20 files changed, 483 insertions, 201 deletions

diff --git a/arch/blackfin/kernel/.gitignore b/arch/blackfin/kernel/.gitignore
new file mode 100644
index 00000000000..c5f676c3c22
--- /dev/null
+++ b/arch/blackfin/kernel/.gitignore
@@ -0,0 +1 @@
+vmlinux.lds
diff --git a/arch/blackfin/kernel/bfin_dma_5xx.c b/arch/blackfin/kernel/bfin_dma_5xx.c
index 8531693fb48..763ed84ba45 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();
 }
 
 /**
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..53e893ff708 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);
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-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/early_printk.c b/arch/blackfin/kernel/early_printk.c
index c8ad051742e..3302719173c 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
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..5fc424803a1 100644
--- a/arch/blackfin/kernel/ipipe.c
+++ b/arch/blackfin/kernel/ipipe.c
@@ -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]))
@@ -310,12 +309,16 @@ 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)
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..80447f99c2b 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();
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