xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
* [PATCH, v3] x86/HVM: RTC periodic timer emulation adjustments
@ 2012-08-31 16:01 Jan Beulich
  2012-08-31 16:50 ` Keir Fraser
  0 siblings, 1 reply; 3+ messages in thread
From: Jan Beulich @ 2012-08-31 16:01 UTC (permalink / raw)
  To: xen-devel

[-- Attachment #1: Type: text/plain, Size: 1837 bytes --]

- don't call rtc_timer_update() on REG_A writes when the value didn't
  change (doing the call always was reported to cause wall clock time
  lagging with the JVM running on Windows)
- don't call rtc_timer_update() on REG_B writes when RTC_PIE didn't
  change

Signed-off-by: Jan Beulich <jbeulich@suse.com>

---
v3: Break out just this change from the previously submitted much
    larger patch. The rest of that one is now planned to go in only
    after 4.2.

--- a/xen/arch/x86/hvm/rtc.c
+++ b/xen/arch/x86/hvm/rtc.c
@@ -365,6 +365,7 @@ static int rtc_ioport_write(void *opaque
 {
     RTCState *s = opaque;
     struct domain *d = vrtc_domain(s);
+    uint32_t orig;
 
     spin_lock(&s->lock);
 
@@ -382,6 +383,7 @@ static int rtc_ioport_write(void *opaque
         return 0;
     }
 
+    orig = s->hw.cmos_data[s->hw.cmos_index];
     switch ( s->hw.cmos_index )
     {
     case RTC_SECONDS_ALARM:
@@ -405,9 +407,9 @@ static int rtc_ioport_write(void *opaque
         break;
     case RTC_REG_A:
         /* UIP bit is read only */
-        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) |
-            (s->hw.cmos_data[RTC_REG_A] & RTC_UIP);
-        rtc_timer_update(s);
+        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) | (orig & RTC_UIP);
+        if ( (data ^ orig) & (RTC_RATE_SELECT | RTC_DIV_CTL) )
+            rtc_timer_update(s);
         break;
     case RTC_REG_B:
         if ( data & RTC_SET )
@@ -436,7 +438,8 @@ static int rtc_ioport_write(void *opaque
                 hvm_isa_irq_assert(d, RTC_IRQ);
             }
         s->hw.cmos_data[RTC_REG_B] = data;
-        rtc_timer_update(s);
+        if ( (data ^ orig) & RTC_PIE )
+            rtc_timer_update(s);
         check_update_timer(s);
         alarm_timer_update(s);
         break;




[-- Attachment #2: x86-hvm-rtc-periodic.patch --]
[-- Type: text/plain, Size: 1884 bytes --]

x86/HVM: RTC periodic timer emulation adjustments

- don't call rtc_timer_update() on REG_A writes when the value didn't
  change (doing the call always was reported to cause wall clock time
  lagging with the JVM running on Windows)
- don't call rtc_timer_update() on REG_B writes when RTC_PIE didn't
  change

Signed-off-by: Jan Beulich <jbeulich@suse.com>

---
v3: Break out just this change from the previously submitted much
    larger patch. The rest of that one is now planned to go in only
    after 4.2.

--- a/xen/arch/x86/hvm/rtc.c
+++ b/xen/arch/x86/hvm/rtc.c
@@ -365,6 +365,7 @@ static int rtc_ioport_write(void *opaque
 {
     RTCState *s = opaque;
     struct domain *d = vrtc_domain(s);
+    uint32_t orig;
 
     spin_lock(&s->lock);
 
@@ -382,6 +383,7 @@ static int rtc_ioport_write(void *opaque
         return 0;
     }
 
+    orig = s->hw.cmos_data[s->hw.cmos_index];
     switch ( s->hw.cmos_index )
     {
     case RTC_SECONDS_ALARM:
@@ -405,9 +407,9 @@ static int rtc_ioport_write(void *opaque
         break;
     case RTC_REG_A:
         /* UIP bit is read only */
-        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) |
-            (s->hw.cmos_data[RTC_REG_A] & RTC_UIP);
-        rtc_timer_update(s);
+        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) | (orig & RTC_UIP);
+        if ( (data ^ orig) & (RTC_RATE_SELECT | RTC_DIV_CTL) )
+            rtc_timer_update(s);
         break;
     case RTC_REG_B:
         if ( data & RTC_SET )
@@ -436,7 +438,8 @@ static int rtc_ioport_write(void *opaque
                 hvm_isa_irq_assert(d, RTC_IRQ);
             }
         s->hw.cmos_data[RTC_REG_B] = data;
-        rtc_timer_update(s);
+        if ( (data ^ orig) & RTC_PIE )
+            rtc_timer_update(s);
         check_update_timer(s);
         alarm_timer_update(s);
         break;

[-- Attachment #3: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH, v3] x86/HVM: RTC periodic timer emulation adjustments
  2012-08-31 16:01 [PATCH, v3] x86/HVM: RTC periodic timer emulation adjustments Jan Beulich
@ 2012-08-31 16:50 ` Keir Fraser
  2012-09-03  6:42   ` Jan Beulich
  0 siblings, 1 reply; 3+ messages in thread
From: Keir Fraser @ 2012-08-31 16:50 UTC (permalink / raw)
  To: Jan Beulich, xen-devel

On 31/08/2012 17:01, "Jan Beulich" <JBeulich@suse.com> wrote:

> - don't call rtc_timer_update() on REG_A writes when the value didn't
>   change (doing the call always was reported to cause wall clock time
>   lagging with the JVM running on Windows)
> - don't call rtc_timer_update() on REG_B writes when RTC_PIE didn't
>   change
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

One comment below in-line.

> ---
> v3: Break out just this change from the previously submitted much
>     larger patch. The rest of that one is now planned to go in only
>     after 4.2.
> 
> --- a/xen/arch/x86/hvm/rtc.c
> +++ b/xen/arch/x86/hvm/rtc.c
> @@ -365,6 +365,7 @@ static int rtc_ioport_write(void *opaque
>  {
>      RTCState *s = opaque;
>      struct domain *d = vrtc_domain(s);
> +    uint32_t orig;
>  
>      spin_lock(&s->lock);
>  
> @@ -382,6 +383,7 @@ static int rtc_ioport_write(void *opaque
>          return 0;
>      }
>  
> +    orig = s->hw.cmos_data[s->hw.cmos_index];
>      switch ( s->hw.cmos_index )
>      {
>      case RTC_SECONDS_ALARM:
> @@ -405,9 +407,9 @@ static int rtc_ioport_write(void *opaque
>          break;
>      case RTC_REG_A:
>          /* UIP bit is read only */
> -        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) |
> -            (s->hw.cmos_data[RTC_REG_A] & RTC_UIP);
> -        rtc_timer_update(s);
> +        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) | (orig & RTC_UIP);
> +        if ( (data ^ orig) & (RTC_RATE_SELECT | RTC_DIV_CTL) )

Please change to 'if ( (data ^ orig) & ~RTC_UIP )'. It is shorter and
matches the style of the immediately preceding line.

Once you make this change:
Acked-by: Keir Fraser <keir@xen.org>

 -- Keir

> +            rtc_timer_update(s);
>          break;
>      case RTC_REG_B:
>          if ( data & RTC_SET )
> @@ -436,7 +438,8 @@ static int rtc_ioport_write(void *opaque
>                  hvm_isa_irq_assert(d, RTC_IRQ);
>              }
>          s->hw.cmos_data[RTC_REG_B] = data;
> -        rtc_timer_update(s);
> +        if ( (data ^ orig) & RTC_PIE )
> +            rtc_timer_update(s);
>          check_update_timer(s);
>          alarm_timer_update(s);
>          break;
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH, v3] x86/HVM: RTC periodic timer emulation adjustments
  2012-08-31 16:50 ` Keir Fraser
@ 2012-09-03  6:42   ` Jan Beulich
  0 siblings, 0 replies; 3+ messages in thread
From: Jan Beulich @ 2012-09-03  6:42 UTC (permalink / raw)
  To: Keir Fraser; +Cc: xen-devel

>>> On 31.08.12 at 18:50, Keir Fraser <keir.xen@gmail.com> wrote:
> On 31/08/2012 17:01, "Jan Beulich" <JBeulich@suse.com> wrote:
>> @@ -405,9 +407,9 @@ static int rtc_ioport_write(void *opaque
>>          break;
>>      case RTC_REG_A:
>>          /* UIP bit is read only */
>> -        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) |
>> -            (s->hw.cmos_data[RTC_REG_A] & RTC_UIP);
>> -        rtc_timer_update(s);
>> +        s->hw.cmos_data[RTC_REG_A] = (data & ~RTC_UIP) | (orig & RTC_UIP);
>> +        if ( (data ^ orig) & (RTC_RATE_SELECT | RTC_DIV_CTL) )
> 
> Please change to 'if ( (data ^ orig) & ~RTC_UIP )'. It is shorter and
> matches the style of the immediately preceding line.
> 
> Once you make this change:
> Acked-by: Keir Fraser <keir@xen.org>

I made the change before committing, albeit I disagree - what
we care about here are explicitly the two fields that the original
version named. Their mask just accidentally happens to be the
complement of RTC_UIP. Whereas in the lines above we
specifically care about that one flag...

Jan

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2012-09-03  6:42 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-08-31 16:01 [PATCH, v3] x86/HVM: RTC periodic timer emulation adjustments Jan Beulich
2012-08-31 16:50 ` Keir Fraser
2012-09-03  6:42   ` Jan Beulich

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).