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
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
|
/*
* Xen time implementation.
*
* This is implemented in terms of a clocksource driver which uses
* the hypervisor clock as a nanosecond timebase, and a clockevent
* driver which uses the hypervisor's timer mechanism.
*
* Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007
*/
#include <linux/kernel.h>
#include <linux/interrupt.h>
#include <linux/clocksource.h>
#include <linux/clockchips.h>
#include <linux/kernel_stat.h>
#include <linux/math64.h>
#include <asm/pvclock.h>
#include <asm/xen/hypervisor.h>
#include <asm/xen/hypercall.h>
#include <xen/events.h>
#include <xen/interface/xen.h>
#include <xen/interface/vcpu.h>
#include "xen-ops.h"
#define XEN_SHIFT 22
/* Xen may fire a timer up to this many ns early */
#define TIMER_SLOP 100000
#define NS_PER_TICK (1000000000LL / HZ)
/* runstate info updated by Xen */
static DEFINE_PER_CPU(struct vcpu_runstate_info, runstate);
/* snapshots of runstate info */
static DEFINE_PER_CPU(struct vcpu_runstate_info, runstate_snapshot);
/* unused ns of stolen and blocked time */
static DEFINE_PER_CPU(u64, residual_stolen);
static DEFINE_PER_CPU(u64, residual_blocked);
/* 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;
/*
* 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 = p32[1];
barrier();
l = p32[0];
barrier();
} while (p32[1] != h);
ret = (((u64)h) << 32) | l;
} else
ret = *p;
return ret;
}
/*
* Runstate accounting
*/
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 = get64(&state->state_entry_time);
barrier();
*res = *state;
barrier();
} while (get64(&state->state_entry_time) != state_time);
}
/* return true when a vcpu could run but has no real cpu to run on */
bool xen_vcpu_stolen(int vcpu)
{
return per_cpu(runstate, vcpu).state == RUNSTATE_runnable;
}
static void setup_runstate_info(int cpu)
{
struct vcpu_register_runstate_memory_area area;
area.addr.v = &per_cpu(runstate, cpu);
if (HYPERVISOR_vcpu_op(VCPUOP_register_runstate_memory_area,
cpu, &area))
BUG();
}
static void do_stolen_accounting(void)
{
struct vcpu_runstate_info state;
struct vcpu_runstate_info *snap;
s64 blocked, runnable, offline, stolen;
cputime_t ticks;
get_runstate_snapshot(&state);
WARN_ON(state.state != RUNSTATE_running);
snap = &__get_cpu_var(runstate_snapshot);
/* work out how much time the VCPU has not been runn*ing* */
blocked = state.time[RUNSTATE_blocked] - snap->time[RUNSTATE_blocked];
runnable = state.time[RUNSTATE_runnable] - snap->time[RUNSTATE_runnable];
offline = state.time[RUNSTATE_offline] - snap->time[RUNSTATE_offline];
*snap = state;
/* Add the appropriate number of ticks of stolen time,
including any left-overs from last time. */
stolen = runnable + offline + __get_cpu_var(residual_stolen);
if (stolen < 0)
stolen = 0;
ticks = iter_div_u64_rem(stolen, NS_PER_TICK, &stolen);
__get_cpu_var(residual_stolen) = stolen;
account_steal_ticks(ticks);
/* Add the appropriate number of ticks of blocked time,
including any left-overs from last time. */
blocked += __get_cpu_var(residual_blocked);
if (blocked < 0)
blocked = 0;
ticks = iter_div_u64_rem(blocked, NS_PER_TICK, &blocked);
__get_cpu_var(residual_blocked) = blocked;
account_idle_ticks(ticks);
}
/*
* Xen sched_clock implementation. Returns the number of unstolen
* nanoseconds, which is nanoseconds the VCPU spent in RUNNING+BLOCKED
* states.
*/
unsigned long long xen_sched_clock(void)
{
struct vcpu_runstate_info state;
cycle_t now;
u64 ret;
s64 offset;
/*
* 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();
now = xen_clocksource_read();
get_runstate_snapshot(&state);
WARN_ON(state.state != RUNSTATE_running);
offset = now - state.state_entry_time;
if (offset < 0)
offset = 0;
ret = state.time[RUNSTATE_blocked] +
state.time[RUNSTATE_running] +
offset;
preempt_enable();
return ret;
}
/* Get the TSC speed from Xen */
unsigned long xen_tsc_khz(void)
{
struct pvclock_vcpu_time_info *info =
&HYPERVISOR_shared_info->vcpu_info[0].time;
return pvclock_tsc_khz(info);
}
cycle_t xen_clocksource_read(void)
{
struct pvclock_vcpu_time_info *src;
cycle_t ret;
src = &get_cpu_var(xen_vcpu)->time;
ret = pvclock_clocksource_read(src);
put_cpu_var(xen_vcpu);
return ret;
}
static cycle_t xen_clocksource_get_cycles(struct clocksource *cs)
{
return xen_clocksource_read();
}
static void xen_read_wallclock(struct timespec *ts)
{
struct shared_info *s = HYPERVISOR_shared_info;
struct pvclock_wall_clock *wall_clock = &(s->wc);
struct pvclock_vcpu_time_info *vcpu_time;
vcpu_time = &get_cpu_var(xen_vcpu)->time;
pvclock_read_wallclock(wall_clock, vcpu_time, ts);
put_cpu_var(xen_vcpu);
}
unsigned long xen_get_wallclock(void)
{
struct timespec ts;
xen_read_wallclock(&ts);
return ts.tv_sec;
}
int xen_set_wallclock(unsigned long now)
{
/* do nothing for domU */
return -1;
}
static struct clocksource xen_clocksource __read_mostly = {
.name = "xen",
.rating = 400,
.read = xen_clocksource_get_cycles,
.mask = ~0,
.mult = 1<<XEN_SHIFT, /* time directly in nanoseconds */
.shift = XEN_SHIFT,
.flags = CLOCK_SOURCE_IS_CONTINUOUS,
};
/*
Xen clockevent implementation
Xen has two clockevent implementations:
The old timer_op one works with all released versions of Xen prior
to version 3.0.4. This version of the hypervisor provides a
single-shot timer with nanosecond resolution. However, sharing the
same event channel is a 100Hz tick which is delivered while the
vcpu is running. We don't care about or use this tick, but it will
cause the core time code to think the timer fired too soon, and
will end up resetting it each time. It could be filtered, but
doing so has complications when the ktime clocksource is not yet
the xen clocksource (ie, at boot time).
The new vcpu_op-based timer interface allows the tick timer period
to be changed or turned off. The tick timer is not useful as a
periodic timer because events are only delivered to running vcpus.
The one-shot timer can report when a timeout is in the past, so
set_next_event is capable of returning -ETIME when appropriate.
This interface is used when available.
*/
/*
Get a hypervisor absolute time. In theory we could maintain an
offset between the kernel's time and the hypervisor's time, and
apply that to a kernel's absolute timeout. Unfortunately the
hypervisor and kernel times can drift even if the kernel is using
the Xen clocksource, because ntp can warp the kernel's clocksource.
*/
static s64 get_abs_timeout(unsigned long delta)
{
return xen_clocksource_read() + delta;
}
static void xen_timerop_set_mode(enum clock_event_mode mode,
struct clock_event_device *evt)
{
switch (mode) {
case CLOCK_EVT_MODE_PERIODIC:
/* unsupported */
WARN_ON(1);
break;
case CLOCK_EVT_MODE_ONESHOT:
case CLOCK_EVT_MODE_RESUME:
break;
case CLOCK_EVT_MODE_UNUSED:
case CLOCK_EVT_MODE_SHUTDOWN:
HYPERVISOR_set_timer_op(0); /* cancel timeout */
break;
}
}
static int xen_timerop_set_next_event(unsigned long delta,
struct clock_event_device *evt)
{
WARN_ON(evt->mode != CLOCK_EVT_MODE_ONESHOT);
if (HYPERVISOR_set_timer_op(get_abs_timeout(delta)) < 0)
BUG();
/* We may have missed the deadline, but there's no real way of
knowing for sure. If the event was in the past, then we'll
get an immediate interrupt. */
return 0;
}
static const struct clock_event_device xen_timerop_clockevent = {
.name = "xen",
.features = CLOCK_EVT_FEAT_ONESHOT,
.max_delta_ns
|