fault.c 16.8 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
11
12
/*
 *  arch/s390/mm/fault.c
 *
 *  S390 version
 *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
 *    Author(s): Hartmut Penner (hp@de.ibm.com)
 *               Ulrich Weigand (uweigand@de.ibm.com)
 *
 *  Derived from "arch/i386/mm/fault.c"
 *    Copyright (C) 1995  Linus Torvalds
 */

13
#include <linux/kernel_stat.h>
14
#include <linux/perf_event.h>
Linus Torvalds's avatar
Linus Torvalds committed
15
16
17
18
19
20
21
22
23
#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/types.h>
#include <linux/ptrace.h>
#include <linux/mman.h>
#include <linux/mm.h>
24
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
25
#include <linux/smp.h>
26
#include <linux/kdebug.h>
Linus Torvalds's avatar
Linus Torvalds committed
27
28
29
30
#include <linux/init.h>
#include <linux/console.h>
#include <linux/module.h>
#include <linux/hardirq.h>
Michael Grundy's avatar
Michael Grundy committed
31
#include <linux/kprobes.h>
32
#include <linux/uaccess.h>
33
#include <linux/hugetlb.h>
34
#include <asm/asm-offsets.h>
Linus Torvalds's avatar
Linus Torvalds committed
35
36
#include <asm/system.h>
#include <asm/pgtable.h>
37
#include <asm/irq.h>
38
#include <asm/mmu_context.h>
39
#include <asm/compat.h>
40
#include "../kernel/entry.h"
Linus Torvalds's avatar
Linus Torvalds committed
41

42
#ifndef CONFIG_64BIT
Linus Torvalds's avatar
Linus Torvalds committed
43
44
45
#define __FAIL_ADDR_MASK 0x7ffff000
#define __SUBCODE_MASK 0x0200
#define __PF_RES_FIELD 0ULL
46
#else /* CONFIG_64BIT */
Linus Torvalds's avatar
Linus Torvalds committed
47
48
49
#define __FAIL_ADDR_MASK -4096L
#define __SUBCODE_MASK 0x0600
#define __PF_RES_FIELD 0x8000000000000000ULL
50
#endif /* CONFIG_64BIT */
Linus Torvalds's avatar
Linus Torvalds committed
51

52
53
54
55
#define VM_FAULT_BADCONTEXT	0x010000
#define VM_FAULT_BADMAP		0x020000
#define VM_FAULT_BADACCESS	0x040000

56
57
58
59
static unsigned long store_indication;

void fault_init(void)
{
60
	if (test_facility(2) && test_facility(75))
61
62
63
		store_indication = 0xc00;
}

64
static inline int notify_page_fault(struct pt_regs *regs)
65
{
66
67
68
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
69
	if (kprobes_built_in() && !user_mode(regs)) {
70
71
72
73
74
75
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
	return ret;
Michael Grundy's avatar
Michael Grundy committed
76
77
}

Linus Torvalds's avatar
Linus Torvalds committed
78
79
80

/*
 * Unlock any spinlocks which will prevent us from getting the
81
 * message out.
Linus Torvalds's avatar
Linus Torvalds committed
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
 */
void bust_spinlocks(int yes)
{
	if (yes) {
		oops_in_progress = 1;
	} else {
		int loglevel_save = console_loglevel;
		console_unblank();
		oops_in_progress = 0;
		/*
		 * OK, the message is on the console.  Now we call printk()
		 * without oops_in_progress set so that printk will give klogd
		 * a poke.  Hold onto your hats...
		 */
		console_loglevel = 15;
		printk(" ");
		console_loglevel = loglevel_save;
	}
}

/*
103
 * Returns the address space associated with the fault.
104
 * Returns 0 for kernel space and 1 for user space.
Linus Torvalds's avatar
Linus Torvalds committed
105
 */
106
static inline int user_space_fault(unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
107
108
{
	/*
109
110
	 * The lowest two bits of the translation exception
	 * identification indicate which paging table was used.
Linus Torvalds's avatar
Linus Torvalds committed
111
	 */
112
113
114
115
	trans_exc_code &= 3;
	if (trans_exc_code == 2)
		/* Access via secondary space, set_fs setting decides */
		return current->thread.mm_segment.ar4;
116
	if (user_mode == HOME_SPACE_MODE)
117
118
119
120
121
122
123
124
125
		/* User space if the access has been done via home space. */
		return trans_exc_code == 3;
	/*
	 * If the user space is not the home space the kernel runs in home
	 * space. Access via secondary space has already been covered,
	 * access via primary space or access register is from user space
	 * and access via home space is from the kernel.
	 */
	return trans_exc_code != 3;
Linus Torvalds's avatar
Linus Torvalds committed
126
127
}

128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
static inline void report_user_fault(struct pt_regs *regs, long int_code,
				     int signr, unsigned long address)
{
	if ((task_pid_nr(current) > 1) && !show_unhandled_signals)
		return;
	if (!unhandled_signal(current, signr))
		return;
	if (!printk_ratelimit())
		return;
	printk("User process fault: interruption code 0x%lX ", int_code);
	print_vma_addr(KERN_CONT "in ", regs->psw.addr & PSW_ADDR_INSN);
	printk("\n");
	printk("failing address: %lX\n", address);
	show_regs(regs);
}

Linus Torvalds's avatar
Linus Torvalds committed
144
145
146
147
/*
 * Send SIGSEGV to task.  This is an external routine
 * to keep the stack usage of do_page_fault small.
 */
148
149
static noinline void do_sigsegv(struct pt_regs *regs, long int_code,
				int si_code, unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
150
151
{
	struct siginfo si;
152
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
153

154
155
156
	address = trans_exc_code & __FAIL_ADDR_MASK;
	current->thread.prot_addr = address;
	current->thread.trap_no = int_code;
157
	report_user_fault(regs, int_code, SIGSEGV, address);
Linus Torvalds's avatar
Linus Torvalds committed
158
159
	si.si_signo = SIGSEGV;
	si.si_code = si_code;
Heiko Carstens's avatar
Heiko Carstens committed
160
	si.si_addr = (void __user *) address;
Linus Torvalds's avatar
Linus Torvalds committed
161
162
163
	force_sig_info(SIGSEGV, &si, current);
}

164
165
static noinline void do_no_context(struct pt_regs *regs, long int_code,
				   unsigned long trans_exc_code)
166
167
{
	const struct exception_table_entry *fixup;
168
	unsigned long address;
169
170

	/* Are we prepared to handle this kernel fault?  */
171
	fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
172
173
174
175
176
177
178
179
180
	if (fixup) {
		regs->psw.addr = fixup->fixup | PSW_ADDR_AMODE;
		return;
	}

	/*
	 * Oops. The kernel tried to access some bad page. We'll have to
	 * terminate things with extreme prejudice.
	 */
181
	address = trans_exc_code & __FAIL_ADDR_MASK;
182
	if (!user_space_fault(trans_exc_code))
183
184
185
186
187
188
		printk(KERN_ALERT "Unable to handle kernel pointer dereference"
		       " at virtual kernel address %p\n", (void *)address);
	else
		printk(KERN_ALERT "Unable to handle kernel paging request"
		       " at virtual user address %p\n", (void *)address);

189
	die("Oops", regs, int_code);
190
191
192
	do_exit(SIGKILL);
}

193
194
static noinline void do_low_address(struct pt_regs *regs, long int_code,
				    unsigned long trans_exc_code)
195
196
197
198
199
{
	/* Low-address protection hit in kernel mode means
	   NULL pointer write access in kernel mode.  */
	if (regs->psw.mask & PSW_MASK_PSTATE) {
		/* Low-address protection hit in user mode 'cannot happen'. */
200
		die ("Low-address protection", regs, int_code);
201
202
203
		do_exit(SIGKILL);
	}

204
	do_no_context(regs, int_code, trans_exc_code);
205
206
}

207
208
static noinline void do_sigbus(struct pt_regs *regs, long int_code,
			       unsigned long trans_exc_code)
209
210
{
	struct task_struct *tsk = current;
211
212
	unsigned long address;
	struct siginfo si;
213
214
215
216
217

	/*
	 * Send a sigbus, regardless of whether we were in kernel
	 * or user mode.
	 */
218
219
	address = trans_exc_code & __FAIL_ADDR_MASK;
	tsk->thread.prot_addr = address;
220
	tsk->thread.trap_no = int_code;
221
222
223
224
225
	si.si_signo = SIGBUS;
	si.si_errno = 0;
	si.si_code = BUS_ADRERR;
	si.si_addr = (void __user *) address;
	force_sig_info(SIGBUS, &si, tsk);
226
227
}

228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
static noinline void do_fault_error(struct pt_regs *regs, long int_code,
				    unsigned long trans_exc_code, int fault)
{
	int si_code;

	switch (fault) {
	case VM_FAULT_BADACCESS:
	case VM_FAULT_BADMAP:
		/* Bad memory access. Check if it is kernel or user space. */
		if (regs->psw.mask & PSW_MASK_PSTATE) {
			/* User mode accesses just cause a SIGSEGV */
			si_code = (fault == VM_FAULT_BADMAP) ?
				SEGV_MAPERR : SEGV_ACCERR;
			do_sigsegv(regs, int_code, si_code, trans_exc_code);
			return;
		}
	case VM_FAULT_BADCONTEXT:
		do_no_context(regs, int_code, trans_exc_code);
		break;
	default: /* fault & VM_FAULT_ERROR */
248
249
250
251
252
253
		if (fault & VM_FAULT_OOM) {
			if (!(regs->psw.mask & PSW_MASK_PSTATE))
				do_no_context(regs, int_code, trans_exc_code);
			else
				pagefault_out_of_memory();
		} else if (fault & VM_FAULT_SIGBUS) {
254
255
256
			/* Kernel mode? Handle exceptions or die */
			if (!(regs->psw.mask & PSW_MASK_PSTATE))
				do_no_context(regs, int_code, trans_exc_code);
257
258
			else
				do_sigbus(regs, int_code, trans_exc_code);
259
260
261
262
263
264
		} else
			BUG();
		break;
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
265
266
267
268
269
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 *
270
 * interruption code (int_code):
Linus Torvalds's avatar
Linus Torvalds committed
271
272
273
274
275
 *   04       Protection           ->  Write-Protection  (suprression)
 *   10       Segment translation  ->  Not present       (nullification)
 *   11       Page translation     ->  Not present       (nullification)
 *   3b       Region third trans.  ->  Not present       (nullification)
 */
276
static inline int do_exception(struct pt_regs *regs, int access,
277
			       unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
278
{
279
280
281
282
	struct task_struct *tsk;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
	unsigned long address;
283
284
	unsigned int flags;
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
285

286
	if (notify_page_fault(regs))
287
		return 0;
Michael Grundy's avatar
Michael Grundy committed
288

289
290
	tsk = current;
	mm = tsk->mm;
Linus Torvalds's avatar
Linus Torvalds committed
291
292
293
294
295
296

	/*
	 * Verify that the fault happened in user space, that
	 * we are not in an interrupt and that there is a 
	 * user context.
	 */
297
	fault = VM_FAULT_BADCONTEXT;
298
	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
299
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
300

301
	address = trans_exc_code & __FAIL_ADDR_MASK;
302
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
303
304
305
	flags = FAULT_FLAG_ALLOW_RETRY;
	if (access == VM_WRITE || (trans_exc_code & store_indication) == 0x400)
		flags |= FAULT_FLAG_WRITE;
306
	down_read(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
307

308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
#ifdef CONFIG_PGSTE
	if (test_tsk_thread_flag(current, TIF_SIE) && S390_lowcore.gmap) {
		address = gmap_fault(address,
				     (struct gmap *) S390_lowcore.gmap);
		if (address == -EFAULT) {
			fault = VM_FAULT_BADMAP;
			goto out_up;
		}
		if (address == -ENOMEM) {
			fault = VM_FAULT_OOM;
			goto out_up;
		}
	}
#endif

retry:
324
	fault = VM_FAULT_BADMAP;
325
326
	vma = find_vma(mm, address);
	if (!vma)
327
		goto out_up;
Gerald Schaefer's avatar
Gerald Schaefer committed
328

329
330
331
332
333
334
335
336
337
338
339
340
	if (unlikely(vma->vm_start > address)) {
		if (!(vma->vm_flags & VM_GROWSDOWN))
			goto out_up;
		if (expand_stack(vma, address))
			goto out_up;
	}

	/*
	 * Ok, we have a good vm_area for this memory access, so
	 * we can handle it..
	 */
	fault = VM_FAULT_BADACCESS;
341
	if (unlikely(!(vma->vm_flags & access)))
342
		goto out_up;
Linus Torvalds's avatar
Linus Torvalds committed
343

344
345
	if (is_vm_hugetlb_page(vma))
		address &= HPAGE_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
346
347
348
349
350
	/*
	 * If for any reason at all we couldn't handle the fault,
	 * make sure we exit gracefully rather than endlessly redo
	 * the fault.
	 */
351
	fault = handle_mm_fault(mm, vma, address, flags);
352
353
354
	if (unlikely(fault & VM_FAULT_ERROR))
		goto out_up;

355
356
357
358
359
360
361
362
	/*
	 * Major/minor page fault accounting is only done on the
	 * initial attempt. If we go through a retry, it is extremely
	 * likely that the page will be found in page cache at that point.
	 */
	if (flags & FAULT_FLAG_ALLOW_RETRY) {
		if (fault & VM_FAULT_MAJOR) {
			tsk->maj_flt++;
363
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1,
364
365
366
				      regs, address);
		} else {
			tsk->min_flt++;
367
			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1,
368
369
370
371
372
373
				      regs, address);
		}
		if (fault & VM_FAULT_RETRY) {
			/* Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
			 * of starvation. */
			flags &= ~FAULT_FLAG_ALLOW_RETRY;
374
			down_read(&mm->mmap_sem);
375
376
			goto retry;
		}
377
	}
Linus Torvalds's avatar
Linus Torvalds committed
378
379
380
381
	/*
	 * The instruction that caused the program check will
	 * be repeated. Don't signal single step via SIGTRAP.
	 */
Martin Schwidefsky's avatar
Martin Schwidefsky committed
382
	clear_tsk_thread_flag(tsk, TIF_PER_TRAP);
383
384
	fault = 0;
out_up:
385
	up_read(&mm->mmap_sem);
386
387
out:
	return fault;
Linus Torvalds's avatar
Linus Torvalds committed
388
389
}

390
391
void __kprobes do_protection_exception(struct pt_regs *regs, long pgm_int_code,
				       unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
392
{
393
	int fault;
394

Lucas De Marchi's avatar
Lucas De Marchi committed
395
	/* Protection exception is suppressing, decrement psw address. */
396
	regs->psw.addr = __rewind_psw(regs->psw, pgm_int_code >> 16);
397
398
399
400
401
	/*
	 * Check for low-address protection.  This needs to be treated
	 * as a special case because the translation exception code
	 * field is not guaranteed to contain valid data in this case.
	 */
402
	if (unlikely(!(trans_exc_code & 4))) {
403
		do_low_address(regs, pgm_int_code, trans_exc_code);
404
405
		return;
	}
406
	fault = do_exception(regs, VM_WRITE, trans_exc_code);
407
408
	if (unlikely(fault))
		do_fault_error(regs, 4, trans_exc_code, fault);
Linus Torvalds's avatar
Linus Torvalds committed
409
410
}

411
412
void __kprobes do_dat_exception(struct pt_regs *regs, long pgm_int_code,
				unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
413
{
414
	int access, fault;
415

416
417
	access = VM_READ | VM_EXEC | VM_WRITE;
	fault = do_exception(regs, access, trans_exc_code);
418
	if (unlikely(fault))
419
		do_fault_error(regs, pgm_int_code & 255, trans_exc_code, fault);
Linus Torvalds's avatar
Linus Torvalds committed
420
421
}

422
#ifdef CONFIG_64BIT
423
424
void __kprobes do_asce_exception(struct pt_regs *regs, long pgm_int_code,
				 unsigned long trans_exc_code)
425
{
426
	struct mm_struct *mm = current->mm;
427
428
	struct vm_area_struct *vma;

429
	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
430
431
432
		goto no_context;

	down_read(&mm->mmap_sem);
433
	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
434
435
436
437
438
439
440
441
442
	up_read(&mm->mmap_sem);

	if (vma) {
		update_mm(mm, current);
		return;
	}

	/* User mode accesses just cause a SIGSEGV */
	if (regs->psw.mask & PSW_MASK_PSTATE) {
443
		do_sigsegv(regs, pgm_int_code, SEGV_MAPERR, trans_exc_code);
444
445
446
447
		return;
	}

no_context:
448
	do_no_context(regs, pgm_int_code, trans_exc_code);
449
450
451
}
#endif

452
int __handle_fault(unsigned long uaddr, unsigned long pgm_int_code, int write)
453
454
455
456
457
458
459
460
461
462
{
	struct pt_regs regs;
	int access, fault;

	regs.psw.mask = psw_kernel_bits;
	if (!irqs_disabled())
		regs.psw.mask |= PSW_MASK_IO | PSW_MASK_EXT;
	regs.psw.addr = (unsigned long) __builtin_return_address(0);
	regs.psw.addr |= PSW_ADDR_AMODE;
	uaddr &= PAGE_MASK;
463
	access = write ? VM_WRITE : VM_READ;
464
465
	fault = do_exception(&regs, access, uaddr | 2);
	if (unlikely(fault)) {
466
467
468
		if (fault & VM_FAULT_OOM)
			return -EFAULT;
		else if (fault & VM_FAULT_SIGBUS)
469
			do_sigbus(&regs, pgm_int_code, uaddr);
470
471
472
473
	}
	return fault ? -EFAULT : 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
474
475
476
477
#ifdef CONFIG_PFAULT 
/*
 * 'pfault' pseudo page faults routines.
 */
478
static int pfault_disable;
Linus Torvalds's avatar
Linus Torvalds committed
479
480
481
482
483
484
485
486
487

static int __init nopfault(char *str)
{
	pfault_disable = 1;
	return 1;
}

__setup("nopfault", nopfault);

Heiko Carstens's avatar
Heiko Carstens committed
488
489
490
491
492
493
494
495
496
497
struct pfault_refbk {
	u16 refdiagc;
	u16 reffcode;
	u16 refdwlen;
	u16 refversn;
	u64 refgaddr;
	u64 refselmk;
	u64 refcmpmk;
	u64 reserved;
} __attribute__ ((packed, aligned(8)));
Linus Torvalds's avatar
Linus Torvalds committed
498
499
500

int pfault_init(void)
{
Heiko Carstens's avatar
Heiko Carstens committed
501
502
503
504
505
506
507
508
509
	struct pfault_refbk refbk = {
		.refdiagc = 0x258,
		.reffcode = 0,
		.refdwlen = 5,
		.refversn = 2,
		.refgaddr = __LC_CURRENT_PID,
		.refselmk = 1ULL << 48,
		.refcmpmk = 1ULL << 48,
		.reserved = __PF_RES_FIELD };
Linus Torvalds's avatar
Linus Torvalds committed
510
511
        int rc;

Heiko Carstens's avatar
Heiko Carstens committed
512
	if (!MACHINE_IS_VM || pfault_disable)
Linus Torvalds's avatar
Linus Torvalds committed
513
		return -1;
514
515
516
517
	asm volatile(
		"	diag	%1,%0,0x258\n"
		"0:	j	2f\n"
		"1:	la	%0,8\n"
Linus Torvalds's avatar
Linus Torvalds committed
518
		"2:\n"
519
520
		EX_TABLE(0b,1b)
		: "=d" (rc) : "a" (&refbk), "m" (refbk) : "cc");
Linus Torvalds's avatar
Linus Torvalds committed
521
522
523
524
525
        return rc;
}

void pfault_fini(void)
{
Heiko Carstens's avatar
Heiko Carstens committed
526
527
528
529
530
531
	struct pfault_refbk refbk = {
		.refdiagc = 0x258,
		.reffcode = 1,
		.refdwlen = 5,
		.refversn = 2,
	};
Linus Torvalds's avatar
Linus Torvalds committed
532

Heiko Carstens's avatar
Heiko Carstens committed
533
	if (!MACHINE_IS_VM || pfault_disable)
Linus Torvalds's avatar
Linus Torvalds committed
534
		return;
535
536
	asm volatile(
		"	diag	%0,0,0x258\n"
Linus Torvalds's avatar
Linus Torvalds committed
537
		"0:\n"
538
539
		EX_TABLE(0b,0b)
		: : "a" (&refbk), "m" (refbk) : "cc");
Linus Torvalds's avatar
Linus Torvalds committed
540
541
}

542
543
544
static DEFINE_SPINLOCK(pfault_lock);
static LIST_HEAD(pfault_list);

545
546
static void pfault_interrupt(unsigned int ext_int_code,
			     unsigned int param32, unsigned long param64)
Linus Torvalds's avatar
Linus Torvalds committed
547
548
549
{
	struct task_struct *tsk;
	__u16 subcode;
550
	pid_t pid;
Linus Torvalds's avatar
Linus Torvalds committed
551
552
553
554
555
556
557

	/*
	 * Get the external interruption subcode & pfault
	 * initial/completion signal bit. VM stores this 
	 * in the 'cpu address' field associated with the
         * external interrupt. 
	 */
558
	subcode = ext_int_code >> 16;
Linus Torvalds's avatar
Linus Torvalds committed
559
560
	if ((subcode & 0xff00) != __SUBCODE_MASK)
		return;
561
	kstat_cpu(smp_processor_id()).irqs[EXTINT_PFL]++;
562
563
564
565
566
567
568
569
570
571
572
573
574
575
	if (subcode & 0x0080) {
		/* Get the token (= pid of the affected task). */
		pid = sizeof(void *) == 4 ? param32 : param64;
		rcu_read_lock();
		tsk = find_task_by_pid_ns(pid, &init_pid_ns);
		if (tsk)
			get_task_struct(tsk);
		rcu_read_unlock();
		if (!tsk)
			return;
	} else {
		tsk = current;
	}
	spin_lock(&pfault_lock);
Linus Torvalds's avatar
Linus Torvalds committed
576
577
	if (subcode & 0x0080) {
		/* signal bit is set -> a page has been swapped in by VM */
578
		if (tsk->thread.pfault_wait == 1) {
Linus Torvalds's avatar
Linus Torvalds committed
579
580
581
582
			/* Initial interrupt was faster than the completion
			 * interrupt. pfault_wait is valid. Set pfault_wait
			 * back to zero and wake up the process. This can
			 * safely be done because the task is still sleeping
583
			 * and can't produce new pfaults. */
Linus Torvalds's avatar
Linus Torvalds committed
584
			tsk->thread.pfault_wait = 0;
585
			list_del(&tsk->thread.list);
Linus Torvalds's avatar
Linus Torvalds committed
586
			wake_up_process(tsk);
587
588
589
590
591
		} else {
			/* Completion interrupt was faster than initial
			 * interrupt. Set pfault_wait to -1 so the initial
			 * interrupt doesn't put the task to sleep. */
			tsk->thread.pfault_wait = -1;
Linus Torvalds's avatar
Linus Torvalds committed
592
		}
593
		put_task_struct(tsk);
Linus Torvalds's avatar
Linus Torvalds committed
594
595
	} else {
		/* signal bit not set -> a real page is missing. */
596
		if (tsk->thread.pfault_wait == -1) {
Linus Torvalds's avatar
Linus Torvalds committed
597
			/* Completion interrupt was faster than the initial
598
599
			 * interrupt (pfault_wait == -1). Set pfault_wait
			 * back to zero and exit. */
Linus Torvalds's avatar
Linus Torvalds committed
600
			tsk->thread.pfault_wait = 0;
601
602
603
604
605
606
		} else {
			/* Initial interrupt arrived before completion
			 * interrupt. Let the task sleep. */
			tsk->thread.pfault_wait = 1;
			list_add(&tsk->thread.list, &pfault_list);
			set_task_state(tsk, TASK_UNINTERRUPTIBLE);
Linus Torvalds's avatar
Linus Torvalds committed
607
			set_tsk_need_resched(tsk);
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
		}
	}
	spin_unlock(&pfault_lock);
}

static int __cpuinit pfault_cpu_notify(struct notifier_block *self,
				       unsigned long action, void *hcpu)
{
	struct thread_struct *thread, *next;
	struct task_struct *tsk;

	switch (action) {
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		spin_lock_irq(&pfault_lock);
		list_for_each_entry_safe(thread, next, &pfault_list, list) {
			thread->pfault_wait = 0;
			list_del(&thread->list);
			tsk = container_of(thread, struct task_struct, thread);
			wake_up_process(tsk);
		}
		spin_unlock_irq(&pfault_lock);
		break;
	default:
		break;
Linus Torvalds's avatar
Linus Torvalds committed
633
	}
634
	return NOTIFY_OK;
Linus Torvalds's avatar
Linus Torvalds committed
635
636
}

637
static int __init pfault_irq_init(void)
Heiko Carstens's avatar
Heiko Carstens committed
638
{
639
	int rc;
Heiko Carstens's avatar
Heiko Carstens committed
640

641
642
643
	if (!MACHINE_IS_VM)
		return 0;
	rc = register_external_interrupt(0x2603, pfault_interrupt);
Heiko Carstens's avatar
Heiko Carstens committed
644
645
646
647
648
	if (rc)
		goto out_extint;
	rc = pfault_init() == 0 ? 0 : -EOPNOTSUPP;
	if (rc)
		goto out_pfault;
649
	service_subclass_irq_register();
Heiko Carstens's avatar
Heiko Carstens committed
650
651
	hotcpu_notifier(pfault_cpu_notify, 0);
	return 0;
Heiko Carstens's avatar
Heiko Carstens committed
652

Heiko Carstens's avatar
Heiko Carstens committed
653
out_pfault:
654
	unregister_external_interrupt(0x2603, pfault_interrupt);
Heiko Carstens's avatar
Heiko Carstens committed
655
656
657
out_extint:
	pfault_disable = 1;
	return rc;
Heiko Carstens's avatar
Heiko Carstens committed
658
}
659
660
early_initcall(pfault_irq_init);

Heiko Carstens's avatar
Heiko Carstens committed
661
#endif /* CONFIG_PFAULT */