cc (sigwait): Fix return value to reflect errno in case of
error according to POSIX. Never return EINTR.
* thread.cc (pthread_kill): Return errno if sig_send failed.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* thread.cc (pthread_getattr_np): Fix memory leak, remove usage of
malloc for small local buffer.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
* miscfuncs.cc (struct pthread_wrapper_arg): Add member guardsize.
(pthread_wrapper): Set thread stack guarantee according to guardsize.
Tweak assembler code so that $rax/$eax is not required by GCC to
prepare the wrapper_arg value.
(CygwinCreateThread): Fix deadzone handling. Drop setting a "POSIX"
guardpage (aka page w/ PAGE_NOACCESS). Always use Windows guard
pages instead. On post-XP systems (providing SetThreadStackGuarantee)
always set up stack Windows like with reserved/commited areas and
movable guard pages. Only on XP set up stack fully commited if the
guardpage size is not the default system guardpage size.
Fill out pthread_wrapper_arg::guardsize. Improve comments.
* resource.cc: Implement RSTACK_LIMIT Linux-like.
(DEFAULT_STACKSIZE): New macro.
(DEFAULT_STACKGUARD): Ditto.
(rlimit_stack_guard): New muto.
(rlimit_stack): New global variable holding current RSTACK_LIMIT values.
(__set_rlimit_stack): Set rlimit_stack under lock.
(__get_rlimit_stack): Initialize rlimit_stack from executable header
and return rlimit_stack values under lock.
(get_rlimit_stack): Filtering function to return useful default
stacksize from rlimit_stack.rlim_cur value.
(getrlimit): Call __get_rlimit_stack in RLIMIT_STACK case.
(setrlimit): Call __set_rlimit_stack in RLIMIT_STACK case.
* thread.cc (pthread::create): Fetch default stacksize calling
get_rlimit_stack.
(pthread_attr::pthread_attr): Fetch default guardsize calling
wincap.def_guard_page_size.
(pthread_attr_getstacksize): Fetch default stacksize calling
get_rlimit_stack.
* thread.h (PTHREAD_DEFAULT_STACKSIZE): Remove.
(PTHREAD_DEFAULT_GUARDSIZE): Remove.
(get_rlimit_stack): Declare.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
to explain the meaning of the possible values.
* cygwait.cc (is_cw_sig_restart): Define.
(is_cw_sig_handle): Check for cw_sig_restart as well.
(cygwait): Restart always if cw_sig_restart is set.
* thread.cc (pthread::join): Call cygwait with cw_sig_restart flag
to avoid having to handle signals at all.
having a new type threadlist_t *. Convert commented out code into an
#if 0. Create thread mutex. Explain why.
(init_cygheap::remove_tls): Drop timeout value. Always wait infinitely
for tls_sentry. Return mutex HANDLE of just deleted threadlist entry.
(init_cygheap::find_tls): New implementation taking tls pointer as
search parameter. Return threadlist_t *.
(init_cygheap::find_tls): Return threadlist_t *. Define ix as auto
variable. Drop exception handling since crash must be made impossible
due to correct synchronization. Return with locked mutex.
* cygheap.h (struct threadlist_t): Define.
(struct init_cygheap): Convert threadlist to threadlist_t type.
(init_cygheap::remove_tls): Align declaration to above change.
(init_cygheap::find_tls): Ditto.
(init_cygheap::unlock_tls): Define.
* cygtls.cc (_cygtls::remove): Unlock and close mutex when finishing.
* exceptions.cc (sigpacket::process): Lock _cygtls area of thread before
accessing it.
* fhandler_termios.cc (fhandler_termios::bg_check): Ditto.
* sigproc.cc (sig_send): Ditto.
* thread.cc (pthread::exit): Ditto. Add comment.
(pthread::cancel): Ditto.
* cygtls.cc (well_known_dlls): Rephrase comment.
(bloda_detect): New function.
(_cygtls::call2): Call init_thread and bloda_detect for non-pthread
threads only.
(_cygtls::remove): Move remove_tls and remove_wq calls up to run first.
* miscfuncs.cc (struct pthread_wrapper_arg): Rename from struct
thread_wrapper_arg.
(pthread_wrapper): Rename from thread_wrapper and drop "static". Fix
comment. Drop call to _cygtls::remove. Call api_fatal rather than
ExitThread. Explain why.
* miscfuncs.h (pthread_wrapper): Declare pthread_wrapper.
* thread.cc (pthread::exit): Add a FIXME comment. Call _cygtls::remove
before calling ExitThread.
handler.
* cygtls.cc (_cygtls::remove): Accommodate the fact that pathbufs
has been moved from _local_storage to _cygtls.
* cygtls.h (class tls_pathbuf): Add comment to hint to gendef usage
of counters. Change type of counters to uint32_t for clarity.
Remove _cygtls as friend class.
(struct _local_storage): Move pathbufs from here...
(struct _cygtls): ...to here, allowing to access it from _sigbe.
(class san): Only define on 32 bit. Remove errno, _c_cnt and _w_cnt
members.
(san::setup): Drop parameter. Don't initialize removed members.
(san::leave): Don't set removed members.
(class myfault): Only define on 32 bit.
(myfault::faulted): Only keep implementation not taking any parameter.
Drop argument in call to sebastian.setup.
(__try/__leave/__except/__endtry): Implement to support real SEH. For
now stick to SJLJ on 32 bit.
* dcrt0.cc (dll_crt0_0): Drop 64 bit call to
exception::install_myfault_handler.
* exception.h (exception_handler): Define with EXCEPTION_DISPOSITION
as return type.
(PDISPATCHER_CONTEXT): Define as void * on 32 bit. Define as pointer
to _DISPATCHER_CONTEXT on 64 bit.
(class exception): Define separately for 32 and 64 bit.
(exception::myfault): Add handler for myfault SEH handling on 64 bit.
(exception::exception): Fix mangled method name to account for change
in type of last parameter.
(exception::install_myfault_handler): Remove.
* exceptions.cc (exception::myfault_handle): Remove.
(exception::myfault): New SEH handler for 64 bit.
* gendef (_sigbe): Set tls_pathbuf counters to 0 explicitely when
returning to the caller.
* ntdll.h: Move a comment to a better place.
(struct _SCOPE_TABLE): Define on 64 bit.
* thread.cc (verifyable_object_isvalid): Remove gcc 4.7 workaround.
* tls_pbuf.cc (tls_pbuf): Fix to accommodate new place of pathbufs.
(tls_pathbuf::destroy): Change type of loop variables to uint32_t.
* tls_pbuf.h (class tmp_pathbuf): Change type of buffer counters to
uint32_t. Accommodate new place of pathbufs.
* tlsoffsets.h: Regenerate.
* tlsoffsets64.h: Regenerate.
Add comment.
(pthread_attr_setstack): Store upper bound address in stackaddr.
Explain why.
(pthread_attr_getstack): Handle stackaddr as upper bound address.
Add comment.
(pthread_attr_setstackaddr): Add comment.
(pthread_attr_getstackaddr): Add comment.
(pthread_attr_getstacksize): Return default stacksize if stacksize has
not been set by the application, just as on Linux. Add comment.
(pthread_getattr_np): Store upper bound address in stackaddr. Explain
why.
* include/pthread.h: Remove outdated comment.
(pthread_attr_getstackaddr): Mark as deprecated, as on Linux.
(pthread_attr_setstackaddr): Ditto.
to PTHREAD_MUTEX_NORMAL.
(pthread_mutex::unlock): Return EPERM if the mutex has no owner and
the mutex type is PTHREAD_MUTEX_ERRORCHECK, as on Linux.
(pthread_mutexattr::pthread_mutexattr): Ditto.
(pthread_mutex_unlock): Do not fail if mutex is a normal mutex
initializer.
* include/pthread.h (PTHREAD_MUTEX_INITIALIZER): Redefine as
PTHREAD_NORMAL_MUTEX_INITIALIZER_NP.
if attr.stacksize is 0.
(pthread_attr::pthread_attr): Initialize stacksize to 0 to align more
closely to Linux.
(pthread_attr_getstack): Fix incorrect stackaddr computation. Return
stackaddr just like pthread_attr_getstackaddr. Remove slightly off
comment.
(pthread_attr_getstackaddr): Remove slightly off comment.
(pthread_getattr_np): Return stackaddr and stacksize based on the full
allocated stackarea.
(enum _SEMAPHORE_INFORMATION_CLASS): Define.
(NtQuerySemaphore): Declare.
* thread.h (class semaphore): Add member startvalue.
(semaphore::fixup_before_fork): New inline method.
(semaphore::_fixup_before_fork): Declare.
* thread.cc (MTinterface::fixup_before_fork): Additionally call
semaphore::fixup_before_fork.
(semaphore::semaphore): Set currentvalue to -1. Set startvalue to
incoming initializer value.
(semaphore::_getvalue): Just query semaphore using NtQuerySemaphore
rather then using WFSO/Release.
(semaphore::_post): Drop setting currentvalue. It's not thread-safe.
(semaphore::_trywait): Ditto.
(semaphore::_timedwait): Ditto.
(semaphore::_wait): Ditto.
(semaphore::_fixup_before_fork): New method, setting currentvalue from
actual windows semaphore right before fork.
(semaphore::_fixup_after_fork): Drop kludge from 2013-07-10. Drop
FIXME comment.
* thread.cc (semaphore::_fixup_after_fork): Report on potential problem
parameter. Make sure that currentvalue is never zero.
(semaphore::init): Make cosmetic change.
warnings between regparm definitions and declarations.
* smallprint.cc (__small_vswprintf): Conditionalize declaration and
setting of l_opt for only x86_64.
* spawn.cc (child_info_spawn::worker): Remove unused 'pid' variable.
* thread.cc (verifyable_object_isvalid): Temporarily define as
non-inline with gcc 4.7+, regardless of target.
* dcrt0.cc (parent_sigmask): Delete.
(dll_crt0_1): Use spawn_info->moreinfo->sigmask rather than saved parent signal
mask.
* thread.cc (pthread::thread_init_wrapper): Add comment stressing the
importance of maintaining ordering of statements.
* cygheap.cc (init_cygheap::find_tls): Rename threadlist_ix -> ix. Only take
one pass through thread list, looking for eligible threads to signal. Set a
new param indicating that function has found a sigwait* mask.
* cygheap.h (init_cygheap::find_tls): Reflect new parameter.
* dcrt0.cc (parent_sigmask): New variable.
(child_info_spawn::handle_spawn): Save parent's signal mask here.
(dll_crt0_1): Restore parent's signal mask to tls sigmask as appropriate. Call
sig_dispatch_pending to flush signal queue when we can finally do something
with signals.
* exceptions.cc (sigpacket::process): Avoid attempting to handle signals if we
haven't finished initializing. Rely on the fact that find_tls will do mask
checking and don't do it again. Delete ill-named 'dummy' variable.
* sigproc.cc (cygheap_exec_info::alloc): Save calling thread's signal mask in
new sigmask field.
(wait_sig): Try to debug when WFSO fails and DEBUGGING is defined.
* thread.cc (pthread::set_tls_self_pointer): Make this a true automatic method
rather than inexplicably relying on a thread parameter.
(pthread::thread_init_wrapper): Accommodate set_tls_self_pointer change to
non-static. Initialize sigmask before setting tid or suffer signal races.
* ehread.h (pthread::set_tls_self_pointer): Make non-static, delete parameter.
that we're looking for the current thread.
(pthread_rwlock::tryrdlock): Eliminate self variable. Accommodate change in
lookup_reader().
(pthread_rwlock::unlock): Ditto.
(pthread_rwlock::rdlock): Ditto. Move add_reader call after writer tests to
more closely mimic old behavior.
(pthread_rwlock::wrlock): Accommodate change in lookup_reader().
* thread.h ((pthread_rwlock::lookup_reader): Eliminate argument.
pointer to allocated RWLOCK_READER structure.
(pthread_rwlock::rdlock): Reorganize to reflect new add_reader functionality.
(pthread_rwlock::tryrdlock): Ditto. Remove unneeded call to lookup_reader().
* thread.h (pthread_rwlock::RWLOCK_READER::RWLOCK_READER): New constructor.
(pthread_rwlock::add_reader): Reflect new functionality.
(select_info): Ditto.
(select_record::dump_select_record): Declare new debugging-only function.
(select_info): Zero all fields.
(select_pipe_info): Ditto.
(select_socket_info): Ditto.
(select_serial_info): Ditto.
(select_mailslot_info): Ditto.
* select.cc (select_record::dump_select_record): Define new debugging-only
function.
(select_stuff::test_and_set): Call dump_select_record when debugging.
* thread.cc (pthread_mutex::unlock): Revert setting of tid to NULL since, in
this context, it is a number, not a pointer.
(pthread_spinlock::unlock): Ditto.
* _cygtls.cc (_cygtls::call2): Use new exit_thread function in place of
ExitThread.
* miscfuncs.cc (thread_wrapper): Ditto.
* thread.cc (pthread::exit): Ditto.
(pthread_mutex::unlock): Set tid to NULL rather than 0.
(pthread_spinlock::unlock): Ditto.
* pinfo.cc (commune_process): Actually call lock_process constructor.
* sigproc.cc (exit_thread): New function.
(wait_sig): Handle __SIGTHREADEXIT case. Don't just block rather than
returning from this function.
* sigproc.h (__SIGTHREADEXIT): New enum.
(exit_thread): Declare.
* sync.cc (muto::release): Accept a tls command-line argument.
* sync.h (muto::release): Accept a tls command-line parameter. Default to
&_my_tls.
* cygserver_ipc.h (ipc_set_proc_info): Use _cygtls::ipc_set_proc_info to set
per-thread signal arrived value.
* cygthread.cc (cygthread::detach): Use per-thread signal_arrived via
set_thread_waiting.
* fork.cc (_cygtls::fixup_after_fork): Clear signal_arrived.
(_cygtls::remove): Close any signal_arrived handle when thread exists.
(_cygtls::find_tls): Remove unneeded function.
* cygtls.h: Update copyright.
(class _cygtls): Reorganize to help avoid rebuilding newlib when structure
changes.
(_cygtls::event): Delete.
(_cygtls::threadkill): Ditto.
(_cygtls::signal_waiting): Declare new bool.
(_cygtls::find_tls): Delete declaration.
(_cygtls::set_threadkill): Ditto.
(_cygtls::reset_threadkill): Ditto.
(_cygtls::set_signal_arrived): Declare new function.
(class set_thread_waiting): Declare new class.
* cygwait.cc (cw_nowait_storage): Define.
(cygwait): Set per-thread signal_arrived via set_thread_waiting. Don't
special-case _main_tls.
* cygwait.h (cw_nowait): Define.
(cw_infinite): Ditto.
(cygwait): Redefine pathological wait-only case.
* dcrt0.cc (dll_crt0_0): Remove call to now-defunct events_init().
(dll_crt0_1): Remove call to now-defunct create_signal_arrived().
* exceptions.cc: Reflect set_signal_mask() argument reordering throughout.
Remove signal mask synchronization throughout.
(events_init): Delete definition.
(mask_sync): Delete now-unneeded mask synchronization.
(set_signal_mask): Reverse order of arguments to "standard" to, from layout.
Rename "newmask" argument to "setmask". Remove debugging.
(sig_handle_tty_stop): Use cancelable_wait rather than WFMO.
(_cygtls::interrupt_setup): Don't treat "threadkill" events specially.
Conditionally set signal_arrived depending on whether the thread has created it
or not.
(sigpacket::process): Reorganize to reflect thread-specific sending of signals
which is more in line with the way it was actually supposed to work.
* fhandler_socket.cc (get_inet_addr): Use cancelable_wait rather than
IsEventSignalled to avoid potential race.
(fhandler_socket::wait_for_events): Set signal_arrived event using
set_thread_waiting().
(fhandler_socket::close): Use cygwait for the case of just waiting 10 ms for a
signal.
* fhandler_tape.cc (fhandler_dev_tape::_lock): Use cancelable_wait rather than
WFMO. Redo switch/case tests accordingly.
* fhandler_termios.cc (fhandler_termios::bg_check): Use cygwait for case of
just waiting 0 ms for a potential signal.
* fhandler_tty.cc (fhandler_pty_master::process_slave_output): Use
cancelable_wait rather than WFSO.
* fhandler_windows.cc (fhandler_windows::read): Set per-thread signal_arrived
via set_thread_waiting().
* flock.cc (lf_setlock): Ditto.
* select.cc (pselect): Ditto. Set per-thread signal_arrived using
set_thread_waiting().
* gendef: Don't special case handling of _cygtls::sig for threads.
* gentls_offsets: Use #pragma once in tlsoffsets.h.
* ntdll.h: Use #pragma once.
* poll.cc: Reflect set_signal_mask() argument reordering.
* posix_ipc.cc (ipc_mutex_lock): Use cancelable_wait rather than WFMO.
(ipc_cond_timedwait): Set perl-thread signal arrived using
set_thread_waiting().
* security.h: Use #pragma once.
* signal.cc (abort): Reflect set_signal_mask() argument reordering.
(clock_nanosleep): Ditto. Change call to cancelable_wait to properly specify
handling of cancel and interrupt.
(sigwaitinfo): Remove handling of per-thread event in favor of per-thread
signal_arrived. Use cancelable_wait rather than WFSO.
* sigproc.cc (signal_arrived): Delete definition.
(create_signal_arrived): Ditto.
* sigproc.h (signal_arrived): Delete declaration.
(set_signal_mask): Avoid defining as a "C" function. Don't conditionally
declare.
(create_signal_arrived): Delete declaration.
* syscalls.cc (rename): Use cygwait() rather than WFSO.
* thread.h (fast_mutex::lock): Use cw_infinite rather than LARGE_NULL.
* wait.cc (wait4): Ditto.
* thread.cc (pthread_mutex::lock): Ditto.
(pthread::join): Ditto.
(semaphore::_wait): Ditto.
(pthread_kill): Remove set_threadkill() accommodation.
* tlsoffsets.h: Regenerate.
(cancelable_wait): Define variant which accepts DWORD time argument.
(cygwait): Use cancelable_wait with DWORD argument.
(cygwait): Use cancelable_wait with DWORD argument and cw_sig_eintr for
timeout-only case.
* exceptions.cc (handle_sigsuspend): Use LARGE_NULL as second argument to
distinguish between cancelable_wait variants.
* thread.cc (pthread_mutex::lock): Ditto.
(pthread::join): Ditto.
(semaphore::_timedwait): Ditto.
* thread.h (fast_mutex::lock): Ditto.
* wait.cc (wait4): Ditto.
newlib doesn't install cleanup handlers. Explain the problem.
(class __cygwin_lock_handler): New class.
(__cygwin_lock_cleanup): New function.
(__cygwin_lock_lock): Push __cygwin_lock_cleanup thread cleanup
handler.
(__cygwin_lock_trylock): Ditto.
(__cygwin_lock_unlock): Pop thread cleanup handler.
(pthread::pop_cleanup_handler): Temporarily allow cleanup function to
destroy cleanup handler so we can pop in another function than we
pushed in.
to PTHREAD_CANCEL_DISABLE from here...
(pthread::pop_all_cleanup_handlers): ...to here, otherwise any explicit
call to pthread_cleanup_pop disables cancellation for this thread.