Re: [PATCH 2/6] sched: add function to execute a function synchronously on a physical cpu

[Juergen Gross]

On 11/03/16 13:57, Peter Zijlstra wrote:
> On Fri, Mar 11, 2016 at 01:48:12PM +0100, Juergen Gross wrote:
>> On 11/03/16 13:42, Peter Zijlstra wrote:
>>> how about something like:
>>>
>>> struct xen_callback_struct {
>>> 	struct work_struct	work;
>>> 	struct completion	done;
> 	int			(*func)(void*);
>>> 	void *			data;
>>> 	int			ret;
>>> };
>>>
>>> static void xen_callback_f(struct work_struct *work)
>>> {
>>> 	struct xen_callback_struct *xcs = container_of(work, struct xen_callback_struct, work);
>>>
>>> 	xcs->ret = xcs->func(xcs->data);
>>>
>>> 	complete(&xcs->done);
>>> }
>>>
>>> xen_call_on_cpu_sync(int cpu, int (*func)(void *), void *data)
>>> {
>>> 	struct xen_callback_state xcs = {
>>> 		.work = __WORK_INITIALIZER(xcs.work, xen_callback_f);
>>> 		.done = COMPLETION_INITIALIZER_ONSTACK(xcs.done),
> 		.func = func,
>>> 		.data = data,
>>> 	};
>>>
>>> 	queue_work_on(&work, cpu);
>>> 	wait_for_completion(&xcs.done);
>>>
>>> 	return xcs.ret;
>>> }
>>>
>>> No mucking about with the scheduler state, no new exported functions
>>> etc..
>>>
>>
>> Hey, I like it. Can't be limited to Xen as on bare metal the function
>> needs to be called on cpu 0, too. But avoiding the scheduler fiddling
>> is much better! As this seems to be required for Dell hardware only,
>> I could add it to some Dell base driver in case you don't want to add
>> it to core code.
> 
> Urgh yeah, saw that in your other mail. It looks like I should go look
> at set_cpus_allowed_ptr() abuse :/
> 
> Not sure where this would fit best, maybe somewhere near workqueue.c or
> smp.c.

At a first glance I think smp.c would be the better choice. I'll have a try.

Thanks,

Juergen