Re: [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation

[Aurelien Jarno <aurelien@aurel32.net>]

On Sat, Jan 29, 2011 at 12:00:47PM +0100, Paolo Bonzini wrote:
> When the QEMU_CLOCK_HOST clock was added, computation of its
> deadline was added to qemu_next_deadline, which is correct but
> incomplete.
> 
> I noticed this while trying to make sense of the rules whereby
> qemu_next_deadline_dyntick is computed, which miss QEMU_CLOCK_HOST
> when use_icount is true.  Looking at the history showed this to
> be just an oversight, as the next patch shows clearly.
> 
> This patch inlines qemu_next_deadline into qemu_next_deadline_dyntick,
> and corrects the logic so that only QEMU_CLOCK_VIRTUAL is skipped for
> use_icount == true.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Jan Kiszka <jan.kiszka@siemens.com>
> ---
>  qemu-timer.c |   15 +++++++++++----
>  1 files changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/qemu-timer.c b/qemu-timer.c
> index db1ec49..174fd0c 100644
> --- a/qemu-timer.c
> +++ b/qemu-timer.c
> @@ -724,11 +724,18 @@ static uint64_t qemu_next_deadline_dyntick(void)
>      int64_t delta;
>      int64_t rtdelta;
>  
> -    if (use_icount)
> +    if (!use_icount && active_timers[QEMU_CLOCK_VIRTUAL]) {
> +        delta = active_timers[QEMU_CLOCK_VIRTUAL]->expire_time -
> +                     qemu_get_clock(vm_clock);
> +    } else {
>          delta = INT32_MAX;
> -    else
> -        delta = (qemu_next_deadline() + 999) / 1000;
> -

This computation handled the fact that the value is returned in ns, and
we want us.

> +    }
> +    if (active_timers[QEMU_CLOCK_HOST]) {
> +        int64_t hdelta = active_timers[QEMU_CLOCK_HOST]->expire_time -
> +                 qemu_get_clock(host_clock);
> +        if (hdelta < delta)
> +            delta = hdelta;
> +    }

And this doesn't appear anymore after your changes.

>      if (active_timers[QEMU_CLOCK_REALTIME]) {
>          rtdelta = (active_timers[QEMU_CLOCK_REALTIME]->expire_time -
>                   qemu_get_clock(rt_clock))*1000;

Here we multiply because the value is returned in ms and we want us.

IMHO we should just use qemu_get_clock_ns() instead, replace
MIN_TIMER_REARM_US by MIN_TIMER_REARM_NS, and return a ns value
instead. Of course dynticks_rearm_timer() has to be adjusted, but it
should not be problematic given it uses ns internally.

Otherwise good catch, the issue is real, and the way to fix it looks
correct.

-- 
Aurelien Jarno	                        GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net