xref: /linux/arch/powerpc/kernel/process.c (revision d8327c784b51b57dac2c26cfad87dce0d68dfd98)
1 /*
2  *  arch/ppc/kernel/process.c
3  *
4  *  Derived from "arch/i386/kernel/process.c"
5  *    Copyright (C) 1995  Linus Torvalds
6  *
7  *  Updated and modified by Cort Dougan (cort@cs.nmt.edu) and
8  *  Paul Mackerras (paulus@cs.anu.edu.au)
9  *
10  *  PowerPC version
11  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
12  *
13  *  This program is free software; you can redistribute it and/or
14  *  modify it under the terms of the GNU General Public License
15  *  as published by the Free Software Foundation; either version
16  *  2 of the License, or (at your option) any later version.
17  */
18 
19 #include <linux/config.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
23 #include <linux/mm.h>
24 #include <linux/smp.h>
25 #include <linux/smp_lock.h>
26 #include <linux/stddef.h>
27 #include <linux/unistd.h>
28 #include <linux/ptrace.h>
29 #include <linux/slab.h>
30 #include <linux/user.h>
31 #include <linux/elf.h>
32 #include <linux/init.h>
33 #include <linux/prctl.h>
34 #include <linux/init_task.h>
35 #include <linux/module.h>
36 #include <linux/kallsyms.h>
37 #include <linux/mqueue.h>
38 #include <linux/hardirq.h>
39 #include <linux/utsname.h>
40 #include <linux/kprobes.h>
41 
42 #include <asm/pgtable.h>
43 #include <asm/uaccess.h>
44 #include <asm/system.h>
45 #include <asm/io.h>
46 #include <asm/processor.h>
47 #include <asm/mmu.h>
48 #include <asm/prom.h>
49 #include <asm/machdep.h>
50 #ifdef CONFIG_PPC64
51 #include <asm/firmware.h>
52 #include <asm/time.h>
53 #endif
54 
55 extern unsigned long _get_SP(void);
56 
57 #ifndef CONFIG_SMP
58 struct task_struct *last_task_used_math = NULL;
59 struct task_struct *last_task_used_altivec = NULL;
60 struct task_struct *last_task_used_spe = NULL;
61 #endif
62 
63 /*
64  * Make sure the floating-point register state in the
65  * the thread_struct is up to date for task tsk.
66  */
67 void flush_fp_to_thread(struct task_struct *tsk)
68 {
69 	if (tsk->thread.regs) {
70 		/*
71 		 * We need to disable preemption here because if we didn't,
72 		 * another process could get scheduled after the regs->msr
73 		 * test but before we have finished saving the FP registers
74 		 * to the thread_struct.  That process could take over the
75 		 * FPU, and then when we get scheduled again we would store
76 		 * bogus values for the remaining FP registers.
77 		 */
78 		preempt_disable();
79 		if (tsk->thread.regs->msr & MSR_FP) {
80 #ifdef CONFIG_SMP
81 			/*
82 			 * This should only ever be called for current or
83 			 * for a stopped child process.  Since we save away
84 			 * the FP register state on context switch on SMP,
85 			 * there is something wrong if a stopped child appears
86 			 * to still have its FP state in the CPU registers.
87 			 */
88 			BUG_ON(tsk != current);
89 #endif
90 			giveup_fpu(current);
91 		}
92 		preempt_enable();
93 	}
94 }
95 
96 void enable_kernel_fp(void)
97 {
98 	WARN_ON(preemptible());
99 
100 #ifdef CONFIG_SMP
101 	if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
102 		giveup_fpu(current);
103 	else
104 		giveup_fpu(NULL);	/* just enables FP for kernel */
105 #else
106 	giveup_fpu(last_task_used_math);
107 #endif /* CONFIG_SMP */
108 }
109 EXPORT_SYMBOL(enable_kernel_fp);
110 
111 int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
112 {
113 	if (!tsk->thread.regs)
114 		return 0;
115 	flush_fp_to_thread(current);
116 
117 	memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs));
118 
119 	return 1;
120 }
121 
122 #ifdef CONFIG_ALTIVEC
123 void enable_kernel_altivec(void)
124 {
125 	WARN_ON(preemptible());
126 
127 #ifdef CONFIG_SMP
128 	if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
129 		giveup_altivec(current);
130 	else
131 		giveup_altivec(NULL);	/* just enable AltiVec for kernel - force */
132 #else
133 	giveup_altivec(last_task_used_altivec);
134 #endif /* CONFIG_SMP */
135 }
136 EXPORT_SYMBOL(enable_kernel_altivec);
137 
138 /*
139  * Make sure the VMX/Altivec register state in the
140  * the thread_struct is up to date for task tsk.
141  */
142 void flush_altivec_to_thread(struct task_struct *tsk)
143 {
144 	if (tsk->thread.regs) {
145 		preempt_disable();
146 		if (tsk->thread.regs->msr & MSR_VEC) {
147 #ifdef CONFIG_SMP
148 			BUG_ON(tsk != current);
149 #endif
150 			giveup_altivec(current);
151 		}
152 		preempt_enable();
153 	}
154 }
155 
156 int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
157 {
158 	flush_altivec_to_thread(current);
159 	memcpy(vrregs, &current->thread.vr[0], sizeof(*vrregs));
160 	return 1;
161 }
162 #endif /* CONFIG_ALTIVEC */
163 
164 #ifdef CONFIG_SPE
165 
166 void enable_kernel_spe(void)
167 {
168 	WARN_ON(preemptible());
169 
170 #ifdef CONFIG_SMP
171 	if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
172 		giveup_spe(current);
173 	else
174 		giveup_spe(NULL);	/* just enable SPE for kernel - force */
175 #else
176 	giveup_spe(last_task_used_spe);
177 #endif /* __SMP __ */
178 }
179 EXPORT_SYMBOL(enable_kernel_spe);
180 
181 void flush_spe_to_thread(struct task_struct *tsk)
182 {
183 	if (tsk->thread.regs) {
184 		preempt_disable();
185 		if (tsk->thread.regs->msr & MSR_SPE) {
186 #ifdef CONFIG_SMP
187 			BUG_ON(tsk != current);
188 #endif
189 			giveup_spe(current);
190 		}
191 		preempt_enable();
192 	}
193 }
194 
195 int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
196 {
197 	flush_spe_to_thread(current);
198 	/* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */
199 	memcpy(evrregs, &current->thread.evr[0], sizeof(u32) * 35);
200 	return 1;
201 }
202 #endif /* CONFIG_SPE */
203 
204 #ifndef CONFIG_SMP
205 /*
206  * If we are doing lazy switching of CPU state (FP, altivec or SPE),
207  * and the current task has some state, discard it.
208  */
209 void discard_lazy_cpu_state(void)
210 {
211 	preempt_disable();
212 	if (last_task_used_math == current)
213 		last_task_used_math = NULL;
214 #ifdef CONFIG_ALTIVEC
215 	if (last_task_used_altivec == current)
216 		last_task_used_altivec = NULL;
217 #endif /* CONFIG_ALTIVEC */
218 #ifdef CONFIG_SPE
219 	if (last_task_used_spe == current)
220 		last_task_used_spe = NULL;
221 #endif
222 	preempt_enable();
223 }
224 #endif /* CONFIG_SMP */
225 
226 #ifdef CONFIG_PPC_MERGE		/* XXX for now */
227 int set_dabr(unsigned long dabr)
228 {
229 	if (ppc_md.set_dabr)
230 		return ppc_md.set_dabr(dabr);
231 
232 	mtspr(SPRN_DABR, dabr);
233 	return 0;
234 }
235 #endif
236 
237 #ifdef CONFIG_PPC64
238 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
239 static DEFINE_PER_CPU(unsigned long, current_dabr);
240 #endif
241 
242 struct task_struct *__switch_to(struct task_struct *prev,
243 	struct task_struct *new)
244 {
245 	struct thread_struct *new_thread, *old_thread;
246 	unsigned long flags;
247 	struct task_struct *last;
248 
249 #ifdef CONFIG_SMP
250 	/* avoid complexity of lazy save/restore of fpu
251 	 * by just saving it every time we switch out if
252 	 * this task used the fpu during the last quantum.
253 	 *
254 	 * If it tries to use the fpu again, it'll trap and
255 	 * reload its fp regs.  So we don't have to do a restore
256 	 * every switch, just a save.
257 	 *  -- Cort
258 	 */
259 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
260 		giveup_fpu(prev);
261 #ifdef CONFIG_ALTIVEC
262 	/*
263 	 * If the previous thread used altivec in the last quantum
264 	 * (thus changing altivec regs) then save them.
265 	 * We used to check the VRSAVE register but not all apps
266 	 * set it, so we don't rely on it now (and in fact we need
267 	 * to save & restore VSCR even if VRSAVE == 0).  -- paulus
268 	 *
269 	 * On SMP we always save/restore altivec regs just to avoid the
270 	 * complexity of changing processors.
271 	 *  -- Cort
272 	 */
273 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
274 		giveup_altivec(prev);
275 #endif /* CONFIG_ALTIVEC */
276 #ifdef CONFIG_SPE
277 	/*
278 	 * If the previous thread used spe in the last quantum
279 	 * (thus changing spe regs) then save them.
280 	 *
281 	 * On SMP we always save/restore spe regs just to avoid the
282 	 * complexity of changing processors.
283 	 */
284 	if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
285 		giveup_spe(prev);
286 #endif /* CONFIG_SPE */
287 
288 #else  /* CONFIG_SMP */
289 #ifdef CONFIG_ALTIVEC
290 	/* Avoid the trap.  On smp this this never happens since
291 	 * we don't set last_task_used_altivec -- Cort
292 	 */
293 	if (new->thread.regs && last_task_used_altivec == new)
294 		new->thread.regs->msr |= MSR_VEC;
295 #endif /* CONFIG_ALTIVEC */
296 #ifdef CONFIG_SPE
297 	/* Avoid the trap.  On smp this this never happens since
298 	 * we don't set last_task_used_spe
299 	 */
300 	if (new->thread.regs && last_task_used_spe == new)
301 		new->thread.regs->msr |= MSR_SPE;
302 #endif /* CONFIG_SPE */
303 
304 #endif /* CONFIG_SMP */
305 
306 #ifdef CONFIG_PPC64	/* for now */
307 	if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
308 		set_dabr(new->thread.dabr);
309 		__get_cpu_var(current_dabr) = new->thread.dabr;
310 	}
311 
312 	flush_tlb_pending();
313 #endif
314 
315 	new_thread = &new->thread;
316 	old_thread = &current->thread;
317 
318 #ifdef CONFIG_PPC64
319 	/*
320 	 * Collect processor utilization data per process
321 	 */
322 	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
323 		struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
324 		long unsigned start_tb, current_tb;
325 		start_tb = old_thread->start_tb;
326 		cu->current_tb = current_tb = mfspr(SPRN_PURR);
327 		old_thread->accum_tb += (current_tb - start_tb);
328 		new_thread->start_tb = current_tb;
329 	}
330 #endif
331 
332 	local_irq_save(flags);
333 	last = _switch(old_thread, new_thread);
334 
335 	local_irq_restore(flags);
336 
337 	return last;
338 }
339 
340 static int instructions_to_print = 16;
341 
342 #ifdef CONFIG_PPC64
343 #define BAD_PC(pc)	((REGION_ID(pc) != KERNEL_REGION_ID) && \
344 		         (REGION_ID(pc) != VMALLOC_REGION_ID))
345 #else
346 #define BAD_PC(pc)	((pc) < KERNELBASE)
347 #endif
348 
349 static void show_instructions(struct pt_regs *regs)
350 {
351 	int i;
352 	unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
353 			sizeof(int));
354 
355 	printk("Instruction dump:");
356 
357 	for (i = 0; i < instructions_to_print; i++) {
358 		int instr;
359 
360 		if (!(i % 8))
361 			printk("\n");
362 
363 		if (BAD_PC(pc) || __get_user(instr, (unsigned int *)pc)) {
364 			printk("XXXXXXXX ");
365 		} else {
366 			if (regs->nip == pc)
367 				printk("<%08x> ", instr);
368 			else
369 				printk("%08x ", instr);
370 		}
371 
372 		pc += sizeof(int);
373 	}
374 
375 	printk("\n");
376 }
377 
378 static struct regbit {
379 	unsigned long bit;
380 	const char *name;
381 } msr_bits[] = {
382 	{MSR_EE,	"EE"},
383 	{MSR_PR,	"PR"},
384 	{MSR_FP,	"FP"},
385 	{MSR_ME,	"ME"},
386 	{MSR_IR,	"IR"},
387 	{MSR_DR,	"DR"},
388 	{0,		NULL}
389 };
390 
391 static void printbits(unsigned long val, struct regbit *bits)
392 {
393 	const char *sep = "";
394 
395 	printk("<");
396 	for (; bits->bit; ++bits)
397 		if (val & bits->bit) {
398 			printk("%s%s", sep, bits->name);
399 			sep = ",";
400 		}
401 	printk(">");
402 }
403 
404 #ifdef CONFIG_PPC64
405 #define REG		"%016lX"
406 #define REGS_PER_LINE	4
407 #define LAST_VOLATILE	13
408 #else
409 #define REG		"%08lX"
410 #define REGS_PER_LINE	8
411 #define LAST_VOLATILE	12
412 #endif
413 
414 void show_regs(struct pt_regs * regs)
415 {
416 	int i, trap;
417 
418 	printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
419 	       regs->nip, regs->link, regs->ctr);
420 	printk("REGS: %p TRAP: %04lx   %s  (%s)\n",
421 	       regs, regs->trap, print_tainted(), system_utsname.release);
422 	printk("MSR: "REG" ", regs->msr);
423 	printbits(regs->msr, msr_bits);
424 	printk("  CR: %08lX  XER: %08lX\n", regs->ccr, regs->xer);
425 	trap = TRAP(regs);
426 	if (trap == 0x300 || trap == 0x600)
427 		printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
428 	printk("TASK = %p[%d] '%s' THREAD: %p",
429 	       current, current->pid, current->comm, task_thread_info(current));
430 
431 #ifdef CONFIG_SMP
432 	printk(" CPU: %d", smp_processor_id());
433 #endif /* CONFIG_SMP */
434 
435 	for (i = 0;  i < 32;  i++) {
436 		if ((i % REGS_PER_LINE) == 0)
437 			printk("\n" KERN_INFO "GPR%02d: ", i);
438 		printk(REG " ", regs->gpr[i]);
439 		if (i == LAST_VOLATILE && !FULL_REGS(regs))
440 			break;
441 	}
442 	printk("\n");
443 #ifdef CONFIG_KALLSYMS
444 	/*
445 	 * Lookup NIP late so we have the best change of getting the
446 	 * above info out without failing
447 	 */
448 	printk("NIP ["REG"] ", regs->nip);
449 	print_symbol("%s\n", regs->nip);
450 	printk("LR ["REG"] ", regs->link);
451 	print_symbol("%s\n", regs->link);
452 #endif
453 	show_stack(current, (unsigned long *) regs->gpr[1]);
454 	if (!user_mode(regs))
455 		show_instructions(regs);
456 }
457 
458 void exit_thread(void)
459 {
460 	kprobe_flush_task(current);
461 	discard_lazy_cpu_state();
462 }
463 
464 void flush_thread(void)
465 {
466 #ifdef CONFIG_PPC64
467 	struct thread_info *t = current_thread_info();
468 
469 	if (t->flags & _TIF_ABI_PENDING)
470 		t->flags ^= (_TIF_ABI_PENDING | _TIF_32BIT);
471 #endif
472 
473 	discard_lazy_cpu_state();
474 
475 #ifdef CONFIG_PPC64	/* for now */
476 	if (current->thread.dabr) {
477 		current->thread.dabr = 0;
478 		set_dabr(0);
479 	}
480 #endif
481 }
482 
483 void
484 release_thread(struct task_struct *t)
485 {
486 }
487 
488 /*
489  * This gets called before we allocate a new thread and copy
490  * the current task into it.
491  */
492 void prepare_to_copy(struct task_struct *tsk)
493 {
494 	flush_fp_to_thread(current);
495 	flush_altivec_to_thread(current);
496 	flush_spe_to_thread(current);
497 }
498 
499 /*
500  * Copy a thread..
501  */
502 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
503 		unsigned long unused, struct task_struct *p,
504 		struct pt_regs *regs)
505 {
506 	struct pt_regs *childregs, *kregs;
507 	extern void ret_from_fork(void);
508 	unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
509 
510 	CHECK_FULL_REGS(regs);
511 	/* Copy registers */
512 	sp -= sizeof(struct pt_regs);
513 	childregs = (struct pt_regs *) sp;
514 	*childregs = *regs;
515 	if ((childregs->msr & MSR_PR) == 0) {
516 		/* for kernel thread, set `current' and stackptr in new task */
517 		childregs->gpr[1] = sp + sizeof(struct pt_regs);
518 #ifdef CONFIG_PPC32
519 		childregs->gpr[2] = (unsigned long) p;
520 #else
521 		clear_tsk_thread_flag(p, TIF_32BIT);
522 #endif
523 		p->thread.regs = NULL;	/* no user register state */
524 	} else {
525 		childregs->gpr[1] = usp;
526 		p->thread.regs = childregs;
527 		if (clone_flags & CLONE_SETTLS) {
528 #ifdef CONFIG_PPC64
529 			if (!test_thread_flag(TIF_32BIT))
530 				childregs->gpr[13] = childregs->gpr[6];
531 			else
532 #endif
533 				childregs->gpr[2] = childregs->gpr[6];
534 		}
535 	}
536 	childregs->gpr[3] = 0;  /* Result from fork() */
537 	sp -= STACK_FRAME_OVERHEAD;
538 
539 	/*
540 	 * The way this works is that at some point in the future
541 	 * some task will call _switch to switch to the new task.
542 	 * That will pop off the stack frame created below and start
543 	 * the new task running at ret_from_fork.  The new task will
544 	 * do some house keeping and then return from the fork or clone
545 	 * system call, using the stack frame created above.
546 	 */
547 	sp -= sizeof(struct pt_regs);
548 	kregs = (struct pt_regs *) sp;
549 	sp -= STACK_FRAME_OVERHEAD;
550 	p->thread.ksp = sp;
551 
552 #ifdef CONFIG_PPC64
553 	if (cpu_has_feature(CPU_FTR_SLB)) {
554 		unsigned long sp_vsid = get_kernel_vsid(sp);
555 		unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
556 
557 		sp_vsid <<= SLB_VSID_SHIFT;
558 		sp_vsid |= SLB_VSID_KERNEL | llp;
559 		p->thread.ksp_vsid = sp_vsid;
560 	}
561 
562 	/*
563 	 * The PPC64 ABI makes use of a TOC to contain function
564 	 * pointers.  The function (ret_from_except) is actually a pointer
565 	 * to the TOC entry.  The first entry is a pointer to the actual
566 	 * function.
567  	 */
568 	kregs->nip = *((unsigned long *)ret_from_fork);
569 #else
570 	kregs->nip = (unsigned long)ret_from_fork;
571 	p->thread.last_syscall = -1;
572 #endif
573 
574 	return 0;
575 }
576 
577 /*
578  * Set up a thread for executing a new program
579  */
580 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
581 {
582 #ifdef CONFIG_PPC64
583 	unsigned long load_addr = regs->gpr[2];	/* saved by ELF_PLAT_INIT */
584 #endif
585 
586 	set_fs(USER_DS);
587 
588 	/*
589 	 * If we exec out of a kernel thread then thread.regs will not be
590 	 * set.  Do it now.
591 	 */
592 	if (!current->thread.regs) {
593 		struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
594 		current->thread.regs = regs - 1;
595 	}
596 
597 	memset(regs->gpr, 0, sizeof(regs->gpr));
598 	regs->ctr = 0;
599 	regs->link = 0;
600 	regs->xer = 0;
601 	regs->ccr = 0;
602 	regs->gpr[1] = sp;
603 
604 #ifdef CONFIG_PPC32
605 	regs->mq = 0;
606 	regs->nip = start;
607 	regs->msr = MSR_USER;
608 #else
609 	if (!test_thread_flag(TIF_32BIT)) {
610 		unsigned long entry, toc;
611 
612 		/* start is a relocated pointer to the function descriptor for
613 		 * the elf _start routine.  The first entry in the function
614 		 * descriptor is the entry address of _start and the second
615 		 * entry is the TOC value we need to use.
616 		 */
617 		__get_user(entry, (unsigned long __user *)start);
618 		__get_user(toc, (unsigned long __user *)start+1);
619 
620 		/* Check whether the e_entry function descriptor entries
621 		 * need to be relocated before we can use them.
622 		 */
623 		if (load_addr != 0) {
624 			entry += load_addr;
625 			toc   += load_addr;
626 		}
627 		regs->nip = entry;
628 		regs->gpr[2] = toc;
629 		regs->msr = MSR_USER64;
630 	} else {
631 		regs->nip = start;
632 		regs->gpr[2] = 0;
633 		regs->msr = MSR_USER32;
634 	}
635 #endif
636 
637 	discard_lazy_cpu_state();
638 	memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
639 	current->thread.fpscr.val = 0;
640 #ifdef CONFIG_ALTIVEC
641 	memset(current->thread.vr, 0, sizeof(current->thread.vr));
642 	memset(&current->thread.vscr, 0, sizeof(current->thread.vscr));
643 	current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
644 	current->thread.vrsave = 0;
645 	current->thread.used_vr = 0;
646 #endif /* CONFIG_ALTIVEC */
647 #ifdef CONFIG_SPE
648 	memset(current->thread.evr, 0, sizeof(current->thread.evr));
649 	current->thread.acc = 0;
650 	current->thread.spefscr = 0;
651 	current->thread.used_spe = 0;
652 #endif /* CONFIG_SPE */
653 }
654 
655 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
656 		| PR_FP_EXC_RES | PR_FP_EXC_INV)
657 
658 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
659 {
660 	struct pt_regs *regs = tsk->thread.regs;
661 
662 	/* This is a bit hairy.  If we are an SPE enabled  processor
663 	 * (have embedded fp) we store the IEEE exception enable flags in
664 	 * fpexc_mode.  fpexc_mode is also used for setting FP exception
665 	 * mode (asyn, precise, disabled) for 'Classic' FP. */
666 	if (val & PR_FP_EXC_SW_ENABLE) {
667 #ifdef CONFIG_SPE
668 		tsk->thread.fpexc_mode = val &
669 			(PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
670 		return 0;
671 #else
672 		return -EINVAL;
673 #endif
674 	}
675 
676 	/* on a CONFIG_SPE this does not hurt us.  The bits that
677 	 * __pack_fe01 use do not overlap with bits used for
678 	 * PR_FP_EXC_SW_ENABLE.  Additionally, the MSR[FE0,FE1] bits
679 	 * on CONFIG_SPE implementations are reserved so writing to
680 	 * them does not change anything */
681 	if (val > PR_FP_EXC_PRECISE)
682 		return -EINVAL;
683 	tsk->thread.fpexc_mode = __pack_fe01(val);
684 	if (regs != NULL && (regs->msr & MSR_FP) != 0)
685 		regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
686 			| tsk->thread.fpexc_mode;
687 	return 0;
688 }
689 
690 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
691 {
692 	unsigned int val;
693 
694 	if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
695 #ifdef CONFIG_SPE
696 		val = tsk->thread.fpexc_mode;
697 #else
698 		return -EINVAL;
699 #endif
700 	else
701 		val = __unpack_fe01(tsk->thread.fpexc_mode);
702 	return put_user(val, (unsigned int __user *) adr);
703 }
704 
705 #define TRUNC_PTR(x)	((typeof(x))(((unsigned long)(x)) & 0xffffffff))
706 
707 int sys_clone(unsigned long clone_flags, unsigned long usp,
708 	      int __user *parent_tidp, void __user *child_threadptr,
709 	      int __user *child_tidp, int p6,
710 	      struct pt_regs *regs)
711 {
712 	CHECK_FULL_REGS(regs);
713 	if (usp == 0)
714 		usp = regs->gpr[1];	/* stack pointer for child */
715 #ifdef CONFIG_PPC64
716 	if (test_thread_flag(TIF_32BIT)) {
717 		parent_tidp = TRUNC_PTR(parent_tidp);
718 		child_tidp = TRUNC_PTR(child_tidp);
719 	}
720 #endif
721  	return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
722 }
723 
724 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
725 	     unsigned long p4, unsigned long p5, unsigned long p6,
726 	     struct pt_regs *regs)
727 {
728 	CHECK_FULL_REGS(regs);
729 	return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
730 }
731 
732 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
733 	      unsigned long p4, unsigned long p5, unsigned long p6,
734 	      struct pt_regs *regs)
735 {
736 	CHECK_FULL_REGS(regs);
737 	return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
738 			regs, 0, NULL, NULL);
739 }
740 
741 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
742 	       unsigned long a3, unsigned long a4, unsigned long a5,
743 	       struct pt_regs *regs)
744 {
745 	int error;
746 	char *filename;
747 
748 	filename = getname((char __user *) a0);
749 	error = PTR_ERR(filename);
750 	if (IS_ERR(filename))
751 		goto out;
752 	flush_fp_to_thread(current);
753 	flush_altivec_to_thread(current);
754 	flush_spe_to_thread(current);
755 	error = do_execve(filename, (char __user * __user *) a1,
756 			  (char __user * __user *) a2, regs);
757 	if (error == 0) {
758 		task_lock(current);
759 		current->ptrace &= ~PT_DTRACE;
760 		task_unlock(current);
761 	}
762 	putname(filename);
763 out:
764 	return error;
765 }
766 
767 static int validate_sp(unsigned long sp, struct task_struct *p,
768 		       unsigned long nbytes)
769 {
770 	unsigned long stack_page = (unsigned long)task_stack_page(p);
771 
772 	if (sp >= stack_page + sizeof(struct thread_struct)
773 	    && sp <= stack_page + THREAD_SIZE - nbytes)
774 		return 1;
775 
776 #ifdef CONFIG_IRQSTACKS
777 	stack_page = (unsigned long) hardirq_ctx[task_cpu(p)];
778 	if (sp >= stack_page + sizeof(struct thread_struct)
779 	    && sp <= stack_page + THREAD_SIZE - nbytes)
780 		return 1;
781 
782 	stack_page = (unsigned long) softirq_ctx[task_cpu(p)];
783 	if (sp >= stack_page + sizeof(struct thread_struct)
784 	    && sp <= stack_page + THREAD_SIZE - nbytes)
785 		return 1;
786 #endif
787 
788 	return 0;
789 }
790 
791 #ifdef CONFIG_PPC64
792 #define MIN_STACK_FRAME	112	/* same as STACK_FRAME_OVERHEAD, in fact */
793 #define FRAME_LR_SAVE	2
794 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
795 #define REGS_MARKER	0x7265677368657265ul
796 #define FRAME_MARKER	12
797 #else
798 #define MIN_STACK_FRAME	16
799 #define FRAME_LR_SAVE	1
800 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
801 #define REGS_MARKER	0x72656773ul
802 #define FRAME_MARKER	2
803 #endif
804 
805 unsigned long get_wchan(struct task_struct *p)
806 {
807 	unsigned long ip, sp;
808 	int count = 0;
809 
810 	if (!p || p == current || p->state == TASK_RUNNING)
811 		return 0;
812 
813 	sp = p->thread.ksp;
814 	if (!validate_sp(sp, p, MIN_STACK_FRAME))
815 		return 0;
816 
817 	do {
818 		sp = *(unsigned long *)sp;
819 		if (!validate_sp(sp, p, MIN_STACK_FRAME))
820 			return 0;
821 		if (count > 0) {
822 			ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
823 			if (!in_sched_functions(ip))
824 				return ip;
825 		}
826 	} while (count++ < 16);
827 	return 0;
828 }
829 EXPORT_SYMBOL(get_wchan);
830 
831 static int kstack_depth_to_print = 64;
832 
833 void show_stack(struct task_struct *tsk, unsigned long *stack)
834 {
835 	unsigned long sp, ip, lr, newsp;
836 	int count = 0;
837 	int firstframe = 1;
838 
839 	sp = (unsigned long) stack;
840 	if (tsk == NULL)
841 		tsk = current;
842 	if (sp == 0) {
843 		if (tsk == current)
844 			asm("mr %0,1" : "=r" (sp));
845 		else
846 			sp = tsk->thread.ksp;
847 	}
848 
849 	lr = 0;
850 	printk("Call Trace:\n");
851 	do {
852 		if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
853 			return;
854 
855 		stack = (unsigned long *) sp;
856 		newsp = stack[0];
857 		ip = stack[FRAME_LR_SAVE];
858 		if (!firstframe || ip != lr) {
859 			printk("["REG"] ["REG"] ", sp, ip);
860 			print_symbol("%s", ip);
861 			if (firstframe)
862 				printk(" (unreliable)");
863 			printk("\n");
864 		}
865 		firstframe = 0;
866 
867 		/*
868 		 * See if this is an exception frame.
869 		 * We look for the "regshere" marker in the current frame.
870 		 */
871 		if (validate_sp(sp, tsk, INT_FRAME_SIZE)
872 		    && stack[FRAME_MARKER] == REGS_MARKER) {
873 			struct pt_regs *regs = (struct pt_regs *)
874 				(sp + STACK_FRAME_OVERHEAD);
875 			printk("--- Exception: %lx", regs->trap);
876 			print_symbol(" at %s\n", regs->nip);
877 			lr = regs->link;
878 			print_symbol("    LR = %s\n", lr);
879 			firstframe = 1;
880 		}
881 
882 		sp = newsp;
883 	} while (count++ < kstack_depth_to_print);
884 }
885 
886 void dump_stack(void)
887 {
888 	show_stack(current, NULL);
889 }
890 EXPORT_SYMBOL(dump_stack);
891 
892 #ifdef CONFIG_PPC64
893 void ppc64_runlatch_on(void)
894 {
895 	unsigned long ctrl;
896 
897 	if (cpu_has_feature(CPU_FTR_CTRL) && !test_thread_flag(TIF_RUNLATCH)) {
898 		HMT_medium();
899 
900 		ctrl = mfspr(SPRN_CTRLF);
901 		ctrl |= CTRL_RUNLATCH;
902 		mtspr(SPRN_CTRLT, ctrl);
903 
904 		set_thread_flag(TIF_RUNLATCH);
905 	}
906 }
907 
908 void ppc64_runlatch_off(void)
909 {
910 	unsigned long ctrl;
911 
912 	if (cpu_has_feature(CPU_FTR_CTRL) && test_thread_flag(TIF_RUNLATCH)) {
913 		HMT_medium();
914 
915 		clear_thread_flag(TIF_RUNLATCH);
916 
917 		ctrl = mfspr(SPRN_CTRLF);
918 		ctrl &= ~CTRL_RUNLATCH;
919 		mtspr(SPRN_CTRLT, ctrl);
920 	}
921 }
922 #endif
923