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/module.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>
41 #include <asm/pgtable.h>
42 #include <asm/uaccess.h>
43 #include <asm/system.h>
45 #include <asm/processor.h>
48 #include <asm/machdep.h>
50 #include <asm/syscalls.h>
52 #include <asm/firmware.h>
54 #include <linux/kprobes.h>
55 #include <linux/kdebug.h>
57 extern unsigned long _get_SP(void);
60 struct task_struct *last_task_used_math = NULL;
61 struct task_struct *last_task_used_altivec = NULL;
62 struct task_struct *last_task_used_vsx = NULL;
63 struct task_struct *last_task_used_spe = NULL;
67 * Make sure the floating-point register state in the
68 * the thread_struct is up to date for task tsk.
70 void flush_fp_to_thread(struct task_struct *tsk)
72 if (tsk->thread.regs) {
74 * We need to disable preemption here because if we didn't,
75 * another process could get scheduled after the regs->msr
76 * test but before we have finished saving the FP registers
77 * to the thread_struct. That process could take over the
78 * FPU, and then when we get scheduled again we would store
79 * bogus values for the remaining FP registers.
82 if (tsk->thread.regs->msr & MSR_FP) {
85 * This should only ever be called for current or
86 * for a stopped child process. Since we save away
87 * the FP register state on context switch on SMP,
88 * there is something wrong if a stopped child appears
89 * to still have its FP state in the CPU registers.
91 BUG_ON(tsk != current);
99 void enable_kernel_fp(void)
101 WARN_ON(preemptible());
104 if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
107 giveup_fpu(NULL); /* just enables FP for kernel */
109 giveup_fpu(last_task_used_math);
110 #endif /* CONFIG_SMP */
112 EXPORT_SYMBOL(enable_kernel_fp);
114 #ifdef CONFIG_ALTIVEC
115 void enable_kernel_altivec(void)
117 WARN_ON(preemptible());
120 if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
121 giveup_altivec(current);
123 giveup_altivec(NULL); /* just enable AltiVec for kernel - force */
125 giveup_altivec(last_task_used_altivec);
126 #endif /* CONFIG_SMP */
128 EXPORT_SYMBOL(enable_kernel_altivec);
131 * Make sure the VMX/Altivec register state in the
132 * the thread_struct is up to date for task tsk.
134 void flush_altivec_to_thread(struct task_struct *tsk)
136 if (tsk->thread.regs) {
138 if (tsk->thread.regs->msr & MSR_VEC) {
140 BUG_ON(tsk != current);
147 #endif /* CONFIG_ALTIVEC */
151 /* not currently used, but some crazy RAID module might want to later */
152 void enable_kernel_vsx(void)
154 WARN_ON(preemptible());
157 if (current->thread.regs && (current->thread.regs->msr & MSR_VSX))
160 giveup_vsx(NULL); /* just enable vsx for kernel - force */
162 giveup_vsx(last_task_used_vsx);
163 #endif /* CONFIG_SMP */
165 EXPORT_SYMBOL(enable_kernel_vsx);
168 void giveup_vsx(struct task_struct *tsk)
175 void flush_vsx_to_thread(struct task_struct *tsk)
177 if (tsk->thread.regs) {
179 if (tsk->thread.regs->msr & MSR_VSX) {
181 BUG_ON(tsk != current);
188 #endif /* CONFIG_VSX */
192 void enable_kernel_spe(void)
194 WARN_ON(preemptible());
197 if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
200 giveup_spe(NULL); /* just enable SPE for kernel - force */
202 giveup_spe(last_task_used_spe);
203 #endif /* __SMP __ */
205 EXPORT_SYMBOL(enable_kernel_spe);
207 void flush_spe_to_thread(struct task_struct *tsk)
209 if (tsk->thread.regs) {
211 if (tsk->thread.regs->msr & MSR_SPE) {
213 BUG_ON(tsk != current);
220 #endif /* CONFIG_SPE */
224 * If we are doing lazy switching of CPU state (FP, altivec or SPE),
225 * and the current task has some state, discard it.
227 void discard_lazy_cpu_state(void)
230 if (last_task_used_math == current)
231 last_task_used_math = NULL;
232 #ifdef CONFIG_ALTIVEC
233 if (last_task_used_altivec == current)
234 last_task_used_altivec = NULL;
235 #endif /* CONFIG_ALTIVEC */
237 if (last_task_used_vsx == current)
238 last_task_used_vsx = NULL;
239 #endif /* CONFIG_VSX */
241 if (last_task_used_spe == current)
242 last_task_used_spe = NULL;
246 #endif /* CONFIG_SMP */
248 void do_dabr(struct pt_regs *regs, unsigned long address,
249 unsigned long error_code)
253 if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
254 11, SIGSEGV) == NOTIFY_STOP)
257 if (debugger_dabr_match(regs))
260 /* Clear the DAC and struct entries. One shot trigger */
261 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
262 mtspr(SPRN_DBCR0, mfspr(SPRN_DBCR0) & ~(DBSR_DAC1R | DBSR_DAC1W
269 /* Deliver the signal to userspace */
270 info.si_signo = SIGTRAP;
272 info.si_code = TRAP_HWBKPT;
273 info.si_addr = (void __user *)address;
274 force_sig_info(SIGTRAP, &info, current);
277 static DEFINE_PER_CPU(unsigned long, current_dabr);
279 int set_dabr(unsigned long dabr)
281 __get_cpu_var(current_dabr) = dabr;
284 return ppc_md.set_dabr(dabr);
286 /* XXX should we have a CPU_FTR_HAS_DABR ? */
287 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
288 mtspr(SPRN_DAC1, dabr);
289 #elif defined(CONFIG_PPC_BOOK3S)
290 mtspr(SPRN_DABR, dabr);
298 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
301 struct task_struct *__switch_to(struct task_struct *prev,
302 struct task_struct *new)
304 struct thread_struct *new_thread, *old_thread;
306 struct task_struct *last;
309 /* avoid complexity of lazy save/restore of fpu
310 * by just saving it every time we switch out if
311 * this task used the fpu during the last quantum.
313 * If it tries to use the fpu again, it'll trap and
314 * reload its fp regs. So we don't have to do a restore
315 * every switch, just a save.
318 if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
320 #ifdef CONFIG_ALTIVEC
322 * If the previous thread used altivec in the last quantum
323 * (thus changing altivec regs) then save them.
324 * We used to check the VRSAVE register but not all apps
325 * set it, so we don't rely on it now (and in fact we need
326 * to save & restore VSCR even if VRSAVE == 0). -- paulus
328 * On SMP we always save/restore altivec regs just to avoid the
329 * complexity of changing processors.
332 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
333 giveup_altivec(prev);
334 #endif /* CONFIG_ALTIVEC */
336 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VSX))
337 /* VMX and FPU registers are already save here */
339 #endif /* CONFIG_VSX */
342 * If the previous thread used spe in the last quantum
343 * (thus changing spe regs) then save them.
345 * On SMP we always save/restore spe regs just to avoid the
346 * complexity of changing processors.
348 if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
350 #endif /* CONFIG_SPE */
352 #else /* CONFIG_SMP */
353 #ifdef CONFIG_ALTIVEC
354 /* Avoid the trap. On smp this this never happens since
355 * we don't set last_task_used_altivec -- Cort
357 if (new->thread.regs && last_task_used_altivec == new)
358 new->thread.regs->msr |= MSR_VEC;
359 #endif /* CONFIG_ALTIVEC */
361 if (new->thread.regs && last_task_used_vsx == new)
362 new->thread.regs->msr |= MSR_VSX;
363 #endif /* CONFIG_VSX */
365 /* Avoid the trap. On smp this this never happens since
366 * we don't set last_task_used_spe
368 if (new->thread.regs && last_task_used_spe == new)
369 new->thread.regs->msr |= MSR_SPE;
370 #endif /* CONFIG_SPE */
372 #endif /* CONFIG_SMP */
374 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
375 /* If new thread DAC (HW breakpoint) is the same then leave it */
376 if (new->thread.dabr)
377 set_dabr(new->thread.dabr);
379 if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr))
380 set_dabr(new->thread.dabr);
384 new_thread = &new->thread;
385 old_thread = ¤t->thread;
389 * Collect processor utilization data per process
391 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
392 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
393 long unsigned start_tb, current_tb;
394 start_tb = old_thread->start_tb;
395 cu->current_tb = current_tb = mfspr(SPRN_PURR);
396 old_thread->accum_tb += (current_tb - start_tb);
397 new_thread->start_tb = current_tb;
401 local_irq_save(flags);
403 account_system_vtime(current);
404 account_process_vtime(current);
405 calculate_steal_time();
408 * We can't take a PMU exception inside _switch() since there is a
409 * window where the kernel stack SLB and the kernel stack are out
410 * of sync. Hard disable here.
413 last = _switch(old_thread, new_thread);
415 local_irq_restore(flags);
420 static int instructions_to_print = 16;
422 static void show_instructions(struct pt_regs *regs)
425 unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
428 printk("Instruction dump:");
430 for (i = 0; i < instructions_to_print; i++) {
436 #if !defined(CONFIG_BOOKE)
437 /* If executing with the IMMU off, adjust pc rather
438 * than print XXXXXXXX.
440 if (!(regs->msr & MSR_IR))
441 pc = (unsigned long)phys_to_virt(pc);
444 /* We use __get_user here *only* to avoid an OOPS on a
445 * bad address because the pc *should* only be a
448 if (!__kernel_text_address(pc) ||
449 __get_user(instr, (unsigned int __user *)pc)) {
453 printk("<%08x> ", instr);
455 printk("%08x ", instr);
464 static struct regbit {
481 static void printbits(unsigned long val, struct regbit *bits)
483 const char *sep = "";
486 for (; bits->bit; ++bits)
487 if (val & bits->bit) {
488 printk("%s%s", sep, bits->name);
496 #define REGS_PER_LINE 4
497 #define LAST_VOLATILE 13
500 #define REGS_PER_LINE 8
501 #define LAST_VOLATILE 12
504 void show_regs(struct pt_regs * regs)
508 printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
509 regs->nip, regs->link, regs->ctr);
510 printk("REGS: %p TRAP: %04lx %s (%s)\n",
511 regs, regs->trap, print_tainted(), init_utsname()->release);
512 printk("MSR: "REG" ", regs->msr);
513 printbits(regs->msr, msr_bits);
514 printk(" CR: %08lx XER: %08lx\n", regs->ccr, regs->xer);
516 if (trap == 0x300 || trap == 0x600)
517 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
518 printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
520 printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
522 printk("TASK = %p[%d] '%s' THREAD: %p",
523 current, task_pid_nr(current), current->comm, task_thread_info(current));
526 printk(" CPU: %d", raw_smp_processor_id());
527 #endif /* CONFIG_SMP */
529 for (i = 0; i < 32; i++) {
530 if ((i % REGS_PER_LINE) == 0)
531 printk("\nGPR%02d: ", i);
532 printk(REG " ", regs->gpr[i]);
533 if (i == LAST_VOLATILE && !FULL_REGS(regs))
537 #ifdef CONFIG_KALLSYMS
539 * Lookup NIP late so we have the best change of getting the
540 * above info out without failing
542 printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
543 printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
545 show_stack(current, (unsigned long *) regs->gpr[1]);
546 if (!user_mode(regs))
547 show_instructions(regs);
550 void exit_thread(void)
552 discard_lazy_cpu_state();
555 void flush_thread(void)
557 discard_lazy_cpu_state();
559 if (current->thread.dabr) {
560 current->thread.dabr = 0;
563 #ifdef CONFIG_PPC_ADV_DEBUG_REGS
564 current->thread.dbcr0 &= ~(DBSR_DAC1R | DBSR_DAC1W);
570 release_thread(struct task_struct *t)
575 * This gets called before we allocate a new thread and copy
576 * the current task into it.
578 void prepare_to_copy(struct task_struct *tsk)
580 flush_fp_to_thread(current);
581 flush_altivec_to_thread(current);
582 flush_vsx_to_thread(current);
583 flush_spe_to_thread(current);
589 int copy_thread(unsigned long clone_flags, unsigned long usp,
590 unsigned long unused, struct task_struct *p,
591 struct pt_regs *regs)
593 struct pt_regs *childregs, *kregs;
594 extern void ret_from_fork(void);
595 unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
597 CHECK_FULL_REGS(regs);
599 sp -= sizeof(struct pt_regs);
600 childregs = (struct pt_regs *) sp;
602 if ((childregs->msr & MSR_PR) == 0) {
603 /* for kernel thread, set `current' and stackptr in new task */
604 childregs->gpr[1] = sp + sizeof(struct pt_regs);
606 childregs->gpr[2] = (unsigned long) p;
608 clear_tsk_thread_flag(p, TIF_32BIT);
610 p->thread.regs = NULL; /* no user register state */
612 childregs->gpr[1] = usp;
613 p->thread.regs = childregs;
614 if (clone_flags & CLONE_SETTLS) {
616 if (!test_thread_flag(TIF_32BIT))
617 childregs->gpr[13] = childregs->gpr[6];
620 childregs->gpr[2] = childregs->gpr[6];
623 childregs->gpr[3] = 0; /* Result from fork() */
624 sp -= STACK_FRAME_OVERHEAD;
627 * The way this works is that at some point in the future
628 * some task will call _switch to switch to the new task.
629 * That will pop off the stack frame created below and start
630 * the new task running at ret_from_fork. The new task will
631 * do some house keeping and then return from the fork or clone
632 * system call, using the stack frame created above.
634 sp -= sizeof(struct pt_regs);
635 kregs = (struct pt_regs *) sp;
636 sp -= STACK_FRAME_OVERHEAD;
638 p->thread.ksp_limit = (unsigned long)task_stack_page(p) +
639 _ALIGN_UP(sizeof(struct thread_info), 16);
641 #ifdef CONFIG_PPC_STD_MMU_64
642 if (cpu_has_feature(CPU_FTR_SLB)) {
643 unsigned long sp_vsid;
644 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
646 if (cpu_has_feature(CPU_FTR_1T_SEGMENT))
647 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
648 << SLB_VSID_SHIFT_1T;
650 sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_256M)
652 sp_vsid |= SLB_VSID_KERNEL | llp;
653 p->thread.ksp_vsid = sp_vsid;
655 #endif /* CONFIG_PPC_STD_MMU_64 */
658 * The PPC64 ABI makes use of a TOC to contain function
659 * pointers. The function (ret_from_except) is actually a pointer
660 * to the TOC entry. The first entry is a pointer to the actual
664 kregs->nip = *((unsigned long *)ret_from_fork);
666 kregs->nip = (unsigned long)ret_from_fork;
673 * Set up a thread for executing a new program
675 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
678 unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
684 * If we exec out of a kernel thread then thread.regs will not be
687 if (!current->thread.regs) {
688 struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
689 current->thread.regs = regs - 1;
692 memset(regs->gpr, 0, sizeof(regs->gpr));
700 * We have just cleared all the nonvolatile GPRs, so make
701 * FULL_REGS(regs) return true. This is necessary to allow
702 * ptrace to examine the thread immediately after exec.
709 regs->msr = MSR_USER;
711 if (!test_thread_flag(TIF_32BIT)) {
712 unsigned long entry, toc;
714 /* start is a relocated pointer to the function descriptor for
715 * the elf _start routine. The first entry in the function
716 * descriptor is the entry address of _start and the second
717 * entry is the TOC value we need to use.
719 __get_user(entry, (unsigned long __user *)start);
720 __get_user(toc, (unsigned long __user *)start+1);
722 /* Check whether the e_entry function descriptor entries
723 * need to be relocated before we can use them.
725 if (load_addr != 0) {
731 regs->msr = MSR_USER64;
735 regs->msr = MSR_USER32;
739 discard_lazy_cpu_state();
741 current->thread.used_vsr = 0;
743 memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
744 current->thread.fpscr.val = 0;
745 #ifdef CONFIG_ALTIVEC
746 memset(current->thread.vr, 0, sizeof(current->thread.vr));
747 memset(¤t->thread.vscr, 0, sizeof(current->thread.vscr));
748 current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
749 current->thread.vrsave = 0;
750 current->thread.used_vr = 0;
751 #endif /* CONFIG_ALTIVEC */
753 memset(current->thread.evr, 0, sizeof(current->thread.evr));
754 current->thread.acc = 0;
755 current->thread.spefscr = 0;
756 current->thread.used_spe = 0;
757 #endif /* CONFIG_SPE */
760 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
761 | PR_FP_EXC_RES | PR_FP_EXC_INV)
763 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
765 struct pt_regs *regs = tsk->thread.regs;
767 /* This is a bit hairy. If we are an SPE enabled processor
768 * (have embedded fp) we store the IEEE exception enable flags in
769 * fpexc_mode. fpexc_mode is also used for setting FP exception
770 * mode (asyn, precise, disabled) for 'Classic' FP. */
771 if (val & PR_FP_EXC_SW_ENABLE) {
773 if (cpu_has_feature(CPU_FTR_SPE)) {
774 tsk->thread.fpexc_mode = val &
775 (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
785 /* on a CONFIG_SPE this does not hurt us. The bits that
786 * __pack_fe01 use do not overlap with bits used for
787 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
788 * on CONFIG_SPE implementations are reserved so writing to
789 * them does not change anything */
790 if (val > PR_FP_EXC_PRECISE)
792 tsk->thread.fpexc_mode = __pack_fe01(val);
793 if (regs != NULL && (regs->msr & MSR_FP) != 0)
794 regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
795 | tsk->thread.fpexc_mode;
799 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
803 if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
805 if (cpu_has_feature(CPU_FTR_SPE))
806 val = tsk->thread.fpexc_mode;
813 val = __unpack_fe01(tsk->thread.fpexc_mode);
814 return put_user(val, (unsigned int __user *) adr);
817 int set_endian(struct task_struct *tsk, unsigned int val)
819 struct pt_regs *regs = tsk->thread.regs;
821 if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
822 (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
828 if (val == PR_ENDIAN_BIG)
829 regs->msr &= ~MSR_LE;
830 else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
838 int get_endian(struct task_struct *tsk, unsigned long adr)
840 struct pt_regs *regs = tsk->thread.regs;
843 if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
844 !cpu_has_feature(CPU_FTR_REAL_LE))
850 if (regs->msr & MSR_LE) {
851 if (cpu_has_feature(CPU_FTR_REAL_LE))
852 val = PR_ENDIAN_LITTLE;
854 val = PR_ENDIAN_PPC_LITTLE;
858 return put_user(val, (unsigned int __user *)adr);
861 int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
863 tsk->thread.align_ctl = val;
867 int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
869 return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
872 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
874 int sys_clone(unsigned long clone_flags, unsigned long usp,
875 int __user *parent_tidp, void __user *child_threadptr,
876 int __user *child_tidp, int p6,
877 struct pt_regs *regs)
879 CHECK_FULL_REGS(regs);
881 usp = regs->gpr[1]; /* stack pointer for child */
883 if (test_thread_flag(TIF_32BIT)) {
884 parent_tidp = TRUNC_PTR(parent_tidp);
885 child_tidp = TRUNC_PTR(child_tidp);
888 return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
891 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
892 unsigned long p4, unsigned long p5, unsigned long p6,
893 struct pt_regs *regs)
895 CHECK_FULL_REGS(regs);
896 return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
899 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
900 unsigned long p4, unsigned long p5, unsigned long p6,
901 struct pt_regs *regs)
903 CHECK_FULL_REGS(regs);
904 return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
905 regs, 0, NULL, NULL);
908 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
909 unsigned long a3, unsigned long a4, unsigned long a5,
910 struct pt_regs *regs)
915 filename = getname((char __user *) a0);
916 error = PTR_ERR(filename);
917 if (IS_ERR(filename))
919 flush_fp_to_thread(current);
920 flush_altivec_to_thread(current);
921 flush_spe_to_thread(current);
922 error = do_execve(filename, (char __user * __user *) a1,
923 (char __user * __user *) a2, regs);
929 #ifdef CONFIG_IRQSTACKS
930 static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
931 unsigned long nbytes)
933 unsigned long stack_page;
934 unsigned long cpu = task_cpu(p);
937 * Avoid crashing if the stack has overflowed and corrupted
938 * task_cpu(p), which is in the thread_info struct.
940 if (cpu < NR_CPUS && cpu_possible(cpu)) {
941 stack_page = (unsigned long) hardirq_ctx[cpu];
942 if (sp >= stack_page + sizeof(struct thread_struct)
943 && sp <= stack_page + THREAD_SIZE - nbytes)
946 stack_page = (unsigned long) softirq_ctx[cpu];
947 if (sp >= stack_page + sizeof(struct thread_struct)
948 && sp <= stack_page + THREAD_SIZE - nbytes)
955 #define valid_irq_stack(sp, p, nb) 0
956 #endif /* CONFIG_IRQSTACKS */
958 int validate_sp(unsigned long sp, struct task_struct *p,
959 unsigned long nbytes)
961 unsigned long stack_page = (unsigned long)task_stack_page(p);
963 if (sp >= stack_page + sizeof(struct thread_struct)
964 && sp <= stack_page + THREAD_SIZE - nbytes)
967 return valid_irq_stack(sp, p, nbytes);
970 EXPORT_SYMBOL(validate_sp);
972 unsigned long get_wchan(struct task_struct *p)
974 unsigned long ip, sp;
977 if (!p || p == current || p->state == TASK_RUNNING)
981 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
985 sp = *(unsigned long *)sp;
986 if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
989 ip = ((unsigned long *)sp)[STACK_FRAME_LR_SAVE];
990 if (!in_sched_functions(ip))
993 } while (count++ < 16);
997 static int kstack_depth_to_print = CONFIG_PRINT_STACK_DEPTH;
999 void show_stack(struct task_struct *tsk, unsigned long *stack)
1001 unsigned long sp, ip, lr, newsp;
1004 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1005 int curr_frame = current->curr_ret_stack;
1006 extern void return_to_handler(void);
1007 unsigned long rth = (unsigned long)return_to_handler;
1008 unsigned long mrth = -1;
1010 extern void mod_return_to_handler(void);
1011 rth = *(unsigned long *)rth;
1012 mrth = (unsigned long)mod_return_to_handler;
1013 mrth = *(unsigned long *)mrth;
1017 sp = (unsigned long) stack;
1022 asm("mr %0,1" : "=r" (sp));
1024 sp = tsk->thread.ksp;
1028 printk("Call Trace:\n");
1030 if (!validate_sp(sp, tsk, STACK_FRAME_OVERHEAD))
1033 stack = (unsigned long *) sp;
1035 ip = stack[STACK_FRAME_LR_SAVE];
1036 if (!firstframe || ip != lr) {
1037 printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip);
1038 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
1039 if ((ip == rth || ip == mrth) && curr_frame >= 0) {
1041 (void *)current->ret_stack[curr_frame].ret);
1046 printk(" (unreliable)");
1052 * See if this is an exception frame.
1053 * We look for the "regshere" marker in the current frame.
1055 if (validate_sp(sp, tsk, STACK_INT_FRAME_SIZE)
1056 && stack[STACK_FRAME_MARKER] == STACK_FRAME_REGS_MARKER) {
1057 struct pt_regs *regs = (struct pt_regs *)
1058 (sp + STACK_FRAME_OVERHEAD);
1060 printk("--- Exception: %lx at %pS\n LR = %pS\n",
1061 regs->trap, (void *)regs->nip, (void *)lr);
1066 } while (count++ < kstack_depth_to_print);
1069 void dump_stack(void)
1071 show_stack(current, NULL);
1073 EXPORT_SYMBOL(dump_stack);
1076 void ppc64_runlatch_on(void)
1080 if (cpu_has_feature(CPU_FTR_CTRL) && !test_thread_flag(TIF_RUNLATCH)) {
1083 ctrl = mfspr(SPRN_CTRLF);
1084 ctrl |= CTRL_RUNLATCH;
1085 mtspr(SPRN_CTRLT, ctrl);
1087 set_thread_flag(TIF_RUNLATCH);
1091 void ppc64_runlatch_off(void)
1095 if (cpu_has_feature(CPU_FTR_CTRL) && test_thread_flag(TIF_RUNLATCH)) {
1098 clear_thread_flag(TIF_RUNLATCH);
1100 ctrl = mfspr(SPRN_CTRLF);
1101 ctrl &= ~CTRL_RUNLATCH;
1102 mtspr(SPRN_CTRLT, ctrl);
1107 #if THREAD_SHIFT < PAGE_SHIFT
1109 static struct kmem_cache *thread_info_cache;
1111 struct thread_info *alloc_thread_info(struct task_struct *tsk)
1113 struct thread_info *ti;
1115 ti = kmem_cache_alloc(thread_info_cache, GFP_KERNEL);
1116 if (unlikely(ti == NULL))
1118 #ifdef CONFIG_DEBUG_STACK_USAGE
1119 memset(ti, 0, THREAD_SIZE);
1124 void free_thread_info(struct thread_info *ti)
1126 kmem_cache_free(thread_info_cache, ti);
1129 void thread_info_cache_init(void)
1131 thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE,
1132 THREAD_SIZE, 0, NULL);
1133 BUG_ON(thread_info_cache == NULL);
1136 #endif /* THREAD_SHIFT < PAGE_SHIFT */
1138 unsigned long arch_align_stack(unsigned long sp)
1140 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
1141 sp -= get_random_int() & ~PAGE_MASK;
1145 static inline unsigned long brk_rnd(void)
1147 unsigned long rnd = 0;
1149 /* 8MB for 32bit, 1GB for 64bit */
1150 if (is_32bit_task())
1151 rnd = (long)(get_random_int() % (1<<(23-PAGE_SHIFT)));
1153 rnd = (long)(get_random_int() % (1<<(30-PAGE_SHIFT)));
1155 return rnd << PAGE_SHIFT;
1158 unsigned long arch_randomize_brk(struct mm_struct *mm)
1160 unsigned long base = mm->brk;
1163 #ifdef CONFIG_PPC_STD_MMU_64
1165 * If we are using 1TB segments and we are allowed to randomise
1166 * the heap, we can put it above 1TB so it is backed by a 1TB
1167 * segment. Otherwise the heap will be in the bottom 1TB
1168 * which always uses 256MB segments and this may result in a
1169 * performance penalty.
1171 if (!is_32bit_task() && (mmu_highuser_ssize == MMU_SEGSIZE_1T))
1172 base = max_t(unsigned long, mm->brk, 1UL << SID_SHIFT_1T);
1175 ret = PAGE_ALIGN(base + brk_rnd());
1183 unsigned long randomize_et_dyn(unsigned long base)
1185 unsigned long ret = PAGE_ALIGN(base + brk_rnd());