_lwp_park(2) [netbsd man page]

_LWP_PARK(2)						      BSD System Calls Manual						      _LWP_PARK(2)

NAME

     _lwp_park -- wait interruptably in the kernel

LIBRARY

     Standard C Library (libc, -lc)

SYNOPSIS

     #include <lwp.h>

     int
     _lwp_park(const struct timespec *abstime, lwpid_t unpark, const void *hint, const void *unparkhint);

DESCRIPTION

     _lwp_park() can be used to synchronize access to resources among multiple light-weight processes.	It causes the calling LWP to wait inter-
     ruptably in the kernel, until one of the following conditions is met:

     o	 The abstime argument is non-NULL, and the absolute UTC time it specifies has passed.

     o	 The LWP receives a directed signal posted using _lwp_kill(), or is elected to handle a signal on behalf of its containing process.

     o	 The LWP is awoken by another LWP in the same process that has made a call to _lwp_wakeup().

     o	 The LWP is awoken by another LWP in the same process that has made a call to _lwp_unpark() or _lwp_unpark_all().

     The preferred method to awaken an LWP sleeping as a result of a call to _lwp_park() is to make a call to _lwp_unpark(), or _lwp_unpark_all().
     The _lwp_wakeup() system call is a more general facility, and requires more resources to execute.

     The optional hint argument specifies the address of object upon which the LWP is synchronizing.  When the hint value is matched between calls
     to _lwp_park() and _lwp_unpark() or _lwp_unpark_all(), it may reduce the time necessary for the system to resume execution of waiting LWPs.

     The unpark and unparkhint arguments can be used to fold a park operation and unpark operation into a single system call.  If unpark is non-
     zero, the system will behave as if the following call had been made before the calling thread begins to wait:

	     _lwp_unpark(unpark, unparkhint);

RETURN VALUES

     _lwp_park() may return a value of 0.  Otherwise, -1 is returned and errno is set to provide more information.

ERRORS

     [EALREADY]
	       A request was made to wake the LWP before it began to wait in the kernel.

     [EINTR]   The LWP has been awoken by a signal or by a call to one of the following functions: _lwp_unpark(), _lwp_unpark_all(),
	       _lwp_wakeup().

     [EINVAL]  The time value specified by abstime is invalid.

     [ESRCH]   No LWP can be found in the current process corresponding to unpark.

     [ETIMEDOUT]
	       The UTC time specified by abstime has passed.

SEE ALSO

     _lwp_unpark(2), _lwp_unpark_all(2), _lwp_wakeup(2)

HISTORY

     The _lwp_park() system call first appeared in NetBSD 5.0.

BSD
								September 25, 2007							       BSD

ttrace_wait(2)							System Calls Manual						    ttrace_wait(2)

NAME

       ttrace_wait - wait for ttrace event

SYNOPSIS

DESCRIPTION

       The system call provides a means to wait for a event to occur.  A tracing process (debugger) will normally invoke after a process or any of
       its threads has been set running.

       synchronizes tracing requests directed at threads within the traced process.  This mechanism differs from the process-oriented synchroniza-
       tion provided by or (see wait(2)).

       The  pid  argument identifies the process-id of a traced process which the debugger expects to stop.  If pid is a positive value, and lwpid
       is zero, then will wait for any thread in the traced process identified by pid to stop in response to an  outstanding  ttrace  event.   The
       information concerning the thread that hit the event point is available in the ttstate_t structure (see ttrace(2)).

       The  lwpid  argument  identifies the Lightweight Process (LWP) id of a thread in the traced process pid for which the debugger must wait to
       validate request completion.  If both pid and lwpid are non-zero values, suspends the calling process until the specified LWP in the traced
       process stops.

       When  multiple  child  processes  are simultaneously traced, can be used to identify the process-id and LWP id of a thread which stopped in
       response to any outstanding request established for the group of traced child processes.  This is achieved by invoking with  both  pid  and
       lwpid set to 0 (zero).

       A zero pid and non-zero lwpid will return an error.

       The  option  argument  must specify either or These values control the synchronizing effect of on the calling process.  The value causes to
       behave in non-blocking mode and return to the calling process immediately whether or not a pre-existing ttrace request completed on  behalf
       of the tracing process.	With suspends the calling process until the requested pid and/or LWP stop.

       As  mentioned  above, the tsp argument references a ttstate_t structure (see ttrace(2)) which provides all the needed information regarding
       the stopped thread.  The size argument specifies the size of the ttstate_t structure referenced by addr.

RETURN VALUE

       If the call succeeds, will return 1 (one) if the event was never waited for, 0 (zero) otherwise.  If the call fails, -1 is returned and	is
       set to the appropriate value.

ERRORS

       The system call fails if one or more of the following is true:

	      [EINVAL]	  pid is zero and lwpid is non-zero.

	      [EINVAL]	  The option is invalid.

	      [EINVAL]	  The lwpid is not controlled by process pid.

	      [ESRCH]	  The pid or lwpid do not identify an existing process (LWP).

	      [EACCES]	  The pid does not identify a process debugged by the invoking process.

	      [ECHILD]	  The process (LWP) died while it was waited for.

	      [EINTR]	  was interrupted by a signal.

	      [EFAULT]	  An invalid address was given for the kernel to write data into.

AUTHOR

       was developed by HP.

SEE ALSO

       ttrace(2), wait(2).

																    ttrace_wait(2)