Merge branch 'ioat' into dmaengine

35 files changed, 422 insertions, 460 deletions

diff --git a/arch/microblaze/Kconfig b/arch/microblaze/Kconfig
index 203ec61c6d4c..76818f926539 100644
--- a/arch/microblaze/Kconfig
+++ b/arch/microblaze/Kconfig
@@ -75,9 +75,6 @@ config LOCKDEP_SUPPORT
 config HAVE_LATENCYTOP_SUPPORT
         def_bool y
 
-config PCI
-        def_bool n
-
 config DTC
         def_bool y
 
diff --git a/arch/microblaze/Makefile b/arch/microblaze/Makefile
index 836832dd9b26..72f6e8583746 100644
--- a/arch/microblaze/Makefile
+++ b/arch/microblaze/Makefile
@@ -84,7 +84,7 @@ define archhelp
   echo '* linux.bin    - Create raw binary'
   echo '  linux.bin.gz - Create compressed raw binary'
   echo '  simpleImage.<dt> - ELF image with $(arch)/boot/dts/<dt>.dts linked in'
-  echo '                   - stripped elf with fdt blob
+  echo '                   - stripped elf with fdt blob'
   echo '  simpleImage.<dt>.unstrip - full ELF image with fdt blob'
   echo '  *_defconfig      - Select default config from arch/microblaze/configs'
   echo ''
@@ -94,3 +94,5 @@ define archhelp
   echo '  name of a dts file from the arch/microblaze/boot/dts/ directory'
   echo '  (minus the .dts extension).'
 endef
+
+MRPROPER_FILES += $(boot)/simpleImage.*
diff --git a/arch/microblaze/boot/Makefile b/arch/microblaze/boot/Makefile
index 902cf9846c3c..57f50c2371c6 100644
--- a/arch/microblaze/boot/Makefile
+++ b/arch/microblaze/boot/Makefile
@@ -23,8 +23,6 @@ $(obj)/system.dtb: $(obj)/$(DTB).dtb
 endif
 
 $(obj)/linux.bin: vmlinux FORCE
-        [ -n $(CONFIG_INITRAMFS_SOURCE) ] && [ ! -e $(CONFIG_INITRAMFS_SOURCE) ] && \
-        touch $(CONFIG_INITRAMFS_SOURCE) || echo "No CPIO image"
         $(call if_changed,objcopy)
         $(call if_changed,uimage)
         @echo 'Kernel: $@ is ready' ' (#'`cat .version`')'
@@ -62,6 +60,4 @@ quiet_cmd_dtc = DTC     $@
 $(obj)/%.dtb: $(dtstree)/%.dts FORCE
         $(call if_changed,dtc)
 
-clean-kernel += linux.bin linux.bin.gz simpleImage.*
-
-clean-files += *.dtb simpleImage.*.unstrip
+clean-files += *.dtb simpleImage.*.unstrip linux.bin.ub
diff --git a/arch/microblaze/include/asm/futex.h b/arch/microblaze/include/asm/futex.h
index 8dbb6e7a03a2..ad3fd61b2fe7 100644
--- a/arch/microblaze/include/asm/futex.h
+++ b/arch/microblaze/include/asm/futex.h
@@ -55,7 +55,7 @@ futex_atomic_op_inuser(int encoded_op, int __user *uaddr)
                 __futex_atomic_op("or %1,%0,%4;", ret, oldval, uaddr, oparg);
                 break;
         case FUTEX_OP_ANDN:
-                __futex_atomic_op("and %1,%0,%4;", ret, oldval, uaddr, oparg);
+                __futex_atomic_op("andn %1,%0,%4;", ret, oldval, uaddr, oparg);
                 break;
         case FUTEX_OP_XOR:
                 __futex_atomic_op("xor %1,%0,%4;", ret, oldval, uaddr, oparg);
diff --git a/arch/microblaze/include/asm/io.h b/arch/microblaze/include/asm/io.h
index 32d621a56aee..e45a6eea92e0 100644
--- a/arch/microblaze/include/asm/io.h
+++ b/arch/microblaze/include/asm/io.h
@@ -108,6 +108,11 @@ static inline void writel(unsigned int v, volatile void __iomem *addr)
 #define iowrite16(v, addr)      __raw_writew((u16)(v), (u16 *)(addr))
 #define iowrite32(v, addr)      __raw_writel((u32)(v), (u32 *)(addr))
 
+#define ioread16be(addr)        __raw_readw((u16 *)(addr))
+#define ioread32be(addr)        __raw_readl((u32 *)(addr))
+#define iowrite16be(v, addr)    __raw_writew((u16)(v), (u16 *)(addr))
+#define iowrite32be(v, addr)    __raw_writel((u32)(v), (u32 *)(addr))
+
 /* These are the definitions for the x86 IO instructions
  * inb/inw/inl/outb/outw/outl, the "string" versions
  * insb/insw/insl/outsb/outsw/outsl, and the "pausing" versions
diff --git a/arch/microblaze/include/asm/processor.h b/arch/microblaze/include/asm/processor.h
index 563c6b9453f0..8eeb09211ece 100644
--- a/arch/microblaze/include/asm/processor.h
+++ b/arch/microblaze/include/asm/processor.h
@@ -14,7 +14,6 @@
 #include <asm/ptrace.h>
 #include <asm/setup.h>
 #include <asm/registers.h>
-#include <asm/segment.h>
 #include <asm/entry.h>
 #include <asm/current.h>
 
diff --git a/arch/microblaze/include/asm/segment.h b/arch/microblaze/include/asm/segment.h
deleted file mode 100644
index 0e7102c3fb11..000000000000
--- a/arch/microblaze/include/asm/segment.h
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (C) 2008-2009 Michal Simek <monstr@monstr.eu>
- * Copyright (C) 2008-2009 PetaLogix
- * Copyright (C) 2006 Atmark Techno, Inc.
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License. See the file "COPYING" in the main directory of this archive
- * for more details.
- */
-
-#ifndef _ASM_MICROBLAZE_SEGMENT_H
-#define _ASM_MICROBLAZE_SEGMENT_H
-
-# ifndef __ASSEMBLY__
-
-typedef struct {
-        unsigned long seg;
-} mm_segment_t;
-
-/*
- * On Microblaze the fs value is actually the top of the corresponding
- * address space.
- *
- * The fs value determines whether argument validity checking should be
- * performed or not. If get_fs() == USER_DS, checking is performed, with
- * get_fs() == KERNEL_DS, checking is bypassed.
- *
- * For historical reasons, these macros are grossly misnamed.
- *
- * For non-MMU arch like Microblaze, KERNEL_DS and USER_DS is equal.
- */
-# define MAKE_MM_SEG(s)       ((mm_segment_t) { (s) })
-
-#  ifndef CONFIG_MMU
-#  define KERNEL_DS     MAKE_MM_SEG(0)
-#  define USER_DS       KERNEL_DS
-#  else
-#  define KERNEL_DS     MAKE_MM_SEG(0xFFFFFFFF)
-#  define USER_DS       MAKE_MM_SEG(TASK_SIZE - 1)
-#  endif
-
-# define get_ds()       (KERNEL_DS)
-# define get_fs()       (current_thread_info()->addr_limit)
-# define set_fs(val)    (current_thread_info()->addr_limit = (val))
-
-# define segment_eq(a, b)       ((a).seg == (b).seg)
-
-# endif /* __ASSEMBLY__ */
-#endif /* _ASM_MICROBLAZE_SEGMENT_H */
diff --git a/arch/microblaze/include/asm/thread_info.h b/arch/microblaze/include/asm/thread_info.h
index 6e92885d381a..b2ca80f64640 100644
--- a/arch/microblaze/include/asm/thread_info.h
+++ b/arch/microblaze/include/asm/thread_info.h
@@ -19,7 +19,6 @@
 #ifndef __ASSEMBLY__
 # include <linux/types.h>
 # include <asm/processor.h>
-# include <asm/segment.h>
 
 /*
  * low level task data that entry.S needs immediate access to
@@ -60,6 +59,10 @@ struct cpu_context {
         __u32   fsr;
 };
 
+typedef struct {
+        unsigned long seg;
+} mm_segment_t;
+
 struct thread_info {
         struct task_struct      *task; /* main task structure */
         struct exec_domain      *exec_domain; /* execution domain */
diff --git a/arch/microblaze/include/asm/tlbflush.h b/arch/microblaze/include/asm/tlbflush.h
index bcb8b41d55af..2e1353c2d18d 100644
--- a/arch/microblaze/include/asm/tlbflush.h
+++ b/arch/microblaze/include/asm/tlbflush.h
@@ -24,6 +24,7 @@ extern void _tlbie(unsigned long address);
 extern void _tlbia(void);
 
 #define __tlbia()       { preempt_disable(); _tlbia(); preempt_enable(); }
+#define __tlbie(x)      { _tlbie(x); }
 
 static inline void local_flush_tlb_all(void)
         { __tlbia(); }
@@ -31,7 +32,7 @@ static inline void local_flush_tlb_mm(struct mm_struct *mm)
         { __tlbia(); }
 static inline void local_flush_tlb_page(struct vm_area_struct *vma,
                                 unsigned long vmaddr)
-        { _tlbie(vmaddr); }
+        { __tlbie(vmaddr); }
 static inline void local_flush_tlb_range(struct vm_area_struct *vma,
                 unsigned long start, unsigned long end)
         { __tlbia(); }
diff --git a/arch/microblaze/include/asm/uaccess.h b/arch/microblaze/include/asm/uaccess.h
index 371bd6e56d9a..446bec29b142 100644
--- a/arch/microblaze/include/asm/uaccess.h
+++ b/arch/microblaze/include/asm/uaccess.h
@@ -22,101 +22,73 @@
 #include <asm/mmu.h>
 #include <asm/page.h>
 #include <asm/pgtable.h>
-#include <asm/segment.h>
 #include <linux/string.h>
 
 #define VERIFY_READ     0
 #define VERIFY_WRITE    1
 
-#define __clear_user(addr, n)   (memset((void *)(addr), 0, (n)), 0)
-
-#ifndef CONFIG_MMU
-
-extern int ___range_ok(unsigned long addr, unsigned long size);
-
-#define __range_ok(addr, size) \
-                ___range_ok((unsigned long)(addr), (unsigned long)(size))
-
-#define access_ok(type, addr, size) (__range_ok((addr), (size)) == 0)
-#define __access_ok(add, size) (__range_ok((addr), (size)) == 0)
-
-/* Undefined function to trigger linker error */
-extern int bad_user_access_length(void);
-
-/* FIXME this is function for optimalization -> memcpy */
-#define __get_user(var, ptr)                            \
-({                                                      \
-        int __gu_err = 0;                               \
-        switch (sizeof(*(ptr))) {                       \
-        case 1:                                         \
-        case 2:                                         \
-        case 4:                                         \
-                (var) = *(ptr);                         \
-                break;                                  \
-        case 8:                                         \
-                memcpy((void *) &(var), (ptr), 8);      \
-                break;                                  \
-        default:                                        \
-                (var) = 0;                              \
-                __gu_err = __get_user_bad();            \
-                break;                                  \
-        }                                               \
-        __gu_err;                                       \
-})
+/*
+ * On Microblaze the fs value is actually the top of the corresponding
+ * address space.
+ *
+ * The fs value determines whether argument validity checking should be
+ * performed or not. If get_fs() == USER_DS, checking is performed, with
+ * get_fs() == KERNEL_DS, checking is bypassed.
+ *
+ * For historical reasons, these macros are grossly misnamed.
+ *
+ * For non-MMU arch like Microblaze, KERNEL_DS and USER_DS is equal.
+ */
+# define MAKE_MM_SEG(s)       ((mm_segment_t) { (s) })
 
-#define __get_user_bad()        (bad_user_access_length(), (-EFAULT))
+#  ifndef CONFIG_MMU
+#  define KERNEL_DS     MAKE_MM_SEG(0)
+#  define USER_DS       KERNEL_DS
+#  else
+#  define KERNEL_DS     MAKE_MM_SEG(0xFFFFFFFF)
+#  define USER_DS       MAKE_MM_SEG(TASK_SIZE - 1)
+#  endif
 
-/* FIXME is not there defined __pu_val */
-#define __put_user(var, ptr)                                    \
-({                                                              \
-        int __pu_err = 0;                                       \
-        switch (sizeof(*(ptr))) {                               \
-        case 1:                                                 \
-        case 2:                                                 \
-        case 4:                                                 \
-                *(ptr) = (var);                                 \
-                break;                                          \
-        case 8: {                                               \
-                typeof(*(ptr)) __pu_val = (var);                \
-                memcpy(ptr, &__pu_val, sizeof(__pu_val));       \
-                }                                               \
-                break;                                          \
-        default:                                                \
-                __pu_err = __put_user_bad();                    \
-                break;                                          \
-        }                                                       \
-        __pu_err;                                               \
-})
+# define get_ds()       (KERNEL_DS)
+# define get_fs()       (current_thread_info()->addr_limit)
+# define set_fs(val)    (current_thread_info()->addr_limit = (val))
 
-#define __put_user_bad()        (bad_user_access_length(), (-EFAULT))
+# define segment_eq(a, b)       ((a).seg == (b).seg)
 
-#define put_user(x, ptr)        __put_user((x), (ptr))
-#define get_user(x, ptr)        __get_user((x), (ptr))
+/*
+ * The exception table consists of pairs of addresses: the first is the
+ * address of an instruction that is allowed to fault, and the second is
+ * the address at which the program should continue. No registers are
+ * modified, so it is entirely up to the continuation code to figure out
+ * what to do.
+ *
+ * All the routines below use bits of fixup code that are out of line
+ * with the main instruction path. This means when everything is well,
+ * we don't even have to jump over them. Further, they do not intrude
+ * on our cache or tlb entries.
+ */
+struct exception_table_entry {
+        unsigned long insn, fixup;
+};
 
-#define copy_to_user(to, from, n)       (memcpy((to), (from), (n)), 0)
-#define copy_from_user(to, from, n)     (memcpy((to), (from), (n)), 0)
+/* Returns 0 if exception not found and fixup otherwise.  */
+extern unsigned long search_exception_table(unsigned long);
 
-#define __copy_to_user(to, from, n)     (copy_to_user((to), (from), (n)))
-#define __copy_from_user(to, from, n)   (copy_from_user((to), (from), (n)))
-#define __copy_to_user_inatomic(to, from, n) \
-                        (__copy_to_user((to), (from), (n)))
-#define __copy_from_user_inatomic(to, from, n) \
-                        (__copy_from_user((to), (from), (n)))
+#ifndef CONFIG_MMU
 
-static inline unsigned long clear_user(void *addr, unsigned long size)
+/* Check against bounds of physical memory */
+static inline int ___range_ok(unsigned long addr, unsigned long size)
 {
-        if (access_ok(VERIFY_WRITE, addr, size))
-                size = __clear_user(addr, size);
-        return size;
+        return ((addr < memory_start) ||
+                ((addr + size) > memory_end));
 }
 
-/* Returns 0 if exception not found and fixup otherwise.  */
-extern unsigned long search_exception_table(unsigned long);
+#define __range_ok(addr, size) \
+                ___range_ok((unsigned long)(addr), (unsigned long)(size))
 
-extern long strncpy_from_user(char *dst, const char *src, long count);
-extern long strnlen_user(const char *src, long count);
+#define access_ok(type, addr, size) (__range_ok((addr), (size)) == 0)
 
-#else /* CONFIG_MMU */
+#else
 
 /*
  * Address is valid if:
@@ -129,24 +101,88 @@ extern long strnlen_user(const char *src, long count);
 /* || printk("access_ok failed for %s at 0x%08lx (size %d), seg 0x%08x\n",
  type?"WRITE":"READ",addr,size,get_fs().seg)) */
 
-/*
- * All the __XXX versions macros/functions below do not perform
- * access checking. It is assumed that the necessary checks have been
- * already performed before the finction (macro) is called.
- */
+#endif
 
-#define get_user(x, ptr)                                                \
-({                                                                      \
-        access_ok(VERIFY_READ, (ptr), sizeof(*(ptr)))                   \
-                ? __get_user((x), (ptr)) : -EFAULT;                     \
-})
+#ifdef CONFIG_MMU
+# define __FIXUP_SECTION        ".section .fixup,\"ax\"\n"
+# define __EX_TABLE_SECTION     ".section __ex_table,\"a\"\n"
+#else
+# define __FIXUP_SECTION        ".section .discard,\"ax\"\n"
+# define __EX_TABLE_SECTION     ".section .discard,\"a\"\n"
+#endif
 
-#define put_user(x, ptr)                                                \
-({                                                                      \
-        access_ok(VERIFY_WRITE, (ptr), sizeof(*(ptr)))                  \
-                ? __put_user((x), (ptr)) : -EFAULT;                     \
+extern unsigned long __copy_tofrom_user(void __user *to,
+                const void __user *from, unsigned long size);
+
+/* Return: number of not copied bytes, i.e. 0 if OK or non-zero if fail. */
+static inline unsigned long __must_check __clear_user(void __user *to,
+                                                        unsigned long n)
+{
+        /* normal memset with two words to __ex_table */
+        __asm__ __volatile__ (                          \
+                        "1:     sb      r0, %2, r0;"    \
+                        "       addik   %0, %0, -1;"    \
+                        "       bneid   %0, 1b;"        \
+                        "       addik   %2, %2, 1;"     \
+                        "2:                     "       \
+                        __EX_TABLE_SECTION              \
+                        ".word  1b,2b;"                 \
+                        ".previous;"                    \
+                : "=r"(n)                               \
+                : "0"(n), "r"(to)
+        );
+        return n;
+}
+
+static inline unsigned long __must_check clear_user(void __user *to,
+                                                        unsigned long n)
+{
+        might_sleep();
+        if (unlikely(!access_ok(VERIFY_WRITE, to, n)))
+                return n;
+
+        return __clear_user(to, n);
+}
+
+/* put_user and get_user macros */
+extern long __user_bad(void);
+
+#define __get_user_asm(insn, __gu_ptr, __gu_val, __gu_err)      \
+({                                                              \
+        __asm__ __volatile__ (                                  \
+                        "1:"    insn    " %1, %2, r0;"          \
+                        "       addk    %0, r0, r0;"            \
+                        "2:                     "               \
+                        __FIXUP_SECTION                         \
+                        "3:     brid    2b;"                    \
+                        "       addik   %0, r0, %3;"            \
+                        ".previous;"                            \
+                        __EX_TABLE_SECTION                      \
+                        ".word  1b,3b;"                         \
+                        ".previous;"                            \
+                : "=&r"(__gu_err), "=r"(__gu_val)               \
+                : "r"(__gu_ptr), "i"(-EFAULT)                   \
+        );                                                      \
 })
 
+/**
+ * get_user: - Get a simple variable from user space.
+ * @x:   Variable to store result.
+ * @ptr: Source address, in user space.
+ *
+ * Context: User context only.  This function may sleep.
+ *
+ * This macro copies a single simple variable from user space to kernel
+ * space.  It supports simple types like char and int, but not larger
+ * data types like structures or arrays.
+ *
+ * @ptr must have pointer-to-simple-variable type, and the result of
+ * dereferencing @ptr must be assignable to @x without a cast.
+ *
+ * Returns zero on success, or -EFAULT on error.
+ * On error, the variable @x is set to zero.
+ */
+
 #define __get_user(x, ptr)                                              \
 ({                                                                      \
         unsigned long __gu_val;                                         \
@@ -163,30 +199,74 @@ extern long strnlen_user(const char *src, long count);
                 __get_user_asm("lw", (ptr), __gu_val, __gu_err);        \
                 break;                                                  \
         default:                                                        \
-                __gu_val = 0; __gu_err = -EINVAL;                       \
+                /* __gu_val = 0; __gu_err = -EINVAL;*/ __gu_err = __user_bad();\
         }                                                               \
         x = (__typeof__(*(ptr))) __gu_val;                              \
         __gu_err;                                                       \
 })
 
-#define __get_user_asm(insn, __gu_ptr, __gu_val, __gu_err)              \
+
+#define get_user(x, ptr)                                                \
 ({                                                                      \
-        __asm__ __volatile__ (                                          \
-                        "1:"    insn    " %1, %2, r0;                   \
-                                addk    %0, r0, r0;                     \
-                        2:                                              \
-                        .section .fixup,\"ax\";                         \
-                        3:      brid    2b;                             \
-                                addik   %0, r0, %3;                     \
-                        .previous;                                      \
-                        .section __ex_table,\"a\";                      \
-                        .word   1b,3b;                                  \
-                        .previous;"                                     \
-                : "=r"(__gu_err), "=r"(__gu_val)                        \
-                : "r"(__gu_ptr), "i"(-EFAULT)                           \
-        );                                                              \
+        access_ok(VERIFY_READ, (ptr), sizeof(*(ptr)))                   \
+                ? __get_user((x), (ptr)) : -EFAULT;                     \
+})
+
+#define __put_user_asm(insn, __gu_ptr, __gu_val, __gu_err)      \
+({                                                              \
+        __asm__ __volatile__ (                                  \
+                        "1:"    insn    " %1, %2, r0;"          \
+                        "       addk    %0, r0, r0;"            \
+                        "2:                     "               \
+                        __FIXUP_SECTION                         \
+                        "3:     brid    2b;"                    \
+                        "       addik   %0, r0, %3;"            \
+                        ".previous;"                            \
+                        __EX_TABLE_SECTION                      \
+                        ".word  1b,3b;"                         \
+                        ".previous;"                            \
+                : "=&r"(__gu_err)                               \
+                : "r"(__gu_val), "r"(__gu_ptr), "i"(-EFAULT)    \
+        );                                                      \
 })
 
+#define __put_user_asm_8(__gu_ptr, __gu_val, __gu_err)          \
+({                                                              \
+        __asm__ __volatile__ (" lwi     %0, %1, 0;"             \
+                        "1:     swi     %0, %2, 0;"             \
+                        "       lwi     %0, %1, 4;"             \
+                        "2:     swi     %0, %2, 4;"             \
+                        "       addk    %0, r0, r0;"            \
+                        "3:                     "               \
+                        __FIXUP_SECTION                         \
+                        "4:     brid    3b;"                    \
+                        "       addik   %0, r0, %3;"            \
+                        ".previous;"                            \
+                        __EX_TABLE_SECTION                      \
+                        ".word  1b,4b,2b,4b;"                   \
+                        ".previous;"                            \
+                : "=&r"(__gu_err)                               \
+                : "r"(&__gu_val), "r"(__gu_ptr), "i"(-EFAULT)   \
+                );                                              \
+})
+
+/**
+ * put_user: - Write a simple value into user space.
+ * @x:   Value to copy to user space.
+ * @ptr: Destination address, in user space.
+ *
+ * Context: User context only.  This function may sleep.
+ *
+ * This macro copies a single simple value from kernel space to user
+ * space.  It supports simple types like char and int, but not larger
+ * data types like structures or arrays.
+ *
+ * @ptr must have pointer-to-simple-variable type, and @x must be assignable
+ * to the result of dereferencing @ptr.
+ *
+ * Returns zero on success, or -EFAULT on error.
+ */
+
 #define __put_user(x, ptr)                                              \
 ({                                                                      \
         __typeof__(*(ptr)) volatile __gu_val = (x);                     \
@@ -195,7 +275,7 @@ extern long strnlen_user(const char *src, long count);
         case 1:                                                         \
                 __put_user_asm("sb", (ptr), __gu_val, __gu_err);        \
                 break;                                                  \
-        case 2:                                                         \
+        case 2:                                                         \
                 __put_user_asm("sh", (ptr), __gu_val, __gu_err);        \
                 break;                                                  \
         case 4:                                                         \
@@ -205,121 +285,82 @@ extern long strnlen_user(const char *src, long count);
                 __put_user_asm_8((ptr), __gu_val, __gu_err);            \
                 break;                                                  \
         default:                                                        \
-                __gu_err = -EINVAL;                                     \
+                /*__gu_err = -EINVAL;*/ __gu_err = __user_bad();        \
         }                                                               \
         __gu_err;                                                       \
 })
 
-#define __put_user_asm_8(__gu_ptr, __gu_val, __gu_err)  \
-({                                                      \
-__asm__ __volatile__ (" lwi     %0, %1, 0;              \
-                1:      swi     %0, %2, 0;              \
-                        lwi     %0, %1, 4;              \
-                2:      swi     %0, %2, 4;              \
-                        addk    %0,r0,r0;               \
-                3:                                      \
-                .section .fixup,\"ax\";                 \
-                4:      brid    3b;                     \
-                        addik   %0, r0, %3;             \
-                .previous;                              \
-                .section __ex_table,\"a\";              \
-                .word   1b,4b,2b,4b;                    \
-                .previous;"                             \
-        : "=&r"(__gu_err)                               \
-        : "r"(&__gu_val),                               \
-        "r"(__gu_ptr), "i"(-EFAULT)                     \
-        );                                              \
-})
+#ifndef CONFIG_MMU
 
-#define __put_user_asm(insn, __gu_ptr, __gu_val, __gu_err)      \
-({                                                              \
-        __asm__ __volatile__ (                                  \
-                        "1:"    insn    " %1, %2, r0;           \
-                                addk    %0, r0, r0;             \
-                        2:                                      \
-                        .section .fixup,\"ax\";                 \
-                        3:      brid    2b;                     \
-                                addik   %0, r0, %3;             \
-                        .previous;                              \
-                        .section __ex_table,\"a\";              \
-                        .word   1b,3b;                          \
-                        .previous;"                             \
-                : "=r"(__gu_err)                                \
-                : "r"(__gu_val), "r"(__gu_ptr), "i"(-EFAULT)    \
-        );                                                      \
-})
+#define put_user(x, ptr)        __put_user((x), (ptr))
 
-/*
- * Return: number of not copied bytes, i.e. 0 if OK or non-zero if fail.
- */
-static inline int clear_user(char *to, int size)
-{
-        if (size && access_ok(VERIFY_WRITE, to, size)) {
-                __asm__ __volatile__ ("                         \
-                                1:                              \
-                                        sb      r0, %2, r0;     \
-                                        addik   %0, %0, -1;     \
-                                        bneid   %0, 1b;         \
-                                        addik   %2, %2, 1;      \
-                                2:                              \
-                                .section __ex_table,\"a\";      \
-                                .word   1b,2b;                  \
-                                .section .text;"                \
-                        : "=r"(size)                            \
-                        : "0"(size), "r"(to)
-                );
-        }
-        return size;
-}
+#else /* CONFIG_MMU */
 
-#define __copy_from_user(to, from, n)   copy_from_user((to), (from), (n))
+#define put_user(x, ptr)                                                \
+({                                                                      \
+        access_ok(VERIFY_WRITE, (ptr), sizeof(*(ptr)))                  \
+                ? __put_user((x), (ptr)) : -EFAULT;                     \
+})
+#endif /* CONFIG_MMU */
+
+/* copy_to_from_user */
+#define __copy_from_user(to, from, n)   \
+        __copy_tofrom_user((__force void __user *)(to), \
+                                (void __user *)(from), (n))
 #define __copy_from_user_inatomic(to, from, n) \
                 copy_from_user((to), (from), (n))
 
-#define copy_to_user(to, from, n)                                       \
-        (access_ok(VERIFY_WRITE, (to), (n)) ?                           \
-                __copy_tofrom_user((void __user *)(to),                 \
-                        (__force const void __user *)(from), (n))       \
-                : -EFAULT)
+static inline long copy_from_user(void *to,
+                const void __user *from, unsigned long n)
+{
+        might_sleep();
+        if (access_ok(VERIFY_READ, from, n))
+                return __copy_from_user(to, from, n);
+        return n;
+}
 
-#define __copy_to_user(to, from, n)     copy_to_user((to), (from), (n))
+#define __copy_to_user(to, from, n)     \
+                __copy_tofrom_user((void __user *)(to), \
+                        (__force const void __user *)(from), (n))
 #define __copy_to_user_inatomic(to, from, n)    copy_to_user((to), (from), (n))
 
-#define copy_from_user(to, from, n)                                     \
-        (access_ok(VERIFY_READ, (from), (n)) ?                          \
-                __copy_tofrom_user((__force void __user *)(to),         \
-                        (void __user *)(from), (n))                     \
-                : -EFAULT)
+static inline long copy_to_user(void __user *to,
+                const void *from, unsigned long n)
+{
+        might_sleep();
+        if (access_ok(VERIFY_WRITE, to, n))
+                return __copy_to_user(to, from, n);
+        return n;
+}
 
+/*
+ * Copy a null terminated string from userspace.
+ */
 extern int __strncpy_user(char *to, const char __user *from, int len);
-extern int __strnlen_user(const char __user *sstr, int len);
 
-#define strncpy_from_user(to, from, len)        \
-                (access_ok(VERIFY_READ, from, 1) ?      \
-                        __strncpy_user(to, from, len) : -EFAULT)
-#define strnlen_user(str, len)  \
-                (access_ok(VERIFY_READ, str, 1) ? __strnlen_user(str, len) : 0)
+#define __strncpy_from_user     __strncpy_user
 
-#endif /* CONFIG_MMU */
-
-extern unsigned long __copy_tofrom_user(void __user *to,
-                const void __user *from, unsigned long size);
+static inline long
+strncpy_from_user(char *dst, const char __user *src, long count)
+{
+        if (!access_ok(VERIFY_READ, src, 1))
+                return -EFAULT;
+        return __strncpy_from_user(dst, src, count);
+}
 
 /*
- * The exception table consists of pairs of addresses: the first is the
- * address of an instruction that is allowed to fault, and the second is
- * the address at which the program should continue. No registers are
- * modified, so it is entirely up to the continuation code to figure out
- * what to do.
+ * Return the size of a string (including the ending 0)
  *
- * All the routines below use bits of fixup code that are out of line
- * with the main instruction path. This means when everything is well,
- * we don't even have to jump over them. Further, they do not intrude
- * on our cache or tlb entries.
+ * Return 0 on exception, a value greater than N if too long
  */
-struct exception_table_entry {
-        unsigned long insn, fixup;
-};
+extern int __strnlen_user(const char __user *sstr, int len);
+
+static inline long strnlen_user(const char __user *src, long n)
+{
+        if (!access_ok(VERIFY_READ, src, 1))
+                return 0;
+        return __strnlen_user(src, n);
+}
 
 #endif  /* __ASSEMBLY__ */
 #endif /* __KERNEL__ */
diff --git a/arch/microblaze/kernel/cpu/cpuinfo.c b/arch/microblaze/kernel/cpu/cpuinfo.c
index 991d71311b0e..255ef880351e 100644
--- a/arch/microblaze/kernel/cpu/cpuinfo.c
+++ b/arch/microblaze/kernel/cpu/cpuinfo.c
@@ -9,7 +9,6 @@
  */
 
 #include <linux/init.h>
-#include <linux/slab.h>
 #include <asm/cpuinfo.h>
 #include <asm/pvr.h>
 
diff --git a/arch/microblaze/kernel/dma.c b/arch/microblaze/kernel/dma.c
index b1084974fccd..ce72dd4967cf 100644
--- a/arch/microblaze/kernel/dma.c
+++ b/arch/microblaze/kernel/dma.c
@@ -8,6 +8,7 @@
 
 #include <linux/device.h>
 #include <linux/dma-mapping.h>
+#include <linux/gfp.h>
 #include <linux/dma-debug.h>
 #include <asm/bug.h>
 #include <asm/cacheflush.h>
@@ -37,7 +38,7 @@ static inline void __dma_sync_page(unsigned long paddr, unsigned long offset,
 
 static unsigned long get_dma_direct_offset(struct device *dev)
 {
-        if (dev)
+        if (likely(dev))
                 return (unsigned long)dev->archdata.dma_data;
 
         return PCI_DRAM_OFFSET; /* FIXME Not sure if is correct */
diff --git a/arch/microblaze/kernel/ftrace.c b/arch/microblaze/kernel/ftrace.c
index 388b31ca65a1..515feb404555 100644
--- a/arch/microblaze/kernel/ftrace.c
+++ b/arch/microblaze/kernel/ftrace.c
@@ -151,13 +151,10 @@ int ftrace_make_nop(struct module *mod,
         return ret;
 }
 
-static int ret_addr; /* initialized as 0 by default */
-
 /* I believe that first is called ftrace_make_nop before this function */
 int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 {
         int ret;
-        ret_addr = addr; /* saving where the barrier jump is */
         pr_debug("%s: addr:0x%x, rec->ip: 0x%x, imm:0x%x\n",
                 __func__, (unsigned int)addr, (unsigned int)rec->ip, imm);
         ret = ftrace_modify_code(rec->ip, imm);
@@ -194,12 +191,9 @@ int ftrace_update_ftrace_func(ftrace_func_t func)
         ret = ftrace_modify_code(ip, upper);
         ret += ftrace_modify_code(ip + 4, lower);
 
-        /* We just need to remove the rtsd r15, 8 by NOP */
-        BUG_ON(!ret_addr);
-        if (ret_addr)
-                ret += ftrace_modify_code(ret_addr, MICROBLAZE_NOP);
-        else
-                ret = 1; /* fault */
+        /* We just need to replace the rtsd r15, 8 with NOP */
+        ret += ftrace_modify_code((unsigned long)&ftrace_caller,
+                                  MICROBLAZE_NOP);
 
         /* All changes are done - lets do caches consistent */
         flush_icache();
diff --git a/arch/microblaze/kernel/head.S b/arch/microblaze/kernel/head.S
index cb7815cfe5ab..da6a5f5dc766 100644
--- a/arch/microblaze/kernel/head.S
+++ b/arch/microblaze/kernel/head.S
@@ -51,6 +51,12 @@ swapper_pg_dir:
 
         .text
 ENTRY(_start)
+#if CONFIG_KERNEL_BASE_ADDR == 0
+        brai    TOPHYS(real_start)
+        .org    0x100
+real_start:
+#endif
+
         mfs     r1, rmsr
         andi    r1, r1, ~2
         mts     rmsr, r1
@@ -99,8 +105,8 @@ no_fdt_arg:
         tophys(r4,r4)                   /* convert to phys address */
         ori     r3, r0, COMMAND_LINE_SIZE - 1 /* number of loops */
 _copy_command_line:
-        lbu     r2, r5, r6 /* r7=r5+r6 - r5 contain pointer to command line */
-        sb      r2, r4, r6              /* addr[r4+r6]= r7*/
+        lbu     r2, r5, r6 /* r2=r5+r6 - r5 contain pointer to command line */
+        sb      r2, r4, r6              /* addr[r4+r6]= r2*/
         addik   r6, r6, 1               /* increment counting */
         bgtid   r3, _copy_command_line  /* loop for all entries       */
         addik   r3, r3, -1              /* descrement loop */
@@ -128,7 +134,7 @@ _copy_bram:
          * virtual to physical.
          */
         nop
-        addik   r3, r0, 63              /* Invalidate all TLB entries */
+        addik   r3, r0, MICROBLAZE_TLB_SIZE -1  /* Invalidate all TLB entries */
 _invalidate:
         mts     rtlbx, r3
         mts     rtlbhi, r0                      /* flush: ensure V is clear   */
diff --git a/arch/microblaze/kernel/hw_exception_handler.S b/arch/microblaze/kernel/hw_exception_handler.S
index 2b86c03aa841..995a2123635b 100644
--- a/arch/microblaze/kernel/hw_exception_handler.S
+++ b/arch/microblaze/kernel/hw_exception_handler.S
@@ -313,13 +313,13 @@ _hw_exception_handler:
         mfs     r5, rmsr;
         nop
         swi     r5, r1, 0;
-        mfs     r3, resr
+        mfs     r4, resr
         nop
-        mfs     r4, rear;
+        mfs     r3, rear;
         nop
 
 #ifndef CONFIG_MMU
-        andi    r5, r3, 0x1000;         /* Check ESR[DS] */
+        andi    r5, r4, 0x1000;         /* Check ESR[DS] */
         beqi    r5, not_in_delay_slot;  /* Branch if ESR[DS] not set */
         mfs     r17, rbtr;      /* ESR[DS] set - return address in BTR */
         nop
@@ -327,13 +327,14 @@ not_in_delay_slot:
         swi     r17, r1, PT_R17
 #endif
 
-        andi    r5, r3, 0x1F;           /* Extract ESR[EXC] */
+        andi    r5, r4, 0x1F;           /* Extract ESR[EXC] */
 
 #ifdef CONFIG_MMU
         /* Calculate exception vector offset = r5 << 2 */
         addk    r6, r5, r5; /* << 1 */
         addk    r6, r6, r6; /* << 2 */
 
+#ifdef DEBUG
 /* counting which exception happen */
         lwi     r5, r0, 0x200 + TOPHYS(r0_ram)
         addi    r5, r5, 1
@@ -341,6 +342,7 @@ not_in_delay_slot:
         lwi     r5, r6, 0x200 + TOPHYS(r0_ram)
         addi    r5, r5, 1
         swi     r5, r6, 0x200 + TOPHYS(r0_ram)
+#endif
 /* end */
         /* Load the HW Exception vector */
         lwi     r6, r6, TOPHYS(_MB_HW_ExceptionVectorTable)
@@ -376,7 +378,7 @@ handle_other_ex: /* Handle Other exceptions here */
         swi     r18, r1, PT_R18
 
         or      r5, r1, r0
-        andi    r6, r3, 0x1F; /* Load ESR[EC] */
+        andi    r6, r4, 0x1F; /* Load ESR[EC] */
         lwi     r7, r0, PER_CPU(KM) /* MS: saving current kernel mode to regs */
         swi     r7, r1, PT_MODE
         mfs     r7, rfsr
@@ -426,11 +428,11 @@ handle_other_ex: /* Handle Other exceptions here */
  */
 handle_unaligned_ex:
         /* Working registers already saved: R3, R4, R5, R6
-         *  R3 = ESR
-         *  R4 = EAR
+         *  R4 = ESR
+         *  R3 = EAR
          */
 #ifdef CONFIG_MMU
-        andi    r6, r3, 0x1000                  /* Check ESR[DS] */
+        andi    r6, r4, 0x1000                  /* Check ESR[DS] */
         beqi    r6, _no_delayslot               /* Branch if ESR[DS] not set */
         mfs     r17, rbtr;      /* ESR[DS] set - return address in BTR */
         nop
@@ -439,7 +441,7 @@ _no_delayslot:
         RESTORE_STATE;
         bri     unaligned_data_trap
 #endif
-        andi    r6, r3, 0x3E0; /* Mask and extract the register operand */
+        andi    r6, r4, 0x3E0; /* Mask and extract the register operand */
         srl     r6, r6; /* r6 >> 5 */
         srl     r6, r6;
         srl     r6, r6;
@@ -448,33 +450,33 @@ _no_delayslot:
         /* Store the register operand in a temporary location */
         sbi     r6, r0, TOPHYS(ex_reg_op);
 
-        andi    r6, r3, 0x400; /* Extract ESR[S] */
+        andi    r6, r4, 0x400; /* Extract ESR[S] */
         bnei    r6, ex_sw;
 ex_lw:
-        andi    r6, r3, 0x800; /* Extract ESR[W] */
+        andi    r6, r4, 0x800; /* Extract ESR[W] */
         beqi    r6, ex_lhw;
-        lbui    r5, r4, 0; /* Exception address in r4 */
+        lbui    r5, r3, 0; /* Exception address in r3 */
         /* Load a word, byte-by-byte from destination address
                 and save it in tmp space */
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_0);
-        lbui    r5, r4, 1;
+        lbui    r5, r3, 1;
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_1);
-        lbui    r5, r4, 2;
+        lbui    r5, r3, 2;
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_2);
-        lbui    r5, r4, 3;
+        lbui    r5, r3, 3;
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_3);
-        /* Get the destination register value into r3 */
-        lwi     r3, r0, TOPHYS(ex_tmp_data_loc_0);
+        /* Get the destination register value into r4 */
+        lwi     r4, r0, TOPHYS(ex_tmp_data_loc_0);
         bri     ex_lw_tail;
 ex_lhw:
-        lbui    r5, r4, 0; /* Exception address in r4 */
+        lbui    r5, r3, 0; /* Exception address in r3 */
         /* Load a half-word, byte-by-byte from destination
                 address and save it in tmp space */
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_0);
-        lbui    r5, r4, 1;
+        lbui    r5, r3, 1;
         sbi     r5, r0, TOPHYS(ex_tmp_data_loc_1);
-        /* Get the destination register value into r3 */
-        lhui    r3, r0, TOPHYS(ex_tmp_data_loc_0);
+        /* Get the destination register value into r4 */
+        lhui    r4, r0, TOPHYS(ex_tmp_data_loc_0);
 ex_lw_tail:
         /* Get the destination register number into r5 */
         lbui    r5, r0, TOPHYS(ex_reg_op);
@@ -502,25 +504,25 @@ ex_sw_tail:
         andi    r6, r6, 0x800; /* Extract ESR[W] */
         beqi    r6, ex_shw;
         /* Get the word - delay slot */
-        swi     r3, r0, TOPHYS(ex_tmp_data_loc_0);
+        swi     r4, r0, TOPHYS(ex_tmp_data_loc_0);
         /* Store the word, byte-by-byte into destination address */
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_0);
-        sbi     r3, r4, 0;
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_1);
-        sbi     r3, r4, 1;
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_2);
-        sbi     r3, r4, 2;
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_3);
-        sbi     r3, r4, 3;
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_0);
+        sbi     r4, r3, 0;
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_1);
+        sbi     r4, r3, 1;
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_2);
+        sbi     r4, r3, 2;
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_3);
+        sbi     r4, r3, 3;
         bri     ex_handler_done;
 
 ex_shw:
         /* Store the lower half-word, byte-by-byte into destination address */
-        swi     r3, r0, TOPHYS(ex_tmp_data_loc_0);
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_2);
-        sbi     r3, r4, 0;
-        lbui    r3, r0, TOPHYS(ex_tmp_data_loc_3);
-        sbi     r3, r4, 1;
+        swi     r4, r0, TOPHYS(ex_tmp_data_loc_0);
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_2);
+        sbi     r4, r3, 0;
+        lbui    r4, r0, TOPHYS(ex_tmp_data_loc_3);
+        sbi     r4, r3, 1;
 ex_sw_end: /* Exception handling of store word, ends. */
 
 ex_handler_done:
@@ -560,21 +562,16 @@ ex_handler_done:
                  */
                 mfs     r11, rpid
                 nop
-                bri     4
-                mfs     r3, rear                /* Get faulting address */
-                nop
                 /* If we are faulting a kernel address, we have to use the
                  * kernel page tables.
                  */
-                ori     r4, r0, CONFIG_KERNEL_START
-                cmpu    r4, r3, r4
-                bgti    r4, ex3
+                ori     r5, r0, CONFIG_KERNEL_START
+                cmpu    r5, r3, r5
+                bgti    r5, ex3
                 /* First, check if it was a zone fault (which means a user
                  * tried to access a kernel or read-protected page - always
                  * a SEGV). All other faults here must be stores, so no
                  * need to check ESR_S as well. */
-                mfs     r4, resr
-                nop
                 andi    r4, r4, 0x800           /* ESR_Z - zone protection */
                 bnei    r4, ex2
 
@@ -589,8 +586,6 @@ ex_handler_done:
                  * tried to access a kernel or read-protected page - always
                  * a SEGV). All other faults here must be stores, so no
                  * need to check ESR_S as well. */
-                mfs     r4, resr
-                nop
                 andi    r4, r4, 0x800           /* ESR_Z */
                 bnei    r4, ex2
                 /* get current task address */
@@ -665,8 +660,6 @@ ex_handler_done:
                  * R3 = ESR
                  */
 
-                mfs     r3, rear                /* Get faulting address */
-                nop
                 RESTORE_STATE;
                 bri     page_fault_instr_trap
 
@@ -677,18 +670,15 @@ ex_handler_done:
          */
         handle_data_tlb_miss_exception:
                 /* Working registers already saved: R3, R4, R5, R6
-                 * R3 = ESR
+                 * R3 = EAR, R4 = ESR
                  */
                 mfs     r11, rpid
                 nop
-                bri     4
-                mfs     r3, rear                /* Get faulting address */
-                nop
 
                 /* If we are faulting a kernel address, we have to use the
                  * kernel page tables. */
-                ori     r4, r0, CONFIG_KERNEL_START
-                cmpu    r4, r3, r4
+                ori     r6, r0, CONFIG_KERNEL_START
+                cmpu    r4, r3, r6
                 bgti    r4, ex5
                 ori     r4, r0, swapper_pg_dir
                 mts     rpid, r0                /* TLB will have 0 TID */
@@ -731,9 +721,8 @@ ex_handler_done:
                  * Many of these bits are software only. Bits we don't set
                  * here we (properly should) assume have the appropriate value.
                  */
+                brid    finish_tlb_load
                 andni   r4, r4, 0x0ce2          /* Make sure 20, 21 are zero */
-
-                bri     finish_tlb_load
         ex7:
                 /* The bailout. Restore registers to pre-exception conditions
                  * and call the heavyweights to help us out.
@@ -754,9 +743,6 @@ ex_handler_done:
                  */
                 mfs     r11, rpid
                 nop
-                bri     4
-                mfs     r3, rear                /* Get faulting address */
-                nop
 
                 /* If we are faulting a kernel address, we have to use the
                  * kernel page tables.
@@ -792,7 +778,7 @@ ex_handler_done:
                 lwi     r4, r5, 0               /* Get Linux PTE */
 
                 andi    r6, r4, _PAGE_PRESENT
-                beqi    r6, ex7
+                beqi    r6, ex10
 
                 ori     r4, r4, _PAGE_ACCESSED
                 swi     r4, r5, 0
@@ -805,9 +791,8 @@ ex_handler_done:
                  * Many of these bits are software only. Bits we don't set
                  * here we (properly should) assume have the appropriate value.
                  */
+                brid    finish_tlb_load
                 andni   r4, r4, 0x0ce2          /* Make sure 20, 21 are zero */
-
-                bri     finish_tlb_load
         ex10:
                 /* The bailout. Restore registers to pre-exception conditions
                  * and call the heavyweights to help us out.
@@ -837,9 +822,9 @@ ex_handler_done:
                 andi    r5, r5, (MICROBLAZE_TLB_SIZE-1)
                 ori     r6, r0, 1
                 cmp     r31, r5, r6
-                blti    r31, sem
+                blti    r31, ex12
                 addik   r5, r6, 1
-        sem:
+        ex12:
                 /* MS: save back current TLB index */
                 swi     r5, r0, TOPHYS(tlb_index)
 
@@ -859,7 +844,6 @@ ex_handler_done:
                 nop
 
                 /* Done...restore registers and get out of here. */
-        ex12:
                 mts     rpid, r11
                 nop
                 bri 4
diff --git a/arch/microblaze/kernel/misc.S b/arch/microblaze/kernel/misc.S
index df16c6287a8e..7cf86498326c 100644
--- a/arch/microblaze/kernel/misc.S
+++ b/arch/microblaze/kernel/misc.S
@@ -26,9 +26,10 @@
  * We avoid flushing the pinned 0, 1 and possibly 2 entries.
  */
 .globl _tlbia;
+.type  _tlbia, @function
 .align 4;
 _tlbia:
-        addik   r12, r0, 63 /* flush all entries (63 - 3) */
+        addik   r12, r0, MICROBLAZE_TLB_SIZE - 1 /* flush all entries (63 - 3) */
         /* isync */
 _tlbia_1:
         mts     rtlbx, r12
@@ -41,11 +42,13 @@ _tlbia_1:
         /* sync */
         rtsd    r15, 8
         nop
+        .size  _tlbia, . - _tlbia
 
 /*
  * Flush MMU TLB for a particular address (in r5)
  */
 .globl _tlbie;
+.type  _tlbie, @function
 .align 4;
 _tlbie:
         mts     rtlbsx, r5 /* look up the address in TLB */
@@ -59,17 +62,20 @@ _tlbie_1:
         rtsd    r15, 8
         nop
 
+        .size  _tlbie, . - _tlbie
+
 /*
  * Allocate TLB entry for early console
  */
 .globl early_console_reg_tlb_alloc;
+.type  early_console_reg_tlb_alloc, @function
 .align 4;
 early_console_reg_tlb_alloc:
         /*
          * Load a TLB entry for the UART, so that microblaze_progress() can use
          * the UARTs nice and early.  We use a 4k real==virtual mapping.
          */
-        ori     r4, r0, 63
+        ori     r4, r0, MICROBLAZE_TLB_SIZE - 1
         mts     rtlbx, r4 /* TLB slot 2 */
 
         or      r4,r5,r0
@@ -86,6 +92,8 @@ early_console_reg_tlb_alloc:
         rtsd    r15, 8
         nop
 
+        .size  early_console_reg_tlb_alloc, . - early_console_reg_tlb_alloc
+
 /*
  * Copy a whole page (4096 bytes).
  */
@@ -104,6 +112,7 @@ early_console_reg_tlb_alloc:
 #define DCACHE_LINE_BYTES (4 * 4)
 
 .globl copy_page;
+.type  copy_page, @function
 .align 4;
 copy_page:
         ori     r11, r0, (PAGE_SIZE/DCACHE_LINE_BYTES) - 1
@@ -118,3 +127,5 @@ _copy_page_loop:
         addik   r11, r11, -1
         rtsd    r15, 8
         nop
+
+        .size  copy_page, . - copy_page
diff --git a/arch/microblaze/kernel/module.c b/arch/microblaze/kernel/module.c
index 5a45b1adfef1..cbecf110dc30 100644
--- a/arch/microblaze/kernel/module.c
+++ b/arch/microblaze/kernel/module.c
@@ -12,7 +12,6 @@
 #include <linux/kernel.h>
 #include <linux/elf.h>
 #include <linux/vmalloc.h>
-#include <linux/slab.h>
 #include <linux/fs.h>
 #include <linux/string.h>
 
diff --git a/arch/microblaze/kernel/of_platform.c b/arch/microblaze/kernel/of_platform.c
index 1c6d684996d7..0dc755286d38 100644
--- a/arch/microblaze/kernel/of_platform.c
+++ b/arch/microblaze/kernel/of_platform.c
@@ -17,7 +17,6 @@
 #include <linux/init.h>
 #include <linux/module.h>
 #include <linux/mod_devicetable.h>
-#include <linux/slab.h>
 #include <linux/pci.h>
 #include <linux/of.h>
 #include <linux/of_device.h>
diff --git a/arch/microblaze/kernel/process.c b/arch/microblaze/kernel/process.c
index 812f1bf06c9e..09bed44dfcd3 100644
--- a/arch/microblaze/kernel/process.c
+++ b/arch/microblaze/kernel/process.c
@@ -15,6 +15,7 @@
 #include <linux/bitops.h>
 #include <asm/system.h>
 #include <asm/pgalloc.h>
+#include <asm/uaccess.h> /* for USER_DS macros */
 #include <asm/cacheflush.h>
 
 void show_regs(struct pt_regs *regs)
@@ -74,7 +75,10 @@ __setup("hlt", hlt_setup);
 
 void default_idle(void)
 {
-        if (!hlt_counter) {
+        if (likely(hlt_counter)) {
+                while (!need_resched())
+                        cpu_relax();
+        } else {
                 clear_thread_flag(TIF_POLLING_NRFLAG);
                 smp_mb__after_clear_bit();
                 local_irq_disable();
@@ -82,9 +86,7 @@ void default_idle(void)
                         cpu_sleep();
                 local_irq_enable();
                 set_thread_flag(TIF_POLLING_NRFLAG);
-        } else
-                while (!need_resched())
-                        cpu_relax();
+        }
 }
 
 void cpu_idle(void)
diff --git a/arch/microblaze/kernel/ptrace.c b/arch/microblaze/kernel/ptrace.c
index 6d6349a145f9..a4a7770c6140 100644
--- a/arch/microblaze/kernel/ptrace.c
+++ b/arch/microblaze/kernel/ptrace.c
@@ -75,7 +75,6 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 {
         int rval;
         unsigned long val = 0;
-        unsigned long copied;
 
         switch (request) {
         /* Read/write the word at location ADDR in the registers. */
diff --git a/arch/microblaze/kernel/setup.c b/arch/microblaze/kernel/setup.c
index f974ec7aa357..17c98dbcec88 100644
--- a/arch/microblaze/kernel/setup.c
+++ b/arch/microblaze/kernel/setup.c
@@ -92,6 +92,12 @@ inline unsigned get_romfs_len(unsigned *addr)
 }
 #endif  /* CONFIG_MTD_UCLINUX_EBSS */
 
+#if defined(CONFIG_EARLY_PRINTK) && defined(CONFIG_SERIAL_UARTLITE_CONSOLE)
+#define eprintk early_printk
+#else
+#define eprintk printk
+#endif
+
 void __init machine_early_init(const char *cmdline, unsigned int ram,
                 unsigned int fdt, unsigned int msr)
 {
@@ -139,32 +145,32 @@ void __init machine_early_init(const char *cmdline, unsigned int ram,
         setup_early_printk(NULL);
 #endif
 
-        early_printk("Ramdisk addr 0x%08x, ", ram);
+        eprintk("Ramdisk addr 0x%08x, ", ram);
         if (fdt)
-                early_printk("FDT at 0x%08x\n", fdt);
+                eprintk("FDT at 0x%08x\n", fdt);
         else
-                early_printk("Compiled-in FDT at 0x%08x\n",
+                eprintk("Compiled-in FDT at 0x%08x\n",
                                         (unsigned int)_fdt_start);
 
 #ifdef CONFIG_MTD_UCLINUX
-        early_printk("Found romfs @ 0x%08x (0x%08x)\n",
+        eprintk("Found romfs @ 0x%08x (0x%08x)\n",
                         romfs_base, romfs_size);
-        early_printk("#### klimit %p ####\n", old_klimit);
+        eprintk("#### klimit %p ####\n", old_klimit);
         BUG_ON(romfs_size < 0); /* What else can we do? */
 
-        early_printk("Moved 0x%08x bytes from 0x%08x to 0x%08x\n",
+        eprintk("Moved 0x%08x bytes from 0x%08x to 0x%08x\n",
                         romfs_size, romfs_base, (unsigned)&_ebss);
 
-        early_printk("New klimit: 0x%08x\n", (unsigned)klimit);
+        eprintk("New klimit: 0x%08x\n", (unsigned)klimit);
 #endif
 
 #if CONFIG_XILINX_MICROBLAZE0_USE_MSR_INSTR
         if (msr)
-                early_printk("!!!Your kernel has setup MSR instruction but "
+                eprintk("!!!Your kernel has setup MSR instruction but "
                                 "CPU don't have it %d\n", msr);
 #else
         if (!msr)
-                early_printk("!!!Your kernel not setup MSR instruction but "
+                eprintk("!!!Your kernel not setup MSR instruction but "
                                 "CPU have it %d\n", msr);
 #endif
 
diff --git a/arch/microblaze/kernel/sys_microblaze.c b/arch/microblaze/kernel/sys_microblaze.c
index 9f3c205fb75b..f4e00b7f1259 100644
--- a/arch/microblaze/kernel/sys_microblaze.c
+++ b/arch/microblaze/kernel/sys_microblaze.c
@@ -30,6 +30,7 @@
 #include <linux/semaphore.h>
 #include <linux/uaccess.h>
 #include <linux/unistd.h>
+#include <linux/slab.h>
 
 #include <asm/syscalls.h>
 
diff --git a/arch/microblaze/kernel/traps.c b/arch/microblaze/kernel/traps.c
index eaaaf805f31b..5e4570ef515c 100644
--- a/arch/microblaze/kernel/traps.c
+++ b/arch/microblaze/kernel/traps.c
@@ -22,13 +22,11 @@ void trap_init(void)
         __enable_hw_exceptions();
 }
 
-static int kstack_depth_to_print = 24;
+static unsigned long kstack_depth_to_print = 24;
 
 static int __init kstack_setup(char *s)
 {
-        kstack_depth_to_print = strict_strtoul(s, 0, NULL);
-
-        return 1;
+        return !strict_strtoul(s, 0, &kstack_depth_to_print);
 }
 __setup("kstack=", kstack_setup);
 
diff --git a/arch/microblaze/lib/Makefile b/arch/microblaze/lib/Makefile
index b579db068c06..4dfe47d3cd91 100644
--- a/arch/microblaze/lib/Makefile
+++ b/arch/microblaze/lib/Makefile
@@ -10,5 +10,4 @@ else
 lib-y += memcpy.o memmove.o
 endif
 
-lib-$(CONFIG_NO_MMU) += uaccess.o
-lib-$(CONFIG_MMU) += uaccess_old.o
+lib-y += uaccess_old.o
diff --git a/arch/microblaze/lib/fastcopy.S b/arch/microblaze/lib/fastcopy.S
index 02e3ab4eddf3..fdc48bb065d8 100644
--- a/arch/microblaze/lib/fastcopy.S
+++ b/arch/microblaze/lib/fastcopy.S
@@ -30,8 +30,9 @@
  */
 
 #include <linux/linkage.h>
-
+        .text
         .globl  memcpy
+        .type  memcpy, @function
         .ent    memcpy
 
 memcpy:
@@ -345,9 +346,11 @@ a_done:
         rtsd    r15, 8
         nop
 
+.size  memcpy, . - memcpy
 .end memcpy
 /*----------------------------------------------------------------------------*/
         .globl  memmove
+        .type  memmove, @function
         .ent    memmove
 
 memmove:
@@ -659,4 +662,5 @@ d_done:
         rtsd    r15, 8
         nop
 
+.size  memmove, . - memmove
 .end memmove
diff --git a/arch/microblaze/lib/memcpy.c b/arch/microblaze/lib/memcpy.c
index cc2108b6b260..014bac92bdff 100644
--- a/arch/microblaze/lib/memcpy.c
+++ b/arch/microblaze/lib/memcpy.c
@@ -53,7 +53,7 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c)
         const uint32_t *i_src;
         uint32_t *i_dst;
 
-        if (c >= 4) {
+        if (likely(c >= 4)) {
                 unsigned  value, buf_hold;
 
                 /* Align the dstination to a word boundry. */
diff --git a/arch/microblaze/lib/memset.c b/arch/microblaze/lib/memset.c
index 4df851d41a29..ecfb663e1fc1 100644
--- a/arch/microblaze/lib/memset.c
+++ b/arch/microblaze/lib/memset.c
@@ -33,22 +33,23 @@
 #ifdef __HAVE_ARCH_MEMSET
 void *memset(void *v_src, int c, __kernel_size_t n)
 {
-
         char *src = v_src;
 #ifdef CONFIG_OPT_LIB_FUNCTION
         uint32_t *i_src;
-        uint32_t w32;
+        uint32_t w32 = 0;
 #endif
         /* Truncate c to 8 bits */
         c = (c & 0xFF);
 
 #ifdef CONFIG_OPT_LIB_FUNCTION
-        /* Make a repeating word out of it */
-        w32 = c;
-        w32 |= w32 << 8;
-        w32 |= w32 << 16;
+        if (unlikely(c)) {
+                /* Make a repeating word out of it */
+                w32 = c;
+                w32 |= w32 << 8;
+                w32 |= w32 << 16;
+        }
 
-        if (n >= 4) {
+        if (likely(n >= 4)) {
                 /* Align the destination to a word boundary */
                 /* This is done in an endian independant manner */
                 switch ((unsigned) src & 3) {
diff --git a/arch/microblaze/lib/uaccess.c b/arch/microblaze/lib/uaccess.c
deleted file mode 100644
index a853fe089c44..000000000000
--- a/arch/microblaze/lib/uaccess.c
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (C) 2006 Atmark Techno, Inc.
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- */
-
-#include <linux/string.h>
-#include <asm/uaccess.h>
-
-#include <asm/bug.h>
-
-long strnlen_user(const char __user *src, long count)
-{
-        return strlen(src) + 1;
-}
-
-#define __do_strncpy_from_user(dst, src, count, res)                    \
-        do {                                                            \
-                char *tmp;                                              \
-                strncpy(dst, src, count);                               \
-                for (tmp = dst; *tmp && count > 0; tmp++, count--)      \
-                        ;                                               \
-                res = (tmp - dst);                                      \
-        } while (0)
-
-long __strncpy_from_user(char *dst, const char __user *src, long count)
-{
-        long res;
-        __do_strncpy_from_user(dst, src, count, res);
-        return res;
-}
-
-long strncpy_from_user(char *dst, const char __user *src, long count)
-{
-        long res = -EFAULT;
-        if (access_ok(VERIFY_READ, src, 1))
-                __do_strncpy_from_user(dst, src, count, res);
-        return res;
-}
-
-unsigned long __copy_tofrom_user(void __user *to,
-                const void __user *from, unsigned long size)
-{
-        memcpy(to, from, size);
-        return 0;
-}
diff --git a/arch/microblaze/lib/uaccess_old.S b/arch/microblaze/lib/uaccess_old.S
index 67f991c14b8a..5810cec54a7a 100644
--- a/arch/microblaze/lib/uaccess_old.S
+++ b/arch/microblaze/lib/uaccess_old.S
@@ -22,6 +22,7 @@
 
         .text
 .globl __strncpy_user;
+.type  __strncpy_user, @function
 .align 4;
 __strncpy_user:
 
@@ -50,7 +51,7 @@ __strncpy_user:
 3:
         rtsd    r15,8
         nop
-
+        .size   __strncpy_user, . - __strncpy_user
 
         .section        .fixup, "ax"
         .align  2
@@ -72,6 +73,7 @@ __strncpy_user:
 
         .text
 .globl __strnlen_user;
+.type  __strnlen_user, @function
 .align 4;
 __strnlen_user:
         addik   r3,r6,0
@@ -90,7 +92,7 @@ __strnlen_user:
 3:
         rtsd    r15,8
         nop
-
+        .size   __strnlen_user, . - __strnlen_user
 
         .section        .fixup,"ax"
 4:
@@ -108,6 +110,7 @@ __strnlen_user:
  */
         .text
 .globl __copy_tofrom_user;
+.type  __copy_tofrom_user, @function
 .align 4;
 __copy_tofrom_user:
         /*
@@ -116,20 +119,34 @@ __copy_tofrom_user:
          * r7, r3 - count
          * r4 - tempval
          */
-        addik   r3,r7,0
-        beqi    r3,3f
-1:
-        lbu     r4,r6,r0
-        addik   r6,r6,1
-2:
-        sb      r4,r5,r0
-        addik   r3,r3,-1
-        bneid   r3,1b
-        addik   r5,r5,1         /* delay slot */
+        beqid   r7, 3f /* zero size is not likely */
+        andi    r3, r7, 0x3 /* filter add count */
+        bneid   r3, 4f /* if is odd value then byte copying */
+        or      r3, r5, r6 /* find if is any to/from unaligned */
+        andi    r3, r3, 0x3 /* mask unaligned */
+        bneid   r3, 1f /* it is unaligned -> then jump */
+        or      r3, r0, r0
+
+/* at least one 4 byte copy */
+5:      lw      r4, r6, r3
+6:      sw      r4, r5, r3
+        addik   r7, r7, -4
+        bneid   r7, 5b
+        addik   r3, r3, 4
+        addik   r3, r7, 0
+        rtsd    r15, 8
+        nop
+4:      or      r3, r0, r0
+1:      lbu     r4,r6,r3
+2:      sb      r4,r5,r3
+        addik   r7,r7,-1
+        bneid   r7,1b
+        addik   r3,r3,1         /* delay slot */
 3:
+        addik   r3,r7,0
         rtsd    r15,8
         nop
-
+        .size   __copy_tofrom_user, . - __copy_tofrom_user
 
         .section        __ex_table,"a"
-        .word   1b,3b,2b,3b
+        .word   1b,3b,2b,3b,5b,3b,6b,3b
diff --git a/arch/microblaze/mm/consistent.c b/arch/microblaze/mm/consistent.c
index a9b443e3fb98..f956e24fe49c 100644
--- a/arch/microblaze/mm/consistent.c
+++ b/arch/microblaze/mm/consistent.c
@@ -32,6 +32,7 @@
 #include <linux/highmem.h>
 #include <linux/pci.h>
 #include <linux/interrupt.h>
+#include <linux/gfp.h>
 
 #include <asm/pgalloc.h>
 #include <linux/io.h>
diff --git a/arch/microblaze/mm/fault.c b/arch/microblaze/mm/fault.c
index d9d249a66ff2..7af87f4b2c2c 100644
--- a/arch/microblaze/mm/fault.c
+++ b/arch/microblaze/mm/fault.c
@@ -106,7 +106,7 @@ void do_page_fault(struct pt_regs *regs, unsigned long address,
         regs->esr = error_code;
 
         /* On a kernel SLB miss we can only check for a valid exception entry */
-        if (kernel_mode(regs) && (address >= TASK_SIZE)) {
+        if (unlikely(kernel_mode(regs) && (address >= TASK_SIZE))) {
                 printk(KERN_WARNING "kernel task_size exceed");
                 _exception(SIGSEGV, regs, code, address);
         }
@@ -122,7 +122,7 @@ void do_page_fault(struct pt_regs *regs, unsigned long address,
         }
 #endif /* CONFIG_KGDB */
 
-        if (in_atomic() || !mm) {
+        if (unlikely(in_atomic() || !mm)) {
                 if (kernel_mode(regs))
                         goto bad_area_nosemaphore;
 
@@ -150,7 +150,7 @@ void do_page_fault(struct pt_regs *regs, unsigned long address,
          * source.  If this is invalid we can skip the address space check,
          * thus avoiding the deadlock.
          */
-        if (!down_read_trylock(&mm->mmap_sem)) {
+        if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
                 if (kernel_mode(regs) && !search_exception_tables(regs->pc))
                         goto bad_area_nosemaphore;
 
@@ -158,16 +158,16 @@ void do_page_fault(struct pt_regs *regs, unsigned long address,
         }
 
         vma = find_vma(mm, address);
-        if (!vma)
+        if (unlikely(!vma))
                 goto bad_area;
 
         if (vma->vm_start <= address)
                 goto good_area;
 
-        if (!(vma->vm_flags & VM_GROWSDOWN))
+        if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
                 goto bad_area;
 
-        if (!is_write)
+        if (unlikely(!is_write))
                 goto bad_area;
 
         /*
@@ -179,7 +179,7 @@ void do_page_fault(struct pt_regs *regs, unsigned long address,
          * before setting the user r1.  Thus we allow the stack to
          * expand to 1MB without further checks.
          */
-        if (address + 0x100000 < vma->vm_end) {
+        if (unlikely(address + 0x100000 < vma->vm_end)) {
 
                 /* get user regs even if this fault is in kernel mode */
                 struct pt_regs *uregs = current->thread.regs;
@@ -209,15 +209,15 @@ good_area:
         code = SEGV_ACCERR;
 
         /* a write */
-        if (is_write) {
-                if (!(vma->vm_flags & VM_WRITE))
+        if (unlikely(is_write)) {
+                if (unlikely(!(vma->vm_flags & VM_WRITE)))
                         goto bad_area;
         /* a read */
         } else {
                 /* protection fault */
-                if (error_code & 0x08000000)
+                if (unlikely(error_code & 0x08000000))
                         goto bad_area;
-                if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
+                if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC))))
                         goto bad_area;
         }
 
@@ -235,7 +235,7 @@ survive:
                         goto do_sigbus;
                 BUG();
         }
-        if (fault & VM_FAULT_MAJOR)
+        if (unlikely(fault & VM_FAULT_MAJOR))
                 current->maj_flt++;
         else
                 current->min_flt++;
diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
index 1608e2e1a44a..f42c2dde8b1c 100644
--- a/arch/microblaze/mm/init.c
+++ b/arch/microblaze/mm/init.c
@@ -15,6 +15,7 @@
 #include <linux/initrd.h>
 #include <linux/pagemap.h>
 #include <linux/pfn.h>
+#include <linux/slab.h>
 #include <linux/swap.h>
 
 #include <asm/page.h>
@@ -165,7 +166,6 @@ void free_init_pages(char *what, unsigned long begin, unsigned long end)
         for (addr = begin; addr < end; addr += PAGE_SIZE) {
                 ClearPageReserved(virt_to_page(addr));
                 init_page_count(virt_to_page(addr));
-                memset((void *)addr, 0xcc, PAGE_SIZE);
                 free_page(addr);
                 totalram_pages++;
         }
@@ -208,14 +208,6 @@ void __init mem_init(void)
 }
 
 #ifndef CONFIG_MMU
-/* Check against bounds of physical memory */
-int ___range_ok(unsigned long addr, unsigned long size)
-{
-        return ((addr < memory_start) ||
-                ((addr + size) > memory_end));
-}
-EXPORT_SYMBOL(___range_ok);
-
 int page_is_ram(unsigned long pfn)
 {
         return __range_ok(pfn, 0);
diff --git a/arch/microblaze/mm/pgtable.c b/arch/microblaze/mm/pgtable.c
index 63a6fd07c48f..d31312cde6ea 100644
--- a/arch/microblaze/mm/pgtable.c
+++ b/arch/microblaze/mm/pgtable.c
@@ -154,7 +154,7 @@ int map_page(unsigned long va, phys_addr_t pa, int flags)
                 err = 0;
                 set_pte_at(&init_mm, va, pg, pfn_pte(pa >> PAGE_SHIFT,
                                 __pgprot(flags)));
-                if (mem_init_done)
+                if (unlikely(mem_init_done))
                         flush_HPTE(0, va, pmd_val(*pd));
                         /* flush_HPTE(0, va, pg); */
         }
diff --git a/arch/microblaze/pci/pci-common.c b/arch/microblaze/pci/pci-common.c
index 0be34350d733..740bb32ec57e 100644
--- a/arch/microblaze/pci/pci-common.c
+++ b/arch/microblaze/pci/pci-common.c
@@ -26,6 +26,7 @@
 #include <linux/syscalls.h>
 #include <linux/irq.h>
 #include <linux/vmalloc.h>
+#include <linux/slab.h>
 
 #include <asm/processor.h>
 #include <asm/io.h>
diff --git a/arch/microblaze/pci/pci_32.c b/arch/microblaze/pci/pci_32.c
index 7e0c94f501cc..3c3d808d7ce0 100644
--- a/arch/microblaze/pci/pci_32.c
+++ b/arch/microblaze/pci/pci_32.c
@@ -14,6 +14,7 @@
 #include <linux/irq.h>
 #include <linux/list.h>
 #include <linux/of.h>
+#include <linux/slab.h>
 
 #include <asm/processor.h>
 #include <asm/io.h>