Thanks for your comments.
>>But what I really wanted is a Common fork()/exit() event handling framework.
>
> Could you expand on this a bit? Especially since you acknowledge that
loadable
> modules are not particularly essential to your work, I am curious as to what
> else you find valuable in such a fork/exit framework.
If we can implement some advanced features (CpuSet, CSA+Job, CKRM, etc...)
as a kernel loadable module, it's best I also think.
But using the hooks in fork()/exit() is better than patching to
fork.c or exit.c for each feature, even though it can't be implemented
as a kernel loadable module. Because we need not modify
kernel/fork.c or kernel/exit.c directly.
For example, we must append individually cpuset_fork() for CpuSet,
pagg_attach() for PAGG(CSA+Job), ckrm_cb_fork() for CKRM in kernel/fork.c
when we try to use those advanced features.
In this case, we need to patch into three points in kernel/fork.c.
But if we have a common purpose hook in kernel/fork.c, those advanced
features does not need to modify kernel/fork.c directly.
They have only to register their own event handler for the fork-hook.
In short, my motivation is to integrate the hooks plugged ramdomly
in kernel/fork.c and so on.
Thanks.
--
Linux Promotion Center, NEC
KaiGai Kohei <kaigai@xxxxxxxxxxxxx>
|