linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [GIT PULL clocksource] Clocksource watchdog commits for v5.15
@ 2021-08-12  0:01 Paul E. McKenney
  2021-08-12 13:46 ` Thomas Gleixner
  0 siblings, 1 reply; 5+ messages in thread
From: Paul E. McKenney @ 2021-08-12  0:01 UTC (permalink / raw)
  To: tglx; +Cc: linux-kernel, john.stultz, kernel-team, ak, rong.a.chen, sboyd

Hello, Thomas,

This pull request contains a single change that prevents clocksource
watchdog testing on systems with HZ < 100, thus preventing the integer
underflow that can occur on leisurely HZed systems.  This has been
posted to LKML:

https://lore.kernel.org/lkml/20210721212755.GA2066078@paulmck-ThinkPad-P17-Gen-1/

This has been tested by Rong Chen of Intel.  It has also been subjected
to subjected to the kbuild test robot and -next testing, and are available
in the git repository based on v5.14-rc2 at:

  git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git clocksource

for you to fetch changes up to 9b073961afabcf70d0804e472ea02fc6c739dcce:

  clocksource: Prohibit clocksource watchdog test when HZ<100 (2021-07-20 13:54:34 -0700)

----------------------------------------------------------------
Paul E. McKenney (1):
      clocksource: Prohibit clocksource watchdog test when HZ<100

 lib/Kconfig.debug | 1 +
 1 file changed, 1 insertion(+)

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [GIT PULL clocksource] Clocksource watchdog commits for v5.15
  2021-08-12  0:01 [GIT PULL clocksource] Clocksource watchdog commits for v5.15 Paul E. McKenney
@ 2021-08-12 13:46 ` Thomas Gleixner
  2021-08-12 16:37   ` Paul E. McKenney
  0 siblings, 1 reply; 5+ messages in thread
From: Thomas Gleixner @ 2021-08-12 13:46 UTC (permalink / raw)
  To: paulmck; +Cc: linux-kernel, john.stultz, kernel-team, ak, rong.a.chen, sboyd

On Wed, Aug 11 2021 at 17:01, Paul E. McKenney wrote:
> This pull request contains a single change that prevents clocksource
> watchdog testing on systems with HZ < 100, thus preventing the integer
> underflow that can occur on leisurely HZed systems.  This has been
> posted to LKML:
>
> https://lore.kernel.org/lkml/20210721212755.GA2066078@paulmck-ThinkPad-P17-Gen-1/

So with HZ < 100 .mult overflows, but why not simply adjusting the
mult, shift value to be

      .mult	= TICK_NSEC,
      .shift	= 0,

which is effectively the same as

      .mult	= TICK_NSEC << 8,
      .shift	= 8,

Hmm?

Thanks,

        tglx

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [GIT PULL clocksource] Clocksource watchdog commits for v5.15
  2021-08-12 13:46 ` Thomas Gleixner
@ 2021-08-12 16:37   ` Paul E. McKenney
  2021-08-12 17:11     ` Thomas Gleixner
  0 siblings, 1 reply; 5+ messages in thread
From: Paul E. McKenney @ 2021-08-12 16:37 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: linux-kernel, john.stultz, kernel-team, ak, rong.a.chen, sboyd

On Thu, Aug 12, 2021 at 03:46:42PM +0200, Thomas Gleixner wrote:
> On Wed, Aug 11 2021 at 17:01, Paul E. McKenney wrote:
> > This pull request contains a single change that prevents clocksource
> > watchdog testing on systems with HZ < 100, thus preventing the integer
> > underflow that can occur on leisurely HZed systems.  This has been
> > posted to LKML:
> >
> > https://lore.kernel.org/lkml/20210721212755.GA2066078@paulmck-ThinkPad-P17-Gen-1/
> 
> So with HZ < 100 .mult overflows, but why not simply adjusting the
> mult, shift value to be
> 
>       .mult	= TICK_NSEC,
>       .shift	= 0,
> 
> which is effectively the same as
> 
>       .mult	= TICK_NSEC << 8,
>       .shift	= 8,
> 
> Hmm?

Another option would be for me to be less lazy and to move this code:

	/* Since jiffies uses a simple TICK_NSEC multiplier
	 * conversion, the .shift value could be zero. However
	 * this would make NTP adjustments impossible as they are
	 * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
	 * shift both the nominator and denominator the same
	 * amount, and give ntp adjustments in units of 1/2^8
	 *
	 * The value 8 is somewhat carefully chosen, as anything
	 * larger can result in overflows. TICK_NSEC grows as HZ
	 * shrinks, so values greater than 8 overflow 32bits when
	 * HZ=100.
	 */
	#if HZ < 34
	#define JIFFIES_SHIFT	6
	#elif HZ < 67
	#define JIFFIES_SHIFT	7
	#else
	#define JIFFIES_SHIFT	8
	#endif

from kernel/time/jiffies.c to include/linux/clocksource.h.

Then remove this from kernel/time/clocksource-wdtest.c:

	/* Assume HZ > 100. */
	#define JIFFIES_SHIFT	8

Then I could get rid of the HZ < 100 restriction.

So how about as shown below?

							Thanx, Paul

------------------------------------------------------------------------

commit 413933be37676419414fc7cd03e333c8eaf8a2db
Author: Paul E. McKenney <paulmck@kernel.org>
Date:   Thu Aug 12 09:31:28 2021 -0700

    clocksource: Make clocksource-wdtest.c safe for slow-HZ systems
    
    Currently, clocksource-wdtest.c sets a local JIFFIES_SHIFT macro for
    operation at HZ>=67, which can cause this test suite to fail on systems
    with HZ<67.  Therefore, move the HZ-based definitions of JIFFIES_SHIFT
    from kernel/time/jiffies.c to include/linux/clocksource.h, allowing
    the local JIFFIES_SHIFT macro to be removed from clocksource-wdtest.c
    in favor of a properly HZ-based definition.  This in turn makes
    clocksource-wdtest.c safe for slow-HZ systems.
    
    Cc: John Stultz <john.stultz@linaro.org>
    Cc: Thomas Gleixner <tglx@linutronix.de>
    Cc: Stephen Boyd <sboyd@kernel.org>
    Cc: Rong Chen <rong.a.chen@intel.com>
    Signed-off-by: Paul E. McKenney <paulmck@kernel.org>

diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
index 1d42d4b173271..61b9a132a7e00 100644
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -294,4 +294,24 @@ static inline void timer_probe(void) {}
 extern ulong max_cswd_read_retries;
 void clocksource_verify_percpu(struct clocksource *cs);
 
+/* Since jiffies uses a simple TICK_NSEC multiplier
+ * conversion, the .shift value could be zero. However
+ * this would make NTP adjustments impossible as they are
+ * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
+ * shift both the nominator and denominator the same
+ * amount, and give ntp adjustments in units of 1/2^8
+ *
+ * The value 8 is somewhat carefully chosen, as anything
+ * larger can result in overflows. TICK_NSEC grows as HZ
+ * shrinks, so values greater than 8 overflow 32bits when
+ * HZ=100.
+ */
+#if HZ < 34
+#define JIFFIES_SHIFT	6
+#elif HZ < 67
+#define JIFFIES_SHIFT	7
+#else
+#define JIFFIES_SHIFT	8
+#endif
+
 #endif /* _LINUX_CLOCKSOURCE_H */
diff --git a/kernel/time/clocksource-wdtest.c b/kernel/time/clocksource-wdtest.c
index b72a969f7b938..781d8dc69be47 100644
--- a/kernel/time/clocksource-wdtest.c
+++ b/kernel/time/clocksource-wdtest.c
@@ -34,9 +34,6 @@ static u64 wdtest_jiffies_read(struct clocksource *cs)
 	return (u64)jiffies;
 }
 
-/* Assume HZ > 100. */
-#define JIFFIES_SHIFT	8
-
 static struct clocksource clocksource_wdtest_jiffies = {
 	.name			= "wdtest-jiffies",
 	.rating			= 1, /* lowest valid rating*/
diff --git a/kernel/time/jiffies.c b/kernel/time/jiffies.c
index 01935aafdb460..74f4b292900d1 100644
--- a/kernel/time/jiffies.c
+++ b/kernel/time/jiffies.c
@@ -12,26 +12,6 @@
 #include "timekeeping.h"
 
 
-/* Since jiffies uses a simple TICK_NSEC multiplier
- * conversion, the .shift value could be zero. However
- * this would make NTP adjustments impossible as they are
- * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
- * shift both the nominator and denominator the same
- * amount, and give ntp adjustments in units of 1/2^8
- *
- * The value 8 is somewhat carefully chosen, as anything
- * larger can result in overflows. TICK_NSEC grows as HZ
- * shrinks, so values greater than 8 overflow 32bits when
- * HZ=100.
- */
-#if HZ < 34
-#define JIFFIES_SHIFT	6
-#elif HZ < 67
-#define JIFFIES_SHIFT	7
-#else
-#define JIFFIES_SHIFT	8
-#endif
-
 static u64 jiffies_read(struct clocksource *cs)
 {
 	return (u64) jiffies;

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [GIT PULL clocksource] Clocksource watchdog commits for v5.15
  2021-08-12 16:37   ` Paul E. McKenney
@ 2021-08-12 17:11     ` Thomas Gleixner
  2021-08-12 20:01       ` Paul E. McKenney
  0 siblings, 1 reply; 5+ messages in thread
From: Thomas Gleixner @ 2021-08-12 17:11 UTC (permalink / raw)
  To: paulmck; +Cc: linux-kernel, john.stultz, kernel-team, ak, rong.a.chen, sboyd

On Thu, Aug 12 2021 at 09:37, Paul E. McKenney wrote:

> On Thu, Aug 12, 2021 at 03:46:42PM +0200, Thomas Gleixner wrote:
>> On Wed, Aug 11 2021 at 17:01, Paul E. McKenney wrote:
>> > This pull request contains a single change that prevents clocksource
>> > watchdog testing on systems with HZ < 100, thus preventing the integer
>> > underflow that can occur on leisurely HZed systems.  This has been
>> > posted to LKML:
>> >
>> > https://lore.kernel.org/lkml/20210721212755.GA2066078@paulmck-ThinkPad-P17-Gen-1/
>> 
>> So with HZ < 100 .mult overflows, but why not simply adjusting the
>> mult, shift value to be
>> 
>>       .mult	= TICK_NSEC,
>>       .shift	= 0,
>> 
>> which is effectively the same as
>> 
>>       .mult	= TICK_NSEC << 8,
>>       .shift	= 8,
>> 
>> Hmm?
>
> Another option would be for me to be less lazy and to move this code:
>
> 	/* Since jiffies uses a simple TICK_NSEC multiplier
> 	 * conversion, the .shift value could be zero. However
> 	 * this would make NTP adjustments impossible as they are
> 	 * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
> 	 * shift both the nominator and denominator the same
> 	 * amount, and give ntp adjustments in units of 1/2^8
> 	 *
> 	 * The value 8 is somewhat carefully chosen, as anything
> 	 * larger can result in overflows. TICK_NSEC grows as HZ
> 	 * shrinks, so values greater than 8 overflow 32bits when
> 	 * HZ=100.
> 	 */
> 	#if HZ < 34
> 	#define JIFFIES_SHIFT	6
> 	#elif HZ < 67
> 	#define JIFFIES_SHIFT	7
> 	#else
> 	#define JIFFIES_SHIFT	8
> 	#endif
>
> from kernel/time/jiffies.c to include/linux/clocksource.h.

No need to expose this globaly.

  kernel/time/tick-internal.h or kernel/time/jiffies.h

Thanks,

        tglx

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [GIT PULL clocksource] Clocksource watchdog commits for v5.15
  2021-08-12 17:11     ` Thomas Gleixner
@ 2021-08-12 20:01       ` Paul E. McKenney
  0 siblings, 0 replies; 5+ messages in thread
From: Paul E. McKenney @ 2021-08-12 20:01 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: linux-kernel, john.stultz, kernel-team, ak, rong.a.chen, sboyd

On Thu, Aug 12, 2021 at 07:11:55PM +0200, Thomas Gleixner wrote:
> On Thu, Aug 12 2021 at 09:37, Paul E. McKenney wrote:
> 
> > On Thu, Aug 12, 2021 at 03:46:42PM +0200, Thomas Gleixner wrote:
> >> On Wed, Aug 11 2021 at 17:01, Paul E. McKenney wrote:
> >> > This pull request contains a single change that prevents clocksource
> >> > watchdog testing on systems with HZ < 100, thus preventing the integer
> >> > underflow that can occur on leisurely HZed systems.  This has been
> >> > posted to LKML:
> >> >
> >> > https://lore.kernel.org/lkml/20210721212755.GA2066078@paulmck-ThinkPad-P17-Gen-1/
> >> 
> >> So with HZ < 100 .mult overflows, but why not simply adjusting the
> >> mult, shift value to be
> >> 
> >>       .mult	= TICK_NSEC,
> >>       .shift	= 0,
> >> 
> >> which is effectively the same as
> >> 
> >>       .mult	= TICK_NSEC << 8,
> >>       .shift	= 8,
> >> 
> >> Hmm?
> >
> > Another option would be for me to be less lazy and to move this code:
> >
> > 	/* Since jiffies uses a simple TICK_NSEC multiplier
> > 	 * conversion, the .shift value could be zero. However
> > 	 * this would make NTP adjustments impossible as they are
> > 	 * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
> > 	 * shift both the nominator and denominator the same
> > 	 * amount, and give ntp adjustments in units of 1/2^8
> > 	 *
> > 	 * The value 8 is somewhat carefully chosen, as anything
> > 	 * larger can result in overflows. TICK_NSEC grows as HZ
> > 	 * shrinks, so values greater than 8 overflow 32bits when
> > 	 * HZ=100.
> > 	 */
> > 	#if HZ < 34
> > 	#define JIFFIES_SHIFT	6
> > 	#elif HZ < 67
> > 	#define JIFFIES_SHIFT	7
> > 	#else
> > 	#define JIFFIES_SHIFT	8
> > 	#endif
> >
> > from kernel/time/jiffies.c to include/linux/clocksource.h.
> 
> No need to expose this globaly.
> 
>   kernel/time/tick-internal.h or kernel/time/jiffies.h

How about like this?

						Thanx, Paul

------------------------------------------------------------------------
 clocksource-wdtest.c |    5 ++---
 jiffies.c            |   21 +--------------------
 tick-internal.h      |   20 ++++++++++++++++++++
 3 files changed, 23 insertions(+), 23 deletions(-)

commit 2419f70beeb885ad96b832c64648e42559d7cf1d
Author: Paul E. McKenney <paulmck@kernel.org>
Date:   Thu Aug 12 09:31:28 2021 -0700

    clocksource: Make clocksource-wdtest.c safe for slow-HZ systems
    
    Currently, clocksource-wdtest.c sets a local JIFFIES_SHIFT macro for
    operation at HZ>=67, which can cause this test suite to fail on systems
    with HZ<67.  Therefore, move the HZ-based definitions of JIFFIES_SHIFT
    from kernel/time/jiffies.c to kernel/time/tick-internal.h, allowing
    the local JIFFIES_SHIFT macro to be removed from clocksource-wdtest.c
    in favor of a properly HZ-based definition.  This in turn makes
    clocksource-wdtest.c safe for slow-HZ systems.
    
    [ paulmck: Moved JIFFIES_SHIFT from include/linux/clocksource.h. ]
    
    Link: https://lore.kernel.org/lkml/20210812000133.GA402890@paulmck-ThinkPad-P17-Gen-1/
    Cc: John Stultz <john.stultz@linaro.org>
    Cc: Thomas Gleixner <tglx@linutronix.de>
    Cc: Stephen Boyd <sboyd@kernel.org>
    Cc: Rong Chen <rong.a.chen@intel.com>
    Signed-off-by: Paul E. McKenney <paulmck@kernel.org>

diff --git a/kernel/time/clocksource-wdtest.c b/kernel/time/clocksource-wdtest.c
index b72a969f7b938..7e82500c400b9 100644
--- a/kernel/time/clocksource-wdtest.c
+++ b/kernel/time/clocksource-wdtest.c
@@ -19,6 +19,8 @@
 #include <linux/prandom.h>
 #include <linux/cpu.h>
 
+#include "tick-internal.h"
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Paul E. McKenney <paulmck@kernel.org>");
 
@@ -34,9 +36,6 @@ static u64 wdtest_jiffies_read(struct clocksource *cs)
 	return (u64)jiffies;
 }
 
-/* Assume HZ > 100. */
-#define JIFFIES_SHIFT	8
-
 static struct clocksource clocksource_wdtest_jiffies = {
 	.name			= "wdtest-jiffies",
 	.rating			= 1, /* lowest valid rating*/
diff --git a/kernel/time/jiffies.c b/kernel/time/jiffies.c
index 01935aafdb460..bc4db9e5ab70c 100644
--- a/kernel/time/jiffies.c
+++ b/kernel/time/jiffies.c
@@ -10,28 +10,9 @@
 #include <linux/init.h>
 
 #include "timekeeping.h"
+#include "tick-internal.h"
 
 
-/* Since jiffies uses a simple TICK_NSEC multiplier
- * conversion, the .shift value could be zero. However
- * this would make NTP adjustments impossible as they are
- * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
- * shift both the nominator and denominator the same
- * amount, and give ntp adjustments in units of 1/2^8
- *
- * The value 8 is somewhat carefully chosen, as anything
- * larger can result in overflows. TICK_NSEC grows as HZ
- * shrinks, so values greater than 8 overflow 32bits when
- * HZ=100.
- */
-#if HZ < 34
-#define JIFFIES_SHIFT	6
-#elif HZ < 67
-#define JIFFIES_SHIFT	7
-#else
-#define JIFFIES_SHIFT	8
-#endif
-
 static u64 jiffies_read(struct clocksource *cs)
 {
 	return (u64) jiffies;
diff --git a/kernel/time/tick-internal.h b/kernel/time/tick-internal.h
index 6a742a29e545f..5459bab2f4f70 100644
--- a/kernel/time/tick-internal.h
+++ b/kernel/time/tick-internal.h
@@ -165,3 +165,23 @@ DECLARE_PER_CPU(struct hrtimer_cpu_base, hrtimer_bases);
 
 extern u64 get_next_timer_interrupt(unsigned long basej, u64 basem);
 void timer_clear_idle(void);
+
+/* Since jiffies uses a simple TICK_NSEC multiplier
+ * conversion, the .shift value could be zero. However
+ * this would make NTP adjustments impossible as they are
+ * in units of 1/2^.shift. Thus we use JIFFIES_SHIFT to
+ * shift both the nominator and denominator the same
+ * amount, and give ntp adjustments in units of 1/2^8
+ *
+ * The value 8 is somewhat carefully chosen, as anything
+ * larger can result in overflows. TICK_NSEC grows as HZ
+ * shrinks, so values greater than 8 overflow 32bits when
+ * HZ=100.
+ */
+#if HZ < 34
+#define JIFFIES_SHIFT	6
+#elif HZ < 67
+#define JIFFIES_SHIFT	7
+#else
+#define JIFFIES_SHIFT	8
+#endif

^ permalink raw reply related	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2021-08-12 20:01 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-12  0:01 [GIT PULL clocksource] Clocksource watchdog commits for v5.15 Paul E. McKenney
2021-08-12 13:46 ` Thomas Gleixner
2021-08-12 16:37   ` Paul E. McKenney
2021-08-12 17:11     ` Thomas Gleixner
2021-08-12 20:01       ` Paul E. McKenney

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).