Merge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm

Pull ARM updates from Russell King:

 - handle inexact watchpoint addresses (Douglas Anderson)

 - decompressor serial debug cleanups (Linus Walleij)

 - update L2 cache prefetch bits (Guillaume Tucker)

 - add text offset and malloc size to the decompressor kexec data

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: add malloc size to decompressor kexec size structure
  ARM: add TEXT_OFFSET to decompressor kexec image structure
  ARM: 9007/1: l2c: fix prefetch bits init in L2X0_AUX_CTRL using DT values
  ARM: 9010/1: uncompress: Print the location of appended DTB
  ARM: 9009/1: uncompress: Enable debug in head.S
  ARM: 9008/1: uncompress: Drop excess whitespace print
  ARM: 9006/1: uncompress: Wait for ready and busy in debug prints
  ARM: 9005/1: debug: Select flow control for all debug UARTs
  ARM: 9004/1: debug: Split waituart to CTS and TXRDY
  ARM: 9003/1: uncompress: Delete unused debug macros
  ARM: 8997/2: hw_breakpoint: Handle inexact watchpoint addresses

34 files changed, 263 insertions, 112 deletions

diff --git a/arch/arm/Kconfig.debug b/arch/arm/Kconfig.debug
index 80000a66a4e3..87912e5c2256 100644
--- a/arch/arm/Kconfig.debug
+++ b/arch/arm/Kconfig.debug
@@ -1546,6 +1546,17 @@ config DEBUG_SIRFSOC_UART
 	bool
 	depends on ARCH_SIRF
 
+config DEBUG_UART_FLOW_CONTROL
+	bool "Enable flow control (CTS) for the debug UART"
+	depends on DEBUG_LL
+	default y if ARCH_EBSA110 || DEBUG_FOOTBRIDGE_COM1 || DEBUG_GEMINI || ARCH_RPC
+	help
+	  Some UART ports are connected to terminals that will use modem
+	  control signals to indicate whether they are ready to receive text.
+	  In practice this means that the terminal is asserting the special
+	  control signal CTS (Clear To Send). If your debug UART supports
+	  this and your debug terminal will require it, enable this option.
+
 config DEBUG_LL_INCLUDE
 	string
 	default "debug/sa1100.S" if DEBUG_SA1100
@@ -1893,11 +1904,6 @@ config DEBUG_UART_8250_PALMCHIP
 	  except for having a different register layout.  Say Y here if
 	  the debug UART is of this type.
 
-config DEBUG_UART_8250_FLOW_CONTROL
-	bool "Enable flow control for 8250 UART"
-	depends on DEBUG_LL_UART_8250 || DEBUG_UART_8250
-	default y if ARCH_EBSA110 || DEBUG_FOOTBRIDGE_COM1 || DEBUG_GEMINI || ARCH_RPC
-
 config DEBUG_UNCOMPRESS
 	bool "Enable decompressor debugging via DEBUG_LL output"
 	depends on ARCH_MULTIPLATFORM || PLAT_SAMSUNG || ARM_SINGLE_ARMV7M
diff --git a/arch/arm/Makefile b/arch/arm/Makefile
index e589da3c8949..c4301437ca72 100644
--- a/arch/arm/Makefile
+++ b/arch/arm/Makefile
@@ -143,6 +143,9 @@ head-y		:= arch/arm/kernel/head$(MMUEXT).o
 
 # Text offset. This list is sorted numerically by address in order to
 # provide a means to avoid/resolve conflicts in multi-arch kernels.
+# Note: the 32kB below this value is reserved for use by the kernel
+# during boot, and this offset is critical to the functioning of
+# kexec-tools.
 textofs-y	:= 0x00008000
 # We don't want the htc bootloader to corrupt kernel during resume
 textofs-$(CONFIG_PM_H1940)      := 0x00108000
diff --git a/arch/arm/boot/compressed/Makefile b/arch/arm/boot/compressed/Makefile
index 58028abd05d9..47f001ca5499 100644
--- a/arch/arm/boot/compressed/Makefile
+++ b/arch/arm/boot/compressed/Makefile
@@ -7,11 +7,11 @@
 
 OBJS		=
 
-AFLAGS_head.o += -DTEXT_OFFSET=$(TEXT_OFFSET)
 HEAD	= head.o
 OBJS	+= misc.o decompress.o
 ifeq ($(CONFIG_DEBUG_UNCOMPRESS),y)
 OBJS	+= debug.o
+AFLAGS_head.o += -DDEBUG
 endif
 FONTC	= $(srctree)/lib/fonts/font_acorn_8x8.c
 
@@ -68,7 +68,12 @@ ZTEXTADDR	:= 0
 ZBSSADDR	:= ALIGN(8)
 endif
 
+MALLOC_SIZE	:= 65536
+
+AFLAGS_head.o += -DTEXT_OFFSET=$(TEXT_OFFSET) -DMALLOC_SIZE=$(MALLOC_SIZE)
 CPPFLAGS_vmlinux.lds := -DTEXT_START="$(ZTEXTADDR)" -DBSS_START="$(ZBSSADDR)"
+CPPFLAGS_vmlinux.lds += -DTEXT_OFFSET="$(TEXT_OFFSET)"
+CPPFLAGS_vmlinux.lds += -DMALLOC_SIZE="$(MALLOC_SIZE)"
 
 compress-$(CONFIG_KERNEL_GZIP) = gzip
 compress-$(CONFIG_KERNEL_LZO)  = lzo
diff --git a/arch/arm/boot/compressed/debug.S b/arch/arm/boot/compressed/debug.S
index 6bf2917a4621..fac40a717fcf 100644
--- a/arch/arm/boot/compressed/debug.S
+++ b/arch/arm/boot/compressed/debug.S
@@ -8,7 +8,10 @@
 
 ENTRY(putc)
 	addruart r1, r2, r3
-	waituart r3, r1
+#ifdef CONFIG_DEBUG_UART_FLOW_CONTROL
+	waituartcts r3, r1
+#endif
+	waituarttxrdy r3, r1
 	senduart r0, r1
 	busyuart r3, r1
 	mov	 pc, lr
diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S
index 434a16982e34..2e04ec5b5446 100644
--- a/arch/arm/boot/compressed/head.S
+++ b/arch/arm/boot/compressed/head.S
@@ -28,19 +28,19 @@
 #if defined(CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K) || defined(CONFIG_CPU_V7)
 		.macro	loadsp, rb, tmp1, tmp2
 		.endm
-		.macro	writeb, ch, rb
+		.macro	writeb, ch, rb, tmp
 		mcr	p14, 0, \ch, c0, c5, 0
 		.endm
 #elif defined(CONFIG_CPU_XSCALE)
 		.macro	loadsp, rb, tmp1, tmp2
 		.endm
-		.macro	writeb, ch, rb
+		.macro	writeb, ch, rb, tmp
 		mcr	p14, 0, \ch, c8, c0, 0
 		.endm
 #else
 		.macro	loadsp, rb, tmp1, tmp2
 		.endm
-		.macro	writeb, ch, rb
+		.macro	writeb, ch, rb, tmp
 		mcr	p14, 0, \ch, c1, c0, 0
 		.endm
 #endif
@@ -49,8 +49,13 @@
 
 #include CONFIG_DEBUG_LL_INCLUDE
 
-		.macro	writeb,	ch, rb
+		.macro	writeb,	ch, rb, tmp
+#ifdef CONFIG_DEBUG_UART_FLOW_CONTROL
+		waituartcts \tmp, \rb
+#endif
+		waituarttxrdy \tmp, \rb
 		senduart \ch, \rb
+		busyuart \tmp, \rb
 		.endm
 
 #if defined(CONFIG_ARCH_SA1100)
@@ -81,42 +86,11 @@
 		bl	phex
 		.endm
 
-		.macro	debug_reloc_start
-#ifdef DEBUG
-		kputc	#'\n'
-		kphex	r6, 8		/* processor id */
-		kputc	#':'
-		kphex	r7, 8		/* architecture id */
-#ifdef CONFIG_CPU_CP15
-		kputc	#':'
-		mrc	p15, 0, r0, c1, c0
-		kphex	r0, 8		/* control reg */
-#endif
-		kputc	#'\n'
-		kphex	r5, 8		/* decompressed kernel start */
-		kputc	#'-'
-		kphex	r9, 8		/* decompressed kernel end  */
-		kputc	#'>'
-		kphex	r4, 8		/* kernel execution address */
-		kputc	#'\n'
-#endif
-		.endm
-
-		.macro	debug_reloc_end
-#ifdef DEBUG
-		kphex	r5, 8		/* end of kernel */
-		kputc	#'\n'
-		mov	r0, r4
-		bl	memdump		/* dump 256 bytes at start of kernel */
-#endif
-		.endm
-
 		/*
 		 * Debug kernel copy by printing the memory addresses involved
 		 */
 		.macro dbgkc, begin, end, cbegin, cend
 #ifdef DEBUG
-		kputc   #'\n'
 		kputc   #'C'
 		kputc   #':'
 		kputc   #'0'
@@ -136,7 +110,28 @@
 		kputc	#'x'
 		kphex	\cend, 8	/* End of kernel copy */
 		kputc	#'\n'
-		kputc	#'\r'
+#endif
+		.endm
+
+		/*
+		 * Debug print of the final appended DTB location
+		 */
+		.macro dbgadtb, begin, end
+#ifdef DEBUG
+		kputc   #'D'
+		kputc   #'T'
+		kputc   #'B'
+		kputc   #':'
+		kputc   #'0'
+		kputc   #'x'
+		kphex   \begin, 8	/* Start of appended DTB */
+		kputc	#' '
+		kputc	#'('
+		kputc	#'0'
+		kputc	#'x'
+		kphex	\end, 8		/* End of appended DTB */
+		kputc	#')'
+		kputc	#'\n'
 #endif
 		.endm
 
@@ -303,7 +298,7 @@ restart:	adr	r0, LC1
 
 #ifndef CONFIG_ZBOOT_ROM
 		/* malloc space is above the relocated stack (64k max) */
-		add	r10, sp, #0x10000
+		add	r10, sp, #MALLOC_SIZE
 #else
 		/*
 		 * With ZBOOT_ROM the bss/stack is non relocatable,
@@ -357,6 +352,7 @@ restart:	adr	r0, LC1
 		mov	r5, r5, ror #8
 		eor	r5, r5, r1, lsr #8
 #endif
+		dbgadtb	r6, r5
 		/* 50% DTB growth should be good enough */
 		add	r5, r5, r5, lsr #1
 		/* preserve 64-bit alignment */
@@ -614,7 +610,7 @@ not_relocated:	mov	r0, #0
  */
 		mov	r0, r4
 		mov	r1, sp			@ malloc space above stack
-		add	r2, sp, #0x10000	@ 64k max
+		add	r2, sp, #MALLOC_SIZE	@ 64k max
 		mov	r3, r7
 		bl	decompress_kernel
 
@@ -1356,7 +1352,7 @@ puts:		loadsp	r3, r2, r1
 1:		ldrb	r2, [r0], #1
 		teq	r2, #0
 		moveq	pc, lr
-2:		writeb	r2, r3
+2:		writeb	r2, r3, r1
 		mov	r1, #0x00020000
 3:		subs	r1, r1, #1
 		bne	3b
diff --git a/arch/arm/boot/compressed/vmlinux.lds.S b/arch/arm/boot/compressed/vmlinux.lds.S
index b914be3a207b..1bcb68ac4b01 100644
--- a/arch/arm/boot/compressed/vmlinux.lds.S
+++ b/arch/arm/boot/compressed/vmlinux.lds.S
@@ -44,10 +44,12 @@ SECTIONS
   }
   .table : ALIGN(4) {
     _table_start = .;
-    LONG(ZIMAGE_MAGIC(4))
+    LONG(ZIMAGE_MAGIC(6))
     LONG(ZIMAGE_MAGIC(0x5a534c4b))
     LONG(ZIMAGE_MAGIC(__piggy_size_addr - _start))
     LONG(ZIMAGE_MAGIC(_kernel_bss_size))
+    LONG(ZIMAGE_MAGIC(TEXT_OFFSET))
+    LONG(ZIMAGE_MAGIC(MALLOC_SIZE))
     LONG(0)
     _table_end = .;
   }
diff --git a/arch/arm/include/debug/8250.S b/arch/arm/include/debug/8250.S
index e4a036f082c2..e3692a37cede 100644
--- a/arch/arm/include/debug/8250.S
+++ b/arch/arm/include/debug/8250.S
@@ -45,10 +45,11 @@
 		bne	1002b
 		.endm
 
-		.macro	waituart,rd,rx
-#ifdef CONFIG_DEBUG_UART_8250_FLOW_CONTROL
+		.macro	waituarttxrdy,rd,rx
+		.endm
+
+		.macro	waituartcts,rd,rx
 1001:		load	\rd, [\rx, #UART_MSR << UART_SHIFT]
 		tst	\rd, #UART_MSR_CTS
 		beq	1001b
-#endif
 		.endm
diff --git a/arch/arm/include/debug/asm9260.S b/arch/arm/include/debug/asm9260.S
index 0da1eb625331..5a0ce145c44a 100644
--- a/arch/arm/include/debug/asm9260.S
+++ b/arch/arm/include/debug/asm9260.S
@@ -11,7 +11,10 @@
 		ldr	\rv, = CONFIG_DEBUG_UART_VIRT
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituarttxrdy,rd,rx
+		.endm
+
+		.macro	waituartcts,rd,rx
 		.endm
 
 		.macro	senduart,rd,rx
diff --git a/arch/arm/include/debug/at91.S b/arch/arm/include/debug/at91.S
index 6c91cbaaa20b..17722824e2f2 100644
--- a/arch/arm/include/debug/at91.S
+++ b/arch/arm/include/debug/at91.S
@@ -19,12 +19,15 @@
 	strb	\rd, [\rx, #(AT91_DBGU_THR)]		@ Write to Transmitter Holding Register
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituarttxrdy,rd,rx
 1001:	ldr	\rd, [\rx, #(AT91_DBGU_SR)]		@ Read Status Register
 	tst	\rd, #AT91_DBGU_TXRDY			@ DBGU_TXRDY = 1 when ready to transmit
 	beq	1001b
 	.endm
 
+	.macro	waituartcts,rd,rx
+	.endm
+
 	.macro	busyuart,rd,rx
 1001:	ldr	\rd, [\rx, #(AT91_DBGU_SR)]		@ Read Status Register
 	tst	\rd, #AT91_DBGU_TXEMPTY			@ DBGU_TXEMPTY = 1 when transmission complete
diff --git a/arch/arm/include/debug/bcm63xx.S b/arch/arm/include/debug/bcm63xx.S
index 06a896227396..da65abb6738d 100644
--- a/arch/arm/include/debug/bcm63xx.S
+++ b/arch/arm/include/debug/bcm63xx.S
@@ -17,12 +17,15 @@
 	strb	\rd, [\rx, #UART_FIFO_REG]
 	.endm
 
-	.macro	waituart, rd, rx
+	.macro	waituarttxrdy, rd, rx
 1001:	ldr	\rd, [\rx, #UART_IR_REG]
 	tst	\rd, #(1 << UART_IR_TXEMPTY)
 	beq	1001b
 	.endm
 
+	.macro	waituartcts, rd, rx
+	.endm
+
 	.macro	busyuart, rd, rx
 1002:	ldr	\rd, [\rx, #UART_IR_REG]
 	tst	\rd, #(1 << UART_IR_TXTRESH)
diff --git a/arch/arm/include/debug/brcmstb.S b/arch/arm/include/debug/brcmstb.S
index 132a20c4a676..7ffe66993029 100644
--- a/arch/arm/include/debug/brcmstb.S
+++ b/arch/arm/include/debug/brcmstb.S
@@ -142,7 +142,10 @@ ARM_BE8(	rev	\rd, \rd )
 		bne	1002b
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituarttxrdy,rd,rx
+		.endm
+
+		.macro	waituartcts,rd,rx
 		.endm
 
 /*
diff --git a/arch/arm/include/debug/clps711x.S b/arch/arm/include/debug/clps711x.S
index 774a67ac3877..a983d12a6515 100644
--- a/arch/arm/include/debug/clps711x.S
+++ b/arch/arm/include/debug/clps711x.S
@@ -20,7 +20,10 @@
 	ldr	\rp, =CLPS711X_UART_PADDR
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 	.endm
 
 	.macro	senduart,rd,rx
diff --git a/arch/arm/include/debug/dc21285.S b/arch/arm/include/debug/dc21285.S
index d7e8c71706ab..4ec0e5e31704 100644
--- a/arch/arm/include/debug/dc21285.S
+++ b/arch/arm/include/debug/dc21285.S
@@ -34,5 +34,8 @@
 		bne	1001b
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 		.endm
diff --git a/arch/arm/include/debug/digicolor.S b/arch/arm/include/debug/digicolor.S
index 256f5f4da275..443674cad76a 100644
--- a/arch/arm/include/debug/digicolor.S
+++ b/arch/arm/include/debug/digicolor.S
@@ -21,7 +21,10 @@
 		strb	\rd, [\rx, #UA0_EMI_REC]
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 		.endm
 
 	.macro	busyuart,rd,rx
diff --git a/arch/arm/include/debug/efm32.S b/arch/arm/include/debug/efm32.S
index 5ed5028306f4..b0083d6e31e8 100644
--- a/arch/arm/include/debug/efm32.S
+++ b/arch/arm/include/debug/efm32.S
@@ -29,7 +29,10 @@
 		strb	\rd, [\rx, #UARTn_TXDATA]
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 1001:		ldr	\rd, [\rx, #UARTn_STATUS]
 		tst	\rd, #UARTn_STATUS_TXBL
 		beq	1001b
diff --git a/arch/arm/include/debug/icedcc.S b/arch/arm/include/debug/icedcc.S
index 74a0dd036a17..d5e65da8a687 100644
--- a/arch/arm/include/debug/icedcc.S
+++ b/arch/arm/include/debug/icedcc.S
@@ -23,7 +23,10 @@
 		beq	1001b
 		.endm
 
-		.macro	waituart, rd, rx
+		.macro	waituartcts, rd, rx
+		.endm
+
+		.macro	waituarttxrdy, rd, rx
 		mov	\rd, #0x2000000
 1001:
 		subs	\rd, \rd, #1
@@ -47,7 +50,10 @@
 		beq	1001b
 		.endm
 
-		.macro	waituart, rd, rx
+		.macro	waituartcts, rd, rx
+		.endm
+
+		.macro	waituarttxrdy, rd, rx
 		mov	\rd, #0x10000000
 1001:
 		subs	\rd, \rd, #1
@@ -72,7 +78,10 @@
 
 		.endm
 
-		.macro	waituart, rd, rx
+		.macro	waituartcts, rd, rx
+		.endm
+
+		.macro	waituarttxrdy, rd, rx
 		mov	\rd, #0x2000000
 1001:
 		subs	\rd, \rd, #1
diff --git a/arch/arm/include/debug/imx.S b/arch/arm/include/debug/imx.S
index 1c1b9d1da4c8..bb7b9550580c 100644
--- a/arch/arm/include/debug/imx.S
+++ b/arch/arm/include/debug/imx.S
@@ -35,7 +35,10 @@
 		str	\rd, [\rx, #0x40]	@ TXDATA
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 		.endm
 
 		.macro	busyuart,rd,rx
diff --git a/arch/arm/include/debug/meson.S b/arch/arm/include/debug/meson.S
index 1e501a0054ae..7b60e4401225 100644
--- a/arch/arm/include/debug/meson.S
+++ b/arch/arm/include/debug/meson.S
@@ -25,7 +25,10 @@
 	beq	1002b
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 1001:	ldr	\rd, [\rx, #MESON_AO_UART_STATUS]
 	tst	\rd, #MESON_AO_UART_TX_FIFO_FULL
 	bne	1001b
diff --git a/arch/arm/include/debug/msm.S b/arch/arm/include/debug/msm.S
index 9405b71461da..530edc74f9a3 100644
--- a/arch/arm/include/debug/msm.S
+++ b/arch/arm/include/debug/msm.S
@@ -17,7 +17,10 @@ ARM_BE8(rev	\rd, \rd )
 	str	\rd, [\rx, #0x70]
 	.endm
 
-	.macro	waituart, rd, rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy, rd, rx
 	@ check for TX_EMT in UARTDM_SR
 	ldr	\rd, [\rx, #0x08]
 ARM_BE8(rev     \rd, \rd )
diff --git a/arch/arm/include/debug/omap2plus.S b/arch/arm/include/debug/omap2plus.S
index b5696a33ba0f..0680be6c79d3 100644
--- a/arch/arm/include/debug/omap2plus.S
+++ b/arch/arm/include/debug/omap2plus.S
@@ -75,5 +75,8 @@ omap_uart_lsr:	.word	0
 		bne	1001b
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 		.endm
diff --git a/arch/arm/include/debug/pl01x.S b/arch/arm/include/debug/pl01x.S
index a2a553afe7b8..0c7bfa4c10db 100644
--- a/arch/arm/include/debug/pl01x.S
+++ b/arch/arm/include/debug/pl01x.S
@@ -26,7 +26,10 @@
 		strb	\rd, [\rx, #UART01x_DR]
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 1001:		ldr	\rd, [\rx, #UART01x_FR]
  ARM_BE8(	rev	\rd, \rd )
 		tst	\rd, #UART01x_FR_TXFF
diff --git a/arch/arm/include/debug/renesas-scif.S b/arch/arm/include/debug/renesas-scif.S
index 25f06663a9a4..8e433e981bbe 100644
--- a/arch/arm/include/debug/renesas-scif.S
+++ b/arch/arm/include/debug/renesas-scif.S
@@ -33,7 +33,10 @@
 	ldr	\rv, =SCIF_VIRT
 	.endm
 
-	.macro	waituart, rd, rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy, rd, rx
 1001:	ldrh	\rd, [\rx, #FSR]
 	tst	\rd, #TDFE
 	beq	1001b
diff --git a/arch/arm/include/debug/sa1100.S b/arch/arm/include/debug/sa1100.S
index 6109e6058e5b..7968ea52df3d 100644
--- a/arch/arm/include/debug/sa1100.S
+++ b/arch/arm/include/debug/sa1100.S
@@ -51,7 +51,10 @@
 		str	\rd, [\rx, #UTDR]
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 1001:		ldr	\rd, [\rx, #UTSR1]
 		tst	\rd, #UTSR1_TNF
 		beq	1001b
diff --git a/arch/arm/include/debug/samsung.S b/arch/arm/include/debug/samsung.S
index 69201d7fb48f..ab474d564a90 100644
--- a/arch/arm/include/debug/samsung.S
+++ b/arch/arm/include/debug/samsung.S
@@ -69,7 +69,10 @@ ARM_BE8(rev \rd, \rd)
 1002:		@ exit busyuart
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 		ldr	\rd, [\rx, # S3C2410_UFCON]
 ARM_BE8(rev \rd, \rd)
 		tst	\rd, #S3C2410_UFCON_FIFOMODE	@ fifo enabled?
diff --git a/arch/arm/include/debug/sirf.S b/arch/arm/include/debug/sirf.S
index e73e4de0a015..3612c7b9cbe7 100644
--- a/arch/arm/include/debug/sirf.S
+++ b/arch/arm/include/debug/sirf.S
@@ -29,7 +29,10 @@
 	.macro	busyuart,rd,rx
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 1001:	ldr	\rd, [\rx, #SIRF_LLUART_TXFIFO_STATUS]
 	tst	\rd, #SIRF_LLUART_TXFIFO_EMPTY
 	beq	1001b
diff --git a/arch/arm/include/debug/sti.S b/arch/arm/include/debug/sti.S
index 6b42c91f217d..72d052511890 100644
--- a/arch/arm/include/debug/sti.S
+++ b/arch/arm/include/debug/sti.S
@@ -45,7 +45,10 @@
                 strb    \rd, [\rx, #ASC_TX_BUF_OFF]
                 .endm
 
-                .macro  waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+                .macro  waituarttxrdy,rd,rx
 1001:           ldr     \rd, [\rx, #ASC_STA_OFF]
                 tst     \rd, #ASC_STA_TX_FULL
                 bne     1001b
diff --git a/arch/arm/include/debug/stm32.S b/arch/arm/include/debug/stm32.S
index f3c4a37210ed..b6d9df30e37d 100644
--- a/arch/arm/include/debug/stm32.S
+++ b/arch/arm/include/debug/stm32.S
@@ -27,7 +27,10 @@
 	strb    \rd, [\rx, #STM32_USART_TDR_OFF]
 .endm
 
-.macro  waituart,rd,rx
+.macro	waituartcts,rd,rx
+.endm
+
+.macro  waituarttxrdy,rd,rx
 1001:	ldr	\rd, [\rx, #(STM32_USART_SR_OFF)]	@ Read Status Register
 	tst	\rd, #STM32_USART_TXE			@ TXE = 1 = tx empty
 	beq	1001b
diff --git a/arch/arm/include/debug/tegra.S b/arch/arm/include/debug/tegra.S
index 2148d0f88591..98daa7f48314 100644
--- a/arch/arm/include/debug/tegra.S
+++ b/arch/arm/include/debug/tegra.S
@@ -178,15 +178,16 @@
 1002:
 		.endm
 
-		.macro	waituart, rd, rx
-#ifdef FLOW_CONTROL
+		.macro	waituartcts, rd, rx
 		cmp	\rx, #0
 		beq	1002f
 1001:		ldrb	\rd, [\rx, #UART_MSR << UART_SHIFT]
 		tst	\rd, #UART_MSR_CTS
 		beq	1001b
 1002:
-#endif
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 		.endm
 
 /*
diff --git a/arch/arm/include/debug/vf.S b/arch/arm/include/debug/vf.S
index 854d9bd82770..035bcbf117ab 100644
--- a/arch/arm/include/debug/vf.S
+++ b/arch/arm/include/debug/vf.S
@@ -29,5 +29,8 @@
 	beq	1001b			@ wait until transmit done
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 	.endm
diff --git a/arch/arm/include/debug/vt8500.S b/arch/arm/include/debug/vt8500.S
index 8dc1df2d91b8..d01094fdbc8c 100644
--- a/arch/arm/include/debug/vt8500.S
+++ b/arch/arm/include/debug/vt8500.S
@@ -28,7 +28,10 @@
 	bne	1001b
 	.endm
 
-	.macro	waituart,rd,rx
+	.macro	waituartcts,rd,rx
+	.endm
+
+	.macro	waituarttxrdy,rd,rx
 	.endm
 
 #endif
diff --git a/arch/arm/include/debug/zynq.S b/arch/arm/include/debug/zynq.S
index 58d77c972fd6..5d42cc35ecf3 100644
--- a/arch/arm/include/debug/zynq.S
+++ b/arch/arm/include/debug/zynq.S
@@ -33,7 +33,10 @@
 		strb	\rd, [\rx, #UART_FIFO_OFFSET]	@ TXDATA
 		.endm
 
-		.macro	waituart,rd,rx
+		.macro	waituartcts,rd,rx
+		.endm
+
+		.macro	waituarttxrdy,rd,rx
 1001:		ldr	\rd, [\rx, #UART_SR_OFFSET]
 ARM_BE8(	rev	\rd, \rd )
 		tst	\rd, #UART_SR_TXEMPTY
diff --git a/arch/arm/kernel/debug.S b/arch/arm/kernel/debug.S
index e112072b579d..d92f44bdf438 100644
--- a/arch/arm/kernel/debug.S
+++ b/arch/arm/kernel/debug.S
@@ -89,11 +89,18 @@ ENTRY(printascii)
 2:		teq     r1, #'\n'
 		bne	3f
 		mov	r1, #'\r'
-		waituart r2, r3
+#ifdef CONFIG_DEBUG_UART_FLOW_CONTROL
+		waituartcts r2, r3
+#endif
+		waituarttxrdy r2, r3
 		senduart r1, r3
 		busyuart r2, r3
 		mov	r1, #'\n'
-3:		waituart r2, r3
+3:
+#ifdef CONFIG_DEBUG_UART_FLOW_CONTROL
+		waituartcts r2, r3
+#endif
+		waituarttxrdy r2, r3
 		senduart r1, r3
 		busyuart r2, r3
 		b	1b
diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
index 7a4853b1213a..08660ae9dcbc 100644
--- a/arch/arm/kernel/hw_breakpoint.c
+++ b/arch/arm/kernel/hw_breakpoint.c
@@ -683,6 +683,40 @@ static void disable_single_step(struct perf_event *bp)
 	arch_install_hw_breakpoint(bp);
 }
 
+/*
+ * Arm32 hardware does not always report a watchpoint hit address that matches
+ * one of the watchpoints set. It can also report an address "near" the
+ * watchpoint if a single instruction access both watched and unwatched
+ * addresses. There is no straight-forward way, short of disassembling the
+ * offending instruction, to map that address back to the watchpoint. This
+ * function computes the distance of the memory access from the watchpoint as a
+ * heuristic for the likelyhood that a given access triggered the watchpoint.
+ *
+ * See this same function in the arm64 platform code, which has the same
+ * problem.
+ *
+ * The function returns the distance of the address from the bytes watched by
+ * the watchpoint. In case of an exact match, it returns 0.
+ */
+static u32 get_distance_from_watchpoint(unsigned long addr, u32 val,
+					struct arch_hw_breakpoint_ctrl *ctrl)
+{
+	u32 wp_low, wp_high;
+	u32 lens, lene;
+
+	lens = __ffs(ctrl->len);
+	lene = __fls(ctrl->len);
+
+	wp_low = val + lens;
+	wp_high = val + lene;
+	if (addr < wp_low)
+		return wp_low - addr;
+	else if (addr > wp_high)
+		return addr - wp_high;
+	else
+		return 0;
+}
+
 static int watchpoint_fault_on_uaccess(struct pt_regs *regs,
 				       struct arch_hw_breakpoint *info)
 {
@@ -692,23 +726,25 @@ static int watchpoint_fault_on_uaccess(struct pt_regs *regs,
 static void watchpoint_handler(unsigned long addr, unsigned int fsr,
 			       struct pt_regs *regs)
 {
-	int i, access;
-	u32 val, ctrl_reg, alignment_mask;
+	int i, access, closest_match = 0;
+	u32 min_dist = -1, dist;
+	u32 val, ctrl_reg;
 	struct perf_event *wp, **slots;
 	struct arch_hw_breakpoint *info;
 	struct arch_hw_breakpoint_ctrl ctrl;
 
 	slots = this_cpu_ptr(wp_on_reg);
 
+	/*
+	 * Find all watchpoints that match the reported address. If no exact
+	 * match is found. Attribute the hit to the closest watchpoint.
+	 */
+	rcu_read_lock();
 	for (i = 0; i < core_num_wrps; ++i) {
-		rcu_read_lock();
-
 		wp = slots[i];
-
 		if (wp == NULL)
-			goto unlock;
+			continue;
 
-		info = counter_arch_bp(wp);
 		/*
 		 * The DFAR is an unknown value on debug architectures prior
 		 * to 7.1. Since we only allow a single watchpoint on these
@@ -717,33 +753,31 @@ static void watchpoint_handler(unsigned long addr, unsigned int fsr,
 		 */
 		if (debug_arch < ARM_DEBUG_ARCH_V7_1) {
 			BUG_ON(i > 0);
+			info = counter_arch_bp(wp);
 			info->trigger = wp->attr.bp_addr;
 		} else {
-			if (info->ctrl.len == ARM_BREAKPOINT_LEN_8)
-				alignment_mask = 0x7;
-			else
-				alignment_mask = 0x3;
-
-			/* Check if the watchpoint value matches. */
-			val = read_wb_reg(ARM_BASE_WVR + i);
-			if (val != (addr & ~alignment_mask))
-				goto unlock;
-
-			/* Possible match, check the byte address select. */
-			ctrl_reg = read_wb_reg(ARM_BASE_WCR + i);
-			decode_ctrl_reg(ctrl_reg, &ctrl);
-			if (!((1 << (addr & alignment_mask)) & ctrl.len))
-				goto unlock;
-
 			/* Check that the access type matches. */
 			if (debug_exception_updates_fsr()) {
 				access = (fsr & ARM_FSR_ACCESS_MASK) ?
 					  HW_BREAKPOINT_W : HW_BREAKPOINT_R;
 				if (!(access & hw_breakpoint_type(wp)))
-					goto unlock;
+					continue;
 			}
 
+			val = read_wb_reg(ARM_BASE_WVR + i);
+			ctrl_reg = read_wb_reg(ARM_BASE_WCR + i);
+			decode_ctrl_reg(ctrl_reg, &ctrl);
+			dist = get_distance_from_watchpoint(addr, val, &ctrl);
+			if (dist < min_dist) {
+				min_dist = dist;
+				closest_match = i;
+			}
+			/* Is this an exact match? */
+			if (dist != 0)
+				continue;
+
 			/* We have a winner. */
+			info = counter_arch_bp(wp);
 			info->trigger = addr;
 		}
 
@@ -765,13 +799,23 @@ static void watchpoint_handler(unsigned long addr, unsigned int fsr,
 		 * we can single-step over the watchpoint trigger.
 		 */
 		if (!is_default_overflow_handler(wp))
-			goto unlock;
-
+			continue;
 step:
 		enable_single_step(wp, instruction_pointer(regs));
-unlock:
-		rcu_read_unlock();
 	}
+
+	if (min_dist > 0 && min_dist != -1) {
+		/* No exact match found. */
+		wp = slots[closest_match];
+		info = counter_arch_bp(wp);
+		info->trigger = addr;
+		pr_debug("watchpoint fired: address = 0x%x\n", info->trigger);
+		perf_bp_event(wp, regs);
+		if (is_default_overflow_handler(wp))
+			enable_single_step(wp, instruction_pointer(regs));
+	}
+
+	rcu_read_unlock();
 }
 
 static void watchpoint_single_step_handler(unsigned long pc)
diff --git a/arch/arm/mm/cache-l2x0.c b/arch/arm/mm/cache-l2x0.c
index 12c26eb88afb..43d91bfd2360 100644
--- a/arch/arm/mm/cache-l2x0.c
+++ b/arch/arm/mm/cache-l2x0.c
@@ -1249,20 +1249,28 @@ static void __init l2c310_of_parse(const struct device_node *np,
 
 	ret = of_property_read_u32(np, "prefetch-data", &val);
 	if (ret == 0) {
-		if (val)
+		if (val) {
 			prefetch |= L310_PREFETCH_CTRL_DATA_PREFETCH;
-		else
+			*aux_val |= L310_PREFETCH_CTRL_DATA_PREFETCH;
+		} else {
 			prefetch &= ~L310_PREFETCH_CTRL_DATA_PREFETCH;
+			*aux_val &= ~L310_PREFETCH_CTRL_DATA_PREFETCH;
+		}
+		*aux_mask &= ~L310_PREFETCH_CTRL_DATA_PREFETCH;
 	} else if (ret != -EINVAL) {
 		pr_err("L2C-310 OF prefetch-data property value is missing\n");
 	}
 
 	ret = of_property_read_u32(np, "prefetch-instr", &val);
 	if (ret == 0) {
-		if (val)
+		if (val) {
 			prefetch |= L310_PREFETCH_CTRL_INSTR_PREFETCH;
-		else
+			*aux_val |= L310_PREFETCH_CTRL_INSTR_PREFETCH;
+		} else {
 			prefetch &= ~L310_PREFETCH_CTRL_INSTR_PREFETCH;
+			*aux_val &= ~L310_PREFETCH_CTRL_INSTR_PREFETCH;
+		}
+		*aux_mask &= ~L310_PREFETCH_CTRL_INSTR_PREFETCH;
 	} else if (ret != -EINVAL) {
 		pr_err("L2C-310 OF prefetch-instr property value is missing\n");
 	}