[linux.git] / drivers / xen / xen-selfballoon.c

/******************************************************************************
 * Xen selfballoon driver (and optional frontswap self-shrinking driver)
 *
 * Copyright (c) 2009-2011, Dan Magenheimer, Oracle Corp.
 *
 * This code complements the cleancache and frontswap patchsets to optimize
 * support for Xen Transcendent Memory ("tmem").  The policy it implements
 * is rudimentary and will likely improve over time, but it does work well
 * enough today.
 *
 * Two functionalities are implemented here which both use "control theory"
 * (feedback) to optimize memory utilization. In a virtualized environment
 * such as Xen, RAM is often a scarce resource and we would like to ensure
 * that each of a possibly large number of virtual machines is using RAM
 * efficiently, i.e. using as little as possible when under light load
 * and obtaining as much as possible when memory demands are high.
 * Since RAM needs vary highly dynamically and sometimes dramatically,
 * "hysteresis" is used, that is, memory target is determined not just
 * on current data but also on past data stored in the system.
 *
 * "Selfballooning" creates memory pressure by managing the Xen balloon
 * driver to decrease and increase available kernel memory, driven
 * largely by the target value of "Committed_AS" (see /proc/meminfo).
 * Since Committed_AS does not account for clean mapped pages (i.e. pages
 * in RAM that are identical to pages on disk), selfballooning has the
 * affect of pushing less frequently used clean pagecache pages out of
 * kernel RAM and, presumably using cleancache, into Xen tmem where
 * Xen can more efficiently optimize RAM utilization for such pages.
 *
 * When kernel memory demand unexpectedly increases faster than Xen, via
 * the selfballoon driver, is able to (or chooses to) provide usable RAM,
 * the kernel may invoke swapping.  In most cases, frontswap is able
 * to absorb this swapping into Xen tmem.  However, due to the fact
 * that the kernel swap subsystem assumes swapping occurs to a disk,
 * swapped pages may sit on the disk for a very long time; even if
 * the kernel knows the page will never be used again.  This is because
 * the disk space costs very little and can be overwritten when
 * necessary.  When such stale pages are in frontswap, however, they
 * are taking up valuable real estate.  "Frontswap selfshrinking" works
 * to resolve this:  When frontswap activity is otherwise stable
 * and the guest kernel is not under memory pressure, the "frontswap
 * selfshrinking" accounts for this by providing pressure to remove some
 * pages from frontswap and return them to kernel memory.
 *
 * For both "selfballooning" and "frontswap-selfshrinking", a worker
 * thread is used and sysfs tunables are provided to adjust the frequency
 * and rate of adjustments to achieve the goal, as well as to disable one
 * or both functions independently.
 *
 * While some argue that this functionality can and should be implemented
 * in userspace, it has been observed that bad things happen (e.g. OOMs).
 *
 * System configuration note: Selfballooning should not be enabled on
 * systems without a sufficiently large swap device configured; for best
 * results, it is recommended that total swap be increased by the size
 * of the guest memory. Note, that selfballooning should be disabled by default
 * if frontswap is not configured.  Similarly selfballooning should be enabled
 * by default if frontswap is configured and can be disabled with the
 * "tmem.selfballooning=0" kernel boot option.  Finally, when frontswap is
 * configured, frontswap-selfshrinking can be disabled  with the
 * "tmem.selfshrink=0" kernel boot option.
 *
 * Selfballooning is disallowed in domain0 and force-disabled.
 *
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/kernel.h>
#include <linux/bootmem.h>
#include <linux/swap.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/module.h>
#include <linux/workqueue.h>
#include <linux/device.h>
#include <xen/balloon.h>
#include <xen/tmem.h>
#include <xen/xen.h>

/* Enable/disable with sysfs. */
static int xen_selfballooning_enabled __read_mostly;

/*
 * Controls rate at which memory target (this iteration) approaches
 * ultimate goal when memory need is increasing (up-hysteresis) or
 * decreasing (down-hysteresis). Higher values of hysteresis cause
 * slower increases/decreases. The default values for the various
 * parameters were deemed reasonable by experimentation, may be
 * workload-dependent, and can all be adjusted via sysfs.
 */
static unsigned int selfballoon_downhysteresis __read_mostly = 8;
static unsigned int selfballoon_uphysteresis __read_mostly = 1;

/* In HZ, controls frequency of worker invocation. */
static unsigned int selfballoon_interval __read_mostly = 5;

/*
 * Minimum usable RAM in MB for selfballooning target for balloon.
 * If non-zero, it is added to totalreserve_pages and self-ballooning
 * will not balloon below the sum.  If zero, a piecewise linear function
 * is calculated as a minimum and added to totalreserve_pages.  Note that
 * setting this value indiscriminately may cause OOMs and crashes.
 */
static unsigned int selfballoon_min_usable_mb;

/*
 * Amount of RAM in MB to add to the target number of pages.
 * Can be used to reserve some more room for caches and the like.
 */
static unsigned int selfballoon_reserved_mb;

static void selfballoon_process(struct work_struct *work);
static DECLARE_DELAYED_WORK(selfballoon_worker, selfballoon_process);

#ifdef CONFIG_FRONTSWAP
#include <linux/frontswap.h>

/* Enable/disable with sysfs. */
static bool frontswap_selfshrinking __read_mostly;

/*
 * The default values for the following parameters were deemed reasonable
 * by experimentation, may be workload-dependent, and can all be
 * adjusted via sysfs.
 */

/* Control rate for frontswap shrinking. Higher hysteresis is slower. */
static unsigned int frontswap_hysteresis __read_mostly = 20;

/*
 * Number of selfballoon worker invocations to wait before observing that
 * frontswap selfshrinking should commence. Note that selfshrinking does
 * not use a separate worker thread.
 */
static unsigned int frontswap_inertia __read_mostly = 3;

/* Countdown to next invocation of frontswap_shrink() */
static unsigned long frontswap_inertia_counter;

/*
 * Invoked by the selfballoon worker thread, uses current number of pages
 * in frontswap (frontswap_curr_pages()), previous status, and control
 * values (hysteresis and inertia) to determine if frontswap should be
 * shrunk and what the new frontswap size should be.  Note that
 * frontswap_shrink is essentially a partial swapoff that immediately
 * transfers pages from the "swap device" (frontswap) back into kernel
 * RAM; despite the name, frontswap "shrinking" is very different from
 * the "shrinker" interface used by the kernel MM subsystem to reclaim
 * memory.
 */
static void frontswap_selfshrink(void)
{
	static unsigned long cur_frontswap_pages;
	static unsigned long last_frontswap_pages;
	static unsigned long tgt_frontswap_pages;

	last_frontswap_pages = cur_frontswap_pages;
	cur_frontswap_pages = frontswap_curr_pages();
	if (!cur_frontswap_pages ||
			(cur_frontswap_pages > last_frontswap_pages)) {
		frontswap_inertia_counter = frontswap_inertia;
		return;
	}
	if (frontswap_inertia_counter && --frontswap_inertia_counter)
		return;
	if (cur_frontswap_pages <= frontswap_hysteresis)
		tgt_frontswap_pages = 0;
	else
		tgt_frontswap_pages = cur_frontswap_pages -
			(cur_frontswap_pages / frontswap_hysteresis);
	frontswap_shrink(tgt_frontswap_pages);
}

#endif /* CONFIG_FRONTSWAP */

#define MB2PAGES(mb)	((mb) << (20 - PAGE_SHIFT))

/*
 * Use current balloon size, the goal (vm_committed_as), and hysteresis
 * parameters to set a new target balloon size
 */
static void selfballoon_process(struct work_struct *work)
{
	unsigned long cur_pages, goal_pages, tgt_pages, floor_pages;
	unsigned long useful_pages;
	bool reset_timer = false;

	if (xen_selfballooning_enabled) {
		cur_pages = totalram_pages;
		tgt_pages = cur_pages; /* default is no change */
		goal_pages = vm_memory_committed() +
				totalreserve_pages +
				MB2PAGES(selfballoon_reserved_mb);
#ifdef CONFIG_FRONTSWAP
		/* allow space for frontswap pages to be repatriated */
		if (frontswap_selfshrinking && frontswap_enabled)
			goal_pages += frontswap_curr_pages();
#endif
		if (cur_pages > goal_pages)
			tgt_pages = cur_pages -
				((cur_pages - goal_pages) /
				  selfballoon_downhysteresis);
		else if (cur_pages < goal_pages)
			tgt_pages = cur_pages +
				((goal_pages - cur_pages) /
				  selfballoon_uphysteresis);
		/* else if cur_pages == goal_pages, no change */
		useful_pages = max_pfn - totalreserve_pages;
		if (selfballoon_min_usable_mb != 0)
			floor_pages = totalreserve_pages +
					MB2PAGES(selfballoon_min_usable_mb);
		/* piecewise linear function ending in ~3% slope */
		else if (useful_pages < MB2PAGES(16))
			floor_pages = max_pfn; /* not worth ballooning */
		else if (useful_pages < MB2PAGES(64))
			floor_pages = totalreserve_pages + MB2PAGES(16) +
					((useful_pages - MB2PAGES(16)) >> 1);
		else if (useful_pages < MB2PAGES(512))
			floor_pages = totalreserve_pages + MB2PAGES(40) +
					((useful_pages - MB2PAGES(40)) >> 3);
		else /* useful_pages >= MB2PAGES(512) */
			floor_pages = totalreserve_pages + MB2PAGES(99) +
					((useful_pages - MB2PAGES(99)) >> 5);
		if (tgt_pages < floor_pages)
			tgt_pages = floor_pages;
		balloon_set_new_target(tgt_pages +
			balloon_stats.current_pages - totalram_pages);
		reset_timer = true;
	}
#ifdef CONFIG_FRONTSWAP
	if (frontswap_selfshrinking && frontswap_enabled) {
		frontswap_selfshrink();
		reset_timer = true;
	}
#endif
	if (reset_timer)
		schedule_delayed_work(&selfballoon_worker,
			selfballoon_interval * HZ);
}

#ifdef CONFIG_SYSFS

#include <linux/capability.h>

#define SELFBALLOON_SHOW(name, format, args...)				\
	static ssize_t show_##name(struct device *dev,	\
					  struct device_attribute *attr, \
					  char *buf) \
	{ \
		return sprintf(buf, format, ##args); \
	}

SELFBALLOON_SHOW(selfballooning, "%d\n", xen_selfballooning_enabled);

static ssize_t store_selfballooning(struct device *dev,
			    struct device_attribute *attr,
			    const char *buf,
			    size_t count)
{
	bool was_enabled = xen_selfballooning_enabled;
	unsigned long tmp;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;

	err = kstrtoul(buf, 10, &tmp);
	if (err)
		return err;
	if ((tmp != 0) && (tmp != 1))
		return -EINVAL;

	xen_selfballooning_enabled = !!tmp;
	if (!was_enabled && xen_selfballooning_enabled)
		schedule_delayed_work(&selfballoon_worker,
			selfballoon_interval * HZ);

	return count;
}

static DEVICE_ATTR(selfballooning, S_IRUGO | S_IWUSR,
		   show_selfballooning, store_selfballooning);

SELFBALLOON_SHOW(selfballoon_interval, "%d\n", selfballoon_interval);

static ssize_t store_selfballoon_interval(struct device *dev,
					  struct device_attribute *attr,
					  const char *buf,
					  size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	selfballoon_interval = val;
	return count;
}

static DEVICE_ATTR(selfballoon_interval, S_IRUGO | S_IWUSR,
		   show_selfballoon_interval, store_selfballoon_interval);

SELFBALLOON_SHOW(selfballoon_downhys, "%d\n", selfballoon_downhysteresis);

static ssize_t store_selfballoon_downhys(struct device *dev,
					 struct device_attribute *attr,
					 const char *buf,
					 size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	selfballoon_downhysteresis = val;
	return count;
}

static DEVICE_ATTR(selfballoon_downhysteresis, S_IRUGO | S_IWUSR,
		   show_selfballoon_downhys, store_selfballoon_downhys);


SELFBALLOON_SHOW(selfballoon_uphys, "%d\n", selfballoon_uphysteresis);

static ssize_t store_selfballoon_uphys(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf,
				       size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	selfballoon_uphysteresis = val;
	return count;
}

static DEVICE_ATTR(selfballoon_uphysteresis, S_IRUGO | S_IWUSR,
		   show_selfballoon_uphys, store_selfballoon_uphys);

SELFBALLOON_SHOW(selfballoon_min_usable_mb, "%d\n",
				selfballoon_min_usable_mb);

static ssize_t store_selfballoon_min_usable_mb(struct device *dev,
					       struct device_attribute *attr,
					       const char *buf,
					       size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	selfballoon_min_usable_mb = val;
	return count;
}

static DEVICE_ATTR(selfballoon_min_usable_mb, S_IRUGO | S_IWUSR,
		   show_selfballoon_min_usable_mb,
		   store_selfballoon_min_usable_mb);

SELFBALLOON_SHOW(selfballoon_reserved_mb, "%d\n",
				selfballoon_reserved_mb);

static ssize_t store_selfballoon_reserved_mb(struct device *dev,
					     struct device_attribute *attr,
					     const char *buf,
					     size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	selfballoon_reserved_mb = val;
	return count;
}

static DEVICE_ATTR(selfballoon_reserved_mb, S_IRUGO | S_IWUSR,
		   show_selfballoon_reserved_mb,
		   store_selfballoon_reserved_mb);


#ifdef CONFIG_FRONTSWAP
SELFBALLOON_SHOW(frontswap_selfshrinking, "%d\n", frontswap_selfshrinking);

static ssize_t store_frontswap_selfshrinking(struct device *dev,
					     struct device_attribute *attr,
					     const char *buf,
					     size_t count)
{
	bool was_enabled = frontswap_selfshrinking;
	unsigned long tmp;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &tmp);
	if (err)
		return err;
	if ((tmp != 0) && (tmp != 1))
		return -EINVAL;
	frontswap_selfshrinking = !!tmp;
	if (!was_enabled && !xen_selfballooning_enabled &&
	     frontswap_selfshrinking)
		schedule_delayed_work(&selfballoon_worker,
			selfballoon_interval * HZ);

	return count;
}

static DEVICE_ATTR(frontswap_selfshrinking, S_IRUGO | S_IWUSR,
		   show_frontswap_selfshrinking, store_frontswap_selfshrinking);

SELFBALLOON_SHOW(frontswap_inertia, "%d\n", frontswap_inertia);

static ssize_t store_frontswap_inertia(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf,
				       size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	frontswap_inertia = val;
	frontswap_inertia_counter = val;
	return count;
}

static DEVICE_ATTR(frontswap_inertia, S_IRUGO | S_IWUSR,
		   show_frontswap_inertia, store_frontswap_inertia);

SELFBALLOON_SHOW(frontswap_hysteresis, "%d\n", frontswap_hysteresis);

static ssize_t store_frontswap_hysteresis(struct device *dev,
					  struct device_attribute *attr,
					  const char *buf,
					  size_t count)
{
	unsigned long val;
	int err;

	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	err = kstrtoul(buf, 10, &val);
	if (err)
		return err;
	if (val == 0)
		return -EINVAL;
	frontswap_hysteresis = val;
	return count;
}

static DEVICE_ATTR(frontswap_hysteresis, S_IRUGO | S_IWUSR,
		   show_frontswap_hysteresis, store_frontswap_hysteresis);

#endif /* CONFIG_FRONTSWAP */

static struct attribute *selfballoon_attrs[] = {
	&dev_attr_selfballooning.attr,
	&dev_attr_selfballoon_interval.attr,
	&dev_attr_selfballoon_downhysteresis.attr,
	&dev_attr_selfballoon_uphysteresis.attr,
	&dev_attr_selfballoon_min_usable_mb.attr,
	&dev_attr_selfballoon_reserved_mb.attr,
#ifdef CONFIG_FRONTSWAP
	&dev_attr_frontswap_selfshrinking.attr,
	&dev_attr_frontswap_hysteresis.attr,
	&dev_attr_frontswap_inertia.attr,
#endif
	NULL
};

static const struct attribute_group selfballoon_group = {
	.name = "selfballoon",
	.attrs = selfballoon_attrs
};
#endif

int register_xen_selfballooning(struct device *dev)
{
	int error = -1;

#ifdef CONFIG_SYSFS
	error = sysfs_create_group(&dev->kobj, &selfballoon_group);
#endif
	return error;
}
EXPORT_SYMBOL(register_xen_selfballooning);

int xen_selfballoon_init(bool use_selfballooning, bool use_frontswap_selfshrink)
{
	bool enable = false;

	if (!xen_domain())
		return -ENODEV;

	if (xen_initial_domain()) {
		pr_info("Xen selfballooning driver disabled for domain0\n");
		return -ENODEV;
	}

	xen_selfballooning_enabled = tmem_enabled && use_selfballooning;
	if (xen_selfballooning_enabled) {
		pr_info("Initializing Xen selfballooning driver\n");
		enable = true;
	}
#ifdef CONFIG_FRONTSWAP
	frontswap_selfshrinking = tmem_enabled && use_frontswap_selfshrink;
	if (frontswap_selfshrinking) {
		pr_info("Initializing frontswap selfshrinking driver\n");
		enable = true;
	}
#endif
	if (!enable)
		return -ENODEV;

	schedule_delayed_work(&selfballoon_worker, selfballoon_interval * HZ);

	return 0;
}
EXPORT_SYMBOL(xen_selfballoon_init);
Commit	Line	Data
a50777c7 DM	1	/******************************************************************************
	2	* Xen selfballoon driver (and optional frontswap self-shrinking driver)
	3	*
	4	* Copyright (c) 2009-2011, Dan Magenheimer, Oracle Corp.
	5	*
	6	* This code complements the cleancache and frontswap patchsets to optimize
	7	* support for Xen Transcendent Memory ("tmem"). The policy it implements
	8	* is rudimentary and will likely improve over time, but it does work well
	9	* enough today.
	10	*
	11	* Two functionalities are implemented here which both use "control theory"
	12	* (feedback) to optimize memory utilization. In a virtualized environment
	13	* such as Xen, RAM is often a scarce resource and we would like to ensure
	14	* that each of a possibly large number of virtual machines is using RAM
	15	* efficiently, i.e. using as little as possible when under light load
	16	* and obtaining as much as possible when memory demands are high.
	17	* Since RAM needs vary highly dynamically and sometimes dramatically,
	18	* "hysteresis" is used, that is, memory target is determined not just
	19	* on current data but also on past data stored in the system.
	20	*
	21	* "Selfballooning" creates memory pressure by managing the Xen balloon
	22	* driver to decrease and increase available kernel memory, driven
	23	* largely by the target value of "Committed_AS" (see /proc/meminfo).
	24	* Since Committed_AS does not account for clean mapped pages (i.e. pages
	25	* in RAM that are identical to pages on disk), selfballooning has the
	26	* affect of pushing less frequently used clean pagecache pages out of
	27	* kernel RAM and, presumably using cleancache, into Xen tmem where
	28	* Xen can more efficiently optimize RAM utilization for such pages.
	29	*
	30	* When kernel memory demand unexpectedly increases faster than Xen, via
	31	* the selfballoon driver, is able to (or chooses to) provide usable RAM,
	32	* the kernel may invoke swapping. In most cases, frontswap is able
	33	* to absorb this swapping into Xen tmem. However, due to the fact
	34	* that the kernel swap subsystem assumes swapping occurs to a disk,
	35	* swapped pages may sit on the disk for a very long time; even if
	36	* the kernel knows the page will never be used again. This is because
	37	* the disk space costs very little and can be overwritten when
	38	* necessary. When such stale pages are in frontswap, however, they
	39	* are taking up valuable real estate. "Frontswap selfshrinking" works
	40	* to resolve this: When frontswap activity is otherwise stable
	41	* and the guest kernel is not under memory pressure, the "frontswap
	42	* selfshrinking" accounts for this by providing pressure to remove some
	43	* pages from frontswap and return them to kernel memory.
	44	*
	45	* For both "selfballooning" and "frontswap-selfshrinking", a worker
	46	* thread is used and sysfs tunables are provided to adjust the frequency
	47	* and rate of adjustments to achieve the goal, as well as to disable one
	48	* or both functions independently.
	49	*
	50	* While some argue that this functionality can and should be implemented
	51	* in userspace, it has been observed that bad things happen (e.g. OOMs).
	52	*
	53	* System configuration note: Selfballooning should not be enabled on
	54	* systems without a sufficiently large swap device configured; for best
	55	* results, it is recommended that total swap be increased by the size
37d46e15 KRW	56	* of the guest memory. Note, that selfballooning should be disabled by default
	57	* if frontswap is not configured. Similarly selfballooning should be enabled
	58	* by default if frontswap is configured and can be disabled with the
	59	* "tmem.selfballooning=0" kernel boot option. Finally, when frontswap is
	60	* configured, frontswap-selfshrinking can be disabled with the
	61	* "tmem.selfshrink=0" kernel boot option.
a50777c7 DM	62	*
	63	* Selfballooning is disallowed in domain0 and force-disabled.
	64	*
	65	*/
	66
283c0972 JP	67	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
283c0972 JP	68
a50777c7	69	#include <linux/kernel.h>
38a1ed4f DM	70	#include <linux/bootmem.h>
38a1ed4f DM	71	#include <linux/swap.h>
a50777c7 DM	72	#include <linux/mm.h>
a50777c7 DM	73	#include <linux/mman.h>
4fec0e0b	74	#include <linux/module.h>
0642d2ed	75	#include <linux/workqueue.h>
cb0c05c5	76	#include <linux/device.h>
a50777c7	77	#include <xen/balloon.h>
a50777c7	78	#include <xen/tmem.h>
0642d2ed	79	#include <xen/xen.h>
a50777c7 DM	80
	81	/* Enable/disable with sysfs. */
	82	static int xen_selfballooning_enabled __read_mostly;
	83
	84	/*
	85	* Controls rate at which memory target (this iteration) approaches
	86	* ultimate goal when memory need is increasing (up-hysteresis) or
	87	* decreasing (down-hysteresis). Higher values of hysteresis cause
	88	* slower increases/decreases. The default values for the various
	89	* parameters were deemed reasonable by experimentation, may be
	90	* workload-dependent, and can all be adjusted via sysfs.
	91	*/
	92	static unsigned int selfballoon_downhysteresis __read_mostly = 8;
	93	static unsigned int selfballoon_uphysteresis __read_mostly = 1;
	94
	95	/* In HZ, controls frequency of worker invocation. */
	96	static unsigned int selfballoon_interval __read_mostly = 5;
	97
38a1ed4f DM	98	/*
	99	* Minimum usable RAM in MB for selfballooning target for balloon.
	100	* If non-zero, it is added to totalreserve_pages and self-ballooning
	101	* will not balloon below the sum. If zero, a piecewise linear function
	102	* is calculated as a minimum and added to totalreserve_pages. Note that
	103	* setting this value indiscriminately may cause OOMs and crashes.
	104	*/
	105	static unsigned int selfballoon_min_usable_mb;
	106
d79d5959 JS	107	/*
	108	* Amount of RAM in MB to add to the target number of pages.
	109	* Can be used to reserve some more room for caches and the like.
	110	*/
	111	static unsigned int selfballoon_reserved_mb;
	112
a50777c7 DM	113	static void selfballoon_process(struct work_struct *work);
	114	static DECLARE_DELAYED_WORK(selfballoon_worker, selfballoon_process);
	115
	116	#ifdef CONFIG_FRONTSWAP
	117	#include <linux/frontswap.h>
	118
	119	/* Enable/disable with sysfs. */
	120	static bool frontswap_selfshrinking __read_mostly;
	121
a50777c7 DM	122	/*
	123	* The default values for the following parameters were deemed reasonable
	124	* by experimentation, may be workload-dependent, and can all be
	125	* adjusted via sysfs.
	126	*/
	127
	128	/* Control rate for frontswap shrinking. Higher hysteresis is slower. */
	129	static unsigned int frontswap_hysteresis __read_mostly = 20;
	130
	131	/*
	132	* Number of selfballoon worker invocations to wait before observing that
	133	* frontswap selfshrinking should commence. Note that selfshrinking does
	134	* not use a separate worker thread.
	135	*/
	136	static unsigned int frontswap_inertia __read_mostly = 3;
	137
	138	/* Countdown to next invocation of frontswap_shrink() */
	139	static unsigned long frontswap_inertia_counter;
	140
	141	/*
	142	* Invoked by the selfballoon worker thread, uses current number of pages
	143	* in frontswap (frontswap_curr_pages()), previous status, and control
	144	* values (hysteresis and inertia) to determine if frontswap should be
	145	* shrunk and what the new frontswap size should be. Note that
	146	* frontswap_shrink is essentially a partial swapoff that immediately
	147	* transfers pages from the "swap device" (frontswap) back into kernel
	148	* RAM; despite the name, frontswap "shrinking" is very different from
	149	* the "shrinker" interface used by the kernel MM subsystem to reclaim
	150	* memory.
	151	*/
	152	static void frontswap_selfshrink(void)
	153	{
	154	static unsigned long cur_frontswap_pages;
	155	static unsigned long last_frontswap_pages;
	156	static unsigned long tgt_frontswap_pages;
	157
	158	last_frontswap_pages = cur_frontswap_pages;
	159	cur_frontswap_pages = frontswap_curr_pages();
	160	if (!cur_frontswap_pages \|\|
	161	(cur_frontswap_pages > last_frontswap_pages)) {
	162	frontswap_inertia_counter = frontswap_inertia;
	163	return;
	164	}
	165	if (frontswap_inertia_counter && --frontswap_inertia_counter)
	166	return;
	167	if (cur_frontswap_pages <= frontswap_hysteresis)
	168	tgt_frontswap_pages = 0;
	169	else
	170	tgt_frontswap_pages = cur_frontswap_pages -
	171	(cur_frontswap_pages / frontswap_hysteresis);
	172	frontswap_shrink(tgt_frontswap_pages);
	173	}
	174
a50777c7 DM	175	#endif /* CONFIG_FRONTSWAP */
a50777c7 DM	176
38a1ed4f DM	177	#define MB2PAGES(mb) ((mb) << (20 - PAGE_SHIFT))
38a1ed4f DM	178
a50777c7 DM	179	/*
	180	* Use current balloon size, the goal (vm_committed_as), and hysteresis
	181	* parameters to set a new target balloon size
	182	*/
	183	static void selfballoon_process(struct work_struct *work)
	184	{
38a1ed4f DM	185	unsigned long cur_pages, goal_pages, tgt_pages, floor_pages;
38a1ed4f DM	186	unsigned long useful_pages;
a50777c7 DM	187	bool reset_timer = false;
	188
	189	if (xen_selfballooning_enabled) {
38a1ed4f	190	cur_pages = totalram_pages;
a50777c7	191	tgt_pages = cur_pages; /* default is no change */
997071bc	192	goal_pages = vm_memory_committed() +
d79d5959 JS	193	totalreserve_pages +
d79d5959 JS	194	MB2PAGES(selfballoon_reserved_mb);
a50777c7 DM	195	#ifdef CONFIG_FRONTSWAP
	196	/* allow space for frontswap pages to be repatriated */
	197	if (frontswap_selfshrinking && frontswap_enabled)
	198	goal_pages += frontswap_curr_pages();
	199	#endif
	200	if (cur_pages > goal_pages)
	201	tgt_pages = cur_pages -
	202	((cur_pages - goal_pages) /
	203	selfballoon_downhysteresis);
	204	else if (cur_pages < goal_pages)
	205	tgt_pages = cur_pages +
	206	((goal_pages - cur_pages) /
	207	selfballoon_uphysteresis);
	208	/* else if cur_pages == goal_pages, no change */
38a1ed4f DM	209	useful_pages = max_pfn - totalreserve_pages;
	210	if (selfballoon_min_usable_mb != 0)
	211	floor_pages = totalreserve_pages +
	212	MB2PAGES(selfballoon_min_usable_mb);
	213	/* piecewise linear function ending in ~3% slope */
	214	else if (useful_pages < MB2PAGES(16))
	215	floor_pages = max_pfn; /* not worth ballooning */
	216	else if (useful_pages < MB2PAGES(64))
	217	floor_pages = totalreserve_pages + MB2PAGES(16) +
	218	((useful_pages - MB2PAGES(16)) >> 1);
	219	else if (useful_pages < MB2PAGES(512))
	220	floor_pages = totalreserve_pages + MB2PAGES(40) +
	221	((useful_pages - MB2PAGES(40)) >> 3);
	222	else /* useful_pages >= MB2PAGES(512) */
	223	floor_pages = totalreserve_pages + MB2PAGES(99) +
	224	((useful_pages - MB2PAGES(99)) >> 5);
	225	if (tgt_pages < floor_pages)
	226	tgt_pages = floor_pages;
	227	balloon_set_new_target(tgt_pages +
	228	balloon_stats.current_pages - totalram_pages);
a50777c7 DM	229	reset_timer = true;
	230	}
	231	#ifdef CONFIG_FRONTSWAP
	232	if (frontswap_selfshrinking && frontswap_enabled) {
	233	frontswap_selfshrink();
	234	reset_timer = true;
	235	}
	236	#endif
	237	if (reset_timer)
	238	schedule_delayed_work(&selfballoon_worker,
	239	selfballoon_interval * HZ);
	240	}
	241
	242	#ifdef CONFIG_SYSFS
	243
a50777c7 DM	244	#include <linux/capability.h>
	245
	246	#define SELFBALLOON_SHOW(name, format, args...) \
07068021 KS	247	static ssize_t show_##name(struct device *dev, \
	248	struct device_attribute *attr, \
	249	char *buf) \
a50777c7 DM	250	{ \
	251	return sprintf(buf, format, ##args); \
	252	}
	253
	254	SELFBALLOON_SHOW(selfballooning, "%d\n", xen_selfballooning_enabled);
	255
07068021 KS	256	static ssize_t store_selfballooning(struct device *dev,
07068021 KS	257	struct device_attribute *attr,
a50777c7 DM	258	const char *buf,
	259	size_t count)
	260	{
	261	bool was_enabled = xen_selfballooning_enabled;
	262	unsigned long tmp;
	263	int err;
	264
	265	if (!capable(CAP_SYS_ADMIN))
	266	return -EPERM;
	267
d3dbd93d JH	268	err = kstrtoul(buf, 10, &tmp);
	269	if (err)
	270	return err;
	271	if ((tmp != 0) && (tmp != 1))
a50777c7 DM	272	return -EINVAL;
	273
	274	xen_selfballooning_enabled = !!tmp;
	275	if (!was_enabled && xen_selfballooning_enabled)
	276	schedule_delayed_work(&selfballoon_worker,
	277	selfballoon_interval * HZ);
	278
	279	return count;
	280	}
	281
07068021	282	static DEVICE_ATTR(selfballooning, S_IRUGO \| S_IWUSR,
a50777c7 DM	283	show_selfballooning, store_selfballooning);
	284
	285	SELFBALLOON_SHOW(selfballoon_interval, "%d\n", selfballoon_interval);
	286
07068021 KS	287	static ssize_t store_selfballoon_interval(struct device *dev,
07068021 KS	288	struct device_attribute *attr,
a50777c7 DM	289	const char *buf,
	290	size_t count)
	291	{
	292	unsigned long val;
	293	int err;
	294
	295	if (!capable(CAP_SYS_ADMIN))
	296	return -EPERM;
d3dbd93d JH	297	err = kstrtoul(buf, 10, &val);
	298	if (err)
	299	return err;
	300	if (val == 0)
a50777c7 DM	301	return -EINVAL;
	302	selfballoon_interval = val;
	303	return count;
	304	}
	305
07068021	306	static DEVICE_ATTR(selfballoon_interval, S_IRUGO \| S_IWUSR,
a50777c7 DM	307	show_selfballoon_interval, store_selfballoon_interval);
	308
	309	SELFBALLOON_SHOW(selfballoon_downhys, "%d\n", selfballoon_downhysteresis);
	310
07068021 KS	311	static ssize_t store_selfballoon_downhys(struct device *dev,
07068021 KS	312	struct device_attribute *attr,
a50777c7 DM	313	const char *buf,
	314	size_t count)
	315	{
	316	unsigned long val;
	317	int err;
	318
	319	if (!capable(CAP_SYS_ADMIN))
	320	return -EPERM;
d3dbd93d JH	321	err = kstrtoul(buf, 10, &val);
	322	if (err)
	323	return err;
	324	if (val == 0)
a50777c7 DM	325	return -EINVAL;
	326	selfballoon_downhysteresis = val;
	327	return count;
	328	}
	329
07068021	330	static DEVICE_ATTR(selfballoon_downhysteresis, S_IRUGO \| S_IWUSR,
a50777c7 DM	331	show_selfballoon_downhys, store_selfballoon_downhys);
	332
	333
	334	SELFBALLOON_SHOW(selfballoon_uphys, "%d\n", selfballoon_uphysteresis);
	335
07068021 KS	336	static ssize_t store_selfballoon_uphys(struct device *dev,
07068021 KS	337	struct device_attribute *attr,
a50777c7 DM	338	const char *buf,
	339	size_t count)
	340	{
	341	unsigned long val;
	342	int err;
	343
	344	if (!capable(CAP_SYS_ADMIN))
	345	return -EPERM;
d3dbd93d JH	346	err = kstrtoul(buf, 10, &val);
	347	if (err)
	348	return err;
	349	if (val == 0)
a50777c7 DM	350	return -EINVAL;
	351	selfballoon_uphysteresis = val;
	352	return count;
	353	}
	354
07068021	355	static DEVICE_ATTR(selfballoon_uphysteresis, S_IRUGO \| S_IWUSR,
a50777c7 DM	356	show_selfballoon_uphys, store_selfballoon_uphys);
a50777c7 DM	357
38a1ed4f DM	358	SELFBALLOON_SHOW(selfballoon_min_usable_mb, "%d\n",
	359	selfballoon_min_usable_mb);
	360
07068021 KS	361	static ssize_t store_selfballoon_min_usable_mb(struct device *dev,
07068021 KS	362	struct device_attribute *attr,
38a1ed4f DM	363	const char *buf,
	364	size_t count)
	365	{
	366	unsigned long val;
	367	int err;
	368
	369	if (!capable(CAP_SYS_ADMIN))
	370	return -EPERM;
d3dbd93d JH	371	err = kstrtoul(buf, 10, &val);
	372	if (err)
	373	return err;
	374	if (val == 0)
38a1ed4f DM	375	return -EINVAL;
	376	selfballoon_min_usable_mb = val;
	377	return count;
	378	}
	379
07068021	380	static DEVICE_ATTR(selfballoon_min_usable_mb, S_IRUGO \| S_IWUSR,
38a1ed4f DM	381	show_selfballoon_min_usable_mb,
	382	store_selfballoon_min_usable_mb);
	383
d79d5959 JS	384	SELFBALLOON_SHOW(selfballoon_reserved_mb, "%d\n",
	385	selfballoon_reserved_mb);
	386
	387	static ssize_t store_selfballoon_reserved_mb(struct device *dev,
	388	struct device_attribute *attr,
	389	const char *buf,
	390	size_t count)
	391	{
	392	unsigned long val;
	393	int err;
	394
	395	if (!capable(CAP_SYS_ADMIN))
	396	return -EPERM;
d3dbd93d JH	397	err = kstrtoul(buf, 10, &val);
	398	if (err)
	399	return err;
	400	if (val == 0)
d79d5959 JS	401	return -EINVAL;
	402	selfballoon_reserved_mb = val;
	403	return count;
	404	}
	405
	406	static DEVICE_ATTR(selfballoon_reserved_mb, S_IRUGO \| S_IWUSR,
	407	show_selfballoon_reserved_mb,
	408	store_selfballoon_reserved_mb);
	409
38a1ed4f	410
a50777c7 DM	411	#ifdef CONFIG_FRONTSWAP
	412	SELFBALLOON_SHOW(frontswap_selfshrinking, "%d\n", frontswap_selfshrinking);
	413
07068021 KS	414	static ssize_t store_frontswap_selfshrinking(struct device *dev,
07068021 KS	415	struct device_attribute *attr,
a50777c7 DM	416	const char *buf,
	417	size_t count)
	418	{
	419	bool was_enabled = frontswap_selfshrinking;
	420	unsigned long tmp;
	421	int err;
	422
	423	if (!capable(CAP_SYS_ADMIN))
	424	return -EPERM;
d3dbd93d JH	425	err = kstrtoul(buf, 10, &tmp);
	426	if (err)
	427	return err;
	428	if ((tmp != 0) && (tmp != 1))
a50777c7 DM	429	return -EINVAL;
	430	frontswap_selfshrinking = !!tmp;
	431	if (!was_enabled && !xen_selfballooning_enabled &&
	432	frontswap_selfshrinking)
	433	schedule_delayed_work(&selfballoon_worker,
	434	selfballoon_interval * HZ);
	435
	436	return count;
	437	}
	438
07068021	439	static DEVICE_ATTR(frontswap_selfshrinking, S_IRUGO \| S_IWUSR,
a50777c7 DM	440	show_frontswap_selfshrinking, store_frontswap_selfshrinking);
	441
	442	SELFBALLOON_SHOW(frontswap_inertia, "%d\n", frontswap_inertia);
	443
07068021 KS	444	static ssize_t store_frontswap_inertia(struct device *dev,
07068021 KS	445	struct device_attribute *attr,
a50777c7 DM	446	const char *buf,
	447	size_t count)
	448	{
	449	unsigned long val;
	450	int err;
	451
	452	if (!capable(CAP_SYS_ADMIN))
	453	return -EPERM;
d3dbd93d JH	454	err = kstrtoul(buf, 10, &val);
	455	if (err)
	456	return err;
	457	if (val == 0)
a50777c7 DM	458	return -EINVAL;
	459	frontswap_inertia = val;
	460	frontswap_inertia_counter = val;
	461	return count;
	462	}
	463
07068021	464	static DEVICE_ATTR(frontswap_inertia, S_IRUGO \| S_IWUSR,
a50777c7 DM	465	show_frontswap_inertia, store_frontswap_inertia);
	466
	467	SELFBALLOON_SHOW(frontswap_hysteresis, "%d\n", frontswap_hysteresis);
	468
07068021 KS	469	static ssize_t store_frontswap_hysteresis(struct device *dev,
07068021 KS	470	struct device_attribute *attr,
a50777c7 DM	471	const char *buf,
	472	size_t count)
	473	{
	474	unsigned long val;
	475	int err;
	476
	477	if (!capable(CAP_SYS_ADMIN))
	478	return -EPERM;
d3dbd93d JH	479	err = kstrtoul(buf, 10, &val);
	480	if (err)
	481	return err;
	482	if (val == 0)
a50777c7 DM	483	return -EINVAL;
	484	frontswap_hysteresis = val;
	485	return count;
	486	}
	487
07068021	488	static DEVICE_ATTR(frontswap_hysteresis, S_IRUGO \| S_IWUSR,
a50777c7 DM	489	show_frontswap_hysteresis, store_frontswap_hysteresis);
	490
	491	#endif /* CONFIG_FRONTSWAP */
	492
	493	static struct attribute *selfballoon_attrs[] = {
07068021 KS	494	&dev_attr_selfballooning.attr,
	495	&dev_attr_selfballoon_interval.attr,
	496	&dev_attr_selfballoon_downhysteresis.attr,
	497	&dev_attr_selfballoon_uphysteresis.attr,
	498	&dev_attr_selfballoon_min_usable_mb.attr,
d79d5959	499	&dev_attr_selfballoon_reserved_mb.attr,
a50777c7	500	#ifdef CONFIG_FRONTSWAP
07068021 KS	501	&dev_attr_frontswap_selfshrinking.attr,
	502	&dev_attr_frontswap_hysteresis.attr,
	503	&dev_attr_frontswap_inertia.attr,
a50777c7 DM	504	#endif
	505	NULL
	506	};
	507
ead1d014	508	static const struct attribute_group selfballoon_group = {
a50777c7 DM	509	.name = "selfballoon",
	510	.attrs = selfballoon_attrs
	511	};
	512	#endif
	513
07068021	514	int register_xen_selfballooning(struct device *dev)
a50777c7 DM	515	{
	516	int error = -1;
	517
	518	#ifdef CONFIG_SYSFS
07068021	519	error = sysfs_create_group(&dev->kobj, &selfballoon_group);
a50777c7 DM	520	#endif
	521	return error;
	522	}
	523	EXPORT_SYMBOL(register_xen_selfballooning);
	524
10a7a077	525	int xen_selfballoon_init(bool use_selfballooning, bool use_frontswap_selfshrink)
a50777c7 DM	526	{
	527	bool enable = false;
	528
	529	if (!xen_domain())
	530	return -ENODEV;
	531
	532	if (xen_initial_domain()) {
283c0972	533	pr_info("Xen selfballooning driver disabled for domain0\n");
a50777c7 DM	534	return -ENODEV;
	535	}
	536
	537	xen_selfballooning_enabled = tmem_enabled && use_selfballooning;
	538	if (xen_selfballooning_enabled) {
283c0972	539	pr_info("Initializing Xen selfballooning driver\n");
a50777c7 DM	540	enable = true;
	541	}
	542	#ifdef CONFIG_FRONTSWAP
	543	frontswap_selfshrinking = tmem_enabled && use_frontswap_selfshrink;
	544	if (frontswap_selfshrinking) {
283c0972	545	pr_info("Initializing frontswap selfshrinking driver\n");
a50777c7 DM	546	enable = true;
	547	}
	548	#endif
	549	if (!enable)
	550	return -ENODEV;
	551
	552	schedule_delayed_work(&selfballoon_worker, selfballoon_interval * HZ);
	553
	554	return 0;
	555	}
10a7a077	556	EXPORT_SYMBOL(xen_selfballoon_init);