x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

25 files changed, 2257 insertions, 0 deletions

diff --git a/arch/csky/abiv2/Makefile b/arch/csky/abiv2/Makefile
new file mode 100644
index 000000000..ea8005fe0
--- /dev/null
+++ b/arch/csky/abiv2/Makefile
@@ -0,0 +1,14 @@
+# SPDX-License-Identifier: GPL-2.0-only
+obj-y				+= cacheflush.o
+obj-$(CONFIG_CPU_HAS_FPU)	+= fpu.o
+obj-y				+= memcmp.o
+ifeq ($(CONFIG_HAVE_EFFICIENT_UNALIGNED_STRING_OPS), y)
+obj-y				+= memcpy.o
+obj-y				+= memmove.o
+obj-y				+= memset.o
+endif
+obj-y				+= strcmp.o
+obj-y				+= strcpy.o
+obj-y				+= strlen.o
+obj-y				+= strksyms.o
+obj-$(CONFIG_FUNCTION_TRACER)	+= mcount.o
diff --git a/arch/csky/abiv2/cacheflush.c b/arch/csky/abiv2/cacheflush.c
new file mode 100644
index 000000000..064b0f0f9
--- /dev/null
+++ b/arch/csky/abiv2/cacheflush.c
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/cache.h>
+#include <linux/highmem.h>
+#include <linux/mm.h>
+#include <asm/cache.h>
+#include <asm/tlbflush.h>
+
+void update_mmu_cache_range(struct vm_fault *vmf, struct vm_area_struct *vma,
+		unsigned long address, pte_t *pte, unsigned int nr)
+{
+	unsigned long pfn = pte_pfn(*pte);
+	struct folio *folio;
+	unsigned int i;
+
+	flush_tlb_page(vma, address);
+
+	if (!pfn_valid(pfn))
+		return;
+
+	folio = page_folio(pfn_to_page(pfn));
+
+	if (test_and_set_bit(PG_dcache_clean, &folio->flags.f))
+		return;
+
+	icache_inv_range(address, address + nr*PAGE_SIZE);
+	for (i = 0; i < folio_nr_pages(folio); i++) {
+		unsigned long addr = (unsigned long) kmap_local_folio(folio,
+								i * PAGE_SIZE);
+
+		dcache_wb_range(addr, addr + PAGE_SIZE);
+		if (vma->vm_flags & VM_EXEC)
+			icache_inv_range(addr, addr + PAGE_SIZE);
+		kunmap_local((void *) addr);
+	}
+}
+
+void flush_icache_deferred(struct mm_struct *mm)
+{
+	unsigned int cpu = smp_processor_id();
+	cpumask_t *mask = &mm->context.icache_stale_mask;
+
+	if (cpumask_test_cpu(cpu, mask)) {
+		cpumask_clear_cpu(cpu, mask);
+		/*
+		 * Ensure the remote hart's writes are visible to this hart.
+		 * This pairs with a barrier in flush_icache_mm.
+		 */
+		smp_mb();
+		local_icache_inv_all(NULL);
+	}
+}
+
+void flush_icache_mm_range(struct mm_struct *mm,
+		unsigned long start, unsigned long end)
+{
+	unsigned int cpu;
+	cpumask_t others, *mask;
+
+	preempt_disable();
+
+#ifdef CONFIG_CPU_HAS_ICACHE_INS
+	if (mm == current->mm) {
+		icache_inv_range(start, end);
+		preempt_enable();
+		return;
+	}
+#endif
+
+	/* Mark every hart's icache as needing a flush for this MM. */
+	mask = &mm->context.icache_stale_mask;
+	cpumask_setall(mask);
+
+	/* Flush this hart's I$ now, and mark it as flushed. */
+	cpu = smp_processor_id();
+	cpumask_clear_cpu(cpu, mask);
+	local_icache_inv_all(NULL);
+
+	/*
+	 * Flush the I$ of other harts concurrently executing, and mark them as
+	 * flushed.
+	 */
+	cpumask_andnot(&others, mm_cpumask(mm), cpumask_of(cpu));
+
+	if (mm != current->active_mm || !cpumask_empty(&others)) {
+		on_each_cpu_mask(&others, local_icache_inv_all, NULL, 1);
+		cpumask_clear(mask);
+	}
+
+	preempt_enable();
+}
diff --git a/arch/csky/abiv2/fpu.c b/arch/csky/abiv2/fpu.c
new file mode 100644
index 000000000..5acc5c2e5
--- /dev/null
+++ b/arch/csky/abiv2/fpu.c
@@ -0,0 +1,270 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/ptrace.h>
+#include <linux/uaccess.h>
+#include <abi/reg_ops.h>
+
+#define MTCR_MASK	0xFC00FFE0
+#define MFCR_MASK	0xFC00FFE0
+#define MTCR_DIST	0xC0006420
+#define MFCR_DIST	0xC0006020
+
+/*
+ * fpu_libc_helper() is to help libc to excute:
+ *  - mfcr %a, cr<1, 2>
+ *  - mfcr %a, cr<2, 2>
+ *  - mtcr %a, cr<1, 2>
+ *  - mtcr %a, cr<2, 2>
+ */
+int fpu_libc_helper(struct pt_regs *regs)
+{
+	int fault;
+	unsigned long instrptr, regx = 0;
+	unsigned long index = 0, tmp = 0;
+	unsigned long tinstr = 0;
+	u16 instr_hi, instr_low;
+
+	instrptr = instruction_pointer(regs);
+	if (instrptr & 1)
+		return 0;
+
+	fault = __get_user(instr_low, (u16 *)instrptr);
+	if (fault)
+		return 0;
+
+	fault = __get_user(instr_hi, (u16 *)(instrptr + 2));
+	if (fault)
+		return 0;
+
+	tinstr = instr_hi | ((unsigned long)instr_low << 16);
+
+	if (((tinstr >> 21) & 0x1F) != 2)
+		return 0;
+
+	if ((tinstr & MTCR_MASK) == MTCR_DIST) {
+		index = (tinstr >> 16) & 0x1F;
+		if (index > 13)
+			return 0;
+
+		tmp = tinstr & 0x1F;
+		if (tmp > 2)
+			return 0;
+
+		regx =  *(&regs->a0 + index);
+
+		if (tmp == 1)
+			mtcr("cr<1, 2>", regx);
+		else if (tmp == 2)
+			mtcr("cr<2, 2>", regx);
+		else
+			return 0;
+
+		regs->pc += 4;
+		return 1;
+	}
+
+	if ((tinstr & MFCR_MASK) == MFCR_DIST) {
+		index = tinstr & 0x1F;
+		if (index > 13)
+			return 0;
+
+		tmp = ((tinstr >> 16) & 0x1F);
+		if (tmp > 2)
+			return 0;
+
+		if (tmp == 1)
+			regx = mfcr("cr<1, 2>");
+		else if (tmp == 2)
+			regx = mfcr("cr<2, 2>");
+		else
+			return 0;
+
+		*(&regs->a0 + index) = regx;
+
+		regs->pc += 4;
+		return 1;
+	}
+
+	return 0;
+}
+
+void fpu_fpe(struct pt_regs *regs)
+{
+	int sig, code;
+	unsigned int fesr;
+
+	fesr = mfcr("cr<2, 2>");
+
+	sig = SIGFPE;
+	code = FPE_FLTUNK;
+
+	if (fesr & FPE_ILLE) {
+		sig = SIGILL;
+		code = ILL_ILLOPC;
+	} else if (fesr & FPE_IDC) {
+		sig = SIGILL;
+		code = ILL_ILLOPN;
+	} else if (fesr & FPE_FEC) {
+		sig = SIGFPE;
+		if (fesr & FPE_IOC)
+			code = FPE_FLTINV;
+		else if (fesr & FPE_DZC)
+			code = FPE_FLTDIV;
+		else if (fesr & FPE_UFC)
+			code = FPE_FLTUND;
+		else if (fesr & FPE_OFC)
+			code = FPE_FLTOVF;
+		else if (fesr & FPE_IXC)
+			code = FPE_FLTRES;
+	}
+
+	force_sig_fault(sig, code, (void __user *)regs->pc);
+}
+
+#define FMFVR_FPU_REGS(vrx, vry)	\
+	"fmfvrl %0, "#vrx"\n"		\
+	"fmfvrh %1, "#vrx"\n"		\
+	"fmfvrl %2, "#vry"\n"		\
+	"fmfvrh %3, "#vry"\n"
+
+#define FMTVR_FPU_REGS(vrx, vry)	\
+	"fmtvrl "#vrx", %0\n"		\
+	"fmtvrh "#vrx", %1\n"		\
+	"fmtvrl "#vry", %2\n"		\
+	"fmtvrh "#vry", %3\n"
+
+#define STW_FPU_REGS(a, b, c, d)	\
+	"stw    %0, (%4, "#a")\n"	\
+	"stw    %1, (%4, "#b")\n"	\
+	"stw    %2, (%4, "#c")\n"	\
+	"stw    %3, (%4, "#d")\n"
+
+#define LDW_FPU_REGS(a, b, c, d)	\
+	"ldw    %0, (%4, "#a")\n"	\
+	"ldw    %1, (%4, "#b")\n"	\
+	"ldw    %2, (%4, "#c")\n"	\
+	"ldw    %3, (%4, "#d")\n"
+
+void save_to_user_fp(struct user_fp *user_fp)
+{
+	unsigned long flg;
+	unsigned long tmp1, tmp2;
+	unsigned long *fpregs;
+
+	local_irq_save(flg);
+
+	tmp1 = mfcr("cr<1, 2>");
+	tmp2 = mfcr("cr<2, 2>");
+
+	user_fp->fcr = tmp1;
+	user_fp->fesr = tmp2;
+
+	fpregs = &user_fp->vr[0];
+#ifdef CONFIG_CPU_HAS_FPUV2
+#ifdef CONFIG_CPU_HAS_VDSP
+	asm volatile(
+		"vstmu.32    vr0-vr3,   (%0)\n"
+		"vstmu.32    vr4-vr7,   (%0)\n"
+		"vstmu.32    vr8-vr11,  (%0)\n"
+		"vstmu.32    vr12-vr15, (%0)\n"
+		"fstmu.64    vr16-vr31, (%0)\n"
+		: "+a"(fpregs)
+		::"memory");
+#else
+	asm volatile(
+		"fstmu.64    vr0-vr31,  (%0)\n"
+		: "+a"(fpregs)
+		::"memory");
+#endif
+#else
+	{
+	unsigned long tmp3, tmp4;
+
+	asm volatile(
+		FMFVR_FPU_REGS(vr0, vr1)
+		STW_FPU_REGS(0, 4, 16, 20)
+		FMFVR_FPU_REGS(vr2, vr3)
+		STW_FPU_REGS(32, 36, 48, 52)
+		FMFVR_FPU_REGS(vr4, vr5)
+		STW_FPU_REGS(64, 68, 80, 84)
+		FMFVR_FPU_REGS(vr6, vr7)
+		STW_FPU_REGS(96, 100, 112, 116)
+		"addi	%4, 128\n"
+		FMFVR_FPU_REGS(vr8, vr9)
+		STW_FPU_REGS(0, 4, 16, 20)
+		FMFVR_FPU_REGS(vr10, vr11)
+		STW_FPU_REGS(32, 36, 48, 52)
+		FMFVR_FPU_REGS(vr12, vr13)
+		STW_FPU_REGS(64, 68, 80, 84)
+		FMFVR_FPU_REGS(vr14, vr15)
+		STW_FPU_REGS(96, 100, 112, 116)
+		: "=a"(tmp1), "=a"(tmp2), "=a"(tmp3),
+		  "=a"(tmp4), "+a"(fpregs)
+		::"memory");
+	}
+#endif
+
+	local_irq_restore(flg);
+}
+
+void restore_from_user_fp(struct user_fp *user_fp)
+{
+	unsigned long flg;
+	unsigned long tmp1, tmp2;
+	unsigned long *fpregs;
+
+	local_irq_save(flg);
+
+	tmp1 = user_fp->fcr;
+	tmp2 = user_fp->fesr;
+
+	mtcr("cr<1, 2>", tmp1);
+	mtcr("cr<2, 2>", tmp2);
+
+	fpregs = &user_fp->vr[0];
+#ifdef CONFIG_CPU_HAS_FPUV2
+#ifdef CONFIG_CPU_HAS_VDSP
+	asm volatile(
+		"vldmu.32    vr0-vr3,   (%0)\n"
+		"vldmu.32    vr4-vr7,   (%0)\n"
+		"vldmu.32    vr8-vr11,  (%0)\n"
+		"vldmu.32    vr12-vr15, (%0)\n"
+		"fldmu.64    vr16-vr31, (%0)\n"
+		: "+a"(fpregs)
+		::"memory");
+#else
+	asm volatile(
+		"fldmu.64    vr0-vr31,  (%0)\n"
+		: "+a"(fpregs)
+		::"memory");
+#endif
+#else
+	{
+	unsigned long tmp3, tmp4;
+
+	asm volatile(
+		LDW_FPU_REGS(0, 4, 16, 20)
+		FMTVR_FPU_REGS(vr0, vr1)
+		LDW_FPU_REGS(32, 36, 48, 52)
+		FMTVR_FPU_REGS(vr2, vr3)
+		LDW_FPU_REGS(64, 68, 80, 84)
+		FMTVR_FPU_REGS(vr4, vr5)
+		LDW_FPU_REGS(96, 100, 112, 116)
+		FMTVR_FPU_REGS(vr6, vr7)
+		"addi	%4, 128\n"
+		LDW_FPU_REGS(0, 4, 16, 20)
+		FMTVR_FPU_REGS(vr8, vr9)
+		LDW_FPU_REGS(32, 36, 48, 52)
+		FMTVR_FPU_REGS(vr10, vr11)
+		LDW_FPU_REGS(64, 68, 80, 84)
+		FMTVR_FPU_REGS(vr12, vr13)
+		LDW_FPU_REGS(96, 100, 112, 116)
+		FMTVR_FPU_REGS(vr14, vr15)
+		: "=a"(tmp1), "=a"(tmp2), "=a"(tmp3),
+		  "=a"(tmp4), "+a"(fpregs)
+		::"memory");
+	}
+#endif
+	local_irq_restore(flg);
+}
diff --git a/arch/csky/abiv2/inc/abi/cacheflush.h b/arch/csky/abiv2/inc/abi/cacheflush.h
new file mode 100644
index 000000000..da51a0f02
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/cacheflush.h
@@ -0,0 +1,61 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_CSKY_CACHEFLUSH_H
+#define __ABI_CSKY_CACHEFLUSH_H
+
+/* Keep includes the same across arches.  */
+#include <linux/mm.h>
+
+/*
+ * The cache doesn't need to be flushed when TLB entries change when
+ * the cache is mapped to physical memory, not virtual memory
+ */
+#define flush_cache_all()			do { } while (0)
+#define flush_cache_mm(mm)			do { } while (0)
+#define flush_cache_dup_mm(mm)			do { } while (0)
+#define flush_cache_range(vma, start, end)	do { } while (0)
+#define flush_cache_page(vma, vmaddr, pfn)	do { } while (0)
+
+#define PG_dcache_clean		PG_arch_1
+
+static inline void flush_dcache_folio(struct folio *folio)
+{
+	if (test_bit(PG_dcache_clean, &folio->flags.f))
+		clear_bit(PG_dcache_clean, &folio->flags.f);
+}
+#define flush_dcache_folio flush_dcache_folio
+
+#define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
+static inline void flush_dcache_page(struct page *page)
+{
+	flush_dcache_folio(page_folio(page));
+}
+
+#define flush_dcache_mmap_lock(mapping)		do { } while (0)
+#define flush_dcache_mmap_unlock(mapping)	do { } while (0)
+
+#define flush_icache_range(start, end)		cache_wbinv_range(start, end)
+
+void flush_icache_mm_range(struct mm_struct *mm,
+			unsigned long start, unsigned long end);
+void flush_icache_deferred(struct mm_struct *mm);
+
+#define flush_cache_vmap(start, end)		do { } while (0)
+#define flush_cache_vmap_early(start, end)	do { } while (0)
+#define flush_cache_vunmap(start, end)		do { } while (0)
+
+#define copy_to_user_page(vma, page, vaddr, dst, src, len) \
+do { \
+	memcpy(dst, src, len); \
+	if (vma->vm_flags & VM_EXEC) { \
+		dcache_wb_range((unsigned long)dst, \
+				(unsigned long)dst + len); \
+		flush_icache_mm_range(current->mm, \
+				(unsigned long)dst, \
+				(unsigned long)dst + len); \
+		} \
+} while (0)
+#define copy_from_user_page(vma, page, vaddr, dst, src, len) \
+	memcpy(dst, src, len)
+
+#endif /* __ABI_CSKY_CACHEFLUSH_H */
diff --git a/arch/csky/abiv2/inc/abi/ckmmu.h b/arch/csky/abiv2/inc/abi/ckmmu.h
new file mode 100644
index 000000000..64215f238
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/ckmmu.h
@@ -0,0 +1,139 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_CKMMUV2_H
+#define __ASM_CSKY_CKMMUV2_H
+
+#include <abi/reg_ops.h>
+#include <asm/barrier.h>
+
+static inline int read_mmu_index(void)
+{
+	return mfcr("cr<0, 15>");
+}
+
+static inline void write_mmu_index(int value)
+{
+	mtcr("cr<0, 15>", value);
+}
+
+static inline int read_mmu_entrylo0(void)
+{
+	return mfcr("cr<2, 15>");
+}
+
+static inline int read_mmu_entrylo1(void)
+{
+	return mfcr("cr<3, 15>");
+}
+
+static inline void write_mmu_pagemask(int value)
+{
+	mtcr("cr<6, 15>", value);
+}
+
+static inline int read_mmu_entryhi(void)
+{
+	return mfcr("cr<4, 15>");
+}
+
+static inline void write_mmu_entryhi(int value)
+{
+	mtcr("cr<4, 15>", value);
+}
+
+static inline unsigned long read_mmu_msa0(void)
+{
+	return mfcr("cr<30, 15>");
+}
+
+static inline void write_mmu_msa0(unsigned long value)
+{
+	mtcr("cr<30, 15>", value);
+}
+
+static inline unsigned long read_mmu_msa1(void)
+{
+	return mfcr("cr<31, 15>");
+}
+
+static inline void write_mmu_msa1(unsigned long value)
+{
+	mtcr("cr<31, 15>", value);
+}
+
+/*
+ * TLB operations.
+ */
+static inline void tlb_probe(void)
+{
+	mtcr("cr<8, 15>", 0x80000000);
+}
+
+static inline void tlb_read(void)
+{
+	mtcr("cr<8, 15>", 0x40000000);
+}
+
+static inline void tlb_invalid_all(void)
+{
+#ifdef CONFIG_CPU_HAS_TLBI
+	sync_is();
+	asm volatile(
+		"tlbi.alls	\n"
+		"sync.i		\n"
+		:
+		:
+		: "memory");
+#else
+	mtcr("cr<8, 15>", 0x04000000);
+#endif
+}
+
+static inline void local_tlb_invalid_all(void)
+{
+#ifdef CONFIG_CPU_HAS_TLBI
+	sync_is();
+	asm volatile(
+		"tlbi.all	\n"
+		"sync.i		\n"
+		:
+		:
+		: "memory");
+#else
+	tlb_invalid_all();
+#endif
+}
+
+static inline void tlb_invalid_indexed(void)
+{
+	mtcr("cr<8, 15>", 0x02000000);
+}
+
+#define NOP32 ".long 0x4820c400\n"
+
+static inline void setup_pgd(pgd_t *pgd, int asid)
+{
+#ifdef CONFIG_CPU_HAS_TLBI
+	sync_is();
+#else
+	mb();
+#endif
+	asm volatile(
+#ifdef CONFIG_CPU_HAS_TLBI
+		"mtcr %1, cr<28, 15>	\n"
+#endif
+		"mtcr %1, cr<29, 15>	\n"
+		"mtcr %0, cr< 4, 15>	\n"
+		".rept 64		\n"
+		NOP32
+		".endr			\n"
+		:
+		:"r"(asid), "r"(__pa(pgd) | BIT(0))
+		:"memory");
+}
+
+static inline pgd_t *get_pgd(void)
+{
+	return __va(mfcr("cr<29, 15>") & ~BIT(0));
+}
+#endif /* __ASM_CSKY_CKMMUV2_H */
diff --git a/arch/csky/abiv2/inc/abi/elf.h b/arch/csky/abiv2/inc/abi/elf.h
new file mode 100644
index 000000000..290f49ef4
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/elf.h
@@ -0,0 +1,43 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_CSKY_ELF_H
+#define __ABI_CSKY_ELF_H
+
+/* The member sort in array pr_reg[x] is defined by GDB. */
+#define ELF_CORE_COPY_REGS(pr_reg, regs) do {	\
+	pr_reg[0] = regs->pc;			\
+	pr_reg[1] = regs->a1;			\
+	pr_reg[2] = regs->a0;			\
+	pr_reg[3] = regs->sr;			\
+	pr_reg[4] = regs->a2;			\
+	pr_reg[5] = regs->a3;			\
+	pr_reg[6] = regs->regs[0];		\
+	pr_reg[7] = regs->regs[1];		\
+	pr_reg[8] = regs->regs[2];		\
+	pr_reg[9] = regs->regs[3];		\
+	pr_reg[10] = regs->regs[4];		\
+	pr_reg[11] = regs->regs[5];		\
+	pr_reg[12] = regs->regs[6];		\
+	pr_reg[13] = regs->regs[7];		\
+	pr_reg[14] = regs->regs[8];		\
+	pr_reg[15] = regs->regs[9];		\
+	pr_reg[16] = regs->usp;			\
+	pr_reg[17] = regs->lr;			\
+	pr_reg[18] = regs->exregs[0];		\
+	pr_reg[19] = regs->exregs[1];		\
+	pr_reg[20] = regs->exregs[2];		\
+	pr_reg[21] = regs->exregs[3];		\
+	pr_reg[22] = regs->exregs[4];		\
+	pr_reg[23] = regs->exregs[5];		\
+	pr_reg[24] = regs->exregs[6];		\
+	pr_reg[25] = regs->exregs[7];		\
+	pr_reg[26] = regs->exregs[8];		\
+	pr_reg[27] = regs->exregs[9];		\
+	pr_reg[28] = regs->exregs[10];		\
+	pr_reg[29] = regs->exregs[11];		\
+	pr_reg[30] = regs->exregs[12];		\
+	pr_reg[31] = regs->exregs[13];		\
+	pr_reg[32] = regs->exregs[14];		\
+	pr_reg[33] = regs->tls;			\
+} while (0);
+#endif /* __ABI_CSKY_ELF_H */
diff --git a/arch/csky/abiv2/inc/abi/entry.h b/arch/csky/abiv2/inc/abi/entry.h
new file mode 100644
index 000000000..cca63e699
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/entry.h
@@ -0,0 +1,314 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_ENTRY_H
+#define __ASM_CSKY_ENTRY_H
+
+#include <asm/setup.h>
+#include <abi/regdef.h>
+
+#define LSAVE_PC	8
+#define LSAVE_PSR	12
+#define LSAVE_A0	24
+#define LSAVE_A1	28
+#define LSAVE_A2	32
+#define LSAVE_A3	36
+#define LSAVE_A4	40
+#define LSAVE_A5	44
+
+#define KSPTOUSP
+#define USPTOKSP
+
+#define usp cr<14, 1>
+
+.macro SAVE_ALL epc_inc
+	subi    sp, 152
+	stw	tls, (sp, 0)
+	stw	lr, (sp, 4)
+
+	RD_MEH	lr
+	WR_MEH	lr
+
+	mfcr	lr, epc
+	movi	tls, \epc_inc
+	add	lr, tls
+	stw	lr, (sp, 8)
+
+	mfcr	lr, epsr
+	stw	lr, (sp, 12)
+	btsti   lr, 31
+	bf      1f
+	addi    lr, sp, 152
+	br	2f
+1:
+	mfcr	lr, usp
+2:
+	stw	lr, (sp, 16)
+
+	stw     a0, (sp, 20)
+	stw     a0, (sp, 24)
+	stw     a1, (sp, 28)
+	stw     a2, (sp, 32)
+	stw     a3, (sp, 36)
+
+	addi	sp, 40
+	stm	r4-r13, (sp)
+
+	addi    sp, 40
+	stm     r16-r30, (sp)
+#ifdef CONFIG_CPU_HAS_HILO
+	mfhi	lr
+	stw	lr, (sp, 60)
+	mflo	lr
+	stw	lr, (sp, 64)
+	mfcr	lr, cr14
+	stw	lr, (sp, 68)
+#endif
+	subi	sp, 80
+.endm
+
+.macro	RESTORE_ALL
+	ldw	tls, (sp, 0)
+	ldw	lr, (sp, 4)
+	ldw	a0, (sp, 8)
+	mtcr	a0, epc
+	ldw	a0, (sp, 12)
+	mtcr	a0, epsr
+	btsti   a0, 31
+	ldw	a0, (sp, 16)
+	mtcr	a0, usp
+	mtcr	a0, ss0
+
+#ifdef CONFIG_CPU_HAS_HILO
+	ldw	a0, (sp, 140)
+	mthi	a0
+	ldw	a0, (sp, 144)
+	mtlo	a0
+	ldw	a0, (sp, 148)
+	mtcr	a0, cr14
+#endif
+
+	ldw     a0, (sp, 24)
+	ldw     a1, (sp, 28)
+	ldw     a2, (sp, 32)
+	ldw     a3, (sp, 36)
+
+	addi	sp, 40
+	ldm	r4-r13, (sp)
+	addi    sp, 40
+	ldm     r16-r30, (sp)
+	addi    sp, 72
+	bf	1f
+	mfcr	sp, ss0
+1:
+	rte
+.endm
+
+.macro SAVE_REGS_FTRACE
+	subi    sp, 152
+	stw	tls, (sp, 0)
+	stw	lr, (sp, 4)
+
+	mfcr	lr, psr
+	stw	lr, (sp, 12)
+
+	addi    lr, sp, 152
+	stw	lr, (sp, 16)
+
+	stw     a0, (sp, 20)
+	stw     a0, (sp, 24)
+	stw     a1, (sp, 28)
+	stw     a2, (sp, 32)
+	stw     a3, (sp, 36)
+
+	addi	sp, 40
+	stm	r4-r13, (sp)
+
+	addi    sp, 40
+	stm     r16-r30, (sp)
+#ifdef CONFIG_CPU_HAS_HILO
+	mfhi	lr
+	stw	lr, (sp, 60)
+	mflo	lr
+	stw	lr, (sp, 64)
+	mfcr	lr, cr14
+	stw	lr, (sp, 68)
+#endif
+	subi	sp, 80
+.endm
+
+.macro	RESTORE_REGS_FTRACE
+	ldw	tls, (sp, 0)
+
+#ifdef CONFIG_CPU_HAS_HILO
+	ldw	a0, (sp, 140)
+	mthi	a0
+	ldw	a0, (sp, 144)
+	mtlo	a0
+	ldw	a0, (sp, 148)
+	mtcr	a0, cr14
+#endif
+
+	ldw     a0, (sp, 24)
+	ldw     a1, (sp, 28)
+	ldw     a2, (sp, 32)
+	ldw     a3, (sp, 36)
+
+	addi	sp, 40
+	ldm	r4-r13, (sp)
+	addi    sp, 40
+	ldm     r16-r30, (sp)
+	addi    sp, 72
+.endm
+
+.macro SAVE_SWITCH_STACK
+	subi    sp, 64
+	stm	r4-r11, (sp)
+	stw	lr,  (sp, 32)
+	stw	r16, (sp, 36)
+	stw	r17, (sp, 40)
+	stw	r26, (sp, 44)
+	stw	r27, (sp, 48)
+	stw	r28, (sp, 52)
+	stw	r29, (sp, 56)
+	stw	r30, (sp, 60)
+#ifdef CONFIG_CPU_HAS_HILO
+	subi	sp, 16
+	mfhi	lr
+	stw	lr, (sp, 0)
+	mflo	lr
+	stw	lr, (sp, 4)
+	mfcr	lr, cr14
+	stw	lr, (sp, 8)
+#endif
+.endm
+
+.macro RESTORE_SWITCH_STACK
+#ifdef CONFIG_CPU_HAS_HILO
+	ldw	lr, (sp, 0)
+	mthi	lr
+	ldw	lr, (sp, 4)
+	mtlo	lr
+	ldw	lr, (sp, 8)
+	mtcr	lr, cr14
+	addi	sp, 16
+#endif
+	ldm	r4-r11, (sp)
+	ldw	lr,  (sp, 32)
+	ldw	r16, (sp, 36)
+	ldw	r17, (sp, 40)
+	ldw	r26, (sp, 44)
+	ldw	r27, (sp, 48)
+	ldw	r28, (sp, 52)
+	ldw	r29, (sp, 56)
+	ldw	r30, (sp, 60)
+	addi	sp, 64
+.endm
+
+/* MMU registers operators. */
+.macro RD_MIR rx
+	mfcr	\rx, cr<0, 15>
+.endm
+
+.macro RD_MEH rx
+	mfcr	\rx, cr<4, 15>
+.endm
+
+.macro RD_MCIR rx
+	mfcr	\rx, cr<8, 15>
+.endm
+
+.macro RD_PGDR rx
+	mfcr	\rx, cr<29, 15>
+.endm
+
+.macro RD_PGDR_K rx
+	mfcr	\rx, cr<28, 15>
+.endm
+
+.macro WR_MEH rx
+	mtcr	\rx, cr<4, 15>
+.endm
+
+.macro WR_MCIR rx
+	mtcr	\rx, cr<8, 15>
+.endm
+
+#ifdef CONFIG_PAGE_OFFSET_80000000
+#define MSA_SET cr<30, 15>
+#define MSA_CLR cr<31, 15>
+#endif
+
+#ifdef CONFIG_PAGE_OFFSET_A0000000
+#define MSA_SET cr<31, 15>
+#define MSA_CLR cr<30, 15>
+#endif
+
+.macro SETUP_MMU
+	/* Init psr and enable ee */
+	lrw	r6, DEFAULT_PSR_VALUE
+	mtcr    r6, psr
+	psrset  ee
+
+	/* Invalid I/Dcache BTB BHT */
+	movi	r6, 7
+	lsli	r6, 16
+	addi	r6, (1<<4) | 3
+	mtcr	r6, cr17
+
+	/* Invalid all TLB */
+	bgeni   r6, 26
+	mtcr	r6, cr<8, 15> /* Set MCIR */
+
+	/* Check MMU on/off */
+	mfcr	r6, cr18
+	btsti	r6, 0
+	bt	1f
+
+	/* MMU off: setup mapping tlb entry */
+	movi	r6, 0
+	mtcr	r6, cr<6, 15> /* Set MPR with 4K page size */
+
+	grs	r6, 1f /* Get current pa by PC */
+	bmaski  r7, (PAGE_SHIFT + 1) /* r7 = 0x1fff */
+	andn    r6, r7
+	mtcr	r6, cr<4, 15> /* Set MEH */
+
+	mov	r8, r6
+	movi    r7, 0x00000006
+	or      r8, r7
+	mtcr	r8, cr<2, 15> /* Set MEL0 */
+	movi    r7, 0x00001006
+	or      r8, r7
+	mtcr	r8, cr<3, 15> /* Set MEL1 */
+
+	bgeni   r8, 28
+	mtcr	r8, cr<8, 15> /* Set MCIR to write TLB */
+
+	br	2f
+1:
+	/*
+	 * MMU on: use origin MSA value from bootloader
+	 *
+	 * cr<30/31, 15> MSA register format:
+	 * 31 - 29 | 28 - 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
+	 *   BA     Reserved  SH  WA  B   SO SEC  C   D   V
+	 */
+	mfcr	r6, MSA_SET /* Get MSA */
+2:
+	lsri	r6, 29
+	lsli	r6, 29
+	addi	r6, 0x1ce
+	mtcr	r6, MSA_SET /* Set MSA */
+
+	movi    r6, 0
+	mtcr	r6, MSA_CLR /* Clr MSA */
+
+	/* enable MMU */
+	mfcr    r6, cr18
+	bseti	r6, 0
+	mtcr    r6, cr18
+
+	jmpi	3f /* jump to va */
+3:
+.endm
+#endif /* __ASM_CSKY_ENTRY_H */
diff --git a/arch/csky/abiv2/inc/abi/fpu.h b/arch/csky/abiv2/inc/abi/fpu.h
new file mode 100644
index 000000000..aabb79355
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/fpu.h
@@ -0,0 +1,66 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_FPU_H
+#define __ASM_CSKY_FPU_H
+
+#include <asm/sigcontext.h>
+#include <asm/ptrace.h>
+
+int fpu_libc_helper(struct pt_regs *regs);
+void fpu_fpe(struct pt_regs *regs);
+
+static inline void init_fpu(void) { mtcr("cr<1, 2>", 0); }
+
+void save_to_user_fp(struct user_fp *user_fp);
+void restore_from_user_fp(struct user_fp *user_fp);
+
+/*
+ * Define the fesr bit for fpe handle.
+ */
+#define  FPE_ILLE  (1 << 16)    /* Illegal instruction  */
+#define  FPE_FEC   (1 << 7)     /* Input float-point arithmetic exception */
+#define  FPE_IDC   (1 << 5)     /* Input denormalized exception */
+#define  FPE_IXC   (1 << 4)     /* Inexact exception */
+#define  FPE_UFC   (1 << 3)     /* Underflow exception */
+#define  FPE_OFC   (1 << 2)     /* Overflow exception */
+#define  FPE_DZC   (1 << 1)     /* Divide by zero exception */
+#define  FPE_IOC   (1 << 0)     /* Invalid operation exception */
+#define  FPE_REGULAR_EXCEPTION (FPE_IXC | FPE_UFC | FPE_OFC | FPE_DZC | FPE_IOC)
+
+#ifdef CONFIG_OPEN_FPU_IDE
+#define IDE_STAT   (1 << 5)
+#else
+#define IDE_STAT   0
+#endif
+
+#ifdef CONFIG_OPEN_FPU_IXE
+#define IXE_STAT   (1 << 4)
+#else
+#define IXE_STAT   0
+#endif
+
+#ifdef CONFIG_OPEN_FPU_UFE
+#define UFE_STAT   (1 << 3)
+#else
+#define UFE_STAT   0
+#endif
+
+#ifdef CONFIG_OPEN_FPU_OFE
+#define OFE_STAT   (1 << 2)
+#else
+#define OFE_STAT   0
+#endif
+
+#ifdef CONFIG_OPEN_FPU_DZE
+#define DZE_STAT   (1 << 1)
+#else
+#define DZE_STAT   0
+#endif
+
+#ifdef CONFIG_OPEN_FPU_IOE
+#define IOE_STAT   (1 << 0)
+#else
+#define IOE_STAT   0
+#endif
+
+#endif /* __ASM_CSKY_FPU_H */
diff --git a/arch/csky/abiv2/inc/abi/page.h b/arch/csky/abiv2/inc/abi/page.h
new file mode 100644
index 000000000..a5a255013
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/page.h
@@ -0,0 +1,6 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+static inline void copy_user_page(void *to, void *from, unsigned long vaddr,
+				  struct page *page)
+{
+	copy_page(to, from);
+}
diff --git a/arch/csky/abiv2/inc/abi/pgtable-bits.h b/arch/csky/abiv2/inc/abi/pgtable-bits.h
new file mode 100644
index 000000000..526152bd2
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/pgtable-bits.h
@@ -0,0 +1,53 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_PGTABLE_BITS_H
+#define __ASM_CSKY_PGTABLE_BITS_H
+
+/* implemented in software */
+#define _PAGE_ACCESSED		(1<<7)
+#define _PAGE_READ		(1<<8)
+#define _PAGE_WRITE		(1<<9)
+#define _PAGE_PRESENT		(1<<10)
+#define _PAGE_MODIFIED		(1<<11)
+
+/* We borrow bit 7 to store the exclusive marker in swap PTEs. */
+#define _PAGE_SWP_EXCLUSIVE	(1<<7)
+
+/* implemented in hardware */
+#define _PAGE_GLOBAL		(1<<0)
+#define _PAGE_VALID		(1<<1)
+#define _PAGE_DIRTY		(1<<2)
+
+#define _PAGE_SO		(1<<5)
+#define _PAGE_BUF		(1<<6)
+#define _PAGE_CACHE		(1<<3)
+#define _CACHE_MASK		_PAGE_CACHE
+
+#define _CACHE_CACHED		(_PAGE_CACHE | _PAGE_BUF)
+#define _CACHE_UNCACHED		(0)
+
+#define _PAGE_PROT_NONE		_PAGE_WRITE
+
+/*
+ * Encode/decode swap entries and swap PTEs. Swap PTEs are all PTEs that
+ * are !pte_none() && !pte_present().
+ *
+ * Format of swap PTE:
+ *     bit          0:    _PAGE_GLOBAL (zero)
+ *     bit          1:    _PAGE_VALID (zero)
+ *     bit      2 - 6:    swap type
+ *     bit          7:    exclusive marker
+ *     bit          8:    swap offset[0]
+ *     bit          9:    _PAGE_WRITE (zero)
+ *     bit         10:    _PAGE_PRESENT (zero)
+ *     bit    11 - 31:    swap offset[1 - 21]
+ */
+#define __swp_type(x)			(((x).val >> 2) & 0x1f)
+#define __swp_offset(x)			((((x).val >> 8) & 0x1) | \
+					(((x).val >> 10) & 0x3ffffe))
+#define __swp_entry(type, offset)	((swp_entry_t) { \
+					((type & 0x1f) << 2) | \
+					((offset & 0x1) << 8) | \
+					((offset & 0x3ffffe) << 10)})
+
+#endif /* __ASM_CSKY_PGTABLE_BITS_H */
diff --git a/arch/csky/abiv2/inc/abi/reg_ops.h b/arch/csky/abiv2/inc/abi/reg_ops.h
new file mode 100644
index 000000000..49ba18a64
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/reg_ops.h
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_REG_OPS_H
+#define __ABI_REG_OPS_H
+#include <asm/reg_ops.h>
+
+static inline unsigned int mfcr_hint(void)
+{
+	return mfcr("cr31");
+}
+
+static inline unsigned int mfcr_ccr2(void)
+{
+	return mfcr("cr23");
+}
+#endif /* __ABI_REG_OPS_H */
diff --git a/arch/csky/abiv2/inc/abi/regdef.h b/arch/csky/abiv2/inc/abi/regdef.h
new file mode 100644
index 000000000..fc08d56cc
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/regdef.h
@@ -0,0 +1,31 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_REGDEF_H
+#define __ASM_CSKY_REGDEF_H
+
+#ifdef __ASSEMBLER__
+#define syscallid	r7
+#else
+#define syscallid	"r7"
+#endif
+
+#define regs_syscallid(regs) regs->regs[3]
+#define regs_fp(regs) regs->regs[4]
+
+/*
+ * PSR format:
+ * | 31 | 30-24 | 23-16 | 15 14 | 13-10 | 9 | 8-0 |
+ *   S              VEC     TM            MM
+ *
+ *   S: Super Mode
+ * VEC: Exception Number
+ *  TM: Trace Mode
+ *  MM: Memory unaligned addr access
+ */
+#define DEFAULT_PSR_VALUE	0x80000200
+
+#define SYSTRACE_SAVENUM	5
+
+#define TRAP0_SIZE		4
+
+#endif /* __ASM_CSKY_REGDEF_H */
diff --git a/arch/csky/abiv2/inc/abi/string.h b/arch/csky/abiv2/inc/abi/string.h
new file mode 100644
index 000000000..f01bad2ac
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/string.h
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_CSKY_STRING_H
+#define __ABI_CSKY_STRING_H
+
+#define __HAVE_ARCH_MEMCMP
+extern int memcmp(const void *, const void *, __kernel_size_t);
+
+#define __HAVE_ARCH_MEMCPY
+extern void *memcpy(void *, const void *, __kernel_size_t);
+
+#define __HAVE_ARCH_MEMMOVE
+extern void *memmove(void *, const void *, __kernel_size_t);
+
+#define __HAVE_ARCH_MEMSET
+extern void *memset(void *, int,  __kernel_size_t);
+
+#define __HAVE_ARCH_STRCMP
+extern int strcmp(const char *, const char *);
+
+#define __HAVE_ARCH_STRCPY
+extern char *strcpy(char *, const char *);
+
+#define __HAVE_ARCH_STRLEN
+extern __kernel_size_t strlen(const char *);
+
+#endif /* __ABI_CSKY_STRING_H */
diff --git a/arch/csky/abiv2/inc/abi/switch_context.h b/arch/csky/abiv2/inc/abi/switch_context.h
new file mode 100644
index 000000000..5dd5c3f4e
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/switch_context.h
@@ -0,0 +1,31 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_CSKY_PTRACE_H
+#define __ABI_CSKY_PTRACE_H
+
+struct switch_stack {
+#ifdef CONFIG_CPU_HAS_HILO
+	unsigned long rhi;
+	unsigned long rlo;
+	unsigned long cr14;
+	unsigned long pad;
+#endif
+	unsigned long r4;
+	unsigned long r5;
+	unsigned long r6;
+	unsigned long r7;
+	unsigned long r8;
+	unsigned long r9;
+	unsigned long r10;
+	unsigned long r11;
+
+	unsigned long r15;
+	unsigned long r16;
+	unsigned long r17;
+	unsigned long r26;
+	unsigned long r27;
+	unsigned long r28;
+	unsigned long r29;
+	unsigned long r30;
+};
+#endif /* __ABI_CSKY_PTRACE_H */
diff --git a/arch/csky/abiv2/inc/abi/vdso.h b/arch/csky/abiv2/inc/abi/vdso.h
new file mode 100644
index 000000000..40fd10d89
--- /dev/null
+++ b/arch/csky/abiv2/inc/abi/vdso.h
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ABI_CSKY_VDSO_H
+#define __ABI_CSKY_VDSO_H
+
+/* movi r7, 173 */
+#define SET_SYSCALL_ID	.long 0x008bea07
+
+#endif /* __ABI_CSKY_VDSO_H */
diff --git a/arch/csky/abiv2/mcount.S b/arch/csky/abiv2/mcount.S
new file mode 100644
index 000000000..d745e10c1
--- /dev/null
+++ b/arch/csky/abiv2/mcount.S
@@ -0,0 +1,211 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include <asm/ftrace.h>
+#include <abi/entry.h>
+#include <asm/asm-offsets.h>
+
+/*
+ * csky-gcc with -pg will put the following asm after prologue:
+ *      push	r15
+ *      jsri	_mcount
+ *
+ * stack layout after mcount_enter in _mcount():
+ *
+ * current sp => 0:+-------+
+ *                 | a0-a3 | -> must save all argument regs
+ *             +16:+-------+
+ *                 | lr    | -> _mcount lr (instrumente function's pc)
+ *             +20:+-------+
+ *                 | fp=r8 | -> instrumented function fp
+ *             +24:+-------+
+ *                 | plr   | -> instrumented function lr (parent's pc)
+ *                 +-------+
+ */
+
+.macro mcount_enter
+	subi	sp, 24
+	stw	a0, (sp, 0)
+	stw	a1, (sp, 4)
+	stw	a2, (sp, 8)
+	stw	a3, (sp, 12)
+	stw	lr, (sp, 16)
+	stw	r8, (sp, 20)
+.endm
+
+.macro mcount_exit
+	ldw	a0, (sp, 0)
+	ldw	a1, (sp, 4)
+	ldw	a2, (sp, 8)
+	ldw	a3, (sp, 12)
+	ldw	t1, (sp, 16)
+	ldw	r8, (sp, 20)
+	ldw	lr, (sp, 24)
+	addi	sp, 28
+	jmp	t1
+.endm
+
+.macro mcount_enter_regs
+	subi	sp, 8
+	stw	lr, (sp, 0)
+	stw	r8, (sp, 4)
+	SAVE_REGS_FTRACE
+.endm
+
+.macro mcount_exit_regs
+	RESTORE_REGS_FTRACE
+	subi	sp, 152
+	ldw	t1, (sp, 4)
+	addi	sp, 152
+	ldw	r8, (sp, 4)
+	ldw	lr, (sp, 8)
+	addi	sp, 12
+	jmp	t1
+.endm
+
+.macro save_return_regs
+	subi	sp, 16
+	stw	a0, (sp, 0)
+	stw	a1, (sp, 4)
+	stw	a2, (sp, 8)
+	stw	a3, (sp, 12)
+.endm
+
+.macro restore_return_regs
+	mov	lr, a0
+	ldw	a0, (sp, 0)
+	ldw	a1, (sp, 4)
+	ldw	a2, (sp, 8)
+	ldw	a3, (sp, 12)
+	addi	sp, 16
+.endm
+
+.macro nop32_stub
+	nop32
+	nop32
+	nop32
+.endm
+
+ENTRY(ftrace_stub)
+	jmp	lr
+END(ftrace_stub)
+
+#ifndef CONFIG_DYNAMIC_FTRACE
+ENTRY(_mcount)
+	mcount_enter
+
+	/* r26 is link register, only used with jsri translation */
+	lrw	r26, ftrace_trace_function
+	ldw	r26, (r26, 0)
+	lrw	a1, ftrace_stub
+	cmpne	r26, a1
+	bf	skip_ftrace
+
+	mov	a0, lr
+	subi	a0, 4
+	ldw	a1, (sp, 24)
+	lrw	a2, function_trace_op
+	ldw	a2, (a2, 0)
+
+	jsr	r26
+
+#ifndef CONFIG_FUNCTION_GRAPH_TRACER
+skip_ftrace:
+	mcount_exit
+#else
+skip_ftrace:
+	lrw	a0, ftrace_graph_return
+	ldw	a0, (a0, 0)
+	lrw	a1, ftrace_stub
+	cmpne	a0, a1
+	bt	ftrace_graph_caller
+
+	lrw	a0, ftrace_graph_entry
+	ldw	a0, (a0, 0)
+	lrw	a1, ftrace_graph_entry_stub
+	cmpne	a0, a1
+	bt	ftrace_graph_caller
+
+	mcount_exit
+#endif
+END(_mcount)
+#else /* CONFIG_DYNAMIC_FTRACE */
+ENTRY(_mcount)
+	mov	t1, lr
+	ldw	lr, (sp, 0)
+	addi	sp, 4
+	jmp	t1
+ENDPROC(_mcount)
+
+ENTRY(ftrace_caller)
+	mcount_enter
+
+	ldw	a0, (sp, 16)
+	subi	a0, 4
+	ldw	a1, (sp, 24)
+	lrw	a2, function_trace_op
+	ldw	a2, (a2, 0)
+
+	nop
+GLOBAL(ftrace_call)
+	nop32_stub
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	nop
+GLOBAL(ftrace_graph_call)
+	nop32_stub
+#endif
+
+	mcount_exit
+ENDPROC(ftrace_caller)
+#endif /* CONFIG_DYNAMIC_FTRACE */
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+ENTRY(ftrace_graph_caller)
+	mov	a0, sp
+	addi	a0, 24
+	ldw	a1, (sp, 16)
+	subi	a1, 4
+	mov	a2, r8
+	lrw	r26, prepare_ftrace_return
+	jsr	r26
+	mcount_exit
+END(ftrace_graph_caller)
+
+ENTRY(return_to_handler)
+	save_return_regs
+	mov	a0, r8
+	jsri	ftrace_return_to_handler
+	restore_return_regs
+	jmp	lr
+END(return_to_handler)
+#endif
+
+#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
+ENTRY(ftrace_regs_caller)
+	mcount_enter_regs
+
+	lrw	t1, PT_FRAME_SIZE
+	add	t1, sp
+
+	ldw	a0, (t1, 0)
+	subi	a0, 4
+	ldw	a1, (t1, 8)
+	lrw	a2, function_trace_op
+	ldw	a2, (a2, 0)
+	mov	a3, sp
+
+	nop
+GLOBAL(ftrace_regs_call)
+	nop32_stub
+
+#ifdef CONFIG_FUNCTION_GRAPH_TRACER
+	nop
+GLOBAL(ftrace_graph_regs_call)
+	nop32_stub
+#endif
+
+	mcount_exit_regs
+ENDPROC(ftrace_regs_caller)
+#endif /* CONFIG_DYNAMIC_FTRACE */
diff --git a/arch/csky/abiv2/memcmp.S b/arch/csky/abiv2/memcmp.S
new file mode 100644
index 000000000..bf0d809f0
--- /dev/null
+++ b/arch/csky/abiv2/memcmp.S
@@ -0,0 +1,152 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+ENTRY(memcmp)
+	/* Test if len less than 4 bytes.  */
+	mov	r3, r0
+	movi	r0, 0
+	mov	r12, r4
+	cmplti	r2, 4
+	bt	.L_compare_by_byte
+
+	andi	r13, r0, 3
+	movi	r19, 4
+
+	/* Test if s1 is not 4 bytes aligned.  */
+	bnez	r13, .L_s1_not_aligned
+
+	LABLE_ALIGN
+.L_s1_aligned:
+	/* If dest is aligned, then copy.  */
+	zext	r18, r2, 31, 4
+	/* Test if len less than 16 bytes.  */
+	bez	r18, .L_compare_by_word
+
+.L_compare_by_4word:
+	/* If aligned, load word each time.  */
+	ldw	r20, (r3, 0)
+	ldw	r21, (r1, 0)
+	/* If s1[i] != s2[i], goto .L_byte_check.  */
+	cmpne	r20, r21
+	bt	.L_byte_check
+
+	ldw	r20, (r3, 4)
+	ldw	r21, (r1, 4)
+	cmpne	r20, r21
+	bt	.L_byte_check
+
+	ldw	r20, (r3, 8)
+	ldw	r21, (r1, 8)
+	cmpne	r20, r21
+	bt	.L_byte_check
+
+	ldw	r20, (r3, 12)
+	ldw	r21, (r1, 12)
+	cmpne	r20, r21
+	bt	.L_byte_check
+
+	PRE_BNEZAD (r18)
+	addi	a3, 16
+	addi	a1, 16
+
+	BNEZAD (r18, .L_compare_by_4word)
+
+.L_compare_by_word:
+	zext	r18, r2, 3, 2
+	bez	r18, .L_compare_by_byte
+.L_compare_by_word_loop:
+	ldw	r20, (r3, 0)
+	ldw	r21, (r1, 0)
+	addi	r3, 4
+	PRE_BNEZAD (r18)
+	cmpne	r20, r21
+	addi    r1, 4
+	bt	.L_byte_check
+	BNEZAD (r18, .L_compare_by_word_loop)
+
+.L_compare_by_byte:
+        zext    r18, r2, 1, 0
+        bez     r18, .L_return
+.L_compare_by_byte_loop:
+        ldb     r0, (r3, 0)
+        ldb     r4, (r1, 0)
+        addi    r3, 1
+        subu    r0, r4
+        PRE_BNEZAD (r18)
+        addi    r1, 1
+        bnez    r0, .L_return
+        BNEZAD (r18, .L_compare_by_byte_loop)
+
+.L_return:
+        mov     r4, r12
+        rts
+
+# ifdef __CSKYBE__
+/* d[i] != s[i] in word, so we check byte 0.  */
+.L_byte_check:
+        xtrb0   r0, r20
+        xtrb0   r2, r21
+        subu    r0, r2
+        bnez    r0, .L_return
+
+        /* check byte 1 */
+        xtrb1   r0, r20
+        xtrb1   r2, r21
+        subu    r0, r2
+        bnez    r0, .L_return
+
+        /* check byte 2 */
+        xtrb2   r0, r20
+        xtrb2   r2, r21
+        subu    r0, r2
+        bnez    r0, .L_return
+
+        /* check byte 3 */
+        xtrb3   r0, r20
+        xtrb3   r2, r21
+        subu    r0, r2
+# else
+/* s1[i] != s2[i] in word, so we check byte 3.  */
+.L_byte_check:
+	xtrb3	r0, r20
+	xtrb3	r2, r21
+        subu    r0, r2
+        bnez    r0, .L_return
+
+	/* check byte 2 */
+	xtrb2	r0, r20
+	xtrb2	r2, r21
+        subu    r0, r2
+        bnez    r0, .L_return
+
+	/* check byte 1 */
+	xtrb1	r0, r20
+	xtrb1	r2, r21
+	subu	r0, r2
+	bnez    r0, .L_return
+
+	/* check byte 0 */
+	xtrb0	r0, r20
+	xtrb0	r2, r21
+	subu	r0, r2
+	br	.L_return
+# endif /* !__CSKYBE__ */
+
+/* Compare when s1 is not aligned.  */
+.L_s1_not_aligned:
+	sub	r13, r19, r13
+	sub	r2, r13
+.L_s1_not_aligned_loop:
+	ldb	r0, (r3, 0)
+	ldb	r4, (r1, 0)
+	addi	r3, 1
+	subu	r0, r4
+	PRE_BNEZAD (r13)
+	addi	r1, 1
+	bnez	r0, .L_return
+	BNEZAD (r13, .L_s1_not_aligned_loop)
+	br	.L_s1_aligned
+ENDPROC(memcmp)
diff --git a/arch/csky/abiv2/memcpy.S b/arch/csky/abiv2/memcpy.S
new file mode 100644
index 000000000..145bf3a93
--- /dev/null
+++ b/arch/csky/abiv2/memcpy.S
@@ -0,0 +1,104 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+ENTRY(__memcpy)
+ENTRY(memcpy)
+	/* Test if len less than 4 bytes.  */
+	mov	r12, r0
+	cmplti	r2, 4
+	bt	.L_copy_by_byte
+
+	andi	r13, r0, 3
+	movi	r19, 4
+	/* Test if dest is not 4 bytes aligned.  */
+	bnez	r13, .L_dest_not_aligned
+
+/* Hardware can handle unaligned access directly.  */
+.L_dest_aligned:
+	/* If dest is aligned, then copy.  */
+	zext	r18, r2, 31, 4
+
+	/* Test if len less than 16 bytes.  */
+	bez	r18, .L_len_less_16bytes
+	movi	r19, 0
+
+	LABLE_ALIGN
+.L_len_larger_16bytes:
+#if defined(__CK860__)
+	ldw	r3, (r1, 0)
+	stw	r3, (r0, 0)
+	ldw	r3, (r1, 4)
+	stw	r3, (r0, 4)
+	ldw	r3, (r1, 8)
+	stw	r3, (r0, 8)
+	ldw	r3, (r1, 12)
+	addi	r1, 16
+	stw	r3, (r0, 12)
+	addi	r0, 16
+#else
+	ldw	r20, (r1, 0)
+	ldw	r21, (r1, 4)
+	ldw	r22, (r1, 8)
+	ldw	r23, (r1, 12)
+	stw	r20, (r0, 0)
+	stw	r21, (r0, 4)
+	stw	r22, (r0, 8)
+	stw	r23, (r0, 12)
+	PRE_BNEZAD (r18)
+	addi	r1, 16
+	addi	r0, 16
+#endif
+	BNEZAD (r18, .L_len_larger_16bytes)
+
+.L_len_less_16bytes:
+	zext	r18, r2, 3, 2
+	bez	r18, .L_copy_by_byte
+.L_len_less_16bytes_loop:
+	ldw	r3, (r1, 0)
+	PRE_BNEZAD (r18)
+	addi	r1, 4
+	stw	r3, (r0, 0)
+	addi	r0, 4
+	BNEZAD (r18, .L_len_less_16bytes_loop)
+
+/* Test if len less than 4 bytes.  */
+.L_copy_by_byte:
+	zext	r18, r2, 1, 0
+	bez	r18, .L_return
+.L_copy_by_byte_loop:
+	ldb	r3, (r1, 0)
+	PRE_BNEZAD (r18)
+	addi	r1, 1
+	stb	r3, (r0, 0)
+	addi	r0, 1
+	BNEZAD (r18, .L_copy_by_byte_loop)
+
+.L_return:
+	mov	r0, r12
+	rts
+
+/*
+ * If dest is not aligned, just copying some bytes makes the
+ * dest align.
+ */
+.L_dest_not_aligned:
+	sub	r13, r19, r13
+	sub	r2, r13
+
+/* Makes the dest align.  */
+.L_dest_not_aligned_loop:
+	ldb	r3, (r1, 0)
+	PRE_BNEZAD (r13)
+	addi	r1, 1
+	stb	r3, (r0, 0)
+	addi	r0, 1
+	BNEZAD (r13, .L_dest_not_aligned_loop)
+	cmplti	r2, 4
+	bt	.L_copy_by_byte
+
+	/* Check whether the src is aligned.  */
+	jbr	.L_dest_aligned
+ENDPROC(__memcpy)
diff --git a/arch/csky/abiv2/memmove.S b/arch/csky/abiv2/memmove.S
new file mode 100644
index 000000000..5721e73ad
--- /dev/null
+++ b/arch/csky/abiv2/memmove.S
@@ -0,0 +1,104 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+	.weak memmove
+ENTRY(__memmove)
+ENTRY(memmove)
+	subu	r3, r0, r1
+	cmphs	r3, r2
+	bt	memcpy
+
+	mov	r12, r0
+	addu	r0, r0, r2
+	addu	r1, r1, r2
+
+	/* Test if len less than 4 bytes.  */
+	cmplti	r2, 4
+	bt	.L_copy_by_byte
+
+	andi	r13, r0, 3
+	/* Test if dest is not 4 bytes aligned.  */
+	bnez	r13, .L_dest_not_aligned
+	/* Hardware can handle unaligned access directly.  */
+.L_dest_aligned:
+	/* If dest is aligned, then copy.  */
+	zext	r18, r2, 31, 4
+	/* Test if len less than 16 bytes.  */
+	bez	r18, .L_len_less_16bytes
+	movi	r19, 0
+
+	/* len > 16 bytes */
+	LABLE_ALIGN
+.L_len_larger_16bytes:
+	subi	r1, 16
+	subi	r0, 16
+#if defined(__CK860__)
+	ldw	r3, (r1, 12)
+	stw	r3, (r0, 12)
+	ldw	r3, (r1, 8)
+	stw	r3, (r0, 8)
+	ldw	r3, (r1, 4)
+	stw	r3, (r0, 4)
+	ldw	r3, (r1, 0)
+	stw	r3, (r0, 0)
+#else
+	ldw	r20, (r1, 0)
+	ldw	r21, (r1, 4)
+	ldw	r22, (r1, 8)
+	ldw	r23, (r1, 12)
+	stw	r20, (r0, 0)
+	stw	r21, (r0, 4)
+	stw	r22, (r0, 8)
+	stw	r23, (r0, 12)
+	PRE_BNEZAD (r18)
+#endif
+	BNEZAD (r18, .L_len_larger_16bytes)
+
+.L_len_less_16bytes:
+	zext	r18, r2, 3, 2
+	bez	r18, .L_copy_by_byte
+.L_len_less_16bytes_loop:
+	subi	r1, 4
+	subi	r0, 4
+	ldw	r3, (r1, 0)
+	PRE_BNEZAD (r18)
+	stw	r3, (r0, 0)
+	BNEZAD (r18, .L_len_less_16bytes_loop)
+
+	/* Test if len less than 4 bytes.  */
+.L_copy_by_byte:
+	zext	r18, r2, 1, 0
+	bez	r18, .L_return
+.L_copy_by_byte_loop:
+	subi	r1, 1
+	subi	r0, 1
+	ldb	r3, (r1, 0)
+	PRE_BNEZAD (r18)
+	stb	r3, (r0, 0)
+	BNEZAD (r18, .L_copy_by_byte_loop)
+
+.L_return:
+	mov	r0, r12
+	rts
+
+	/* If dest is not aligned, just copy some bytes makes the dest
+	   align.  */
+.L_dest_not_aligned:
+	sub	r2, r13
+.L_dest_not_aligned_loop:
+	subi	r1, 1
+	subi	r0, 1
+	/* Makes the dest align.  */
+	ldb	r3, (r1, 0)
+	PRE_BNEZAD (r13)
+	stb	r3, (r0, 0)
+	BNEZAD (r13, .L_dest_not_aligned_loop)
+	cmplti	r2, 4
+	bt	.L_copy_by_byte
+	/* Check whether the src is aligned.  */
+	jbr	.L_dest_aligned
+ENDPROC(memmove)
+ENDPROC(__memmove)
diff --git a/arch/csky/abiv2/memset.S b/arch/csky/abiv2/memset.S
new file mode 100644
index 000000000..a7e7d994b
--- /dev/null
+++ b/arch/csky/abiv2/memset.S
@@ -0,0 +1,83 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+	.weak memset
+ENTRY(__memset)
+ENTRY(memset)
+	/* Test if len less than 4 bytes.  */
+	mov	r12, r0
+	cmplti	r2, 8
+	bt	.L_set_by_byte
+
+	andi	r13, r0, 3
+	movi	r19, 4
+	/* Test if dest is not 4 bytes aligned.  */
+	bnez	r13, .L_dest_not_aligned
+	/* Hardware can handle unaligned access directly.  */
+.L_dest_aligned:
+        zextb   r3, r1
+        lsli    r1, 8
+        or      r1, r3
+        lsli    r3, r1, 16
+        or      r3, r1
+
+	/* If dest is aligned, then copy.  */
+	zext	r18, r2, 31, 4
+	/* Test if len less than 16 bytes.  */
+	bez	r18, .L_len_less_16bytes
+
+	LABLE_ALIGN
+.L_len_larger_16bytes:
+	stw	r3, (r0, 0)
+	stw	r3, (r0, 4)
+	stw	r3, (r0, 8)
+	stw	r3, (r0, 12)
+	PRE_BNEZAD (r18)
+	addi	r0, 16
+	BNEZAD (r18, .L_len_larger_16bytes)
+
+.L_len_less_16bytes:
+	zext	r18, r2, 3, 2
+	andi	r2, 3
+	bez	r18, .L_set_by_byte
+.L_len_less_16bytes_loop:
+	stw	r3, (r0, 0)
+	PRE_BNEZAD (r18)
+	addi	r0, 4
+	BNEZAD (r18, .L_len_less_16bytes_loop)
+
+	/* Test if len less than 4 bytes.  */
+.L_set_by_byte:
+	zext	r18, r2, 2, 0
+	bez	r18, .L_return
+.L_set_by_byte_loop:
+	stb	r1, (r0, 0)
+	PRE_BNEZAD (r18)
+	addi	r0, 1
+	BNEZAD (r18, .L_set_by_byte_loop)
+
+.L_return:
+	mov	r0, r12
+	rts
+
+	/* If dest is not aligned, just set some bytes makes the dest
+	   align.  */
+
+.L_dest_not_aligned:
+	sub	r13, r19, r13
+	sub	r2, r13
+.L_dest_not_aligned_loop:
+	/* Makes the dest align.  */
+	stb	r1, (r0, 0)
+	PRE_BNEZAD (r13)
+	addi	r0, 1
+	BNEZAD (r13, .L_dest_not_aligned_loop)
+	cmplti	r2, 8
+	bt	.L_set_by_byte
+	/* Check whether the src is aligned.  */
+	jbr	.L_dest_aligned
+ENDPROC(memset)
+ENDPROC(__memset)
diff --git a/arch/csky/abiv2/strcmp.S b/arch/csky/abiv2/strcmp.S
new file mode 100644
index 000000000..f8403f4d8
--- /dev/null
+++ b/arch/csky/abiv2/strcmp.S
@@ -0,0 +1,168 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+ENTRY(strcmp)
+	mov	a3, a0
+	/* Check if the s1 addr is aligned.  */
+	xor	a2, a3, a1
+	andi	a2, 0x3
+	bnez	a2, 7f
+	andi	t1, a0, 0x3
+	bnez	t1, 5f
+
+1:
+	/* If aligned, load word each time.  */
+	ldw	t0, (a3, 0)
+	ldw	t1, (a1, 0)
+	/* If s1[i] != s2[i], goto 2f.  */
+	cmpne   t0, t1
+	bt      2f
+	/* If s1[i] == s2[i], check if s1 or s2 is at the end.  */
+	tstnbz	t0
+	/* If at the end, goto 3f (finish comparing).  */
+	bf	3f
+
+	ldw	t0, (a3, 4)
+	ldw	t1, (a1, 4)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 8)
+	ldw	t1, (a1, 8)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 12)
+	ldw	t1, (a1, 12)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 16)
+	ldw	t1, (a1, 16)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 20)
+	ldw	t1, (a1, 20)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 24)
+	ldw	t1, (a1, 24)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	ldw	t0, (a3, 28)
+	ldw	t1, (a1, 28)
+	cmpne	t0, t1
+	bt	2f
+	tstnbz	t0
+	bf	3f
+
+	addi	a3, 32
+	addi	a1, 32
+
+	br	1b
+
+# ifdef __CSKYBE__
+	/* d[i] != s[i] in word, so we check byte 0.  */
+2:
+	xtrb0   a0, t0
+	xtrb0   a2, t1
+	subu    a0, a2
+	bez     a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 1 */
+	xtrb1   a0, t0
+	xtrb1   a2, t1
+	subu    a0, a2
+	bez     a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 2 */
+	xtrb2   a0, t0
+	xtrb2   a2, t1
+	subu    a0, a2
+	bez     a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 3 */
+	xtrb3   a0, t0
+	xtrb3   a2, t1
+	subu    a0, a2
+# else
+	/* s1[i] != s2[i] in word, so we check byte 3.  */
+2:
+	xtrb3	a0, t0
+	xtrb3	a2, t1
+	subu    a0, a2
+	bez     a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 2 */
+	xtrb2	a0, t0
+	xtrb2	a2, t1
+	subu    a0, a2
+	bez     a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 1 */
+	xtrb1	a0, t0
+	xtrb1	a2, t1
+	subu	a0, a2
+	bez	a2, 4f
+	bnez    a0, 4f
+
+	/* check byte 0 */
+	xtrb0	a0, t0
+	xtrb0	a2, t1
+	subu	a0, a2
+
+# endif /* !__CSKYBE__ */
+	jmp     lr
+3:
+	movi	a0, 0
+4:
+	jmp     lr
+
+	/* Compare when s1 or s2 is not aligned.  */
+5:
+	subi    t1, 4
+6:
+	ldb	a0, (a3, 0)
+	ldb	a2, (a1, 0)
+	subu	a0, a2
+	bez	a2, 4b
+	bnez	a0, 4b
+	addi    t1, 1
+	addi	a1, 1
+	addi	a3, 1
+	bnez	t1, 6b
+	br	1b
+
+7:
+	ldb	a0, (a3, 0)
+	addi	a3, 1
+	ldb	a2, (a1, 0)
+	addi	a1, 1
+	subu    a0, a2
+	bnez    a0, 4b
+	bnez	a2, 7b
+	jmp	r15
+ENDPROC(strcmp)
diff --git a/arch/csky/abiv2/strcpy.S b/arch/csky/abiv2/strcpy.S
new file mode 100644
index 000000000..3c6d3f6a5
--- /dev/null
+++ b/arch/csky/abiv2/strcpy.S
@@ -0,0 +1,123 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+ENTRY(strcpy)
+	mov	a3, a0
+	/* Check if the src addr is aligned.  */
+        andi    t0, a1, 3
+        bnez	t0, 11f
+1:
+	/* Check if all the bytes in the word are not zero.  */
+	ldw	a2, (a1)
+	tstnbz	a2
+	bf	9f
+	stw	a2, (a3)
+
+	ldw	a2, (a1, 4)
+	tstnbz	a2
+	bf	2f
+	stw	a2, (a3, 4)
+
+	ldw	a2, (a1, 8)
+	tstnbz	a2
+	bf	3f
+	stw	a2, (a3, 8)
+
+	ldw	a2, (a1, 12)
+	tstnbz	a2
+	bf	4f
+	stw	a2, (a3, 12)
+
+	ldw	a2, (a1, 16)
+	tstnbz	a2
+	bf	5f
+	stw	a2, (a3, 16)
+
+	ldw	a2, (a1, 20)
+	tstnbz	a2
+	bf	6f
+	stw	a2, (a3, 20)
+
+	ldw	a2, (a1, 24)
+	tstnbz	a2
+	bf	7f
+	stw	a2, (a3, 24)
+
+	ldw	a2, (a1, 28)
+	tstnbz	a2
+	bf	8f
+	stw	a2, (a3, 28)
+
+	addi	a3, 32
+	addi	a1, 32
+	br	1b
+
+
+2:
+	addi	a3, 4
+	br	9f
+
+3:
+	addi	a3, 8
+	br	9f
+
+4:
+	addi	a3, 12
+	br	9f
+
+5:
+	addi	a3, 16
+	br	9f
+
+6:
+	addi	a3, 20
+	br	9f
+
+7:
+	addi	a3, 24
+	br	9f
+
+8:
+	addi	a3, 28
+9:
+# ifdef __CSKYBE__
+	xtrb0	t0, a2
+	st.b	t0, (a3)
+	bez	t0, 10f
+	xtrb1	t0, a2
+	st.b	t0, (a3, 1)
+	bez	t0, 10f
+	xtrb2	t0, a2
+	st.b	t0, (a3, 2)
+	bez	t0, 10f
+	stw	a2, (a3)
+# else
+	xtrb3	t0, a2
+	st.b	t0, (a3)
+	bez	t0, 10f
+	xtrb2	t0, a2
+	st.b	t0, (a3, 1)
+	bez	t0, 10f
+	xtrb1	t0, a2
+	st.b	t0, (a3, 2)
+	bez	t0, 10f
+	stw	a2, (a3)
+# endif	/* !__CSKYBE__ */
+10:
+	jmp	lr
+
+11:
+	subi    t0, 4
+12:
+        ld.b    a2, (a1)
+        st.b	a2, (a3)
+        bez	a2, 10b
+	addi    t0, 1
+        addi    a1, a1, 1
+        addi    a3, a3, 1
+	bnez	t0, 12b
+	jbr	1b
+ENDPROC(strcpy)
diff --git a/arch/csky/abiv2/strksyms.c b/arch/csky/abiv2/strksyms.c
new file mode 100644
index 000000000..8d1fd28c6
--- /dev/null
+++ b/arch/csky/abiv2/strksyms.c
@@ -0,0 +1,14 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/module.h>
+
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_STRING_OPS
+EXPORT_SYMBOL(memcpy);
+EXPORT_SYMBOL(memset);
+EXPORT_SYMBOL(memmove);
+#endif
+EXPORT_SYMBOL(memcmp);
+EXPORT_SYMBOL(strcmp);
+EXPORT_SYMBOL(strcpy);
+EXPORT_SYMBOL(strlen);
diff --git a/arch/csky/abiv2/strlen.S b/arch/csky/abiv2/strlen.S
new file mode 100644
index 000000000..bcdd70764
--- /dev/null
+++ b/arch/csky/abiv2/strlen.S
@@ -0,0 +1,97 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+// Copyright (C) 2018 Hangzhou C-SKY Microsystems co.,ltd.
+
+#include <linux/linkage.h>
+#include "sysdep.h"
+
+ENTRY(strlen)
+	/* Check if the start addr is aligned.  */
+	mov	r3, r0
+	andi	r1, r0, 3
+	movi	r2, 4
+	movi	r0, 0
+	bnez	r1, .L_start_not_aligned
+
+	LABLE_ALIGN
+.L_start_addr_aligned:
+	/* Check if all the bytes in the word are not zero.  */
+	ldw	r1, (r3)
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 4)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 8)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 12)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 16)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 20)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 24)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	ldw	r1, (r3, 28)
+	addi	r0, 4
+	tstnbz	r1
+	bf	.L_string_tail
+
+	addi	r0, 4
+	addi	r3, 32
+	br	.L_start_addr_aligned
+
+.L_string_tail:
+# ifdef __CSKYBE__
+	xtrb0	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+	xtrb1	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+	xtrb2	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+# else
+	xtrb3	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+	xtrb2	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+	xtrb1	r3, r1
+	bez	r3, .L_return
+	addi	r0, 1
+# endif	/* !__CSKYBE__ */
+
+.L_return:
+	rts
+
+.L_start_not_aligned:
+	sub	r2, r2, r1
+.L_start_not_aligned_loop:
+	ldb	r1, (r3)
+	PRE_BNEZAD (r2)
+	addi	r3, 1
+	bez	r1, .L_return
+	addi	r0, 1
+	BNEZAD (r2, .L_start_not_aligned_loop)
+	br	.L_start_addr_aligned
+ENDPROC(strlen)
diff --git a/arch/csky/abiv2/sysdep.h b/arch/csky/abiv2/sysdep.h
new file mode 100644
index 000000000..61abe9201
--- /dev/null
+++ b/arch/csky/abiv2/sysdep.h
@@ -0,0 +1,29 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __SYSDEP_H
+#define __SYSDEP_H
+
+#ifdef __ASSEMBLER__
+
+#if defined(__CK860__)
+#define LABLE_ALIGN	\
+	.balignw 16, 0x6c03
+
+#define PRE_BNEZAD(R)
+
+#define BNEZAD(R, L)	\
+	bnezad	R, L
+#else
+#define LABLE_ALIGN	\
+	.balignw 8, 0x6c03
+
+#define PRE_BNEZAD(R)	\
+	subi	R, 1
+
+#define BNEZAD(R, L)	\
+	bnez	R, L
+#endif
+
+#endif
+
+#endif