Merge master.kernel.org:/home/rmk/linux-2.6-arm

19 files changed, 337 insertions, 484 deletions

diff --git a/Documentation/dontdiff b/Documentation/dontdiff
index b974cf595d0..96bea278bbf 100644
--- a/Documentation/dontdiff
+++ b/Documentation/dontdiff
@@ -104,6 +104,7 @@ logo_*.c
 logo_*_clut224.c
 logo_*_mono.c
 lxdialog
+mach-types
 mach-types.h
 make_times_h
 map
diff --git a/arch/arm/mach-ixp4xx/coyote-setup.c b/arch/arm/mach-ixp4xx/coyote-setup.c
index 4ff4393ef0e..7f58afb27e7 100644
--- a/arch/arm/mach-ixp4xx/coyote-setup.c
+++ b/arch/arm/mach-ixp4xx/coyote-setup.c
@@ -61,7 +61,7 @@ static struct plat_serial8250_port coyote_uart_data[] = {
 		.mapbase	= IXP4XX_UART2_BASE_PHYS,
 		.membase	= (char *)IXP4XX_UART2_BASE_VIRT + REG_OFFSET,
 		.irq		= IRQ_IXP4XX_UART2,
-		.flags		= UPF_BOOT_AUTOCONF,
+		.flags		= UPF_BOOT_AUTOCONF | UPF_SKIP_TEST,
 		.iotype		= UPIO_MEM,
 		.regshift	= 2,
 		.uartclk	= IXP4XX_UART_XTAL,
diff --git a/arch/arm/mach-ixp4xx/gtwx5715-setup.c b/arch/arm/mach-ixp4xx/gtwx5715-setup.c
index 8ba1cd9406e..65e356bd10d 100644
--- a/arch/arm/mach-ixp4xx/gtwx5715-setup.c
+++ b/arch/arm/mach-ixp4xx/gtwx5715-setup.c
@@ -83,7 +83,7 @@ static struct plat_serial8250_port gtwx5715_uart_platform_data[] = {
 	.mapbase	= IXP4XX_UART2_BASE_PHYS,
 	.membase	= (char *)IXP4XX_UART2_BASE_VIRT + REG_OFFSET,
 	.irq		= IRQ_IXP4XX_UART2,
-	.flags		= UPF_BOOT_AUTOCONF,
+	.flags		= UPF_BOOT_AUTOCONF | UPF_SKIP_TEST,
 	.iotype		= UPIO_MEM,
 	.regshift	= 2,
 	.uartclk	= IXP4XX_UART_XTAL,
diff --git a/arch/arm/mach-ixp4xx/ixdp425-setup.c b/arch/arm/mach-ixp4xx/ixdp425-setup.c
index c2ba759e994..4633470a6a3 100644
--- a/arch/arm/mach-ixp4xx/ixdp425-setup.c
+++ b/arch/arm/mach-ixp4xx/ixdp425-setup.c
@@ -82,7 +82,7 @@ static struct plat_serial8250_port ixdp425_uart_data[] = {
 		.mapbase	= IXP4XX_UART1_BASE_PHYS,
 		.membase	= (char *)IXP4XX_UART1_BASE_VIRT + REG_OFFSET,
 		.irq		= IRQ_IXP4XX_UART1,
-		.flags		= UPF_BOOT_AUTOCONF,
+		.flags		= UPF_BOOT_AUTOCONF | UPF_SKIP_TEST,
 		.iotype		= UPIO_MEM,
 		.regshift	= 2,
 		.uartclk	= IXP4XX_UART_XTAL,
@@ -91,7 +91,7 @@ static struct plat_serial8250_port ixdp425_uart_data[] = {
 		.mapbase	= IXP4XX_UART2_BASE_PHYS,
 		.membase	= (char *)IXP4XX_UART2_BASE_VIRT + REG_OFFSET,
 		.irq		= IRQ_IXP4XX_UART1,
-		.flags		= UPF_BOOT_AUTOCONF,
+		.flags		= UPF_BOOT_AUTOCONF | UPF_SKIP_TEST,
 		.iotype		= UPIO_MEM,
 		.regshift	= 2,
 		.uartclk	= IXP4XX_UART_XTAL,
diff --git a/arch/arm/mach-s3c2410/mach-bast.c b/arch/arm/mach-s3c2410/mach-bast.c
index 1e7f343822d..e9182242da9 100644
--- a/arch/arm/mach-s3c2410/mach-bast.c
+++ b/arch/arm/mach-s3c2410/mach-bast.c
@@ -30,6 +30,7 @@
  *     28-Jun-2005 BJD  Moved pm functionality out to common code
  *     17-Jul-2005 BJD  Changed to platform device for SuperIO 16550s
  *     25-Jul-2005 BJD  Removed ASIX static mappings
+ *     27-Jul-2005 BJD  Ensure maximum frequency of i2c bus
 */
 
 #include <linux/kernel.h>
@@ -60,6 +61,7 @@
 #include <asm/arch/regs-mem.h>
 #include <asm/arch/regs-lcd.h>
 #include <asm/arch/nand.h>
+#include <asm/arch/iic.h>
 
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/nand.h>
@@ -304,7 +306,7 @@ static void bast_nand_select(struct s3c2410_nand_set *set, int slot)
 }
 
 static struct s3c2410_platform_nand bast_nand_info = {
-	.tacls		= 80,
+	.tacls		= 40,
 	.twrph0		= 80,
 	.twrph1		= 80,
 	.nr_sets	= ARRAY_SIZE(bast_nand_sets),
@@ -385,6 +387,17 @@ static struct platform_device bast_sio = {
 	},
 };
 
+/* we have devices on the bus which cannot work much over the
+ * standard 100KHz i2c bus frequency
+*/
+
+static struct s3c2410_platform_i2c bast_i2c_info = {
+	.flags		= 0,
+	.slave_addr	= 0x10,
+	.bus_freq	= 100*1000,
+	.max_freq	= 130*1000,
+};
+
 /* Standard BAST devices */
 
 static struct platform_device *bast_devices[] __initdata = {
@@ -431,6 +444,7 @@ void __init bast_map_io(void)
 	s3c24xx_uclk.parent  = &s3c24xx_clkout1;
 
 	s3c_device_nand.dev.platform_data = &bast_nand_info;
+	s3c_device_i2c.dev.platform_data = &bast_i2c_info;
 
 	s3c24xx_init_io(bast_iodesc, ARRAY_SIZE(bast_iodesc));
 	s3c24xx_init_clocks(0);
diff --git a/arch/arm/mm/proc-xscale.S b/arch/arm/mm/proc-xscale.S
index 2d977b4eeea..b88de270014 100644
--- a/arch/arm/mm/proc-xscale.S
+++ b/arch/arm/mm/proc-xscale.S
@@ -370,142 +370,6 @@ ENTRY(cpu_xscale_dcache_clean_area)
 	bhi	1b
 	mov	pc, lr
 
-/* ================================ CACHE LOCKING============================
- *
- * The XScale MicroArchitecture implements support for locking entries into
- * the data and instruction cache.  The following functions implement the core
- * low level instructions needed to accomplish the locking.  The developer's
- * manual states that the code that performs the locking must be in non-cached
- * memory.  To accomplish this, the code in xscale-cache-lock.c copies the
- * following functions from the cache into a non-cached memory region that
- * is allocated through consistent_alloc().
- *
- */
-	.align	5
-/*
- * xscale_icache_lock
- *
- * r0: starting address to lock
- * r1: end address to lock
- */
-ENTRY(xscale_icache_lock)
-
-iLockLoop:
-	bic	r0, r0, #CACHELINESIZE - 1
-	mcr	p15, 0, r0, c9, c1, 0	@ lock into cache
-	cmp	r0, r1			@ are we done?
-	add	r0, r0, #CACHELINESIZE	@ advance to next cache line
-	bls	iLockLoop
-	mov	pc, lr
-
-/*
- * xscale_icache_unlock
- */
-ENTRY(xscale_icache_unlock)
-	mcr	p15, 0, r0, c9, c1, 1	@ Unlock icache
-	mov	pc, lr
-
-/*
- * xscale_dcache_lock
- *
- * r0: starting address to lock
- * r1: end address to lock
- */
-ENTRY(xscale_dcache_lock)
-	mcr	p15, 0, ip, c7, c10, 4		@ Drain Write (& Fill) Buffer
-	mov	r2, #1
-	mcr	p15, 0, r2, c9, c2, 0	@ Put dcache in lock mode
-	cpwait	ip			@ Wait for completion
-
-	mrs	r2, cpsr
-	orr	r3, r2, #PSR_F_BIT | PSR_I_BIT
-dLockLoop:
-	msr	cpsr_c, r3
-	mcr	p15, 0, r0, c7, c10, 1	@ Write back line if it is dirty
-	mcr	p15, 0, r0, c7, c6, 1	@ Flush/invalidate line
-	msr	cpsr_c, r2
-	ldr	ip, [r0], #CACHELINESIZE @ Preload 32 bytes into cache from
-					@ location [r0]. Post-increment
-					@ r3 to next cache line
-	cmp	r0, r1			@ Are we done?
-	bls	dLockLoop
-
-	mcr	p15, 0, ip, c7, c10, 4		@ Drain Write (& Fill) Buffer
-	mov	r2, #0
-	mcr	p15, 0, r2, c9, c2, 0	@ Get out of lock mode
-	cpwait_ret lr, ip
-
-/*
- * xscale_dcache_unlock
- */
-ENTRY(xscale_dcache_unlock)
-	mcr	p15, 0, ip, c7, c10, 4		@ Drain Write (& Fill) Buffer
-	mcr	p15, 0, ip, c9, c2, 1	@ Unlock cache
-	mov	pc, lr
-
-/*
- * Needed to determine the length of the code that needs to be copied.
- */
-	.align	5
-ENTRY(xscale_cache_dummy)
-	mov	pc, lr
-
-/* ================================ TLB LOCKING==============================
- *
- * The XScale MicroArchitecture implements support for locking entries into
- * the Instruction and Data TLBs.  The following functions provide the
- * low level support for supporting these under Linux.  xscale-lock.c
- * implements some higher level management code.  Most of the following
- * is taken straight out of the Developer's Manual.
- */
-
-/*
- * Lock I-TLB entry
- *
- * r0: Virtual address to translate and lock
- */
-	.align	5
-ENTRY(xscale_itlb_lock)
-	mrs	r2, cpsr
-	orr	r3, r2, #PSR_F_BIT | PSR_I_BIT
-	msr	cpsr_c, r3			@ Disable interrupts
-	mcr	p15, 0, r0, c8, c5, 1		@ Invalidate I-TLB entry
-	mcr	p15, 0, r0, c10, c4, 0		@ Translate and lock
-	msr	cpsr_c, r2			@ Restore interrupts
-	cpwait_ret lr, ip
-
-/*
- * Lock D-TLB entry
- *
- * r0: Virtual address to translate and lock
- */
-	.align	5
-ENTRY(xscale_dtlb_lock)
-	mrs	r2, cpsr
-	orr	r3, r2, #PSR_F_BIT | PSR_I_BIT
-	msr	cpsr_c, r3			@ Disable interrupts
-	mcr	p15, 0, r0, c8, c6, 1		@ Invalidate D-TLB entry
-	mcr	p15, 0, r0, c10, c8, 0		@ Translate and lock
-	msr	cpsr_c, r2			@ Restore interrupts
-	cpwait_ret lr, ip
-
-/*
- * Unlock all I-TLB entries
- */
-	.align	5
-ENTRY(xscale_itlb_unlock)
-	mcr	p15, 0, ip, c10, c4, 1		@ Unlock I-TLB
-	mcr	p15, 0, ip, c8, c5, 0		@ Invalidate I-TLB
-	cpwait_ret lr, ip
-
-/*
- * Unlock all D-TLB entries
- */
-ENTRY(xscale_dtlb_unlock)
-	mcr	p15, 0, ip, c10, c8, 1		@ Unlock D-TBL
-	mcr	p15, 0, ip, c8, c6, 0		@ Invalidate D-TLB
-	cpwait_ret lr, ip
-
 /* =============================== PageTable ============================== */
 
 #define PTE_CACHE_WRITE_ALLOCATE 0
diff --git a/arch/arm/nwfpe/double_cpdo.c b/arch/arm/nwfpe/double_cpdo.c
index 7ffd8cb9bc9..c51d1386a97 100644
--- a/arch/arm/nwfpe/double_cpdo.c
+++ b/arch/arm/nwfpe/double_cpdo.c
@@ -40,17 +40,17 @@ float64 float64_arccos(float64 rFm);
 float64 float64_pow(float64 rFn, float64 rFm);
 float64 float64_pol(float64 rFn, float64 rFm);
 
-static float64 float64_rsf(float64 rFn, float64 rFm)
+static float64 float64_rsf(struct roundingData *roundData, float64 rFn, float64 rFm)
 {
-	return float64_sub(rFm, rFn);
+	return float64_sub(roundData, rFm, rFn);
 }
 
-static float64 float64_rdv(float64 rFn, float64 rFm)
+static float64 float64_rdv(struct roundingData *roundData, float64 rFn, float64 rFm)
 {
-	return float64_div(rFm, rFn);
+	return float64_div(roundData, rFm, rFn);
 }
 
-static float64 (*const dyadic_double[16])(float64 rFn, float64 rFm) = {
+static float64 (*const dyadic_double[16])(struct roundingData*, float64 rFn, float64 rFm) = {
 	[ADF_CODE >> 20] = float64_add,
 	[MUF_CODE >> 20] = float64_mul,
 	[SUF_CODE >> 20] = float64_sub,
@@ -65,12 +65,12 @@ static float64 (*const dyadic_double[16])(float64 rFn, float64 rFm) = {
 	[FRD_CODE >> 20] = float64_rdv,
 };
 
-static float64 float64_mvf(float64 rFm)
+static float64 float64_mvf(struct roundingData *roundData,float64 rFm)
 {
 	return rFm;
 }
 
-static float64 float64_mnf(float64 rFm)
+static float64 float64_mnf(struct roundingData *roundData,float64 rFm)
 {
 	union float64_components u;
 
@@ -84,7 +84,7 @@ static float64 float64_mnf(float64 rFm)
 	return u.f64;
 }
 
-static float64 float64_abs(float64 rFm)
+static float64 float64_abs(struct roundingData *roundData,float64 rFm)
 {
 	union float64_components u;
 
@@ -98,7 +98,7 @@ static float64 float64_abs(float64 rFm)
 	return u.f64;
 }
 
-static float64 (*const monadic_double[16])(float64 rFm) = {
+static float64 (*const monadic_double[16])(struct roundingData *, float64 rFm) = {
 	[MVF_CODE >> 20] = float64_mvf,
 	[MNF_CODE >> 20] = float64_mnf,
 	[ABS_CODE >> 20] = float64_abs,
@@ -108,7 +108,7 @@ static float64 (*const monadic_double[16])(float64 rFm) = {
 	[NRM_CODE >> 20] = float64_mvf,
 };
 
-unsigned int DoubleCPDO(const unsigned int opcode, FPREG * rFd)
+unsigned int DoubleCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	float64 rFm;
@@ -151,13 +151,13 @@ unsigned int DoubleCPDO(const unsigned int opcode, FPREG * rFd)
 		}
 
 		if (dyadic_double[opc_mask_shift]) {
-			rFd->fDouble = dyadic_double[opc_mask_shift](rFn, rFm);
+			rFd->fDouble = dyadic_double[opc_mask_shift](roundData, rFn, rFm);
 		} else {
 			return 0;
 		}
 	} else {
 		if (monadic_double[opc_mask_shift]) {
-			rFd->fDouble = monadic_double[opc_mask_shift](rFm);
+			rFd->fDouble = monadic_double[opc_mask_shift](roundData, rFm);
 		} else {
 			return 0;
 		}
diff --git a/arch/arm/nwfpe/extended_cpdo.c b/arch/arm/nwfpe/extended_cpdo.c
index c39f68a3449..65a279ba927 100644
--- a/arch/arm/nwfpe/extended_cpdo.c
+++ b/arch/arm/nwfpe/extended_cpdo.c
@@ -35,17 +35,17 @@ floatx80 floatx80_arccos(floatx80 rFm);
 floatx80 floatx80_pow(floatx80 rFn, floatx80 rFm);
 floatx80 floatx80_pol(floatx80 rFn, floatx80 rFm);
 
-static floatx80 floatx80_rsf(floatx80 rFn, floatx80 rFm)
+static floatx80 floatx80_rsf(struct roundingData *roundData, floatx80 rFn, floatx80 rFm)
 {
-	return floatx80_sub(rFm, rFn);
+	return floatx80_sub(roundData, rFm, rFn);
 }
 
-static floatx80 floatx80_rdv(floatx80 rFn, floatx80 rFm)
+static floatx80 floatx80_rdv(struct roundingData *roundData, floatx80 rFn, floatx80 rFm)
 {
-	return floatx80_div(rFm, rFn);
+	return floatx80_div(roundData, rFm, rFn);
 }
 
-static floatx80 (*const dyadic_extended[16])(floatx80 rFn, floatx80 rFm) = {
+static floatx80 (*const dyadic_extended[16])(struct roundingData*, floatx80 rFn, floatx80 rFm) = {
 	[ADF_CODE >> 20] = floatx80_add,
 	[MUF_CODE >> 20] = floatx80_mul,
 	[SUF_CODE >> 20] = floatx80_sub,
@@ -60,24 +60,24 @@ static floatx80 (*const dyadic_extended[16])(floatx80 rFn, floatx80 rFm) = {
 	[FRD_CODE >> 20] = floatx80_rdv,
 };
 
-static floatx80 floatx80_mvf(floatx80 rFm)
+static floatx80 floatx80_mvf(struct roundingData *roundData, floatx80 rFm)
 {
 	return rFm;
 }
 
-static floatx80 floatx80_mnf(floatx80 rFm)
+static floatx80 floatx80_mnf(struct roundingData *roundData, floatx80 rFm)
 {
 	rFm.high ^= 0x8000;
 	return rFm;
 }
 
-static floatx80 floatx80_abs(floatx80 rFm)
+static floatx80 floatx80_abs(struct roundingData *roundData, floatx80 rFm)
 {
 	rFm.high &= 0x7fff;
 	return rFm;
 }
 
-static floatx80 (*const monadic_extended[16])(floatx80 rFm) = {
+static floatx80 (*const monadic_extended[16])(struct roundingData*, floatx80 rFm) = {
 	[MVF_CODE >> 20] = floatx80_mvf,
 	[MNF_CODE >> 20] = floatx80_mnf,
 	[ABS_CODE >> 20] = floatx80_abs,
@@ -87,7 +87,7 @@ static floatx80 (*const monadic_extended[16])(floatx80 rFm) = {
 	[NRM_CODE >> 20] = floatx80_mvf,
 };
 
-unsigned int ExtendedCPDO(const unsigned int opcode, FPREG * rFd)
+unsigned int ExtendedCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	floatx80 rFm;
@@ -138,13 +138,13 @@ unsigned int ExtendedCPDO(const unsigned int opcode, FPREG * rFd)
 		}
 
 		if (dyadic_extended[opc_mask_shift]) {
-			rFd->fExtended = dyadic_extended[opc_mask_shift](rFn, rFm);
+			rFd->fExtended = dyadic_extended[opc_mask_shift](roundData, rFn, rFm);
 		} else {
 			return 0;
 		}
 	} else {
 		if (monadic_extended[opc_mask_shift]) {
-			rFd->fExtended = monadic_extended[opc_mask_shift](rFm);
+			rFd->fExtended = monadic_extended[opc_mask_shift](roundData, rFm);
 		} else {
 			return 0;
 		}
diff --git a/arch/arm/nwfpe/fpa11.c b/arch/arm/nwfpe/fpa11.c
index bf61696865e..7690f731ee8 100644
--- a/arch/arm/nwfpe/fpa11.c
+++ b/arch/arm/nwfpe/fpa11.c
@@ -51,48 +51,42 @@ static void resetFPA11(void)
 	fpa11->fpsr = FP_EMULATOR | BIT_AC;
 }
 
-void SetRoundingMode(const unsigned int opcode)
+int8 SetRoundingMode(const unsigned int opcode)
 {
 	switch (opcode & MASK_ROUNDING_MODE) {
 	default:
 	case ROUND_TO_NEAREST:
-		float_rounding_mode = float_round_nearest_even;
-		break;
+		return float_round_nearest_even;
 
 	case ROUND_TO_PLUS_INFINITY:
-		float_rounding_mode = float_round_up;
-		break;
+		return float_round_up;
 
 	case ROUND_TO_MINUS_INFINITY:
-		float_rounding_mode = float_round_down;
-		break;
+		return float_round_down;
 
 	case ROUND_TO_ZERO:
-		float_rounding_mode = float_round_to_zero;
-		break;
+		return float_round_to_zero;
 	}
 }
 
-void SetRoundingPrecision(const unsigned int opcode)
+int8 SetRoundingPrecision(const unsigned int opcode)
 {
 #ifdef CONFIG_FPE_NWFPE_XP
 	switch (opcode & MASK_ROUNDING_PRECISION) {
 	case ROUND_SINGLE:
-		floatx80_rounding_precision = 32;
-		break;
+		return 32;
 
 	case ROUND_DOUBLE:
-		floatx80_rounding_precision = 64;
-		break;
+		return 64;
 
 	case ROUND_EXTENDED:
-		floatx80_rounding_precision = 80;
-		break;
+		return 80;
 
 	default:
-		floatx80_rounding_precision = 80;
+		return 80;
 	}
 #endif
+	return 80;
 }
 
 void nwfpe_init_fpa(union fp_state *fp)
@@ -103,8 +97,6 @@ void nwfpe_init_fpa(union fp_state *fp)
 #endif
  	memset(fpa11, 0, sizeof(FPA11));
 	resetFPA11();
-	SetRoundingMode(ROUND_TO_NEAREST);
-	SetRoundingPrecision(ROUND_EXTENDED);
 	fpa11->initflag = 1;
 }
 
diff --git a/arch/arm/nwfpe/fpa11.h b/arch/arm/nwfpe/fpa11.h
index e4a61aea534..93523ae4b7a 100644
--- a/arch/arm/nwfpe/fpa11.h
+++ b/arch/arm/nwfpe/fpa11.h
@@ -37,6 +37,13 @@
 /* includes */
 #include "fpsr.h"		/* FP control and status register definitions */
 #include "milieu.h"
+
+struct roundingData {
+    int8 mode;
+    int8 precision;
+    signed char exception;
+};
+
 #include "softfloat.h"
 
 #define		typeNone		0x00
@@ -84,8 +91,8 @@ typedef struct tagFPA11 {
 				   initialised. */
 } FPA11;
 
-extern void SetRoundingMode(const unsigned int);
-extern void SetRoundingPrecision(const unsigned int);
+extern int8 SetRoundingMode(const unsigned int);
+extern int8 SetRoundingPrecision(const unsigned int);
 extern void nwfpe_init_fpa(union fp_state *fp);
 
 #endif
diff --git a/arch/arm/nwfpe/fpa11_cpdo.c b/arch/arm/nwfpe/fpa11_cpdo.c
index 1bea67437b6..4a31dfd9406 100644
--- a/arch/arm/nwfpe/fpa11_cpdo.c
+++ b/arch/arm/nwfpe/fpa11_cpdo.c
@@ -24,15 +24,16 @@
 #include "fpa11.h"
 #include "fpopcode.h"
 
-unsigned int SingleCPDO(const unsigned int opcode, FPREG * rFd);
-unsigned int DoubleCPDO(const unsigned int opcode, FPREG * rFd);
-unsigned int ExtendedCPDO(const unsigned int opcode, FPREG * rFd);
+unsigned int SingleCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd);
+unsigned int DoubleCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd);
+unsigned int ExtendedCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd);
 
 unsigned int EmulateCPDO(const unsigned int opcode)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	FPREG *rFd;
 	unsigned int nType, nDest, nRc;
+	struct roundingData roundData;
 
 	/* Get the destination size.  If not valid let Linux perform
 	   an invalid instruction trap. */
@@ -40,7 +41,9 @@ unsigned int EmulateCPDO(const unsigned int opcode)
 	if (typeNone == nDest)
 		return 0;
 
-	SetRoundingMode(opcode);
+	roundData.mode = SetRoundingMode(opcode);
+	roundData.precision = SetRoundingPrecision(opcode);
+	roundData.exception = 0;
 
 	/* Compare the size of the operands in Fn and Fm.
 	   Choose the largest size and perform operations in that size,
@@ -63,14 +66,14 @@ unsigned int EmulateCPDO(const unsigned int opcode)
 
 	switch (nType) {
 	case typeSingle:
-		nRc = SingleCPDO(opcode, rFd);
+		nRc = SingleCPDO(&roundData, opcode, rFd);
 		break;
 	case typeDouble:
-		nRc = DoubleCPDO(opcode, rFd);
+		nRc = DoubleCPDO(&roundData, opcode, rFd);
 		break;
 #ifdef CONFIG_FPE_NWFPE_XP
 	case typeExtended:
-		nRc = ExtendedCPDO(opcode, rFd);
+		nRc = ExtendedCPDO(&roundData, opcode, rFd);
 		break;
 #endif
 	default:
@@ -93,9 +96,9 @@ unsigned int EmulateCPDO(const unsigned int opcode)
 			case typeSingle:
 				{
 					if (typeDouble == nType)
-						rFd->fSingle = float64_to_float32(rFd->fDouble);
+						rFd->fSingle = float64_to_float32(&roundData, rFd->fDouble);
 					else
-						rFd->fSingle = floatx80_to_float32(rFd->fExtended);
+						rFd->fSingle = floatx80_to_float32(&roundData, rFd->fExtended);
 				}
 				break;
 
@@ -104,7 +107,7 @@ unsigned int EmulateCPDO(const unsigned int opcode)
 					if (typeSingle == nType)
 						rFd->fDouble = float32_to_float64(rFd->fSingle);
 					else
-						rFd->fDouble = floatx80_to_float64(rFd->fExtended);
+						rFd->fDouble = floatx80_to_float64(&roundData, rFd->fExtended);
 				}
 				break;
 
@@ -121,12 +124,15 @@ unsigned int EmulateCPDO(const unsigned int opcode)
 #else
 		if (nDest != nType) {
 			if (nDest == typeSingle)
-				rFd->fSingle = float64_to_float32(rFd->fDouble);
+				rFd->fSingle = float64_to_float32(&roundData, rFd->fDouble);
 			else
 				rFd->fDouble = float32_to_float64(rFd->fSingle);
 		}
 #endif
 	}
 
+	if (roundData.exception)
+		float_raise(roundData.exception);
+
 	return nRc;
 }
diff --git a/arch/arm/nwfpe/fpa11_cpdt.c b/arch/arm/nwfpe/fpa11_cpdt.c
index 95fb63fa9d1..b0db5cbcc3b 100644
--- a/arch/arm/nwfpe/fpa11_cpdt.c
+++ b/arch/arm/nwfpe/fpa11_cpdt.c
@@ -96,7 +96,7 @@ static inline void loadMultiple(const unsigned int Fn, const unsigned int __user
 	}
 }
 
-static inline void storeSingle(const unsigned int Fn, unsigned int __user *pMem)
+static inline void storeSingle(struct roundingData *roundData, const unsigned int Fn, unsigned int __user *pMem)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	union {
@@ -106,12 +106,12 @@ static inline void storeSingle(const unsigned int Fn, unsigned int __user *pMem)
 
 	switch (fpa11->fType[Fn]) {
 	case typeDouble:
-		val.f = float64_to_float32(fpa11->fpreg[Fn].fDouble);
+		val.f = float64_to_float32(roundData, fpa11->fpreg[Fn].fDouble);
 		break;
 
 #ifdef CONFIG_FPE_NWFPE_XP
 	case typeExtended:
-		val.f = floatx80_to_float32(fpa11->fpreg[Fn].fExtended);
+		val.f = floatx80_to_float32(roundData, fpa11->fpreg[Fn].fExtended);
 		break;
 #endif
 
@@ -122,7 +122,7 @@ static inline void storeSingle(const unsigned int Fn, unsigned int __user *pMem)
 	put_user(val.i[0], pMem);
 }
 
-static inline void storeDouble(const unsigned int Fn, unsigned int __user *pMem)
+static inline void storeDouble(struct roundingData *roundData, const unsigned int Fn, unsigned int __user *pMem)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	union {
@@ -137,7 +137,7 @@ static inline void storeDouble(const unsigned int Fn, unsigned int __user *pMem)
 
 #ifdef CONFIG_FPE_NWFPE_XP
 	case typeExtended:
-		val.f = floatx80_to_float64(fpa11->fpreg[Fn].fExtended);
+		val.f = floatx80_to_float64(roundData, fpa11->fpreg[Fn].fExtended);
 		break;
 #endif
 
@@ -259,8 +259,11 @@ unsigned int PerformSTF(const unsigned int opcode)
 {
 	unsigned int __user *pBase, *pAddress, *pFinal;
 	unsigned int nRc = 1, write_back = WRITE_BACK(opcode);
+	struct roundingData roundData;
 
-	SetRoundingMode(ROUND_TO_NEAREST);
+	roundData.mode = SetRoundingMode(opcode);
+	roundData.precision = SetRoundingPrecision(opcode);
+	roundData.exception = 0;
 
 	pBase = (unsigned int __user *) readRegister(getRn(opcode));
 	if (REG_PC == getRn(opcode)) {
@@ -281,10 +284,10 @@ unsigned int PerformSTF(const unsigned int opcode)
 
 	switch (opcode & MASK_TRANSFER_LENGTH) {
 	case TRANSFER_SINGLE:
-		storeSingle(getFd(opcode), pAddress);
+		storeSingle(&roundData, getFd(opcode), pAddress);
 		break;
 	case TRANSFER_DOUBLE:
-		storeDouble(getFd(opcode), pAddress);
+		storeDouble(&roundData, getFd(opcode), pAddress);
 		break;
 #ifdef CONFIG_FPE_NWFPE_XP
 	case TRANSFER_EXTENDED:
@@ -295,6 +298,9 @@ unsigned int PerformSTF(const unsigned int opcode)
 		nRc = 0;
 	}
 
+	if (roundData.exception)
+		float_raise(roundData.exception);
+
 	if (write_back)
 		writeRegister(getRn(opcode), (unsigned long) pFinal);
 	return nRc;
diff --git a/arch/arm/nwfpe/fpa11_cprt.c b/arch/arm/nwfpe/fpa11_cprt.c
index db01fbc9721..adf8d300054 100644
--- a/arch/arm/nwfpe/fpa11_cprt.c
+++ b/arch/arm/nwfpe/fpa11_cprt.c
@@ -33,8 +33,6 @@ extern flag floatx80_is_nan(floatx80);
 extern flag float64_is_nan(float64);
 extern flag float32_is_nan(float32);
 
-void SetRoundingMode(const unsigned int opcode);
-
 unsigned int PerformFLT(const unsigned int opcode);
 unsigned int PerformFIX(const unsigned int opcode);
 
@@ -77,14 +75,17 @@ unsigned int EmulateCPRT(const unsigned int opcode)
 unsigned int PerformFLT(const unsigned int opcode)
 {
 	FPA11 *fpa11 = GET_FPA11();
-	SetRoundingMode(opcode);
-	SetRoundingPrecision(opcode);
+	struct roundingData roundData;
+
+	roundData.mode = SetRoundingMode(opcode);
+	roundData.precision = SetRoundingPrecision(opcode);
+	roundData.exception = 0;
 
 	switch (opcode & MASK_ROUNDING_PRECISION) {
 	case ROUND_SINGLE:
 		{
 			fpa11->fType[getFn(opcode)] = typeSingle;
-			fpa11->fpreg[getFn(opcode)].fSingle = int32_to_float32(readRegister(getRd(opcode)));
+			fpa11->fpreg[getFn(opcode)].fSingle = int32_to_float32(&roundData, readRegister(getRd(opcode)));
 		}
 		break;
 
@@ -108,6 +109,9 @@ unsigned int PerformFLT(const unsigned int opcode)
 		return 0;
 	}
 
+	if (roundData.exception)
+		float_raise(roundData.exception);
+
 	return 1;
 }
 
@@ -115,26 +119,29 @@ unsigned int PerformFIX(const unsigned int opcode)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	unsigned int Fn = getFm(opcode);
+	struct roundingData roundData;
 
-	SetRoundingMode(opcode);
+	roundData.mode = SetRoundingMode(opcode);
+	roundData.precision = SetRoundingPrecision(opcode);
+	roundData.exception = 0;
 
 	switch (fpa11->fType[Fn]) {
 	case typeSingle:
 		{
-			writeRegister(getRd(opcode), float32_to_int32(fpa11->fpreg[Fn].fSingle));
+			writeRegister(getRd(opcode), float32_to_int32(&roundData, fpa11->fpreg[Fn].fSingle));
 		}
 		break;
 
 	case typeDouble:
 		{
-			writeRegister(getRd(opcode), float64_to_int32(fpa11->fpreg[Fn].fDouble));
+			writeRegister(getRd(opcode), float64_to_int32(&roundData, fpa11->fpreg[Fn].fDouble));
 		}
 		break;
 
 #ifdef CONFIG_FPE_NWFPE_XP
 	case typeExtended:
 		{
-			writeRegister(getRd(opcode), floatx80_to_int32(fpa11->fpreg[Fn].fExtended));
+			writeRegister(getRd(opcode), floatx80_to_int32(&roundData, fpa11->fpreg[Fn].fExtended));
 		}
 		break;
 #endif
@@ -143,6 +150,9 @@ unsigned int PerformFIX(const unsigned int opcode)
 		return 0;
 	}
 
+	if (roundData.exception)
+		float_raise(roundData.exception);
+
 	return 1;
 }
 
diff --git a/arch/arm/nwfpe/fpmodule.c b/arch/arm/nwfpe/fpmodule.c
index 12885f31d34..2dfe1ac42ee 100644
--- a/arch/arm/nwfpe/fpmodule.c
+++ b/arch/arm/nwfpe/fpmodule.c
@@ -116,8 +116,6 @@ fpmodule.c to integrate with the NetBSD kernel (I hope!).
 code to access data in user space in some other source files at the 
 moment (grep for get_user / put_user calls).  --philb]
 
-float_exception_flags is a global variable in SoftFloat.
-
 This function is called by the SoftFloat routines to raise a floating
 point exception.  We check the trap enable byte in the FPSR, and raise
 a SIGFPE exception if necessary.  If not the relevant bits in the 
@@ -129,15 +127,14 @@ void float_raise(signed char flags)
 	register unsigned int fpsr, cumulativeTraps;
 
 #ifdef CONFIG_DEBUG_USER
-	printk(KERN_DEBUG
-	       "NWFPE: %s[%d] takes exception %08x at %p from %08lx\n",
-	       current->comm, current->pid, flags,
-	       __builtin_return_address(0), GET_USERREG()->ARM_pc);
+ 	/* Ignore inexact errors as there are far too many of them to log */
+ 	if (flags & ~BIT_IXC)
+ 		printk(KERN_DEBUG
+		       "NWFPE: %s[%d] takes exception %08x at %p from %08lx\n",
+		       current->comm, current->pid, flags,
+		       __builtin_return_address(0), GET_USERREG()->ARM_pc);
 #endif
 
-	/* Keep SoftFloat exception flags up to date.  */
-	float_exception_flags |= flags;
-
 	/* Read fpsr and initialize the cumulativeTraps.  */
 	fpsr = readFPSR();
 	cumulativeTraps = 0;
diff --git a/arch/arm/nwfpe/single_cpdo.c b/arch/arm/nwfpe/single_cpdo.c
index 705808e88d9..c66981d682c 100644
--- a/arch/arm/nwfpe/single_cpdo.c
+++ b/arch/arm/nwfpe/single_cpdo.c
@@ -36,17 +36,17 @@ float32 float32_arccos(float32 rFm);
 float32 float32_pow(float32 rFn, float32 rFm);
 float32 float32_pol(float32 rFn, float32 rFm);
 
-static float32 float32_rsf(float32 rFn, float32 rFm)
+static float32 float32_rsf(struct roundingData *roundData, float32 rFn, float32 rFm)
 {
-	return float32_sub(rFm, rFn);
+	return float32_sub(roundData, rFm, rFn);
 }
 
-static float32 float32_rdv(float32 rFn, float32 rFm)
+static float32 float32_rdv(struct roundingData *roundData, float32 rFn, float32 rFm)
 {
-	return float32_div(rFm, rFn);
+	return float32_div(roundData, rFm, rFn);
 }
 
-static float32 (*const dyadic_single[16])(float32 rFn, float32 rFm) = {
+static float32 (*const dyadic_single[16])(struct roundingData *, float32 rFn, float32 rFm) = {
 	[ADF_CODE >> 20] = float32_add,
 	[MUF_CODE >> 20] = float32_mul,
 	[SUF_CODE >> 20] = float32_sub,
@@ -60,22 +60,22 @@ static float32 (*const dyadic_single[16])(float32 rFn, float32 rFm) = {
 	[FRD_CODE >> 20] = float32_rdv,
 };
 
-static float32 float32_mvf(float32 rFm)
+static float32 float32_mvf(struct roundingData *roundData, float32 rFm)
 {
 	return rFm;
 }
 
-static float32 float32_mnf(float32 rFm)
+static float32 float32_mnf(struct roundingData *roundData, float32 rFm)
 {
 	return rFm ^ 0x80000000;
 }
 
-static float32 float32_abs(float32 rFm)
+static float32 float32_abs(struct roundingData *roundData, float32 rFm)
 {
 	return rFm & 0x7fffffff;
 }
 
-static float32 (*const monadic_single[16])(float32 rFm) = {
+static float32 (*const monadic_single[16])(struct roundingData*, float32 rFm) = {
 	[MVF_CODE >> 20] = float32_mvf,
 	[MNF_CODE >> 20] = float32_mnf,
 	[ABS_CODE >> 20] = float32_abs,
@@ -85,7 +85,7 @@ static float32 (*const monadic_single[16])(float32 rFm) = {
 	[NRM_CODE >> 20] = float32_mvf,
 };
 
-unsigned int SingleCPDO(const unsigned int opcode, FPREG * rFd)
+unsigned int SingleCPDO(struct roundingData *roundData, const unsigned int opcode, FPREG * rFd)
 {
 	FPA11 *fpa11 = GET_FPA11();
 	float32 rFm;
@@ -108,13 +108,13 @@ unsigned int SingleCPDO(const unsigned int opcode, FPREG * rFd)
 		if (fpa11->fType[Fn] == typeSingle &&
 		    dyadic_single[opc_mask_shift]) {
 			rFn = fpa11->fpreg[Fn].fSingle;
-			rFd->fSingle = dyadic_single[opc_mask_shift](rFn, rFm);
+			rFd->fSingle = dyadic_single[opc_mask_shift](roundDat