d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
55b270e68c
freebsd-nq/lib/libthr/thread/thr_sig.c

763 lines
19 KiB
C
Raw Normal View History

- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
/*
Update copyright.
2005-12-17 09:42:45 +00:00
* Copyright (c) 2005, David Xu <davidxu@freebsd.org>
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
Update copyright.
2005-12-17 09:42:45 +00:00
* notice unmodified, this list of conditions, and the following
* disclaimer.
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
Update copyright.
2005-12-17 09:42:45 +00:00
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
*
* $FreeBSD$
*/
WARNS level 4 cleanup.
2006-04-04 02:57:49 +00:00
#include "namespace.h"
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
#include <sys/param.h>
#include <sys/types.h>
#include <sys/signalvar.h>
Pre-resolve symbols required for the deferred signal processing. This avoids recursion into rtld when leaving libthr critical section for the deferred signal delivery. For the same reason, use syscall(2) instead of referencing __sys_sigreturn(2). Syscall() is already pre-resolved for fork() interceptor. Tested by: Andre Meiser <ortadur@web.de> Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-08-10 17:02:42 +00:00
#include <sys/syscall.h>
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
#include <signal.h>
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
#include <errno.h>
Use getcontextx(3) internal API instead of getcontext(2) to provide the signal handlers with the context information in the deferrred case. Only enable the use of getcontextx(3) in the deferred signal delivery code on amd64 and i386. Sparc64 seems to have some undetermined issues with interaction of alloca(3) and signal delivery. Tested by: flo (who also provided sparc64 harware access for me), pho Discussed with: marius MFC after: 1 month
2012-01-21 18:06:18 +00:00
#include <stdlib.h>
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
#include <string.h>
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
#include <pthread.h>
WARNS level 4 cleanup.
2006-04-04 02:57:49 +00:00
#include "un-namespace.h"
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
#include "libc_private.h"
Fix the wrapper function around signals so that a signal handling thread on one of the mutex or condition variable queues is removed from those queues before the real signal handler is called.
2003-12-09 11:12:11 +00:00
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
#include "libc_private.h"
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
#include "thr_private.h"
/* #define DEBUG_SIGNAL */
#ifdef DEBUG_SIGNAL
#define DBG_MSG stdout_debug
#else
#define DBG_MSG(x...)
#endif
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct usigaction {
struct sigaction sigact;
struct urwlock lock;
};
static struct usigaction _thr_sigact[_SIG_MAXSIG];
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
static inline struct usigaction *
__libc_sigaction_slot(int signo)
{
return (&_thr_sigact[signo - 1]);
}
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
static void thr_sighandler(int, siginfo_t *, void *);
static void handle_signal(struct sigaction *, int, siginfo_t *, ucontext_t *);
static void check_deferred_signal(struct pthread *);
static void check_suspend(struct pthread *);
static void check_cancel(struct pthread *curthread, ucontext_t *ucp);
Add some function prototypes.
2007-11-21 05:23:54 +00:00
int _sigtimedwait(const sigset_t *set, siginfo_t *info,
const struct timespec * timeout);
int _sigwaitinfo(const sigset_t *set, siginfo_t *info);
int _sigwait(const sigset_t *set, int *sig);
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
int _setcontext(const ucontext_t *);
int _swapcontext(ucontext_t *, const ucontext_t *);
WARNS level 4 cleanup.
2006-04-04 02:57:49 +00:00
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
static const sigset_t _thr_deferset={{
0xffffffff & ~(_SIG_BIT(SIGBUS)|_SIG_BIT(SIGILL)|_SIG_BIT(SIGFPE)|
_SIG_BIT(SIGSEGV)|_SIG_BIT(SIGTRAP)|_SIG_BIT(SIGSYS)),
0xffffffff,
0xffffffff,
0xffffffff}};
static const sigset_t _thr_maskset={{
0xffffffff,
0xffffffff,
0xffffffff,
0xffffffff}};
void
_thr_signal_block(struct pthread *curthread)
{
if (curthread->sigblock > 0) {
curthread->sigblock++;
return;
}
__sys_sigprocmask(SIG_BLOCK, &_thr_maskset, &curthread->sigmask);
curthread->sigblock++;
}
void
_thr_signal_unblock(struct pthread *curthread)
{
if (--curthread->sigblock == 0)
__sys_sigprocmask(SIG_SETMASK, &curthread->sigmask, NULL);
}
int
_thr_send_sig(struct pthread *thread, int sig)
{
return thr_kill(thread->tid, sig);
}
static inline void
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
remove_thr_signals(sigset_t *set)
{
if (SIGISMEMBER(*set, SIGCANCEL))
SIGDELSET(*set, SIGCANCEL);
}
static const sigset_t *
thr_remove_thr_signals(const sigset_t *set, sigset_t *newset)
{
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
*newset = *set;
remove_thr_signals(newset);
return (newset);
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
}
Remove umtx_t definition, use type long directly, add wrapper function _thr_umtx_wait_uint() for umtx operation UMTX_OP_WAIT_UINT, use the function in semaphore operations, this fixed compiler warnings.
2007-11-21 05:21:58 +00:00
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
static void
WARNS level 4 cleanup.
2006-04-04 02:57:49 +00:00
sigcancel_handler(int sig __unused,
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
siginfo_t *info __unused, ucontext_t *ucp)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
struct pthread *curthread = _get_curthread();
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
int err;
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
if (THR_IN_CRITICAL(curthread))
return;
err = errno;
check_suspend(curthread);
check_cancel(curthread, ucp);
errno = err;
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
typedef void (*ohandler)(int sig, int code, struct sigcontext *scp,
char *addr, __sighandler_t *catcher);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/*
* The signal handler wrapper is entered with all signal masked.
*/
static void
thr_sighandler(int sig, siginfo_t *info, void *_ucp)
{
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
struct pthread *curthread;
ucontext_t *ucp;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct sigaction act;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
struct usigaction *usa;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
int err;
err = errno;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
curthread = _get_curthread();
ucp = _ucp;
usa = __libc_sigaction_slot(sig);
_thr_rwl_rdlock(&usa->lock);
act = usa->sigact;
_thr_rwl_unlock(&usa->lock);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
errno = err;
If check_deferred_signal() execution needs binding of PLT symbol, unlocking the rtld bind lock results in the processing of ast and recursing into the check_deferred_signal(). Nested execution of check_deferred_signal() delivers the signal to user code and clears si_signo. On return, top-level check_deferred_signal() frame continues delivering the same signal one more time, but now with zero si_signo. Fix this by adding a flag to indicate that deferred delivery is running, so check_deferred_signal() should avoid doing anything. Since user signal handler is allowed to modify the passed machine context to make return from the signal handler to cause arbitrary jump, or do longjmp(). For this case, also clear the flag in thr_sighandler(), since kernel signal delivery means that nested delivery code should not run right now. Reported by: Vitaly Magerya <vmagerya@gmail.com> Reviewed by: davidxu, jilles Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-11-23 15:48:17 +00:00
curthread->deferred_run = 0;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/*
* if a thread is in critical region, for example it holds low level locks,
* try to defer the signal processing, however if the signal is synchronous
* signal, it means a bad thing has happened, this is a programming error,
* resuming fault point can not help anything (normally causes deadloop),
* so here we let user code handle it immediately.
*/
if (THR_IN_CRITICAL(curthread) && SIGISMEMBER(_thr_deferset, sig)) {
memcpy(&curthread->deferred_sigact, &act, sizeof(struct sigaction));
memcpy(&curthread->deferred_siginfo, info, sizeof(siginfo_t));
curthread->deferred_sigmask = ucp->uc_sigmask;
/* mask all signals, we will restore it later. */
ucp->uc_sigmask = _thr_deferset;
return;
}
handle_signal(&act, sig, info, ucp);
}
static void
handle_signal(struct sigaction *actp, int sig, siginfo_t *info, ucontext_t *ucp)
{
struct pthread *curthread = _get_curthread();
ucontext_t uc2;
__siginfohandler_t *sigfunc;
int cancel_point;
int cancel_async;
int cancel_enable;
int in_sigsuspend;
int err;
/* add previous level mask */
SIGSETOR(actp->sa_mask, ucp->uc_sigmask);
/* add this signal's mask */
if (!(actp->sa_flags & SA_NODEFER))
SIGADDSET(actp->sa_mask, sig);
in_sigsuspend = curthread->in_sigsuspend;
curthread->in_sigsuspend = 0;
/*
Partially apply the capitalization of the heading word of the sequence and fix typo. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-05-27 18:45:45 +00:00
* If thread is in deferred cancellation mode, disable cancellation
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
* in signal handler.
Partially apply the capitalization of the heading word of the sequence and fix typo. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-05-27 18:45:45 +00:00
* If user signal handler calls a cancellation point function, e.g,
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
* it calls write() to write data to file, because write() is a
* cancellation point, the thread is immediately cancelled if
* cancellation is pending, to avoid this problem while thread is in
* deferring mode, cancellation is temporarily disabled.
*/
cancel_point = curthread->cancel_point;
cancel_async = curthread->cancel_async;
cancel_enable = curthread->cancel_enable;
curthread->cancel_point = 0;
if (!cancel_async)
curthread->cancel_enable = 0;
/* restore correct mask before calling user handler */
__sys_sigprocmask(SIG_SETMASK, &actp->sa_mask, NULL);
sigfunc = actp->sa_sigaction;
/*
* We have already reset cancellation point flags, so if user's code
* longjmp()s out of its signal handler, wish its jmpbuf was set
* outside of a cancellation point, in most cases, this would be
Partially apply the capitalization of the heading word of the sequence and fix typo. Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-05-27 18:45:45 +00:00
* true. However, there is no way to save cancel_enable in jmpbuf,
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
* so after setjmps() returns once more, the user code may need to
* re-set cancel_enable flag by calling pthread_setcancelstate().
*/
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
if ((actp->sa_flags & SA_SIGINFO) != 0) {
sigfunc(sig, info, ucp);
} else {
((ohandler)sigfunc)(sig, info->si_code,
(struct sigcontext *)ucp, info->si_addr,
(__sighandler_t *)sigfunc);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
err = errno;
curthread->in_sigsuspend = in_sigsuspend;
curthread->cancel_point = cancel_point;
curthread->cancel_enable = cancel_enable;
memcpy(&uc2, ucp, sizeof(uc2));
SIGDELSET(uc2.uc_sigmask, SIGCANCEL);
/* reschedule cancellation */
check_cancel(curthread, &uc2);
errno = err;
Pre-resolve symbols required for the deferred signal processing. This avoids recursion into rtld when leaving libthr critical section for the deferred signal delivery. For the same reason, use syscall(2) instead of referencing __sys_sigreturn(2). Syscall() is already pre-resolved for fork() interceptor. Tested by: Andre Meiser <ortadur@web.de> Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-08-10 17:02:42 +00:00
syscall(SYS_sigreturn, &uc2);
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
}
void
_thr_ast(struct pthread *curthread)
{
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
if (!THR_IN_CRITICAL(curthread)) {
check_deferred_signal(curthread);
check_suspend(curthread);
check_cancel(curthread, NULL);
}
}
/* reschedule cancellation */
static void
check_cancel(struct pthread *curthread, ucontext_t *ucp)
{
If we are at cancellation point, always work as deferred mode despite whether asynchronous mode is turned on or not, this always gives us a chance to decide whether thread should be canceled or not in cancellation points.
2010-09-21 06:47:04 +00:00
if (__predict_true(!curthread->cancel_pending ||
!curthread->cancel_enable || curthread->no_cancel))
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
return;
If we are at cancellation point, always work as deferred mode despite whether asynchronous mode is turned on or not, this always gives us a chance to decide whether thread should be canceled or not in cancellation points.
2010-09-21 06:47:04 +00:00
/*
* Otherwise, we are in defer mode, and we are at
* cancel point, tell kernel to not block the current
* thread on next cancelable system call.
*
* There are three cases we should call thr_wake() to
* turn on TDP_WAKEUP or send SIGCANCEL in kernel:
* 1) we are going to call a cancelable system call,
* non-zero cancel_point means we are already in
* cancelable state, next system call is cancelable.
* 2) because _thr_ast() may be called by
* THR_CRITICAL_LEAVE() which is used by rtld rwlock
* and any libthr internal locks, when rtld rwlock
Fix typo in comment. MFC after: 3 days
2015-06-14 19:19:46 +00:00
* is used, it is mostly caused by an unresolved PLT.
* Those routines may clear the TDP_WAKEUP flag by
If we are at cancellation point, always work as deferred mode despite whether asynchronous mode is turned on or not, this always gives us a chance to decide whether thread should be canceled or not in cancellation points.
2010-09-21 06:47:04 +00:00
* invoking some system calls, in those cases, we
* also should reenable the flag.
* 3) thread is in sigsuspend(), and the syscall insists
* on getting a signal before it agrees to return.
*/
if (curthread->cancel_point) {
if (curthread->in_sigsuspend && ucp) {
SIGADDSET(ucp->uc_sigmask, SIGCANCEL);
curthread->unblock_sigcancel = 1;
_thr_send_sig(curthread, SIGCANCEL);
} else
thr_wake(curthread->tid);
} else if (curthread->cancel_async) {
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/*
If we are at cancellation point, always work as deferred mode despite whether asynchronous mode is turned on or not, this always gives us a chance to decide whether thread should be canceled or not in cancellation points.
2010-09-21 06:47:04 +00:00
* asynchronous cancellation mode, act upon
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
* immediately.
If we are at cancellation point, always work as deferred mode despite whether asynchronous mode is turned on or not, this always gives us a chance to decide whether thread should be canceled or not in cancellation points.
2010-09-21 06:47:04 +00:00
*/
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_pthread_exit_mask(PTHREAD_CANCELED,
ucp? &ucp->uc_sigmask : NULL);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
}
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
static void
check_deferred_signal(struct pthread *curthread)
{
Use getcontextx(3) internal API instead of getcontext(2) to provide the signal handlers with the context information in the deferrred case. Only enable the use of getcontextx(3) in the deferred signal delivery code on amd64 and i386. Sparc64 seems to have some undetermined issues with interaction of alloca(3) and signal delivery. Tested by: flo (who also provided sparc64 harware access for me), pho Discussed with: marius MFC after: 1 month
2012-01-21 18:06:18 +00:00
ucontext_t *uc;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct sigaction act;
siginfo_t info;
Since the cause of the problems with the __fillcontextx() was identified, unify the code of check_deferred_signal() for all architectures, making the variant under #ifdef x86 common. Tested by: marius (sparc64) Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-06-03 04:22:42 +00:00
int uc_len;
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
If check_deferred_signal() execution needs binding of PLT symbol, unlocking the rtld bind lock results in the processing of ast and recursing into the check_deferred_signal(). Nested execution of check_deferred_signal() delivers the signal to user code and clears si_signo. On return, top-level check_deferred_signal() frame continues delivering the same signal one more time, but now with zero si_signo. Fix this by adding a flag to indicate that deferred delivery is running, so check_deferred_signal() should avoid doing anything. Since user signal handler is allowed to modify the passed machine context to make return from the signal handler to cause arbitrary jump, or do longjmp(). For this case, also clear the flag in thr_sighandler(), since kernel signal delivery means that nested delivery code should not run right now. Reported by: Vitaly Magerya <vmagerya@gmail.com> Reviewed by: davidxu, jilles Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-11-23 15:48:17 +00:00
if (__predict_true(curthread->deferred_siginfo.si_signo == 0 ||
curthread->deferred_run))
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
return;
Use getcontextx(3) internal API instead of getcontext(2) to provide the signal handlers with the context information in the deferrred case. Only enable the use of getcontextx(3) in the deferred signal delivery code on amd64 and i386. Sparc64 seems to have some undetermined issues with interaction of alloca(3) and signal delivery. Tested by: flo (who also provided sparc64 harware access for me), pho Discussed with: marius MFC after: 1 month
2012-01-21 18:06:18 +00:00
If check_deferred_signal() execution needs binding of PLT symbol, unlocking the rtld bind lock results in the processing of ast and recursing into the check_deferred_signal(). Nested execution of check_deferred_signal() delivers the signal to user code and clears si_signo. On return, top-level check_deferred_signal() frame continues delivering the same signal one more time, but now with zero si_signo. Fix this by adding a flag to indicate that deferred delivery is running, so check_deferred_signal() should avoid doing anything. Since user signal handler is allowed to modify the passed machine context to make return from the signal handler to cause arbitrary jump, or do longjmp(). For this case, also clear the flag in thr_sighandler(), since kernel signal delivery means that nested delivery code should not run right now. Reported by: Vitaly Magerya <vmagerya@gmail.com> Reviewed by: davidxu, jilles Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-11-23 15:48:17 +00:00
curthread->deferred_run = 1;
The getcontext() from the __fillcontextx() call in the check_deferred_signal() returns twice, since handle_signal() emulates the return from the normal signal handler by sigreturn(2)ing the passed context. Second return is performed on the destroyed stack frame, because __fillcontextx() has already returned. This causes undefined and bad behaviour, usually the victim thread gets SIGSEGV. Avoid nested frame and the need to return from it by doing direct call to getcontext() in the check_deferred_signal() and using a new private libc helper __fillcontextx2() to complement the context with the extended CPU state if the deferred signal is still present. The __fillcontextx() is now unused, but is kept to allow older libthr.so to be used with the new libc. Mark __fillcontextx() as returning twice [1]. Reported by: pgj Pointy hat to: kib Discussed with: dim Tested by: pgj, dim Suggested by: jilles [1] MFC after: 1 week
2013-05-28 04:54:16 +00:00
uc_len = __getcontextx_size();
uc = alloca(uc_len);
getcontext(uc);
If check_deferred_signal() execution needs binding of PLT symbol, unlocking the rtld bind lock results in the processing of ast and recursing into the check_deferred_signal(). Nested execution of check_deferred_signal() delivers the signal to user code and clears si_signo. On return, top-level check_deferred_signal() frame continues delivering the same signal one more time, but now with zero si_signo. Fix this by adding a flag to indicate that deferred delivery is running, so check_deferred_signal() should avoid doing anything. Since user signal handler is allowed to modify the passed machine context to make return from the signal handler to cause arbitrary jump, or do longjmp(). For this case, also clear the flag in thr_sighandler(), since kernel signal delivery means that nested delivery code should not run right now. Reported by: Vitaly Magerya <vmagerya@gmail.com> Reviewed by: davidxu, jilles Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-11-23 15:48:17 +00:00
if (curthread->deferred_siginfo.si_signo == 0) {
curthread->deferred_run = 0;
The getcontext() from the __fillcontextx() call in the check_deferred_signal() returns twice, since handle_signal() emulates the return from the normal signal handler by sigreturn(2)ing the passed context. Second return is performed on the destroyed stack frame, because __fillcontextx() has already returned. This causes undefined and bad behaviour, usually the victim thread gets SIGSEGV. Avoid nested frame and the need to return from it by doing direct call to getcontext() in the check_deferred_signal() and using a new private libc helper __fillcontextx2() to complement the context with the extended CPU state if the deferred signal is still present. The __fillcontextx() is now unused, but is kept to allow older libthr.so to be used with the new libc. Mark __fillcontextx() as returning twice [1]. Reported by: pgj Pointy hat to: kib Discussed with: dim Tested by: pgj, dim Suggested by: jilles [1] MFC after: 1 week
2013-05-28 04:54:16 +00:00
return;
If check_deferred_signal() execution needs binding of PLT symbol, unlocking the rtld bind lock results in the processing of ast and recursing into the check_deferred_signal(). Nested execution of check_deferred_signal() delivers the signal to user code and clears si_signo. On return, top-level check_deferred_signal() frame continues delivering the same signal one more time, but now with zero si_signo. Fix this by adding a flag to indicate that deferred delivery is running, so check_deferred_signal() should avoid doing anything. Since user signal handler is allowed to modify the passed machine context to make return from the signal handler to cause arbitrary jump, or do longjmp(). For this case, also clear the flag in thr_sighandler(), since kernel signal delivery means that nested delivery code should not run right now. Reported by: Vitaly Magerya <vmagerya@gmail.com> Reviewed by: davidxu, jilles Tested by: pho Sponsored by: The FreeBSD Foundation MFC after: 1 week
2013-11-23 15:48:17 +00:00
}
The getcontext() from the __fillcontextx() call in the check_deferred_signal() returns twice, since handle_signal() emulates the return from the normal signal handler by sigreturn(2)ing the passed context. Second return is performed on the destroyed stack frame, because __fillcontextx() has already returned. This causes undefined and bad behaviour, usually the victim thread gets SIGSEGV. Avoid nested frame and the need to return from it by doing direct call to getcontext() in the check_deferred_signal() and using a new private libc helper __fillcontextx2() to complement the context with the extended CPU state if the deferred signal is still present. The __fillcontextx() is now unused, but is kept to allow older libthr.so to be used with the new libc. Mark __fillcontextx() as returning twice [1]. Reported by: pgj Pointy hat to: kib Discussed with: dim Tested by: pgj, dim Suggested by: jilles [1] MFC after: 1 week
2013-05-28 04:54:16 +00:00
__fillcontextx2((char *)uc);
Since the cause of the problems with the __fillcontextx() was identified, unify the code of check_deferred_signal() for all architectures, making the variant under #ifdef x86 common. Tested by: marius (sparc64) Sponsored by: The FreeBSD Foundation MFC after: 2 weeks
2013-06-03 04:22:42 +00:00
act = curthread->deferred_sigact;
uc->uc_sigmask = curthread->deferred_sigmask;
memcpy(&info, &curthread->deferred_siginfo, sizeof(siginfo_t));
/* remove signal */
curthread->deferred_siginfo.si_signo = 0;
handle_signal(&act, info.si_signo, &info, uc);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
}
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
static void
check_suspend(struct pthread *curthread)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
Use UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE to save time in kernel(avoid VM lookup).
2008-04-29 03:58:18 +00:00
uint32_t cycle;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
if (__predict_true((curthread->flags &
(THR_FLAGS_NEED_SUSPEND | THR_FLAGS_SUSPENDED))
!= THR_FLAGS_NEED_SUSPEND))
return;
In suspend_common(), don't wait for a thread which is in creation, because pthread_suspend_all_np() may have already suspended its parent thread. Add locking code in pthread_suspend_all_np() to only allow one thread to suspend other threads, this eliminates a deadlock where two or more threads try to suspend each others.
2012-08-27 03:09:39 +00:00
if (curthread == _single_thread)
return;
- Copy signal mask out before THR_UNLOCK(), because THR_UNLOCK() may call _thr_suspend_check() which messes sigmask saved in thread structure. - Don't suspend a thread has force_exit set. - In pthread_exit(), if there is a suspension flag set, wake up waiting- thread after setting PS_DEAD, this causes waiting-thread to break loop in suspend_common().
2008-03-18 02:06:51 +00:00
if (curthread->force_exit)
return;
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
/*
* Blocks SIGCANCEL which other threads must send.
*/
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
_thr_signal_block(curthread);
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
/*
* Increase critical_count, here we don't use THR_LOCK/UNLOCK
* because we are leaf code, we don't want to recursively call
* ourself.
*/
curthread->critical_count++;
Replace internal usage of struct umtx with umutex which can supports real-time if we want, no functionality is changed.
2006-09-06 04:04:10 +00:00
THR_UMUTEX_LOCK(curthread, &(curthread)->lock);
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
while ((curthread->flags & (THR_FLAGS_NEED_SUSPEND |
THR_FLAGS_SUSPENDED)) == THR_FLAGS_NEED_SUSPEND) {
curthread->cycle++;
cycle = curthread->cycle;
/* Wake the thread suspending us. */
Use UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE to save time in kernel(avoid VM lookup).
2008-04-29 03:58:18 +00:00
_thr_umtx_wake(&curthread->cycle, INT_MAX, 0);
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
/*
* if we are from pthread_exit, we don't want to
* suspend, just go and die.
*/
if (curthread->state == PS_DEAD)
break;
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
curthread->flags |= THR_FLAGS_SUSPENDED;
Replace internal usage of struct umtx with umutex which can supports real-time if we want, no functionality is changed.
2006-09-06 04:04:10 +00:00
THR_UMUTEX_UNLOCK(curthread, &(curthread)->lock);
Use UMTX_OP_WAIT_UINT_PRIVATE and UMTX_OP_WAKE_PRIVATE to save time in kernel(avoid VM lookup).
2008-04-29 03:58:18 +00:00
_thr_umtx_wait_uint(&curthread->cycle, cycle, NULL, 0);
Replace internal usage of struct umtx with umutex which can supports real-time if we want, no functionality is changed.
2006-09-06 04:04:10 +00:00
THR_UMUTEX_LOCK(curthread, &(curthread)->lock);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
curthread->flags &= ~THR_FLAGS_SUSPENDED;
}
Replace internal usage of struct umtx with umutex which can supports real-time if we want, no functionality is changed.
2006-09-06 04:04:10 +00:00
THR_UMUTEX_UNLOCK(curthread, &(curthread)->lock);
Refine thread suspension code, now thread suspension is a blockable operation, the caller is blocked util target threads are really suspended, also avoid suspending a thread when it is holding a critical lock. Fix a bug in _thr_ref_delete which tests a never set flag.
2006-01-05 13:51:22 +00:00
curthread->critical_count--;
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
_thr_signal_unblock(curthread);
}
void
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
_thr_signal_init(int dlopened)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
struct sigaction act, nact, oact;
struct usigaction *usa;
sigset_t oldset;
int sig, error;
if (dlopened) {
__sys_sigprocmask(SIG_SETMASK, &_thr_maskset, &oldset);
for (sig = 1; sig <= _SIG_MAXSIG; sig++) {
if (sig == SIGCANCEL)
continue;
error = __sys_sigaction(sig, NULL, &oact);
if (error == -1 || oact.sa_handler == SIG_DFL ||
oact.sa_handler == SIG_IGN)
continue;
usa = __libc_sigaction_slot(sig);
usa->sigact = oact;
nact = oact;
remove_thr_signals(&usa->sigact.sa_mask);
nact.sa_flags &= ~SA_NODEFER;
nact.sa_flags |= SA_SIGINFO;
nact.sa_sigaction = thr_sighandler;
nact.sa_mask = _thr_maskset;
(void)__sys_sigaction(sig, &nact, NULL);
}
__sys_sigprocmask(SIG_SETMASK, &oldset, NULL);
}
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/* Install SIGCANCEL handler. */
SIGFILLSET(act.sa_mask);
act.sa_flags = SA_SIGINFO;
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
act.sa_sigaction = (__siginfohandler_t *)&sigcancel_handler;
__sys_sigaction(SIGCANCEL, &act, NULL);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/* Unblock SIGCANCEL */
SIGEMPTYSET(act.sa_mask);
SIGADDSET(act.sa_mask, SIGCANCEL);
__sys_sigprocmask(SIG_UNBLOCK, &act.sa_mask, NULL);
}
void
_thr_sigact_unload(struct dl_phdr_info *phdr_info)
{
Fix off-by-one error in function _thr_sigact_unload, also disable the function, it seems some gnome application tends to crash if we unregister sigaction automatically.
2010-09-06 03:00:54 +00:00
#if 0
Remove incorrect comments, also make sure signal is disabled when unregistering sigaction.
2010-09-01 13:22:55 +00:00
struct pthread *curthread = _get_curthread();
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct urwlock *rwlp;
struct sigaction *actp;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
struct usigaction *usa;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct sigaction kact;
void (*handler)(int);
int sig;
Remove incorrect comments, also make sure signal is disabled when unregistering sigaction.
2010-09-01 13:22:55 +00:00
_thr_signal_block(curthread);
Fix off-by-one error in function _thr_sigact_unload, also disable the function, it seems some gnome application tends to crash if we unregister sigaction automatically.
2010-09-06 03:00:54 +00:00
for (sig = 1; sig <= _SIG_MAXSIG; sig++) {
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
usa = __libc_sigaction_slot(sig);
actp = &usa->sigact;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
retry:
handler = actp->sa_handler;
if (handler != SIG_DFL && handler != SIG_IGN &&
__elf_phdr_match_addr(phdr_info, handler)) {
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
rwlp = &usa->lock;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_rwl_wrlock(rwlp);
if (handler != actp->sa_handler) {
_thr_rwl_unlock(rwlp);
goto retry;
}
actp->sa_handler = SIG_DFL;
actp->sa_flags = SA_SIGINFO;
SIGEMPTYSET(actp->sa_mask);
if (__sys_sigaction(sig, NULL, &kact) == 0 &&
kact.sa_handler != SIG_DFL &&
kact.sa_handler != SIG_IGN)
__sys_sigaction(sig, actp, NULL);
Remove incorrect comments, also make sure signal is disabled when unregistering sigaction.
2010-09-01 13:22:55 +00:00
_thr_rwl_unlock(rwlp);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
}
Remove incorrect comments, also make sure signal is disabled when unregistering sigaction.
2010-09-01 13:22:55 +00:00
_thr_signal_unblock(curthread);
Fix off-by-one error in function _thr_sigact_unload, also disable the function, it seems some gnome application tends to crash if we unregister sigaction automatically.
2010-09-06 03:00:54 +00:00
#endif
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
void
_thr_signal_prefork(void)
{
int i;
libthr: In the atfork handlers for signals, do not skip the last signal. _SIG_MAXSIG works a bit unexpectedly: signals 1 till _SIG_MAXSIG are valid, both bounds inclusive. Reviewed by: davidxu MFC after: 1 week
2012-03-26 17:05:26 +00:00
for (i = 1; i <= _SIG_MAXSIG; ++i)
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
_thr_rwl_rdlock(&__libc_sigaction_slot(i)->lock);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
void
_thr_signal_postfork(void)
{
int i;
libthr: In the atfork handlers for signals, do not skip the last signal. _SIG_MAXSIG works a bit unexpectedly: signals 1 till _SIG_MAXSIG are valid, both bounds inclusive. Reviewed by: davidxu MFC after: 1 week
2012-03-26 17:05:26 +00:00
for (i = 1; i <= _SIG_MAXSIG; ++i)
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
_thr_rwl_unlock(&__libc_sigaction_slot(i)->lock);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
void
_thr_signal_postfork_child(void)
{
int i;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
for (i = 1; i <= _SIG_MAXSIG; ++i) {
bzero(&__libc_sigaction_slot(i) -> lock,
sizeof(struct urwlock));
}
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
}
void
_thr_signal_deinit(void)
{
}
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigaction(int sig, const struct sigaction *act, struct sigaction *oact)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct sigaction newact, oldact, oldact2;
sigset_t oldset;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
struct usigaction *usa;
int ret, err;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
Use _SIG_VALID instead of expanded form of the macro. Submitted by: Garrett Cooper <yanegomi gmail com> MFC after: 1 week
2010-07-12 10:15:33 +00:00
if (!_SIG_VALID(sig) || sig == SIGCANCEL) {
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
errno = EINVAL;
return (-1);
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
ret = 0;
err = 0;
usa = __libc_sigaction_slot(sig);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
__sys_sigprocmask(SIG_SETMASK, &_thr_maskset, &oldset);
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
_thr_rwl_wrlock(&usa->lock);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
if (act != NULL) {
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
oldact2 = usa->sigact;
newact = *act;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
/*
* if a new sig handler is SIG_DFL or SIG_IGN,
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
* don't remove old handler from __libc_sigact[],
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
* so deferred signals still can use the handlers,
* multiple threads invoking sigaction itself is
* a race condition, so it is not a problem.
*/
if (newact.sa_handler != SIG_DFL &&
newact.sa_handler != SIG_IGN) {
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
usa->sigact = newact;
remove_thr_signals(&usa->sigact.sa_mask);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
newact.sa_flags &= ~SA_NODEFER;
newact.sa_flags |= SA_SIGINFO;
newact.sa_sigaction = thr_sighandler;
newact.sa_mask = _thr_maskset; /* mask all signals */
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
ret = __sys_sigaction(sig, &newact, &oldact);
if (ret == -1) {
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
err = errno;
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
usa->sigact = oldact2;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
} else if (oact != NULL) {
ret = __sys_sigaction(sig, NULL, &oldact);
err = errno;
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
if (oldact.sa_handler != SIG_DFL && oldact.sa_handler != SIG_IGN) {
Return previous sigaction correctly. Submitted by: avg
2010-10-29 09:35:36 +00:00
if (act != NULL)
oldact = oldact2;
else if (oact != NULL)
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
oldact = usa->sigact;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
_thr_rwl_unlock(&usa->lock);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
__sys_sigprocmask(SIG_SETMASK, &oldset, NULL);
if (ret == 0) {
if (oact != NULL)
*oact = oldact;
} else {
errno = err;
}
return (ret);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
}
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigprocmask(int how, const sigset_t *set, sigset_t *oset)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
const sigset_t *p = set;
sigset_t newset;
if (how != SIG_UNBLOCK) {
if (set != NULL) {
newset = *set;
SIGDELSET(newset, SIGCANCEL);
p = &newset;
}
}
return (__sys_sigprocmask(how, p, oset));
}
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
__weak_reference(_pthread_sigmask, pthread_sigmask);
int
_pthread_sigmask(int how, const sigset_t *set, sigset_t *oset)
{
Fix bug in r276630. Do not allow pthread_sigmask() to block SIGCANCEL. Reported and tested by: royger Sponsored by: The FreeBSD Foundation MFC after: 3 days
2015-01-21 16:13:37 +00:00
if (__thr_sigprocmask(how, set, oset))
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
return (errno);
return (0);
}
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
int
_sigsuspend(const sigset_t * set)
{
sigset_t newset;
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
return (__sys_sigsuspend(thr_remove_thr_signals(set, &newset)));
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
}
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigsuspend(const sigset_t * set)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
struct pthread *curthread;
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
sigset_t newset;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
int ret, old;
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
curthread = _get_curthread();
old = curthread->in_sigsuspend;
curthread->in_sigsuspend = 1;
Eliminate atomic operations in thread cancellation functions, it should reduce overheads of cancellation points.
2006-11-24 09:57:38 +00:00
_thr_cancel_enter(curthread);
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
ret = __sys_sigsuspend(thr_remove_thr_signals(set, &newset));
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_cancel_leave(curthread, 1);
curthread->in_sigsuspend = old;
if (curthread->unblock_sigcancel) {
curthread->unblock_sigcancel = 0;
SIGEMPTYSET(newset);
SIGADDSET(newset, SIGCANCEL);
__sys_sigprocmask(SIG_UNBLOCK, &newset, NULL);
}
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
return (ret);
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
}
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
int
_sigtimedwait(const sigset_t *set, siginfo_t *info,
const struct timespec * timeout)
{
sigset_t newset;
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
return (__sys_sigtimedwait(thr_remove_thr_signals(set, &newset), info,
timeout));
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
}
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
/*
* Cancellation behavior:
* Thread may be canceled at start, if thread got signal,
* it is not canceled.
*/
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigtimedwait(const sigset_t *set, siginfo_t *info,
const struct timespec * timeout)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
struct pthread *curthread = _get_curthread();
sigset_t newset;
int ret;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_cancel_enter(curthread);
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
ret = __sys_sigtimedwait(thr_remove_thr_signals(set, &newset), info,
timeout);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_cancel_leave(curthread, (ret == -1));
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
return (ret);
}
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
int
_sigwaitinfo(const sigset_t *set, siginfo_t *info)
{
sigset_t newset;
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
return (__sys_sigwaitinfo(thr_remove_thr_signals(set, &newset), info));
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
}
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
/*
* Cancellation behavior:
* Thread may be canceled at start, if thread got signal,
* it is not canceled.
*/
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigwaitinfo(const sigset_t *set, siginfo_t *info)
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
{
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
struct pthread *curthread = _get_curthread();
sigset_t newset;
int ret;
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_cancel_enter(curthread);
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
ret = __sys_sigwaitinfo(thr_remove_thr_signals(set, &newset), info);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
_thr_cancel_leave(curthread, ret == -1);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
return (ret);
}
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
int
_sigwait(const sigset_t *set, int *sig)
{
sigset_t newset;
Reduce redundant code. Submitted by: kib
2010-08-20 13:42:48 +00:00
return (__sys_sigwait(thr_remove_thr_signals(set, &newset), sig));
1. Don't override underscore version of aio_suspend(), system(), wait(), waitpid() and usleep(), they are internal versions and should not be cancellation points. 2. Make wait3() as a cancellation point. 3. Move raise() and pause() into file thr_sig.c. 4. Add functions _sigsuspend, _sigwait, _sigtimedwait and _sigwaitinfo, remove SIGCANCEL bit in wait-set for those functions, the signal is used internally to implement thread cancellation.
2006-07-25 12:50:05 +00:00
}
In current implementation, thread cancellation is done in signal handler, which does not know what is the state of interrupted system call, for example, open() system call opened a file and the thread is still cancelled, result is descriptor leak, there are other problems which can cause resource leak or undeterminable side effect when a thread is cancelled. However, this is no longer true in new implementation. In defering mode, a thread is canceled if cancellation request is pending and later the thread enters a cancellation point, otherwise, a later pthread_cancel() just causes SIGCANCEL to be sent to the target thread, and causes target thread to abort system call, userland code in libthr then checks cancellation state, and cancels the thread if needed. For example, the cancellation point open(), the thread may be canceled at start, but later, if it opened a file descriptor, it is not canceled, this avoids file handle leak. Another example is read(), a thread may be canceled at start of the function, but later, if it read some bytes from a socket, the thread is not canceled, the caller then can decide if it should still enable cancelling or disable it and continue reading data until it thinks it has read all bytes of a packet, and keeps a protocol stream in health state, if user ignores partly reading of a packet without disabling cancellation, then second iteration of read loop cause the thread to be cancelled. An exception is that the close() cancellation point always closes a file handle despite whether the thread is cancelled or not. The old mechanism is still kept, for a functions which is not so easily to fix a cancellation problem, the rough mechanism is used. Reviewed by: kib@
2010-08-20 05:15:39 +00:00
/*
* Cancellation behavior:
* Thread may be canceled at start, if thread got signal,
* it is not canceled.
*/
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_sigwait(const sigset_t *set, int *sig)
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
{
struct pthread *curthread = _get_curthread();
sigset_t newset;
int ret;
Because POSIX does not allow EINTR to be returned from sigwait(), add a wrapper for it in libc and rework the code in libthr, the system call still can return EINTR, we keep this feature. Discussed on: thread Reviewed by: jilles
2010-09-10 01:47:37 +00:00
do {
_thr_cancel_enter(curthread);
ret = __sys_sigwait(thr_remove_thr_signals(set, &newset), sig);
_thr_cancel_leave(curthread, (ret != 0));
} while (ret == EINTR);
Import my recent 1:1 threading working. some features improved includes: 1. fast simple type mutex. 2. __thread tls works. 3. asynchronous cancellation works ( using signal ). 4. thread synchronization is fully based on umtx, mainly, condition variable and other synchronization objects were rewritten by using umtx directly. those objects can be shared between processes via shared memory, it has to change ABI which does not happen yet. 5. default stack size is increased to 1M on 32 bits platform, 2M for 64 bits platform. As the result, some mysql super-smack benchmarks show performance is improved massivly. Okayed by: jeff, mtm, rwatson, scottl
2005-04-02 01:20:00 +00:00
return (ret);
- Add libthr but don't hook it up to the regular build yet. This is an adaptation of libc_r for the thr system call interface. This is beta quality code.
2003-04-01 03:46:29 +00:00
}
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_setcontext(const ucontext_t *ucp)
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
{
ucontext_t uc;
Fix return value for setcontext and swapcontext.
2013-05-09 04:41:03 +00:00
if (ucp == NULL) {
errno = EINVAL;
return (-1);
}
Avoid copying memory if SIGCANCEL is not masked.
2013-04-18 05:56:00 +00:00
if (!SIGISMEMBER(uc.uc_sigmask, SIGCANCEL))
return __sys_setcontext(ucp);
Add signal handler wrapper, the reason to add it becauses there are some cases we want to improve: 1) if a thread signal got a signal while in cancellation point, it is possible the TDP_WAKEUP may be eaten by signal handler if the handler called some interruptibly system calls. 2) In signal handler, we want to disable cancellation. 3) When thread holding some low level locks, it is better to disable signal, those code need not to worry reentrancy, sigprocmask system call is avoided because it is a bit expensive. The signal handler wrapper works in this way: 1) libthr installs its signal handler if user code invokes sigaction to install its handler, the user handler is recorded in internal array. 2) when a signal is delivered, libthr's signal handler is invoke, libthr checks if thread holds some low level lock or is in critical region, if it is true, the signal is buffered, and all signals are masked, once the thread leaves critical region, correct signal mask is restored and buffered signal is processed. 3) before user signal handler is invoked, cancellation is temporarily disabled, after user signal handler is returned, cancellation state is restored, and pending cancellation is rescheduled.
2010-09-01 02:18:33 +00:00
(void) memcpy(&uc, ucp, sizeof(uc));
Avoid copying memory if SIGCANCEL is not masked.
2013-04-18 05:56:00 +00:00
SIGDELSET(uc.uc_sigmask, SIGCANCEL);
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
return (__sys_setcontext(&uc));
Add wrapper for setcontext() and swapcontext(), the wrappers unblock SIGCANCEL which is needed by thread cancellation.
2010-08-24 09:57:06 +00:00
}
Revert revision 249323, the PR/177624 is confusing, that bug is caused by using buggy getcontext/setcontext on same stack, while swapcontext normally works on different stack, there is no such a problem.
2013-04-18 05:12:11 +00:00
int
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
__thr_swapcontext(ucontext_t *oucp, const ucontext_t *ucp)
Revert revision 249323, the PR/177624 is confusing, that bug is caused by using buggy getcontext/setcontext on same stack, while swapcontext normally works on different stack, there is no such a problem.
2013-04-18 05:12:11 +00:00
{
ucontext_t uc;
Fix return value for setcontext and swapcontext.
2013-05-09 04:41:03 +00:00
if (oucp == NULL || ucp == NULL) {
errno = EINVAL;
return (-1);
}
Avoid copying memory if SIGCANCEL is not masked.
2013-04-18 05:56:00 +00:00
if (SIGISMEMBER(ucp->uc_sigmask, SIGCANCEL)) {
(void) memcpy(&uc, ucp, sizeof(uc));
SIGDELSET(uc.uc_sigmask, SIGCANCEL);
ucp = &uc;
}
Fix known issues which blow up the process after dlopen("libthr.so") (or loading a dso linked to libthr.so into process which was not linked against threading library). - Remove libthr interposers of the libc functions, including __error(). Instead, functions calls are indirected through the interposing table, similar to how pthread stubs in libc are already done. Libc by default points either to syscall trampolines or to existing libc implementations. On libthr load, libthr rewrites the pointers to the cancellable implementations already in libthr. The interposition table is separate from pthreads stubs indirection table to not pull pthreads stubs into static binaries. - Postpone the malloc(3) internal mutexes initialization until libthr is loaded. This avoids recursion between calloc(3) and static pthread_mutex_t initialization. - Reinstall signal handlers with wrapper on libthr load. The _rtld_is_dlopened(3) is used to avoid useless calls to sigaction(2) when libthr is statically referenced from the main binary. In the process, fix openat(2), swapcontext(2) and setcontext(2) interposing. The libc symbols were exported at different versions than libthr interposers. Export both libc and libthr versions from libc now, with default set to the higher version from libthr. Remove unused and disconnected swapcontext(3) userspace implementation from libc/gen. No objections from: deischen Tested by: pho, antoine (exp-run) (previous versions) Sponsored by: The FreeBSD Foundation MFC after: 1 week
2015-01-03 18:38:46 +00:00
return (__sys_swapcontext(oucp, ucp));
Revert revision 249323, the PR/177624 is confusing, that bug is caused by using buggy getcontext/setcontext on same stack, while swapcontext normally works on different stack, there is no such a problem.
2013-04-18 05:12:11 +00:00
}
Reference in New Issue Copy Permalink