[linux-2.6-block.git] / drivers / xen / time.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Xen stolen ticks accounting.
 */
#include <linux/kernel.h>
#include <linux/kernel_stat.h>
#include <linux/math64.h>
#include <linux/gfp.h>
#include <linux/slab.h>

#include <asm/paravirt.h>
#include <asm/xen/hypervisor.h>
#include <asm/xen/hypercall.h>

#include <xen/events.h>
#include <xen/features.h>
#include <xen/interface/xen.h>
#include <xen/interface/vcpu.h>
#include <xen/xen-ops.h>

/* runstate info updated by Xen */
static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate);

static DEFINE_PER_CPU(u64[4], old_runstate_time);

/* return an consistent snapshot of 64-bit time/counter value */
static u64 get64(const u64 *p)
{
	u64 ret;

	if (BITS_PER_LONG < 64) {
		u32 *p32 = (u32 *)p;
		u32 h, l, h2;

		/*
		 * Read high then low, and then make sure high is
		 * still the same; this will only loop if low wraps
		 * and carries into high.
		 * XXX some clean way to make this endian-proof?
		 */
		do {
			h = READ_ONCE(p32[1]);
			l = READ_ONCE(p32[0]);
			h2 = READ_ONCE(p32[1]);
		} while(h2 != h);

		ret = (((u64)h) << 32) | l;
	} else
		ret = READ_ONCE(*p);

	return ret;
}

static void xen_get_runstate_snapshot_cpu_delta(
			      struct vcpu_runstate_info *res, unsigned int cpu)
{
	u64 state_time;
	struct vcpu_runstate_info *state;

	BUG_ON(preemptible());

	state = per_cpu_ptr(&xen_runstate, cpu);

	do {
		state_time = get64(&state->state_entry_time);
		rmb();	/* Hypervisor might update data. */
		*res = READ_ONCE(*state);
		rmb();	/* Hypervisor might update data. */
	} while (get64(&state->state_entry_time) != state_time ||
		 (state_time & XEN_RUNSTATE_UPDATE));
}

static void xen_get_runstate_snapshot_cpu(struct vcpu_runstate_info *res,
					  unsigned int cpu)
{
	int i;

	xen_get_runstate_snapshot_cpu_delta(res, cpu);

	for (i = 0; i < 4; i++)
		res->time[i] += per_cpu(old_runstate_time, cpu)[i];
}

void xen_manage_runstate_time(int action)
{
	static struct vcpu_runstate_info *runstate_delta;
	struct vcpu_runstate_info state;
	int cpu, i;

	switch (action) {
	case -1: /* backup runstate time before suspend */
		if (unlikely(runstate_delta))
			pr_warn_once("%s: memory leak as runstate_delta is not NULL\n",
					__func__);

		runstate_delta = kmalloc_array(num_possible_cpus(),
					sizeof(*runstate_delta),
					GFP_ATOMIC);
		if (unlikely(!runstate_delta)) {
			pr_warn("%s: failed to allocate runstate_delta\n",
					__func__);
			return;
		}

		for_each_possible_cpu(cpu) {
			xen_get_runstate_snapshot_cpu_delta(&state, cpu);
			memcpy(runstate_delta[cpu].time, state.time,
					sizeof(runstate_delta[cpu].time));
		}

		break;

	case 0: /* backup runstate time after resume */
		if (unlikely(!runstate_delta)) {
			pr_warn("%s: cannot accumulate runstate time as runstate_delta is NULL\n",
					__func__);
			return;
		}

		for_each_possible_cpu(cpu) {
			for (i = 0; i < 4; i++)
				per_cpu(old_runstate_time, cpu)[i] +=
					runstate_delta[cpu].time[i];
		}

		break;

	default: /* do not accumulate runstate time for checkpointing */
		break;
	}

	if (action != -1 && runstate_delta) {
		kfree(runstate_delta);
		runstate_delta = NULL;
	}
}

/*
 * Runstate accounting
 */
void xen_get_runstate_snapshot(struct vcpu_runstate_info *res)
{
	xen_get_runstate_snapshot_cpu(res, smp_processor_id());
}

/* return true when a vcpu could run but has no real cpu to run on */
bool xen_vcpu_stolen(int vcpu)
{
	return per_cpu(xen_runstate, vcpu).state == RUNSTATE_runnable;
}

u64 xen_steal_clock(int cpu)
{
	struct vcpu_runstate_info state;

	xen_get_runstate_snapshot_cpu(&state, cpu);
	return state.time[RUNSTATE_runnable] + state.time[RUNSTATE_offline];
}

void xen_setup_runstate_info(int cpu)
{
	struct vcpu_register_runstate_memory_area area;

	area.addr.v = &per_cpu(xen_runstate, cpu);

	if (HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area,
			       xen_vcpu_nr(cpu), &area))
		BUG();
}

void __init xen_time_setup_guest(void)
{
	bool xen_runstate_remote;

	xen_runstate_remote = !HYPERVISOR_vm_assist(VMASST_CMD_enable,
					VMASST_TYPE_runstate_update_flag);

	pv_ops.time.steal_clock = xen_steal_clock;

	static_key_slow_inc(&paravirt_steal_enabled);
	if (xen_runstate_remote)
		static_key_slow_inc(&paravirt_steal_rq_enabled);
}
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
4ccefbe5 SS	2	/*
	3	* Xen stolen ticks accounting.
	4	*/
	5	#include <linux/kernel.h>
	6	#include <linux/kernel_stat.h>
	7	#include <linux/math64.h>
	8	#include <linux/gfp.h>
5e25f5db	9	#include <linux/slab.h>
4ccefbe5	10
ecb23dc6	11	#include <asm/paravirt.h>
4ccefbe5 SS	12	#include <asm/xen/hypervisor.h>
	13	#include <asm/xen/hypercall.h>
	14
	15	#include <xen/events.h>
	16	#include <xen/features.h>
	17	#include <xen/interface/xen.h>
	18	#include <xen/interface/vcpu.h>
	19	#include <xen/xen-ops.h>
	20
	21	/* runstate info updated by Xen */
	22	static DEFINE_PER_CPU(struct vcpu_runstate_info, xen_runstate);
	23
5e25f5db DZ	24	static DEFINE_PER_CPU(u64[4], old_runstate_time);
5e25f5db DZ	25
4ccefbe5 SS	26	/* return an consistent snapshot of 64-bit time/counter value */
	27	static u64 get64(const u64 *p)
	28	{
	29	u64 ret;
	30
	31	if (BITS_PER_LONG < 64) {
	32	u32 p32 = (u32 )p;
2dd887e3	33	u32 h, l, h2;
4ccefbe5 SS	34
	35	/*
	36	* Read high then low, and then make sure high is
	37	* still the same; this will only loop if low wraps
	38	* and carries into high.
	39	* XXX some clean way to make this endian-proof?
	40	*/
	41	do {
2dd887e3 SS	42	h = READ_ONCE(p32[1]);
	43	l = READ_ONCE(p32[0]);
	44	h2 = READ_ONCE(p32[1]);
	45	} while(h2 != h);
4ccefbe5 SS	46
	47	ret = (((u64)h) << 32) \| l;
	48	} else
2dd887e3	49	ret = READ_ONCE(*p);
4ccefbe5 SS	50
	51	return ret;
	52	}
	53
5e25f5db DZ	54	static void xen_get_runstate_snapshot_cpu_delta(
5e25f5db DZ	55	struct vcpu_runstate_info *res, unsigned int cpu)
4ccefbe5 SS	56	{
	57	u64 state_time;
	58	struct vcpu_runstate_info *state;
	59
	60	BUG_ON(preemptible());
	61
6ba286ad	62	state = per_cpu_ptr(&xen_runstate, cpu);
4ccefbe5	63
4ccefbe5 SS	64	do {
4ccefbe5 SS	65	state_time = get64(&state->state_entry_time);
6ba286ad	66	rmb(); /* Hypervisor might update data. */
2dd887e3	67	res = READ_ONCE(state);
6ba286ad JG	68	rmb(); /* Hypervisor might update data. */
	69	} while (get64(&state->state_entry_time) != state_time \|\|
	70	(state_time & XEN_RUNSTATE_UPDATE));
	71	}
	72
5e25f5db DZ	73	static void xen_get_runstate_snapshot_cpu(struct vcpu_runstate_info *res,
	74	unsigned int cpu)
	75	{
	76	int i;
	77
	78	xen_get_runstate_snapshot_cpu_delta(res, cpu);
	79
	80	for (i = 0; i < 4; i++)
	81	res->time[i] += per_cpu(old_runstate_time, cpu)[i];
	82	}
	83
	84	void xen_manage_runstate_time(int action)
	85	{
	86	static struct vcpu_runstate_info *runstate_delta;
	87	struct vcpu_runstate_info state;
	88	int cpu, i;
	89
	90	switch (action) {
	91	case -1: /* backup runstate time before suspend */
	92	if (unlikely(runstate_delta))
	93	pr_warn_once("%s: memory leak as runstate_delta is not NULL\n",
	94	__func__);
	95
	96	runstate_delta = kmalloc_array(num_possible_cpus(),
	97	sizeof(*runstate_delta),
	98	GFP_ATOMIC);
	99	if (unlikely(!runstate_delta)) {
	100	pr_warn("%s: failed to allocate runstate_delta\n",
	101	__func__);
	102	return;
	103	}
	104
	105	for_each_possible_cpu(cpu) {
	106	xen_get_runstate_snapshot_cpu_delta(&state, cpu);
	107	memcpy(runstate_delta[cpu].time, state.time,
	108	sizeof(runstate_delta[cpu].time));
	109	}
	110
	111	break;
	112
	113	case 0: /* backup runstate time after resume */
	114	if (unlikely(!runstate_delta)) {
	115	pr_warn("%s: cannot accumulate runstate time as runstate_delta is NULL\n",
	116	__func__);
	117	return;
	118	}
	119
	120	for_each_possible_cpu(cpu) {
	121	for (i = 0; i < 4; i++)
	122	per_cpu(old_runstate_time, cpu)[i] +=
	123	runstate_delta[cpu].time[i];
	124	}
	125
	126	break;
	127
	128	default: /* do not accumulate runstate time for checkpointing */
	129	break;
	130	}
	131
	132	if (action != -1 && runstate_delta) {
	133	kfree(runstate_delta);
	134	runstate_delta = NULL;
	135	}
	136	}
137
6ba286ad JG	138	/*
	139	* Runstate accounting
	140	*/
	141	void xen_get_runstate_snapshot(struct vcpu_runstate_info *res)
	142	{
	143	xen_get_runstate_snapshot_cpu(res, smp_processor_id());
4ccefbe5 SS	144	}
	145
	146	/* return true when a vcpu could run but has no real cpu to run on */
	147	bool xen_vcpu_stolen(int vcpu)
	148	{
	149	return per_cpu(xen_runstate, vcpu).state == RUNSTATE_runnable;
	150	}
	151
d34c30cc	152	u64 xen_steal_clock(int cpu)
ecb23dc6 JG	153	{
	154	struct vcpu_runstate_info state;
	155
6ba286ad	156	xen_get_runstate_snapshot_cpu(&state, cpu);
ecb23dc6 JG	157	return state.time[RUNSTATE_runnable] + state.time[RUNSTATE_offline];
	158	}
	159
4ccefbe5 SS	160	void xen_setup_runstate_info(int cpu)
	161	{
	162	struct vcpu_register_runstate_memory_area area;
	163
	164	area.addr.v = &per_cpu(xen_runstate, cpu);
	165
	166	if (HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area,
ad5475f9	167	xen_vcpu_nr(cpu), &area))
4ccefbe5 SS	168	BUG();
	169	}
	170
ecb23dc6 JG	171	void __init xen_time_setup_guest(void)
ecb23dc6 JG	172	{
6ba286ad JG	173	bool xen_runstate_remote;
	174
	175	xen_runstate_remote = !HYPERVISOR_vm_assist(VMASST_CMD_enable,
	176	VMASST_TYPE_runstate_update_flag);
	177
5c83511b	178	pv_ops.time.steal_clock = xen_steal_clock;
ecb23dc6 JG	179
ecb23dc6 JG	180	static_key_slow_inc(&paravirt_steal_enabled);
6ba286ad JG	181	if (xen_runstate_remote)
6ba286ad JG	182	static_key_slow_inc(&paravirt_steal_rq_enabled);
ecb23dc6	183	}