riscv: update misaligned memory access exception handling

Support for more situations: floating point, compressed instructions,
etc. Add support for redirect exception to S-Mode.


Change-Id: I9983d56245eab1d458a84cb1432aeb805df7a49f
Signed-off-by: Xiang Wang <wxjstz@126.com>
Reviewed-on: https://review.coreboot.org/27972
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>

5 files changed, 652 insertions, 68 deletions

diff --git a/src/arch/riscv/Makefile.inc b/src/arch/riscv/Makefile.inc
index ae327f2f169c..f4b69bb95e31 100644
--- a/src/arch/riscv/Makefile.inc
+++ b/src/arch/riscv/Makefile.inc
@@ -43,6 +43,8 @@ ifeq ($(CONFIG_ARCH_BOOTBLOCK_RISCV),y)
 bootblock-y = bootblock.S
 bootblock-y += trap_util.S
 bootblock-y += trap_handler.c
+bootblock-y += fp_asm.S
+bootblock-y += misaligned.c
 bootblock-y += mcall.c
 bootblock-y += virtual_memory.c
 bootblock-y += boot.c
@@ -101,6 +103,8 @@ ramstage-y += ramstage.S
 ramstage-y += mcall.c
 ramstage-y += trap_util.S
 ramstage-y += trap_handler.c
+ramstage-y += fp_asm.S
+ramstage-y += misaligned.c
 ramstage-y += virtual_memory.c
 ramstage-y += stages.c
 ramstage-y += misc.c
diff --git a/src/arch/riscv/fp_asm.S b/src/arch/riscv/fp_asm.S
new file mode 100644
index 000000000000..9c6cc650d842
--- /dev/null
+++ b/src/arch/riscv/fp_asm.S
@@ -0,0 +1,362 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2018 HardenedLinux
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+/*
+ * This file define some function used to swap value between memory
+ * and float register. This will be used in misaligned access exception
+ * handling.
+ */
+
+#if defined(__riscv_flen)
+#if __riscv_flen >= 32
+
+	.text
+
+/* void read_f32(int regnum, uint32_t* v)
+ *    regnum : which register want to read
+ *    v : address to hold bits info of reading
+ */
+	.align	1
+	.globl read_f32
+read_f32:
+	la	a2, .Lr32_t
+	andi	a0, a0, 31
+	slli	a0, a0, 1
+	add	a0, a0, a2
+	lbu	a0, 0(a0)
+	add	a0, a0, a2
+	jr	a0
+	.align	2
+.Lr32_t:
+	.half	.Lr32_f0  - .Lr32_t
+	.half	.Lr32_f1  - .Lr32_t
+	.half	.Lr32_f2  - .Lr32_t
+	.half	.Lr32_f3  - .Lr32_t
+	.half	.Lr32_f4  - .Lr32_t
+	.half	.Lr32_f5  - .Lr32_t
+	.half	.Lr32_f6  - .Lr32_t
+	.half	.Lr32_f7  - .Lr32_t
+	.half	.Lr32_f8  - .Lr32_t
+	.half	.Lr32_f9  - .Lr32_t
+	.half	.Lr32_f10 - .Lr32_t
+	.half	.Lr32_f11 - .Lr32_t
+	.half	.Lr32_f12 - .Lr32_t
+	.half	.Lr32_f13 - .Lr32_t
+	.half	.Lr32_f14 - .Lr32_t
+	.half	.Lr32_f15 - .Lr32_t
+	.half	.Lr32_f16 - .Lr32_t
+	.half	.Lr32_f17 - .Lr32_t
+	.half	.Lr32_f18 - .Lr32_t
+	.half	.Lr32_f19 - .Lr32_t
+	.half	.Lr32_f20 - .Lr32_t
+	.half	.Lr32_f21 - .Lr32_t
+	.half	.Lr32_f22 - .Lr32_t
+	.half	.Lr32_f23 - .Lr32_t
+	.half	.Lr32_f24 - .Lr32_t
+	.half	.Lr32_f25 - .Lr32_t
+	.half	.Lr32_f26 - .Lr32_t
+	.half	.Lr32_f27 - .Lr32_t
+	.half	.Lr32_f28 - .Lr32_t
+	.half	.Lr32_f29 - .Lr32_t
+	.half	.Lr32_f30 - .Lr32_t
+	.half	.Lr32_f31 - .Lr32_t
+#define  read32(which)	.Lr32_##which: fsw which, 0(a1); ret
+	read32(f0)
+	read32(f1)
+	read32(f2)
+	read32(f3)
+	read32(f4)
+	read32(f5)
+	read32(f6)
+	read32(f7)
+	read32(f8)
+	read32(f9)
+	read32(f10)
+	read32(f11)
+	read32(f12)
+	read32(f13)
+	read32(f14)
+	read32(f15)
+	read32(f16)
+	read32(f17)
+	read32(f18)
+	read32(f19)
+	read32(f20)
+	read32(f21)
+	read32(f22)
+	read32(f23)
+	read32(f24)
+	read32(f25)
+	read32(f26)
+	read32(f27)
+	read32(f28)
+	read32(f29)
+	read32(f30)
+	read32(f31)
+
+/* void write_f32(int regnum, uint32_t* v)
+ *    regnum: which register want to write
+ *    v : address to hold bits info of writing
+ */
+	.align	1
+	.globl write_f32
+write_f32:
+	la	a2, .Lw32_t
+	andi	a0, a0, 31
+	slli	a0, a0, 1
+	add	a0, a0, a2
+	lbu	a0, 0(a0)
+	add	a0, a0, a2
+	jr	a0
+	.align	2
+.Lw32_t:
+	.half	.Lw32_f0  - .Lw32_t
+	.half	.Lw32_f1  - .Lw32_t
+	.half	.Lw32_f2  - .Lw32_t
+	.half	.Lw32_f3  - .Lw32_t
+	.half	.Lw32_f4  - .Lw32_t
+	.half	.Lw32_f5  - .Lw32_t
+	.half	.Lw32_f6  - .Lw32_t
+	.half	.Lw32_f7  - .Lw32_t
+	.half	.Lw32_f8  - .Lw32_t
+	.half	.Lw32_f9  - .Lw32_t
+	.half	.Lw32_f10 - .Lw32_t
+	.half	.Lw32_f11 - .Lw32_t
+	.half	.Lw32_f12 - .Lw32_t
+	.half	.Lw32_f13 - .Lw32_t
+	.half	.Lw32_f14 - .Lw32_t
+	.half	.Lw32_f15 - .Lw32_t
+	.half	.Lw32_f16 - .Lw32_t
+	.half	.Lw32_f17 - .Lw32_t
+	.half	.Lw32_f18 - .Lw32_t
+	.half	.Lw32_f19 - .Lw32_t
+	.half	.Lw32_f20 - .Lw32_t
+	.half	.Lw32_f21 - .Lw32_t
+	.half	.Lw32_f22 - .Lw32_t
+	.half	.Lw32_f23 - .Lw32_t
+	.half	.Lw32_f24 - .Lw32_t
+	.half	.Lw32_f25 - .Lw32_t
+	.half	.Lw32_f26 - .Lw32_t
+	.half	.Lw32_f27 - .Lw32_t
+	.half	.Lw32_f28 - .Lw32_t
+	.half	.Lw32_f29 - .Lw32_t
+	.half	.Lw32_f30 - .Lw32_t
+	.half	.Lw32_f31 - .Lw32_t
+#define write32(which)	.Lw32_##which: flw which, 0(a1); ret
+	write32(f0)
+	write32(f1)
+	write32(f2)
+	write32(f3)
+	write32(f4)
+	write32(f5)
+	write32(f6)
+	write32(f7)
+	write32(f8)
+	write32(f9)
+	write32(f10)
+	write32(f11)
+	write32(f12)
+	write32(f13)
+	write32(f14)
+	write32(f15)
+	write32(f16)
+	write32(f17)
+	write32(f18)
+	write32(f19)
+	write32(f20)
+	write32(f21)
+	write32(f22)
+	write32(f23)
+	write32(f24)
+	write32(f25)
+	write32(f26)
+	write32(f27)
+	write32(f28)
+	write32(f29)
+	write32(f30)
+	write32(f31)
+#endif // __riscv_flen >= 32
+
+#if __riscv_flen >= 64
+
+	.text
+
+/* void read_f64(int regnum, uint64_t* v)
+ *    regnum : which register want to read
+ *    v : address to hold bits info of reading
+ */
+	.align	1
+	.globl read_f64
+read_f64:
+	la	a2, .Lr64_t
+	andi	a0, a0, 31
+	slli	a0, a0, 1
+	add	a0, a0, a2
+	lbu	a0, 0(a0)
+	add	a0, a0, a2
+	jr	a0
+	.align	2
+.Lr64_t:
+	.half	.Lr64_f0  - .Lr64_t
+	.half	.Lr64_f1  - .Lr64_t
+	.half	.Lr64_f2  - .Lr64_t
+	.half	.Lr64_f3  - .Lr64_t
+	.half	.Lr64_f4  - .Lr64_t
+	.half	.Lr64_f5  - .Lr64_t
+	.half	.Lr64_f6  - .Lr64_t
+	.half	.Lr64_f7  - .Lr64_t
+	.half	.Lr64_f8  - .Lr64_t
+	.half	.Lr64_f9  - .Lr64_t
+	.half	.Lr64_f10 - .Lr64_t
+	.half	.Lr64_f11 - .Lr64_t
+	.half	.Lr64_f12 - .Lr64_t
+	.half	.Lr64_f13 - .Lr64_t
+	.half	.Lr64_f14 - .Lr64_t
+	.half	.Lr64_f15 - .Lr64_t
+	.half	.Lr64_f16 - .Lr64_t
+	.half	.Lr64_f17 - .Lr64_t
+	.half	.Lr64_f18 - .Lr64_t
+	.half	.Lr64_f19 - .Lr64_t
+	.half	.Lr64_f20 - .Lr64_t
+	.half	.Lr64_f21 - .Lr64_t
+	.half	.Lr64_f22 - .Lr64_t
+	.half	.Lr64_f23 - .Lr64_t
+	.half	.Lr64_f24 - .Lr64_t
+	.half	.Lr64_f25 - .Lr64_t
+	.half	.Lr64_f26 - .Lr64_t
+	.half	.Lr64_f27 - .Lr64_t
+	.half	.Lr64_f28 - .Lr64_t
+	.half	.Lr64_f29 - .Lr64_t
+	.half	.Lr64_f30 - .Lr64_t
+	.half	.Lr64_f31 - .Lr64_t
+#define  read64(which)	.Lr64_##which: fsd which, 0(a1); ret
+	read64(f0)
+	read64(f1)
+	read64(f2)
+	read64(f3)
+	read64(f4)
+	read64(f5)
+	read64(f6)
+	read64(f7)
+	read64(f8)
+	read64(f9)
+	read64(f10)
+	read64(f11)
+	read64(f12)
+	read64(f13)
+	read64(f14)
+	read64(f15)
+	read64(f16)
+	read64(f17)
+	read64(f18)
+	read64(f19)
+	read64(f20)
+	read64(f21)
+	read64(f22)
+	read64(f23)
+	read64(f24)
+	read64(f25)
+	read64(f26)
+	read64(f27)
+	read64(f28)
+	read64(f29)
+	read64(f30)
+	read64(f31)
+
+/* void write_f64(int regnum, uint64_t* v)
+ *    regnum: which register want to write
+ *    v : address to hold bits info of writing
+ */
+	.align	1
+	.globl write_f64
+write_f64:
+	la	a2, .Lw64_t
+	andi	a0, a0, 31
+	slli	a0, a0, 1
+	add	a0, a0, a2
+	lbu	a0, 0(a0)
+	add	a0, a0, a2
+	jr	a0
+	.align	2
+.Lw64_t:
+	.half	.Lw64_f0  - .Lw64_t
+	.half	.Lw64_f1  - .Lw64_t
+	.half	.Lw64_f2  - .Lw64_t
+	.half	.Lw64_f3  - .Lw64_t
+	.half	.Lw64_f4  - .Lw64_t
+	.half	.Lw64_f5  - .Lw64_t
+	.half	.Lw64_f6  - .Lw64_t
+	.half	.Lw64_f7  - .Lw64_t
+	.half	.Lw64_f8  - .Lw64_t
+	.half	.Lw64_f9  - .Lw64_t
+	.half	.Lw64_f10 - .Lw64_t
+	.half	.Lw64_f11 - .Lw64_t
+	.half	.Lw64_f12 - .Lw64_t
+	.half	.Lw64_f13 - .Lw64_t
+	.half	.Lw64_f14 - .Lw64_t
+	.half	.Lw64_f15 - .Lw64_t
+	.half	.Lw64_f16 - .Lw64_t
+	.half	.Lw64_f17 - .Lw64_t
+	.half	.Lw64_f18 - .Lw64_t
+	.half	.Lw64_f19 - .Lw64_t
+	.half	.Lw64_f20 - .Lw64_t
+	.half	.Lw64_f21 - .Lw64_t
+	.half	.Lw64_f22 - .Lw64_t
+	.half	.Lw64_f23 - .Lw64_t
+	.half	.Lw64_f24 - .Lw64_t
+	.half	.Lw64_f25 - .Lw64_t
+	.half	.Lw64_f26 - .Lw64_t
+	.half	.Lw64_f27 - .Lw64_t
+	.half	.Lw64_f28 - .Lw64_t
+	.half	.Lw64_f29 - .Lw64_t
+	.half	.Lw64_f30 - .Lw64_t
+	.half	.Lw64_f31 - .Lw64_t
+#define write64(which)	.Lw64_##which: fld which, 0(a1); ret
+	write64(f0)
+	write64(f1)
+	write64(f2)
+	write64(f3)
+	write64(f4)
+	write64(f5)
+	write64(f6)
+	write64(f7)
+	write64(f8)
+	write64(f9)
+	write64(f10)
+	write64(f11)
+	write64(f12)
+	write64(f13)
+	write64(f14)
+	write64(f15)
+	write64(f16)
+	write64(f17)
+	write64(f18)
+	write64(f19)
+	write64(f20)
+	write64(f21)
+	write64(f22)
+	write64(f23)
+	write64(f24)
+	write64(f25)
+	write64(f26)
+	write64(f27)
+	write64(f28)
+	write64(f29)
+	write64(f30)
+	write64(f31)
+
+#endif // __riscv_flen >= 64
+
+#endif // defined(__riscv_flen)
diff --git a/src/arch/riscv/include/arch/exception.h b/src/arch/riscv/include/arch/exception.h
index fc57b3b55c50..6fbbdf0a89a7 100644
--- a/src/arch/riscv/include/arch/exception.h
+++ b/src/arch/riscv/include/arch/exception.h
@@ -32,8 +32,7 @@
 
 #include <stdint.h>
 
-typedef struct
-{
+typedef struct {
 	uintptr_t gpr[32];
 	uintptr_t status;
 	uintptr_t epc;
@@ -53,9 +52,9 @@ static inline void exception_init(void)
 {
 }
 
-void trap_handler(trapframe* tf);
-void handle_supervisor_call(trapframe* tf);
-void handle_misaligned_load(trapframe *tf);
-void handle_misaligned_store(trapframe *tf);
+void redirect_trap(void);
+void trap_handler(trapframe *tf);
+void handle_supervisor_call(trapframe *tf);
+void handle_misaligned(trapframe *tf);
 
 #endif
diff --git a/src/arch/riscv/misaligned.c b/src/arch/riscv/misaligned.c
new file mode 100644
index 000000000000..cb045b83237d
--- /dev/null
+++ b/src/arch/riscv/misaligned.c
@@ -0,0 +1,267 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2018 HardenedLinux
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include <vm.h>
+#include <arch/exception.h>
+#include <commonlib/helpers.h>
+
+/*  these functions are defined in src/arch/riscv/fp_asm.S */
+#if defined(__riscv_flen)
+#if __riscv_flen >= 32
+extern void  read_f32(int regnum, uint32_t *v);
+extern void write_f32(int regnum, uint32_t *v);
+#endif // __riscv_flen >= 32
+#if __riscv_flen >= 64
+extern void  read_f64(int regnum, uint64_t *v);
+extern void write_f64(int regnum, uint64_t *v);
+#endif // __riscv_flen >= 64
+#endif // defined(__riscv_flen)
+
+/* This union makes it easy to read multibyte types by byte operations. */
+union endian_buf {
+	uint8_t   b[8];
+	uint16_t  h[4];
+	uint32_t  w[2];
+	uint64_t  d[1];
+	uintptr_t v;
+};
+
+/* This struct hold info of load/store instruction */
+struct memory_instruction_info {
+	/* opcode/mask used to identify instruction,
+	 * (instruction_val) & mask == opcode */
+	uint32_t opcode;
+	uint32_t mask;
+	/* reg_shift/reg_mask/reg_addition used to get register number
+	 * ((instruction_val >> reg_shift) & reg_mask) + reg_addition */
+	unsigned int reg_shift;
+	unsigned int reg_mask;
+	unsigned int reg_addition;
+	unsigned int is_fp : 1;   /* mark as a float operation */
+	unsigned int is_load : 1; /* mark as a load operation */
+	unsigned int width : 8;   /* Record the memory width of the operation */
+	unsigned int sign_extend : 1; /* mark need to be sign extended */
+};
+
+static struct memory_instruction_info insn_info[] = {
+#if __riscv_xlen == 128
+	{ 0x00002000, 0x0000e003,  2,  7, 8, 0, 1, 16, 1}, // C.LQ
+#else
+	{ 0x00002000, 0x0000e003,  2,  7, 8, 1, 1,  8, 0}, // C.FLD
+#endif
+	{ 0x00004000, 0x0000e003,  2,  7, 8, 0, 1,  4, 1}, // C.LW
+#if __riscv_xlen == 32
+	{ 0x00006000, 0x0000e003,  2,  7, 8, 1, 1,  4, 0}, // C.FLW
+#else
+	{ 0x00006000, 0x0000e003,  2,  7, 8, 0, 1,  8, 1}, // C.LD
+#endif
+
+#if __riscv_xlen == 128
+	{ 0x0000a000, 0x0000e003,  2,  7, 8, 0, 0, 16, 0}, // C.SQ
+#else
+	{ 0x0000a000, 0x0000e003,  2,  7, 8, 1, 0,  8, 0}, // C.FSD
+#endif
+	{ 0x0000c000, 0x0000e003,  2,  7, 8, 0, 0,  4, 0}, // C.SW
+#if __riscv_xlen == 32
+	{ 0x0000e000, 0x0000e003,  2,  7, 8, 1, 0,  4, 0}, // C.FSW
+#else
+	{ 0x0000e000, 0x0000e003,  2,  7, 8, 0, 0,  8, 0}, // C.SD
+#endif
+
+#if __riscv_xlen == 128
+	{ 0x00002002, 0x0000e003,  7, 15, 0, 0, 1, 16, 1}, // C.LQSP
+#else
+	{ 0x00002002, 0x0000e003,  7, 15, 0, 1, 1,  8, 0}, // C.FLDSP
+#endif
+	{ 0x00004002, 0x0000e003,  7, 15, 0, 0, 1,  4, 1}, // C.LWSP
+#if __riscv_xlen == 32
+	{ 0x00006002, 0x0000e003,  7, 15, 0, 1, 1,  4, 0}, // C.FLWSP
+#else
+	{ 0x00006002, 0x0000e003,  7, 15, 0, 0, 1,  8, 1}, // C.LDSP
+#endif
+
+#if __riscv_xlen == 128
+	{ 0x0000a002, 0x0000e003,  2, 15, 0, 0, 0, 16, 0}, // C.SQSP
+#else
+	{ 0x0000a002, 0x0000e003,  2, 15, 0, 1, 0,  8, 0}, // C.FSDSP
+#endif
+	{ 0x0000c002, 0x0000e003,  2, 15, 0, 0, 0,  4, 0}, // C.SWSP
+#if __riscv_xlen == 32
+	{ 0x0000e002, 0x0000e003,  2, 15, 0, 1, 0,  4, 0}, // C.FSWSP
+#else
+	{ 0x0000e002, 0x0000e003,  2, 15, 0, 0, 0,  8, 0}, // C.SDSP
+#endif
+
+	{ 0x00000003, 0x0000707f,  7, 15, 0, 0, 1,  1, 1}, // LB
+	{ 0x00001003, 0x0000707f,  7, 15, 0, 0, 1,  2, 1}, // LH
+	{ 0x00002003, 0x0000707f,  7, 15, 0, 0, 1,  4, 1}, // LW
+#if __riscv_xlen > 32
+	{ 0x00003003, 0x0000707f,  7, 15, 0, 0, 1,  8, 1}, // LD
+#endif
+	{ 0x00004003, 0x0000707f,  7, 15, 0, 0, 1,  1, 0}, // LBU
+	{ 0x00005003, 0x0000707f,  7, 15, 0, 0, 1,  2, 0}, // LHU
+	{ 0x00006003, 0x0000707f,  7, 15, 0, 0, 1,  4, 0}, // LWU
+
+	{ 0x00000023, 0x0000707f, 20, 15, 0, 0, 0,  1, 0}, // SB
+	{ 0x00001023, 0x0000707f, 20, 15, 0, 0, 0,  2, 0}, // SH
+	{ 0x00002023, 0x0000707f, 20, 15, 0, 0, 0,  4, 0}, // SW
+#if __riscv_xlen > 32
+	{ 0x00003023, 0x0000707f, 20, 15, 0, 0, 0,  8, 0}, // SD
+#endif
+
+#if defined(__riscv_flen)
+#if __riscv_flen >= 32
+	{ 0x00002007, 0x0000707f,  7, 15, 0, 1, 1,  4, 0}, // FLW
+	{ 0x00003007, 0x0000707f,  7, 15, 0, 1, 1,  8, 0}, // FLD
+#endif // __riscv_flen >= 32
+
+#if __riscv_flen >= 64
+	{ 0x00002027, 0x0000707f, 20, 15, 0, 1, 0,  4, 0}, // FSW
+	{ 0x00003027, 0x0000707f, 20, 15, 0, 1, 0,  8, 0}, // FSD
+#endif // __riscv_flen >= 64
+#endif // defined(__riscv_flen)
+};
+
+static struct memory_instruction_info *match_instruction(uintptr_t insn)
+{
+	int i;
+	for (i = 0; i < ARRAY_SIZE(insn_info); i++)
+		if ((insn_info[i].mask & insn) == insn_info[i].opcode)
+			return &(insn_info[i]);
+	return NULL;
+}
+
+
+static int fetch_16bit_instruction(uintptr_t vaddr, uintptr_t *insn)
+{
+	uint16_t ins = mprv_read_mxr_u16((uint16_t *)vaddr);
+	if (EXTRACT_FIELD(ins, 0x3) != 3) {
+		*insn = ins;
+		return 0;
+	}
+	return -1;
+}
+
+static int fetch_32bit_instruction(uintptr_t vaddr, uintptr_t *insn)
+{
+	uint32_t l = (uint32_t)mprv_read_mxr_u16((uint16_t *)vaddr + 0);
+	uint32_t h = (uint32_t)mprv_read_mxr_u16((uint16_t *)vaddr + 2);
+	uint32_t ins = (h << 16) | l;
+	if ((EXTRACT_FIELD(ins, 0x3) == 3) &&
+		(EXTRACT_FIELD(ins, 0x1c) != 0x7)) {
+		*insn = ins;
+		return 0;
+	}
+	return -1;
+}
+
+
+void handle_misaligned(trapframe *tf)
+{
+	uintptr_t insn = 0;
+	union endian_buf buff;
+
+	/* try to fetch 16/32 bits instruction */
+	if (fetch_16bit_instruction(tf->epc, &insn))
+		if (fetch_32bit_instruction(tf->epc, &insn))
+			redirect_trap();
+
+	/* matching instruction */
+	struct memory_instruction_info *match = match_instruction(insn);
+
+	if (!match) {
+		redirect_trap();
+		return;
+	}
+
+	int regnum;
+	regnum = ((insn >> match->reg_shift) & match->reg_mask);
+	regnum = regnum + match->reg_addition;
+	buff.v = 0;
+	if (match->is_load) {
+		/* load operation */
+
+		/* reading from memory by bytes prevents misaligned
+		 * memory access */
+		for (int i = 0; i < match->width; i++) {
+			uint8_t *addr = (uint8_t *)(tf->badvaddr + i);
+			buff.b[i] = mprv_read_u8(addr);
+		}
+
+		/* sign extend for signed integer loading */
+		if (match->sign_extend)
+			if (buff.v >> (8 * match->width - 1))
+				buff.v |= -1 << (8 * match->width);
+
+		/* write to register */
+		if (match->is_fp) {
+			int done = 0;
+#if defined(__riscv_flen)
+#if __riscv_flen >= 32
+			/* single-precision floating-point */
+			if (match->width == 4) {
+				write_f32(regnum, buff.w);
+				done = 1;
+			}
+#endif // __riscv_flen >= 32
+#if __riscv_flen >= 64
+			/* double-precision floating-point */
+			if (match->width == 8) {
+				write_f64(regnum, buff.d);
+				done = 1;
+			}
+#endif // __riscv_flen >= 64
+#endif // defined(__riscv_flen)
+			if (!done)
+				redirect_trap();
+		} else {
+			tf->gpr[regnum] = buff.v;
+		}
+	} else {
+		/* store operation */
+
+		/* reading from register */
+		if (match->is_fp) {
+			int done = 0;
+#if defined(__riscv_flen)
+#if __riscv_flen >= 32
+			if (match->width == 4) {
+				read_f32(regnum, buff.w);
+				done = 1;
+			}
+#endif // __riscv_flen >= 32
+#if __riscv_flen >= 64
+			if (match->width == 8) {
+				read_f64(regnum, buff.d);
+				done = 1;
+			}
+#endif // __riscv_flen >= 64
+#endif // defined(__riscv_flen)
+			if (!done)
+				redirect_trap();
+		} else {
+			buff.v = tf->gpr[regnum];
+		}
+
+		/* writing to memory by bytes prevents misaligned
+		 * memory access */
+		for (int i = 0; i < match->width; i++) {
+			uint8_t *addr = (uint8_t *)(tf->badvaddr + i);
+			mprv_write_u8(addr, buff.b[i]);
+		}
+	}
+}
diff --git a/src/arch/riscv/trap_handler.c b/src/arch/riscv/trap_handler.c
index 7b35c2e7cd4e..772be64f4073 100644
--- a/src/arch/riscv/trap_handler.c
+++ b/src/arch/riscv/trap_handler.c
@@ -166,12 +166,9 @@ void trap_handler(trapframe *tf)
 			print_trap_information(tf);
 			break;
 		case CAUSE_MISALIGNED_LOAD:
-			print_trap_information(tf);
-			handle_misaligned_load(tf);
-			return;
 		case CAUSE_MISALIGNED_STORE:
 			print_trap_information(tf);
-			handle_misaligned_store(tf);
+			handle_misaligned(tf);
 			return;
 		default:
 			printk(BIOS_EMERG, "================================\n");
@@ -184,62 +181,17 @@ void trap_handler(trapframe *tf)
 	die("Can't recover from trap. Halting.\n");
 }
 
-static uint32_t fetch_instruction(uintptr_t vaddr) {
-	printk(BIOS_SPEW, "fetching instruction at 0x%016zx\n", (size_t)vaddr);
-	return mprv_read_u32((uint32_t *) vaddr);
-}
-
-void handle_misaligned_load(trapframe *tf) {
-	printk(BIOS_DEBUG, "Trapframe ptr:      %p\n", tf);
-	uintptr_t faultingInstructionAddr = tf->epc;
-	insn_t faultingInstruction = fetch_instruction(faultingInstructionAddr);
-	printk(BIOS_DEBUG, "Faulting instruction: 0x%x\n", faultingInstruction);
-	insn_t widthMask = 0x7000;
-	insn_t memWidth = (faultingInstruction & widthMask) >> 12;
-	insn_t destMask = 0xF80;
-	insn_t destRegister = (faultingInstruction & destMask) >> 7;
-	printk(BIOS_DEBUG, "Width:              %d bits\n", (1 << memWidth) * 8);
-	if (memWidth == 3) {
-		// load double, handle the issue
-		void* badAddress = (void*) tf->badvaddr;
-		uint64_t value = 0;
-		for (int i = 0; i < 8; i++) {
-			value <<= 8;
-			value += mprv_read_u8(badAddress+i);
-		}
-		tf->gpr[destRegister] = value;
-	} else {
-		// panic, this should not have happened
-		die("Code should not reach this path, misaligned on a non-64 bit store/load\n");
-	}
-
-	// return to where we came from
-	write_csr(mepc, read_csr(mepc) + 4);
-}
-
-void handle_misaligned_store(trapframe *tf) {
-	printk(BIOS_DEBUG, "Trapframe ptr:      %p\n", tf);
-	uintptr_t faultingInstructionAddr = tf->epc;
-	insn_t faultingInstruction = fetch_instruction(faultingInstructionAddr);
-	printk(BIOS_DEBUG, "Faulting instruction: 0x%x\n", faultingInstruction);
-	insn_t widthMask = 0x7000;
-	insn_t memWidth = (faultingInstruction & widthMask) >> 12;
-	insn_t srcMask = 0x1F00000;
-	insn_t srcRegister = (faultingInstruction & srcMask) >> 20;
-	printk(BIOS_DEBUG, "Width:              %d bits\n", (1 << memWidth) * 8);
-	if (memWidth == 3) {
-		// store double, handle the issue
-		void* badAddress = (void*) tf->badvaddr;
-		uint64_t value = tf->gpr[srcRegister];
-		for (int i = 0; i < 8; i++) {
-			mprv_write_u8(badAddress+i, value);
-			value >>= 8;
-		}
-	} else {
-		// panic, this should not have happened
-		die("Code should not reach this path, misaligned on a non-64 bit store/load\n");
-	}
-
-	// return to where we came from
-	write_csr(mepc, read_csr(mepc) + 4);
+/* This function used to redirect trap to s-mode. */
+void redirect_trap(void)
+{
+	write_csr(sbadaddr, read_csr(mbadaddr));
+	write_csr(sepc, read_csr(mepc));
+	write_csr(scause, read_csr(mcause));
+	write_csr(mepc, read_csr(stvec));
+
+	uintptr_t status = read_csr(mstatus);
+	uintptr_t mpp = EXTRACT_FIELD(status, MSTATUS_MPP);
+	status = INSERT_FIELD(status, MSTATUS_MPP, 1);
+	status = INSERT_FIELD(status, MSTATUS_SPP, mpp & 1);
+	write_csr(mstatus, status);
 }