All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dmitry Osipenko <digetx@gmail.com>
To: Peter Crosthwaite <crosthwaitepeter@gmail.com>
Cc: Peter Maydell <peter.maydell@linaro.org>,
	qemu-arm@nongnu.org, QEMU Developers <qemu-devel@nongnu.org>
Subject: Re: [Qemu-arm] [PATCH v8 2/4] hw/ptimer: Perform tick and counter wrap around if timer already expired
Date: Wed, 6 Jan 2016 16:12:07 +0300	[thread overview]
Message-ID: <568D12A7.3010304@gmail.com> (raw)
In-Reply-To: <20160106121711.GC4227@pcrost-box>

06.01.2016 15:17, Peter Crosthwaite пишет:
> On Tue, Jan 05, 2016 at 05:33:27AM +0300, Dmitry Osipenko wrote:
>> ptimer_get_count() might be called while QEMU timer already been expired.
>> In that case ptimer would return counter = 0, which might be undesirable
>> in case of polled timer. Do counter wrap around for periodic timer to keep
>> it distributed.
>>
>> In addition, there is no reason to keep expired timer tick deferred, so
>> just perform the tick from ptimer_get_count().
>>
>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>> ---
>>   hw/core/ptimer.c | 35 +++++++++++++++++++++++++++++------
>>   1 file changed, 29 insertions(+), 6 deletions(-)
>>
>> diff --git a/hw/core/ptimer.c b/hw/core/ptimer.c
>> index 035af97..96a6c7a 100644
>> --- a/hw/core/ptimer.c
>> +++ b/hw/core/ptimer.c
>> @@ -85,15 +85,21 @@ static void ptimer_tick(void *opaque)
>>
>>   uint64_t ptimer_get_count(ptimer_state *s)
>>   {
>> +    int enabled = s->enabled;
>
> Variable localisation not needed.
>
>>       int64_t now;
>> +    int64_t next;
>>       uint64_t counter;
>> +    int expired;
>> +    int oneshot;
>
> Variable defs can be localised to the if (enabled) (even though now
> in original code doesn't do that).
>

Yeah, I just tried to keep original style here.

>>
>> -    if (s->enabled) {
>> +    if (enabled) {
>>           now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
>> +        next = s->next_event;
>> +        expired = (now - next >= 0);
>> +        oneshot = (enabled == 2);
>>           /* Figure out the current counter value.  */
>
> This comment is now out of place.
>

Okay, will fix it.

>> -        if (now - s->next_event > 0
>> -            || s->period == 0) {
>> -            /* Prevent timer underflowing if it should already have
>> +        if (s->period == 0 || (expired && oneshot)) {
>> +            /* Prevent one-shot timer underflowing if it should already have
>>                  triggered.  */
>>               counter = 0;
>>           } else {
>> @@ -114,12 +120,12 @@ uint64_t ptimer_get_count(ptimer_state *s)
>>                  backwards.
>>               */
>>
>> -            if ((s->enabled == 1) && (s->limit * period < 10000)) {
>> +            if (!oneshot && (s->limit * period < 10000)) {
>>                   period = 10000 / s->limit;
>>                   period_frac = 0;
>>               }
>>
>> -            rem = s->next_event - now;
>> +            rem = expired ? now - next : next - now;
>>               div = period;
>>
>>               clz1 = clz64(rem);
>> @@ -139,6 +145,23 @@ uint64_t ptimer_get_count(ptimer_state *s)
>>                       div += 1;
>>               }
>>               counter = rem / div;
>> +
>> +            if (expired) {
>> +                /* Wrap around periodic counter.  */
>> +                counter = s->delta = s->limit - counter % s->limit;
>
> Why do you update the delta here?
>

Because we would want to schedule next tick based on current wrapped around 
counter value and not some arbitrary delta.

> Also can you just get ptimer_reload to do the modulo math for you? If the
> timer is !oneshot and expired, then you call ptimer_reload anyway,
> which will update next_event. When the expired test returns false
> you can just reliably use the original logic involving now and next.
>

Yes, that's what I changed in V9. Have you received it?

https://lists.nongnu.org/archive/html/qemu-devel/2016-01/msg00272.html

>> +            }
>> +        }
>> +
>> +        if (expired) {
>> +            if (oneshot) {
>
> This if-else has a lot of common structure with the one above. I think
> they could be merged.
>

That's a good suggestion, will do it in V10. Thanks.

> Regards,
> Peter
>
>> +                ptimer_tick(s);
>> +            } else {
>> +                /* Don't use ptimer_tick() for the periodic timer since it
>> +                 * would reset the delta value.
>> +                 */
>> +                ptimer_trigger(s);
>> +                ptimer_reload(s);
>> +            }
>>           }
>>       } else {
>>           counter = s->delta;
>> --
>> 2.6.4
>>


-- 
Dmitry

WARNING: multiple messages have this Message-ID (diff)
From: Dmitry Osipenko <digetx@gmail.com>
To: Peter Crosthwaite <crosthwaitepeter@gmail.com>
Cc: Peter Maydell <peter.maydell@linaro.org>,
	qemu-arm@nongnu.org, QEMU Developers <qemu-devel@nongnu.org>
Subject: Re: [Qemu-devel] [PATCH v8 2/4] hw/ptimer: Perform tick and counter wrap around if timer already expired
Date: Wed, 6 Jan 2016 16:12:07 +0300	[thread overview]
Message-ID: <568D12A7.3010304@gmail.com> (raw)
In-Reply-To: <20160106121711.GC4227@pcrost-box>

06.01.2016 15:17, Peter Crosthwaite пишет:
> On Tue, Jan 05, 2016 at 05:33:27AM +0300, Dmitry Osipenko wrote:
>> ptimer_get_count() might be called while QEMU timer already been expired.
>> In that case ptimer would return counter = 0, which might be undesirable
>> in case of polled timer. Do counter wrap around for periodic timer to keep
>> it distributed.
>>
>> In addition, there is no reason to keep expired timer tick deferred, so
>> just perform the tick from ptimer_get_count().
>>
>> Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
>> ---
>>   hw/core/ptimer.c | 35 +++++++++++++++++++++++++++++------
>>   1 file changed, 29 insertions(+), 6 deletions(-)
>>
>> diff --git a/hw/core/ptimer.c b/hw/core/ptimer.c
>> index 035af97..96a6c7a 100644
>> --- a/hw/core/ptimer.c
>> +++ b/hw/core/ptimer.c
>> @@ -85,15 +85,21 @@ static void ptimer_tick(void *opaque)
>>
>>   uint64_t ptimer_get_count(ptimer_state *s)
>>   {
>> +    int enabled = s->enabled;
>
> Variable localisation not needed.
>
>>       int64_t now;
>> +    int64_t next;
>>       uint64_t counter;
>> +    int expired;
>> +    int oneshot;
>
> Variable defs can be localised to the if (enabled) (even though now
> in original code doesn't do that).
>

Yeah, I just tried to keep original style here.

>>
>> -    if (s->enabled) {
>> +    if (enabled) {
>>           now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
>> +        next = s->next_event;
>> +        expired = (now - next >= 0);
>> +        oneshot = (enabled == 2);
>>           /* Figure out the current counter value.  */
>
> This comment is now out of place.
>

Okay, will fix it.

>> -        if (now - s->next_event > 0
>> -            || s->period == 0) {
>> -            /* Prevent timer underflowing if it should already have
>> +        if (s->period == 0 || (expired && oneshot)) {
>> +            /* Prevent one-shot timer underflowing if it should already have
>>                  triggered.  */
>>               counter = 0;
>>           } else {
>> @@ -114,12 +120,12 @@ uint64_t ptimer_get_count(ptimer_state *s)
>>                  backwards.
>>               */
>>
>> -            if ((s->enabled == 1) && (s->limit * period < 10000)) {
>> +            if (!oneshot && (s->limit * period < 10000)) {
>>                   period = 10000 / s->limit;
>>                   period_frac = 0;
>>               }
>>
>> -            rem = s->next_event - now;
>> +            rem = expired ? now - next : next - now;
>>               div = period;
>>
>>               clz1 = clz64(rem);
>> @@ -139,6 +145,23 @@ uint64_t ptimer_get_count(ptimer_state *s)
>>                       div += 1;
>>               }
>>               counter = rem / div;
>> +
>> +            if (expired) {
>> +                /* Wrap around periodic counter.  */
>> +                counter = s->delta = s->limit - counter % s->limit;
>
> Why do you update the delta here?
>

Because we would want to schedule next tick based on current wrapped around 
counter value and not some arbitrary delta.

> Also can you just get ptimer_reload to do the modulo math for you? If the
> timer is !oneshot and expired, then you call ptimer_reload anyway,
> which will update next_event. When the expired test returns false
> you can just reliably use the original logic involving now and next.
>

Yes, that's what I changed in V9. Have you received it?

https://lists.nongnu.org/archive/html/qemu-devel/2016-01/msg00272.html

>> +            }
>> +        }
>> +
>> +        if (expired) {
>> +            if (oneshot) {
>
> This if-else has a lot of common structure with the one above. I think
> they could be merged.
>

That's a good suggestion, will do it in V10. Thanks.

> Regards,
> Peter
>
>> +                ptimer_tick(s);
>> +            } else {
>> +                /* Don't use ptimer_tick() for the periodic timer since it
>> +                 * would reset the delta value.
>> +                 */
>> +                ptimer_trigger(s);
>> +                ptimer_reload(s);
>> +            }
>>           }
>>       } else {
>>           counter = s->delta;
>> --
>> 2.6.4
>>


-- 
Dmitry

  reply	other threads:[~2016-01-06 13:12 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-01-05  2:33 [Qemu-devel] [PATCH v8 0/4] PTimer fixes and ARM MPTimer conversion Dmitry Osipenko
2016-01-05  2:33 ` [Qemu-devel] [PATCH v8 1/4] hw/ptimer: Fix issues caused by the adjusted timer limit value Dmitry Osipenko
2016-01-06 12:15   ` [Qemu-arm] " Peter Crosthwaite
2016-01-06 12:15     ` [Qemu-devel] " Peter Crosthwaite
2016-01-06 13:25     ` [Qemu-arm] " Dmitry Osipenko
2016-01-06 13:25       ` [Qemu-devel] " Dmitry Osipenko
2016-01-06 13:38       ` Peter Crosthwaite
2016-01-05  2:33 ` [Qemu-devel] [PATCH v8 2/4] hw/ptimer: Perform tick and counter wrap around if timer already expired Dmitry Osipenko
2016-01-06 12:17   ` [Qemu-arm] " Peter Crosthwaite
2016-01-06 12:17     ` [Qemu-devel] " Peter Crosthwaite
2016-01-06 13:12     ` Dmitry Osipenko [this message]
2016-01-06 13:12       ` Dmitry Osipenko
2016-01-06 13:59       ` Peter Crosthwaite
2016-01-06 20:52         ` [Qemu-arm] " Dmitry Osipenko
2016-01-06 20:52           ` [Qemu-devel] " Dmitry Osipenko
2016-01-05  2:33 ` [Qemu-devel] [PATCH v8 3/4] hw/ptimer: Update .delta on period/freq change Dmitry Osipenko
2016-01-06 12:17   ` Peter Crosthwaite
2016-01-05  2:33 ` [Qemu-devel] [PATCH v8 4/4] arm_mptimer: Convert to use ptimer Dmitry Osipenko
2016-01-06 13:17   ` Peter Crosthwaite
2016-01-07 14:40     ` [Qemu-arm] " Dmitry Osipenko
2016-01-07 14:40       ` [Qemu-devel] " Dmitry Osipenko
2016-01-07 17:34     ` [Qemu-arm] " Dmitry Osipenko
2016-01-07 17:34       ` [Qemu-devel] " Dmitry Osipenko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=568D12A7.3010304@gmail.com \
    --to=digetx@gmail.com \
    --cc=crosthwaitepeter@gmail.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-arm@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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.