]> pilppa.org Git - linux-2.6-omap-h63xx.git/blobdiff - arch/x86/kernel/ptrace.c
x86: EFI runtime code mapping enhancement
[linux-2.6-omap-h63xx.git] / arch / x86 / kernel / ptrace.c
index eaec75a4094b0d42662b902694cb3c6140e0236b..702c33efea84dfad40b15d62cd7775958d653979 100644 (file)
@@ -2,6 +2,9 @@
 /*
  * Pentium III FXSR, SSE support
  *     Gareth Hughes <gareth@valinux.com>, May 2000
+ *
+ * BTS tracing
+ *     Markus Metzger <markus.t.metzger@intel.com>, Dec 2007
  */
 
 #include <linux/kernel.h>
@@ -10,7 +13,9 @@
 #include <linux/smp.h>
 #include <linux/errno.h>
 #include <linux/ptrace.h>
+#include <linux/regset.h>
 #include <linux/user.h>
+#include <linux/elf.h>
 #include <linux/security.h>
 #include <linux/audit.h>
 #include <linux/seccomp.h>
 #include <asm/debugreg.h>
 #include <asm/ldt.h>
 #include <asm/desc.h>
+#include <asm/prctl.h>
+#include <asm/proto.h>
+#include <asm/ds.h>
+
+#include "tls.h"
+
+enum x86_regset {
+       REGSET_GENERAL,
+       REGSET_FP,
+       REGSET_XFP,
+       REGSET_TLS,
+};
 
 /*
  * does not yet catch signals sent when the child dies.
                                  X86_EFLAGS_DF | X86_EFLAGS_OF |       \
                                  X86_EFLAGS_RF | X86_EFLAGS_AC))
 
+/*
+ * Determines whether a value may be installed in a segment register.
+ */
+static inline bool invalid_selector(u16 value)
+{
+       return unlikely(value != 0 && (value & SEGMENT_RPL_MASK) != USER_RPL);
+}
+
+#ifdef CONFIG_X86_32
+
 #define FLAG_MASK              FLAG_MASK_32
 
 static long *pt_regs_access(struct pt_regs *regs, unsigned long regno)
@@ -73,12 +100,29 @@ static int set_segment_reg(struct task_struct *task,
        /*
         * The value argument was already truncated to 16 bits.
         */
-       if (value && (value & 3) != 3)
+       if (invalid_selector(value))
                return -EIO;
 
-       if (offset != offsetof(struct user_regs_struct, gs))
+       /*
+        * For %cs and %ss we cannot permit a null selector.
+        * We can permit a bogus selector as long as it has USER_RPL.
+        * Null selectors are fine for other segment registers, but
+        * we will never get back to user mode with invalid %cs or %ss
+        * and will take the trap in iret instead.  Much code relies
+        * on user_mode() to distinguish a user trap frame (which can
+        * safely use invalid selectors) from a kernel trap frame.
+        */
+       switch (offset) {
+       case offsetof(struct user_regs_struct, cs):
+       case offsetof(struct user_regs_struct, ss):
+               if (unlikely(value == 0))
+                       return -EIO;
+
+       default:
                *pt_regs_access(task_pt_regs(task), offset) = value;
-       else {
+               break;
+
+       case offsetof(struct user_regs_struct, gs):
                task->thread.gs = value;
                if (task == current)
                        /*
@@ -91,6 +135,146 @@ static int set_segment_reg(struct task_struct *task,
        return 0;
 }
 
+static unsigned long debugreg_addr_limit(struct task_struct *task)
+{
+       return TASK_SIZE - 3;
+}
+
+#else  /* CONFIG_X86_64 */
+
+#define FLAG_MASK              (FLAG_MASK_32 | X86_EFLAGS_NT)
+
+static unsigned long *pt_regs_access(struct pt_regs *regs, unsigned long offset)
+{
+       BUILD_BUG_ON(offsetof(struct pt_regs, r15) != 0);
+       return &regs->r15 + (offset / sizeof(regs->r15));
+}
+
+static u16 get_segment_reg(struct task_struct *task, unsigned long offset)
+{
+       /*
+        * Returning the value truncates it to 16 bits.
+        */
+       unsigned int seg;
+
+       switch (offset) {
+       case offsetof(struct user_regs_struct, fs):
+               if (task == current) {
+                       /* Older gas can't assemble movq %?s,%r?? */
+                       asm("movl %%fs,%0" : "=r" (seg));
+                       return seg;
+               }
+               return task->thread.fsindex;
+       case offsetof(struct user_regs_struct, gs):
+               if (task == current) {
+                       asm("movl %%gs,%0" : "=r" (seg));
+                       return seg;
+               }
+               return task->thread.gsindex;
+       case offsetof(struct user_regs_struct, ds):
+               if (task == current) {
+                       asm("movl %%ds,%0" : "=r" (seg));
+                       return seg;
+               }
+               return task->thread.ds;
+       case offsetof(struct user_regs_struct, es):
+               if (task == current) {
+                       asm("movl %%es,%0" : "=r" (seg));
+                       return seg;
+               }
+               return task->thread.es;
+
+       case offsetof(struct user_regs_struct, cs):
+       case offsetof(struct user_regs_struct, ss):
+               break;
+       }
+       return *pt_regs_access(task_pt_regs(task), offset);
+}
+
+static int set_segment_reg(struct task_struct *task,
+                          unsigned long offset, u16 value)
+{
+       /*
+        * The value argument was already truncated to 16 bits.
+        */
+       if (invalid_selector(value))
+               return -EIO;
+
+       switch (offset) {
+       case offsetof(struct user_regs_struct,fs):
+               /*
+                * If this is setting fs as for normal 64-bit use but
+                * setting fs_base has implicitly changed it, leave it.
+                */
+               if ((value == FS_TLS_SEL && task->thread.fsindex == 0 &&
+                    task->thread.fs != 0) ||
+                   (value == 0 && task->thread.fsindex == FS_TLS_SEL &&
+                    task->thread.fs == 0))
+                       break;
+               task->thread.fsindex = value;
+               if (task == current)
+                       loadsegment(fs, task->thread.fsindex);
+               break;
+       case offsetof(struct user_regs_struct,gs):
+               /*
+                * If this is setting gs as for normal 64-bit use but
+                * setting gs_base has implicitly changed it, leave it.
+                */
+               if ((value == GS_TLS_SEL && task->thread.gsindex == 0 &&
+                    task->thread.gs != 0) ||
+                   (value == 0 && task->thread.gsindex == GS_TLS_SEL &&
+                    task->thread.gs == 0))
+                       break;
+               task->thread.gsindex = value;
+               if (task == current)
+                       load_gs_index(task->thread.gsindex);
+               break;
+       case offsetof(struct user_regs_struct,ds):
+               task->thread.ds = value;
+               if (task == current)
+                       loadsegment(ds, task->thread.ds);
+               break;
+       case offsetof(struct user_regs_struct,es):
+               task->thread.es = value;
+               if (task == current)
+                       loadsegment(es, task->thread.es);
+               break;
+
+               /*
+                * Can't actually change these in 64-bit mode.
+                */
+       case offsetof(struct user_regs_struct,cs):
+               if (unlikely(value == 0))
+                       return -EIO;
+#ifdef CONFIG_IA32_EMULATION
+               if (test_tsk_thread_flag(task, TIF_IA32))
+                       task_pt_regs(task)->cs = value;
+#endif
+               break;
+       case offsetof(struct user_regs_struct,ss):
+               if (unlikely(value == 0))
+                       return -EIO;
+#ifdef CONFIG_IA32_EMULATION
+               if (test_tsk_thread_flag(task, TIF_IA32))
+                       task_pt_regs(task)->ss = value;
+#endif
+               break;
+       }
+
+       return 0;
+}
+
+static unsigned long debugreg_addr_limit(struct task_struct *task)
+{
+#ifdef CONFIG_IA32_EMULATION
+       if (test_tsk_thread_flag(task, TIF_IA32))
+               return IA32_PAGE_OFFSET - 3;
+#endif
+       return TASK_SIZE64 - 7;
+}
+
+#endif /* CONFIG_X86_32 */
+
 static unsigned long get_flags(struct task_struct *task)
 {
        unsigned long retval = task_pt_regs(task)->flags;
@@ -137,6 +321,29 @@ static int putreg(struct task_struct *child,
 
        case offsetof(struct user_regs_struct, flags):
                return set_flags(child, value);
+
+#ifdef CONFIG_X86_64
+       case offsetof(struct user_regs_struct,fs_base):
+               if (value >= TASK_SIZE_OF(child))
+                       return -EIO;
+               /*
+                * When changing the segment base, use do_arch_prctl
+                * to set either thread.fs or thread.fsindex and the
+                * corresponding GDT slot.
+                */
+               if (child->thread.fs != value)
+                       return do_arch_prctl(child, ARCH_SET_FS, value);
+               return 0;
+       case offsetof(struct user_regs_struct,gs_base):
+               /*
+                * Exactly the same here as the %fs handling above.
+                */
+               if (value >= TASK_SIZE_OF(child))
+                       return -EIO;
+               if (child->thread.gs != value)
+                       return do_arch_prctl(child, ARCH_SET_GS, value);
+               return 0;
+#endif
        }
 
        *pt_regs_access(task_pt_regs(child), offset) = value;
@@ -156,11 +363,95 @@ static unsigned long getreg(struct task_struct *task, unsigned long offset)
 
        case offsetof(struct user_regs_struct, flags):
                return get_flags(task);
+
+#ifdef CONFIG_X86_64
+       case offsetof(struct user_regs_struct, fs_base): {
+               /*
+                * do_arch_prctl may have used a GDT slot instead of
+                * the MSR.  To userland, it appears the same either
+                * way, except the %fs segment selector might not be 0.
+                */
+               unsigned int seg = task->thread.fsindex;
+               if (task->thread.fs != 0)
+                       return task->thread.fs;
+               if (task == current)
+                       asm("movl %%fs,%0" : "=r" (seg));
+               if (seg != FS_TLS_SEL)
+                       return 0;
+               return get_desc_base(&task->thread.tls_array[FS_TLS]);
+       }
+       case offsetof(struct user_regs_struct, gs_base): {
+               /*
+                * Exactly the same here as the %fs handling above.
+                */
+               unsigned int seg = task->thread.gsindex;
+               if (task->thread.gs != 0)
+                       return task->thread.gs;
+               if (task == current)
+                       asm("movl %%gs,%0" : "=r" (seg));
+               if (seg != GS_TLS_SEL)
+                       return 0;
+               return get_desc_base(&task->thread.tls_array[GS_TLS]);
+       }
+#endif
        }
 
        return *pt_regs_access(task_pt_regs(task), offset);
 }
 
+static int genregs_get(struct task_struct *target,
+                      const struct user_regset *regset,
+                      unsigned int pos, unsigned int count,
+                      void *kbuf, void __user *ubuf)
+{
+       if (kbuf) {
+               unsigned long *k = kbuf;
+               while (count > 0) {
+                       *k++ = getreg(target, pos);
+                       count -= sizeof(*k);
+                       pos += sizeof(*k);
+               }
+       } else {
+               unsigned long __user *u = ubuf;
+               while (count > 0) {
+                       if (__put_user(getreg(target, pos), u++))
+                               return -EFAULT;
+                       count -= sizeof(*u);
+                       pos += sizeof(*u);
+               }
+       }
+
+       return 0;
+}
+
+static int genregs_set(struct task_struct *target,
+                      const struct user_regset *regset,
+                      unsigned int pos, unsigned int count,
+                      const void *kbuf, const void __user *ubuf)
+{
+       int ret = 0;
+       if (kbuf) {
+               const unsigned long *k = kbuf;
+               while (count > 0 && !ret) {
+                       ret = putreg(target, pos, *k++);
+                       count -= sizeof(*k);
+                       pos += sizeof(*k);
+               }
+       } else {
+               const unsigned long  __user *u = ubuf;
+               while (count > 0 && !ret) {
+                       unsigned long word;
+                       ret = __get_user(word, u++);
+                       if (ret)
+                               break;
+                       ret = putreg(target, pos, word);
+                       count -= sizeof(*u);
+                       pos += sizeof(*u);
+               }
+       }
+       return ret;
+}
+
 /*
  * This function is trivial and will be inlined by the compiler.
  * Having it separates the implementation details of debug
@@ -187,7 +478,7 @@ static int ptrace_set_debugreg(struct task_struct *child,
        if (unlikely(n == 4 || n == 5))
                return -EIO;
 
-       if (n < 4 && unlikely(data >= TASK_SIZE - 3))
+       if (n < 4 && unlikely(data >= debugreg_addr_limit(child)))
                return -EIO;
 
        switch (n) {
@@ -197,6 +488,8 @@ static int ptrace_set_debugreg(struct task_struct *child,
        case 3:         child->thread.debugreg3 = data; break;
 
        case 6:
+               if ((data & ~0xffffffffUL) != 0)
+                       return -EIO;
                child->thread.debugreg6 = data;
                break;
 
@@ -215,7 +508,7 @@ static int ptrace_set_debugreg(struct task_struct *child,
                 * data in the watchpoint case.
                 *
                 * The invalid values are:
-                * - LENi == 0x10 (undefined), so mask |= 0x0f00.
+                * - LENi == 0x10 (undefined), so mask |= 0x0f00.       [32-bit]
                 * - R/Wi == 0x10 (break on I/O reads or writes), so
                 *   mask |= 0x4444.
                 * - R/Wi == 0x00 && LENi != 0x00, so we have mask |=
@@ -231,9 +524,14 @@ static int ptrace_set_debugreg(struct task_struct *child,
                 * 64-bit kernel), so the x86_64 mask value is 0x5454.
                 * See the AMD manual no. 24593 (AMD64 System Programming)
                 */
+#ifdef CONFIG_X86_32
+#define        DR7_MASK        0x5f54
+#else
+#define        DR7_MASK        0x5554
+#endif
                data &= ~DR_CONTROL_RESERVED;
                for (i = 0; i < 4; i++)
-                       if ((0x5f54 >> ((data >> (16 + 4*i)) & 0xf)) & 1)
+                       if ((DR7_MASK >> ((data >> (16 + 4*i)) & 0xf)) & 1)
                                return -EIO;
                child->thread.debugreg7 = data;
                if (data)
@@ -246,6 +544,289 @@ static int ptrace_set_debugreg(struct task_struct *child,
        return 0;
 }
 
+static int ptrace_bts_get_size(struct task_struct *child)
+{
+       if (!child->thread.ds_area_msr)
+               return -ENXIO;
+
+       return ds_get_bts_index((void *)child->thread.ds_area_msr);
+}
+
+static int ptrace_bts_read_record(struct task_struct *child,
+                                 long index,
+                                 struct bts_struct __user *out)
+{
+       struct bts_struct ret;
+       int retval;
+       int bts_end;
+       int bts_index;
+
+       if (!child->thread.ds_area_msr)
+               return -ENXIO;
+
+       if (index < 0)
+               return -EINVAL;
+
+       bts_end = ds_get_bts_end((void *)child->thread.ds_area_msr);
+       if (bts_end <= index)
+               return -EINVAL;
+
+       /* translate the ptrace bts index into the ds bts index */
+       bts_index = ds_get_bts_index((void *)child->thread.ds_area_msr);
+       bts_index -= (index + 1);
+       if (bts_index < 0)
+               bts_index += bts_end;
+
+       retval = ds_read_bts((void *)child->thread.ds_area_msr,
+                            bts_index, &ret);
+       if (retval < 0)
+               return retval;
+
+       if (copy_to_user(out, &ret, sizeof(ret)))
+               return -EFAULT;
+
+       return sizeof(ret);
+}
+
+static int ptrace_bts_write_record(struct task_struct *child,
+                                  const struct bts_struct *in)
+{
+       int retval;
+
+       if (!child->thread.ds_area_msr)
+               return -ENXIO;
+
+       retval = ds_write_bts((void *)child->thread.ds_area_msr, in);
+       if (retval)
+               return retval;
+
+       return sizeof(*in);
+}
+
+static int ptrace_bts_clear(struct task_struct *child)
+{
+       if (!child->thread.ds_area_msr)
+               return -ENXIO;
+
+       return ds_clear((void *)child->thread.ds_area_msr);
+}
+
+static int ptrace_bts_drain(struct task_struct *child,
+                           long size,
+                           struct bts_struct __user *out)
+{
+       int end, i;
+       void *ds = (void *)child->thread.ds_area_msr;
+
+       if (!ds)
+               return -ENXIO;
+
+       end = ds_get_bts_index(ds);
+       if (end <= 0)
+               return end;
+
+       if (size < (end * sizeof(struct bts_struct)))
+               return -EIO;
+
+       for (i = 0; i < end; i++, out++) {
+               struct bts_struct ret;
+               int retval;
+
+               retval = ds_read_bts(ds, i, &ret);
+               if (retval < 0)
+                       return retval;
+
+               if (copy_to_user(out, &ret, sizeof(ret)))
+                       return -EFAULT;
+       }
+
+       ds_clear(ds);
+
+       return end;
+}
+
+static int ptrace_bts_realloc(struct task_struct *child,
+                             int size, int reduce_size)
+{
+       unsigned long rlim, vm;
+       int ret, old_size;
+
+       if (size < 0)
+               return -EINVAL;
+
+       old_size = ds_get_bts_size((void *)child->thread.ds_area_msr);
+       if (old_size < 0)
+               return old_size;
+
+       ret = ds_free((void **)&child->thread.ds_area_msr);
+       if (ret < 0)
+               goto out;
+
+       size >>= PAGE_SHIFT;
+       old_size >>= PAGE_SHIFT;
+
+       current->mm->total_vm  -= old_size;
+       current->mm->locked_vm -= old_size;
+
+       if (size == 0)
+               goto out;
+
+       rlim = current->signal->rlim[RLIMIT_AS].rlim_cur >> PAGE_SHIFT;
+       vm = current->mm->total_vm  + size;
+       if (rlim < vm) {
+               ret = -ENOMEM;
+
+               if (!reduce_size)
+                       goto out;
+
+               size = rlim - current->mm->total_vm;
+               if (size <= 0)
+                       goto out;
+       }
+
+       rlim = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >> PAGE_SHIFT;
+       vm = current->mm->locked_vm  + size;
+       if (rlim < vm) {
+               ret = -ENOMEM;
+
+               if (!reduce_size)
+                       goto out;
+
+               size = rlim - current->mm->locked_vm;
+               if (size <= 0)
+                       goto out;
+       }
+
+       ret = ds_allocate((void **)&child->thread.ds_area_msr,
+                         size << PAGE_SHIFT);
+       if (ret < 0)
+               goto out;
+
+       current->mm->total_vm  += size;
+       current->mm->locked_vm += size;
+
+out:
+       if (child->thread.ds_area_msr)
+               set_tsk_thread_flag(child, TIF_DS_AREA_MSR);
+       else
+               clear_tsk_thread_flag(child, TIF_DS_AREA_MSR);
+
+       return ret;
+}
+
+static int ptrace_bts_config(struct task_struct *child,
+                            long cfg_size,
+                            const struct ptrace_bts_config __user *ucfg)
+{
+       struct ptrace_bts_config cfg;
+       int bts_size, ret = 0;
+       void *ds;
+
+       if (cfg_size < sizeof(cfg))
+               return -EIO;
+
+       if (copy_from_user(&cfg, ucfg, sizeof(cfg)))
+               return -EFAULT;
+
+       if ((int)cfg.size < 0)
+               return -EINVAL;
+
+       bts_size = 0;
+       ds = (void *)child->thread.ds_area_msr;
+       if (ds) {
+               bts_size = ds_get_bts_size(ds);
+               if (bts_size < 0)
+                       return bts_size;
+       }
+       cfg.size = PAGE_ALIGN(cfg.size);
+
+       if (bts_size != cfg.size) {
+               ret = ptrace_bts_realloc(child, cfg.size,
+                                        cfg.flags & PTRACE_BTS_O_CUT_SIZE);
+               if (ret < 0)
+                       goto errout;
+
+               ds = (void *)child->thread.ds_area_msr;
+       }
+
+       if (cfg.flags & PTRACE_BTS_O_SIGNAL)
+               ret = ds_set_overflow(ds, DS_O_SIGNAL);
+       else
+               ret = ds_set_overflow(ds, DS_O_WRAP);
+       if (ret < 0)
+               goto errout;
+
+       if (cfg.flags & PTRACE_BTS_O_TRACE)
+               child->thread.debugctlmsr |= ds_debugctl_mask();
+       else
+               child->thread.debugctlmsr &= ~ds_debugctl_mask();
+
+       if (cfg.flags & PTRACE_BTS_O_SCHED)
+               set_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+       else
+               clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+
+       ret = sizeof(cfg);
+
+out:
+       if (child->thread.debugctlmsr)
+               set_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
+       else
+               clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
+
+       return ret;
+
+errout:
+       child->thread.debugctlmsr &= ~ds_debugctl_mask();
+       clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+       goto out;
+}
+
+static int ptrace_bts_status(struct task_struct *child,
+                            long cfg_size,
+                            struct ptrace_bts_config __user *ucfg)
+{
+       void *ds = (void *)child->thread.ds_area_msr;
+       struct ptrace_bts_config cfg;
+
+       if (cfg_size < sizeof(cfg))
+               return -EIO;
+
+       memset(&cfg, 0, sizeof(cfg));
+
+       if (ds) {
+               cfg.size = ds_get_bts_size(ds);
+
+               if (ds_get_overflow(ds) == DS_O_SIGNAL)
+                       cfg.flags |= PTRACE_BTS_O_SIGNAL;
+
+               if (test_tsk_thread_flag(child, TIF_DEBUGCTLMSR) &&
+                   child->thread.debugctlmsr & ds_debugctl_mask())
+                       cfg.flags |= PTRACE_BTS_O_TRACE;
+
+               if (test_tsk_thread_flag(child, TIF_BTS_TRACE_TS))
+                       cfg.flags |= PTRACE_BTS_O_SCHED;
+       }
+
+       cfg.bts_size = sizeof(struct bts_struct);
+
+       if (copy_to_user(ucfg, &cfg, sizeof(cfg)))
+               return -EFAULT;
+
+       return sizeof(cfg);
+}
+
+void ptrace_bts_take_timestamp(struct task_struct *tsk,
+                              enum bts_qualifier qualifier)
+{
+       struct bts_struct rec = {
+               .qualifier = qualifier,
+               .variant.jiffies = jiffies_64
+       };
+
+       ptrace_bts_write_record(tsk, &rec);
+}
+
 /*
  * Called by kernel/ptrace.c when detaching..
  *
@@ -254,172 +835,576 @@ static int ptrace_set_debugreg(struct task_struct *child,
 void ptrace_disable(struct task_struct *child)
 {
        user_disable_single_step(child);
+#ifdef TIF_SYSCALL_EMU
        clear_tsk_thread_flag(child, TIF_SYSCALL_EMU);
+#endif
+       if (child->thread.ds_area_msr) {
+               ptrace_bts_realloc(child, 0, 0);
+               child->thread.debugctlmsr &= ~ds_debugctl_mask();
+               if (!child->thread.debugctlmsr)
+                       clear_tsk_thread_flag(child, TIF_DEBUGCTLMSR);
+               clear_tsk_thread_flag(child, TIF_BTS_TRACE_TS);
+       }
 }
 
+#if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
+static const struct user_regset_view user_x86_32_view; /* Initialized below. */
+#endif
+
 long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 {
-       struct user * dummy = NULL;
-       int i, ret;
+       int ret;
        unsigned long __user *datap = (unsigned long __user *)data;
 
        switch (request) {
-       /* when I and D space are separate, these will need to be fixed. */
-       case PTRACE_PEEKTEXT: /* read word at location addr. */
-       case PTRACE_PEEKDATA:
-               ret = generic_ptrace_peekdata(child, addr, data);
-               break;
-
        /* read the word at location addr in the USER area. */
        case PTRACE_PEEKUSR: {
                unsigned long tmp;
 
                ret = -EIO;
-               if ((addr & 3) || addr < 0 ||
-                   addr > sizeof(struct user) - 3)
+               if ((addr & (sizeof(data) - 1)) || addr < 0 ||
+                   addr >= sizeof(struct user))
                        break;
 
                tmp = 0;  /* Default return condition */
-               if(addr < FRAME_SIZE*sizeof(long))
+               if (addr < sizeof(struct user_regs_struct))
                        tmp = getreg(child, addr);
-               if(addr >= (long) &dummy->u_debugreg[0] &&
-                  addr <= (long) &dummy->u_debugreg[7]){
-                       addr -= (long) &dummy->u_debugreg[0];
-                       addr = addr >> 2;
-                       tmp = ptrace_get_debugreg(child, addr);
+               else if (addr >= offsetof(struct user, u_debugreg[0]) &&
+                        addr <= offsetof(struct user, u_debugreg[7])) {
+                       addr -= offsetof(struct user, u_debugreg[0]);
+                       tmp = ptrace_get_debugreg(child, addr / sizeof(data));
                }
                ret = put_user(tmp, datap);
                break;
        }
 
-       /* when I and D space are separate, this will have to be fixed. */
-       case PTRACE_POKETEXT: /* write the word at location addr. */
-       case PTRACE_POKEDATA:
-               ret = generic_ptrace_pokedata(child, addr, data);
-               break;
-
        case PTRACE_POKEUSR: /* write the word at location addr in the USER area */
                ret = -EIO;
-               if ((addr & 3) || addr < 0 ||
-                   addr > sizeof(struct user) - 3)
+               if ((addr & (sizeof(data) - 1)) || addr < 0 ||
+                   addr >= sizeof(struct user))
                        break;
 
-               if (addr < FRAME_SIZE*sizeof(long)) {
+               if (addr < sizeof(struct user_regs_struct))
                        ret = putreg(child, addr, data);
-                       break;
-               }
-               /* We need to be very careful here.  We implicitly
-                  want to modify a portion of the task_struct, and we
-                  have to be selective about what portions we allow someone
-                  to modify. */
-
-                 ret = -EIO;
-                 if(addr >= (long) &dummy->u_debugreg[0] &&
-                    addr <= (long) &dummy->u_debugreg[7]){
-                         addr -= (long) &dummy->u_debugreg;
-                         addr = addr >> 2;
-                         ret = ptrace_set_debugreg(child, addr, data);
-                 }
-                 break;
-
-       case PTRACE_GETREGS: { /* Get all gp regs from the child. */
-               if (!access_ok(VERIFY_WRITE, datap, FRAME_SIZE*sizeof(long))) {
-                       ret = -EIO;
-                       break;
-               }
-               for ( i = 0; i < FRAME_SIZE*sizeof(long); i += sizeof(long) ) {
-                       __put_user(getreg(child, i), datap);
-                       datap++;
+               else if (addr >= offsetof(struct user, u_debugreg[0]) &&
+                        addr <= offsetof(struct user, u_debugreg[7])) {
+                       addr -= offsetof(struct user, u_debugreg[0]);
+                       ret = ptrace_set_debugreg(child,
+                                                 addr / sizeof(data), data);
                }
-               ret = 0;
                break;
-       }
 
-       case PTRACE_SETREGS: { /* Set all gp regs in the child. */
-               unsigned long tmp;
-               if (!access_ok(VERIFY_READ, datap, FRAME_SIZE*sizeof(long))) {
-                       ret = -EIO;
-                       break;
-               }
-               for ( i = 0; i < FRAME_SIZE*sizeof(long); i += sizeof(long) ) {
-                       __get_user(tmp, datap);
-                       putreg(child, i, tmp);
-                       datap++;
-               }
-               ret = 0;
+       case PTRACE_GETREGS:    /* Get all gp regs from the child. */
+               return copy_regset_to_user(child,
+                                          task_user_regset_view(current),
+                                          REGSET_GENERAL,
+                                          0, sizeof(struct user_regs_struct),
+                                          datap);
+
+       case PTRACE_SETREGS:    /* Set all gp regs in the child. */
+               return copy_regset_from_user(child,
+                                            task_user_regset_view(current),
+                                            REGSET_GENERAL,
+                                            0, sizeof(struct user_regs_struct),
+                                            datap);
+
+       case PTRACE_GETFPREGS:  /* Get the child FPU state. */
+               return copy_regset_to_user(child,
+                                          task_user_regset_view(current),
+                                          REGSET_FP,
+                                          0, sizeof(struct user_i387_struct),
+                                          datap);
+
+       case PTRACE_SETFPREGS:  /* Set the child FPU state. */
+               return copy_regset_from_user(child,
+                                            task_user_regset_view(current),
+                                            REGSET_FP,
+                                            0, sizeof(struct user_i387_struct),
+                                            datap);
+
+#ifdef CONFIG_X86_32
+       case PTRACE_GETFPXREGS: /* Get the child extended FPU state. */
+               return copy_regset_to_user(child, &user_x86_32_view,
+                                          REGSET_XFP,
+                                          0, sizeof(struct user_fxsr_struct),
+                                          datap);
+
+       case PTRACE_SETFPXREGS: /* Set the child extended FPU state. */
+               return copy_regset_from_user(child, &user_x86_32_view,
+                                            REGSET_XFP,
+                                            0, sizeof(struct user_fxsr_struct),
+                                            datap);
+#endif
+
+#if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
+       case PTRACE_GET_THREAD_AREA:
+               if (addr < 0)
+                       return -EIO;
+               ret = do_get_thread_area(child, addr,
+                                        (struct user_desc __user *) data);
+               break;
+
+       case PTRACE_SET_THREAD_AREA:
+               if (addr < 0)
+                       return -EIO;
+               ret = do_set_thread_area(child, addr,
+                                        (struct user_desc __user *) data, 0);
+               break;
+#endif
+
+#ifdef CONFIG_X86_64
+               /* normal 64bit interface to access TLS data.
+                  Works just like arch_prctl, except that the arguments
+                  are reversed. */
+       case PTRACE_ARCH_PRCTL:
+               ret = do_arch_prctl(child, data, addr);
+               break;
+#endif
+
+       case PTRACE_BTS_CONFIG:
+               ret = ptrace_bts_config
+                       (child, data, (struct ptrace_bts_config __user *)addr);
+               break;
+
+       case PTRACE_BTS_STATUS:
+               ret = ptrace_bts_status
+                       (child, data, (struct ptrace_bts_config __user *)addr);
+               break;
+
+       case PTRACE_BTS_SIZE:
+               ret = ptrace_bts_get_size(child);
+               break;
+
+       case PTRACE_BTS_GET:
+               ret = ptrace_bts_read_record
+                       (child, data, (struct bts_struct __user *) addr);
+               break;
+
+       case PTRACE_BTS_CLEAR:
+               ret = ptrace_bts_clear(child);
+               break;
+
+       case PTRACE_BTS_DRAIN:
+               ret = ptrace_bts_drain
+                       (child, data, (struct bts_struct __user *) addr);
+               break;
+
+       default:
+               ret = ptrace_request(child, request, addr, data);
                break;
        }
 
-       case PTRACE_GETFPREGS: { /* Get the child FPU state. */
-               if (!access_ok(VERIFY_WRITE, datap,
-                              sizeof(struct user_i387_struct))) {
-                       ret = -EIO;
-                       break;
-               }
-               ret = 0;
-               if (!tsk_used_math(child))
-                       init_fpu(child);
-               get_fpregs((struct user_i387_struct __user *)data, child);
+       return ret;
+}
+
+#ifdef CONFIG_IA32_EMULATION
+
+#include <linux/compat.h>
+#include <linux/syscalls.h>
+#include <asm/ia32.h>
+#include <asm/user32.h>
+
+#define R32(l,q)                                                       \
+       case offsetof(struct user32, regs.l):                           \
+               regs->q = value; break
+
+#define SEG32(rs)                                                      \
+       case offsetof(struct user32, regs.rs):                          \
+               return set_segment_reg(child,                           \
+                                      offsetof(struct user_regs_struct, rs), \
+                                      value);                          \
+               break
+
+static int putreg32(struct task_struct *child, unsigned regno, u32 value)
+{
+       struct pt_regs *regs = task_pt_regs(child);
+
+       switch (regno) {
+
+       SEG32(cs);
+       SEG32(ds);
+       SEG32(es);
+       SEG32(fs);
+       SEG32(gs);
+       SEG32(ss);
+
+       R32(ebx, bx);
+       R32(ecx, cx);
+       R32(edx, dx);
+       R32(edi, di);
+       R32(esi, si);
+       R32(ebp, bp);
+       R32(eax, ax);
+       R32(orig_eax, orig_ax);
+       R32(eip, ip);
+       R32(esp, sp);
+
+       case offsetof(struct user32, regs.eflags):
+               return set_flags(child, value);
+
+       case offsetof(struct user32, u_debugreg[0]) ...
+               offsetof(struct user32, u_debugreg[7]):
+               regno -= offsetof(struct user32, u_debugreg[0]);
+               return ptrace_set_debugreg(child, regno / 4, value);
+
+       default:
+               if (regno > sizeof(struct user32) || (regno & 3))
+                       return -EIO;
+
+               /*
+                * Other dummy fields in the virtual user structure
+                * are ignored
+                */
                break;
        }
+       return 0;
+}
 
-       case PTRACE_SETFPREGS: { /* Set the child FPU state. */
-               if (!access_ok(VERIFY_READ, datap,
-                              sizeof(struct user_i387_struct))) {
-                       ret = -EIO;
-                       break;
-               }
-               set_stopped_child_used_math(child);
-               set_fpregs(child, (struct user_i387_struct __user *)data);
-               ret = 0;
+#undef R32
+#undef SEG32
+
+#define R32(l,q)                                                       \
+       case offsetof(struct user32, regs.l):                           \
+               *val = regs->q; break
+
+#define SEG32(rs)                                                      \
+       case offsetof(struct user32, regs.rs):                          \
+               *val = get_segment_reg(child,                           \
+                                      offsetof(struct user_regs_struct, rs)); \
+               break
+
+static int getreg32(struct task_struct *child, unsigned regno, u32 *val)
+{
+       struct pt_regs *regs = task_pt_regs(child);
+
+       switch (regno) {
+
+       SEG32(ds);
+       SEG32(es);
+       SEG32(fs);
+       SEG32(gs);
+
+       R32(cs, cs);
+       R32(ss, ss);
+       R32(ebx, bx);
+       R32(ecx, cx);
+       R32(edx, dx);
+       R32(edi, di);
+       R32(esi, si);
+       R32(ebp, bp);
+       R32(eax, ax);
+       R32(orig_eax, orig_ax);
+       R32(eip, ip);
+       R32(esp, sp);
+
+       case offsetof(struct user32, regs.eflags):
+               *val = get_flags(child);
+               break;
+
+       case offsetof(struct user32, u_debugreg[0]) ...
+               offsetof(struct user32, u_debugreg[7]):
+               regno -= offsetof(struct user32, u_debugreg[0]);
+               *val = ptrace_get_debugreg(child, regno / 4);
+               break;
+
+       default:
+               if (regno > sizeof(struct user32) || (regno & 3))
+                       return -EIO;
+
+               /*
+                * Other dummy fields in the virtual user structure
+                * are ignored
+                */
+               *val = 0;
                break;
        }
+       return 0;
+}
 
-       case PTRACE_GETFPXREGS: { /* Get the child extended FPU state. */
-               if (!access_ok(VERIFY_WRITE, datap,
-                              sizeof(struct user_fxsr_struct))) {
-                       ret = -EIO;
-                       break;
+#undef R32
+#undef SEG32
+
+static int genregs32_get(struct task_struct *target,
+                        const struct user_regset *regset,
+                        unsigned int pos, unsigned int count,
+                        void *kbuf, void __user *ubuf)
+{
+       if (kbuf) {
+               compat_ulong_t *k = kbuf;
+               while (count > 0) {
+                       getreg32(target, pos, k++);
+                       count -= sizeof(*k);
+                       pos += sizeof(*k);
+               }
+       } else {
+               compat_ulong_t __user *u = ubuf;
+               while (count > 0) {
+                       compat_ulong_t word;
+                       getreg32(target, pos, &word);
+                       if (__put_user(word, u++))
+                               return -EFAULT;
+                       count -= sizeof(*u);
+                       pos += sizeof(*u);
                }
-               if (!tsk_used_math(child))
-                       init_fpu(child);
-               ret = get_fpxregs((struct user_fxsr_struct __user *)data, child);
-               break;
        }
 
-       case PTRACE_SETFPXREGS: { /* Set the child extended FPU state. */
-               if (!access_ok(VERIFY_READ, datap,
-                              sizeof(struct user_fxsr_struct))) {
-                       ret = -EIO;
-                       break;
+       return 0;
+}
+
+static int genregs32_set(struct task_struct *target,
+                        const struct user_regset *regset,
+                        unsigned int pos, unsigned int count,
+                        const void *kbuf, const void __user *ubuf)
+{
+       int ret = 0;
+       if (kbuf) {
+               const compat_ulong_t *k = kbuf;
+               while (count > 0 && !ret) {
+                       ret = putreg(target, pos, *k++);
+                       count -= sizeof(*k);
+                       pos += sizeof(*k);
+               }
+       } else {
+               const compat_ulong_t __user *u = ubuf;
+               while (count > 0 && !ret) {
+                       compat_ulong_t word;
+                       ret = __get_user(word, u++);
+                       if (ret)
+                               break;
+                       ret = putreg(target, pos, word);
+                       count -= sizeof(*u);
+                       pos += sizeof(*u);
                }
-               set_stopped_child_used_math(child);
-               ret = set_fpxregs(child, (struct user_fxsr_struct __user *)data);
-               break;
        }
+       return ret;
+}
 
+static long ptrace32_siginfo(unsigned request, u32 pid, u32 addr, u32 data)
+{
+       siginfo_t __user *si = compat_alloc_user_space(sizeof(siginfo_t));
+       compat_siginfo_t __user *si32 = compat_ptr(data);
+       siginfo_t ssi;
+       int ret;
+
+       if (request == PTRACE_SETSIGINFO) {
+               memset(&ssi, 0, sizeof(siginfo_t));
+               ret = copy_siginfo_from_user32(&ssi, si32);
+               if (ret)
+                       return ret;
+               if (copy_to_user(si, &ssi, sizeof(siginfo_t)))
+                       return -EFAULT;
+       }
+       ret = sys_ptrace(request, pid, addr, (unsigned long)si);
+       if (ret)
+               return ret;
+       if (request == PTRACE_GETSIGINFO) {
+               if (copy_from_user(&ssi, si, sizeof(siginfo_t)))
+                       return -EFAULT;
+               ret = copy_siginfo_to_user32(si32, &ssi);
+       }
+       return ret;
+}
+
+asmlinkage long sys32_ptrace(long request, u32 pid, u32 addr, u32 data)
+{
+       struct task_struct *child;
+       struct pt_regs *childregs;
+       void __user *datap = compat_ptr(data);
+       int ret;
+       __u32 val;
+
+       switch (request) {
+       case PTRACE_TRACEME:
+       case PTRACE_ATTACH:
+       case PTRACE_KILL:
+       case PTRACE_CONT:
+       case PTRACE_SINGLESTEP:
+       case PTRACE_SINGLEBLOCK:
+       case PTRACE_DETACH:
+       case PTRACE_SYSCALL:
+       case PTRACE_OLDSETOPTIONS:
+       case PTRACE_SETOPTIONS:
+       case PTRACE_SET_THREAD_AREA:
        case PTRACE_GET_THREAD_AREA:
-               if (addr < 0)
-                       return -EIO;
-               ret = do_get_thread_area(child, addr,
-                                        (struct user_desc __user *) data);
+       case PTRACE_BTS_CONFIG:
+       case PTRACE_BTS_STATUS:
+       case PTRACE_BTS_SIZE:
+       case PTRACE_BTS_GET:
+       case PTRACE_BTS_CLEAR:
+       case PTRACE_BTS_DRAIN:
+               return sys_ptrace(request, pid, addr, data);
+
+       default:
+               return -EINVAL;
+
+       case PTRACE_PEEKTEXT:
+       case PTRACE_PEEKDATA:
+       case PTRACE_POKEDATA:
+       case PTRACE_POKETEXT:
+       case PTRACE_POKEUSR:
+       case PTRACE_PEEKUSR:
+       case PTRACE_GETREGS:
+       case PTRACE_SETREGS:
+       case PTRACE_SETFPREGS:
+       case PTRACE_GETFPREGS:
+       case PTRACE_SETFPXREGS:
+       case PTRACE_GETFPXREGS:
+       case PTRACE_GETEVENTMSG:
                break;
 
-       case PTRACE_SET_THREAD_AREA:
-               if (addr < 0)
-                       return -EIO;
-               ret = do_set_thread_area(child, addr,
-                                        (struct user_desc __user *) data, 0);
+       case PTRACE_SETSIGINFO:
+       case PTRACE_GETSIGINFO:
+               return ptrace32_siginfo(request, pid, addr, data);
+       }
+
+       child = ptrace_get_task_struct(pid);
+       if (IS_ERR(child))
+               return PTR_ERR(child);
+
+       ret = ptrace_check_attach(child, request == PTRACE_KILL);
+       if (ret < 0)
+               goto out;
+
+       childregs = task_pt_regs(child);
+
+       switch (request) {
+       case PTRACE_PEEKUSR:
+               ret = getreg32(child, addr, &val);
+               if (ret == 0)
+                       ret = put_user(val, (__u32 __user *)datap);
                break;
 
-       default:
-               ret = ptrace_request(child, request, addr, data);
+       case PTRACE_POKEUSR:
+               ret = putreg32(child, addr, data);
                break;
+
+       case PTRACE_GETREGS:    /* Get all gp regs from the child. */
+               return copy_regset_to_user(child, &user_x86_32_view,
+                                          REGSET_GENERAL,
+                                          0, sizeof(struct user_regs_struct32),
+                                          datap);
+
+       case PTRACE_SETREGS:    /* Set all gp regs in the child. */
+               return copy_regset_from_user(child, &user_x86_32_view,
+                                            REGSET_GENERAL, 0,
+                                            sizeof(struct user_regs_struct32),
+                                            datap);
+
+       case PTRACE_GETFPREGS:  /* Get the child FPU state. */
+               return copy_regset_to_user(child, &user_x86_32_view,
+                                          REGSET_FP, 0,
+                                          sizeof(struct user_i387_ia32_struct),
+                                          datap);
+
+       case PTRACE_SETFPREGS:  /* Set the child FPU state. */
+               return copy_regset_from_user(
+                       child, &user_x86_32_view, REGSET_FP,
+                       0, sizeof(struct user_i387_ia32_struct), datap);
+
+       case PTRACE_GETFPXREGS: /* Get the child extended FPU state. */
+               return copy_regset_to_user(child, &user_x86_32_view,
+                                          REGSET_XFP, 0,
+                                          sizeof(struct user32_fxsr_struct),
+                                          datap);
+
+       case PTRACE_SETFPXREGS: /* Set the child extended FPU state. */
+               return copy_regset_from_user(child, &user_x86_32_view,
+                                            REGSET_XFP, 0,
+                                            sizeof(struct user32_fxsr_struct),
+                                            datap);
+
+       default:
+               return compat_ptrace_request(child, request, addr, data);
        }
 
+ out:
+       put_task_struct(child);
        return ret;
 }
 
+#endif /* CONFIG_IA32_EMULATION */
+
+#ifdef CONFIG_X86_64
+
+static const struct user_regset x86_64_regsets[] = {
+       [REGSET_GENERAL] = {
+               .core_note_type = NT_PRSTATUS,
+               .n = sizeof(struct user_regs_struct) / sizeof(long),
+               .size = sizeof(long), .align = sizeof(long),
+               .get = genregs_get, .set = genregs_set
+       },
+       [REGSET_FP] = {
+               .core_note_type = NT_PRFPREG,
+               .n = sizeof(struct user_i387_struct) / sizeof(long),
+               .size = sizeof(long), .align = sizeof(long),
+               .active = xfpregs_active, .get = xfpregs_get, .set = xfpregs_set
+       },
+};
+
+static const struct user_regset_view user_x86_64_view = {
+       .name = "x86_64", .e_machine = EM_X86_64,
+       .regsets = x86_64_regsets, .n = ARRAY_SIZE(x86_64_regsets)
+};
+
+#else  /* CONFIG_X86_32 */
+
+#define user_regs_struct32     user_regs_struct
+#define genregs32_get          genregs_get
+#define genregs32_set          genregs_set
+
+#endif /* CONFIG_X86_64 */
+
+#if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
+static const struct user_regset x86_32_regsets[] = {
+       [REGSET_GENERAL] = {
+               .core_note_type = NT_PRSTATUS,
+               .n = sizeof(struct user_regs_struct32) / sizeof(u32),
+               .size = sizeof(u32), .align = sizeof(u32),
+               .get = genregs32_get, .set = genregs32_set
+       },
+       [REGSET_FP] = {
+               .core_note_type = NT_PRFPREG,
+               .n = sizeof(struct user_i387_struct) / sizeof(u32),
+               .size = sizeof(u32), .align = sizeof(u32),
+               .active = fpregs_active, .get = fpregs_get, .set = fpregs_set
+       },
+       [REGSET_XFP] = {
+               .core_note_type = NT_PRXFPREG,
+               .n = sizeof(struct user_i387_struct) / sizeof(u32),
+               .size = sizeof(u32), .align = sizeof(u32),
+               .active = xfpregs_active, .get = xfpregs_get, .set = xfpregs_set
+       },
+       [REGSET_TLS] = {
+               .core_note_type = NT_386_TLS,
+               .n = GDT_ENTRY_TLS_ENTRIES, .bias = GDT_ENTRY_TLS_MIN,
+               .size = sizeof(struct user_desc),
+               .align = sizeof(struct user_desc),
+               .active = regset_tls_active,
+               .get = regset_tls_get, .set = regset_tls_set
+       },
+};
+
+static const struct user_regset_view user_x86_32_view = {
+       .name = "i386", .e_machine = EM_386,
+       .regsets = x86_32_regsets, .n = ARRAY_SIZE(x86_32_regsets)
+};
+#endif
+
+const struct user_regset_view *task_user_regset_view(struct task_struct *task)
+{
+#ifdef CONFIG_IA32_EMULATION
+       if (test_tsk_thread_flag(task, TIF_IA32))
+#endif
+#if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
+               return &user_x86_32_view;
+#endif
+#ifdef CONFIG_X86_64
+       return &user_x86_64_view;
+#endif
+}
+
+#ifdef CONFIG_X86_32
+
 void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code)
 {
        struct siginfo info;
@@ -517,3 +1502,65 @@ out:
                audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax);
        return 1;
 }
+
+#else  /* CONFIG_X86_64 */
+
+static void syscall_trace(struct pt_regs *regs)
+{
+
+#if 0
+       printk("trace %s ip %lx sp %lx ax %d origrax %d caller %lx tiflags %x ptrace %x\n",
+              current->comm,
+              regs->ip, regs->sp, regs->ax, regs->orig_ax, __builtin_return_address(0),
+              current_thread_info()->flags, current->ptrace);
+#endif
+
+       ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD)
+                               ? 0x80 : 0));
+       /*
+        * this isn't the same as continuing with a signal, but it will do
+        * for normal use.  strace only continues with a signal if the
+        * stopping signal is not SIGTRAP.  -brl
+        */
+       if (current->exit_code) {
+               send_sig(current->exit_code, current, 1);
+               current->exit_code = 0;
+       }
+}
+
+asmlinkage void syscall_trace_enter(struct pt_regs *regs)
+{
+       /* do the secure computing check first */
+       secure_computing(regs->orig_ax);
+
+       if (test_thread_flag(TIF_SYSCALL_TRACE)
+           && (current->ptrace & PT_PTRACED))
+               syscall_trace(regs);
+
+       if (unlikely(current->audit_context)) {
+               if (test_thread_flag(TIF_IA32)) {
+                       audit_syscall_entry(AUDIT_ARCH_I386,
+                                           regs->orig_ax,
+                                           regs->bx, regs->cx,
+                                           regs->dx, regs->si);
+               } else {
+                       audit_syscall_entry(AUDIT_ARCH_X86_64,
+                                           regs->orig_ax,
+                                           regs->di, regs->si,
+                                           regs->dx, regs->r10);
+               }
+       }
+}
+
+asmlinkage void syscall_trace_leave(struct pt_regs *regs)
+{
+       if (unlikely(current->audit_context))
+               audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax);
+
+       if ((test_thread_flag(TIF_SYSCALL_TRACE)
+            || test_thread_flag(TIF_SINGLESTEP))
+           && (current->ptrace & PT_PTRACED))
+               syscall_trace(regs);
+}
+
+#endif /* CONFIG_X86_32 */