xref: /linux/arch/powerpc/kernel/process.c (revision ba6e8564f459211117ce300eae2c7fdd23befe34)
1 /*
2  *  Derived from "arch/i386/kernel/process.c"
3  *    Copyright (C) 1995  Linus Torvalds
4  *
5  *  Updated and modified by Cort Dougan (cort@cs.nmt.edu) and
6  *  Paul Mackerras (paulus@cs.anu.edu.au)
7  *
8  *  PowerPC version
9  *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
10  *
11  *  This program is free software; you can redistribute it and/or
12  *  modify it under the terms of the GNU General Public License
13  *  as published by the Free Software Foundation; either version
14  *  2 of the License, or (at your option) any later version.
15  */
16 
17 #include <linux/errno.h>
18 #include <linux/sched.h>
19 #include <linux/kernel.h>
20 #include <linux/mm.h>
21 #include <linux/smp.h>
22 #include <linux/smp_lock.h>
23 #include <linux/stddef.h>
24 #include <linux/unistd.h>
25 #include <linux/ptrace.h>
26 #include <linux/slab.h>
27 #include <linux/user.h>
28 #include <linux/elf.h>
29 #include <linux/init.h>
30 #include <linux/prctl.h>
31 #include <linux/init_task.h>
32 #include <linux/module.h>
33 #include <linux/kallsyms.h>
34 #include <linux/mqueue.h>
35 #include <linux/hardirq.h>
36 #include <linux/utsname.h>
37 
38 #include <asm/pgtable.h>
39 #include <asm/uaccess.h>
40 #include <asm/system.h>
41 #include <asm/io.h>
42 #include <asm/processor.h>
43 #include <asm/mmu.h>
44 #include <asm/prom.h>
45 #include <asm/machdep.h>
46 #include <asm/time.h>
47 #include <asm/syscalls.h>
48 #ifdef CONFIG_PPC64
49 #include <asm/firmware.h>
50 #endif
51 
52 extern unsigned long _get_SP(void);
53 
54 #ifndef CONFIG_SMP
55 struct task_struct *last_task_used_math = NULL;
56 struct task_struct *last_task_used_altivec = NULL;
57 struct task_struct *last_task_used_spe = NULL;
58 #endif
59 
60 /*
61  * Make sure the floating-point register state in the
62  * the thread_struct is up to date for task tsk.
63  */
64 void flush_fp_to_thread(struct task_struct *tsk)
65 {
66 	if (tsk->thread.regs) {
67 		/*
68 		 * We need to disable preemption here because if we didn't,
69 		 * another process could get scheduled after the regs->msr
70 		 * test but before we have finished saving the FP registers
71 		 * to the thread_struct.  That process could take over the
72 		 * FPU, and then when we get scheduled again we would store
73 		 * bogus values for the remaining FP registers.
74 		 */
75 		preempt_disable();
76 		if (tsk->thread.regs->msr & MSR_FP) {
77 #ifdef CONFIG_SMP
78 			/*
79 			 * This should only ever be called for current or
80 			 * for a stopped child process.  Since we save away
81 			 * the FP register state on context switch on SMP,
82 			 * there is something wrong if a stopped child appears
83 			 * to still have its FP state in the CPU registers.
84 			 */
85 			BUG_ON(tsk != current);
86 #endif
87 			giveup_fpu(current);
88 		}
89 		preempt_enable();
90 	}
91 }
92 
93 void enable_kernel_fp(void)
94 {
95 	WARN_ON(preemptible());
96 
97 #ifdef CONFIG_SMP
98 	if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
99 		giveup_fpu(current);
100 	else
101 		giveup_fpu(NULL);	/* just enables FP for kernel */
102 #else
103 	giveup_fpu(last_task_used_math);
104 #endif /* CONFIG_SMP */
105 }
106 EXPORT_SYMBOL(enable_kernel_fp);
107 
108 int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
109 {
110 	if (!tsk->thread.regs)
111 		return 0;
112 	flush_fp_to_thread(current);
113 
114 	memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs));
115 
116 	return 1;
117 }
118 
119 #ifdef CONFIG_ALTIVEC
120 void enable_kernel_altivec(void)
121 {
122 	WARN_ON(preemptible());
123 
124 #ifdef CONFIG_SMP
125 	if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
126 		giveup_altivec(current);
127 	else
128 		giveup_altivec(NULL);	/* just enable AltiVec for kernel - force */
129 #else
130 	giveup_altivec(last_task_used_altivec);
131 #endif /* CONFIG_SMP */
132 }
133 EXPORT_SYMBOL(enable_kernel_altivec);
134 
135 /*
136  * Make sure the VMX/Altivec register state in the
137  * the thread_struct is up to date for task tsk.
138  */
139 void flush_altivec_to_thread(struct task_struct *tsk)
140 {
141 	if (tsk->thread.regs) {
142 		preempt_disable();
143 		if (tsk->thread.regs->msr & MSR_VEC) {
144 #ifdef CONFIG_SMP
145 			BUG_ON(tsk != current);
146 #endif
147 			giveup_altivec(current);
148 		}
149 		preempt_enable();
150 	}
151 }
152 
153 int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
154 {
155 	flush_altivec_to_thread(current);
156 	memcpy(vrregs, &current->thread.vr[0], sizeof(*vrregs));
157 	return 1;
158 }
159 #endif /* CONFIG_ALTIVEC */
160 
161 #ifdef CONFIG_SPE
162 
163 void enable_kernel_spe(void)
164 {
165 	WARN_ON(preemptible());
166 
167 #ifdef CONFIG_SMP
168 	if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
169 		giveup_spe(current);
170 	else
171 		giveup_spe(NULL);	/* just enable SPE for kernel - force */
172 #else
173 	giveup_spe(last_task_used_spe);
174 #endif /* __SMP __ */
175 }
176 EXPORT_SYMBOL(enable_kernel_spe);
177 
178 void flush_spe_to_thread(struct task_struct *tsk)
179 {
180 	if (tsk->thread.regs) {
181 		preempt_disable();
182 		if (tsk->thread.regs->msr & MSR_SPE) {
183 #ifdef CONFIG_SMP
184 			BUG_ON(tsk != current);
185 #endif
186 			giveup_spe(current);
187 		}
188 		preempt_enable();
189 	}
190 }
191 
192 int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
193 {
194 	flush_spe_to_thread(current);
195 	/* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */
196 	memcpy(evrregs, &current->thread.evr[0], sizeof(u32) * 35);
197 	return 1;
198 }
199 #endif /* CONFIG_SPE */
200 
201 #ifndef CONFIG_SMP
202 /*
203  * If we are doing lazy switching of CPU state (FP, altivec or SPE),
204  * and the current task has some state, discard it.
205  */
206 void discard_lazy_cpu_state(void)
207 {
208 	preempt_disable();
209 	if (last_task_used_math == current)
210 		last_task_used_math = NULL;
211 #ifdef CONFIG_ALTIVEC
212 	if (last_task_used_altivec == current)
213 		last_task_used_altivec = NULL;
214 #endif /* CONFIG_ALTIVEC */
215 #ifdef CONFIG_SPE
216 	if (last_task_used_spe == current)
217 		last_task_used_spe = NULL;
218 #endif
219 	preempt_enable();
220 }
221 #endif /* CONFIG_SMP */
222 
223 #ifdef CONFIG_PPC_MERGE		/* XXX for now */
224 int set_dabr(unsigned long dabr)
225 {
226 	if (ppc_md.set_dabr)
227 		return ppc_md.set_dabr(dabr);
228 
229 	mtspr(SPRN_DABR, dabr);
230 	return 0;
231 }
232 #endif
233 
234 #ifdef CONFIG_PPC64
235 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
236 static DEFINE_PER_CPU(unsigned long, current_dabr);
237 #endif
238 
239 struct task_struct *__switch_to(struct task_struct *prev,
240 	struct task_struct *new)
241 {
242 	struct thread_struct *new_thread, *old_thread;
243 	unsigned long flags;
244 	struct task_struct *last;
245 
246 #ifdef CONFIG_SMP
247 	/* avoid complexity of lazy save/restore of fpu
248 	 * by just saving it every time we switch out if
249 	 * this task used the fpu during the last quantum.
250 	 *
251 	 * If it tries to use the fpu again, it'll trap and
252 	 * reload its fp regs.  So we don't have to do a restore
253 	 * every switch, just a save.
254 	 *  -- Cort
255 	 */
256 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
257 		giveup_fpu(prev);
258 #ifdef CONFIG_ALTIVEC
259 	/*
260 	 * If the previous thread used altivec in the last quantum
261 	 * (thus changing altivec regs) then save them.
262 	 * We used to check the VRSAVE register but not all apps
263 	 * set it, so we don't rely on it now (and in fact we need
264 	 * to save & restore VSCR even if VRSAVE == 0).  -- paulus
265 	 *
266 	 * On SMP we always save/restore altivec regs just to avoid the
267 	 * complexity of changing processors.
268 	 *  -- Cort
269 	 */
270 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
271 		giveup_altivec(prev);
272 #endif /* CONFIG_ALTIVEC */
273 #ifdef CONFIG_SPE
274 	/*
275 	 * If the previous thread used spe in the last quantum
276 	 * (thus changing spe regs) then save them.
277 	 *
278 	 * On SMP we always save/restore spe regs just to avoid the
279 	 * complexity of changing processors.
280 	 */
281 	if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
282 		giveup_spe(prev);
283 #endif /* CONFIG_SPE */
284 
285 #else  /* CONFIG_SMP */
286 #ifdef CONFIG_ALTIVEC
287 	/* Avoid the trap.  On smp this this never happens since
288 	 * we don't set last_task_used_altivec -- Cort
289 	 */
290 	if (new->thread.regs && last_task_used_altivec == new)
291 		new->thread.regs->msr |= MSR_VEC;
292 #endif /* CONFIG_ALTIVEC */
293 #ifdef CONFIG_SPE
294 	/* Avoid the trap.  On smp this this never happens since
295 	 * we don't set last_task_used_spe
296 	 */
297 	if (new->thread.regs && last_task_used_spe == new)
298 		new->thread.regs->msr |= MSR_SPE;
299 #endif /* CONFIG_SPE */
300 
301 #endif /* CONFIG_SMP */
302 
303 #ifdef CONFIG_PPC64	/* for now */
304 	if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
305 		set_dabr(new->thread.dabr);
306 		__get_cpu_var(current_dabr) = new->thread.dabr;
307 	}
308 
309 	flush_tlb_pending();
310 #endif
311 
312 	new_thread = &new->thread;
313 	old_thread = &current->thread;
314 
315 #ifdef CONFIG_PPC64
316 	/*
317 	 * Collect processor utilization data per process
318 	 */
319 	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
320 		struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
321 		long unsigned start_tb, current_tb;
322 		start_tb = old_thread->start_tb;
323 		cu->current_tb = current_tb = mfspr(SPRN_PURR);
324 		old_thread->accum_tb += (current_tb - start_tb);
325 		new_thread->start_tb = current_tb;
326 	}
327 #endif
328 
329 	local_irq_save(flags);
330 
331 	account_system_vtime(current);
332 	account_process_vtime(current);
333 	calculate_steal_time();
334 
335 	last = _switch(old_thread, new_thread);
336 
337 	local_irq_restore(flags);
338 
339 	return last;
340 }
341 
342 static int instructions_to_print = 16;
343 
344 static void show_instructions(struct pt_regs *regs)
345 {
346 	int i;
347 	unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
348 			sizeof(int));
349 
350 	printk("Instruction dump:");
351 
352 	for (i = 0; i < instructions_to_print; i++) {
353 		int instr;
354 
355 		if (!(i % 8))
356 			printk("\n");
357 
358 		/* We use __get_user here *only* to avoid an OOPS on a
359 		 * bad address because the pc *should* only be a
360 		 * kernel address.
361 		 */
362 		if (!__kernel_text_address(pc) ||
363 		     __get_user(instr, (unsigned int __user *)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(), init_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 	discard_lazy_cpu_state();
461 }
462 
463 void flush_thread(void)
464 {
465 #ifdef CONFIG_PPC64
466 	struct thread_info *t = current_thread_info();
467 
468 	if (test_ti_thread_flag(t, TIF_ABI_PENDING)) {
469 		clear_ti_thread_flag(t, TIF_ABI_PENDING);
470 		if (test_ti_thread_flag(t, TIF_32BIT))
471 			clear_ti_thread_flag(t, TIF_32BIT);
472 		else
473 			set_ti_thread_flag(t, TIF_32BIT);
474 	}
475 #endif
476 
477 	discard_lazy_cpu_state();
478 
479 #ifdef CONFIG_PPC64	/* for now */
480 	if (current->thread.dabr) {
481 		current->thread.dabr = 0;
482 		set_dabr(0);
483 	}
484 #endif
485 }
486 
487 void
488 release_thread(struct task_struct *t)
489 {
490 }
491 
492 /*
493  * This gets called before we allocate a new thread and copy
494  * the current task into it.
495  */
496 void prepare_to_copy(struct task_struct *tsk)
497 {
498 	flush_fp_to_thread(current);
499 	flush_altivec_to_thread(current);
500 	flush_spe_to_thread(current);
501 }
502 
503 /*
504  * Copy a thread..
505  */
506 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
507 		unsigned long unused, struct task_struct *p,
508 		struct pt_regs *regs)
509 {
510 	struct pt_regs *childregs, *kregs;
511 	extern void ret_from_fork(void);
512 	unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
513 
514 	CHECK_FULL_REGS(regs);
515 	/* Copy registers */
516 	sp -= sizeof(struct pt_regs);
517 	childregs = (struct pt_regs *) sp;
518 	*childregs = *regs;
519 	if ((childregs->msr & MSR_PR) == 0) {
520 		/* for kernel thread, set `current' and stackptr in new task */
521 		childregs->gpr[1] = sp + sizeof(struct pt_regs);
522 #ifdef CONFIG_PPC32
523 		childregs->gpr[2] = (unsigned long) p;
524 #else
525 		clear_tsk_thread_flag(p, TIF_32BIT);
526 #endif
527 		p->thread.regs = NULL;	/* no user register state */
528 	} else {
529 		childregs->gpr[1] = usp;
530 		p->thread.regs = childregs;
531 		if (clone_flags & CLONE_SETTLS) {
532 #ifdef CONFIG_PPC64
533 			if (!test_thread_flag(TIF_32BIT))
534 				childregs->gpr[13] = childregs->gpr[6];
535 			else
536 #endif
537 				childregs->gpr[2] = childregs->gpr[6];
538 		}
539 	}
540 	childregs->gpr[3] = 0;  /* Result from fork() */
541 	sp -= STACK_FRAME_OVERHEAD;
542 
543 	/*
544 	 * The way this works is that at some point in the future
545 	 * some task will call _switch to switch to the new task.
546 	 * That will pop off the stack frame created below and start
547 	 * the new task running at ret_from_fork.  The new task will
548 	 * do some house keeping and then return from the fork or clone
549 	 * system call, using the stack frame created above.
550 	 */
551 	sp -= sizeof(struct pt_regs);
552 	kregs = (struct pt_regs *) sp;
553 	sp -= STACK_FRAME_OVERHEAD;
554 	p->thread.ksp = sp;
555 
556 #ifdef CONFIG_PPC64
557 	if (cpu_has_feature(CPU_FTR_SLB)) {
558 		unsigned long sp_vsid = get_kernel_vsid(sp);
559 		unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
560 
561 		sp_vsid <<= SLB_VSID_SHIFT;
562 		sp_vsid |= SLB_VSID_KERNEL | llp;
563 		p->thread.ksp_vsid = sp_vsid;
564 	}
565 
566 	/*
567 	 * The PPC64 ABI makes use of a TOC to contain function
568 	 * pointers.  The function (ret_from_except) is actually a pointer
569 	 * to the TOC entry.  The first entry is a pointer to the actual
570 	 * function.
571  	 */
572 	kregs->nip = *((unsigned long *)ret_from_fork);
573 #else
574 	kregs->nip = (unsigned long)ret_from_fork;
575 	p->thread.last_syscall = -1;
576 #endif
577 
578 	return 0;
579 }
580 
581 /*
582  * Set up a thread for executing a new program
583  */
584 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
585 {
586 #ifdef CONFIG_PPC64
587 	unsigned long load_addr = regs->gpr[2];	/* saved by ELF_PLAT_INIT */
588 #endif
589 
590 	set_fs(USER_DS);
591 
592 	/*
593 	 * If we exec out of a kernel thread then thread.regs will not be
594 	 * set.  Do it now.
595 	 */
596 	if (!current->thread.regs) {
597 		struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
598 		current->thread.regs = regs - 1;
599 	}
600 
601 	memset(regs->gpr, 0, sizeof(regs->gpr));
602 	regs->ctr = 0;
603 	regs->link = 0;
604 	regs->xer = 0;
605 	regs->ccr = 0;
606 	regs->gpr[1] = sp;
607 
608 #ifdef CONFIG_PPC32
609 	regs->mq = 0;
610 	regs->nip = start;
611 	regs->msr = MSR_USER;
612 #else
613 	if (!test_thread_flag(TIF_32BIT)) {
614 		unsigned long entry, toc;
615 
616 		/* start is a relocated pointer to the function descriptor for
617 		 * the elf _start routine.  The first entry in the function
618 		 * descriptor is the entry address of _start and the second
619 		 * entry is the TOC value we need to use.
620 		 */
621 		__get_user(entry, (unsigned long __user *)start);
622 		__get_user(toc, (unsigned long __user *)start+1);
623 
624 		/* Check whether the e_entry function descriptor entries
625 		 * need to be relocated before we can use them.
626 		 */
627 		if (load_addr != 0) {
628 			entry += load_addr;
629 			toc   += load_addr;
630 		}
631 		regs->nip = entry;
632 		regs->gpr[2] = toc;
633 		regs->msr = MSR_USER64;
634 	} else {
635 		regs->nip = start;
636 		regs->gpr[2] = 0;
637 		regs->msr = MSR_USER32;
638 	}
639 #endif
640 
641 	discard_lazy_cpu_state();
642 	memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
643 	current->thread.fpscr.val = 0;
644 #ifdef CONFIG_ALTIVEC
645 	memset(current->thread.vr, 0, sizeof(current->thread.vr));
646 	memset(&current->thread.vscr, 0, sizeof(current->thread.vscr));
647 	current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
648 	current->thread.vrsave = 0;
649 	current->thread.used_vr = 0;
650 #endif /* CONFIG_ALTIVEC */
651 #ifdef CONFIG_SPE
652 	memset(current->thread.evr, 0, sizeof(current->thread.evr));
653 	current->thread.acc = 0;
654 	current->thread.spefscr = 0;
655 	current->thread.used_spe = 0;
656 #endif /* CONFIG_SPE */
657 }
658 
659 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
660 		| PR_FP_EXC_RES | PR_FP_EXC_INV)
661 
662 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
663 {
664 	struct pt_regs *regs = tsk->thread.regs;
665 
666 	/* This is a bit hairy.  If we are an SPE enabled  processor
667 	 * (have embedded fp) we store the IEEE exception enable flags in
668 	 * fpexc_mode.  fpexc_mode is also used for setting FP exception
669 	 * mode (asyn, precise, disabled) for 'Classic' FP. */
670 	if (val & PR_FP_EXC_SW_ENABLE) {
671 #ifdef CONFIG_SPE
672 		tsk->thread.fpexc_mode = val &
673 			(PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
674 		return 0;
675 #else
676 		return -EINVAL;
677 #endif
678 	}
679 
680 	/* on a CONFIG_SPE this does not hurt us.  The bits that
681 	 * __pack_fe01 use do not overlap with bits used for
682 	 * PR_FP_EXC_SW_ENABLE.  Additionally, the MSR[FE0,FE1] bits
683 	 * on CONFIG_SPE implementations are reserved so writing to
684 	 * them does not change anything */
685 	if (val > PR_FP_EXC_PRECISE)
686 		return -EINVAL;
687 	tsk->thread.fpexc_mode = __pack_fe01(val);
688 	if (regs != NULL && (regs->msr & MSR_FP) != 0)
689 		regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
690 			| tsk->thread.fpexc_mode;
691 	return 0;
692 }
693 
694 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
695 {
696 	unsigned int val;
697 
698 	if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
699 #ifdef CONFIG_SPE
700 		val = tsk->thread.fpexc_mode;
701 #else
702 		return -EINVAL;
703 #endif
704 	else
705 		val = __unpack_fe01(tsk->thread.fpexc_mode);
706 	return put_user(val, (unsigned int __user *) adr);
707 }
708 
709 int set_endian(struct task_struct *tsk, unsigned int val)
710 {
711 	struct pt_regs *regs = tsk->thread.regs;
712 
713 	if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
714 	    (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
715 		return -EINVAL;
716 
717 	if (regs == NULL)
718 		return -EINVAL;
719 
720 	if (val == PR_ENDIAN_BIG)
721 		regs->msr &= ~MSR_LE;
722 	else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
723 		regs->msr |= MSR_LE;
724 	else
725 		return -EINVAL;
726 
727 	return 0;
728 }
729 
730 int get_endian(struct task_struct *tsk, unsigned long adr)
731 {
732 	struct pt_regs *regs = tsk->thread.regs;
733 	unsigned int val;
734 
735 	if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
736 	    !cpu_has_feature(CPU_FTR_REAL_LE))
737 		return -EINVAL;
738 
739 	if (regs == NULL)
740 		return -EINVAL;
741 
742 	if (regs->msr & MSR_LE) {
743 		if (cpu_has_feature(CPU_FTR_REAL_LE))
744 			val = PR_ENDIAN_LITTLE;
745 		else
746 			val = PR_ENDIAN_PPC_LITTLE;
747 	} else
748 		val = PR_ENDIAN_BIG;
749 
750 	return put_user(val, (unsigned int __user *)adr);
751 }
752 
753 int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
754 {
755 	tsk->thread.align_ctl = val;
756 	return 0;
757 }
758 
759 int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
760 {
761 	return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
762 }
763 
764 #define TRUNC_PTR(x)	((typeof(x))(((unsigned long)(x)) & 0xffffffff))
765 
766 int sys_clone(unsigned long clone_flags, unsigned long usp,
767 	      int __user *parent_tidp, void __user *child_threadptr,
768 	      int __user *child_tidp, int p6,
769 	      struct pt_regs *regs)
770 {
771 	CHECK_FULL_REGS(regs);
772 	if (usp == 0)
773 		usp = regs->gpr[1];	/* stack pointer for child */
774 #ifdef CONFIG_PPC64
775 	if (test_thread_flag(TIF_32BIT)) {
776 		parent_tidp = TRUNC_PTR(parent_tidp);
777 		child_tidp = TRUNC_PTR(child_tidp);
778 	}
779 #endif
780  	return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
781 }
782 
783 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
784 	     unsigned long p4, unsigned long p5, unsigned long p6,
785 	     struct pt_regs *regs)
786 {
787 	CHECK_FULL_REGS(regs);
788 	return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
789 }
790 
791 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
792 	      unsigned long p4, unsigned long p5, unsigned long p6,
793 	      struct pt_regs *regs)
794 {
795 	CHECK_FULL_REGS(regs);
796 	return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
797 			regs, 0, NULL, NULL);
798 }
799 
800 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
801 	       unsigned long a3, unsigned long a4, unsigned long a5,
802 	       struct pt_regs *regs)
803 {
804 	int error;
805 	char *filename;
806 
807 	filename = getname((char __user *) a0);
808 	error = PTR_ERR(filename);
809 	if (IS_ERR(filename))
810 		goto out;
811 	flush_fp_to_thread(current);
812 	flush_altivec_to_thread(current);
813 	flush_spe_to_thread(current);
814 	error = do_execve(filename, (char __user * __user *) a1,
815 			  (char __user * __user *) a2, regs);
816 	if (error == 0) {
817 		task_lock(current);
818 		current->ptrace &= ~PT_DTRACE;
819 		task_unlock(current);
820 	}
821 	putname(filename);
822 out:
823 	return error;
824 }
825 
826 int validate_sp(unsigned long sp, struct task_struct *p,
827 		       unsigned long nbytes)
828 {
829 	unsigned long stack_page = (unsigned long)task_stack_page(p);
830 
831 	if (sp >= stack_page + sizeof(struct thread_struct)
832 	    && sp <= stack_page + THREAD_SIZE - nbytes)
833 		return 1;
834 
835 #ifdef CONFIG_IRQSTACKS
836 	stack_page = (unsigned long) hardirq_ctx[task_cpu(p)];
837 	if (sp >= stack_page + sizeof(struct thread_struct)
838 	    && sp <= stack_page + THREAD_SIZE - nbytes)
839 		return 1;
840 
841 	stack_page = (unsigned long) softirq_ctx[task_cpu(p)];
842 	if (sp >= stack_page + sizeof(struct thread_struct)
843 	    && sp <= stack_page + THREAD_SIZE - nbytes)
844 		return 1;
845 #endif
846 
847 	return 0;
848 }
849 
850 #ifdef CONFIG_PPC64
851 #define MIN_STACK_FRAME	112	/* same as STACK_FRAME_OVERHEAD, in fact */
852 #define FRAME_LR_SAVE	2
853 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
854 #define REGS_MARKER	0x7265677368657265ul
855 #define FRAME_MARKER	12
856 #else
857 #define MIN_STACK_FRAME	16
858 #define FRAME_LR_SAVE	1
859 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
860 #define REGS_MARKER	0x72656773ul
861 #define FRAME_MARKER	2
862 #endif
863 
864 EXPORT_SYMBOL(validate_sp);
865 
866 unsigned long get_wchan(struct task_struct *p)
867 {
868 	unsigned long ip, sp;
869 	int count = 0;
870 
871 	if (!p || p == current || p->state == TASK_RUNNING)
872 		return 0;
873 
874 	sp = p->thread.ksp;
875 	if (!validate_sp(sp, p, MIN_STACK_FRAME))
876 		return 0;
877 
878 	do {
879 		sp = *(unsigned long *)sp;
880 		if (!validate_sp(sp, p, MIN_STACK_FRAME))
881 			return 0;
882 		if (count > 0) {
883 			ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
884 			if (!in_sched_functions(ip))
885 				return ip;
886 		}
887 	} while (count++ < 16);
888 	return 0;
889 }
890 
891 static int kstack_depth_to_print = 64;
892 
893 void show_stack(struct task_struct *tsk, unsigned long *stack)
894 {
895 	unsigned long sp, ip, lr, newsp;
896 	int count = 0;
897 	int firstframe = 1;
898 
899 	sp = (unsigned long) stack;
900 	if (tsk == NULL)
901 		tsk = current;
902 	if (sp == 0) {
903 		if (tsk == current)
904 			asm("mr %0,1" : "=r" (sp));
905 		else
906 			sp = tsk->thread.ksp;
907 	}
908 
909 	lr = 0;
910 	printk("Call Trace:\n");
911 	do {
912 		if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
913 			return;
914 
915 		stack = (unsigned long *) sp;
916 		newsp = stack[0];
917 		ip = stack[FRAME_LR_SAVE];
918 		if (!firstframe || ip != lr) {
919 			printk("["REG"] ["REG"] ", sp, ip);
920 			print_symbol("%s", ip);
921 			if (firstframe)
922 				printk(" (unreliable)");
923 			printk("\n");
924 		}
925 		firstframe = 0;
926 
927 		/*
928 		 * See if this is an exception frame.
929 		 * We look for the "regshere" marker in the current frame.
930 		 */
931 		if (validate_sp(sp, tsk, INT_FRAME_SIZE)
932 		    && stack[FRAME_MARKER] == REGS_MARKER) {
933 			struct pt_regs *regs = (struct pt_regs *)
934 				(sp + STACK_FRAME_OVERHEAD);
935 			printk("--- Exception: %lx", regs->trap);
936 			print_symbol(" at %s\n", regs->nip);
937 			lr = regs->link;
938 			print_symbol("    LR = %s\n", lr);
939 			firstframe = 1;
940 		}
941 
942 		sp = newsp;
943 	} while (count++ < kstack_depth_to_print);
944 }
945 
946 void dump_stack(void)
947 {
948 	show_stack(current, NULL);
949 }
950 EXPORT_SYMBOL(dump_stack);
951 
952 #ifdef CONFIG_PPC64
953 void ppc64_runlatch_on(void)
954 {
955 	unsigned long ctrl;
956 
957 	if (cpu_has_feature(CPU_FTR_CTRL) && !test_thread_flag(TIF_RUNLATCH)) {
958 		HMT_medium();
959 
960 		ctrl = mfspr(SPRN_CTRLF);
961 		ctrl |= CTRL_RUNLATCH;
962 		mtspr(SPRN_CTRLT, ctrl);
963 
964 		set_thread_flag(TIF_RUNLATCH);
965 	}
966 }
967 
968 void ppc64_runlatch_off(void)
969 {
970 	unsigned long ctrl;
971 
972 	if (cpu_has_feature(CPU_FTR_CTRL) && test_thread_flag(TIF_RUNLATCH)) {
973 		HMT_medium();
974 
975 		clear_thread_flag(TIF_RUNLATCH);
976 
977 		ctrl = mfspr(SPRN_CTRLF);
978 		ctrl &= ~CTRL_RUNLATCH;
979 		mtspr(SPRN_CTRLT, ctrl);
980 	}
981 }
982 #endif
983