28 files changed, 204 insertions, 127 deletions

diff --git a/arch/alpha/kernel/machvec_impl.h b/arch/alpha/kernel/machvec_impl.h
index 4959b7a3e1e..11f996f24fd 100644
--- a/arch/alpha/kernel/machvec_impl.h
+++ b/arch/alpha/kernel/machvec_impl.h
@@ -41,7 +41,7 @@
 #define CAT1(x,y)  x##y
 #define CAT(x,y)   CAT1(x,y)
 
-#define DO_DEFAULT_RTC rtc_port: 0x70
+#define DO_DEFAULT_RTC .rtc_port = 0x70
 
 #define DO_EV4_MMU                                                      \
         .max_asn =                      EV4_MAX_ASN,                    \
diff --git a/arch/arm/kernel/calls.S b/arch/arm/kernel/calls.S
index 2ad4aa2a153..55076a75e5b 100644
--- a/arch/arm/kernel/calls.S
+++ b/arch/arm/kernel/calls.S
@@ -131,7 +131,7 @@ __syscall_start:
                 .long   sys_wait4
 /* 115 */       .long   sys_swapoff
                 .long   sys_sysinfo
-                .long   sys_ipc_wrapper
+                .long   sys_ipc
                 .long   sys_fsync
                 .long   sys_sigreturn_wrapper
 /* 120 */       .long   sys_clone_wrapper
@@ -254,7 +254,7 @@ __syscall_start:
                 .long   sys_fremovexattr
                 .long   sys_tkill
                 .long   sys_sendfile64
-/* 240 */       .long   sys_futex_wrapper
+/* 240 */       .long   sys_futex
                 .long   sys_sched_setaffinity
                 .long   sys_sched_getaffinity
                 .long   sys_io_setup
@@ -284,7 +284,7 @@ __syscall_start:
                 .long   sys_fstatfs64
                 .long   sys_tgkill
                 .long   sys_utimes
-/* 270 */       .long   sys_arm_fadvise64_64_wrapper
+/* 270 */       .long   sys_arm_fadvise64_64
                 .long   sys_pciconfig_iobase
                 .long   sys_pciconfig_read
                 .long   sys_pciconfig_write
@@ -333,7 +333,7 @@ __syscall_start:
                 .long   sys_inotify_init
                 .long   sys_inotify_add_watch
                 .long   sys_inotify_rm_watch
-                .long   sys_mbind_wrapper
+                .long   sys_mbind
 /* 320 */       .long   sys_get_mempolicy
                 .long   sys_set_mempolicy
 __syscall_end:
diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index d9fb819bf7c..2a8d27e18fa 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -614,6 +614,47 @@ __kuser_helper_start:
 /*
  * Reference prototype:
  *
+ *      void __kernel_memory_barrier(void)
+ *
+ * Input:
+ *
+ *      lr = return address
+ *
+ * Output:
+ *
+ *      none
+ *
+ * Clobbered:
+ *
+ *      the Z flag might be lost
+ *
+ * Definition and user space usage example:
+ *
+ *      typedef void (__kernel_dmb_t)(void);
+ *      #define __kernel_dmb (*(__kernel_dmb_t *)0xffff0fa0)
+ *
+ * Apply any needed memory barrier to preserve consistency with data modified
+ * manually and __kuser_cmpxchg usage.
+ *
+ * This could be used as follows:
+ *
+ * #define __kernel_dmb() \
+ *         asm volatile ( "mov r0, #0xffff0fff; mov lr, pc; sub pc, r0, #95" \
+ *              : : : "lr","cc" )
+ */
+
+__kuser_memory_barrier:                         @ 0xffff0fa0
+
+#if __LINUX_ARM_ARCH__ >= 6 && defined(CONFIG_SMP)
+        mcr     p15, 0, r0, c7, c10, 5  @ dmb
+#endif
+        mov     pc, lr
+
+        .align  5
+
+/*
+ * Reference prototype:
+ *
  *      int __kernel_cmpxchg(int oldval, int newval, int *ptr)
  *
  * Input:
@@ -642,6 +683,8 @@ __kuser_helper_start:
  * The C flag is also set if *ptr was changed to allow for assembly
  * optimization in the calling code.
  *
+ * Note: this routine already includes memory barriers as needed.
+ *
  * For example, a user space atomic_add implementation could look like this:
  *
  * #define atomic_add(ptr, val) \
@@ -698,10 +741,16 @@ __kuser_cmpxchg:				@ 0xffff0fc0
 
 #else
 
+#ifdef CONFIG_SMP
+        mcr     p15, 0, r0, c7, c10, 5  @ dmb
+#endif
         ldrex   r3, [r2]
         subs    r3, r3, r0
         strexeq r3, r1, [r2]
         rsbs    r0, r3, #0
+#ifdef CONFIG_SMP
+        mcr     p15, 0, r0, c7, c10, 5  @ dmb
+#endif
         mov     pc, lr
 
 #endif
diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S
index f7f18307523..e2b42997ad3 100644
--- a/arch/arm/kernel/entry-common.S
+++ b/arch/arm/kernel/entry-common.S
@@ -145,7 +145,7 @@ ENTRY(vector_swi)
 #endif
         enable_irq
 
-        str     r4, [sp, #-S_OFF]!              @ push fifth arg
+        stmdb   sp!, {r4, r5}                   @ push fifth and sixth args
 
         get_thread_info tsk
         ldr     ip, [tsk, #TI_FLAGS]            @ check for syscall tracing
@@ -204,7 +204,7 @@ ENTRY(sys_call_table)
  * Special system call wrappers
  */
 @ r0 = syscall number
-@ r5 = syscall table
+@ r8 = syscall table
                 .type   sys_syscall, #function
 sys_syscall:
                 eor     scno, r0, #__NR_SYSCALL_BASE
@@ -255,22 +255,6 @@ sys_sigaltstack_wrapper:
                 ldr     r2, [sp, #S_OFF + S_SP]
                 b       do_sigaltstack
 
-sys_futex_wrapper:
-                str     r5, [sp, #4]            @ push sixth arg
-                b       sys_futex
-
-sys_arm_fadvise64_64_wrapper:
-                str     r5, [sp, #4]            @ push r5 to stack
-                b       sys_arm_fadvise64_64
-
-sys_mbind_wrapper:
-                str     r5, [sp, #4]
-                b       sys_mbind
-
-sys_ipc_wrapper:
-                str     r5, [sp, #4]            @ push sixth arg
-                b       sys_ipc
-
 /*
  * Note: off_4k (r5) is always units of 4K.  If we can't do the requested
  * offset, we return EINVAL.
diff --git a/arch/i386/mm/ioremap.c b/arch/i386/mm/ioremap.c
index 8498b5ac395..247fde76aae 100644
--- a/arch/i386/mm/ioremap.c
+++ b/arch/i386/mm/ioremap.c
@@ -245,7 +245,7 @@ void iounmap(volatile void __iomem *addr)
                         addr < phys_to_virt(ISA_END_ADDRESS))
                 return;
 
-        addr = (volatile void *)(PAGE_MASK & (unsigned long __force)addr);
+        addr = (volatile void __iomem *)(PAGE_MASK & (unsigned long __force)addr);
 
         /* Use the vm area unlocked, assuming the caller
            ensures there isn't another iounmap for the same address
diff --git a/arch/i386/pci/mmconfig.c b/arch/i386/pci/mmconfig.c
index 08a08490121..4bb4d4b0f73 100644
--- a/arch/i386/pci/mmconfig.c
+++ b/arch/i386/pci/mmconfig.c
@@ -155,7 +155,7 @@ static __init void unreachable_devices(void)
                 addr = get_base_addr(0, 0, PCI_DEVFN(i, 0));
                 if (addr != 0)
                         pci_exp_set_dev_base(addr, 0, PCI_DEVFN(i, 0));
-                if (addr == 0 || readl((u32 *)addr) != val1)
+                if (addr == 0 || readl((u32 __iomem *)mmcfg_virt_addr) != val1)
                         set_bit(i, fallback_slots);
                 spin_unlock_irqrestore(&pci_config_lock, flags);
         }
diff --git a/arch/ia64/configs/sn2_defconfig b/arch/ia64/configs/sn2_defconfig
index e1924cc9687..ff8bb3770c9 100644
--- a/arch/ia64/configs/sn2_defconfig
+++ b/arch/ia64/configs/sn2_defconfig
@@ -113,7 +113,7 @@ CONFIG_IOSAPIC=y
 CONFIG_IA64_SGI_SN_XP=m
 CONFIG_FORCE_MAX_ZONEORDER=17
 CONFIG_SMP=y
-CONFIG_NR_CPUS=512
+CONFIG_NR_CPUS=1024
 # CONFIG_HOTPLUG_CPU is not set
 CONFIG_SCHED_SMT=y
 CONFIG_PREEMPT=y
diff --git a/arch/ia64/kernel/time.c b/arch/ia64/kernel/time.c
index 5b7e736f3b4..028a2b95936 100644
--- a/arch/ia64/kernel/time.c
+++ b/arch/ia64/kernel/time.c
@@ -249,3 +249,32 @@ time_init (void)
          */
         set_normalized_timespec(&wall_to_monotonic, -xtime.tv_sec, -xtime.tv_nsec);
 }
+
+#define SMALLUSECS 100
+
+void
+udelay (unsigned long usecs)
+{
+        unsigned long start;
+        unsigned long cycles;
+        unsigned long smallusecs;
+
+        /*
+         * Execute the non-preemptible delay loop (because the ITC might
+         * not be synchronized between CPUS) in relatively short time
+         * chunks, allowing preemption between the chunks.
+         */
+        while (usecs > 0) {
+                smallusecs = (usecs > SMALLUSECS) ? SMALLUSECS : usecs;
+                preempt_disable();
+                cycles = smallusecs*local_cpu_data->cyc_per_usec;
+                start = ia64_get_itc();
+
+                while (ia64_get_itc() - start < cycles)
+                        cpu_relax();
+
+                preempt_enable();
+                usecs -= smallusecs;
+        }
+}
+EXPORT_SYMBOL(udelay);
diff --git a/arch/ia64/kernel/uncached.c b/arch/ia64/kernel/uncached.c
index c6d40446c2c..b631cf86ed4 100644
--- a/arch/ia64/kernel/uncached.c
+++ b/arch/ia64/kernel/uncached.c
@@ -53,7 +53,7 @@ static void uncached_ipi_visibility(void *data)
         if ((status != PAL_VISIBILITY_OK) &&
             (status != PAL_VISIBILITY_OK_REMOTE_NEEDED))
                 printk(KERN_DEBUG "pal_prefetch_visibility() returns %i on "
-                       "CPU %i\n", status, get_cpu());
+                       "CPU %i\n", status, raw_smp_processor_id());
 }
 
 
@@ -63,7 +63,7 @@ static void uncached_ipi_mc_drain(void *data)
         status = ia64_pal_mc_drain();
         if (status)
                 printk(KERN_WARNING "ia64_pal_mc_drain() failed with %i on "
-                       "CPU %i\n", status, get_cpu());
+                       "CPU %i\n", status, raw_smp_processor_id());
 }
 
 
@@ -105,7 +105,7 @@ uncached_get_new_chunk(struct gen_pool *poolp)
         status = ia64_pal_prefetch_visibility(PAL_VISIBILITY_PHYSICAL);
 
         dprintk(KERN_INFO "pal_prefetch_visibility() returns %i on cpu %i\n",
-                status, get_cpu());
+                status, raw_smp_processor_id());
 
         if (!status) {
                 status = smp_call_function(uncached_ipi_visibility, NULL, 0, 1);
diff --git a/arch/ia64/kernel/vmlinux.lds.S b/arch/ia64/kernel/vmlinux.lds.S
index 30d8564e960..73af6267d2e 100644
--- a/arch/ia64/kernel/vmlinux.lds.S
+++ b/arch/ia64/kernel/vmlinux.lds.S
@@ -177,6 +177,9 @@ SECTIONS
         }
   . = ALIGN(PAGE_SIZE);         /* make sure the gate page doesn't expose kernel data */
 
+  .data.read_mostly : AT(ADDR(.data.read_mostly) - LOAD_OFFSET)
+        { *(.data.read_mostly) }
+
   .data.cacheline_aligned : AT(ADDR(.data.cacheline_aligned) - LOAD_OFFSET)
         { *(.data.cacheline_aligned) }
 
diff --git a/arch/ia64/sn/kernel/sn2/sn2_smp.c b/arch/ia64/sn/kernel/sn2/sn2_smp.c
index 5d54f5f4e92..471bbaa65d1 100644
--- a/arch/ia64/sn/kernel/sn2/sn2_smp.c
+++ b/arch/ia64/sn/kernel/sn2/sn2_smp.c
@@ -202,7 +202,7 @@ sn2_global_tlb_purge(struct mm_struct *mm, unsigned long start,
                      unsigned long end, unsigned long nbits)
 {
         int i, opt, shub1, cnode, mynasid, cpu, lcpu = 0, nasid, flushed = 0;
-        int mymm = (mm == current->active_mm);
+        int mymm = (mm == current->active_mm && current->mm);
         volatile unsigned long *ptc0, *ptc1;
         unsigned long itc, itc2, flags, data0 = 0, data1 = 0, rr_value;
         short nasids[MAX_NUMNODES], nix;
diff --git a/arch/ia64/sn/pci/pcibr/pcibr_reg.c b/arch/ia64/sn/pci/pcibr/pcibr_reg.c
index 5d534091262..79fdb91d725 100644
--- a/arch/ia64/sn/pci/pcibr/pcibr_reg.c
+++ b/arch/ia64/sn/pci/pcibr/pcibr_reg.c
@@ -25,7 +25,7 @@ union br_ptr {
  */
 void pcireg_control_bit_clr(struct pcibus_info *pcibus_info, uint64_t bits)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -38,14 +38,14 @@ void pcireg_control_bit_clr(struct pcibus_info *pcibus_info, uint64_t bits)
                 default:
                         panic
                             ("pcireg_control_bit_clr: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
 
 void pcireg_control_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -58,7 +58,7 @@ void pcireg_control_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
                 default:
                         panic
                             ("pcireg_control_bit_set: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
@@ -68,7 +68,7 @@ void pcireg_control_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
  */
 uint64_t pcireg_tflush_get(struct pcibus_info *pcibus_info)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
         uint64_t ret = 0;
 
         if (pcibus_info) {
@@ -82,7 +82,7 @@ uint64_t pcireg_tflush_get(struct pcibus_info *pcibus_info)
                 default:
                         panic
                             ("pcireg_tflush_get: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 
@@ -98,7 +98,7 @@ uint64_t pcireg_tflush_get(struct pcibus_info *pcibus_info)
  */
 uint64_t pcireg_intr_status_get(struct pcibus_info * pcibus_info)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
         uint64_t ret = 0;
 
         if (pcibus_info) {
@@ -112,7 +112,7 @@ uint64_t pcireg_intr_status_get(struct pcibus_info * pcibus_info)
                 default:
                         panic
                             ("pcireg_intr_status_get: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
         return ret;
@@ -123,7 +123,7 @@ uint64_t pcireg_intr_status_get(struct pcibus_info * pcibus_info)
  */
 void pcireg_intr_enable_bit_clr(struct pcibus_info *pcibus_info, uint64_t bits)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -136,14 +136,14 @@ void pcireg_intr_enable_bit_clr(struct pcibus_info *pcibus_info, uint64_t bits)
                 default:
                         panic
                             ("pcireg_intr_enable_bit_clr: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
 
 void pcireg_intr_enable_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -156,7 +156,7 @@ void pcireg_intr_enable_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
                 default:
                         panic
                             ("pcireg_intr_enable_bit_set: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
@@ -167,7 +167,7 @@ void pcireg_intr_enable_bit_set(struct pcibus_info *pcibus_info, uint64_t bits)
 void pcireg_intr_addr_addr_set(struct pcibus_info *pcibus_info, int int_n,
                                uint64_t addr)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -186,7 +186,7 @@ void pcireg_intr_addr_addr_set(struct pcibus_info *pcibus_info, int int_n,
                 default:
                         panic
                             ("pcireg_intr_addr_addr_get: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
@@ -196,7 +196,7 @@ void pcireg_intr_addr_addr_set(struct pcibus_info *pcibus_info, int int_n,
  */
 void pcireg_force_intr_set(struct pcibus_info *pcibus_info, int int_n)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -209,7 +209,7 @@ void pcireg_force_intr_set(struct pcibus_info *pcibus_info, int int_n)
                 default:
                         panic
                             ("pcireg_force_intr_set: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
@@ -219,7 +219,7 @@ void pcireg_force_intr_set(struct pcibus_info *pcibus_info, int int_n)
  */
 uint64_t pcireg_wrb_flush_get(struct pcibus_info *pcibus_info, int device)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
         uint64_t ret = 0;
 
         if (pcibus_info) {
@@ -233,7 +233,7 @@ uint64_t pcireg_wrb_flush_get(struct pcibus_info *pcibus_info, int device)
                             __sn_readq_relaxed(&ptr->pic.p_wr_req_buf[device]);
                         break;
                 default:
-                      panic("pcireg_wrb_flush_get: unknown bridgetype bridge 0x%p", (void *)ptr);
+                      panic("pcireg_wrb_flush_get: unknown bridgetype bridge 0x%p", ptr);
                 }
 
         }
@@ -244,7 +244,7 @@ uint64_t pcireg_wrb_flush_get(struct pcibus_info *pcibus_info, int device)
 void pcireg_int_ate_set(struct pcibus_info *pcibus_info, int ate_index,
                         uint64_t val)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -257,15 +257,15 @@ void pcireg_int_ate_set(struct pcibus_info *pcibus_info, int ate_index,
                 default:
                         panic
                             ("pcireg_int_ate_set: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
 }
 
-uint64_t *pcireg_int_ate_addr(struct pcibus_info *pcibus_info, int ate_index)
+uint64_t __iomem *pcireg_int_ate_addr(struct pcibus_info *pcibus_info, int ate_index)
 {
-        union br_ptr *ptr = (union br_ptr *)pcibus_info->pbi_buscommon.bs_base;
-        uint64_t *ret = (uint64_t *) 0;
+        union br_ptr __iomem *ptr = (union br_ptr __iomem *)pcibus_info->pbi_buscommon.bs_base;
+        uint64_t __iomem *ret = NULL;
 
         if (pcibus_info) {
                 switch (pcibus_info->pbi_bridge_type) {
@@ -278,7 +278,7 @@ uint64_t *pcireg_int_ate_addr(struct pcibus_info *pcibus_info, int ate_index)
                 default:
                         panic
                             ("pcireg_int_ate_addr: unknown bridgetype bridge 0x%p",
-                             (void *)ptr);
+                             ptr);
                 }
         }
         return ret;
diff --git a/arch/ia64/sn/pci/tioca_provider.c b/arch/ia64/sn/pci/tioca_provider.c
index 46b646a6d34..27aa1842dac 100644
--- a/arch/ia64/sn/pci/tioca_provider.c
+++ b/arch/ia64/sn/pci/tioca_provider.c
@@ -38,10 +38,10 @@ tioca_gart_init(struct tioca_kernel *tioca_kern)
         uint64_t offset;
         struct page *tmp;
         struct tioca_common *tioca_common;
-        struct tioca *ca_base;
+        struct tioca __iomem *ca_base;
 
         tioca_common = tioca_kern->ca_common;
-        ca_base = (struct tioca *)tioca_common->ca_common.bs_base;
+        ca_base = (struct tioca __iomem *)tioca_common->ca_common.bs_base;
 
         if (list_empty(tioca_kern->ca_devices))
                 return 0;
@@ -215,7 +215,7 @@ tioca_fastwrite_enable(struct tioca_kernel *tioca_kern)
 {
         int cap_ptr;
         uint32_t reg;
-        struct tioca *tioca_base;
+        struct tioca __iomem *tioca_base;
         struct pci_dev *pdev;
         struct tioca_common *common;
 
@@ -257,7 +257,7 @@ tioca_fastwrite_enable(struct tioca_kernel *tioca_kern)
          * Set ca's fw to match
          */
 
-        tioca_base = (struct tioca *)common->ca_common.bs_base;
+        tioca_base = (struct tioca __iomem*)common->ca_common.bs_base;
         __sn_setq_relaxed(&tioca_base->ca_control1, CA_AGP_FW_ENABLE);
 }
 
@@ -322,7 +322,7 @@ static uint64_t
 tioca_dma_d48(struct pci_dev *pdev, uint64_t paddr)
 {
         struct tioca_common *tioca_common;
-        struct tioca *ca_base;
+        struct tioca __iomem *ca_base;
         uint64_t ct_addr;
         dma_addr_t bus_addr;
         uint32_t node_upper;
@@ -330,7 +330,7 @@ tioca_dma_d48(struct pci_dev *pdev, uint64_t paddr)
         struct pcidev_info *pcidev_info = SN_PCIDEV_INFO(pdev);
 
         tioca_common = (struct tioca_common *)pcidev_info->pdi_pcibus_info;
-        ca_base = (struct tioca *)tioca_common->ca_common.bs_base;
+        ca_base = (struct tioca __iomem *)tioca_common->ca_common.bs_base;
 
         ct_addr = PHYS_TO_TIODMA(paddr);
         if (!ct_addr)
diff --git a/arch/powerpc/kernel/syscalls.c b/arch/powerpc/kernel/syscalls.c
index f72ced11212..91b93d917b6 100644
--- a/arch/powerpc/kernel/syscalls.c
+++ b/arch/powerpc/kernel/syscalls.c
@@ -247,7 +247,7 @@ long ppc64_personality(unsigned long personality)
 #define OVERRIDE_MACHINE    0
 #endif
 
-static inline int override_machine(char *mach)
+static inline int override_machine(char __user *mach)
 {
         if (OVERRIDE_MACHINE) {
                 /* change ppc64 to ppc */
diff --git a/arch/ppc/syslib/ppc4xx_dma.c b/arch/ppc/syslib/ppc4xx_dma.c
index f15e64285f9..05ccd598dd4 100644
--- a/arch/ppc/syslib/ppc4xx_dma.c
+++ b/arch/ppc/syslib/ppc4xx_dma.c
@@ -30,6 +30,7 @@
 
 #include <asm/system.h>
 #include <asm/io.h>
+#include <asm/dma.h>
 #include <asm/ppc4xx_dma.h>
 
 ppc_dma_ch_t dma_channels[MAX_PPC4xx_DMA_CHANNELS];
diff --git a/arch/sparc/kernel/ebus.c b/arch/sparc/kernel/ebus.c
index 1754192c69d..5c3529ceb5d 100644
--- a/arch/sparc/kernel/ebus.c
+++ b/arch/sparc/kernel/ebus.c
@@ -22,7 +22,7 @@
 #include <asm/oplib.h>
 #include <asm/bpp.h>
 
-struct linux_ebus *ebus_chain = 0;
+struct linux_ebus *ebus_chain = NULL;
 
 /* We are together with pcic.c under CONFIG_PCI. */
 extern unsigned int pcic_pin_to_irq(unsigned int, char *name);
@@ -46,7 +46,7 @@ static struct ebus_device_irq je1_1[] = {
         { "SUNW,CS4231",         0 },
         { "parallel",            0 },
         { "se",                  2 },
-        { 0, 0 }
+        { NULL, 0 }
 };
 
 /*
@@ -55,7 +55,7 @@ static struct ebus_device_irq je1_1[] = {
  */
 static struct ebus_system_entry ebus_blacklist[] = {
         { "SUNW,JavaEngine1", je1_1 },
-        { 0, 0 }
+        { NULL, NULL }
 };
 
 static struct ebus_device_irq *ebus_blackp = NULL;
@@ -233,7 +233,7 @@ void __init fill_ebus_device(int node, struct linux_ebus_device *dev)
                         ebus_alloc(sizeof(struct linux_ebus_child));
 
                 child = dev->children;
-                child->next = 0;
+                child->next = NULL;
                 child->parent = dev;
                 child->bus = dev->bus;
                 fill_ebus_child(node, &regs[0], child);
@@ -243,7 +243,7 @@ void __init fill_ebus_device(int node, struct linux_ebus_device *dev)
                                 ebus_alloc(sizeof(struct linux_ebus_child));
 
                         child = child->next;
-                        child->next = 0;
+                        child->next = NULL;
                         child->parent = dev;
                         child->bus = dev->bus;
                         fill_ebus_child(node, &regs[0], child);
@@ -275,7 +275,7 @@ void __init ebus_init(void)
                 }
         }
 
-        pdev = pci_get_device(PCI_VENDOR_ID_SUN, PCI_DEVICE_ID_SUN_EBUS, 0);
+        pdev = pci_get_device(PCI_VENDOR_ID_SUN, PCI_DEVICE_ID_SUN_EBUS, NULL);
         if (!pdev) {
                 return;
         }
@@ -284,7 +284,7 @@ void __init ebus_init(void)
 
         ebus_chain = ebus = (struct linux_ebus *)
                         ebus_alloc(sizeof(struct linux_ebus));
-        ebus->next = 0;
+        ebus->next = NULL;
 
         while (ebusnd) {
 
@@ -325,8 +325,8 @@ void __init ebus_init(void)
                                 ebus_alloc(sizeof(struct linux_ebus_device));
 
                 dev = ebus->devices;
-                dev->next = 0;
-                dev->children = 0;
+                dev->next = NULL;
+                dev->children = NULL;
                 dev->bus = ebus;
                 fill_ebus_device(nd, dev);
 
@@ -335,8 +335,8 @@ void __init ebus_init(void)
                                 ebus_alloc(sizeof(struct linux_ebus_device));
 
                         dev = dev->next;
-                        dev->next = 0;
-                        dev->children = 0;
+                        dev->next = NULL;
+                        dev->children = NULL;
                         dev->bus = ebus;
                         fill_ebus_device(nd, dev);
                 }
@@ -353,7 +353,7 @@ void __init ebus_init(void)
                 ebus->next = (struct linux_ebus *)
                         ebus_alloc(sizeof(struct linux_ebus));
                 ebus = ebus->next;
-                ebus->next = 0;
+                ebus->next = NULL;
                 ++num_ebus;
         }
         if (pdev)
diff --git a/arch/sparc/kernel/led.c b/arch/sparc/kernel/led.c
index 2a3afca453c..313d1620ae8 100644
--- a/arch/sparc/kernel/led.c
+++ b/arch/sparc/kernel/led.c
@@ -55,7 +55,7 @@ static int led_read_proc(char *buf, char **start, off_t offset, int count,
         return len;
 }
 
-static int led_write_proc(struct file *file, const char *buffer,
+static int led_write_proc(struct file *file, const char __user *buffer,
                           unsigned long count, void *data)
 {
         char *buf = NULL;
diff --git a/arch/sparc/kernel/pcic.c b/arch/sparc/kernel/pcic.c
index cccfc12802e..42002b742de 100644
--- a/arch/sparc/kernel/pcic.c
+++ b/arch/sparc/kernel/pcic.c
@@ -161,7 +161,7 @@ static struct pcic_sn2list pcic_known_sysnames[] = {
 static int pcic0_up;
 static struct linux_pcic pcic0;
 
-void * __iomem pcic_regs;
+void __iomem *pcic_regs;
 volatile int pcic_speculative;
 volatile int pcic_trapped;
 
diff --git a/arch/sparc/kernel/time.c b/arch/sparc/kernel/time.c
index 24814d58f9e..7dadcdb4ca4 100644
--- a/arch/sparc/kernel/time.c
+++ b/arch/sparc/kernel/time.c
@@ -49,7 +49,7 @@ DEFINE_SPINLOCK(rtc_lock);
 enum sparc_clock_type sp_clock_typ;
 DEFINE_SPINLOCK(mostek_lock);
 void __iomem *mstk48t02_regs = NULL;
-static struct mostek48t08 *mstk48t08_regs = NULL;
+static struct mostek48t08 __iomem *mstk48t08_regs = NULL;
 static int set_rtc_mmss(unsigned long);
 static int sbus_do_settimeofday(struct timespec *tv);
 
@@ -342,7 +342,7 @@ static __inline__ void clock_probe(void)
                 /* XXX r/o attribute is somewhere in r.flags */
                 r.flags = clk_reg[0].which_io;
                 r.start = clk_reg[0].phys_addr;
-                mstk48t08_regs = (struct mostek48t08 *) sbus_ioremap(&r, 0,
+                mstk48t08_regs = sbus_ioremap(&r, 0,
                     sizeof(struct mostek48t08), "mk48t08");
 
                 mstk48t02_regs = &mstk48t08_regs->regs;
diff --git a/arch/sparc/mm/sun4c.c b/arch/sparc/mm/sun4c.c
index 1d560390e28..731f19603ca 100644
--- a/arch/sparc/mm/sun4c.c
+++ b/arch/sparc/mm/sun4c.c
@@ -497,7 +497,7 @@ static void __init sun4c_probe_mmu(void)
         patch_kernel_fault_handler();
 }
 
-volatile unsigned long *sun4c_memerr_reg = NULL;
+volatile unsigned long __iomem *sun4c_memerr_reg = NULL;
 
 void __init sun4c_probe_memerr_reg(void)
 {
diff --git a/arch/um/Kconfig b/arch/um/Kconfig
index 563301fe5df..1eb21de9d1b 100644
--- a/arch/um/Kconfig
+++ b/arch/um/Kconfig
@@ -289,6 +289,8 @@ source "arch/um/Kconfig.net"
 
 source "drivers/net/Kconfig"
 
+source "drivers/connector/Kconfig"
+
 source "fs/Kconfig"
 
 source "security/Kconfig"
diff --git a/arch/um/Makefile-x86_64 b/arch/um/Makefile-x86_64
index 4f118d5cc2e..38df311e75d 100644
--- a/arch/um/Makefile-x86_64
+++ b/arch/um/Makefile-x86_64
@@ -12,3 +12,7 @@ CHECKFLAGS  += -m64
 
 ELF_ARCH := i386:x86-64
 ELF_FORMAT := elf64-x86-64
+
+# Not on all 64-bit distros /lib is a symlink to /lib64. PLD is an example.
+
+LINK-$(CONFIG_LD_SCRIPT_DYN) += -Wl,-rpath,/lib64
diff --git a/arch/um/include/sysdep-i386/stub.h b/arch/um/include/sysdep-i386/stub.h
index 6ba8cbbe0d3..b492b12b4a1 100644
--- a/arch/um/include/sysdep-i386/stub.h
+++ b/arch/um/include/sysdep-i386/stub.h
@@ -6,8 +6,12 @@
 #ifndef __SYSDEP_STUB_H
 #define __SYSDEP_STUB_H
 
+#include <sys/mman.h>
 #include <asm/ptrace.h>
 #include <asm/unistd.h>
+#include "stub-data.h"
+#include "kern_constants.h"
+#include "uml-config.h"
 
 extern void stub_segv_handler(int sig);
 extern void stub_clone_handler(void);
@@ -76,23 +80,22 @@ static inline long stub_syscall5(long syscall, long arg1, long arg2, long arg3,
         return ret;
 }
 
-static inline long stub_syscall6(long syscall, long arg1, long arg2, long arg3,
-                                 long arg4, long arg5, long arg6)
+static inline void trap_myself(void)
 {
-        long ret;
-
-        __asm__ volatile ("push %%ebp ; movl %%eax,%%ebp ; movl %1,%%eax ; "
-                        "int $0x80 ; pop %%ebp"
-                        : "=a" (ret)
-                        : "g" (syscall), "b" (arg1), "c" (arg2), "d" (arg3),
-                          "S" (arg4), "D" (arg5), "0" (arg6));
-
-        return ret;
+        __asm("int3");
 }
 
-static inline void trap_myself(void)
+static inline void remap_stack(int fd, unsigned long offset)
 {
-        __asm("int3");
+        __asm__ volatile ("movl %%eax,%%ebp ; movl %0,%%eax ; int $0x80 ;"
+                          "movl %7, %%ebx ; movl %%eax, (%%ebx)"
+                          : : "g" (STUB_MMAP_NR), "b" (UML_CONFIG_STUB_DATA), 
+                            "c" (UM_KERN_PAGE_SIZE), 
+                            "d" (PROT_READ | PROT_WRITE),
+                            "S" (MAP_FIXED | MAP_SHARED), "D" (fd), 
+                            "a" (offset), 
+                            "i" (&((struct stub_data *) UML_CONFIG_STUB_DATA)->err) 
+                          : "memory");
 }
 
 #endif
diff --git a/arch/um/include/sysdep-x86_64/stub.h b/arch/um/include/sysdep-x86_64/stub.h
index c41689c13dc..92e989f8176 100644
--- a/arch/um/include/sysdep-x86_64/stub.h
+++ b/arch/um/include/sysdep-x86_64/stub.h
@@ -6,8 +6,12 @@
 #ifndef __SYSDEP_STUB_H
 #define __SYSDEP_STUB_H
 
+#include <sys/mman.h>
 #include <asm/unistd.h>
 #include <sysdep/ptrace_user.h>
+#include "stub-data.h"
+#include "kern_constants.h"
+#include "uml-config.h"
 
 extern void stub_segv_handler(int sig);
 extern void stub_clone_handler(void);
@@ -81,23 +85,23 @@ static inline long stub_syscall5(long syscall, long arg1, long arg2, long arg3,
         return ret;
 }
 
-static inline long stub_syscall6(long syscall, long arg1, long arg2, long arg3,
-                                 long arg4, long arg5, long arg6)
+static inline void trap_myself(void)
 {
-        long ret;
-
-        __asm__ volatile ("movq %5,%%r10 ; movq %6,%%r8 ; "
-                "movq %7, %%r9; " __syscall : "=a" (ret)
-                : "0" (syscall), "D" (arg1), "S" (arg2), "d" (arg3),
-                  "g" (arg4), "g" (arg5), "g" (arg6)
-                : __syscall_clobber, "r10", "r8", "r9" );
-
-        return ret;
+        __asm("int3");
 }
 
-static inline void trap_myself(void)
+static inline void remap_stack(long fd, unsigned long offset)
 {
-        __asm("int3");
+        __asm__ volatile ("movq %4,%%r10 ; movq %5,%%r8 ; "
+                          "movq %6, %%r9; " __syscall "; movq %7, %%rbx ; "
+                          "movq %%rax, (%%rbx)": 
+                          : "a" (STUB_MMAP_NR), "D" (UML_CONFIG_STUB_DATA), 
+                            "S" (UM_KERN_PAGE_SIZE), 
+                            "d" (PROT_READ | PROT_WRITE), 
+                            "g" (MAP_FIXED | MAP_SHARED), "g" (fd), 
+                            "g" (offset),
+                            "i" (&((struct stub_data *) UML_CONFIG_STUB_DATA)->err)
+                          : __syscall_clobber, "r10", "r8", "r9" );
 }
 
 #endif
diff --git a/arch/um/kernel/skas/clone.c b/arch/um/kernel/skas/clone.c
index cb37ce9124a..47b812b3bca 100644
--- a/arch/um/kernel/skas/clone.c
+++ b/arch/um/kernel/skas/clone.c
@@ -18,11 +18,10 @@
  * on some systems.
  */
 
-#define STUB_DATA(field) (((struct stub_data *) UML_CONFIG_STUB_DATA)->field)
-
 void __attribute__ ((__section__ (".__syscall_stub")))
 stub_clone_handler(void)
 {
+        struct stub_data *data = (struct stub_data *) UML_CONFIG_STUB_DATA;
         long err;
 
         err = stub_syscall2(__NR_clone, CLONE_PARENT | CLONE_FILES | SIGCHLD,
@@ -35,17 +34,21 @@ stub_clone_handler(void)
         if(err)
                 goto out;
 
-        err = stub_syscall3(__NR_setitimer, ITIMER_VIRTUAL,
-                            (long) &STUB_DATA(timer), 0);
+        err = stub_syscall3(__NR_setitimer, ITIMER_VIRTUAL, 
+                            (long) &data->timer, 0);
         if(err)
                 goto out;
 
-        err = stub_syscall6(STUB_MMAP_NR, UML_CONFIG_STUB_DATA,
-                            UM_KERN_PAGE_SIZE, PROT_READ | PROT_WRITE,
-                            MAP_FIXED | MAP_SHARED, STUB_DATA(fd),
-                            STUB_DATA(offset));
+        remap_stack(data->fd, data->offset);
+        goto done;
+
  out:
-        /* save current result. Parent: pid; child: retcode of mmap */
-        STUB_DATA(err) = err;
+        /* save current result. 
+         * Parent: pid; 
+         * child: retcode of mmap already saved and it jumps around this 
+         * assignment
+         */
+        data->err = err;
+ done:
         trap_myself();
 }
diff --git a/arch/um/scripts/Makefile.rules b/arch/um/scripts/Makefile.rules
index b3fbf125709..2e41cabd3d9 100644
--- a/arch/um/scripts/Makefile.rules
+++ b/arch/um/scripts/Makefile.rules
@@ -21,11 +21,6 @@ define unprofile
 endef
 
 
-# The stubs and unmap.o can't try to call mcount or update basic block data
-define unprofile
-        $(patsubst -pg,,$(patsubst -fprofile-arcs -ftest-coverage,,$(1)))
-endef
-
 # cmd_make_link checks to see if the $(foo-dir) variable starts with a /.  If
 # so, it's considered to be a path relative to $(srcdir) rather than
 # $(srcdir)/arch/$(SUBARCH).  This is because x86_64 wants to get ldt.c from
diff --git a/arch/x86_64/mm/ioremap.c b/arch/x86_64/mm/ioremap.c
index 0d260e4492f..ae207064201 100644
--- a/arch/x86_64/mm/ioremap.c
+++ b/arch/x86_64/mm/ioremap.c
@@ -263,7 +263,7 @@ void iounmap(volatile void __iomem *addr)
                 addr < phys_to_virt(ISA_END_ADDRESS))
                 return;
 
-        addr = (volatile void *)(PAGE_MASK & (unsigned long __force)addr);
+        addr = (volatile void __iomem *)(PAGE_MASK & (unsigned long __force)addr);
         /* Use the vm area unlocked, assuming the caller
            ensures there isn't another iounmap for the same address
            in parallel. Reuse of the virtual address is prevented by
diff --git a/arch/x86_64/pci/mmconfig.c b/arch/x86_64/pci/mmconfig.c
index 9c4f907e301..f16c0d57c55 100644
--- a/arch/x86_64/pci/mmconfig.c
+++ b/arch/x86_64/pci/mmconfig.c
@@ -18,11 +18,11 @@ static DECLARE_BITMAP(fallback_slots, 32);
 /* Static virtual mapping of the MMCONFIG aperture */
 struct mmcfg_virt {
         struct acpi_table_mcfg_config *cfg;
-        char *virt;
+        char __iomem *virt;
 };
 static struct mmcfg_virt *pci_mmcfg_virt;
 
-static char *get_virt(unsigned int seg, unsigned bus)
+static char __iomem *get_virt(unsigned int seg, unsigned bus)
 {
         int cfg_num = -1;
         struct acpi_table_mcfg_config *cfg;
@@ -43,9 +43,9 @@ static char *get_virt(unsigned int seg, unsigned bus)
         }
 }
 
-static char *pci_dev_base(unsigned int seg, unsigned int bus, unsigned int devfn)
+static char __iomem *pci_dev_base(unsigned int seg, unsigned int bus, unsigned int devfn)
 {
-        char *addr;
+        char __iomem *addr;
         if (seg == 0 && bus == 0 && test_bit(PCI_SLOT(devfn), &fallback_slots))
                 return NULL;
         addr = get_virt(seg, bus);
@@ -57,7 +57,7 @@ static char *pci_dev_base(unsigned int seg, unsigned int bus, unsigned int devfn
 static int pci_mmcfg_read(unsigned int seg, unsigned int bus,
                           unsigned int devfn, int reg, int len, u32 *value)
 {
-        char *addr;
+        char __iomem *addr;
 
         /* Why do we have this when nobody checks it. How about a BUG()!? -AK */
         if (unlikely(!value || (bus > 255) || (devfn > 255) || (reg > 4095)))
@@ -85,7 +85,7 @@ static int pci_mmcfg_read(unsigned int seg, unsigned int bus,
 static int pci_mmcfg_write(unsigned int seg, unsigned int bus,
                            unsigned int devfn, int reg, int len, u32 value)
 {
-        char *addr;
+        char __iomem *addr;
 
         /* Why do we have this when nobody checks it. How about a BUG()!? -AK */
         if (unlikely((bus > 255) || (devfn > 255) || (reg > 4095)))
@@ -127,7 +127,7 @@ static __init void unreachable_devices(void)
         int i;
         for (i = 0; i < 32; i++) {
                 u32 val1;
-                char *addr;
+                char __iomem *addr;
 
                 pci_conf1_read(0, 0, PCI_DEVFN(i,0), 0, 4, &val1);
                 if (val1 == 0xffffffff)