xref: /openbmc/linux/arch/powerpc/kernel/process.c (revision cdfce539)
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/stddef.h>
23 #include <linux/unistd.h>
24 #include <linux/ptrace.h>
25 #include <linux/slab.h>
26 #include <linux/user.h>
27 #include <linux/elf.h>
28 #include <linux/init.h>
29 #include <linux/prctl.h>
30 #include <linux/init_task.h>
31 #include <linux/export.h>
32 #include <linux/kallsyms.h>
33 #include <linux/mqueue.h>
34 #include <linux/hardirq.h>
35 #include <linux/utsname.h>
36 #include <linux/ftrace.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/personality.h>
39 #include <linux/random.h>
40 #include <linux/hw_breakpoint.h>
41 
42 #include <asm/pgtable.h>
43 #include <asm/uaccess.h>
44 #include <asm/io.h>
45 #include <asm/processor.h>
46 #include <asm/mmu.h>
47 #include <asm/prom.h>
48 #include <asm/machdep.h>
49 #include <asm/time.h>
50 #include <asm/runlatch.h>
51 #include <asm/syscalls.h>
52 #include <asm/switch_to.h>
53 #include <asm/tm.h>
54 #include <asm/debug.h>
55 #ifdef CONFIG_PPC64
56 #include <asm/firmware.h>
57 #endif
58 #include <linux/kprobes.h>
59 #include <linux/kdebug.h>
60 
61 /* Transactional Memory debug */
62 #ifdef TM_DEBUG_SW
63 #define TM_DEBUG(x...) printk(KERN_INFO x)
64 #else
65 #define TM_DEBUG(x...) do { } while(0)
66 #endif
67 
68 extern unsigned long _get_SP(void);
69 
70 #ifndef CONFIG_SMP
71 struct task_struct *last_task_used_math = NULL;
72 struct task_struct *last_task_used_altivec = NULL;
73 struct task_struct *last_task_used_vsx = NULL;
74 struct task_struct *last_task_used_spe = NULL;
75 #endif
76 
77 /*
78  * Make sure the floating-point register state in the
79  * the thread_struct is up to date for task tsk.
80  */
81 void flush_fp_to_thread(struct task_struct *tsk)
82 {
83 	if (tsk->thread.regs) {
84 		/*
85 		 * We need to disable preemption here because if we didn't,
86 		 * another process could get scheduled after the regs->msr
87 		 * test but before we have finished saving the FP registers
88 		 * to the thread_struct.  That process could take over the
89 		 * FPU, and then when we get scheduled again we would store
90 		 * bogus values for the remaining FP registers.
91 		 */
92 		preempt_disable();
93 		if (tsk->thread.regs->msr & MSR_FP) {
94 #ifdef CONFIG_SMP
95 			/*
96 			 * This should only ever be called for current or
97 			 * for a stopped child process.  Since we save away
98 			 * the FP register state on context switch on SMP,
99 			 * there is something wrong if a stopped child appears
100 			 * to still have its FP state in the CPU registers.
101 			 */
102 			BUG_ON(tsk != current);
103 #endif
104 			giveup_fpu(tsk);
105 		}
106 		preempt_enable();
107 	}
108 }
109 EXPORT_SYMBOL_GPL(flush_fp_to_thread);
110 
111 void enable_kernel_fp(void)
112 {
113 	WARN_ON(preemptible());
114 
115 #ifdef CONFIG_SMP
116 	if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
117 		giveup_fpu(current);
118 	else
119 		giveup_fpu(NULL);	/* just enables FP for kernel */
120 #else
121 	giveup_fpu(last_task_used_math);
122 #endif /* CONFIG_SMP */
123 }
124 EXPORT_SYMBOL(enable_kernel_fp);
125 
126 #ifdef CONFIG_ALTIVEC
127 void enable_kernel_altivec(void)
128 {
129 	WARN_ON(preemptible());
130 
131 #ifdef CONFIG_SMP
132 	if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
133 		giveup_altivec(current);
134 	else
135 		giveup_altivec_notask();
136 #else
137 	giveup_altivec(last_task_used_altivec);
138 #endif /* CONFIG_SMP */
139 }
140 EXPORT_SYMBOL(enable_kernel_altivec);
141 
142 /*
143  * Make sure the VMX/Altivec register state in the
144  * the thread_struct is up to date for task tsk.
145  */
146 void flush_altivec_to_thread(struct task_struct *tsk)
147 {
148 	if (tsk->thread.regs) {
149 		preempt_disable();
150 		if (tsk->thread.regs->msr & MSR_VEC) {
151 #ifdef CONFIG_SMP
152 			BUG_ON(tsk != current);
153 #endif
154 			giveup_altivec(tsk);
155 		}
156 		preempt_enable();
157 	}
158 }
159 EXPORT_SYMBOL_GPL(flush_altivec_to_thread);
160 #endif /* CONFIG_ALTIVEC */
161 
162 #ifdef CONFIG_VSX
163 #if 0
164 /* not currently used, but some crazy RAID module might want to later */
165 void enable_kernel_vsx(void)
166 {
167 	WARN_ON(preemptible());
168 
169 #ifdef CONFIG_SMP
170 	if (current->thread.regs && (current->thread.regs->msr & MSR_VSX))
171 		giveup_vsx(current);
172 	else
173 		giveup_vsx(NULL);	/* just enable vsx for kernel - force */
174 #else
175 	giveup_vsx(last_task_used_vsx);
176 #endif /* CONFIG_SMP */
177 }
178 EXPORT_SYMBOL(enable_kernel_vsx);
179 #endif
180 
181 void giveup_vsx(struct task_struct *tsk)
182 {
183 	giveup_fpu(tsk);
184 	giveup_altivec(tsk);
185 	__giveup_vsx(tsk);
186 }
187 
188 void flush_vsx_to_thread(struct task_struct *tsk)
189 {
190 	if (tsk->thread.regs) {
191 		preempt_disable();
192 		if (tsk->thread.regs->msr & MSR_VSX) {
193 #ifdef CONFIG_SMP
194 			BUG_ON(tsk != current);
195 #endif
196 			giveup_vsx(tsk);
197 		}
198 		preempt_enable();
199 	}
200 }
201 EXPORT_SYMBOL_GPL(flush_vsx_to_thread);
202 #endif /* CONFIG_VSX */
203 
204 #ifdef CONFIG_SPE
205 
206 void enable_kernel_spe(void)
207 {
208 	WARN_ON(preemptible());
209 
210 #ifdef CONFIG_SMP
211 	if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
212 		giveup_spe(current);
213 	else
214 		giveup_spe(NULL);	/* just enable SPE for kernel - force */
215 #else
216 	giveup_spe(last_task_used_spe);
217 #endif /* __SMP __ */
218 }
219 EXPORT_SYMBOL(enable_kernel_spe);
220 
221 void flush_spe_to_thread(struct task_struct *tsk)
222 {
223 	if (tsk->thread.regs) {
224 		preempt_disable();
225 		if (tsk->thread.regs->msr & MSR_SPE) {
226 #ifdef CONFIG_SMP
227 			BUG_ON(tsk != current);
228 #endif
229 			tsk->thread.spefscr = mfspr(SPRN_SPEFSCR);
230 			giveup_spe(tsk);
231 		}
232 		preempt_enable();
233 	}
234 }
235 #endif /* CONFIG_SPE */
236 
237 #ifndef CONFIG_SMP
238 /*
239  * If we are doing lazy switching of CPU state (FP, altivec or SPE),
240  * and the current task has some state, discard it.
241  */
242 void discard_lazy_cpu_state(void)
243 {
244 	preempt_disable();
245 	if (last_task_used_math == current)
246 		last_task_used_math = NULL;
247 #ifdef CONFIG_ALTIVEC
248 	if (last_task_used_altivec == current)
249 		last_task_used_altivec = NULL;
250 #endif /* CONFIG_ALTIVEC */
251 #ifdef CONFIG_VSX
252 	if (last_task_used_vsx == current)
253 		last_task_used_vsx = NULL;
254 #endif /* CONFIG_VSX */
255 #ifdef CONFIG_SPE
256 	if (last_task_used_spe == current)
257 		last_task_used_spe = NULL;
258 #endif
259 	preempt_enable();
260 }
261 #endif /* CONFIG_SMP */
262 
263 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
264 void do_send_trap(struct pt_regs *regs, unsigned long address,
265 		  unsigned long error_code, int signal_code, int breakpt)
266 {
267 	siginfo_t info;
268 
269 	current->thread.trap_nr = signal_code;
270 	if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
271 			11, SIGSEGV) == NOTIFY_STOP)
272 		return;
273 
274 	/* Deliver the signal to userspace */
275 	info.si_signo = SIGTRAP;
276 	info.si_errno = breakpt;	/* breakpoint or watchpoint id */
277 	info.si_code = signal_code;
278 	info.si_addr = (void __user *)address;
279 	force_sig_info(SIGTRAP, &info, current);
280 }
281 #else	/* !CONFIG_PPC_ADV_DEBUG_REGS */
282 void do_break (struct pt_regs *regs, unsigned long address,
283 		    unsigned long error_code)
284 {
285 	siginfo_t info;
286 
287 	current->thread.trap_nr = TRAP_HWBKPT;
288 	if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
289 			11, SIGSEGV) == NOTIFY_STOP)
290 		return;
291 
292 	if (debugger_break_match(regs))
293 		return;
294 
295 	/* Clear the breakpoint */
296 	hw_breakpoint_disable();
297 
298 	/* Deliver the signal to userspace */
299 	info.si_signo = SIGTRAP;
300 	info.si_errno = 0;
301 	info.si_code = TRAP_HWBKPT;
302 	info.si_addr = (void __user *)address;
303 	force_sig_info(SIGTRAP, &info, current);
304 }
305 #endif	/* CONFIG_PPC_ADV_DEBUG_REGS */
306 
307 static DEFINE_PER_CPU(struct arch_hw_breakpoint, current_brk);
308 
309 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
310 /*
311  * Set the debug registers back to their default "safe" values.
312  */
313 static void set_debug_reg_defaults(struct thread_struct *thread)
314 {
315 	thread->iac1 = thread->iac2 = 0;
316 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
317 	thread->iac3 = thread->iac4 = 0;
318 #endif
319 	thread->dac1 = thread->dac2 = 0;
320 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
321 	thread->dvc1 = thread->dvc2 = 0;
322 #endif
323 	thread->dbcr0 = 0;
324 #ifdef CONFIG_BOOKE
325 	/*
326 	 * Force User/Supervisor bits to b11 (user-only MSR[PR]=1)
327 	 */
328 	thread->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US |	\
329 			DBCR1_IAC3US | DBCR1_IAC4US;
330 	/*
331 	 * Force Data Address Compare User/Supervisor bits to be User-only
332 	 * (0b11 MSR[PR]=1) and set all other bits in DBCR2 register to be 0.
333 	 */
334 	thread->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
335 #else
336 	thread->dbcr1 = 0;
337 #endif
338 }
339 
340 static void prime_debug_regs(struct thread_struct *thread)
341 {
342 	/*
343 	 * We could have inherited MSR_DE from userspace, since
344 	 * it doesn't get cleared on exception entry.  Make sure
345 	 * MSR_DE is clear before we enable any debug events.
346 	 */
347 	mtmsr(mfmsr() & ~MSR_DE);
348 
349 	mtspr(SPRN_IAC1, thread->iac1);
350 	mtspr(SPRN_IAC2, thread->iac2);
351 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
352 	mtspr(SPRN_IAC3, thread->iac3);
353 	mtspr(SPRN_IAC4, thread->iac4);
354 #endif
355 	mtspr(SPRN_DAC1, thread->dac1);
356 	mtspr(SPRN_DAC2, thread->dac2);
357 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
358 	mtspr(SPRN_DVC1, thread->dvc1);
359 	mtspr(SPRN_DVC2, thread->dvc2);
360 #endif
361 	mtspr(SPRN_DBCR0, thread->dbcr0);
362 	mtspr(SPRN_DBCR1, thread->dbcr1);
363 #ifdef CONFIG_BOOKE
364 	mtspr(SPRN_DBCR2, thread->dbcr2);
365 #endif
366 }
367 /*
368  * Unless neither the old or new thread are making use of the
369  * debug registers, set the debug registers from the values
370  * stored in the new thread.
371  */
372 static void switch_booke_debug_regs(struct thread_struct *new_thread)
373 {
374 	if ((current->thread.dbcr0 & DBCR0_IDM)
375 		|| (new_thread->dbcr0 & DBCR0_IDM))
376 			prime_debug_regs(new_thread);
377 }
378 #else	/* !CONFIG_PPC_ADV_DEBUG_REGS */
379 #ifndef CONFIG_HAVE_HW_BREAKPOINT
380 static void set_debug_reg_defaults(struct thread_struct *thread)
381 {
382 	thread->hw_brk.address = 0;
383 	thread->hw_brk.type = 0;
384 	set_breakpoint(&thread->hw_brk);
385 }
386 #endif /* !CONFIG_HAVE_HW_BREAKPOINT */
387 #endif	/* CONFIG_PPC_ADV_DEBUG_REGS */
388 
389 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
390 static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
391 {
392 	mtspr(SPRN_DAC1, dabr);
393 #ifdef CONFIG_PPC_47x
394 	isync();
395 #endif
396 	return 0;
397 }
398 #elif defined(CONFIG_PPC_BOOK3S)
399 static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
400 {
401 	mtspr(SPRN_DABR, dabr);
402 	mtspr(SPRN_DABRX, dabrx);
403 	return 0;
404 }
405 #else
406 static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
407 {
408 	return -EINVAL;
409 }
410 #endif
411 
412 static inline int set_dabr(struct arch_hw_breakpoint *brk)
413 {
414 	unsigned long dabr, dabrx;
415 
416 	dabr = brk->address | (brk->type & HW_BRK_TYPE_DABR);
417 	dabrx = ((brk->type >> 3) & 0x7);
418 
419 	if (ppc_md.set_dabr)
420 		return ppc_md.set_dabr(dabr, dabrx);
421 
422 	return __set_dabr(dabr, dabrx);
423 }
424 
425 static inline int set_dawr(struct arch_hw_breakpoint *brk)
426 {
427 	unsigned long dawr, dawrx, mrd;
428 
429 	dawr = brk->address;
430 
431 	dawrx  = (brk->type & (HW_BRK_TYPE_READ | HW_BRK_TYPE_WRITE)) \
432 		                   << (63 - 58); //* read/write bits */
433 	dawrx |= ((brk->type & (HW_BRK_TYPE_TRANSLATE)) >> 2) \
434 		                   << (63 - 59); //* translate */
435 	dawrx |= (brk->type & (HW_BRK_TYPE_PRIV_ALL)) \
436 		                   >> 3; //* PRIM bits */
437 	/* dawr length is stored in field MDR bits 48:53.  Matches range in
438 	   doublewords (64 bits) baised by -1 eg. 0b000000=1DW and
439 	   0b111111=64DW.
440 	   brk->len is in bytes.
441 	   This aligns up to double word size, shifts and does the bias.
442 	*/
443 	mrd = ((brk->len + 7) >> 3) - 1;
444 	dawrx |= (mrd & 0x3f) << (63 - 53);
445 
446 	if (ppc_md.set_dawr)
447 		return ppc_md.set_dawr(dawr, dawrx);
448 	mtspr(SPRN_DAWR, dawr);
449 	mtspr(SPRN_DAWRX, dawrx);
450 	return 0;
451 }
452 
453 int set_breakpoint(struct arch_hw_breakpoint *brk)
454 {
455 	__get_cpu_var(current_brk) = *brk;
456 
457 	if (cpu_has_feature(CPU_FTR_DAWR))
458 		return set_dawr(brk);
459 
460 	return set_dabr(brk);
461 }
462 
463 #ifdef CONFIG_PPC64
464 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
465 #endif
466 
467 static inline bool hw_brk_match(struct arch_hw_breakpoint *a,
468 			      struct arch_hw_breakpoint *b)
469 {
470 	if (a->address != b->address)
471 		return false;
472 	if (a->type != b->type)
473 		return false;
474 	if (a->len != b->len)
475 		return false;
476 	return true;
477 }
478 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
479 static inline void tm_reclaim_task(struct task_struct *tsk)
480 {
481 	/* We have to work out if we're switching from/to a task that's in the
482 	 * middle of a transaction.
483 	 *
484 	 * In switching we need to maintain a 2nd register state as
485 	 * oldtask->thread.ckpt_regs.  We tm_reclaim(oldproc); this saves the
486 	 * checkpointed (tbegin) state in ckpt_regs and saves the transactional
487 	 * (current) FPRs into oldtask->thread.transact_fpr[].
488 	 *
489 	 * We also context switch (save) TFHAR/TEXASR/TFIAR in here.
490 	 */
491 	struct thread_struct *thr = &tsk->thread;
492 
493 	if (!thr->regs)
494 		return;
495 
496 	if (!MSR_TM_ACTIVE(thr->regs->msr))
497 		goto out_and_saveregs;
498 
499 	/* Stash the original thread MSR, as giveup_fpu et al will
500 	 * modify it.  We hold onto it to see whether the task used
501 	 * FP & vector regs.
502 	 */
503 	thr->tm_orig_msr = thr->regs->msr;
504 
505 	TM_DEBUG("--- tm_reclaim on pid %d (NIP=%lx, "
506 		 "ccr=%lx, msr=%lx, trap=%lx)\n",
507 		 tsk->pid, thr->regs->nip,
508 		 thr->regs->ccr, thr->regs->msr,
509 		 thr->regs->trap);
510 
511 	tm_reclaim(thr, thr->regs->msr, TM_CAUSE_RESCHED);
512 
513 	TM_DEBUG("--- tm_reclaim on pid %d complete\n",
514 		 tsk->pid);
515 
516 out_and_saveregs:
517 	/* Always save the regs here, even if a transaction's not active.
518 	 * This context-switches a thread's TM info SPRs.  We do it here to
519 	 * be consistent with the restore path (in recheckpoint) which
520 	 * cannot happen later in _switch().
521 	 */
522 	tm_save_sprs(thr);
523 }
524 
525 static inline void tm_recheckpoint_new_task(struct task_struct *new)
526 {
527 	unsigned long msr;
528 
529 	if (!cpu_has_feature(CPU_FTR_TM))
530 		return;
531 
532 	/* Recheckpoint the registers of the thread we're about to switch to.
533 	 *
534 	 * If the task was using FP, we non-lazily reload both the original and
535 	 * the speculative FP register states.  This is because the kernel
536 	 * doesn't see if/when a TM rollback occurs, so if we take an FP
537 	 * unavoidable later, we are unable to determine which set of FP regs
538 	 * need to be restored.
539 	 */
540 	if (!new->thread.regs)
541 		return;
542 
543 	/* The TM SPRs are restored here, so that TEXASR.FS can be set
544 	 * before the trecheckpoint and no explosion occurs.
545 	 */
546 	tm_restore_sprs(&new->thread);
547 
548 	if (!MSR_TM_ACTIVE(new->thread.regs->msr))
549 		return;
550 	msr = new->thread.tm_orig_msr;
551 	/* Recheckpoint to restore original checkpointed register state. */
552 	TM_DEBUG("*** tm_recheckpoint of pid %d "
553 		 "(new->msr 0x%lx, new->origmsr 0x%lx)\n",
554 		 new->pid, new->thread.regs->msr, msr);
555 
556 	/* This loads the checkpointed FP/VEC state, if used */
557 	tm_recheckpoint(&new->thread, msr);
558 
559 	/* This loads the speculative FP/VEC state, if used */
560 	if (msr & MSR_FP) {
561 		do_load_up_transact_fpu(&new->thread);
562 		new->thread.regs->msr |=
563 			(MSR_FP | new->thread.fpexc_mode);
564 	}
565 #ifdef CONFIG_ALTIVEC
566 	if (msr & MSR_VEC) {
567 		do_load_up_transact_altivec(&new->thread);
568 		new->thread.regs->msr |= MSR_VEC;
569 	}
570 #endif
571 	/* We may as well turn on VSX too since all the state is restored now */
572 	if (msr & MSR_VSX)
573 		new->thread.regs->msr |= MSR_VSX;
574 
575 	TM_DEBUG("*** tm_recheckpoint of pid %d complete "
576 		 "(kernel msr 0x%lx)\n",
577 		 new->pid, mfmsr());
578 }
579 
580 static inline void __switch_to_tm(struct task_struct *prev)
581 {
582 	if (cpu_has_feature(CPU_FTR_TM)) {
583 		tm_enable();
584 		tm_reclaim_task(prev);
585 	}
586 }
587 #else
588 #define tm_recheckpoint_new_task(new)
589 #define __switch_to_tm(prev)
590 #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
591 
592 struct task_struct *__switch_to(struct task_struct *prev,
593 	struct task_struct *new)
594 {
595 	struct thread_struct *new_thread, *old_thread;
596 	unsigned long flags;
597 	struct task_struct *last;
598 #ifdef CONFIG_PPC_BOOK3S_64
599 	struct ppc64_tlb_batch *batch;
600 #endif
601 
602 	__switch_to_tm(prev);
603 
604 #ifdef CONFIG_SMP
605 	/* avoid complexity of lazy save/restore of fpu
606 	 * by just saving it every time we switch out if
607 	 * this task used the fpu during the last quantum.
608 	 *
609 	 * If it tries to use the fpu again, it'll trap and
610 	 * reload its fp regs.  So we don't have to do a restore
611 	 * every switch, just a save.
612 	 *  -- Cort
613 	 */
614 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
615 		giveup_fpu(prev);
616 #ifdef CONFIG_ALTIVEC
617 	/*
618 	 * If the previous thread used altivec in the last quantum
619 	 * (thus changing altivec regs) then save them.
620 	 * We used to check the VRSAVE register but not all apps
621 	 * set it, so we don't rely on it now (and in fact we need
622 	 * to save & restore VSCR even if VRSAVE == 0).  -- paulus
623 	 *
624 	 * On SMP we always save/restore altivec regs just to avoid the
625 	 * complexity of changing processors.
626 	 *  -- Cort
627 	 */
628 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
629 		giveup_altivec(prev);
630 #endif /* CONFIG_ALTIVEC */
631 #ifdef CONFIG_VSX
632 	if (prev->thread.regs && (prev->thread.regs->msr & MSR_VSX))
633 		/* VMX and FPU registers are already save here */
634 		__giveup_vsx(prev);
635 #endif /* CONFIG_VSX */
636 #ifdef CONFIG_SPE
637 	/*
638 	 * If the previous thread used spe in the last quantum
639 	 * (thus changing spe regs) then save them.
640 	 *
641 	 * On SMP we always save/restore spe regs just to avoid the
642 	 * complexity of changing processors.
643 	 */
644 	if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
645 		giveup_spe(prev);
646 #endif /* CONFIG_SPE */
647 
648 #else  /* CONFIG_SMP */
649 #ifdef CONFIG_ALTIVEC
650 	/* Avoid the trap.  On smp this this never happens since
651 	 * we don't set last_task_used_altivec -- Cort
652 	 */
653 	if (new->thread.regs && last_task_used_altivec == new)
654 		new->thread.regs->msr |= MSR_VEC;
655 #endif /* CONFIG_ALTIVEC */
656 #ifdef CONFIG_VSX
657 	if (new->thread.regs && last_task_used_vsx == new)
658 		new->thread.regs->msr |= MSR_VSX;
659 #endif /* CONFIG_VSX */
660 #ifdef CONFIG_SPE
661 	/* Avoid the trap.  On smp this this never happens since
662 	 * we don't set last_task_used_spe
663 	 */
664 	if (new->thread.regs && last_task_used_spe == new)
665 		new->thread.regs->msr |= MSR_SPE;
666 #endif /* CONFIG_SPE */
667 
668 #endif /* CONFIG_SMP */
669 
670 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
671 	switch_booke_debug_regs(&new->thread);
672 #else
673 /*
674  * For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
675  * schedule DABR
676  */
677 #ifndef CONFIG_HAVE_HW_BREAKPOINT
678 	if (unlikely(hw_brk_match(&__get_cpu_var(current_brk), &new->thread.hw_brk)))
679 		set_breakpoint(&new->thread.hw_brk);
680 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
681 #endif
682 
683 
684 	new_thread = &new->thread;
685 	old_thread = &current->thread;
686 
687 #ifdef CONFIG_PPC64
688 	/*
689 	 * Collect processor utilization data per process
690 	 */
691 	if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
692 		struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
693 		long unsigned start_tb, current_tb;
694 		start_tb = old_thread->start_tb;
695 		cu->current_tb = current_tb = mfspr(SPRN_PURR);
696 		old_thread->accum_tb += (current_tb - start_tb);
697 		new_thread->start_tb = current_tb;
698 	}
699 #endif /* CONFIG_PPC64 */
700 
701 #ifdef CONFIG_PPC_BOOK3S_64
702 	batch = &__get_cpu_var(ppc64_tlb_batch);
703 	if (batch->active) {
704 		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
705 		if (batch->index)
706 			__flush_tlb_pending(batch);
707 		batch->active = 0;
708 	}
709 #endif /* CONFIG_PPC_BOOK3S_64 */
710 
711 	local_irq_save(flags);
712 
713 	/*
714 	 * We can't take a PMU exception inside _switch() since there is a
715 	 * window where the kernel stack SLB and the kernel stack are out
716 	 * of sync. Hard disable here.
717 	 */
718 	hard_irq_disable();
719 
720 	tm_recheckpoint_new_task(new);
721 
722 	last = _switch(old_thread, new_thread);
723 
724 #ifdef CONFIG_PPC_BOOK3S_64
725 	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
726 		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
727 		batch = &__get_cpu_var(ppc64_tlb_batch);
728 		batch->active = 1;
729 	}
730 #endif /* CONFIG_PPC_BOOK3S_64 */
731 
732 	local_irq_restore(flags);
733 
734 	return last;
735 }
736 
737 static int instructions_to_print = 16;
738 
739 static void show_instructions(struct pt_regs *regs)
740 {
741 	int i;
742 	unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
743 			sizeof(int));
744 
745 	printk("Instruction dump:");
746 
747 	for (i = 0; i < instructions_to_print; i++) {
748 		int instr;
749 
750 		if (!(i % 8))
751 			printk("\n");
752 
753 #if !defined(CONFIG_BOOKE)
754 		/* If executing with the IMMU off, adjust pc rather
755 		 * than print XXXXXXXX.
756 		 */
757 		if (!(regs->msr & MSR_IR))
758 			pc = (unsigned long)phys_to_virt(pc);
759 #endif
760 
761 		/* We use __get_user here *only* to avoid an OOPS on a
762 		 * bad address because the pc *should* only be a
763 		 * kernel address.
764 		 */
765 		if (!__kernel_text_address(pc) ||
766 		     __get_user(instr, (unsigned int __user *)pc)) {
767 			printk(KERN_CONT "XXXXXXXX ");
768 		} else {
769 			if (regs->nip == pc)
770 				printk(KERN_CONT "<%08x> ", instr);
771 			else
772 				printk(KERN_CONT "%08x ", instr);
773 		}
774 
775 		pc += sizeof(int);
776 	}
777 
778 	printk("\n");
779 }
780 
781 static struct regbit {
782 	unsigned long bit;
783 	const char *name;
784 } msr_bits[] = {
785 #if defined(CONFIG_PPC64) && !defined(CONFIG_BOOKE)
786 	{MSR_SF,	"SF"},
787 	{MSR_HV,	"HV"},
788 #endif
789 	{MSR_VEC,	"VEC"},
790 	{MSR_VSX,	"VSX"},
791 #ifdef CONFIG_BOOKE
792 	{MSR_CE,	"CE"},
793 #endif
794 	{MSR_EE,	"EE"},
795 	{MSR_PR,	"PR"},
796 	{MSR_FP,	"FP"},
797 	{MSR_ME,	"ME"},
798 #ifdef CONFIG_BOOKE
799 	{MSR_DE,	"DE"},
800 #else
801 	{MSR_SE,	"SE"},
802 	{MSR_BE,	"BE"},
803 #endif
804 	{MSR_IR,	"IR"},
805 	{MSR_DR,	"DR"},
806 	{MSR_PMM,	"PMM"},
807 #ifndef CONFIG_BOOKE
808 	{MSR_RI,	"RI"},
809 	{MSR_LE,	"LE"},
810 #endif
811 	{0,		NULL}
812 };
813 
814 static void printbits(unsigned long val, struct regbit *bits)
815 {
816 	const char *sep = "";
817 
818 	printk("<");
819 	for (; bits->bit; ++bits)
820 		if (val & bits->bit) {
821 			printk("%s%s", sep, bits->name);
822 			sep = ",";
823 		}
824 	printk(">");
825 }
826 
827 #ifdef CONFIG_PPC64
828 #define REG		"%016lx"
829 #define REGS_PER_LINE	4
830 #define LAST_VOLATILE	13
831 #else
832 #define REG		"%08lx"
833 #define REGS_PER_LINE	8
834 #define LAST_VOLATILE	12
835 #endif
836 
837 void show_regs(struct pt_regs * regs)
838 {
839 	int i, trap;
840 
841 	show_regs_print_info(KERN_DEFAULT);
842 
843 	printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
844 	       regs->nip, regs->link, regs->ctr);
845 	printk("REGS: %p TRAP: %04lx   %s  (%s)\n",
846 	       regs, regs->trap, print_tainted(), init_utsname()->release);
847 	printk("MSR: "REG" ", regs->msr);
848 	printbits(regs->msr, msr_bits);
849 	printk("  CR: %08lx  XER: %08lx\n", regs->ccr, regs->xer);
850 #ifdef CONFIG_PPC64
851 	printk("SOFTE: %ld\n", regs->softe);
852 #endif
853 	trap = TRAP(regs);
854 	if ((regs->trap != 0xc00) && cpu_has_feature(CPU_FTR_CFAR))
855 		printk("CFAR: "REG"\n", regs->orig_gpr3);
856 	if (trap == 0x300 || trap == 0x600)
857 #if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)
858 		printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
859 #else
860 		printk("DAR: "REG", DSISR: %08lx\n", regs->dar, regs->dsisr);
861 #endif
862 
863 	for (i = 0;  i < 32;  i++) {
864 		if ((i % REGS_PER_LINE) == 0)
865 			printk("\nGPR%02d: ", i);
866 		printk(REG " ", regs->gpr[i]);
867 		if (i == LAST_VOLATILE && !FULL_REGS(regs))
868 			break;
869 	}
870 	printk("\n");
871 #ifdef CONFIG_KALLSYMS
872 	/*
873 	 * Lookup NIP late so we have the best change of getting the
874 	 * above info out without failing
875 	 */
876 	printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
877 	printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
878 #endif
879 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
880 	printk("PACATMSCRATCH [%llx]\n", get_paca()->tm_scratch);
881 #endif
882 	show_stack(current, (unsigned long *) regs->gpr[1]);
883 	if (!user_mode(regs))
884 		show_instructions(regs);
885 }
886 
887 void exit_thread(void)
888 {
889 	discard_lazy_cpu_state();
890 }
891 
892 void flush_thread(void)
893 {
894 	discard_lazy_cpu_state();
895 
896 #ifdef CONFIG_HAVE_HW_BREAKPOINT
897 	flush_ptrace_hw_breakpoint(current);
898 #else /* CONFIG_HAVE_HW_BREAKPOINT */
899 	set_debug_reg_defaults(&current->thread);
900 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
901 }
902 
903 void
904 release_thread(struct task_struct *t)
905 {
906 }
907 
908 /*
909  * this gets called so that we can store coprocessor state into memory and
910  * copy the current task into the new thread.
911  */
912 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
913 {
914 	flush_fp_to_thread(src);
915 	flush_altivec_to_thread(src);
916 	flush_vsx_to_thread(src);
917 	flush_spe_to_thread(src);
918 	*dst = *src;
919 	return 0;
920 }
921 
922 /*
923  * Copy a thread..
924  */
925 extern unsigned long dscr_default; /* defined in arch/powerpc/kernel/sysfs.c */
926 
927 int copy_thread(unsigned long clone_flags, unsigned long usp,
928 		unsigned long arg, struct task_struct *p)
929 {
930 	struct pt_regs *childregs, *kregs;
931 	extern void ret_from_fork(void);
932 	extern void ret_from_kernel_thread(void);
933 	void (*f)(void);
934 	unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
935 
936 	/* Copy registers */
937 	sp -= sizeof(struct pt_regs);
938 	childregs = (struct pt_regs *) sp;
939 	if (unlikely(p->flags & PF_KTHREAD)) {
940 		struct thread_info *ti = (void *)task_stack_page(p);
941 		memset(childregs, 0, sizeof(struct pt_regs));
942 		childregs->gpr[1] = sp + sizeof(struct pt_regs);
943 		childregs->gpr[14] = usp;	/* function */
944 #ifdef CONFIG_PPC64
945 		clear_tsk_thread_flag(p, TIF_32BIT);
946 		childregs->softe = 1;
947 #endif
948 		childregs->gpr[15] = arg;
949 		p->thread.regs = NULL;	/* no user register state */
950 		ti->flags |= _TIF_RESTOREALL;
951 		f = ret_from_kernel_thread;
952 	} else {
953 		struct pt_regs *regs = current_pt_regs();
954 		CHECK_FULL_REGS(regs);
955 		*childregs = *regs;
956 		if (usp)
957 			childregs->gpr[1] = usp;
958 		p->thread.regs = childregs;
959 		childregs->gpr[3] = 0;  /* Result from fork() */
960 		if (clone_flags & CLONE_SETTLS) {
961 #ifdef CONFIG_PPC64
962 			if (!is_32bit_task())
963 				childregs->gpr[13] = childregs->gpr[6];
964 			else
965 #endif
966 				childregs->gpr[2] = childregs->gpr[6];
967 		}
968 
969 		f = ret_from_fork;
970 	}
971 	sp -= STACK_FRAME_OVERHEAD;
972 
973 	/*
974 	 * The way this works is that at some point in the future
975 	 * some task will call _switch to switch to the new task.
976 	 * That will pop off the stack frame created below and start
977 	 * the new task running at ret_from_fork.  The new task will
978 	 * do some house keeping and then return from the fork or clone
979 	 * system call, using the stack frame created above.
980 	 */
981 	((unsigned long *)sp)[0] = 0;
982 	sp -= sizeof(struct pt_regs);
983 	kregs = (struct pt_regs *) sp;
984 	sp -= STACK_FRAME_OVERHEAD;
985 	p->thread.ksp = sp;
986 	p->thread.ksp_limit = (unsigned long)task_stack_page(p) +
987 				_ALIGN_UP(sizeof(struct thread_info), 16);
988 
989 #ifdef CONFIG_HAVE_HW_BREAKPOINT
990 	p->thread.ptrace_bps[0] = NULL;
991 #endif
992 
993 #ifdef CONFIG_PPC_STD_MMU_64
994 	if (mmu_has_feature(MMU_FTR_SLB)) {
995 		unsigned long sp_vsid;
996 		unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
997 
998 		if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
999 			sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
1000 				<< SLB_VSID_SHIFT_1T;
1001 		else
1002 			sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_256M)
1003 				<< SLB_VSID_SHIFT;
1004 		sp_vsid |= SLB_VSID_KERNEL | llp;
1005 		p->thread.ksp_vsid = sp_vsid;
1006 	}
1007 #endif /* CONFIG_PPC_STD_MMU_64 */
1008 #ifdef CONFIG_PPC64
1009 	if (cpu_has_feature(CPU_FTR_DSCR)) {
1010 		p->thread.dscr_inherit = current->thread.dscr_inherit;
1011 		p->thread.dscr = current->thread.dscr;
1012 	}
1013 	if (cpu_has_feature(CPU_FTR_HAS_PPR))
1014 		p->thread.ppr = INIT_PPR;
1015 #endif
1016 	/*
1017 	 * The PPC64 ABI makes use of a TOC to contain function
1018 	 * pointers.  The function (ret_from_except) is actually a pointer
1019 	 * to the TOC entry.  The first entry is a pointer to the actual
1020 	 * function.
1021 	 */
1022 #ifdef CONFIG_PPC64
1023 	kregs->nip = *((unsigned long *)f);
1024 #else
1025 	kregs->nip = (unsigned long)f;
1026 #endif
1027 	return 0;
1028 }
1029 
1030 /*
1031  * Set up a thread for executing a new program
1032  */
1033 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
1034 {
1035 #ifdef CONFIG_PPC64
1036 	unsigned long load_addr = regs->gpr[2];	/* saved by ELF_PLAT_INIT */
1037 #endif
1038 
1039 	/*
1040 	 * If we exec out of a kernel thread then thread.regs will not be
1041 	 * set.  Do it now.
1042 	 */
1043 	if (!current->thread.regs) {
1044 		struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
1045 		current->thread.regs = regs - 1;
1046 	}
1047 
1048 	memset(regs->gpr, 0, sizeof(regs->gpr));
1049 	regs->ctr = 0;
1050 	regs->link = 0;
1051 	regs->xer = 0;
1052 	regs->ccr = 0;
1053 	regs->gpr[1] = sp;
1054 
1055 	/*
1056 	 * We have just cleared all the nonvolatile GPRs, so make
1057 	 * FULL_REGS(regs) return true.  This is necessary to allow
1058 	 * ptrace to examine the thread immediately after exec.
1059 	 */
1060 	regs->trap &= ~1UL;
1061 
1062 #ifdef CONFIG_PPC32
1063 	regs->mq = 0;
1064 	regs->nip = start;
1065 	regs->msr = MSR_USER;
1066 #else
1067 	if (!is_32bit_task()) {
1068 		unsigned long entry, toc;
1069 
1070 		/* start is a relocated pointer to the function descriptor for
1071 		 * the elf _start routine.  The first entry in the function
1072 		 * descriptor is the entry address of _start and the second
1073 		 * entry is the TOC value we need to use.
1074 		 */
1075 		__get_user(entry, (unsigned long __user *)start);
1076 		__get_user(toc, (unsigned long __user *)start+1);
1077 
1078 		/* Check whether the e_entry function descriptor entries
1079 		 * need to be relocated before we can use them.
1080 		 */
1081 		if (load_addr != 0) {
1082 			entry += load_addr;
1083 			toc   += load_addr;
1084 		}
1085 		regs->nip = entry;
1086 		regs->gpr[2] = toc;
1087 		regs->msr = MSR_USER64;
1088 	} else {
1089 		regs->nip = start;
1090 		regs->gpr[2] = 0;
1091 		regs->msr = MSR_USER32;
1092 	}
1093 #endif
1094 	discard_lazy_cpu_state();
1095 #ifdef CONFIG_VSX
1096 	current->thread.used_vsr = 0;
1097 #endif
1098 	memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
1099 	current->thread.fpscr.val = 0;
1100 #ifdef CONFIG_ALTIVEC
1101 	memset(current->thread.vr, 0, sizeof(current->thread.vr));
1102 	memset(&current->thread.vscr, 0, sizeof(current->thread.vscr));
1103 	current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
1104 	current->thread.vrsave = 0;
1105 	current->thread.used_vr = 0;
1106 #endif /* CONFIG_ALTIVEC */
1107 #ifdef CONFIG_SPE
1108 	memset(current->thread.evr, 0, sizeof(current->thread.evr));
1109 	current->thread.acc = 0;
1110 	current->thread.spefscr = 0;
1111 	current->thread.used_spe = 0;
1112 #endif /* CONFIG_SPE */
1113 #ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1114 	if (cpu_has_feature(CPU_FTR_TM))
1115 		regs->msr |= MSR_TM;
1116 	current->thread.tm_tfhar = 0;
1117 	current->thread.tm_texasr = 0;
1118 	current->thread.tm_tfiar = 0;
1119 #endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
1120 }
1121 
1122 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
1123 		| PR_FP_EXC_RES | PR_FP_EXC_INV)
1124 
1125 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
1126 {
1127 	struct pt_regs *regs = tsk->thread.regs;
1128 
1129 	/* This is a bit hairy.  If we are an SPE enabled  processor
1130 	 * (have embedded fp) we store the IEEE exception enable flags in
1131 	 * fpexc_mode.  fpexc_mode is also used for setting FP exception
1132 	 * mode (asyn, precise, disabled) for 'Classic' FP. */
1133 	if (val & PR_FP_EXC_SW_ENABLE) {
1134 #ifdef CONFIG_SPE
1135 		if (cpu_has_feature(CPU_FTR_SPE)) {
1136 			tsk->thread.fpexc_mode = val &
1137 				(PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
1138 			return 0;
1139 		} else {
1140 			return -EINVAL;
1141 		}
1142 #else
1143 		return -EINVAL;
1144 #endif
1145 	}
1146 
1147 	/* on a CONFIG_SPE this does not hurt us.  The bits that
1148 	 * __pack_fe01 use do not overlap with bits used for
1149 	 * PR_FP_EXC_SW_ENABLE.  Additionally, the MSR[FE0,FE1] bits
1150 	 * on CONFIG_SPE implementations are reserved so writing to
1151 	 * them does not change anything */
1152 	if (val > PR_FP_EXC_PRECISE)
1153 		return -EINVAL;
1154 	tsk->thread.fpexc_mode = __pack_fe01(val);
1155 	if (regs != NULL && (regs->msr & MSR_FP) != 0)
1156 		regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
1157 			| tsk->thread.fpexc_mode;
1158 	return 0;
1159 }
1160 
1161 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
1162 {
1163 	unsigned int val;
1164 
1165 	if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
1166 #ifdef CONFIG_SPE
1167 		if (cpu_has_feature(CPU_FTR_SPE))
1168 			val = tsk->thread.fpexc_mode;
1169 		else
1170 			return -EINVAL;
1171 #else
1172 		return -EINVAL;
1173 #endif
1174 	else
1175 		val = __unpack_fe01(tsk->thread.fpexc_mode);
1176 	return put_user(val, (unsigned int __user *) adr);
1177 }
1178 
1179 int set_endian(struct task_struct *tsk, unsigned int val)
1180 {
1181 	struct pt_regs *regs = tsk->thread.regs;
1182 
1183 	if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
1184 	    (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
1185 		return -EINVAL;
1186 
1187 	if (regs == NULL)
1188 		return -EINVAL;
1189 
1190 	if (val == PR_ENDIAN_BIG)
1191 		regs->msr &= ~MSR_LE;
1192 	else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
1193 		regs->msr |= MSR_LE;
1194 	else
1195 		return -EINVAL;
1196 
1197 	return 0;
1198 }
1199 
1200 int get_endian(struct task_struct *tsk, unsigned long adr)
1201 {
1202 	struct pt_regs *regs = tsk->thread.regs;
1203 	unsigned int val;
1204 
1205 	if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
1206 	    !cpu_has_feature(CPU_FTR_REAL_LE))
1207 		return -EINVAL;
1208 
1209 	if (regs == NULL)
1210 		return -EINVAL;
1211 
1212 	if (regs->msr & MSR_LE) {
1213 		if (cpu_has_feature(CPU_FTR_REAL_LE))
1214 			val = PR_ENDIAN_LITTLE;
1215 		else
1216 			val = PR_ENDIAN_PPC_LITTLE;
1217 	} else
1218 		val = PR_ENDIAN_BIG;
1219 
1220 	return put_user(val, (unsigned int __user *)adr);
1221 }
1222 
1223 int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
1224 {
1225 	tsk->thread.align_ctl = val;
1226 	return 0;
1227 }
1228 
1229 int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
1230 {
1231 	return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
1232 }
1233 
1234 static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
1235 				  unsigned long nbytes)
1236 {
1237 	unsigned long stack_page;
1238 	unsigned long cpu = task_cpu(p);
1239 
1240 	/*
1241 	 * Avoid crashing if the stack has overflowed and corrupted
1242 	 * task_cpu(p), which is in the thread_info struct.
1243 	 */
1244 	if (cpu < NR_CPUS && cpu_possible(cpu)) {
1245 		stack_page = (unsigned long) hardirq_ctx[cpu];
1246 		if (sp >= stack_page + sizeof(struct thread_struct)
1247 		    && sp <= stack_page + THREAD_SIZE - nbytes)
1248 			return 1;
1249 
1250 		stack_page = (unsigned long) softirq_ctx[cpu];
1251 		if (sp >= stack_page + sizeof(struct thread_struct)
1252 		    && sp <= stack_page + THREAD_SIZE - nbytes)
1253 			return 1;
1254 	}
1255 	return 0;
1256 }
1257 
1258 int validate_sp(unsigned long sp, struct task_struct *p,
1259 		       unsigned long nbytes)
1260 {
1261 	unsigned long stack_page = (unsigned long)task_stack_page(p);
1262 
1263 	if (sp >= stack_page + sizeof(struct thread_struct)
1264 	    && sp <= stack_page + THREAD_SIZE - nbytes)
1265 		return 1;
1266 
1267 	return valid_irq_stack(sp, p, nbytes);
1268 }
1269 
1270 EXPORT_SYMBOL(validate_sp);
1271 
1272 unsigned long get_wchan(struct task_struct *p)
1273 {
1274 	unsigned long ip, sp;
1275 	int count = 0;
1276 
1277 	if (!p || p == current || p->state == TASK_RUNNING)
1278 		return 0;
1279 
1280 	sp = p->thread.ksp;
1281 	if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
1282 		return 0;
1283 
1284 	do {
1285 		sp = *(unsigned long *)sp;
1286 		if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
1287 			return 0;
1288 		if (count > 0) {
1289 			ip = ((unsigned long *)sp)[STACK_FRAME_LR_SAVE];
1290 			if (!in_sched_functions(ip))
1291 				return ip;
1292 		}
1293 	} while (count++ < 16);
1294 	return 0;
1295 }
1296 
1297 static int kstack_depth_to_print = CONFIG_PRINT_STACK_DEPTH;
1298 
1299 void show_stack(struct task_struct *tsk, unsigned long *stack)
1300 {
1301 	unsigned long sp, ip, lr, newsp;
1302 	int count = 0;
1303 	int firstframe = 1;
1304 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1305 	int curr_frame = current->curr_ret_stack;
1306 	extern void return_to_handler(void);
1307 	unsigned long rth = (unsigned long)return_to_handler;
1308 	unsigned long mrth = -1;
1309 #ifdef CONFIG_PPC64
1310 	extern void mod_return_to_handler(void);
1311 	rth = *(unsigned long *)rth;
1312 	mrth = (unsigned long)mod_return_to_handler;
1313 	mrth = *(unsigned long *)mrth;
1314 #endif
1315 #endif
1316 
1317 	sp = (unsigned long) stack;
1318 	if (tsk == NULL)
1319 		tsk = current;
1320 	if (sp == 0) {
1321 		if (tsk == current)
1322 			asm("mr %0,1" : "=r" (sp));
1323 		else
1324 			sp = tsk->thread.ksp;
1325 	}
1326 
1327 	lr = 0;
1328 	printk("Call Trace:\n");
1329 	do {
1330 		if (!validate_sp(sp, tsk, STACK_FRAME_OVERHEAD))
1331 			return;
1332 
1333 		stack = (unsigned long *) sp;
1334 		newsp = stack[0];
1335 		ip = stack[STACK_FRAME_LR_SAVE];
1336 		if (!firstframe || ip != lr) {
1337 			printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip);
1338 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1339 			if ((ip == rth || ip == mrth) && curr_frame >= 0) {
1340 				printk(" (%pS)",
1341 				       (void *)current->ret_stack[curr_frame].ret);
1342 				curr_frame--;
1343 			}
1344 #endif
1345 			if (firstframe)
1346 				printk(" (unreliable)");
1347 			printk("\n");
1348 		}
1349 		firstframe = 0;
1350 
1351 		/*
1352 		 * See if this is an exception frame.
1353 		 * We look for the "regshere" marker in the current frame.
1354 		 */
1355 		if (validate_sp(sp, tsk, STACK_INT_FRAME_SIZE)
1356 		    && stack[STACK_FRAME_MARKER] == STACK_FRAME_REGS_MARKER) {
1357 			struct pt_regs *regs = (struct pt_regs *)
1358 				(sp + STACK_FRAME_OVERHEAD);
1359 			lr = regs->link;
1360 			printk("--- Exception: %lx at %pS\n    LR = %pS\n",
1361 			       regs->trap, (void *)regs->nip, (void *)lr);
1362 			firstframe = 1;
1363 		}
1364 
1365 		sp = newsp;
1366 	} while (count++ < kstack_depth_to_print);
1367 }
1368 
1369 #ifdef CONFIG_PPC64
1370 /* Called with hard IRQs off */
1371 void __ppc64_runlatch_on(void)
1372 {
1373 	struct thread_info *ti = current_thread_info();
1374 	unsigned long ctrl;
1375 
1376 	ctrl = mfspr(SPRN_CTRLF);
1377 	ctrl |= CTRL_RUNLATCH;
1378 	mtspr(SPRN_CTRLT, ctrl);
1379 
1380 	ti->local_flags |= _TLF_RUNLATCH;
1381 }
1382 
1383 /* Called with hard IRQs off */
1384 void __ppc64_runlatch_off(void)
1385 {
1386 	struct thread_info *ti = current_thread_info();
1387 	unsigned long ctrl;
1388 
1389 	ti->local_flags &= ~_TLF_RUNLATCH;
1390 
1391 	ctrl = mfspr(SPRN_CTRLF);
1392 	ctrl &= ~CTRL_RUNLATCH;
1393 	mtspr(SPRN_CTRLT, ctrl);
1394 }
1395 #endif /* CONFIG_PPC64 */
1396 
1397 unsigned long arch_align_stack(unsigned long sp)
1398 {
1399 	if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
1400 		sp -= get_random_int() & ~PAGE_MASK;
1401 	return sp & ~0xf;
1402 }
1403 
1404 static inline unsigned long brk_rnd(void)
1405 {
1406         unsigned long rnd = 0;
1407 
1408 	/* 8MB for 32bit, 1GB for 64bit */
1409 	if (is_32bit_task())
1410 		rnd = (long)(get_random_int() % (1<<(23-PAGE_SHIFT)));
1411 	else
1412 		rnd = (long)(get_random_int() % (1<<(30-PAGE_SHIFT)));
1413 
1414 	return rnd << PAGE_SHIFT;
1415 }
1416 
1417 unsigned long arch_randomize_brk(struct mm_struct *mm)
1418 {
1419 	unsigned long base = mm->brk;
1420 	unsigned long ret;
1421 
1422 #ifdef CONFIG_PPC_STD_MMU_64
1423 	/*
1424 	 * If we are using 1TB segments and we are allowed to randomise
1425 	 * the heap, we can put it above 1TB so it is backed by a 1TB
1426 	 * segment. Otherwise the heap will be in the bottom 1TB
1427 	 * which always uses 256MB segments and this may result in a
1428 	 * performance penalty.
1429 	 */
1430 	if (!is_32bit_task() && (mmu_highuser_ssize == MMU_SEGSIZE_1T))
1431 		base = max_t(unsigned long, mm->brk, 1UL << SID_SHIFT_1T);
1432 #endif
1433 
1434 	ret = PAGE_ALIGN(base + brk_rnd());
1435 
1436 	if (ret < mm->brk)
1437 		return mm->brk;
1438 
1439 	return ret;
1440 }
1441 
1442 unsigned long randomize_et_dyn(unsigned long base)
1443 {
1444 	unsigned long ret = PAGE_ALIGN(base + brk_rnd());
1445 
1446 	if (ret < base)
1447 		return base;
1448 
1449 	return ret;
1450 }
1451