fault.c 15.4 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/perf_event.h>
Linus Torvalds's avatar
Linus Torvalds committed
14
15
16
17
18
19
20
21
22
#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>
23
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
24
#include <linux/smp.h>
25
#include <linux/kdebug.h>
Linus Torvalds's avatar
Linus Torvalds committed
26
27
28
29
#include <linux/init.h>
#include <linux/console.h>
#include <linux/module.h>
#include <linux/hardirq.h>
Michael Grundy's avatar
Michael Grundy committed
30
#include <linux/kprobes.h>
31
#include <linux/uaccess.h>
32
#include <linux/hugetlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
33
34
#include <asm/system.h>
#include <asm/pgtable.h>
Heiko Carstens's avatar
Heiko Carstens committed
35
#include <asm/s390_ext.h>
36
#include <asm/mmu_context.h>
37
#include <asm/compat.h>
38
#include "../kernel/entry.h"
Linus Torvalds's avatar
Linus Torvalds committed
39

40
#ifndef CONFIG_64BIT
Linus Torvalds's avatar
Linus Torvalds committed
41
42
43
#define __FAIL_ADDR_MASK 0x7ffff000
#define __SUBCODE_MASK 0x0200
#define __PF_RES_FIELD 0ULL
44
#else /* CONFIG_64BIT */
Linus Torvalds's avatar
Linus Torvalds committed
45
46
47
#define __FAIL_ADDR_MASK -4096L
#define __SUBCODE_MASK 0x0600
#define __PF_RES_FIELD 0x8000000000000000ULL
48
#endif /* CONFIG_64BIT */
Linus Torvalds's avatar
Linus Torvalds committed
49
50
51
52
53

#ifdef CONFIG_SYSCTL
extern int sysctl_userprocess_debug;
#endif

54
55
56
57
#define VM_FAULT_BADCONTEXT	0x010000
#define VM_FAULT_BADMAP		0x020000
#define VM_FAULT_BADACCESS	0x040000

58
static inline int notify_page_fault(struct pt_regs *regs)
59
{
60
61
	int ret = 0;

62
#ifdef CONFIG_KPROBES
63
64
65
66
67
68
69
	/* kprobe_running() needs smp_processor_id() */
	if (!user_mode(regs)) {
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}
70
#endif
71
	return ret;
Michael Grundy's avatar
Michael Grundy committed
72
73
}

Linus Torvalds's avatar
Linus Torvalds committed
74
75
76

/*
 * Unlock any spinlocks which will prevent us from getting the
77
 * message out.
Linus Torvalds's avatar
Linus Torvalds committed
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
 */
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;
	}
}

/*
99
 * Returns the address space associated with the fault.
100
 * Returns 0 for kernel space and 1 for user space.
Linus Torvalds's avatar
Linus Torvalds committed
101
 */
102
static inline int user_space_fault(unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
103
104
{
	/*
105
106
	 * The lowest two bits of the translation exception
	 * identification indicate which paging table was used.
Linus Torvalds's avatar
Linus Torvalds committed
107
	 */
108
109
110
111
	trans_exc_code &= 3;
	if (trans_exc_code == 2)
		/* Access via secondary space, set_fs setting decides */
		return current->thread.mm_segment.ar4;
112
	if (user_mode == HOME_SPACE_MODE)
113
114
115
116
117
118
119
120
121
		/* 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
122
123
124
125
126
127
}

/*
 * Send SIGSEGV to task.  This is an external routine
 * to keep the stack usage of do_page_fault small.
 */
128
129
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
130
131
{
	struct siginfo si;
132
	unsigned long address;
Linus Torvalds's avatar
Linus Torvalds committed
133

134
135
136
	address = trans_exc_code & __FAIL_ADDR_MASK;
	current->thread.prot_addr = address;
	current->thread.trap_no = int_code;
Linus Torvalds's avatar
Linus Torvalds committed
137
138
139
140
141
142
#if defined(CONFIG_SYSCTL) || defined(CONFIG_PROCESS_DEBUG)
#if defined(CONFIG_SYSCTL)
	if (sysctl_userprocess_debug)
#endif
	{
		printk("User process fault: interruption code 0x%lX\n",
143
		       int_code);
Linus Torvalds's avatar
Linus Torvalds committed
144
145
146
147
148
149
		printk("failing address: %lX\n", address);
		show_regs(regs);
	}
#endif
	si.si_signo = SIGSEGV;
	si.si_code = si_code;
Heiko Carstens's avatar
Heiko Carstens committed
150
	si.si_addr = (void __user *) address;
Linus Torvalds's avatar
Linus Torvalds committed
151
152
153
	force_sig_info(SIGSEGV, &si, current);
}

154
155
static noinline void do_no_context(struct pt_regs *regs, long int_code,
				   unsigned long trans_exc_code)
156
157
{
	const struct exception_table_entry *fixup;
158
	unsigned long address;
159
160

	/* Are we prepared to handle this kernel fault?  */
161
	fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
162
163
164
165
166
167
168
169
170
	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.
	 */
171
	address = trans_exc_code & __FAIL_ADDR_MASK;
172
	if (!user_space_fault(trans_exc_code))
173
174
175
176
177
178
		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);

179
	die("Oops", regs, int_code);
180
181
182
	do_exit(SIGKILL);
}

183
184
static noinline void do_low_address(struct pt_regs *regs, long int_code,
				    unsigned long trans_exc_code)
185
186
187
188
189
{
	/* 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'. */
190
		die ("Low-address protection", regs, int_code);
191
192
193
		do_exit(SIGKILL);
	}

194
	do_no_context(regs, int_code, trans_exc_code);
195
196
}

197
198
static noinline void do_sigbus(struct pt_regs *regs, long int_code,
			       unsigned long trans_exc_code)
199
200
201
202
203
204
205
{
	struct task_struct *tsk = current;

	/*
	 * Send a sigbus, regardless of whether we were in kernel
	 * or user mode.
	 */
206
	tsk->thread.prot_addr = trans_exc_code & __FAIL_ADDR_MASK;
207
	tsk->thread.trap_no = int_code;
208
209
210
	force_sig(SIGBUS, tsk);
}

Gerald Schaefer's avatar
Gerald Schaefer committed
211
#ifdef CONFIG_S390_EXEC_PROTECT
212
213
static noinline int signal_return(struct pt_regs *regs, long int_code,
				  unsigned long trans_exc_code)
Gerald Schaefer's avatar
Gerald Schaefer committed
214
{
215
	u16 instruction;
Heiko Carstens's avatar
Heiko Carstens committed
216
	int rc;
217
218
219

	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);

220
221
222
223
224
225
226
227
228
229
230
231
232
233
	if (!rc && instruction == 0x0a77) {
		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
		if (is_compat_task())
			sys32_sigreturn();
		else
			sys_sigreturn();
	} else if (!rc && instruction == 0x0aad) {
		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
		if (is_compat_task())
			sys32_rt_sigreturn();
		else
			sys_rt_sigreturn();
	} else
		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
Gerald Schaefer's avatar
Gerald Schaefer committed
234
235
236
237
	return 0;
}
#endif /* CONFIG_S390_EXEC_PROTECT */

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
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:
#ifdef CONFIG_S390_EXEC_PROTECT
		if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
		    (trans_exc_code & 3) == 0) {
			signal_return(regs, int_code, trans_exc_code);
			break;
		}
#endif /* CONFIG_S390_EXEC_PROTECT */
	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 */
		if (fault & VM_FAULT_OOM)
			pagefault_out_of_memory();
		else if (fault & VM_FAULT_SIGBUS) {
			do_sigbus(regs, int_code, trans_exc_code);
			/* Kernel mode? Handle exceptions or die */
			if (!(regs->psw.mask & PSW_MASK_PSTATE))
				do_no_context(regs, int_code, trans_exc_code);
		} else
			BUG();
		break;
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
278
279
280
281
282
/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 *
283
 * interruption code (int_code):
Linus Torvalds's avatar
Linus Torvalds committed
284
285
286
287
288
 *   04       Protection           ->  Write-Protection  (suprression)
 *   10       Segment translation  ->  Not present       (nullification)
 *   11       Page translation     ->  Not present       (nullification)
 *   3b       Region third trans.  ->  Not present       (nullification)
 */
289
static inline int do_exception(struct pt_regs *regs, int access,
290
			       unsigned long trans_exc_code)
Linus Torvalds's avatar
Linus Torvalds committed
291
{
292
293
294
295
	struct task_struct *tsk;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
	unsigned long address;
Nick Piggin's avatar
Nick Piggin committed
296
	int fault;
Linus Torvalds's avatar
Linus Torvalds committed
297

298
	if (notify_page_fault(regs))
299
		return 0;
Michael Grundy's avatar
Michael Grundy committed
300

301
302
	tsk = current;
	mm = tsk->mm;
Linus Torvalds's avatar
Linus Torvalds committed
303
304
305
306
307
308

	/*
	 * Verify that the fault happened in user space, that
	 * we are not in an interrupt and that there is a 
	 * user context.
	 */
309
	fault = VM_FAULT_BADCONTEXT;
310
	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
311
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
312

313
	address = trans_exc_code & __FAIL_ADDR_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
314
315
316
317
318
319
	/*
	 * When we get here, the fault happened in the current
	 * task's user address space, so we can switch on the
	 * interrupts again and then search the VMAs
	 */
	local_irq_enable();
320
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
321
	down_read(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
322

323
	fault = VM_FAULT_BADMAP;
324
325
	vma = find_vma(mm, address);
	if (!vma)
326
		goto out_up;
Gerald Schaefer's avatar
Gerald Schaefer committed
327

328
329
330
331
332
333
334
335
336
337
338
339
	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;
340
	if (unlikely(!(vma->vm_flags & access)))
341
		goto out_up;
Linus Torvalds's avatar
Linus Torvalds committed
342

343
344
	if (is_vm_hugetlb_page(vma))
		address &= HPAGE_MASK;
Linus Torvalds's avatar
Linus Torvalds committed
345
346
347
348
349
	/*
	 * If for any reason at all we couldn't handle the fault,
	 * make sure we exit gracefully rather than endlessly redo
	 * the fault.
	 */
350
351
	fault = handle_mm_fault(mm, vma, address,
				(access == VM_WRITE) ? FAULT_FLAG_WRITE : 0);
352
353
354
	if (unlikely(fault & VM_FAULT_ERROR))
		goto out_up;

355
	if (fault & VM_FAULT_MAJOR) {
Nick Piggin's avatar
Nick Piggin committed
356
		tsk->maj_flt++;
357
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
358
359
				     regs, address);
	} else {
Nick Piggin's avatar
Nick Piggin committed
360
		tsk->min_flt++;
361
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
362
363
				     regs, address);
	}
Linus Torvalds's avatar
Linus Torvalds committed
364
365
366
367
	/*
	 * The instruction that caused the program check will
	 * be repeated. Don't signal single step via SIGTRAP.
	 */
368
	clear_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
369
370
	fault = 0;
out_up:
371
	up_read(&mm->mmap_sem);
372
373
out:
	return fault;
Linus Torvalds's avatar
Linus Torvalds committed
374
375
}

376
void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
Linus Torvalds's avatar
Linus Torvalds committed
377
{
378
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
379
	int fault;
380

381
	/* Protection exception is supressing, decrement psw address. */
382
	regs->psw.addr -= (int_code >> 16);
383
384
385
386
387
	/*
	 * 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.
	 */
388
	if (unlikely(!(trans_exc_code & 4))) {
389
		do_low_address(regs, int_code, trans_exc_code);
390
391
		return;
	}
392
	fault = do_exception(regs, VM_WRITE, trans_exc_code);
393
394
	if (unlikely(fault))
		do_fault_error(regs, 4, trans_exc_code, fault);
Linus Torvalds's avatar
Linus Torvalds committed
395
396
}

397
void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
Linus Torvalds's avatar
Linus Torvalds committed
398
{
399
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
400
	int access, fault;
401

402
403
404
405
406
407
408
	access = VM_READ | VM_EXEC | VM_WRITE;
#ifdef CONFIG_S390_EXEC_PROTECT
	if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
	    (trans_exc_code & 3) == 0)
		access = VM_EXEC;
#endif
	fault = do_exception(regs, access, trans_exc_code);
409
410
	if (unlikely(fault))
		do_fault_error(regs, int_code & 255, trans_exc_code, fault);
Linus Torvalds's avatar
Linus Torvalds committed
411
412
}

413
#ifdef CONFIG_64BIT
414
void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
415
{
416
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
417
	struct mm_struct *mm = current->mm;
418
419
	struct vm_area_struct *vma;

420
	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
421
422
423
424
425
		goto no_context;

	local_irq_enable();

	down_read(&mm->mmap_sem);
426
	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
427
428
429
430
431
432
433
434
435
	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) {
436
		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
437
438
439
440
		return;
	}

no_context:
441
	do_no_context(regs, int_code, trans_exc_code);
442
443
444
}
#endif

445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
{
	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;
	access = write_user ? VM_WRITE : VM_READ;
	fault = do_exception(&regs, access, uaddr | 2);
	if (unlikely(fault)) {
		if (fault & VM_FAULT_OOM) {
			pagefault_out_of_memory();
			fault = 0;
		} else if (fault & VM_FAULT_SIGBUS)
			do_sigbus(&regs, int_code, uaddr);
	}
	return fault ? -EFAULT : 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
468
469
470
471
#ifdef CONFIG_PFAULT 
/*
 * 'pfault' pseudo page faults routines.
 */
Heiko Carstens's avatar
Heiko Carstens committed
472
static ext_int_info_t ext_int_pfault;
Linus Torvalds's avatar
Linus Torvalds committed
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
static int pfault_disable = 0;

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

__setup("nopfault", nopfault);

typedef struct {
	__u16 refdiagc;
	__u16 reffcode;
	__u16 refdwlen;
	__u16 refversn;
	__u64 refgaddr;
	__u64 refselmk;
	__u64 refcmpmk;
	__u64 reserved;
492
} __attribute__ ((packed, aligned(8))) pfault_refbk_t;
Linus Torvalds's avatar
Linus Torvalds committed
493
494
495
496
497
498
499
500

int pfault_init(void)
{
	pfault_refbk_t refbk =
		{ 0x258, 0, 5, 2, __LC_CURRENT, 1ULL << 48, 1ULL << 48,
		  __PF_RES_FIELD };
        int rc;

Heiko Carstens's avatar
Heiko Carstens committed
501
	if (!MACHINE_IS_VM || pfault_disable)
Linus Torvalds's avatar
Linus Torvalds committed
502
		return -1;
503
504
505
506
	asm volatile(
		"	diag	%1,%0,0x258\n"
		"0:	j	2f\n"
		"1:	la	%0,8\n"
Linus Torvalds's avatar
Linus Torvalds committed
507
		"2:\n"
508
509
		EX_TABLE(0b,1b)
		: "=d" (rc) : "a" (&refbk), "m" (refbk) : "cc");
Linus Torvalds's avatar
Linus Torvalds committed
510
511
512
513
514
515
516
517
518
        __ctl_set_bit(0, 9);
        return rc;
}

void pfault_fini(void)
{
	pfault_refbk_t refbk =
	{ 0x258, 1, 5, 2, 0ULL, 0ULL, 0ULL, 0ULL };

Heiko Carstens's avatar
Heiko Carstens committed
519
	if (!MACHINE_IS_VM || pfault_disable)
Linus Torvalds's avatar
Linus Torvalds committed
520
521
		return;
	__ctl_clear_bit(0,9);
522
523
	asm volatile(
		"	diag	%0,0,0x258\n"
Linus Torvalds's avatar
Linus Torvalds committed
524
		"0:\n"
525
526
		EX_TABLE(0b,0b)
		: : "a" (&refbk), "m" (refbk) : "cc");
Linus Torvalds's avatar
Linus Torvalds committed
527
528
}

529
static void pfault_interrupt(__u16 int_code)
Linus Torvalds's avatar
Linus Torvalds committed
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
{
	struct task_struct *tsk;
	__u16 subcode;

	/*
	 * Get the external interruption subcode & pfault
	 * initial/completion signal bit. VM stores this 
	 * in the 'cpu address' field associated with the
         * external interrupt. 
	 */
	subcode = S390_lowcore.cpu_addr;
	if ((subcode & 0xff00) != __SUBCODE_MASK)
		return;

	/*
	 * Get the token (= address of the task structure of the affected task).
	 */
	tsk = *(struct task_struct **) __LC_PFAULT_INTPARM;

	if (subcode & 0x0080) {
		/* signal bit is set -> a page has been swapped in by VM */
		if (xchg(&tsk->thread.pfault_wait, -1) != 0) {
			/* 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
556
			 * and can't produce new pfaults. */
Linus Torvalds's avatar
Linus Torvalds committed
557
558
			tsk->thread.pfault_wait = 0;
			wake_up_process(tsk);
559
			put_task_struct(tsk);
Linus Torvalds's avatar
Linus Torvalds committed
560
561
562
		}
	} else {
		/* signal bit not set -> a real page is missing. */
563
		get_task_struct(tsk);
Linus Torvalds's avatar
Linus Torvalds committed
564
565
566
567
568
569
570
571
572
		set_task_state(tsk, TASK_UNINTERRUPTIBLE);
		if (xchg(&tsk->thread.pfault_wait, 1) != 0) {
			/* Completion interrupt was faster than the initial
			 * interrupt (swapped in a -1 for pfault_wait). Set
			 * pfault_wait back to zero and exit. This can be
			 * done safely because tsk is running in kernel 
			 * mode and can't produce new pfaults. */
			tsk->thread.pfault_wait = 0;
			set_task_state(tsk, TASK_RUNNING);
573
			put_task_struct(tsk);
Linus Torvalds's avatar
Linus Torvalds committed
574
575
576
577
578
		} else
			set_tsk_need_resched(tsk);
	}
}

Heiko Carstens's avatar
Heiko Carstens committed
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
void __init pfault_irq_init(void)
{
	if (!MACHINE_IS_VM)
		return;

	/*
	 * Try to get pfault pseudo page faults going.
	 */
	if (register_early_external_interrupt(0x2603, pfault_interrupt,
					      &ext_int_pfault) != 0)
		panic("Couldn't request external interrupt 0x2603");

	if (pfault_init() == 0)
		return;

	/* Tough luck, no pfault. */
	pfault_disable = 1;
	unregister_early_external_interrupt(0x2603, pfault_interrupt,
					    &ext_int_pfault);
}
#endif