xref: /linux/arch/powerpc/kernel/process.c (revision eb2bce7f5e7ac1ca6da434461217fadf3c688d2c)
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 #endif /* CONFIG_PPC64 */
309 
310 	new_thread = &new->thread;
311 	old_thread = &current->thread;
312 
313 #ifdef CONFIG_PPC64
314 	/*
315 	 * Collect processor utilization data per process
316 	 */
317 	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
318 		struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
319 		long unsigned start_tb, current_tb;
320 		start_tb = old_thread->start_tb;
321 		cu->current_tb = current_tb = mfspr(SPRN_PURR);
322 		old_thread->accum_tb += (current_tb - start_tb);
323 		new_thread->start_tb = current_tb;
324 	}
325 #endif
326 
327 	local_irq_save(flags);
328 
329 	account_system_vtime(current);
330 	account_process_vtime(current);
331 	calculate_steal_time();
332 
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 static void show_instructions(struct pt_regs *regs)
343 {
344 	int i;
345 	unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
346 			sizeof(int));
347 
348 	printk("Instruction dump:");
349 
350 	for (i = 0; i < instructions_to_print; i++) {
351 		int instr;
352 
353 		if (!(i % 8))
354 			printk("\n");
355 
356 		/* We use __get_user here *only* to avoid an OOPS on a
357 		 * bad address because the pc *should* only be a
358 		 * kernel address.
359 		 */
360 		if (!__kernel_text_address(pc) ||
361 		     __get_user(instr, (unsigned int __user *)pc)) {
362 			printk("XXXXXXXX ");
363 		} else {
364 			if (regs->nip == pc)
365 				printk("<%08x> ", instr);
366 			else
367 				printk("%08x ", instr);
368 		}
369 
370 		pc += sizeof(int);
371 	}
372 
373 	printk("\n");
374 }
375 
376 static struct regbit {
377 	unsigned long bit;
378 	const char *name;
379 } msr_bits[] = {
380 	{MSR_EE,	"EE"},
381 	{MSR_PR,	"PR"},
382 	{MSR_FP,	"FP"},
383 	{MSR_ME,	"ME"},
384 	{MSR_IR,	"IR"},
385 	{MSR_DR,	"DR"},
386 	{0,		NULL}
387 };
388 
389 static void printbits(unsigned long val, struct regbit *bits)
390 {
391 	const char *sep = "";
392 
393 	printk("<");
394 	for (; bits->bit; ++bits)
395 		if (val & bits->bit) {
396 			printk("%s%s", sep, bits->name);
397 			sep = ",";
398 		}
399 	printk(">");
400 }
401 
402 #ifdef CONFIG_PPC64
403 #define REG		"%016lx"
404 #define REGS_PER_LINE	4
405 #define LAST_VOLATILE	13
406 #else
407 #define REG		"%08lx"
408 #define REGS_PER_LINE	8
409 #define LAST_VOLATILE	12
410 #endif
411 
412 void show_regs(struct pt_regs * regs)
413 {
414 	int i, trap;
415 
416 	printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
417 	       regs->nip, regs->link, regs->ctr);
418 	printk("REGS: %p TRAP: %04lx   %s  (%s)\n",
419 	       regs, regs->trap, print_tainted(), init_utsname()->release);
420 	printk("MSR: "REG" ", regs->msr);
421 	printbits(regs->msr, msr_bits);
422 	printk("  CR: %08lx  XER: %08lx\n", regs->ccr, regs->xer);
423 	trap = TRAP(regs);
424 	if (trap == 0x300 || trap == 0x600)
425 		printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
426 	printk("TASK = %p[%d] '%s' THREAD: %p",
427 	       current, current->pid, current->comm, task_thread_info(current));
428 
429 #ifdef CONFIG_SMP
430 	printk(" CPU: %d", smp_processor_id());
431 #endif /* CONFIG_SMP */
432 
433 	for (i = 0;  i < 32;  i++) {
434 		if ((i % REGS_PER_LINE) == 0)
435 			printk("\n" KERN_INFO "GPR%02d: ", i);
436 		printk(REG " ", regs->gpr[i]);
437 		if (i == LAST_VOLATILE && !FULL_REGS(regs))
438 			break;
439 	}
440 	printk("\n");
441 #ifdef CONFIG_KALLSYMS
442 	/*
443 	 * Lookup NIP late so we have the best change of getting the
444 	 * above info out without failing
445 	 */
446 	printk("NIP ["REG"] ", regs->nip);
447 	print_symbol("%s\n", regs->nip);
448 	printk("LR ["REG"] ", regs->link);
449 	print_symbol("%s\n", regs->link);
450 #endif
451 	show_stack(current, (unsigned long *) regs->gpr[1]);
452 	if (!user_mode(regs))
453 		show_instructions(regs);
454 }
455 
456 void exit_thread(void)
457 {
458 	discard_lazy_cpu_state();
459 }
460 
461 void flush_thread(void)
462 {
463 #ifdef CONFIG_PPC64
464 	struct thread_info *t = current_thread_info();
465 
466 	if (test_ti_thread_flag(t, TIF_ABI_PENDING)) {
467 		clear_ti_thread_flag(t, TIF_ABI_PENDING);
468 		if (test_ti_thread_flag(t, TIF_32BIT))
469 			clear_ti_thread_flag(t, TIF_32BIT);
470 		else
471 			set_ti_thread_flag(t, TIF_32BIT);
472 	}
473 #endif
474 
475 	discard_lazy_cpu_state();
476 
477 #ifdef CONFIG_PPC64	/* for now */
478 	if (current->thread.dabr) {
479 		current->thread.dabr = 0;
480 		set_dabr(0);
481 	}
482 #endif
483 }
484 
485 void
486 release_thread(struct task_struct *t)
487 {
488 }
489 
490 /*
491  * This gets called before we allocate a new thread and copy
492  * the current task into it.
493  */
494 void prepare_to_copy(struct task_struct *tsk)
495 {
496 	flush_fp_to_thread(current);
497 	flush_altivec_to_thread(current);
498 	flush_spe_to_thread(current);
499 }
500 
501 /*
502  * Copy a thread..
503  */
504 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
505 		unsigned long unused, struct task_struct *p,
506 		struct pt_regs *regs)
507 {
508 	struct pt_regs *childregs, *kregs;
509 	extern void ret_from_fork(void);
510 	unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
511 
512 	CHECK_FULL_REGS(regs);
513 	/* Copy registers */
514 	sp -= sizeof(struct pt_regs);
515 	childregs = (struct pt_regs *) sp;
516 	*childregs = *regs;
517 	if ((childregs->msr & MSR_PR) == 0) {
518 		/* for kernel thread, set `current' and stackptr in new task */
519 		childregs->gpr[1] = sp + sizeof(struct pt_regs);
520 #ifdef CONFIG_PPC32
521 		childregs->gpr[2] = (unsigned long) p;
522 #else
523 		clear_tsk_thread_flag(p, TIF_32BIT);
524 #endif
525 		p->thread.regs = NULL;	/* no user register state */
526 	} else {
527 		childregs->gpr[1] = usp;
528 		p->thread.regs = childregs;
529 		if (clone_flags & CLONE_SETTLS) {
530 #ifdef CONFIG_PPC64
531 			if (!test_thread_flag(TIF_32BIT))
532 				childregs->gpr[13] = childregs->gpr[6];
533 			else
534 #endif
535 				childregs->gpr[2] = childregs->gpr[6];
536 		}
537 	}
538 	childregs->gpr[3] = 0;  /* Result from fork() */
539 	sp -= STACK_FRAME_OVERHEAD;
540 
541 	/*
542 	 * The way this works is that at some point in the future
543 	 * some task will call _switch to switch to the new task.
544 	 * That will pop off the stack frame created below and start
545 	 * the new task running at ret_from_fork.  The new task will
546 	 * do some house keeping and then return from the fork or clone
547 	 * system call, using the stack frame created above.
548 	 */
549 	sp -= sizeof(struct pt_regs);
550 	kregs = (struct pt_regs *) sp;
551 	sp -= STACK_FRAME_OVERHEAD;
552 	p->thread.ksp = sp;
553 
554 #ifdef CONFIG_PPC64
555 	if (cpu_has_feature(CPU_FTR_SLB)) {
556 		unsigned long sp_vsid = get_kernel_vsid(sp);
557 		unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
558 
559 		sp_vsid <<= SLB_VSID_SHIFT;
560 		sp_vsid |= SLB_VSID_KERNEL | llp;
561 		p->thread.ksp_vsid = sp_vsid;
562 	}
563 
564 	/*
565 	 * The PPC64 ABI makes use of a TOC to contain function
566 	 * pointers.  The function (ret_from_except) is actually a pointer
567 	 * to the TOC entry.  The first entry is a pointer to the actual
568 	 * function.
569  	 */
570 	kregs->nip = *((unsigned long *)ret_from_fork);
571 #else
572 	kregs->nip = (unsigned long)ret_from_fork;
573 #endif
574 
575 	return 0;
576 }
577 
578 /*
579  * Set up a thread for executing a new program
580  */
581 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
582 {
583 #ifdef CONFIG_PPC64
584 	unsigned long load_addr = regs->gpr[2];	/* saved by ELF_PLAT_INIT */
585 #endif
586 
587 	set_fs(USER_DS);
588 
589 	/*
590 	 * If we exec out of a kernel thread then thread.regs will not be
591 	 * set.  Do it now.
592 	 */
593 	if (!current->thread.regs) {
594 		struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
595 		current->thread.regs = regs - 1;
596 	}
597 
598 	memset(regs->gpr, 0, sizeof(regs->gpr));
599 	regs->ctr = 0;
600 	regs->link = 0;
601 	regs->xer = 0;
602 	regs->ccr = 0;
603 	regs->gpr[1] = sp;
604 
605 #ifdef CONFIG_PPC32
606 	regs->mq = 0;
607 	regs->nip = start;
608 	regs->msr = MSR_USER;
609 #else
610 	if (!test_thread_flag(TIF_32BIT)) {
611 		unsigned long entry, toc;
612 
613 		/* start is a relocated pointer to the function descriptor for
614 		 * the elf _start routine.  The first entry in the function
615 		 * descriptor is the entry address of _start and the second
616 		 * entry is the TOC value we need to use.
617 		 */
618 		__get_user(entry, (unsigned long __user *)start);
619 		__get_user(toc, (unsigned long __user *)start+1);
620 
621 		/* Check whether the e_entry function descriptor entries
622 		 * need to be relocated before we can use them.
623 		 */
624 		if (load_addr != 0) {
625 			entry += load_addr;
626 			toc   += load_addr;
627 		}
628 		regs->nip = entry;
629 		regs->gpr[2] = toc;
630 		regs->msr = MSR_USER64;
631 	} else {
632 		regs->nip = start;
633 		regs->gpr[2] = 0;
634 		regs->msr = MSR_USER32;
635 	}
636 #endif
637 
638 	discard_lazy_cpu_state();
639 	memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
640 	current->thread.fpscr.val = 0;
641 #ifdef CONFIG_ALTIVEC
642 	memset(current->thread.vr, 0, sizeof(current->thread.vr));
643 	memset(&current->thread.vscr, 0, sizeof(current->thread.vscr));
644 	current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
645 	current->thread.vrsave = 0;
646 	current->thread.used_vr = 0;
647 #endif /* CONFIG_ALTIVEC */
648 #ifdef CONFIG_SPE
649 	memset(current->thread.evr, 0, sizeof(current->thread.evr));
650 	current->thread.acc = 0;
651 	current->thread.spefscr = 0;
652 	current->thread.used_spe = 0;
653 #endif /* CONFIG_SPE */
654 }
655 
656 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
657 		| PR_FP_EXC_RES | PR_FP_EXC_INV)
658 
659 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
660 {
661 	struct pt_regs *regs = tsk->thread.regs;
662 
663 	/* This is a bit hairy.  If we are an SPE enabled  processor
664 	 * (have embedded fp) we store the IEEE exception enable flags in
665 	 * fpexc_mode.  fpexc_mode is also used for setting FP exception
666 	 * mode (asyn, precise, disabled) for 'Classic' FP. */
667 	if (val & PR_FP_EXC_SW_ENABLE) {
668 #ifdef CONFIG_SPE
669 		tsk->thread.fpexc_mode = val &
670 			(PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
671 		return 0;
672 #else
673 		return -EINVAL;
674 #endif
675 	}
676 
677 	/* on a CONFIG_SPE this does not hurt us.  The bits that
678 	 * __pack_fe01 use do not overlap with bits used for
679 	 * PR_FP_EXC_SW_ENABLE.  Additionally, the MSR[FE0,FE1] bits
680 	 * on CONFIG_SPE implementations are reserved so writing to
681 	 * them does not change anything */
682 	if (val > PR_FP_EXC_PRECISE)
683 		return -EINVAL;
684 	tsk->thread.fpexc_mode = __pack_fe01(val);
685 	if (regs != NULL && (regs->msr & MSR_FP) != 0)
686 		regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
687 			| tsk->thread.fpexc_mode;
688 	return 0;
689 }
690 
691 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
692 {
693 	unsigned int val;
694 
695 	if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
696 #ifdef CONFIG_SPE
697 		val = tsk->thread.fpexc_mode;
698 #else
699 		return -EINVAL;
700 #endif
701 	else
702 		val = __unpack_fe01(tsk->thread.fpexc_mode);
703 	return put_user(val, (unsigned int __user *) adr);
704 }
705 
706 int set_endian(struct task_struct *tsk, unsigned int val)
707 {
708 	struct pt_regs *regs = tsk->thread.regs;
709 
710 	if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
711 	    (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
712 		return -EINVAL;
713 
714 	if (regs == NULL)
715 		return -EINVAL;
716 
717 	if (val == PR_ENDIAN_BIG)
718 		regs->msr &= ~MSR_LE;
719 	else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
720 		regs->msr |= MSR_LE;
721 	else
722 		return -EINVAL;
723 
724 	return 0;
725 }
726 
727 int get_endian(struct task_struct *tsk, unsigned long adr)
728 {
729 	struct pt_regs *regs = tsk->thread.regs;
730 	unsigned int val;
731 
732 	if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
733 	    !cpu_has_feature(CPU_FTR_REAL_LE))
734 		return -EINVAL;
735 
736 	if (regs == NULL)
737 		return -EINVAL;
738 
739 	if (regs->msr & MSR_LE) {
740 		if (cpu_has_feature(CPU_FTR_REAL_LE))
741 			val = PR_ENDIAN_LITTLE;
742 		else
743 			val = PR_ENDIAN_PPC_LITTLE;
744 	} else
745 		val = PR_ENDIAN_BIG;
746 
747 	return put_user(val, (unsigned int __user *)adr);
748 }
749 
750 int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
751 {
752 	tsk->thread.align_ctl = val;
753 	return 0;
754 }
755 
756 int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
757 {
758 	return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
759 }
760 
761 #define TRUNC_PTR(x)	((typeof(x))(((unsigned long)(x)) & 0xffffffff))
762 
763 int sys_clone(unsigned long clone_flags, unsigned long usp,
764 	      int __user *parent_tidp, void __user *child_threadptr,
765 	      int __user *child_tidp, int p6,
766 	      struct pt_regs *regs)
767 {
768 	CHECK_FULL_REGS(regs);
769 	if (usp == 0)
770 		usp = regs->gpr[1];	/* stack pointer for child */
771 #ifdef CONFIG_PPC64
772 	if (test_thread_flag(TIF_32BIT)) {
773 		parent_tidp = TRUNC_PTR(parent_tidp);
774 		child_tidp = TRUNC_PTR(child_tidp);
775 	}
776 #endif
777  	return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
778 }
779 
780 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
781 	     unsigned long p4, unsigned long p5, unsigned long p6,
782 	     struct pt_regs *regs)
783 {
784 	CHECK_FULL_REGS(regs);
785 	return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
786 }
787 
788 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
789 	      unsigned long p4, unsigned long p5, unsigned long p6,
790 	      struct pt_regs *regs)
791 {
792 	CHECK_FULL_REGS(regs);
793 	return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
794 			regs, 0, NULL, NULL);
795 }
796 
797 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
798 	       unsigned long a3, unsigned long a4, unsigned long a5,
799 	       struct pt_regs *regs)
800 {
801 	int error;
802 	char *filename;
803 
804 	filename = getname((char __user *) a0);
805 	error = PTR_ERR(filename);
806 	if (IS_ERR(filename))
807 		goto out;
808 	flush_fp_to_thread(current);
809 	flush_altivec_to_thread(current);
810 	flush_spe_to_thread(current);
811 	error = do_execve(filename, (char __user * __user *) a1,
812 			  (char __user * __user *) a2, regs);
813 	if (error == 0) {
814 		task_lock(current);
815 		current->ptrace &= ~PT_DTRACE;
816 		task_unlock(current);
817 	}
818 	putname(filename);
819 out:
820 	return error;
821 }
822 
823 #ifdef CONFIG_IRQSTACKS
824 static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
825 				  unsigned long nbytes)
826 {
827 	unsigned long stack_page;
828 	unsigned long cpu = task_cpu(p);
829 
830 	/*
831 	 * Avoid crashing if the stack has overflowed and corrupted
832 	 * task_cpu(p), which is in the thread_info struct.
833 	 */
834 	if (cpu < NR_CPUS && cpu_possible(cpu)) {
835 		stack_page = (unsigned long) hardirq_ctx[cpu];
836 		if (sp >= stack_page + sizeof(struct thread_struct)
837 		    && sp <= stack_page + THREAD_SIZE - nbytes)
838 			return 1;
839 
840 		stack_page = (unsigned long) softirq_ctx[cpu];
841 		if (sp >= stack_page + sizeof(struct thread_struct)
842 		    && sp <= stack_page + THREAD_SIZE - nbytes)
843 			return 1;
844 	}
845 	return 0;
846 }
847 
848 #else
849 #define valid_irq_stack(sp, p, nb)	0
850 #endif /* CONFIG_IRQSTACKS */
851 
852 int validate_sp(unsigned long sp, struct task_struct *p,
853 		       unsigned long nbytes)
854 {
855 	unsigned long stack_page = (unsigned long)task_stack_page(p);
856 
857 	if (sp >= stack_page + sizeof(struct thread_struct)
858 	    && sp <= stack_page + THREAD_SIZE - nbytes)
859 		return 1;
860 
861 	return valid_irq_stack(sp, p, nbytes);
862 }
863 
864 #ifdef CONFIG_PPC64
865 #define MIN_STACK_FRAME	112	/* same as STACK_FRAME_OVERHEAD, in fact */
866 #define FRAME_LR_SAVE	2
867 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
868 #define REGS_MARKER	0x7265677368657265ul
869 #define FRAME_MARKER	12
870 #else
871 #define MIN_STACK_FRAME	16
872 #define FRAME_LR_SAVE	1
873 #define INT_FRAME_SIZE	(sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
874 #define REGS_MARKER	0x72656773ul
875 #define FRAME_MARKER	2
876 #endif
877 
878 EXPORT_SYMBOL(validate_sp);
879 
880 unsigned long get_wchan(struct task_struct *p)
881 {
882 	unsigned long ip, sp;
883 	int count = 0;
884 
885 	if (!p || p == current || p->state == TASK_RUNNING)
886 		return 0;
887 
888 	sp = p->thread.ksp;
889 	if (!validate_sp(sp, p, MIN_STACK_FRAME))
890 		return 0;
891 
892 	do {
893 		sp = *(unsigned long *)sp;
894 		if (!validate_sp(sp, p, MIN_STACK_FRAME))
895 			return 0;
896 		if (count > 0) {
897 			ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
898 			if (!in_sched_functions(ip))
899 				return ip;
900 		}
901 	} while (count++ < 16);
902 	return 0;
903 }
904 
905 static int kstack_depth_to_print = 64;
906 
907 void show_stack(struct task_struct *tsk, unsigned long *stack)
908 {
909 	unsigned long sp, ip, lr, newsp;
910 	int count = 0;
911 	int firstframe = 1;
912 
913 	sp = (unsigned long) stack;
914 	if (tsk == NULL)
915 		tsk = current;
916 	if (sp == 0) {
917 		if (tsk == current)
918 			asm("mr %0,1" : "=r" (sp));
919 		else
920 			sp = tsk->thread.ksp;
921 	}
922 
923 	lr = 0;
924 	printk("Call Trace:\n");
925 	do {
926 		if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
927 			return;
928 
929 		stack = (unsigned long *) sp;
930 		newsp = stack[0];
931 		ip = stack[FRAME_LR_SAVE];
932 		if (!firstframe || ip != lr) {
933 			printk("["REG"] ["REG"] ", sp, ip);
934 			print_symbol("%s", ip);
935 			if (firstframe)
936 				printk(" (unreliable)");
937 			printk("\n");
938 		}
939 		firstframe = 0;
940 
941 		/*
942 		 * See if this is an exception frame.
943 		 * We look for the "regshere" marker in the current frame.
944 		 */
945 		if (validate_sp(sp, tsk, INT_FRAME_SIZE)
946 		    && stack[FRAME_MARKER] == REGS_MARKER) {
947 			struct pt_regs *regs = (struct pt_regs *)
948 				(sp + STACK_FRAME_OVERHEAD);
949 			printk("--- Exception: %lx", regs->trap);
950 			print_symbol(" at %s\n", regs->nip);
951 			lr = regs->link;
952 			print_symbol("    LR = %s\n", lr);
953 			firstframe = 1;
954 		}
955 
956 		sp = newsp;
957 	} while (count++ < kstack_depth_to_print);
958 }
959 
960 void dump_stack(void)
961 {
962 	show_stack(current, NULL);
963 }
964 EXPORT_SYMBOL(dump_stack);
965 
966 #ifdef CONFIG_PPC64
967 void ppc64_runlatch_on(void)
968 {
969 	unsigned long ctrl;
970 
971 	if (cpu_has_feature(CPU_FTR_CTRL) && !test_thread_flag(TIF_RUNLATCH)) {
972 		HMT_medium();
973 
974 		ctrl = mfspr(SPRN_CTRLF);
975 		ctrl |= CTRL_RUNLATCH;
976 		mtspr(SPRN_CTRLT, ctrl);
977 
978 		set_thread_flag(TIF_RUNLATCH);
979 	}
980 }
981 
982 void ppc64_runlatch_off(void)
983 {
984 	unsigned long ctrl;
985 
986 	if (cpu_has_feature(CPU_FTR_CTRL) && test_thread_flag(TIF_RUNLATCH)) {
987 		HMT_medium();
988 
989 		clear_thread_flag(TIF_RUNLATCH);
990 
991 		ctrl = mfspr(SPRN_CTRLF);
992 		ctrl &= ~CTRL_RUNLATCH;
993 		mtspr(SPRN_CTRLT, ctrl);
994 	}
995 }
996 #endif
997