Function kickoff

Placement in the overall structure

Module: thread
Time of creation: Task 4
Modified in: Task 5
Files: kickoff.cc

Description

For the very first activation of a Coroutine, a "return address" to a function must be known, from which the transition from the C to the C++ level can then be performed. This function should be kickoff.

Activation of kickoff

When a Coroutine is initialized with the toc_settle() function, not only is the toc struct prepared for the first coroutine switch, but the address of the kickoff function is also placed on the stack as the return address. The first coroutine switch with toc_switch therefore "jumps back" to the beginning of kickoff, where the execution starts.

Task description

void kickoff (void *dummy1, void *dummy2, void *dummy3, void *dummy4, void *dummy5, void *dummy6, Coroutine* object): The most essential function of kickoff is of course still to make the jump from the C to C++ by calling the action() method of the coroutine object found in the 7th parameter. The six dummy parameters are necessary because on x86_64 the first six parameters are passed via registers and only the seventh parameter is expected on the stack.
Furthermore, kickoff allows the same initialization to be used for all coroutine objects. For example, context switches are always triggered from a critical section. Therefore, the first action of the new coroutine must be to release the critical section.

Notes

Since this function is not really called, but only jumped to by a clever initialization of the stack of the coroutine, it must never terminate. Otherwise a meaningless value would be interpreted as a return address and the computer would crash.