[patch 0/5] cpuidle haltpoll driver and governor (v4)

[patch 0/5] cpuidle haltpoll driver and governor (v4)

[Marcelo Tosatti]

The cpuidle-haltpoll driver with haltpoll governor allows the guest
vcpus to poll for a specified amount of time before halting. 
This provides the following benefits to host side polling:

         1) The POLL flag is set while polling is performed, which allows
            a remote vCPU to avoid sending an IPI (and the associated
            cost of handling the IPI) when performing a wakeup.

         2) The VM-exit cost can be avoided.

The downside of guest side polling is that polling is performed
even with other runnable tasks in the host.

Results comparing halt_poll_ns and server/client application
where a small packet is ping-ponged:

host                                        --> 31.33
halt_poll_ns=300000 / no guest busy spin    --> 33.40   (93.8%)
halt_poll_ns=0 / guest_halt_poll_ns=300000  --> 32.73   (95.7%)

For the SAP HANA benchmarks (where idle_spin is a parameter
of the previous version of the patch, results should be the
same):

hpns == halt_poll_ns

                           idle_spin=0/   idle_spin=800/    idle_spin=0/
                           hpns=200000    hpns=0            hpns=800000
DeleteC06T03 (100 thread) 1.76           1.71 (-3%)        1.78   (+1%)
InsertC16T02 (100 thread) 2.14           2.07 (-3%)        2.18   (+1.8%)
DeleteC00T01 (1 thread)   1.34           1.28 (-4.5%)      1.29   (-3.7%)
UpdateC00T03 (1 thread)   4.72           4.18 (-12%)       4.53   (-5%)

V2:

- Move from x86 to generic code (Paolo/Christian)
- Add auto-tuning logic (Paolo)
- Add MSR to disable host side polling (Paolo)

V3:

- Do not be specific about HLT VM-exit in the documentation (Ankur Arora)
- Mark tuning parameters static and __read_mostly (Andrea Arcangeli)
- Add WARN_ON if host does not support poll control (Joao Martins)
- Use sched_clock and cleanup haltpoll_enter_idle (Peter Zijlstra)
- Mark certain functions in kvm.c as static (kernel test robot)
- Remove tracepoints as they use RCU from extended quiescent state (kernel
test robot)

V4:
- Use a haltpoll governor, use poll_state.c poll code (Rafael J. Wysocki)

[patch 1/5] drivers/cpuidle: add cpuidle-haltpoll driver

[Marcelo Tosatti]

Add a cpuidle driver that calls the architecture default_idle routine.

To be used in conjunction with the haltpoll governor.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

---
 arch/x86/kernel/process.c          |    2 -
 drivers/cpuidle/Kconfig            |    9 +++++
 drivers/cpuidle/Makefile           |    1 
 drivers/cpuidle/cpuidle-haltpoll.c |   65 +++++++++++++++++++++++++++++++++++++
 4 files changed, 76 insertions(+), 1 deletion(-)

Index: linux-2.6.git/arch/x86/kernel/process.c
===================================================================
--- linux-2.6.git.orig/arch/x86/kernel/process.c	2019-06-13 16:19:27.877064340 -0400
+++ linux-2.6.git/arch/x86/kernel/process.c	2019-06-13 16:19:48.795544892 -0400
@@ -580,7 +580,7 @@
 	safe_halt();
 	trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
 }
-#ifdef CONFIG_APM_MODULE
+#if defined(CONFIG_APM_MODULE) || defined(CONFIG_HALTPOLL_CPUIDLE_MODULE)
 EXPORT_SYMBOL(default_idle);
 #endif
 
Index: linux-2.6.git/drivers/cpuidle/Kconfig
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/Kconfig	2019-06-13 16:19:27.878064316 -0400
+++ linux-2.6.git/drivers/cpuidle/Kconfig	2019-06-13 18:41:40.599912671 -0400
@@ -51,6 +51,15 @@
 source "drivers/cpuidle/Kconfig.powerpc"
 endmenu
 
+config HALTPOLL_CPUIDLE
+       tristate "Halt poll cpuidle driver"
+       depends on X86
+       default y
+       help
+         This option enables halt poll cpuidle driver, which allows to poll
+         before halting in the guest (more efficient than polling in the
+         host via halt_poll_ns for some scenarios).
+
 endif
 
 config ARCH_NEEDS_CPU_IDLE_COUPLED
Index: linux-2.6.git/drivers/cpuidle/Makefile
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/Makefile	2019-06-13 16:19:27.878064316 -0400
+++ linux-2.6.git/drivers/cpuidle/Makefile	2019-06-13 16:19:48.796544867 -0400
@@ -7,6 +7,7 @@
 obj-$(CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED) += coupled.o
 obj-$(CONFIG_DT_IDLE_STATES)		  += dt_idle_states.o
 obj-$(CONFIG_ARCH_HAS_CPU_RELAX)	  += poll_state.o
+obj-$(CONFIG_HALTPOLL_CPUIDLE)		  += cpuidle-haltpoll.o
 
 ##################################################################################
 # ARM SoC drivers
Index: linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c	2019-06-13 18:41:39.305933413 -0400
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * cpuidle driver for haltpoll governor.
+ *
+ * Copyright 2019 Red Hat, Inc. and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ * Authors: Marcelo Tosatti <mtosatti@redhat.com>
+ */
+
+#include <linux/init.h>
+#include <linux/cpuidle.h>
+#include <linux/module.h>
+#include <linux/sched/idle.h>
+
+static int default_enter_idle(struct cpuidle_device *dev,
+			      struct cpuidle_driver *drv, int index)
+{
+	if (current_clr_polling_and_test()) {
+		local_irq_enable();
+		return index;
+	}
+	default_idle();
+	return index;
+}
+
+static struct cpuidle_driver haltpoll_driver = {
+	.name = "haltpoll",
+	.owner = THIS_MODULE,
+	.states = {
+		{ /* entry 0 is for polling */ },
+		{
+			.enter			= default_enter_idle,
+			.exit_latency		= 0,
+			.target_residency	= 0,
+			.power_usage		= -1,
+			.name			= "haltpoll idle",
+			.desc			= "default architecture idle",
+		},
+	},
+	.safe_state_index = 0,
+	.state_count = 2,
+};
+
+static int __init haltpoll_init(void)
+{
+	struct cpuidle_driver *drv = &haltpoll_driver;
+
+	cpuidle_poll_state_init(drv);
+
+	return cpuidle_register(&haltpoll_driver, NULL);
+}
+
+static void __exit haltpoll_exit(void)
+{
+	cpuidle_unregister(&haltpoll_driver);
+}
+
+module_init(haltpoll_init);
+module_exit(haltpoll_exit);
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Marcelo Tosatti <mtosatti@redhat.com>");
+

Re: [patch 1/5] drivers/cpuidle: add cpuidle-haltpoll driver

[Rafael J. Wysocki]

On Fri, Jun 14, 2019 at 12:55 AM Marcelo Tosatti <mtosatti@redhat.com> wrote:
>
> Add a cpuidle driver that calls the architecture default_idle routine.
>
> To be used in conjunction with the haltpoll governor.
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
> ---
>  arch/x86/kernel/process.c          |    2 -
>  drivers/cpuidle/Kconfig            |    9 +++++
>  drivers/cpuidle/Makefile           |    1
>  drivers/cpuidle/cpuidle-haltpoll.c |   65 +++++++++++++++++++++++++++++++++++++
>  4 files changed, 76 insertions(+), 1 deletion(-)
>
> Index: linux-2.6.git/arch/x86/kernel/process.c
> ===================================================================
> --- linux-2.6.git.orig/arch/x86/kernel/process.c        2019-06-13 16:19:27.877064340 -0400
> +++ linux-2.6.git/arch/x86/kernel/process.c     2019-06-13 16:19:48.795544892 -0400
> @@ -580,7 +580,7 @@
>         safe_halt();
>         trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
>  }
> -#ifdef CONFIG_APM_MODULE
> +#if defined(CONFIG_APM_MODULE) || defined(CONFIG_HALTPOLL_CPUIDLE_MODULE)
>  EXPORT_SYMBOL(default_idle);
>  #endif
>
> Index: linux-2.6.git/drivers/cpuidle/Kconfig
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/Kconfig  2019-06-13 16:19:27.878064316 -0400
> +++ linux-2.6.git/drivers/cpuidle/Kconfig       2019-06-13 18:41:40.599912671 -0400
> @@ -51,6 +51,15 @@
>  source "drivers/cpuidle/Kconfig.powerpc"
>  endmenu
>
> +config HALTPOLL_CPUIDLE
> +       tristate "Halt poll cpuidle driver"
> +       depends on X86
> +       default y
> +       help
> +         This option enables halt poll cpuidle driver, which allows to poll
> +         before halting in the guest (more efficient than polling in the
> +         host via halt_poll_ns for some scenarios).
> +
>  endif
>
>  config ARCH_NEEDS_CPU_IDLE_COUPLED
> Index: linux-2.6.git/drivers/cpuidle/Makefile
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/Makefile 2019-06-13 16:19:27.878064316 -0400
> +++ linux-2.6.git/drivers/cpuidle/Makefile      2019-06-13 16:19:48.796544867 -0400
> @@ -7,6 +7,7 @@
>  obj-$(CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED) += coupled.o
>  obj-$(CONFIG_DT_IDLE_STATES)             += dt_idle_states.o
>  obj-$(CONFIG_ARCH_HAS_CPU_RELAX)         += poll_state.o
> +obj-$(CONFIG_HALTPOLL_CPUIDLE)           += cpuidle-haltpoll.o
>
>  ##################################################################################
>  # ARM SoC drivers
> Index: linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c
> ===================================================================
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c    2019-06-13 18:41:39.305933413 -0400
> @@ -0,0 +1,65 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * cpuidle driver for haltpoll governor.
> + *
> + * Copyright 2019 Red Hat, Inc. and/or its affiliates.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.  See
> + * the COPYING file in the top-level directory.
> + *
> + * Authors: Marcelo Tosatti <mtosatti@redhat.com>
> + */
> +
> +#include <linux/init.h>
> +#include <linux/cpuidle.h>
> +#include <linux/module.h>
> +#include <linux/sched/idle.h>
> +
> +static int default_enter_idle(struct cpuidle_device *dev,
> +                             struct cpuidle_driver *drv, int index)
> +{
> +       if (current_clr_polling_and_test()) {
> +               local_irq_enable();
> +               return index;
> +       }
> +       default_idle();
> +       return index;
> +}
> +
> +static struct cpuidle_driver haltpoll_driver = {
> +       .name = "haltpoll",
> +       .owner = THIS_MODULE,
> +       .states = {
> +               { /* entry 0 is for polling */ },
> +               {
> +                       .enter                  = default_enter_idle,
> +                       .exit_latency           = 0,
> +                       .target_residency       = 0,
> +                       .power_usage            = -1,
> +                       .name                   = "haltpoll idle",
> +                       .desc                   = "default architecture idle",
> +               },
> +       },
> +       .safe_state_index = 0,
> +       .state_count = 2,
> +};
> +
> +static int __init haltpoll_init(void)
> +{
> +       struct cpuidle_driver *drv = &haltpoll_driver;
> +
> +       cpuidle_poll_state_init(drv);
> +
> +       return cpuidle_register(&haltpoll_driver, NULL);

Would it be practical to prevent this driver from loading on non-virt?

> +}
> +
> +static void __exit haltpoll_exit(void)
> +{
> +       cpuidle_unregister(&haltpoll_driver);
> +}
> +
> +module_init(haltpoll_init);
> +module_exit(haltpoll_exit);
> +MODULE_LICENSE("GPL");
> +MODULE_AUTHOR("Marcelo Tosatti <mtosatti@redhat.com>");
> +

Apart from the above it is simple enough.

Re: [patch 1/5] drivers/cpuidle: add cpuidle-haltpoll driver

[Rafael J. Wysocki]

On Fri, Jun 14, 2019 at 12:55 AM Marcelo Tosatti <mtosatti@redhat.com> wrote:
>
> Add a cpuidle driver that calls the architecture default_idle routine.
>
> To be used in conjunction with the haltpoll governor.
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
> ---
>  arch/x86/kernel/process.c          |    2 -
>  drivers/cpuidle/Kconfig            |    9 +++++
>  drivers/cpuidle/Makefile           |    1
>  drivers/cpuidle/cpuidle-haltpoll.c |   65 +++++++++++++++++++++++++++++++++++++
>  4 files changed, 76 insertions(+), 1 deletion(-)
>
> Index: linux-2.6.git/arch/x86/kernel/process.c
> ===================================================================
> --- linux-2.6.git.orig/arch/x86/kernel/process.c        2019-06-13 16:19:27.877064340 -0400
> +++ linux-2.6.git/arch/x86/kernel/process.c     2019-06-13 16:19:48.795544892 -0400
> @@ -580,7 +580,7 @@
>         safe_halt();
>         trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, smp_processor_id());
>  }
> -#ifdef CONFIG_APM_MODULE
> +#if defined(CONFIG_APM_MODULE) || defined(CONFIG_HALTPOLL_CPUIDLE_MODULE)
>  EXPORT_SYMBOL(default_idle);
>  #endif
>
> Index: linux-2.6.git/drivers/cpuidle/Kconfig
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/Kconfig  2019-06-13 16:19:27.878064316 -0400
> +++ linux-2.6.git/drivers/cpuidle/Kconfig       2019-06-13 18:41:40.599912671 -0400
> @@ -51,6 +51,15 @@
>  source "drivers/cpuidle/Kconfig.powerpc"
>  endmenu
>
> +config HALTPOLL_CPUIDLE
> +       tristate "Halt poll cpuidle driver"
> +       depends on X86
> +       default y
> +       help
> +         This option enables halt poll cpuidle driver, which allows to poll
> +         before halting in the guest (more efficient than polling in the
> +         host via halt_poll_ns for some scenarios).
> +
>  endif
>
>  config ARCH_NEEDS_CPU_IDLE_COUPLED
> Index: linux-2.6.git/drivers/cpuidle/Makefile
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/Makefile 2019-06-13 16:19:27.878064316 -0400
> +++ linux-2.6.git/drivers/cpuidle/Makefile      2019-06-13 16:19:48.796544867 -0400
> @@ -7,6 +7,7 @@
>  obj-$(CONFIG_ARCH_NEEDS_CPU_IDLE_COUPLED) += coupled.o
>  obj-$(CONFIG_DT_IDLE_STATES)             += dt_idle_states.o
>  obj-$(CONFIG_ARCH_HAS_CPU_RELAX)         += poll_state.o
> +obj-$(CONFIG_HALTPOLL_CPUIDLE)           += cpuidle-haltpoll.o
>
>  ##################################################################################
>  # ARM SoC drivers
> Index: linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c
> ===================================================================
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c    2019-06-13 18:41:39.305933413 -0400
> @@ -0,0 +1,65 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * cpuidle driver for haltpoll governor.
> + *
> + * Copyright 2019 Red Hat, Inc. and/or its affiliates.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.  See
> + * the COPYING file in the top-level directory.
> + *
> + * Authors: Marcelo Tosatti <mtosatti@redhat.com>
> + */
> +
> +#include <linux/init.h>
> +#include <linux/cpuidle.h>
> +#include <linux/module.h>
> +#include <linux/sched/idle.h>
> +
> +static int default_enter_idle(struct cpuidle_device *dev,
> +                             struct cpuidle_driver *drv, int index)
> +{
> +       if (current_clr_polling_and_test()) {
> +               local_irq_enable();
> +               return index;
> +       }
> +       default_idle();
> +       return index;
> +}
> +
> +static struct cpuidle_driver haltpoll_driver = {
> +       .name = "haltpoll",
> +       .owner = THIS_MODULE,
> +       .states = {
> +               { /* entry 0 is for polling */ },
> +               {
> +                       .enter                  = default_enter_idle,
> +                       .exit_latency           = 0,
> +                       .target_residency       = 0,

On a second thought, I think that you need a target residency of at
least 1 (and an exit latency of at least 1) here to distinguish this
state from the polling one.

And if they both really were 0, polling wouldn't be needed at all ...

> +                       .power_usage            = -1,
> +                       .name                   = "haltpoll idle",
> +                       .desc                   = "default architecture idle",
> +               },
> +       },
> +       .safe_state_index = 0,
> +       .state_count = 2,
> +};
> +
> +static int __init haltpoll_init(void)
> +{
> +       struct cpuidle_driver *drv = &haltpoll_driver;
> +
> +       cpuidle_poll_state_init(drv);
> +
> +       return cpuidle_register(&haltpoll_driver, NULL);
> +}
> +
> +static void __exit haltpoll_exit(void)
> +{
> +       cpuidle_unregister(&haltpoll_driver);
> +}
> +
> +module_init(haltpoll_init);
> +module_exit(haltpoll_exit);
> +MODULE_LICENSE("GPL");
> +MODULE_AUTHOR("Marcelo Tosatti <mtosatti@redhat.com>");
> +
>
>

[patch 2/5] cpuidle: add get_poll_time callback

[Marcelo Tosatti]

Add a "get_poll_time" callback to the cpuidle_governor structure,
and change poll state to poll for that amount of time.

Provide a default method for it, while allowing individual governors
to override it.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

---
 drivers/cpuidle/cpuidle.c    |   40 ++++++++++++++++++++++++++++++++++++++++
 drivers/cpuidle/poll_state.c |   11 ++---------
 include/linux/cpuidle.h      |    8 ++++++++
 3 files changed, 50 insertions(+), 9 deletions(-)

Index: linux-2.6.git/drivers/cpuidle/cpuidle.c
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/cpuidle.c	2019-06-13 17:57:33.111185824 -0400
+++ linux-2.6.git/drivers/cpuidle/cpuidle.c	2019-06-13 18:09:48.158500660 -0400
@@ -362,6 +362,46 @@
 }
 
 /**
+ * cpuidle_default_poll_time - default routine used to return poll time
+ * governors can override it if necessary
+ *
+ * @drv:   the cpuidle driver tied with the cpu
+ * @dev:   the cpuidle device
+ *
+ */
+static u64 cpuidle_default_poll_time(struct cpuidle_driver *drv,
+				     struct cpuidle_device *dev)
+{
+	int i;
+
+	for (i = 1; i < drv->state_count; i++) {
+		if (drv->states[i].disabled || dev->states_usage[i].disable)
+			continue;
+
+		return (u64)drv->states[i].target_residency * NSEC_PER_USEC;
+	}
+
+	return TICK_NSEC;
+}
+
+/**
+ * cpuidle_get_poll_time - tell the polling driver how much time to poll,
+ *			   in nanoseconds.
+ *
+ * @drv: the cpuidle driver tied with the cpu
+ * @dev: the cpuidle device
+ *
+ */
+u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
+			  struct cpuidle_device *dev)
+{
+	if (cpuidle_curr_governor->get_poll_time)
+		return cpuidle_curr_governor->get_poll_time(drv, dev);
+
+	return cpuidle_default_poll_time(drv, dev);
+}
+
+/**
  * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
  */
 void cpuidle_install_idle_handler(void)
Index: linux-2.6.git/drivers/cpuidle/poll_state.c
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/poll_state.c	2019-06-13 17:57:33.111185824 -0400
+++ linux-2.6.git/drivers/cpuidle/poll_state.c	2019-06-13 18:01:19.846944820 -0400
@@ -20,16 +20,9 @@
 	local_irq_enable();
 	if (!current_set_polling_and_test()) {
 		unsigned int loop_count = 0;
-		u64 limit = TICK_NSEC;
-		int i;
+		u64 limit;
 
-		for (i = 1; i < drv->state_count; i++) {
-			if (drv->states[i].disabled || dev->states_usage[i].disable)
-				continue;
-
-			limit = (u64)drv->states[i].target_residency * NSEC_PER_USEC;
-			break;
-		}
+		limit = cpuidle_get_poll_time(drv, dev);
 
 		while (!need_resched()) {
 			cpu_relax();
Index: linux-2.6.git/include/linux/cpuidle.h
===================================================================
--- linux-2.6.git.orig/include/linux/cpuidle.h	2019-06-13 17:57:33.111185824 -0400
+++ linux-2.6.git/include/linux/cpuidle.h	2019-06-13 18:01:19.846944820 -0400
@@ -132,6 +132,8 @@
 extern int cpuidle_enter(struct cpuidle_driver *drv,
 			 struct cpuidle_device *dev, int index);
 extern void cpuidle_reflect(struct cpuidle_device *dev, int index);
+extern u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
+				 struct cpuidle_device *dev);
 
 extern int cpuidle_register_driver(struct cpuidle_driver *drv);
 extern struct cpuidle_driver *cpuidle_get_driver(void);
@@ -166,6 +168,9 @@
 				struct cpuidle_device *dev, int index)
 {return -ENODEV; }
 static inline void cpuidle_reflect(struct cpuidle_device *dev, int index) { }
+extern u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
+				 struct cpuidle_device *dev)
+{return 0; }
 static inline int cpuidle_register_driver(struct cpuidle_driver *drv)
 {return -ENODEV; }
 static inline struct cpuidle_driver *cpuidle_get_driver(void) {return NULL; }
@@ -246,6 +251,9 @@
 					struct cpuidle_device *dev,
 					bool *stop_tick);
 	void (*reflect)		(struct cpuidle_device *dev, int index);
+
+	u64 (*get_poll_time)	(struct cpuidle_driver *drv,
+					struct cpuidle_device *dev);
 };
 
 #ifdef CONFIG_CPU_IDLE

Re: [patch 2/5] cpuidle: add get_poll_time callback

[Rafael J. Wysocki]

On Fri, Jun 14, 2019 at 12:55 AM Marcelo Tosatti <mtosatti@redhat.com> wrote:
>
> Add a "get_poll_time" callback to the cpuidle_governor structure,
> and change poll state to poll for that amount of time.
>
> Provide a default method for it, while allowing individual governors
> to override it.
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Fair enough:

Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

>
> ---
>  drivers/cpuidle/cpuidle.c    |   40 ++++++++++++++++++++++++++++++++++++++++
>  drivers/cpuidle/poll_state.c |   11 ++---------
>  include/linux/cpuidle.h      |    8 ++++++++
>  3 files changed, 50 insertions(+), 9 deletions(-)
>
> Index: linux-2.6.git/drivers/cpuidle/cpuidle.c
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/cpuidle.c        2019-06-13 17:57:33.111185824 -0400
> +++ linux-2.6.git/drivers/cpuidle/cpuidle.c     2019-06-13 18:09:48.158500660 -0400
> @@ -362,6 +362,46 @@
>  }
>
>  /**
> + * cpuidle_default_poll_time - default routine used to return poll time
> + * governors can override it if necessary
> + *
> + * @drv:   the cpuidle driver tied with the cpu
> + * @dev:   the cpuidle device
> + *
> + */
> +static u64 cpuidle_default_poll_time(struct cpuidle_driver *drv,
> +                                    struct cpuidle_device *dev)
> +{
> +       int i;
> +
> +       for (i = 1; i < drv->state_count; i++) {
> +               if (drv->states[i].disabled || dev->states_usage[i].disable)
> +                       continue;
> +
> +               return (u64)drv->states[i].target_residency * NSEC_PER_USEC;
> +       }
> +
> +       return TICK_NSEC;
> +}
> +
> +/**
> + * cpuidle_get_poll_time - tell the polling driver how much time to poll,
> + *                        in nanoseconds.
> + *
> + * @drv: the cpuidle driver tied with the cpu
> + * @dev: the cpuidle device
> + *
> + */
> +u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
> +                         struct cpuidle_device *dev)
> +{
> +       if (cpuidle_curr_governor->get_poll_time)
> +               return cpuidle_curr_governor->get_poll_time(drv, dev);
> +
> +       return cpuidle_default_poll_time(drv, dev);
> +}
> +
> +/**
>   * cpuidle_install_idle_handler - installs the cpuidle idle loop handler
>   */
>  void cpuidle_install_idle_handler(void)
> Index: linux-2.6.git/drivers/cpuidle/poll_state.c
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/poll_state.c     2019-06-13 17:57:33.111185824 -0400
> +++ linux-2.6.git/drivers/cpuidle/poll_state.c  2019-06-13 18:01:19.846944820 -0400
> @@ -20,16 +20,9 @@
>         local_irq_enable();
>         if (!current_set_polling_and_test()) {
>                 unsigned int loop_count = 0;
> -               u64 limit = TICK_NSEC;
> -               int i;
> +               u64 limit;
>
> -               for (i = 1; i < drv->state_count; i++) {
> -                       if (drv->states[i].disabled || dev->states_usage[i].disable)
> -                               continue;
> -
> -                       limit = (u64)drv->states[i].target_residency * NSEC_PER_USEC;
> -                       break;
> -               }
> +               limit = cpuidle_get_poll_time(drv, dev);
>
>                 while (!need_resched()) {
>                         cpu_relax();
> Index: linux-2.6.git/include/linux/cpuidle.h
> ===================================================================
> --- linux-2.6.git.orig/include/linux/cpuidle.h  2019-06-13 17:57:33.111185824 -0400
> +++ linux-2.6.git/include/linux/cpuidle.h       2019-06-13 18:01:19.846944820 -0400
> @@ -132,6 +132,8 @@
>  extern int cpuidle_enter(struct cpuidle_driver *drv,
>                          struct cpuidle_device *dev, int index);
>  extern void cpuidle_reflect(struct cpuidle_device *dev, int index);
> +extern u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
> +                                struct cpuidle_device *dev);
>
>  extern int cpuidle_register_driver(struct cpuidle_driver *drv);
>  extern struct cpuidle_driver *cpuidle_get_driver(void);
> @@ -166,6 +168,9 @@
>                                 struct cpuidle_device *dev, int index)
>  {return -ENODEV; }
>  static inline void cpuidle_reflect(struct cpuidle_device *dev, int index) { }
> +extern u64 cpuidle_get_poll_time(struct cpuidle_driver *drv,
> +                                struct cpuidle_device *dev)
> +{return 0; }
>  static inline int cpuidle_register_driver(struct cpuidle_driver *drv)
>  {return -ENODEV; }
>  static inline struct cpuidle_driver *cpuidle_get_driver(void) {return NULL; }
> @@ -246,6 +251,9 @@
>                                         struct cpuidle_device *dev,
>                                         bool *stop_tick);
>         void (*reflect)         (struct cpuidle_device *dev, int index);
> +
> +       u64 (*get_poll_time)    (struct cpuidle_driver *drv,
> +                                       struct cpuidle_device *dev);
>  };
>
>  #ifdef CONFIG_CPU_IDLE
>
>

[patch 3/5] cpuidle: add haltpoll governor

[Marcelo Tosatti]

The cpuidle_haltpoll governor, in conjunction with the haltpoll cpuidle
driver, allows guest vcpus to poll for a specified amount of time before
halting.
This provides the following benefits to host side polling:

        1) The POLL flag is set while polling is performed, which allows
           a remote vCPU to avoid sending an IPI (and the associated
           cost of handling the IPI) when performing a wakeup.

        2) The VM-exit cost can be avoided.

The downside of guest side polling is that polling is performed
even with other runnable tasks in the host.

Results comparing halt_poll_ns and server/client application
where a small packet is ping-ponged:

host                                        --> 31.33
halt_poll_ns=300000 / no guest busy spin    --> 33.40   (93.8%)
halt_poll_ns=0 / guest_halt_poll_ns=300000  --> 32.73   (95.7%)

For the SAP HANA benchmarks (where idle_spin is a parameter
of the previous version of the patch, results should be the
same):

hpns == halt_poll_ns

                          idle_spin=0/   idle_spin=800/    idle_spin=0/
                          hpns=200000    hpns=0            hpns=800000
DeleteC06T03 (100 thread) 1.76           1.71 (-3%)        1.78   (+1%)
InsertC16T02 (100 thread) 2.14           2.07 (-3%)        2.18   (+1.8%)
DeleteC00T01 (1 thread)   1.34           1.28 (-4.5%)      1.29   (-3.7%)
UpdateC00T03 (1 thread)   4.72           4.18 (-12%)       4.53   (-5%)

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>


---
 Documentation/virtual/guest-halt-polling.txt |   79 ++++++++++++
 drivers/cpuidle/Kconfig                      |   11 +
 drivers/cpuidle/governors/Makefile           |    1 
 drivers/cpuidle/governors/haltpoll.c         |  175 +++++++++++++++++++++++++++
 4 files changed, 266 insertions(+)

Index: linux-2.6.git/drivers/cpuidle/Kconfig
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/Kconfig	2019-06-13 18:05:46.456294042 -0400
+++ linux-2.6.git/drivers/cpuidle/Kconfig	2019-06-13 18:14:58.981570277 -0400
@@ -33,6 +33,17 @@
 	  Some workloads benefit from using it and it generally should be safe
 	  to use.  Say Y here if you are not happy with the alternatives.
 
+config CPU_IDLE_GOV_HALTPOLL
+	bool "Haltpoll governor (for virtualized systems)"
+	depends on KVM_GUEST
+	help
+	  This governor implements haltpoll idle state selection, to be
+	  used in conjunction with the haltpoll cpuidle driver, allowing
+	  for polling for a certain amount of time before entering idle
+	  state.
+
+	  Some virtualized workloads benefit from using it.
+
 config DT_IDLE_STATES
 	bool
 
Index: linux-2.6.git/drivers/cpuidle/governors/Makefile
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/governors/Makefile	2019-06-13 18:05:46.456294042 -0400
+++ linux-2.6.git/drivers/cpuidle/governors/Makefile	2019-06-13 18:10:53.861444033 -0400
@@ -6,3 +6,4 @@
 obj-$(CONFIG_CPU_IDLE_GOV_LADDER) += ladder.o
 obj-$(CONFIG_CPU_IDLE_GOV_MENU) += menu.o
 obj-$(CONFIG_CPU_IDLE_GOV_TEO) += teo.o
+obj-$(CONFIG_CPU_IDLE_GOV_HALTPOLL) += haltpoll.o
Index: linux-2.6.git/drivers/cpuidle/governors/haltpoll.c
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6.git/drivers/cpuidle/governors/haltpoll.c	2019-06-13 18:12:46.581615748 -0400
@@ -0,0 +1,175 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * haltpoll.c - haltpoll idle governor
+ *
+ * Copyright 2019 Red Hat, Inc. and/or its affiliates.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ * Authors: Marcelo Tosatti <mtosatti@redhat.com>
+ */
+
+#include <linux/kernel.h>
+#include <linux/cpuidle.h>
+#include <linux/time.h>
+#include <linux/ktime.h>
+#include <linux/hrtimer.h>
+#include <linux/tick.h>
+#include <linux/sched.h>
+#include <linux/module.h>
+#include <linux/kvm_para.h>
+
+static unsigned int guest_halt_poll_us __read_mostly = 200;
+module_param(guest_halt_poll_us, uint, 0644);
+
+/* division factor to shrink halt_poll_us */
+static unsigned int guest_halt_poll_shrink __read_mostly = 2;
+module_param(guest_halt_poll_shrink, uint, 0644);
+
+/* multiplication factor to grow per-cpu halt_poll_us */
+static unsigned int guest_halt_poll_grow __read_mostly = 2;
+module_param(guest_halt_poll_grow, uint, 0644);
+
+/* value in us to start growing per-cpu halt_poll_us */
+static unsigned int guest_halt_poll_grow_start __read_mostly = 50;
+module_param(guest_halt_poll_grow_start, uint, 0644);
+
+/* allow shrinking guest halt poll */
+static bool guest_halt_poll_allow_shrink __read_mostly = true;
+module_param(guest_halt_poll_allow_shrink, bool, 0644);
+
+struct haltpoll_device {
+	int		last_state_idx;
+	unsigned int	halt_poll_us;
+};
+
+static DEFINE_PER_CPU_ALIGNED(struct haltpoll_device, hpoll_devices);
+
+/**
+ * haltpoll_select - selects the next idle state to enter
+ * @drv: cpuidle driver containing state data
+ * @dev: the CPU
+ * @stop_tick: indication on whether or not to stop the tick
+ */
+static int haltpoll_select(struct cpuidle_driver *drv,
+			   struct cpuidle_device *dev,
+			   bool *stop_tick)
+{
+	struct haltpoll_device *hdev = this_cpu_ptr(&hpoll_devices);
+
+	if (!drv->state_count) {
+		*stop_tick = false;
+		return 0;
+	}
+
+	if (hdev->halt_poll_us == 0)
+		return 1;
+
+	/* Last state was poll? */
+	if (hdev->last_state_idx == 0) {
+		/* Halt if no event occurred on poll window */
+		if (dev->poll_time_limit == true)
+			return 1;
+
+		*stop_tick = false;
+		/* Otherwise, poll again */
+		return 0;
+	}
+
+	*stop_tick = false;
+	/* Last state was halt: poll */
+	return 0;
+}
+
+static void adjust_haltpoll_us(unsigned int block_us,
+			       struct haltpoll_device *dev)
+{
+	unsigned int val;
+
+	/* Grow cpu_halt_poll_us if
+	 * cpu_halt_poll_us < block_ns < guest_halt_poll_us
+	 */
+	if (block_us > dev->halt_poll_us && block_us <= guest_halt_poll_us) {
+		val = dev->halt_poll_us * guest_halt_poll_grow;
+
+		if (val < guest_halt_poll_grow_start)
+			val = guest_halt_poll_grow_start;
+		if (val > guest_halt_poll_us)
+			val = guest_halt_poll_us;
+
+		dev->halt_poll_us = val;
+	} else if (block_us > guest_halt_poll_us &&
+		   guest_halt_poll_allow_shrink) {
+		unsigned int shrink = guest_halt_poll_shrink;
+
+		val = dev->halt_poll_us;
+		if (shrink == 0)
+			val = 0;
+		else
+			val /= shrink;
+		dev->halt_poll_us = val;
+	}
+}
+
+/**
+ * haltpoll_reflect - update variables and update poll time
+ * @dev: the CPU
+ * @index: the index of actual entered state
+ */
+static void haltpoll_reflect(struct cpuidle_device *dev, int index)
+{
+	struct haltpoll_device *hdev = this_cpu_ptr(&hpoll_devices);
+
+	hdev->last_state_idx = index;
+
+	if (index != 0)
+		adjust_haltpoll_us(dev->last_residency, hdev);
+}
+
+/**
+ * haltpoll_enable_device - scans a CPU's states and does setup
+ * @drv: cpuidle driver
+ * @dev: the CPU
+ */
+static int haltpoll_enable_device(struct cpuidle_driver *drv,
+				  struct cpuidle_device *dev)
+{
+	struct haltpoll_device *hdev = &per_cpu(hpoll_devices, dev->cpu);
+
+	memset(hdev, 0, sizeof(struct haltpoll_device));
+
+	return 0;
+}
+
+/**
+ * haltpoll_get_poll_time - return amount of poll time
+ * @drv: cpuidle driver
+ * @dev: the CPU
+ */
+static u64 haltpoll_get_poll_time(struct cpuidle_driver *drv,
+				struct cpuidle_device *dev)
+{
+	struct haltpoll_device *hdev = &per_cpu(hpoll_devices, dev->cpu);
+
+	return hdev->halt_poll_us * NSEC_PER_USEC;
+}
+
+static struct cpuidle_governor haltpoll_governor = {
+	.name =			"haltpoll",
+	.rating =		21,
+	.enable =		haltpoll_enable_device,
+	.select =		haltpoll_select,
+	.reflect =		haltpoll_reflect,
+	.get_poll_time =	haltpoll_get_poll_time,
+};
+
+static int __init init_haltpoll(void)
+{
+	if (kvm_para_available())
+		return cpuidle_register_governor(&haltpoll_governor);
+
+	return 0;
+}
+
+postcore_initcall(init_haltpoll);
Index: linux-2.6.git/Documentation/virtual/guest-halt-polling.txt
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6.git/Documentation/virtual/guest-halt-polling.txt	2019-06-13 18:16:22.414262777 -0400
@@ -0,0 +1,79 @@
+Guest halt polling
+==================
+
+The cpuidle_haltpoll driver, with the haltpoll governor, allows
+the guest vcpus to poll for a specified amount of time before
+halting.
+This provides the following benefits to host side polling:
+
+	1) The POLL flag is set while polling is performed, which allows
+	   a remote vCPU to avoid sending an IPI (and the associated
+ 	   cost of handling the IPI) when performing a wakeup.
+
+	2) The VM-exit cost can be avoided.
+
+The downside of guest side polling is that polling is performed
+even with other runnable tasks in the host.
+
+The basic logic as follows: A global value, guest_halt_poll_us,
+is configured by the user, indicating the maximum amount of
+time polling is allowed. This value is fixed.
+
+Each vcpu has an adjustable guest_halt_poll_us
+("per-cpu guest_halt_poll_us"), which is adjusted by the algorithm
+in response to events (explained below).
+
+Module Parameters
+=================
+
+The haltpoll governor has 5 tunable module parameters:
+
+1) guest_halt_poll_us:
+Maximum amount of time, in microseconds, that polling is
+performed before halting.
+
+Default: 200
+
+2) guest_halt_poll_shrink:
+Division factor used to shrink per-cpu guest_halt_poll_us when
+wakeup event occurs after the global guest_halt_poll_us.
+
+Default: 2
+
+3) guest_halt_poll_grow:
+Multiplication factor used to grow per-cpu guest_halt_poll_us
+when event occurs after per-cpu guest_halt_poll_us
+but before global guest_halt_poll_us.
+
+Default: 2
+
+4) guest_halt_poll_grow_start:
+The per-cpu guest_halt_poll_us eventually reaches zero
+in case of an idle system. This value sets the initial
+per-cpu guest_halt_poll_us when growing. This can
+be increased from 10, to avoid misses during the initial
+growth stage:
+
+10, 20, 40, ... (example assumes guest_halt_poll_grow=2).
+
+Default: 50
+
+5) guest_halt_poll_allow_shrink:
+
+Bool parameter which allows shrinking. Set to N
+to avoid it (per-cpu guest_halt_poll_us will remain
+high once achieves global guest_halt_poll_us value).
+
+Default: Y
+
+The module parameters can be set from the debugfs files in:
+
+	/sys/module/haltpoll/parameters/
+
+Further Notes
+=============
+
+- Care should be taken when setting the guest_halt_poll_us parameter as a
+large value has the potential to drive the cpu usage to 100% on a machine which
+would be almost entirely idle otherwise.
+

RE: [patch 3/5] cpuidle: add haltpoll governor

[Doug Smythies]

Hi,

I tried your patch set, but only to check
that they didn't cause any regression for situations
where idle state 0 (Poll) is used a lot (teo governor).

They didn't (my testing was not thorough).

I do not know if the below matters or not.

On 2019.06.13 15:46 Marcelo Tosatti wrote:

... [snip] ...

> Index: linux-2.6.git/Documentation/virtual/guest-halt-polling.txt
> ===================================================================
> --- /dev/null	1970-01-01 00:00:00.000000000 +0000
> +++ linux-2.6.git/Documentation/virtual/guest-halt-polling.txt	2019-06-13 18:16:22.414262777 -0400
> @@ -0,0 +1,79 @@
> +Guest halt polling
> +==================
> +
> +The cpuidle_haltpoll driver, with the haltpoll governor, allows
> +the guest vcpus to poll for a specified amount of time before
> +halting.
> +This provides the following benefits to host side polling:
> +
> +	1) The POLL flag is set while polling is performed, which allows
> +	   a remote vCPU to avoid sending an IPI (and the associated
> + 	   cost of handling the IPI) when performing a wakeup.
   ^
   |_ While applying the patches, git complains about this space character before the TAB.

It also complains about a few patches with a blank line before EOF.

... Doug

Re: [patch 3/5] cpuidle: add haltpoll governor

[Marcelo Tosatti]

On Thu, Jun 20, 2019 at 05:34:46PM -0700, Doug Smythies wrote:
> Hi,
> 
> I tried your patch set, but only to check
> that they didn't cause any regression for situations
> where idle state 0 (Poll) is used a lot (teo governor).
> 
> They didn't (my testing was not thorough).
> 
> I do not know if the below matters or not.
> 
> On 2019.06.13 15:46 Marcelo Tosatti wrote:
> 
> ... [snip] ...
> 
> > Index: linux-2.6.git/Documentation/virtual/guest-halt-polling.txt
> > ===================================================================
> > --- /dev/null	1970-01-01 00:00:00.000000000 +0000
> > +++ linux-2.6.git/Documentation/virtual/guest-halt-polling.txt	2019-06-13 18:16:22.414262777 -0400
> > @@ -0,0 +1,79 @@
> > +Guest halt polling
> > +==================
> > +
> > +The cpuidle_haltpoll driver, with the haltpoll governor, allows
> > +the guest vcpus to poll for a specified amount of time before
> > +halting.
> > +This provides the following benefits to host side polling:
> > +
> > +	1) The POLL flag is set while polling is performed, which allows
> > +	   a remote vCPU to avoid sending an IPI (and the associated
> > + 	   cost of handling the IPI) when performing a wakeup.
>    ^
>    |_ While applying the patches, git complains about this space character before the TAB.
> 
> It also complains about a few patches with a blank line before EOF.
> 
> ... Doug

Hi Doug,

Will fix those, thanks.

Re: [patch 3/5] cpuidle: add haltpoll governor

[Rafael J. Wysocki]

On Fri, Jun 14, 2019 at 12:55 AM Marcelo Tosatti <mtosatti@redhat.com> wrote:
>
> The cpuidle_haltpoll governor, in conjunction with the haltpoll cpuidle
> driver, allows guest vcpus to poll for a specified amount of time before
> halting.
> This provides the following benefits to host side polling:
>
>         1) The POLL flag is set while polling is performed, which allows
>            a remote vCPU to avoid sending an IPI (and the associated
>            cost of handling the IPI) when performing a wakeup.
>
>         2) The VM-exit cost can be avoided.
>
> The downside of guest side polling is that polling is performed
> even with other runnable tasks in the host.
>
> Results comparing halt_poll_ns and server/client application
> where a small packet is ping-ponged:
>
> host                                        --> 31.33
> halt_poll_ns=300000 / no guest busy spin    --> 33.40   (93.8%)
> halt_poll_ns=0 / guest_halt_poll_ns=300000  --> 32.73   (95.7%)
>
> For the SAP HANA benchmarks (where idle_spin is a parameter
> of the previous version of the patch, results should be the
> same):
>
> hpns == halt_poll_ns
>
>                           idle_spin=0/   idle_spin=800/    idle_spin=0/
>                           hpns=200000    hpns=0            hpns=800000
> DeleteC06T03 (100 thread) 1.76           1.71 (-3%)        1.78   (+1%)
> InsertC16T02 (100 thread) 2.14           2.07 (-3%)        2.18   (+1.8%)
> DeleteC00T01 (1 thread)   1.34           1.28 (-4.5%)      1.29   (-3.7%)
> UpdateC00T03 (1 thread)   4.72           4.18 (-12%)       4.53   (-5%)
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
>
> ---
>  Documentation/virtual/guest-halt-polling.txt |   79 ++++++++++++
>  drivers/cpuidle/Kconfig                      |   11 +
>  drivers/cpuidle/governors/Makefile           |    1
>  drivers/cpuidle/governors/haltpoll.c         |  175 +++++++++++++++++++++++++++
>  4 files changed, 266 insertions(+)
>
> Index: linux-2.6.git/drivers/cpuidle/Kconfig
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/Kconfig  2019-06-13 18:05:46.456294042 -0400
> +++ linux-2.6.git/drivers/cpuidle/Kconfig       2019-06-13 18:14:58.981570277 -0400
> @@ -33,6 +33,17 @@
>           Some workloads benefit from using it and it generally should be safe
>           to use.  Say Y here if you are not happy with the alternatives.
>
> +config CPU_IDLE_GOV_HALTPOLL
> +       bool "Haltpoll governor (for virtualized systems)"
> +       depends on KVM_GUEST
> +       help
> +         This governor implements haltpoll idle state selection, to be
> +         used in conjunction with the haltpoll cpuidle driver, allowing
> +         for polling for a certain amount of time before entering idle
> +         state.
> +
> +         Some virtualized workloads benefit from using it.
> +
>  config DT_IDLE_STATES
>         bool
>
> Index: linux-2.6.git/drivers/cpuidle/governors/Makefile
> ===================================================================
> --- linux-2.6.git.orig/drivers/cpuidle/governors/Makefile       2019-06-13 18:05:46.456294042 -0400
> +++ linux-2.6.git/drivers/cpuidle/governors/Makefile    2019-06-13 18:10:53.861444033 -0400
> @@ -6,3 +6,4 @@
>  obj-$(CONFIG_CPU_IDLE_GOV_LADDER) += ladder.o
>  obj-$(CONFIG_CPU_IDLE_GOV_MENU) += menu.o
>  obj-$(CONFIG_CPU_IDLE_GOV_TEO) += teo.o
> +obj-$(CONFIG_CPU_IDLE_GOV_HALTPOLL) += haltpoll.o
> Index: linux-2.6.git/drivers/cpuidle/governors/haltpoll.c
> ===================================================================
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ linux-2.6.git/drivers/cpuidle/governors/haltpoll.c  2019-06-13 18:12:46.581615748 -0400
> @@ -0,0 +1,175 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * haltpoll.c - haltpoll idle governor
> + *
> + * Copyright 2019 Red Hat, Inc. and/or its affiliates.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.  See
> + * the COPYING file in the top-level directory.
> + *
> + * Authors: Marcelo Tosatti <mtosatti@redhat.com>
> + */
> +
> +#include <linux/kernel.h>
> +#include <linux/cpuidle.h>
> +#include <linux/time.h>
> +#include <linux/ktime.h>
> +#include <linux/hrtimer.h>
> +#include <linux/tick.h>
> +#include <linux/sched.h>
> +#include <linux/module.h>
> +#include <linux/kvm_para.h>
> +
> +static unsigned int guest_halt_poll_us __read_mostly = 200;
> +module_param(guest_halt_poll_us, uint, 0644);
> +
> +/* division factor to shrink halt_poll_us */
> +static unsigned int guest_halt_poll_shrink __read_mostly = 2;
> +module_param(guest_halt_poll_shrink, uint, 0644);
> +
> +/* multiplication factor to grow per-cpu halt_poll_us */
> +static unsigned int guest_halt_poll_grow __read_mostly = 2;
> +module_param(guest_halt_poll_grow, uint, 0644);
> +
> +/* value in us to start growing per-cpu halt_poll_us */
> +static unsigned int guest_halt_poll_grow_start __read_mostly = 50;
> +module_param(guest_halt_poll_grow_start, uint, 0644);
> +
> +/* allow shrinking guest halt poll */
> +static bool guest_halt_poll_allow_shrink __read_mostly = true;
> +module_param(guest_halt_poll_allow_shrink, bool, 0644);
> +
> +struct haltpoll_device {
> +       int             last_state_idx;
> +       unsigned int    halt_poll_us;
> +};
> +
> +static DEFINE_PER_CPU_ALIGNED(struct haltpoll_device, hpoll_devices);
> +
> +/**
> + * haltpoll_select - selects the next idle state to enter
> + * @drv: cpuidle driver containing state data
> + * @dev: the CPU
> + * @stop_tick: indication on whether or not to stop the tick
> + */
> +static int haltpoll_select(struct cpuidle_driver *drv,
> +                          struct cpuidle_device *dev,
> +                          bool *stop_tick)
> +{
> +       struct haltpoll_device *hdev = this_cpu_ptr(&hpoll_devices);

What about PM QoS constraints?

Especially if the resume latency limit is set to 0?

> +
> +       if (!drv->state_count) {
> +               *stop_tick = false;
> +               return 0;
> +       }
> +
> +       if (hdev->halt_poll_us == 0)
> +               return 1;
> +
> +       /* Last state was poll? */
> +       if (hdev->last_state_idx == 0) {
> +               /* Halt if no event occurred on poll window */
> +               if (dev->poll_time_limit == true)
> +                       return 1;
> +
> +               *stop_tick = false;
> +               /* Otherwise, poll again */
> +               return 0;
> +       }
> +
> +       *stop_tick = false;
> +       /* Last state was halt: poll */
> +       return 0;
> +}
> +
> +static void adjust_haltpoll_us(unsigned int block_us,
> +                              struct haltpoll_device *dev)
> +{
> +       unsigned int val;
> +
> +       /* Grow cpu_halt_poll_us if
> +        * cpu_halt_poll_us < block_ns < guest_halt_poll_us
> +        */
> +       if (block_us > dev->halt_poll_us && block_us <= guest_halt_poll_us) {
> +               val = dev->halt_poll_us * guest_halt_poll_grow;
> +
> +               if (val < guest_halt_poll_grow_start)
> +                       val = guest_halt_poll_grow_start;
> +               if (val > guest_halt_poll_us)
> +                       val = guest_halt_poll_us;
> +
> +               dev->halt_poll_us = val;
> +       } else if (block_us > guest_halt_poll_us &&
> +                  guest_halt_poll_allow_shrink) {
> +               unsigned int shrink = guest_halt_poll_shrink;
> +
> +               val = dev->halt_poll_us;
> +               if (shrink == 0)
> +                       val = 0;
> +               else
> +                       val /= shrink;
> +               dev->halt_poll_us = val;
> +       }
> +}
> +
> +/**
> + * haltpoll_reflect - update variables and update poll time
> + * @dev: the CPU
> + * @index: the index of actual entered state
> + */
> +static void haltpoll_reflect(struct cpuidle_device *dev, int index)
> +{
> +       struct haltpoll_device *hdev = this_cpu_ptr(&hpoll_devices);
> +
> +       hdev->last_state_idx = index;
> +
> +       if (index != 0)
> +               adjust_haltpoll_us(dev->last_residency, hdev);
> +}
> +
> +/**
> + * haltpoll_enable_device - scans a CPU's states and does setup
> + * @drv: cpuidle driver
> + * @dev: the CPU
> + */
> +static int haltpoll_enable_device(struct cpuidle_driver *drv,
> +                                 struct cpuidle_device *dev)
> +{
> +       struct haltpoll_device *hdev = &per_cpu(hpoll_devices, dev->cpu);
> +
> +       memset(hdev, 0, sizeof(struct haltpoll_device));
> +
> +       return 0;
> +}
> +
> +/**
> + * haltpoll_get_poll_time - return amount of poll time
> + * @drv: cpuidle driver
> + * @dev: the CPU
> + */
> +static u64 haltpoll_get_poll_time(struct cpuidle_driver *drv,
> +                               struct cpuidle_device *dev)
> +{
> +       struct haltpoll_device *hdev = &per_cpu(hpoll_devices, dev->cpu);
> +
> +       return hdev->halt_poll_us * NSEC_PER_USEC;
> +}
> +
> +static struct cpuidle_governor haltpoll_governor = {
> +       .name =                 "haltpoll",
> +       .rating =               21,
> +       .enable =               haltpoll_enable_device,
> +       .select =               haltpoll_select,
> +       .reflect =              haltpoll_reflect,
> +       .get_poll_time =        haltpoll_get_poll_time,
> +};
> +
> +static int __init init_haltpoll(void)
> +{
> +       if (kvm_para_available())
> +               return cpuidle_register_governor(&haltpoll_governor);
> +
> +       return 0;
> +}
> +
> +postcore_initcall(init_haltpoll);
> Index: linux-2.6.git/Documentation/virtual/guest-halt-polling.txt
> ===================================================================
> --- /dev/null   1970-01-01 00:00:00.000000000 +0000
> +++ linux-2.6.git/Documentation/virtual/guest-halt-polling.txt  2019-06-13 18:16:22.414262777 -0400
> @@ -0,0 +1,79 @@
> +Guest halt polling
> +==================
> +
> +The cpuidle_haltpoll driver, with the haltpoll governor, allows
> +the guest vcpus to poll for a specified amount of time before
> +halting.
> +This provides the following benefits to host side polling:
> +
> +       1) The POLL flag is set while polling is performed, which allows
> +          a remote vCPU to avoid sending an IPI (and the associated
> +          cost of handling the IPI) when performing a wakeup.
> +
> +       2) The VM-exit cost can be avoided.
> +
> +The downside of guest side polling is that polling is performed
> +even with other runnable tasks in the host.
> +
> +The basic logic as follows: A global value, guest_halt_poll_us,
> +is configured by the user, indicating the maximum amount of
> +time polling is allowed. This value is fixed.
> +
> +Each vcpu has an adjustable guest_halt_poll_us
> +("per-cpu guest_halt_poll_us"), which is adjusted by the algorithm
> +in response to events (explained below).
> +
> +Module Parameters
> +=================
> +
> +The haltpoll governor has 5 tunable module parameters:
> +
> +1) guest_halt_poll_us:
> +Maximum amount of time, in microseconds, that polling is
> +performed before halting.
> +
> +Default: 200
> +
> +2) guest_halt_poll_shrink:
> +Division factor used to shrink per-cpu guest_halt_poll_us when
> +wakeup event occurs after the global guest_halt_poll_us.
> +
> +Default: 2
> +
> +3) guest_halt_poll_grow:
> +Multiplication factor used to grow per-cpu guest_halt_poll_us
> +when event occurs after per-cpu guest_halt_poll_us
> +but before global guest_halt_poll_us.
> +
> +Default: 2
> +
> +4) guest_halt_poll_grow_start:
> +The per-cpu guest_halt_poll_us eventually reaches zero
> +in case of an idle system. This value sets the initial
> +per-cpu guest_halt_poll_us when growing. This can
> +be increased from 10, to avoid misses during the initial
> +growth stage:
> +
> +10, 20, 40, ... (example assumes guest_halt_poll_grow=2).
> +
> +Default: 50
> +
> +5) guest_halt_poll_allow_shrink:
> +
> +Bool parameter which allows shrinking. Set to N
> +to avoid it (per-cpu guest_halt_poll_us will remain
> +high once achieves global guest_halt_poll_us value).
> +
> +Default: Y
> +
> +The module parameters can be set from the debugfs files in:
> +
> +       /sys/module/haltpoll/parameters/
> +
> +Further Notes
> +=============
> +
> +- Care should be taken when setting the guest_halt_poll_us parameter as a
> +large value has the potential to drive the cpu usage to 100% on a machine which
> +would be almost entirely idle otherwise.
> +
>
>

[patch 4/5] kvm: x86: add host poll control msrs

[Marcelo Tosatti]

Add an MSRs which allows the guest to disable 
host polling (specifically the cpuidle-haltpoll, 
when performing polling in the guest, disables
host side polling).

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

---
 Documentation/virtual/kvm/msr.txt    |    9 +++++++++
 arch/x86/include/asm/kvm_host.h      |    2 ++
 arch/x86/include/uapi/asm/kvm_para.h |    2 ++
 arch/x86/kvm/Kconfig                 |    1 +
 arch/x86/kvm/cpuid.c                 |    3 ++-
 arch/x86/kvm/x86.c                   |   23 +++++++++++++++++++++++
 6 files changed, 39 insertions(+), 1 deletion(-)

Index: linux-2.6.git/Documentation/virtual/kvm/msr.txt
===================================================================
--- linux-2.6.git.orig/Documentation/virtual/kvm/msr.txt	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/Documentation/virtual/kvm/msr.txt	2019-06-13 18:42:06.149503132 -0400
@@ -273,3 +273,12 @@
 	guest must both read the least significant bit in the memory area and
 	clear it using a single CPU instruction, such as test and clear, or
 	compare and exchange.
+
+MSR_KVM_POLL_CONTROL: 0x4b564d05
+	Control host side polling.
+
+	data: Bit 0 enables (1) or disables (0) host halt poll
+	logic.
+	KVM guests can disable host halt polling when performing
+	polling themselves.
+
Index: linux-2.6.git/arch/x86/include/asm/kvm_host.h
===================================================================
--- linux-2.6.git.orig/arch/x86/include/asm/kvm_host.h	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/arch/x86/include/asm/kvm_host.h	2019-06-13 18:42:06.149503132 -0400
@@ -755,6 +755,8 @@
 		struct gfn_to_hva_cache data;
 	} pv_eoi;
 
+	u64 msr_kvm_poll_control;
+
 	/*
 	 * Indicate whether the access faults on its page table in guest
 	 * which is set when fix page fault and used to detect unhandeable
Index: linux-2.6.git/arch/x86/include/uapi/asm/kvm_para.h
===================================================================
--- linux-2.6.git.orig/arch/x86/include/uapi/asm/kvm_para.h	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/arch/x86/include/uapi/asm/kvm_para.h	2019-06-13 18:42:06.150503116 -0400
@@ -29,6 +29,7 @@
 #define KVM_FEATURE_PV_TLB_FLUSH	9
 #define KVM_FEATURE_ASYNC_PF_VMEXIT	10
 #define KVM_FEATURE_PV_SEND_IPI	11
+#define KVM_FEATURE_POLL_CONTROL	12
 
 #define KVM_HINTS_REALTIME      0
 
@@ -47,6 +48,7 @@
 #define MSR_KVM_ASYNC_PF_EN 0x4b564d02
 #define MSR_KVM_STEAL_TIME  0x4b564d03
 #define MSR_KVM_PV_EOI_EN      0x4b564d04
+#define MSR_KVM_POLL_CONTROL	0x4b564d05
 
 struct kvm_steal_time {
 	__u64 steal;
Index: linux-2.6.git/arch/x86/kvm/Kconfig
===================================================================
--- linux-2.6.git.orig/arch/x86/kvm/Kconfig	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/arch/x86/kvm/Kconfig	2019-06-13 18:42:06.150503116 -0400
@@ -41,6 +41,7 @@
 	select PERF_EVENTS
 	select HAVE_KVM_MSI
 	select HAVE_KVM_CPU_RELAX_INTERCEPT
+	select HAVE_KVM_NO_POLL
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_VFIO
 	select SRCU
Index: linux-2.6.git/arch/x86/kvm/cpuid.c
===================================================================
--- linux-2.6.git.orig/arch/x86/kvm/cpuid.c	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/arch/x86/kvm/cpuid.c	2019-06-13 18:42:06.150503116 -0400
@@ -643,7 +643,8 @@
 			     (1 << KVM_FEATURE_PV_UNHALT) |
 			     (1 << KVM_FEATURE_PV_TLB_FLUSH) |
 			     (1 << KVM_FEATURE_ASYNC_PF_VMEXIT) |
-			     (1 << KVM_FEATURE_PV_SEND_IPI);
+			     (1 << KVM_FEATURE_PV_SEND_IPI) |
+			     (1 << KVM_FEATURE_POLL_CONTROL);
 
 		if (sched_info_on())
 			entry->eax |= (1 << KVM_FEATURE_STEAL_TIME);
Index: linux-2.6.git/arch/x86/kvm/x86.c
===================================================================
--- linux-2.6.git.orig/arch/x86/kvm/x86.c	2019-06-13 18:41:40.127920237 -0400
+++ linux-2.6.git/arch/x86/kvm/x86.c	2019-06-13 18:42:06.151503100 -0400
@@ -1177,6 +1177,7 @@
 	MSR_IA32_POWER_CTL,
 
 	MSR_K7_HWCR,
+	MSR_KVM_POLL_CONTROL,
 };
 
 static unsigned num_emulated_msrs;
@@ -2628,6 +2629,14 @@
 			return 1;
 		break;
 
+	case MSR_KVM_POLL_CONTROL:
+		/* only enable bit supported */
+		if (data & (-1ULL << 1))
+			return 1;
+
+		vcpu->arch.msr_kvm_poll_control = data;
+		break;
+
 	case MSR_IA32_MCG_CTL:
 	case MSR_IA32_MCG_STATUS:
 	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
@@ -2877,6 +2886,9 @@
 	case MSR_KVM_PV_EOI_EN:
 		msr_info->data = vcpu->arch.pv_eoi.msr_val;
 		break;
+	case MSR_KVM_POLL_CONTROL:
+		msr_info->data = vcpu->arch.msr_kvm_poll_control;
+		break;
 	case MSR_IA32_P5_MC_ADDR:
 	case MSR_IA32_P5_MC_TYPE:
 	case MSR_IA32_MCG_CAP:
@@ -8877,6 +8889,10 @@
 	msr.host_initiated = true;
 	kvm_write_tsc(vcpu, &msr);
 	vcpu_put(vcpu);
+
+	/* poll control enabled by default */
+	vcpu->arch.msr_kvm_poll_control = 1;
+
 	mutex_unlock(&vcpu->mutex);
 
 	if (!kvmclock_periodic_sync)
@@ -9951,6 +9967,13 @@
 }
 EXPORT_SYMBOL_GPL(kvm_vector_hashing_enabled);
 
+bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
+{
+	return (vcpu->arch.msr_kvm_poll_control & 1) == 0;
+}
+EXPORT_SYMBOL_GPL(kvm_arch_no_poll);
+
+
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_exit);
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_fast_mmio);
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_inj_virq);

[patch 5/5] cpuidle-haltpoll: disable host side polling when kvm virtualized

[Marcelo Tosatti]

When performing guest side polling, it is not necessary to 
also perform host side polling. 

So disable host side polling, via the new MSR interface, 
when loading cpuidle-haltpoll driver.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

---
 arch/x86/Kconfig                        |    7 +++++
 arch/x86/include/asm/cpuidle_haltpoll.h |    8 ++++++
 arch/x86/kernel/kvm.c                   |   42 ++++++++++++++++++++++++++++++++
 drivers/cpuidle/cpuidle-haltpoll.c      |   10 ++++++-
 include/linux/cpuidle_haltpoll.h        |   16 ++++++++++++
 5 files changed, 82 insertions(+), 1 deletion(-)

Index: linux-2.6.git/arch/x86/include/asm/cpuidle_haltpoll.h
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6.git/arch/x86/include/asm/cpuidle_haltpoll.h	2019-06-13 18:42:06.891491238 -0400
@@ -0,0 +1,8 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ARCH_HALTPOLL_H
+#define _ARCH_HALTPOLL_H
+
+void arch_haltpoll_enable(void);
+void arch_haltpoll_disable(void);
+
+#endif
Index: linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c
===================================================================
--- linux-2.6.git.orig/drivers/cpuidle/cpuidle-haltpoll.c	2019-06-13 18:41:39.305933413 -0400
+++ linux-2.6.git/drivers/cpuidle/cpuidle-haltpoll.c	2019-06-13 18:42:06.892491222 -0400
@@ -14,6 +14,7 @@
 #include <linux/cpuidle.h>
 #include <linux/module.h>
 #include <linux/sched/idle.h>
+#include <linux/cpuidle_haltpoll.h>
 
 static int default_enter_idle(struct cpuidle_device *dev,
 			      struct cpuidle_driver *drv, int index)
@@ -46,15 +47,22 @@
 
 static int __init haltpoll_init(void)
 {
+	int ret;
 	struct cpuidle_driver *drv = &haltpoll_driver;
 
 	cpuidle_poll_state_init(drv);
 
-	return cpuidle_register(&haltpoll_driver, NULL);
+	ret = cpuidle_register(&haltpoll_driver, NULL);
+
+	if (ret == 0)
+		arch_haltpoll_enable();
+
+	return ret;
 }
 
 static void __exit haltpoll_exit(void)
 {
+	arch_haltpoll_disable();
 	cpuidle_unregister(&haltpoll_driver);
 }
 
Index: linux-2.6.git/include/linux/cpuidle_haltpoll.h
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ linux-2.6.git/include/linux/cpuidle_haltpoll.h	2019-06-13 18:42:06.892491222 -0400
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _CPUIDLE_HALTPOLL_H
+#define _CPUIDLE_HALTPOLL_H
+
+#ifdef CONFIG_ARCH_CPUIDLE_HALTPOLL
+#include <asm/cpuidle_haltpoll.h>
+#else
+static inline void arch_haltpoll_enable(void)
+{
+}
+
+static inline void arch_haltpoll_disable(void)
+{
+}
+#endif
+#endif
Index: linux-2.6.git/arch/x86/Kconfig
===================================================================
--- linux-2.6.git.orig/arch/x86/Kconfig	2019-06-13 18:41:39.305933413 -0400
+++ linux-2.6.git/arch/x86/Kconfig	2019-06-13 18:42:06.893491206 -0400
@@ -787,6 +787,7 @@
 	bool "KVM Guest support (including kvmclock)"
 	depends on PARAVIRT
 	select PARAVIRT_CLOCK
+	select ARCH_CPUIDLE_HALTPOLL
 	default y
 	---help---
 	  This option enables various optimizations for running under the KVM
@@ -795,6 +796,12 @@
 	  underlying device model, the host provides the guest with
 	  timing infrastructure such as time of day, and system time
 
+config ARCH_CPUIDLE_HALTPOLL
+        def_bool n
+        prompt "Disable host haltpoll when loading haltpoll driver"
+        help
+	  If virtualized under KVM, disable host haltpoll.
+
 config PVH
 	bool "Support for running PVH guests"
 	---help---
Index: linux-2.6.git/arch/x86/kernel/kvm.c
===================================================================
--- linux-2.6.git.orig/arch/x86/kernel/kvm.c	2019-06-13 18:41:39.305933413 -0400
+++ linux-2.6.git/arch/x86/kernel/kvm.c	2019-06-13 18:42:06.893491206 -0400
@@ -853,3 +853,45 @@
 }
 
 #endif	/* CONFIG_PARAVIRT_SPINLOCKS */
+
+#ifdef CONFIG_ARCH_CPUIDLE_HALTPOLL
+
+static void kvm_disable_host_haltpoll(void *i)
+{
+	wrmsrl(MSR_KVM_POLL_CONTROL, 0);
+}
+
+static void kvm_enable_host_haltpoll(void *i)
+{
+	wrmsrl(MSR_KVM_POLL_CONTROL, 1);
+}
+
+void arch_haltpoll_enable(void)
+{
+	if (!kvm_para_has_feature(KVM_FEATURE_POLL_CONTROL)) {
+		printk(KERN_ERR "kvm: host does not support poll control\n");
+		printk(KERN_ERR "kvm: host upgrade recommended\n");
+		return;
+	}
+
+	preempt_disable();
+	/* Enable guest halt poll disables host halt poll */
+	kvm_disable_host_haltpoll(NULL);
+	smp_call_function(kvm_disable_host_haltpoll, NULL, 1);
+	preempt_enable();
+}
+EXPORT_SYMBOL_GPL(arch_haltpoll_enable);
+
+void arch_haltpoll_disable(void)
+{
+	if (!kvm_para_has_feature(KVM_FEATURE_POLL_CONTROL))
+		return;
+
+	preempt_disable();
+	/* Enable guest halt poll disables host halt poll */
+	kvm_enable_host_haltpoll(NULL);
+	smp_call_function(kvm_enable_host_haltpoll, NULL, 1);
+	preempt_enable();
+}
+EXPORT_SYMBOL_GPL(arch_haltpoll_disable);
+#endif