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/config.h>
18 #include <linux/errno.h>
19 #include <linux/sched.h>
20 #include <linux/kernel.h>
22 #include <linux/smp.h>
23 #include <linux/smp_lock.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/ptrace.h>
27 #include <linux/slab.h>
28 #include <linux/user.h>
29 #include <linux/elf.h>
30 #include <linux/init.h>
31 #include <linux/prctl.h>
32 #include <linux/init_task.h>
33 #include <linux/module.h>
34 #include <linux/kallsyms.h>
35 #include <linux/mqueue.h>
36 #include <linux/hardirq.h>
37 #include <linux/utsname.h>
38 #include <linux/kprobes.h>
40 #include <asm/pgtable.h>
41 #include <asm/uaccess.h>
42 #include <asm/system.h>
44 #include <asm/processor.h>
47 #include <asm/machdep.h>
49 #include <asm/firmware.h>
53 extern unsigned long _get_SP(void);
56 struct task_struct *last_task_used_math = NULL;
57 struct task_struct *last_task_used_altivec = NULL;
58 struct task_struct *last_task_used_spe = NULL;
62 * Make sure the floating-point register state in the
63 * the thread_struct is up to date for task tsk.
65 void flush_fp_to_thread(struct task_struct *tsk)
67 if (tsk->thread.regs) {
69 * We need to disable preemption here because if we didn't,
70 * another process could get scheduled after the regs->msr
71 * test but before we have finished saving the FP registers
72 * to the thread_struct. That process could take over the
73 * FPU, and then when we get scheduled again we would store
74 * bogus values for the remaining FP registers.
77 if (tsk->thread.regs->msr & MSR_FP) {
80 * This should only ever be called for current or
81 * for a stopped child process. Since we save away
82 * the FP register state on context switch on SMP,
83 * there is something wrong if a stopped child appears
84 * to still have its FP state in the CPU registers.
86 BUG_ON(tsk != current);
94 void enable_kernel_fp(void)
96 WARN_ON(preemptible());
99 if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
102 giveup_fpu(NULL); /* just enables FP for kernel */
104 giveup_fpu(last_task_used_math);
105 #endif /* CONFIG_SMP */
107 EXPORT_SYMBOL(enable_kernel_fp);
109 int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
111 if (!tsk->thread.regs)
113 flush_fp_to_thread(current);
115 memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs));
120 #ifdef CONFIG_ALTIVEC
121 void enable_kernel_altivec(void)
123 WARN_ON(preemptible());
126 if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
127 giveup_altivec(current);
129 giveup_altivec(NULL); /* just enable AltiVec for kernel - force */
131 giveup_altivec(last_task_used_altivec);
132 #endif /* CONFIG_SMP */
134 EXPORT_SYMBOL(enable_kernel_altivec);
137 * Make sure the VMX/Altivec register state in the
138 * the thread_struct is up to date for task tsk.
140 void flush_altivec_to_thread(struct task_struct *tsk)
142 if (tsk->thread.regs) {
144 if (tsk->thread.regs->msr & MSR_VEC) {
146 BUG_ON(tsk != current);
148 giveup_altivec(current);
154 int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
156 flush_altivec_to_thread(current);
157 memcpy(vrregs, ¤t->thread.vr[0], sizeof(*vrregs));
160 #endif /* CONFIG_ALTIVEC */
164 void enable_kernel_spe(void)
166 WARN_ON(preemptible());
169 if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
172 giveup_spe(NULL); /* just enable SPE for kernel - force */
174 giveup_spe(last_task_used_spe);
175 #endif /* __SMP __ */
177 EXPORT_SYMBOL(enable_kernel_spe);
179 void flush_spe_to_thread(struct task_struct *tsk)
181 if (tsk->thread.regs) {
183 if (tsk->thread.regs->msr & MSR_SPE) {
185 BUG_ON(tsk != current);
193 int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
195 flush_spe_to_thread(current);
196 /* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */
197 memcpy(evrregs, ¤t->thread.evr[0], sizeof(u32) * 35);
200 #endif /* CONFIG_SPE */
204 * If we are doing lazy switching of CPU state (FP, altivec or SPE),
205 * and the current task has some state, discard it.
207 void discard_lazy_cpu_state(void)
210 if (last_task_used_math == current)
211 last_task_used_math = NULL;
212 #ifdef CONFIG_ALTIVEC
213 if (last_task_used_altivec == current)
214 last_task_used_altivec = NULL;
215 #endif /* CONFIG_ALTIVEC */
217 if (last_task_used_spe == current)
218 last_task_used_spe = NULL;
222 #endif /* CONFIG_SMP */
224 #ifdef CONFIG_PPC_MERGE /* XXX for now */
225 int set_dabr(unsigned long dabr)
228 return ppc_md.set_dabr(dabr);
230 mtspr(SPRN_DABR, dabr);
236 DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
237 static DEFINE_PER_CPU(unsigned long, current_dabr);
240 struct task_struct *__switch_to(struct task_struct *prev,
241 struct task_struct *new)
243 struct thread_struct *new_thread, *old_thread;
245 struct task_struct *last;
248 /* avoid complexity of lazy save/restore of fpu
249 * by just saving it every time we switch out if
250 * this task used the fpu during the last quantum.
252 * If it tries to use the fpu again, it'll trap and
253 * reload its fp regs. So we don't have to do a restore
254 * every switch, just a save.
257 if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
259 #ifdef CONFIG_ALTIVEC
261 * If the previous thread used altivec in the last quantum
262 * (thus changing altivec regs) then save them.
263 * We used to check the VRSAVE register but not all apps
264 * set it, so we don't rely on it now (and in fact we need
265 * to save & restore VSCR even if VRSAVE == 0). -- paulus
267 * On SMP we always save/restore altivec regs just to avoid the
268 * complexity of changing processors.
271 if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
272 giveup_altivec(prev);
273 #endif /* CONFIG_ALTIVEC */
276 * If the previous thread used spe in the last quantum
277 * (thus changing spe regs) then save them.
279 * On SMP we always save/restore spe regs just to avoid the
280 * complexity of changing processors.
282 if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
284 #endif /* CONFIG_SPE */
286 #else /* CONFIG_SMP */
287 #ifdef CONFIG_ALTIVEC
288 /* Avoid the trap. On smp this this never happens since
289 * we don't set last_task_used_altivec -- Cort
291 if (new->thread.regs && last_task_used_altivec == new)
292 new->thread.regs->msr |= MSR_VEC;
293 #endif /* CONFIG_ALTIVEC */
295 /* Avoid the trap. On smp this this never happens since
296 * we don't set last_task_used_spe
298 if (new->thread.regs && last_task_used_spe == new)
299 new->thread.regs->msr |= MSR_SPE;
300 #endif /* CONFIG_SPE */
302 #endif /* CONFIG_SMP */
304 #ifdef CONFIG_PPC64 /* for now */
305 if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
306 set_dabr(new->thread.dabr);
307 __get_cpu_var(current_dabr) = new->thread.dabr;
313 new_thread = &new->thread;
314 old_thread = ¤t->thread;
318 * Collect processor utilization data per process
320 if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
321 struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
322 long unsigned start_tb, current_tb;
323 start_tb = old_thread->start_tb;
324 cu->current_tb = current_tb = mfspr(SPRN_PURR);
325 old_thread->accum_tb += (current_tb - start_tb);
326 new_thread->start_tb = current_tb;
330 local_irq_save(flags);
331 last = _switch(old_thread, new_thread);
333 local_irq_restore(flags);
338 static int instructions_to_print = 16;
341 #define BAD_PC(pc) ((REGION_ID(pc) != KERNEL_REGION_ID) && \
342 (REGION_ID(pc) != VMALLOC_REGION_ID))
344 #define BAD_PC(pc) ((pc) < KERNELBASE)
347 static void show_instructions(struct pt_regs *regs)
350 unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
353 printk("Instruction dump:");
355 for (i = 0; i < instructions_to_print; i++) {
361 if (BAD_PC(pc) || __get_user(instr, (unsigned int *)pc)) {
365 printk("<%08x> ", instr);
367 printk("%08x ", instr);
376 static struct regbit {
389 static void printbits(unsigned long val, struct regbit *bits)
391 const char *sep = "";
394 for (; bits->bit; ++bits)
395 if (val & bits->bit) {
396 printk("%s%s", sep, bits->name);
404 #define REGS_PER_LINE 4
405 #define LAST_VOLATILE 13
408 #define REGS_PER_LINE 8
409 #define LAST_VOLATILE 12
412 void show_regs(struct pt_regs * regs)
416 printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
417 regs->nip, regs->link, regs->ctr);
418 printk("REGS: %p TRAP: %04lx %s (%s)\n",
419 regs, regs->trap, print_tainted(), system_utsname.release);
420 printk("MSR: "REG" ", regs->msr);
421 printbits(regs->msr, msr_bits);
422 printk(" CR: %08lX XER: %08lX\n", regs->ccr, regs->xer);
424 if (trap == 0x300 || trap == 0x600)
425 printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
426 printk("TASK = %p[%d] '%s' THREAD: %p",
427 current, current->pid, current->comm, task_thread_info(current));
430 printk(" CPU: %d", smp_processor_id());
431 #endif /* CONFIG_SMP */
433 for (i = 0; i < 32; i++) {
434 if ((i % REGS_PER_LINE) == 0)
435 printk("\n" KERN_INFO "GPR%02d: ", i);
436 printk(REG " ", regs->gpr[i]);
437 if (i == LAST_VOLATILE && !FULL_REGS(regs))
441 #ifdef CONFIG_KALLSYMS
443 * Lookup NIP late so we have the best change of getting the
444 * above info out without failing
446 printk("NIP ["REG"] ", regs->nip);
447 print_symbol("%s\n", regs->nip);
448 printk("LR ["REG"] ", regs->link);
449 print_symbol("%s\n", regs->link);
451 show_stack(current, (unsigned long *) regs->gpr[1]);
452 if (!user_mode(regs))
453 show_instructions(regs);
456 void exit_thread(void)
458 kprobe_flush_task(current);
459 discard_lazy_cpu_state();
462 void flush_thread(void)
465 struct thread_info *t = current_thread_info();
467 if (t->flags & _TIF_ABI_PENDING)
468 t->flags ^= (_TIF_ABI_PENDING | _TIF_32BIT);
471 discard_lazy_cpu_state();
473 #ifdef CONFIG_PPC64 /* for now */
474 if (current->thread.dabr) {
475 current->thread.dabr = 0;
482 release_thread(struct task_struct *t)
487 * This gets called before we allocate a new thread and copy
488 * the current task into it.
490 void prepare_to_copy(struct task_struct *tsk)
492 flush_fp_to_thread(current);
493 flush_altivec_to_thread(current);
494 flush_spe_to_thread(current);
500 int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
501 unsigned long unused, struct task_struct *p,
502 struct pt_regs *regs)
504 struct pt_regs *childregs, *kregs;
505 extern void ret_from_fork(void);
506 unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
508 CHECK_FULL_REGS(regs);
510 sp -= sizeof(struct pt_regs);
511 childregs = (struct pt_regs *) sp;
513 if ((childregs->msr & MSR_PR) == 0) {
514 /* for kernel thread, set `current' and stackptr in new task */
515 childregs->gpr[1] = sp + sizeof(struct pt_regs);
517 childregs->gpr[2] = (unsigned long) p;
519 clear_tsk_thread_flag(p, TIF_32BIT);
521 p->thread.regs = NULL; /* no user register state */
523 childregs->gpr[1] = usp;
524 p->thread.regs = childregs;
525 if (clone_flags & CLONE_SETTLS) {
527 if (!test_thread_flag(TIF_32BIT))
528 childregs->gpr[13] = childregs->gpr[6];
531 childregs->gpr[2] = childregs->gpr[6];
534 childregs->gpr[3] = 0; /* Result from fork() */
535 sp -= STACK_FRAME_OVERHEAD;
538 * The way this works is that at some point in the future
539 * some task will call _switch to switch to the new task.
540 * That will pop off the stack frame created below and start
541 * the new task running at ret_from_fork. The new task will
542 * do some house keeping and then return from the fork or clone
543 * system call, using the stack frame created above.
545 sp -= sizeof(struct pt_regs);
546 kregs = (struct pt_regs *) sp;
547 sp -= STACK_FRAME_OVERHEAD;
551 if (cpu_has_feature(CPU_FTR_SLB)) {
552 unsigned long sp_vsid = get_kernel_vsid(sp);
553 unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
555 sp_vsid <<= SLB_VSID_SHIFT;
556 sp_vsid |= SLB_VSID_KERNEL | llp;
557 p->thread.ksp_vsid = sp_vsid;
561 * The PPC64 ABI makes use of a TOC to contain function
562 * pointers. The function (ret_from_except) is actually a pointer
563 * to the TOC entry. The first entry is a pointer to the actual
566 kregs->nip = *((unsigned long *)ret_from_fork);
568 kregs->nip = (unsigned long)ret_from_fork;
569 p->thread.last_syscall = -1;
576 * Set up a thread for executing a new program
578 void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
581 unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
587 * If we exec out of a kernel thread then thread.regs will not be
590 if (!current->thread.regs) {
591 struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
592 current->thread.regs = regs - 1;
595 memset(regs->gpr, 0, sizeof(regs->gpr));
605 regs->msr = MSR_USER;
607 if (!test_thread_flag(TIF_32BIT)) {
608 unsigned long entry, toc;
610 /* start is a relocated pointer to the function descriptor for
611 * the elf _start routine. The first entry in the function
612 * descriptor is the entry address of _start and the second
613 * entry is the TOC value we need to use.
615 __get_user(entry, (unsigned long __user *)start);
616 __get_user(toc, (unsigned long __user *)start+1);
618 /* Check whether the e_entry function descriptor entries
619 * need to be relocated before we can use them.
621 if (load_addr != 0) {
627 regs->msr = MSR_USER64;
631 regs->msr = MSR_USER32;
635 discard_lazy_cpu_state();
636 memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
637 current->thread.fpscr.val = 0;
638 #ifdef CONFIG_ALTIVEC
639 memset(current->thread.vr, 0, sizeof(current->thread.vr));
640 memset(¤t->thread.vscr, 0, sizeof(current->thread.vscr));
641 current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
642 current->thread.vrsave = 0;
643 current->thread.used_vr = 0;
644 #endif /* CONFIG_ALTIVEC */
646 memset(current->thread.evr, 0, sizeof(current->thread.evr));
647 current->thread.acc = 0;
648 current->thread.spefscr = 0;
649 current->thread.used_spe = 0;
650 #endif /* CONFIG_SPE */
653 #define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
654 | PR_FP_EXC_RES | PR_FP_EXC_INV)
656 int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
658 struct pt_regs *regs = tsk->thread.regs;
660 /* This is a bit hairy. If we are an SPE enabled processor
661 * (have embedded fp) we store the IEEE exception enable flags in
662 * fpexc_mode. fpexc_mode is also used for setting FP exception
663 * mode (asyn, precise, disabled) for 'Classic' FP. */
664 if (val & PR_FP_EXC_SW_ENABLE) {
666 tsk->thread.fpexc_mode = val &
667 (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
674 /* on a CONFIG_SPE this does not hurt us. The bits that
675 * __pack_fe01 use do not overlap with bits used for
676 * PR_FP_EXC_SW_ENABLE. Additionally, the MSR[FE0,FE1] bits
677 * on CONFIG_SPE implementations are reserved so writing to
678 * them does not change anything */
679 if (val > PR_FP_EXC_PRECISE)
681 tsk->thread.fpexc_mode = __pack_fe01(val);
682 if (regs != NULL && (regs->msr & MSR_FP) != 0)
683 regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
684 | tsk->thread.fpexc_mode;
688 int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
692 if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
694 val = tsk->thread.fpexc_mode;
699 val = __unpack_fe01(tsk->thread.fpexc_mode);
700 return put_user(val, (unsigned int __user *) adr);
703 #define TRUNC_PTR(x) ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
705 int sys_clone(unsigned long clone_flags, unsigned long usp,
706 int __user *parent_tidp, void __user *child_threadptr,
707 int __user *child_tidp, int p6,
708 struct pt_regs *regs)
710 CHECK_FULL_REGS(regs);
712 usp = regs->gpr[1]; /* stack pointer for child */
714 if (test_thread_flag(TIF_32BIT)) {
715 parent_tidp = TRUNC_PTR(parent_tidp);
716 child_tidp = TRUNC_PTR(child_tidp);
719 return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
722 int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
723 unsigned long p4, unsigned long p5, unsigned long p6,
724 struct pt_regs *regs)
726 CHECK_FULL_REGS(regs);
727 return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
730 int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
731 unsigned long p4, unsigned long p5, unsigned long p6,
732 struct pt_regs *regs)
734 CHECK_FULL_REGS(regs);
735 return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
736 regs, 0, NULL, NULL);
739 int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
740 unsigned long a3, unsigned long a4, unsigned long a5,
741 struct pt_regs *regs)
746 filename = getname((char __user *) a0);
747 error = PTR_ERR(filename);
748 if (IS_ERR(filename))
750 flush_fp_to_thread(current);
751 flush_altivec_to_thread(current);
752 flush_spe_to_thread(current);
753 error = do_execve(filename, (char __user * __user *) a1,
754 (char __user * __user *) a2, regs);
757 current->ptrace &= ~PT_DTRACE;
758 task_unlock(current);
765 static int validate_sp(unsigned long sp, struct task_struct *p,
766 unsigned long nbytes)
768 unsigned long stack_page = (unsigned long)task_stack_page(p);
770 if (sp >= stack_page + sizeof(struct thread_struct)
771 && sp <= stack_page + THREAD_SIZE - nbytes)
774 #ifdef CONFIG_IRQSTACKS
775 stack_page = (unsigned long) hardirq_ctx[task_cpu(p)];
776 if (sp >= stack_page + sizeof(struct thread_struct)
777 && sp <= stack_page + THREAD_SIZE - nbytes)
780 stack_page = (unsigned long) softirq_ctx[task_cpu(p)];
781 if (sp >= stack_page + sizeof(struct thread_struct)
782 && sp <= stack_page + THREAD_SIZE - nbytes)
790 #define MIN_STACK_FRAME 112 /* same as STACK_FRAME_OVERHEAD, in fact */
791 #define FRAME_LR_SAVE 2
792 #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
793 #define REGS_MARKER 0x7265677368657265ul
794 #define FRAME_MARKER 12
796 #define MIN_STACK_FRAME 16
797 #define FRAME_LR_SAVE 1
798 #define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
799 #define REGS_MARKER 0x72656773ul
800 #define FRAME_MARKER 2
803 unsigned long get_wchan(struct task_struct *p)
805 unsigned long ip, sp;
808 if (!p || p == current || p->state == TASK_RUNNING)
812 if (!validate_sp(sp, p, MIN_STACK_FRAME))
816 sp = *(unsigned long *)sp;
817 if (!validate_sp(sp, p, MIN_STACK_FRAME))
820 ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
821 if (!in_sched_functions(ip))
824 } while (count++ < 16);
827 EXPORT_SYMBOL(get_wchan);
829 static int kstack_depth_to_print = 64;
831 void show_stack(struct task_struct *tsk, unsigned long *stack)
833 unsigned long sp, ip, lr, newsp;
837 sp = (unsigned long) stack;
842 asm("mr %0,1" : "=r" (sp));
844 sp = tsk->thread.ksp;
848 printk("Call Trace:\n");
850 if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
853 stack = (unsigned long *) sp;
855 ip = stack[FRAME_LR_SAVE];
856 if (!firstframe || ip != lr) {
857 printk("["REG"] ["REG"] ", sp, ip);
858 print_symbol("%s", ip);
860 printk(" (unreliable)");
866 * See if this is an exception frame.
867 * We look for the "regshere" marker in the current frame.
869 if (validate_sp(sp, tsk, INT_FRAME_SIZE)
870 && stack[FRAME_MARKER] == REGS_MARKER) {
871 struct pt_regs *regs = (struct pt_regs *)
872 (sp + STACK_FRAME_OVERHEAD);
873 printk("--- Exception: %lx", regs->trap);
874 print_symbol(" at %s\n", regs->nip);
876 print_symbol(" LR = %s\n", lr);
881 } while (count++ < kstack_depth_to_print);
884 void dump_stack(void)
886 show_stack(current, NULL);
888 EXPORT_SYMBOL(dump_stack);