Setting a small stack size can fail due to having 3rd-party
libraries (e.g. libkqueue) loaded, if those libraries use
thread-local-storage (__thread) heavily. This causes
pthread_create to fail with small stacks; even if our
timer_thread function does not hit any of the TLS-using code
paths.
Today, some RubyGems are capable of using libkqueue (or __thread
storage directly), and future versions of Ruby may use kqueue
internally.
cf. https://www.akkadia.org/drepper/tls.pdf
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58874 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (ruby_init_stack): check stack bounds even if
get_main_stack succeeded, on the "co-routine" case.
https://github.com/ruby/ruby/commit/53953ee#commitcomment-18887413
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@56060 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (gvl_atfork): used in rb_thread_atfork_internal
only if HAVE_WORKING_FORK is defined.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@55085 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (rb_thread_setname): defer setting native thread name
set in initialize until the native thread is created.
[ruby-core:74963] [Bug #12290]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@54600 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
pthread_getthrds_np() is wrong on AIX. Use
__pi_stackend - __pi_stackaddr instead.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@54000 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This fixes an occasional [ASYNC BUG] failure in
bootstraptest/test_fork.rb '[ruby-dev:37934]'
which tests fork/pthread_create failure by setting
RLIMIT_NPROC to 1 and triggering EAGAIN on pthread_create
when attempting to recreate the timer thread.
The problem timeline is as follows:
thread 1 thread 2
---------------------------------------------------------------
rb_thread_create_timer_thread
setup_communication_pipe
rb_thread_wakeup_timer_thread_low
pthread_create fails pipe looks valid, write!
CLOSE_INVALIDATE (x4) EBADF -> ASYNC BUG
The checks in rb_thread_wakeup_timer_thread_low only tried to
guarantee proper ordering with native_stop_timer_thread, not
rb_thread_create_timer_thread :x
Now, this should allow rb_thread_create_timer_thread to
synchronize properly with rb_thread_wakeup_timer_thread_low by
delaying the validation marking of the timer_thread_pipe until
we are certain the timer thread is alive.
In this version, rb_thread_wakeup_timer_thread_low becomes a
noop. Threading is still completely broken with NPROC==1, but
there's not much we can do about it beside warn the user.
We no longer spew a scary [ASYNC BUG] message or dump core
on them.
* thread_pthread.c (setup_communication_pipe): delay setting owner
(rb_thread_create_timer_thread): until thread creation succeeds
[ruby-core:72590] [Bug #11922]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@53373 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* configure.in: separate SET_CURRENT_THREAD_NAME, which can set
the name of current thread only, and SET_ANOTHER_THREAD_NAME,
which can set the name of other threads.
* thread.c (rb_thread_setname): use SET_ANOTHER_THREAD_NAME. OS X
is not possible to set another thread name.
* thread_pthread.c (native_set_thread_name, thread_timer): use
SET_CURRENT_THREAD_NAME.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@52866 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (native_stop_timer_thread): explicit type of
old value to suppress a pointer-sign warning.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51578 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Shutting down the timer thread now always closes pipes to free FDs.
In fact, we close the write ends of the pipes is done in the main
RubyVM to signal the timer thread shutdown.
To effectively close pipes, we implement userspace locks via
atomics to force the pipe closing thread to wait on any signal
handlers which may be waking up.
While we're at it, improve robustness during resource exhaustion and
allow it to limp along non-fatally if restarting a timer thread
fails.
This reverts r51268
Note: this change is tested with VM_CHECK_MODE 1 in vm_core.h
* process.c (close_unless_reserved): add extra check
(dup2_with_divert): remove
(redirect_dup2): use dup2 without divert
(before_exec_non_async_signal_safe): adjust call + comment
(rb_f_exec): stop timer thread for all OSes
(rb_exec_without_timer_thread): remove
* eval.c (ruby_cleanup): adjust call
* thread.c (rb_thread_stop_timer_thread): always close pipes
* thread_pthread.c (struct timer_thread_pipe): add writing field,
mark owner_process volatile for signal handlers
(rb_thread_wakeup_timer_thread_fd): check valid FD
(rb_thread_wakeup_timer_thread): set writing flag to prevent close
(rb_thread_wakeup_timer_thread_low): ditto
(CLOSE_INVALIDATE): new macro
(close_invalidate): new function
(close_communication_pipe): removed
(setup_communication_pipe_internal): make errors non-fatal
(setup_communication_pipe): ditto
(thread_timer): close reading ends inside timer thread
(rb_thread_create_timer_thread): make errors non-fatal
(native_stop_timer_thread): close write ends only, always,
wait for signal handlers to finish
(rb_divert_reserved_fd): remove
* thread_win32.c (native_stop_timer_thread): adjust (untested)
(rb_divert_reserved_fd): remove
* vm_core.h: adjust prototype
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51576 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
with one of the timer thread FDs, the internal FD is diverted.
[Bug #11336] [ruby-core:69886] [Bug #11350] [ruby-core:69961]
* process.c (dup2_with_divert): new function for the above purpose.
* thread_pthread.c (rb_divert_reserved_fd): new function for
diverting reserved FD. If the given FD is the same as one of the
reserved FDs, the reserved FD number is internally changed.
It returns -1 when error. Otherwise, returns 0. It also returns
0 if there is no need to change reserved FD number.
* thread_win32.c (rb_divert_reserved_fd): always returns 0 because
of no reserved FDs.
* internal.h (rb_divert_reserved_fd): prototype declaration.
It is Ruby internal use only.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51268 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
reserved FD. It should be closed in the exec system call due to the
O_CLOEXEC or FD_CLOEXEC flag. [Bug #11353] [ruby-core:69977]
* process.c (close_unless_reserved): new function to close FD unless
it is reserved for internal communication.
* thread_pthread.c (rb_reserved_fd_p): should check owner_process pid
to avoid false positive in forked child process.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@51255 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
for FreeBSD, and don't define it if both pthread_setname_np
and pthread_set_name_np don't exist.
* thread_pthread.c (SET_THREAD_NAME): don't define if they don't exist.
* thread_pthread.c (native_set_thread_name): run if SET_THREAD_NAME
is defined.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50831 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (reserve_stack): keep sp safe zone to get rid
of crash by -fstack-check. [ruby-core:68740] [Bug #11030]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50316 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (timer_thread_pipe.owner_process): process ID
should be pid_t.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49796 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (reserve_stack): fix intermittent SIGBUS on
Linux, by reserving the stack virtual address space at process
start up so that it will not clash with the heap space.
[Fix GH-822]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49452 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread.c (struct join_arg): restructure and make smaller
(thread_join_sleep): avoid timeofday() call if forever
(thread_join): pass join_arg.delay directly
(rb_thread_inspect_msg): remove, inline into rb_thread_inspect
(rb_thread_inspect): reduce branching and string creation
* thread_pthread.c (native_set_thread_name): create string directly
to avoid reparsing. [Misc #10723]
This reduces time in benchmark/bm_vm_thread_create_join.rb by
a few percent.
Minor improvements only:
target 0: 2.1.5 (ruby 2.1.5p273 (2014-11-13 revision 48405) [x86_64-linux])
target 1: trunk (ruby 2.3.0dev (2015-01-16 trunk 49282) [x86_64-linux])
target 2: built (ruby 2.3.0dev (2015-01-16 trunk 49282) [x86_64-linux])
benchmark results:
minimum results in each 3 measurements.
Execution time (sec)
name 2.1.5 trunk built
vm_thread_create_join 1.049 1.242 1.138
Speedup ratio: compare with the result of `2.1.5' (greater is better)
name trunk built
vm_thread_create_join 0.845 0.923
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@49430 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
set thread name visible with ps command on GNU/Linux.
Ex. ps -o %c -L
* thread.c (thread_start_func_2): Call native_set_thread_name at
beginning.
(rb_thread_inspect_msg): Extract from rb_thread_inspect.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47670 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
This will make it easier for us to try alternative mutex/condvar
implementations while still using pthreads for thread management.
[Feature #10134]
* thread_pthread.h: define RB_NATIVETHREAD_LOCK_INIT and
RB_NATIVETHREAD_COND_INIT macros
* thread_pthread.c (native_mutex_lock, native_mutex_unlock,
native_mutex_trylock, native_mutex_initialize,
native_mutex_destroy, native_cond_wait):
use rb_nativethread_lock_t instead of pthread_mutex_t
* thread_pthread.c (native_mutex_debug): make argument type-agnostic
to avoid later cast.
* thread_pthread.c (register_cached_thread_and_wait):
replace PTHREAD_COND_INITIALIZER with RB_NATIVETHREAD_COND_INIT,
use native_mutex_{lock,unlock}
* thread_pthread.c (use_cached_thread):
use native_mutex_{lock,unlock}
* thread_pthread.c (native_sleep):
use rb_nativethread_lock_t to match th->interrupt_lock,
use native_mutex_{lock,unlock}
* thread_pthread.c (timer_thread_lock): use rb_nativethread_lock_t type
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@47185 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* configure.in (rb_cv_scalar_pthread_t): pthread_t is not required
to be a scalar type.
* thread.c (fill_thread_id_string, thread_id_str): dump pthread_t
in hexadecimal form if it is not a scalar type, assume it can be
represented in a pointer form otherwise. based on the patch by
Rei Odaira at [ruby-core:62867]. [ruby-core:62857] [Bug #9884]
* thread_pthread.c (Init_native_thread, thread_start_func_1),
(native_thread_create): set thread_id_str if needed.
* vm_core.h (rb_thread_t): add thread_id_string if needed.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46406 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (timer_thread): add a flag to tell timer thread
is created, since 0 may be a valid value as pthread_t.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@46405 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (rb_thread_wakeup_timer_thread_fd): use a local
variable for errno.
* thread_pthread.c (consume_communication_pipe): ditto. add
EWOULDBLOCK case.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45909 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (ubf_select): compare thread_id with
pthread_equal() but not directly.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45906 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (timer_thread_sleep): use a local variable for
errno.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@45898 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* configure.in: check if pthread_setname_np is available.
* thread_pthread.c: pthread_setname_np is not available on old
Darwins. [ruby-core:60524] [Bug #9492]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44925 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* vm_core.h (rb_thread_struct): aggregate cpu stuff into a struct,
so that a debugger can show its content at once.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44722 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (ruby_stack_overflowed_p): range of rlim_cur
may be larger than size_t. [ruby-core:60113] [Bug #9454]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44715 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c: get current main thread stack size, which may
be expanded than allocated size at initialization, by rlimit().
[ruby-core:60113] [Bug #9454]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44712 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (rb_thread_create_timer_thread): fix for platforms
where PTHREAD_STACK_MIN is a dynamic value and not a compile-time
constant. [ruby-dev:47911] [Bug #9436]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44675 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (rb_thread_create_timer_thread): expand timer
thread stack size to get rid of segfault on FreeBSD/powerpc64.
based on the patch by Steve Wills at [ruby-core:59923].
[ruby-core:56590] [Bug #8783]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44671 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* thread_pthread.c (ruby_init_stack): set stack_start properly by
get_main_stack() if possible.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@43706 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
nobu
normal
naruse
odaira
nagachika
hsbt
kosaki
ngoto
akr