SYNOPSIS

#include <pcl.h>

---

coroutine_t co_create(void *func, void *data, void *stack, int stacksize);
void co_delete(coroutine_t co);
void co_call(coroutine_t co);
void co_resume(void);
void co_exit_to(coroutine_t co);
void co_exit(void);
coroutine_t co_current(void);

DESCRIPTION

The Portable Coroutine Library (PCL) implements the low level functionality for coroutines. For a definition of the term coroutine see The Art of Computer Programming by Donald E. Knuth. Coroutines are a very simple cooperative multitasking environment where the switch from one task to another is done explicitly by a function call. Coroutines are a lot faster than processes or threads switch, since there is no OS kernel involvement for the operation. This document defines an API for the low level handling of coroutines i.e. creating and deleting coroutines and switching between them. Higher level functionality (scheduler, etc.) is not covered.

Functions

The following functions are defined:

coroutine_t co_create(void *func, void *data, void *stack, int stacksize);

This function creates a new coroutine. func is the entry point of the coroutine. It will be called with one arg, a void *, which holds the data passed through the data parameter. If func terminates, the associated coroutine is deleted. stack is the base of the stack this coroutine will use and stacksize its size in bytes. You may pass a NULL pointer for stack in which case the memory will be allocated by co_create itself. Both, stack and stacksize are aligned to system requirements. A stacksize of less then 4096 bytes will be rejected. You have to make sure, that the stack is large enough for your coroutine and possible signal handlers (see below). The stack will not grow! (Exception: the main coroutine uses the standard system stack which may still grow) On success, a handle (coroutine_t) for a new coroutine is returned, otherwise NULL.

void co_delete(coroutine_t co);

This function deletes the given coroutine co. If the stack for this coroutine was allocated by co_create it will be freed. After a coroutine handle was passed to co_delete it is invalid and may not be used any more. It is invalid for a coroutine to delete itself with this function.

void co_call(coroutine_t co);

This function passes execution to the given coroutine co. The first time the coroutine is executed, its entry point func is called, and the data parameter used during the call to co_create is passed to func. The current coroutine is suspended until another one restarts it with a co_call or co_resume call. Calling oneself returns immediately.

void co_resume(void);

This function passes execution back to the coroutine which either initially started this one or restarted it after a prior co_resume.

void co_exit_to(coroutine_t co);

This function does the same a co_delete(co_current()) followed by a co_call would do. That is, it deletes itself and then passes execution to another coroutine co.

void co_exit(void);

This function does the same a co_delete(co_current()) followed by a co_resume would do. That is, it deletes itself and then passes execution back to the coroutine which either initially started this one or restarted it after a prior co_resume.

coroutine_t co_current(void);

This function returns the currently running coroutine.

Notes

Some interactions with other parts of the system are covered here.

Signals

First, a signal handler is not defined to run in any specific coroutine. The only way to leave the signal handler is by a return statement.

Second, the signal handler may run with the stack of any coroutine, even with the stack of library internal coroutines which have an undefined stack size (just enough to perform a kernel call). Using and alternate stack for signal processing (see sigaltstack(2)) is recommended!

Conclusion: avoid signals like a plague. The only thing you may do reliable is setting some global variables and return. Simple kernel calls may work too, but nowadays it's pretty hairy to tell, which function really is a kernel call. (Btw, all this applies to normal C programs, too. The coroutines just add one more problem)

setjmp/longjmp

The use of setjmp(2)/longjmp(2) is limited to jumping inside one coroutine. Never try to jump from one coroutine to another with longjmp(2).

DIAGNOSTICS

Some fatal errors are caught by the library. If one occurs, a short message is written to file descriptor 2 (stderr) and a segmentation violation is generated.

[PCL]: Cannot delete itself

A coroutine has called co_delete with it's own handle.

[PCL]: Resume to deleted coroutine

A coroutine has deleted itself with co_exit or co_exit_to and the coroutine that was activated by the exit tried a co_resume.

[PCL]: Stale coroutine called

Someone tried to active a coroutine that has already been deleted. This error is only detected, if the stack of the deleted coroutine is still resident in memory.

[PCL]: Context switch failed

Low level error generated by the library in case a context switch between two coroutines failes.

AUTHOR

Developed by Davide Libenzi < [email protected] >. Ideas and man page base source taken by the coroutine library developed by E. Toernig < [email protected] >. Also some code and ideas comes from the GNU Pth library available at http://www.gnu.org/software/pth/ .