summaryrefslogtreecommitdiffstats
path: root/arch/ia64/xen/time.c
blob: dbeadb9c8e20550fd0e4a6ad4d2c84e7420430b9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
/******************************************************************************
 * arch/ia64/xen/time.c
 *
 * Copyright (c) 2008 Isaku Yamahata <yamahata at valinux co jp>
 *                    VA Linux Systems Japan K.K.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 */

#include <linux/delay.h>
#include <linux/kernel_stat.h>
#include <linux/posix-timers.h>
#include <linux/irq.h>
#include <linux/clocksource.h>

#include <asm/timex.h>

#include <asm/xen/hypervisor.h>

#include <xen/interface/vcpu.h>

#include "../kernel/fsyscall_gtod_data.h"

DEFINE_PER_CPU(struct vcpu_runstate_info, runstate);
DEFINE_PER_CPU(unsigned long, processed_stolen_time);
DEFINE_PER_CPU(unsigned long, processed_blocked_time);

/* taken from i386/kernel/time-xen.c */
static void xen_init_missing_ticks_accounting(int cpu)
{
	struct vcpu_register_runstate_memory_area area;
	struct vcpu_runstate_info *runstate = &per_cpu(runstate, cpu);
	int rc;

	memset(runstate, 0, sizeof(*runstate));

	area.addr.v = runstate;
	rc = HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area, cpu,
				&area);
	WARN_ON(rc && rc != -ENOSYS);

	per_cpu(processed_blocked_time, cpu) = runstate->time[RUNSTATE_blocked];
	per_cpu(processed_stolen_time, cpu) = runstate->time[RUNSTATE_runnable]
					    + runstate->time[RUNSTATE_offline];
}

/*
 * Runstate accounting
 */
/* stolen from arch/x86/xen/time.c */
static void get_runstate_snapshot(struct vcpu_runstate_info *res)
{
	u64 state_time;
	struct vcpu_runstate_info *state;

	BUG_ON(preemptible());

	state = &__get_cpu_var(runstate);

	/*
	 * The runstate info is always updated by the hypervisor on
	 * the current CPU, so there's no need to use anything
	 * stronger than a compiler barrier when fetching it.
	 */
	do {
		state_time = state->state_entry_time;
		rmb();
		*res = *state;
		rmb();
	} while (state->state_entry_time != state_time);
}

#define NS_PER_TICK (1000000000LL/HZ)

static unsigned long
consider_steal_time(unsigned long new_itm)
{
	unsigned long stolen, blocked;
	unsigned long delta_itm = 0, stolentick = 0;
	int cpu = smp_processor_id();
	struct vcpu_runstate_info runstate;
	struct task_struct *p = current;

	get_runstate_snapshot(&runstate);

	/*
	 * Check for vcpu migration effect
	 * In this case, itc value is reversed.
	 * This causes huge stolen value.
	 * This function just checks and reject this effect.
	 */
	if (!time_after_eq(runstate.time[RUNSTATE_blocked],
			   per_cpu(processed_blocked_time, cpu)))
		blocked = 0;

	if (!time_after_eq(runstate.time[RUNSTATE_runnable] +
			   runstate.time[RUNSTATE_offline],
			   per_cpu(processed_stolen_time, cpu)))
		stolen = 0;

	if (!time_after(delta_itm + new_itm, ia64_get_itc()))
		stolentick = ia64_get_itc() - new_itm;

	do_div(stolentick, NS_PER_TICK);
	stolentick++;

	do_div(stolen, NS_PER_TICK);

	if (stolen > stolentick)
		stolen = stolentick;

	stolentick -= stolen;
	do_div(blocked, NS_PER_TICK);

	if (blocked > stolentick)
		blocked = stolentick;

	if (stolen > 0 || blocked > 0) {
		account_steal_ticks(stolen);
		account_idle_ticks(blocked);
		run_local_timers();

		rcu_check_callbacks(cpu, user_mode(get_irq_regs()));

		scheduler_tick();
		run_posix_cpu_timers(p);
		delta_itm += local_cpu_data->itm_delta * (stolen + blocked);

		if (cpu == time_keeper_id) {
			write_seqlock(&xtime_lock);
			do_timer(stolen + blocked);
			local_cpu_data->itm_next = delta_itm + new_itm;
			write_sequnlock(&xtime_lock);
		} else {
			local_cpu_data->itm_next = delta_itm + new_itm;
		}
		per_cpu(processed_stolen_time, cpu) += NS_PER_TICK * stolen;
		per_cpu(processed_blocked_time, cpu) += NS_PER_TICK * blocked;
	}
	return delta_itm;
}

static int xen_do_steal_accounting(unsigned long *new_itm)
{
	unsigned long delta_itm;
	delta_itm = consider_steal_time(*new_itm);
	*new_itm += delta_itm;
	if (time_after(*new_itm, ia64_get_itc()) && delta_itm)
		return 1;

	return 0;
}

static void xen_itc_jitter_data_reset(void)
{
	u64 lcycle, ret;

	do {
		lcycle = itc_jitter_data.itc_lastcycle;
		ret = cmpxchg(&itc_jitter_data.itc_lastcycle, lcycle, 0);
	} while (unlikely(ret != lcycle));
}

/* based on xen_sched_clock() in arch/x86/xen/time.c. */
/*
 * This relies on HAVE_UNSTABLE_SCHED_CLOCK. If it can't be defined,
 * something similar logic should be implemented here.
 */
/*
 * Xen sched_clock implementation.  Returns the number of unstolen
 * nanoseconds, which is nanoseconds the VCPU spent in RUNNING+BLOCKED
 * states.
 */
static unsigned long long xen_sched_clock(void)
{
	struct vcpu_runstate_info runstate;

	unsigned long long now;
	unsigned long long offset;
	unsigned long long ret;

	/*
	 * Ideally sched_clock should be called on a per-cpu basis
	 * anyway, so preempt should already be disabled, but that's
	 * not current practice at the moment.
	 */
	preempt_disable();

	/*
	 * both ia64_native_sched_clock() and xen's runstate are
	 * based on mAR.ITC. So difference of them makes sense.
	 */
	now = ia64_native_sched_clock();

	get_runstate_snapshot(&runstate);

	WARN_ON(runstate.state != RUNSTATE_running);

	offset = 0;
	if (now > runstate.state_entry_time)
		offset = now - runstate.state_entry_time;
	ret = runstate.time[RUNSTATE_blocked] +
		runstate.time[RUNSTATE_running] +
		offset;

	preempt_enable();

	return ret;
}

struct pv_time_ops xen_time_ops __initdata = {
	.init_missing_ticks_accounting	= xen_init_missing_ticks_accounting,
	.do_steal_accounting		= xen_do_steal_accounting,
	.clocksource_resume		= xen_itc_jitter_data_reset,
	.sched_clock			= xen_sched_clock,
};

/* Called after suspend, to resume time.  */
static void xen_local_tick_resume(void)
{
	/* Just trigger a tick.  */
	ia64_cpu_local_tick();
	touch_softlockup_watchdog();
}

void
xen_timer_resume(void)
{
	unsigned int cpu;

	xen_local_tick_resume();

	for_each_online_cpu(cpu)
		xen_init_missing_ticks_accounting(cpu);
}

static void ia64_cpu_local_tick_fn(void *unused)
{
	xen_local_tick_resume();
	xen_init_missing_ticks_accounting(smp_processor_id());
}

void
xen_timer_resume_on_aps(void)
{
	smp_call_function(&ia64_cpu_local_tick_fn, NULL, 1);
}