smp.c 17.8 KB
Newer Older
J
Jeremy Fitzhardinge 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * Xen SMP support
 *
 * This file implements the Xen versions of smp_ops.  SMP under Xen is
 * very straightforward.  Bringing a CPU up is simply a matter of
 * loading its initial context and setting it running.
 *
 * IPIs are handled through the Xen event mechanism.
 *
 * Because virtual CPUs can be scheduled onto any real CPU, there's no
 * useful topology information for the kernel to make use of.  As a
 * result, all CPUs are treated as if they're single-core and
 * single-threaded.
 */
#include <linux/sched.h>
#include <linux/err.h>
17
#include <linux/slab.h>
J
Jeremy Fitzhardinge 已提交
18
#include <linux/smp.h>
19
#include <linux/irq_work.h>
20
#include <linux/tick.h>
J
Jeremy Fitzhardinge 已提交
21 22 23 24 25 26 27 28 29 30 31 32

#include <asm/paravirt.h>
#include <asm/desc.h>
#include <asm/pgtable.h>
#include <asm/cpu.h>

#include <xen/interface/xen.h>
#include <xen/interface/vcpu.h>

#include <asm/xen/interface.h>
#include <asm/xen/hypercall.h>

33
#include <xen/xen.h>
J
Jeremy Fitzhardinge 已提交
34 35 36
#include <xen/page.h>
#include <xen/events.h>

37
#include <xen/hvc-console.h>
J
Jeremy Fitzhardinge 已提交
38 39 40
#include "xen-ops.h"
#include "mmu.h"

41
cpumask_var_t xen_cpu_initialized_map;
J
Jeremy Fitzhardinge 已提交
42

43 44 45 46
struct xen_common_irq {
	int irq;
	char *name;
};
47 48 49 50
static DEFINE_PER_CPU(struct xen_common_irq, xen_resched_irq) = { .irq = -1 };
static DEFINE_PER_CPU(struct xen_common_irq, xen_callfunc_irq) = { .irq = -1 };
static DEFINE_PER_CPU(struct xen_common_irq, xen_callfuncsingle_irq) = { .irq = -1 };
static DEFINE_PER_CPU(struct xen_common_irq, xen_irq_work) = { .irq = -1 };
51
static DEFINE_PER_CPU(struct xen_common_irq, xen_debug_irq) = { .irq = -1 };
J
Jeremy Fitzhardinge 已提交
52 53

static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id);
54
static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id);
55
static irqreturn_t xen_irq_work_interrupt(int irq, void *dev_id);
J
Jeremy Fitzhardinge 已提交
56 57

/*
58
 * Reschedule call back.
J
Jeremy Fitzhardinge 已提交
59 60 61
 */
static irqreturn_t xen_reschedule_interrupt(int irq, void *dev_id)
{
62
	inc_irq_stat(irq_resched_count);
63
	scheduler_ipi();
64

J
Jeremy Fitzhardinge 已提交
65 66 67
	return IRQ_HANDLED;
}

68
static void cpu_bringup(void)
J
Jeremy Fitzhardinge 已提交
69
{
70
	int cpu;
J
Jeremy Fitzhardinge 已提交
71 72

	cpu_init();
A
Alex Nixon 已提交
73
	touch_softlockup_watchdog();
74 75
	preempt_disable();

76
	xen_enable_sysenter();
77
	xen_enable_syscall();
J
Jeremy Fitzhardinge 已提交
78

79 80 81 82
	cpu = smp_processor_id();
	smp_store_cpu_info(cpu);
	cpu_data(cpu).x86_max_cores = 1;
	set_cpu_sibling_map(cpu);
J
Jeremy Fitzhardinge 已提交
83 84 85

	xen_setup_cpu_clockevents();

86 87
	notify_cpu_starting(cpu);

88
	set_cpu_online(cpu, true);
89

90
	this_cpu_write(cpu_state, CPU_ONLINE);
91

92 93
	wmb();

J
Jeremy Fitzhardinge 已提交
94 95 96 97
	/* We can take interrupts now: we're officially "up". */
	local_irq_enable();

	wmb();			/* make sure everything is out */
A
Alex Nixon 已提交
98 99
}

100
static void cpu_bringup_and_idle(void)
A
Alex Nixon 已提交
101 102
{
	cpu_bringup();
T
Thomas Gleixner 已提交
103
	cpu_startup_entry(CPUHP_ONLINE);
J
Jeremy Fitzhardinge 已提交
104 105
}

106 107
static void xen_smp_intr_free(unsigned int cpu)
{
108
	if (per_cpu(xen_resched_irq, cpu).irq >= 0) {
109
		unbind_from_irqhandler(per_cpu(xen_resched_irq, cpu).irq, NULL);
110
		per_cpu(xen_resched_irq, cpu).irq = -1;
111 112
		kfree(per_cpu(xen_resched_irq, cpu).name);
		per_cpu(xen_resched_irq, cpu).name = NULL;
113 114
	}
	if (per_cpu(xen_callfunc_irq, cpu).irq >= 0) {
115
		unbind_from_irqhandler(per_cpu(xen_callfunc_irq, cpu).irq, NULL);
116
		per_cpu(xen_callfunc_irq, cpu).irq = -1;
117 118
		kfree(per_cpu(xen_callfunc_irq, cpu).name);
		per_cpu(xen_callfunc_irq, cpu).name = NULL;
119 120
	}
	if (per_cpu(xen_debug_irq, cpu).irq >= 0) {
121
		unbind_from_irqhandler(per_cpu(xen_debug_irq, cpu).irq, NULL);
122
		per_cpu(xen_debug_irq, cpu).irq = -1;
123 124
		kfree(per_cpu(xen_debug_irq, cpu).name);
		per_cpu(xen_debug_irq, cpu).name = NULL;
125 126
	}
	if (per_cpu(xen_callfuncsingle_irq, cpu).irq >= 0) {
127
		unbind_from_irqhandler(per_cpu(xen_callfuncsingle_irq, cpu).irq,
128
				       NULL);
129
		per_cpu(xen_callfuncsingle_irq, cpu).irq = -1;
130 131
		kfree(per_cpu(xen_callfuncsingle_irq, cpu).name);
		per_cpu(xen_callfuncsingle_irq, cpu).name = NULL;
132
	}
133 134 135
	if (xen_hvm_domain())
		return;

136
	if (per_cpu(xen_irq_work, cpu).irq >= 0) {
137
		unbind_from_irqhandler(per_cpu(xen_irq_work, cpu).irq, NULL);
138
		per_cpu(xen_irq_work, cpu).irq = -1;
139 140
		kfree(per_cpu(xen_irq_work, cpu).name);
		per_cpu(xen_irq_work, cpu).name = NULL;
141
	}
142
};
J
Jeremy Fitzhardinge 已提交
143 144 145
static int xen_smp_intr_init(unsigned int cpu)
{
	int rc;
146
	char *resched_name, *callfunc_name, *debug_name;
J
Jeremy Fitzhardinge 已提交
147 148 149 150 151 152 153 154 155 156

	resched_name = kasprintf(GFP_KERNEL, "resched%d", cpu);
	rc = bind_ipi_to_irqhandler(XEN_RESCHEDULE_VECTOR,
				    cpu,
				    xen_reschedule_interrupt,
				    IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
				    resched_name,
				    NULL);
	if (rc < 0)
		goto fail;
157
	per_cpu(xen_resched_irq, cpu).irq = rc;
158
	per_cpu(xen_resched_irq, cpu).name = resched_name;
J
Jeremy Fitzhardinge 已提交
159 160 161 162 163 164 165 166 167 168

	callfunc_name = kasprintf(GFP_KERNEL, "callfunc%d", cpu);
	rc = bind_ipi_to_irqhandler(XEN_CALL_FUNCTION_VECTOR,
				    cpu,
				    xen_call_function_interrupt,
				    IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
				    callfunc_name,
				    NULL);
	if (rc < 0)
		goto fail;
169
	per_cpu(xen_callfunc_irq, cpu).irq = rc;
170
	per_cpu(xen_callfunc_irq, cpu).name = callfunc_name;
J
Jeremy Fitzhardinge 已提交
171

172 173 174 175 176 177
	debug_name = kasprintf(GFP_KERNEL, "debug%d", cpu);
	rc = bind_virq_to_irqhandler(VIRQ_DEBUG, cpu, xen_debug_interrupt,
				     IRQF_DISABLED | IRQF_PERCPU | IRQF_NOBALANCING,
				     debug_name, NULL);
	if (rc < 0)
		goto fail;
178
	per_cpu(xen_debug_irq, cpu).irq = rc;
179
	per_cpu(xen_debug_irq, cpu).name = debug_name;
180

181 182 183 184 185 186 187 188 189
	callfunc_name = kasprintf(GFP_KERNEL, "callfuncsingle%d", cpu);
	rc = bind_ipi_to_irqhandler(XEN_CALL_FUNCTION_SINGLE_VECTOR,
				    cpu,
				    xen_call_function_single_interrupt,
				    IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
				    callfunc_name,
				    NULL);
	if (rc < 0)
		goto fail;
190
	per_cpu(xen_callfuncsingle_irq, cpu).irq = rc;
191
	per_cpu(xen_callfuncsingle_irq, cpu).name = callfunc_name;
192

193 194 195 196 197 198 199
	/*
	 * The IRQ worker on PVHVM goes through the native path and uses the
	 * IPI mechanism.
	 */
	if (xen_hvm_domain())
		return 0;

200 201 202 203 204 205 206 207 208
	callfunc_name = kasprintf(GFP_KERNEL, "irqwork%d", cpu);
	rc = bind_ipi_to_irqhandler(XEN_IRQ_WORK_VECTOR,
				    cpu,
				    xen_irq_work_interrupt,
				    IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
				    callfunc_name,
				    NULL);
	if (rc < 0)
		goto fail;
209
	per_cpu(xen_irq_work, cpu).irq = rc;
210
	per_cpu(xen_irq_work, cpu).name = callfunc_name;
211

J
Jeremy Fitzhardinge 已提交
212 213 214
	return 0;

 fail:
215
	xen_smp_intr_free(cpu);
J
Jeremy Fitzhardinge 已提交
216 217 218
	return rc;
}

219
static void __init xen_fill_possible_map(void)
J
Jeremy Fitzhardinge 已提交
220 221 222
{
	int i, rc;

223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
	if (xen_initial_domain())
		return;

	for (i = 0; i < nr_cpu_ids; i++) {
		rc = HYPERVISOR_vcpu_op(VCPUOP_is_up, i, NULL);
		if (rc >= 0) {
			num_processors++;
			set_cpu_possible(i, true);
		}
	}
}

static void __init xen_filter_cpu_maps(void)
{
	int i, rc;
238
	unsigned int subtract = 0;
239 240 241 242

	if (!xen_initial_domain())
		return;

243 244
	num_processors = 0;
	disabled_cpus = 0;
245
	for (i = 0; i < nr_cpu_ids; i++) {
J
Jeremy Fitzhardinge 已提交
246
		rc = HYPERVISOR_vcpu_op(VCPUOP_is_up, i, NULL);
J
Jeremy Fitzhardinge 已提交
247 248
		if (rc >= 0) {
			num_processors++;
249
			set_cpu_possible(i, true);
250 251 252
		} else {
			set_cpu_possible(i, false);
			set_cpu_present(i, false);
253
			subtract++;
J
Jeremy Fitzhardinge 已提交
254
		}
J
Jeremy Fitzhardinge 已提交
255
	}
256 257 258 259 260 261 262 263 264 265 266 267 268
#ifdef CONFIG_HOTPLUG_CPU
	/* This is akin to using 'nr_cpus' on the Linux command line.
	 * Which is OK as when we use 'dom0_max_vcpus=X' we can only
	 * have up to X, while nr_cpu_ids is greater than X. This
	 * normally is not a problem, except when CPU hotplugging
	 * is involved and then there might be more than X CPUs
	 * in the guest - which will not work as there is no
	 * hypercall to expand the max number of VCPUs an already
	 * running guest has. So cap it up to X. */
	if (subtract)
		nr_cpu_ids = nr_cpu_ids - subtract;
#endif

J
Jeremy Fitzhardinge 已提交
269 270
}

271
static void __init xen_smp_prepare_boot_cpu(void)
J
Jeremy Fitzhardinge 已提交
272 273 274 275
{
	BUG_ON(smp_processor_id() != 0);
	native_smp_prepare_boot_cpu();

276 277 278 279
	if (xen_pv_domain()) {
		/* We've switched to the "real" per-cpu gdt, so make sure the
		   old memory can be recycled */
		make_lowmem_page_readwrite(xen_initial_gdt);
280

281 282 283 284 285 286 287 288 289
		xen_filter_cpu_maps();
		xen_setup_vcpu_info_placement();
	}
	/*
	 * The alternative logic (which patches the unlock/lock) runs before
	 * the smp bootup up code is activated. Hence we need to set this up
	 * the core kernel is being patched. Otherwise we will have only
	 * modules patched but not core code.
	 */
290
	xen_init_spinlocks();
J
Jeremy Fitzhardinge 已提交
291 292
}

293
static void __init xen_smp_prepare_cpus(unsigned int max_cpus)
J
Jeremy Fitzhardinge 已提交
294 295
{
	unsigned cpu;
A
Andrew Jones 已提交
296
	unsigned int i;
J
Jeremy Fitzhardinge 已提交
297

298 299 300 301 302 303 304 305 306
	if (skip_ioapic_setup) {
		char *m = (max_cpus == 0) ?
			"The nosmp parameter is incompatible with Xen; " \
			"use Xen dom0_max_vcpus=1 parameter" :
			"The noapic parameter is incompatible with Xen";

		xen_raw_printk(m);
		panic(m);
	}
307 308
	xen_init_lock_cpu(0);

309
	smp_store_boot_cpu_info();
310
	cpu_data(0).x86_max_cores = 1;
A
Andrew Jones 已提交
311 312 313 314 315 316

	for_each_possible_cpu(i) {
		zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
		zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
		zalloc_cpumask_var(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL);
	}
J
Jeremy Fitzhardinge 已提交
317 318 319 320 321
	set_cpu_sibling_map(0);

	if (xen_smp_intr_init(0))
		BUG();

322 323 324 325
	if (!alloc_cpumask_var(&xen_cpu_initialized_map, GFP_KERNEL))
		panic("could not allocate xen_cpu_initialized_map\n");

	cpumask_copy(xen_cpu_initialized_map, cpumask_of(0));
J
Jeremy Fitzhardinge 已提交
326 327 328

	/* Restrict the possible_map according to max_cpus. */
	while ((num_possible_cpus() > 1) && (num_possible_cpus() > max_cpus)) {
329
		for (cpu = nr_cpu_ids - 1; !cpu_possible(cpu); cpu--)
J
Jeremy Fitzhardinge 已提交
330
			continue;
331
		set_cpu_possible(cpu, false);
J
Jeremy Fitzhardinge 已提交
332 333
	}

334
	for_each_possible_cpu(cpu)
335
		set_cpu_present(cpu, true);
J
Jeremy Fitzhardinge 已提交
336 337
}

338
static int
J
Jeremy Fitzhardinge 已提交
339 340 341
cpu_initialize_context(unsigned int cpu, struct task_struct *idle)
{
	struct vcpu_guest_context *ctxt;
342
	struct desc_struct *gdt;
343
	unsigned long gdt_mfn;
J
Jeremy Fitzhardinge 已提交
344

345
	if (cpumask_test_and_set_cpu(cpu, xen_cpu_initialized_map))
J
Jeremy Fitzhardinge 已提交
346 347 348 349 350 351
		return 0;

	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
	if (ctxt == NULL)
		return -ENOMEM;

352 353
	gdt = get_cpu_gdt_table(cpu);

J
Jeremy Fitzhardinge 已提交
354 355
	ctxt->flags = VGCF_IN_KERNEL;
	ctxt->user_regs.ss = __KERNEL_DS;
356 357
#ifdef CONFIG_X86_32
	ctxt->user_regs.fs = __KERNEL_PERCPU;
358
	ctxt->user_regs.gs = __KERNEL_STACK_CANARY;
359 360
#else
	ctxt->gs_base_kernel = per_cpu_offset(cpu);
361
#endif
J
Jeremy Fitzhardinge 已提交
362 363 364 365
	ctxt->user_regs.eip = (unsigned long)cpu_bringup_and_idle;

	memset(&ctxt->fpu_ctxt, 0, sizeof(ctxt->fpu_ctxt));

366 367 368 369
	{
		ctxt->user_regs.eflags = 0x1000; /* IOPL_RING1 */
		ctxt->user_regs.ds = __USER_DS;
		ctxt->user_regs.es = __USER_DS;
J
Jeremy Fitzhardinge 已提交
370

371
		xen_copy_trap_info(ctxt->trap_ctxt);
J
Jeremy Fitzhardinge 已提交
372

373
		ctxt->ldt_ents = 0;
374

375
		BUG_ON((unsigned long)gdt & ~PAGE_MASK);
J
Jeremy Fitzhardinge 已提交
376

377 378 379
		gdt_mfn = arbitrary_virt_to_mfn(gdt);
		make_lowmem_page_readonly(gdt);
		make_lowmem_page_readonly(mfn_to_virt(gdt_mfn));
J
Jeremy Fitzhardinge 已提交
380

381 382
		ctxt->gdt_frames[0] = gdt_mfn;
		ctxt->gdt_ents      = GDT_ENTRIES;
J
Jeremy Fitzhardinge 已提交
383

384 385
		ctxt->kernel_ss = __KERNEL_DS;
		ctxt->kernel_sp = idle->thread.sp0;
J
Jeremy Fitzhardinge 已提交
386

387
#ifdef CONFIG_X86_32
388 389
		ctxt->event_callback_cs     = __KERNEL_CS;
		ctxt->failsafe_callback_cs  = __KERNEL_CS;
390
#endif
391 392 393 394 395 396 397
		ctxt->event_callback_eip    =
					(unsigned long)xen_hypervisor_callback;
		ctxt->failsafe_callback_eip =
					(unsigned long)xen_failsafe_callback;
	}
	ctxt->user_regs.cs = __KERNEL_CS;
	ctxt->user_regs.esp = idle->thread.sp0 - sizeof(struct pt_regs);
J
Jeremy Fitzhardinge 已提交
398 399 400 401 402 403 404 405 406 407 408

	per_cpu(xen_cr3, cpu) = __pa(swapper_pg_dir);
	ctxt->ctrlreg[3] = xen_pfn_to_cr3(virt_to_mfn(swapper_pg_dir));

	if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt))
		BUG();

	kfree(ctxt);
	return 0;
}

409
static int xen_cpu_up(unsigned int cpu, struct task_struct *idle)
J
Jeremy Fitzhardinge 已提交
410 411 412
{
	int rc;

413
	per_cpu(current_task, cpu) = idle;
414
#ifdef CONFIG_X86_32
J
Jeremy Fitzhardinge 已提交
415
	irq_ctx_init(cpu);
416 417
#else
	clear_tsk_thread_flag(idle, TIF_FORK);
418 419 420
	per_cpu(kernel_stack, cpu) =
		(unsigned long)task_stack_page(idle) -
		KERNEL_STACK_OFFSET + THREAD_SIZE;
421
#endif
422
	xen_setup_runstate_info(cpu);
J
Jeremy Fitzhardinge 已提交
423
	xen_setup_timer(cpu);
424
	xen_init_lock_cpu(cpu);
J
Jeremy Fitzhardinge 已提交
425

426 427
	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;

J
Jeremy Fitzhardinge 已提交
428 429 430 431 432 433 434 435
	/* make sure interrupts start blocked */
	per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1;

	rc = cpu_initialize_context(cpu, idle);
	if (rc)
		return rc;

	if (num_online_cpus() == 1)
436 437
		/* Just in case we booted with a single CPU. */
		alternatives_enable_smp();
J
Jeremy Fitzhardinge 已提交
438 439 440 441 442 443 444 445

	rc = xen_smp_intr_init(cpu);
	if (rc)
		return rc;

	rc = HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL);
	BUG_ON(rc);

446
	while(per_cpu(cpu_state, cpu) != CPU_ONLINE) {
H
Hannes Eder 已提交
447
		HYPERVISOR_sched_op(SCHEDOP_yield, NULL);
448 449 450
		barrier();
	}

J
Jeremy Fitzhardinge 已提交
451 452 453
	return 0;
}

454
static void xen_smp_cpus_done(unsigned int max_cpus)
J
Jeremy Fitzhardinge 已提交
455 456 457
{
}

458
#ifdef CONFIG_HOTPLUG_CPU
459
static int xen_cpu_disable(void)
A
Alex Nixon 已提交
460 461 462 463 464 465 466 467 468 469 470
{
	unsigned int cpu = smp_processor_id();
	if (cpu == 0)
		return -EBUSY;

	cpu_disable_common();

	load_cr3(swapper_pg_dir);
	return 0;
}

471
static void xen_cpu_die(unsigned int cpu)
A
Alex Nixon 已提交
472
{
473
	while (xen_pv_domain() && HYPERVISOR_vcpu_op(VCPUOP_is_up, cpu, NULL)) {
A
Alex Nixon 已提交
474 475 476
		current->state = TASK_UNINTERRUPTIBLE;
		schedule_timeout(HZ/10);
	}
477
	xen_smp_intr_free(cpu);
A
Alex Nixon 已提交
478 479 480 481
	xen_uninit_lock_cpu(cpu);
	xen_teardown_timer(cpu);
}

482
static void xen_play_dead(void) /* used only with HOTPLUG_CPU */
A
Alex Nixon 已提交
483 484 485 486
{
	play_dead_common();
	HYPERVISOR_vcpu_op(VCPUOP_down, smp_processor_id(), NULL);
	cpu_bringup();
487 488 489 490 491 492 493
	/*
	 * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu down)
	 * clears certain data that the cpu_idle loop (which called us
	 * and that we return from) expects. The only way to get that
	 * data back is to call:
	 */
	tick_nohz_idle_enter();
A
Alex Nixon 已提交
494 495
}

496
#else /* !CONFIG_HOTPLUG_CPU */
497
static int xen_cpu_disable(void)
498 499 500 501
{
	return -ENOSYS;
}

502
static void xen_cpu_die(unsigned int cpu)
503 504 505 506
{
	BUG();
}

507
static void xen_play_dead(void)
508 509 510 511 512
{
	BUG();
}

#endif
J
Jeremy Fitzhardinge 已提交
513 514 515 516 517 518 519 520
static void stop_self(void *v)
{
	int cpu = smp_processor_id();

	/* make sure we're not pinning something down */
	load_cr3(swapper_pg_dir);
	/* should set up a minimal gdt */

521 522
	set_cpu_online(cpu, false);

J
Jeremy Fitzhardinge 已提交
523 524 525 526
	HYPERVISOR_vcpu_op(VCPUOP_down, cpu, NULL);
	BUG();
}

527
static void xen_stop_other_cpus(int wait)
J
Jeremy Fitzhardinge 已提交
528
{
529
	smp_call_function(stop_self, NULL, wait);
J
Jeremy Fitzhardinge 已提交
530 531
}

532
static void xen_smp_send_reschedule(int cpu)
J
Jeremy Fitzhardinge 已提交
533 534 535 536
{
	xen_send_IPI_one(cpu, XEN_RESCHEDULE_VECTOR);
}

B
Ben Guthro 已提交
537 538
static void __xen_send_IPI_mask(const struct cpumask *mask,
			      int vector)
J
Jeremy Fitzhardinge 已提交
539 540 541
{
	unsigned cpu;

542
	for_each_cpu_and(cpu, mask, cpu_online_mask)
J
Jeremy Fitzhardinge 已提交
543 544 545
		xen_send_IPI_one(cpu, vector);
}

546
static void xen_smp_send_call_function_ipi(const struct cpumask *mask)
547 548 549
{
	int cpu;

B
Ben Guthro 已提交
550
	__xen_send_IPI_mask(mask, XEN_CALL_FUNCTION_VECTOR);
551 552

	/* Make sure other vcpus get a chance to run if they need to. */
553
	for_each_cpu(cpu, mask) {
554
		if (xen_vcpu_stolen(cpu)) {
H
Hannes Eder 已提交
555
			HYPERVISOR_sched_op(SCHEDOP_yield, NULL);
556 557 558 559 560
			break;
		}
	}
}

561
static void xen_smp_send_call_function_single_ipi(int cpu)
562
{
B
Ben Guthro 已提交
563
	__xen_send_IPI_mask(cpumask_of(cpu),
564
			  XEN_CALL_FUNCTION_SINGLE_VECTOR);
565 566
}

B
Ben Guthro 已提交
567 568 569 570 571 572 573 574 575 576 577 578 579 580
static inline int xen_map_vector(int vector)
{
	int xen_vector;

	switch (vector) {
	case RESCHEDULE_VECTOR:
		xen_vector = XEN_RESCHEDULE_VECTOR;
		break;
	case CALL_FUNCTION_VECTOR:
		xen_vector = XEN_CALL_FUNCTION_VECTOR;
		break;
	case CALL_FUNCTION_SINGLE_VECTOR:
		xen_vector = XEN_CALL_FUNCTION_SINGLE_VECTOR;
		break;
581 582 583
	case IRQ_WORK_VECTOR:
		xen_vector = XEN_IRQ_WORK_VECTOR;
		break;
584 585 586 587 588 589
#ifdef CONFIG_X86_64
	case NMI_VECTOR:
	case APIC_DM_NMI: /* Some use that instead of NMI_VECTOR */
		xen_vector = XEN_NMI_VECTOR;
		break;
#endif
B
Ben Guthro 已提交
590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
	default:
		xen_vector = -1;
		printk(KERN_ERR "xen: vector 0x%x is not implemented\n",
			vector);
	}

	return xen_vector;
}

void xen_send_IPI_mask(const struct cpumask *mask,
			      int vector)
{
	int xen_vector = xen_map_vector(vector);

	if (xen_vector >= 0)
		__xen_send_IPI_mask(mask, xen_vector);
}

void xen_send_IPI_all(int vector)
{
	int xen_vector = xen_map_vector(vector);

	if (xen_vector >= 0)
		__xen_send_IPI_mask(cpu_online_mask, xen_vector);
}

void xen_send_IPI_self(int vector)
{
	int xen_vector = xen_map_vector(vector);

	if (xen_vector >= 0)
		xen_send_IPI_one(smp_processor_id(), xen_vector);
}

void xen_send_IPI_mask_allbutself(const struct cpumask *mask,
				int vector)
{
	unsigned cpu;
	unsigned int this_cpu = smp_processor_id();
S
Stefan Bader 已提交
629
	int xen_vector = xen_map_vector(vector);
B
Ben Guthro 已提交
630

S
Stefan Bader 已提交
631
	if (!(num_online_cpus() > 1) || (xen_vector < 0))
B
Ben Guthro 已提交
632 633 634 635 636 637
		return;

	for_each_cpu_and(cpu, mask, cpu_online_mask) {
		if (this_cpu == cpu)
			continue;

S
Stefan Bader 已提交
638
		xen_send_IPI_one(cpu, xen_vector);
B
Ben Guthro 已提交
639 640 641 642 643
	}
}

void xen_send_IPI_allbutself(int vector)
{
S
Stefan Bader 已提交
644
	xen_send_IPI_mask_allbutself(cpu_online_mask, vector);
B
Ben Guthro 已提交
645 646
}

J
Jeremy Fitzhardinge 已提交
647 648 649
static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id)
{
	irq_enter();
650
	generic_smp_call_function_interrupt();
651
	inc_irq_stat(irq_call_count);
J
Jeremy Fitzhardinge 已提交
652 653 654 655 656
	irq_exit();

	return IRQ_HANDLED;
}

657
static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id)
J
Jeremy Fitzhardinge 已提交
658
{
659 660
	irq_enter();
	generic_smp_call_function_single_interrupt();
661
	inc_irq_stat(irq_call_count);
662
	irq_exit();
J
Jeremy Fitzhardinge 已提交
663

664
	return IRQ_HANDLED;
J
Jeremy Fitzhardinge 已提交
665
}
666

667 668 669 670 671 672 673 674 675 676
static irqreturn_t xen_irq_work_interrupt(int irq, void *dev_id)
{
	irq_enter();
	irq_work_run();
	inc_irq_stat(apic_irq_work_irqs);
	irq_exit();

	return IRQ_HANDLED;
}

677
static const struct smp_ops xen_smp_ops __initconst = {
678 679 680 681
	.smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu,
	.smp_prepare_cpus = xen_smp_prepare_cpus,
	.smp_cpus_done = xen_smp_cpus_done,

A
Alex Nixon 已提交
682 683 684 685 686
	.cpu_up = xen_cpu_up,
	.cpu_die = xen_cpu_die,
	.cpu_disable = xen_cpu_disable,
	.play_dead = xen_play_dead,

687
	.stop_other_cpus = xen_stop_other_cpus,
688 689 690 691 692 693 694 695 696
	.smp_send_reschedule = xen_smp_send_reschedule,

	.send_call_func_ipi = xen_smp_send_call_function_ipi,
	.send_call_func_single_ipi = xen_smp_send_call_function_single_ipi,
};

void __init xen_smp_init(void)
{
	smp_ops = xen_smp_ops;
697
	xen_fill_possible_map();
698
}
699 700 701 702 703 704 705 706 707

static void __init xen_hvm_smp_prepare_cpus(unsigned int max_cpus)
{
	native_smp_prepare_cpus(max_cpus);
	WARN_ON(xen_smp_intr_init(0));

	xen_init_lock_cpu(0);
}

708
static int xen_hvm_cpu_up(unsigned int cpu, struct task_struct *tidle)
709 710
{
	int rc;
711 712 713 714 715 716 717 718 719
	/*
	 * xen_smp_intr_init() needs to run before native_cpu_up()
	 * so that IPI vectors are set up on the booting CPU before
	 * it is marked online in native_cpu_up().
	*/
	rc = xen_smp_intr_init(cpu);
	WARN_ON(rc);
	if (!rc)
		rc =  native_cpu_up(cpu, tidle);
720 721 722 723 724 725 726 727 728

	/*
	 * We must initialize the slowpath CPU kicker _after_ the native
	 * path has executed. If we initialized it before none of the
	 * unlocker IPI kicks would reach the booting CPU as the booting
	 * CPU had not set itself 'online' in cpu_online_mask. That mask
	 * is checked when IPIs are sent (on HVM at least).
	 */
	xen_init_lock_cpu(cpu);
729 730 731 732 733
	return rc;
}

static void xen_hvm_cpu_die(unsigned int cpu)
{
734
	xen_cpu_die(cpu);
735 736 737 738 739
	native_cpu_die(cpu);
}

void __init xen_hvm_smp_init(void)
{
740 741
	if (!xen_have_vector_callback)
		return;
742 743 744 745 746 747
	smp_ops.smp_prepare_cpus = xen_hvm_smp_prepare_cpus;
	smp_ops.smp_send_reschedule = xen_smp_send_reschedule;
	smp_ops.cpu_up = xen_hvm_cpu_up;
	smp_ops.cpu_die = xen_hvm_cpu_die;
	smp_ops.send_call_func_ipi = xen_smp_send_call_function_ipi;
	smp_ops.send_call_func_single_ipi = xen_smp_send_call_function_single_ipi;
748
	smp_ops.smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu;
749
}