* [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation
@ 2011-01-29 11:00 Paolo Bonzini
2011-01-29 11:00 ` [Qemu-devel] [PATCH 2/2] Unify " Paolo Bonzini
2011-01-30 16:46 ` [Qemu-devel] [PATCH 1/2] Correct " Aurelien Jarno
0 siblings, 2 replies; 3+ messages in thread
From: Paolo Bonzini @ 2011-01-29 11:00 UTC (permalink / raw)
To: qemu-devel; +Cc: Jan Kiszka
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;
-
+ }
+ 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;
+ }
if (active_timers[QEMU_CLOCK_REALTIME]) {
rtdelta = (active_timers[QEMU_CLOCK_REALTIME]->expire_time -
qemu_get_clock(rt_clock))*1000;
--
1.7.3.4
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [Qemu-devel] [PATCH 2/2] Unify alarm deadline computation
2011-01-29 11:00 [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation Paolo Bonzini
@ 2011-01-29 11:00 ` Paolo Bonzini
2011-01-30 16:46 ` [Qemu-devel] [PATCH 1/2] Correct " Aurelien Jarno
1 sibling, 0 replies; 3+ messages in thread
From: Paolo Bonzini @ 2011-01-29 11:00 UTC (permalink / raw)
To: qemu-devel
Now that qemu_next_deadline_dyntick uses the correct formula, we can
simplify the code in host_alarm_handler and eliminate useless duplication.
Also, qemu_next_deadline_dyntick is always used in signed arithmetic,
so change its type.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
qemu-timer.c | 28 +++++++++++-----------------
roms/vgabios | 2 +-
2 files changed, 12 insertions(+), 18 deletions(-)
diff --git a/qemu-timer.c b/qemu-timer.c
index 174fd0c..1842d7a 100644
--- a/qemu-timer.c
+++ b/qemu-timer.c
@@ -635,6 +635,8 @@ void qemu_run_all_timers(void)
qemu_run_timers(host_clock);
}
+static int64_t qemu_next_alarm_deadline(void);
+
#ifdef _WIN32
static void CALLBACK host_alarm_handler(UINT uTimerID, UINT uMsg,
DWORD_PTR dwUser, DWORD_PTR dw1,
@@ -677,14 +679,7 @@ static void host_alarm_handler(int host_signum)
}
#endif
if (alarm_has_dynticks(t) ||
- (!use_icount &&
- qemu_timer_expired(active_timers[QEMU_CLOCK_VIRTUAL],
- qemu_get_clock(vm_clock))) ||
- qemu_timer_expired(active_timers[QEMU_CLOCK_REALTIME],
- qemu_get_clock(rt_clock)) ||
- qemu_timer_expired(active_timers[QEMU_CLOCK_HOST],
- qemu_get_clock(host_clock))) {
-
+ qemu_next_alarm_deadline () <= 0) {
t->expired = alarm_has_dynticks(t);
t->pending = 1;
qemu_notify_event();
@@ -715,11 +710,7 @@ int64_t qemu_next_deadline(void)
#ifndef _WIN32
-#if defined(__linux__)
-
-#define RTC_FREQ 1024
-
-static uint64_t qemu_next_deadline_dyntick(void)
+static int64_t qemu_next_alarm_deadline(void)
{
int64_t delta;
int64_t rtdelta;
@@ -743,12 +734,13 @@ static uint64_t qemu_next_deadline_dyntick(void)
delta = rtdelta;
}
- if (delta < MIN_TIMER_REARM_US)
- delta = MIN_TIMER_REARM_US;
-
return delta;
}
+#if defined(__linux__)
+
+#define RTC_FREQ 1024
+
static void enable_sigio_timer(int fd)
{
struct sigaction act;
@@ -903,7 +895,9 @@ static void dynticks_rearm_timer(struct qemu_alarm_timer *t)
!active_timers[QEMU_CLOCK_HOST])
return;
- nearest_delta_us = qemu_next_deadline_dyntick();
+ nearest_delta_us = qemu_next_alarm_deadline();
+ if (nearest_delta_us < MIN_TIMER_REARM_US)
+ nearest_delta_us = MIN_TIMER_REARM_US;
/* check whether a timer is already running */
if (timer_gettime(host_timer, &timeout)) {
--
1.7.3.4
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation
2011-01-29 11:00 [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation Paolo Bonzini
2011-01-29 11:00 ` [Qemu-devel] [PATCH 2/2] Unify " Paolo Bonzini
@ 2011-01-30 16:46 ` Aurelien Jarno
1 sibling, 0 replies; 3+ messages in thread
From: Aurelien Jarno @ 2011-01-30 16:46 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: Jan Kiszka, qemu-devel
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
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2011-01-30 16:46 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-01-29 11:00 [Qemu-devel] [PATCH 1/2] Correct alarm deadline computation Paolo Bonzini
2011-01-29 11:00 ` [Qemu-devel] [PATCH 2/2] Unify " Paolo Bonzini
2011-01-30 16:46 ` [Qemu-devel] [PATCH 1/2] Correct " Aurelien Jarno
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.