This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / arch / powerpc / kernel / process.c
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
new file mode 100644 (file)
index 0000000..2dd47d2
--- /dev/null
@@ -0,0 +1,929 @@
+/*
+ *  Derived from "arch/i386/kernel/process.c"
+ *    Copyright (C) 1995  Linus Torvalds
+ *
+ *  Updated and modified by Cort Dougan (cort@cs.nmt.edu) and
+ *  Paul Mackerras (paulus@cs.anu.edu.au)
+ *
+ *  PowerPC version
+ *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
+ *
+ *  This program is free software; you can redistribute it and/or
+ *  modify it under the terms of the GNU General Public License
+ *  as published by the Free Software Foundation; either version
+ *  2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/config.h>
+#include <linux/errno.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/smp_lock.h>
+#include <linux/stddef.h>
+#include <linux/unistd.h>
+#include <linux/ptrace.h>
+#include <linux/slab.h>
+#include <linux/user.h>
+#include <linux/elf.h>
+#include <linux/init.h>
+#include <linux/prctl.h>
+#include <linux/init_task.h>
+#include <linux/module.h>
+#include <linux/kallsyms.h>
+#include <linux/mqueue.h>
+#include <linux/hardirq.h>
+#include <linux/utsname.h>
+
+#include <asm/pgtable.h>
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/io.h>
+#include <asm/processor.h>
+#include <asm/mmu.h>
+#include <asm/prom.h>
+#include <asm/machdep.h>
+#include <asm/time.h>
+#include <asm/syscalls.h>
+#ifdef CONFIG_PPC64
+#include <asm/firmware.h>
+#endif
+
+extern unsigned long _get_SP(void);
+
+#ifndef CONFIG_SMP
+struct task_struct *last_task_used_math = NULL;
+struct task_struct *last_task_used_altivec = NULL;
+struct task_struct *last_task_used_spe = NULL;
+#endif
+
+/*
+ * Make sure the floating-point register state in the
+ * the thread_struct is up to date for task tsk.
+ */
+void flush_fp_to_thread(struct task_struct *tsk)
+{
+       if (tsk->thread.regs) {
+               /*
+                * We need to disable preemption here because if we didn't,
+                * another process could get scheduled after the regs->msr
+                * test but before we have finished saving the FP registers
+                * to the thread_struct.  That process could take over the
+                * FPU, and then when we get scheduled again we would store
+                * bogus values for the remaining FP registers.
+                */
+               preempt_disable();
+               if (tsk->thread.regs->msr & MSR_FP) {
+#ifdef CONFIG_SMP
+                       /*
+                        * This should only ever be called for current or
+                        * for a stopped child process.  Since we save away
+                        * the FP register state on context switch on SMP,
+                        * there is something wrong if a stopped child appears
+                        * to still have its FP state in the CPU registers.
+                        */
+                       BUG_ON(tsk != current);
+#endif
+                       giveup_fpu(current);
+               }
+               preempt_enable();
+       }
+}
+
+void enable_kernel_fp(void)
+{
+       WARN_ON(preemptible());
+
+#ifdef CONFIG_SMP
+       if (current->thread.regs && (current->thread.regs->msr & MSR_FP))
+               giveup_fpu(current);
+       else
+               giveup_fpu(NULL);       /* just enables FP for kernel */
+#else
+       giveup_fpu(last_task_used_math);
+#endif /* CONFIG_SMP */
+}
+EXPORT_SYMBOL(enable_kernel_fp);
+
+int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
+{
+       if (!tsk->thread.regs)
+               return 0;
+       flush_fp_to_thread(current);
+
+       memcpy(fpregs, &tsk->thread.fpr[0], sizeof(*fpregs));
+
+       return 1;
+}
+
+#ifdef CONFIG_ALTIVEC
+void enable_kernel_altivec(void)
+{
+       WARN_ON(preemptible());
+
+#ifdef CONFIG_SMP
+       if (current->thread.regs && (current->thread.regs->msr & MSR_VEC))
+               giveup_altivec(current);
+       else
+               giveup_altivec(NULL);   /* just enable AltiVec for kernel - force */
+#else
+       giveup_altivec(last_task_used_altivec);
+#endif /* CONFIG_SMP */
+}
+EXPORT_SYMBOL(enable_kernel_altivec);
+
+/*
+ * Make sure the VMX/Altivec register state in the
+ * the thread_struct is up to date for task tsk.
+ */
+void flush_altivec_to_thread(struct task_struct *tsk)
+{
+       if (tsk->thread.regs) {
+               preempt_disable();
+               if (tsk->thread.regs->msr & MSR_VEC) {
+#ifdef CONFIG_SMP
+                       BUG_ON(tsk != current);
+#endif
+                       giveup_altivec(current);
+               }
+               preempt_enable();
+       }
+}
+
+int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
+{
+       flush_altivec_to_thread(current);
+       memcpy(vrregs, &current->thread.vr[0], sizeof(*vrregs));
+       return 1;
+}
+#endif /* CONFIG_ALTIVEC */
+
+#ifdef CONFIG_SPE
+
+void enable_kernel_spe(void)
+{
+       WARN_ON(preemptible());
+
+#ifdef CONFIG_SMP
+       if (current->thread.regs && (current->thread.regs->msr & MSR_SPE))
+               giveup_spe(current);
+       else
+               giveup_spe(NULL);       /* just enable SPE for kernel - force */
+#else
+       giveup_spe(last_task_used_spe);
+#endif /* __SMP __ */
+}
+EXPORT_SYMBOL(enable_kernel_spe);
+
+void flush_spe_to_thread(struct task_struct *tsk)
+{
+       if (tsk->thread.regs) {
+               preempt_disable();
+               if (tsk->thread.regs->msr & MSR_SPE) {
+#ifdef CONFIG_SMP
+                       BUG_ON(tsk != current);
+#endif
+                       giveup_spe(current);
+               }
+               preempt_enable();
+       }
+}
+
+int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
+{
+       flush_spe_to_thread(current);
+       /* We copy u32 evr[32] + u64 acc + u32 spefscr -> 35 */
+       memcpy(evrregs, &current->thread.evr[0], sizeof(u32) * 35);
+       return 1;
+}
+#endif /* CONFIG_SPE */
+
+#ifndef CONFIG_SMP
+/*
+ * If we are doing lazy switching of CPU state (FP, altivec or SPE),
+ * and the current task has some state, discard it.
+ */
+void discard_lazy_cpu_state(void)
+{
+       preempt_disable();
+       if (last_task_used_math == current)
+               last_task_used_math = NULL;
+#ifdef CONFIG_ALTIVEC
+       if (last_task_used_altivec == current)
+               last_task_used_altivec = NULL;
+#endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_SPE
+       if (last_task_used_spe == current)
+               last_task_used_spe = NULL;
+#endif
+       preempt_enable();
+}
+#endif /* CONFIG_SMP */
+
+#ifdef CONFIG_PPC_MERGE                /* XXX for now */
+int set_dabr(unsigned long dabr)
+{
+       if (ppc_md.set_dabr)
+               return ppc_md.set_dabr(dabr);
+
+       mtspr(SPRN_DABR, dabr);
+       return 0;
+}
+#endif
+
+#ifdef CONFIG_PPC64
+DEFINE_PER_CPU(struct cpu_usage, cpu_usage_array);
+static DEFINE_PER_CPU(unsigned long, current_dabr);
+#endif
+
+struct task_struct *__switch_to(struct task_struct *prev,
+       struct task_struct *new)
+{
+       struct thread_struct *new_thread, *old_thread;
+       unsigned long flags;
+       struct task_struct *last;
+
+#ifdef CONFIG_SMP
+       /* avoid complexity of lazy save/restore of fpu
+        * by just saving it every time we switch out if
+        * this task used the fpu during the last quantum.
+        *
+        * If it tries to use the fpu again, it'll trap and
+        * reload its fp regs.  So we don't have to do a restore
+        * every switch, just a save.
+        *  -- Cort
+        */
+       if (prev->thread.regs && (prev->thread.regs->msr & MSR_FP))
+               giveup_fpu(prev);
+#ifdef CONFIG_ALTIVEC
+       /*
+        * If the previous thread used altivec in the last quantum
+        * (thus changing altivec regs) then save them.
+        * We used to check the VRSAVE register but not all apps
+        * set it, so we don't rely on it now (and in fact we need
+        * to save & restore VSCR even if VRSAVE == 0).  -- paulus
+        *
+        * On SMP we always save/restore altivec regs just to avoid the
+        * complexity of changing processors.
+        *  -- Cort
+        */
+       if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
+               giveup_altivec(prev);
+#endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_SPE
+       /*
+        * If the previous thread used spe in the last quantum
+        * (thus changing spe regs) then save them.
+        *
+        * On SMP we always save/restore spe regs just to avoid the
+        * complexity of changing processors.
+        */
+       if ((prev->thread.regs && (prev->thread.regs->msr & MSR_SPE)))
+               giveup_spe(prev);
+#endif /* CONFIG_SPE */
+
+#else  /* CONFIG_SMP */
+#ifdef CONFIG_ALTIVEC
+       /* Avoid the trap.  On smp this this never happens since
+        * we don't set last_task_used_altivec -- Cort
+        */
+       if (new->thread.regs && last_task_used_altivec == new)
+               new->thread.regs->msr |= MSR_VEC;
+#endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_SPE
+       /* Avoid the trap.  On smp this this never happens since
+        * we don't set last_task_used_spe
+        */
+       if (new->thread.regs && last_task_used_spe == new)
+               new->thread.regs->msr |= MSR_SPE;
+#endif /* CONFIG_SPE */
+
+#endif /* CONFIG_SMP */
+
+#ifdef CONFIG_PPC64    /* for now */
+       if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
+               set_dabr(new->thread.dabr);
+               __get_cpu_var(current_dabr) = new->thread.dabr;
+       }
+
+       flush_tlb_pending();
+#endif
+
+       new_thread = &new->thread;
+       old_thread = &current->thread;
+
+#ifdef CONFIG_PPC64
+       /*
+        * Collect processor utilization data per process
+        */
+       if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
+               struct cpu_usage *cu = &__get_cpu_var(cpu_usage_array);
+               long unsigned start_tb, current_tb;
+               start_tb = old_thread->start_tb;
+               cu->current_tb = current_tb = mfspr(SPRN_PURR);
+               old_thread->accum_tb += (current_tb - start_tb);
+               new_thread->start_tb = current_tb;
+       }
+#endif
+
+       local_irq_save(flags);
+
+       account_system_vtime(current);
+       account_process_vtime(current);
+       calculate_steal_time();
+
+       last = _switch(old_thread, new_thread);
+
+       local_irq_restore(flags);
+
+       return last;
+}
+
+static int instructions_to_print = 16;
+
+#ifdef CONFIG_PPC64
+#define BAD_PC(pc)     ((REGION_ID(pc) != KERNEL_REGION_ID) && \
+                        (REGION_ID(pc) != VMALLOC_REGION_ID))
+#else
+#define BAD_PC(pc)     ((pc) < KERNELBASE)
+#endif
+
+static void show_instructions(struct pt_regs *regs)
+{
+       int i;
+       unsigned long pc = regs->nip - (instructions_to_print * 3 / 4 *
+                       sizeof(int));
+
+       printk("Instruction dump:");
+
+       for (i = 0; i < instructions_to_print; i++) {
+               int instr;
+
+               if (!(i % 8))
+                       printk("\n");
+
+               /* We use __get_user here *only* to avoid an OOPS on a
+                * bad address because the pc *should* only be a
+                * kernel address.
+                */
+               if (BAD_PC(pc) || __get_user(instr, (unsigned int __user *)pc)) {
+                       printk("XXXXXXXX ");
+               } else {
+                       if (regs->nip == pc)
+                               printk("<%08x> ", instr);
+                       else
+                               printk("%08x ", instr);
+               }
+
+               pc += sizeof(int);
+       }
+
+       printk("\n");
+}
+
+static struct regbit {
+       unsigned long bit;
+       const char *name;
+} msr_bits[] = {
+       {MSR_EE,        "EE"},
+       {MSR_PR,        "PR"},
+       {MSR_FP,        "FP"},
+       {MSR_ME,        "ME"},
+       {MSR_IR,        "IR"},
+       {MSR_DR,        "DR"},
+       {0,             NULL}
+};
+
+static void printbits(unsigned long val, struct regbit *bits)
+{
+       const char *sep = "";
+
+       printk("<");
+       for (; bits->bit; ++bits)
+               if (val & bits->bit) {
+                       printk("%s%s", sep, bits->name);
+                       sep = ",";
+               }
+       printk(">");
+}
+
+#ifdef CONFIG_PPC64
+#define REG            "%016lX"
+#define REGS_PER_LINE  4
+#define LAST_VOLATILE  13
+#else
+#define REG            "%08lX"
+#define REGS_PER_LINE  8
+#define LAST_VOLATILE  12
+#endif
+
+void show_regs(struct pt_regs * regs)
+{
+       int i, trap;
+
+       printk("NIP: "REG" LR: "REG" CTR: "REG"\n",
+              regs->nip, regs->link, regs->ctr);
+       printk("REGS: %p TRAP: %04lx   %s  (%s)\n",
+              regs, regs->trap, print_tainted(), system_utsname.release);
+       printk("MSR: "REG" ", regs->msr);
+       printbits(regs->msr, msr_bits);
+       printk("  CR: %08lX  XER: %08lX\n", regs->ccr, regs->xer);
+       trap = TRAP(regs);
+       if (trap == 0x300 || trap == 0x600)
+               printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
+       printk("TASK = %p[%d] '%s' THREAD: %p",
+              current, current->pid, current->comm, task_thread_info(current));
+
+#ifdef CONFIG_SMP
+       printk(" CPU: %d", smp_processor_id());
+#endif /* CONFIG_SMP */
+
+       for (i = 0;  i < 32;  i++) {
+               if ((i % REGS_PER_LINE) == 0)
+                       printk("\n" KERN_INFO "GPR%02d: ", i);
+               printk(REG " ", regs->gpr[i]);
+               if (i == LAST_VOLATILE && !FULL_REGS(regs))
+                       break;
+       }
+       printk("\n");
+#ifdef CONFIG_KALLSYMS
+       /*
+        * Lookup NIP late so we have the best change of getting the
+        * above info out without failing
+        */
+       printk("NIP ["REG"] ", regs->nip);
+       print_symbol("%s\n", regs->nip);
+       printk("LR ["REG"] ", regs->link);
+       print_symbol("%s\n", regs->link);
+#endif
+       show_stack(current, (unsigned long *) regs->gpr[1]);
+       if (!user_mode(regs))
+               show_instructions(regs);
+}
+
+void exit_thread(void)
+{
+       discard_lazy_cpu_state();
+}
+
+void flush_thread(void)
+{
+#ifdef CONFIG_PPC64
+       struct thread_info *t = current_thread_info();
+
+       if (t->flags & _TIF_ABI_PENDING)
+               t->flags ^= (_TIF_ABI_PENDING | _TIF_32BIT);
+#endif
+
+       discard_lazy_cpu_state();
+
+#ifdef CONFIG_PPC64    /* for now */
+       if (current->thread.dabr) {
+               current->thread.dabr = 0;
+               set_dabr(0);
+       }
+#endif
+}
+
+void
+release_thread(struct task_struct *t)
+{
+}
+
+/*
+ * This gets called before we allocate a new thread and copy
+ * the current task into it.
+ */
+void prepare_to_copy(struct task_struct *tsk)
+{
+       flush_fp_to_thread(current);
+       flush_altivec_to_thread(current);
+       flush_spe_to_thread(current);
+}
+
+/*
+ * Copy a thread..
+ */
+int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
+               unsigned long unused, struct task_struct *p,
+               struct pt_regs *regs)
+{
+       struct pt_regs *childregs, *kregs;
+       extern void ret_from_fork(void);
+       unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
+
+       CHECK_FULL_REGS(regs);
+       /* Copy registers */
+       sp -= sizeof(struct pt_regs);
+       childregs = (struct pt_regs *) sp;
+       *childregs = *regs;
+       if ((childregs->msr & MSR_PR) == 0) {
+               /* for kernel thread, set `current' and stackptr in new task */
+               childregs->gpr[1] = sp + sizeof(struct pt_regs);
+#ifdef CONFIG_PPC32
+               childregs->gpr[2] = (unsigned long) p;
+#else
+               clear_tsk_thread_flag(p, TIF_32BIT);
+#endif
+               p->thread.regs = NULL;  /* no user register state */
+       } else {
+               childregs->gpr[1] = usp;
+               p->thread.regs = childregs;
+               if (clone_flags & CLONE_SETTLS) {
+#ifdef CONFIG_PPC64
+                       if (!test_thread_flag(TIF_32BIT))
+                               childregs->gpr[13] = childregs->gpr[6];
+                       else
+#endif
+                               childregs->gpr[2] = childregs->gpr[6];
+               }
+       }
+       childregs->gpr[3] = 0;  /* Result from fork() */
+       sp -= STACK_FRAME_OVERHEAD;
+
+       /*
+        * The way this works is that at some point in the future
+        * some task will call _switch to switch to the new task.
+        * That will pop off the stack frame created below and start
+        * the new task running at ret_from_fork.  The new task will
+        * do some house keeping and then return from the fork or clone
+        * system call, using the stack frame created above.
+        */
+       sp -= sizeof(struct pt_regs);
+       kregs = (struct pt_regs *) sp;
+       sp -= STACK_FRAME_OVERHEAD;
+       p->thread.ksp = sp;
+
+#ifdef CONFIG_PPC64
+       if (cpu_has_feature(CPU_FTR_SLB)) {
+               unsigned long sp_vsid = get_kernel_vsid(sp);
+               unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
+
+               sp_vsid <<= SLB_VSID_SHIFT;
+               sp_vsid |= SLB_VSID_KERNEL | llp;
+               p->thread.ksp_vsid = sp_vsid;
+       }
+
+       /*
+        * The PPC64 ABI makes use of a TOC to contain function 
+        * pointers.  The function (ret_from_except) is actually a pointer
+        * to the TOC entry.  The first entry is a pointer to the actual
+        * function.
+        */
+       kregs->nip = *((unsigned long *)ret_from_fork);
+#else
+       kregs->nip = (unsigned long)ret_from_fork;
+       p->thread.last_syscall = -1;
+#endif
+
+       return 0;
+}
+
+/*
+ * Set up a thread for executing a new program
+ */
+void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
+{
+#ifdef CONFIG_PPC64
+       unsigned long load_addr = regs->gpr[2]; /* saved by ELF_PLAT_INIT */
+#endif
+
+       set_fs(USER_DS);
+
+       /*
+        * If we exec out of a kernel thread then thread.regs will not be
+        * set.  Do it now.
+        */
+       if (!current->thread.regs) {
+               struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
+               current->thread.regs = regs - 1;
+       }
+
+       memset(regs->gpr, 0, sizeof(regs->gpr));
+       regs->ctr = 0;
+       regs->link = 0;
+       regs->xer = 0;
+       regs->ccr = 0;
+       regs->gpr[1] = sp;
+
+#ifdef CONFIG_PPC32
+       regs->mq = 0;
+       regs->nip = start;
+       regs->msr = MSR_USER;
+#else
+       if (!test_thread_flag(TIF_32BIT)) {
+               unsigned long entry, toc;
+
+               /* start is a relocated pointer to the function descriptor for
+                * the elf _start routine.  The first entry in the function
+                * descriptor is the entry address of _start and the second
+                * entry is the TOC value we need to use.
+                */
+               __get_user(entry, (unsigned long __user *)start);
+               __get_user(toc, (unsigned long __user *)start+1);
+
+               /* Check whether the e_entry function descriptor entries
+                * need to be relocated before we can use them.
+                */
+               if (load_addr != 0) {
+                       entry += load_addr;
+                       toc   += load_addr;
+               }
+               regs->nip = entry;
+               regs->gpr[2] = toc;
+               regs->msr = MSR_USER64;
+       } else {
+               regs->nip = start;
+               regs->gpr[2] = 0;
+               regs->msr = MSR_USER32;
+       }
+#endif
+
+       discard_lazy_cpu_state();
+       memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
+       current->thread.fpscr.val = 0;
+#ifdef CONFIG_ALTIVEC
+       memset(current->thread.vr, 0, sizeof(current->thread.vr));
+       memset(&current->thread.vscr, 0, sizeof(current->thread.vscr));
+       current->thread.vscr.u[3] = 0x00010000; /* Java mode disabled */
+       current->thread.vrsave = 0;
+       current->thread.used_vr = 0;
+#endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_SPE
+       memset(current->thread.evr, 0, sizeof(current->thread.evr));
+       current->thread.acc = 0;
+       current->thread.spefscr = 0;
+       current->thread.used_spe = 0;
+#endif /* CONFIG_SPE */
+}
+
+#define PR_FP_ALL_EXCEPT (PR_FP_EXC_DIV | PR_FP_EXC_OVF | PR_FP_EXC_UND \
+               | PR_FP_EXC_RES | PR_FP_EXC_INV)
+
+int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
+{
+       struct pt_regs *regs = tsk->thread.regs;
+
+       /* This is a bit hairy.  If we are an SPE enabled  processor
+        * (have embedded fp) we store the IEEE exception enable flags in
+        * fpexc_mode.  fpexc_mode is also used for setting FP exception
+        * mode (asyn, precise, disabled) for 'Classic' FP. */
+       if (val & PR_FP_EXC_SW_ENABLE) {
+#ifdef CONFIG_SPE
+               tsk->thread.fpexc_mode = val &
+                       (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
+               return 0;
+#else
+               return -EINVAL;
+#endif
+       }
+
+       /* on a CONFIG_SPE this does not hurt us.  The bits that
+        * __pack_fe01 use do not overlap with bits used for
+        * PR_FP_EXC_SW_ENABLE.  Additionally, the MSR[FE0,FE1] bits
+        * on CONFIG_SPE implementations are reserved so writing to
+        * them does not change anything */
+       if (val > PR_FP_EXC_PRECISE)
+               return -EINVAL;
+       tsk->thread.fpexc_mode = __pack_fe01(val);
+       if (regs != NULL && (regs->msr & MSR_FP) != 0)
+               regs->msr = (regs->msr & ~(MSR_FE0|MSR_FE1))
+                       | tsk->thread.fpexc_mode;
+       return 0;
+}
+
+int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
+{
+       unsigned int val;
+
+       if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
+#ifdef CONFIG_SPE
+               val = tsk->thread.fpexc_mode;
+#else
+               return -EINVAL;
+#endif
+       else
+               val = __unpack_fe01(tsk->thread.fpexc_mode);
+       return put_user(val, (unsigned int __user *) adr);
+}
+
+#define TRUNC_PTR(x)   ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
+
+int sys_clone(unsigned long clone_flags, unsigned long usp,
+             int __user *parent_tidp, void __user *child_threadptr,
+             int __user *child_tidp, int p6,
+             struct pt_regs *regs)
+{
+       CHECK_FULL_REGS(regs);
+       if (usp == 0)
+               usp = regs->gpr[1];     /* stack pointer for child */
+#ifdef CONFIG_PPC64
+       if (test_thread_flag(TIF_32BIT)) {
+               parent_tidp = TRUNC_PTR(parent_tidp);
+               child_tidp = TRUNC_PTR(child_tidp);
+       }
+#endif
+       return do_fork(clone_flags, usp, regs, 0, parent_tidp, child_tidp);
+}
+
+int sys_fork(unsigned long p1, unsigned long p2, unsigned long p3,
+            unsigned long p4, unsigned long p5, unsigned long p6,
+            struct pt_regs *regs)
+{
+       CHECK_FULL_REGS(regs);
+       return do_fork(SIGCHLD, regs->gpr[1], regs, 0, NULL, NULL);
+}
+
+int sys_vfork(unsigned long p1, unsigned long p2, unsigned long p3,
+             unsigned long p4, unsigned long p5, unsigned long p6,
+             struct pt_regs *regs)
+{
+       CHECK_FULL_REGS(regs);
+       return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->gpr[1],
+                       regs, 0, NULL, NULL);
+}
+
+int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
+              unsigned long a3, unsigned long a4, unsigned long a5,
+              struct pt_regs *regs)
+{
+       int error;
+       char *filename;
+
+       filename = getname((char __user *) a0);
+       error = PTR_ERR(filename);
+       if (IS_ERR(filename))
+               goto out;
+       flush_fp_to_thread(current);
+       flush_altivec_to_thread(current);
+       flush_spe_to_thread(current);
+       error = do_execve(filename, (char __user * __user *) a1,
+                         (char __user * __user *) a2, regs);
+       if (error == 0) {
+               task_lock(current);
+               current->ptrace &= ~PT_DTRACE;
+               task_unlock(current);
+       }
+       putname(filename);
+out:
+       return error;
+}
+
+int validate_sp(unsigned long sp, struct task_struct *p,
+                      unsigned long nbytes)
+{
+       unsigned long stack_page = (unsigned long)task_stack_page(p);
+
+       if (sp >= stack_page + sizeof(struct thread_struct)
+           && sp <= stack_page + THREAD_SIZE - nbytes)
+               return 1;
+
+#ifdef CONFIG_IRQSTACKS
+       stack_page = (unsigned long) hardirq_ctx[task_cpu(p)];
+       if (sp >= stack_page + sizeof(struct thread_struct)
+           && sp <= stack_page + THREAD_SIZE - nbytes)
+               return 1;
+
+       stack_page = (unsigned long) softirq_ctx[task_cpu(p)];
+       if (sp >= stack_page + sizeof(struct thread_struct)
+           && sp <= stack_page + THREAD_SIZE - nbytes)
+               return 1;
+#endif
+
+       return 0;
+}
+
+#ifdef CONFIG_PPC64
+#define MIN_STACK_FRAME        112     /* same as STACK_FRAME_OVERHEAD, in fact */
+#define FRAME_LR_SAVE  2
+#define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD + 288)
+#define REGS_MARKER    0x7265677368657265ul
+#define FRAME_MARKER   12
+#else
+#define MIN_STACK_FRAME        16
+#define FRAME_LR_SAVE  1
+#define INT_FRAME_SIZE (sizeof(struct pt_regs) + STACK_FRAME_OVERHEAD)
+#define REGS_MARKER    0x72656773ul
+#define FRAME_MARKER   2
+#endif
+
+EXPORT_SYMBOL(validate_sp);
+
+unsigned long get_wchan(struct task_struct *p)
+{
+       unsigned long ip, sp;
+       int count = 0;
+
+       if (!p || p == current || p->state == TASK_RUNNING)
+               return 0;
+
+       sp = p->thread.ksp;
+       if (!validate_sp(sp, p, MIN_STACK_FRAME))
+               return 0;
+
+       do {
+               sp = *(unsigned long *)sp;
+               if (!validate_sp(sp, p, MIN_STACK_FRAME))
+                       return 0;
+               if (count > 0) {
+                       ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
+                       if (!in_sched_functions(ip))
+                               return ip;
+               }
+       } while (count++ < 16);
+       return 0;
+}
+
+static int kstack_depth_to_print = 64;
+
+void show_stack(struct task_struct *tsk, unsigned long *stack)
+{
+       unsigned long sp, ip, lr, newsp;
+       int count = 0;
+       int firstframe = 1;
+
+       sp = (unsigned long) stack;
+       if (tsk == NULL)
+               tsk = current;
+       if (sp == 0) {
+               if (tsk == current)
+                       asm("mr %0,1" : "=r" (sp));
+               else
+                       sp = tsk->thread.ksp;
+       }
+
+       lr = 0;
+       printk("Call Trace:\n");
+       do {
+               if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
+                       return;
+
+               stack = (unsigned long *) sp;
+               newsp = stack[0];
+               ip = stack[FRAME_LR_SAVE];
+               if (!firstframe || ip != lr) {
+                       printk("["REG"] ["REG"] ", sp, ip);
+                       print_symbol("%s", ip);
+                       if (firstframe)
+                               printk(" (unreliable)");
+                       printk("\n");
+               }
+               firstframe = 0;
+
+               /*
+                * See if this is an exception frame.
+                * We look for the "regshere" marker in the current frame.
+                */
+               if (validate_sp(sp, tsk, INT_FRAME_SIZE)
+                   && stack[FRAME_MARKER] == REGS_MARKER) {
+                       struct pt_regs *regs = (struct pt_regs *)
+                               (sp + STACK_FRAME_OVERHEAD);
+                       printk("--- Exception: %lx", regs->trap);
+                       print_symbol(" at %s\n", regs->nip);
+                       lr = regs->link;
+                       print_symbol("    LR = %s\n", lr);
+                       firstframe = 1;
+               }
+
+               sp = newsp;
+       } while (count++ < kstack_depth_to_print);
+}
+
+void dump_stack(void)
+{
+       show_stack(current, NULL);
+}
+EXPORT_SYMBOL(dump_stack);
+
+#ifdef CONFIG_PPC64
+void ppc64_runlatch_on(void)
+{
+       unsigned long ctrl;
+
+       if (cpu_has_feature(CPU_FTR_CTRL) && !test_thread_flag(TIF_RUNLATCH)) {
+               HMT_medium();
+
+               ctrl = mfspr(SPRN_CTRLF);
+               ctrl |= CTRL_RUNLATCH;
+               mtspr(SPRN_CTRLT, ctrl);
+
+               set_thread_flag(TIF_RUNLATCH);
+       }
+}
+
+void ppc64_runlatch_off(void)
+{
+       unsigned long ctrl;
+
+       if (cpu_has_feature(CPU_FTR_CTRL) && test_thread_flag(TIF_RUNLATCH)) {
+               HMT_medium();
+
+               clear_thread_flag(TIF_RUNLATCH);
+
+               ctrl = mfspr(SPRN_CTRLF);
+               ctrl &= ~CTRL_RUNLATCH;
+               mtspr(SPRN_CTRLT, ctrl);
+       }
+}
+#endif