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>
45 #include <asm/processor.h>
48 #include <asm/machdep.h>
50 #include <asm/runlatch.h>
51 #include <asm/syscalls.h>
52 #include <asm/switch_to.h>
53 #include <asm/debug.h>
55 #include <asm/firmware.h>
57 #include <linux/kprobes.h>
58 #include <linux/kdebug.h>
60 extern unsigned long _get_SP(void);
63 struct task_struct *last_task_used_math = NULL;
64 struct task_struct *last_task_used_altivec = NULL;
65 struct task_struct *last_task_used_vsx = NULL;
66 struct task_struct *last_task_used_spe = NULL;
70 * Make sure the floating-point register state in the
71 * the thread_struct is up to date for task tsk.
73 void flush_fp_to_thread(struct task_struct *tsk)
75 if (tsk->thread.regs) {
77 * We need to disable preemption here because if we didn't,
78 * another process could get scheduled after the regs->msr
79 * test but before we have finished saving the FP registers
80 * to the thread_struct. That process could take over the
81 * FPU, and then when we get scheduled again we would store
82 * bogus values for the remaining FP registers.
85 if (tsk->thread.regs->msr & MSR_FP) {
88 * This should only ever be called for current or
89 * for a stopped child process. Since we save away
90 * the FP register state on context switch on SMP,
91 * there is something wrong if a stopped child appears
92 * to still have its FP state in the CPU registers.
94 BUG_ON(tsk != current);
101 EXPORT_SYMBOL_GPL(flush_fp_to_thread);
103 void enable_kernel_fp(void)
105 WARN_ON(preemptible());
108 if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
111 giveup_fpu(NULL); /* just enables FP for kernel */
113 giveup_fpu(last_task_used_math);
114 #endif /* CONFIG_SMP */
116 EXPORT_SYMBOL(enable_kernel_fp);
118 #ifdef CONFIG_ALTIVEC
119 void enable_kernel_altivec(void)
121 WARN_ON(preemptible());
124 if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
125 giveup_altivec(current);
127 giveup_altivec_notask();
129 giveup_altivec(last_task_used_altivec);
130 #endif /* CONFIG_SMP */
132 EXPORT_SYMBOL(enable_kernel_altivec);
135 * Make sure the VMX/Altivec register state in the
136 * the thread_struct is up to date for task tsk.
138 void flush_altivec_to_thread(struct task_struct *tsk)
140 if (tsk->thread.regs) {
142 if (tsk->thread.regs->msr & MSR_VEC) {
144 BUG_ON(tsk != current);
151 EXPORT_SYMBOL_GPL(flush_altivec_to_thread);
152 #endif /* CONFIG_ALTIVEC */
156 /* not currently used, but some crazy RAID module might want to later */
157 void enable_kernel_vsx(void)
159 WARN_ON(preemptible());
162 if (current->thread.regs && (current->thread.regs->msr & MSR_VSX))
165 giveup_vsx(NULL); /* just enable vsx for kernel - force */
167 giveup_vsx(last_task_used_vsx);
168 #endif /* CONFIG_SMP */
170 EXPORT_SYMBOL(enable_kernel_vsx);
173 void giveup_vsx(struct task_struct *tsk)
180 void flush_vsx_to_thread(struct task_struct *tsk)
182 if (tsk->thread.regs) {
184 if (tsk->thread.regs->msr & MSR_VSX) {
186 BUG_ON(tsk != current);
193 EXPORT_SYMBOL_GPL(flush_vsx_to_thread);
194 #endif /* CONFIG_VSX */
198 void enable_kernel_spe(void)
200 WARN_ON(preemptible());
203 if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
206 giveup_spe(NULL); /* just enable SPE for kernel - force */
208 giveup_spe(last_task_used_spe);
209 #endif /* __SMP __ */
211 EXPORT_SYMBOL(enable_kernel_spe);
213 void flush_spe_to_thread(struct task_struct *tsk)
215 if (tsk->thread.regs) {
217 if (tsk->thread.regs->msr & MSR_SPE) {
219 BUG_ON(tsk != current);
221 tsk->thread.spefscr = mfspr(SPRN_SPEFSCR);
227 #endif /* CONFIG_SPE */
231 * If we are doing lazy switching of CPU state (FP, altivec or SPE),
232 * and the current task has some state, discard it.
234 void discard_lazy_cpu_state(void)
237 if (last_task_used_math == current)
238 last_task_used_math = NULL;
239 #ifdef CONFIG_ALTIVEC
240 if (last_task_used_altivec == current)
241 last_task_used_altivec = NULL;
242 #endif /* CONFIG_ALTIVEC */
244 if (last_task_used_vsx == current)
245 last_task_used_vsx = NULL;
246 #endif /* CONFIG_VSX */
248 if (last_task_used_spe == current)
249 last_task_used_spe = NULL;
253 #endif /* CONFIG_SMP */
255 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
256 void do_send_trap(struct pt_regs *regs, unsigned long address,
257 unsigned long error_code, int signal_code, int breakpt)
261 current->thread.trap_nr = signal_code;
262 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
263 11, SIGSEGV) == NOTIFY_STOP)
266 /* Deliver the signal to userspace */
267 info.si_signo = SIGTRAP;
268 info.si_errno = breakpt; /* breakpoint or watchpoint id */
269 info.si_code = signal_code;
270 info.si_addr = (void __user *)address;
271 force_sig_info(SIGTRAP, &info, current);
273 #else /* !CONFIG_PPC_ADV_DEBUG_REGS */
274 void do_dabr(struct pt_regs *regs, unsigned long address,
275 unsigned long error_code)
279 current->thread.trap_nr = TRAP_HWBKPT;
280 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
281 11, SIGSEGV) == NOTIFY_STOP)
284 if (debugger_dabr_match(regs))
290 /* Deliver the signal to userspace */
291 info.si_signo = SIGTRAP;
293 info.si_code = TRAP_HWBKPT;
294 info.si_addr = (void __user *)address;
295 force_sig_info(SIGTRAP, &info, current);
297 #endif /* CONFIG_PPC_ADV_DEBUG_REGS */
299 static DEFINE_PER_CPU(unsigned long, current_dabr);
301 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
303 * Set the debug registers back to their default "safe" values.
305 static void set_debug_reg_defaults(struct thread_struct *thread)
307 thread->iac1 = thread->iac2 = 0;
308 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
309 thread->iac3 = thread->iac4 = 0;
311 thread->dac1 = thread->dac2 = 0;
312 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
313 thread->dvc1 = thread->dvc2 = 0;
318 * Force User/Supervisor bits to b11 (user-only MSR[PR]=1)
320 thread->dbcr1 = DBCR1_IAC1US | DBCR1_IAC2US | \
321 DBCR1_IAC3US | DBCR1_IAC4US;
323 * Force Data Address Compare User/Supervisor bits to be User-only
324 * (0b11 MSR[PR]=1) and set all other bits in DBCR2 register to be 0.
326 thread->dbcr2 = DBCR2_DAC1US | DBCR2_DAC2US;
332 static void prime_debug_regs(struct thread_struct *thread)
334 mtspr(SPRN_IAC1, thread->iac1);
335 mtspr(SPRN_IAC2, thread->iac2);
336 #if CONFIG_PPC_ADV_DEBUG_IACS > 2
337 mtspr(SPRN_IAC3, thread->iac3);
338 mtspr(SPRN_IAC4, thread->iac4);
340 mtspr(SPRN_DAC1, thread->dac1);
341 mtspr(SPRN_DAC2, thread->dac2);
342 #if CONFIG_PPC_ADV_DEBUG_DVCS > 0
343 mtspr(SPRN_DVC1, thread->dvc1);
344 mtspr(SPRN_DVC2, thread->dvc2);
346 mtspr(SPRN_DBCR0, thread->dbcr0);
347 mtspr(SPRN_DBCR1, thread->dbcr1);
349 mtspr(SPRN_DBCR2, thread->dbcr2);
353 * Unless neither the old or new thread are making use of the
354 * debug registers, set the debug registers from the values
355 * stored in the new thread.
357 static void switch_booke_debug_regs(struct thread_struct *new_thread)
359 if ((current->thread.dbcr0 & DBCR0_IDM)
360 || (new_thread->dbcr0 & DBCR0_IDM))
361 prime_debug_regs(new_thread);
363 #else /* !CONFIG_PPC_ADV_DEBUG_REGS */
364 #ifndef CONFIG_HAVE_HW_BREAKPOINT
365 static void set_debug_reg_defaults(struct thread_struct *thread)
373 #endif /* !CONFIG_HAVE_HW_BREAKPOINT */
374 #endif /* CONFIG_PPC_ADV_DEBUG_REGS */
376 int set_dabr(unsigned long dabr, unsigned long dabrx)
378 __get_cpu_var(current_dabr) = dabr;
381 return ppc_md.set_dabr(dabr, dabrx);
383 /* XXX should we have a CPU_FTR_HAS_DABR ? */
384 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
385 mtspr(SPRN_DAC1, dabr);
386 #ifdef CONFIG_PPC_47x
389 #elif defined(CONFIG_PPC_BOOK3S)
390 mtspr(SPRN_DABR, dabr);
391 mtspr(SPRN_DABRX, dabrx);
397 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
400 struct task_struct *__switch_to(struct task_struct *prev,
401 struct task_struct *new)
403 struct thread_struct *new_thread, *old_thread;
405 struct task_struct *last;
406 #ifdef CONFIG_PPC_BOOK3S_64
407 struct ppc64_tlb_batch *batch;
411 /* avoid complexity of lazy save/restore of fpu
412 * by just saving it every time we switch out if
413 * this task used the fpu during the last quantum.
415 * If it tries to use the fpu again, it'll trap and
416 * reload its fp regs. So we don't have to do a restore
417 * every switch, just a save.
420 if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
422 #ifdef CONFIG_ALTIVEC
424 * If the previous thread used altivec in the last quantum
425 * (thus changing altivec regs) then save them.
426 * We used to check the VRSAVE register but not all apps
427 * set it, so we don't rely on it now (and in fact we need
428 * to save & restore VSCR even if VRSAVE == 0). -- paulus
430 * On SMP we always save/restore altivec regs just to avoid the
431 * complexity of changing processors.
434 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
435 giveup_altivec(prev);
436 #endif /* CONFIG_ALTIVEC */
438 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VSX))
439 /* VMX and FPU registers are already save here */
441 #endif /* CONFIG_VSX */
444 * If the previous thread used spe in the last quantum
445 * (thus changing spe regs) then save them.
447 * On SMP we always save/restore spe regs just to avoid the
448 * complexity of changing processors.
450 if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
452 #endif /* CONFIG_SPE */
454 #else /* CONFIG_SMP */
455 #ifdef CONFIG_ALTIVEC
456 /* Avoid the trap. On smp this this never happens since
457 * we don't set last_task_used_altivec -- Cort
459 if (new->thread.regs && last_task_used_altivec == new)
460 new->thread.regs->msr |= MSR_VEC;
461 #endif /* CONFIG_ALTIVEC */
463 if (new->thread.regs && last_task_used_vsx == new)
464 new->thread.regs->msr |= MSR_VSX;
465 #endif /* CONFIG_VSX */
467 /* Avoid the trap. On smp this this never happens since
468 * we don't set last_task_used_spe
470 if (new->thread.regs && last_task_used_spe == new)
471 new->thread.regs->msr |= MSR_SPE;
472 #endif /* CONFIG_SPE */
474 #endif /* CONFIG_SMP */
476 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
477 switch_booke_debug_regs(&new->thread);
480 * For PPC_BOOK3S_64, we use the hw-breakpoint interfaces that would
483 #ifndef CONFIG_HAVE_HW_BREAKPOINT
484 if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr))
485 set_dabr(new->thread.dabr, new->thread.dabrx);
486 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
490 new_thread = &new->thread;
491 old_thread = ¤t->thread;
495 * Collect processor utilization data per process
497 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
498 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
499 long unsigned start_tb, current_tb;
500 start_tb = old_thread->start_tb;
501 cu->current_tb = current_tb = mfspr(SPRN_PURR);
502 old_thread->accum_tb += (current_tb - start_tb);
503 new_thread->start_tb = current_tb;
505 #endif /* CONFIG_PPC64 */
507 #ifdef CONFIG_PPC_BOOK3S_64
508 batch = &__get_cpu_var(ppc64_tlb_batch);
510 current_thread_info()->local_flags |= _TLF_LAZY_MMU;
512 __flush_tlb_pending(batch);
515 #endif /* CONFIG_PPC_BOOK3S_64 */
517 local_irq_save(flags);
520 * We can't take a PMU exception inside _switch() since there is a
521 * window where the kernel stack SLB and the kernel stack are out
522 * of sync. Hard disable here.
525 last = _switch(old_thread, new_thread);
527 #ifdef CONFIG_PPC_BOOK3S_64
528 if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
529 current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
530 batch = &__get_cpu_var(ppc64_tlb_batch);
533 #endif /* CONFIG_PPC_BOOK3S_64 */
535 local_irq_restore(flags);
540 static int instructions_to_print = 16;
542 static void show_instructions(struct pt_regs *regs)
545 unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
548 printk("Instruction dump:");
550 for (i = 0; i < instructions_to_print; i++) {
556 #if !defined(CONFIG_BOOKE)
557 /* If executing with the IMMU off, adjust pc rather
558 * than print XXXXXXXX.
560 if (!(regs->msr & MSR_IR))
561 pc = (unsigned long)phys_to_virt(pc);
564 /* We use __get_user here *only* to avoid an OOPS on a
565 * bad address because the pc *should* only be a
568 if (!__kernel_text_address(pc) ||
569 __get_user(instr, (unsigned int __user *)pc)) {
570 printk(KERN_CONT "XXXXXXXX ");
573 printk(KERN_CONT "<%08x> ", instr);
575 printk(KERN_CONT "%08x ", instr);
584 static struct regbit {
588 #if defined(CONFIG_PPC64) && !defined(CONFIG_BOOKE)
617 static void printbits(unsigned long val, struct regbit *bits)
619 const char *sep = "";
622 for (; bits->bit; ++bits)
623 if (val & bits->bit) {
624 printk("%s%s", sep, bits->name);
632 #define REGS_PER_LINE 4
633 #define LAST_VOLATILE 13
636 #define REGS_PER_LINE 8
637 #define LAST_VOLATILE 12
640 void show_regs(struct pt_regs * regs)
644 printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
645 regs->nip, regs->link, regs->ctr);
646 printk("REGS: %p TRAP: %04lx %s (%s)\n",
647 regs, regs->trap, print_tainted(), init_utsname()->release);
648 printk("MSR: "REG" ", regs->msr);
649 printbits(regs->msr, msr_bits);
650 printk(" CR: %08lx XER: %08lx\n", regs->ccr, regs->xer);
652 printk("SOFTE: %ld\n", regs->softe);
655 if ((regs->trap != 0xc00) && cpu_has_feature(CPU_FTR_CFAR))
656 printk("CFAR: "REG"\n", regs->orig_gpr3);
657 if (trap == 0x300 || trap == 0x600)
658 #if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)
659 printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
661 printk("DAR: "REG", DSISR: %08lx\n", regs->dar, regs->dsisr);
663 printk("TASK = %p[%d] '%s' THREAD: %p",
664 current, task_pid_nr(current), current->comm, task_thread_info(current));
667 printk(" CPU: %d", raw_smp_processor_id());
668 #endif /* CONFIG_SMP */
670 for (i = 0; i < 32; i++) {
671 if ((i % REGS_PER_LINE) == 0)
672 printk("\nGPR%02d: ", i);
673 printk(REG " ", regs->gpr[i]);
674 if (i == LAST_VOLATILE && !FULL_REGS(regs))
678 #ifdef CONFIG_KALLSYMS
680 * Lookup NIP late so we have the best change of getting the
681 * above info out without failing
683 printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
684 printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
686 show_stack(current, (unsigned long *) regs->gpr[1]);
687 if (!user_mode(regs))
688 show_instructions(regs);
691 void exit_thread(void)
693 discard_lazy_cpu_state();
696 void flush_thread(void)
698 discard_lazy_cpu_state();
700 #ifdef CONFIG_HAVE_HW_BREAKPOINT
701 flush_ptrace_hw_breakpoint(current);
702 #else /* CONFIG_HAVE_HW_BREAKPOINT */
703 set_debug_reg_defaults(¤t->thread);
704 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
708 release_thread(struct task_struct *t)
713 * this gets called so that we can store coprocessor state into memory and
714 * copy the current task into the new thread.
716 int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
718 flush_fp_to_thread(src);
719 flush_altivec_to_thread(src);
720 flush_vsx_to_thread(src);
721 flush_spe_to_thread(src);
722 #ifdef CONFIG_HAVE_HW_BREAKPOINT
723 flush_ptrace_hw_breakpoint(src);
724 #endif /* CONFIG_HAVE_HW_BREAKPOINT */
733 extern unsigned long dscr_default; /* defined in arch/powerpc/kernel/sysfs.c */
735 int copy_thread(unsigned long clone_flags, unsigned long usp,
736 unsigned long arg, struct task_struct *p,
737 struct pt_regs *regs)
739 struct pt_regs *childregs, *kregs;
740 extern void ret_from_fork(void);
741 extern void ret_from_kernel_thread(void);
743 unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
746 sp -= sizeof(struct pt_regs);
747 childregs = (struct pt_regs *) sp;
749 /* for kernel thread, set `current' and stackptr in new task */
750 memset(childregs, 0, sizeof(struct pt_regs));
751 childregs->gpr[1] = sp + sizeof(struct pt_regs);
753 childregs->gpr[14] = *(unsigned long *)usp;
754 childregs->gpr[2] = ((unsigned long *)usp)[1],
755 clear_tsk_thread_flag(p, TIF_32BIT);
757 childregs->gpr[14] = usp; /* function */
758 childregs->gpr[2] = (unsigned long) p;
760 childregs->gpr[15] = arg;
761 p->thread.regs = NULL; /* no user register state */
762 f = ret_from_kernel_thread;
764 CHECK_FULL_REGS(regs);
766 childregs->gpr[1] = usp;
767 p->thread.regs = childregs;
768 childregs->gpr[3] = 0; /* Result from fork() */
769 if (clone_flags & CLONE_SETTLS) {
771 if (!is_32bit_task())
772 childregs->gpr[13] = childregs->gpr[6];
775 childregs->gpr[2] = childregs->gpr[6];
780 sp -= STACK_FRAME_OVERHEAD;
783 * The way this works is that at some point in the future
784 * some task will call _switch to switch to the new task.
785 * That will pop off the stack frame created below and start
786 * the new task running at ret_from_fork. The new task will
787 * do some house keeping and then return from the fork or clone
788 * system call, using the stack frame created above.
790 sp -= sizeof(struct pt_regs);
791 kregs = (struct pt_regs *) sp;
792 sp -= STACK_FRAME_OVERHEAD;
794 p->thread.ksp_limit = (unsigned long)task_stack_page(p) +
795 _ALIGN_UP(sizeof(struct thread_info), 16);
797 #ifdef CONFIG_PPC_STD_MMU_64
798 if (mmu_has_feature(MMU_FTR_SLB)) {
799 unsigned long sp_vsid;
800 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
802 if (mmu_has_feature(MMU_FTR_1T_SEGMENT))
803 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
804 << SLB_VSID_SHIFT_1T;
806 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_256M)
808 sp_vsid |= SLB_VSID_KERNEL | llp;
809 p->thread.ksp_vsid = sp_vsid;
811 #endif /* CONFIG_PPC_STD_MMU_64 */
813 if (cpu_has_feature(CPU_FTR_DSCR)) {
814 p->thread.dscr_inherit = current->thread.dscr_inherit;
815 p->thread.dscr = current->thread.dscr;
819 * The PPC64 ABI makes use of a TOC to contain function
820 * pointers. The function (ret_from_except) is actually a pointer
821 * to the TOC entry. The first entry is a pointer to the actual
825 kregs->nip = *((unsigned long *)f);
827 kregs->nip = (unsigned long)f;
833 * Set up a thread for executing a new program
835 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
838 unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
842 * If we exec out of a kernel thread then thread.regs will not be
845 if (!current->thread.regs) {
846 struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
847 current->thread.regs = regs - 1;
850 memset(regs->gpr, 0, sizeof(regs->gpr));
858 * We have just cleared all the nonvolatile GPRs, so make
859 * FULL_REGS(regs) return true. This is necessary to allow
860 * ptrace to examine the thread immediately after exec.
867 regs->msr = MSR_USER;
869 if (!is_32bit_task()) {
870 unsigned long entry, toc;
872 /* start is a relocated pointer to the function descriptor for
873 * the elf _start routine. The first entry in the function
874 * descriptor is the entry address of _start and the second
875 * entry is the TOC value we need to use.
877 __get_user(entry, (unsigned long __user *)start);
878 __get_user(toc, (unsigned long __user *)start+1);
880 /* Check whether the e_entry function descriptor entries
881 * need to be relocated before we can use them.
883 if (load_addr != 0) {
889 regs->msr = MSR_USER64;
893 regs->msr = MSR_USER32;
897 discard_lazy_cpu_state();
899 current->thread.used_vsr = 0;
901 memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
902 current->thread.fpscr.val = 0;
903 #ifdef CONFIG_ALTIVEC
904 memset(current->thread.vr, 0, sizeof(current->thread.vr));
905 memset(¤t->thread.vscr, 0, sizeof(current->thread.vscr));
906 current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
907 current->thread.vrsave = 0;
908 current->thread.used_vr = 0;
909 #endif /* CONFIG_ALTIVEC */
911 memset(current->thread.evr, 0, sizeof(current->thread.evr));
912 current->thread.acc = 0;
913 current->thread.spefscr = 0;
914 current->thread.used_spe = 0;
915 #endif /* CONFIG_SPE */
918 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
919 | PR_FP_EXC_RES | PR_FP_EXC_INV)
921 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
923 struct pt_regs *regs = tsk->thread.regs;
925 /* This is a bit hairy. If we are an SPE enabled processor
926 * (have embedded fp) we store the IEEE exception enable flags in
927 * fpexc_mode. fpexc_mode is also used for setting FP exception
928 * mode (asyn, precise, disabled) for 'Classic' FP. */
929 if (val & PR_FP_EXC_SW_ENABLE) {
931 if (cpu_has_feature(CPU_FTR_SPE)) {
932 tsk->thread.fpexc_mode = val &
933 (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
943 /* on a CONFIG_SPE this does not hurt us. The bits that
944 * __pack_fe01 use do not overlap with bits used for
945 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
946 * on CONFIG_SPE implementations are reserved so writing to
947 * them does not change anything */
948 if (val > PR_FP_EXC_PRECISE)
950 tsk->thread.fpexc_mode = __pack_fe01(val);
951 if (regs != NULL && (regs->msr & MSR_FP) != 0)
952 regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
953 | tsk->thread.fpexc_mode;
957 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
961 if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
963 if (cpu_has_feature(CPU_FTR_SPE))
964 val = tsk->thread.fpexc_mode;
971 val = __unpack_fe01(tsk->thread.fpexc_mode);
972 return put_user(val, (unsigned int __user *) adr);
975 int set_endian(struct task_struct *tsk, unsigned int val)
977 struct pt_regs *regs = tsk->thread.regs;
979 if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
980 (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
986 if (val == PR_ENDIAN_BIG)
987 regs->msr &= ~MSR_LE;
988 else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
996 int get_endian(struct task_struct *tsk, unsigned long adr)
998 struct pt_regs *regs = tsk->thread.regs;
1001 if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
1002 !cpu_has_feature(CPU_FTR_REAL_LE))
1008 if (regs->msr & MSR_LE) {
1009 if (cpu_has_feature(CPU_FTR_REAL_LE))
1010 val = PR_ENDIAN_LITTLE;
1012 val = PR_ENDIAN_PPC_LITTLE;
1014 val = PR_ENDIAN_BIG;
1016 return put_user(val, (unsigned int __user *)adr);
1019 int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
1021 tsk->thread.align_ctl = val;
1025 int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
1027 return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
1030 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
1032 int sys_clone(unsigned long clone_flags, unsigned long usp,
1033 int __user *parent_tidp, void __user *child_threadptr,
1034 int __user *child_tidp, int p6,
1035 struct pt_regs *regs)
1037 CHECK_FULL_REGS(regs);
1039 usp = regs->gpr[1]; /* stack pointer for child */
1041 if (is_32bit_task()) {
1042 parent_tidp = TRUNC_PTR(parent_tidp);
1043 child_tidp = TRUNC_PTR(child_tidp);
1046 return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
1049 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
1050 unsigned long p4, unsigned long p5, unsigned long p6,
1051 struct pt_regs *regs)
1053 CHECK_FULL_REGS(regs);
1054 return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
1057 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
1058 unsigned long p4, unsigned long p5, unsigned long p6,
1059 struct pt_regs *regs)
1061 CHECK_FULL_REGS(regs);
1062 return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
1063 regs, 0, NULL, NULL);
1066 void __ret_from_kernel_execve(struct pt_regs *normal)
1069 void ret_from_kernel_execve(struct pt_regs *normal)
1071 set_thread_flag(TIF_RESTOREALL);
1072 __ret_from_kernel_execve(normal);
1075 static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
1076 unsigned long nbytes)
1078 unsigned long stack_page;
1079 unsigned long cpu = task_cpu(p);
1082 * Avoid crashing if the stack has overflowed and corrupted
1083 * task_cpu(p), which is in the thread_info struct.
1085 if (cpu < NR_CPUS && cpu_possible(cpu)) {
1086 stack_page = (unsigned long) hardirq_ctx[cpu];
1087 if (sp >= stack_page + sizeof(struct thread_struct)
1088 && sp <= stack_page + THREAD_SIZE - nbytes)
1091 stack_page = (unsigned long) softirq_ctx[cpu];
1092 if (sp >= stack_page + sizeof(struct thread_struct)
1093 && sp <= stack_page + THREAD_SIZE - nbytes)
1099 int validate_sp(unsigned long sp, struct task_struct *p,
1100 unsigned long nbytes)
1102 unsigned long stack_page = (unsigned long)task_stack_page(p);
1104 if (sp >= stack_page + sizeof(struct thread_struct)
1105 && sp <= stack_page + THREAD_SIZE - nbytes)
1108 return valid_irq_stack(sp, p, nbytes);
1111 EXPORT_SYMBOL(validate_sp);
1113 unsigned long get_wchan(struct task_struct *p)
1115 unsigned long ip, sp;
1118 if (!p || p == current || p->state == TASK_RUNNING)
1122 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
1126 sp = *(unsigned long *)sp;
1127 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
1130 ip = ((unsigned long *)sp)[STACK_FRAME_LR_SAVE];
1131 if (!in_sched_functions(ip))
1134 } while (count++ < 16);
1138 static int kstack_depth_to_print = CONFIG_PRINT_STACK_DEPTH;
1140 void show_stack(struct task_struct *tsk, unsigned long *stack)
1142 unsigned long sp, ip, lr, newsp;
1145 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1146 int curr_frame = current->curr_ret_stack;
1147 extern void return_to_handler(void);
1148 unsigned long rth = (unsigned long)return_to_handler;
1149 unsigned long mrth = -1;
1151 extern void mod_return_to_handler(void);
1152 rth = *(unsigned long *)rth;
1153 mrth = (unsigned long)mod_return_to_handler;
1154 mrth = *(unsigned long *)mrth;
1158 sp = (unsigned long) stack;
1163 asm("mr %0,1" : "=r" (sp));
1165 sp = tsk->thread.ksp;
1169 printk("Call Trace:\n");
1171 if (!validate_sp(sp, tsk, STACK_FRAME_OVERHEAD))
1174 stack = (unsigned long *) sp;
1176 ip = stack[STACK_FRAME_LR_SAVE];
1177 if (!firstframe || ip != lr) {
1178 printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip);
1179 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1180 if ((ip == rth || ip == mrth) && curr_frame >= 0) {
1182 (void *)current->ret_stack[curr_frame].ret);
1187 printk(" (unreliable)");
1193 * See if this is an exception frame.
1194 * We look for the "regshere" marker in the current frame.
1196 if (validate_sp(sp, tsk, STACK_INT_FRAME_SIZE)
1197 && stack[STACK_FRAME_MARKER] == STACK_FRAME_REGS_MARKER) {
1198 struct pt_regs *regs = (struct pt_regs *)
1199 (sp + STACK_FRAME_OVERHEAD);
1201 printk("--- Exception: %lx at %pS\n LR = %pS\n",
1202 regs->trap, (void *)regs->nip, (void *)lr);
1207 } while (count++ < kstack_depth_to_print);
1210 void dump_stack(void)
1212 show_stack(current, NULL);
1214 EXPORT_SYMBOL(dump_stack);
1217 /* Called with hard IRQs off */
1218 void __ppc64_runlatch_on(void)
1220 struct thread_info *ti = current_thread_info();
1223 ctrl = mfspr(SPRN_CTRLF);
1224 ctrl |= CTRL_RUNLATCH;
1225 mtspr(SPRN_CTRLT, ctrl);
1227 ti->local_flags |= _TLF_RUNLATCH;
1230 /* Called with hard IRQs off */
1231 void __ppc64_runlatch_off(void)
1233 struct thread_info *ti = current_thread_info();
1236 ti->local_flags &= ~_TLF_RUNLATCH;
1238 ctrl = mfspr(SPRN_CTRLF);
1239 ctrl &= ~CTRL_RUNLATCH;
1240 mtspr(SPRN_CTRLT, ctrl);
1242 #endif /* CONFIG_PPC64 */
1244 unsigned long arch_align_stack(unsigned long sp)
1246 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
1247 sp -= get_random_int() & ~PAGE_MASK;
1251 static inline unsigned long brk_rnd(void)
1253 unsigned long rnd = 0;
1255 /* 8MB for 32bit, 1GB for 64bit */
1256 if (is_32bit_task())
1257 rnd = (long)(get_random_int() % (1<<(23-PAGE_SHIFT)));
1259 rnd = (long)(get_random_int() % (1<<(30-PAGE_SHIFT)));
1261 return rnd << PAGE_SHIFT;
1264 unsigned long arch_randomize_brk(struct mm_struct *mm)
1266 unsigned long base = mm->brk;
1269 #ifdef CONFIG_PPC_STD_MMU_64
1271 * If we are using 1TB segments and we are allowed to randomise
1272 * the heap, we can put it above 1TB so it is backed by a 1TB
1273 * segment. Otherwise the heap will be in the bottom 1TB
1274 * which always uses 256MB segments and this may result in a
1275 * performance penalty.
1277 if (!is_32bit_task() && (mmu_highuser_ssize == MMU_SEGSIZE_1T))
1278 base = max_t(unsigned long, mm->brk, 1UL << SID_SHIFT_1T);
1281 ret = PAGE_ALIGN(base + brk_rnd());
1289 unsigned long randomize_et_dyn(unsigned long base)
1291 unsigned long ret = PAGE_ALIGN(base + brk_rnd());