2 * arch/ppc/kernel/process.c
4 * Derived from "arch/i386/kernel/process.c"
5 * Copyright (C) 1995 Linus Torvalds
7 * Updated and modified by Cort Dougan (cort@cs.nmt.edu) and
8 * Paul Mackerras (paulus@cs.anu.edu.au)
11 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
19 #include <linux/config.h>
20 #include <linux/errno.h>
21 #include <linux/sched.h>
22 #include <linux/kernel.h>
24 #include <linux/smp.h>
25 #include <linux/smp_lock.h>
26 #include <linux/stddef.h>
27 #include <linux/unistd.h>
28 #include <linux/ptrace.h>
29 #include <linux/slab.h>
30 #include <linux/user.h>
31 #include <linux/elf.h>
32 #include <linux/init.h>
33 #include <linux/prctl.h>
34 #include <linux/init_task.h>
35 #include <linux/module.h>
36 #include <linux/kallsyms.h>
37 #include <linux/mqueue.h>
38 #include <linux/hardirq.h>
39 #include <linux/utsname.h>
40 #include <linux/kprobes.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/firmware.h>
55 extern unsigned long _get_SP(void);
58 struct task_struct *last_task_used_math = NULL;
59 struct task_struct *last_task_used_altivec = NULL;
60 struct task_struct *last_task_used_spe = NULL;
64 * Make sure the floating-point register state in the
65 * the thread_struct is up to date for task tsk.
67 void flush_fp_to_thread(struct task_struct *tsk)
69 if (tsk->thread.regs) {
71 * We need to disable preemption here because if we didn't,
72 * another process could get scheduled after the regs->msr
73 * test but before we have finished saving the FP registers
74 * to the thread_struct. That process could take over the
75 * FPU, and then when we get scheduled again we would store
76 * bogus values for the remaining FP registers.
79 if (tsk->thread.regs->msr & MSR_FP) {
82 * This should only ever be called for current or
83 * for a stopped child process. Since we save away
84 * the FP register state on context switch on SMP,
85 * there is something wrong if a stopped child appears
86 * to still have its FP state in the CPU registers.
88 BUG_ON(tsk != current);
96 void enable_kernel_fp(void)
98 WARN_ON(preemptible());
101 if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
104 giveup_fpu(NULL); /* just enables FP for kernel */
106 giveup_fpu(last_task_used_math);
107 #endif /* CONFIG_SMP */
109 EXPORT_SYMBOL(enable_kernel_fp);
111 int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
113 if (!tsk->thread.regs)
115 flush_fp_to_thread(current);
117 memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs));
122 #ifdef CONFIG_ALTIVEC
123 void enable_kernel_altivec(void)
125 WARN_ON(preemptible());
128 if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
129 giveup_altivec(current);
131 giveup_altivec(NULL); /* just enable AltiVec for kernel - force */
133 giveup_altivec(last_task_used_altivec);
134 #endif /* CONFIG_SMP */
136 EXPORT_SYMBOL(enable_kernel_altivec);
139 * Make sure the VMX/Altivec register state in the
140 * the thread_struct is up to date for task tsk.
142 void flush_altivec_to_thread(struct task_struct *tsk)
144 if (tsk->thread.regs) {
146 if (tsk->thread.regs->msr & MSR_VEC) {
148 BUG_ON(tsk != current);
150 giveup_altivec(current);
156 int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
158 flush_altivec_to_thread(current);
159 memcpy(vrregs, ¤t->thread.vr[0], sizeof(*vrregs));
162 #endif /* CONFIG_ALTIVEC */
166 void enable_kernel_spe(void)
168 WARN_ON(preemptible());
171 if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
174 giveup_spe(NULL); /* just enable SPE for kernel - force */
176 giveup_spe(last_task_used_spe);
177 #endif /* __SMP __ */
179 EXPORT_SYMBOL(enable_kernel_spe);
181 void flush_spe_to_thread(struct task_struct *tsk)
183 if (tsk->thread.regs) {
185 if (tsk->thread.regs->msr & MSR_SPE) {
187 BUG_ON(tsk != current);
195 int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
197 flush_spe_to_thread(current);
198 /* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */
199 memcpy(evrregs, ¤t->thread.evr[0], sizeof(u32) * 35);
202 #endif /* CONFIG_SPE */
204 int set_dabr(unsigned long dabr)
207 return ppc_md.set_dabr(dabr);
209 mtspr(SPRN_DABR, dabr);
214 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
215 static DEFINE_PER_CPU(unsigned long, current_dabr);
218 struct task_struct *__switch_to(struct task_struct *prev,
219 struct task_struct *new)
221 struct thread_struct *new_thread, *old_thread;
223 struct task_struct *last;
226 /* avoid complexity of lazy save/restore of fpu
227 * by just saving it every time we switch out if
228 * this task used the fpu during the last quantum.
230 * If it tries to use the fpu again, it'll trap and
231 * reload its fp regs. So we don't have to do a restore
232 * every switch, just a save.
235 if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
237 #ifdef CONFIG_ALTIVEC
239 * If the previous thread used altivec in the last quantum
240 * (thus changing altivec regs) then save them.
241 * We used to check the VRSAVE register but not all apps
242 * set it, so we don't rely on it now (and in fact we need
243 * to save & restore VSCR even if VRSAVE == 0). -- paulus
245 * On SMP we always save/restore altivec regs just to avoid the
246 * complexity of changing processors.
249 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
250 giveup_altivec(prev);
251 #endif /* CONFIG_ALTIVEC */
254 * If the previous thread used spe in the last quantum
255 * (thus changing spe regs) then save them.
257 * On SMP we always save/restore spe regs just to avoid the
258 * complexity of changing processors.
260 if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
262 #endif /* CONFIG_SPE */
264 #else /* CONFIG_SMP */
265 #ifdef CONFIG_ALTIVEC
266 /* Avoid the trap. On smp this this never happens since
267 * we don't set last_task_used_altivec -- Cort
269 if (new->thread.regs && last_task_used_altivec == new)
270 new->thread.regs->msr |= MSR_VEC;
271 #endif /* CONFIG_ALTIVEC */
273 /* Avoid the trap. On smp this this never happens since
274 * we don't set last_task_used_spe
276 if (new->thread.regs && last_task_used_spe == new)
277 new->thread.regs->msr |= MSR_SPE;
278 #endif /* CONFIG_SPE */
280 #endif /* CONFIG_SMP */
282 #ifdef CONFIG_PPC64 /* for now */
283 if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
284 set_dabr(new->thread.dabr);
285 __get_cpu_var(current_dabr) = new->thread.dabr;
291 new_thread = &new->thread;
292 old_thread = ¤t->thread;
296 * Collect processor utilization data per process
298 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
299 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
300 long unsigned start_tb, current_tb;
301 start_tb = old_thread->start_tb;
302 cu->current_tb = current_tb = mfspr(SPRN_PURR);
303 old_thread->accum_tb += (current_tb - start_tb);
304 new_thread->start_tb = current_tb;
308 local_irq_save(flags);
309 last = _switch(old_thread, new_thread);
311 local_irq_restore(flags);
316 static int instructions_to_print = 16;
319 #define BAD_PC(pc) ((REGION_ID(pc) != KERNEL_REGION_ID) && \
320 (REGION_ID(pc) != VMALLOC_REGION_ID))
322 #define BAD_PC(pc) ((pc) < KERNELBASE)
325 static void show_instructions(struct pt_regs *regs)
328 unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
331 printk("Instruction dump:");
333 for (i = 0; i < instructions_to_print; i++) {
339 if (BAD_PC(pc) || __get_user(instr, (unsigned int *)pc)) {
343 printk("<%08x> ", instr);
345 printk("%08x ", instr);
354 static struct regbit {
367 static void printbits(unsigned long val, struct regbit *bits)
369 const char *sep = "";
372 for (; bits->bit; ++bits)
373 if (val & bits->bit) {
374 printk("%s%s", sep, bits->name);
382 #define REGS_PER_LINE 4
383 #define LAST_VOLATILE 13
386 #define REGS_PER_LINE 8
387 #define LAST_VOLATILE 12
390 void show_regs(struct pt_regs * regs)
394 printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
395 regs->nip, regs->link, regs->ctr);
396 printk("REGS: %p TRAP: %04lx %s (%s)\n",
397 regs, regs->trap, print_tainted(), system_utsname.release);
398 printk("MSR: "REG" ", regs->msr);
399 printbits(regs->msr, msr_bits);
400 printk(" CR: %08lX XER: %08lX\n", regs->ccr, regs->xer);
402 if (trap == 0x300 || trap == 0x600)
403 printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
404 printk("TASK = %p[%d] '%s' THREAD: %p",
405 current, current->pid, current->comm, current->thread_info);
408 printk(" CPU: %d", smp_processor_id());
409 #endif /* CONFIG_SMP */
411 for (i = 0; i < 32; i++) {
412 if ((i % REGS_PER_LINE) == 0)
413 printk("\n" KERN_INFO "GPR%02d: ", i);
414 printk(REG " ", regs->gpr[i]);
415 if (i == LAST_VOLATILE && !FULL_REGS(regs))
419 #ifdef CONFIG_KALLSYMS
421 * Lookup NIP late so we have the best change of getting the
422 * above info out without failing
424 printk("NIP ["REG"] ", regs->nip);
425 print_symbol("%s\n", regs->nip);
426 printk("LR ["REG"] ", regs->link);
427 print_symbol("%s\n", regs->link);
429 show_stack(current, (unsigned long *) regs->gpr[1]);
430 if (!user_mode(regs))
431 show_instructions(regs);
434 void exit_thread(void)
436 kprobe_flush_task(current);
439 if (last_task_used_math == current)
440 last_task_used_math = NULL;
441 #ifdef CONFIG_ALTIVEC
442 if (last_task_used_altivec == current)
443 last_task_used_altivec = NULL;
444 #endif /* CONFIG_ALTIVEC */
446 if (last_task_used_spe == current)
447 last_task_used_spe = NULL;
449 #endif /* CONFIG_SMP */
452 void flush_thread(void)
455 struct thread_info *t = current_thread_info();
457 if (t->flags & _TIF_ABI_PENDING)
458 t->flags ^= (_TIF_ABI_PENDING | _TIF_32BIT);
462 if (last_task_used_math == current)
463 last_task_used_math = NULL;
464 #ifdef CONFIG_ALTIVEC
465 if (last_task_used_altivec == current)
466 last_task_used_altivec = NULL;
467 #endif /* CONFIG_ALTIVEC */
469 if (last_task_used_spe == current)
470 last_task_used_spe = NULL;
472 #endif /* CONFIG_SMP */
474 #ifdef CONFIG_PPC64 /* for now */
475 if (current->thread.dabr) {
476 current->thread.dabr = 0;
483 release_thread(struct task_struct *t)
488 * This gets called before we allocate a new thread and copy
489 * the current task into it.
491 void prepare_to_copy(struct task_struct *tsk)
493 flush_fp_to_thread(current);
494 flush_altivec_to_thread(current);
495 flush_spe_to_thread(current);
501 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
502 unsigned long unused, struct task_struct *p,
503 struct pt_regs *regs)
505 struct pt_regs *childregs, *kregs;
506 extern void ret_from_fork(void);
507 unsigned long sp = (unsigned long)p->thread_info + THREAD_SIZE;
509 CHECK_FULL_REGS(regs);
511 sp -= sizeof(struct pt_regs);
512 childregs = (struct pt_regs *) sp;
514 if ((childregs->msr & MSR_PR) == 0) {
515 /* for kernel thread, set `current' and stackptr in new task */
516 childregs->gpr[1] = sp + sizeof(struct pt_regs);
518 childregs->gpr[2] = (unsigned long) p;
520 clear_ti_thread_flag(p->thread_info, TIF_32BIT);
522 p->thread.regs = NULL; /* no user register state */
524 childregs->gpr[1] = usp;
525 p->thread.regs = childregs;
526 if (clone_flags & CLONE_SETTLS) {
528 if (!test_thread_flag(TIF_32BIT))
529 childregs->gpr[13] = childregs->gpr[6];
532 childregs->gpr[2] = childregs->gpr[6];
535 childregs->gpr[3] = 0; /* Result from fork() */
536 sp -= STACK_FRAME_OVERHEAD;
539 * The way this works is that at some point in the future
540 * some task will call _switch to switch to the new task.
541 * That will pop off the stack frame created below and start
542 * the new task running at ret_from_fork. The new task will
543 * do some house keeping and then return from the fork or clone
544 * system call, using the stack frame created above.
546 sp -= sizeof(struct pt_regs);
547 kregs = (struct pt_regs *) sp;
548 sp -= STACK_FRAME_OVERHEAD;
552 if (cpu_has_feature(CPU_FTR_SLB)) {
553 unsigned long sp_vsid = get_kernel_vsid(sp);
554 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
556 sp_vsid <<= SLB_VSID_SHIFT;
557 sp_vsid |= SLB_VSID_KERNEL | llp;
558 p->thread.ksp_vsid = sp_vsid;
562 * The PPC64 ABI makes use of a TOC to contain function
563 * pointers. The function (ret_from_except) is actually a pointer
564 * to the TOC entry. The first entry is a pointer to the actual
567 kregs->nip = *((unsigned long *)ret_from_fork);
569 kregs->nip = (unsigned long)ret_from_fork;
570 p->thread.last_syscall = -1;
577 * Set up a thread for executing a new program
579 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
582 unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
588 * If we exec out of a kernel thread then thread.regs will not be
591 if (!current->thread.regs) {
592 unsigned long childregs = (unsigned long)current->thread_info +
594 childregs -= sizeof(struct pt_regs);
595 current->thread.regs = (struct pt_regs *)childregs;
598 memset(regs->gpr, 0, sizeof(regs->gpr));
608 regs->msr = MSR_USER;
610 if (!test_thread_flag(TIF_32BIT)) {
611 unsigned long entry, toc;
613 /* start is a relocated pointer to the function descriptor for
614 * the elf _start routine. The first entry in the function
615 * descriptor is the entry address of _start and the second
616 * entry is the TOC value we need to use.
618 __get_user(entry, (unsigned long __user *)start);
619 __get_user(toc, (unsigned long __user *)start+1);
621 /* Check whether the e_entry function descriptor entries
622 * need to be relocated before we can use them.
624 if (load_addr != 0) {
630 regs->msr = MSR_USER64;
634 regs->msr = MSR_USER32;
639 if (last_task_used_math == current)
640 last_task_used_math = NULL;
641 #ifdef CONFIG_ALTIVEC
642 if (last_task_used_altivec == current)
643 last_task_used_altivec = NULL;
646 if (last_task_used_spe == current)
647 last_task_used_spe = NULL;
649 #endif /* CONFIG_SMP */
650 memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
651 current->thread.fpscr.val = 0;
652 #ifdef CONFIG_ALTIVEC
653 memset(current->thread.vr, 0, sizeof(current->thread.vr));
654 memset(¤t->thread.vscr, 0, sizeof(current->thread.vscr));
655 current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
656 current->thread.vrsave = 0;
657 current->thread.used_vr = 0;
658 #endif /* CONFIG_ALTIVEC */
660 memset(current->thread.evr, 0, sizeof(current->thread.evr));
661 current->thread.acc = 0;
662 current->thread.spefscr = 0;
663 current->thread.used_spe = 0;
664 #endif /* CONFIG_SPE */
667 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
668 | PR_FP_EXC_RES | PR_FP_EXC_INV)
670 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
672 struct pt_regs *regs = tsk->thread.regs;
674 /* This is a bit hairy. If we are an SPE enabled processor
675 * (have embedded fp) we store the IEEE exception enable flags in
676 * fpexc_mode. fpexc_mode is also used for setting FP exception
677 * mode (asyn, precise, disabled) for 'Classic' FP. */
678 if (val & PR_FP_EXC_SW_ENABLE) {
680 tsk->thread.fpexc_mode = val &
681 (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
688 /* on a CONFIG_SPE this does not hurt us. The bits that
689 * __pack_fe01 use do not overlap with bits used for
690 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
691 * on CONFIG_SPE implementations are reserved so writing to
692 * them does not change anything */
693 if (val > PR_FP_EXC_PRECISE)
695 tsk->thread.fpexc_mode = __pack_fe01(val);
696 if (regs != NULL && (regs->msr & MSR_FP) != 0)
697 regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
698 | tsk->thread.fpexc_mode;
702 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
706 if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
708 val = tsk->thread.fpexc_mode;
713 val = __unpack_fe01(tsk->thread.fpexc_mode);
714 return put_user(val, (unsigned int __user *) adr);
717 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
719 int sys_clone(unsigned long clone_flags, unsigned long usp,
720 int __user *parent_tidp, void __user *child_threadptr,
721 int __user *child_tidp, int p6,
722 struct pt_regs *regs)
724 CHECK_FULL_REGS(regs);
726 usp = regs->gpr[1]; /* stack pointer for child */
728 if (test_thread_flag(TIF_32BIT)) {
729 parent_tidp = TRUNC_PTR(parent_tidp);
730 child_tidp = TRUNC_PTR(child_tidp);
733 return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
736 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
737 unsigned long p4, unsigned long p5, unsigned long p6,
738 struct pt_regs *regs)
740 CHECK_FULL_REGS(regs);
741 return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
744 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
745 unsigned long p4, unsigned long p5, unsigned long p6,
746 struct pt_regs *regs)
748 CHECK_FULL_REGS(regs);
749 return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
750 regs, 0, NULL, NULL);
753 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
754 unsigned long a3, unsigned long a4, unsigned long a5,
755 struct pt_regs *regs)
760 filename = getname((char __user *) a0);
761 error = PTR_ERR(filename);
762 if (IS_ERR(filename))
764 flush_fp_to_thread(current);
765 flush_altivec_to_thread(current);
766 flush_spe_to_thread(current);
767 error = do_execve(filename, (char __user * __user *) a1,
768 (char __user * __user *) a2, regs);
771 current->ptrace &= ~PT_DTRACE;
772 task_unlock(current);
779 static int validate_sp(unsigned long sp, struct task_struct *p,
780 unsigned long nbytes)
782 unsigned long stack_page = (unsigned long)p->thread_info;
784 if (sp >= stack_page + sizeof(struct thread_struct)
785 && sp <= stack_page + THREAD_SIZE - nbytes)
788 #ifdef CONFIG_IRQSTACKS
789 stack_page = (unsigned long) hardirq_ctx[task_cpu(p)];
790 if (sp >= stack_page + sizeof(struct thread_struct)
791 && sp <= stack_page + THREAD_SIZE - nbytes)
794 stack_page = (unsigned long) softirq_ctx[task_cpu(p)];
795 if (sp >= stack_page + sizeof(struct thread_struct)
796 && sp <= stack_page + THREAD_SIZE - nbytes)
804 #define MIN_STACK_FRAME 112 /* same as STACK_FRAME_OVERHEAD, in fact */
805 #define FRAME_LR_SAVE 2
806 #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
807 #define REGS_MARKER 0x7265677368657265ul
808 #define FRAME_MARKER 12
810 #define MIN_STACK_FRAME 16
811 #define FRAME_LR_SAVE 1
812 #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
813 #define REGS_MARKER 0x72656773ul
814 #define FRAME_MARKER 2
817 unsigned long get_wchan(struct task_struct *p)
819 unsigned long ip, sp;
822 if (!p || p == current || p->state == TASK_RUNNING)
826 if (!validate_sp(sp, p, MIN_STACK_FRAME))
830 sp = *(unsigned long *)sp;
831 if (!validate_sp(sp, p, MIN_STACK_FRAME))
834 ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
835 if (!in_sched_functions(ip))
838 } while (count++ < 16);
841 EXPORT_SYMBOL(get_wchan);
843 static int kstack_depth_to_print = 64;
845 void show_stack(struct task_struct *tsk, unsigned long *stack)
847 unsigned long sp, ip, lr, newsp;
851 sp = (unsigned long) stack;
856 asm("mr %0,1" : "=r" (sp));
858 sp = tsk->thread.ksp;
862 printk("Call Trace:\n");
864 if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
867 stack = (unsigned long *) sp;
869 ip = stack[FRAME_LR_SAVE];
870 if (!firstframe || ip != lr) {
871 printk("["REG"] ["REG"] ", sp, ip);
872 print_symbol("%s", ip);
874 printk(" (unreliable)");
880 * See if this is an exception frame.
881 * We look for the "regshere" marker in the current frame.
883 if (validate_sp(sp, tsk, INT_FRAME_SIZE)
884 && stack[FRAME_MARKER] == REGS_MARKER) {
885 struct pt_regs *regs = (struct pt_regs *)
886 (sp + STACK_FRAME_OVERHEAD);
887 printk("--- Exception: %lx", regs->trap);
888 print_symbol(" at %s\n", regs->nip);
890 print_symbol(" LR = %s\n", lr);
895 } while (count++ < kstack_depth_to_print);
898 void dump_stack(void)
900 show_stack(current, NULL);
902 EXPORT_SYMBOL(dump_stack);