Hi On Thu, Oct 16, 2008 at 3:50 AM, Daniel Baluta <daniel.baluta@xxxxxxxxx> wrote: > Hello , > > I want to understand how kernel handles signals delivering between processes. > Can you guide me through kernel code , or provide some basics ideas. For basic idea, I could propose this kind of concept: think of it like posting a message to destination (in this case...a task). After it is posted to a target, rescheduling is called to make sure the target process pick it up ASAP. of course there is no guarantee that process will run right away since it's possible for example higher prio runnable process is sitting in the run queue. Let's assume the process is selected by scheduler and executed. Also let's assume it's going from kernel space back to user space. Kernel will notice there is "a message" sitting and determine what that is. If it's SIGKILL or any other unmodifiable signal (no way to change the handle), kernel handler will be engaged. If kernel notice it needs to call user space handler, it will jump into the related EIP in user space. At the same time, kernel will modify the stack of the target process so that when it finished executing the handler, sigreturn will be called so task will jump back into the kernel path. And then it returns to normal flow, that is going back to the path before the task entered kernel space. I might missed something, so feel free to CMIIW people... regards, Mulyadi. -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
