[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [RFC] [PATCHv2] aio / timers pt2: Replace main_loop_tlg
From: |
Stefan Hajnoczi |
Subject: |
Re: [Qemu-devel] [RFC] [PATCHv2] aio / timers pt2: Replace main_loop_tlg with qemu_dummy_timer_ctx |
Date: |
Thu, 15 Aug 2013 14:16:52 +0200 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
On Sat, Aug 10, 2013 at 12:21:27PM +0100, Alex Bligh wrote:
> Currently we use a separate timer list group (main_loop_tlg)
> for the main loop. This patch replaces this with a dummy AioContext
> used just for timers in the main loop.
Things get interesting when we make main loop qemu_set_fd_handler() and
timers just use AioContext.
Basically, we are distilling out the main-loop.c stuff which is a
low-level glib-style event loop from the AioContext fd handlers, timers,
and BHs. This is a good step in that direction.
I'm in favor of this although it current really is still a bit of a
hack.
> @@ -486,6 +499,20 @@ void init_clocks(void)
> qemu_clock_init(type);
> }
>
> + /* Make a dummy context for timers in the main loop.
> + *
> + * This context should not call the AioContext's
> + * supplied notifier function (which calls
> + * aio_notify) as it needs to call qemu_notify()
> + * instead, as there's nothing actually using the
> + * AioContext. This is a bit of a hack.
> + */
> + qemu_dummy_timer_ctx = aio_context_new();
> + for (type = 0; type < QEMU_CLOCK_MAX; type++) {
> + qemu_dummy_timer_ctx->tlg[type]->notify_cb = NULL;
> + qemu_dummy_timer_ctx->tlg[type]->notify_opaque = NULL;
> + }
> +
IIRC your previous patch series has something like:
if (timer_list->notify_cb == NULL) {
qemu_notify_event();
} else {
timer_list->notify_cb(timer_list->notify_opaque);
}
So this should work.