MIPS: microMIPS: Add support for exception handling.

All exceptions must be taken in microMIPS mode, never in classic
MIPS mode or the kernel falls apart. A few NOP instructions are
used to maintain the correct alignment of microMIPS versions of
the exception vectors.

Signed-off-by: Steven J. Hill <Steven.Hill@imgtec.com>

5 files changed, 277 insertions, 101 deletions

diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c
index 5fe66a0c3224..4bbffdb9024f 100644
--- a/arch/mips/kernel/cpu-probe.c
+++ b/arch/mips/kernel/cpu-probe.c
@@ -470,6 +470,9 @@ static inline unsigned int decode_config3(struct cpuinfo_mips *c)
 		c->options |= MIPS_CPU_ULRI;
 	if (config3 & MIPS_CONF3_ISA)
 		c->options |= MIPS_CPU_MICROMIPS;
+#ifdef CONFIG_CPU_MICROMIPS
+	write_c0_config3(read_c0_config3() | MIPS_CONF3_ISA_OE);
+#endif
 	if (config3 & MIPS_CONF3_VZ)
 		c->ases |= MIPS_ASE_VZ;
 
diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S
index 5360b1db337d..5c2ba9f08a80 100644
--- a/arch/mips/kernel/genex.S
+++ b/arch/mips/kernel/genex.S
@@ -5,8 +5,8 @@
  *
  * Copyright (C) 1994 - 2000, 2001, 2003 Ralf Baechle
  * Copyright (C) 1999, 2000 Silicon Graphics, Inc.
- * Copyright (C) 2001 MIPS Technologies, Inc.
  * Copyright (C) 2002, 2007  Maciej W. Rozycki
+ * Copyright (C) 2001, 2012 MIPS Technologies, Inc.  All rights reserved.
  */
 #include <linux/init.h>
 
@@ -21,8 +21,10 @@
 #include <asm/war.h>
 #include <asm/thread_info.h>
 
+#ifdef CONFIG_MIPS_MT_SMTC
 #define PANIC_PIC(msg)					\
-		.set push;				\
+		.set	push;				\
+		.set	nomicromips;			\
 		.set	reorder;			\
 		PTR_LA	a0,8f;				\
 		.set	noat;				\
@@ -31,17 +33,10 @@
 9:		b	9b;				\
 		.set	pop;				\
 		TEXT(msg)
+#endif
 
 	__INIT
 
-NESTED(except_vec0_generic, 0, sp)
-	PANIC_PIC("Exception vector 0 called")
-	END(except_vec0_generic)
-
-NESTED(except_vec1_generic, 0, sp)
-	PANIC_PIC("Exception vector 1 called")
-	END(except_vec1_generic)
-
 /*
  * General exception vector for all other CPUs.
  *
@@ -138,12 +133,19 @@ LEAF(r4k_wait)
 	 nop
 	nop
 	nop
+#ifdef CONFIG_CPU_MICROMIPS
+	nop
+	nop
+	nop
+	nop
+#endif
 	.set	mips3
 	wait
 	/* end of rollback region (the region size must be power of two) */
-	.set	pop
 1:
 	jr	ra
+	nop
+	.set	pop
 	END(r4k_wait)
 
 	.macro	BUILD_ROLLBACK_PROLOGUE handler
@@ -201,7 +203,11 @@ NESTED(handle_int, PT_SIZE, sp)
 	LONG_L	s0, TI_REGS($28)
 	LONG_S	sp, TI_REGS($28)
 	PTR_LA	ra, ret_from_irq
-	j	plat_irq_dispatch
+	PTR_LA  v0, plat_irq_dispatch
+	jr	v0
+#ifdef CONFIG_CPU_MICROMIPS
+	nop
+#endif
 	END(handle_int)
 
 	__INIT
@@ -222,11 +228,14 @@ NESTED(except_vec4, 0, sp)
 /*
  * EJTAG debug exception handler.
  * The EJTAG debug exception entry point is 0xbfc00480, which
- * normally is in the boot PROM, so the boot PROM must do a
+ * normally is in the boot PROM, so the boot PROM must do an
  * unconditional jump to this vector.
  */
 NESTED(except_vec_ejtag_debug, 0, sp)
 	j	ejtag_debug_handler
+#ifdef CONFIG_CPU_MICROMIPS
+	 nop
+#endif
 	END(except_vec_ejtag_debug)
 
 	__FINIT
@@ -251,9 +260,10 @@ NESTED(except_vec_vi, 0, sp)
 FEXPORT(except_vec_vi_mori)
 	ori	a0, $0, 0
 #endif /* CONFIG_MIPS_MT_SMTC */
+	PTR_LA	v1, except_vec_vi_handler
 FEXPORT(except_vec_vi_lui)
 	lui	v0, 0		/* Patched */
-	j	except_vec_vi_handler
+	jr	v1
 FEXPORT(except_vec_vi_ori)
 	 ori	v0, 0		/* Patched */
 	.set	pop
@@ -354,6 +364,9 @@ EXPORT(ejtag_debug_buffer)
  */
 NESTED(except_vec_nmi, 0, sp)
 	j	nmi_handler
+#ifdef CONFIG_CPU_MICROMIPS
+	 nop
+#endif
 	END(except_vec_nmi)
 
 	__FINIT
@@ -500,13 +513,35 @@ NESTED(nmi_handler, PT_SIZE, sp)
 	.set	push
 	.set	noat
 	.set	noreorder
-	/* 0x7c03e83b: rdhwr v1,$29 */
+	/* MIPS32:    0x7c03e83b: rdhwr v1,$29 */
+	/* microMIPS: 0x007d6b3c: rdhwr v1,$29 */
 	MFC0	k1, CP0_EPC
-	lui	k0, 0x7c03
-	lw	k1, (k1)
-	ori	k0, 0xe83b
-	.set	reorder
+#if defined(CONFIG_CPU_MICROMIPS) || defined(CONFIG_CPU_MIPS32_R2) || defined(CONFIG_CPU_MIPS64_R2)
+	and     k0, k1, 1
+	beqz    k0, 1f
+	xor     k1, k0
+	lhu     k0, (k1)
+	lhu     k1, 2(k1)
+	ins     k1, k0, 16, 16
+	lui     k0, 0x007d
+	b       docheck
+	ori     k0, 0x6b3c
+1:
+	lui     k0, 0x7c03
+	lw      k1, (k1)
+	ori     k0, 0xe83b
+#else
+	andi    k0, k1, 1
+	bnez    k0, handle_ri
+	lui     k0, 0x7c03
+	lw      k1, (k1)
+	ori     k0, 0xe83b
+#endif
+	.set    reorder
+docheck:
 	bne	k0, k1, handle_ri	/* if not ours */
+
+isrdhwr:
 	/* The insn is rdhwr.  No need to check CAUSE.BD here. */
 	get_saved_sp	/* k1 := current_thread_info */
 	.set	noreorder
diff --git a/arch/mips/kernel/scall32-o32.S b/arch/mips/kernel/scall32-o32.S
index 9ea29649fc28..9b36424b03c5 100644
--- a/arch/mips/kernel/scall32-o32.S
+++ b/arch/mips/kernel/scall32-o32.S
@@ -138,9 +138,18 @@ stackargs:
 5:	jr	t1
 	 sw	t5, 16(sp)		# argument #5 to ksp
 
+#ifdef CONFIG_CPU_MICROMIPS
 	sw	t8, 28(sp)		# argument #8 to ksp
+	nop
 	sw	t7, 24(sp)		# argument #7 to ksp
+	nop
 	sw	t6, 20(sp)		# argument #6 to ksp
+	nop
+#else
+	sw	t8, 28(sp)		# argument #8 to ksp
+	sw	t7, 24(sp)		# argument #7 to ksp
+	sw	t6, 20(sp)		# argument #6 to ksp
+#endif
 6:	j	stack_done		# go back
 	 nop
 	.set	pop
diff --git a/arch/mips/kernel/smtc-asm.S b/arch/mips/kernel/smtc-asm.S
index 76016ac0a9c8..2866863a39df 100644
--- a/arch/mips/kernel/smtc-asm.S
+++ b/arch/mips/kernel/smtc-asm.S
@@ -49,6 +49,9 @@ CAN WE PROVE THAT WE WON'T DO THIS IF INTS DISABLED??
 	.text
 	.align 5
 FEXPORT(__smtc_ipi_vector)
+#ifdef CONFIG_CPU_MICROMIPS
+	nop
+#endif
 	.set	noat
 	/* Disable thread scheduling to make Status update atomic */
 	DMT	27					# dmt	k1
diff --git a/arch/mips/kernel/traps.c b/arch/mips/kernel/traps.c
index 333782b83164..571a69c57d82 100644
--- a/arch/mips/kernel/traps.c
+++ b/arch/mips/kernel/traps.c
@@ -8,8 +8,8 @@
  * Copyright (C) 1998 Ulf Carlsson
  * Copyright (C) 1999 Silicon Graphics, Inc.
  * Kevin D. Kissell, kevink@mips.com and Carsten Langgaard, carstenl@mips.com
- * Copyright (C) 2000, 01 MIPS Technologies, Inc.
  * Copyright (C) 2002, 2003, 2004, 2005, 2007  Maciej W. Rozycki
+ * Copyright (C) 2000, 2001, 2012 MIPS Technologies, Inc.  All rights reserved.
  */
 #include <linux/bug.h>
 #include <linux/compiler.h>
@@ -83,10 +83,6 @@ extern asmlinkage void handle_dsp(void);
 extern asmlinkage void handle_mcheck(void);
 extern asmlinkage void handle_reserved(void);
 
-extern int fpu_emulator_cop1Handler(struct pt_regs *xcp,
-				    struct mips_fpu_struct *ctx, int has_fpu,
-				    void *__user *fault_addr);
-
 void (*board_be_init)(void);
 int (*board_be_handler)(struct pt_regs *regs, int is_fixup);
 void (*board_nmi_handler_setup)(void);
@@ -495,6 +491,12 @@ asmlinkage void do_be(struct pt_regs *regs)
 #define SYNC   0x0000000f
 #define RDHWR  0x0000003b
 
+/*  microMIPS definitions   */
+#define MM_POOL32A_FUNC 0xfc00ffff
+#define MM_RDHWR        0x00006b3c
+#define MM_RS           0x001f0000
+#define MM_RT           0x03e00000
+
 /*
  * The ll_bit is cleared by r*_switch.S
  */
@@ -609,42 +611,62 @@ static int simulate_llsc(struct pt_regs *regs, unsigned int opcode)
  * Simulate trapping 'rdhwr' instructions to provide user accessible
  * registers not implemented in hardware.
  */
-static int simulate_rdhwr(struct pt_regs *regs, unsigned int opcode)
+static int simulate_rdhwr(struct pt_regs *regs, int rd, int rt)
 {
 	struct thread_info *ti = task_thread_info(current);
 
+	perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS,
+			1, regs, 0);
+	switch (rd) {
+	case 0:		/* CPU number */
+		regs->regs[rt] = smp_processor_id();
+		return 0;
+	case 1:		/* SYNCI length */
+		regs->regs[rt] = min(current_cpu_data.dcache.linesz,
+				     current_cpu_data.icache.linesz);
+		return 0;
+	case 2:		/* Read count register */
+		regs->regs[rt] = read_c0_count();
+		return 0;
+	case 3:		/* Count register resolution */
+		switch (current_cpu_data.cputype) {
+		case CPU_20KC:
+		case CPU_25KF:
+			regs->regs[rt] = 1;
+			break;
+		default:
+			regs->regs[rt] = 2;
+		}
+		return 0;
+	case 29:
+		regs->regs[rt] = ti->tp_value;
+		return 0;
+	default:
+		return -1;
+	}
+}
+
+static int simulate_rdhwr_normal(struct pt_regs *regs, unsigned int opcode)
+{
 	if ((opcode & OPCODE) == SPEC3 && (opcode & FUNC) == RDHWR) {
 		int rd = (opcode & RD) >> 11;
 		int rt = (opcode & RT) >> 16;
-		perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS,
-				1, regs, 0);
-		switch (rd) {
-		case 0:		/* CPU number */
-			regs->regs[rt] = smp_processor_id();
-			return 0;
-		case 1:		/* SYNCI length */
-			regs->regs[rt] = min(current_cpu_data.dcache.linesz,
-					     current_cpu_data.icache.linesz);
-			return 0;
-		case 2:		/* Read count register */
-			regs->regs[rt] = read_c0_count();
-			return 0;
-		case 3:		/* Count register resolution */
-			switch (current_cpu_data.cputype) {
-			case CPU_20KC:
-			case CPU_25KF:
-				regs->regs[rt] = 1;
-				break;
-			default:
-				regs->regs[rt] = 2;
-			}
-			return 0;
-		case 29:
-			regs->regs[rt] = ti->tp_value;
-			return 0;
-		default:
-			return -1;
-		}
+
+		simulate_rdhwr(regs, rd, rt);
+		return 0;
+	}
+
+	/* Not ours.  */
+	return -1;
+}
+
+static int simulate_rdhwr_mm(struct pt_regs *regs, unsigned short opcode)
+{
+	if ((opcode & MM_POOL32A_FUNC) == MM_RDHWR) {
+		int rd = (opcode & MM_RS) >> 16;
+		int rt = (opcode & MM_RT) >> 21;
+		simulate_rdhwr(regs, rd, rt);
+		return 0;
 	}
 
 	/* Not ours.  */
@@ -826,9 +848,29 @@ static void do_trap_or_bp(struct pt_regs *regs, unsigned int code,
 asmlinkage void do_bp(struct pt_regs *regs)
 {
 	unsigned int opcode, bcode;
-
-	if (__get_user(opcode, (unsigned int __user *) exception_epc(regs)))
-		goto out_sigsegv;
+	unsigned long epc;
+	u16 instr[2];
+
+	if (get_isa16_mode(regs->cp0_epc)) {
+		/* Calculate EPC. */
+		epc = exception_epc(regs);
+		if (cpu_has_mmips) {
+			if ((__get_user(instr[0], (u16 __user *)msk_isa16_mode(epc)) ||
+			    (__get_user(instr[1], (u16 __user *)msk_isa16_mode(epc + 2)))))
+				goto out_sigsegv;
+		    opcode = (instr[0] << 16) | instr[1];
+		} else {
+		    /* MIPS16e mode */
+		    if (__get_user(instr[0], (u16 __user *)msk_isa16_mode(epc)))
+				goto out_sigsegv;
+		    bcode = (instr[0] >> 6) & 0x3f;
+		    do_trap_or_bp(regs, bcode, "Break");
+		    return;
+		}
+	} else {
+		if (__get_user(opcode, (unsigned int __user *) exception_epc(regs)))
+			goto out_sigsegv;
+	}
 
 	/*
 	 * There is the ancient bug in the MIPS assemblers that the break
@@ -869,13 +911,22 @@ out_sigsegv:
 asmlinkage void do_tr(struct pt_regs *regs)
 {
 	unsigned int opcode, tcode = 0;
+	u16 instr[2];
+	unsigned long epc = exception_epc(regs);
 
-	if (__get_user(opcode, (unsigned int __user *) exception_epc(regs)))
-		goto out_sigsegv;
+	if ((__get_user(instr[0], (u16 __user *)msk_isa16_mode(epc))) ||
+		(__get_user(instr[1], (u16 __user *)msk_isa16_mode(epc + 2))))
+			goto out_sigsegv;
+	opcode = (instr[0] << 16) | instr[1];
 
 	/* Immediate versions don't provide a code.  */
-	if (!(opcode & OPCODE))
-		tcode = ((opcode >> 6) & ((1 << 10) - 1));
+	if (!(opcode & OPCODE)) {
+		if (get_isa16_mode(regs->cp0_epc))
+			/* microMIPS */
+			tcode = (opcode >> 12) & 0x1f;
+		else
+			tcode = ((opcode >> 6) & ((1 << 10) - 1));
+	}
 
 	do_trap_or_bp(regs, tcode, "Trap");
 	return;
@@ -888,6 +939,7 @@ asmlinkage void do_ri(struct pt_regs *regs)
 {
 	unsigned int __user *epc = (unsigned int __user *)exception_epc(regs);
 	unsigned long old_epc = regs->cp0_epc;
+	unsigned long old31 = regs->regs[31];
 	unsigned int opcode = 0;
 	int status = -1;
 
@@ -900,23 +952,37 @@ asmlinkage void do_ri(struct pt_regs *regs)
 	if (unlikely(compute_return_epc(regs) < 0))
 		return;
 
-	if (unlikely(get_user(opcode, epc) < 0))
-		status = SIGSEGV;
+	if (get_isa16_mode(regs->cp0_epc)) {
+		unsigned short mmop[2] = { 0 };
 
-	if (!cpu_has_llsc && status < 0)
-		status = simulate_llsc(regs, opcode);
+		if (unlikely(get_user(mmop[0], epc) < 0))
+			status = SIGSEGV;
+		if (unlikely(get_user(mmop[1], epc) < 0))
+			status = SIGSEGV;
+		opcode = (mmop[0] << 16) | mmop[1];
 
-	if (status < 0)
-		status = simulate_rdhwr(regs, opcode);
+		if (status < 0)
+			status = simulate_rdhwr_mm(regs, opcode);
+	} else {
+		if (unlikely(get_user(opcode, epc) < 0))
+			status = SIGSEGV;
 
-	if (status < 0)
-		status = simulate_sync(regs, opcode);
+		if (!cpu_has_llsc && status < 0)
+			status = simulate_llsc(regs, opcode);
+
+		if (status < 0)
+			status = simulate_rdhwr_normal(regs, opcode);
+
+		if (status < 0)
+			status = simulate_sync(regs, opcode);
+	}
 
 	if (status < 0)
 		status = SIGILL;
 
 	if (unlikely(status > 0)) {
 		regs->cp0_epc = old_epc;		/* Undo skip-over.  */
+		regs->regs[31] = old31;
 		force_sig(status, current);
 	}
 }
@@ -986,7 +1052,7 @@ static int default_cu2_call(struct notifier_block *nfb, unsigned long action,
 asmlinkage void do_cpu(struct pt_regs *regs)
 {
 	unsigned int __user *epc;
-	unsigned long old_epc;
+	unsigned long old_epc, old31;
 	unsigned int opcode;
 	unsigned int cpid;
 	int status;
@@ -1000,26 +1066,41 @@ asmlinkage void do_cpu(struct pt_regs *regs)
 	case 0:
 		epc = (unsigned int __user *)exception_epc(regs);
 		old_epc = regs->cp0_epc;
+		old31 = regs->regs[31];
 		opcode = 0;
 		status = -1;
 
 		if (unlikely(compute_return_epc(regs) < 0))
 			return;
 
-		if (unlikely(get_user(opcode, epc) < 0))
-			status = SIGSEGV;
+		if (get_isa16_mode(regs->cp0_epc)) {
+			unsigned short mmop[2] = { 0 };
 
-		if (!cpu_has_llsc && status < 0)
-			status = simulate_llsc(regs, opcode);
+			if (unlikely(get_user(mmop[0], epc) < 0))
+				status = SIGSEGV;
+			if (unlikely(get_user(mmop[1], epc) < 0))
+				status = SIGSEGV;
+			opcode = (mmop[0] << 16) | mmop[1];
 
-		if (status < 0)
-			status = simulate_rdhwr(regs, opcode);
+			if (status < 0)
+				status = simulate_rdhwr_mm(regs, opcode);
+		} else {
+			if (unlikely(get_user(opcode, epc) < 0))
+				status = SIGSEGV;
+
+			if (!cpu_has_llsc && status < 0)
+				status = simulate_llsc(regs, opcode);
+
+			if (status < 0)
+				status = simulate_rdhwr_normal(regs, opcode);
+		}
 
 		if (status < 0)
 			status = SIGILL;
 
 		if (unlikely(status > 0)) {
 			regs->cp0_epc = old_epc;	/* Undo skip-over.  */
+			regs->regs[31] = old31;
 			force_sig(status, current);
 		}
 
@@ -1333,7 +1414,7 @@ asmlinkage void cache_parity_error(void)
 void ejtag_exception_handler(struct pt_regs *regs)
 {
 	const int field = 2 * sizeof(unsigned long);
-	unsigned long depc, old_epc;
+	unsigned long depc, old_epc, old_ra;
 	unsigned int debug;
 
 	printk(KERN_DEBUG "SDBBP EJTAG debug exception - not handled yet, just ignored!\n");
@@ -1348,10 +1429,12 @@ void ejtag_exception_handler(struct pt_regs *regs)
 		 * calculation.
 		 */
 		old_epc = regs->cp0_epc;
+		old_ra = regs->regs[31];
 		regs->cp0_epc = depc;
-		__compute_return_epc(regs);
+		compute_return_epc(regs);
 		depc = regs->cp0_epc;
 		regs->cp0_epc = old_epc;
+		regs->regs[31] = old_ra;
 	} else
 		depc += 4;
 	write_c0_depc(depc);
@@ -1392,9 +1475,24 @@ void __init *set_except_vector(int n, void *addr)
 	unsigned long handler = (unsigned long) addr;
 	unsigned long old_handler = exception_handlers[n];
 
+#ifdef CONFIG_CPU_MICROMIPS
+	/*
+	 * Only the TLB handlers are cache aligned with an even
+	 * address. All other handlers are on an odd address and
+	 * require no modification. Otherwise, MIPS32 mode will
+	 * be entered when handling any TLB exceptions. That
+	 * would be bad...since we must stay in microMIPS mode.
+	 */
+	if (!(handler & 0x1))
+		handler |= 1;
+#endif
 	exception_handlers[n] = handler;
 	if (n == 0 && cpu_has_divec) {
+#ifdef CONFIG_CPU_MICROMIPS
+		unsigned long jump_mask = ~((1 << 27) - 1);
+#else
 		unsigned long jump_mask = ~((1 << 28) - 1);
+#endif
 		u32 *buf = (u32 *)(ebase + 0x200);
 		unsigned int k0 = 26;
 		if ((handler & jump_mask) == ((ebase + 0x200) & jump_mask)) {
@@ -1421,17 +1519,18 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs)
 	unsigned long handler;
 	unsigned long old_handler = vi_handlers[n];
 	int srssets = current_cpu_data.srsets;
-	u32 *w;
+	u16 *h;
 	unsigned char *b;
 
 	BUG_ON(!cpu_has_veic && !cpu_has_vint);
+	BUG_ON((n < 0) && (n > 9));
 
 	if (addr == NULL) {
 		handler = (unsigned long) do_default_vi;
 		srs = 0;
 	} else
 		handler = (unsigned long) addr;
-	vi_handlers[n] = (unsigned long) addr;
+	vi_handlers[n] = handler;
 
 	b = (unsigned char *)(ebase + 0x200 + n*VECTORSPACING);
 
@@ -1450,9 +1549,8 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs)
 	if (srs == 0) {
 		/*
 		 * If no shadow set is selected then use the default handler
-		 * that does normal register saving and a standard interrupt exit
+		 * that does normal register saving and standard interrupt exit
 		 */
-
 		extern char except_vec_vi, except_vec_vi_lui;
 		extern char except_vec_vi_ori, except_vec_vi_end;
 		extern char rollback_except_vec_vi;
@@ -1465,11 +1563,20 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs)
 		 * Status.IM bit to be masked before going there.
 		 */
 		extern char except_vec_vi_mori;
+#if defined(CONFIG_CPU_MICROMIPS) || defined(CONFIG_CPU_BIG_ENDIAN)
+		const int mori_offset = &except_vec_vi_mori - vec_start + 2;
+#else
 		const int mori_offset = &except_vec_vi_mori - vec_start;
+#endif
 #endif /* CONFIG_MIPS_MT_SMTC */
-		const int handler_len = &except_vec_vi_end - vec_start;
+#if defined(CONFIG_CPU_MICROMIPS) || defined(CONFIG_CPU_BIG_ENDIAN)
+		const int lui_offset = &except_vec_vi_lui - vec_start + 2;
+		const int ori_offset = &except_vec_vi_ori - vec_start + 2;
+#else
 		const int lui_offset = &except_vec_vi_lui - vec_start;
 		const int ori_offset = &except_vec_vi_ori - vec_start;
+#endif
+		const int handler_len = &except_vec_vi_end - vec_start;
 
 		if (handler_len > VECTORSPACING) {
 			/*
@@ -1479,30 +1586,44 @@ static void *set_vi_srs_handler(int n, vi_handler_t addr, int srs)
 			panic("VECTORSPACING too small");
 		}
 
-		memcpy(b, vec_start, handler_len);
+		set_handler(((unsigned long)b - ebase), vec_start,
+#ifdef CONFIG_CPU_MICROMIPS
+				(handler_len - 1));
+#else
+				handler_len);
+#endif
 #ifdef CONFIG_MIPS_MT_SMTC
 		BUG_ON(n > 7);	/* Vector index %d exceeds SMTC maximum. */
 
-		w = (u32 *)(b + mori_offset);
-		*w = (*w & 0xffff0000) | (0x100 << n);
+		h = (u16 *)(b + mori_offset);
+		*h = (0x100 << n);
 #endif /* CONFIG_MIPS_MT_SMTC */
-		w = (u32 *)(b + lui_offset);
-		*w = (*w & 0xffff0000) | (((u32)handler >> 16) & 0xffff);
-		w = (u32 *)(b + ori_offset);
-		*w = (*w & 0xffff0000) | ((u32)handler & 0xffff);
+		h = (u16 *)(b + lui_offset);
+		*h = (handler >> 16) & 0xffff;
+		h = (u16 *)(b + ori_offset);
+		*h = (handler & 0xffff);
 		local_flush_icache_range((unsigned long)b,
 					 (unsigned long)(b+handler_len));
 	}
 	else {
 		/*
-		 * In other cases jump directly to the interrupt handler
-		 *
-		 * It is the handlers responsibility to save registers if required
-		 * (eg hi/lo) and return from the exception using "eret"
+		 * In other cases jump directly to the interrupt handler. It
+		 * is the handler's responsibility to save registers if required
+		 * (eg hi/lo) and return from the exception using "eret".
 		 */
-		w = (u32 *)b;
-		*w++ = 0x08000000 | (((u32)handler >> 2) & 0x03fffff); /* j handler */
-		*w = 0;
+		u32 insn;
+
+		h = (u16 *)b;
+		/* j handler */
+#ifdef CONFIG_CPU_MICROMIPS
+		insn = 0xd4000000 | (((u32)handler & 0x07ffffff) >> 1);
+#else
+		insn = 0x08000000 | (((u32)handler & 0x0fffffff) >> 2);
+#endif
+		h[0] = (insn >> 16) & 0xffff;
+		h[1] = insn & 0xffff;
+		h[2] = 0;
+		h[3] = 0;
 		local_flush_icache_range((unsigned long)b,
 					 (unsigned long)(b+8));
 	}
@@ -1663,7 +1784,11 @@ void __cpuinit per_cpu_trap_init(bool is_boot_cpu)
 /* Install CPU exception handler */
 void __cpuinit set_handler(unsigned long offset, void *addr, unsigned long size)
 {
+#ifdef CONFIG_CPU_MICROMIPS
+	memcpy((void *)(ebase + offset), ((unsigned char *)addr - 1), size);
+#else
 	memcpy((void *)(ebase + offset), addr, size);
+#endif
 	local_flush_icache_range(ebase + offset, ebase + offset + size);
 }
 
@@ -1697,8 +1822,9 @@ __setup("rdhwr_noopt", set_rdhwr_noopt);
 
 void __init trap_init(void)
 {
-	extern char except_vec3_generic, except_vec3_r4000;
+	extern char except_vec3_generic;
 	extern char except_vec4;
+	extern char except_vec3_r4000;
 	unsigned long i;
 	int rollback;
 
@@ -1831,11 +1957,11 @@ void __init trap_init(void)
 
 	if (cpu_has_vce)
 		/* Special exception: R4[04]00 uses also the divec space. */
-		memcpy((void *)(ebase + 0x180), &except_vec3_r4000, 0x100);
+		set_handler(0x180, &except_vec3_r4000, 0x100);
 	else if (cpu_has_4kex)
-		memcpy((void *)(ebase + 0x180), &except_vec3_generic, 0x80);
+		set_handler(0x180, &except_vec3_generic, 0x80);
 	else
-		memcpy((void *)(ebase + 0x080), &except_vec3_generic, 0x80);
+		set_handler(0x080, &except_vec3_generic, 0x80);
 
 	local_flush_icache_range(ebase, ebase + 0x400);
 	flush_tlb_handlers();