State Threads Library Reference

Types
st_thread_t
st_cond_t
st_mutex_t
st_utime_t
st_netfd_t
st_switch_cb_t

Error Handling

Library Initialization

st_init()
st_getfdlimit()
st_set_eventsys()
st_get_eventsys()
st_get_eventsys_name()
st_set_utime_function()
st_timecache_set()
st_randomize_stacks()

st_switch_cb_t type
st_set_switch_in_cb()
st_set_switch_out_cb()

Thread Control and Identification

st_thread_t type
st_thread_create()
st_thread_exit()
st_thread_join()
st_thread_self()
st_thread_interrupt()
st_sleep()
st_usleep()
st_randomize_stacks()

Per-Thread Private Data

st_key_create()
st_key_getlimit()
st_thread_setspecific()
st_thread_getspecific()

Synchronization

st_cond_t type
st_cond_new()
st_cond_destroy()
st_cond_wait()
st_cond_timedwait()
st_cond_signal()
st_cond_broadcast()

st_mutex_t type
st_mutex_new()
st_mutex_destroy()
st_mutex_lock()
st_mutex_trylock()
st_mutex_unlock()

Timing

st_utime_t type
st_utime()
st_set_utime_function()
st_timecache_set()
st_time()

I/O Functions

st_netfd_t type
st_netfd_open()
st_netfd_open_socket()
st_netfd_free()
st_netfd_close()
st_netfd_fileno()
st_netfd_setspecific()
st_netfd_getspecific()
st_netfd_serialize_accept()
st_netfd_poll()

st_accept()
st_connect()
st_read()
st_read_fully()
st_read_resid()
st_readv()
st_readv_resid()
st_write()
st_write_resid()
st_writev()
st_writev_resid()
st_recvfrom()
st_sendto()
st_recvmsg()
st_sendmsg()

st_open()
st_poll()

Program Structure

List of Blocking Functions


Types

The State Thread library defines the following types in the st.h header file:

st_thread_t
st_cond_t
st_mutex_t
st_utime_t
st_netfd_t


st_thread_t

Thread type.

Syntax
#include <st.h>

typedef void *  st_thread_t;

Description
A thread is represented and identified by a pointer to an opaque data structure. This pointer is a required parameter for most of the functions that operate on threads.

The thread identifier remains valid until the thread returns from its root function and, if the thread was created joinable, is joined.


st_cond_t

Condition variable type.

Syntax
#include <st.h>

typedef void *  st_cond_t;

Description
A condition variable is an opaque object identified by a pointer. Condition variables provide synchronization primitives to wait for or wake up threads waiting for certain conditions to be satisfied.

In the State Threads library there is no need to lock a mutex before waiting on a condition variable.


st_mutex_t

Mutex type.

Syntax
#include <st.h>

typedef void *  st_mutex_t;

Description
A mutex is an opaque object identified by a pointer. Mutual exclusion locks (mutexes) are used to serialize the execution of threads through critical sections of code.

If application using the State Threads library is written with no I/O or control yielding in critical sections (that is no blocking functions in critical sections), then there is no need for mutexes.

These mutexes can only be used for intra-process thread synchronization. They cannot be used for inter-process synchronization.


st_utime_t

High resolution time type ("u" stands for "micro").

Syntax
#include <st.h>

typedef unsigned long long  st_utime_t;

Description
This datatype (unsigned 64-bit integer) represents high-resolution real time expressed in microseconds since some arbitrary time in the past. It is not correlated in any way to the time of day.


st_netfd_t

File descriptor type.

Syntax
#include <st.h>

typedef void *  st_netfd_t;

Description
This datatype typically represents any open end point of network communication (socket, end point of a pipe, FIFO, etc.) but can encapsulate any open file descriptor. Objects of this type are identified by a pointer to an opaque data structure.


st_switch_cb_t

Context switch callback function type.

Syntax
#include <st.h>

typedef void (*st_switch_cb_t)(void);

Description
This datatype is a convenience type for describing a pointer to a function that will be called when a thread is set to stop or set to run. This feature is available only when ST_SWITCH_CB is defined in <st.h>.


Error Handling

All State Threads library non-void functions return on success either a non-negative integer or a pointer to a newly created object (constructor-type functions). On failure they return either -1 or a NULL pointer respectively and set global errno to indicate the error. It is safe to use errno because it is set right before the function return and only one thread at a time can modify its value.

The perror(3) function can be used to produce an error message on the standard error output.


Library Initialization

st_init()
st_getfdlimit()
st_set_eventsys()
st_get_eventsys()
st_get_eventsys_name()

These functions operate on a callback function of type st_switch_cb_t:

st_set_switch_in_cb()
st_set_switch_out_cb()


st_init()

Initializes the runtime.

Syntax
#include <st.h>

int st_init(void);

Parameters
None.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error.

Description
This function initializes the library runtime. It should be called near the beginning of the application's main() function before any other State Threads library function is called.

Among other things, this function limits the number of open file descriptors to the OS imposed per-process maximum number or, if select(2) is used, to FD_SETSIZE, whichever is less (getrlimit(2)). This limit can be retrieved by st_getfdlimit(). It also sets the disposition of the SIGPIPE signal to SIG_IGN (to be ignored) (signal(5)).

Unlike POSIX threads, a new process created by the fork(2) system call is an exact copy of the calling process and all state threads which are running in the parent do exist in the child. That means that st_init() may be called either before or after multiple processes are created by fork(2).

If the library runtime is not properly initialized (e.g., st_init() is accidentally omitted), then the process will receive either an arithmetic exception (SIGFPE or SIGTRAP) or segmentation fault (SIGSEGV) signal upon new thread creation or the first context switch, respectively.


st_getfdlimit()

Returns the maximum number of file descriptors that the calling process can open.

Syntax
#include <st.h>

int st_getfdlimit(void);

Parameters
None.

Returns
The maximum number of file descriptors that the calling process can open. If this function is called before the library is successfully initialized by st_init(), a value of -1 is returned.

Description
This function returns the limit on the number of open file descriptors which is set by the st_init() function.


st_set_eventsys()

Sets event notification mechanism.

Syntax
#include <st.h>

int st_set_eventsys(int eventsys);

Parameters
st_set_eventsys() has the following parameter:

eventsys

An integer value identifying selected event notification mechanism. The following values are defined in the st.h header file:

ST_EVENTSYS_DEFAULT Use default event notification mechanism. Usually it's select(2) but if the library was compiled with the USE_POLL macro defined then the default is poll(2).
ST_EVENTSYS_SELECT Use select(2) as an event notification mechanism.
ST_EVENTSYS_POLL Use poll(2) as an event notification mechanism.
ST_EVENTSYS_ALT Use an alternative event notification mechanism. The actual mechanism selected depends on OS support. For example, epoll(4) will be used on Linux if supported and kqueue(2) will be used on FreeBSD/OpenBSD. If the OS supports no alternative event notification mechanism, setting ST_EVENTSYS_ALT has no effect and the ST_EVENTSYS_DEFAULT mechanism will be used.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINVAL The supplied eventsys parameter has an invalid value.
EBUSY The event notification mechanism has already been set.

Description
This function sets the event notification mechanism that will be used by the State Threads library. To have any effect, it must be called before the st_init() function which performs the actual initialization. If st_set_eventsys() is not called, st_init() will set the ST_EVENTSYS_DEFAULT mechanism. The mechanism cannot be changed once set.

There are no strict rules for selecting an event notification mechanism. The "best" one depends on how your application behaves. Try a few to see which one works best for you. As a rule of thumb, you should use the ST_EVENTSYS_ALT mechanism if your application deals with a very large number of network connections of which only a few are active at once.


st_get_eventsys()

Returns the integer value identifying the event notification mechanism being used by the State Threads library.

Syntax
#include <st.h>

int st_get_eventsys(void);

Parameters
None.

Returns
The integer value identifying the current event notification mechanism. This value can be one of the following (see st_set_eventsys()): ST_EVENTSYS_SELECT, ST_EVENTSYS_POLL, or ST_EVENTSYS_ALT. Future versions of the library may return other values. If a mechanism hasn't been set yet, a value of -1 is returned.

Description
This function returns the integer value identifying the event notification mechanism which is actually being used by the State Threads library.


st_get_eventsys_name()

Returns the name of the event notification mechanism being used by the State Threads library.

Syntax
#include <st.h>

const char *st_get_eventsys_name(void);

Parameters
None.

Returns
The string identifying the current event notification mechanism. If a mechanism hasn't been set yet (see st_set_eventsys()), an empty string is returned. Possible return values are "select", "poll", "kqueue", or "epoll". Future versions of the library may return other values.

Description
This function returns the string identifying the event notification mechanism which is actually being used by the State Threads library.


st_set_switch_in_cb()

st_set_switch_out_cb()

Set the optional callback function for thread switches.

Syntax
#include <st.h>

st_switch_cb_t st_set_switch_in_cb(st_switch_cb_t cb);
st_switch_cb_t st_set_switch_out_cb(st_switch_cb_t cb);

Parameters
st_set_switch_in_cb() and st_set_switch_out_cb() have the following parameter:

cb

A function to be called when a thread is resumed and stopped respectively.

Returns
The previous callback function pointer.

Description
These functions set the callback for when a thread is resumed and stopped respectively. After being called any thread switch will call the callback. Use a NULL pointer to disable the callback (this is the default). Use st_thread_self() or thread specific data to differentiate between threads.

These functions can be called at any time.

This feature is available only when ST_SWITCH_CB is defined in <st.h>.


Thread Control and Identification

These functions operate on a thread object of type st_thread_t.

st_thread_create()
st_thread_exit()
st_thread_join()
st_thread_self()
st_thread_interrupt()
st_sleep()
st_usleep()
st_randomize_stacks()


st_thread_create()

Creates a new thread.

Syntax
#include <st.h>

st_thread_t st_thread_create(void *(*start)(void *arg), void *arg,
                             int joinable, int stack_size);

Parameters
st_thread_create() has the following parameters:

start

A pointer to the thread's start function, which is called as the root of the new thread. Return from this function terminates a thread.

arg

A pointer to the root function's only parameter.

joinable

Specifies whether the thread is joinable or unjoinable. If this parameter is zero, the thread is unjoinable. Otherwise, it is joinable. See also st_thread_join().

stack_size

Specifies your preference for the size of the stack, in bytes, associated with the newly created thread. If you pass zero in this parameter, the default stack size will be used. The default stack size is 128 KB on IA-64 and 64 KB on all other platforms. On IA-64 only a half of stack_size bytes is used for the memory stack. The other half is used for the register stack backing store.

Returns
Upon successful completion, a new thread identifier is returned (this identifier remains valid until the thread returns from its start function). Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new thread. Note that the total number of threads created by the application is limited by the amount of swap space available. Upon thread creation, stack_size bytes are reserved on the swap space. The stack pages are not actually used (valid) until touched by the application.


st_thread_exit()

Terminates the calling thread.

Syntax
#include <st.h>

void st_thread_exit(void *retval);

Parameters
st_thread_exit() has the following parameters:

retval

If the thread is joinable, then the value retval may be retrieved by st_thread_join(). If a thread returns from its start function, it acts as if it had called st_thread_exit() with retval as the value returned.

Returns
Nothing.

Description
This function terminates the calling thread. When a thread exits, per-thread private data is destroyed by invoking the destructor function for any non-NULL thread specific values associated with active keys (see st_key_create()). This function is implicitly called when a thread returns from its start function.

When the last thread terminates the process exits with a zero status value.


st_thread_join()

Blocks the calling thread until a specified thread terminates.

Syntax
#include <st.h>

int st_thread_join(st_thread_t thread, void **retvalp);

Parameters
st_thread_join() has the following parameters:

thread

A valid identifier for the thread that is to be joined.

retvalp

If this parameter is not NULL, then the exit value of the thread will be placed in the location referenced by this parameter (see st_thread_exit()).

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINVALTarget thread is unjoinable.
EINVALOther thread already waits on the same joinable thread.
EDEADLKTarget thread is the same as the calling thread.
EINTRCurrent thread was interrupted by st_thread_interrupt().

Description
This function is used to synchronize the termination of a thread and possibly retrieve its exit value. Several threads cannot wait for the same thread to complete - one of the calling threads operates successfully, and the others terminate with the error. The calling thread is not blocked if the target thread has already terminated.


st_thread_self()

Identifies the calling thread.

Syntax
#include <st.h>

st_thread_t st_thread_self(void);

Parameters
None.

Returns
Always returns a valid reference to the calling thread - a self-identity.

Description
This function identifies the calling thread. This is the same identifier that the creating thread obtains from st_thread_create().


st_thread_interrupt()

Interrupts a target thread.

Syntax
#include <st.h>

void st_thread_interrupt(st_thread_t thread);

Parameters
st_thread_interrupt() has the following parameters:

thread

A valid identifier for the thread being interrupted.

Returns
Nothing.

Description
This function interrupts (unblocks) a target thread that is blocked in one of the blocking functions. A function that was interrupted returns an error and sets errno to EINTR. It is up to the target thread to act upon an interrupt (e.g., it may exit or just abort the current transaction).

Note: State Threads library functions are never interrupted by a caught signal. A blocking library function returns an error and sets errno to EINTR only if the current thread was interrupted via st_thread_interrupt().

If a target thread is already runnable or running (e.g., it is a newly created thread or calling thread itself), this function will prevent it from subsequent blocking. In other words, the interrupt will be "delivered" only when a target thread is about to block.


st_sleep(), st_usleep()

Suspends current thread for a specified amount of time.

Syntax
#include <st.h>

int st_sleep(int secs);

int st_usleep(st_utime_t usecs);

Parameters
st_sleep() has the following parameters:

secs

The number of seconds you want the thread to sleep for.

st_usleep() has the following parameters:

usecs

The number of microseconds you want the thread to sleep for. This parameter is a variable of type st_utime_t.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINTRThe current thread was interrupted by st_thread_interrupt().

Description
These functions suspend the calling thread from execution for a specified number of seconds (st_sleep()) or microseconds (st_usleep()).

If zero is passed as a parameter to st_sleep(), or ST_UTIME_NO_WAIT (0) is passed to st_usleep(), the calling thread yields, thus potentially allowing another thread to run.

If -1 is passed as a parameter to st_sleep(), or ST_UTIME_NO_TIMEOUT (-1) is passed to st_usleep(), the calling thread will be suspended permanently. It can be resumed again by interrupting it via st_thread_interrupt().


st_randomize_stacks()

Turns stack base address randomization on or off.

Syntax
#include <st.h>

int st_randomize_stacks(int on);

Parameters
st_randomize_stacks() has the following parameters:

on

If this parameter has a non-zero value, the State Threads library randomizes the base addresses of stacks allocated for threads created after this call. Otherwise new threads' stacks are typically page aligned.

Returns
The previous state of stack randomization (a value of 0 if it was off and a non-zero value otherwise).

Description
Randomizing state threads' stack bases may improve cache performance on some systems when large numbers of state threads all perform roughly the same work, as when they all start from the same root function. On many modern systems the performance increase is negligible. You should compare your application's performance with this feature on and off to see if you really need it.

When randomization is enabled, new stacks are allocated one page larger to accomodate the randomization.

This call affects only threads created afterward. It has no effect on existing threads.


Per-Thread Private Data

These functions allow to associate private data with each of the threads in a process.

st_key_create()
st_key_getlimit()
st_thread_setspecific()
st_thread_getspecific()


st_key_create()

Creates a key (non-negative integer) that can be used by all threads in the process to get and set thread-specific data.

Syntax
#include <st.h>

int st_key_create(int *keyp, void (*destructor)(void *));

Parameters
st_key_create() has the following parameters:

keyp

The newly created key is returned in the memory pointed to by this parameter. The new key can be used with st_thread_setspecific() and st_thread_getspecific().

destructor

Specifies an optional destructor function for the private data associated with the key. This function can be specified as NULL. Upon thread exit (see st_thread_exit()), if a key has a non-NULL destructor and has a non-NULL value associated with that key, then the destructor function will be called with the associated value.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EAGAINThe limit on the total number of keys per process has been exceeded (see st_key_getlimit()).

Description
If this function is successful, every thread in the same process is capable of associating private data with the new key. After a new key is created, all active threads have the value NULL associated with that key. After a new thread is created, the value NULL is associated with all keys for that thread. If a non-NULL destructor function is registered with a new key, it will be called at one of two times, as long as the private data is not NULL:

The key maintains independent data values for each binding thread. A thread can get access only to its own thread-specific data. There is no way to deallocate a private data key once it is allocated.


st_key_getlimit()

Returns the key limit.

Syntax
#include <st.h>

int st_key_getlimit(void);

Parameters
None.

Returns
The limit on the total number of keys per process.

Description
This function can be used to obtain the limit on the total number of keys per process (see st_key_create()).


st_thread_setspecific()

Sets per-thread private data.

Syntax
#include <st.h>

int st_thread_setspecific(int key, void *value);

Parameters
st_thread_setspecific() has the following parameters:

key

This parameter represents a key with which thread-specific data is associated.

value

The per-thread private data, or more likely, a pointer to the data which is associated with key.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINVALThe specified key is invalid.

Description
This function associates a thread-specific value with key. Different threads may bind different values to the same key.

If the thread already has non-NULL private data associated with key, and if the destructor function for that key is not NULL, this destructor function will be called before setting the new data value.


st_thread_getspecific()

Retrieves the per-thread private data for the current thread.

Syntax
#include <st.h>

void *st_thread_getspecific(int key);

Parameters
st_thread_getspecific() has the following parameters:

key

This parameter represents a key with which thread-specific data is associated.

Returns
The thread-specific data associated with key. If no data is associated with key, then NULL is returned.

Description
This function returns the calling thread's value that is bound to the specified key (see st_thread_setspecific()).


Synchronization

These functions operate on condition variables and mutual exclusion locks (mutexes).

Functions are provided to wait on a condition variable and to wake up (signal) threads that are waiting on the condition variable.

st_cond_new()
st_cond_destroy()
st_cond_wait()
st_cond_timedwait()
st_cond_signal()
st_cond_broadcast()

st_mutex_new()
st_mutex_destroy()
st_mutex_lock()
st_mutex_trylock()
st_mutex_unlock()


st_cond_new()

Creates a new condition variable.

Syntax
#include <st.h>

st_cond_t st_cond_new(void);

Parameters
None.

Returns
Upon successful completion, a new condition variable identifier is returned. Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new condition variable.


st_cond_destroy()

Destroys a condition variable.

Syntax
#include <st.h>

int st_cond_destroy(st_cond_t cvar);

Parameters
st_cond_destroy() has the following parameters:

cvar

An identifier of the condition variable object to be destroyed.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EBUSYThe condition variable is currently being used by one or more threads.

Description
This function destroys a condition variable. The caller is responsible for ensuring that the condition variable is no longer in use.


st_cond_wait()

Waits on a condition.

Syntax
#include <st.h>

int st_cond_wait(st_cond_t cvar);

Parameters
st_cond_wait() has the following parameters:

cvar

The condition variable on which to wait.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINTRThe current thread was interrupted by st_thread_interrupt().

Description
This function is used to block on a condition variable. A return from this function does not guarantee that the condition or event for which the caller was waiting actually occurred. It is the responsibility of the caller to recheck the condition wait predicate before proceeding.

Note: The State Threads library scheduling guarantees that the condition cannot change between the checking and blocking, therefore there is no need for mutex protection. You must not call any blocking functions between the condition checking and the st_cond_wait() call.


st_cond_timedwait()

Waits on a condition.

Syntax
#include <st.h>

int st_cond_timedwait(st_cond_t cvar, st_utime_t timeout);

Parameters
st_cond_timedwait() has the following parameters:

cvar

The condition variable on which to wait.

timeout

If the number of microseconds specified by this parameter passes before the waiting thread is signalled, an error is returned. This parameter is a variable of type st_utime_t. Note that this time value is a time delta; it is not an absolute time. Also note that timeouts are measured since the last context switch.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred before the thread was awakened by st_cond_signal() or st_cond_broadcast().

Description
This function works the same way as st_cond_wait(), except that an error is returned if the number of microseconds specified by timeout passes before the waiting thread is signalled.


st_cond_signal()

Unblocks a thread waiting on a condition variable.

Syntax
#include <st.h>

int st_cond_signal(st_cond_t cvar);

Parameters
st_cond_signal() has the following parameters:

cvar

The condition variable to signal.

Returns
Always zero.

Description
This function unblocks (signals) one of the threads that are blocked on cvar at the time of the call. If no thread is waiting on the condition variable, the signal operation is a no-op.


st_cond_broadcast()

Unblocks all threads waiting on a condition variable.

Syntax
#include <st.h>

int st_cond_broadcast(st_cond_t cvar);

Parameters
st_cond_broadcast() has the following parameters:

cvar

The condition variable to broadcast.

Returns
Always zero.

Description
This function unblocks all threads blocked on the specified condition variable at the time of the call. If no threads are waiting, this operation is a no-op.


st_mutex_new()

Creates a new mutual exclusion lock (mutex).

Syntax
#include <st.h>

st_mutex_t st_mutex_new(void);

Parameters
None.

Returns
Upon successful completion, a new mutex identifier is returned. Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new opaque mutual exclusion lock (see st_mutex_t).


st_mutex_destroy()

Destroys a specified mutex object.

Syntax
#include <st.h>

int st_mutex_destroy(st_mutex_t lock);

Parameters
st_mutex_destroy() has the following parameters:

lock

An identifier of the mutex object to be destroyed.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EBUSYThe mutex is currently being used by other threads.

Description
This function destroys a mutex. The caller is responsible for ensuring that the mutex is no longer in use.


st_mutex_lock()

Locks a specified mutex object.

Syntax
#include <st.h>

int st_mutex_lock(st_mutex_t lock);

Parameters
st_mutex_lock() has the following parameters:

lock

An identifier of the mutex object to be locked.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EDEADLKThe current thread already owns the mutex.
EINTRThe current thread was interrupted by st_thread_interrupt().

Description
A thread that calls this function will block until it can gain exclusive ownership of a mutex, and retains ownership until it calls st_mutex_unlock().


st_mutex_trylock()

Attempts to acquire a mutex.

Syntax
#include <st.h>

int st_mutex_trylock(st_mutex_t lock);

Parameters
st_mutex_trylock() has the following parameters:

lock

An identifier of the mutex object to be locked.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EBUSYThe mutex is currently held by another thread.

Description
This function attempts to acquire a mutex. If the mutex object is locked (by any thread, including the current thread), the call returns immediately with an error.


st_mutex_unlock()

Releases a specified mutex object.

Syntax
#include <st.h>

int st_mutex_unlock(st_mutex_t lock);

Parameters
st_mutex_unlock() has the following parameters:

lock

An identifier of the mutex object to be unlocked.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EPERMThe current thread does not own the mutex.

Description
This function releases a specified mutex object previously acquired by st_mutex_lock() or st_mutex_trylock(). Only the thread that locked a mutex should unlock it.


Timing

st_utime()
st_set_utime_function()
st_timecache_set()
st_time()


st_utime()

Returns current high-resolution time.

Syntax
#include <st.h>

st_utime_t st_utime(void);

Parameters
None.

Returns
Current high-resolution time value of type st_utime_t.

Description
This function returns the current high-resolution time. Time is expressed as microseconds since some arbitrary time in the past. It is not correlated in any way to the time of day. See also st_utime_t and st_time().


st_set_utime_function()

Set high-resolution time function.

Syntax
#include <st.h>

int st_set_utime_function(st_utime_t (*func)(void));

Parameters
st_set_utime_function() has the following parameters:

func

This function will be called to get high-resolution time instead of the default st_utime() function. It must return number of microseconds since some arbitrary time in the past.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to EINVAL to indicate the error.

Description
This function may be called to replace the default implementation of the st_utime() function. It must be called before the ST library has been initialized (see st_init()). The user-provided function func will be invoked whenever st_utime() is called to obtain current high-resolution time. Replacing default implementation may be useful, for example, for taking advantage of high performance CPU cycle counters.


st_timecache_set()

Turns the time caching on or off.

Syntax
#include <st.h>

int st_timecache_set(int on);

Parameters
st_timecache_set() has the following parameters:

on

If this parameter has a non-zero value, the time caching is turned on (enabled). Otherwise, the time caching is turned off (disabled). By default time caching is disabled.

Returns
The previous state of time caching (a value of 0 if it was off and a value of 1 otherwise).

Description
The State Threads library has the ability to "cache" the time value that is reported by the time(2) system call. If the time caching is enabled by calling this function with a non-zero argument, then the result value of time(2) will be stored and updated at most once per second. The cached time can be retrieved by st_time(). By default time caching is disabled. You may enable or disable time caching at any time but generally you enable it once (if desired) during program initialization.

Note: There are some pathological cases (e.g., very heavy loads during application benchmarking) when a single thread runs for a long time without giving up control and the cached time value is not updated properly. If you always need "real-time" time values, don't enable the time caching.


st_time()

Returns the value of time in seconds since 00:00:00 UTC, January 1, 1970.

Syntax
#include <st.h>

time_t st_time(void);

Parameters
None.

Returns
The value of time in seconds since 00:00:00 UTC, January 1, 1970 as reported by the time(2) system call.

Description
If the time caching was enabled by st_timecache_set(), then this function returns the cached result. Otherwise, it just calls time(2).


I/O Functions

Most State Threads library I/O functions look like corresponding C library functions with two exceptions:

st_netfd_open()
st_netfd_open_socket()
st_netfd_free()
st_netfd_close()
st_netfd_fileno()
st_netfd_setspecific()
st_netfd_getspecific()
st_netfd_serialize_accept()
st_netfd_poll()

st_accept()
st_connect()
st_read()
st_read_fully()
st_read_resid()
st_readv()
st_read_resid()
st_write()
st_write_resid()
st_writev()
st_writev_resid()
st_recvfrom()
st_sendto()
st_recvmsg()
st_sendmsg()
st_open()
st_poll()


st_netfd_open()

Creates a new file descriptor object.

Syntax
#include <st.h>

st_netfd_t st_netfd_open(int osfd);

Parameters
st_netfd_open() has the following parameters:

osfd

Any open OS file descriptor; can be obtained from calls to functions including, but not restricted to, pipe(2), socket(3), socketpair(3), fcntl(2), dup(2), etc.

Returns
Upon successful completion, a new file descriptor object identifier is returned. Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new file descriptor object of type st_netfd_t.

Note: Among other things, this function sets a non-blocking flag on the underlying OS file descriptor. You should not modify this flag directly. Also, once an st_netfd_t has been created with a given file descriptor, you should avoid passing that descriptor to normal I/O or stdio functions. Since the O_NONBLOCK flag is shared across dup(2), this applies to dup()'ed file descriptors as well - for instance, if you pass standard output or standard input to st_netfd_open(), then you should use st_write() instead of write or fprintf when writing to standard error as well - since all three descriptors could point to the same terminal. If necessary, you can still use write directly if you remember to check errno for EAGAIN, but fprintf and other stdio functions should be avoided completely because, at least on Linux, the stdio library cannot be made to work reliably with non-blocking files. (This only applies to file descriptors which are passed to st_netfd_open() or st_netfd_open_socket(), or which are related to such descriptors through dup(); other file descriptors are untouched by State Threads.)


st_netfd_open_socket()

Creates a new file descriptor object from a socket.

Syntax
#include <st.h>

st_netfd_t st_netfd_open_socket(int osfd);

Parameters
st_netfd_open_socket() has the following parameters:

osfd

An open OS file descriptor which is a socket initially obtained from a socket(3) or socketpair(3) call.

Returns
Upon successful completion, a new file descriptor object identifier is returned. Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new file descriptor object of type st_netfd_t which represents an open end point of network communication.

Unlike the st_netfd_open() function which may be used on OS file descriptors of any origin, st_netfd_open_socket() must be used only on sockets. It is slightly more efficient than st_netfd_open().

Note: Among other things, this function sets a non-blocking flag on the underlying OS socket. You should not modify this flag directly. See st_netfd_open().


st_netfd_free()

Frees a file descriptor object without closing the underlying OS file descriptor.

Syntax
#include <st.h>

void st_netfd_free(st_netfd_t fd);

Parameters
st_netfd_free() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

Returns
Nothing.

Description
This function frees the memory and other resources identified by the fd parameter without closing the underlying OS file descriptor. Any non-NULL descriptor-specific data is destroyed by invoking the specified destructor function (see st_netfd_setspecific()).

A thread should not free file descriptor objects that are in use by other threads because it may lead to unpredictable results (e.g., a freed file descriptor may be reused without other threads knowing that).


st_netfd_close()

Closes a file descriptor.

Syntax
#include <st.h>

int st_netfd_close(st_netfd_t fd);

Parameters
st_netfd_close() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error.

Description
This function closes the underlying OS file descriptor, frees the memory and other resources identified by the fd parameter. Any non-NULL descriptor-specific data is destroyed by invoking the specified destructor function (see st_netfd_setspecific()).

A thread should not close file descriptor objects that are in use by other threads because it may lead to unpredictable results (e.g., a closed file descriptor may be reused without other threads knowing that).


st_netfd_fileno()

Returns an underlying OS file descriptor.

Syntax
#include <st.h>

int st_netfd_fileno(st_netfd_t fd);

Parameters
st_netfd_fileno() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

Returns
An underlying OS file descriptor.

Description
This function returns the integer OS file descriptor associated with the named file descriptor object.


st_netfd_setspecific()

Sets per-descriptor private data.

Syntax
#include <st.h>

void st_netfd_setspecific(st_netfd_t fd, void *value,
                          void (*destructor)(void *));

Parameters
st_netfd_setspecific() has the following parameters:

fd

A valid file descriptor object identifier (see st_netfd_t).

value

The per-descriptor private data, or more likely, a pointer to the data which is being associated with the named file descriptor object.

destructor

Specifies an optional destructor function for the private data associated with fd. This function can be specified as NULL. If value is not NULL, then this destructor function will be called with value as an argument upon freeing the file descriptor object (see st_netfd_free() and st_netfd_close()).

Returns
Nothing.

Description
This function allows to associate any data with the specified file descriptor object (network connection). If a non-NULL destructor function is registered, it will be called at one of two times, as long as the associated data is not NULL:


st_netfd_getspecific()

Retrieves the per-descriptor private data.

Syntax
#include <st.h>

void *st_netfd_getspecific(st_netfd_t fd);

Parameters
st_netfd_getspecific() has the following parameters:

fd

A valid file descriptor object identifier (see st_netfd_t).

Returns
The data associated with the named file descriptor object. If no data is associated with fd, then NULL is returned.

Description
This function allows to retrieve the data that was associated with the specified file descriptor object (see st_netfd_setspecific()).


st_netfd_serialize_accept()

Serializes all subsequent accept(3) calls on a specified file descriptor object.

Syntax
#include <st.h>

int st_netfd_serialize_accept(st_netfd_t fd);

Parameters
st_netfd_serialize_accept() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) which has been successfully created from a valid listening socket by st_netfd_open() or st_netfd_open_socket().

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error.

Description
On some platforms (e.g., Solaris 2.5 and possibly other SVR4 implementations) accept(3) calls from different processes on the same listening socket (see bind(3), listen(3)) must be serialized. This function causes all subsequent accept(3) calls made by st_accept() on the specified file descriptor object to be serialized.

st_netfd_serialize_accept() must be called before creating multiple server processes via fork(2). If the application does not create multiple processes to accept network connections on the same listening socket, there is no need to call this function.

Deciding whether or not to serialize accepts is tricky. On some platforms (IRIX, Linux) it's not needed at all and st_netfd_serialize_accept() is a no-op. On other platforms it depends on the version of the OS (Solaris 2.6 doesn't need it but earlier versions do). Serializing accepts does incur a slight performance penalty so you want to enable it only if necessary. Read your system's manual pages for accept(2) and select(2) to see if accept serialization is necessary on your system.

st_netfd_serialize_accept() allocates resources that are freed upon freeing of the specified file descriptor object (see st_netfd_free() and st_netfd_close()).


st_netfd_poll()

Waits for I/O on a single file descriptor object.

Syntax
#include <st.h>

int st_netfd_poll(st_netfd_t fd, int how, st_utime_t timeout);

Parameters
st_netfd_poll() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

how

Specifies I/O events of interest. This parameter can be constructed by OR-ing any combination of the following event flags which are defined in the poll.h header file:

POLLINfd is readable.
POLLOUTfd is is writable.
POLLPRIfd has an exception condition.

timeout

Amount of time in microseconds the call will block waiting for I/O to become ready. This parameter is a variable of type st_utime_t. If this time expires without any I/O becoming ready, st_netfd_poll() returns an error and sets errno to ETIME. Note that timeouts are measured since the last context switch.

Returns
If the named file descriptor object is ready for I/O within the specified amount of time, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error:

EBADFThe underlying OS file descriptor is invalid.
EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred without any I/O becoming ready.

Description
This function returns as soon as I/O is ready on the named file descriptor object or the specified amount of time expires. The how parameter should be set to the I/O events (readable, writable, exception, or some combination) that the caller is interested in. If the value of timeout is ST_UTIME_NO_TIMEOUT (-1), this function blocks until a requested I/O event occurs or until the call is interrupted by st_thread_interrupt().

Despite having an interface like poll(2), this function uses the same event notification mechanism as the rest of the library. For instance if an alternative event nofication mechanism was set using st_set_eventsys(), this function uses that mechanism to check for events.

Note: if kqueue(2) is used as an alternative event notification mechanism (see st_set_eventsys()), the POLLPRI event flag is not supported and st_netfd_poll() will return an error if it's set (errno will be set to EINVAL).


st_accept()

Accepts a connection on a specified file descriptor object.

Syntax
#include <st.h>

st_netfd_t st_accept(st_netfd_t fd, struct sockaddr *addr, int *addrlen,
                     st_utime_t timeout);

Parameters
st_accept() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing the rendezvous socket on which the caller is willing to accept new connections. This object has been created from a valid listening socket by st_netfd_open() or st_netfd_open_socket().

addr

If this value is non-zero, it is a result parameter that is filled in with the address of the connecting entity, as known to the communications layer (see accept(3)).

addrlen

This parameter should initially contain the amount of space pointed to by addr; on return it will contain the actual length (in bytes) of the address returned (see accept(3)).

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the accept operation. Note that timeouts are measured since the last context switch.

Returns
Upon successful completion, a new file descriptor object identifier representing the newly accepted connection is returned. Otherwise, NULL is returned and errno is set to indicate the error. Possible errno values are the same as set by the accept(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no pending connection was accepted.

Description
This function accepts the first connection from the queue of pending connections and creates a new file descriptor object for the newly accepted connection. The rendezvous socket can still be used to accept more connections.

st_accept() blocks the calling thread until either a new connection is successfully accepted or an error occurs. If no pending connection can be accepted before the time limit, this function returns NULL and sets errno to ETIME.


st_connect()

Initiates a connection on a specified file descriptor object.

Syntax
#include <st.h>

int st_connect(st_netfd_t fd, struct sockaddr *addr, int addrlen,
               st_utime_t timeout);

Parameters
st_connect() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing a socket.

addr

A pointer to the address of the peer to which the socket is to be connected.

addrlen

This parameter specifies the amount of space pointed to by addr.

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the connect operation. Note that timeouts are measured since the last context switch.

Returns
Upon successful completion, a value of 0 is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the connect(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and connection setup was not completed.

Description
This function is usually invoked on a file descriptor object representing a TCP socket. Upon completion it establishes a TCP connection to the peer. If the underlying OS socket is not bound, it will be bound to an arbitrary local address (see connect(3)).

st_connect() blocks the calling thread until either the connection is successfully established or an error occurs. If the connection setup cannot complete before the specified time limit, this function fails with errno set to ETIME.


st_read()

Reads data from a specified file descriptor object.

Syntax
#include <st.h>

ssize_t st_read(st_netfd_t fd, void *buf, size_t nbyte, st_utime_t timeout);

Parameters
st_read() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

buf

A pointer to a buffer to hold the data read in. On output the buffer contains the data.

nbyte

The size of buf in bytes.

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the read operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes actually read is returned (a value of 0 means the network connection is closed or end of file is reached). Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the read(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was read.

Description
This function blocks the calling thread until it encounters an end-of-stream indication, some positive number of bytes (but no more than nbyte bytes) are read in, a timeout occurs, or an error occurs.


st_read_fully()

Reads the specified amount of data in full from a file descriptor object.

Syntax
#include <st.h>

ssize_t st_read_fully(st_netfd_t fd, void *buf, size_t nbyte,
                      st_utime_t timeout);

Parameters
st_read_fully() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

buf

A pointer to a buffer to hold the data read in. On output the buffer contains the data.

nbyte

The amount of data to be read in full (in bytes). It must not exceed the size of buf.

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes actually read is returned (a value less than nbyte means the network connection is closed or end of file is reached). Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the read(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

Description
This function blocks the calling thread until the specified amount of data is read in full, it encounters an end-of-stream indication, a timeout occurs, or an error occurs.


st_read_resid()

Reads the specified amount of data in full from a file descriptor object.

Syntax
#include <st.h>

int st_read_resid(st_netfd_t fd, void *buf, size_t *resid,
		  st_utime_t timeout);

Parameters
st_read_resid() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

buf

A pointer to a buffer to hold the data read in. On output the buffer contains the data.

resid

A pointer to a number of bytes. On entry, the amount of data to be read in full. It must not exceed the size of buf. On return, the amount of data remaining to be read. (A non-zero returned value means some but not all of the data was read.)

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success, zero is returned. *resid may be zero, indicating a complete read, or non-zero, indicating the network connection is closed or end of file is reached.

Otherwise, a value of -1 is returned, *resid is non-zero, and errno is set to indicate the error. Possible errno values are the same as set by the read(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

Description
This function blocks the calling thread until the specified amount of data is read in full, it encounters an end-of-stream indication, a timeout occurs, or an error occurs. It differs from st_read_fully() only in that it allows the caller to know how many bytes were transferred before an error occurred.


st_readv()

Reads data from a specified file descriptor object into multiple buffers.

Syntax
#include <st.h>

ssize_t st_readv(st_netfd_t fd, const struct iovec *iov, int iov_size,
		 st_utime_t timeout);

Parameters
st_readv() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

iov

An array of iovec structures that identify the buffers for holding the data read in. On return the buffers contain the data.

iov_size

The number of iovec structures in the iov array.

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the read operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes actually read is returned (a value of 0 means the network connection is closed or end of file is reached). Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the readv(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was read.

Description
This function blocks the calling thread until it encounters an end-of-stream indication, some positive number of bytes (but no more than fit in the buffers) are read in, a timeout occurs, or an error occurs.


st_readv_resid()

Reads the specified amount of data in full from a file descriptor object into multiple buffers.

Syntax
#include <st.h>

int st_readv_resid(st_netfd_t fd, struct iovec **iov, int *iov_size,
		   st_utime_t timeout);

Parameters
st_readv_resid() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

iov

A pointer to an array of iovec structures. On entry, the iovecs identify the buffers for holding the data read in. On return, the incomplete iovecs. This function modifies both the pointer and the array to which it points.

iov_size

A pointer to a number of iovec structures. On entry, the number of iovec structures pointed to by *iov. On return, the number of incomplete or unused iovec structures. (A non-zero returned value means some but not all of the data was read.)

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success, zero is returned. *iov_size may be zero, indicating a complete read, or non-zero, indicating the network connection is closed or end of file is reached. *iov points to the first iovec after the end of the original array on a complete read, or to the first incomplete iovec on an incomplete read.

Otherwise, a value of -1 is returned, *iov_size is non-zero, and errno is set to indicate the error. *iov points to the first unused iovec. Possible errno values are the same as set by the readv(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

All of the iovecs before *iov are modified such that iov_base points to the end of the original buffer and iov_len is zero.

Description
This function blocks the calling thread until the specified amount of data is read in full, it encounters an end-of-stream indication, a timeout occurs, or an error occurs. Like st_read_resid() it blocks the thread until all of the requested data is read or an error occurs. Use st_readv() to read up to the requested amount of data.


st_write()

Writes a buffer of data to a specified file descriptor object.

Syntax
#include <st.h>

ssize_t st_write(st_netfd_t fd, const void *buf, size_t nbyte,
                 st_utime_t timeout);

Parameters
st_write() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

buf

A pointer to the buffer holding the data to be written.

nbyte

The amount of data in bytes to be written from the buffer.

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer equal to nbyte is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the write(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

Description
This function blocks the calling thread until all the data is written, a timeout occurs, or the write operation fails. The return value is equal to either nbyte (on success) or -1 (on failure). Note that if st_write() returns -1, some data (less than nbyte bytes) may have been written before an error occurred.


st_write_resid()

Writes a buffer of data to a specified file descriptor object.

Syntax
#include <st.h>

int st_write_resid(st_netfd_t fd, const void *buf, size_t *resid,
                   st_utime_t timeout);

Parameters
st_write_resid() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

buf

A pointer to the buffer holding the data to be written.

resid

A pointer to a number of bytes. On entry, the amount of data to be written from the buffer. On return, the amount of data remaining to be written. (A non-zero returned value means some but not all of the data was written.)

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success, zero is returned and *resid is zero. Otherwise, a value of -1 is returned, *resid is non-zero, and errno is set to indicate the error. Possible errno values are the same as set by the write(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

Description
This function blocks the calling thread until all the data is written, a timeout occurs, or the write operation fails. It differs from st_write() only in that it allows the caller to know how many bytes were transferred before an error occurred.


st_writev()

Writes data to a specified file descriptor object from multiple buffers.

Syntax
#include <st.h>

ssize_t st_writev(st_netfd_t fd, const struct iovec *iov, int iov_size,
                  st_utime_t timeout);

Parameters
st_writev() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

iov

An array of iovec structures that describe the buffers to write from (see writev(2)).

iov_size

Number of iovec structures in the iov array.

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer equal to the sum of all the buffer lengths is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the writev(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

Description
This function blocks the calling thread until all the data is written, a timeout occurs, or the write operation fails. The return value is equal to either the sum of all the buffer lengths (on success) or -1 (on failure). Note that if st_writev() returns -1, part of the data may have been written before an error occurred.


st_writev_resid()

Writes multiple buffers of data to a specified file descriptor object.

Syntax
#include <st.h>

int st_writev_resid(st_netfd_t fd, struct iovec **iov, int *iov_size,
		    st_utime_t timeout);

Parameters
st_writev_resid() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t).

iov

A pointer to an array of iovec structures. On entry, the iovecs identify the buffers holding the data to write. On return, the incomplete iovecs. This function modifies both the pointer and the array to which it points.

iov_size

A pointer to a number of iovec structures. On entry, the number of iovec structures pointed to by *iov. On return, the number of incomplete or unused iovec structures. (A non-zero returned value means some but not all of the data was written.)

timeout

A value of type st_utime_t specifying the inactivity timeout (in microseconds). Note that timeouts are measured since the last context switch.

Returns
On success, zero is returned, *iov_size is zero, and *iov points to the first iovec after the end of the original array. Otherwise, a value of -1 is returned, *iov_size is non-zero, *iov points to the first incomplete iovec, and errno is set to indicate the error. Possible errno values are the same as set by the writev(2) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred.

All of the iovecs before *iov are modified such that iov_base points to the end of the original buffer and iov_len is zero.

Description
This function blocks the calling thread until all the data is written, a timeout occurs, or the write operation fails. It differs from st_writev() only in that it allows the caller to know how many bytes were transferred before an error occurred.


st_recvfrom()

Receives bytes from a file descriptor object and stores the sending peer's address.

Syntax
#include <st.h>

int st_recvfrom(st_netfd_t fd, void *buf, int len, struct sockaddr *from,
                int *fromlen, st_utime_t timeout);

Parameters
st_recvfrom() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing a UDP socket.

buf

A pointer to a buffer to hold the data received.

len

The size of buf in bytes.

from

If this parameter is not a NULL pointer, the source address of the message is filled in (see recvfrom(3)).

fromlen

This is a value-result parameter, initialized to the size of the buffer associated with from, and modified on return to indicate the actual size of the address stored there.

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the receive operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the length of the received message in bytes is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the recvfrom(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was received.

Description
This function receives up to a specified number of bytes from the specified file descriptor object representing a UDP socket.

st_recvfrom() blocks the calling thread until one or more bytes are transferred, a timeout has occurred, or there is an error. No more than len bytes will be transferred.


st_sendto()

Sends bytes to a specified destination.

Syntax
#include <st.h>

int st_sendto(st_netfd_t fd, const void *msg, int len, struct sockaddr *to,
              int tolen, st_utime_t timeout);

Parameters
st_sendto() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing a UDP socket.

msg

A pointer to a buffer containing the message to be sent.

len

The length of the message to be sent (in bytes).

to

A pointer to the address of the destination (see sendto(3)).

tolen

This parameter specifies the size of the destination address.

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the send operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes sent is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the sendto(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was sent.

Description
This function sends a specified number of bytes from a file descriptor object representing a UDP socket to the specified destination address. If no buffer space is available at the underlying OS socket to hold the message to be transmitted, then st_sendto() blocks the calling thread until the space becomes available, a timeout occurs, or an error occurs.


st_recvmsg()

Receives a message from a file descriptor object.

Syntax
#include <st.h>

int st_recvmsg(st_netfd_t fd, struct msghdr *msg, int flags,
               st_utime_t timeout);

Parameters
st_recvmsg() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing a UDP socket.

msg

A pointer to a msghdr structure to describe the data received.

flags

Control flags for recvmsg(3).

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the receive operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes received is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the recvmsg(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was received.

Description
This function receives bytes from the specified file descriptor object representing a UDP socket. The operation is controlled by the in/out msg parameter.

st_recvmsg() blocks the calling thread until one or more bytes are transferred, a timeout has occurred, or there is an error.


st_sendmsg()

Sends a message to a file descriptor object.

Syntax
#include <st.h>

int st_sendmsg(st_netfd_t fd, const struct msghdr *msg, int flags,
               st_utime_t timeout);

Parameters
st_sendmsg() has the following parameters:

fd

A file descriptor object identifier (see st_netfd_t) representing a UDP socket.

msg

A pointer to a msghdr structure describing the message to be sent.

flags

Control flags for sendmsg(3).

timeout

A value of type st_utime_t specifying the time limit in microseconds for completion of the send operation. Note that timeouts are measured since the last context switch.

Returns
On success a non-negative integer indicating the number of bytes sent is returned. Otherwise, a value of -1 is returned and errno is set to indicate the error. Possible errno values are the same as set by the sendmsg(3) call with two exceptions:

EINTRThe current thread was interrupted by st_thread_interrupt().
ETIMEThe timeout occurred and no data was sent.

Description
This function sends bytes to a file descriptor object representing a UDP socket. The operation is controlled by the msg parameter. If no buffer space is available at the underlying OS socket to hold the message to be transmitted, then st_sendmsg() blocks the calling thread until the space becomes available, a timeout occurs, or an error occurs.


st_open()

Opens a file for reading, writing, or both.

Syntax
#include <st.h>

st_netfd_t st_open(const char *path, int oflags, mode_t mode);

Parameters
st_open() has the following parameters:

path

The pathname of the file to be opened.

oflags

File status flags. These are the same flags that are used by the open(2) system call.

mode

Access permission bits of the file mode, if the file is created when O_CREAT is set in oflags (see open(2)).

Returns
Upon successful completion, a new file descriptor object identifier is returned. Otherwise, NULL is returned and errno is set to indicate the error.

Description
This function creates a new file descriptor object of type st_netfd_t for the file with the pathname path. This object can be freed by st_netfd_free() or st_netfd_close().

The primary purpose of this function is to open FIFOs (named pipes) or other special files in order to create an end point of communication. However, it can be used on regular files as well.

Among other things, this function always sets a non-blocking flag on the underlying OS file descriptor, so there is no need to include that flag in oflags.


st_poll()

Detects when I/O is ready for a set of OS file descriptors.

Syntax
#include <st.h>

int st_poll(struct pollfd *pds, int npds, st_utime_t timeout);

Parameters
st_poll() has the following parameters:

pds

A pointer to an array of pollfd structures (see poll(2)).

npds

The number of elements in the pds array.

timeout

A value of type st_utime_t specifying the amount of time in microseconds the call will block waiting for I/O to become ready. If this time expires without any I/O becoming ready, st_poll() returns zero. Note that timeouts are measured since the last context switch.

Returns
Upon successful completion, a non-negative value is returned. A positive value indicates the total number of OS file descriptors in pds that have events. A value of 0 indicates that the call timed out. Upon failure, a value of -1 is returned and errno is set to indicate the error:

EINTRThe current thread was interrupted by st_thread_interrupt().

If an alternative event notification mechanism has been set by st_set_eventsys(), other values of errno could be set upon failure as well. The values depend on the specific mechanism in use.

Description
This function returns as soon as I/O is ready on one or more of the specified OS file descriptors. A count of the number of ready descriptors is returned unless a timeout occurs, in which case zero is returned.

The pollfd structure is defined in the poll.h header file and contains the following members:

    int fd;             /* OS file descriptor */
    short events;       /* requested events   */
    short revents;      /* returned events    */
The events field should be set to the I/O events (readable, writable, exception, or some combination) that the caller is interested in. On return, the revents field is set to indicate what kind of I/O is ready on the respective descriptor.

The events and revents fields are constructed by OR-ing any combination of the following event flags (defined in poll.h):

POLLINfd is readable.
POLLOUTfd is is writable.
POLLPRIfd has an exception condition.
POLLNVALfd is bad.

The POLLNVAL flag is only valid in the revents field; it is not used in the events field.

Despite having an interface like poll(2), this function uses the same event notification mechanism as the rest of the library. For instance if an alternative event nofication mechanism was set using st_set_eventsys(), this function uses that mechanism to check for events.

Note that unlike the poll(2) call, this function has the timeout parameter expressed in microseconds. If the value of timeout is ST_UTIME_NO_TIMEOUT (-1), this function blocks until a requested I/O event occurs or until the call is interrupted by st_thread_interrupt().

Note: if kqueue(2) is used as an alternative event notification mechanism (see st_set_eventsys()), the POLLPRI event flag is not supported and st_poll() will return an error if it's set (errno will be set to EINVAL).


Program Structure

Generally, the following steps should be followed when writing an application using the State Threads library:

  1. Configure the library by calling these pre-init functions, if desired.
  2. Initialize the library by calling st_init().
  3. Configure the library by calling these post-init functions, if desired.
  4. Create resources that will be shared among different processes: create and bind listening sockets (see socket(3), bind(3), listen(3), st_netfd_open_socket(), and possibly st_netfd_serialize_accept()), create shared memory segments, inter-process communication (IPC) channels and synchronization primitives (if any).
  5. Create several processes via fork(2). The parent process should either exit or become a "watchdog" (e.g., it starts a new process when an existing one crashes, does a cleanup upon application termination, etc.).
  6. In each child process create a pool of threads (see st_thread_create()) to handle user connections. Each thread in the pool may accept client connections (st_accept()), connect to other servers (st_connect()), perform various network I/O (st_read(), st_write(), etc.).

Note that only State Threads library I/O functions should be used for a network I/O: any other I/O calls may block the calling process indefinitely. For example, standard I/O functions (fgets(3), fread(3), fwrite(3), fprintf(3), etc.) call read(2) and write(2) directly and therefore should not be used on sockets or pipes.

Also note that for short timeouts to work the program should do context switches (for example by calling st_usleep()) on a regular basis.


List of Blocking Functions

The thread context switch (process state change) can only happen in a well-known set of blocking functions. Only the following functions can block the calling thread:

st_thread_join()
st_sleep()
st_usleep()
st_cond_wait()
st_cond_timedwait()
st_mutex_lock()
st_netfd_poll()
st_accept()
st_connect()
st_read()
st_read_fully()
st_read_resid()
st_readv()
st_readv_resid()
st_write()
st_write_resid()
st_writev()
st_writev_resid()
st_recvfrom()
st_sendto()
st_recvmsg()
st_sendmsg()
st_poll()