]>
git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blob - arch/powerpc/kernel/process.c
2 * Derived from "arch/i386/kernel/process.c"
3 * Copyright (C) 1995 Linus Torvalds
5 * Updated and modified by Cort Dougan (cort@cs.nmt.edu) and
6 * Paul Mackerras (paulus@cs.anu.edu.au)
9 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
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.
17 #include <linux/errno.h>
18 #include <linux/sched.h>
19 #include <linux/kernel.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>
42 #include <asm/pgtable.h>
43 #include <asm/uaccess.h>
44 #include <asm/system.h>
46 #include <asm/processor.h>
49 #include <asm/machdep.h>
51 #include <asm/syscalls.h>
53 #include <asm/firmware.h>
55 #include <linux/kprobes.h>
56 #include <linux/kdebug.h>
58 extern unsigned long _get_SP(void);
61 struct task_struct
*last_task_used_math
= NULL
;
62 struct task_struct
*last_task_used_altivec
= NULL
;
63 struct task_struct
*last_task_used_vsx
= NULL
;
64 struct task_struct
*last_task_used_spe
= NULL
;
68 * Make sure the floating-point register state in the
69 * the thread_struct is up to date for task tsk.
71 void flush_fp_to_thread(struct task_struct
*tsk
)
73 if (tsk
->thread
.regs
) {
75 * We need to disable preemption here because if we didn't,
76 * another process could get scheduled after the regs->msr
77 * test but before we have finished saving the FP registers
78 * to the thread_struct. That process could take over the
79 * FPU, and then when we get scheduled again we would store
80 * bogus values for the remaining FP registers.
83 if (tsk
->thread
.regs
->msr
& MSR_FP
) {
86 * This should only ever be called for current or
87 * for a stopped child process. Since we save away
88 * the FP register state on context switch on SMP,
89 * there is something wrong if a stopped child appears
90 * to still have its FP state in the CPU registers.
92 BUG_ON(tsk
!= current
);
99 EXPORT_SYMBOL_GPL(flush_fp_to_thread
);
101 void enable_kernel_fp(void)
103 WARN_ON(preemptible());
106 if (current
->thread
.regs
&& (current
->thread
.regs
->msr
& MSR_FP
))
109 giveup_fpu(NULL
); /* just enables FP for kernel */
111 giveup_fpu(last_task_used_math
);
112 #endif /* CONFIG_SMP */
114 EXPORT_SYMBOL(enable_kernel_fp
);
116 #ifdef CONFIG_ALTIVEC
117 void enable_kernel_altivec(void)
119 WARN_ON(preemptible());
122 if (current
->thread
.regs
&& (current
->thread
.regs
->msr
& MSR_VEC
))
123 giveup_altivec(current
);
125 giveup_altivec(NULL
); /* just enable AltiVec for kernel - force */
127 giveup_altivec(last_task_used_altivec
);
128 #endif /* CONFIG_SMP */
130 EXPORT_SYMBOL(enable_kernel_altivec
);
133 * Make sure the VMX/Altivec register state in the
134 * the thread_struct is up to date for task tsk.
136 void flush_altivec_to_thread(struct task_struct
*tsk
)
138 if (tsk
->thread
.regs
) {
140 if (tsk
->thread
.regs
->msr
& MSR_VEC
) {
142 BUG_ON(tsk
!= current
);
149 EXPORT_SYMBOL_GPL(flush_altivec_to_thread
);
150 #endif /* CONFIG_ALTIVEC */
154 /* not currently used, but some crazy RAID module might want to later */
155 void enable_kernel_vsx(void)
157 WARN_ON(preemptible());
160 if (current
->thread
.regs
&& (current
->thread
.regs
->msr
& MSR_VSX
))
163 giveup_vsx(NULL
); /* just enable vsx for kernel - force */
165 giveup_vsx(last_task_used_vsx
);
166 #endif /* CONFIG_SMP */
168 EXPORT_SYMBOL(enable_kernel_vsx
);
171 void giveup_vsx(struct task_struct
*tsk
)
178 void flush_vsx_to_thread(struct task_struct
*tsk
)
180 if (tsk
->thread
.regs
) {
182 if (tsk
->thread
.regs
->msr
& MSR_VSX
) {
184 BUG_ON(tsk
!= current
);
191 EXPORT_SYMBOL_GPL(flush_vsx_to_thread
);
192 #endif /* CONFIG_VSX */
196 void enable_kernel_spe(void)
198 WARN_ON(preemptible());
201 if (current
->thread
.regs
&& (current
->thread
.regs
->msr
& MSR_SPE
))
204 giveup_spe(NULL
); /* just enable SPE for kernel - force */
206 giveup_spe(last_task_used_spe
);
207 #endif /* __SMP __ */
209 EXPORT_SYMBOL(enable_kernel_spe
);
211 void flush_spe_to_thread(struct task_struct
*tsk
)
213 if (tsk
->thread
.regs
) {
215 if (tsk
->thread
.regs
->msr
& MSR_SPE
) {
217 BUG_ON(tsk
!= current
);
219 tsk
->thread
.spefscr
= mfspr(SPRN_SPEFSCR
);
225 #endif /* CONFIG_SPE */
229 * If we are doing lazy switching of CPU state (FP, altivec or SPE),
230 * and the current task has some state, discard it.
232 void discard_lazy_cpu_state(void)
235 if (last_task_used_math
== current
)
236 last_task_used_math
= NULL
;
237 #ifdef CONFIG_ALTIVEC
238 if (last_task_used_altivec
== current
)
239 last_task_used_altivec
= NULL
;
240 #endif /* CONFIG_ALTIVEC */
242 if (last_task_used_vsx
== current
)
243 last_task_used_vsx
= NULL
;
244 #endif /* CONFIG_VSX */
246 if (last_task_used_spe
== current
)
247 last_task_used_spe
= NULL
;
251 #endif /* CONFIG_SMP */
253 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
254 void do_send_trap(struct pt_regs
*regs
, unsigned long address
,
255 unsigned long error_code
, int signal_code
, int breakpt
)
259 if (notify_die(DIE_DABR_MATCH
, "dabr_match", regs
, error_code
,
260 11, SIGSEGV
) == NOTIFY_STOP
)
263 /* Deliver the signal to userspace */
264 info
.si_signo
= SIGTRAP
;
265 info
.si_errno
= breakpt
; /* breakpoint or watchpoint id */
266 info
.si_code
= signal_code
;
267 info
.si_addr
= (void __user
*)address
;
268 force_sig_info(SIGTRAP
, &info
, current
);
270 #else /* !CONFIG_PPC_ADV_DEBUG_REGS */
271 void do_dabr(struct pt_regs
*regs
, unsigned long address
,
272 unsigned long error_code
)
276 if (notify_die(DIE_DABR_MATCH
, "dabr_match", regs
, error_code
,
277 11, SIGSEGV
) == NOTIFY_STOP
)
280 if (debugger_dabr_match(regs
))
286 /* Deliver the signal to userspace */
287 info
.si_signo
= SIGTRAP
;
289 info
.si_code
= TRAP_HWBKPT
;
290 info
.si_addr
= (void __user
*)address
;
291 force_sig_info(SIGTRAP
, &info
, current
);
293 #endif /* CONFIG_PPC_ADV_DEBUG_REGS */
295 static DEFINE_PER_CPU(unsigned long, current_dabr
);
297 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
299 * Set the debug registers back to their default "safe" values.
301 static void set_debug_reg_defaults(struct thread_struct
*thread
)
303 thread
->iac1
= thread
->iac2
= 0;
304 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
305 thread
->iac3
= thread
->iac4
= 0;
307 thread
->dac1
= thread
->dac2
= 0;
308 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
309 thread
->dvc1
= thread
->dvc2
= 0;
314 * Force User/Supervisor bits to b11 (user-only MSR[PR]=1)
316 thread
->dbcr1
= DBCR1_IAC1US
| DBCR1_IAC2US
| \
317 DBCR1_IAC3US
| DBCR1_IAC4US
;
319 * Force Data Address Compare User/Supervisor bits to be User-only
320 * (0b11 MSR[PR]=1) and set all other bits in DBCR2 register to be 0.
322 thread
->dbcr2
= DBCR2_DAC1US
| DBCR2_DAC2US
;
328 static void prime_debug_regs(struct thread_struct
*thread
)
330 mtspr(SPRN_IAC1
, thread
->iac1
);
331 mtspr(SPRN_IAC2
, thread
->iac2
);
332 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
333 mtspr(SPRN_IAC3
, thread
->iac3
);
334 mtspr(SPRN_IAC4
, thread
->iac4
);
336 mtspr(SPRN_DAC1
, thread
->dac1
);
337 mtspr(SPRN_DAC2
, thread
->dac2
);
338 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
339 mtspr(SPRN_DVC1
, thread
->dvc1
);
340 mtspr(SPRN_DVC2
, thread
->dvc2
);
342 mtspr(SPRN_DBCR0
, thread
->dbcr0
);
343 mtspr(SPRN_DBCR1
, thread
->dbcr1
);
345 mtspr(SPRN_DBCR2
, thread
->dbcr2
);
349 * Unless neither the old or new thread are making use of the
350 * debug registers, set the debug registers from the values
351 * stored in the new thread.
353 static void switch_booke_debug_regs(struct thread_struct
*new_thread
)
355 if ((current
->thread
.dbcr0
& DBCR0_IDM
)
356 || (new_thread
->dbcr0
& DBCR0_IDM
))
357 prime_debug_regs(new_thread
);
359 #else /* !CONFIG_PPC_ADV_DEBUG_REGS */
360 #ifndef CONFIG_HAVE_HW_BREAKPOINT
361 static void set_debug_reg_defaults(struct thread_struct
*thread
)
368 #endif /* !CONFIG_HAVE_HW_BREAKPOINT */
369 #endif /* CONFIG_PPC_ADV_DEBUG_REGS */
371 int set_dabr(unsigned long dabr
)
373 __get_cpu_var(current_dabr
) = dabr
;
376 return ppc_md
.set_dabr(dabr
);
378 /* XXX should we have a CPU_FTR_HAS_DABR ? */
379 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
380 mtspr(SPRN_DAC1
, dabr
);
381 #ifdef CONFIG_PPC_47x
384 #elif defined(CONFIG_PPC_BOOK3S)
385 mtspr(SPRN_DABR
, dabr
);
393 DEFINE_PER_CPU(struct cpu_usage
, cpu_usage_array
);
396 struct task_struct
*__switch_to(struct task_struct
*prev
,
397 struct task_struct
*new)
399 struct thread_struct
*new_thread
, *old_thread
;
401 struct task_struct
*last
;
402 #ifdef CONFIG_PPC_BOOK3S_64
403 struct ppc64_tlb_batch
*batch
;
407 /* avoid complexity of lazy save/restore of fpu
408 * by just saving it every time we switch out if
409 * this task used the fpu during the last quantum.
411 * If it tries to use the fpu again, it'll trap and
412 * reload its fp regs. So we don't have to do a restore
413 * every switch, just a save.
416 if (prev
->thread
.regs
&& (prev
->thread
.regs
->msr
& MSR_FP
))
418 #ifdef CONFIG_ALTIVEC
420 * If the previous thread used altivec in the last quantum
421 * (thus changing altivec regs) then save them.
422 * We used to check the VRSAVE register but not all apps
423 * set it, so we don't rely on it now (and in fact we need
424 * to save & restore VSCR even if VRSAVE == 0). -- paulus
426 * On SMP we always save/restore altivec regs just to avoid the
427 * complexity of changing processors.
430 if (prev
->thread
.regs
&& (prev
->thread
.regs
->msr
& MSR_VEC
))
431 giveup_altivec(prev
);
432 #endif /* CONFIG_ALTIVEC */
434 if (prev
->thread
.regs
&& (prev
->thread
.regs
->msr
& MSR_VSX
))
435 /* VMX and FPU registers are already save here */
437 #endif /* CONFIG_VSX */
440 * If the previous thread used spe in the last quantum
441 * (thus changing spe regs) then save them.
443 * On SMP we always save/restore spe regs just to avoid the
444 * complexity of changing processors.
446 if ((prev
->thread
.regs
&& (prev
->thread
.regs
->msr
& MSR_SPE
)))
448 #endif /* CONFIG_SPE */
450 #else /* CONFIG_SMP */
451 #ifdef CONFIG_ALTIVEC
452 /* Avoid the trap. On smp this this never happens since
453 * we don't set last_task_used_altivec -- Cort
455 if (new->thread
.regs
&& last_task_used_altivec
== new)
456 new->thread
.regs
->msr
|= MSR_VEC
;
457 #endif /* CONFIG_ALTIVEC */
459 if (new->thread
.regs
&& last_task_used_vsx
== new)
460 new->thread
.regs
->msr
|= MSR_VSX
;
461 #endif /* CONFIG_VSX */
463 /* Avoid the trap. On smp this this never happens since
464 * we don't set last_task_used_spe
466 if (new->thread
.regs
&& last_task_used_spe
== new)
467 new->thread
.regs
->msr
|= MSR_SPE
;
468 #endif /* CONFIG_SPE */
470 #endif /* CONFIG_SMP */
472 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
473 switch_booke_debug_regs(&new->thread
);
476 * For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
479 #ifndef CONFIG_HAVE_HW_BREAKPOINT
480 if (unlikely(__get_cpu_var(current_dabr
) != new->thread
.dabr
))
481 set_dabr(new->thread
.dabr
);
482 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
486 new_thread
= &new->thread
;
487 old_thread
= ¤t
->thread
;
491 * Collect processor utilization data per process
493 if (firmware_has_feature(FW_FEATURE_SPLPAR
)) {
494 struct cpu_usage
*cu
= &__get_cpu_var(cpu_usage_array
);
495 long unsigned start_tb
, current_tb
;
496 start_tb
= old_thread
->start_tb
;
497 cu
->current_tb
= current_tb
= mfspr(SPRN_PURR
);
498 old_thread
->accum_tb
+= (current_tb
- start_tb
);
499 new_thread
->start_tb
= current_tb
;
501 #endif /* CONFIG_PPC64 */
503 #ifdef CONFIG_PPC_BOOK3S_64
504 batch
= &__get_cpu_var(ppc64_tlb_batch
);
506 current_thread_info()->local_flags
|= _TLF_LAZY_MMU
;
508 __flush_tlb_pending(batch
);
511 #endif /* CONFIG_PPC_BOOK3S_64 */
513 local_irq_save(flags
);
515 account_system_vtime(current
);
516 account_process_vtime(current
);
519 * We can't take a PMU exception inside _switch() since there is a
520 * window where the kernel stack SLB and the kernel stack are out
521 * of sync. Hard disable here.
524 last
= _switch(old_thread
, new_thread
);
526 #ifdef CONFIG_PPC_BOOK3S_64
527 if (current_thread_info()->local_flags
& _TLF_LAZY_MMU
) {
528 current_thread_info()->local_flags
&= ~_TLF_LAZY_MMU
;
529 batch
= &__get_cpu_var(ppc64_tlb_batch
);
532 #endif /* CONFIG_PPC_BOOK3S_64 */
534 local_irq_restore(flags
);
539 static int instructions_to_print
= 16;
541 static void show_instructions(struct pt_regs
*regs
)
544 unsigned long pc
= regs
->nip
- (instructions_to_print
* 3 / 4 *
547 printk("Instruction dump:");
549 for (i
= 0; i
< instructions_to_print
; i
++) {
555 #if !defined(CONFIG_BOOKE)
556 /* If executing with the IMMU off, adjust pc rather
557 * than print XXXXXXXX.
559 if (!(regs
->msr
& MSR_IR
))
560 pc
= (unsigned long)phys_to_virt(pc
);
563 /* We use __get_user here *only* to avoid an OOPS on a
564 * bad address because the pc *should* only be a
567 if (!__kernel_text_address(pc
) ||
568 __get_user(instr
, (unsigned int __user
*)pc
)) {
572 printk("<%08x> ", instr
);
574 printk("%08x ", instr
);
583 static struct regbit
{
600 static void printbits(unsigned long val
, struct regbit
*bits
)
602 const char *sep
= "";
605 for (; bits
->bit
; ++bits
)
606 if (val
& bits
->bit
) {
607 printk("%s%s", sep
, bits
->name
);
615 #define REGS_PER_LINE 4
616 #define LAST_VOLATILE 13
619 #define REGS_PER_LINE 8
620 #define LAST_VOLATILE 12
623 void show_regs(struct pt_regs
* regs
)
627 printk("NIP: "REG
" LR: "REG
" CTR: "REG
"\n",
628 regs
->nip
, regs
->link
, regs
->ctr
);
629 printk("REGS: %p TRAP: %04lx %s (%s)\n",
630 regs
, regs
->trap
, print_tainted(), init_utsname()->release
);
631 printk("MSR: "REG
" ", regs
->msr
);
632 printbits(regs
->msr
, msr_bits
);
633 printk(" CR: %08lx XER: %08lx\n", regs
->ccr
, regs
->xer
);
635 if ((regs
->trap
!= 0xc00) && cpu_has_feature(CPU_FTR_CFAR
))
636 printk("CFAR: "REG
"\n", regs
->orig_gpr3
);
637 if (trap
== 0x300 || trap
== 0x600)
638 #if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)
639 printk("DEAR: "REG
", ESR: "REG
"\n", regs
->dar
, regs
->dsisr
);
641 printk("DAR: "REG
", DSISR: %08lx\n", regs
->dar
, regs
->dsisr
);
643 printk("TASK = %p[%d] '%s' THREAD: %p",
644 current
, task_pid_nr(current
), current
->comm
, task_thread_info(current
));
647 printk(" CPU: %d", raw_smp_processor_id());
648 #endif /* CONFIG_SMP */
650 for (i
= 0; i
< 32; i
++) {
651 if ((i
% REGS_PER_LINE
) == 0)
652 printk("\nGPR%02d: ", i
);
653 printk(REG
" ", regs
->gpr
[i
]);
654 if (i
== LAST_VOLATILE
&& !FULL_REGS(regs
))
658 #ifdef CONFIG_KALLSYMS
660 * Lookup NIP late so we have the best change of getting the
661 * above info out without failing
663 printk("NIP ["REG
"] %pS\n", regs
->nip
, (void *)regs
->nip
);
664 printk("LR ["REG
"] %pS\n", regs
->link
, (void *)regs
->link
);
666 show_stack(current
, (unsigned long *) regs
->gpr
[1]);
667 if (!user_mode(regs
))
668 show_instructions(regs
);
671 void exit_thread(void)
673 discard_lazy_cpu_state();
676 void flush_thread(void)
678 discard_lazy_cpu_state();
680 #ifdef CONFIG_HAVE_HW_BREAKPOINT
681 flush_ptrace_hw_breakpoint(current
);
682 #else /* CONFIG_HAVE_HW_BREAKPOINT */
683 set_debug_reg_defaults(¤t
->thread
);
684 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
688 release_thread(struct task_struct
*t
)
693 * This gets called before we allocate a new thread and copy
694 * the current task into it.
696 void prepare_to_copy(struct task_struct
*tsk
)
698 flush_fp_to_thread(current
);
699 flush_altivec_to_thread(current
);
700 flush_vsx_to_thread(current
);
701 flush_spe_to_thread(current
);
702 #ifdef CONFIG_HAVE_HW_BREAKPOINT
703 flush_ptrace_hw_breakpoint(tsk
);
704 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
710 extern unsigned long dscr_default
; /* defined in arch/powerpc/kernel/sysfs.c */
712 int copy_thread(unsigned long clone_flags
, unsigned long usp
,
713 unsigned long unused
, struct task_struct
*p
,
714 struct pt_regs
*regs
)
716 struct pt_regs
*childregs
, *kregs
;
717 extern void ret_from_fork(void);
718 unsigned long sp
= (unsigned long)task_stack_page(p
) + THREAD_SIZE
;
720 CHECK_FULL_REGS(regs
);
722 sp
-= sizeof(struct pt_regs
);
723 childregs
= (struct pt_regs
*) sp
;
725 if ((childregs
->msr
& MSR_PR
) == 0) {
726 /* for kernel thread, set `current' and stackptr in new task */
727 childregs
->gpr
[1] = sp
+ sizeof(struct pt_regs
);
729 childregs
->gpr
[2] = (unsigned long) p
;
731 clear_tsk_thread_flag(p
, TIF_32BIT
);
733 p
->thread
.regs
= NULL
; /* no user register state */
735 childregs
->gpr
[1] = usp
;
736 p
->thread
.regs
= childregs
;
737 if (clone_flags
& CLONE_SETTLS
) {
739 if (!is_32bit_task())
740 childregs
->gpr
[13] = childregs
->gpr
[6];
743 childregs
->gpr
[2] = childregs
->gpr
[6];
746 childregs
->gpr
[3] = 0; /* Result from fork() */
747 sp
-= STACK_FRAME_OVERHEAD
;
750 * The way this works is that at some point in the future
751 * some task will call _switch to switch to the new task.
752 * That will pop off the stack frame created below and start
753 * the new task running at ret_from_fork. The new task will
754 * do some house keeping and then return from the fork or clone
755 * system call, using the stack frame created above.
757 sp
-= sizeof(struct pt_regs
);
758 kregs
= (struct pt_regs
*) sp
;
759 sp
-= STACK_FRAME_OVERHEAD
;
761 p
->thread
.ksp_limit
= (unsigned long)task_stack_page(p
) +
762 _ALIGN_UP(sizeof(struct thread_info
), 16);
764 #ifdef CONFIG_PPC_STD_MMU_64
765 if (mmu_has_feature(MMU_FTR_SLB
)) {
766 unsigned long sp_vsid
;
767 unsigned long llp
= mmu_psize_defs
[mmu_linear_psize
].sllp
;
769 if (mmu_has_feature(MMU_FTR_1T_SEGMENT
))
770 sp_vsid
= get_kernel_vsid(sp
, MMU_SEGSIZE_1T
)
771 << SLB_VSID_SHIFT_1T
;
773 sp_vsid
= get_kernel_vsid(sp
, MMU_SEGSIZE_256M
)
775 sp_vsid
|= SLB_VSID_KERNEL
| llp
;
776 p
->thread
.ksp_vsid
= sp_vsid
;
778 #endif /* CONFIG_PPC_STD_MMU_64 */
780 if (cpu_has_feature(CPU_FTR_DSCR
)) {
781 if (current
->thread
.dscr_inherit
) {
782 p
->thread
.dscr_inherit
= 1;
783 p
->thread
.dscr
= current
->thread
.dscr
;
784 } else if (0 != dscr_default
) {
785 p
->thread
.dscr_inherit
= 1;
786 p
->thread
.dscr
= dscr_default
;
788 p
->thread
.dscr_inherit
= 0;
795 * The PPC64 ABI makes use of a TOC to contain function
796 * pointers. The function (ret_from_except) is actually a pointer
797 * to the TOC entry. The first entry is a pointer to the actual
801 kregs
->nip
= *((unsigned long *)ret_from_fork
);
803 kregs
->nip
= (unsigned long)ret_from_fork
;
810 * Set up a thread for executing a new program
812 void start_thread(struct pt_regs
*regs
, unsigned long start
, unsigned long sp
)
815 unsigned long load_addr
= regs
->gpr
[2]; /* saved by ELF_PLAT_INIT */
819 * If we exec out of a kernel thread then thread.regs will not be
822 if (!current
->thread
.regs
) {
823 struct pt_regs
*regs
= task_stack_page(current
) + THREAD_SIZE
;
824 current
->thread
.regs
= regs
- 1;
827 memset(regs
->gpr
, 0, sizeof(regs
->gpr
));
835 * We have just cleared all the nonvolatile GPRs, so make
836 * FULL_REGS(regs) return true. This is necessary to allow
837 * ptrace to examine the thread immediately after exec.
844 regs
->msr
= MSR_USER
;
846 if (!is_32bit_task()) {
847 unsigned long entry
, toc
;
849 /* start is a relocated pointer to the function descriptor for
850 * the elf _start routine. The first entry in the function
851 * descriptor is the entry address of _start and the second
852 * entry is the TOC value we need to use.
854 __get_user(entry
, (unsigned long __user
*)start
);
855 __get_user(toc
, (unsigned long __user
*)start
+1);
857 /* Check whether the e_entry function descriptor entries
858 * need to be relocated before we can use them.
860 if (load_addr
!= 0) {
866 regs
->msr
= MSR_USER64
;
870 regs
->msr
= MSR_USER32
;
874 discard_lazy_cpu_state();
876 current
->thread
.used_vsr
= 0;
878 memset(current
->thread
.fpr
, 0, sizeof(current
->thread
.fpr
));
879 current
->thread
.fpscr
.val
= 0;
880 #ifdef CONFIG_ALTIVEC
881 memset(current
->thread
.vr
, 0, sizeof(current
->thread
.vr
));
882 memset(¤t
->thread
.vscr
, 0, sizeof(current
->thread
.vscr
));
883 current
->thread
.vscr
.u
[3] = 0x00010000; /* Java mode disabled */
884 current
->thread
.vrsave
= 0;
885 current
->thread
.used_vr
= 0;
886 #endif /* CONFIG_ALTIVEC */
888 memset(current
->thread
.evr
, 0, sizeof(current
->thread
.evr
));
889 current
->thread
.acc
= 0;
890 current
->thread
.spefscr
= 0;
891 current
->thread
.used_spe
= 0;
892 #endif /* CONFIG_SPE */
895 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
896 | PR_FP_EXC_RES | PR_FP_EXC_INV)
898 int set_fpexc_mode(struct task_struct
*tsk
, unsigned int val
)
900 struct pt_regs
*regs
= tsk
->thread
.regs
;
902 /* This is a bit hairy. If we are an SPE enabled processor
903 * (have embedded fp) we store the IEEE exception enable flags in
904 * fpexc_mode. fpexc_mode is also used for setting FP exception
905 * mode (asyn, precise, disabled) for 'Classic' FP. */
906 if (val
& PR_FP_EXC_SW_ENABLE
) {
908 if (cpu_has_feature(CPU_FTR_SPE
)) {
909 tsk
->thread
.fpexc_mode
= val
&
910 (PR_FP_EXC_SW_ENABLE
| PR_FP_ALL_EXCEPT
);
920 /* on a CONFIG_SPE this does not hurt us. The bits that
921 * __pack_fe01 use do not overlap with bits used for
922 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
923 * on CONFIG_SPE implementations are reserved so writing to
924 * them does not change anything */
925 if (val
> PR_FP_EXC_PRECISE
)
927 tsk
->thread
.fpexc_mode
= __pack_fe01(val
);
928 if (regs
!= NULL
&& (regs
->msr
& MSR_FP
) != 0)
929 regs
->msr
= (regs
->msr
& ~(MSR_FE0
|MSR_FE1
))
930 | tsk
->thread
.fpexc_mode
;
934 int get_fpexc_mode(struct task_struct
*tsk
, unsigned long adr
)
938 if (tsk
->thread
.fpexc_mode
& PR_FP_EXC_SW_ENABLE
)
940 if (cpu_has_feature(CPU_FTR_SPE
))
941 val
= tsk
->thread
.fpexc_mode
;
948 val
= __unpack_fe01(tsk
->thread
.fpexc_mode
);
949 return put_user(val
, (unsigned int __user
*) adr
);
952 int set_endian(struct task_struct
*tsk
, unsigned int val
)
954 struct pt_regs
*regs
= tsk
->thread
.regs
;
956 if ((val
== PR_ENDIAN_LITTLE
&& !cpu_has_feature(CPU_FTR_REAL_LE
)) ||
957 (val
== PR_ENDIAN_PPC_LITTLE
&& !cpu_has_feature(CPU_FTR_PPC_LE
)))
963 if (val
== PR_ENDIAN_BIG
)
964 regs
->msr
&= ~MSR_LE
;
965 else if (val
== PR_ENDIAN_LITTLE
|| val
== PR_ENDIAN_PPC_LITTLE
)
973 int get_endian(struct task_struct
*tsk
, unsigned long adr
)
975 struct pt_regs
*regs
= tsk
->thread
.regs
;
978 if (!cpu_has_feature(CPU_FTR_PPC_LE
) &&
979 !cpu_has_feature(CPU_FTR_REAL_LE
))
985 if (regs
->msr
& MSR_LE
) {
986 if (cpu_has_feature(CPU_FTR_REAL_LE
))
987 val
= PR_ENDIAN_LITTLE
;
989 val
= PR_ENDIAN_PPC_LITTLE
;
993 return put_user(val
, (unsigned int __user
*)adr
);
996 int set_unalign_ctl(struct task_struct
*tsk
, unsigned int val
)
998 tsk
->thread
.align_ctl
= val
;
1002 int get_unalign_ctl(struct task_struct
*tsk
, unsigned long adr
)
1004 return put_user(tsk
->thread
.align_ctl
, (unsigned int __user
*)adr
);
1007 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
1009 int sys_clone(unsigned long clone_flags
, unsigned long usp
,
1010 int __user
*parent_tidp
, void __user
*child_threadptr
,
1011 int __user
*child_tidp
, int p6
,
1012 struct pt_regs
*regs
)
1014 CHECK_FULL_REGS(regs
);
1016 usp
= regs
->gpr
[1]; /* stack pointer for child */
1018 if (is_32bit_task()) {
1019 parent_tidp
= TRUNC_PTR(parent_tidp
);
1020 child_tidp
= TRUNC_PTR(child_tidp
);
1023 return do_fork(clone_flags
, usp
, regs
, 0, parent_tidp
, child_tidp
);
1026 int sys_fork(unsigned long p1
, unsigned long p2
, unsigned long p3
,
1027 unsigned long p4
, unsigned long p5
, unsigned long p6
,
1028 struct pt_regs
*regs
)
1030 CHECK_FULL_REGS(regs
);
1031 return do_fork(SIGCHLD
, regs
->gpr
[1], regs
, 0, NULL
, NULL
);
1034 int sys_vfork(unsigned long p1
, unsigned long p2
, unsigned long p3
,
1035 unsigned long p4
, unsigned long p5
, unsigned long p6
,
1036 struct pt_regs
*regs
)
1038 CHECK_FULL_REGS(regs
);
1039 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, regs
->gpr
[1],
1040 regs
, 0, NULL
, NULL
);
1043 int sys_execve(unsigned long a0
, unsigned long a1
, unsigned long a2
,
1044 unsigned long a3
, unsigned long a4
, unsigned long a5
,
1045 struct pt_regs
*regs
)
1050 filename
= getname((const char __user
*) a0
);
1051 error
= PTR_ERR(filename
);
1052 if (IS_ERR(filename
))
1054 flush_fp_to_thread(current
);
1055 flush_altivec_to_thread(current
);
1056 flush_spe_to_thread(current
);
1057 error
= do_execve(filename
,
1058 (const char __user
*const __user
*) a1
,
1059 (const char __user
*const __user
*) a2
, regs
);
1065 static inline int valid_irq_stack(unsigned long sp
, struct task_struct
*p
,
1066 unsigned long nbytes
)
1068 unsigned long stack_page
;
1069 unsigned long cpu
= task_cpu(p
);
1072 * Avoid crashing if the stack has overflowed and corrupted
1073 * task_cpu(p), which is in the thread_info struct.
1075 if (cpu
< NR_CPUS
&& cpu_possible(cpu
)) {
1076 stack_page
= (unsigned long) hardirq_ctx
[cpu
];
1077 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1078 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1081 stack_page
= (unsigned long) softirq_ctx
[cpu
];
1082 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1083 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1089 int validate_sp(unsigned long sp
, struct task_struct
*p
,
1090 unsigned long nbytes
)
1092 unsigned long stack_page
= (unsigned long)task_stack_page(p
);
1094 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1095 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1098 return valid_irq_stack(sp
, p
, nbytes
);
1101 EXPORT_SYMBOL(validate_sp
);
1103 unsigned long get_wchan(struct task_struct
*p
)
1105 unsigned long ip
, sp
;
1108 if (!p
|| p
== current
|| p
->state
== TASK_RUNNING
)
1112 if (!validate_sp(sp
, p
, STACK_FRAME_OVERHEAD
))
1116 sp
= *(unsigned long *)sp
;
1117 if (!validate_sp(sp
, p
, STACK_FRAME_OVERHEAD
))
1120 ip
= ((unsigned long *)sp
)[STACK_FRAME_LR_SAVE
];
1121 if (!in_sched_functions(ip
))
1124 } while (count
++ < 16);
1128 static int kstack_depth_to_print
= CONFIG_PRINT_STACK_DEPTH
;
1130 void show_stack(struct task_struct
*tsk
, unsigned long *stack
)
1132 unsigned long sp
, ip
, lr
, newsp
;
1135 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1136 int curr_frame
= current
->curr_ret_stack
;
1137 extern void return_to_handler(void);
1138 unsigned long rth
= (unsigned long)return_to_handler
;
1139 unsigned long mrth
= -1;
1141 extern void mod_return_to_handler(void);
1142 rth
= *(unsigned long *)rth
;
1143 mrth
= (unsigned long)mod_return_to_handler
;
1144 mrth
= *(unsigned long *)mrth
;
1148 sp
= (unsigned long) stack
;
1153 asm("mr %0,1" : "=r" (sp
));
1155 sp
= tsk
->thread
.ksp
;
1159 printk("Call Trace:\n");
1161 if (!validate_sp(sp
, tsk
, STACK_FRAME_OVERHEAD
))
1164 stack
= (unsigned long *) sp
;
1166 ip
= stack
[STACK_FRAME_LR_SAVE
];
1167 if (!firstframe
|| ip
!= lr
) {
1168 printk("["REG
"] ["REG
"] %pS", sp
, ip
, (void *)ip
);
1169 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1170 if ((ip
== rth
|| ip
== mrth
) && curr_frame
>= 0) {
1172 (void *)current
->ret_stack
[curr_frame
].ret
);
1177 printk(" (unreliable)");
1183 * See if this is an exception frame.
1184 * We look for the "regshere" marker in the current frame.
1186 if (validate_sp(sp
, tsk
, STACK_INT_FRAME_SIZE
)
1187 && stack
[STACK_FRAME_MARKER
] == STACK_FRAME_REGS_MARKER
) {
1188 struct pt_regs
*regs
= (struct pt_regs
*)
1189 (sp
+ STACK_FRAME_OVERHEAD
);
1191 printk("--- Exception: %lx at %pS\n LR = %pS\n",
1192 regs
->trap
, (void *)regs
->nip
, (void *)lr
);
1197 } while (count
++ < kstack_depth_to_print
);
1200 void dump_stack(void)
1202 show_stack(current
, NULL
);
1204 EXPORT_SYMBOL(dump_stack
);
1207 void ppc64_runlatch_on(void)
1211 if (cpu_has_feature(CPU_FTR_CTRL
) && !test_thread_flag(TIF_RUNLATCH
)) {
1214 ctrl
= mfspr(SPRN_CTRLF
);
1215 ctrl
|= CTRL_RUNLATCH
;
1216 mtspr(SPRN_CTRLT
, ctrl
);
1218 set_thread_flag(TIF_RUNLATCH
);
1222 void __ppc64_runlatch_off(void)
1228 clear_thread_flag(TIF_RUNLATCH
);
1230 ctrl
= mfspr(SPRN_CTRLF
);
1231 ctrl
&= ~CTRL_RUNLATCH
;
1232 mtspr(SPRN_CTRLT
, ctrl
);
1236 #if THREAD_SHIFT < PAGE_SHIFT
1238 static struct kmem_cache
*thread_info_cache
;
1240 struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
, int node
)
1242 struct thread_info
*ti
;
1244 ti
= kmem_cache_alloc_node(thread_info_cache
, GFP_KERNEL
, node
);
1245 if (unlikely(ti
== NULL
))
1247 #ifdef CONFIG_DEBUG_STACK_USAGE
1248 memset(ti
, 0, THREAD_SIZE
);
1253 void free_thread_info(struct thread_info
*ti
)
1255 kmem_cache_free(thread_info_cache
, ti
);
1258 void thread_info_cache_init(void)
1260 thread_info_cache
= kmem_cache_create("thread_info", THREAD_SIZE
,
1261 THREAD_SIZE
, 0, NULL
);
1262 BUG_ON(thread_info_cache
== NULL
);
1265 #endif /* THREAD_SHIFT < PAGE_SHIFT */
1267 unsigned long arch_align_stack(unsigned long sp
)
1269 if (!(current
->personality
& ADDR_NO_RANDOMIZE
) && randomize_va_space
)
1270 sp
-= get_random_int() & ~PAGE_MASK
;
1274 static inline unsigned long brk_rnd(void)
1276 unsigned long rnd
= 0;
1278 /* 8MB for 32bit, 1GB for 64bit */
1279 if (is_32bit_task())
1280 rnd
= (long)(get_random_int() % (1<<(23-PAGE_SHIFT
)));
1282 rnd
= (long)(get_random_int() % (1<<(30-PAGE_SHIFT
)));
1284 return rnd
<< PAGE_SHIFT
;
1287 unsigned long arch_randomize_brk(struct mm_struct
*mm
)
1289 unsigned long base
= mm
->brk
;
1292 #ifdef CONFIG_PPC_STD_MMU_64
1294 * If we are using 1TB segments and we are allowed to randomise
1295 * the heap, we can put it above 1TB so it is backed by a 1TB
1296 * segment. Otherwise the heap will be in the bottom 1TB
1297 * which always uses 256MB segments and this may result in a
1298 * performance penalty.
1300 if (!is_32bit_task() && (mmu_highuser_ssize
== MMU_SEGSIZE_1T
))
1301 base
= max_t(unsigned long, mm
->brk
, 1UL << SID_SHIFT_1T
);
1304 ret
= PAGE_ALIGN(base
+ brk_rnd());
1312 unsigned long randomize_et_dyn(unsigned long base
)
1314 unsigned long ret
= PAGE_ALIGN(base
+ brk_rnd());