]>
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
;
489 #if defined(CONFIG_PPC_BOOK3E_64)
490 /* XXX Current Book3E code doesn't deal with kernel side DBCR0,
491 * we always hold the user values, so we set it now.
493 * However, we ensure the kernel MSR:DE is appropriately cleared too
494 * to avoid spurrious single step exceptions in the kernel.
496 * This will have to change to merge with the ppc32 code at some point,
497 * but I don't like much what ppc32 is doing today so there's some
498 * thinking needed there
500 if ((new_thread
->dbcr0
| old_thread
->dbcr0
) & DBCR0_IDM
) {
503 mtmsr(mfmsr() & ~MSR_DE
);
505 dbcr0
= mfspr(SPRN_DBCR0
);
506 dbcr0
= (dbcr0
& DBCR0_EDM
) | new_thread
->dbcr0
;
507 mtspr(SPRN_DBCR0
, dbcr0
);
509 #endif /* CONFIG_PPC64_BOOK3E */
513 * Collect processor utilization data per process
515 if (firmware_has_feature(FW_FEATURE_SPLPAR
)) {
516 struct cpu_usage
*cu
= &__get_cpu_var(cpu_usage_array
);
517 long unsigned start_tb
, current_tb
;
518 start_tb
= old_thread
->start_tb
;
519 cu
->current_tb
= current_tb
= mfspr(SPRN_PURR
);
520 old_thread
->accum_tb
+= (current_tb
- start_tb
);
521 new_thread
->start_tb
= current_tb
;
523 #endif /* CONFIG_PPC64 */
525 #ifdef CONFIG_PPC_BOOK3S_64
526 batch
= &__get_cpu_var(ppc64_tlb_batch
);
528 current_thread_info()->local_flags
|= _TLF_LAZY_MMU
;
530 __flush_tlb_pending(batch
);
533 #endif /* CONFIG_PPC_BOOK3S_64 */
535 local_irq_save(flags
);
537 account_system_vtime(current
);
538 account_process_vtime(current
);
541 * We can't take a PMU exception inside _switch() since there is a
542 * window where the kernel stack SLB and the kernel stack are out
543 * of sync. Hard disable here.
546 last
= _switch(old_thread
, new_thread
);
548 #ifdef CONFIG_PPC_BOOK3S_64
549 if (current_thread_info()->local_flags
& _TLF_LAZY_MMU
) {
550 current_thread_info()->local_flags
&= ~_TLF_LAZY_MMU
;
551 batch
= &__get_cpu_var(ppc64_tlb_batch
);
554 #endif /* CONFIG_PPC_BOOK3S_64 */
556 local_irq_restore(flags
);
561 static int instructions_to_print
= 16;
563 static void show_instructions(struct pt_regs
*regs
)
566 unsigned long pc
= regs
->nip
- (instructions_to_print
* 3 / 4 *
569 printk("Instruction dump:");
571 for (i
= 0; i
< instructions_to_print
; i
++) {
577 #if !defined(CONFIG_BOOKE)
578 /* If executing with the IMMU off, adjust pc rather
579 * than print XXXXXXXX.
581 if (!(regs
->msr
& MSR_IR
))
582 pc
= (unsigned long)phys_to_virt(pc
);
585 /* We use __get_user here *only* to avoid an OOPS on a
586 * bad address because the pc *should* only be a
589 if (!__kernel_text_address(pc
) ||
590 __get_user(instr
, (unsigned int __user
*)pc
)) {
594 printk("<%08x> ", instr
);
596 printk("%08x ", instr
);
605 static struct regbit
{
622 static void printbits(unsigned long val
, struct regbit
*bits
)
624 const char *sep
= "";
627 for (; bits
->bit
; ++bits
)
628 if (val
& bits
->bit
) {
629 printk("%s%s", sep
, bits
->name
);
637 #define REGS_PER_LINE 4
638 #define LAST_VOLATILE 13
641 #define REGS_PER_LINE 8
642 #define LAST_VOLATILE 12
645 void show_regs(struct pt_regs
* regs
)
649 printk("NIP: "REG
" LR: "REG
" CTR: "REG
"\n",
650 regs
->nip
, regs
->link
, regs
->ctr
);
651 printk("REGS: %p TRAP: %04lx %s (%s)\n",
652 regs
, regs
->trap
, print_tainted(), init_utsname()->release
);
653 printk("MSR: "REG
" ", regs
->msr
);
654 printbits(regs
->msr
, msr_bits
);
655 printk(" CR: %08lx XER: %08lx\n", regs
->ccr
, regs
->xer
);
657 if ((regs
->trap
!= 0xc00) && cpu_has_feature(CPU_FTR_CFAR
))
658 printk("CFAR: "REG
"\n", regs
->orig_gpr3
);
659 if (trap
== 0x300 || trap
== 0x600)
660 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
661 printk("DEAR: "REG
", ESR: "REG
"\n", regs
->dar
, regs
->dsisr
);
663 printk("DAR: "REG
", DSISR: %08lx\n", regs
->dar
, regs
->dsisr
);
665 printk("TASK = %p[%d] '%s' THREAD: %p",
666 current
, task_pid_nr(current
), current
->comm
, task_thread_info(current
));
669 printk(" CPU: %d", raw_smp_processor_id());
670 #endif /* CONFIG_SMP */
672 for (i
= 0; i
< 32; i
++) {
673 if ((i
% REGS_PER_LINE
) == 0)
674 printk("\nGPR%02d: ", i
);
675 printk(REG
" ", regs
->gpr
[i
]);
676 if (i
== LAST_VOLATILE
&& !FULL_REGS(regs
))
680 #ifdef CONFIG_KALLSYMS
682 * Lookup NIP late so we have the best change of getting the
683 * above info out without failing
685 printk("NIP ["REG
"] %pS\n", regs
->nip
, (void *)regs
->nip
);
686 printk("LR ["REG
"] %pS\n", regs
->link
, (void *)regs
->link
);
688 show_stack(current
, (unsigned long *) regs
->gpr
[1]);
689 if (!user_mode(regs
))
690 show_instructions(regs
);
693 void exit_thread(void)
695 discard_lazy_cpu_state();
698 void flush_thread(void)
700 discard_lazy_cpu_state();
702 #ifdef CONFIG_HAVE_HW_BREAKPOINT
703 flush_ptrace_hw_breakpoint(current
);
704 #else /* CONFIG_HAVE_HW_BREAKPOINT */
705 set_debug_reg_defaults(¤t
->thread
);
706 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
710 release_thread(struct task_struct
*t
)
715 * This gets called before we allocate a new thread and copy
716 * the current task into it.
718 void prepare_to_copy(struct task_struct
*tsk
)
720 flush_fp_to_thread(current
);
721 flush_altivec_to_thread(current
);
722 flush_vsx_to_thread(current
);
723 flush_spe_to_thread(current
);
724 #ifdef CONFIG_HAVE_HW_BREAKPOINT
725 flush_ptrace_hw_breakpoint(tsk
);
726 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
732 extern unsigned long dscr_default
; /* defined in arch/powerpc/kernel/sysfs.c */
734 int copy_thread(unsigned long clone_flags
, unsigned long usp
,
735 unsigned long unused
, struct task_struct
*p
,
736 struct pt_regs
*regs
)
738 struct pt_regs
*childregs
, *kregs
;
739 extern void ret_from_fork(void);
740 unsigned long sp
= (unsigned long)task_stack_page(p
) + THREAD_SIZE
;
742 CHECK_FULL_REGS(regs
);
744 sp
-= sizeof(struct pt_regs
);
745 childregs
= (struct pt_regs
*) sp
;
747 if ((childregs
->msr
& MSR_PR
) == 0) {
748 /* for kernel thread, set `current' and stackptr in new task */
749 childregs
->gpr
[1] = sp
+ sizeof(struct pt_regs
);
751 childregs
->gpr
[2] = (unsigned long) p
;
753 clear_tsk_thread_flag(p
, TIF_32BIT
);
755 p
->thread
.regs
= NULL
; /* no user register state */
757 childregs
->gpr
[1] = usp
;
758 p
->thread
.regs
= childregs
;
759 if (clone_flags
& CLONE_SETTLS
) {
761 if (!is_32bit_task())
762 childregs
->gpr
[13] = childregs
->gpr
[6];
765 childregs
->gpr
[2] = childregs
->gpr
[6];
768 childregs
->gpr
[3] = 0; /* Result from fork() */
769 sp
-= STACK_FRAME_OVERHEAD
;
772 * The way this works is that at some point in the future
773 * some task will call _switch to switch to the new task.
774 * That will pop off the stack frame created below and start
775 * the new task running at ret_from_fork. The new task will
776 * do some house keeping and then return from the fork or clone
777 * system call, using the stack frame created above.
779 sp
-= sizeof(struct pt_regs
);
780 kregs
= (struct pt_regs
*) sp
;
781 sp
-= STACK_FRAME_OVERHEAD
;
783 p
->thread
.ksp_limit
= (unsigned long)task_stack_page(p
) +
784 _ALIGN_UP(sizeof(struct thread_info
), 16);
786 #ifdef CONFIG_PPC_STD_MMU_64
787 if (mmu_has_feature(MMU_FTR_SLB
)) {
788 unsigned long sp_vsid
;
789 unsigned long llp
= mmu_psize_defs
[mmu_linear_psize
].sllp
;
791 if (mmu_has_feature(MMU_FTR_1T_SEGMENT
))
792 sp_vsid
= get_kernel_vsid(sp
, MMU_SEGSIZE_1T
)
793 << SLB_VSID_SHIFT_1T
;
795 sp_vsid
= get_kernel_vsid(sp
, MMU_SEGSIZE_256M
)
797 sp_vsid
|= SLB_VSID_KERNEL
| llp
;
798 p
->thread
.ksp_vsid
= sp_vsid
;
800 #endif /* CONFIG_PPC_STD_MMU_64 */
802 if (cpu_has_feature(CPU_FTR_DSCR
)) {
803 if (current
->thread
.dscr_inherit
) {
804 p
->thread
.dscr_inherit
= 1;
805 p
->thread
.dscr
= current
->thread
.dscr
;
806 } else if (0 != dscr_default
) {
807 p
->thread
.dscr_inherit
= 1;
808 p
->thread
.dscr
= dscr_default
;
810 p
->thread
.dscr_inherit
= 0;
817 * The PPC64 ABI makes use of a TOC to contain function
818 * pointers. The function (ret_from_except) is actually a pointer
819 * to the TOC entry. The first entry is a pointer to the actual
823 kregs
->nip
= *((unsigned long *)ret_from_fork
);
825 kregs
->nip
= (unsigned long)ret_from_fork
;
832 * Set up a thread for executing a new program
834 void start_thread(struct pt_regs
*regs
, unsigned long start
, unsigned long sp
)
837 unsigned long load_addr
= regs
->gpr
[2]; /* saved by ELF_PLAT_INIT */
841 * If we exec out of a kernel thread then thread.regs will not be
844 if (!current
->thread
.regs
) {
845 struct pt_regs
*regs
= task_stack_page(current
) + THREAD_SIZE
;
846 current
->thread
.regs
= regs
- 1;
849 memset(regs
->gpr
, 0, sizeof(regs
->gpr
));
857 * We have just cleared all the nonvolatile GPRs, so make
858 * FULL_REGS(regs) return true. This is necessary to allow
859 * ptrace to examine the thread immediately after exec.
866 regs
->msr
= MSR_USER
;
868 if (!is_32bit_task()) {
869 unsigned long entry
, toc
;
871 /* start is a relocated pointer to the function descriptor for
872 * the elf _start routine. The first entry in the function
873 * descriptor is the entry address of _start and the second
874 * entry is the TOC value we need to use.
876 __get_user(entry
, (unsigned long __user
*)start
);
877 __get_user(toc
, (unsigned long __user
*)start
+1);
879 /* Check whether the e_entry function descriptor entries
880 * need to be relocated before we can use them.
882 if (load_addr
!= 0) {
888 regs
->msr
= MSR_USER64
;
892 regs
->msr
= MSR_USER32
;
896 discard_lazy_cpu_state();
898 current
->thread
.used_vsr
= 0;
900 memset(current
->thread
.fpr
, 0, sizeof(current
->thread
.fpr
));
901 current
->thread
.fpscr
.val
= 0;
902 #ifdef CONFIG_ALTIVEC
903 memset(current
->thread
.vr
, 0, sizeof(current
->thread
.vr
));
904 memset(¤t
->thread
.vscr
, 0, sizeof(current
->thread
.vscr
));
905 current
->thread
.vscr
.u
[3] = 0x00010000; /* Java mode disabled */
906 current
->thread
.vrsave
= 0;
907 current
->thread
.used_vr
= 0;
908 #endif /* CONFIG_ALTIVEC */
910 memset(current
->thread
.evr
, 0, sizeof(current
->thread
.evr
));
911 current
->thread
.acc
= 0;
912 current
->thread
.spefscr
= 0;
913 current
->thread
.used_spe
= 0;
914 #endif /* CONFIG_SPE */
917 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
918 | PR_FP_EXC_RES | PR_FP_EXC_INV)
920 int set_fpexc_mode(struct task_struct
*tsk
, unsigned int val
)
922 struct pt_regs
*regs
= tsk
->thread
.regs
;
924 /* This is a bit hairy. If we are an SPE enabled processor
925 * (have embedded fp) we store the IEEE exception enable flags in
926 * fpexc_mode. fpexc_mode is also used for setting FP exception
927 * mode (asyn, precise, disabled) for 'Classic' FP. */
928 if (val
& PR_FP_EXC_SW_ENABLE
) {
930 if (cpu_has_feature(CPU_FTR_SPE
)) {
931 tsk
->thread
.fpexc_mode
= val
&
932 (PR_FP_EXC_SW_ENABLE
| PR_FP_ALL_EXCEPT
);
942 /* on a CONFIG_SPE this does not hurt us. The bits that
943 * __pack_fe01 use do not overlap with bits used for
944 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
945 * on CONFIG_SPE implementations are reserved so writing to
946 * them does not change anything */
947 if (val
> PR_FP_EXC_PRECISE
)
949 tsk
->thread
.fpexc_mode
= __pack_fe01(val
);
950 if (regs
!= NULL
&& (regs
->msr
& MSR_FP
) != 0)
951 regs
->msr
= (regs
->msr
& ~(MSR_FE0
|MSR_FE1
))
952 | tsk
->thread
.fpexc_mode
;
956 int get_fpexc_mode(struct task_struct
*tsk
, unsigned long adr
)
960 if (tsk
->thread
.fpexc_mode
& PR_FP_EXC_SW_ENABLE
)
962 if (cpu_has_feature(CPU_FTR_SPE
))
963 val
= tsk
->thread
.fpexc_mode
;
970 val
= __unpack_fe01(tsk
->thread
.fpexc_mode
);
971 return put_user(val
, (unsigned int __user
*) adr
);
974 int set_endian(struct task_struct
*tsk
, unsigned int val
)
976 struct pt_regs
*regs
= tsk
->thread
.regs
;
978 if ((val
== PR_ENDIAN_LITTLE
&& !cpu_has_feature(CPU_FTR_REAL_LE
)) ||
979 (val
== PR_ENDIAN_PPC_LITTLE
&& !cpu_has_feature(CPU_FTR_PPC_LE
)))
985 if (val
== PR_ENDIAN_BIG
)
986 regs
->msr
&= ~MSR_LE
;
987 else if (val
== PR_ENDIAN_LITTLE
|| val
== PR_ENDIAN_PPC_LITTLE
)
995 int get_endian(struct task_struct
*tsk
, unsigned long adr
)
997 struct pt_regs
*regs
= tsk
->thread
.regs
;
1000 if (!cpu_has_feature(CPU_FTR_PPC_LE
) &&
1001 !cpu_has_feature(CPU_FTR_REAL_LE
))
1007 if (regs
->msr
& MSR_LE
) {
1008 if (cpu_has_feature(CPU_FTR_REAL_LE
))
1009 val
= PR_ENDIAN_LITTLE
;
1011 val
= PR_ENDIAN_PPC_LITTLE
;
1013 val
= PR_ENDIAN_BIG
;
1015 return put_user(val
, (unsigned int __user
*)adr
);
1018 int set_unalign_ctl(struct task_struct
*tsk
, unsigned int val
)
1020 tsk
->thread
.align_ctl
= val
;
1024 int get_unalign_ctl(struct task_struct
*tsk
, unsigned long adr
)
1026 return put_user(tsk
->thread
.align_ctl
, (unsigned int __user
*)adr
);
1029 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
1031 int sys_clone(unsigned long clone_flags
, unsigned long usp
,
1032 int __user
*parent_tidp
, void __user
*child_threadptr
,
1033 int __user
*child_tidp
, int p6
,
1034 struct pt_regs
*regs
)
1036 CHECK_FULL_REGS(regs
);
1038 usp
= regs
->gpr
[1]; /* stack pointer for child */
1040 if (is_32bit_task()) {
1041 parent_tidp
= TRUNC_PTR(parent_tidp
);
1042 child_tidp
= TRUNC_PTR(child_tidp
);
1045 return do_fork(clone_flags
, usp
, regs
, 0, parent_tidp
, child_tidp
);
1048 int sys_fork(unsigned long p1
, unsigned long p2
, unsigned long p3
,
1049 unsigned long p4
, unsigned long p5
, unsigned long p6
,
1050 struct pt_regs
*regs
)
1052 CHECK_FULL_REGS(regs
);
1053 return do_fork(SIGCHLD
, regs
->gpr
[1], regs
, 0, NULL
, NULL
);
1056 int sys_vfork(unsigned long p1
, unsigned long p2
, unsigned long p3
,
1057 unsigned long p4
, unsigned long p5
, unsigned long p6
,
1058 struct pt_regs
*regs
)
1060 CHECK_FULL_REGS(regs
);
1061 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, regs
->gpr
[1],
1062 regs
, 0, NULL
, NULL
);
1065 int sys_execve(unsigned long a0
, unsigned long a1
, unsigned long a2
,
1066 unsigned long a3
, unsigned long a4
, unsigned long a5
,
1067 struct pt_regs
*regs
)
1072 filename
= getname((const char __user
*) a0
);
1073 error
= PTR_ERR(filename
);
1074 if (IS_ERR(filename
))
1076 flush_fp_to_thread(current
);
1077 flush_altivec_to_thread(current
);
1078 flush_spe_to_thread(current
);
1079 error
= do_execve(filename
,
1080 (const char __user
*const __user
*) a1
,
1081 (const char __user
*const __user
*) a2
, regs
);
1087 static inline int valid_irq_stack(unsigned long sp
, struct task_struct
*p
,
1088 unsigned long nbytes
)
1090 unsigned long stack_page
;
1091 unsigned long cpu
= task_cpu(p
);
1094 * Avoid crashing if the stack has overflowed and corrupted
1095 * task_cpu(p), which is in the thread_info struct.
1097 if (cpu
< NR_CPUS
&& cpu_possible(cpu
)) {
1098 stack_page
= (unsigned long) hardirq_ctx
[cpu
];
1099 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1100 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1103 stack_page
= (unsigned long) softirq_ctx
[cpu
];
1104 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1105 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1111 int validate_sp(unsigned long sp
, struct task_struct
*p
,
1112 unsigned long nbytes
)
1114 unsigned long stack_page
= (unsigned long)task_stack_page(p
);
1116 if (sp
>= stack_page
+ sizeof(struct thread_struct
)
1117 && sp
<= stack_page
+ THREAD_SIZE
- nbytes
)
1120 return valid_irq_stack(sp
, p
, nbytes
);
1123 EXPORT_SYMBOL(validate_sp
);
1125 unsigned long get_wchan(struct task_struct
*p
)
1127 unsigned long ip
, sp
;
1130 if (!p
|| p
== current
|| p
->state
== TASK_RUNNING
)
1134 if (!validate_sp(sp
, p
, STACK_FRAME_OVERHEAD
))
1138 sp
= *(unsigned long *)sp
;
1139 if (!validate_sp(sp
, p
, STACK_FRAME_OVERHEAD
))
1142 ip
= ((unsigned long *)sp
)[STACK_FRAME_LR_SAVE
];
1143 if (!in_sched_functions(ip
))
1146 } while (count
++ < 16);
1150 static int kstack_depth_to_print
= CONFIG_PRINT_STACK_DEPTH
;
1152 void show_stack(struct task_struct
*tsk
, unsigned long *stack
)
1154 unsigned long sp
, ip
, lr
, newsp
;
1157 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1158 int curr_frame
= current
->curr_ret_stack
;
1159 extern void return_to_handler(void);
1160 unsigned long rth
= (unsigned long)return_to_handler
;
1161 unsigned long mrth
= -1;
1163 extern void mod_return_to_handler(void);
1164 rth
= *(unsigned long *)rth
;
1165 mrth
= (unsigned long)mod_return_to_handler
;
1166 mrth
= *(unsigned long *)mrth
;
1170 sp
= (unsigned long) stack
;
1175 asm("mr %0,1" : "=r" (sp
));
1177 sp
= tsk
->thread
.ksp
;
1181 printk("Call Trace:\n");
1183 if (!validate_sp(sp
, tsk
, STACK_FRAME_OVERHEAD
))
1186 stack
= (unsigned long *) sp
;
1188 ip
= stack
[STACK_FRAME_LR_SAVE
];
1189 if (!firstframe
|| ip
!= lr
) {
1190 printk("["REG
"] ["REG
"] %pS", sp
, ip
, (void *)ip
);
1191 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1192 if ((ip
== rth
|| ip
== mrth
) && curr_frame
>= 0) {
1194 (void *)current
->ret_stack
[curr_frame
].ret
);
1199 printk(" (unreliable)");
1205 * See if this is an exception frame.
1206 * We look for the "regshere" marker in the current frame.
1208 if (validate_sp(sp
, tsk
, STACK_INT_FRAME_SIZE
)
1209 && stack
[STACK_FRAME_MARKER
] == STACK_FRAME_REGS_MARKER
) {
1210 struct pt_regs
*regs
= (struct pt_regs
*)
1211 (sp
+ STACK_FRAME_OVERHEAD
);
1213 printk("--- Exception: %lx at %pS\n LR = %pS\n",
1214 regs
->trap
, (void *)regs
->nip
, (void *)lr
);
1219 } while (count
++ < kstack_depth_to_print
);
1222 void dump_stack(void)
1224 show_stack(current
, NULL
);
1226 EXPORT_SYMBOL(dump_stack
);
1229 void ppc64_runlatch_on(void)
1233 if (cpu_has_feature(CPU_FTR_CTRL
) && !test_thread_flag(TIF_RUNLATCH
)) {
1236 ctrl
= mfspr(SPRN_CTRLF
);
1237 ctrl
|= CTRL_RUNLATCH
;
1238 mtspr(SPRN_CTRLT
, ctrl
);
1240 set_thread_flag(TIF_RUNLATCH
);
1244 void __ppc64_runlatch_off(void)
1250 clear_thread_flag(TIF_RUNLATCH
);
1252 ctrl
= mfspr(SPRN_CTRLF
);
1253 ctrl
&= ~CTRL_RUNLATCH
;
1254 mtspr(SPRN_CTRLT
, ctrl
);
1258 #if THREAD_SHIFT < PAGE_SHIFT
1260 static struct kmem_cache
*thread_info_cache
;
1262 struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
, int node
)
1264 struct thread_info
*ti
;
1266 ti
= kmem_cache_alloc_node(thread_info_cache
, GFP_KERNEL
, node
);
1267 if (unlikely(ti
== NULL
))
1269 #ifdef CONFIG_DEBUG_STACK_USAGE
1270 memset(ti
, 0, THREAD_SIZE
);
1275 void free_thread_info(struct thread_info
*ti
)
1277 kmem_cache_free(thread_info_cache
, ti
);
1280 void thread_info_cache_init(void)
1282 thread_info_cache
= kmem_cache_create("thread_info", THREAD_SIZE
,
1283 THREAD_SIZE
, 0, NULL
);
1284 BUG_ON(thread_info_cache
== NULL
);
1287 #endif /* THREAD_SHIFT < PAGE_SHIFT */
1289 unsigned long arch_align_stack(unsigned long sp
)
1291 if (!(current
->personality
& ADDR_NO_RANDOMIZE
) && randomize_va_space
)
1292 sp
-= get_random_int() & ~PAGE_MASK
;
1296 static inline unsigned long brk_rnd(void)
1298 unsigned long rnd
= 0;
1300 /* 8MB for 32bit, 1GB for 64bit */
1301 if (is_32bit_task())
1302 rnd
= (long)(get_random_int() % (1<<(23-PAGE_SHIFT
)));
1304 rnd
= (long)(get_random_int() % (1<<(30-PAGE_SHIFT
)));
1306 return rnd
<< PAGE_SHIFT
;
1309 unsigned long arch_randomize_brk(struct mm_struct
*mm
)
1311 unsigned long base
= mm
->brk
;
1314 #ifdef CONFIG_PPC_STD_MMU_64
1316 * If we are using 1TB segments and we are allowed to randomise
1317 * the heap, we can put it above 1TB so it is backed by a 1TB
1318 * segment. Otherwise the heap will be in the bottom 1TB
1319 * which always uses 256MB segments and this may result in a
1320 * performance penalty.
1322 if (!is_32bit_task() && (mmu_highuser_ssize
== MMU_SEGSIZE_1T
))
1323 base
= max_t(unsigned long, mm
->brk
, 1UL << SID_SHIFT_1T
);
1326 ret
= PAGE_ALIGN(base
+ brk_rnd());
1334 unsigned long randomize_et_dyn(unsigned long base
)
1336 unsigned long ret
= PAGE_ALIGN(base
+ brk_rnd());