Merge tag 'mmc-updates-for-3.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel...
[deliverable/linux.git] / arch / powerpc / kernel / process.c
CommitLineData
14cf11af 1/*
14cf11af
PM
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
14cf11af
PM
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>
14cf11af
PM
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>
4b16f8e2 31#include <linux/export.h>
14cf11af
PM
32#include <linux/kallsyms.h>
33#include <linux/mqueue.h>
34#include <linux/hardirq.h>
06d67d54 35#include <linux/utsname.h>
6794c782 36#include <linux/ftrace.h>
79741dd3 37#include <linux/kernel_stat.h>
d839088c
AB
38#include <linux/personality.h>
39#include <linux/random.h>
5aae8a53 40#include <linux/hw_breakpoint.h>
14cf11af
PM
41
42#include <asm/pgtable.h>
43#include <asm/uaccess.h>
14cf11af
PM
44#include <asm/io.h>
45#include <asm/processor.h>
46#include <asm/mmu.h>
47#include <asm/prom.h>
76032de8 48#include <asm/machdep.h>
c6622f63 49#include <asm/time.h>
ae3a197e 50#include <asm/runlatch.h>
a7f31841 51#include <asm/syscalls.h>
ae3a197e 52#include <asm/switch_to.h>
fb09692e 53#include <asm/tm.h>
ae3a197e 54#include <asm/debug.h>
06d67d54
PM
55#ifdef CONFIG_PPC64
56#include <asm/firmware.h>
06d67d54 57#endif
d6a61bfc
LM
58#include <linux/kprobes.h>
59#include <linux/kdebug.h>
14cf11af 60
8b3c34cf
MN
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
14cf11af
PM
68extern unsigned long _get_SP(void);
69
70#ifndef CONFIG_SMP
71struct task_struct *last_task_used_math = NULL;
72struct task_struct *last_task_used_altivec = NULL;
ce48b210 73struct task_struct *last_task_used_vsx = NULL;
14cf11af
PM
74struct task_struct *last_task_used_spe = NULL;
75#endif
76
037f0eed 77#ifdef CONFIG_PPC_FPU
14cf11af
PM
78/*
79 * Make sure the floating-point register state in the
80 * the thread_struct is up to date for task tsk.
81 */
82void flush_fp_to_thread(struct task_struct *tsk)
83{
84 if (tsk->thread.regs) {
85 /*
86 * We need to disable preemption here because if we didn't,
87 * another process could get scheduled after the regs->msr
88 * test but before we have finished saving the FP registers
89 * to the thread_struct. That process could take over the
90 * FPU, and then when we get scheduled again we would store
91 * bogus values for the remaining FP registers.
92 */
93 preempt_disable();
94 if (tsk->thread.regs->msr & MSR_FP) {
95#ifdef CONFIG_SMP
96 /*
97 * This should only ever be called for current or
98 * for a stopped child process. Since we save away
99 * the FP register state on context switch on SMP,
100 * there is something wrong if a stopped child appears
101 * to still have its FP state in the CPU registers.
102 */
103 BUG_ON(tsk != current);
104#endif
0ee6c15e 105 giveup_fpu(tsk);
14cf11af
PM
106 }
107 preempt_enable();
108 }
109}
de56a948 110EXPORT_SYMBOL_GPL(flush_fp_to_thread);
037f0eed 111#endif
14cf11af
PM
112
113void enable_kernel_fp(void)
114{
115 WARN_ON(preemptible());
116
117#ifdef CONFIG_SMP
118 if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
119 giveup_fpu(current);
120 else
121 giveup_fpu(NULL); /* just enables FP for kernel */
122#else
123 giveup_fpu(last_task_used_math);
124#endif /* CONFIG_SMP */
125}
126EXPORT_SYMBOL(enable_kernel_fp);
127
14cf11af
PM
128#ifdef CONFIG_ALTIVEC
129void enable_kernel_altivec(void)
130{
131 WARN_ON(preemptible());
132
133#ifdef CONFIG_SMP
134 if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
135 giveup_altivec(current);
136 else
35000870 137 giveup_altivec_notask();
14cf11af
PM
138#else
139 giveup_altivec(last_task_used_altivec);
140#endif /* CONFIG_SMP */
141}
142EXPORT_SYMBOL(enable_kernel_altivec);
143
144/*
145 * Make sure the VMX/Altivec register state in the
146 * the thread_struct is up to date for task tsk.
147 */
148void flush_altivec_to_thread(struct task_struct *tsk)
149{
150 if (tsk->thread.regs) {
151 preempt_disable();
152 if (tsk->thread.regs->msr & MSR_VEC) {
153#ifdef CONFIG_SMP
154 BUG_ON(tsk != current);
155#endif
0ee6c15e 156 giveup_altivec(tsk);
14cf11af
PM
157 }
158 preempt_enable();
159 }
160}
de56a948 161EXPORT_SYMBOL_GPL(flush_altivec_to_thread);
14cf11af
PM
162#endif /* CONFIG_ALTIVEC */
163
ce48b210
MN
164#ifdef CONFIG_VSX
165#if 0
166/* not currently used, but some crazy RAID module might want to later */
167void enable_kernel_vsx(void)
168{
169 WARN_ON(preemptible());
170
171#ifdef CONFIG_SMP
172 if (current->thread.regs && (current->thread.regs->msr & MSR_VSX))
173 giveup_vsx(current);
174 else
175 giveup_vsx(NULL); /* just enable vsx for kernel - force */
176#else
177 giveup_vsx(last_task_used_vsx);
178#endif /* CONFIG_SMP */
179}
180EXPORT_SYMBOL(enable_kernel_vsx);
181#endif
182
7c292170
MN
183void giveup_vsx(struct task_struct *tsk)
184{
185 giveup_fpu(tsk);
186 giveup_altivec(tsk);
187 __giveup_vsx(tsk);
188}
189
ce48b210
MN
190void flush_vsx_to_thread(struct task_struct *tsk)
191{
192 if (tsk->thread.regs) {
193 preempt_disable();
194 if (tsk->thread.regs->msr & MSR_VSX) {
195#ifdef CONFIG_SMP
196 BUG_ON(tsk != current);
197#endif
198 giveup_vsx(tsk);
199 }
200 preempt_enable();
201 }
202}
de56a948 203EXPORT_SYMBOL_GPL(flush_vsx_to_thread);
ce48b210
MN
204#endif /* CONFIG_VSX */
205
14cf11af
PM
206#ifdef CONFIG_SPE
207
208void enable_kernel_spe(void)
209{
210 WARN_ON(preemptible());
211
212#ifdef CONFIG_SMP
213 if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
214 giveup_spe(current);
215 else
216 giveup_spe(NULL); /* just enable SPE for kernel - force */
217#else
218 giveup_spe(last_task_used_spe);
219#endif /* __SMP __ */
220}
221EXPORT_SYMBOL(enable_kernel_spe);
222
223void flush_spe_to_thread(struct task_struct *tsk)
224{
225 if (tsk->thread.regs) {
226 preempt_disable();
227 if (tsk->thread.regs->msr & MSR_SPE) {
228#ifdef CONFIG_SMP
229 BUG_ON(tsk != current);
230#endif
685659ee 231 tsk->thread.spefscr = mfspr(SPRN_SPEFSCR);
0ee6c15e 232 giveup_spe(tsk);
14cf11af
PM
233 }
234 preempt_enable();
235 }
236}
14cf11af
PM
237#endif /* CONFIG_SPE */
238
5388fb10 239#ifndef CONFIG_SMP
48abec07
PM
240/*
241 * If we are doing lazy switching of CPU state (FP, altivec or SPE),
242 * and the current task has some state, discard it.
243 */
5388fb10 244void discard_lazy_cpu_state(void)
48abec07 245{
48abec07
PM
246 preempt_disable();
247 if (last_task_used_math == current)
248 last_task_used_math = NULL;
249#ifdef CONFIG_ALTIVEC
250 if (last_task_used_altivec == current)
251 last_task_used_altivec = NULL;
252#endif /* CONFIG_ALTIVEC */
ce48b210
MN
253#ifdef CONFIG_VSX
254 if (last_task_used_vsx == current)
255 last_task_used_vsx = NULL;
256#endif /* CONFIG_VSX */
48abec07
PM
257#ifdef CONFIG_SPE
258 if (last_task_used_spe == current)
259 last_task_used_spe = NULL;
260#endif
261 preempt_enable();
48abec07 262}
5388fb10 263#endif /* CONFIG_SMP */
48abec07 264
3bffb652
DK
265#ifdef CONFIG_PPC_ADV_DEBUG_REGS
266void do_send_trap(struct pt_regs *regs, unsigned long address,
267 unsigned long error_code, int signal_code, int breakpt)
268{
269 siginfo_t info;
270
41ab5266 271 current->thread.trap_nr = signal_code;
3bffb652
DK
272 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
273 11, SIGSEGV) == NOTIFY_STOP)
274 return;
275
276 /* Deliver the signal to userspace */
277 info.si_signo = SIGTRAP;
278 info.si_errno = breakpt; /* breakpoint or watchpoint id */
279 info.si_code = signal_code;
280 info.si_addr = (void __user *)address;
281 force_sig_info(SIGTRAP, &info, current);
282}
283#else /* !CONFIG_PPC_ADV_DEBUG_REGS */
9422de3e 284void do_break (struct pt_regs *regs, unsigned long address,
d6a61bfc
LM
285 unsigned long error_code)
286{
287 siginfo_t info;
288
41ab5266 289 current->thread.trap_nr = TRAP_HWBKPT;
d6a61bfc
LM
290 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
291 11, SIGSEGV) == NOTIFY_STOP)
292 return;
293
9422de3e 294 if (debugger_break_match(regs))
d6a61bfc
LM
295 return;
296
9422de3e
MN
297 /* Clear the breakpoint */
298 hw_breakpoint_disable();
d6a61bfc
LM
299
300 /* Deliver the signal to userspace */
301 info.si_signo = SIGTRAP;
302 info.si_errno = 0;
303 info.si_code = TRAP_HWBKPT;
304 info.si_addr = (void __user *)address;
305 force_sig_info(SIGTRAP, &info, current);
306}
3bffb652 307#endif /* CONFIG_PPC_ADV_DEBUG_REGS */
d6a61bfc 308
9422de3e 309static DEFINE_PER_CPU(struct arch_hw_breakpoint, current_brk);
a2ceff5e 310
3bffb652
DK
311#ifdef CONFIG_PPC_ADV_DEBUG_REGS
312/*
313 * Set the debug registers back to their default "safe" values.
314 */
315static void set_debug_reg_defaults(struct thread_struct *thread)
316{
51ae8d4a 317 thread->debug.iac1 = thread->debug.iac2 = 0;
3bffb652 318#if CONFIG_PPC_ADV_DEBUG_IACS > 2
51ae8d4a 319 thread->debug.iac3 = thread->debug.iac4 = 0;
3bffb652 320#endif
51ae8d4a 321 thread->debug.dac1 = thread->debug.dac2 = 0;
3bffb652 322#if CONFIG_PPC_ADV_DEBUG_DVCS > 0
51ae8d4a 323 thread->debug.dvc1 = thread->debug.dvc2 = 0;
3bffb652 324#endif
51ae8d4a 325 thread->debug.dbcr0 = 0;
3bffb652
DK
326#ifdef CONFIG_BOOKE
327 /*
328 * Force User/Supervisor bits to b11 (user-only MSR[PR]=1)
329 */
51ae8d4a 330 thread->debug.dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US |
3bffb652
DK
331 DBCR1_IAC3US | DBCR1_IAC4US;
332 /*
333 * Force Data Address Compare User/Supervisor bits to be User-only
334 * (0b11 MSR[PR]=1) and set all other bits in DBCR2 register to be 0.
335 */
51ae8d4a 336 thread->debug.dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
3bffb652 337#else
51ae8d4a 338 thread->debug.dbcr1 = 0;
3bffb652
DK
339#endif
340}
341
342static void prime_debug_regs(struct thread_struct *thread)
343{
6cecf76b
SW
344 /*
345 * We could have inherited MSR_DE from userspace, since
346 * it doesn't get cleared on exception entry. Make sure
347 * MSR_DE is clear before we enable any debug events.
348 */
349 mtmsr(mfmsr() & ~MSR_DE);
350
51ae8d4a
BB
351 mtspr(SPRN_IAC1, thread->debug.iac1);
352 mtspr(SPRN_IAC2, thread->debug.iac2);
3bffb652 353#if CONFIG_PPC_ADV_DEBUG_IACS > 2
51ae8d4a
BB
354 mtspr(SPRN_IAC3, thread->debug.iac3);
355 mtspr(SPRN_IAC4, thread->debug.iac4);
3bffb652 356#endif
51ae8d4a
BB
357 mtspr(SPRN_DAC1, thread->debug.dac1);
358 mtspr(SPRN_DAC2, thread->debug.dac2);
3bffb652 359#if CONFIG_PPC_ADV_DEBUG_DVCS > 0
51ae8d4a
BB
360 mtspr(SPRN_DVC1, thread->debug.dvc1);
361 mtspr(SPRN_DVC2, thread->debug.dvc2);
3bffb652 362#endif
51ae8d4a
BB
363 mtspr(SPRN_DBCR0, thread->debug.dbcr0);
364 mtspr(SPRN_DBCR1, thread->debug.dbcr1);
3bffb652 365#ifdef CONFIG_BOOKE
51ae8d4a 366 mtspr(SPRN_DBCR2, thread->debug.dbcr2);
3bffb652
DK
367#endif
368}
369/*
370 * Unless neither the old or new thread are making use of the
371 * debug registers, set the debug registers from the values
372 * stored in the new thread.
373 */
3743c9b8 374void switch_booke_debug_regs(struct thread_struct *new_thread)
3bffb652 375{
51ae8d4a
BB
376 if ((current->thread.debug.dbcr0 & DBCR0_IDM)
377 || (new_thread->debug.dbcr0 & DBCR0_IDM))
3bffb652
DK
378 prime_debug_regs(new_thread);
379}
3743c9b8 380EXPORT_SYMBOL_GPL(switch_booke_debug_regs);
3bffb652 381#else /* !CONFIG_PPC_ADV_DEBUG_REGS */
e0780b72 382#ifndef CONFIG_HAVE_HW_BREAKPOINT
3bffb652
DK
383static void set_debug_reg_defaults(struct thread_struct *thread)
384{
9422de3e
MN
385 thread->hw_brk.address = 0;
386 thread->hw_brk.type = 0;
b9818c33 387 set_breakpoint(&thread->hw_brk);
3bffb652 388}
e0780b72 389#endif /* !CONFIG_HAVE_HW_BREAKPOINT */
3bffb652
DK
390#endif /* CONFIG_PPC_ADV_DEBUG_REGS */
391
172ae2e7 392#ifdef CONFIG_PPC_ADV_DEBUG_REGS
9422de3e
MN
393static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
394{
d6a61bfc 395 mtspr(SPRN_DAC1, dabr);
221c185d
DK
396#ifdef CONFIG_PPC_47x
397 isync();
398#endif
9422de3e
MN
399 return 0;
400}
c6c9eace 401#elif defined(CONFIG_PPC_BOOK3S)
9422de3e
MN
402static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
403{
c6c9eace 404 mtspr(SPRN_DABR, dabr);
82a9f16a
MN
405 if (cpu_has_feature(CPU_FTR_DABRX))
406 mtspr(SPRN_DABRX, dabrx);
cab0af98 407 return 0;
14cf11af 408}
9422de3e
MN
409#else
410static inline int __set_dabr(unsigned long dabr, unsigned long dabrx)
411{
412 return -EINVAL;
413}
414#endif
415
416static inline int set_dabr(struct arch_hw_breakpoint *brk)
417{
418 unsigned long dabr, dabrx;
419
420 dabr = brk->address | (brk->type & HW_BRK_TYPE_DABR);
421 dabrx = ((brk->type >> 3) & 0x7);
422
423 if (ppc_md.set_dabr)
424 return ppc_md.set_dabr(dabr, dabrx);
425
426 return __set_dabr(dabr, dabrx);
427}
428
bf99de36
MN
429static inline int set_dawr(struct arch_hw_breakpoint *brk)
430{
05d694ea 431 unsigned long dawr, dawrx, mrd;
bf99de36
MN
432
433 dawr = brk->address;
434
435 dawrx = (brk->type & (HW_BRK_TYPE_READ | HW_BRK_TYPE_WRITE)) \
436 << (63 - 58); //* read/write bits */
437 dawrx |= ((brk->type & (HW_BRK_TYPE_TRANSLATE)) >> 2) \
438 << (63 - 59); //* translate */
439 dawrx |= (brk->type & (HW_BRK_TYPE_PRIV_ALL)) \
440 >> 3; //* PRIM bits */
05d694ea
MN
441 /* dawr length is stored in field MDR bits 48:53. Matches range in
442 doublewords (64 bits) baised by -1 eg. 0b000000=1DW and
443 0b111111=64DW.
444 brk->len is in bytes.
445 This aligns up to double word size, shifts and does the bias.
446 */
447 mrd = ((brk->len + 7) >> 3) - 1;
448 dawrx |= (mrd & 0x3f) << (63 - 53);
bf99de36
MN
449
450 if (ppc_md.set_dawr)
451 return ppc_md.set_dawr(dawr, dawrx);
452 mtspr(SPRN_DAWR, dawr);
453 mtspr(SPRN_DAWRX, dawrx);
454 return 0;
455}
456
b9818c33 457int set_breakpoint(struct arch_hw_breakpoint *brk)
9422de3e
MN
458{
459 __get_cpu_var(current_brk) = *brk;
460
bf99de36
MN
461 if (cpu_has_feature(CPU_FTR_DAWR))
462 return set_dawr(brk);
463
9422de3e
MN
464 return set_dabr(brk);
465}
14cf11af 466
06d67d54
PM
467#ifdef CONFIG_PPC64
468DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
06d67d54 469#endif
14cf11af 470
9422de3e
MN
471static inline bool hw_brk_match(struct arch_hw_breakpoint *a,
472 struct arch_hw_breakpoint *b)
473{
474 if (a->address != b->address)
475 return false;
476 if (a->type != b->type)
477 return false;
478 if (a->len != b->len)
479 return false;
480 return true;
481}
fb09692e
MN
482#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
483static inline void tm_reclaim_task(struct task_struct *tsk)
484{
485 /* We have to work out if we're switching from/to a task that's in the
486 * middle of a transaction.
487 *
488 * In switching we need to maintain a 2nd register state as
489 * oldtask->thread.ckpt_regs. We tm_reclaim(oldproc); this saves the
490 * checkpointed (tbegin) state in ckpt_regs and saves the transactional
491 * (current) FPRs into oldtask->thread.transact_fpr[].
492 *
493 * We also context switch (save) TFHAR/TEXASR/TFIAR in here.
494 */
495 struct thread_struct *thr = &tsk->thread;
496
497 if (!thr->regs)
498 return;
499
500 if (!MSR_TM_ACTIVE(thr->regs->msr))
501 goto out_and_saveregs;
502
503 /* Stash the original thread MSR, as giveup_fpu et al will
504 * modify it. We hold onto it to see whether the task used
505 * FP & vector regs.
506 */
507 thr->tm_orig_msr = thr->regs->msr;
508
509 TM_DEBUG("--- tm_reclaim on pid %d (NIP=%lx, "
510 "ccr=%lx, msr=%lx, trap=%lx)\n",
511 tsk->pid, thr->regs->nip,
512 thr->regs->ccr, thr->regs->msr,
513 thr->regs->trap);
514
515 tm_reclaim(thr, thr->regs->msr, TM_CAUSE_RESCHED);
516
517 TM_DEBUG("--- tm_reclaim on pid %d complete\n",
518 tsk->pid);
519
520out_and_saveregs:
521 /* Always save the regs here, even if a transaction's not active.
522 * This context-switches a thread's TM info SPRs. We do it here to
523 * be consistent with the restore path (in recheckpoint) which
524 * cannot happen later in _switch().
525 */
526 tm_save_sprs(thr);
527}
528
bc2a9408 529static inline void tm_recheckpoint_new_task(struct task_struct *new)
fb09692e
MN
530{
531 unsigned long msr;
532
533 if (!cpu_has_feature(CPU_FTR_TM))
534 return;
535
536 /* Recheckpoint the registers of the thread we're about to switch to.
537 *
538 * If the task was using FP, we non-lazily reload both the original and
539 * the speculative FP register states. This is because the kernel
540 * doesn't see if/when a TM rollback occurs, so if we take an FP
541 * unavoidable later, we are unable to determine which set of FP regs
542 * need to be restored.
543 */
544 if (!new->thread.regs)
545 return;
546
547 /* The TM SPRs are restored here, so that TEXASR.FS can be set
548 * before the trecheckpoint and no explosion occurs.
549 */
550 tm_restore_sprs(&new->thread);
551
552 if (!MSR_TM_ACTIVE(new->thread.regs->msr))
553 return;
554 msr = new->thread.tm_orig_msr;
555 /* Recheckpoint to restore original checkpointed register state. */
556 TM_DEBUG("*** tm_recheckpoint of pid %d "
557 "(new->msr 0x%lx, new->origmsr 0x%lx)\n",
558 new->pid, new->thread.regs->msr, msr);
559
560 /* This loads the checkpointed FP/VEC state, if used */
561 tm_recheckpoint(&new->thread, msr);
562
563 /* This loads the speculative FP/VEC state, if used */
564 if (msr & MSR_FP) {
565 do_load_up_transact_fpu(&new->thread);
566 new->thread.regs->msr |=
567 (MSR_FP | new->thread.fpexc_mode);
568 }
f110c0c1 569#ifdef CONFIG_ALTIVEC
fb09692e
MN
570 if (msr & MSR_VEC) {
571 do_load_up_transact_altivec(&new->thread);
572 new->thread.regs->msr |= MSR_VEC;
573 }
f110c0c1 574#endif
fb09692e
MN
575 /* We may as well turn on VSX too since all the state is restored now */
576 if (msr & MSR_VSX)
577 new->thread.regs->msr |= MSR_VSX;
578
579 TM_DEBUG("*** tm_recheckpoint of pid %d complete "
580 "(kernel msr 0x%lx)\n",
581 new->pid, mfmsr());
582}
583
584static inline void __switch_to_tm(struct task_struct *prev)
585{
586 if (cpu_has_feature(CPU_FTR_TM)) {
587 tm_enable();
588 tm_reclaim_task(prev);
589 }
590}
591#else
592#define tm_recheckpoint_new_task(new)
593#define __switch_to_tm(prev)
594#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
9422de3e 595
14cf11af
PM
596struct task_struct *__switch_to(struct task_struct *prev,
597 struct task_struct *new)
598{
599 struct thread_struct *new_thread, *old_thread;
14cf11af 600 struct task_struct *last;
d6bf29b4
PZ
601#ifdef CONFIG_PPC_BOOK3S_64
602 struct ppc64_tlb_batch *batch;
603#endif
14cf11af 604
7ba5fef7
MN
605 WARN_ON(!irqs_disabled());
606
c2d52644
MN
607 /* Back up the TAR across context switches.
608 * Note that the TAR is not available for use in the kernel. (To
609 * provide this, the TAR should be backed up/restored on exception
610 * entry/exit instead, and be in pt_regs. FIXME, this should be in
611 * pt_regs anyway (for debug).)
612 * Save the TAR here before we do treclaim/trecheckpoint as these
613 * will change the TAR.
614 */
615 save_tar(&prev->thread);
616
bc2a9408
MN
617 __switch_to_tm(prev);
618
14cf11af
PM
619#ifdef CONFIG_SMP
620 /* avoid complexity of lazy save/restore of fpu
621 * by just saving it every time we switch out if
622 * this task used the fpu during the last quantum.
623 *
624 * If it tries to use the fpu again, it'll trap and
625 * reload its fp regs. So we don't have to do a restore
626 * every switch, just a save.
627 * -- Cort
628 */
629 if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
630 giveup_fpu(prev);
631#ifdef CONFIG_ALTIVEC
632 /*
633 * If the previous thread used altivec in the last quantum
634 * (thus changing altivec regs) then save them.
635 * We used to check the VRSAVE register but not all apps
636 * set it, so we don't rely on it now (and in fact we need
637 * to save & restore VSCR even if VRSAVE == 0). -- paulus
638 *
639 * On SMP we always save/restore altivec regs just to avoid the
640 * complexity of changing processors.
641 * -- Cort
642 */
643 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
644 giveup_altivec(prev);
14cf11af 645#endif /* CONFIG_ALTIVEC */
ce48b210
MN
646#ifdef CONFIG_VSX
647 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VSX))
7c292170
MN
648 /* VMX and FPU registers are already save here */
649 __giveup_vsx(prev);
ce48b210 650#endif /* CONFIG_VSX */
14cf11af
PM
651#ifdef CONFIG_SPE
652 /*
653 * If the previous thread used spe in the last quantum
654 * (thus changing spe regs) then save them.
655 *
656 * On SMP we always save/restore spe regs just to avoid the
657 * complexity of changing processors.
658 */
659 if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
660 giveup_spe(prev);
c0c0d996
PM
661#endif /* CONFIG_SPE */
662
663#else /* CONFIG_SMP */
664#ifdef CONFIG_ALTIVEC
665 /* Avoid the trap. On smp this this never happens since
666 * we don't set last_task_used_altivec -- Cort
667 */
668 if (new->thread.regs && last_task_used_altivec == new)
669 new->thread.regs->msr |= MSR_VEC;
670#endif /* CONFIG_ALTIVEC */
ce48b210
MN
671#ifdef CONFIG_VSX
672 if (new->thread.regs && last_task_used_vsx == new)
673 new->thread.regs->msr |= MSR_VSX;
674#endif /* CONFIG_VSX */
c0c0d996 675#ifdef CONFIG_SPE
14cf11af
PM
676 /* Avoid the trap. On smp this this never happens since
677 * we don't set last_task_used_spe
678 */
679 if (new->thread.regs && last_task_used_spe == new)
680 new->thread.regs->msr |= MSR_SPE;
681#endif /* CONFIG_SPE */
c0c0d996 682
14cf11af
PM
683#endif /* CONFIG_SMP */
684
172ae2e7 685#ifdef CONFIG_PPC_ADV_DEBUG_REGS
3bffb652 686 switch_booke_debug_regs(&new->thread);
c6c9eace 687#else
5aae8a53
P
688/*
689 * For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
690 * schedule DABR
691 */
692#ifndef CONFIG_HAVE_HW_BREAKPOINT
9422de3e 693 if (unlikely(hw_brk_match(&__get_cpu_var(current_brk), &new->thread.hw_brk)))
b9818c33 694 set_breakpoint(&new->thread.hw_brk);
5aae8a53 695#endif /* CONFIG_HAVE_HW_BREAKPOINT */
d6a61bfc
LM
696#endif
697
c6c9eace 698
14cf11af
PM
699 new_thread = &new->thread;
700 old_thread = &current->thread;
06d67d54
PM
701
702#ifdef CONFIG_PPC64
703 /*
704 * Collect processor utilization data per process
705 */
706 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
707 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
708 long unsigned start_tb, current_tb;
709 start_tb = old_thread->start_tb;
710 cu->current_tb = current_tb = mfspr(SPRN_PURR);
711 old_thread->accum_tb += (current_tb - start_tb);
712 new_thread->start_tb = current_tb;
713 }
d6bf29b4
PZ
714#endif /* CONFIG_PPC64 */
715
716#ifdef CONFIG_PPC_BOOK3S_64
717 batch = &__get_cpu_var(ppc64_tlb_batch);
718 if (batch->active) {
719 current_thread_info()->local_flags |= _TLF_LAZY_MMU;
720 if (batch->index)
721 __flush_tlb_pending(batch);
722 batch->active = 0;
723 }
724#endif /* CONFIG_PPC_BOOK3S_64 */
06d67d54 725
44387e9f
AB
726 /*
727 * We can't take a PMU exception inside _switch() since there is a
728 * window where the kernel stack SLB and the kernel stack are out
729 * of sync. Hard disable here.
730 */
731 hard_irq_disable();
bc2a9408
MN
732
733 tm_recheckpoint_new_task(new);
734
14cf11af
PM
735 last = _switch(old_thread, new_thread);
736
d6bf29b4
PZ
737#ifdef CONFIG_PPC_BOOK3S_64
738 if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
739 current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
740 batch = &__get_cpu_var(ppc64_tlb_batch);
741 batch->active = 1;
742 }
743#endif /* CONFIG_PPC_BOOK3S_64 */
744
14cf11af
PM
745 return last;
746}
747
06d67d54
PM
748static int instructions_to_print = 16;
749
06d67d54
PM
750static void show_instructions(struct pt_regs *regs)
751{
752 int i;
753 unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
754 sizeof(int));
755
756 printk("Instruction dump:");
757
758 for (i = 0; i < instructions_to_print; i++) {
759 int instr;
760
761 if (!(i % 8))
762 printk("\n");
763
0de2d820
SW
764#if !defined(CONFIG_BOOKE)
765 /* If executing with the IMMU off, adjust pc rather
766 * than print XXXXXXXX.
767 */
768 if (!(regs->msr & MSR_IR))
769 pc = (unsigned long)phys_to_virt(pc);
770#endif
771
af308377
SR
772 /* We use __get_user here *only* to avoid an OOPS on a
773 * bad address because the pc *should* only be a
774 * kernel address.
775 */
00ae36de
AB
776 if (!__kernel_text_address(pc) ||
777 __get_user(instr, (unsigned int __user *)pc)) {
40c8cefa 778 printk(KERN_CONT "XXXXXXXX ");
06d67d54
PM
779 } else {
780 if (regs->nip == pc)
40c8cefa 781 printk(KERN_CONT "<%08x> ", instr);
06d67d54 782 else
40c8cefa 783 printk(KERN_CONT "%08x ", instr);
06d67d54
PM
784 }
785
786 pc += sizeof(int);
787 }
788
789 printk("\n");
790}
791
792static struct regbit {
793 unsigned long bit;
794 const char *name;
795} msr_bits[] = {
3bfd0c9c
AB
796#if defined(CONFIG_PPC64) && !defined(CONFIG_BOOKE)
797 {MSR_SF, "SF"},
798 {MSR_HV, "HV"},
799#endif
800 {MSR_VEC, "VEC"},
801 {MSR_VSX, "VSX"},
802#ifdef CONFIG_BOOKE
803 {MSR_CE, "CE"},
804#endif
06d67d54
PM
805 {MSR_EE, "EE"},
806 {MSR_PR, "PR"},
807 {MSR_FP, "FP"},
808 {MSR_ME, "ME"},
3bfd0c9c 809#ifdef CONFIG_BOOKE
1b98326b 810 {MSR_DE, "DE"},
3bfd0c9c
AB
811#else
812 {MSR_SE, "SE"},
813 {MSR_BE, "BE"},
814#endif
06d67d54
PM
815 {MSR_IR, "IR"},
816 {MSR_DR, "DR"},
3bfd0c9c
AB
817 {MSR_PMM, "PMM"},
818#ifndef CONFIG_BOOKE
819 {MSR_RI, "RI"},
820 {MSR_LE, "LE"},
821#endif
06d67d54
PM
822 {0, NULL}
823};
824
825static void printbits(unsigned long val, struct regbit *bits)
826{
827 const char *sep = "";
828
829 printk("<");
830 for (; bits->bit; ++bits)
831 if (val & bits->bit) {
832 printk("%s%s", sep, bits->name);
833 sep = ",";
834 }
835 printk(">");
836}
837
838#ifdef CONFIG_PPC64
f6f7dde3 839#define REG "%016lx"
06d67d54
PM
840#define REGS_PER_LINE 4
841#define LAST_VOLATILE 13
842#else
f6f7dde3 843#define REG "%08lx"
06d67d54
PM
844#define REGS_PER_LINE 8
845#define LAST_VOLATILE 12
846#endif
847
14cf11af
PM
848void show_regs(struct pt_regs * regs)
849{
850 int i, trap;
851
a43cb95d
TH
852 show_regs_print_info(KERN_DEFAULT);
853
06d67d54
PM
854 printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
855 regs->nip, regs->link, regs->ctr);
856 printk("REGS: %p TRAP: %04lx %s (%s)\n",
96b644bd 857 regs, regs->trap, print_tainted(), init_utsname()->release);
06d67d54
PM
858 printk("MSR: "REG" ", regs->msr);
859 printbits(regs->msr, msr_bits);
f6f7dde3 860 printk(" CR: %08lx XER: %08lx\n", regs->ccr, regs->xer);
7230c564
BH
861#ifdef CONFIG_PPC64
862 printk("SOFTE: %ld\n", regs->softe);
863#endif
14cf11af 864 trap = TRAP(regs);
5115a026
MN
865 if ((regs->trap != 0xc00) && cpu_has_feature(CPU_FTR_CFAR))
866 printk("CFAR: "REG"\n", regs->orig_gpr3);
14cf11af 867 if (trap == 0x300 || trap == 0x600)
ba28c9aa 868#if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)
14170789
KG
869 printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
870#else
7071854b 871 printk("DAR: "REG", DSISR: %08lx\n", regs->dar, regs->dsisr);
14170789 872#endif
14cf11af
PM
873
874 for (i = 0; i < 32; i++) {
06d67d54 875 if ((i % REGS_PER_LINE) == 0)
a2367194 876 printk("\nGPR%02d: ", i);
06d67d54
PM
877 printk(REG " ", regs->gpr[i]);
878 if (i == LAST_VOLATILE && !FULL_REGS(regs))
14cf11af
PM
879 break;
880 }
881 printk("\n");
882#ifdef CONFIG_KALLSYMS
883 /*
884 * Lookup NIP late so we have the best change of getting the
885 * above info out without failing
886 */
058c78f4
BH
887 printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
888 printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
14cf11af 889#endif
afc07701
MN
890#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
891 printk("PACATMSCRATCH [%llx]\n", get_paca()->tm_scratch);
892#endif
14cf11af 893 show_stack(current, (unsigned long *) regs->gpr[1]);
06d67d54
PM
894 if (!user_mode(regs))
895 show_instructions(regs);
14cf11af
PM
896}
897
898void exit_thread(void)
899{
48abec07 900 discard_lazy_cpu_state();
14cf11af
PM
901}
902
903void flush_thread(void)
904{
48abec07 905 discard_lazy_cpu_state();
14cf11af 906
e0780b72 907#ifdef CONFIG_HAVE_HW_BREAKPOINT
5aae8a53 908 flush_ptrace_hw_breakpoint(current);
e0780b72 909#else /* CONFIG_HAVE_HW_BREAKPOINT */
3bffb652 910 set_debug_reg_defaults(&current->thread);
e0780b72 911#endif /* CONFIG_HAVE_HW_BREAKPOINT */
14cf11af
PM
912}
913
914void
915release_thread(struct task_struct *t)
916{
917}
918
919/*
55ccf3fe
SS
920 * this gets called so that we can store coprocessor state into memory and
921 * copy the current task into the new thread.
14cf11af 922 */
55ccf3fe 923int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
14cf11af 924{
55ccf3fe
SS
925 flush_fp_to_thread(src);
926 flush_altivec_to_thread(src);
927 flush_vsx_to_thread(src);
928 flush_spe_to_thread(src);
330a1eb7 929
55ccf3fe 930 *dst = *src;
330a1eb7
ME
931
932 clear_task_ebb(dst);
933
55ccf3fe 934 return 0;
14cf11af
PM
935}
936
937/*
938 * Copy a thread..
939 */
efcac658
AK
940extern unsigned long dscr_default; /* defined in arch/powerpc/kernel/sysfs.c */
941
6f2c55b8 942int copy_thread(unsigned long clone_flags, unsigned long usp,
afa86fc4 943 unsigned long arg, struct task_struct *p)
14cf11af
PM
944{
945 struct pt_regs *childregs, *kregs;
946 extern void ret_from_fork(void);
58254e10
AV
947 extern void ret_from_kernel_thread(void);
948 void (*f)(void);
0cec6fd1 949 unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
14cf11af 950
14cf11af
PM
951 /* Copy registers */
952 sp -= sizeof(struct pt_regs);
953 childregs = (struct pt_regs *) sp;
ab75819d 954 if (unlikely(p->flags & PF_KTHREAD)) {
138d1ce8 955 struct thread_info *ti = (void *)task_stack_page(p);
58254e10 956 memset(childregs, 0, sizeof(struct pt_regs));
14cf11af 957 childregs->gpr[1] = sp + sizeof(struct pt_regs);
53b50f94 958 childregs->gpr[14] = usp; /* function */
58254e10 959#ifdef CONFIG_PPC64
b5e2fc1c 960 clear_tsk_thread_flag(p, TIF_32BIT);
138d1ce8 961 childregs->softe = 1;
06d67d54 962#endif
58254e10 963 childregs->gpr[15] = arg;
14cf11af 964 p->thread.regs = NULL; /* no user register state */
138d1ce8 965 ti->flags |= _TIF_RESTOREALL;
58254e10 966 f = ret_from_kernel_thread;
14cf11af 967 } else {
afa86fc4 968 struct pt_regs *regs = current_pt_regs();
58254e10
AV
969 CHECK_FULL_REGS(regs);
970 *childregs = *regs;
ea516b11
AV
971 if (usp)
972 childregs->gpr[1] = usp;
14cf11af 973 p->thread.regs = childregs;
58254e10 974 childregs->gpr[3] = 0; /* Result from fork() */
06d67d54
PM
975 if (clone_flags & CLONE_SETTLS) {
976#ifdef CONFIG_PPC64
9904b005 977 if (!is_32bit_task())
06d67d54
PM
978 childregs->gpr[13] = childregs->gpr[6];
979 else
980#endif
981 childregs->gpr[2] = childregs->gpr[6];
982 }
58254e10
AV
983
984 f = ret_from_fork;
14cf11af 985 }
14cf11af 986 sp -= STACK_FRAME_OVERHEAD;
14cf11af
PM
987
988 /*
989 * The way this works is that at some point in the future
990 * some task will call _switch to switch to the new task.
991 * That will pop off the stack frame created below and start
992 * the new task running at ret_from_fork. The new task will
993 * do some house keeping and then return from the fork or clone
994 * system call, using the stack frame created above.
995 */
af945cf4 996 ((unsigned long *)sp)[0] = 0;
14cf11af
PM
997 sp -= sizeof(struct pt_regs);
998 kregs = (struct pt_regs *) sp;
999 sp -= STACK_FRAME_OVERHEAD;
1000 p->thread.ksp = sp;
cbc9565e 1001#ifdef CONFIG_PPC32
85218827
KG
1002 p->thread.ksp_limit = (unsigned long)task_stack_page(p) +
1003 _ALIGN_UP(sizeof(struct thread_info), 16);
cbc9565e 1004#endif
28d170ab
ON
1005#ifdef CONFIG_HAVE_HW_BREAKPOINT
1006 p->thread.ptrace_bps[0] = NULL;
1007#endif
1008
18461960
PM
1009 p->thread.fp_save_area = NULL;
1010#ifdef CONFIG_ALTIVEC
1011 p->thread.vr_save_area = NULL;
1012#endif
1013
94491685 1014#ifdef CONFIG_PPC_STD_MMU_64
44ae3ab3 1015 if (mmu_has_feature(MMU_FTR_SLB)) {
1189be65 1016 unsigned long sp_vsid;
3c726f8d 1017 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
06d67d54 1018
44ae3ab3 1019 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
1189be65
PM
1020 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
1021 << SLB_VSID_SHIFT_1T;
1022 else
1023 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_256M)
1024 << SLB_VSID_SHIFT;
3c726f8d 1025 sp_vsid |= SLB_VSID_KERNEL | llp;
06d67d54
PM
1026 p->thread.ksp_vsid = sp_vsid;
1027 }
747bea91 1028#endif /* CONFIG_PPC_STD_MMU_64 */
efcac658
AK
1029#ifdef CONFIG_PPC64
1030 if (cpu_has_feature(CPU_FTR_DSCR)) {
1021cb26
AB
1031 p->thread.dscr_inherit = current->thread.dscr_inherit;
1032 p->thread.dscr = current->thread.dscr;
efcac658 1033 }
92779245
HM
1034 if (cpu_has_feature(CPU_FTR_HAS_PPR))
1035 p->thread.ppr = INIT_PPR;
efcac658 1036#endif
06d67d54
PM
1037 /*
1038 * The PPC64 ABI makes use of a TOC to contain function
1039 * pointers. The function (ret_from_except) is actually a pointer
1040 * to the TOC entry. The first entry is a pointer to the actual
1041 * function.
58254e10 1042 */
747bea91 1043#ifdef CONFIG_PPC64
58254e10 1044 kregs->nip = *((unsigned long *)f);
06d67d54 1045#else
58254e10 1046 kregs->nip = (unsigned long)f;
06d67d54 1047#endif
14cf11af
PM
1048 return 0;
1049}
1050
1051/*
1052 * Set up a thread for executing a new program
1053 */
06d67d54 1054void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
14cf11af 1055{
90eac727
ME
1056#ifdef CONFIG_PPC64
1057 unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
1058#endif
1059
06d67d54
PM
1060 /*
1061 * If we exec out of a kernel thread then thread.regs will not be
1062 * set. Do it now.
1063 */
1064 if (!current->thread.regs) {
0cec6fd1
AV
1065 struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
1066 current->thread.regs = regs - 1;
06d67d54
PM
1067 }
1068
14cf11af
PM
1069 memset(regs->gpr, 0, sizeof(regs->gpr));
1070 regs->ctr = 0;
1071 regs->link = 0;
1072 regs->xer = 0;
1073 regs->ccr = 0;
14cf11af 1074 regs->gpr[1] = sp;
06d67d54 1075
474f8196
RM
1076 /*
1077 * We have just cleared all the nonvolatile GPRs, so make
1078 * FULL_REGS(regs) return true. This is necessary to allow
1079 * ptrace to examine the thread immediately after exec.
1080 */
1081 regs->trap &= ~1UL;
1082
06d67d54
PM
1083#ifdef CONFIG_PPC32
1084 regs->mq = 0;
1085 regs->nip = start;
14cf11af 1086 regs->msr = MSR_USER;
06d67d54 1087#else
9904b005 1088 if (!is_32bit_task()) {
90eac727 1089 unsigned long entry, toc;
06d67d54
PM
1090
1091 /* start is a relocated pointer to the function descriptor for
1092 * the elf _start routine. The first entry in the function
1093 * descriptor is the entry address of _start and the second
1094 * entry is the TOC value we need to use.
1095 */
1096 __get_user(entry, (unsigned long __user *)start);
1097 __get_user(toc, (unsigned long __user *)start+1);
1098
1099 /* Check whether the e_entry function descriptor entries
1100 * need to be relocated before we can use them.
1101 */
1102 if (load_addr != 0) {
1103 entry += load_addr;
1104 toc += load_addr;
1105 }
1106 regs->nip = entry;
1107 regs->gpr[2] = toc;
1108 regs->msr = MSR_USER64;
d4bf9a78
SR
1109 } else {
1110 regs->nip = start;
1111 regs->gpr[2] = 0;
1112 regs->msr = MSR_USER32;
06d67d54
PM
1113 }
1114#endif
48abec07 1115 discard_lazy_cpu_state();
ce48b210
MN
1116#ifdef CONFIG_VSX
1117 current->thread.used_vsr = 0;
1118#endif
de79f7b9 1119 memset(&current->thread.fp_state, 0, sizeof(current->thread.fp_state));
18461960 1120 current->thread.fp_save_area = NULL;
14cf11af 1121#ifdef CONFIG_ALTIVEC
de79f7b9
PM
1122 memset(&current->thread.vr_state, 0, sizeof(current->thread.vr_state));
1123 current->thread.vr_state.vscr.u[3] = 0x00010000; /* Java mode disabled */
18461960 1124 current->thread.vr_save_area = NULL;
14cf11af
PM
1125 current->thread.vrsave = 0;
1126 current->thread.used_vr = 0;
1127#endif /* CONFIG_ALTIVEC */
1128#ifdef CONFIG_SPE
1129 memset(current->thread.evr, 0, sizeof(current->thread.evr));
1130 current->thread.acc = 0;
1131 current->thread.spefscr = 0;
1132 current->thread.used_spe = 0;
1133#endif /* CONFIG_SPE */
bc2a9408
MN
1134#ifdef CONFIG_PPC_TRANSACTIONAL_MEM
1135 if (cpu_has_feature(CPU_FTR_TM))
1136 regs->msr |= MSR_TM;
1137 current->thread.tm_tfhar = 0;
1138 current->thread.tm_texasr = 0;
1139 current->thread.tm_tfiar = 0;
1140#endif /* CONFIG_PPC_TRANSACTIONAL_MEM */
14cf11af
PM
1141}
1142
1143#define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
1144 | PR_FP_EXC_RES | PR_FP_EXC_INV)
1145
1146int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
1147{
1148 struct pt_regs *regs = tsk->thread.regs;
1149
1150 /* This is a bit hairy. If we are an SPE enabled processor
1151 * (have embedded fp) we store the IEEE exception enable flags in
1152 * fpexc_mode. fpexc_mode is also used for setting FP exception
1153 * mode (asyn, precise, disabled) for 'Classic' FP. */
1154 if (val & PR_FP_EXC_SW_ENABLE) {
1155#ifdef CONFIG_SPE
5e14d21e
KG
1156 if (cpu_has_feature(CPU_FTR_SPE)) {
1157 tsk->thread.fpexc_mode = val &
1158 (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
1159 return 0;
1160 } else {
1161 return -EINVAL;
1162 }
14cf11af
PM
1163#else
1164 return -EINVAL;
1165#endif
14cf11af 1166 }
06d67d54
PM
1167
1168 /* on a CONFIG_SPE this does not hurt us. The bits that
1169 * __pack_fe01 use do not overlap with bits used for
1170 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
1171 * on CONFIG_SPE implementations are reserved so writing to
1172 * them does not change anything */
1173 if (val > PR_FP_EXC_PRECISE)
1174 return -EINVAL;
1175 tsk->thread.fpexc_mode = __pack_fe01(val);
1176 if (regs != NULL && (regs->msr & MSR_FP) != 0)
1177 regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
1178 | tsk->thread.fpexc_mode;
14cf11af
PM
1179 return 0;
1180}
1181
1182int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
1183{
1184 unsigned int val;
1185
1186 if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
1187#ifdef CONFIG_SPE
5e14d21e
KG
1188 if (cpu_has_feature(CPU_FTR_SPE))
1189 val = tsk->thread.fpexc_mode;
1190 else
1191 return -EINVAL;
14cf11af
PM
1192#else
1193 return -EINVAL;
1194#endif
1195 else
1196 val = __unpack_fe01(tsk->thread.fpexc_mode);
1197 return put_user(val, (unsigned int __user *) adr);
1198}
1199
fab5db97
PM
1200int set_endian(struct task_struct *tsk, unsigned int val)
1201{
1202 struct pt_regs *regs = tsk->thread.regs;
1203
1204 if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
1205 (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
1206 return -EINVAL;
1207
1208 if (regs == NULL)
1209 return -EINVAL;
1210
1211 if (val == PR_ENDIAN_BIG)
1212 regs->msr &= ~MSR_LE;
1213 else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
1214 regs->msr |= MSR_LE;
1215 else
1216 return -EINVAL;
1217
1218 return 0;
1219}
1220
1221int get_endian(struct task_struct *tsk, unsigned long adr)
1222{
1223 struct pt_regs *regs = tsk->thread.regs;
1224 unsigned int val;
1225
1226 if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
1227 !cpu_has_feature(CPU_FTR_REAL_LE))
1228 return -EINVAL;
1229
1230 if (regs == NULL)
1231 return -EINVAL;
1232
1233 if (regs->msr & MSR_LE) {
1234 if (cpu_has_feature(CPU_FTR_REAL_LE))
1235 val = PR_ENDIAN_LITTLE;
1236 else
1237 val = PR_ENDIAN_PPC_LITTLE;
1238 } else
1239 val = PR_ENDIAN_BIG;
1240
1241 return put_user(val, (unsigned int __user *)adr);
1242}
1243
e9370ae1
PM
1244int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
1245{
1246 tsk->thread.align_ctl = val;
1247 return 0;
1248}
1249
1250int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
1251{
1252 return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
1253}
1254
bb72c481
PM
1255static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
1256 unsigned long nbytes)
1257{
1258 unsigned long stack_page;
1259 unsigned long cpu = task_cpu(p);
1260
1261 /*
1262 * Avoid crashing if the stack has overflowed and corrupted
1263 * task_cpu(p), which is in the thread_info struct.
1264 */
1265 if (cpu < NR_CPUS && cpu_possible(cpu)) {
1266 stack_page = (unsigned long) hardirq_ctx[cpu];
1267 if (sp >= stack_page + sizeof(struct thread_struct)
1268 && sp <= stack_page + THREAD_SIZE - nbytes)
1269 return 1;
1270
1271 stack_page = (unsigned long) softirq_ctx[cpu];
1272 if (sp >= stack_page + sizeof(struct thread_struct)
1273 && sp <= stack_page + THREAD_SIZE - nbytes)
1274 return 1;
1275 }
1276 return 0;
1277}
1278
2f25194d 1279int validate_sp(unsigned long sp, struct task_struct *p,
14cf11af
PM
1280 unsigned long nbytes)
1281{
0cec6fd1 1282 unsigned long stack_page = (unsigned long)task_stack_page(p);
14cf11af
PM
1283
1284 if (sp >= stack_page + sizeof(struct thread_struct)
1285 && sp <= stack_page + THREAD_SIZE - nbytes)
1286 return 1;
1287
bb72c481 1288 return valid_irq_stack(sp, p, nbytes);
14cf11af
PM
1289}
1290
2f25194d
AB
1291EXPORT_SYMBOL(validate_sp);
1292
14cf11af
PM
1293unsigned long get_wchan(struct task_struct *p)
1294{
1295 unsigned long ip, sp;
1296 int count = 0;
1297
1298 if (!p || p == current || p->state == TASK_RUNNING)
1299 return 0;
1300
1301 sp = p->thread.ksp;
ec2b36b9 1302 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
14cf11af
PM
1303 return 0;
1304
1305 do {
1306 sp = *(unsigned long *)sp;
ec2b36b9 1307 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
14cf11af
PM
1308 return 0;
1309 if (count > 0) {
ec2b36b9 1310 ip = ((unsigned long *)sp)[STACK_FRAME_LR_SAVE];
14cf11af
PM
1311 if (!in_sched_functions(ip))
1312 return ip;
1313 }
1314 } while (count++ < 16);
1315 return 0;
1316}
06d67d54 1317
c4d04be1 1318static int kstack_depth_to_print = CONFIG_PRINT_STACK_DEPTH;
06d67d54
PM
1319
1320void show_stack(struct task_struct *tsk, unsigned long *stack)
1321{
1322 unsigned long sp, ip, lr, newsp;
1323 int count = 0;
1324 int firstframe = 1;
6794c782
SR
1325#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1326 int curr_frame = current->curr_ret_stack;
1327 extern void return_to_handler(void);
9135c3cc
SR
1328 unsigned long rth = (unsigned long)return_to_handler;
1329 unsigned long mrth = -1;
6794c782 1330#ifdef CONFIG_PPC64
9135c3cc
SR
1331 extern void mod_return_to_handler(void);
1332 rth = *(unsigned long *)rth;
1333 mrth = (unsigned long)mod_return_to_handler;
1334 mrth = *(unsigned long *)mrth;
6794c782
SR
1335#endif
1336#endif
06d67d54
PM
1337
1338 sp = (unsigned long) stack;
1339 if (tsk == NULL)
1340 tsk = current;
1341 if (sp == 0) {
1342 if (tsk == current)
1343 asm("mr %0,1" : "=r" (sp));
1344 else
1345 sp = tsk->thread.ksp;
1346 }
1347
1348 lr = 0;
1349 printk("Call Trace:\n");
1350 do {
ec2b36b9 1351 if (!validate_sp(sp, tsk, STACK_FRAME_OVERHEAD))
06d67d54
PM
1352 return;
1353
1354 stack = (unsigned long *) sp;
1355 newsp = stack[0];
ec2b36b9 1356 ip = stack[STACK_FRAME_LR_SAVE];
06d67d54 1357 if (!firstframe || ip != lr) {
058c78f4 1358 printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip);
6794c782 1359#ifdef CONFIG_FUNCTION_GRAPH_TRACER
9135c3cc 1360 if ((ip == rth || ip == mrth) && curr_frame >= 0) {
6794c782
SR
1361 printk(" (%pS)",
1362 (void *)current->ret_stack[curr_frame].ret);
1363 curr_frame--;
1364 }
1365#endif
06d67d54
PM
1366 if (firstframe)
1367 printk(" (unreliable)");
1368 printk("\n");
1369 }
1370 firstframe = 0;
1371
1372 /*
1373 * See if this is an exception frame.
1374 * We look for the "regshere" marker in the current frame.
1375 */
ec2b36b9
BH
1376 if (validate_sp(sp, tsk, STACK_INT_FRAME_SIZE)
1377 && stack[STACK_FRAME_MARKER] == STACK_FRAME_REGS_MARKER) {
06d67d54
PM
1378 struct pt_regs *regs = (struct pt_regs *)
1379 (sp + STACK_FRAME_OVERHEAD);
06d67d54 1380 lr = regs->link;
058c78f4
BH
1381 printk("--- Exception: %lx at %pS\n LR = %pS\n",
1382 regs->trap, (void *)regs->nip, (void *)lr);
06d67d54
PM
1383 firstframe = 1;
1384 }
1385
1386 sp = newsp;
1387 } while (count++ < kstack_depth_to_print);
1388}
1389
cb2c9b27 1390#ifdef CONFIG_PPC64
fe1952fc 1391/* Called with hard IRQs off */
0e37739b 1392void notrace __ppc64_runlatch_on(void)
cb2c9b27 1393{
fe1952fc 1394 struct thread_info *ti = current_thread_info();
cb2c9b27
AB
1395 unsigned long ctrl;
1396
fe1952fc
BH
1397 ctrl = mfspr(SPRN_CTRLF);
1398 ctrl |= CTRL_RUNLATCH;
1399 mtspr(SPRN_CTRLT, ctrl);
cb2c9b27 1400
fae2e0fb 1401 ti->local_flags |= _TLF_RUNLATCH;
cb2c9b27
AB
1402}
1403
fe1952fc 1404/* Called with hard IRQs off */
0e37739b 1405void notrace __ppc64_runlatch_off(void)
cb2c9b27 1406{
fe1952fc 1407 struct thread_info *ti = current_thread_info();
cb2c9b27
AB
1408 unsigned long ctrl;
1409
fae2e0fb 1410 ti->local_flags &= ~_TLF_RUNLATCH;
cb2c9b27 1411
4138d653
AB
1412 ctrl = mfspr(SPRN_CTRLF);
1413 ctrl &= ~CTRL_RUNLATCH;
1414 mtspr(SPRN_CTRLT, ctrl);
cb2c9b27 1415}
fe1952fc 1416#endif /* CONFIG_PPC64 */
f6a61680 1417
d839088c
AB
1418unsigned long arch_align_stack(unsigned long sp)
1419{
1420 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
1421 sp -= get_random_int() & ~PAGE_MASK;
1422 return sp & ~0xf;
1423}
912f9ee2
AB
1424
1425static inline unsigned long brk_rnd(void)
1426{
1427 unsigned long rnd = 0;
1428
1429 /* 8MB for 32bit, 1GB for 64bit */
1430 if (is_32bit_task())
1431 rnd = (long)(get_random_int() % (1<<(23-PAGE_SHIFT)));
1432 else
1433 rnd = (long)(get_random_int() % (1<<(30-PAGE_SHIFT)));
1434
1435 return rnd << PAGE_SHIFT;
1436}
1437
1438unsigned long arch_randomize_brk(struct mm_struct *mm)
1439{
8bbde7a7
AB
1440 unsigned long base = mm->brk;
1441 unsigned long ret;
1442
ce7a35c7 1443#ifdef CONFIG_PPC_STD_MMU_64
8bbde7a7
AB
1444 /*
1445 * If we are using 1TB segments and we are allowed to randomise
1446 * the heap, we can put it above 1TB so it is backed by a 1TB
1447 * segment. Otherwise the heap will be in the bottom 1TB
1448 * which always uses 256MB segments and this may result in a
1449 * performance penalty.
1450 */
1451 if (!is_32bit_task() && (mmu_highuser_ssize == MMU_SEGSIZE_1T))
1452 base = max_t(unsigned long, mm->brk, 1UL << SID_SHIFT_1T);
1453#endif
1454
1455 ret = PAGE_ALIGN(base + brk_rnd());
912f9ee2
AB
1456
1457 if (ret < mm->brk)
1458 return mm->brk;
1459
1460 return ret;
1461}
501cb16d
AB
1462
1463unsigned long randomize_et_dyn(unsigned long base)
1464{
1465 unsigned long ret = PAGE_ALIGN(base + brk_rnd());
1466
1467 if (ret < base)
1468 return base;
1469
1470 return ret;
1471}
This page took 0.624612 seconds and 5 git commands to generate.