ipipe.c 14.6 KB
Newer Older
1
/* -*- linux-c -*-
2
 * linux/arch/arm64/kernel/ipipe.c
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
 *
 * Copyright (C) 2002-2005 Philippe Gerum.
 * Copyright (C) 2004 Wolfgang Grandegger (Adeos/arm port over 2.4).
 * Copyright (C) 2005 Heikki Lindholm (PowerPC 970 fixes).
 * Copyright (C) 2005 Stelian Pop.
 * Copyright (C) 2006-2008 Gilles Chanteperdrix.
 * Copyright (C) 2010 Philippe Gerum (SMP port).
 *
 * 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, Inc., 675 Mass Ave, Cambridge MA 02139,
 * USA; 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.
 *
 * Architecture-dependent I-PIPE support for ARM.
 */

#include <linux/kernel.h>
#include <linux/smp.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/bitops.h>
#include <linux/interrupt.h>
#include <linux/module.h>
#include <linux/errno.h>
#include <linux/kallsyms.h>
#include <linux/kprobes.h>
#include <linux/ipipe_trace.h>
#include <linux/irq.h>
#include <linux/irqnr.h>
#include <linux/prefetch.h>
#include <linux/cpu.h>
#include <linux/ipipe_domain.h>
#include <linux/ipipe_tickdev.h>
#include <asm/atomic.h>
#include <asm/hardirq.h>
#include <asm/io.h>
#include <asm/unistd.h>
#include <asm/mmu_context.h>
#include <asm/exception.h>

53
54
55
56
#ifndef CONFIG_IPIPE_ARM_KUSER_TSC
#include <asm/arch_timer.h>
#endif

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
static void __ipipe_do_IRQ(unsigned irq, void *cookie);

#ifdef CONFIG_IPIPE_DEBUG_INTERNAL
void (*__ipipe_mach_hrtimer_debug)(unsigned irq);
#endif

#ifdef CONFIG_SMP

struct __ipipe_vnmidata {
	void (*fn)(void *);
	void *arg;
	cpumask_t cpumask;
};

static struct __ipipe_vnmislot {
	ipipe_spinlock_t lock;
	struct __ipipe_vnmidata *data;
	ipipe_rwlock_t data_lock;
} __ipipe_vnmi __cacheline_aligned_in_smp = {
	.lock		= IPIPE_SPIN_LOCK_UNLOCKED,
	.data		= NULL,
	.data_lock	= IPIPE_RW_LOCK_UNLOCKED,
};

void __ipipe_early_core_setup(void)
{
	__ipipe_mach_init_platform();
}

void ipipe_stall_root(void)
{
	unsigned long flags;

	ipipe_root_only();
	flags = hard_smp_local_irq_save();
	__set_bit(IPIPE_STALL_FLAG, &__ipipe_root_status);
	hard_smp_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(ipipe_stall_root);

unsigned long ipipe_test_and_stall_root(void)
{
	unsigned long flags;
	int x;

	ipipe_root_only();
	flags = hard_smp_local_irq_save();
	x = __test_and_set_bit(IPIPE_STALL_FLAG, &__ipipe_root_status);
	hard_smp_local_irq_restore(flags);

	return x;
}
EXPORT_SYMBOL_GPL(ipipe_test_and_stall_root);

unsigned long ipipe_test_root(void)
{
	unsigned long flags;
	int x;

	flags = hard_smp_local_irq_save();
	x = test_bit(IPIPE_STALL_FLAG, &__ipipe_root_status);
	hard_smp_local_irq_restore(flags);

	return x;
}
EXPORT_SYMBOL_GPL(ipipe_test_root);

void __ipipe_do_vnmi(unsigned int irq, void *cookie)
{
	int cpu = ipipe_processor_id();
	struct __ipipe_vnmidata *data;

	read_lock(&__ipipe_vnmi.data_lock);

	data = __ipipe_vnmi.data;
	if (likely(data && cpumask_test_cpu(cpu, &data->cpumask))) {
		data->fn(data->arg);
		cpu_clear(cpu, data->cpumask);
	}

	read_unlock(&__ipipe_vnmi.data_lock);
}

static inline void
hook_internal_ipi(struct ipipe_domain *ipd, int virq,
		  void (*handler)(unsigned int irq, void *cookie))
{
	ipd->irqs[virq].ackfn = NULL;
	ipd->irqs[virq].handler = handler;
	ipd->irqs[virq].cookie = NULL;
	/* Immediately handle in the current domain but *never* pass */
	ipd->irqs[virq].control = IPIPE_HANDLE_MASK|IPIPE_STICKY_MASK;
}

void __ipipe_hook_critical_ipi(struct ipipe_domain *ipd)
{
	__ipipe_ipis_alloc();
	hook_internal_ipi(ipd, IPIPE_CRITICAL_IPI, __ipipe_do_critical_sync);
	hook_internal_ipi(ipd, IPIPE_SERVICE_VNMI, __ipipe_do_vnmi);
}

void ipipe_set_irq_affinity(unsigned int irq, cpumask_t cpumask)
{
	if (WARN_ON_ONCE(irq_get_chip(irq)->irq_set_affinity == NULL))
		return;

	cpus_and(cpumask, cpumask, *cpu_online_mask);
	if (WARN_ON_ONCE(cpus_empty(cpumask)))
		return;

	irq_get_chip(irq)->irq_set_affinity(irq_get_irq_data(irq), &cpumask, true);
}
EXPORT_SYMBOL_GPL(ipipe_set_irq_affinity);

void __ipipe_send_vnmi(void (*fn)(void *), cpumask_t cpumask, void *arg)
{
	struct __ipipe_vnmidata data;
	unsigned long flags;
	int cpu;

	data.fn = fn;
	data.arg = arg;
	data.cpumask = cpumask;

	while (!spin_trylock_irqsave(&__ipipe_vnmi.lock, flags)) {
		if (hard_irqs_disabled())
			__ipipe_do_vnmi(IPIPE_SERVICE_VNMI, NULL);
		cpu_relax();
	}

	cpu = ipipe_processor_id();
	cpu_clear(cpu, data.cpumask);
	if (cpus_empty(data.cpumask)) {
		spin_unlock_irqrestore(&__ipipe_vnmi.lock, flags);
		return;
	}

	write_lock(&__ipipe_vnmi.data_lock);
	__ipipe_vnmi.data = &data;
	write_unlock(&__ipipe_vnmi.data_lock);

	ipipe_send_ipi(IPIPE_SERVICE_VNMI, data.cpumask);
	while (!cpus_empty(data.cpumask))
		cpu_relax();

	write_lock(&__ipipe_vnmi.data_lock);
	__ipipe_vnmi.data = NULL;
	write_unlock(&__ipipe_vnmi.data_lock);

	spin_unlock_irqrestore(&__ipipe_vnmi.lock, flags);
}
EXPORT_SYMBOL_GPL(__ipipe_send_vnmi);
#endif	/* CONFIG_SMP */

#ifdef CONFIG_SMP_ON_UP
struct static_key __ipipe_smp_key = STATIC_KEY_INIT_TRUE;

unsigned __ipipe_processor_id(void)
{
	return raw_smp_processor_id();
}
EXPORT_SYMBOL_GPL(__ipipe_processor_id);

static int ipipe_disable_smp(void)
{
	if (num_online_cpus() == 1) {
		unsigned long flags;

		printk("I-pipe: disabling SMP code\n");

		flags = hard_local_irq_save();
		static_key_slow_dec(&__ipipe_smp_key);
		hard_local_irq_restore(flags);
	}
	return 0;
}
arch_initcall(ipipe_disable_smp);
#endif /* SMP_ON_UP */

int ipipe_get_sysinfo(struct ipipe_sysinfo *info)
{
	info->sys_nr_cpus = num_online_cpus();
	info->sys_cpu_freq = __ipipe_hrclock_freq;
	info->sys_hrtimer_irq = per_cpu(ipipe_percpu.hrtimer_irq, 0);
	info->sys_hrtimer_freq = __ipipe_hrtimer_freq;
	info->sys_hrclock_freq = __ipipe_hrclock_freq;
	__ipipe_mach_get_tscinfo(&info->arch.tsc);

	return 0;
}
EXPORT_SYMBOL_GPL(ipipe_get_sysinfo);

struct ipipe_mach_pic_muter ipipe_pic_muter;
EXPORT_SYMBOL_GPL(ipipe_pic_muter);

void ipipe_pic_muter_register(struct ipipe_mach_pic_muter *muter)
{
	ipipe_pic_muter = *muter;
}

void __ipipe_enable_irqdesc(struct ipipe_domain *ipd, unsigned irq)
{
	/* With sparse IRQs, some irqs may not have a descriptor */
	if (irq_to_desc(irq) == NULL)
		return;

	if (ipipe_pic_muter.enable_irqdesc)
		ipipe_pic_muter.enable_irqdesc(ipd, irq);
}
EXPORT_SYMBOL_GPL(__ipipe_enable_irqdesc);

void __ipipe_disable_irqdesc(struct ipipe_domain *ipd, unsigned irq)
{
	if (ipipe_pic_muter.disable_irqdesc)
		ipipe_pic_muter.disable_irqdesc(ipd, irq);
}
EXPORT_SYMBOL_GPL(__ipipe_disable_irqdesc);

/*
 * __ipipe_enable_pipeline() -- We are running on the boot CPU, hw
 * interrupts are off, and secondary CPUs are still lost in space.
 */
void __ipipe_enable_pipeline(void)
{
	unsigned long flags;
	unsigned int irq;

	flags = ipipe_critical_enter(NULL);

	/* virtualize all interrupts from the root domain. */
	for (irq = 0; irq < IPIPE_NR_ROOT_IRQS; irq++)
		ipipe_request_irq(ipipe_root_domain,
				  irq,
				  (ipipe_irq_handler_t)__ipipe_do_IRQ,
				  NULL, NULL);

#ifdef CONFIG_SMP
	__ipipe_ipis_request();
#endif /* CONFIG_SMP */

	ipipe_critical_exit(flags);
}

#ifdef CONFIG_IPIPE_DEBUG_INTERNAL
unsigned asmlinkage __ipipe_bugon_irqs_enabled(unsigned x)
{
	BUG_ON(!hard_irqs_disabled());
	return x;		/* Preserve r0 */
}
#endif

asmlinkage int __ipipe_check_root_interruptible(void)
{
	return __ipipe_root_p && !irqs_disabled();
}

__kprobes int
__ipipe_switch_to_notifier_call_chain(struct atomic_notifier_head *nh,
				      unsigned long val, void *v)
{
	unsigned long flags;
	int ret;

	local_irq_save(flags);
	ret = atomic_notifier_call_chain(nh, val, v);
	__ipipe_restore_root_nosync(flags);

	return ret;
}

#define fast_irq_disable()			\
	({					\
		hard_local_irq_disable();	\
		0;				\
	})
#define fast_irq_enable(flags)			\
	({					\
		hard_local_irq_enable();	\
		(void)(flags);			\
	})

338
339
340
341
#ifndef __NR_SYSCALL_BASE
#define __NR_SYSCALL_BASE 0
#endif

342
343
344
345
asmlinkage int __ipipe_syscall_root(unsigned long scno, struct pt_regs *regs)
{
	struct task_struct *const task = current;
	struct ipipe_percpu_domain_data *p;
346
	unsigned long orig_x8;
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
	unsigned long flags;
	int ret = 0;

#ifdef CONFIG_IPIPE_DEBUG_INTERNAL
	WARN_ON_ONCE(hard_irqs_disabled());
#endif

	/*
	 * This routine either returns:
	 * 0 -- if the syscall is to be passed to Linux;
	 * <0 -- if the syscall should not be passed to Linux, and no
	 * tail work should be performed;
	 * >0 -- if the syscall should not be passed to Linux but the
	 * tail work has to be performed (for handling signals etc).
	 */

	scno += __NR_SYSCALL_BASE;
	if (!__ipipe_syscall_watched_p(task, scno))
		goto out;

367
368
369
370
371
	/*
	 * We use x8 to pass the syscall number to the other domains.
	 */
	orig_x8 = regs->regs[8];
	regs->regs[8] = scno;
372
373
374

	ret = __ipipe_notify_syscall(regs);

375
	regs->regs[8] = orig_x8;
376
377
378
379
380
381
382

	flags = fast_irq_disable();

	/*
	 * This is the end of the syscall path, so we may
	 * safely assume a valid Linux task stack here.
	 */
383
384
	if (ipipe_test_thread_flag(TIP_MAYDAY)) {
		ipipe_clear_thread_flag(TIP_MAYDAY);
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
		__ipipe_notify_trap(IPIPE_TRAP_MAYDAY, regs);
	}

	if (!__ipipe_root_p)
		ret = -1;
	else {
		p = ipipe_this_cpu_root_context();
		if (__ipipe_ipending_p(p))
			__ipipe_sync_stage();
	}

	fast_irq_enable(flags);
out:
	return ret;
}

void __ipipe_exit_irq(struct pt_regs *regs)
{
403
404
405
406
407
	/*
	 * Testing for user_regs() eliminates foreign stack contexts,
	 * including from legacy domains which did not set the foreign
	 * stack bit (foreign stacks are always kernel-based).
	 */
408
	if (user_mode(regs) &&
409
	    ipipe_test_thread_flag(TIP_MAYDAY)) {
410
		/*
411
412
413
		 * MAYDAY is never raised under normal circumstances,
		 * so prefer test then maybe clear over
		 * test_and_clear.
414
		 */
415
		ipipe_clear_thread_flag(TIP_MAYDAY);
416
417
418
419
420
421
422
		__ipipe_notify_trap(IPIPE_TRAP_MAYDAY, regs);
	}
}

/* hw irqs off */
asmlinkage void __exception __ipipe_grab_irq(int irq, struct pt_regs *regs)
{
423
	struct ipipe_percpu_data *p = __ipipe_raw_cpu_ptr(&ipipe_percpu);
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442

	ipipe_trace_irq_entry(irq);

	if (p->hrtimer_irq == -1)
		goto copy_regs;

	if (irq == p->hrtimer_irq) {
		/*
		 * Given our deferred dispatching model for regular IRQs, we
		 * only record CPU regs for the last timer interrupt, so that
		 * the timer handler charges CPU times properly. It is assumed
		 * that other interrupt handlers don't actually care for such
		 * information.
		 */
#ifdef CONFIG_IPIPE_DEBUG_INTERNAL
		if (__ipipe_mach_hrtimer_debug)
			__ipipe_mach_hrtimer_debug(irq);
#endif /* CONFIG_IPIPE_DEBUG_INTERNAL */
	  copy_regs:
443
		p->tick_regs.pstate =
444
			(p->curr == &p->root
445
446
447
			 ? regs->pstate
			 : regs->pstate | PSR_I_BIT);
		p->tick_regs.pc = regs->pc;
448
449
450
451
452
453
454
455
456
457
458
	}

	__ipipe_dispatch_irq(irq, 0);

	ipipe_trace_irq_exit(irq);

	__ipipe_exit_irq(regs);
}

static void __ipipe_do_IRQ(unsigned irq, void *cookie)
{
459
	struct pt_regs *regs = raw_cpu_ptr(&ipipe_percpu.tick_regs);
460
	__handle_domain_irq(NULL, irq, false, regs);
461
462
463
464
465
466
467
}

#ifdef CONFIG_MMU
void __switch_mm_inner(struct mm_struct *prev, struct mm_struct *next,
		       struct task_struct *tsk)
{
	struct mm_struct ** const active_mm =
468
		raw_cpu_ptr(&ipipe_percpu.active_mm);
469
	int ret;
470
471
#ifdef CONFIG_IPIPE_WANT_PREEMPTIBLE_SWITCH
	struct thread_info *const tip = current_thread_info();
472
473
	unsigned long flags;

474
475
476
477
478
	prev = *active_mm;
	clear_bit(TIF_MMSWITCH_INT, &tip->flags);
	barrier();
	*active_mm = NULL;
	barrier();
479

480
	for (;;) {
481
		ret = __do_switch_mm(prev, next, tsk, true);
482
483
484
485
486
487
		/*
		 * Reading thread_info flags and setting active_mm
		 * must be done atomically.
		 */
		flags = hard_local_irq_save();
		if (__test_and_clear_bit(TIF_MMSWITCH_INT, &tip->flags) == 0) {
488
			*active_mm = ret < 0 ? prev : next;
489
490
491
492
493
			hard_local_irq_restore(flags);
			return;
		}
		hard_local_irq_restore(flags);

494
		if (ret < 0)
495
496
497
498
499
500
501
502
503
504
			/*
			 * We were interrupted by head domain, which
			 * may have changed the mm context, mm context
			 * is now unknown, but will be switched in
			 * deferred_switch_mm
			 */
			return;

		prev = NULL;
	}
505
#else
506
507
508
	ret = __do_switch_mm(prev, next, tsk, true);
	*active_mm = ret < 0 ? prev : next;
#endif	/* CONFIG_IPIPE_WANT_PREEMPTIBLE_SWITCH */
509
510
511
512
513
514
}

#ifdef finish_arch_post_lock_switch
void deferred_switch_mm(struct mm_struct *next)
{
	struct mm_struct ** const active_mm =
515
		raw_cpu_ptr(&ipipe_percpu.active_mm);
516
517
518
	struct mm_struct *prev = *active_mm;
#ifdef CONFIG_IPIPE_WANT_PREEMPTIBLE_SWITCH
	struct thread_info *const tip = current_thread_info();
519
520
	unsigned long flags;

521
522
523
524
	clear_bit(TIF_MMSWITCH_INT, &tip->flags);
	barrier();
	*active_mm = NULL;
	barrier();
525

526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
	for (;;) {
		__do_switch_mm(prev, next, NULL, false);
		/*
		 * Reading thread_info flags and setting active_mm
		 * must be done atomically.
		 */
		flags = hard_local_irq_save();
		if (__test_and_clear_bit(TIF_MMSWITCH_INT, &tip->flags) == 0) {
			*active_mm = next;
			hard_local_irq_restore(flags);
			return;
		}
		hard_local_irq_restore(flags);
		prev = NULL;
	}
541
#else
542
	__do_switch_mm(prev, next, NULL, false);
543
	*active_mm = next;
544
#endif	/* CONFIG_IPIPE_WANT_PREEMPTIBLE_SWITCH */
545
}
546
#endif	/* finish_arch_post_lock_switch */
547
548
#endif /* CONFIG_MMU */

549
550
551
552
553
554
#ifndef CONFIG_IPIPE_ARM_KUSER_TSC
static struct __ipipe_tscinfo tsc_info;

void __init __ipipe_tsc_register(struct __ipipe_tscinfo *info)
{
	tsc_info = *info;
555
	__ipipe_hrclock_freq = info->freq;
556
557
558
559
560
}
void __ipipe_mach_get_tscinfo(struct __ipipe_tscinfo *info)
{
	*info = tsc_info;
}
561
562
563
unsigned long long __ipipe_mach_get_tsc(void) {
	return arch_counter_get_cntvct();
}
564
565
#endif

566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
#if defined(CONFIG_IPIPE_DEBUG) && defined(CONFIG_DEBUG_LL)
void printascii(const char *s);

static IPIPE_DEFINE_SPINLOCK(serial_debug_lock);

void __ipipe_serial_debug(const char *fmt, ...)
{
	unsigned long flags;
	char buf[128];
	va_list ap;
	int n;

	va_start(ap, fmt);
	n = vsnprintf(buf, sizeof(buf) - 2, fmt, ap);
	va_end(ap);

	if (n > 0 && buf[n - 1] == '\n') {
		buf[n] = '\r';
		buf[n+1] = '\0';
	}

	spin_lock_irqsave(&serial_debug_lock, flags);
	printascii(buf);
	spin_unlock_irqrestore(&serial_debug_lock, flags);
}

#ifndef CONFIG_SERIAL_8250_CONSOLE
EXPORT_SYMBOL_GPL(__ipipe_serial_debug);
#endif

#endif

EXPORT_SYMBOL_GPL(do_munmap);
EXPORT_SYMBOL_GPL(show_stack);
EXPORT_SYMBOL_GPL(init_mm);
#ifndef MULTI_CPU
EXPORT_SYMBOL_GPL(cpu_do_switch_mm);
#endif
#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK)
EXPORT_SYMBOL_GPL(tasklist_lock);
#endif /* CONFIG_SMP || CONFIG_DEBUG_SPINLOCK */

#ifndef CONFIG_SPARSE_IRQ
EXPORT_SYMBOL_GPL(irq_desc);
#endif