Merge branch 'linus' into x86/urgent

Merge reason: pull in latest to fix a bug in it.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

25 files changed, 868 insertions, 331 deletions

diff --git a/arch/blackfin/kernel/Makefile b/arch/blackfin/kernel/Makefile
index fd4d4328a0f..3731088e181 100644
--- a/arch/blackfin/kernel/Makefile
+++ b/arch/blackfin/kernel/Makefile
@@ -15,6 +15,10 @@ else
     obj-y += time.o
 endif
 
+obj-$(CONFIG_FUNCTION_TRACER)        += ftrace-entry.o
+obj-$(CONFIG_FUNCTION_GRAPH_TRACER)  += ftrace.o
+CFLAGS_REMOVE_ftrace.o = -pg
+
 obj-$(CONFIG_IPIPE)                  += ipipe.o
 obj-$(CONFIG_IPIPE_TRACE_MCOUNT)     += mcount.o
 obj-$(CONFIG_BFIN_GPTIMERS)          += gptimers.o
@@ -23,6 +27,7 @@ obj-$(CONFIG_MODULES)                += module.o
 obj-$(CONFIG_KGDB)                   += kgdb.o
 obj-$(CONFIG_KGDB_TESTS)             += kgdb_test.o
 obj-$(CONFIG_EARLY_PRINTK)           += early_printk.o
+obj-$(CONFIG_STACKTRACE)             += stacktrace.o
 
 # the kgdb test puts code into L2 and without linker
 # relaxation, we need to force long calls to/from it
diff --git a/arch/blackfin/kernel/bfin_dma_5xx.c b/arch/blackfin/kernel/bfin_dma_5xx.c
index 8531693fb48..e0bf8cc0690 100644
--- a/arch/blackfin/kernel/bfin_dma_5xx.c
+++ b/arch/blackfin/kernel/bfin_dma_5xx.c
@@ -20,6 +20,11 @@
 #include <asm/dma.h>
 #include <asm/uaccess.h>
 
+/*
+ * To make sure we work around 05000119 - we always check DMA_DONE bit,
+ * never the DMA_RUN bit
+ */
+
 struct dma_channel dma_ch[MAX_DMA_CHANNELS];
 EXPORT_SYMBOL(dma_ch);
 
@@ -232,6 +237,87 @@ void blackfin_dma_resume(void)
 void __init blackfin_dma_early_init(void)
 {
 	bfin_write_MDMA_S0_CONFIG(0);
+	bfin_write_MDMA_S1_CONFIG(0);
+}
+
+void __init early_dma_memcpy(void *pdst, const void *psrc, size_t size)
+{
+	unsigned long dst = (unsigned long)pdst;
+	unsigned long src = (unsigned long)psrc;
+	struct dma_register *dst_ch, *src_ch;
+
+	/* We assume that everything is 4 byte aligned, so include
+	 * a basic sanity check
+	 */
+	BUG_ON(dst % 4);
+	BUG_ON(src % 4);
+	BUG_ON(size % 4);
+
+	/* Force a sync in case a previous config reset on this channel
+	 * occurred.  This is needed so subsequent writes to DMA registers
+	 * are not spuriously lost/corrupted.
+	 */
+	__builtin_bfin_ssync();
+
+	src_ch = 0;
+	/* Find an avalible memDMA channel */
+	while (1) {
+		if (!src_ch || src_ch == (struct dma_register *)MDMA_S1_NEXT_DESC_PTR) {
+			dst_ch = (struct dma_register *)MDMA_D0_NEXT_DESC_PTR;
+			src_ch = (struct dma_register *)MDMA_S0_NEXT_DESC_PTR;
+		} else {
+			dst_ch = (struct dma_register *)MDMA_D1_NEXT_DESC_PTR;
+			src_ch = (struct dma_register *)MDMA_S1_NEXT_DESC_PTR;
+		}
+
+		if (!bfin_read16(&src_ch->cfg)) {
+			break;
+		} else {
+			if (bfin_read16(&src_ch->irq_status) & DMA_DONE)
+				bfin_write16(&src_ch->cfg, 0);
+		}
+
+	}
+
+	/* Destination */
+	bfin_write32(&dst_ch->start_addr, dst);
+	bfin_write16(&dst_ch->x_count, size >> 2);
+	bfin_write16(&dst_ch->x_modify, 1 << 2);
+	bfin_write16(&dst_ch->irq_status, DMA_DONE | DMA_ERR);
+
+	/* Source */
+	bfin_write32(&src_ch->start_addr, src);
+	bfin_write16(&src_ch->x_count, size >> 2);
+	bfin_write16(&src_ch->x_modify, 1 << 2);
+	bfin_write16(&src_ch->irq_status, DMA_DONE | DMA_ERR);
+
+	/* Enable */
+	bfin_write16(&src_ch->cfg, DMAEN | WDSIZE_32);
+	bfin_write16(&dst_ch->cfg, WNR | DI_EN | DMAEN | WDSIZE_32);
+
+	/* Since we are atomic now, don't use the workaround ssync */
+	__builtin_bfin_ssync();
+}
+
+void __init early_dma_memcpy_done(void)
+{
+	while ((bfin_read_MDMA_S0_CONFIG() && !(bfin_read_MDMA_D0_IRQ_STATUS() & DMA_DONE)) ||
+	       (bfin_read_MDMA_S1_CONFIG() && !(bfin_read_MDMA_D1_IRQ_STATUS() & DMA_DONE)))
+		continue;
+
+	bfin_write_MDMA_D0_IRQ_STATUS(DMA_DONE | DMA_ERR);
+	bfin_write_MDMA_D1_IRQ_STATUS(DMA_DONE | DMA_ERR);
+	/*
+	 * Now that DMA is done, we would normally flush cache, but
+	 * i/d cache isn't running this early, so we don't bother,
+	 * and just clear out the DMA channel for next time
+	 */
+	bfin_write_MDMA_S0_CONFIG(0);
+	bfin_write_MDMA_S1_CONFIG(0);
+	bfin_write_MDMA_D0_CONFIG(0);
+	bfin_write_MDMA_D1_CONFIG(0);
+
+	__builtin_bfin_ssync();
 }
 
 /**
@@ -367,10 +453,10 @@ void *dma_memcpy(void *pdst, const void *psrc, size_t size)
 	unsigned long src = (unsigned long)psrc;
 	size_t bulk, rest;
 
-	if (bfin_addr_dcachable(src))
+	if (bfin_addr_dcacheable(src))
 		blackfin_dcache_flush_range(src, src + size);
 
-	if (bfin_addr_dcachable(dst))
+	if (bfin_addr_dcacheable(dst))
 		blackfin_dcache_invalidate_range(dst, dst + size);
 
 	bulk = size & ~0xffff;
diff --git a/arch/blackfin/kernel/bfin_gpio.c b/arch/blackfin/kernel/bfin_gpio.c
index a0678da4053..beffa00a93c 100644
--- a/arch/blackfin/kernel/bfin_gpio.c
+++ b/arch/blackfin/kernel/bfin_gpio.c
@@ -313,15 +313,6 @@ inline void portmux_setup(unsigned short per)
 # define portmux_setup(...)  do { } while (0)
 #endif
 
-static int __init bfin_gpio_init(void)
-{
-	printk(KERN_INFO "Blackfin GPIO Controller\n");
-
-	return 0;
-}
-arch_initcall(bfin_gpio_init);
-
-
 #ifndef CONFIG_BF54x
 /***********************************************************
 *
@@ -1021,15 +1012,6 @@ int bfin_gpio_irq_request(unsigned gpio, const char *label)
 
 	local_irq_save_hw(flags);
 
-	if (unlikely(reserved_gpio_irq_map[gpio_bank(gpio)] & gpio_bit(gpio))) {
-		if (system_state == SYSTEM_BOOTING)
-			dump_stack();
-		printk(KERN_ERR
-		       "bfin-gpio: GPIO %d is already reserved as gpio-irq !\n",
-		       gpio);
-		local_irq_restore_hw(flags);
-		return -EBUSY;
-	}
 	if (unlikely(reserved_peri_map[gpio_bank(gpio)] & gpio_bit(gpio))) {
 		if (system_state == SYSTEM_BOOTING)
 			dump_stack();
diff --git a/arch/blackfin/kernel/bfin_ksyms.c b/arch/blackfin/kernel/bfin_ksyms.c
index 01f917d58b5..aa05e638fb7 100644
--- a/arch/blackfin/kernel/bfin_ksyms.c
+++ b/arch/blackfin/kernel/bfin_ksyms.c
@@ -16,7 +16,6 @@ EXPORT_SYMBOL(bfin_return_from_exception);
 
 /* All the Blackfin cache functions: mach-common/cache.S */
 EXPORT_SYMBOL(blackfin_dcache_invalidate_range);
-EXPORT_SYMBOL(blackfin_icache_dcache_flush_range);
 EXPORT_SYMBOL(blackfin_icache_flush_range);
 EXPORT_SYMBOL(blackfin_dcache_flush_range);
 EXPORT_SYMBOL(blackfin_dflush_page);
@@ -104,3 +103,8 @@ EXPORT_SYMBOL(__raw_smp_mark_barrier_asm);
 EXPORT_SYMBOL(__raw_smp_check_barrier_asm);
 #endif
 #endif
+
+#ifdef CONFIG_FUNCTION_TRACER
+extern void _mcount(void);
+EXPORT_SYMBOL(_mcount);
+#endif
diff --git a/arch/blackfin/kernel/cplb-mpu/cacheinit.c b/arch/blackfin/kernel/cplb-mpu/cacheinit.c
index c6ff947f9d3..d5a86c3017f 100644
--- a/arch/blackfin/kernel/cplb-mpu/cacheinit.c
+++ b/arch/blackfin/kernel/cplb-mpu/cacheinit.c
@@ -55,7 +55,14 @@ void __cpuinit bfin_dcache_init(struct cplb_entry *dcplb_tbl)
 	}
 
 	ctrl = bfin_read_DMEM_CONTROL();
-	ctrl |= DMEM_CNTR;
+
+	/*
+	 *  Anomaly notes:
+	 *  05000287 - We implement workaround #2 - Change the DMEM_CONTROL
+	 *  register, so that the port preferences for DAG0 and DAG1 are set
+	 *  to port B
+	 */
+	ctrl |= DMEM_CNTR | PORT_PREF0 | (ANOMALY_05000287 ? PORT_PREF1 : 0);
 	bfin_write_DMEM_CONTROL(ctrl);
 	SSYNC();
 }
diff --git a/arch/blackfin/kernel/cplb-mpu/cplbinit.c b/arch/blackfin/kernel/cplb-mpu/cplbinit.c
index 3e329a6ce04..c006a44527b 100644
--- a/arch/blackfin/kernel/cplb-mpu/cplbinit.c
+++ b/arch/blackfin/kernel/cplb-mpu/cplbinit.c
@@ -64,7 +64,7 @@ void __init generate_cplb_tables_cpu(unsigned int cpu)
 	dcplb_tbl[cpu][i_d++].data = SDRAM_OOPS | PAGE_SIZE_1KB;
 
 	icplb_tbl[cpu][i_i].addr = 0;
-	icplb_tbl[cpu][i_i++].data = i_cache | CPLB_USER_RD | PAGE_SIZE_1KB;
+	icplb_tbl[cpu][i_i++].data = CPLB_VALID | i_cache | CPLB_USER_RD | PAGE_SIZE_1KB;
 
 	/* Cover kernel memory with 4M pages.  */
 	addr = 0;
diff --git a/arch/blackfin/kernel/cplb-mpu/cplbmgr.c b/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
index 87463ce87f5..784923e52a9 100644
--- a/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
+++ b/arch/blackfin/kernel/cplb-mpu/cplbmgr.c
@@ -151,7 +151,7 @@ static noinline int dcplb_miss(unsigned int cpu)
 
 	d_data = CPLB_SUPV_WR | CPLB_VALID | CPLB_DIRTY | PAGE_SIZE_4KB;
 #ifdef CONFIG_BFIN_DCACHE
-	if (bfin_addr_dcachable(addr)) {
+	if (bfin_addr_dcacheable(addr)) {
 		d_data |= CPLB_L1_CHBL | ANOMALY_05000158_WORKAROUND;
 #ifdef CONFIG_BFIN_WT
 		d_data |= CPLB_L1_AOW | CPLB_WT;
diff --git a/arch/blackfin/kernel/cplb-nompu/cacheinit.c b/arch/blackfin/kernel/cplb-nompu/cacheinit.c
index c6ff947f9d3..d5a86c3017f 100644
--- a/arch/blackfin/kernel/cplb-nompu/cacheinit.c
+++ b/arch/blackfin/kernel/cplb-nompu/cacheinit.c
@@ -55,7 +55,14 @@ void __cpuinit bfin_dcache_init(struct cplb_entry *dcplb_tbl)
 	}
 
 	ctrl = bfin_read_DMEM_CONTROL();
-	ctrl |= DMEM_CNTR;
+
+	/*
+	 *  Anomaly notes:
+	 *  05000287 - We implement workaround #2 - Change the DMEM_CONTROL
+	 *  register, so that the port preferences for DAG0 and DAG1 are set
+	 *  to port B
+	 */
+	ctrl |= DMEM_CNTR | PORT_PREF0 | (ANOMALY_05000287 ? PORT_PREF1 : 0);
 	bfin_write_DMEM_CONTROL(ctrl);
 	SSYNC();
 }
diff --git a/arch/blackfin/kernel/cplb-nompu/cplbmgr.c b/arch/blackfin/kernel/cplb-nompu/cplbmgr.c
index 8cbb47c7b66..12b030842fd 100644
--- a/arch/blackfin/kernel/cplb-nompu/cplbmgr.c
+++ b/arch/blackfin/kernel/cplb-nompu/cplbmgr.c
@@ -28,6 +28,7 @@
 #include <asm/cplbinit.h>
 #include <asm/cplb.h>
 #include <asm/mmu_context.h>
+#include <asm/traps.h>
 
 /*
  * WARNING
@@ -100,28 +101,6 @@ static inline void write_icplb_data(int cpu, int idx, unsigned long data,
 #endif
 }
 
-/*
- * Given the contents of the status register, return the index of the
- * CPLB that caused the fault.
- */
-static inline int faulting_cplb_index(int status)
-{
-	int signbits = __builtin_bfin_norm_fr1x32(status & 0xFFFF);
-	return 30 - signbits;
-}
-
-/*
- * Given the contents of the status register and the DCPLB_DATA contents,
- * return true if a write access should be permitted.
- */
-static inline int write_permitted(int status, unsigned long data)
-{
-	if (status & FAULT_USERSUPV)
-		return !!(data & CPLB_SUPV_WR);
-	else
-		return !!(data & CPLB_USER_WR);
-}
-
 /* Counters to implement round-robin replacement.  */
 static int icplb_rr_index[NR_CPUS] PDT_ATTR;
 static int dcplb_rr_index[NR_CPUS] PDT_ATTR;
@@ -245,43 +224,16 @@ MGR_ATTR static int dcplb_miss(int cpu)
 	return CPLB_RELOADED;
 }
 
-MGR_ATTR static noinline int dcplb_protection_fault(int cpu)
-{
-	int status = bfin_read_DCPLB_STATUS();
-
-	nr_dcplb_prot[cpu]++;
-
-	if (likely(status & FAULT_RW)) {
-		int idx = faulting_cplb_index(status);
-		unsigned long regaddr = DCPLB_DATA0 + idx * 4;
-		unsigned long data = bfin_read32(regaddr);
-
-		/* Check if fault is to dirty a clean page */
-		if (!(data & CPLB_WT) && !(data & CPLB_DIRTY) &&
-		    write_permitted(status, data)) {
-
-			dcplb_tbl[cpu][idx].data = data;
-			bfin_write32(regaddr, data);
-			return CPLB_RELOADED;
-		}
-	}
-
-	return CPLB_PROT_VIOL;
-}
-
 MGR_ATTR int cplb_hdr(int seqstat, struct pt_regs *regs)
 {
 	int cause = seqstat & 0x3f;
 	unsigned int cpu = smp_processor_id();
 	switch (cause) {
-	case 0x2C:
+	case VEC_CPLB_I_M:
 		return icplb_miss(cpu);
-	case 0x26:
+	case VEC_CPLB_M:
 		return dcplb_miss(cpu);
 	default:
-		if (unlikely(cause == 0x23))
-			return dcplb_protection_fault(cpu);
-
 		return CPLB_UNKNOWN_ERR;
 	}
 }
diff --git a/arch/blackfin/kernel/early_printk.c b/arch/blackfin/kernel/early_printk.c
index c8ad051742e..2ab56811841 100644
--- a/arch/blackfin/kernel/early_printk.c
+++ b/arch/blackfin/kernel/early_printk.c
@@ -178,25 +178,15 @@ int __init setup_early_printk(char *buf)
 
 asmlinkage void __init init_early_exception_vectors(void)
 {
+	u32 evt;
 	SSYNC();
 
 	/* cannot program in software:
 	 * evt0 - emulation (jtag)
 	 * evt1 - reset
 	 */
-	bfin_write_EVT2(early_trap);
-	bfin_write_EVT3(early_trap);
-	bfin_write_EVT5(early_trap);
-	bfin_write_EVT6(early_trap);
-	bfin_write_EVT7(early_trap);
-	bfin_write_EVT8(early_trap);
-	bfin_write_EVT9(early_trap);
-	bfin_write_EVT10(early_trap);
-	bfin_write_EVT11(early_trap);
-	bfin_write_EVT12(early_trap);
-	bfin_write_EVT13(early_trap);
-	bfin_write_EVT14(early_trap);
-	bfin_write_EVT15(early_trap);
+	for (evt = EVT2; evt <= EVT15; evt += 4)
+		bfin_write32(evt, early_trap);
 	CSYNC();
 
 	/* Set all the return from interrupt, exception, NMI to a known place
@@ -212,11 +202,15 @@ asmlinkage void __init init_early_exception_vectors(void)
 asmlinkage void __init early_trap_c(struct pt_regs *fp, void *retaddr)
 {
 	/* This can happen before the uart is initialized, so initialize
-	 * the UART now
+	 * the UART now (but only if we are running on the processor we think
+	 * we are compiled for - otherwise we write to MMRs that don't exist,
+	 * and cause other problems. Nothing comes out the UART, but it does
+	 * end up in the __buf_log.
 	 */
-	if (likely(early_console == NULL))
+	if (likely(early_console == NULL) && CPUID == bfin_cpuid())
 		setup_early_printk(DEFAULT_EARLY_PORT);
 
+	printk(KERN_EMERG "Early panic\n");
 	dump_bfin_mem(fp);
 	show_regs(fp);
 	dump_bfin_trace_buffer();
diff --git a/arch/blackfin/kernel/ftrace-entry.S b/arch/blackfin/kernel/ftrace-entry.S
new file mode 100644
index 00000000000..6980b7a0615
--- /dev/null
+++ b/arch/blackfin/kernel/ftrace-entry.S
@@ -0,0 +1,140 @@
+/*
+ * mcount and friends -- ftrace stuff
+ *
+ * Copyright (C) 2009 Analog Devices Inc.
+ * Licensed under the GPL-2 or later.
+ */
+
+#include <linux/linkage.h>
+#include <asm/ftrace.h>
+
+.text
+
+/* GCC will have called us before setting up the function prologue, so we
+ * can clobber the normal scratch registers, but we need to make sure to
+ * save/restore the registers used for argument passing (R0-R2) in case
+ * the profiled function is using them.  With data registers, R3 is the
+ * only one we can blow away.  With pointer registers, we have P0-P2.
+ *
+ * Upon entry, the RETS will point to the top of the current profiled
+ * function.  And since GCC setup the frame for us, the previous function
+ * will be waiting there.  mmmm pie.
+ */
+ENTRY(__mcount)
+	/* save third function arg early so we can do testing below */
+	[--sp] = r2;
+
+	/* load the function pointer to the tracer */
+	p0.l = _ftrace_trace_function;
+	p0.h = _ftrace_trace_function;
+	r3 = [p0];
+
+	/* optional micro optimization: don't call the stub tracer */
+	r2.l = _ftrace_stub;
+	r2.h = _ftrace_stub;
+	cc = r2 == r3;
+	if ! cc jump .Ldo_trace;
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	/* if the ftrace_graph_return function pointer is not set to
+	 * the ftrace_stub entry, call prepare_ftrace_return().
+	 */
+	p0.l = _ftrace_graph_return;
+	p0.h = _ftrace_graph_return;
+	r3 = [p0];
+	cc = r2 == r3;
+	if ! cc jump _ftrace_graph_caller;
+
+	/* similarly, if the ftrace_graph_entry function pointer is not
+	 * set to the ftrace_graph_entry_stub entry, ...
+	 */
+	p0.l = _ftrace_graph_entry;
+	p0.h = _ftrace_graph_entry;
+	r2.l = _ftrace_graph_entry_stub;
+	r2.h = _ftrace_graph_entry_stub;
+	r3 = [p0];
+	cc = r2 == r3;
+	if ! cc jump _ftrace_graph_caller;
+#endif
+
+	r2 = [sp++];
+	rts;
+
+.Ldo_trace:
+
+	/* save first/second function arg and the return register */
+	[--sp] = r0;
+	[--sp] = r1;
+	[--sp] = rets;
+
+	/* setup the tracer function */
+	p0 = r3;
+
+	/* tracer(ulong frompc, ulong selfpc):
+	 *  frompc: the pc that did the call to ...
+	 *  selfpc: ... this location
+	 * the selfpc itself will need adjusting for the mcount call
+	 */
+	r1 = rets;
+	r0 = [fp + 4];
+	r1 += -MCOUNT_INSN_SIZE;
+
+	/* call the tracer */
+	call (p0);
+
+	/* restore state and get out of dodge */
+.Lfinish_trace:
+	rets = [sp++];
+	r1 = [sp++];
+	r0 = [sp++];
+	r2 = [sp++];
+
+.globl _ftrace_stub
+_ftrace_stub:
+	rts;
+ENDPROC(__mcount)
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+/* The prepare_ftrace_return() function is similar to the trace function
+ * except it takes a pointer to the location of the frompc.  This is so
+ * the prepare_ftrace_return() can hijack it temporarily for probing
+ * purposes.
+ */
+ENTRY(_ftrace_graph_caller)
+	/* save first/second function arg and the return register */
+	[--sp] = r0;
+	[--sp] = r1;
+	[--sp] = rets;
+
+	r0 = fp;
+	r1 = rets;
+	r0 += 4;
+	r1 += -MCOUNT_INSN_SIZE;
+	call _prepare_ftrace_return;
+
+	jump .Lfinish_trace;
+ENDPROC(_ftrace_graph_caller)
+
+/* Undo the rewrite caused by ftrace_graph_caller().  The common function
+ * ftrace_return_to_handler() will return the original rets so we can
+ * restore it and be on our way.
+ */
+ENTRY(_return_to_handler)
+	/* make sure original return values are saved */
+	[--sp] = p0;
+	[--sp] = r0;
+	[--sp] = r1;
+
+	/* get original return address */
+	call _ftrace_return_to_handler;
+	rets = r0;
+
+	/* anomaly 05000371 - make sure we have at least three instructions
+	 * between rets setting and the return
+	 */
+	r1 = [sp++];
+	r0 = [sp++];
+	p0 = [sp++];
+	rts;
+ENDPROC(_return_to_handler)
+#endif
diff --git a/arch/blackfin/kernel/ftrace.c b/arch/blackfin/kernel/ftrace.c
new file mode 100644
index 00000000000..905bfc40a00
--- /dev/null
+++ b/arch/blackfin/kernel/ftrace.c
@@ -0,0 +1,42 @@
+/*
+ * ftrace graph code
+ *
+ * Copyright (C) 2009 Analog Devices Inc.
+ * Licensed under the GPL-2 or later.
+ */
+
+#include <linux/ftrace.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <asm/atomic.h>
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+
+/*
+ * 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)
+{
+	struct ftrace_graph_ent trace;
+	unsigned long return_hooker = (unsigned long)&return_to_handler;
+
+	if (unlikely(atomic_read(&current->tracing_graph_pause)))
+		return;
+
+	if (ftrace_push_return_trace(*parent, self_addr, &trace.depth) == -EBUSY)
+		return;
+
+	trace.func = self_addr;
+
+	/* Only trace if the calling function expects to */
+	if (!ftrace_graph_entry(&trace)) {
+		current->curr_ret_stack--;
+		return;
+	}
+
+	/* all is well in the world !  hijack RETS ... */
+	*parent = return_hooker;
+}
+
+#endif
diff --git a/arch/blackfin/kernel/gptimers.c b/arch/blackfin/kernel/gptimers.c
index 3a3e9615b00..7281a91d26b 100644
--- a/arch/blackfin/kernel/gptimers.c
+++ b/arch/blackfin/kernel/gptimers.c
@@ -189,10 +189,10 @@ void set_gptimer_status(int group, uint32_t value)
 }
 EXPORT_SYMBOL(set_gptimer_status);
 
-uint16_t get_gptimer_intr(int timer_id)
+int get_gptimer_intr(int timer_id)
 {
 	tassert(timer_id < MAX_BLACKFIN_GPTIMERS);
-	return (group_regs[BFIN_TIMER_OCTET(timer_id)]->status & timil_mask[timer_id]) ? 1 : 0;
+	return !!(group_regs[BFIN_TIMER_OCTET(timer_id)]->status & timil_mask[timer_id]);
 }
 EXPORT_SYMBOL(get_gptimer_intr);
 
@@ -203,10 +203,10 @@ void clear_gptimer_intr(int timer_id)
 }
 EXPORT_SYMBOL(clear_gptimer_intr);
 
-uint16_t get_gptimer_over(int timer_id)
+int get_gptimer_over(int timer_id)
 {
 	tassert(timer_id < MAX_BLACKFIN_GPTIMERS);
-	return (group_regs[BFIN_TIMER_OCTET(timer_id)]->status & tovf_mask[timer_id]) ? 1 : 0;
+	return !!(group_regs[BFIN_TIMER_OCTET(timer_id)]->status & tovf_mask[timer_id]);
 }
 EXPORT_SYMBOL(get_gptimer_over);
 
@@ -217,6 +217,13 @@ void clear_gptimer_over(int timer_id)
 }
 EXPORT_SYMBOL(clear_gptimer_over);
 
+int get_gptimer_run(int timer_id)
+{
+	tassert(timer_id < MAX_BLACKFIN_GPTIMERS);
+	return !!(group_regs[BFIN_TIMER_OCTET(timer_id)]->status & trun_mask[timer_id]);
+}
+EXPORT_SYMBOL(get_gptimer_run);
+
 void set_gptimer_config(int timer_id, uint16_t config)
 {
 	tassert(timer_id < MAX_BLACKFIN_GPTIMERS);
@@ -244,7 +251,7 @@ void enable_gptimers(uint16_t mask)
 }
 EXPORT_SYMBOL(enable_gptimers);
 
-void disable_gptimers(uint16_t mask)
+static void _disable_gptimers(uint16_t mask)
 {
 	int i;
 	uint16_t m = mask;
@@ -253,6 +260,12 @@ void disable_gptimers(uint16_t mask)
 		group_regs[i]->disable = m & 0xFF;
 		m >>= 8;
 	}
+}
+
+void disable_gptimers(uint16_t mask)
+{
+	int i;
+	_disable_gptimers(mask);
 	for (i = 0; i < MAX_BLACKFIN_GPTIMERS; ++i)
 		if (mask & (1 << i))
 			group_regs[BFIN_TIMER_OCTET(i)]->status |= trun_mask[i];
@@ -260,6 +273,13 @@ void disable_gptimers(uint16_t mask)
 }
 EXPORT_SYMBOL(disable_gptimers);
 
+void disable_gptimers_sync(uint16_t mask)
+{
+	_disable_gptimers(mask);
+	SSYNC();
+}
+EXPORT_SYMBOL(disable_gptimers_sync);
+
 void set_gptimer_pulse_hi(int timer_id)
 {
 	tassert(timer_id < MAX_BLACKFIN_GPTIMERS);
diff --git a/arch/blackfin/kernel/ipipe.c b/arch/blackfin/kernel/ipipe.c
index a5de8d45424..d8cde1fc5cb 100644
--- a/arch/blackfin/kernel/ipipe.c
+++ b/arch/blackfin/kernel/ipipe.c
@@ -99,7 +99,7 @@ void __ipipe_handle_irq(unsigned irq, struct pt_regs *regs)
 	 * interrupt.
 	 */
 	m_ack = (regs == NULL || irq == IRQ_SYSTMR || irq == IRQ_CORETMR);
-	this_domain = ipipe_current_domain;
+	this_domain = __ipipe_current_domain;
 
 	if (unlikely(test_bit(IPIPE_STICKY_FLAG, &this_domain->irqs[irq].control)))
 		head = &this_domain->p_link;
@@ -167,7 +167,7 @@ int __ipipe_check_root(void)
 void __ipipe_enable_irqdesc(struct ipipe_domain *ipd, unsigned irq)
 {
 	struct irq_desc *desc = irq_to_desc(irq);
-	int prio = desc->ic_prio;
+	int prio = __ipipe_get_irq_priority(irq);
 
 	desc->depth = 0;
 	if (ipd != &ipipe_root &&
@@ -178,8 +178,7 @@ EXPORT_SYMBOL(__ipipe_enable_irqdesc);
 
 void __ipipe_disable_irqdesc(struct ipipe_domain *ipd, unsigned irq)
 {
-	struct irq_desc *desc = irq_to_desc(irq);
-	int prio = desc->ic_prio;
+	int prio = __ipipe_get_irq_priority(irq);
 
 	if (ipd != &ipipe_root &&
 	    atomic_dec_and_test(&__ipipe_irq_lvdepth[prio]))
@@ -213,7 +212,9 @@ void __ipipe_unstall_root_raw(void)
 
 int __ipipe_syscall_root(struct pt_regs *regs)
 {
+	struct ipipe_percpu_domain_data *p;
 	unsigned long flags;
+	int ret;
 
 	/*
 	 * We need to run the IRQ tail hook whenever we don't
@@ -232,29 +233,31 @@ int __ipipe_syscall_root(struct pt_regs *regs)
 	/*
 	 * This routine either returns:
 	 * 0 -- if the syscall is to be passed to Linux;
-	 * 1 -- if the syscall should not be passed to Linux, and no
+	 * >0 -- if the syscall should not be passed to Linux, and no
 	 * tail work should be performed;
-	 * -1 -- if the syscall should not be passed to Linux but the
+	 * <0 -- if the syscall should not be passed to Linux but the
 	 * tail work has to be performed (for handling signals etc).
 	 */
 
-	if (__ipipe_event_monitored_p(IPIPE_EVENT_SYSCALL) &&
-	    __ipipe_dispatch_event(IPIPE_EVENT_SYSCALL, regs) > 0) {
-		if (ipipe_root_domain_p && !in_atomic()) {
-			/*
-			 * Sync pending VIRQs before _TIF_NEED_RESCHED
-			 * is tested.
-			 */
-			local_irq_save_hw(flags);
-			if ((ipipe_root_cpudom_var(irqpend_himask) & IPIPE_IRQMASK_VIRT) != 0)
-				__ipipe_sync_pipeline(IPIPE_IRQMASK_VIRT);
-			local_irq_restore_hw(flags);
-			return -1;
-		}
+	if (!__ipipe_event_monitored_p(IPIPE_EVENT_SYSCALL))
+		return 0;
+
+	ret = __ipipe_dispatch_event(IPIPE_EVENT_SYSCALL, regs);
+
+	local_irq_save_hw(flags);
+
+	if (!__ipipe_root_domain_p) {
+		local_irq_restore_hw(flags);
 		return 1;
 	}
 
-	return 0;
+	p = ipipe_root_cpudom_ptr();
+	if ((p->irqpend_himask & IPIPE_IRQMASK_VIRT) != 0)
+		__ipipe_sync_pipeline(IPIPE_IRQMASK_VIRT);
+
+	local_irq_restore_hw(flags);
+
+	return -ret;
 }
 
 unsigned long ipipe_critical_enter(void (*syncfn) (void))
@@ -310,12 +313,16 @@ int ipipe_trigger_irq(unsigned irq)
 
 asmlinkage void __ipipe_sync_root(void)
 {
+	void (*irq_tail_hook)(void) = (void (*)(void))__ipipe_irq_tail_hook;
 	unsigned long flags;
 
 	BUG_ON(irqs_disabled());
 
 	local_irq_save_hw(flags);
 
+	if (irq_tail_hook)
+		irq_tail_hook();
+
 	clear_thread_flag(TIF_IRQ_SYNC);
 
 	if (ipipe_root_cpudom_var(irqpend_himask) != 0)
@@ -326,9 +333,7 @@ asmlinkage void __ipipe_sync_root(void)
 
 void ___ipipe_sync_pipeline(unsigned long syncmask)
 {
-	struct ipipe_domain *ipd = ipipe_current_domain;
-
-	if (ipd == ipipe_root_domain) {
+	if (__ipipe_root_domain_p) {
 		if (test_bit(IPIPE_SYNCDEFER_FLAG, &ipipe_root_cpudom_var(status)))
 			return;
 	}
diff --git a/arch/blackfin/kernel/irqchip.c b/arch/blackfin/kernel/irqchip.c
index 401bd32aa49..6e31e935bb3 100644
--- a/arch/blackfin/kernel/irqchip.c
+++ b/arch/blackfin/kernel/irqchip.c
@@ -59,12 +59,14 @@ static struct irq_chip bad_chip = {
 	.unmask = dummy_mask_unmask_irq,
 };
 
+static int bad_stats;
 static struct irq_desc bad_irq_desc = {
 	.status = IRQ_DISABLED,
 	.chip = &bad_chip,
 	.handle_irq = handle_bad_irq,
 	.depth = 1,
 	.lock = __SPIN_LOCK_UNLOCKED(irq_desc->lock),
+	.kstat_irqs = &bad_stats,
 #ifdef CONFIG_SMP
 	.affinity = CPU_MASK_ALL
 #endif
diff --git a/arch/blackfin/kernel/kgdb.c b/arch/blackfin/kernel/kgdb.c
index b163f6d3330..da28f796ad7 100644
--- a/arch/blackfin/kernel/kgdb.c
+++ b/arch/blackfin/kernel/kgdb.c
@@ -466,7 +466,7 @@ static int validate_memory_access_address(unsigned long addr, int size)
 	int cpu = raw_smp_processor_id();
 
 	if (size < 0)
-		return EFAULT;
+		return -EFAULT;
 	if (addr >= 0x1000 && (addr + size) <= physical_mem_end)
 		return 0;
 	if (addr >= SYSMMR_BASE)
@@ -498,7 +498,7 @@ static int validate_memory_access_address(unsigned long addr, int size)
 	if (IN_MEM(addr, size, L2_START, L2_LENGTH))
 		return 0;
 
-	return EFAULT;
+	return -EFAULT;
 }
 
 /*
@@ -508,14 +508,15 @@ static int validate_memory_access_address(unsigned long addr, int size)
 int kgdb_mem2hex(char *mem, char *buf, int count)
 {
 	char *tmp;
-	int err = 0;
+	int err;
 	unsigned char *pch;
 	unsigned short mmr16;
 	unsigned long mmr32;
 	int cpu = raw_smp_processor_id();
 
-	if (validate_memory_access_address((unsigned long)mem, count))
-		return EFAULT;
+	err = validate_memory_access_address((unsigned long)mem, count);
+	if (err)
+		return err;
 
 	/*
 	 * We use the upper half of buf as an intermediate buffer for the
@@ -533,7 +534,7 @@ int kgdb_mem2hex(char *mem, char *buf, int count)
 				*tmp++ = *pch++;
 				tmp -= 2;
 			} else
-				err = EFAULT;
+				err = -EFAULT;
 			break;
 		case 4:
 			if ((unsigned int)mem % 4 == 0) {
@@ -545,10 +546,10 @@ int kgdb_mem2hex(char *mem, char *buf, int count)
 				*tmp++ = *pch++;
 				tmp -= 4;
 			} else
-				err = EFAULT;
+				err = -EFAULT;
 			break;
 		default:
-			err = EFAULT;
+			err = -EFAULT;
 		}
 	} else if ((cpu == 0 && IN_MEM(mem, count, L1_CODE_START, L1_CODE_LENGTH))
 #ifdef CONFIG_SMP
@@ -557,7 +558,7 @@ int kgdb_mem2hex(char *mem, char *buf, int count)
 		) {
 		/* access L1 instruction SRAM*/
 		if (dma_memcpy(tmp, mem, count) == NULL)
-			err = EFAULT;
+			err = -EFAULT;
 	} else
 		err = probe_kernel_read(tmp, mem, count);
 
@@ -585,24 +586,24 @@ int kgdb_ebin2mem(char *buf, char *mem, int count)
 	char *tmp_new;
 	unsigned short *mmr16;
 	unsigned long *mmr32;
-	int err = 0;
-	int size = 0;
+	int err;
+	int size;
 	int cpu = raw_smp_processor_id();
 
 	tmp_old = tmp_new = buf;
 
-	while (count-- > 0) {
+	for (size = 0; size < count; ++size) {
 		if (*tmp_old == 0x7d)
 			*tmp_new = *(++tmp_old) ^ 0x20;
 		else
 			*tmp_new = *tmp_old;
 		tmp_new++;
 		tmp_old++;
-		size++;
 	}
 
-	if (validate_memory_access_address((unsigned long)mem, size))
-		return EFAULT;
+	err = validate_memory_access_address((unsigned long)mem, size);
+	if (err)
+		return err;
 
 	if ((unsigned int)mem >= SYSMMR_BASE) { /*access MMR registers*/
 		switch (size) {
@@ -611,17 +612,17 @@ int kgdb_ebin2mem(char *buf, char *mem, int count)
 				mmr16 = (unsigned short *)buf;
 				*(unsigned short *)mem = *mmr16;
 			} else
-				return EFAULT;
+				err = -EFAULT;
 			break;
 		case 4:
 			if ((unsigned int)mem % 4 == 0) {
 				mmr32 = (unsigned long *)buf;
 				*(unsigned long *)mem = *mmr32;
 			} else
-				return EFAULT;
+				err = -EFAULT;
 			break;
 		default:
-			return EFAULT;
+			err = -EFAULT;
 		}
 	} else if ((cpu == 0 && IN_MEM(mem, count, L1_CODE_START, L1_CODE_LENGTH))
 #ifdef CONFIG_SMP
@@ -630,7 +631,7 @@ int kgdb_ebin2mem(char *buf, char *mem, int count)
 		) {
 		/* access L1 instruction SRAM */
 		if (dma_memcpy(mem, buf, size) == NULL)
-			err = EFAULT;
+			err = -EFAULT;
 	} else
 		err = probe_kernel_write(mem, buf, size);
 
@@ -648,10 +649,12 @@ int kgdb_hex2mem(char *buf, char *mem, int count)
 	char *tmp_hex;
 	unsigned short *mmr16;
 	unsigned long *mmr32;
+	int err;
 	int cpu = raw_smp_processor_id();
 
-	if (validate_memory_access_address((unsigned long)mem, count))
-		return EFAULT;
+	err = validate_memory_access_address((unsigned long)mem, count);
+	if (err)
+		return err;
 
 	/*
 	 * We use the upper half of buf as an intermediate buffer for the
@@ -673,17 +676,17 @@ int kgdb_hex2mem(char *buf, char *mem, int count)
 				mmr16 = (unsigned short *)tmp_raw;
 				*(unsigned short *)mem = *mmr16;
 			} else
-				return EFAULT;
+				err = -EFAULT;
 			break;
 		case 4:
 			if ((unsigned int)mem % 4 == 0) {
 				mmr32 = (unsigned long *)tmp_raw;
 				*(unsigned long *)mem = *mmr32;
 			} else
-				return EFAULT;
+				err = -EFAULT;
 			break;
 		default:
-			return EFAULT;
+			err = -EFAULT;
 		}
 	} else if ((cpu == 0 && IN_MEM(mem, count, L1_CODE_START, L1_CODE_LENGTH))
 #ifdef CONFIG_SMP
@@ -692,10 +695,11 @@ int kgdb_hex2mem(char *buf, char *mem, int count)
 		) {
 		/* access L1 instruction SRAM */
 		if (dma_memcpy(mem, tmp_raw, count) == NULL)
-			return EFAULT;
+			err = -EFAULT;
 	} else
-		return probe_kernel_write(mem, tmp_raw, count);
-	return 0;
+		err = probe_kernel_write(mem, tmp_raw, count);
+
+	return err;
 }
 
 int kgdb_validate_break_address(unsigned long addr)
@@ -715,7 +719,7 @@ int kgdb_validate_break_address(unsigned long addr)
 	if (IN_MEM(addr, BREAK_INSTR_SIZE, L2_START, L2_LENGTH))
 		return 0;
 
-	return EFAULT;
+	return -EFAULT;
 }
 
 int kgdb_arch_set_breakpoint(unsigned long addr, char *saved_instr)
diff --git a/arch/blackfin/kernel/module.c b/arch/blackfin/kernel/module.c
index 1bd7f2d018a..d5aee362668 100644
--- a/arch/blackfin/kernel/module.c
+++ b/arch/blackfin/kernel/module.c
@@ -201,8 +201,8 @@ apply_relocate(Elf_Shdr * sechdrs, const char *strtab,
 /*            Arithmetic relocations are handled.                        */
 /*            We do not expect LSETUP to be split and hence is not       */
 /*            handled.                                                   */
-/*            R_byte and R_byte2 are also not handled as the gas         */
-/*            does not generate it.                                      */
+/*            R_BFIN_BYTE and R_BFIN_BYTE2 are also not handled as the   */
+/*            gas does not generate it.                                  */
 /*************************************************************************/
 int
 apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
@@ -243,8 +243,8 @@ apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
 #endif
 		switch (ELF32_R_TYPE(rel[i].r_info)) {
 
-		case R_pcrel24:
-		case R_pcrel24_jump_l:
+		case R_BFIN_PCREL24:
+		case R_BFIN_PCREL24_JUMP_L:
 			/* Add the value, subtract its postition */
 			location16 =
 			    (uint16_t *) (sechdrs[sechdrs[relsec].sh_info].
@@ -266,18 +266,18 @@ apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
 			    (*location16 & 0xff00) | (value >> 16 & 0x00ff);
 			*(location16 + 1) = value & 0xffff;
 			break;
-		case R_pcrel12_jump:
-		case R_pcrel12_jump_s:
+		case R_BFIN_PCREL12_JUMP:
+		case R_BFIN_PCREL12_JUMP_S:
 			value -= (uint32_t) location32;
 			value >>= 1;
 			*location16 = (value & 0xfff);
 			break;
-		case R_pcrel10:
+		case R_BFIN_PCREL10:
 			value -= (uint32_t) location32;
 			value >>= 1;
 			*location16 = (value & 0x3ff);
 			break;
-		case R_luimm16:
+		case R_BFIN_LUIMM16:
 			pr_debug("before %x after %x\n", *location16,
 				       (value & 0xffff));
 			tmp = (value & 0xffff);
@@ -286,7 +286,7 @@ apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
 			} else
 				*location16 = tmp;
 			break;
-		case R_huimm16:
+		case R_BFIN_HUIMM16:
 			pr_debug("before %x after %x\n", *location16,
 				       ((value >> 16) & 0xffff));
 			tmp = ((value >> 16) & 0xffff);
@@ -295,10 +295,10 @@ apply_relocate_add(Elf_Shdr * sechdrs, const char *strtab,
 			} else
 				*location16 = tmp;
 			break;
-		case R_rimm16:
+		case R_BFIN_RIMM16:
 			*location16 = (value & 0xffff);
 			break;
-		case R_byte4_data:
+		case R_BFIN_BYTE4_DATA:
 			pr_debug("before %x after %x\n", *location32, value);
 			*location32 = value;
 			break;
diff --git a/arch/blackfin/kernel/process.c b/arch/blackfin/kernel/process.c
index e040e03335e..30d0843ed70 100644
--- a/arch/blackfin/kernel/process.c
+++ b/arch/blackfin/kernel/process.c
@@ -322,6 +322,9 @@ void finish_atomic_sections (struct pt_regs *regs)
 }
 
 #if defined(CONFIG_ACCESS_CHECK)
+#ifdef CONFIG_ACCESS_OK_L1
+__attribute__((l1_text))
+#endif
 /* Return 1 if access to memory range is OK, 0 otherwise */
 int _access_ok(unsigned long addr, unsigned long size)
 {
diff --git a/arch/blackfin/kernel/setup.c b/arch/blackfin/kernel/setup.c
index a58687bdee6..6454babdfaf 100644
--- a/arch/blackfin/kernel/setup.c
+++ b/arch/blackfin/kernel/setup.c
@@ -18,9 +18,12 @@
 #include <linux/tty.h>
 #include <linux/pfn.h>
 
+#ifdef CONFIG_MTD_UCLINUX
+#include <linux/mtd/map.h>
 #include <linux/ext2_fs.h>
 #include <linux/cramfs_fs.h>
 #include <linux/romfs_fs.h>
+#endif
 
 #include <asm/cplb.h>
 #include <asm/cacheflush.h>
@@ -45,6 +48,7 @@ EXPORT_SYMBOL(_ramend);
 EXPORT_SYMBOL(reserved_mem_dcache_on);
 
 #ifdef CONFIG_MTD_UCLINUX
+extern struct map_info uclinux_ram_map;
 unsigned long memory_mtd_end, memory_mtd_start, mtd_size;
 unsigned long _ebss;
 EXPORT_SYMBOL(memory_mtd_end);
@@ -150,40 +154,45 @@ void __init bfin_relocate_l1_mem(void)
 	unsigned long l1_data_b_length;
 	unsigned long l2_length;
 
+	/*
+	 * due to the ALIGN(4) in the arch/blackfin/kernel/vmlinux.lds.S
+	 * we know that everything about l1 text/data is nice and aligned,
+	 * so copy by 4 byte chunks, and don't worry about overlapping
+	 * src/dest.
+	 *
+	 * We can't use the dma_memcpy functions, since they can call
+	 * scheduler functions which might be in L1 :( and core writes
+	 * into L1 instruction cause bad access errors, so we are stuck,
+	 * we are required to use DMA, but can't use the common dma
+	 * functions. We can't use memcpy either - since that might be
+	 * going to be in the relocated L1
+	 */
+
 	blackfin_dma_early_init();
 
+	/* if necessary, copy _stext_l1 to _etext_l1 to L1 instruction SRAM */
 	l1_code_length = _etext_l1 - _stext_l1;
-	if (l1_code_length > L1_CODE_LENGTH)
-		panic("L1 Instruction SRAM Overflow\n");
-	/* cannot complain as printk is not available as yet.
-	 * But we can continue booting and complain later!
-	 */
-
-	/* Copy _stext_l1 to _etext_l1 to L1 instruction SRAM */
-	dma_memcpy(_stext_l1, _l1_lma_start, l1_code_length);
+	if (l1_code_length)
+		early_dma_memcpy(_stext_l1, _l1_lma_start, l1_code_length);
 
+	/* if necessary, copy _sdata_l1 to _sbss_l1 to L1 data bank A SRAM */
 	l1_data_a_length = _sbss_l1 - _sdata_l1;
-	if (l1_data_a_length > L1_DATA_A_LENGTH)
-		panic("L1 Data SRAM Bank A Overflow\n");
-
-	/* Copy _sdata_l1 to _sbss_l1 to L1 data bank A SRAM */
-	dma_memcpy(_sdata_l1, _l1_lma_start + l1_code_length, l1_data_a_length);
+	if (l1_data_a_length)
+		early_dma_memcpy(_sdata_l1, _l1_lma_start + l1_code_length, l1_data_a_length);
 
+	/* if necessary, copy _sdata_b_l1 to _sbss_b_l1 to L1 data bank B SRAM */
 	l1_data_b_length = _sbss_b_l1 - _sdata_b_l1;
-	if (l1_data_b_length > L1_DATA_B_LENGTH)
-		panic("L1 Data SRAM Bank B Overflow\n");
-
-	/* Copy _sdata_b_l1 to _sbss_b_l1 to L1 data bank B SRAM */
-	dma_memcpy(_sdata_b_l1, _l1_lma_start + l1_code_length +
+	if (l1_data_b_length)
+		early_dma_memcpy(_sdata_b_l1, _l1_lma_start + l1_code_length +
 			l1_data_a_length, l1_data_b_length);
 
+	early_dma_memcpy_done();
+
+	/* if necessary, copy _stext_l2 to _edata_l2 to L2 SRAM */
 	if (L2_LENGTH != 0) {
 		l2_length = _sbss_l2 - _stext_l2;
-		if (l2_length > L2_LENGTH)
-			panic("L2 SRAM Overflow\n");
-
-		/* Copy _stext_l2 to _edata_l2 to L2 SRAM */
-		dma_memcpy(_stext_l2, _l2_lma_start, l2_length);
+		if (l2_length)
+			memcpy(_stext_l2, _l2_lma_start, l2_length);
 	}
 }
 
@@ -472,7 +481,7 @@ static __init void memory_setup(void)
 
 	if (DMA_UNCACHED_REGION > (_ramend - _ramstart)) {
 		console_init();
-		panic("DMA region exceeds memory limit: %lu.\n",
+		panic("DMA region exceeds memory limit: %lu.",
 			_ramend - _ramstart);
 	}
 	memory_end = _ramend - DMA_UNCACHED_REGION;
@@ -526,14 +535,13 @@ static __init void memory_setup(void)
 
 	if (mtd_size == 0) {
 		console_init();
-		panic("Don't boot kernel without rootfs attached.\n");
+		panic("Don't boot kernel without rootfs attached.");
 	}
 
 	/* Relocate MTD image to the top of memory after the uncached memory area */
-	dma_memcpy((char *)memory_end, _end, mtd_size);
-
-	memory_mtd_start = memory_end;
-	_ebss = memory_mtd_start;	/* define _ebss for compatible */
+	uclinux_ram_map.phys = memory_mtd_start = memory_end;
+	uclinux_ram_map.size = mtd_size;
+	dma_memcpy((void *)uclinux_ram_map.phys, _end, uclinux_ram_map.size);
 #endif				/* CONFIG_MTD_UCLINUX */
 
 #if (defined(CONFIG_BFIN_ICACHE) && ANOMALY_05000263)
@@ -796,10 +804,8 @@ void __init setup_arch(char **cmdline_p)
 	cclk = get_cclk();
 	sclk = get_sclk();
 
-#if !defined(CONFIG_BFIN_KERNEL_CLOCK)
-	if (ANOMALY_05000273 && cclk == sclk)
-		panic("ANOMALY 05000273, SCLK can not be same as CCLK");
-#endif
+	if ((ANOMALY_05000273 || ANOMALY_05000274) && (cclk >> 1) < sclk)
+		panic("ANOMALY 05000273 or 05000274: CCLK must be >= 2*SCLK");
 
 #ifdef BF561_FAMILY
 	if (ANOMALY_05000266) {
@@ -881,7 +887,7 @@ void __init setup_arch(char **cmdline_p)
 				printk(KERN_ERR "Warning: Compiled for Rev %d, but running on Rev %d\n",
 				       bfin_compiled_revid(), bfin_revid());
 				if (bfin_compiled_revid() > bfin_revid())
-					panic("Error: you are missing anomaly workarounds for this rev\n");
+					panic("Error: you are missing anomaly workarounds for this rev");
 			}
 		}
 		if (bfin_revid() < CONFIG_BF_REV_MIN || bfin_revid() > CONFIG_BF_REV_MAX)
@@ -891,16 +897,13 @@ void __init setup_arch(char **cmdline_p)
 
 	/* We can't run on BF548-0.1 due to ANOMALY 05000448 */
 	if (bfin_cpuid() == 0x27de && bfin_revid() == 1)
-		panic("You can't run on this processor due to 05000448\n");
+		panic("You can't run on this processor due to 05000448");
 
 	printk(KERN_INFO "Blackfin Linux support by http://blackfin.uclinux.org/\n");
 
 	printk(KERN_INFO "Processor Speed: %lu MHz core clock and %lu MHz System Clock\n",
 	       cclk / 1000000, sclk / 1000000);
 
-	if (ANOMALY_05000273 && (cclk >> 1) <= sclk)
-		printk("\n\n\nANOMALY_05000273: CCLK must be >= 2*SCLK !!!\n\n\n");
-
 	setup_bootmem_allocator();
 
 	paging_init();
@@ -1095,7 +1098,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 			CPUID, bfin_cpuid());
 
 	seq_printf(m, "model name\t: ADSP-%s %lu(MHz CCLK) %lu(MHz SCLK) (%s)\n"
-		"stepping\t: %d\n",
+		"stepping\t: %d ",
 		cpu, cclk/1000000, sclk/1000000,
 #ifdef CONFIG_MPU
 		"mpu on",
@@ -1104,7 +1107,16 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 #endif
 		revid);
 
-	seq_printf(m, "cpu MHz\t\t: %lu.%03lu/%lu.%03lu\n",
+	if (bfin_revid() != bfin_compiled_revid()) {
+		if (bfin_compiled_revid() == -1)
+			seq_printf(m, "(Compiled for Rev none)");
+		else if (bfin_compiled_revid() == 0xffff)
+			seq_printf(m, "(Compiled for Rev any)");
+		else
+			seq_printf(m, "(Compiled for Rev %d)", bfin_compiled_revid());
+	}
+
+	seq_printf(m, "\ncpu MHz\t\t: %lu.%03lu/%lu.%03lu\n",
 		cclk/1000000, cclk%1000000,
 		sclk/1000000, sclk%1000000);
 	seq_printf(m, "bogomips\t: %lu.%02lu\n"
@@ -1169,6 +1181,9 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 #ifdef __ARCH_SYNC_CORE_DCACHE
 	seq_printf(m, "SMP Dcache Flushes\t: %lu\n\n", cpudata->dcache_invld_count);
 #endif
+#ifdef __ARCH_SYNC_CORE_ICACHE
+	seq_printf(m, "SMP Icache Flushes\t: %lu\n\n", cpudata->icache_invld_count);
+#endif
 #ifdef CONFIG_BFIN_ICACHE_LOCK
 	switch ((cpudata->imemctl >> 3) & WAYALL_L) {
 	case WAY0_L:
diff --git a/arch/blackfin/kernel/stacktrace.c b/arch/blackfin/kernel/stacktrace.c
new file mode 100644
index 00000000000..30301e1eace
--- /dev/null
+++ b/arch/blackfin/kernel/stacktrace.c
@@ -0,0 +1,53 @@
+/*
+ * Blackfin stacktrace code (mostly copied from avr32)
+ *
+ * Copyright 2009 Analog Devices Inc.
+ * Licensed under the GPL-2 or later.
+ */
+
+#include <linux/sched.h>
+#include <linux/stacktrace.h>
+#include <linux/thread_info.h>
+#include <linux/module.h>
+
+register unsigned long current_frame_pointer asm("FP");
+
+struct stackframe {
+	unsigned long fp;
+	unsigned long rets;
+};
+
+/*
+ * Save stack-backtrace addresses into a stack_trace buffer.
+ */
+void save_stack_trace(struct stack_trace *trace)
+{
+	unsigned long low, high;
+	unsigned long fp;
+	struct stackframe *frame;
+	int skip = trace->skip;
+
+	low = (unsigned long)task_stack_page(current);
+	high = low + THREAD_SIZE;
+	fp = current_frame_pointer;
+
+	while (fp >= low && fp <= (high - sizeof(*frame))) {
+		frame = (struct stackframe *)fp;
+
+		if (skip) {
+			skip--;
+		} else {
+			trace->entries[trace->nr_entries++] = frame->rets;
+			if (trace->nr_entries >= trace->max_entries)
+				break;
+		}
+
+		/*
+		 * The next frame must be at a higher address than the
+		 * current frame.
+		 */
+		low = fp + sizeof(*frame);
+		fp = frame->fp;
+	}
+}
+EXPORT_SYMBOL_GPL(save_stack_trace);
diff --git a/arch/blackfin/kernel/sys_bfin.c b/arch/blackfin/kernel/sys_bfin.c
index fce49d7cf00..a8f1329c15a 100644
--- a/arch/blackfin/kernel/sys_bfin.c
+++ b/arch/blackfin/kernel/sys_bfin.c
@@ -78,11 +78,6 @@ asmlinkage long sys_mmap2(unsigned long addr, unsigned long len,
 	return do_mmap2(addr, len, prot, flags, fd, pgoff);
 }
 
-asmlinkage int sys_getpagesize(void)
-{
-	return PAGE_SIZE;
-}
-
 asmlinkage void *sys_sram_alloc(size_t size, unsigned long flags)
 {
 	return sram_alloc_with_lsl(size, flags);
diff --git a/arch/blackfin/kernel/time-ts.c b/arch/blackfin/kernel/time-ts.c
index 27646121280..0791eba40d9 100644
--- a/arch/blackfin/kernel/time-ts.c
+++ b/arch/blackfin/kernel/time-ts.c
@@ -20,8 +20,9 @@
 
 #include <asm/blackfin.h>
 #include <asm/time.h>
+#include <asm/gptimers.h>
 
-#ifdef CONFIG_CYCLES_CLOCKSOURCE
+#if defined(CONFIG_CYCLES_CLOCKSOURCE)
 
 /* Accelerators for sched_clock()
  * convert from cycles(64bits) => nanoseconds (64bits)
@@ -58,15 +59,15 @@ static inline unsigned long long cycles_2_ns(cycle_t cyc)
 	return (cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR;
 }
 
-static cycle_t read_cycles(struct clocksource *cs)
+static cycle_t bfin_read_cycles(struct clocksource *cs)
 {
 	return __bfin_cycles_off + (get_cycles() << __bfin_cycles_mod);
 }
 
-static struct clocksource clocksource_bfin = {
-	.name		= "bfin_cycles",
+static struct clocksource bfin_cs_cycles = {
+	.name		= "bfin_cs_cycles",
 	.rating		= 350,
-	.read		= read_cycles,
+	.read		= bfin_read_cycles,
 	.mask		= CLOCKSOURCE_MASK(64),
 	.shift		= 22,
 	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
@@ -74,53 +75,198 @@ static struct clocksource clocksource_bfin = {
 
 unsigned long long sched_clock(void)
 {
-	return cycles_2_ns(read_cycles(&clocksource_bfin));
+	return cycles_2_ns(bfin_read_cycles(&bfin_cs_cycles));
 }
 
-static int __init bfin_clocksource_init(void)
+static int __init bfin_cs_cycles_init(void)
 {
 	set_cyc2ns_scale(get_cclk() / 1000);
 
-	clocksource_bfin.mult = clocksource_hz2mult(get_cclk(), clocksource_bfin.shift);
+	bfin_cs_cycles.mult = \
+		clocksource_hz2mult(get_cclk(), bfin_cs_cycles.shift);
 
-	if (clocksource_register(&clocksource_bfin))
+	if (clocksource_register(&bfin_cs_cycles))
 		panic("failed to register clocksource");
 
 	return 0;
 }
+#else
+# define bfin_cs_cycles_init()
+#endif
+
+#ifdef CONFIG_GPTMR0_CLOCKSOURCE
+
+void __init setup_gptimer0(void)
+{
+	disable_gptimers(TIMER0bit);
+
+	set_gptimer_config(TIMER0_id, \
+		TIMER_OUT_DIS | TIMER_PERIOD_CNT | TIMER_MODE_PWM);
+	set_gptimer_period(TIMER0_id, -1);
+	set_gptimer_pwidth(TIMER0_id, -2);
+	SSYNC();
+	enable_gptimers(TIMER0bit);
+}
+
+static cycle_t bfin_read_gptimer0(void)
+{
+	return bfin_read_TIMER0_COUNTER();
+}
+
+static struct clocksource bfin_cs_gptimer0 = {
+	.name		= "bfin_cs_gptimer0",
+	.rating		= 400,
+	.read		= bfin_read_gptimer0,
+	.mask		= CLOCKSOURCE_MASK(32),
+	.shift		= 22,
+	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
+};
+
+static int __init bfin_cs_gptimer0_init(void)
+{
+	setup_gptimer0();
 
+	bfin_cs_gptimer0.mult = \
+		clocksource_hz2mult(get_sclk(), bfin_cs_gptimer0.shift);
+
+	if (clocksource_register(&bfin_cs_gptimer0))
+		panic("failed to register clocksource");
+
+	return 0;
+}
 #else
-# define bfin_clocksource_init()
+# define bfin_cs_gptimer0_init()
 #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,
                                      struct clock_event_device *evt)
 {
+	disable_gptimers(TIMER0bit);
+
+	/* it starts counting three SCLK cycles after the TIMENx bit is set */
+	set_gptimer_pwidth(TIMER0_id, cycles - 3);
+	enable_gptimers(TIMER0bit);
+	return 0;
+}
+
+static void bfin_timer_set_mode(enum clock_event_mode mode,
+				struct clock_event_device *evt)
+{
+	switch (mode) {
+	case CLOCK_EVT_MODE_PERIODIC: {
+		set_gptimer_config(TIMER0_id, \
+			TIMER_OUT_DIS | TIMER_IRQ_ENA | \
+			TIMER_PERIOD_CNT | TIMER_MODE_PWM);
+		set_gptimer_period(TIMER0_id, get_sclk() / HZ);
+		set_gptimer_pwidth(TIMER0_id, get_sclk() / HZ - 1);
+		enable_gptimers(TIMER0bit);
+		break;
+	}
+	case CLOCK_EVT_MODE_ONESHOT:
+		disable_gptimers(TIMER0bit);
+		set_gptimer_config(TIMER0_id, \
+			TIMER_OUT_DIS | TIMER_IRQ_ENA | TIMER_MODE_PWM);
+		set_gptimer_period(TIMER0_id, 0);
+		break;
+	case CLOCK_EVT_MODE_UNUSED:
+	case CLOCK_EVT_MODE_SHUTDOWN:
+		disable_gptimers(TIMER0bit);
+		break;
+	case CLOCK_EVT_MODE_RESUME:
+		break;
+	}
+}
+
+static void bfin_timer_ack(void)
+{
+	set_gptimer_status(TIMER_GROUP1, TIMER_STATUS_TIMIL0);
+}
+
+static void __init bfin_timer_init(void)
+{
+	disable_gptimers(TIMER0bit);
+}
+
+static unsigned long  __init bfin_clockevent_check(void)
+{
+	setup_irq(IRQ_TIMER0, &bfin_timer_irq);
+	return get_sclk();
+}
+
+#else /* CONFIG_TICKSOURCE_CORETMR */
+
+static int bfin_timer_set_next_event(unsigned long cycles,
+				struct clock_event_device *evt)
+{
+	bfin_write_TCNTL(TMPWR);
+	CSYNC();
 	bfin_write_TCOUNT(cycles);
 	CSYNC();
+	bfin_write_TCNTL(TMPWR | TMREN);
 	return 0;
 }
 
 static void bfin_timer_set_mode(enum clock_event_mode mode,
-                                struct clock_event_device *evt)
+				struct clock_event_device *evt)
 {
 	switch (mode) {
 	case CLOCK_EVT_MODE_PERIODIC: {
 		unsigned long tcount = ((get_cclk() / (HZ * TIME_SCALE)) - 1);
 		bfin_write_TCNTL(TMPWR);
-		bfin_write_TSCALE(TIME_SCALE - 1);
 		CSYNC();
+		bfin_write_TSCALE(TIME_SCALE - 1);
 		bfin_write_TPERIOD(tcount);
 		bfin_write_TCOUNT(tcount);
-		bfin_write_TCNTL(TMPWR | TMREN | TAUTORLD);
 		CSYNC();
+		bfin_write_TCNTL(TMPWR | TMREN | TAUTORLD);
 		break;
 	}
 	case CLOCK_EVT_MODE_ONESHOT:
+		bfin_write_TCNTL(TMPWR);
+		CSYNC();
 		bfin_write_TSCALE(TIME_SCALE - 1);
+		bfin_write_TPERIOD(0);
 		bfin_write_TCOUNT(0);
-		bfin_write_TCNTL(TMPWR | TMREN);
-		CSYNC();
 		break;
 	case CLOCK_EVT_MODE_UNUSED:
 	case CLOCK_EVT_MODE_SHUTDOWN:
@@ -132,6 +278,10 @@ static void bfin_timer_set_mode(enum clock_event_mode mode,
 	}
 }
 
+static void bfin_timer_ack(void)
+{
+}
+
 static void __init bfin_timer_init(void)
 {
 	/* power up the timer, but don't enable it just yet */
@@ -145,38 +295,32 @@ static void __init bfin_timer_init(void)
 	bfin_write_TPERIOD(0);
 	bfin_write_TCOUNT(0);
 
-	/* now enable the timer */
 	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)
+{
+	bfin_timer_init();
+	bfin_timer_set_mode(CLOCK_EVT_MODE_PERIODIC, NULL);
+}
+#endif /* CONFIG_TICKSOURCE_GPTMR0 */
+
 /*
  * timer_interrupt() needs to keep up the real-time clock,
  * as well as call the "do_timer()" routine every clocktick
  */
-#ifdef CONFIG_CORE_TIMER_IRQ_L1
-__attribute__((l1_text))
-#endif
-irqreturn_t timer_interrupt(int irq, void *dev_id);
-
-static struct clock_event_device clockevent_bfin = {
-	.name		= "bfin_core_timer",
-	.features	= CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT,
-	.shift		= 32,
-	.set_next_event = bfin_timer_set_next_event,
-	.set_mode	= bfin_timer_set_mode,
-};
-
-static struct irqaction bfin_timer_irq = {
-	.name		= "Blackfin Core Timer",
-	.flags		= IRQF_DISABLED | IRQF_TIMER | IRQF_IRQPOLL,
-	.handler	= timer_interrupt,
-	.dev_id		= &clockevent_bfin,
-};
-
 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;
 }
 
@@ -184,9 +328,8 @@ static int __init bfin_clockevent_init(void)
 {
 	unsigned long timer_clk;
 
-	timer_clk = get_cclk() / TIME_SCALE;
+	timer_clk = bfin_clockevent_check();
 
-	setup_irq(IRQ_CORETMR, &bfin_timer_irq);
 	bfin_timer_init();
 
 	clockevent_bfin.mult = div_sc(timer_clk, NSEC_PER_SEC, clockevent_bfin.shift);
@@ -218,6 +361,7 @@ void __init time_init(void)
 	xtime.tv_nsec = 0;
 	set_normalized_timespec(&wall_to_monotonic, -xtime.tv_sec, -xtime.tv_nsec);
 
-	bfin_clocksource_init();
+	bfin_cs_cycles_init();
+	bfin_cs_gptimer0_init();
 	bfin_clockevent_init();
 }
diff --git a/arch/blackfin/kernel/time.c b/arch/blackfin/kernel/time.c
index 1bbacfbd4c5..adb54aa7d7c 100644
--- a/arch/blackfin/kernel/time.c
+++ b/arch/blackfin/kernel/time.c
@@ -24,14 +24,10 @@
 
 static struct irqaction bfin_timer_irq = {
 	.name = "Blackfin Timer Tick",
-#ifdef CONFIG_IRQ_PER_CPU
-	.flags = IRQF_DISABLED | IRQF_PERCPU,
-#else
 	.flags = IRQF_DISABLED
-#endif
 };
 
-#if defined(CONFIG_TICK_SOURCE_SYSTMR0) || defined(CONFIG_IPIPE)
+#if defined(CONFIG_IPIPE)
 void __init setup_system_timer0(void)
 {
 	/* Power down the core timer, just to play safe. */
@@ -74,7 +70,7 @@ void __init setup_core_timer(void)
 static void __init
 time_sched_init(irqreturn_t(*timer_routine) (int, void *))
 {
-#if defined(CONFIG_TICK_SOURCE_SYSTMR0) || defined(CONFIG_IPIPE)
+#if defined(CONFIG_IPIPE)
 	setup_system_timer0();
 	bfin_timer_irq.handler = timer_routine;
 	setup_irq(IRQ_TIMER0, &bfin_timer_irq);
@@ -94,7 +90,7 @@ static unsigned long gettimeoffset(void)
 	unsigned long offset;
 	unsigned long clocks_per_jiffy;
 
-#if defined(CONFIG_TICK_SOURCE_SYSTMR0) || defined(CONFIG_IPIPE)
+#if defined(CONFIG_IPIPE)
 	clocks_per_jiffy = bfin_read_TIMER0_PERIOD();
 	offset = bfin_read_TIMER0_COUNTER() / \
 		(((clocks_per_jiffy + 1) * HZ) / USEC_PER_SEC);
@@ -133,36 +129,25 @@ irqreturn_t timer_interrupt(int irq, void *dummy)
 	static long last_rtc_update;
 
 	write_seqlock(&xtime_lock);
-#if defined(CONFIG_TICK_SOURCE_SYSTMR0) && !defined(CONFIG_IPIPE)
+	do_timer(1);
+
 	/*
-	 * TIMIL0 is latched in __ipipe_grab_irq() when the I-Pipe is
-	 * enabled.
+	 * If we have an externally synchronized Linux clock, then update
+	 * CMOS clock accordingly every ~11 minutes. Set_rtc_mmss() has to be
+	 * called as close as possible to 500 ms before the new second starts.
 	 */
-	if (get_gptimer_status(0) & TIMER_STATUS_TIMIL0) {
-#endif
-		do_timer(1);
-
-		/*
-		 * If we have an externally synchronized Linux clock, then update
-		 * CMOS clock accordingly every ~11 minutes. Set_rtc_mmss() has to be
-		 * called as close as possible to 500 ms before the new second starts.
-		 */
-		if (ntp_synced() &&
-		    xtime.tv_sec > last_rtc_update + 660 &&
-		    (xtime.tv_nsec / NSEC_PER_USEC) >=
-		    500000 - ((unsigned)TICK_SIZE) / 2
-		    && (xtime.tv_nsec / NSEC_PER_USEC) <=
-		    500000 + ((unsigned)TICK_SIZE) / 2) {
-			if (set_rtc_mmss(xtime.tv_sec) == 0)
-				last_rtc_update = xtime.tv_sec;
-			else
-				/* Do it again in 60s. */
-				last_rtc_update = xtime.tv_sec - 600;
-		}
-#if defined(CONFIG_TICK_SOURCE_SYSTMR0) && !defined(CONFIG_IPIPE)
-		set_gptimer_status(0, TIMER_STATUS_TIMIL0);
+	if (ntp_synced() &&
+	    xtime.tv_sec > last_rtc_update + 660 &&
+	    (xtime.tv_nsec / NSEC_PER_USEC) >=
+	    500000 - ((unsigned)TICK_SIZE) / 2
+	    && (xtime.tv_nsec / NSEC_PER_USEC) <=
+	    500000 + ((unsigned)TICK_SIZE) / 2) {
+		if (set_rtc_mmss(xtime.tv_sec) == 0)
+			last_rtc_update = xtime.tv_sec;
+		else
+			/* Do it again in 60s. */
+			last_rtc_update = xtime.tv_sec - 600;
 	}
-#endif
 	write_sequnlock(&xtime_lock);
 
 #ifdef CONFIG_IPIPE
diff --git a/arch/blackfin/kernel/traps.c b/arch/blackfin/kernel/traps.c
index ffe7fb53ecc..d279552fe9b 100644
--- a/arch/blackfin/kernel/traps.c
+++ b/arch/blackfin/kernel/traps.c
@@ -27,6 +27,7 @@
  * 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  */
 
+#include <linux/bug.h>
 #include <linux/uaccess.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
@@ -68,6 +69,13 @@
 	({ if (0) printk(fmt, ##arg); 0; })
 #endif
 
+#if defined(CONFIG_DEBUG_MMRS) || defined(CONFIG_DEBUG_MMRS_MODULE)
+u32 last_seqstat;
+#ifdef CONFIG_DEBUG_MMRS_MODULE
+EXPORT_SYMBOL(last_seqstat);
+#endif
+#endif
+
 /* Initiate the event table handler */
 void __init trap_init(void)
 {
@@ -79,7 +87,6 @@ void __init trap_init(void)
 static void decode_address(char *buf, unsigned long address)
 {
 #ifdef CONFIG_DEBUG_VERBOSE
-	struct vm_list_struct *vml;
 	struct task_struct *p;
 	struct mm_struct *mm;
 	unsigned long flags, offset;
@@ -196,6 +203,11 @@ done:
 
 asmlinkage void double_fault_c(struct pt_regs *fp)
 {
+#ifdef CONFIG_DEBUG_BFIN_HWTRACE_ON
+	int j;
+	trace_buffer_save(j);
+#endif
+
 	console_verbose();
 	oops_in_progress = 1;
 #ifdef CONFIG_DEBUG_VERBOSE
@@ -220,10 +232,16 @@ asmlinkage void double_fault_c(struct pt_regs *fp)
 		dump_bfin_process(fp);
 		dump_bfin_mem(fp);
 		show_regs(fp);
+		dump_bfin_trace_buffer();
 	}
 #endif
-	panic("Double Fault - unrecoverable event\n");
+	panic("Double Fault - unrecoverable event");
+
+}
 
+static int kernel_mode_regs(struct pt_regs *regs)
+{
+	return regs->ipend & 0xffc0;
 }
 
 asmlinkage void trap_c(struct pt_regs *fp)
@@ -234,37 +252,24 @@ asmlinkage void trap_c(struct pt_regs *fp)
 #ifdef CONFIG_DEBUG_HUNT_FOR_ZERO
 	unsigned int cpu = smp_processor_id();
 #endif
+	const char *strerror = NULL;
 	int sig = 0;
 	siginfo_t info;
 	unsigned long trapnr = fp->seqstat & SEQSTAT_EXCAUSE;
 
 	trace_buffer_save(j);
+#if defined(CONFIG_DEBUG_MMRS) || defined(CONFIG_DEBUG_MMRS_MODULE)
+	last_seqstat = (u32)fp->seqstat;
+#endif
 
 	/* Important - be very careful dereferncing pointers - will lead to
 	 * double faults if the stack has become corrupt
 	 */
 
-	/* If the fault was caused by a kernel thread, or interrupt handler
-	 * we will kernel panic, so the system reboots.
-	 * If KGDB is enabled, don't set this for kernel breakpoints
-	*/
-
-	/* TODO: check to see if we are in some sort of deferred HWERR
-	 * that we should be able to recover from, not kernel panic
-	 */
-	if ((bfin_read_IPEND() & 0xFFC0) && (trapnr != VEC_STEP)
-#ifdef CONFIG_KGDB
-		&& (trapnr != VEC_EXCPT02)
+#ifndef CONFIG_KGDB
+	/* IPEND is skipped if KGDB isn't enabled (see entry code) */
+	fp->ipend = bfin_read_IPEND();
 #endif
-	){
-		console_verbose();
-		oops_in_progress = 1;
-	} else if (current) {
-		if (current->mm == NULL) {
-			console_verbose();
-			oops_in_progress = 1;
-		}
-	}
 
 	/* trap_c() will be called for exceptions. During exceptions
 	 * processing, the pc value should be set with retx value.
@@ -292,15 +297,15 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		sig = SIGTRAP;
 		CHK_DEBUGGER_TRAP_MAYBE();
 		/* Check if this is a breakpoint in kernel space */
-		if (fp->ipend & 0xffc0)
-			return;
+		if (kernel_mode_regs(fp))
+			goto traps_done;
 		else
 			break;
 	/* 0x03 - User Defined, userspace stack overflow */
 	case VEC_EXCPT03:
 		info.si_code = SEGV_STACKFLOW;
 		sig = SIGSEGV;
-		verbose_printk(KERN_NOTICE EXC_0x03(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x03(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x02 - KGDB initial connection and break signal trap */
@@ -309,7 +314,7 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		info.si_code = TRAP_ILLTRAP;
 		sig = SIGTRAP;
 		CHK_DEBUGGER_TRAP();
-		return;
+		goto traps_done;
 #endif
 	/* 0x04 - User Defined */
 	/* 0x05 - User Defined */
@@ -329,7 +334,7 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	case VEC_EXCPT04 ... VEC_EXCPT15:
 		info.si_code = ILL_ILLPARAOP;
 		sig = SIGILL;
-		verbose_printk(KERN_NOTICE EXC_0x04(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x04(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x10 HW Single step, handled here */
@@ -338,15 +343,15 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		sig = SIGTRAP;
 		CHK_DEBUGGER_TRAP_MAYBE();
 		/* Check if this is a single step in kernel space */
-		if (fp->ipend & 0xffc0)
-			return;
+		if (kernel_mode_regs(fp))
+			goto traps_done;
 		else
 			break;
 	/* 0x11 - Trace Buffer Full, handled here */
 	case VEC_OVFLOW:
 		info.si_code = TRAP_TRACEFLOW;
 		sig = SIGTRAP;
-		verbose_printk(KERN_NOTICE EXC_0x11(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x11(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x12 - Reserved, Caught by default */
@@ -366,37 +371,54 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	/* 0x20 - Reserved, Caught by default */
 	/* 0x21 - Undefined Instruction, handled here */
 	case VEC_UNDEF_I:
+#ifdef CONFIG_BUG
+		if (kernel_mode_regs(fp)) {
+			switch (report_bug(fp->pc, fp)) {
+			case BUG_TRAP_TYPE_NONE:
+				break;
+			case BUG_TRAP_TYPE_WARN:
+				dump_bfin_trace_buffer();
+				fp->pc += 2;
+				goto traps_done;
+			case BUG_TRAP_TYPE_BUG:
+				/* call to panic() will dump trace, and it is
+				 * off at this point, so it won't be clobbered
+				 */
+				panic("BUG()");
+			}
+		}
+#endif
 		info.si_code = ILL_ILLOPC;
 		sig = SIGILL;
-		verbose_printk(KERN_NOTICE EXC_0x21(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x21(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x22 - Illegal Instruction Combination, handled here */
 	case VEC_ILGAL_I:
 		info.si_code = ILL_ILLPARAOP;
 		sig = SIGILL;
-		verbose_printk(KERN_NOTICE EXC_0x22(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x22(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x23 - Data CPLB protection violation, handled here */
 	case VEC_CPLB_VL:
 		info.si_code = ILL_CPLB_VI;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x23(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x23(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x24 - Data access misaligned, handled here */
 	case VEC_MISALI_D:
 		info.si_code = BUS_ADRALN;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x24(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x24(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x25 - Unrecoverable Event, handled here */
 	case VEC_UNCOV:
 		info.si_code = ILL_ILLEXCPT;
 		sig = SIGILL;
-		verbose_printk(KERN_NOTICE EXC_0x25(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x25(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x26 - Data CPLB Miss, normal case is handled in _cplb_hdr,
@@ -404,7 +426,7 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	case VEC_CPLB_M:
 		info.si_code = BUS_ADRALN;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x26(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x26(KERN_NOTICE);
 		break;
 	/* 0x27 - Data CPLB Multiple Hits - Linux Trap Zero, handled here */
 	case VEC_CPLB_MHIT:
@@ -412,10 +434,10 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		sig = SIGSEGV;
 #ifdef CONFIG_DEBUG_HUNT_FOR_ZERO
 		if (cpu_pda[cpu].dcplb_fault_addr < FIXED_CODE_START)
-			verbose_printk(KERN_NOTICE "NULL pointer access\n");
+			strerror = KERN_NOTICE "NULL pointer access\n";
 		else
 #endif
-			verbose_printk(KERN_NOTICE EXC_0x27(KERN_NOTICE));
+			strerror = KERN_NOTICE EXC_0x27(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x28 - Emulation Watchpoint, handled here */
@@ -425,8 +447,8 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		pr_debug(EXC_0x28(KERN_DEBUG));
 		CHK_DEBUGGER_TRAP_MAYBE();
 		/* Check if this is a watchpoint in kernel space */
-		if (fp->ipend & 0xffc0)
-			return;
+		if (kernel_mode_regs(fp))
+			goto traps_done;
 		else
 			break;
 #ifdef CONFIG_BF535
@@ -434,7 +456,7 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	case VEC_ISTRU_VL:      /* ADSP-BF535 only (MH) */
 		info.si_code = BUS_OPFETCH;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE "BF535: VEC_ISTRU_VL\n");
+		strerror = KERN_NOTICE "BF535: VEC_ISTRU_VL\n";
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 #else
@@ -444,21 +466,21 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	case VEC_MISALI_I:
 		info.si_code = BUS_ADRALN;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x2A(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x2A(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x2B - Instruction CPLB protection violation, handled here */
 	case VEC_CPLB_I_VL:
 		info.si_code = ILL_CPLB_VI;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x2B(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x2B(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x2C - Instruction CPLB miss, handled in _cplb_hdr */
 	case VEC_CPLB_I_M:
 		info.si_code = ILL_CPLB_MISS;
 		sig = SIGBUS;
-		verbose_printk(KERN_NOTICE EXC_0x2C(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x2C(KERN_NOTICE);
 		break;
 	/* 0x2D - Instruction CPLB Multiple Hits, handled here */
 	case VEC_CPLB_I_MHIT:
@@ -466,17 +488,17 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		sig = SIGSEGV;
 #ifdef CONFIG_DEBUG_HUNT_FOR_ZERO
 		if (cpu_pda[cpu].icplb_fault_addr < FIXED_CODE_START)
-			verbose_printk(KERN_NOTICE "Jump to NULL address\n");
+			strerror = KERN_NOTICE "Jump to NULL address\n";
 		else
 #endif
-			verbose_printk(KERN_NOTICE EXC_0x2D(KERN_NOTICE));
+			strerror = KERN_NOTICE EXC_0x2D(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x2E - Illegal use of Supervisor Resource, handled here */
 	case VEC_ILL_RES:
 		info.si_code = ILL_PRVOPC;
 		sig = SIGILL;
-		verbose_printk(KERN_NOTICE EXC_0x2E(KERN_NOTICE));
+		strerror = KERN_NOTICE EXC_0x2E(KERN_NOTICE);
 		CHK_DEBUGGER_TRAP_MAYBE();
 		break;
 	/* 0x2F - Reserved, Caught by default */
@@ -504,17 +526,17 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		case (SEQSTAT_HWERRCAUSE_SYSTEM_MMR):
 			info.si_code = BUS_ADRALN;
 			sig = SIGBUS;
-			verbose_printk(KERN_NOTICE HWC_x2(KERN_NOTICE));
+			strerror = KERN_NOTICE HWC_x2(KERN_NOTICE);
 			break;
 		/* External Memory Addressing Error */
 		case (SEQSTAT_HWERRCAUSE_EXTERN_ADDR):
 			info.si_code = BUS_ADRERR;
 			sig = SIGBUS;
-			verbose_printk(KERN_NOTICE HWC_x3(KERN_NOTICE));
+			strerror = KERN_NOTICE HWC_x3(KERN_NOTICE);
 			break;
 		/* Performance Monitor Overflow */
 		case (SEQSTAT_HWERRCAUSE_PERF_FLOW):
-			verbose_printk(KERN_NOTICE HWC_x12(KERN_NOTICE));
+			strerror = KERN_NOTICE HWC_x12(KERN_NOTICE);
 			break;
 		/* RAISE 5 instruction */
 		case (SEQSTAT_HWERRCAUSE_RAISE_5):
@@ -531,7 +553,6 @@ asmlinkage void trap_c(struct pt_regs *fp)
 	 * if we get here we hit a reserved one, so panic
 	 */
 	default:
-		oops_in_progress = 1;
 		info.si_code = ILL_ILLPARAOP;
 		sig = SIGILL;
 		verbose_printk(KERN_EMERG "Caught Unhandled Exception, code = %08lx\n",
@@ -542,6 +563,16 @@ asmlinkage void trap_c(struct pt_regs *fp)
 
 	BUG_ON(sig == 0);
 
+	/* If the fault was caused by a kernel thread, or interrupt handler
+	 * we will kernel panic, so the system reboots.
+	 */
+	if (kernel_mode_regs(fp) || (current && !current->mm)) {
+		console_verbose();
+		oops_in_progress = 1;
+		if (strerror)
+			verbose_printk(strerror);
+	}
+
 	if (sig != SIGTRAP) {
 		dump_bfin_process(fp);
 		dump_bfin_mem(fp);
@@ -588,8 +619,11 @@ asmlinkage void trap_c(struct pt_regs *fp)
 		force_sig_info(sig, &info, current);
 	}
 
+	if (ANOMALY_05000461 && trapnr == VEC_HWERR && !access_ok(VERIFY_READ, fp->pc, 8))
+		fp->pc = SAFE_USER_INSTRUCTION;
+
+ traps_done:
 	trace_buffer_restore(j);
-	return;
 }
 
 /* Typical exception handling routines	*/
@@ -774,6 +808,18 @@ void dump_bfin_trace_buffer(void)
 }
 EXPORT_SYMBOL(dump_bfin_trace_buffer);
 
+#ifdef CONFIG_BUG
+int is_valid_bugaddr(unsigned long addr)
+{
+	unsigned short opcode;
+
+	if (!get_instruction(&opcode, (unsigned short *)addr))
+		return 0;
+
+	return opcode == BFIN_BUG_OPCODE;
+}
+#endif
+
 /*
  * Checks to see if the address pointed to is either a
  * 16-bit CALL instruction, or a 32-bit CALL instruction
@@ -832,6 +878,11 @@ void show_stack(struct task_struct *task, unsigned long *stack)
 	decode_address(buf, (unsigned int)stack);
 	printk(KERN_NOTICE " SP: [0x%p] %s\n", stack, buf);
 
+	if (!access_ok(VERIFY_READ, stack, (unsigned int)endstack - (unsigned int)stack)) {
+		printk(KERN_NOTICE "Invalid stack pointer\n");
+		return;
+	}
+
 	/* First thing is to look for a frame pointer */
 	for (addr = (unsigned int *)((unsigned int)stack & ~0xF); addr < endstack; addr++) {
 		if (*addr & 0x1)
@@ -1066,6 +1117,29 @@ void show_regs(struct pt_regs *fp)
 	unsigned int cpu = smp_processor_id();
 	unsigned char in_atomic = (bfin_read_IPEND() & 0x10) || in_atomic();
 
+	verbose_printk(KERN_NOTICE "\n");
+	if (CPUID != bfin_cpuid())
+		verbose_printk(KERN_NOTICE "Compiled for cpu family 0x%04x (Rev %d), "
+			"but running on:0x%04x (Rev %d)\n",
+			CPUID, bfin_compiled_revid(), bfin_cpuid(), bfin_revid());
+
+	verbose_printk(KERN_NOTICE "ADSP-%s-0.%d",
+		CPU, bfin_compiled_revid());
+
+	if (bfin_compiled_revid() !=  bfin_revid())
+		verbose_printk("(Detected 0.%d)", bfin_revid());
+
+	verbose_printk(" %lu(MHz CCLK) %lu(MHz SCLK) (%s)\n",
+		get_cclk()/1000000, get_sclk()/1000000,
+#ifdef CONFIG_MPU
+		"mpu on"
+#else
+		"mpu off"
+#endif
+		);
+
+	verbose_printk(KERN_NOTICE "%s", linux_banner);
+
 	verbose_printk(KERN_NOTICE "\n" KERN_NOTICE "SEQUENCER STATUS:\t\t%s\n", print_tainted());
 	verbose_printk(KERN_NOTICE " SEQSTAT: %08lx  IPEND: %04lx  SYSCFG: %04lx\n",
 		(long)fp->seqstat, fp->ipend, fp->syscfg);
@@ -1246,5 +1320,5 @@ void panic_cplb_error(int cplb_panic, struct pt_regs *fp)
 	dump_bfin_mem(fp);
 	show_regs(fp);
 	dump_stack();
-	panic("Unrecoverable event\n");
+	panic("Unrecoverable event");
 }
diff --git a/arch/blackfin/kernel/vmlinux.lds.S b/arch/blackfin/kernel/vmlinux.lds.S
index 27952ae047d..6ac307ca0d8 100644
--- a/arch/blackfin/kernel/vmlinux.lds.S
+++ b/arch/blackfin/kernel/vmlinux.lds.S
@@ -50,8 +50,11 @@ SECTIONS
 		_text = .;
 		__stext = .;
 		TEXT_TEXT
+#ifndef CONFIG_SCHEDULE_L1
 		SCHED_TEXT
+#endif
 		LOCK_TEXT
+		IRQENTRY_TEXT
 		KPROBES_TEXT
 		*(.text.*)
 		*(.fixup)
@@ -164,6 +167,20 @@ SECTIONS
 	}
 	PERCPU(4)
 	SECURITY_INIT
+
+	/* 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...)
+	 */
+	.exit.text :
+	{
+		EXIT_TEXT
+	}
+	.exit.data :
+	{
+		EXIT_DATA
+	}
+
 	.init.ramfs :
 	{
 		. = ALIGN(4);
@@ -180,6 +197,9 @@ SECTIONS
 		. = ALIGN(4);
 		__stext_l1 = .;
 		*(.l1.text)
+#ifdef CONFIG_SCHEDULE_L1
+		SCHED_TEXT
+#endif
 		. = ALIGN(4);
 		__etext_l1 = .;
 	}
@@ -259,8 +279,6 @@ SECTIONS
 
 	/DISCARD/ :
 	{
-		EXIT_TEXT
-		EXIT_DATA
 		*(.exitcall.exit)
 	}
 }