powerpc64: convert to dynamic percpu allocator
[safe/jmp/linux-2.6] / arch / powerpc / kernel / process.c
index 047da1a..892a9f2 100644 (file)
@@ -1,6 +1,4 @@
 /*
- *  arch/ppc/kernel/process.c
- *
  *  Derived from "arch/i386/kernel/process.c"
  *    Copyright (C) 1995  Linus Torvalds
  *
  *  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/mqueue.h>
 #include <linux/hardirq.h>
 #include <linux/utsname.h>
-#include <linux/kprobes.h>
+#include <linux/ftrace.h>
+#include <linux/kernel_stat.h>
+#include <linux/personality.h>
+#include <linux/random.h>
 
 #include <asm/pgtable.h>
 #include <asm/uaccess.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>
-#include <asm/plpar_wrappers.h>
-#include <asm/time.h>
 #endif
+#include <linux/kprobes.h>
+#include <linux/kdebug.h>
 
 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_vsx = NULL;
 struct task_struct *last_task_used_spe = NULL;
 #endif
 
@@ -87,7 +90,7 @@ void flush_fp_to_thread(struct task_struct *tsk)
                         */
                        BUG_ON(tsk != current);
 #endif
-                       giveup_fpu(current);
+                       giveup_fpu(tsk);
                }
                preempt_enable();
        }
@@ -108,17 +111,6 @@ void enable_kernel_fp(void)
 }
 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)
 {
@@ -147,19 +139,53 @@ void flush_altivec_to_thread(struct task_struct *tsk)
 #ifdef CONFIG_SMP
                        BUG_ON(tsk != current);
 #endif
-                       giveup_altivec(current);
+                       giveup_altivec(tsk);
                }
                preempt_enable();
        }
 }
+#endif /* CONFIG_ALTIVEC */
 
-int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
+#ifdef CONFIG_VSX
+#if 0
+/* not currently used, but some crazy RAID module might want to later */
+void enable_kernel_vsx(void)
 {
-       flush_altivec_to_thread(current);
-       memcpy(vrregs, &current->thread.vr[0], sizeof(*vrregs));
-       return 1;
+       WARN_ON(preemptible());
+
+#ifdef CONFIG_SMP
+       if (current->thread.regs && (current->thread.regs->msr & MSR_VSX))
+               giveup_vsx(current);
+       else
+               giveup_vsx(NULL);       /* just enable vsx for kernel - force */
+#else
+       giveup_vsx(last_task_used_vsx);
+#endif /* CONFIG_SMP */
+}
+EXPORT_SYMBOL(enable_kernel_vsx);
+#endif
+
+void giveup_vsx(struct task_struct *tsk)
+{
+       giveup_fpu(tsk);
+       giveup_altivec(tsk);
+       __giveup_vsx(tsk);
 }
-#endif /* CONFIG_ALTIVEC */
+
+void flush_vsx_to_thread(struct task_struct *tsk)
+{
+       if (tsk->thread.regs) {
+               preempt_disable();
+               if (tsk->thread.regs->msr & MSR_VSX) {
+#ifdef CONFIG_SMP
+                       BUG_ON(tsk != current);
+#endif
+                       giveup_vsx(tsk);
+               }
+               preempt_enable();
+       }
+}
+#endif /* CONFIG_VSX */
 
 #ifdef CONFIG_SPE
 
@@ -186,47 +212,91 @@ void flush_spe_to_thread(struct task_struct *tsk)
 #ifdef CONFIG_SMP
                        BUG_ON(tsk != current);
 #endif
-                       giveup_spe(current);
+                       giveup_spe(tsk);
                }
                preempt_enable();
        }
 }
+#endif /* CONFIG_SPE */
 
-int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
+#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)
 {
-       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;
+       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_VSX
+       if (last_task_used_vsx == current)
+               last_task_used_vsx = NULL;
+#endif /* CONFIG_VSX */
+#ifdef CONFIG_SPE
+       if (last_task_used_spe == current)
+               last_task_used_spe = NULL;
+#endif
+       preempt_enable();
 }
-#endif /* CONFIG_SPE */
+#endif /* CONFIG_SMP */
 
-static void set_dabr_spr(unsigned long val)
+void do_dabr(struct pt_regs *regs, unsigned long address,
+                   unsigned long error_code)
 {
-       mtspr(SPRN_DABR, val);
+       siginfo_t info;
+
+       if (notify_die(DIE_DABR_MATCH, "dabr_match", regs, error_code,
+                       11, SIGSEGV) == NOTIFY_STOP)
+               return;
+
+       if (debugger_dabr_match(regs))
+               return;
+
+       /* Clear the DAC and struct entries.  One shot trigger */
+#if defined(CONFIG_BOOKE)
+       mtspr(SPRN_DBCR0, mfspr(SPRN_DBCR0) & ~(DBSR_DAC1R | DBSR_DAC1W
+                                                       | DBCR0_IDM));
+#endif
+
+       /* Clear the DABR */
+       set_dabr(0);
+
+       /* Deliver the signal to userspace */
+       info.si_signo = SIGTRAP;
+       info.si_errno = 0;
+       info.si_code = TRAP_HWBKPT;
+       info.si_addr = (void __user *)address;
+       force_sig_info(SIGTRAP, &info, current);
 }
 
+static DEFINE_PER_CPU(unsigned long, current_dabr);
+
 int set_dabr(unsigned long dabr)
 {
-       int ret = 0;
+       __get_cpu_var(current_dabr) = dabr;
 
-#ifdef CONFIG_PPC64
-       if (firmware_has_feature(FW_FEATURE_XDABR)) {
-               /* We want to catch accesses from kernel and userspace */
-               unsigned long flags = H_DABRX_KERNEL|H_DABRX_USER;
-               ret = plpar_set_xdabr(dabr, flags);
-       } else if (firmware_has_feature(FW_FEATURE_DABR)) {
-               ret = plpar_set_dabr(dabr);
-       } else
+       if (ppc_md.set_dabr)
+               return ppc_md.set_dabr(dabr);
+
+       /* XXX should we have a CPU_FTR_HAS_DABR ? */
+#if defined(CONFIG_PPC64) || defined(CONFIG_6xx)
+       mtspr(SPRN_DABR, dabr);
 #endif
-               set_dabr_spr(dabr);
 
-       return ret;
+#if defined(CONFIG_BOOKE)
+       mtspr(SPRN_DAC1, dabr);
+#endif
+
+       return 0;
 }
 
 #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,
@@ -263,6 +333,11 @@ struct task_struct *__switch_to(struct task_struct *prev,
        if (prev->thread.regs && (prev->thread.regs->msr & MSR_VEC))
                giveup_altivec(prev);
 #endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_VSX
+       if (prev->thread.regs && (prev->thread.regs->msr & MSR_VSX))
+               /* VMX and FPU registers are already save here */
+               __giveup_vsx(prev);
+#endif /* CONFIG_VSX */
 #ifdef CONFIG_SPE
        /*
         * If the previous thread used spe in the last quantum
@@ -283,6 +358,10 @@ struct task_struct *__switch_to(struct task_struct *prev,
        if (new->thread.regs && last_task_used_altivec == new)
                new->thread.regs->msr |= MSR_VEC;
 #endif /* CONFIG_ALTIVEC */
+#ifdef CONFIG_VSX
+       if (new->thread.regs && last_task_used_vsx == new)
+               new->thread.regs->msr |= MSR_VSX;
+#endif /* CONFIG_VSX */
 #ifdef CONFIG_SPE
        /* Avoid the trap.  On smp this this never happens since
         * we don't set last_task_used_spe
@@ -293,13 +372,13 @@ struct task_struct *__switch_to(struct task_struct *prev,
 
 #endif /* CONFIG_SMP */
 
-#ifdef CONFIG_PPC64    /* for now */
-       if (unlikely(__get_cpu_var(current_dabr) != new->thread.dabr)) {
+       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();
+#if defined(CONFIG_BOOKE)
+       /* If new thread DAC (HW breakpoint) is the same then leave it */
+       if (new->thread.dabr)
+               set_dabr(new->thread.dabr);
 #endif
 
        new_thread = &new->thread;
@@ -320,6 +399,17 @@ struct task_struct *__switch_to(struct task_struct *prev,
 #endif
 
        local_irq_save(flags);
+
+       account_system_vtime(current);
+       account_process_vtime(current);
+       calculate_steal_time();
+
+       /*
+        * We can't take a PMU exception inside _switch() since there is a
+        * window where the kernel stack SLB and the kernel stack are out
+        * of sync. Hard disable here.
+        */
+       hard_irq_disable();
        last = _switch(old_thread, new_thread);
 
        local_irq_restore(flags);
@@ -329,13 +419,6 @@ struct task_struct *__switch_to(struct task_struct *prev,
 
 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;
@@ -350,7 +433,20 @@ static void show_instructions(struct pt_regs *regs)
                if (!(i % 8))
                        printk("\n");
 
-               if (BAD_PC(pc) || __get_user(instr, (unsigned int *)pc)) {
+#if !defined(CONFIG_BOOKE)
+               /* If executing with the IMMU off, adjust pc rather
+                * than print XXXXXXXX.
+                */
+               if (!(regs->msr & MSR_IR))
+                       pc = (unsigned long)phys_to_virt(pc);
+#endif
+
+               /* We use __get_user here *only* to avoid an OOPS on a
+                * bad address because the pc *should* only be a
+                * kernel address.
+                */
+               if (!__kernel_text_address(pc) ||
+                    __get_user(instr, (unsigned int __user *)pc)) {
                        printk("XXXXXXXX ");
                } else {
                        if (regs->nip == pc)
@@ -372,7 +468,11 @@ static struct regbit {
        {MSR_EE,        "EE"},
        {MSR_PR,        "PR"},
        {MSR_FP,        "FP"},
+       {MSR_VEC,       "VEC"},
+       {MSR_VSX,       "VSX"},
        {MSR_ME,        "ME"},
+       {MSR_CE,        "CE"},
+       {MSR_DE,        "DE"},
        {MSR_IR,        "IR"},
        {MSR_DR,        "DR"},
        {0,             NULL}
@@ -392,11 +492,11 @@ static void printbits(unsigned long val, struct regbit *bits)
 }
 
 #ifdef CONFIG_PPC64
-#define REG            "%016lX"
+#define REG            "%016lx"
 #define REGS_PER_LINE  4
 #define LAST_VOLATILE  13
 #else
-#define REG            "%08lX"
+#define REG            "%08lx"
 #define REGS_PER_LINE  8
 #define LAST_VOLATILE  12
 #endif
@@ -408,23 +508,27 @@ void show_regs(struct pt_regs * regs)
        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);
+              regs, regs->trap, print_tainted(), init_utsname()->release);
        printk("MSR: "REG" ", regs->msr);
        printbits(regs->msr, msr_bits);
-       printk("  CR: %08lX  XER: %08lX\n", regs->ccr, regs->xer);
+       printk("  CR: %08lx  XER: %08lx\n", regs->ccr, regs->xer);
        trap = TRAP(regs);
        if (trap == 0x300 || trap == 0x600)
+#if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)
+               printk("DEAR: "REG", ESR: "REG"\n", regs->dar, regs->dsisr);
+#else
                printk("DAR: "REG", DSISR: "REG"\n", regs->dar, regs->dsisr);
+#endif
        printk("TASK = %p[%d] '%s' THREAD: %p",
-              current, current->pid, current->comm, current->thread_info);
+              current, task_pid_nr(current), current->comm, task_thread_info(current));
 
 #ifdef CONFIG_SMP
-       printk(" CPU: %d", smp_processor_id());
+       printk(" CPU: %d", raw_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("\nGPR%02d: ", i);
                printk(REG " ", regs->gpr[i]);
                if (i == LAST_VOLATILE && !FULL_REGS(regs))
                        break;
@@ -435,10 +539,8 @@ void show_regs(struct pt_regs * regs)
         * 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);
+       printk("NIP ["REG"] %pS\n", regs->nip, (void *)regs->nip);
+       printk("LR ["REG"] %pS\n", regs->link, (void *)regs->link);
 #endif
        show_stack(current, (unsigned long *) regs->gpr[1]);
        if (!user_mode(regs))
@@ -447,20 +549,7 @@ void show_regs(struct pt_regs * regs)
 
 void exit_thread(void)
 {
-       kprobe_flush_task(current);
-
-#ifndef CONFIG_SMP
-       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
-#endif /* CONFIG_SMP */
+       discard_lazy_cpu_state();
 }
 
 void flush_thread(void)
@@ -468,30 +557,25 @@ 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);
+       if (test_ti_thread_flag(t, TIF_ABI_PENDING)) {
+               clear_ti_thread_flag(t, TIF_ABI_PENDING);
+               if (test_ti_thread_flag(t, TIF_32BIT))
+                       clear_ti_thread_flag(t, TIF_32BIT);
+               else
+                       set_ti_thread_flag(t, TIF_32BIT);
+       }
 #endif
-       kprobe_flush_task(current);
 
-#ifndef CONFIG_SMP
-       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
-#endif /* CONFIG_SMP */
+       discard_lazy_cpu_state();
 
-#ifdef CONFIG_PPC64    /* for now */
        if (current->thread.dabr) {
                current->thread.dabr = 0;
                set_dabr(0);
-       }
+
+#if defined(CONFIG_BOOKE)
+               current->thread.dbcr0 &= ~(DBSR_DAC1R | DBSR_DAC1W);
 #endif
+       }
 }
 
 void
@@ -507,19 +591,20 @@ void prepare_to_copy(struct task_struct *tsk)
 {
        flush_fp_to_thread(current);
        flush_altivec_to_thread(current);
+       flush_vsx_to_thread(current);
        flush_spe_to_thread(current);
 }
 
 /*
  * Copy a thread..
  */
-int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
+int copy_thread(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)p->thread_info + THREAD_SIZE;
+       unsigned long sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
 
        CHECK_FULL_REGS(regs);
        /* Copy registers */
@@ -532,7 +617,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 #ifdef CONFIG_PPC32
                childregs->gpr[2] = (unsigned long) p;
 #else
-               clear_ti_thread_flag(p->thread_info, TIF_32BIT);
+               clear_tsk_thread_flag(p, TIF_32BIT);
 #endif
                p->thread.regs = NULL;  /* no user register state */
        } else {
@@ -562,16 +647,21 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
        kregs = (struct pt_regs *) sp;
        sp -= STACK_FRAME_OVERHEAD;
        p->thread.ksp = sp;
+       p->thread.ksp_limit = (unsigned long)task_stack_page(p) +
+                               _ALIGN_UP(sizeof(struct thread_info), 16);
 
-#ifdef CONFIG_PPC64
+#ifdef CONFIG_PPC_STD_MMU_64
        if (cpu_has_feature(CPU_FTR_SLB)) {
-               unsigned long sp_vsid = get_kernel_vsid(sp);
-
-               sp_vsid <<= SLB_VSID_SHIFT;
-               sp_vsid |= SLB_VSID_KERNEL;
-               if (cpu_has_feature(CPU_FTR_16M_PAGE))
-                       sp_vsid |= SLB_VSID_L;
+               unsigned long sp_vsid;
+               unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
 
+               if (cpu_has_feature(CPU_FTR_1T_SEGMENT))
+                       sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_1T)
+                               << SLB_VSID_SHIFT_1T;
+               else
+                       sp_vsid = get_kernel_vsid(sp, MMU_SEGSIZE_256M)
+                               << SLB_VSID_SHIFT;
+               sp_vsid |= SLB_VSID_KERNEL | llp;
                p->thread.ksp_vsid = sp_vsid;
        }
 
@@ -584,7 +674,6 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
        kregs->nip = *((unsigned long *)ret_from_fork);
 #else
        kregs->nip = (unsigned long)ret_from_fork;
-       p->thread.last_syscall = -1;
 #endif
 
        return 0;
@@ -606,10 +695,8 @@ void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
         * set.  Do it now.
         */
        if (!current->thread.regs) {
-               unsigned long childregs = (unsigned long)current->thread_info +
-                                               THREAD_SIZE;
-               childregs -= sizeof(struct pt_regs);
-               current->thread.regs = (struct pt_regs *)childregs;
+               struct pt_regs *regs = task_stack_page(current) + THREAD_SIZE;
+               current->thread.regs = regs - 1;
        }
 
        memset(regs->gpr, 0, sizeof(regs->gpr));
@@ -619,6 +706,13 @@ void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
        regs->ccr = 0;
        regs->gpr[1] = sp;
 
+       /*
+        * We have just cleared all the nonvolatile GPRs, so make
+        * FULL_REGS(regs) return true.  This is necessary to allow
+        * ptrace to examine the thread immediately after exec.
+        */
+       regs->trap &= ~1UL;
+
 #ifdef CONFIG_PPC32
        regs->mq = 0;
        regs->nip = start;
@@ -652,20 +746,12 @@ void start_thread(struct pt_regs *regs, unsigned long start, unsigned long sp)
        }
 #endif
 
-#ifndef CONFIG_SMP
-       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
-#ifdef CONFIG_SPE
-       if (last_task_used_spe == current)
-               last_task_used_spe = NULL;
+       discard_lazy_cpu_state();
+#ifdef CONFIG_VSX
+       current->thread.used_vsr = 0;
 #endif
-#endif /* CONFIG_SMP */
        memset(current->thread.fpr, 0, sizeof(current->thread.fpr));
-       current->thread.fpscr = 0;
+       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));
@@ -694,9 +780,13 @@ int set_fpexc_mode(struct task_struct *tsk, unsigned int val)
         * 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;
+               if (cpu_has_feature(CPU_FTR_SPE)) {
+                       tsk->thread.fpexc_mode = val &
+                               (PR_FP_EXC_SW_ENABLE | PR_FP_ALL_EXCEPT);
+                       return 0;
+               } else {
+                       return -EINVAL;
+               }
 #else
                return -EINVAL;
 #endif
@@ -722,7 +812,10 @@ int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
 
        if (tsk->thread.fpexc_mode & PR_FP_EXC_SW_ENABLE)
 #ifdef CONFIG_SPE
-               val = tsk->thread.fpexc_mode;
+               if (cpu_has_feature(CPU_FTR_SPE))
+                       val = tsk->thread.fpexc_mode;
+               else
+                       return -EINVAL;
 #else
                return -EINVAL;
 #endif
@@ -731,6 +824,61 @@ int get_fpexc_mode(struct task_struct *tsk, unsigned long adr)
        return put_user(val, (unsigned int __user *) adr);
 }
 
+int set_endian(struct task_struct *tsk, unsigned int val)
+{
+       struct pt_regs *regs = tsk->thread.regs;
+
+       if ((val == PR_ENDIAN_LITTLE && !cpu_has_feature(CPU_FTR_REAL_LE)) ||
+           (val == PR_ENDIAN_PPC_LITTLE && !cpu_has_feature(CPU_FTR_PPC_LE)))
+               return -EINVAL;
+
+       if (regs == NULL)
+               return -EINVAL;
+
+       if (val == PR_ENDIAN_BIG)
+               regs->msr &= ~MSR_LE;
+       else if (val == PR_ENDIAN_LITTLE || val == PR_ENDIAN_PPC_LITTLE)
+               regs->msr |= MSR_LE;
+       else
+               return -EINVAL;
+
+       return 0;
+}
+
+int get_endian(struct task_struct *tsk, unsigned long adr)
+{
+       struct pt_regs *regs = tsk->thread.regs;
+       unsigned int val;
+
+       if (!cpu_has_feature(CPU_FTR_PPC_LE) &&
+           !cpu_has_feature(CPU_FTR_REAL_LE))
+               return -EINVAL;
+
+       if (regs == NULL)
+               return -EINVAL;
+
+       if (regs->msr & MSR_LE) {
+               if (cpu_has_feature(CPU_FTR_REAL_LE))
+                       val = PR_ENDIAN_LITTLE;
+               else
+                       val = PR_ENDIAN_PPC_LITTLE;
+       } else
+               val = PR_ENDIAN_BIG;
+
+       return put_user(val, (unsigned int __user *)adr);
+}
+
+int set_unalign_ctl(struct task_struct *tsk, unsigned int val)
+{
+       tsk->thread.align_ctl = val;
+       return 0;
+}
+
+int get_unalign_ctl(struct task_struct *tsk, unsigned long adr)
+{
+       return put_user(tsk->thread.align_ctl, (unsigned int __user *)adr);
+}
+
 #define TRUNC_PTR(x)   ((typeof(x))(((unsigned long)(x)) & 0xffffffff))
 
 int sys_clone(unsigned long clone_flags, unsigned long usp,
@@ -783,53 +931,53 @@ int sys_execve(unsigned long a0, unsigned long a1, unsigned long a2,
        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;
 }
 
-static int validate_sp(unsigned long sp, struct task_struct *p,
-                      unsigned long nbytes)
+#ifdef CONFIG_IRQSTACKS
+static inline int valid_irq_stack(unsigned long sp, struct task_struct *p,
+                                 unsigned long nbytes)
 {
-       unsigned long stack_page = (unsigned long)p->thread_info;
+       unsigned long stack_page;
+       unsigned long cpu = task_cpu(p);
 
-       if (sp >= stack_page + sizeof(struct thread_struct)
-           && sp <= stack_page + THREAD_SIZE - nbytes)
-               return 1;
+       /*
+        * Avoid crashing if the stack has overflowed and corrupted
+        * task_cpu(p), which is in the thread_info struct.
+        */
+       if (cpu < NR_CPUS && cpu_possible(cpu)) {
+               stack_page = (unsigned long) hardirq_ctx[cpu];
+               if (sp >= stack_page + sizeof(struct thread_struct)
+                   && sp <= stack_page + THREAD_SIZE - nbytes)
+                       return 1;
+
+               stack_page = (unsigned long) softirq_ctx[cpu];
+               if (sp >= stack_page + sizeof(struct thread_struct)
+                   && sp <= stack_page + THREAD_SIZE - nbytes)
+                       return 1;
+       }
+       return 0;
+}
 
-#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;
+#else
+#define valid_irq_stack(sp, p, nb)     0
+#endif /* CONFIG_IRQSTACKS */
+
+int validate_sp(unsigned long sp, struct task_struct *p,
+                      unsigned long nbytes)
+{
+       unsigned long stack_page = (unsigned long)task_stack_page(p);
 
-       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;
+       return valid_irq_stack(sp, p, nbytes);
 }
 
-#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)
 {
@@ -840,30 +988,37 @@ unsigned long get_wchan(struct task_struct *p)
                return 0;
 
        sp = p->thread.ksp;
-       if (!validate_sp(sp, p, MIN_STACK_FRAME))
+       if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
                return 0;
 
        do {
                sp = *(unsigned long *)sp;
-               if (!validate_sp(sp, p, MIN_STACK_FRAME))
+               if (!validate_sp(sp, p, STACK_FRAME_OVERHEAD))
                        return 0;
                if (count > 0) {
-                       ip = ((unsigned long *)sp)[FRAME_LR_SAVE];
+                       ip = ((unsigned long *)sp)[STACK_FRAME_LR_SAVE];
                        if (!in_sched_functions(ip))
                                return ip;
                }
        } while (count++ < 16);
        return 0;
 }
-EXPORT_SYMBOL(get_wchan);
 
-static int kstack_depth_to_print = 64;
+static int kstack_depth_to_print = CONFIG_PRINT_STACK_DEPTH;
 
 void show_stack(struct task_struct *tsk, unsigned long *stack)
 {
        unsigned long sp, ip, lr, newsp;
        int count = 0;
        int firstframe = 1;
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+       int curr_frame = current->curr_ret_stack;
+       extern void return_to_handler(void);
+       unsigned long addr = (unsigned long)return_to_handler;
+#ifdef CONFIG_PPC64
+       addr = *(unsigned long*)addr;
+#endif
+#endif
 
        sp = (unsigned long) stack;
        if (tsk == NULL)
@@ -878,15 +1033,21 @@ void show_stack(struct task_struct *tsk, unsigned long *stack)
        lr = 0;
        printk("Call Trace:\n");
        do {
-               if (!validate_sp(sp, tsk, MIN_STACK_FRAME))
+               if (!validate_sp(sp, tsk, STACK_FRAME_OVERHEAD))
                        return;
 
                stack = (unsigned long *) sp;
                newsp = stack[0];
-               ip = stack[FRAME_LR_SAVE];
+               ip = stack[STACK_FRAME_LR_SAVE];
                if (!firstframe || ip != lr) {
-                       printk("["REG"] ["REG"] ", sp, ip);
-                       print_symbol("%s", ip);
+                       printk("["REG"] ["REG"] %pS", sp, ip, (void *)ip);
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+                       if (ip == addr && curr_frame >= 0) {
+                               printk(" (%pS)",
+                                      (void *)current->ret_stack[curr_frame].ret);
+                               curr_frame--;
+                       }
+#endif
                        if (firstframe)
                                printk(" (unreliable)");
                        printk("\n");
@@ -897,14 +1058,13 @@ void show_stack(struct task_struct *tsk, unsigned long *stack)
                 * 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) {
+               if (validate_sp(sp, tsk, STACK_INT_FRAME_SIZE)
+                   && stack[STACK_FRAME_MARKER] == STACK_FRAME_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);
+                       printk("--- Exception: %lx at %pS\n    LR = %pS\n",
+                              regs->trap, (void *)regs->nip, (void *)lr);
                        firstframe = 1;
                }
 
@@ -917,3 +1077,106 @@ 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
+
+#if THREAD_SHIFT < PAGE_SHIFT
+
+static struct kmem_cache *thread_info_cache;
+
+struct thread_info *alloc_thread_info(struct task_struct *tsk)
+{
+       struct thread_info *ti;
+
+       ti = kmem_cache_alloc(thread_info_cache, GFP_KERNEL);
+       if (unlikely(ti == NULL))
+               return NULL;
+#ifdef CONFIG_DEBUG_STACK_USAGE
+       memset(ti, 0, THREAD_SIZE);
+#endif
+       return ti;
+}
+
+void free_thread_info(struct thread_info *ti)
+{
+       kmem_cache_free(thread_info_cache, ti);
+}
+
+void thread_info_cache_init(void)
+{
+       thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE,
+                                             THREAD_SIZE, 0, NULL);
+       BUG_ON(thread_info_cache == NULL);
+}
+
+#endif /* THREAD_SHIFT < PAGE_SHIFT */
+
+unsigned long arch_align_stack(unsigned long sp)
+{
+       if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
+               sp -= get_random_int() & ~PAGE_MASK;
+       return sp & ~0xf;
+}
+
+static inline unsigned long brk_rnd(void)
+{
+        unsigned long rnd = 0;
+
+       /* 8MB for 32bit, 1GB for 64bit */
+       if (is_32bit_task())
+               rnd = (long)(get_random_int() % (1<<(23-PAGE_SHIFT)));
+       else
+               rnd = (long)(get_random_int() % (1<<(30-PAGE_SHIFT)));
+
+       return rnd << PAGE_SHIFT;
+}
+
+unsigned long arch_randomize_brk(struct mm_struct *mm)
+{
+       unsigned long ret = PAGE_ALIGN(mm->brk + brk_rnd());
+
+       if (ret < mm->brk)
+               return mm->brk;
+
+       return ret;
+}
+
+unsigned long randomize_et_dyn(unsigned long base)
+{
+       unsigned long ret = PAGE_ALIGN(base + brk_rnd());
+
+       if (ret < base)
+               return base;
+
+       return ret;
+}