Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6

21 files changed, 820 insertions, 367 deletions

diff --git a/arch/blackfin/kernel/Makefile b/arch/blackfin/kernel/Makefile
index a8ddbc8ed5a..346a421f156 100644
--- a/arch/blackfin/kernel/Makefile
+++ b/arch/blackfin/kernel/Makefile
@@ -25,6 +25,7 @@ obj-$(CONFIG_CPLB_INFO)              += cplbinfo.o
 obj-$(CONFIG_MODULES)                += module.o
 obj-$(CONFIG_KGDB)                   += kgdb.o
 obj-$(CONFIG_KGDB_TESTS)             += kgdb_test.o
+obj-$(CONFIG_NMI_WATCHDOG)           += nmi.o
 obj-$(CONFIG_EARLY_PRINTK)           += early_printk.o
 obj-$(CONFIG_EARLY_PRINTK)           += shadow_console.o
 obj-$(CONFIG_STACKTRACE)             += stacktrace.o
diff --git a/arch/blackfin/kernel/bfin_dma_5xx.c b/arch/blackfin/kernel/bfin_dma_5xx.c
index 924c00286ba..26403d1c9e6 100644
--- a/arch/blackfin/kernel/bfin_dma_5xx.c
+++ b/arch/blackfin/kernel/bfin_dma_5xx.c
@@ -91,7 +91,7 @@ late_initcall(proc_dma_init);
  */
 int request_dma(unsigned int channel, const char *device_id)
 {
-	pr_debug("request_dma() : BEGIN \n");
+	pr_debug("request_dma() : BEGIN\n");
 
 	if (device_id == NULL)
 		printk(KERN_WARNING "request_dma(%u): no device_id given\n", channel);
@@ -107,7 +107,7 @@ int request_dma(unsigned int channel, const char *device_id)
 #endif
 
 	if (atomic_cmpxchg(&dma_ch[channel].chan_status, 0, 1)) {
-		pr_debug("DMA CHANNEL IN USE  \n");
+		pr_debug("DMA CHANNEL IN USE\n");
 		return -EBUSY;
 	}
 
@@ -131,7 +131,7 @@ int request_dma(unsigned int channel, const char *device_id)
 	 * you have to request DMA, before doing any operations on
 	 * descriptor/channel
 	 */
-	pr_debug("request_dma() : END  \n");
+	pr_debug("request_dma() : END\n");
 	return 0;
 }
 EXPORT_SYMBOL(request_dma);
@@ -171,7 +171,7 @@ static void clear_dma_buffer(unsigned int channel)
 
 void free_dma(unsigned int channel)
 {
-	pr_debug("freedma() : BEGIN \n");
+	pr_debug("freedma() : BEGIN\n");
 	BUG_ON(channel >= MAX_DMA_CHANNELS ||
 			!atomic_read(&dma_ch[channel].chan_status));
 
@@ -185,7 +185,7 @@ void free_dma(unsigned int channel)
 	/* Clear the DMA Variable in the Channel */
 	atomic_set(&dma_ch[channel].chan_status, 0);
 
-	pr_debug("freedma() : END \n");
+	pr_debug("freedma() : END\n");
 }
 EXPORT_SYMBOL(free_dma);
 
diff --git a/arch/blackfin/kernel/bfin_gpio.c b/arch/blackfin/kernel/bfin_gpio.c
index a174596cc00..e35e20f00d9 100644
--- a/arch/blackfin/kernel/bfin_gpio.c
+++ b/arch/blackfin/kernel/bfin_gpio.c
@@ -1289,44 +1289,50 @@ __initcall(gpio_register_proc);
 #endif
 
 #ifdef CONFIG_GPIOLIB
-int bfin_gpiolib_direction_input(struct gpio_chip *chip, unsigned gpio)
+static int bfin_gpiolib_direction_input(struct gpio_chip *chip, unsigned gpio)
 {
 	return bfin_gpio_direction_input(gpio);
 }
 
-int bfin_gpiolib_direction_output(struct gpio_chip *chip, unsigned gpio, int level)
+static int bfin_gpiolib_direction_output(struct gpio_chip *chip, unsigned gpio, int level)
 {
 	return bfin_gpio_direction_output(gpio, level);
 }
 
-int bfin_gpiolib_get_value(struct gpio_chip *chip, unsigned gpio)
+static int bfin_gpiolib_get_value(struct gpio_chip *chip, unsigned gpio)
 {
 	return bfin_gpio_get_value(gpio);
 }
 
-void bfin_gpiolib_set_value(struct gpio_chip *chip, unsigned gpio, int value)
+static void bfin_gpiolib_set_value(struct gpio_chip *chip, unsigned gpio, int value)
 {
 	return bfin_gpio_set_value(gpio, value);
 }
 
-int bfin_gpiolib_gpio_request(struct gpio_chip *chip, unsigned gpio)
+static int bfin_gpiolib_gpio_request(struct gpio_chip *chip, unsigned gpio)
 {
 	return bfin_gpio_request(gpio, chip->label);
 }
 
-void bfin_gpiolib_gpio_free(struct gpio_chip *chip, unsigned gpio)
+static void bfin_gpiolib_gpio_free(struct gpio_chip *chip, unsigned gpio)
 {
 	return bfin_gpio_free(gpio);
 }
 
+static int bfin_gpiolib_gpio_to_irq(struct gpio_chip *chip, unsigned gpio)
+{
+	return gpio + GPIO_IRQ_BASE;
+}
+
 static struct gpio_chip bfin_chip = {
-	.label			= "Blackfin-GPIOlib",
+	.label			= "BFIN-GPIO",
 	.direction_input	= bfin_gpiolib_direction_input,
 	.get			= bfin_gpiolib_get_value,
 	.direction_output	= bfin_gpiolib_direction_output,
 	.set			= bfin_gpiolib_set_value,
 	.request		= bfin_gpiolib_gpio_request,
 	.free			= bfin_gpiolib_gpio_free,
+	.to_irq			= bfin_gpiolib_gpio_to_irq,
 	.base			= 0,
 	.ngpio			= MAX_BLACKFIN_GPIOS,
 };
diff --git a/arch/blackfin/kernel/cplb-mpu/cplbinit.c b/arch/blackfin/kernel/cplb-mpu/cplbinit.c
index 8d42b9e50df..30fd6417f06 100644
--- a/arch/blackfin/kernel/cplb-mpu/cplbinit.c
+++ b/arch/blackfin/kernel/cplb-mpu/cplbinit.c
@@ -64,6 +64,15 @@ void __init generate_cplb_tables_cpu(unsigned int cpu)
 		icplb_tbl[cpu][i_i++].data = i_data | (addr == 0 ? CPLB_USER_RD : 0);
 	}
 
+#ifdef CONFIG_ROMKERNEL
+	/* Cover kernel XIP flash area */
+	addr = CONFIG_ROM_BASE & ~(4 * 1024 * 1024 - 1);
+	dcplb_tbl[cpu][i_d].addr = addr;
+	dcplb_tbl[cpu][i_d++].data = d_data | CPLB_USER_RD;
+	icplb_tbl[cpu][i_i].addr = addr;
+	icplb_tbl[cpu][i_i++].data = i_data | CPLB_USER_RD;
+#endif
+
 	/* Cover L1 memory.  One 4M area for code and data each is enough.  */
 #if L1_DATA_A_LENGTH > 0 || L1_DATA_B_LENGTH > 0
 	dcplb_tbl[cpu][i_d].addr = get_l1_data_a_start_cpu(cpu);
diff --git a/arch/blackfin/kernel/cplb-mpu/cplbmgr.c b/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
index 930c01c0681..87b25b1b30e 100644
--- a/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
+++ b/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
@@ -31,6 +31,12 @@ int nr_dcplb_miss[NR_CPUS], nr_icplb_miss[NR_CPUS];
 int nr_icplb_supv_miss[NR_CPUS], nr_dcplb_prot[NR_CPUS];
 int nr_cplb_flush[NR_CPUS];
 
+#ifdef CONFIG_EXCPT_IRQ_SYSC_L1
+#define MGR_ATTR __attribute__((l1_text))
+#else
+#define MGR_ATTR
+#endif
+
 /*
  * Given the contents of the status register, return the index of the
  * CPLB that caused the fault.
@@ -59,7 +65,7 @@ static int icplb_rr_index[NR_CPUS], dcplb_rr_index[NR_CPUS];
 /*
  * Find an ICPLB entry to be evicted and return its index.
  */
-static int evict_one_icplb(unsigned int cpu)
+MGR_ATTR static int evict_one_icplb(unsigned int cpu)
 {
 	int i;
 	for (i = first_switched_icplb; i < MAX_CPLBS; i++)
@@ -74,7 +80,7 @@ static int evict_one_icplb(unsigned int cpu)
 	return i;
 }
 
-static int evict_one_dcplb(unsigned int cpu)
+MGR_ATTR static int evict_one_dcplb(unsigned int cpu)
 {
 	int i;
 	for (i = first_switched_dcplb; i < MAX_CPLBS; i++)
@@ -89,7 +95,7 @@ static int evict_one_dcplb(unsigned int cpu)
 	return i;
 }
 
-static noinline int dcplb_miss(unsigned int cpu)
+MGR_ATTR static noinline int dcplb_miss(unsigned int cpu)
 {
 	unsigned long addr = bfin_read_DCPLB_FAULT_ADDR();
 	int status = bfin_read_DCPLB_STATUS();
@@ -114,10 +120,15 @@ static noinline int dcplb_miss(unsigned int cpu)
 		d_data = L2_DMEMORY;
 	} else if (addr >= physical_mem_end) {
 		if (addr >= ASYNC_BANK0_BASE && addr < ASYNC_BANK3_BASE + ASYNC_BANK3_SIZE) {
-			addr &= ~(4 * 1024 * 1024 - 1);
-			d_data &= ~PAGE_SIZE_4KB;
-			d_data |= PAGE_SIZE_4MB;
-			d_data |= CPLB_USER_RD | CPLB_USER_WR;
+			mask = current_rwx_mask[cpu];
+			if (mask) {
+				int page = (addr - (ASYNC_BANK0_BASE - _ramend)) >> PAGE_SHIFT;
+				int idx = page >> 5;
+				int bit = 1 << (page & 31);
+
+				if (mask[idx] & bit)
+					d_data |= CPLB_USER_RD;
+			}
 		} else if (addr >= BOOT_ROM_START && addr < BOOT_ROM_START + BOOT_ROM_LENGTH
 		    && (status & (FAULT_RW | FAULT_USERSUPV)) == FAULT_USERSUPV) {
 			addr &= ~(1 * 1024 * 1024 - 1);
@@ -126,7 +137,9 @@ static noinline int dcplb_miss(unsigned int cpu)
 		} else
 			return CPLB_PROT_VIOL;
 	} else if (addr >= _ramend) {
-	    d_data |= CPLB_USER_RD | CPLB_USER_WR;
+		d_data |= CPLB_USER_RD | CPLB_USER_WR;
+		if (reserved_mem_dcache_on)
+			d_data |= CPLB_L1_CHBL;
 	} else {
 		mask = current_rwx_mask[cpu];
 		if (mask) {
@@ -156,7 +169,7 @@ static noinline int dcplb_miss(unsigned int cpu)
 	return 0;
 }
 
-static noinline int icplb_miss(unsigned int cpu)
+MGR_ATTR static noinline int icplb_miss(unsigned int cpu)
 {
 	unsigned long addr = bfin_read_ICPLB_FAULT_ADDR();
 	int status = bfin_read_ICPLB_STATUS();
@@ -204,10 +217,19 @@ static noinline int icplb_miss(unsigned int cpu)
 		i_data = L2_IMEMORY;
 	} else if (addr >= physical_mem_end) {
 		if (addr >= ASYNC_BANK0_BASE && addr < ASYNC_BANK3_BASE + ASYNC_BANK3_SIZE) {
-			addr &= ~(4 * 1024 * 1024 - 1);
-			i_data &= ~PAGE_SIZE_4KB;
-			i_data |= PAGE_SIZE_4MB;
-			i_data |= CPLB_USER_RD;
+			if (!(status & FAULT_USERSUPV)) {
+				unsigned long *mask = current_rwx_mask[cpu];
+
+				if (mask) {
+					int page = (addr - (ASYNC_BANK0_BASE - _ramend)) >> PAGE_SHIFT;
+					int idx = page >> 5;
+					int bit = 1 << (page & 31);
+
+					mask += 2 * page_mask_nelts;
+					if (mask[idx] & bit)
+						i_data |= CPLB_USER_RD;
+				}
+			}
 		} else if (addr >= BOOT_ROM_START && addr < BOOT_ROM_START + BOOT_ROM_LENGTH
 		    && (status & FAULT_USERSUPV)) {
 			addr &= ~(1 * 1024 * 1024 - 1);
@@ -217,6 +239,8 @@ static noinline int icplb_miss(unsigned int cpu)
 		    return CPLB_PROT_VIOL;
 	} else if (addr >= _ramend) {
 		i_data |= CPLB_USER_RD;
+		if (reserved_mem_icache_on)
+			i_data |= CPLB_L1_CHBL;
 	} else {
 		/*
 		 * Two cases to distinguish - a supervisor access must
@@ -251,7 +275,7 @@ static noinline int icplb_miss(unsigned int cpu)
 	return 0;
 }
 
-static noinline int dcplb_protection_fault(unsigned int cpu)
+MGR_ATTR static noinline int dcplb_protection_fault(unsigned int cpu)
 {
 	int status = bfin_read_DCPLB_STATUS();
 
@@ -271,7 +295,7 @@ static noinline int dcplb_protection_fault(unsigned int cpu)
 	return CPLB_PROT_VIOL;
 }
 
-int cplb_hdr(int seqstat, struct pt_regs *regs)
+MGR_ATTR int cplb_hdr(int seqstat, struct pt_regs *regs)
 {
 	int cause = seqstat & 0x3f;
 	unsigned int cpu = raw_smp_processor_id();
diff --git a/arch/blackfin/kernel/cplb-nompu/cplbinit.c b/arch/blackfin/kernel/cplb-nompu/cplbinit.c
index 282a7919821..bfe75af4e8b 100644
--- a/arch/blackfin/kernel/cplb-nompu/cplbinit.c
+++ b/arch/blackfin/kernel/cplb-nompu/cplbinit.c
@@ -56,6 +56,15 @@ void __init generate_cplb_tables_cpu(unsigned int cpu)
 		i_tbl[i_i++].data = SDRAM_IGENERIC | PAGE_SIZE_4MB;
 	}
 
+#ifdef CONFIG_ROMKERNEL
+	/* Cover kernel XIP flash area */
+	addr = CONFIG_ROM_BASE & ~(4 * 1024 * 1024 - 1);
+	d_tbl[i_d].addr = addr;
+	d_tbl[i_d++].data = SDRAM_DGENERIC | PAGE_SIZE_4MB;
+	i_tbl[i_i].addr = addr;
+	i_tbl[i_i++].data = SDRAM_IGENERIC | PAGE_SIZE_4MB;
+#endif
+
 	/* Cover L1 memory.  One 4M area for code and data each is enough.  */
 	if (cpu == 0) {
 		if (L1_DATA_A_LENGTH || L1_DATA_B_LENGTH) {
diff --git a/arch/blackfin/kernel/dma-mapping.c b/arch/blackfin/kernel/dma-mapping.c
index e937f323d82..04ddcfeb798 100644
--- a/arch/blackfin/kernel/dma-mapping.c
+++ b/arch/blackfin/kernel/dma-mapping.c
@@ -116,7 +116,7 @@ EXPORT_SYMBOL(dma_free_coherent);
 void __dma_sync(dma_addr_t addr, size_t size,
 		enum dma_data_direction dir)
 {
-	_dma_sync(addr, size, dir);
+	__dma_sync_inline(addr, size, dir);
 }
 EXPORT_SYMBOL(__dma_sync);
 
diff --git a/arch/blackfin/kernel/entry.S b/arch/blackfin/kernel/entry.S
index f27dc2292e1..686478f5f66 100644
--- a/arch/blackfin/kernel/entry.S
+++ b/arch/blackfin/kernel/entry.S
@@ -44,7 +44,7 @@ ENTRY(_ret_from_fork)
 	sti r4;
 #endif /* CONFIG_IPIPE */
 	SP += -12;
-	call _schedule_tail;
+	pseudo_long_call _schedule_tail, p5;
 	SP += 12;
 	r0 = [sp + PT_IPEND];
 	cc = bittst(r0,1);
@@ -79,7 +79,7 @@ ENTRY(_sys_vfork)
 	r0 += 24;
 	[--sp] = rets;
 	SP += -12;
-	call _bfin_vfork;
+	pseudo_long_call _bfin_vfork, p2;
 	SP += 12;
 	rets = [sp++];
 	rts;
@@ -90,7 +90,7 @@ ENTRY(_sys_clone)
 	r0 += 24;
 	[--sp] = rets;
 	SP += -12;
-	call _bfin_clone;
+	pseudo_long_call _bfin_clone, p2;
 	SP += 12;
 	rets = [sp++];
 	rts;
@@ -101,7 +101,7 @@ ENTRY(_sys_rt_sigreturn)
 	r0 += 24;
 	[--sp] = rets;
 	SP += -12;
-	call _do_rt_sigreturn;
+	pseudo_long_call _do_rt_sigreturn, p2;
 	SP += 12;
 	rets = [sp++];
 	rts;
diff --git a/arch/blackfin/kernel/ftrace-entry.S b/arch/blackfin/kernel/ftrace-entry.S
index 76dd4fbcd17..d66446b572c 100644
--- a/arch/blackfin/kernel/ftrace-entry.S
+++ b/arch/blackfin/kernel/ftrace-entry.S
@@ -1,7 +1,7 @@
 /*
  * mcount and friends -- ftrace stuff
  *
- * Copyright (C) 2009 Analog Devices Inc.
+ * Copyright (C) 2009-2010 Analog Devices Inc.
  * Licensed under the GPL-2 or later.
  */
 
@@ -21,6 +21,15 @@
  * function will be waiting there.  mmmm pie.
  */
 ENTRY(__mcount)
+#ifdef CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST
+	/* optional micro optimization: return if stopped */
+	p1.l = _function_trace_stop;
+	p1.h = _function_trace_stop;
+	r3 = [p1];
+	cc = r3 == 0;
+	if ! cc jump _ftrace_stub (bp);
+#endif
+
 	/* save third function arg early so we can do testing below */
 	[--sp] = r2;
 
@@ -106,9 +115,12 @@ ENTRY(_ftrace_graph_caller)
 	[--sp] = r1;
 	[--sp] = rets;
 
-	/* prepare_ftrace_return(unsigned long *parent, unsigned long self_addr) */
-	r0 = sp;
-	r1 = rets;
+	/* prepare_ftrace_return(parent, self_addr, frame_pointer) */
+	r0 = sp;	/* unsigned long *parent */
+	r1 = rets;	/* unsigned long self_addr */
+#ifdef CONFIG_HAVE_FUNCTION_GRAPH_FP_TEST
+	r2 = fp;	/* unsigned long frame_pointer */
+#endif
 	r0 += 16;	/* skip the 4 local regs on stack */
 	r1 += -MCOUNT_INSN_SIZE;
 	call _prepare_ftrace_return;
@@ -127,6 +139,9 @@ ENTRY(_return_to_handler)
 	[--sp] = r1;
 
 	/* get original return address */
+#ifdef CONFIG_HAVE_FUNCTION_GRAPH_FP_TEST
+	r0 = fp;	/* Blackfin is sane, so omit this */
+#endif
 	call _ftrace_return_to_handler;
 	rets = r0;
 
diff --git a/arch/blackfin/kernel/ftrace.c b/arch/blackfin/kernel/ftrace.c
index f2c85ac6f2d..a61d948ea92 100644
--- a/arch/blackfin/kernel/ftrace.c
+++ b/arch/blackfin/kernel/ftrace.c
@@ -16,7 +16,8 @@
  * Hook the return address and push it in the stack of return addrs
  * in current thread info.
  */
-void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
+void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr,
+                           unsigned long frame_pointer)
 {
 	struct ftrace_graph_ent trace;
 	unsigned long return_hooker = (unsigned long)&return_to_handler;
@@ -24,7 +25,8 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
 	if (unlikely(atomic_read(&current->tracing_graph_pause)))
 		return;
 
-	if (ftrace_push_return_trace(*parent, self_addr, &trace.depth, 0) == -EBUSY)
+	if (ftrace_push_return_trace(*parent, self_addr, &trace.depth,
+	                             frame_pointer) == -EBUSY)
 		return;
 
 	trace.func = self_addr;
diff --git a/arch/blackfin/kernel/init_task.c b/arch/blackfin/kernel/init_task.c
index 118c5b9deda..d3970e8acd1 100644
--- a/arch/blackfin/kernel/init_task.c
+++ b/arch/blackfin/kernel/init_task.c
@@ -28,5 +28,5 @@ EXPORT_SYMBOL(init_task);
  * "init_task" linker map entry.
  */
 union thread_union init_thread_union
-    __attribute__ ((__section__(".init_task.data"))) = {
+    __init_task_data = {
 INIT_THREAD_INFO(init_task)};
diff --git a/arch/blackfin/kernel/ipipe.c b/arch/blackfin/kernel/ipipe.c
index a77307a4473..1a496cd71ba 100644
--- a/arch/blackfin/kernel/ipipe.c
+++ b/arch/blackfin/kernel/ipipe.c
@@ -27,7 +27,6 @@
 #include <linux/interrupt.h>
 #include <linux/percpu.h>
 #include <linux/bitops.h>
-#include <linux/slab.h>
 #include <linux/errno.h>
 #include <linux/kthread.h>
 #include <linux/unistd.h>
diff --git a/arch/blackfin/kernel/kgdb.c b/arch/blackfin/kernel/kgdb.c
index 34c7c3ed2c9..2c501ceb1e5 100644
--- a/arch/blackfin/kernel/kgdb.c
+++ b/arch/blackfin/kernel/kgdb.c
@@ -145,7 +145,7 @@ void gdb_regs_to_pt_regs(unsigned long *gdb_regs, struct pt_regs *regs)
 #endif
 }
 
-struct hw_breakpoint {
+static struct hw_breakpoint {
 	unsigned int occupied:1;
 	unsigned int skip:1;
 	unsigned int enabled:1;
@@ -155,7 +155,7 @@ struct hw_breakpoint {
 	unsigned int addr;
 } breakinfo[HW_WATCHPOINT_NUM];
 
-int bfin_set_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
+static int bfin_set_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
 {
 	int breakno;
 	int bfin_type;
@@ -202,7 +202,7 @@ int bfin_set_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
 	return -ENOSPC;
 }
 
-int bfin_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
+static int bfin_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
 {
 	int breakno;
 	int bfin_type;
@@ -230,7 +230,7 @@ int bfin_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype type)
 	return 0;
 }
 
-void bfin_remove_all_hw_break(void)
+static void bfin_remove_all_hw_break(void)
 {
 	int breakno;
 
@@ -242,7 +242,7 @@ void bfin_remove_all_hw_break(void)
 		breakinfo[breakno].type = TYPE_DATA_WATCHPOINT;
 }
 
-void bfin_correct_hw_break(void)
+static void bfin_correct_hw_break(void)
 {
 	int breakno;
 	unsigned int wpiactl = 0;
diff --git a/arch/blackfin/kernel/nmi.c b/arch/blackfin/kernel/nmi.c
new file mode 100644
index 00000000000..0b5f72f17fd
--- /dev/null
+++ b/arch/blackfin/kernel/nmi.c
@@ -0,0 +1,299 @@
+/*
+ * Blackfin nmi_watchdog Driver
+ *
+ * Originally based on bfin_wdt.c
+ * Copyright 2010-2010 Analog Devices Inc.
+ *		Graff Yang <graf.yang@analog.com>
+ *
+ * Enter bugs at http://blackfin.uclinux.org/
+ *
+ * Licensed under the GPL-2 or later.
+ */
+
+#include <linux/bitops.h>
+#include <linux/hardirq.h>
+#include <linux/sysdev.h>
+#include <linux/pm.h>
+#include <linux/nmi.h>
+#include <linux/smp.h>
+#include <linux/timer.h>
+#include <asm/blackfin.h>
+#include <asm/atomic.h>
+#include <asm/cacheflush.h>
+#include <asm/bfin_watchdog.h>
+
+#define DRV_NAME "nmi-wdt"
+
+#define NMI_WDT_TIMEOUT 5          /* 5 seconds */
+#define NMI_CHECK_TIMEOUT (4 * HZ) /* 4 seconds in jiffies */
+static int nmi_wdt_cpu = 1;
+
+static unsigned int timeout = NMI_WDT_TIMEOUT;
+static int nmi_active;
+
+static unsigned short wdoga_ctl;
+static unsigned int wdoga_cnt;
+static struct corelock_slot saved_corelock;
+static atomic_t nmi_touched[NR_CPUS];
+static struct timer_list ntimer;
+
+enum {
+	COREA_ENTER_NMI = 0,
+	COREA_EXIT_NMI,
+	COREB_EXIT_NMI,
+
+	NMI_EVENT_NR,
+};
+static unsigned long nmi_event __attribute__ ((__section__(".l2.bss")));
+
+/* we are in nmi, non-atomic bit ops is safe */
+static inline void set_nmi_event(int event)
+{
+	__set_bit(event, &nmi_event);
+}
+
+static inline void wait_nmi_event(int event)
+{
+	while (!test_bit(event, &nmi_event))
+		barrier();
+	__clear_bit(event, &nmi_event);
+}
+
+static inline void send_corea_nmi(void)
+{
+	wdoga_ctl = bfin_read_WDOGA_CTL();
+	wdoga_cnt = bfin_read_WDOGA_CNT();
+
+	bfin_write_WDOGA_CTL(WDEN_DISABLE);
+	bfin_write_WDOGA_CNT(0);
+	bfin_write_WDOGA_CTL(WDEN_ENABLE | ICTL_NMI);
+}
+
+static inline void restore_corea_nmi(void)
+{
+	bfin_write_WDOGA_CTL(WDEN_DISABLE);
+	bfin_write_WDOGA_CTL(WDOG_EXPIRED | WDEN_DISABLE | ICTL_NONE);
+
+	bfin_write_WDOGA_CNT(wdoga_cnt);
+	bfin_write_WDOGA_CTL(wdoga_ctl);
+}
+
+static inline void save_corelock(void)
+{
+	saved_corelock = corelock;
+	corelock.lock = 0;
+}
+
+static inline void restore_corelock(void)
+{
+	corelock = saved_corelock;
+}
+
+
+static inline void nmi_wdt_keepalive(void)
+{
+	bfin_write_WDOGB_STAT(0);
+}
+
+static inline void nmi_wdt_stop(void)
+{
+	bfin_write_WDOGB_CTL(WDEN_DISABLE);
+}
+
+/* before calling this function, you must stop the WDT */
+static inline void nmi_wdt_clear(void)
+{
+	/* clear TRO bit, disable event generation */
+	bfin_write_WDOGB_CTL(WDOG_EXPIRED | WDEN_DISABLE | ICTL_NONE);
+}
+
+static inline void nmi_wdt_start(void)
+{
+	bfin_write_WDOGB_CTL(WDEN_ENABLE | ICTL_NMI);
+}
+
+static inline int nmi_wdt_running(void)
+{
+	return ((bfin_read_WDOGB_CTL() & WDEN_MASK) != WDEN_DISABLE);
+}
+
+static inline int nmi_wdt_set_timeout(unsigned long t)
+{
+	u32 cnt, max_t, sclk;
+	int run;
+
+	sclk = get_sclk();
+	max_t = -1 / sclk;
+	cnt = t * sclk;
+	if (t > max_t) {
+		pr_warning("NMI: timeout value is too large\n");
+		return -EINVAL;
+	}
+
+	run = nmi_wdt_running();
+	nmi_wdt_stop();
+	bfin_write_WDOGB_CNT(cnt);
+	if (run)
+		nmi_wdt_start();
+
+	timeout = t;
+
+	return 0;
+}
+
+int check_nmi_wdt_touched(void)
+{
+	unsigned int this_cpu = smp_processor_id();
+	unsigned int cpu;
+
+	cpumask_t mask = cpu_online_map;
+
+	if (!atomic_read(&nmi_touched[this_cpu]))
+		return 0;
+
+	atomic_set(&nmi_touched[this_cpu], 0);
+
+	cpu_clear(this_cpu, mask);
+	for_each_cpu_mask(cpu, mask) {
+		invalidate_dcache_range((unsigned long)(&nmi_touched[cpu]),
+				(unsigned long)(&nmi_touched[cpu]));
+		if (!atomic_read(&nmi_touched[cpu]))
+			return 0;
+		atomic_set(&nmi_touched[cpu], 0);
+	}
+
+	return 1;
+}
+
+static void nmi_wdt_timer(unsigned long data)
+{
+	if (check_nmi_wdt_touched())
+		nmi_wdt_keepalive();
+
+	mod_timer(&ntimer, jiffies + NMI_CHECK_TIMEOUT);
+}
+
+static int __init init_nmi_wdt(void)
+{
+	nmi_wdt_set_timeout(timeout);
+	nmi_wdt_start();
+	nmi_active = true;
+
+	init_timer(&ntimer);
+	ntimer.function = nmi_wdt_timer;
+	ntimer.expires = jiffies + NMI_CHECK_TIMEOUT;
+	add_timer(&ntimer);
+
+	pr_info("nmi_wdt: initialized: timeout=%d sec\n", timeout);
+	return 0;
+}
+device_initcall(init_nmi_wdt);
+
+void touch_nmi_watchdog(void)
+{
+	atomic_set(&nmi_touched[smp_processor_id()], 1);
+}
+
+/* Suspend/resume support */
+#ifdef CONFIG_PM
+static int nmi_wdt_suspend(struct sys_device *dev, pm_message_t state)
+{
+	nmi_wdt_stop();
+	return 0;
+}
+
+static int nmi_wdt_resume(struct sys_device *dev)
+{
+	if (nmi_active)
+		nmi_wdt_start();
+	return 0;
+}
+
+static struct sysdev_class nmi_sysclass = {
+	.name		= DRV_NAME,
+	.resume		= nmi_wdt_resume,
+	.suspend	= nmi_wdt_suspend,
+};
+
+static struct sys_device device_nmi_wdt = {
+	.id	= 0,
+	.cls	= &nmi_sysclass,
+};
+
+static int __init init_nmi_wdt_sysfs(void)
+{
+	int error;
+
+	if (!nmi_active)
+		return 0;
+
+	error = sysdev_class_register(&nmi_sysclass);
+	if (!error)
+		error = sysdev_register(&device_nmi_wdt);
+	return error;
+}
+late_initcall(init_nmi_wdt_sysfs);
+
+#endif	/* CONFIG_PM */
+
+
+asmlinkage notrace void do_nmi(struct pt_regs *fp)
+{
+	unsigned int cpu = smp_processor_id();
+	nmi_enter();
+
+	cpu_pda[cpu].__nmi_count += 1;
+
+	if (cpu == nmi_wdt_cpu) {
+		/* CoreB goes here first */
+
+		/* reload the WDOG_STAT */
+		nmi_wdt_keepalive();
+
+		/* clear nmi interrupt for CoreB */
+		nmi_wdt_stop();
+		nmi_wdt_clear();
+
+		/* trigger NMI interrupt of CoreA */
+		send_corea_nmi();
+
+		/* waiting CoreB to enter NMI */
+		wait_nmi_event(COREA_ENTER_NMI);
+
+		/* recover WDOGA's settings */
+		restore_corea_nmi();
+
+		save_corelock();
+
+		/* corelock is save/cleared, CoreA is dummping messages */
+
+		wait_nmi_event(COREA_EXIT_NMI);
+	} else {
+		/* OK, CoreA entered NMI */
+		set_nmi_event(COREA_ENTER_NMI);
+	}
+
+	pr_emerg("\nNMI Watchdog detected LOCKUP, dump for CPU %d\n", cpu);
+	dump_bfin_process(fp);
+	dump_bfin_mem(fp);
+	show_regs(fp);
+	dump_bfin_trace_buffer();
+	show_stack(current, (unsigned long *)fp);
+
+	if (cpu == nmi_wdt_cpu) {
+		pr_emerg("This fault is not recoverable, sorry!\n");
+
+		/* CoreA dump finished, restore the corelock */
+		restore_corelock();
+
+		set_nmi_event(COREB_EXIT_NMI);
+	} else {
+		/* CoreB dump finished, notice the CoreA we are done */
+		set_nmi_event(COREA_EXIT_NMI);
+
+		/* synchronize with CoreA */
+		wait_nmi_event(COREB_EXIT_NMI);
+	}
+
+	nmi_exit();
+}
diff --git a/arch/blackfin/kernel/process.c b/arch/blackfin/kernel/process.c
index b56b0e485e0..93ec07da2e5 100644
--- a/arch/blackfin/kernel/process.c
+++ b/arch/blackfin/kernel/process.c
@@ -11,6 +11,7 @@
 #include <linux/unistd.h>
 #include <linux/user.h>
 #include <linux/uaccess.h>
+#include <linux/slab.h>
 #include <linux/sched.h>
 #include <linux/tick.h>
 #include <linux/fs.h>
@@ -98,13 +99,6 @@ void cpu_idle(void)
 	}
 }
 
-/* Fill in the fpu structure for a core dump.  */
-
-int dump_fpu(struct pt_regs *regs, elf_fpregset_t * fpregs)
-{
-	return 1;
-}
-
 /*
  * This gets run with P1 containing the
  * function to call, and R1 containing
diff --git a/arch/blackfin/kernel/ptrace.c b/arch/blackfin/kernel/ptrace.c
index 65567dc4b9f..43eb969405d 100644
--- a/arch/blackfin/kernel/ptrace.c
+++ b/arch/blackfin/kernel/ptrace.c
@@ -1,6 +1,6 @@
 /*
  * linux/kernel/ptrace.c is by Ross Biro 1/23/92, edited by Linus Torvalds
- * these modifications are Copyright 2004-2009 Analog Devices Inc.
+ * these modifications are Copyright 2004-2010 Analog Devices Inc.
  *
  * Licensed under the GPL-2
  */
@@ -9,10 +9,13 @@
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/smp.h>
+#include <linux/elf.h>
 #include <linux/errno.h>
 #include <linux/ptrace.h>
 #include <linux/user.h>
+#include <linux/regset.h>
 #include <linux/signal.h>
+#include <linux/tracehook.h>
 #include <linux/uaccess.h>
 
 #include <asm/page.h>
@@ -25,90 +28,57 @@
 #include <asm/cacheflush.h>
 #include <asm/mem_map.h>
 
-#define TEXT_OFFSET 0
 /*
  * does not yet catch signals sent when the child dies.
  * in exit.c or in signal.c.
  */
 
-/* determines which bits in the SYSCFG reg the user has access to. */
-/* 1 = access 0 = no access */
-#define SYSCFG_MASK 0x0007	/* SYSCFG reg */
-/* sets the trace bits. */
-#define TRACE_BITS 0x0001
-
-/* Find the stack offset for a register, relative to thread.esp0. */
-#define PT_REG(reg)	((long)&((struct pt_regs *)0)->reg)
-
-/*
- * Get the address of the live pt_regs for the specified task.
- * These are saved onto the top kernel stack when the process
- * is not running.
- *
- * Note: if a user thread is execve'd from kernel space, the
- * kernel stack will not be empty on entry to the kernel, so
- * ptracing these tasks will fail.
- */
-static inline struct pt_regs *get_user_regs(struct task_struct *task)
-{
-	return (struct pt_regs *)
-	    ((unsigned long)task_stack_page(task) +
-	     (THREAD_SIZE - sizeof(struct pt_regs)));
-}
-
-/*
- * Get all user integer registers.
- */
-static inline int ptrace_getregs(struct task_struct *tsk, void __user *uregs)
-{
-	struct pt_regs regs;
-	memcpy(&regs, get_user_regs(tsk), sizeof(regs));
-	regs.usp = tsk->thread.usp;
-	return copy_to_user(uregs, &regs, sizeof(struct pt_regs)) ? -EFAULT : 0;
-}
-
-/* Mapping from PT_xxx to the stack offset at which the register is
- * saved.  Notice that usp has no stack-slot and needs to be treated
- * specially (see get_reg/put_reg below).
- */
-
 /*
  * Get contents of register REGNO in task TASK.
  */
-static inline long get_reg(struct task_struct *task, int regno)
+static inline long
+get_reg(struct task_struct *task, long regno, unsigned long __user *datap)
 {
-	unsigned char *reg_ptr;
+	long tmp;
+	struct pt_regs *regs = task_pt_regs(task);
 
-	struct pt_regs *regs =
-	    (struct pt_regs *)((unsigned long)task_stack_page(task) +
-			       (THREAD_SIZE - sizeof(struct pt_regs)));
-	reg_ptr = (char *)regs;
+	if (regno & 3 || regno > PT_LAST_PSEUDO || regno < 0)
+		return -EIO;
 
 	switch (regno) {
+	case PT_TEXT_ADDR:
+		tmp = task->mm->start_code;
+		break;
+	case PT_TEXT_END_ADDR:
+		tmp = task->mm->end_code;
+		break;
+	case PT_DATA_ADDR:
+		tmp = task->mm->start_data;
+		break;
 	case PT_USP:
-		return task->thread.usp;
+		tmp = task->thread.usp;
+		break;
 	default:
-		if (regno <= 216)
-			return *(long *)(reg_ptr + regno);
+		if (regno < sizeof(*regs)) {
+			void *reg_ptr = regs;
+			tmp = *(long *)(reg_ptr + regno);
+		} else
+			return -EIO;
 	}
-	/* slight mystery ... never seems to come here but kernel misbehaves without this code! */
 
-	printk(KERN_WARNING "Request to get for unknown register %d\n", regno);
-	return 0;
+	return put_user(tmp, datap);
 }
 
 /*
  * Write contents of register REGNO in task TASK.
  */
 static inline int
-put_reg(struct task_struct *task, int regno, unsigned long data)
+put_reg(struct task_struct *task, long regno, unsigned long data)
 {
-	char *reg_ptr;
+	struct pt_regs *regs = task_pt_regs(task);
 
-	struct pt_regs *regs =
-	    (struct pt_regs *)((unsigned long)task_stack_page(task) +
-			       (THREAD_SIZE - sizeof(struct pt_regs)));
-	reg_ptr = (char *)regs;
+	if (regno & 3 || regno > PT_LAST_PSEUDO || regno < 0)
+		return -EIO;
 
 	switch (regno) {
 	case PT_PC:
@@ -125,10 +95,18 @@ put_reg(struct task_struct *task, int regno, unsigned long data)
 		regs->usp = data;
 		task->thread.usp = data;
 		break;
+	case PT_SYSCFG:	/* don't let userspace screw with this */
+		if ((data & ~1) != 0x6)
+			pr_warning("ptrace: ignore syscfg write of %#lx\n", data);
+		break;		/* regs->syscfg = data; break; */
 	default:
-		if (regno <= 216)
-			*(long *)(reg_ptr + regno) = data;
+		if (regno < sizeof(*regs)) {
+			void *reg_offset = regs;
+			*(long *)(reg_offset + regno) = data;
+		}
+		/* Ignore writes to pseudo registers */
 	}
+
 	return 0;
 }
 
@@ -160,24 +138,98 @@ static inline int is_user_addr_valid(struct task_struct *child,
 	return -EIO;
 }
 
-void ptrace_enable(struct task_struct *child)
+/*
+ * retrieve the contents of Blackfin userspace general registers
+ */
+static int genregs_get(struct task_struct *target,
+		       const struct user_regset *regset,
+		       unsigned int pos, unsigned int count,
+		       void *kbuf, void __user *ubuf)
 {
-	unsigned long tmp;
-	tmp = get_reg(child, PT_SYSCFG) | (TRACE_BITS);
-	put_reg(child, PT_SYSCFG, tmp);
+	struct pt_regs *regs = task_pt_regs(target);
+	int ret;
+
+	/* This sucks ... */
+	regs->usp = target->thread.usp;
+
+	ret = user_regset_copyout(&pos, &count, &kbuf, &ubuf,
+				  regs, 0, sizeof(*regs));
+	if (ret < 0)
+		return ret;
+
+	return user_regset_copyout_zero(&pos, &count, &kbuf, &ubuf,
+					sizeof(*regs), -1);
 }
 
 /*
- * Called by kernel/ptrace.c when detaching..
- *
- * Make sure the single step bit is not set.
+ * update the contents of the Blackfin userspace general registers
+ */
+static int genregs_set(struct task_struct *target,
+		       const struct user_regset *regset,
+		       unsigned int pos, unsigned int count,
+		       const void *kbuf, const void __user *ubuf)
+{
+	struct pt_regs *regs = task_pt_regs(target);
+	int ret;
+
+	/* Don't let people set SYSCFG (it's at the end of pt_regs) */
+	ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf,
+				 regs, 0, PT_SYSCFG);
+	if (ret < 0)
+		return ret;
+
+	/* This sucks ... */
+	target->thread.usp = regs->usp;
+	/* regs->retx = regs->pc; */
+
+	return user_regset_copyin_ignore(&pos, &count, &kbuf, &ubuf,
+					PT_SYSCFG, -1);
+}
+
+/*
+ * Define the register sets available on the Blackfin under Linux
  */
-void ptrace_disable(struct task_struct *child)
+enum bfin_regset {
+	REGSET_GENERAL,
+};
+
+static const struct user_regset bfin_regsets[] = {
+	[REGSET_GENERAL] = {
+		.core_note_type = NT_PRSTATUS,
+		.n              = sizeof(struct pt_regs) / sizeof(long),
+		.size           = sizeof(long),
+		.align          = sizeof(long),
+		.get            = genregs_get,
+		.set            = genregs_set,
+	},
+};
+
+static const struct user_regset_view user_bfin_native_view = {
+	.name      = "Blackfin",
+	.e_machine = EM_BLACKFIN,
+	.regsets   = bfin_regsets,
+	.n         = ARRAY_SIZE(bfin_regsets),
+};
+
+const struct user_regset_view *task_user_regset_view(struct task_struct *task)
+{
+	return &user_bfin_native_view;
+}
+
+void user_enable_single_step(struct task_struct *child)
+{
+	struct pt_regs *regs = task_pt_regs(child);
+	regs->syscfg |= SYSCFG_SSSTEP;
+
+	set_tsk_thread_flag(child, TIF_SINGLESTEP);
+}
+
+void user_disable_single_step(struct task_struct *child)
 {
-	unsigned long tmp;
-	/* make sure the single step bit is not set. */
-	tmp = get_reg(child, PT_SYSCFG) & ~TRACE_BITS;
-	put_reg(child, PT_SYSCFG, tmp);
+	struct pt_regs *regs = task_pt_regs(child);
+	regs->syscfg &= ~SYSCFG_SSSTEP;
+
+	clear_tsk_thread_flag(child, TIF_SINGLESTEP);
 }
 
 long arch_ptrace(struct task_struct *child, long request, long addr, long data)
@@ -240,40 +292,6 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 			break;
 		}
 
-		/* read the word at location addr in the USER area. */
-	case PTRACE_PEEKUSR:
-		{
-			unsigned long tmp;
-			ret = -EIO;
-			tmp = 0;
-			if ((addr & 3) || (addr > (sizeof(struct pt_regs) + 16))) {
-				printk(KERN_WARNING "ptrace error : PEEKUSR : temporarily returning "
-				                    "0 - %x sizeof(pt_regs) is %lx\n",
-				     (int)addr, sizeof(struct pt_regs));
-				break;
-			}
-			if (addr == sizeof(struct pt_regs)) {
-				/* PT_TEXT_ADDR */
-				tmp = child->mm->start_code + TEXT_OFFSET;
-			} else if (addr == (sizeof(struct pt_regs) + 4)) {
-				/* PT_TEXT_END_ADDR */
-				tmp = child->mm->end_code;
-			} else if (addr == (sizeof(struct pt_regs) + 8)) {
-				/* PT_DATA_ADDR */
-				tmp = child->mm->start_data;
-#ifdef CONFIG_BINFMT_ELF_FDPIC
-			} else if (addr == (sizeof(struct pt_regs) + 12)) {
-				goto case_PTRACE_GETFDPIC_EXEC;
-			} else if (addr == (sizeof(struct pt_regs) + 16)) {
-				goto case_PTRACE_GETFDPIC_INTERP;
-#endif
-			} else {
-				tmp = get_reg(child, addr);
-			}
-			ret = put_user(tmp, datap);
-			break;
-		}
-
 #ifdef CONFIG_BINFMT_ELF_FDPIC
 	case PTRACE_GETFDPIC: {
 		unsigned long tmp = 0;
@@ -336,78 +354,36 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 			break;
 		}
 
-	case PTRACE_POKEUSR:	/* write the word at location addr in the USER area */
-		ret = -EIO;
-		if ((addr & 3) || (addr > (sizeof(struct pt_regs) + 16))) {
-			printk(KERN_WARNING "ptrace error : POKEUSR: temporarily returning 0\n");
-			break;
-		}
-
-		if (addr >= (sizeof(struct pt_regs))) {
-			ret = 0;
-			break;
-		}
-		if (addr == PT_SYSCFG) {
-			data &= SYSCFG_MASK;
-			data |= get_reg(child, PT_SYSCFG);
+	case PTRACE_PEEKUSR:
+		switch (addr) {
+#ifdef CONFIG_BINFMT_ELF_FDPIC	/* backwards compat */
+		case PT_FDPIC_EXEC:   goto case_PTRACE_GETFDPIC_EXEC;
+		case PT_FDPIC_INTERP: goto case_PTRACE_GETFDPIC_INTERP;
+#endif
+		default:
+			ret = get_reg(child, addr, datap);
 		}
-		ret = put_reg(child, addr, data);
+		pr_debug("ptrace: PEEKUSR reg %li with %#lx = %i\n", addr, data, ret);
 		break;
 
-	case PTRACE_SYSCALL:	/* continue and stop at next (return from) syscall */
-	case PTRACE_CONT:	/* restart after signal. */
-		pr_debug("ptrace: syscall/cont\n");
-
-		ret = -EIO;
-		if (!valid_signal(data))
-			break;
-		if (request == PTRACE_SYSCALL)
-			set_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
-		else
-			clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
-		child->exit_code = data;
-		ptrace_disable(child);
-		pr_debug("ptrace: before wake_up_process\n");
-		wake_up_process(child);
-		ret = 0;
-		break;
-
-	/*
-	 * make the child exit.  Best I can do is send it a sigkill.
-	 * perhaps it should be put in the status that it wants to
-	 * exit.
-	 */
-	case PTRACE_KILL:
-		ret = 0;
-		if (child->exit_state == EXIT_ZOMBIE)	/* already dead */
-			break;
-		child->exit_code = SIGKILL;
-		ptrace_disable(child);
-		wake_up_process(child);
-		break;
-
-	case PTRACE_SINGLESTEP:	/* set the trap flag. */
-		pr_debug("ptrace: single step\n");
-		ret = -EIO;
-		if (!valid_signal(data))
-			break;
-		clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE);
-		ptrace_enable(child);
-		child->exit_code = data;
-		wake_up_process(child);
-		ret = 0;
+	case PTRACE_POKEUSR:
+		ret = put_reg(child, addr, data);
+		pr_debug("ptrace: POKEUSR reg %li with %li = %i\n", addr, data, ret);
 		break;
 
 	case PTRACE_GETREGS:
-		/* Get all gp regs from the child. */
-		ret = ptrace_getregs(child, datap);
-		break;
+		pr_debug("ptrace: PTRACE_GETREGS\n");
+		return copy_regset_to_user(child, &user_bfin_native_view,
+					   REGSET_GENERAL,
+					   0, sizeof(struct pt_regs),
+					   (void __user *)data);
 
 	case PTRACE_SETREGS:
-		printk(KERN_WARNING "ptrace: SETREGS: **** NOT IMPLEMENTED ***\n");
-		/* Set all gp regs in the child. */
-		ret = 0;
-		break;
+		pr_debug("ptrace: PTRACE_SETREGS\n");
+		return copy_regset_from_user(child, &user_bfin_native_view,
+					     REGSET_GENERAL,
+					     0, sizeof(struct pt_regs),
+					     (const void __user *)data);
 
 	default:
 		ret = ptrace_request(child, request, addr, data);
@@ -417,27 +393,21 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 	return ret;
 }
 
-asmlinkage void syscall_trace(void)
+asmlinkage int syscall_trace_enter(struct pt_regs *regs)
 {
-	if (!test_thread_flag(TIF_SYSCALL_TRACE))
-		return;
-
-	if (!(current->ptrace & PT_PTRACED))
-		return;
-
-	/* the 0x80 provides a way for the tracing parent to distinguish
-	 * between a syscall stop and SIGTRAP delivery
-	 */
-	ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD)
-				 ? 0x80 : 0));
-
-	/*
-	 * this isn't the same as continuing with a signal, but it will do
-	 * for normal use.  strace only continues with a signal if the
-	 * stopping signal is not SIGTRAP.  -brl
-	 */
-	if (current->exit_code) {
-		send_sig(current->exit_code, current, 1);
-		current->exit_code = 0;
-	}
+	int ret = 0;
+
+	if (test_thread_flag(TIF_SYSCALL_TRACE))
+		ret = tracehook_report_syscall_entry(regs);
+
+	return ret;
+}
+
+asmlinkage void syscall_trace_leave(struct pt_regs *regs)
+{
+	int step;
+
+	step = test_thread_flag(TIF_SINGLESTEP);
+	if (step || test_thread_flag(TIF_SYSCALL_TRACE))
+		tracehook_report_syscall_exit(regs, step);
 }
diff --git a/arch/blackfin/kernel/setup.c b/arch/blackfin/kernel/setup.c
index 95448ae9c43..8e2efceb364 100644
--- a/arch/blackfin/kernel/setup.c
+++ b/arch/blackfin/kernel/setup.c
@@ -220,6 +220,16 @@ void __init bfin_relocate_l1_mem(void)
 		memcpy(_stext_l2, _l2_lma, l2_len);
 }
 
+#ifdef CONFIG_ROMKERNEL
+void __init bfin_relocate_xip_data(void)
+{
+	early_shadow_stamp();
+
+	memcpy(_sdata, _data_lma, (unsigned long)_data_len - THREAD_SIZE + sizeof(struct thread_info));
+	memcpy(_sinitdata, _init_data_lma, (unsigned long)_init_data_len);
+}
+#endif
+
 /* add_memory_region to memmap */
 static void __init add_memory_region(unsigned long long start,
 			      unsigned long long size, int type)
@@ -504,7 +514,7 @@ static __init void memory_setup(void)
 #endif
 	unsigned long max_mem;
 
-	_rambase = (unsigned long)_stext;
+	_rambase = CONFIG_BOOT_LOAD;
 	_ramstart = (unsigned long)_end;
 
 	if (DMA_UNCACHED_REGION > (_ramend - _ramstart)) {
@@ -597,7 +607,12 @@ static __init void memory_setup(void)
 	}
 
 #ifdef CONFIG_MPU
+#if defined(CONFIG_ROMFS_ON_MTD) && defined(CONFIG_MTD_ROM)
+	page_mask_nelts = (((_ramend + ASYNC_BANK3_BASE + ASYNC_BANK3_SIZE -
+					ASYNC_BANK0_BASE) >> PAGE_SHIFT) + 31) / 32;
+#else
 	page_mask_nelts = ((_ramend >> PAGE_SHIFT) + 31) / 32;
+#endif
 	page_mask_order = get_order(3 * page_mask_nelts * sizeof(long));
 #endif
 
@@ -630,7 +645,7 @@ static __init void memory_setup(void)
 		__bss_start, __bss_stop,
 		_sdata, _edata,
 		(void *)&init_thread_union,
-		(void *)((int)(&init_thread_union) + 0x2000),
+		(void *)((int)(&init_thread_union) + THREAD_SIZE),
 		__init_begin, __init_end,
 		(void *)_ramstart, (void *)memory_end
 #ifdef CONFIG_MTD_UCLINUX
@@ -792,10 +807,17 @@ static inline int __init get_mem_size(void)
 	BUG();
 }
 
+__attribute__((weak))
+void __init native_machine_early_platform_add_devices(void)
+{
+}
+
 void __init setup_arch(char **cmdline_p)
 {
 	unsigned long sclk, cclk;
 
+	native_machine_early_platform_add_devices();
+
 	enable_shadow_console();
 
 	/* Check to make sure we are running on the right processor */
@@ -1217,10 +1239,10 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 		   dsup_banks, BFIN_DSUBBANKS, BFIN_DWAYS,
 		   BFIN_DLINES);
 #ifdef __ARCH_SYNC_CORE_DCACHE
-	seq_printf(m, "SMP Dcache Flushes\t: %lu\n\n", cpudata->dcache_invld_count);
+	seq_printf(m, "SMP Dcache Flushes\t: %lu\n\n", dcache_invld_count[cpu_num]);
 #endif
 #ifdef __ARCH_SYNC_CORE_ICACHE
-	seq_printf(m, "SMP Icache Flushes\t: %lu\n\n", cpudata->icache_invld_count);
+	seq_printf(m, "SMP Icache Flushes\t: %lu\n\n", icache_invld_count[cpu_num]);
 #endif
 
 	if (cpu_num != num_possible_cpus() - 1)
@@ -1249,8 +1271,8 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 	seq_printf(m, "board memory\t: %ld kB (0x%p -> 0x%p)\n",
 		 physical_mem_end >> 10, (void *)0, (void *)physical_mem_end);
 	seq_printf(m, "kernel memory\t: %d kB (0x%p -> 0x%p)\n",
-		((int)memory_end - (int)_stext) >> 10,
-		_stext,
+		((int)memory_end - (int)_rambase) >> 10,
+		(void *)_rambase,
 		(void *)memory_end);
 	seq_printf(m, "\n");
 
diff --git a/arch/blackfin/kernel/signal.c b/arch/blackfin/kernel/signal.c
index e0fd63e9e38..d536f35d1f4 100644
--- a/arch/blackfin/kernel/signal.c
+++ b/arch/blackfin/kernel/signal.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2004-2009 Analog Devices Inc.
+ * Copyright 2004-2010 Analog Devices Inc.
  *
  * Licensed under the GPL-2 or later
  */
@@ -17,6 +17,7 @@
 #include <asm/cacheflush.h>
 #include <asm/ucontext.h>
 #include <asm/fixed_code.h>
+#include <asm/syscall.h>
 
 #define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
 
@@ -50,6 +51,9 @@ rt_restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *p
 	unsigned long usp = 0;
 	int err = 0;
 
+	/* Always make any pending restarted system calls return -EINTR */
+	current_thread_info()->restart_block.fn = do_no_restart_syscall;
+
 #define RESTORE(x) err |= __get_user(regs->x, &sc->sc_##x)
 
 	/* restore passed registers */
@@ -206,16 +210,6 @@ setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t * info,
 	regs->r1 = (unsigned long)(&frame->info);
 	regs->r2 = (unsigned long)(&frame->uc);
 
-	/*
-	 * Clear the trace flag when entering the signal handler, but
-	 * notify any tracer that was single-stepping it. The tracer
-	 * may want to single-step inside the handler too.
-	 */
-	if (regs->syscfg & TRACE_BITS) {
-		regs->syscfg &= ~TRACE_BITS;
-		ptrace_notify(SIGTRAP);
-	}
-
 	return 0;
 
  give_sigsegv:
@@ -247,6 +241,11 @@ handle_restart(struct pt_regs *regs, struct k_sigaction *ka, int has_handler)
 		regs->r0 = regs->orig_r0;
 		regs->pc -= 2;
 		break;
+
+	case -ERESTART_RESTARTBLOCK:
+		regs->p0 = __NR_restart_syscall;
+		regs->pc -= 2;
+		break;
 	}
 }
 
@@ -315,6 +314,9 @@ asmlinkage void do_signal(struct pt_regs *regs)
 			 * clear the TIF_RESTORE_SIGMASK flag */
 			if (test_thread_flag(TIF_RESTORE_SIGMASK))
 				clear_thread_flag(TIF_RESTORE_SIGMASK);
+
+			tracehook_signal_handler(signr, &info, &ka, regs,
+				test_thread_flag(TIF_SINGLESTEP));
 		}
 
 		return;
diff --git a/arch/blackfin/kernel/time-ts.c b/arch/blackfin/kernel/time-ts.c
index 17c38c5b5b2..cb7a01d4f00 100644
--- a/arch/blackfin/kernel/time-ts.c
+++ b/arch/blackfin/kernel/time-ts.c
@@ -21,6 +21,7 @@
 #include <asm/blackfin.h>
 #include <asm/time.h>
 #include <asm/gptimers.h>
+#include <asm/nmi.h>
 
 /* Accelerators for sched_clock()
  * convert from cycles(64bits) => nanoseconds (64bits)
@@ -50,7 +51,11 @@
 
 static notrace cycle_t bfin_read_cycles(struct clocksource *cs)
 {
+#ifdef CONFIG_CPU_FREQ
 	return __bfin_cycles_off + (get_cycles() << __bfin_cycles_mod);
+#else
+	return get_cycles();
+#endif
 }
 
 static struct clocksource bfin_cs_cycles = {
@@ -132,7 +137,6 @@ static int __init bfin_cs_gptimer0_init(void)
 # define bfin_cs_gptimer0_init()
 #endif
 
-
 #if defined(CONFIG_GPTMR0_CLOCKSOURCE) || defined(CONFIG_CYCLES_CLOCKSOURCE)
 /* prefer to use cycles since it has higher rating */
 notrace unsigned long long sched_clock(void)
@@ -145,47 +149,8 @@ notrace unsigned long long sched_clock(void)
 }
 #endif
 
-#ifdef CONFIG_CORE_TIMER_IRQ_L1
-__attribute__((l1_text))
-#endif
-irqreturn_t timer_interrupt(int irq, void *dev_id);
-
-static int bfin_timer_set_next_event(unsigned long, \
-		struct clock_event_device *);
-
-static void bfin_timer_set_mode(enum clock_event_mode, \
-		struct clock_event_device *);
-
-static struct clock_event_device clockevent_bfin = {
-#if defined(CONFIG_TICKSOURCE_GPTMR0)
-	.name		= "bfin_gptimer0",
-	.rating		= 300,
-	.irq		= IRQ_TIMER0,
-#else
-	.name		= "bfin_core_timer",
-	.rating		= 350,
-	.irq		= IRQ_CORETMR,
-#endif
-	.shift		= 32,
-	.features	= CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT,
-	.set_next_event = bfin_timer_set_next_event,
-	.set_mode	= bfin_timer_set_mode,
-};
-
-static struct irqaction bfin_timer_irq = {
-#if defined(CONFIG_TICKSOURCE_GPTMR0)
-	.name		= "Blackfin GPTimer0",
-#else
-	.name		= "Blackfin CoreTimer",
-#endif
-	.flags		= IRQF_DISABLED | IRQF_TIMER | \
-			  IRQF_IRQPOLL | IRQF_PERCPU,
-	.handler	= timer_interrupt,
-	.dev_id		= &clockevent_bfin,
-};
-
 #if defined(CONFIG_TICKSOURCE_GPTMR0)
-static int bfin_timer_set_next_event(unsigned long cycles,
+static int bfin_gptmr0_set_next_event(unsigned long cycles,
                                      struct clock_event_device *evt)
 {
 	disable_gptimers(TIMER0bit);
@@ -196,7 +161,7 @@ static int bfin_timer_set_next_event(unsigned long cycles,
 	return 0;
 }
 
-static void bfin_timer_set_mode(enum clock_event_mode mode,
+static void bfin_gptmr0_set_mode(enum clock_event_mode mode,
 				struct clock_event_device *evt)
 {
 	switch (mode) {
@@ -224,25 +189,65 @@ static void bfin_timer_set_mode(enum clock_event_mode mode,
 	}
 }
 
-static void bfin_timer_ack(void)
+static void bfin_gptmr0_ack(void)
 {
 	set_gptimer_status(TIMER_GROUP1, TIMER_STATUS_TIMIL0);
 }
 
-static void __init bfin_timer_init(void)
+static void __init bfin_gptmr0_init(void)
 {
 	disable_gptimers(TIMER0bit);
 }
 
-static unsigned long  __init bfin_clockevent_check(void)
+#ifdef CONFIG_CORE_TIMER_IRQ_L1
+__attribute__((l1_text))
+#endif
+irqreturn_t bfin_gptmr0_interrupt(int irq, void *dev_id)
 {
-	setup_irq(IRQ_TIMER0, &bfin_timer_irq);
-	return get_sclk();
+	struct clock_event_device *evt = dev_id;
+	smp_mb();
+	evt->event_handler(evt);
+	bfin_gptmr0_ack();
+	return IRQ_HANDLED;
 }
 
-#else /* CONFIG_TICKSOURCE_CORETMR */
+static struct irqaction gptmr0_irq = {
+	.name		= "Blackfin GPTimer0",
+	.flags		= IRQF_DISABLED | IRQF_TIMER | \
+			  IRQF_IRQPOLL | IRQF_PERCPU,
+	.handler	= bfin_gptmr0_interrupt,
+};
 
-static int bfin_timer_set_next_event(unsigned long cycles,
+static struct clock_event_device clockevent_gptmr0 = {
+	.name		= "bfin_gptimer0",
+	.rating		= 300,
+	.irq		= IRQ_TIMER0,
+	.shift		= 32,
+	.features	= CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT,
+	.set_next_event = bfin_gptmr0_set_next_event,
+	.set_mode	= bfin_gptmr0_set_mode,
+};
+
+static void __init bfin_gptmr0_clockevent_init(struct clock_event_device *evt)
+{
+	unsigned long clock_tick;
+
+	clock_tick = get_sclk();
+	evt->mult = div_sc(clock_tick, NSEC_PER_SEC, evt->shift);
+	evt->max_delta_ns = clockevent_delta2ns(-1, evt);
+	evt->min_delta_ns = clockevent_delta2ns(100, evt);
+
+	evt->cpumask = cpumask_of(0);
+
+	clockevents_register_device(evt);
+}
+#endif /* CONFIG_TICKSOURCE_GPTMR0 */
+
+#if defined(CONFIG_TICKSOURCE_CORETMR)
+/* per-cpu local core timer */
+static DEFINE_PER_CPU(struct clock_event_device, coretmr_events);
+
+static int bfin_coretmr_set_next_event(unsigned long cycles,
 				struct clock_event_device *evt)
 {
 	bfin_write_TCNTL(TMPWR);
@@ -253,7 +258,7 @@ static int bfin_timer_set_next_event(unsigned long cycles,
 	return 0;
 }
 
-static void bfin_timer_set_mode(enum clock_event_mode mode,
+static void bfin_coretmr_set_mode(enum clock_event_mode mode,
 				struct clock_event_device *evt)
 {
 	switch (mode) {
@@ -285,19 +290,13 @@ static void bfin_timer_set_mode(enum clock_event_mode mode,
 	}
 }
 
-static void bfin_timer_ack(void)
-{
-}
-
-static void __init bfin_timer_init(void)
+void bfin_coretmr_init(void)
 {
 	/* power up the timer, but don't enable it just yet */
 	bfin_write_TCNTL(TMPWR);
 	CSYNC();
 
-	/*
-	 * the TSCALE prescaler counter.
-	 */
+	/* the TSCALE prescaler counter. */
 	bfin_write_TSCALE(TIME_SCALE - 1);
 	bfin_write_TPERIOD(0);
 	bfin_write_TCOUNT(0);
@@ -305,48 +304,54 @@ static void __init bfin_timer_init(void)
 	CSYNC();
 }
 
-static unsigned long  __init bfin_clockevent_check(void)
-{
-	setup_irq(IRQ_CORETMR, &bfin_timer_irq);
-	return get_cclk() / TIME_SCALE;
-}
-
-void __init setup_core_timer(void)
+#ifdef CONFIG_CORE_TIMER_IRQ_L1
+__attribute__((l1_text))
+#endif
+irqreturn_t bfin_coretmr_interrupt(int irq, void *dev_id)
 {
-	bfin_timer_init();
-	bfin_timer_set_mode(CLOCK_EVT_MODE_PERIODIC, NULL);
-}
-#endif /* CONFIG_TICKSOURCE_GPTMR0 */
+	int cpu = smp_processor_id();
+	struct clock_event_device *evt = &per_cpu(coretmr_events, cpu);
 
-/*
- * timer_interrupt() needs to keep up the real-time clock,
- * as well as call the "do_timer()" routine every clocktick
- */
-irqreturn_t timer_interrupt(int irq, void *dev_id)
-{
-	struct clock_event_device *evt = dev_id;
 	smp_mb();
 	evt->event_handler(evt);
-	bfin_timer_ack();
-	return IRQ_HANDLED;
-}
-
-static int __init bfin_clockevent_init(void)
-{
-	unsigned long timer_clk;
 
-	timer_clk = bfin_clockevent_check();
+	touch_nmi_watchdog();
 
-	bfin_timer_init();
+	return IRQ_HANDLED;
+}
 
-	clockevent_bfin.mult = div_sc(timer_clk, NSEC_PER_SEC, clockevent_bfin.shift);
-	clockevent_bfin.max_delta_ns = clockevent_delta2ns(-1, &clockevent_bfin);
-	clockevent_bfin.min_delta_ns = clockevent_delta2ns(100, &clockevent_bfin);
-	clockevent_bfin.cpumask = cpumask_of(0);
-	clockevents_register_device(&clockevent_bfin);
+static struct irqaction coretmr_irq = {
+	.name		= "Blackfin CoreTimer",
+	.flags		= IRQF_DISABLED | IRQF_TIMER | \
+			  IRQF_IRQPOLL | IRQF_PERCPU,
+	.handler	= bfin_coretmr_interrupt,
+};
 
-	return 0;
+void bfin_coretmr_clockevent_init(void)
+{
+	unsigned long clock_tick;
+	unsigned int cpu = smp_processor_id();
+	struct clock_event_device *evt = &per_cpu(coretmr_events, cpu);
+
+	evt->name = "bfin_core_timer";
+	evt->rating = 350;
+	evt->irq = -1;
+	evt->shift = 32;
+	evt->features = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT;
+	evt->set_next_event = bfin_coretmr_set_next_event;
+	evt->set_mode = bfin_coretmr_set_mode;
+
+	clock_tick = get_cclk() / TIME_SCALE;
+	evt->mult = div_sc(clock_tick, NSEC_PER_SEC, evt->shift);
+	evt->max_delta_ns = clockevent_delta2ns(-1, evt);
+	evt->min_delta_ns = clockevent_delta2ns(100, evt);
+
+	evt->cpumask = cpumask_of(cpu);
+
+	clockevents_register_device(evt);
 }
+#endif /* CONFIG_TICKSOURCE_CORETMR */
+
 
 void __init time_init(void)
 {
@@ -370,5 +375,21 @@ void __init time_init(void)
 
 	bfin_cs_cycles_init();
 	bfin_cs_gptimer0_init();
-	bfin_clockevent_init();
+
+#if defined(CONFIG_TICKSOURCE_CORETMR)
+	bfin_coretmr_init();
+	setup_irq(IRQ_CORETMR, &coretmr_irq);
+	bfin_coretmr_clockevent_init();
+#endif
+
+#if defined(CONFIG_TICKSOURCE_GPTMR0)
+	bfin_gptmr0_init();
+	setup_irq(IRQ_TIMER0, &gptmr0_irq);
+	gptmr0_irq.dev_id = &clockevent_gptmr0;
+	bfin_gptmr0_clockevent_init(&clockevent_gptmr0);
+#endif
+
+#if !defined(CONFIG_TICKSOURCE_CORETMR) && !defined(CONFIG_TICKSOURCE_GPTMR0)
+# error at least one clock event device is required
+#endif
 }
diff --git a/arch/blackfin/kernel/traps.c b/arch/blackfin/kernel/traps.c
index d3cbcd6bd98..ba70c4bc269 100644
--- a/arch/blackfin/kernel/traps.c
+++ b/arch/blackfin/kernel/traps.c
@@ -138,6 +138,12 @@ static void decode_address(char *buf, unsigned long address)
 		if (!mm)
 			continue;
 
+		if (!down_read_trylock(&mm->mmap_sem)) {
+			if (!in_atomic)
+				mmput(mm);
+			continue;
+		}
+
 		for (n = rb_first(&mm->mm_rb); n; n = rb_next(n)) {
 			struct vm_area_struct *vma;
 
@@ -177,6 +183,7 @@ static void decode_address(char *buf, unsigned long address)
 					sprintf(buf, "[ %s vma:0x%lx-0x%lx]",
 						name, vma->vm_start, vma->vm_end);
 
+				up_read(&mm->mmap_sem);
 				if (!in_atomic)
 					mmput(mm);
 
@@ -186,11 +193,16 @@ static void decode_address(char *buf, unsigned long address)
 				goto done;
 			}
 		}
+
+		up_read(&mm->mmap_sem);
 		if (!in_atomic)
 			mmput(mm);
 	}
 
-	/* we were unable to find this address anywhere */
+	/*
+	 * we were unable to find this address anywhere,
+	 * or some MMs were skipped because they were in use.
+	 */
 	sprintf(buf, "/* kernel dynamic memory */");
 
 done:
@@ -248,9 +260,7 @@ asmlinkage notrace void trap_c(struct pt_regs *fp)
 #ifdef CONFIG_DEBUG_BFIN_HWTRACE_ON
 	int j;
 #endif
-#ifdef CONFIG_DEBUG_HUNT_FOR_ZERO
 	unsigned int cpu = raw_smp_processor_id();
-#endif
 	const char *strerror = NULL;
 	int sig = 0;
 	siginfo_t info;
@@ -639,7 +649,17 @@ asmlinkage notrace void trap_c(struct pt_regs *fp)
 	{
 		info.si_signo = sig;
 		info.si_errno = 0;
-		info.si_addr = (void __user *)fp->pc;
+		switch (trapnr) {
+		case VEC_CPLB_VL:
+		case VEC_MISALI_D:
+		case VEC_CPLB_M:
+		case VEC_CPLB_MHIT:
+			info.si_addr = (void __user *)cpu_pda[cpu].dcplb_fault_addr;
+			break;
+		default:
+			info.si_addr = (void __user *)fp->pc;
+			break;
+		}
 		force_sig_info(sig, &info, current);
 	}
 
@@ -712,7 +732,7 @@ static void decode_instruction(unsigned short *address)
 			verbose_printk("RTE");
 		else if (opcode == 0x0025)
 			verbose_printk("EMUEXCPT");
-		else if (opcode == 0x0040 && opcode <= 0x0047)
+		else if (opcode >= 0x0040 && opcode <= 0x0047)
 			verbose_printk("STI R%i", opcode & 7);
 		else if (opcode >= 0x0050 && opcode <= 0x0057)
 			verbose_printk("JUMP (P%i)", opcode & 7);
@@ -1096,7 +1116,7 @@ void dump_bfin_mem(struct pt_regs *fp)
 			/* And the last RETI points to the current userspace context */
 			if ((fp + 1)->pc >= current->mm->start_code &&
 			    (fp + 1)->pc <= current->mm->end_code) {
-				verbose_printk(KERN_NOTICE "It might be better to look around here : \n");
+				verbose_printk(KERN_NOTICE "It might be better to look around here :\n");
 				verbose_printk(KERN_NOTICE "-------------------------------------------\n");
 				show_regs(fp + 1);
 				verbose_printk(KERN_NOTICE "-------------------------------------------\n");
diff --git a/arch/blackfin/kernel/vmlinux.lds.S b/arch/blackfin/kernel/vmlinux.lds.S
index 66799e763dc..984c7817239 100644
--- a/arch/blackfin/kernel/vmlinux.lds.S
+++ b/arch/blackfin/kernel/vmlinux.lds.S
@@ -15,7 +15,12 @@ _jiffies = _jiffies_64;
 
 SECTIONS
 {
+#ifdef CONFIG_RAMKERNEL
 	. = CONFIG_BOOT_LOAD;
+#else
+	. = CONFIG_ROM_BASE;
+#endif
+
 	/* Neither the text, ro_data or bss section need to be aligned
 	 * So pack them back to back
 	 */
@@ -31,6 +36,12 @@ SECTIONS
 		LOCK_TEXT
 		IRQENTRY_TEXT
 		KPROBES_TEXT
+#ifdef CONFIG_ROMKERNEL
+		__sinittext = .;
+		INIT_TEXT
+		__einittext = .;
+		EXIT_TEXT
+#endif
 		*(.text.*)
 		*(.fixup)
 
@@ -50,8 +61,14 @@ SECTIONS
 
 	/* Just in case the first read only is a 32-bit access */
 	RO_DATA(4)
+	__rodata_end = .;
 
+#ifdef CONFIG_ROMKERNEL
+	. = CONFIG_BOOT_LOAD;
+	.bss : AT(__rodata_end)
+#else
 	.bss :
+#endif
 	{
 		. = ALIGN(4);
 		___bss_start = .;
@@ -67,7 +84,11 @@ SECTIONS
 		___bss_stop = .;
 	}
 
+#if defined(CONFIG_ROMKERNEL)
+	.data : AT(LOADADDR(.bss) + SIZEOF(.bss))
+#else
 	.data :
+#endif
 	{
 		__sdata = .;
 		/* This gets done first, so the glob doesn't suck it in */
@@ -94,6 +115,8 @@ SECTIONS
 
 		__edata = .;
 	}
+	__data_lma = LOADADDR(.data);
+	__data_len = SIZEOF(.data);
 
 	/* The init section should be last, so when we free it, it goes into
 	 * the general memory pool, and (hopefully) will decrease fragmentation
@@ -103,25 +126,58 @@ SECTIONS
 	. = ALIGN(PAGE_SIZE);
 	___init_begin = .;
 
+#ifdef CONFIG_RAMKERNEL
 	INIT_TEXT_SECTION(PAGE_SIZE)
-	. = ALIGN(16);
-	INIT_DATA_SECTION(16)
-	PERCPU(4)
 
-	/* we have to discard exit text and such at runtime, not link time, to
+	/* We have to discard exit text and such at runtime, not link time, to
 	 * handle embedded cross-section references (alt instructions, bug
-	 * table, eh_frame, etc...)
+	 * table, eh_frame, etc...).  We need all of our .text up front and
+	 * .data after it for PCREL call issues.
 	 */
 	.exit.text :
 	{
 		EXIT_TEXT
 	}
+
+	. = ALIGN(16);
+	INIT_DATA_SECTION(16)
+	PERCPU(4)
+
 	.exit.data :
 	{
 		EXIT_DATA
 	}
 
 	.text_l1 L1_CODE_START : AT(LOADADDR(.exit.data) + SIZEOF(.exit.data))
+#else
+	.init.data : AT(__data_lma + __data_len)
+	{
+		__sinitdata = .;
+		INIT_DATA
+		INIT_SETUP(16)
+		INIT_CALLS
+		CON_INITCALL
+		SECURITY_INITCALL
+		INIT_RAM_FS
+
+		. = ALIGN(4);
+		___per_cpu_load = .;
+		___per_cpu_start = .;
+		*(.data.percpu.first)
+		*(.data.percpu.page_aligned)
+		*(.data.percpu)
+		*(.data.percpu.shared_aligned)
+		___per_cpu_end = .;
+
+		EXIT_DATA
+		__einitdata = .;
+	}
+	__init_data_lma = LOADADDR(.init.data);
+	__init_data_len = SIZEOF(.init.data);
+	__init_data_end = .;
+
+	.text_l1 L1_CODE_START : AT(__init_data_lma + __init_data_len)
+#endif
 	{
 		. = ALIGN(4);
 		__stext_l1 = .;
@@ -202,7 +258,11 @@ SECTIONS
 	/* Force trailing alignment of our init section so that when we
 	 * free our init memory, we don't leave behind a partial page.
 	 */
+#ifdef CONFIG_RAMKERNEL
 	. = __l2_lma + __l2_len;
+#else
+	. = __init_data_end;
+#endif
 	. = ALIGN(PAGE_SIZE);
 	___init_end = .;