20 files changed, 819 insertions, 366 deletions

diff --git a/arch/blackfin/kernel/Makefile b/arch/blackfin/kernel/Makefile
index a8ddbc8ed5af..346a421f1562 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 924c00286bab..26403d1c9e65 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 a174596cc009..e35e20f00d9b 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 8d42b9e50dfa..30fd6417f069 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 930c01c06813..87b25b1b30ed 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 282a7919821b..bfe75af4e8bd 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 e937f323d82c..04ddcfeb7981 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 f27dc2292e1b..686478f5f66b 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 76dd4fbcd17a..d66446b572c0 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 f2c85ac6f2da..a61d948ea925 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 118c5b9dedac..d3970e8acd1a 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/kgdb.c b/arch/blackfin/kernel/kgdb.c
index 34c7c3ed2c9c..2c501ceb1e55 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 000000000000..0b5f72f17fd0
--- /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 b56b0e485e0b..29705cec91de 100644
--- a/arch/blackfin/kernel/process.c
+++ b/arch/blackfin/kernel/process.c
@@ -98,13 +98,6 @@ 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 65567dc4b9f5..43eb969405d1 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.
  */