patch-2.4.20 linux-2.4.20/arch/mips/math-emu/cp1emu.c

Next file: linux-2.4.20/arch/mips/math-emu/dp_add.c
Previous file: linux-2.4.20/arch/mips/math-emu/Makefile
Back to the patch index
Back to the overall index

diff -urN linux-2.4.19/arch/mips/math-emu/cp1emu.c linux-2.4.20/arch/mips/math-emu/cp1emu.c
@@ -1,6 +1,6 @@
 /*
  * cp1emu.c: a MIPS coprocessor 1 (fpu) instruction emulator
- * 
+ *
  * MIPS floating point support
  * Copyright (C) 1994-2000 Algorithmics Ltd.  All rights reserved.
  * http://www.algor.co.uk
@@ -23,38 +23,30 @@
  *
  * A complete emulator for MIPS coprocessor 1 instructions.  This is
  * required for #float(switch) or #float(trap), where it catches all
- * COP1 instructions via the "CoProcessor Unusable" exception.  
+ * COP1 instructions via the "CoProcessor Unusable" exception.
  *
  * More surprisingly it is also required for #float(ieee), to help out
  * the hardware fpu at the boundaries of the IEEE-754 representation
  * (denormalised values, infinities, underflow, etc).  It is made
  * quite nasty because emulation of some non-COP1 instructions is
  * required, e.g. in branch delay slots.
- * 
- * Note if you know that you won't have an fpu, then you'll get much 
+ *
+ * Note if you know that you won't have an fpu, then you'll get much
  * better performance by compiling with -msoft-float!
  */
-#include <linux/compiler.h>
-#include <linux/mm.h>
-#include <linux/signal.h>
-#include <linux/smp.h>
-#include <linux/smp_lock.h>
-
-#include <asm/asm.h>
-#include <asm/branch.h>
+#include <asm/inst.h>
 #include <asm/bootinfo.h>
-#include <asm/byteorder.h>
 #include <asm/cpu.h>
-#include <asm/inst.h>
-#include <asm/uaccess.h>
 #include <asm/processor.h>
+#include <asm/ptrace.h>
+#include <asm/signal.h>
 #include <asm/mipsregs.h>
-#include <asm/system.h>
-#include <asm/pgtable.h>
-
 #include <asm/fpu_emulator.h>
+#include <asm/uaccess.h>
+#include <asm/branch.h>
 
 #include "ieee754.h"
+#include "dsemul.h"
 
 /* Strap kernel emulator for full MIPS IV emulation */
 
@@ -63,20 +55,14 @@
 #endif
 #define __mips 4
 
-typedef void *vaddr_t;
-
-/* Function which emulates the instruction in a branch delay slot. */
-
-static int mips_dsemul(struct pt_regs *, mips_instruction, unsigned long);
-
 /* Function which emulates a floating point instruction. */
 
 static int fpu_emu(struct pt_regs *, struct mips_fpu_soft_struct *,
-	 mips_instruction);
+	mips_instruction);
 
 #if __mips >= 4 && __mips != 32
 static int fpux_emu(struct pt_regs *,
-		    struct mips_fpu_soft_struct *, mips_instruction);
+	struct mips_fpu_soft_struct *, mips_instruction);
 #endif
 
 /* Further private data for which no space exists in mips_fpu_soft_struct */
@@ -108,8 +94,7 @@
 #endif
 
 
-
-/* 
+/*
  * Redundant with logic already in kernel/branch.c,
  * embedded in compute_return_epc.  At some point,
  * a single subroutine should be used across both
@@ -165,46 +150,70 @@
 	return 0;
 }
 
-#define REG_TO_VA (vaddr_t)
-#define VA_TO_REG (unsigned long)
-
 /*
  * In the Linux kernel, we support selection of FPR format on the
  * basis of the Status.FR bit.  This does imply that, if a full 32
  * FPRs are desired, there needs to be a flip-flop that can be written
- * to one at that bit position.  In any case, normal MIPS ABI uses
+ * to one at that bit position.  In any case, O32 MIPS ABI uses
  * only the even FPRs (Status.FR = 0).
  */
 
 #define CP0_STATUS_FR_SUPPORT
 
+#ifdef CP0_STATUS_FR_SUPPORT
+#define FR_BIT ST0_FR
+#else
+#define FR_BIT 0
+#endif
+
+#define SIFROMREG(si,x)	((si) = \
+			(xcp->cp0_status & FR_BIT) || !(x & 1) ? \
+			(int)ctx->regs[x] : \
+			(int)(ctx->regs[x & ~1] >> 32 ))
+#define SITOREG(si,x)	(ctx->regs[x & ~((xcp->cp0_status & FR_BIT) == 0)] = \
+			(xcp->cp0_status & FR_BIT) || !(x & 1) ? \
+			ctx->regs[x & ~1] >> 32 << 32 | (u32)(si) : \
+			ctx->regs[x & ~1] << 32 >> 32 | (u64)(si) << 32)
+
+#define DIFROMREG(di,x)	((di) = \
+			ctx->regs[x & ~((xcp->cp0_status & FR_BIT) == 0)])
+#define DITOREG(di,x)	(ctx->regs[x & ~((xcp->cp0_status & FR_BIT) == 0)] \
+			= (di))
+#define DIFROMREG(di,x)	((di) = \
+			ctx->regs[x & ~((xcp->cp0_status & FR_BIT) == 0)])
+#define DITOREG(di,x)	(ctx->regs[x & ~((xcp->cp0_status & FR_BIT) == 0)] \
+			= (di))
+
+#define SPFROMREG(sp,x)	SIFROMREG((sp).bits,x)
+#define SPTOREG(sp,x)	SITOREG((sp).bits,x)
+#define DPFROMREG(dp,x)	DIFROMREG((dp).bits,x)
+#define DPTOREG(dp,x)	DITOREG((dp).bits,x)
+
 /*
  * Emulate the single floating point instruction pointed at by EPC.
  * Two instructions if the instruction is in a branch delay slot.
  */
 
-static int cop1Emulate(struct pt_regs *regs, struct mips_fpu_soft_struct *ctx)
+static int cop1Emulate(struct pt_regs *xcp, struct mips_fpu_soft_struct *ctx)
 {
-	vaddr_t emulpc, contpc;
 	mips_instruction ir;
+	vaddr_t emulpc, contpc;
 	unsigned int cond;
-	int err = 0;
 
-	err = get_user(ir, (mips_instruction *) regs->cp0_epc);
-	if (err) {
+	if (get_user(ir, (mips_instruction *) REG_TO_VA xcp->cp0_epc)) {
 		fpuemuprivate.stats.errors++;
 		return SIGBUS;
 	}
 
 	/* XXX NEC Vr54xx bug workaround */
-	if ((regs->cp0_cause & CAUSEF_BD) && !isBranchInstr(&ir))
-		regs->cp0_cause &= ~CAUSEF_BD;
+	if ((xcp->cp0_cause & CAUSEF_BD) && !isBranchInstr(&ir))
+		xcp->cp0_cause &= ~CAUSEF_BD;
 
-	if (regs->cp0_cause & CAUSEF_BD) {
+	if (xcp->cp0_cause & CAUSEF_BD) {
 		/*
 		 * The instruction to be emulated is in a branch delay slot
 		 * which means that we have to  emulate the branch instruction
-		 * BEFORE we do the cop1 instruction. 
+		 * BEFORE we do the cop1 instruction.
 		 *
 		 * This branch could be a COP1 branch, but in that case we
 		 * would have had a trap for that instruction, and would not
@@ -213,368 +222,206 @@
 		 * Linux MIPS branch emulator operates on context, updating the
 		 * cp0_epc.
 		 */
-		emulpc = REG_TO_VA(regs->cp0_epc + 4);	/* Snapshot emulation target */
+		emulpc = REG_TO_VA(xcp->cp0_epc + 4);	/* Snapshot emulation target */
 
-		if (__compute_return_epc(regs)) {
+		if (__compute_return_epc(xcp)) {
 #ifdef CP1DBG
 			printk("failed to emulate branch at %p\n",
-				    REG_TO_VA(regs->cp0_epc));
+				REG_TO_VA(xcp->cp0_epc));
 #endif
 			return SIGILL;
 		}
-		err = get_user(ir, (mips_instruction *) emulpc);
-		if (err) {
+		if (get_user(ir, (mips_instruction *) emulpc)) {
 			fpuemuprivate.stats.errors++;
 			return SIGBUS;
 		}
 		/* __computer_return_epc() will have updated cp0_epc */
-		contpc = REG_TO_VA regs->cp0_epc;
+		contpc = REG_TO_VA xcp->cp0_epc;
 		/* In order not to confuse ptrace() et al, tweak context */
-		regs->cp0_epc = VA_TO_REG emulpc - 4;
-	} else {
-		emulpc = REG_TO_VA regs->cp0_epc;
-		contpc = REG_TO_VA regs->cp0_epc + 4;
+		xcp->cp0_epc = VA_TO_REG emulpc - 4;
+	}
+	else {
+		emulpc = REG_TO_VA xcp->cp0_epc;
+		contpc = REG_TO_VA(xcp->cp0_epc + 4);
 	}
 
-emul:
+      emul:
 	fpuemuprivate.stats.emulated++;
 	switch (MIPSInst_OPCODE(ir)) {
-#ifdef CP0_STATUS_FR_SUPPORT
-		/* R4000+ 64-bit fpu registers */
 #ifndef SINGLE_ONLY_FPU
-	case ldc1_op:
-		{
-			u64 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			          MIPSInst_SIMM(ir);
-			int ft = MIPSInst_RT(ir);
-
-			if (!(regs->cp0_status & ST0_FR))
-				ft &= ~1;
-			err = get_user(ctx->regs[ft], va);
-			fpuemuprivate.stats.loads++;
-			if (err) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
+	case ldc1_op:{
+		u64 *va = REG_TO_VA(xcp->regs[MIPSInst_RS(ir)] +
+			MIPSInst_SIMM(ir));
+		u64 val;
+
+		fpuemuprivate.stats.loads++;
+		if (get_user(val, va)) {
+			fpuemuprivate.stats.errors++;
+			return SIGBUS;
 		}
+		DITOREG(val, MIPSInst_RT(ir));
 		break;
+	}
 
-	case sdc1_op:
-		{
-			fpureg_t *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			               MIPSInst_SIMM(ir);
-			int ft = MIPSInst_RT(ir);
-
-			if (!(regs->cp0_status & ST0_FR))
-				ft &= ~1;
-			fpuemuprivate.stats.stores++;
-			if (put_user(ctx->regs[ft], va)) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
+	case sdc1_op:{
+		u64 *va = REG_TO_VA(xcp->regs[MIPSInst_RS(ir)] +
+			MIPSInst_SIMM(ir));
+		u64 val;
+
+		fpuemuprivate.stats.stores++;
+		DIFROMREG(val, MIPSInst_RT(ir));
+		if (put_user(val, va)) {
+			fpuemuprivate.stats.errors++;
+			return SIGBUS;
 		}
 		break;
+	}
 #endif
 
-	case lwc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			               MIPSInst_SIMM(ir);
-			int ft = MIPSInst_RT(ir);
-			u32 val;
+	case lwc1_op:{
+		u32 *va = REG_TO_VA(xcp->regs[MIPSInst_RS(ir)] +
+			MIPSInst_SIMM(ir));
+		u32 val;
 
-			fpuemuprivate.stats.loads++;
-			err = get_user(val, va);
-			if (err) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
-			if (regs->cp0_status & ST0_FR) {
-				/* load whole register */
-				ctx->regs[ft] = (s64) val;
-			} else if (ft & 1) {
-				/* load to m.s. 32 bits */
-#ifdef SINGLE_ONLY_FPU
-				/* illegal register in single-float mode */
-				return SIGILL;
-#else
-				ctx->regs[(ft & ~1)] &= 0xffffffff;
-				ctx->regs[(ft & ~1)] |= (fpureg_t) val << 32;
-#endif
-			} else {
-				/* load to l.s. 32 bits */
-				ctx->regs[ft] &= ~0xffffffffLL;
-				ctx->regs[ft] |= val;
-			}
+		fpuemuprivate.stats.loads++;
+		if (get_user(val, va)) {
+			fpuemuprivate.stats.errors++;
+			return SIGBUS;
 		}
-		break;
-
-	case swc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			          MIPSInst_SIMM(ir);
-			unsigned int val;
-			int ft = MIPSInst_RT(ir);
-
-			fpuemuprivate.stats.stores++;
-			if (regs->cp0_status & ST0_FR) {
-				/* store whole register */
-				val = ctx->regs[ft];
-			} else if (ft & 1) {
 #ifdef SINGLE_ONLY_FPU
-				/* illegal register in single-float mode */
-				return SIGILL;
-#else
-				/* store from m.s. 32 bits */
-				val = ctx->regs[(ft & ~1)] >> 32;
-#endif
-			} else {
-				/* store from l.s. 32 bits */
-				val = ctx->regs[ft];
-			}
-			if (put_user(val, va)) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
-		}
-		break;
-#else				/* old 32-bit fpu registers */
-	case lwc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			          MIPSInst_SIMM(ir);
-			err = get_user(ctx->regs[MIPSInst_RT(ir)], va);
-
-			fpuemuprivate.stats.loads++;
-			if (err) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
-		}
-		break;
-
-	case swc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			          MIPSInst_SIMM(ir);
-			fpuemuprivate.stats.stores++;
-			if (put_user(ctx->regs[MIPSInst_RT(ir)], va)) {
-				fpuemuprivate.stats.errors++;
-				return SIGBUS;
-			}
+		if (MIPSInst_RT(ir) & 1) {
+			/* illegal register in single-float mode */
+			return SIGILL;
 		}
-		break;
-	case ldc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)])
-			    + MIPSInst_SIMM(ir);
-			unsigned int rt = MIPSInst_RT(ir) & ~1;
-			int errs = 0;
-			fpuemuprivate.stats.loads++;
-#if (defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN) || defined(__MIPSEB__)
-			err = get_user(ctx->regs[rt + 1], va + 0);
-			err |= get_user(ctx->regs[rt + 0], va + 1);
-#else
-			err = get_user(ctx->regs[rt + 0], va + 0);
-			err |= get_user(ctx->regs[rt + 1], va + 1);
 #endif
-			if (err)
-				return SIGBUS;
-		}
+		SITOREG(val, MIPSInst_RT(ir));
 		break;
+	}
 
-	case sdc1_op:
-		{
-			u32 *va = REG_TO_VA(regs->regs[MIPSInst_RS(ir)]) +
-			          MIPSInst_SIMM(ir);
-			unsigned int rt = MIPSInst_RT(ir) & ~1;
+	case swc1_op:{
+		u32 *va = REG_TO_VA(xcp->regs[MIPSInst_RS(ir)] +
+			MIPSInst_SIMM(ir));
+		u32 val;
 
-			fpuemuprivate.stats.stores++;
-#if (defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN) || defined(__MIPSEB__)
-			if (put_user(ctx->regs[rt + 1], va + 0)
-			    || put_user(ctx->regs[rt + 0], va + 1))
-				return SIGBUS;
-#else
-			if (put_user(ctx->regs[rt + 0], va + 0)
-			    || put_user(ctx->regs[rt + 1], va + 1))
-				return SIGBUS;
+		fpuemuprivate.stats.stores++;
+#ifdef SINGLE_ONLY_FPU
+		if (MIPSInst_RT(ir) & 1) {
+			/* illegal register in single-float mode */
+			return SIGILL;
+		}
 #endif
+		SIFROMREG(val, MIPSInst_RT(ir));
+		if (put_user(val, va)) {
+			fpuemuprivate.stats.errors++;
+			return SIGBUS;
 		}
 		break;
-#endif
+	}
 
 	case cop1_op:
 		switch (MIPSInst_RS(ir)) {
 
-#ifdef CP0_STATUS_FR_SUPPORT
 #if __mips64 && !defined(SINGLE_ONLY_FPU)
 		case dmfc_op:
 			/* copregister fs -> gpr[rt] */
 			if (MIPSInst_RT(ir) != 0) {
-				int fs = MIPSInst_RD(ir);
-				if (!(regs->cp0_status & ST0_FR))
-					fs &= ~1;
-				regs->regs[MIPSInst_RT(ir)] = ctx->regs[fs];
+				DIFROMREG(xcp->regs[MIPSInst_RT(ir)],
+					MIPSInst_RD(ir));
 			}
 			break;
 
-		case dmtc_op: {
+		case dmtc_op:
 			/* copregister fs <- rt */
-				fpureg_t value;
-				int fs = MIPSInst_RD(ir);
-				if (!(regs->cp0_status & ST0_FR))
-					fs &= ~1;
-				value =
-				    (MIPSInst_RT(ir) ==
-				     0) ? 0 : regs->regs[MIPSInst_RT(ir)];
-				ctx->regs[fs] = value;
+			DITOREG(xcp->regs[MIPSInst_RT(ir)], MIPSInst_RD(ir));
 			break;
-		}
 #endif
 
 		case mfc_op:
 			/* copregister rd -> gpr[rt] */
-			if (MIPSInst_RT(ir) != 0) {
-				/* default value from l.s. 32 bits */
-				int value = ctx->regs[MIPSInst_RD(ir)];
-				if (MIPSInst_RD(ir) & 1) {
 #ifdef SINGLE_ONLY_FPU
-					/* illegal register in single-float mode */
-					return SIGILL;
-#else
-					if (!(regs->cp0_status & ST0_FR)) {
-						/* move from m.s. 32 bits */
-						value =
-						    ctx->
-						    regs[MIPSInst_RD(ir) &
-							 ~1] >> 32;
-					}
-#endif
-				}
-				regs->regs[MIPSInst_RT(ir)] = value;
+			if (MIPSInst_RD(ir) & 1) {
+				/* illegal register in single-float mode */
+				return SIGILL;
 			}
-			break;
-
-		case mtc_op:
-			/* copregister rd <- rt */
-			{
-				fpureg_t value;
-				if (MIPSInst_RT(ir) == 0)
-					value = 0;
-				else
-					value =
-					    (unsigned int) regs->
-					    regs[MIPSInst_RT(ir)];
-				if (MIPSInst_RD(ir) & 1) {
-#ifdef SINGLE_ONLY_FPU
-					/* illegal register in single-float mode */
-					return SIGILL;
-#else
-					if (!(regs->cp0_status & ST0_FR)) {
-						/* move to m.s. 32 bits */
-						ctx->
-						    regs[
-							 (MIPSInst_RD(ir) &
-							  ~1)] &=
-						    0xffffffff;
-						ctx->
-						    regs[
-							 (MIPSInst_RD(ir) &
-							  ~1)] |=
-						    value << 32;
-						break;
-					}
 #endif
-				}
-				/* move to l.s. 32 bits */
-				ctx->regs[MIPSInst_RD(ir)] &=
-				    ~0xffffffffLL;
-				ctx->regs[MIPSInst_RD(ir)] |= value;
-			}
-			break;
-#else
-
-		case mfc_op:
-			/* copregister rd -> gpr[rt] */
 			if (MIPSInst_RT(ir) != 0) {
-				unsigned value =
-				    ctx->regs[MIPSInst_RD(ir)];
-				regs->regs[MIPSInst_RT(ir)] = value;
+				SIFROMREG(xcp->regs[MIPSInst_RT(ir)],
+					MIPSInst_RD(ir));
 			}
 			break;
 
 		case mtc_op:
 			/* copregister rd <- rt */
-			{
-				unsigned value;
-				value =
-				    (MIPSInst_RT(ir) ==
-				     0) ? 0 : regs->regs[MIPSInst_RT(ir)];
-				ctx->regs[MIPSInst_RD(ir)] = value;
+#ifdef SINGLE_ONLY_FPU
+			if (MIPSInst_RD(ir) & 1) {
+				/* illegal register in single-float mode */
+				return SIGILL;
 			}
-			break;
 #endif
+			SITOREG(xcp->regs[MIPSInst_RT(ir)], MIPSInst_RD(ir));
+			break;
 
-		case cfc_op:
+		case cfc_op:{
 			/* cop control register rd -> gpr[rt] */
-			{
-				unsigned value;
+			u32 value;
 
-				if (MIPSInst_RD(ir) == FPCREG_CSR) {
-					value = ctx->sr;
+			if (ir == CP1UNDEF) {
+				return do_dsemulret(xcp);
+			}
+			if (MIPSInst_RD(ir) == FPCREG_CSR) {
+				value = ctx->sr;
 #ifdef CSRTRACE
-					printk
-					    ("%p gpr[%d]<-csr=%08x\n",
-					     REG_TO_VA(regs->cp0_epc),
-					     MIPSInst_RT(ir), value);
+				printk("%p gpr[%d]<-csr=%08x\n",
+					REG_TO_VA(xcp->cp0_epc),
+					MIPSInst_RT(ir), value);
 #endif
-				} else if (MIPSInst_RD(ir) == FPCREG_RID)
-					value = 0;
-				else
-					value = 0;
-				if (MIPSInst_RT(ir))
-					regs->regs[MIPSInst_RT(ir)] = value;
 			}
+			else if (MIPSInst_RD(ir) == FPCREG_RID)
+				value = 0;
+			else
+				value = 0;
+			if (MIPSInst_RT(ir))
+				xcp->regs[MIPSInst_RT(ir)] = value;
 			break;
+		}
 
-		case ctc_op:
+		case ctc_op:{
 			/* copregister rd <- rt */
-			{
-				unsigned value;
-
-				if (MIPSInst_RT(ir) == 0)
-					value = 0;
-				else
-					value = regs->regs[MIPSInst_RT(ir)];
+			u32 value;
 
-				/* we only have one writable control reg
-				 */
-				if (MIPSInst_RD(ir) == FPCREG_CSR) {
+			if (MIPSInst_RT(ir) == 0)
+				value = 0;
+			else
+				value = xcp->regs[MIPSInst_RT(ir)];
+
+			/* we only have one writable control reg
+			 */
+			if (MIPSInst_RD(ir) == FPCREG_CSR) {
 #ifdef CSRTRACE
-					printk
-					    ("%p gpr[%d]->csr=%08x\n",
-					     REG_TO_VA(regs->cp0_epc),
-					     MIPSInst_RT(ir), value);
-#endif
-					ctx->sr = value;
-					/* copy new rounding mode and
-					   flush bit to ieee library state! */
-					ieee754_csr.nod =
-					    (ctx->sr & 0x1000000) != 0;
-					ieee754_csr.rm =
-					    ieee_rm[value & 0x3];
-				}
+				printk("%p gpr[%d]->csr=%08x\n",
+					REG_TO_VA(xcp->cp0_epc),
+					MIPSInst_RT(ir), value);
+#endif
+				ctx->sr = value;
+				/* copy new rounding mode and
+				   flush bit to ieee library state! */
+				ieee754_csr.nod = (ctx->sr & 0x1000000) != 0;
+				ieee754_csr.rm = ieee_rm[value & 0x3];
+			}
+			if ((ctx->sr >> 5) & ctx->sr & FPU_CSR_ALL_E) {
+				return SIGFPE;
 			}
 			break;
+		}
 
-		case bc_op: {
+		case bc_op:{
 			int likely = 0;
 
-			if (regs->cp0_cause & CAUSEF_BD)
+			if (xcp->cp0_cause & CAUSEF_BD)
 				return SIGILL;
 
 #if __mips >= 4
-			cond = ctx-> sr & fpucondbit[MIPSInst_RT(ir) >> 2];
+			cond = ctx->sr & fpucondbit[MIPSInst_RT(ir) >> 2];
 #else
 			cond = ctx->sr & FPU_CSR_COND;
 #endif
@@ -593,16 +440,18 @@
 				return SIGILL;
 			}
 
-			regs->cp0_cause |= CAUSEF_BD;
+			xcp->cp0_cause |= CAUSEF_BD;
 			if (cond) {
-				/* branch taken: emulate dslot instruction */
-				regs->cp0_epc += 4;
-				contpc = REG_TO_VA regs->cp0_epc +
-				                   (MIPSInst_SIMM(ir) << 2);
-
-				err = get_user(ir,
-					(mips_instruction *)regs->cp0_epc);
-				if (err) {
+				/* branch taken: emulate dslot
+				 * instruction
+				 */
+				xcp->cp0_epc += 4;
+				contpc = REG_TO_VA
+					(xcp->cp0_epc +
+					(MIPSInst_SIMM(ir) << 2));
+
+				if (get_user(ir, (mips_instruction *)
+						REG_TO_VA xcp->cp0_epc)) {
 					fpuemuprivate.stats.errors++;
 					return SIGBUS;
 				}
@@ -629,49 +478,50 @@
 				}
 
 				/*
-				 * Single step the non-cp1 instruction in the
-				 * dslot
+				 * Single step the non-cp1
+				 * instruction in the dslot
 				 */
-				return mips_dsemul(regs, ir, contpc);
-			} else {
+				return mips_dsemul(xcp, ir, VA_TO_REG contpc);
+			}
+			else {
 				/* branch not taken */
 				if (likely) {
 					/*
-					 * branch likely nullifies dslot if not
-					 * taken
+					 * branch likely nullifies
+					 * dslot if not taken
 					 */
-					regs->cp0_epc += 4;
+					xcp->cp0_epc += 4;
 					contpc += 4;
 					/*
-					 * else continue & execute dslot as
-					 * normal insn
+					 * else continue & execute
+					 * dslot as normal insn
 					 */
 				}
 			}
 			break;
 		}
 
-		default: {
-			int sig;
-
+		default:
 			if (!(MIPSInst_RS(ir) & 0x10))
 				return SIGILL;
+			{
+				int sig;
 
-			/* a real fpu computation instruction */
-				if ((sig = fpu_emu(regs, ctx, ir)))
+				/* a real fpu computation instruction */
+				if ((sig = fpu_emu(xcp, ctx, ir)))
 					return sig;
 			}
 		}
 		break;
 
 #if __mips >= 4 && __mips != 32
-	case cop1x_op:
-		{
-			int sig;
-			if ((sig = fpux_emu(regs, ctx, ir)))
-				return sig;
-		}
+	case cop1x_op:{
+		int sig;
+
+		if ((sig = fpux_emu(xcp, ctx, ir)))
+			return sig;
 		break;
+	}
 #endif
 
 #if __mips >= 4
@@ -681,7 +531,7 @@
 		cond = fpucondbit[MIPSInst_RT(ir) >> 2];
 		if (((ctx->sr & cond) != 0) != ((MIPSInst_RT(ir) & 1) != 0))
 			return 0;
-		regs->regs[MIPSInst_RD(ir)] = regs->regs[MIPSInst_RS(ir)];
+		xcp->regs[MIPSInst_RD(ir)] = xcp->regs[MIPSInst_RS(ir)];
 		break;
 #endif
 
@@ -690,194 +540,26 @@
 	}
 
 	/* we did it !! */
-	regs->cp0_epc = VA_TO_REG(contpc);
-	regs->cp0_cause &= ~CAUSEF_BD;
+	xcp->cp0_epc = VA_TO_REG(contpc);
+	xcp->cp0_cause &= ~CAUSEF_BD;
 	return 0;
 }
 
 /*
- * Emulate the arbritrary instruction ir at xcp->cp0_epc.  Required when
- * we have to emulate the instruction in a COP1 branch delay slot.  Do
- * not change cp0_epc due to the instruction
- *
- * According to the spec:
- * 1) it shouldnt be a branch :-)
- * 2) it can be a COP instruction :-(
- * 3) if we are tring to run a protected memory space we must take
- *    special care on memory access instructions :-(
- */
-
-/*
- * "Trampoline" return routine to catch exception following
- *  execution of delay-slot instruction execution.
- */
-
-/* Instruction inserted following delay slot instruction to force trap */
-#define AdELOAD 0x8c000001	/* lw $0,1($0) */
-
-/* Instruction inserted following the AdELOAD to further tag the sequence */
-#define BD_COOKIE 0x0000bd36	/* tne $0,$0 with baggage */
-
-struct emuframe {
-	mips_instruction	emul;
-	mips_instruction	adel;
-	mips_instruction	cookie;
-	unsigned long	epc;
-};
-
-int do_dsemulret(struct pt_regs *xcp)
-{
-	struct emuframe *fr;
-	unsigned long epc;
-	u32 insn, cookie;
-	int err = 0;
-
-	fr = (struct emuframe *) (xcp->cp0_epc - sizeof(mips_instruction));
-
-	/* 
-	 * If we can't even access the area, something is very wrong, but we'll
-	 * leave that to the default handling
-	 */
-	if (verify_area(VERIFY_READ, fr, sizeof(struct emuframe)))
-		return 0;
-
-	/*
-	 * Do some sanity checking on the stackframe:
-	 *
-	 *  - Is the instruction pointed to by the EPC an AdELOAD?
-	 *  - Is the following memory word the BD_COOKIE?
-	 */
-	err = __get_user(insn, &fr->adel);
-	err |= __get_user(cookie, &fr->cookie);
-
-	if (unlikely(err || (insn != AdELOAD) || (cookie != BD_COOKIE))) {
-		fpuemuprivate.stats.errors++;
-
-		return 0;
-	}
-
-	/* 
-	 * At this point, we are satisfied that it's a BD emulation trap.  Yes,
-	 * a user might have deliberately put two malformed and useless
-	 * instructions in a row in his program, in which case he's in for a
-	 * nasty surprise - the next instruction will be treated as a
-	 * continuation address!  Alas, this seems to be the only way that we
-	 * can handle signals, recursion, and longjmps() in the context of
-	 * emulating the branch delay instruction.
-	 */
-
-#ifdef DSEMUL_TRACE
-	printk("desemulret\n");
-#endif
-	if (__get_user(epc, &fr->epc)) {		/* Saved EPC */
-		/* This is not a good situation to be in */
-		force_sig(SIGBUS, current);
-
-		return 1;
-	}
-
-	/* Set EPC to return to post-branch instruction */
-	xcp->cp0_epc = epc;
-
-	return 1;
-}
-
-static int mips_dsemul(struct pt_regs *regs, mips_instruction ir,
-	unsigned long cpc)
-{
-	extern asmlinkage void handle_dsemulret(void);
-	mips_instruction *dsemul_insns;
-	struct emuframe *fr;
-	int err;
-
-	if (ir == 0) {		/* a nop is easy */
-		regs->cp0_epc = cpc;
-		regs->cp0_cause &= ~CAUSEF_BD;
-		return 0;
-	}
-#ifdef DSEMUL_TRACE
-	printk("desemul %lx %lx\n", regs->cp0_epc, cpc);
-
-#endif
- 
-	/* 
-	 * The strategy is to push the instruction onto the user stack 
-	 * and put a trap after it which we can catch and jump to 
-	 * the required address any alternative apart from full 
-	 * instruction emulation!!.
-	 *
-	 * Algorithmics used a system call instruction, and
-	 * borrowed that vector.  MIPS/Linux version is a bit
-	 * more heavyweight in the interests of portability and
-	 * multiprocessor support.  We flag the thread for special
-	 * handling in the unaligned access handler and force an
-	 * address error excpetion.
-	 */
-
-	/* Ensure that the two instructions are in the same cache line */
-	dsemul_insns = (mips_instruction *) (regs->regs[29] & ~0xf);
-	dsemul_insns -= 4;	/* Retain 16-byte alignment */
-	fr = (struct emuframe *) dsemul_insns;
-
-	/* Verify that the stack pointer is not competely insane */
-	if (unlikely(verify_area(VERIFY_WRITE, fr, sizeof(struct emuframe))))
-		return SIGBUS;
-
-	err = __put_user(ir, &dsemul_insns[0]);
-	err |= __put_user((mips_instruction)AdELOAD, &fr->adel);
-	err |= __put_user((mips_instruction)BD_COOKIE, &fr->cookie);
-	err |= __put_user(cpc, &fr->epc);
-
-	if (unlikely(err)) {
-		fpuemuprivate.stats.errors++;
-		return SIGBUS;
-	}
-
-	regs->cp0_epc = VA_TO_REG & dsemul_insns[0];
-
-	flush_cache_sigtramp((unsigned long)&fr->adel);
-
-	return SIGILL;		/* force out of emulation loop */
-}
-
-/* 
  * Conversion table from MIPS compare ops 48-63
- * cond = ieee754dp_cmp(x,y,IEEE754_UN);
+ * cond = ieee754dp_cmp(x,y,IEEE754_UN,sig);
  */
 static const unsigned char cmptab[8] = {
-	0,					/* cmp_0 (sig) cmp_sf */
-	IEEE754_CUN,				/* cmp_un (sig) cmp_ngle */
-	IEEE754_CEQ,				/* cmp_eq (sig) cmp_seq */
-	IEEE754_CEQ | IEEE754_CUN,		/* cmp_ueq (sig) cmp_ngl  */
-	IEEE754_CLT,				/* cmp_olt (sig) cmp_lt */
-	IEEE754_CLT | IEEE754_CUN,		/* cmp_ult (sig) cmp_nge */
-	IEEE754_CLT | IEEE754_CEQ,		/* cmp_ole (sig) cmp_le */
-	IEEE754_CLT | IEEE754_CEQ | IEEE754_CUN, /* cmp_ule (sig) cmp_ngt */
+	0,			/* cmp_0 (sig) cmp_sf */
+	IEEE754_CUN,		/* cmp_un (sig) cmp_ngle */
+	IEEE754_CEQ,		/* cmp_eq (sig) cmp_seq */
+	IEEE754_CEQ | IEEE754_CUN,	/* cmp_ueq (sig) cmp_ngl  */
+	IEEE754_CLT,		/* cmp_olt (sig) cmp_lt */
+	IEEE754_CLT | IEEE754_CUN,	/* cmp_ult (sig) cmp_nge */
+	IEEE754_CLT | IEEE754_CEQ,	/* cmp_ole (sig) cmp_le */
+	IEEE754_CLT | IEEE754_CEQ | IEEE754_CUN,	/* cmp_ule (sig) cmp_ngt */
 };
 
-#define SIFROMREG(si,x)	((si) = ctx->regs[x])
-#define SITOREG(si,x)	(ctx->regs[x] = (int)(si))
-
-#if __mips64 && !defined(SINGLE_ONLY_FPU)
-#define DIFROMREG(di,x)	((di) = ctx->regs[x])
-#define DITOREG(di,x)	(ctx->regs[x] = (di))
-#endif
-
-#define SPFROMREG(sp,x)	((sp).bits = ctx->regs[x])
-#define SPTOREG(sp,x)	(ctx->regs[x] = (sp).bits)
-
-#ifdef CP0_STATUS_FR_SUPPORT
-#define DPFROMREG(dp,x)	((dp).bits = \
-	ctx->regs[(xcp->cp0_status & ST0_FR) ? x : (x & ~1)])
-#define DPTOREG(dp,x)	(ctx->regs[(xcp->cp0_status & ST0_FR) ? x : (x & ~1)] \
-	= (dp).bits)
-#else
-/* Beware: MIPS COP1 doubles are always little_word endian in registers */
-#define DPFROMREG(dp,x)	\
-	((dp).bits = ((u64)ctx->regs[(x)+1] << 32) | ctx->regs[x])
-#define DPTOREG(dp,x) \
-	(ctx->regs[x] = (dp).bits, ctx->regs[(x)+1] = (dp).bits >> 32)
-#endif
 
 #if __mips >= 4 && __mips != 32
 
@@ -889,17 +571,17 @@
 static ieee754##p fpemu_##p##_##name (ieee754##p r, ieee754##p s, \
     ieee754##p t) \
 { \
-    struct ieee754_csr ieee754_csr_save; \
-    s = f1 (s, t); \
-    ieee754_csr_save = ieee754_csr; \
-    s = f2 (s, r); \
-    ieee754_csr_save.cx |= ieee754_csr.cx; \
-    ieee754_csr_save.sx |= ieee754_csr.sx; \
-    s = f3 (s); \
-    ieee754_csr.cx |= ieee754_csr_save.cx; \
-    ieee754_csr.sx |= ieee754_csr_save.sx; \
-    return s; \
-}    
+	struct ieee754_csr ieee754_csr_save; \
+	s = f1 (s, t); \
+	ieee754_csr_save = ieee754_csr; \
+	s = f2 (s, r); \
+	ieee754_csr_save.cx |= ieee754_csr.cx; \
+	ieee754_csr_save.sx |= ieee754_csr.sx; \
+	s = f3 (s); \
+	ieee754_csr.cx |= ieee754_csr_save.cx; \
+	ieee754_csr.sx |= ieee754_csr_save.sx; \
+	return s; \
+}
 
 static ieee754dp fpemu_dp_recip(ieee754dp d)
 {
@@ -921,233 +603,184 @@
 	return ieee754sp_div(ieee754sp_one(0), ieee754sp_sqrt(s));
 }
 
-DEF3OP(madd, sp, ieee754sp_mul, ieee754sp_add, );
-DEF3OP(msub, sp, ieee754sp_mul, ieee754sp_sub, );
+DEF3OP(madd, sp, ieee754sp_mul, ieee754sp_add,);
+DEF3OP(msub, sp, ieee754sp_mul, ieee754sp_sub,);
 DEF3OP(nmadd, sp, ieee754sp_mul, ieee754sp_add, ieee754sp_neg);
 DEF3OP(nmsub, sp, ieee754sp_mul, ieee754sp_sub, ieee754sp_neg);
-DEF3OP(madd, dp, ieee754dp_mul, ieee754dp_add, );
-DEF3OP(msub, dp, ieee754dp_mul, ieee754dp_sub, );
+DEF3OP(madd, dp, ieee754dp_mul, ieee754dp_add,);
+DEF3OP(msub, dp, ieee754dp_mul, ieee754dp_sub,);
 DEF3OP(nmadd, dp, ieee754dp_mul, ieee754dp_add, ieee754dp_neg);
 DEF3OP(nmsub, dp, ieee754dp_mul, ieee754dp_sub, ieee754dp_neg);
 
 static int fpux_emu(struct pt_regs *xcp, struct mips_fpu_soft_struct *ctx,
-                    mips_instruction ir)
+	mips_instruction ir)
 {
 	unsigned rcsr = 0;	/* resulting csr */
 
 	fpuemuprivate.stats.cp1xops++;
 
 	switch (MIPSInst_FMA_FFMT(ir)) {
-	case s_fmt:		/* 0 */
-		{
-			ieee754sp(*handler) (ieee754sp, ieee754sp,
-					     ieee754sp);
-			ieee754sp fd, fr, fs, ft;
-
-			switch (MIPSInst_FUNC(ir)) {
-			case lwxc1_op:
-				{
-					u32 *va = REG_TO_VA(
-						xcp->regs[MIPSInst_FR(ir)] +
-						xcp->regs[MIPSInst_FT(ir)]);
-					fpureg_t val;
-
-					if (get_user(val, va)) {
-						fpuemuprivate.stats.errors++;
-						return SIGBUS;
-					}
-					if (xcp->cp0_status & ST0_FR) {
-						/* load whole register */
-						ctx->regs[MIPSInst_FD(ir)] =
-						    val;
-					} else if (MIPSInst_FD(ir) & 1) {
-						/* load to m.s. 32 bits */
-#if defined(SINGLE_ONLY_FPU)
-						/* illegal register in single-float mode */
-						return SIGILL;
-#else
-						ctx->
-						    regs[
-							 (MIPSInst_FD(ir) &
-							  ~1)] &=
-						    0xffffffff;
-						ctx->
-						    regs[
-							 (MIPSInst_FD(ir) &
-							  ~1)] |=
-						    val << 32;
-#endif
-					} else {
-						/* load to l.s. 32 bits */
-						ctx->
-						    regs[MIPSInst_FD(ir)]
-						    &= ~0xffffffffLL;
-						ctx->
-						    regs[MIPSInst_FD(ir)]
-						    |= val;
-					}
-				}
-				break;
+	case s_fmt:{		/* 0 */
 
-			case swxc1_op:
-				{
-					u32 *va = REG_TO_VA(
-						xcp->regs[MIPSInst_FR(ir)] +
-						xcp->regs[MIPSInst_FT(ir)]);
-					unsigned int val;
-
-					if (xcp->cp0_status & ST0_FR) {
-						/* store whole register */
-						val = ctx->regs[MIPSInst_FS(ir)];
-					} else if (MIPSInst_FS(ir) & 1) {
-#if defined(SINGLE_ONLY_FPU)
-						/* illegal register in single-float mode */
-						return SIGILL;
-#else
-						/* store from m.s. 32 bits */
-						val = ctx->regs[(MIPSInst_FS(ir) & ~1)] >> 32;
+		ieee754sp(*handler) (ieee754sp, ieee754sp, ieee754sp);
+		ieee754sp fd, fr, fs, ft;
+		u32 *va;
+		u32 val;
+
+		switch (MIPSInst_FUNC(ir)) {
+		case lwxc1_op:
+			va = REG_TO_VA(xcp->regs[MIPSInst_FR(ir)] +
+				xcp->regs[MIPSInst_FT(ir)]);
+
+			fpuemuprivate.stats.loads++;
+			if (get_user(val, va)) {
+				fpuemuprivate.stats.errors++;
+				return SIGBUS;
+			}
+#ifdef SINGLE_ONLY_FPU
+			if (MIPSInst_FD(ir) & 1) {
+				/* illegal register in single-float
+				 * mode
+				 */
+				return SIGILL;
+			}
 #endif
-					} else {
-						/* store from l.s. 32 bits */
-						val =
-						    ctx->
-						    regs[MIPSInst_FS(ir)];
-					}
-					if (put_user(val, va)) {
-						fpuemuprivate.stats.errors++;
-						return SIGBUS;
-					}
-				}
-				break;
+			SITOREG(val, MIPSInst_FD(ir));
+			break;
 
-			case madd_s_op:
-				handler = fpemu_sp_madd;
-				goto scoptop;
-			case msub_s_op:
-				handler = fpemu_sp_msub;
-				goto scoptop;
-			case nmadd_s_op:
-				handler = fpemu_sp_nmadd;
-				goto scoptop;
-			case nmsub_s_op:
-				handler = fpemu_sp_nmsub;
-				goto scoptop;
+		case swxc1_op:
+			va = REG_TO_VA(xcp->regs[MIPSInst_FR(ir)] +
+				xcp->regs[MIPSInst_FT(ir)]);
 
-			      scoptop:
-				SPFROMREG(fr, MIPSInst_FR(ir));
-				SPFROMREG(fs, MIPSInst_FS(ir));
-				SPFROMREG(ft, MIPSInst_FT(ir));
-				fd = (*handler) (fr, fs, ft);
-				SPTOREG(fd, MIPSInst_FD(ir));
+			fpuemuprivate.stats.stores++;
+#ifdef SINGLE_ONLY_FPU
+			if (MIPSInst_FS(ir) & 1) {
+				/* illegal register in single-float
+				 * mode
+				 */
+				return SIGILL;
+			}
+#endif
 
-			      copcsr:
-				if (ieee754_cxtest(IEEE754_INEXACT))
-					rcsr |=
-					    FPU_CSR_INE_X | FPU_CSR_INE_S;
-				if (ieee754_cxtest(IEEE754_UNDERFLOW))
-					rcsr |=
-					    FPU_CSR_UDF_X | FPU_CSR_UDF_S;
-				if (ieee754_cxtest(IEEE754_OVERFLOW))
-					rcsr |=
-					    FPU_CSR_OVF_X | FPU_CSR_OVF_S;
-				if (ieee754_cxtest
-				    (IEEE754_INVALID_OPERATION)) rcsr |=
-					    FPU_CSR_INV_X | FPU_CSR_INV_S;
-
-				ctx->sr =
-				    (ctx->sr & ~FPU_CSR_ALL_X) | rcsr;
-				if (ieee754_csr.nod)
-				    ctx->sr |= 0x1000000;
-				if ((ctx->sr >> 5) & ctx->
-				    sr & FPU_CSR_ALL_E) {
-		/*printk ("SIGFPE: fpu csr = %08x\n",ctx->sr); */
-					return SIGFPE;
-				}
+			SIFROMREG(val, MIPSInst_FS(ir));
+			if (put_user(val, va)) {
+				fpuemuprivate.stats.errors++;
+				return SIGBUS;
+			}
+			break;
 
-				break;
+		case madd_s_op:
+			handler = fpemu_sp_madd;
+			goto scoptop;
+		case msub_s_op:
+			handler = fpemu_sp_msub;
+			goto scoptop;
+		case nmadd_s_op:
+			handler = fpemu_sp_nmadd;
+			goto scoptop;
+		case nmsub_s_op:
+			handler = fpemu_sp_nmsub;
+			goto scoptop;
 
-			default:
-				return SIGILL;
+		      scoptop:
+			SPFROMREG(fr, MIPSInst_FR(ir));
+			SPFROMREG(fs, MIPSInst_FS(ir));
+			SPFROMREG(ft, MIPSInst_FT(ir));
+			fd = (*handler) (fr, fs, ft);
+			SPTOREG(fd, MIPSInst_FD(ir));
+
+		      copcsr:
+			if (ieee754_cxtest(IEEE754_INEXACT))
+				rcsr |= FPU_CSR_INE_X | FPU_CSR_INE_S;
+			if (ieee754_cxtest(IEEE754_UNDERFLOW))
+				rcsr |= FPU_CSR_UDF_X | FPU_CSR_UDF_S;
+			if (ieee754_cxtest(IEEE754_OVERFLOW))
+				rcsr |= FPU_CSR_OVF_X | FPU_CSR_OVF_S;
+			if (ieee754_cxtest(IEEE754_INVALID_OPERATION))
+				rcsr |= FPU_CSR_INV_X | FPU_CSR_INV_S;
+
+			ctx->sr = (ctx->sr & ~FPU_CSR_ALL_X) | rcsr;
+			if (ieee754_csr.nod)
+				ctx->sr |= 0x1000000;
+			if ((ctx->sr >> 5) & ctx->sr & FPU_CSR_ALL_E) {
+				/*printk ("SIGFPE: fpu csr = %08x\n",
+				   ctx->sr); */
+				return SIGFPE;
 			}
+
+			break;
+
+		default:
+			return SIGILL;
 		}
 		break;
+	}
 
-#if !defined(SINGLE_ONLY_FPU)
-	case d_fmt:		/* 1 */
-		{
-			ieee754dp(*handler) (ieee754dp, ieee754dp,
-					     ieee754dp);
-			ieee754dp fd, fr, fs, ft;
-
-			switch (MIPSInst_FUNC(ir)) {
-			case ldxc1_op:
-				{
-					u64 *va = REG_TO_VA(
-						xcp->regs[MIPSInst_FR(ir)] +
-						xcp->regs[MIPSInst_FT(ir)]);
-					u64 val;
-
-					if (get_user(val, va)) {
-						fpuemuprivate.stats.errors++;
-						return SIGBUS;
-					}
-					ctx->regs[MIPSInst_FD(ir)] = val;
-				}
-				break;
+#ifndef SINGLE_ONLY_FPU
+	case d_fmt:{		/* 1 */
+		ieee754dp(*handler) (ieee754dp, ieee754dp, ieee754dp);
+		ieee754dp fd, fr, fs, ft;
+		u64 *va;
+		u64 val;
 
-			case sdxc1_op:
-				{
-					u64 *va = REG_TO_VA(
-						xcp->regs[MIPSInst_FR(ir)] +
-						xcp->regs[MIPSInst_FT(ir)]);
-					u64 val;
-
-					val = ctx->regs[MIPSInst_FS(ir)];
-					if (put_user(val, va)) {
-						fpuemuprivate.stats.errors++;
-						return SIGBUS;
-					}
-				}
-				break;
+		switch (MIPSInst_FUNC(ir)) {
+		case ldxc1_op:
+			va = REG_TO_VA(xcp->regs[MIPSInst_FR(ir)] +
+				xcp->regs[MIPSInst_FT(ir)]);
 
-			case madd_d_op:
-				handler = fpemu_dp_madd;
-				goto dcoptop;
-
-			case msub_d_op:
-				handler = fpemu_dp_msub;
-				goto dcoptop;
-
-			case nmadd_d_op:
-				handler = fpemu_dp_nmadd;
-				goto dcoptop;
-
-			case nmsub_d_op:
-				handler = fpemu_dp_nmsub;
-				goto dcoptop;
+			fpuemuprivate.stats.loads++;
+			if (get_user(val, va)) {
+				fpuemuprivate.stats.errors++;
+				return SIGBUS;
+			}
+			DITOREG(val, MIPSInst_FD(ir));
+			break;
 
-			      dcoptop:
-				DPFROMREG(fr, MIPSInst_FR(ir));
-				DPFROMREG(fs, MIPSInst_FS(ir));
-				DPFROMREG(ft, MIPSInst_FT(ir));
-				fd = (*handler) (fr, fs, ft);
-				DPTOREG(fd, MIPSInst_FD(ir));
-				goto copcsr;
+		case sdxc1_op:
+			va = REG_TO_VA(xcp->regs[MIPSInst_FR(ir)] +
+				xcp->regs[MIPSInst_FT(ir)]);
 
-			default:
-				return SIGILL;
+			fpuemuprivate.stats.stores++;
+			DIFROMREG(val, MIPSInst_FS(ir));
+			if (put_user(val, va)) {
+				fpuemuprivate.stats.errors++;
+				return SIGBUS;
 			}
+			break;
+
+		case madd_d_op:
+			handler = fpemu_dp_madd;
+			goto dcoptop;
+		case msub_d_op:
+			handler = fpemu_dp_msub;
+			goto dcoptop;
+		case nmadd_d_op:
+			handler = fpemu_dp_nmadd;
+			goto dcoptop;
+		case nmsub_d_op:
+			handler = fpemu_dp_nmsub;
+			goto dcoptop;
+
+		      dcoptop:
+			DPFROMREG(fr, MIPSInst_FR(ir));
+			DPFROMREG(fs, MIPSInst_FS(ir));
+			DPFROMREG(ft, MIPSInst_FT(ir));
+			fd = (*handler) (fr, fs, ft);
+			DPTOREG(fd, MIPSInst_FD(ir));
+			goto copcsr;
+
+		default:
+			return SIGILL;
 		}
 		break;
+	}
 #endif
 
 	case 0x7:		/* 7 */
-		{
-			if (MIPSInst_FUNC(ir) != pfetch_op) {
-				return SIGILL;
-			}
-			/* ignore prefx operation */
+		if (MIPSInst_FUNC(ir) != pfetch_op) {
+			return SIGILL;
 		}
+		/* ignore prefx operation */
 		break;
 
 	default:
@@ -1164,7 +797,7 @@
  * Emulate a single COP1 arithmetic instruction.
  */
 static int fpu_emu(struct pt_regs *xcp, struct mips_fpu_soft_struct *ctx,
-                   mips_instruction ir)
+	mips_instruction ir)
 {
 	int rfmt;		/* resulting format */
 	unsigned rcsr = 0;	/* resulting csr */
@@ -1174,49 +807,52 @@
 		ieee754sp s;
 		int w;
 #if __mips64
-		long long l;
+		s64 l;
 #endif
 	} rv;			/* resulting value */
 
 	fpuemuprivate.stats.cp1ops++;
 	switch (rfmt = (MIPSInst_FFMT(ir) & 0xf)) {
-	case s_fmt: {		/* 0 */
-		ieee754sp(*handler) ();
+	case s_fmt:{		/* 0 */
+		union {
+			ieee754sp(*b) (ieee754sp, ieee754sp);
+			ieee754sp(*u) (ieee754sp);
+		} handler;
 
 		switch (MIPSInst_FUNC(ir)) {
 			/* binary ops */
 		case fadd_op:
-			handler = ieee754sp_add;
+			handler.b = ieee754sp_add;
 			goto scopbop;
 		case fsub_op:
-			handler = ieee754sp_sub;
+			handler.b = ieee754sp_sub;
 			goto scopbop;
 		case fmul_op:
-			handler = ieee754sp_mul;
+			handler.b = ieee754sp_mul;
 			goto scopbop;
 		case fdiv_op:
-			handler = ieee754sp_div;
+			handler.b = ieee754sp_div;
 			goto scopbop;
 
 			/* unary  ops */
 #if __mips >= 2 || __mips64
 		case fsqrt_op:
-			handler = ieee754sp_sqrt;
+			handler.u = ieee754sp_sqrt;
 			goto scopuop;
 #endif
 #if __mips >= 4 && __mips != 32
 		case frsqrt_op:
-			handler = fpemu_sp_rsqrt;
+			handler.u = fpemu_sp_rsqrt;
 			goto scopuop;
 		case frecip_op:
-			handler = fpemu_sp_recip;
+			handler.u = fpemu_sp_recip;
 			goto scopuop;
 #endif
 #if __mips >= 4
 		case fmovc_op:
 			cond = fpucondbit[MIPSInst_FT(ir) >> 2];
 			if (((ctx->sr & cond) != 0) !=
-			    ((MIPSInst_FT(ir) & 1) != 0))
+				((MIPSInst_FT(ir) & 1) != 0))
 				return 0;
 			SPFROMREG(rv.s, MIPSInst_FS(ir));
 			break;
@@ -1232,10 +868,10 @@
 			break;
 #endif
 		case fabs_op:
-			handler = ieee754sp_abs;
+			handler.u = ieee754sp_abs;
 			goto scopuop;
 		case fneg_op:
-			handler = ieee754sp_neg;
+			handler.u = ieee754sp_neg;
 			goto scopuop;
 		case fmov_op:
 			/* an easy one */
@@ -1243,25 +879,25 @@
 			goto copcsr;
 
 			/* binary op on handler */
-scopbop:
+		      scopbop:
 			{
 				ieee754sp fs, ft;
 
 				SPFROMREG(fs, MIPSInst_FS(ir));
 				SPFROMREG(ft, MIPSInst_FT(ir));
 
-				rv.s = (*handler) (fs, ft);
+				rv.s = (*handler.b) (fs, ft);
 				goto copcsr;
 			}
-scopuop:
+		      scopuop:
 			{
 				ieee754sp fs;
 
 				SPFROMREG(fs, MIPSInst_FS(ir));
-				rv.s = (*handler) (fs);
+				rv.s = (*handler.u) (fs);
 				goto copcsr;
 			}
-copcsr:
+		      copcsr:
 			if (ieee754_cxtest(IEEE754_INEXACT))
 				rcsr |= FPU_CSR_INE_X | FPU_CSR_INE_S;
 			if (ieee754_cxtest(IEEE754_UNDERFLOW))
@@ -1270,15 +906,14 @@
 				rcsr |= FPU_CSR_OVF_X | FPU_CSR_OVF_S;
 			if (ieee754_cxtest(IEEE754_ZERO_DIVIDE))
 				rcsr |= FPU_CSR_DIV_X | FPU_CSR_DIV_S;
-			if (ieee754_cxtest
-				(IEEE754_INVALID_OPERATION)) rcsr |=
-					    FPU_CSR_INV_X | FPU_CSR_INV_S;
-				break;
+			if (ieee754_cxtest(IEEE754_INVALID_OPERATION))
+				rcsr |= FPU_CSR_INV_X | FPU_CSR_INV_S;
+			break;
 
-				/* unary conv ops */
+			/* unary conv ops */
 		case fcvts_op:
 			return SIGILL;	/* not defined */
-		case fcvtd_op: {
+		case fcvtd_op:{
 #ifdef SINGLE_ONLY_FPU
 			return SIGILL;	/* not defined */
 #else
@@ -1290,7 +925,7 @@
 			goto copcsr;
 		}
 #endif
-		case fcvtw_op: {
+		case fcvtw_op:{
 			ieee754sp fs;
 
 			SPFROMREG(fs, MIPSInst_FS(ir));
@@ -1303,7 +938,7 @@
 		case fround_op:
 		case ftrunc_op:
 		case fceil_op:
-		case ffloor_op: {
+		case ffloor_op:{
 			unsigned int oldrm = ieee754_csr.rm;
 			ieee754sp fs;
 
@@ -1317,7 +952,7 @@
 #endif /* __mips >= 2 */
 
 #if __mips64 && !defined(SINGLE_ONLY_FPU)
-		case fcvtl_op: {
+		case fcvtl_op:{
 			ieee754sp fs;
 
 			SPFROMREG(fs, MIPSInst_FS(ir));
@@ -1329,7 +964,7 @@
 		case froundl_op:
 		case ftruncl_op:
 		case fceill_op:
-		case ffloorl_op: {
+		case ffloorl_op:{
 			unsigned int oldrm = ieee754_csr.rm;
 			ieee754sp fs;
 
@@ -1349,14 +984,17 @@
 
 				SPFROMREG(fs, MIPSInst_FS(ir));
 				SPFROMREG(ft, MIPSInst_FT(ir));
-				rv.w = ieee754sp_cmp(fs, ft, cmptab[cmpop & 0x7]);
+				rv.w = ieee754sp_cmp(fs, ft,
+					cmptab[cmpop & 0x7], cmpop & 0x8);
 				rfmt = -1;
-				if ((cmpop & 0x8) && ieee754_cxtest(IEEE754_INVALID_OPERATION))
+				if ((cmpop & 0x8) && ieee754_cxtest
+					(IEEE754_INVALID_OPERATION))
 					rcsr = FPU_CSR_INV_X | FPU_CSR_INV_S;
 				else
 					goto copcsr;
 
-			} else {
+			}
+			else {
 				return SIGILL;
 			}
 			break;
@@ -1364,43 +1002,47 @@
 		break;
 	}
 
-#if !defined(SINGLE_ONLY_FPU)
-	case d_fmt: {
-		ieee754dp(*handler) ();
+#ifndef SINGLE_ONLY_FPU
+	case d_fmt:{
+		union {
+			ieee754dp(*b) (ieee754dp, ieee754dp);
+			ieee754dp(*u) (ieee754dp);
+		} handler;
 
 		switch (MIPSInst_FUNC(ir)) {
 			/* binary ops */
 		case fadd_op:
-			handler = ieee754dp_add;
+			handler.b = ieee754dp_add;
 			goto dcopbop;
 		case fsub_op:
-			handler = ieee754dp_sub;
+			handler.b = ieee754dp_sub;
 			goto dcopbop;
 		case fmul_op:
-			handler = ieee754dp_mul;
+			handler.b = ieee754dp_mul;
 			goto dcopbop;
 		case fdiv_op:
-			handler = ieee754dp_div;
+			handler.b = ieee754dp_div;
 			goto dcopbop;
 
 			/* unary  ops */
 #if __mips >= 2 || __mips64
 		case fsqrt_op:
-			handler = ieee754dp_sqrt;
+			handler.u = ieee754dp_sqrt;
 			goto dcopuop;
 #endif
 #if __mips >= 4 && __mips != 32
 		case frsqrt_op:
-			handler = fpemu_dp_rsqrt;
+			handler.u = fpemu_dp_rsqrt;
 			goto dcopuop;
 		case frecip_op:
-			handler = fpemu_dp_recip;
+			handler.u = fpemu_dp_recip;
 			goto dcopuop;
 #endif
 #if __mips >= 4
 		case fmovc_op:
 			cond = fpucondbit[MIPSInst_FT(ir) >> 2];
-			if (((ctx->sr & cond) != 0) != ((MIPSInst_FT(ir) & 1) != 0))
+			if (((ctx->sr & cond) != 0) !=
+				((MIPSInst_FT(ir) & 1) != 0))
 				return 0;
 			DPFROMREG(rv.d, MIPSInst_FS(ir));
 			break;
@@ -1416,11 +1058,11 @@
 			break;
 #endif
 		case fabs_op:
-			handler = ieee754dp_abs;
+			handler.u = ieee754dp_abs;
 			goto dcopuop;
 
 		case fneg_op:
-			handler = ieee754dp_neg;
+			handler.u = ieee754dp_neg;
 			goto dcopuop;
 
 		case fmov_op:
@@ -1429,27 +1071,25 @@
 			goto copcsr;
 
 			/* binary op on handler */
-dcopbop:
-			{
+		      dcopbop:{
 				ieee754dp fs, ft;
 
 				DPFROMREG(fs, MIPSInst_FS(ir));
 				DPFROMREG(ft, MIPSInst_FT(ir));
 
-				rv.d = (*handler) (fs, ft);
+				rv.d = (*handler.b) (fs, ft);
 				goto copcsr;
 			}
-dcopuop:
-			{
+		      dcopuop:{
 				ieee754dp fs;
 
 				DPFROMREG(fs, MIPSInst_FS(ir));
-				rv.d = (*handler) (fs);
+				rv.d = (*handler.u) (fs);
 				goto copcsr;
 			}
 
-		/* unary conv ops */
-		case fcvts_op: {
+			/* unary conv ops */
+		case fcvts_op:{
 			ieee754dp fs;
 
 			DPFROMREG(fs, MIPSInst_FS(ir));
@@ -1460,7 +1100,7 @@
 		case fcvtd_op:
 			return SIGILL;	/* not defined */
 
-		case fcvtw_op: {
+		case fcvtw_op:{
 			ieee754dp fs;
 
 			DPFROMREG(fs, MIPSInst_FS(ir));
@@ -1473,7 +1113,7 @@
 		case fround_op:
 		case ftrunc_op:
 		case fceil_op:
-		case ffloor_op: {
+		case ffloor_op:{
 			unsigned int oldrm = ieee754_csr.rm;
 			ieee754dp fs;
 
@@ -1487,7 +1127,7 @@
 #endif
 
 #if __mips64 && !defined(SINGLE_ONLY_FPU)
-		case fcvtl_op: {
+		case fcvtl_op:{
 			ieee754dp fs;
 
 			DPFROMREG(fs, MIPSInst_FS(ir));
@@ -1499,7 +1139,7 @@
 		case froundl_op:
 		case ftruncl_op:
 		case fceill_op:
-		case ffloorl_op: {
+		case ffloorl_op:{
 			unsigned int oldrm = ieee754_csr.rm;
 			ieee754dp fs;
 
@@ -1519,33 +1159,42 @@
 
 				DPFROMREG(fs, MIPSInst_FS(ir));
 				DPFROMREG(ft, MIPSInst_FT(ir));
-				rv.w = ieee754dp_cmp(fs, ft, cmptab[cmpop & 0x7]);
+				rv.w = ieee754dp_cmp(fs, ft,
+					cmptab[cmpop & 0x7], cmpop & 0x8);
 				rfmt = -1;
-				if ((cmpop & 0x8) && ieee754_cxtest (IEEE754_INVALID_OPERATION))
+				if ((cmpop & 0x8)
+					&&
+					ieee754_cxtest
+					(IEEE754_INVALID_OPERATION))
 					rcsr = FPU_CSR_INV_X | FPU_CSR_INV_S;
 				else
 					goto copcsr;
 
-			} else {
+			}
+			else {
 				return SIGILL;
 			}
 			break;
 		}
 		break;
 	}
-#endif /* !defined(SINGLE_ONLY_FPU) */
+#endif /* ifndef SINGLE_ONLY_FPU */
+
+	case w_fmt:{
+		ieee754sp fs;
 
-	case w_fmt: {
 		switch (MIPSInst_FUNC(ir)) {
 		case fcvts_op:
 			/* convert word to single precision real */
-			rv.s = ieee754sp_fint(ctx-> regs[MIPSInst_FS(ir)]);
+			SPFROMREG(fs, MIPSInst_FS(ir));
+			rv.s = ieee754sp_fint(fs.bits);
 			rfmt = s_fmt;
 			goto copcsr;
-#if !defined(SINGLE_ONLY_FPU)
+#ifndef SINGLE_ONLY_FPU
 		case fcvtd_op:
 			/* convert word to double precision real */
-			rv.d = ieee754dp_fint(ctx-> regs[MIPSInst_FS(ir)]);
+			SPFROMREG(fs, MIPSInst_FS(ir));
+			rv.d = ieee754dp_fint(fs.bits);
 			rfmt = d_fmt;
 			goto copcsr;
 #endif
@@ -1556,16 +1205,16 @@
 	}
 
 #if __mips64 && !defined(SINGLE_ONLY_FPU)
-	case l_fmt: {
+	case l_fmt:{
 		switch (MIPSInst_FUNC(ir)) {
 		case fcvts_op:
 			/* convert long to single precision real */
-			rv.s = ieee754sp_flong(ctx-> regs[MIPSInst_FS(ir)]);
+			rv.s = ieee754sp_flong(ctx->regs[MIPSInst_FS(ir)]);
 			rfmt = s_fmt;
 			goto copcsr;
 		case fcvtd_op:
 			/* convert long to double precision real */
-			rv.d = ieee754dp_flong(ctx-> regs[MIPSInst_FS(ir)]);
+			rv.d = ieee754dp_flong(ctx->regs[MIPSInst_FS(ir)]);
 			rfmt = d_fmt;
 			goto copcsr;
 		default:
@@ -1592,11 +1241,11 @@
 		return SIGFPE;
 	}
 
-	/* 
+	/*
 	 * Now we can safely write the result back to the register file.
 	 */
 	switch (rfmt) {
-	case -1: {
+	case -1:{
 #if __mips >= 4
 		cond = fpucondbit[MIPSInst_FD(ir) >> 2];
 #else
@@ -1608,7 +1257,7 @@
 			ctx->sr &= ~cond;
 		break;
 	}
-#if !defined(SINGLE_ONLY_FPU)
+#ifndef SINGLE_ONLY_FPU
 	case d_fmt:
 		DPTOREG(rv.d, MIPSInst_FD(ir));
 		break;
@@ -1631,16 +1280,10 @@
 	return 0;
 }
 
-
-/*
- * Emulate the floating point instruction at EPC, and continue to run until we
- * hit a non-fp instruction, or a backward branch.  This cuts down dramatically
- * on the per instruction exception overhead.
- */
-int fpu_emulator_cop1Handler(struct pt_regs *xcp)
+int fpu_emulator_cop1Handler(int xcptno, struct pt_regs *xcp,
+	struct mips_fpu_soft_struct *ctx)
 {
-	struct mips_fpu_soft_struct *ctx = &current->thread.fpu.soft;
-	unsigned long oldepc, prevepc;
+	gpreg_t oldepc, prevepc;
 	mips_instruction insn, *insnp;
 	int sig = 0;
 
@@ -1652,9 +1295,10 @@
 		 * This is a braindead way to do it but the only sane way I
 		 * found to keep the 64-bit egcs 1.1.2 from crashing.
 		 */
-		insnp = (mips_instruction *) xcp->cp0_epc;
+		insnp = (mips_instruction *) REG_TO_VA xcp->cp0_epc;
+
 		if (verify_area(VERIFY_READ, insnp, 4) ||
-		    __get_user(insn, insnp)) {
+			__get_user(insn, insnp)) {
 			fpuemuprivate.stats.errors++;
 			return SIGBUS;
 		}
@@ -1673,9 +1317,13 @@
 			break;
 		if (sig)
 			break;
-
+#ifdef STANDALONE_EMULATOR
+		if (xcptno == EX_FPE)
+			break;
+#else
 		if (current->need_resched)
 			schedule();
+#endif
 	} while (xcp->cp0_epc > prevepc);
 
 	/* SIGILL indicates a non-fpu instruction */

FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen (who was at: slshen@lbl.gov)