On Wed, 2009-07-15 at 18:46 +0800, Andi Kleen wrote:
> Zhang Rui <rui.zhang@xxxxxxxxx> writes:
>
> > i8042 controller support device async actions.
> >
> > If boot option "dev_async_action" is added,
> > i8042 controller and its child devices can be
> > suspended/resumed/shutdown asynchronously.
> >
>
> From a quick look at the i8042 driver it still
> seems to do a lot of slow actions without actually sleeping
> or worse holding locks. e.g. the delay loop in i8042_flush()
>
> Did you measure how long that one takes?
>
> Due to the locks even preempt kernels couldn't do
> something during that time.
>
> The spinlocks are probably needed when the code is executed
> during interrupts, but perhaps the suspend variant
> could use a different code path.
>
> Perhaps it only makes sense to do this when this
> code is converted to sleep during delays too?
>
most of the psmouse suspend/shutdown time was cost by a ps2 reset
command.
In my test, the serio2 (psmouse) total suspend time is 0.42s, while
the psmouse_reset takes 0.4s.
i.e.
drivers/input/serio/libps2.c
int ps2_command(struct ps2dev *ps2dev, unsigned char *param, int
command)
{
...
/*
* The reset command takes a long time to execute.
*/
timeout = msecs_to_jiffies(command == PS2_CMD_RESET_BAT ? 4000 : 500);
timeout = wait_event_timeout(ps2dev->wait,
!(ps2dev->flags & PS2_FLAG_CMD1), timeout);
...
}
so the patch could save suepend/shutdown time a lot, without any other
changes in psmouse driver.
thanks,
rui
_______________________________________________
linux-pm mailing list
linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linux-foundation.org/mailman/listinfo/linux-pm
