[PATCH v3 0/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[PATCH v3 0/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[Prasad Sodagudi]

These patches are introducing support for global deferrable timers
and this feature is very useful for the mobile chipsets. All comments
in initial discussion [1] are addressed. Here is the latest discussion - [2]
as these  patches are in downstream kernel for quite a long time.

[1]- https://lore.kernel.org/patchwork/patch/500541/
[2]- https://lkml.org/lkml/2020/3/16/147

Changelog:
v2 -> v3:
 - Fixed 'kbuild test robot' warnings.

v1 -> v2:
 - Fixed tglx comments about warning.
 - Thanks Pavan for your suggestion.

Joonwoo Park (1):
  timer: make deferrable cpu unbound timers really not bound to a cpu

Prasad Sodagudi (1):
  sched: Add a check for cpu unbound deferrable timers

 include/linux/timer.h    |  3 +++
 kernel/time/tick-sched.c |  8 +++++-
 kernel/time/timer.c      | 69 ++++++++++++++++++++++++++++++++++++++++++++----
 3 files changed, 74 insertions(+), 6 deletions(-)

-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[Prasad Sodagudi]

From: Joonwoo Park <joonwoop@codeaurora.org>

When a deferrable work (INIT_DEFERRABLE_WORK, etc.) is queued via
queue_delayed_work() it's probably intended to run the work item on any
CPU that isn't idle. However, we queue the work to run at a later time
by starting a deferrable timer that binds to whatever CPU the work is
queued on which is same with queue_delayed_work_on(smp_processor_id())
effectively.

As a result WORK_CPU_UNBOUND work items aren't really cpu unbound now.
In fact this is perfectly fine with UP kernel and also won't affect much a
system without dyntick with SMP kernel too as every cpus run timers
periodically.  But on SMP systems with dyntick current implementation leads
deferrable timers not very scalable because the timer's base which has
queued the deferrable timer won't wake up till next non-deferrable timer
expires even though there are possible other non idle cpus are running
which are able to run expired deferrable timers.

The deferrable work is a good example of the current implementation's
victim like below.

INIT_DEFERRABLE_WORK(&dwork, fn);
CPU 0                                 CPU 1
queue_delayed_work(wq, &dwork, HZ);
    queue_delayed_work_on(WORK_CPU_UNBOUND);
        ...
	__mod_timer() -> queues timer to the
			 current cpu's timer
			 base.
	...
tick_nohz_idle_enter() -> cpu enters idle.
A second later
cpu 0 is now in idle.                 cpu 1 exits idle or wasn't in idle so
                                      now it's in active but won't
cpu 0 won't wake up till next         handle cpu unbound deferrable timer
non-deferrable timer expires.         as it's in cpu 0's timer base.

To make all cpu unbound deferrable timers are scalable, introduce a common
timer base which is only for cpu unbound deferrable timers to make those
are indeed cpu unbound so that can be scheduled by any of non idle cpus.
This common timer fixes scalability issue of delayed work and all other cpu
unbound deferrable timer using implementations.

Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
Signed-off-by: Prasad Sodagudi <psodagud@codeaurora.org>
---
 kernel/time/timer.c | 42 +++++++++++++++++++++++++++++++++++++-----
 1 file changed, 37 insertions(+), 5 deletions(-)

diff --git a/kernel/time/timer.c b/kernel/time/timer.c
index a5221ab..1bf9b49 100644
--- a/kernel/time/timer.c
+++ b/kernel/time/timer.c
@@ -220,6 +220,7 @@ static void timer_update_keys(struct work_struct *work);
 static DECLARE_WORK(timer_update_work, timer_update_keys);
 
 #ifdef CONFIG_SMP
+struct timer_base timer_base_deferrable;
 unsigned int sysctl_timer_migration = 1;
 
 DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
@@ -841,8 +842,14 @@ static inline struct timer_base *get_timer_cpu_base(u32 tflags, u32 cpu)
 	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
 	 * to use the deferrable base.
 	 */
-	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
-		base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
+	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) {
+#ifdef CONFIG_SMP
+		base = &timer_base_deferrable;
+#endif
+		if (tflags & TIMER_PINNED)
+			base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
+	}
+
 	return base;
 }
 
@@ -854,8 +861,14 @@ static inline struct timer_base *get_timer_this_cpu_base(u32 tflags)
 	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
 	 * to use the deferrable base.
 	 */
-	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
-		base = this_cpu_ptr(&timer_bases[BASE_DEF]);
+	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) {
+#ifdef CONFIG_SMP
+		base = &timer_base_deferrable;
+#endif
+		if (tflags & TIMER_PINNED)
+			base = this_cpu_ptr(&timer_bases[BASE_DEF]);
+	}
+
 	return base;
 }
 
@@ -1785,8 +1798,14 @@ static __latent_entropy void run_timer_softirq(struct softirq_action *h)
 	struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
 
 	__run_timers(base);
-	if (IS_ENABLED(CONFIG_NO_HZ_COMMON))
+	if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
 		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
+#ifdef CONFIG_SMP
+		if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
+				tick_do_timer_cpu == smp_processor_id())
+			__run_timers(&timer_base_deferrable);
+#endif
+	}
 }
 
 /*
@@ -2025,6 +2044,16 @@ static void __init init_timer_cpu(int cpu)
 	}
 }
 
+#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
+static void __init init_timer_deferrable_global(void)
+{
+	timer_base_deferrable.cpu = nr_cpu_ids;
+	raw_spin_lock_init(&timer_base_deferrable.lock);
+	timer_base_deferrable.clk = jiffies;
+	timer_base_init_expiry_lock(&timer_base_deferrable);
+}
+#endif
+
 static void __init init_timer_cpus(void)
 {
 	int cpu;
@@ -2036,6 +2065,9 @@ static void __init init_timer_cpus(void)
 void __init init_timers(void)
 {
 	init_timer_cpus();
+#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
+	init_timer_deferrable_global();
+#endif
 	open_softirq(TIMER_SOFTIRQ, run_timer_softirq);
 }
 
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

[PATCH v3 2/2] sched: Add a check for cpu unbound deferrable timers

[Prasad Sodagudi]

Add a check to find expired unbound deferrable timers
and trigger softirq for handling timers. This way a CPU
can process all the expired deferrable timers whenever
it is out off idle state due to an interrupt.

Signed-off-by: Prasad Sodagudi <psodagud@codeaurora.org>
---
 include/linux/timer.h    |  3 +++
 kernel/time/tick-sched.c |  8 +++++++-
 kernel/time/timer.c      | 29 ++++++++++++++++++++++++++++-
 3 files changed, 38 insertions(+), 2 deletions(-)

diff --git a/include/linux/timer.h b/include/linux/timer.h
index 0dc19a8..e85dd2d 100644
--- a/include/linux/timer.h
+++ b/include/linux/timer.h
@@ -172,6 +172,9 @@ extern int del_timer(struct timer_list * timer);
 extern int mod_timer(struct timer_list *timer, unsigned long expires);
 extern int mod_timer_pending(struct timer_list *timer, unsigned long expires);
 extern int timer_reduce(struct timer_list *timer, unsigned long expires);
+#ifdef CONFIG_SMP
+extern bool check_pending_deferrable_timers(int cpu);
+#endif
 
 /*
  * The jiffies value which is added to now, when there is no timer
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 3e2dc9b..16aec80 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -23,6 +23,7 @@
 #include <linux/module.h>
 #include <linux/irq_work.h>
 #include <linux/posix-timers.h>
+#include <linux/timer.h>
 #include <linux/context_tracking.h>
 #include <linux/mm.h>
 
@@ -1274,8 +1275,13 @@ static inline void tick_nohz_irq_enter(void)
 	now = ktime_get();
 	if (ts->idle_active)
 		tick_nohz_stop_idle(ts, now);
-	if (ts->tick_stopped)
+	if (ts->tick_stopped) {
 		tick_nohz_update_jiffies(now);
+#ifdef CONFIG_SMP
+		if (check_pending_deferrable_timers(smp_processor_id()))
+			raise_softirq_irqoff(TIMER_SOFTIRQ);
+#endif
+	}
 }
 
 #else
diff --git a/kernel/time/timer.c b/kernel/time/timer.c
index 1bf9b49..5947c63 100644
--- a/kernel/time/timer.c
+++ b/kernel/time/timer.c
@@ -221,6 +221,7 @@ static DECLARE_WORK(timer_update_work, timer_update_keys);
 
 #ifdef CONFIG_SMP
 struct timer_base timer_base_deferrable;
+static atomic_t deferrable_pending;
 unsigned int sysctl_timer_migration = 1;
 
 DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
@@ -1610,6 +1611,31 @@ static u64 cmp_next_hrtimer_event(u64 basem, u64 expires)
 	return DIV_ROUND_UP_ULL(nextevt, TICK_NSEC) * TICK_NSEC;
 }
 
+
+#ifdef CONFIG_SMP
+/*
+ * check_pending_deferrable_timers - Check for unbound deferrable timer expiry
+ * @cpu - Current CPU
+ *
+ * The function checks whether any global deferrable pending timers
+ * are exipired or not. This function does not check cpu bounded
+ * diferrable pending timers expiry.
+ *
+ * The function returns true when a cpu unbounded deferrable timer is expired.
+ */
+bool check_pending_deferrable_timers(int cpu)
+{
+	if (cpu == tick_do_timer_cpu ||
+		tick_do_timer_cpu == TICK_DO_TIMER_NONE) {
+		if (time_after_eq(jiffies, timer_base_deferrable.clk)
+			&& !atomic_cmpxchg(&deferrable_pending, 0, 1)) {
+			return true;
+		}
+	}
+	return false;
+}
+#endif
+
 /**
  * get_next_timer_interrupt - return the time (clock mono) of the next timer
  * @basej:	base time jiffies
@@ -1801,7 +1827,8 @@ static __latent_entropy void run_timer_softirq(struct softirq_action *h)
 	if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
 		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
 #ifdef CONFIG_SMP
-		if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
+		if ((atomic_cmpxchg(&deferrable_pending, 1, 0) &&
+				tick_do_timer_cpu == TICK_DO_TIMER_NONE) ||
 				tick_do_timer_cpu == smp_processor_id())
 			__run_timers(&timer_base_deferrable);
 #endif
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 2895 bytes --]

Hi Prasad,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on tip/timers/core]
[also build test ERROR on tip/auto-latest tip/timers/nohz v5.7-rc4 next-20200501]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Prasad-Sodagudi/timer-make-deferrable-cpu-unbound-timers-really-not-bound-to-a-cpu/20200503-025049
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 4479730e9263befbb9ce9563a09563db2acb8f7c
config: riscv-nommu_virt_defconfig (attached as .config)
compiler: riscv64-linux-gcc (GCC) 9.3.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day GCC_VERSION=9.3.0 make.cross ARCH=riscv 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   kernel/time/timer.c: In function 'get_timer_cpu_base':
>> kernel/time/timer.c:847:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     847 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c:847:11: note: each undeclared identifier is reported only once for each function it appears in
   kernel/time/timer.c: In function 'get_timer_this_cpu_base':
   kernel/time/timer.c:866:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     866 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c: In function 'run_timer_softirq':
   kernel/time/timer.c:1805:18: error: 'timer_base_deferrable' undeclared (first use in this function)
    1805 |    __run_timers(&timer_base_deferrable);
         |                  ^~~~~~~~~~~~~~~~~~~~~

vim +/timer_base_deferrable +847 kernel/time/timer.c

   836	
   837	static inline struct timer_base *get_timer_cpu_base(u32 tflags, u32 cpu)
   838	{
   839		struct timer_base *base = per_cpu_ptr(&timer_bases[BASE_STD], cpu);
   840	
   841		/*
   842		 * If the timer is deferrable and NO_HZ_COMMON is set then we need
   843		 * to use the deferrable base.
   844		 */
   845		if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) {
   846	#ifdef CONFIG_SMP
 > 847			base = &timer_base_deferrable;
   848	#endif
   849			if (tflags & TIMER_PINNED)
   850				base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
   851		}
   852	
   853		return base;
   854	}
   855	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 6789 bytes --]

Re: [PATCH v3 2/2] sched: Add a check for cpu unbound deferrable timers

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3174 bytes --]

Hi Prasad,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on tip/timers/core]
[also build test ERROR on tip/auto-latest tip/timers/nohz v5.7-rc4 next-20200501]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Prasad-Sodagudi/timer-make-deferrable-cpu-unbound-timers-really-not-bound-to-a-cpu/20200503-025049
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 4479730e9263befbb9ce9563a09563db2acb8f7c
config: riscv-nommu_virt_defconfig (attached as .config)
compiler: riscv64-linux-gcc (GCC) 9.3.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day GCC_VERSION=9.3.0 make.cross ARCH=riscv 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   kernel/time/timer.c: In function 'get_timer_cpu_base':
   kernel/time/timer.c:848:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     848 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c:848:11: note: each undeclared identifier is reported only once for each function it appears in
   kernel/time/timer.c: In function 'get_timer_this_cpu_base':
   kernel/time/timer.c:867:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     867 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c: In function 'run_timer_softirq':
>> kernel/time/timer.c:1829:24: error: 'deferrable_pending' undeclared (first use in this function)
    1829 |   if ((atomic_cmpxchg(&deferrable_pending, 1, 0) &&
         |                        ^~~~~~~~~~~~~~~~~~
   kernel/time/timer.c:1832:18: error: 'timer_base_deferrable' undeclared (first use in this function)
    1832 |    __run_timers(&timer_base_deferrable);
         |                  ^~~~~~~~~~~~~~~~~~~~~

vim +/deferrable_pending +1829 kernel/time/timer.c

  1817	
  1818	/*
  1819	 * This function runs timers and the timer-tq in bottom half context.
  1820	 */
  1821	static __latent_entropy void run_timer_softirq(struct softirq_action *h)
  1822	{
  1823		struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
  1824	
  1825		__run_timers(base);
  1826		if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
  1827			__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
  1828	#ifdef CONFIG_SMP
> 1829			if ((atomic_cmpxchg(&deferrable_pending, 1, 0) &&
  1830					tick_do_timer_cpu == TICK_DO_TIMER_NONE) ||
  1831					tick_do_timer_cpu == smp_processor_id())
  1832				__run_timers(&timer_base_deferrable);
  1833	#endif
  1834		}
  1835	}
  1836	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 6789 bytes --]

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3360 bytes --]

Hi Prasad,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on tip/timers/core]
[also build test ERROR on tip/auto-latest tip/timers/nohz v5.7-rc4 next-20200504]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Prasad-Sodagudi/timer-make-deferrable-cpu-unbound-timers-really-not-bound-to-a-cpu/20200503-025049
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git 4479730e9263befbb9ce9563a09563db2acb8f7c
config: ia64-randconfig-a001-20200505 (attached as .config)
compiler: ia64-linux-gcc (GCC) 9.3.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day GCC_VERSION=9.3.0 make.cross ARCH=ia64 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   kernel/time/timer.c: In function 'get_timer_cpu_base':
   kernel/time/timer.c:847:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     847 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c:847:11: note: each undeclared identifier is reported only once for each function it appears in
   kernel/time/timer.c: In function 'get_timer_this_cpu_base':
   kernel/time/timer.c:866:11: error: 'timer_base_deferrable' undeclared (first use in this function)
     866 |   base = &timer_base_deferrable;
         |           ^~~~~~~~~~~~~~~~~~~~~
   kernel/time/timer.c: In function 'run_timer_softirq':
>> kernel/time/timer.c:1803:7: error: 'tick_do_timer_cpu' undeclared (first use in this function); did you mean 'tick_dep_clear_cpu'?
    1803 |   if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
         |       ^~~~~~~~~~~~~~~~~
         |       tick_dep_clear_cpu
>> kernel/time/timer.c:1803:28: error: 'TICK_DO_TIMER_NONE' undeclared (first use in this function)
    1803 |   if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
         |                            ^~~~~~~~~~~~~~~~~~
   kernel/time/timer.c:1805:18: error: 'timer_base_deferrable' undeclared (first use in this function)
    1805 |    __run_timers(&timer_base_deferrable);
         |                  ^~~~~~~~~~~~~~~~~~~~~

vim +1803 kernel/time/timer.c

  1791	
  1792	/*
  1793	 * This function runs timers and the timer-tq in bottom half context.
  1794	 */
  1795	static __latent_entropy void run_timer_softirq(struct softirq_action *h)
  1796	{
  1797		struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
  1798	
  1799		__run_timers(base);
  1800		if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
  1801			__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
  1802	#ifdef CONFIG_SMP
> 1803			if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
  1804					tick_do_timer_cpu == smp_processor_id())
  1805				__run_timers(&timer_base_deferrable);
  1806	#endif
  1807		}
  1808	}
  1809	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 31009 bytes --]

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[Thomas Gleixner]

Prasad Sodagudi <psodagud@codeaurora.org> writes:
> To make all cpu unbound deferrable timers are scalable, introduce a common
> timer base which is only for cpu unbound deferrable timers to make those
> are indeed cpu unbound so that can be scheduled by any of non idle cpus.
> This common timer fixes scalability issue of delayed work and all other cpu
> unbound deferrable timer using implementations.

Scalability? That's really the wrong term here. A global timer base is
the opposite and you really want to explain why this is not creating a
scalability problem on large systems.

>  #ifdef CONFIG_SMP
> +struct timer_base timer_base_deferrable;
>  unsigned int sysctl_timer_migration = 1;
>  
>  DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
> @@ -841,8 +842,14 @@ static inline struct timer_base *get_timer_cpu_base(u32 tflags, u32 cpu)
>  	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
>  	 * to use the deferrable base.
>  	 */
> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
> -		base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) {
> +#ifdef CONFIG_SMP
> +		base = &timer_base_deferrable;
> +#endif

There are definitely smarter ways of solving this than sprinkling
#ifdef's around the code.

> +		if (tflags & TIMER_PINNED)
> +			base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
> +	}
> +
>  	return base;
>  }
> @@ -1785,8 +1798,14 @@ static __latent_entropy void run_timer_softirq(struct softirq_action *h)
>  	struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
>  
>  	__run_timers(base);
> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON))
> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
>  		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
> +#ifdef CONFIG_SMP
> +		if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
> +				tick_do_timer_cpu == smp_processor_id())
> +			__run_timers(&timer_base_deferrable);
> +#endif

Again, this can be solved in readable ways. Just slapping #ifdefs all
over the place is sloppy and lazy.

Aside of that accessing the tick internals here open coded is just a
layering violation.

> +	}
>  }
>  
>  /*
> @@ -2025,6 +2044,16 @@ static void __init init_timer_cpu(int cpu)
>  	}
>  }
>  
> +#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
> +static void __init init_timer_deferrable_global(void)
> +{
> +	timer_base_deferrable.cpu = nr_cpu_ids;

This was obviously never tested with CONFIG_DEBUG_PER_CPU_MAPS=y as this
will simply result in out of bounds accesses.

>  static void __init init_timer_cpus(void)
>  {
>  	int cpu;
> @@ -2036,6 +2065,9 @@ static void __init init_timer_cpus(void)
>  void __init init_timers(void)
>  {
>  	init_timer_cpus();
> +#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
> +	init_timer_deferrable_global();
> +#endif

Stub functions exist to avoid this unreadable #ifdef garbage.

Thanks,

        tglx

Re: [PATCH v3 2/2] sched: Add a check for cpu unbound deferrable timers

[Thomas Gleixner]

Prasad Sodagudi <psodagud@codeaurora.org> writes:

> Subject: sched: Add...

How is this related to the scheduler?

> Add a check to find expired unbound deferrable timers
> and trigger softirq for handling timers. This way a CPU
> can process all the expired deferrable timers whenever
> it is out off idle state due to an interrupt.

A little bit more context would be useful.

> Signed-off-by: Prasad Sodagudi <psodagud@codeaurora.org>
> ---
>  include/linux/timer.h    |  3 +++
>  kernel/time/tick-sched.c |  8 +++++++-
>  kernel/time/timer.c      | 29 ++++++++++++++++++++++++++++-
>  3 files changed, 38 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/timer.h b/include/linux/timer.h
> index 0dc19a8..e85dd2d 100644
> --- a/include/linux/timer.h
> +++ b/include/linux/timer.h
> @@ -172,6 +172,9 @@ extern int del_timer(struct timer_list * timer);
>  extern int mod_timer(struct timer_list *timer, unsigned long expires);
>  extern int mod_timer_pending(struct timer_list *timer, unsigned long expires);
>  extern int timer_reduce(struct timer_list *timer, unsigned long expires);
> +#ifdef CONFIG_SMP

This #ifdef is useless because ...

> +extern bool check_pending_deferrable_timers(int cpu);
> +#endif
>  
>  /*
>   * The jiffies value which is added to now, when there is no timer
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 3e2dc9b..16aec80 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -23,6 +23,7 @@
>  #include <linux/module.h>
>  #include <linux/irq_work.h>
>  #include <linux/posix-timers.h>
> +#include <linux/timer.h>
>  #include <linux/context_tracking.h>
>  #include <linux/mm.h>
>  
> @@ -1274,8 +1275,13 @@ static inline void tick_nohz_irq_enter(void)
>  	now = ktime_get();
>  	if (ts->idle_active)
>  		tick_nohz_stop_idle(ts, now);
> -	if (ts->tick_stopped)
> +	if (ts->tick_stopped) {
>  		tick_nohz_update_jiffies(now);
> +#ifdef CONFIG_SMP
> +		if (check_pending_deferrable_timers(smp_processor_id()))
> +			raise_softirq_irqoff(TIMER_SOFTIRQ);
> +#endif

... you failed to provide a stub function which avoids this #ifdef

> +	}
>  }
>  
>  #else
> diff --git a/kernel/time/timer.c b/kernel/time/timer.c
> index 1bf9b49..5947c63 100644
> --- a/kernel/time/timer.c
> +++ b/kernel/time/timer.c
> @@ -221,6 +221,7 @@ static DECLARE_WORK(timer_update_work, timer_update_keys);
>  
>  #ifdef CONFIG_SMP
>  struct timer_base timer_base_deferrable;
> +static atomic_t deferrable_pending;
>  unsigned int sysctl_timer_migration = 1;
>  
>  DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
> @@ -1610,6 +1611,31 @@ static u64 cmp_next_hrtimer_event(u64 basem, u64 expires)
>  	return DIV_ROUND_UP_ULL(nextevt, TICK_NSEC) * TICK_NSEC;
>  }
>  
> +
> +#ifdef CONFIG_SMP
> +/*
> + * check_pending_deferrable_timers - Check for unbound deferrable timer expiry
> + * @cpu - Current CPU

Bogus doc format.

> + *
> + * The function checks whether any global deferrable pending timers
> + * are exipired or not. This function does not check cpu bounded
> + * diferrable pending timers expiry.

Editors have spell check.

> + *
> + * The function returns true when a cpu unbounded deferrable timer is expired.
> + */
> +bool check_pending_deferrable_timers(int cpu)
> +{
> +	if (cpu == tick_do_timer_cpu ||
> +		tick_do_timer_cpu == TICK_DO_TIMER_NONE) {

The second line conditional wants to be aligned with the first line
conditional.

> +		if (time_after_eq(jiffies, timer_base_deferrable.clk)
> +			&& !atomic_cmpxchg(&deferrable_pending, 0, 1)) {

Ditto. Aside of that this deferrable pending magic lacks any form of
explanation.

> +			return true;
> +		}
> +	}
> +	return false;
> +}
> +#endif
> +
>  /**
>   * get_next_timer_interrupt - return the time (clock mono) of the next timer
>   * @basej:	base time jiffies
> @@ -1801,7 +1827,8 @@ static __latent_entropy void run_timer_softirq(struct softirq_action *h)
>  	if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
>  		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
>  #ifdef CONFIG_SMP
> -		if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
> +		if ((atomic_cmpxchg(&deferrable_pending, 1, 0) &&

How is that supposed to compile with NOHZ=n?

For every version of these patches the 0-day robot is complaining about
exactly the same problem:

    Your testing is solely done with one config, i.e. the config you are
    interested in.

Is it really so hard to compile test for a total of 4 combinations of
SMP and NOHZ?

Thanks,

        tglx

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[psodagud]

Hi Tglx,

On 2020-05-06 06:28, Thomas Gleixner wrote:
> Prasad Sodagudi <psodagud@codeaurora.org> writes:
>> To make all cpu unbound deferrable timers are scalable, introduce a 
>> common
>> timer base which is only for cpu unbound deferrable timers to make 
>> those
>> are indeed cpu unbound so that can be scheduled by any of non idle 
>> cpus.
>> This common timer fixes scalability issue of delayed work and all 
>> other cpu
>> unbound deferrable timer using implementations.
> 
> Scalability? That's really the wrong term here. A global timer base is
> the opposite and you really want to explain why this is not creating a
> scalability problem on large systems.
> 
>>  #ifdef CONFIG_SMP
>> +struct timer_base timer_base_deferrable;
>>  unsigned int sysctl_timer_migration = 1;
>> 
>>  DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
>> @@ -841,8 +842,14 @@ static inline struct timer_base 
>> *get_timer_cpu_base(u32 tflags, u32 cpu)
>>  	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
>>  	 * to use the deferrable base.
>>  	 */
>> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
>> -		base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
>> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) 
>> {
>> +#ifdef CONFIG_SMP
>> +		base = &timer_base_deferrable;
>> +#endif
> 
> There are definitely smarter ways of solving this than sprinkling
> #ifdef's around the code.

I am able to understand all other comments and I will address all those 
comments in the next patch set.
It is not clear to me how to avoid #ifdef's in this case. Could you 
please share an example here?


> 
>> +		if (tflags & TIMER_PINNED)
>> +			base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
>> +	}
>> +
>>  	return base;
>>  }
>> @@ -1785,8 +1798,14 @@ static __latent_entropy void 
>> run_timer_softirq(struct softirq_action *h)
>>  	struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
>> 
>>  	__run_timers(base);
>> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON))
>> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON)) {
>>  		__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
>> +#ifdef CONFIG_SMP
>> +		if (tick_do_timer_cpu == TICK_DO_TIMER_NONE ||
>> +				tick_do_timer_cpu == smp_processor_id())
>> +			__run_timers(&timer_base_deferrable);
>> +#endif
> 
> Again, this can be solved in readable ways. Just slapping #ifdefs all
> over the place is sloppy and lazy.
Sorry. I will try to address this.  It is not clear to me how to avoid 
#ifdefs in this case too.
Please provide me more information.

> 
> Aside of that accessing the tick internals here open coded is just a
> layering violation.
I will fix this and avoid using referring to tick internals here.

> 
>> +	}
>>  }
>> 
>>  /*
>> @@ -2025,6 +2044,16 @@ static void __init init_timer_cpu(int cpu)
>>  	}
>>  }
>> 
>> +#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
>> +static void __init init_timer_deferrable_global(void)
>> +{
>> +	timer_base_deferrable.cpu = nr_cpu_ids;
> 
> This was obviously never tested with CONFIG_DEBUG_PER_CPU_MAPS=y as 
> this
> will simply result in out of bounds accesses.

Sure. I will test this CONFIG_DEBUG_PER_CPU_MAPS=y enabled before 
pushing the next patch set.
> 
>>  static void __init init_timer_cpus(void)
>>  {
>>  	int cpu;
>> @@ -2036,6 +2065,9 @@ static void __init init_timer_cpus(void)
>>  void __init init_timers(void)
>>  {
>>  	init_timer_cpus();
>> +#if defined(CONFIG_NO_HZ_COMMON) && defined(CONFIG_SMP)
>> +	init_timer_deferrable_global();
>> +#endif
> 
> Stub functions exist to avoid this unreadable #ifdef garbage.
Got it. I will fix this in the next patch set.

> 
> Thanks,
> 
>         tglx

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[Thomas Gleixner]

psodagud@codeaurora.org writes:
> On 2020-05-06 06:28, Thomas Gleixner wrote:
>>>  #ifdef CONFIG_SMP
>>> +struct timer_base timer_base_deferrable;
>>>  unsigned int sysctl_timer_migration = 1;
>>> 
>>>  DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
>>> @@ -841,8 +842,14 @@ static inline struct timer_base 
>>> *get_timer_cpu_base(u32 tflags, u32 cpu)
>>>  	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
>>>  	 * to use the deferrable base.
>>>  	 */
>>> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
>>> -		base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
>>> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE)) 
>>> {
>>> +#ifdef CONFIG_SMP
>>> +		base = &timer_base_deferrable;
>>> +#endif
>> 
>> There are definitely smarter ways of solving this than sprinkling
>> #ifdef's around the code.
>
> I am able to understand all other comments and I will address all those 
> comments in the next patch set.
> It is not clear to me how to avoid #ifdef's in this case. Could you 
> please share an example here?

The answer is further down already:
 
>> Stub functions exist to avoid this unreadable #ifdef garbage.

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[psodagud]

On 2020-05-13 13:28, Thomas Gleixner wrote:
> psodagud@codeaurora.org writes:
>> On 2020-05-06 06:28, Thomas Gleixner wrote:
>>>>  #ifdef CONFIG_SMP
>>>> +struct timer_base timer_base_deferrable;
>>>>  unsigned int sysctl_timer_migration = 1;
>>>> 
>>>>  DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
>>>> @@ -841,8 +842,14 @@ static inline struct timer_base
>>>> *get_timer_cpu_base(u32 tflags, u32 cpu)
>>>>  	 * If the timer is deferrable and NO_HZ_COMMON is set then we need
>>>>  	 * to use the deferrable base.
>>>>  	 */
>>>> -	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & 
>>>> TIMER_DEFERRABLE))
>>>> -		base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
>>>> +	if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & 
>>>> TIMER_DEFERRABLE))
>>>> {
>>>> +#ifdef CONFIG_SMP
>>>> +		base = &timer_base_deferrable;
>>>> +#endif
>>> 
>>> There are definitely smarter ways of solving this than sprinkling
>>> #ifdef's around the code.
>> 
>> I am able to understand all other comments and I will address all 
>> those
>> comments in the next patch set.
>> It is not clear to me how to avoid #ifdef's in this case. Could you
>> please share an example here?
> 
> The answer is further down already:

Thanks Tglx for quick response.

I think, you are referring stub functions. Yes. I can reduce some of the 
#ifdefs with stub functions as you mentioned and not all the cases 
right?
I have introduced two variables timer_base_deferrable and 
deferrable_pending and I can put stub function where ever is possible. 
But it may not be appropriate to have stub function for all the 
references of these variables right? Correct me if my understanding is 
wrong.

-Thanks, Prasad

> 
>>> Stub functions exist to avoid this unreadable #ifdef garbage.

Re: [PATCH v3 1/2] timer: make deferrable cpu unbound timers really not bound to a cpu

[Thomas Gleixner]

Prasad,

psodagud@codeaurora.org writes:
> On 2020-05-13 13:28, Thomas Gleixner wrote:
>> psodagud@codeaurora.org writes:
>>> It is not clear to me how to avoid #ifdef's in this case. Could you
>>> please share an example here?
>> 
>> The answer is further down already:
>
> I think, you are referring stub functions. Yes. I can reduce some of the 
> #ifdefs with stub functions as you mentioned and not all the cases 
> right?
> I have introduced two variables timer_base_deferrable and 
> deferrable_pending and I can put stub function where ever is possible. 
> But it may not be appropriate to have stub function for all the 
> references of these variables right? Correct me if my understanding is 
> wrong.

Is this a quiz or are you expecting me to make your homework?

Thanks,

        Thomas