[POWERPC] Don't touch PT_DTRACE in exec
[safe/jmp/linux-2.6] / arch / powerpc / kernel / process.c
index 105d560..7c8e3da 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>
@@ -37,7 +33,6 @@
 #include <linux/mqueue.h>
 #include <linux/hardirq.h>
 #include <linux/utsname.h>
-#include <linux/kprobes.h>
 
 #include <asm/pgtable.h>
 #include <asm/uaccess.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/time.h>
 #endif
 
 extern unsigned long _get_SP(void);
@@ -87,7 +83,7 @@ void flush_fp_to_thread(struct task_struct *tsk)
                         */
                        BUG_ON(tsk != current);
 #endif
-                       giveup_fpu(current);
+                       giveup_fpu(tsk);
                }
                preempt_enable();
        }
@@ -147,16 +143,38 @@ 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();
        }
 }
 
-int dump_task_altivec(struct pt_regs *regs, elf_vrregset_t *vrregs)
+int dump_task_altivec(struct task_struct *tsk, elf_vrregset_t *vrregs)
 {
-       flush_altivec_to_thread(current);
-       memcpy(vrregs, &current->thread.vr[0], sizeof(*vrregs));
+       /* ELF_NVRREG includes the VSCR and VRSAVE which we need to save
+        * separately, see below */
+       const int nregs = ELF_NVRREG - 2;
+       elf_vrreg_t *reg;
+       u32 *dest;
+
+       if (tsk == current)
+               flush_altivec_to_thread(tsk);
+
+       reg = (elf_vrreg_t *)vrregs;
+
+       /* copy the 32 vr registers */
+       memcpy(reg, &tsk->thread.vr[0], nregs * sizeof(*reg));
+       reg += nregs;
+
+       /* copy the vscr */
+       memcpy(reg, &tsk->thread.vscr, sizeof(*reg));
+       reg++;
+
+       /* vrsave is stored in the high 32bit slot of the final 128bits */
+       memset(reg, 0, sizeof(*reg));
+       dest = (u32 *)reg;
+       *dest = tsk->thread.vrsave;
+
        return 1;
 }
 #endif /* CONFIG_ALTIVEC */
@@ -186,7 +204,7 @@ 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();
        }
@@ -201,13 +219,13 @@ int dump_spe(struct pt_regs *regs, elf_vrregset_t *evrregs)
 }
 #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.
  */
-static inline void discard_lazy_cpu_state(void)
+void discard_lazy_cpu_state(void)
 {
-#ifndef CONFIG_SMP
        preempt_disable();
        if (last_task_used_math == current)
                last_task_used_math = NULL;
@@ -220,23 +238,29 @@ static inline void discard_lazy_cpu_state(void)
                last_task_used_spe = NULL;
 #endif
        preempt_enable();
-#endif /* CONFIG_SMP */
 }
+#endif /* CONFIG_SMP */
 
 int set_dabr(unsigned long dabr)
 {
+#ifdef CONFIG_PPC_MERGE                /* XXX for now */
        if (ppc_md.set_dabr)
                return ppc_md.set_dabr(dabr);
+#endif
 
+       /* XXX should we have a CPU_FTR_HAS_DABR ? */
+#if defined(CONFIG_PPC64) || defined(CONFIG_6xx)
        mtspr(SPRN_DABR, 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
 
+static DEFINE_PER_CPU(unsigned long, current_dabr);
+
 struct task_struct *__switch_to(struct task_struct *prev,
        struct task_struct *new)
 {
@@ -301,15 +325,11 @@ 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)) {
                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;
 
@@ -328,6 +348,11 @@ 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();
+
        last = _switch(old_thread, new_thread);
 
        local_irq_restore(flags);
@@ -337,13 +362,6 @@ 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;
@@ -358,7 +376,20 @@ 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)
@@ -400,11 +431,11 @@ 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
@@ -416,18 +447,22 @@ 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++) {
@@ -455,7 +490,6 @@ void show_regs(struct pt_regs * regs)
 
 void exit_thread(void)
 {
-       kprobe_flush_task(current);
        discard_lazy_cpu_state();
 }
 
@@ -464,18 +498,21 @@ 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
 
        discard_lazy_cpu_state();
 
-#ifdef CONFIG_PPC64    /* for now */
        if (current->thread.dabr) {
                current->thread.dabr = 0;
                set_dabr(0);
        }
-#endif
 }
 
 void
@@ -503,7 +540,7 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 {
        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 */
@@ -516,7 +553,7 @@ 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 {
@@ -549,10 +586,15 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
 
 #ifdef CONFIG_PPC64
        if (cpu_has_feature(CPU_FTR_SLB)) {
-               unsigned long sp_vsid = get_kernel_vsid(sp);
+               unsigned long sp_vsid;
                unsigned long llp = mmu_psize_defs[mmu_linear_psize].sllp;
 
-               sp_vsid <<= SLB_VSID_SHIFT;
+               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;
        }
@@ -566,7 +608,6 @@ 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;
@@ -588,10 +629,8 @@ 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));
@@ -601,6 +640,13 @@ 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;
@@ -665,9 +711,13 @@ 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
@@ -693,7 +743,10 @@ 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
@@ -702,6 +755,61 @@ 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,
@@ -754,38 +862,50 @@ 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
@@ -802,6 +922,8 @@ static int validate_sp(unsigned long sp, struct task_struct *p,
 #define FRAME_MARKER   2
 #endif
 
+EXPORT_SYMBOL(validate_sp);
+
 unsigned long get_wchan(struct task_struct *p)
 {
        unsigned long ip, sp;
@@ -826,7 +948,6 @@ unsigned long get_wchan(struct task_struct *p)
        } while (count++ < 16);
        return 0;
 }
-EXPORT_SYMBOL(get_wchan);
 
 static int kstack_depth_to_print = 64;
 
@@ -888,3 +1009,35 @@ 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