diff --git a/arch/sw_64/Kconfig b/arch/sw_64/Kconfig
index 562bc448becc516fbe027e3e11057c6e281b8fba..deb266745a048076cf4e41b96fa442083ed8225f 100644
--- a/arch/sw_64/Kconfig
+++ b/arch/sw_64/Kconfig
@@ -714,8 +714,9 @@ config ARCH_HIBERNATION_POSSIBLE
 
 config SW64_POWERCAP
 	bool "Sunway powercap driver"
+	select IPMI_HANDLER
 	select IPMI_SI
-	depends on SW64 && CPU_FREQ && ACPI && IPMI_HANDLER
+	depends on SW64_CPUFREQ && ACPI
 	help
 	  This enables support for the sunway powercap driver
 	  based on BMC and IPMI system interface.
diff --git a/arch/sw_64/Makefile b/arch/sw_64/Makefile
index 4208979ceced182d7b8fde48c465f539c1a040fa..c2307116427e2c46c302161eca7381f66a1bd233 100644
--- a/arch/sw_64/Makefile
+++ b/arch/sw_64/Makefile
@@ -17,18 +17,25 @@ archheaders:
 
 NM := $(NM) -B
 CCVERSION := $(shell $(CC) -dumpversion)
-LDFLAGS_vmlinux	:= -static -N #-relax
+LDFLAGS_vmlinux	:= -static -N
 CHECKFLAGS	+= -D__sw__
 
 ifeq ($(CONFIG_RELOCATABLE),y)
 LDFLAGS_vmlinux                        += --emit-relocs
 endif
 
+ifeq ($(CONFIG_FUNCTION_TRACER),y)
+	LDFLAGS_vmlinux += -relax
+endif
+
 CHECKFLAGS	+= -D__sw__
 cflags-y	:= -pipe -ffixed-8 -mno-fp-regs #-msmall-data
 ifneq ($(CONFIG_SUBARCH_C3B),y)
 	cflags-y += -fsw-rev -fsw-unalign-byte
 endif
+cflags-y	+= $(call cc-option, -fsw-memcpy)
+cflags-y	+= $(call cc-option, -fsw-memclr)
+cflags-y	+= $(call cc-option, -fsw-leaf-optimize-fp)
 cflags-y	+= $(call cc-option, -fno-jump-tables)
 
 cflags-y				+= $(cpuflags-y)
diff --git a/arch/sw_64/include/asm/bitops.h b/arch/sw_64/include/asm/bitops.h
index 46b242f4b4babb25863ed3597cdd74e690352995..b48ccfda23a275c8f8ba01e77756e39ab6404b3e 100644
--- a/arch/sw_64/include/asm/bitops.h
+++ b/arch/sw_64/include/asm/bitops.h
@@ -534,6 +534,95 @@ static inline unsigned int __arch_hweight8(unsigned int w)
 	return __arch_hweight64(w & 0xff);
 }
 
+#define find_next_bit_le find_next_bit_le
+#define find_next_bit __find_next_set_bit
+static inline unsigned long __find_next_set_bit(const unsigned long *addr,
+		unsigned long size,
+		unsigned long offset)
+{
+	unsigned long word;
+
+	if (unlikely(offset >= size))
+		return size;
+
+	word = addr[offset / BITS_PER_LONG];
+
+	/* Handle 1st word. */
+	word = word >> (offset % BITS_PER_LONG) << (offset % BITS_PER_LONG);
+
+	offset &= ~(BITS_PER_LONG - 1);
+
+	while (!word) {
+		offset += BITS_PER_LONG;
+		if (offset >= size)
+			return size;
+
+		word = addr[offset / BITS_PER_LONG];
+	}
+
+	return (offset + __ffs(word)) < size ? (offset + __ffs(word)) : size;
+}
+
+#define find_next_and_bit __find_next_and_bit
+static inline unsigned long __find_next_and_bit(const unsigned long *addr1,
+		const unsigned long *addr2, unsigned long size,
+		unsigned long offset)
+{
+	unsigned long word;
+
+	if (unlikely(offset >= size))
+		return size;
+
+	word = addr1[offset / BITS_PER_LONG];
+	word &= addr2[offset / BITS_PER_LONG];
+
+	/* Handle 1st word. */
+	word = word >> (offset % BITS_PER_LONG) << (offset % BITS_PER_LONG);
+
+	offset &= ~(BITS_PER_LONG - 1);
+
+	while (!word) {
+		offset += BITS_PER_LONG;
+		if (offset >= size)
+			return size;
+
+		word = addr1[offset / BITS_PER_LONG];
+		word &= addr2[offset / BITS_PER_LONG];
+	}
+
+	return (offset + __ffs(word)) < size ? (offset + __ffs(word)) : size;
+}
+
+#define find_next_zero_bit_le find_next_zero_bit_le
+#define find_next_zero_bit __find_next_zero_bit
+static inline unsigned long __find_next_zero_bit(const unsigned long *addr, unsigned
+		long size, unsigned long offset)
+{
+	unsigned long word;
+
+	if (unlikely(offset >= size))
+		return size;
+
+	word = addr[offset / BITS_PER_LONG];
+	word ^= ~(0UL);
+
+	/* Handle 1st word. */
+	word = word >> (offset % BITS_PER_LONG) << (offset % BITS_PER_LONG);
+
+	offset &= ~(BITS_PER_LONG - 1);
+
+	while (!word) {
+		offset += BITS_PER_LONG;
+		if (offset >= size)
+			return size;
+
+		word = addr[offset / BITS_PER_LONG];
+		word ^= ~(0UL);
+	}
+
+	return (offset + __ffs(word)) < size ? (offset + __ffs(word)) : size;
+}
+
 #include <asm-generic/bitops/const_hweight.h>
 
 #endif /* __KERNEL__ */
diff --git a/arch/sw_64/include/asm/mmu.h b/arch/sw_64/include/asm/mmu.h
index f24219fac654bb3381ea0e85e5faeeb3486d23d3..3f4a394915fb6940aa98bfd0394b2e8cc3c6ca73 100644
--- a/arch/sw_64/include/asm/mmu.h
+++ b/arch/sw_64/include/asm/mmu.h
@@ -7,4 +7,9 @@ typedef struct {
 	unsigned long asid[NR_CPUS];
 	void *vdso;
 } mm_context_t;
+
+static inline void clear_asid(mm_context_t *mm_context)
+{
+	memset(mm_context->asid, 0, sizeof(mm_context->asid[0]) * nr_cpu_ids);
+}
 #endif /* _ASM_SW64_MMU_H */
diff --git a/arch/sw_64/include/asm/mmu_context.h b/arch/sw_64/include/asm/mmu_context.h
index bfc3dbc5f12fd09ae9361784eff8d830459176f6..a62041aaa4e0ec15df222c2148cce132ec6facd0 100644
--- a/arch/sw_64/include/asm/mmu_context.h
+++ b/arch/sw_64/include/asm/mmu_context.h
@@ -93,10 +93,7 @@ switch_mm(struct mm_struct *prev_mm, struct mm_struct *next_mm,
 static inline int init_new_context(struct task_struct *tsk,
 				   struct mm_struct *mm)
 {
-	int i;
-
-	for_each_possible_cpu(i)
-		mm->context.asid[i] = 0;
+	clear_asid(&mm->context);
 	return 0;
 }
 
diff --git a/arch/sw_64/include/asm/syscall.h b/arch/sw_64/include/asm/syscall.h
index a821bf68be1643d7f457ad738d9cde9cc1930d77..33b4101a81e464b17d3a1a89f0ed07cb6458de20 100644
--- a/arch/sw_64/include/asm/syscall.h
+++ b/arch/sw_64/include/asm/syscall.h
@@ -13,7 +13,7 @@ extern syscall_fn_t sys_call_table[];
 static inline int syscall_get_nr(struct task_struct *task,
 				 struct pt_regs *regs)
 {
-	return regs->regs[0];
+	return regs->orig_r0;
 }
 
 static inline long
diff --git a/arch/sw_64/include/asm/traps.h b/arch/sw_64/include/asm/traps.h
new file mode 100644
index 0000000000000000000000000000000000000000..17064049632f1a3d7e4f2af0ed66f0a8f8ad295b
--- /dev/null
+++ b/arch/sw_64/include/asm/traps.h
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_SW64_TRAPS_H
+#define _ASM_SW64_TRAPS_H
+
+#define ENT_IDX_INT		0
+#define ENT_IDX_ARITH		1
+#define ENT_IDX_MM		2
+#define ENT_IDX_IF		3
+#define ENT_IDX_UNA		4
+#define ENT_IDX_SYS		5
+
+#if defined(CONFIG_SUBARCH_C3B)
+/* C3B does not support NMI, so invalid idx */
+# define ENT_IDX_NMI		(-1)
+# define ENT_IDX_SUSPEND	6
+#elif defined(CONFIG_SUBARCH_C4)
+# define ENT_IDX_NMI		6
+# define ENT_IDX_SUSPEND	7
+#endif
+
+#define TRAP_CAUSE_NMI		(-2)
+#define TRAP_CAUSE_INT		(-1)
+#define TRAP_CAUSE_ARITH	1
+#define TRAP_CAUSE_MM		2
+#define TRAP_CAUSE_IF		3
+#define TRAP_CAUSE_UNA		4
+
+#endif /* _ASM_SW64_TRAPS_H */
diff --git a/arch/sw_64/include/asm/xchg.h b/arch/sw_64/include/asm/xchg.h
index d69dfe893ec4f45626ac4f7a0b9da29aa77d1e3a..0c44f871e2a7b57d539faeb0e0d41020940d09b1 100644
--- a/arch/sw_64/include/asm/xchg.h
+++ b/arch/sw_64/include/asm/xchg.h
@@ -310,6 +310,7 @@ ____xchg(_u32, volatile int *m, unsigned long val)
 	".previous"
 	: [old_v]"=&r"(val), [tmp1]"=&r"(tmp1), [tmp2]"=&r"(tmp2)
 	: [new_v]"r"(val), [m]"r"(m)
+	: "memory"
 	);
 
 	return val;
@@ -332,6 +333,7 @@ ____xchg(_u64, volatile long *m, unsigned long val)
 	".previous"
 	: [old_v]"=&r"(val), [tmp1]"=&r"(tmp1), [tmp2]"=&r"(tmp2)
 	: [new_v]"r"(val), [m]"r"(m)
+	: "memory"
 	);
 
 	return val;
diff --git a/arch/sw_64/include/uapi/asm/kvm.h b/arch/sw_64/include/uapi/asm/kvm.h
index e43f382156f65e667b03f5f39d1cfca8eb1dc8d7..855858e5beca9d9904f15922ecc6d1cad33c8f0d 100644
--- a/arch/sw_64/include/uapi/asm/kvm.h
+++ b/arch/sw_64/include/uapi/asm/kvm.h
@@ -2,6 +2,8 @@
 #ifndef _UAPI_ASM_SW64_KVM_H
 #define _UAPI_ASM_SW64_KVM_H
 
+#include <asm/ptrace.h>
+
 /*
  * KVM SW specific structures and definitions.
  */
@@ -22,6 +24,20 @@ enum SW64_KVM_IRQ {
 
 #define KVM_NR_IRQCHIPS		1
 
+#ifndef __KERNEL__
+struct kvm_regs {
+	union {
+		struct user_pt_regs regs;
+		struct {
+			unsigned long r[31];
+			unsigned long pc;
+			unsigned long ps;
+		};
+	};
+	struct user_fpsimd_state fpstate;
+};
+#endif
+
 /*
  * for KVM_GET_FPU and KVM_SET_FPU
  */
diff --git a/arch/sw_64/kernel/chip_setup.c b/arch/sw_64/kernel/chip_setup.c
index df98e141b14f4f3ed39c4e6a465390521ba11f0c..d4e5b1736bc14d61ba4c1924cbb8ccf83939fa62 100644
--- a/arch/sw_64/kernel/chip_setup.c
+++ b/arch/sw_64/kernel/chip_setup.c
@@ -90,6 +90,7 @@ static void i2c_srst(void)
 	writeq(0x1, spbu_base + OFFSET_I2C2_SRST_L);
 }
 
+#ifdef CONFIG_PCI
 static void pcie_save(void)
 {
 	struct pci_controller *hose;
@@ -130,6 +131,17 @@ static void pcie_restore(void)
 	}
 
 }
+#else
+static void pcie_save(void)
+{
+
+}
+
+static void pcie_restore(void)
+{
+
+}
+#endif
 
 static unsigned long saved_dvc_int, saved_long_time;
 
diff --git a/arch/sw_64/kernel/cpu.c b/arch/sw_64/kernel/cpu.c
index 84b4a66a06e9a32fb9a20267bcde77eb38e9e54d..89961119fa03cd16ec07aa70cc464e051e561758 100644
--- a/arch/sw_64/kernel/cpu.c
+++ b/arch/sw_64/kernel/cpu.c
@@ -70,7 +70,7 @@ static int cpuinfo_cpu_online(unsigned int cpu)
 	return 0;
 }
 
-static const char * __init dmi_get_string(const struct dmi_header *dm, u8 s)
+static __maybe_unused const char * __init dmi_get_string(const struct dmi_header *dm, u8 s)
 {
 	const u8 *bp = ((u8 *) dm) + dm->length;
 	const u8 *nsp;
@@ -90,7 +90,7 @@ static const char * __init dmi_get_string(const struct dmi_header *dm, u8 s)
 	return "";
 }
 
-static void __init find_dmi_processor_version(const struct dmi_header *dm,
+static __maybe_unused void __init find_dmi_processor_version(const struct dmi_header *dm,
 		void *private)
 {
 	char *dmi_data = (char *)dm;
@@ -113,10 +113,12 @@ static void __init get_model_id(void)
 	int i;
 	unsigned long val;
 
+#ifdef CONFIG_EFI
 	/* Prefer model id from SMBIOS */
 	if (!IS_ENABLED(CONFIG_SUBARCH_C3B) &&
 			BIOS_SUPPORT_RESET_CLALLBACK((void *)bios_version))
 		dmi_walk(find_dmi_processor_version, NULL);
+#endif
 
 	if (strlen(model_id) > 0)
 		return;
@@ -128,7 +130,7 @@ static void __init get_model_id(void)
 	}
 }
 
-static void __init find_dmi_processor_manufacturer(const struct dmi_header *dm,
+static __maybe_unused void __init find_dmi_processor_manufacturer(const struct dmi_header *dm,
 		void *private)
 {
 	char *dmi_data = (char *)dm;
@@ -151,10 +153,12 @@ static void __init get_vendor_id(void)
 	int i;
 	unsigned long val;
 
+#ifdef CONFIG_EFI
 	/* Prefer vendor id from SMBIOS */
 	if (!IS_ENABLED(CONFIG_SUBARCH_C3B) &&
 			BIOS_SUPPORT_RESET_CLALLBACK((void *)bios_version))
 		dmi_walk(find_dmi_processor_manufacturer, NULL);
+#endif
 
 	if (strlen(vendor_id) > 0)
 		return;
@@ -166,7 +170,7 @@ static void __init get_vendor_id(void)
 	}
 }
 
-static void __init find_dmi_processor_family(const struct dmi_header *dm,
+static __maybe_unused void __init find_dmi_processor_family(const struct dmi_header *dm,
 		void *private)
 {
 	char *dmi_data = (char *)dm;
@@ -182,10 +186,12 @@ static void __init find_dmi_processor_family(const struct dmi_header *dm,
 
 static void __init get_family(void)
 {
+#ifdef CONFIG_EFI
 	/* Prefer processor family from SMBIOS */
 	if (!IS_ENABLED(CONFIG_SUBARCH_C3B) &&
 			BIOS_SUPPORT_RESET_CLALLBACK((void *)bios_version))
 		dmi_walk(find_dmi_processor_family, NULL);
+#endif
 
 	if (family)
 		return;
diff --git a/arch/sw_64/kernel/irq.c b/arch/sw_64/kernel/irq.c
index 78cac3ce2056c7ab344864974b65401a4b66dcd8..e1476375854397177adb0609b00cd977b2625f64 100644
--- a/arch/sw_64/kernel/irq.c
+++ b/arch/sw_64/kernel/irq.c
@@ -20,6 +20,7 @@
 
 #include <asm/cpu.h>
 #include <asm/irq_impl.h>
+#include <asm/traps.h>
 
 volatile unsigned long irq_err_count;
 DEFINE_PER_CPU(unsigned long, irq_pmi_count);
@@ -173,10 +174,10 @@ void __init init_IRQ(void)
 	unsigned long nmi_stack __maybe_unused = nmi_stack_page ?
 		(unsigned long)page_address(nmi_stack_page) : 0;
 
-	wrent(entInt, 0);
+	wrent(entInt, ENT_IDX_INT);
 	if (IS_ENABLED(CONFIG_SUBARCH_C4) && is_in_host()) {
 		sw64_write_csr_imb(nmi_stack + THREAD_SIZE, CSR_NMI_STACK);
-		wrent(entNMI, 6);
+		wrent(entNMI, ENT_IDX_NMI);
 		set_nmi(INT_PC);
 	}
 
diff --git a/arch/sw_64/kernel/ptrace.c b/arch/sw_64/kernel/ptrace.c
index b137101419a956d2d1d266fcd33d94188ecb3fa7..607a2484dfc2638f7da0c18552d111381f5c7ae4 100644
--- a/arch/sw_64/kernel/ptrace.c
+++ b/arch/sw_64/kernel/ptrace.c
@@ -147,6 +147,20 @@ put_reg(struct task_struct *task, unsigned long regno, unsigned long data)
 		return PTR_ERR(addr);
 	else {
 		*addr = data;
+
+		/*
+		 * Any modifications made to $0 and $19 at syscall entry
+		 * should also be reflected on orig_r0 and orig_r19.
+		 *
+		 * See syscall_set() for how to modify orig_r0 only.
+		 */
+		if (task->ptrace_message == PTRACE_EVENTMSG_SYSCALL_ENTRY) {
+			if (regno == PT_REG_BASE)
+				task_pt_regs(task)->orig_r0 = data;
+			if (regno == (PT_REG_BASE + 19))
+				task_pt_regs(task)->orig_r19 = data;
+		}
+
 		return 0;
 	}
 }
@@ -424,7 +438,6 @@ long arch_ptrace(struct task_struct *child, long request,
 
 asmlinkage unsigned long syscall_trace_enter(void)
 {
-	unsigned long ret = 0;
 	struct pt_regs *regs = current_pt_regs();
 
 	if (test_thread_flag(TIF_SYSCALL_TRACE) &&
@@ -438,9 +451,10 @@ asmlinkage unsigned long syscall_trace_enter(void)
 #endif
 
 	if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
-		trace_sys_enter(regs, regs->regs[0]);
-	audit_syscall_entry(regs->regs[0], regs->regs[16], regs->regs[17], regs->regs[18], regs->regs[19]);
-	return ret ?: regs->regs[0];
+		trace_sys_enter(regs, regs->orig_r0);
+	audit_syscall_entry(regs->orig_r0, regs->regs[16],
+			regs->regs[17], regs->regs[18], regs->regs[19]);
+	return regs->orig_r0;
 }
 
 asmlinkage void
diff --git a/arch/sw_64/kernel/setup.c b/arch/sw_64/kernel/setup.c
index 7d22671749fde244576c74c55f1296076cf7c2aa..506e8d7b4010bb3baf696c40029b458da0333308 100644
--- a/arch/sw_64/kernel/setup.c
+++ b/arch/sw_64/kernel/setup.c
@@ -591,7 +591,7 @@ static void __init setup_firmware_fdt(void)
 
 static void __init setup_cpu_caps(void)
 {
-	if (!IS_ENABLED(CONFIG_SUBARCH_C3B) && !is_junzhang_v1())
+	if (cpuid(GET_FEATURES, 0) & CPU_FEAT_UNA)
 		static_branch_enable(&hw_una_enabled);
 }
 
diff --git a/arch/sw_64/kernel/smp.c b/arch/sw_64/kernel/smp.c
index f0ad2cf3c3e774a7bed6df1e51d8a80bae8b6c59..2c8fd8b6efc1f66d6809985d7b8021324001eb16 100644
--- a/arch/sw_64/kernel/smp.c
+++ b/arch/sw_64/kernel/smp.c
@@ -18,6 +18,7 @@
 #include <asm/sw64_init.h>
 #include <asm/topology.h>
 #include <asm/timer.h>
+#include <asm/traps.h>
 
 #include "proto.h"
 
@@ -37,18 +38,20 @@ void *idle_task_pointer[NR_CPUS];
 /* State of each CPU */
 DEFINE_PER_CPU(int, cpu_state) = { 0 };
 
-/* A collection of single bit ipi messages.  */
-static struct {
-	unsigned long bits ____cacheline_aligned;
-} ipi_data[NR_CPUS] __cacheline_aligned;
-
 enum ipi_message_type {
 	IPI_RESCHEDULE,
 	IPI_CALL_FUNC,
 	IPI_CPU_STOP,
 	IPI_IRQ_WORK,
+	IPI_MAX,
 };
 
+#define IPI_PENDING_TYPE	u64
+/* A collection of single bit ipi messages.  */
+static struct {
+	IPI_PENDING_TYPE pending[IPI_MAX] ____cacheline_aligned;
+} ipi_data[CONFIG_NR_CPUS] __cacheline_aligned;
+
 int smp_num_cpus = 1;		/* Number that came online.  */
 EXPORT_SYMBOL(smp_num_cpus);
 
@@ -64,6 +67,22 @@ enum core_version {
 	CORE_VERSION_RESERVED = 3 /* 3 and greater are reserved */
 };
 
+static inline void get_ipi_pending(int cpu, IPI_PENDING_TYPE *dst)
+{
+	int i;
+
+	/* make sure we can see pending for this ipi */
+	smp_mb();
+	for (i = 0; i < IPI_MAX; i++) {
+		if (READ_ONCE(ipi_data[cpu].pending[i])) {
+			WRITE_ONCE(dst[i], 1);
+			WRITE_ONCE(ipi_data[cpu].pending[i], 0);
+		}
+	}
+	/* make sure we won't miss any pending set after ipi */
+	smp_mb();
+}
+
 #ifdef CONFIG_SUBARCH_C4
 #define OFFSET_CLU_LV2_SELH	0x3a00UL
 #define OFFSET_CLU_LV2_SELL	0x3b00UL
@@ -152,7 +171,7 @@ void smp_callin(void)
 	trap_init();
 
 	/* Set interrupt vector.  */
-	wrent(entInt, 0);
+	wrent(entInt, ENT_IDX_INT);
 
 	/* Get our local ticker going. */
 	sw64_setup_timer();
@@ -174,7 +193,7 @@ void smp_callin(void)
 		nmi_stack = nmi_stack_page ?
 			(unsigned long)page_address(nmi_stack_page) : 0;
 		sw64_write_csr_imb(nmi_stack + THREAD_SIZE, CSR_NMI_STACK);
-		wrent(entNMI, 6);
+		wrent(entNMI, ENT_IDX_NMI);
 		set_nmi(INT_PC);
 	}
 
@@ -593,13 +612,15 @@ static void send_ipi_message(const struct cpumask *to_whom, enum ipi_message_typ
 {
 	int i;
 
-	mb();
-	for_each_cpu(i, to_whom)
-		set_bit(operation, &ipi_data[i].bits);
-
-	mb();
-	for_each_cpu(i, to_whom)
-		send_ipi(i, II_II0);
+	/* make sure ipi data are written before setting pending */
+	smp_mb();
+	for_each_cpu(i, to_whom) {
+		WRITE_ONCE(ipi_data[i].pending[operation], 1);
+		/* make sure pending is set before sending ipi */
+		smp_mb();
+		if (READ_ONCE(ipi_data[i].pending[operation]))
+			send_ipi(i, II_II0);
+	}
 }
 
 static void ipi_cpu_stop(int cpu)
@@ -620,20 +641,13 @@ void arch_irq_work_raise(void)
 void handle_ipi(struct pt_regs *regs)
 {
 	int cpu = smp_processor_id();
-	unsigned long *pending_ipis = &ipi_data[cpu].bits;
-	unsigned long ops;
-
-	mb();	/* Order interrupt and bit testing. */
-	while ((ops = xchg(pending_ipis, 0)) != 0) {
-		mb();	/* Order bit clearing and data access. */
-		do {
-			unsigned long which;
-
-			which = ops & -ops;
-			ops &= ~which;
-			which = __ffs(which);
+	IPI_PENDING_TYPE pending[IPI_MAX] = { 0 };
+	int i;
 
-			switch (which) {
+	get_ipi_pending(cpu, pending);
+	for (i = 0; i < IPI_MAX; i++) {
+		if (pending[i]) {
+			switch (i) {
 			case IPI_RESCHEDULE:
 				scheduler_ipi();
 				break;
@@ -651,12 +665,10 @@ void handle_ipi(struct pt_regs *regs)
 				break;
 
 			default:
-				pr_crit("Unknown IPI on CPU %d: %lu\n", cpu, which);
+				pr_crit("Unknown IPI on CPU %d\n", cpu);
 				break;
 			}
-		} while (ops);
-
-		mb();	/* Order data access and bit testing. */
+		}
 	}
 
 	cpu_data[cpu].ipi_count++;
@@ -723,7 +735,6 @@ static void ipi_flush_tlb_mm(void *x)
 
 void flush_tlb_mm(struct mm_struct *mm)
 {
-
 	/* happens as a result of exit_mmap()
 	 * Shall we clear mm->context.asid[] here?
 	 */
@@ -735,12 +746,11 @@ void flush_tlb_mm(struct mm_struct *mm)
 	if (atomic_read(&mm->mm_users) != 1 || mm != current->mm) {
 		on_each_cpu_mask(mm_cpumask(mm), ipi_flush_tlb_mm, mm, 1);
 	} else {
-		int cpu, this_cpu = smp_processor_id();
+		int cpu = smp_processor_id();
+		unsigned long asid = mm->context.asid[cpu];
 
-		for_each_online_cpu(cpu) {
-			if (cpu != this_cpu && mm->context.asid[cpu])
-				mm->context.asid[cpu] = 0;
-		}
+		clear_asid(&mm->context);
+		mm->context.asid[cpu] = asid;
 		local_flush_tlb_mm(mm);
 	}
 
@@ -775,12 +785,11 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 		};
 		on_each_cpu_mask(mm_cpumask(mm), ipi_flush_tlb_page, &info, 1);
 	} else {
-		int cpu, this_cpu = smp_processor_id();
+		int cpu = smp_processor_id();
+		unsigned long asid = mm->context.asid[cpu];
 
-		for_each_online_cpu(cpu) {
-			if (cpu != this_cpu && mm->context.asid[cpu])
-				mm->context.asid[cpu] = 0;
-		}
+		clear_asid(&mm->context);
+		mm->context.asid[cpu] = asid;
 		local_flush_tlb_page(vma, addr);
 	}
 
diff --git a/arch/sw_64/kernel/sys_sw64.c b/arch/sw_64/kernel/sys_sw64.c
index 1397d86ac7a20e8afb5dfb49b176e5f5c5e89344..09289be850fcf7a12bea2c46755a32803bd013e8 100644
--- a/arch/sw_64/kernel/sys_sw64.c
+++ b/arch/sw_64/kernel/sys_sw64.c
@@ -346,7 +346,7 @@ asmlinkage void noinstr do_entSys(struct pt_regs *regs)
 
 	regs->orig_r0 = regs->regs[0];
 	regs->orig_r19 = regs->regs[19];
-	nr = regs->regs[0];
+	nr = regs->orig_r0;
 
 	if (ti_flags & _TIF_SYSCALL_WORK) {
 		nr = syscall_trace_enter();
@@ -369,8 +369,6 @@ asmlinkage void noinstr do_entSys(struct pt_regs *regs)
 			syscall_set_return_value(current, regs, -ENOSYS, 0);
 		if (nr == NO_SYSCALL)
 			goto syscall_out;
-		regs->orig_r0 = regs->regs[0];
-		regs->orig_r19 = regs->regs[19];
 	}
 
 	if (nr < __NR_syscalls) {
diff --git a/arch/sw_64/kernel/traps.c b/arch/sw_64/kernel/traps.c
index 2fd63a95d6cbc4ce65a841c97741b4e94d5e89e4..a0403130835af4d5258fe1c04824099cd7930da1 100644
--- a/arch/sw_64/kernel/traps.c
+++ b/arch/sw_64/kernel/traps.c
@@ -31,6 +31,7 @@
 #include <asm/debug.h>
 #include <asm/efi.h>
 #include <asm/unistd.h>
+#include <asm/traps.h>
 
 #include "proto.h"
 
@@ -410,13 +411,13 @@ trap_init(void)
 	register unsigned long gptr __asm__("$29");
 	wrkgp(gptr);
 
-	wrent(entArith, 1);
-	wrent(entMM, 2);
-	wrent(entIF, 3);
-	wrent(entUna, 4);
-	wrent(entSys, 5);
+	wrent(entArith, ENT_IDX_ARITH);
+	wrent(entMM, ENT_IDX_MM);
+	wrent(entIF, ENT_IDX_IF);
+	wrent(entUna, ENT_IDX_UNA);
+	wrent(entSys, ENT_IDX_SYS);
 #ifdef CONFIG_EFI
 	if (smp_processor_id() == 0)
-		wrent((void *)entSuspend, 6);
+		wrent((void *)entSuspend, ENT_IDX_SUSPEND);
 #endif
 }
diff --git a/arch/sw_64/lib/iomap.c b/arch/sw_64/lib/iomap.c
index 8377313e1a719dc968cab6a21988dada862eced2..ee8607fa95ecd2425d0701e567da8d499041ff83 100644
--- a/arch/sw_64/lib/iomap.c
+++ b/arch/sw_64/lib/iomap.c
@@ -427,8 +427,8 @@ EXPORT_SYMBOL(_memset_c_io);
 void __iomem *ioport_map(unsigned long port, unsigned int size)
 {
 	if (port >= 0x100000)
-		return ioremap(port, size);
+		return __va(port);
 
-	return ioremap((port << legacy_io_shift) | legacy_io_base, size);
+	return __va((port << legacy_io_shift) | legacy_io_base);
 }
 EXPORT_SYMBOL(ioport_map);
diff --git a/drivers/irqchip/irq-sunway-cpu.c b/drivers/irqchip/irq-sunway-cpu.c
index c0c83833ab9d1d2ff201eb948825195c9cb94ebf..0ba130d979c6b35f2a50543eb52ad29f3509c4c5 100644
--- a/drivers/irqchip/irq-sunway-cpu.c
+++ b/drivers/irqchip/irq-sunway-cpu.c
@@ -11,6 +11,7 @@
 #include <asm/irq_impl.h>
 #include <asm/pmc.h>
 #include <asm/sw64_init.h>
+#include <asm/traps.h>
 
 /**
  * The topology of interrupt controllers of SW64 is as follows:
@@ -87,7 +88,7 @@ asmlinkage void noinstr do_entInt(struct pt_regs *regs)
 	/* restart idle routine if it is interrupted */
 	if (regs->pc > (u64)__idle_start && regs->pc < (u64)__idle_end)
 		regs->pc = (u64)__idle_start;
-	if (unlikely(regs->cause == -2 && IS_ENABLED(CONFIG_SUBARCH_C4)))
+	if (unlikely(regs->cause == TRAP_CAUSE_NMI && IS_ENABLED(CONFIG_SUBARCH_C4)))
 		nmi_enter();
 	else
 		irq_enter();
@@ -176,7 +177,7 @@ asmlinkage void noinstr do_entInt(struct pt_regs *regs)
 
 out:
 	set_irq_regs(old_regs);
-	if (unlikely(regs->cause == -2 && IS_ENABLED(CONFIG_SUBARCH_C4)))
+	if (unlikely(regs->cause == TRAP_CAUSE_NMI && IS_ENABLED(CONFIG_SUBARCH_C4)))
 		nmi_exit();
 	else
 		irq_exit();
diff --git a/drivers/irqchip/irq-sunway-msi-v2.c b/drivers/irqchip/irq-sunway-msi-v2.c
index 119859b5551a1493e9c3b04d2fc214222655fe0f..b32c277c781f84728c269279019e983f6f605aed 100644
--- a/drivers/irqchip/irq-sunway-msi-v2.c
+++ b/drivers/irqchip/irq-sunway-msi-v2.c
@@ -234,6 +234,9 @@ static int __assign_irq_vector(int virq, unsigned int nr_irqs,
 	if (irqd_affinity_is_managed(irq_data)) {
 		mask = irq_data_get_affinity_mask(irq_data);
 		cpumask_and(&searchmask, mask, cpu_online_mask);
+
+		if (cpumask_empty(&searchmask))
+			return -EINVAL;
 	} else {
 		node = irq_data_get_node(irq_data);
 		cpumask_copy(&searchmask, cpumask_of_node(node));
diff --git a/drivers/irqchip/irq-sunway-msi.c b/drivers/irqchip/irq-sunway-msi.c
index 8c678c034c5a1f574bcd28d69b44eedb7f92ba71..17be578d75b634a732440fd769ba131f3c747ac5 100644
--- a/drivers/irqchip/irq-sunway-msi.c
+++ b/drivers/irqchip/irq-sunway-msi.c
@@ -213,6 +213,9 @@ static int __assign_irq_vector(int virq, unsigned int nr_irqs,
 	if (irqd_affinity_is_managed(irq_data)) {
 		mask = irq_data_get_affinity_mask(irq_data);
 		cpumask_and(&searchmask, mask, cpu_online_mask);
+
+		if (cpumask_empty(&searchmask))
+			return -EINVAL;
 	} else {
 		node = irq_data_get_node(irq_data);
 		cpumask_copy(&searchmask, cpumask_of_node(node));
diff --git a/drivers/irqchip/irq-sunway-pci-intx.c b/drivers/irqchip/irq-sunway-pci-intx.c
index 47a42776e736dbd9497a71bce8b58ae9d85aee36..5b3393c26bde72e3e5eba98cb5774d736b4930a4 100644
--- a/drivers/irqchip/irq-sunway-pci-intx.c
+++ b/drivers/irqchip/irq-sunway-pci-intx.c
@@ -60,6 +60,7 @@ static int __assign_piu_intx_config(struct intx_chip_data *chip_data,
 	unsigned int cpu;
 	int thread, node, core, rcid;
 	unsigned int i;
+	struct irq_data *irq_data;
 
 	if (is_guest_or_emul())
 		return 0;
@@ -89,6 +90,9 @@ static int __assign_piu_intx_config(struct intx_chip_data *chip_data,
 					(i << PCI_INTXCONFIG_OFFSET));
 		chip_data->intxconfig[i] = intxconfig;
 	}
+	irq_data = irq_get_irq_data(hose->int_irq);
+	irq_data_update_effective_affinity(irq_data, cpumask_of(cpu));
+
 	return 0;
 }
 
diff --git a/drivers/platform/sw64/powercap.c b/drivers/platform/sw64/powercap.c
index 3f380ace560172536bb13b66f53c8120e5154813..b1856ca24ddc282b3c5a04826e1cfd8911555d85 100644
--- a/drivers/platform/sw64/powercap.c
+++ b/drivers/platform/sw64/powercap.c
@@ -18,6 +18,8 @@
 
 #define SUNWAY_POWERCAP_ACPI_NOTIFY_VALUE 0x84
 
+#define POLL_INTERVAL_UNIT_10MS 10
+
 enum sunway_powercap_version {
 	SUNWAY_POWERCAP_V1 = 1,
 	SUNWAY_POWERCAP_VERSION_MAX,
@@ -243,6 +245,16 @@ static inline bool is_powercap_cpu_match(const struct sunway_powercap_cpu *data,
 	return true;
 }
 
+static inline bool is_powercap_enabled(const struct sunway_powercap_freq *freq)
+{
+	return !!(freq->flags & FREQ_FLAG_ENABLE);
+}
+
+static inline bool is_powercap_no_limit(const struct sunway_powercap_freq *freq)
+{
+	return !!(freq->flags & FREQ_FLAG_FREE);
+}
+
 static int sunway_powercap_validate_freq(const struct sunway_powercap_freq *freq,
 		struct sunway_powercap_ack *ack)
 {
@@ -268,6 +280,10 @@ static int sunway_powercap_validate_freq(const struct sunway_powercap_freq *freq
 		ack->flags |= ACK_FLAG_VALID_NODE;
 		ack->flags |= ACK_FLAG_VALID_CORE;
 
+		/* Make sure that the target freq field is meaningful */
+		if (!is_powercap_enabled(freq) || is_powercap_no_limit(freq))
+			return 0;
+
 		if (cpufreq_frequency_table_get_index(policy, target_freq) < 0) {
 			pr_err("invalid target freq %u\n", target_freq);
 			return -EINVAL;
@@ -284,16 +300,6 @@ static int sunway_powercap_validate_freq(const struct sunway_powercap_freq *freq
 	return -EINVAL;
 }
 
-static inline bool is_powercap_enabled(const struct sunway_powercap_freq *freq)
-{
-	return !!(freq->flags & FREQ_FLAG_ENABLE);
-}
-
-static inline bool is_powercap_no_limit(const struct sunway_powercap_freq *freq)
-{
-	return !!(freq->flags & FREQ_FLAG_FREE);
-}
-
 static inline bool
 is_powercap_freq_data_terminate(const struct sunway_powercap_freq *freq)
 {
@@ -527,7 +533,7 @@ static int sunway_powercap_setup_cfg(const struct sunway_powercap_cfg *cfg)
 
 	driver_data.version = cfg->version;
 	driver_data.mode = cfg->mode;
-	driver_data.poll_interval = cfg->poll_interval;
+	driver_data.poll_interval = cfg->poll_interval * POLL_INTERVAL_UNIT_10MS;
 
 	if (is_poll_mode) {
 		timer_setup(&driver_data.timer, sunway_powercap_poll_func, 0);
@@ -698,10 +704,10 @@ static int sunway_powercap_probe(struct platform_device *pdev)
 		struct freq_qos_request *req;
 
 		/* Initial state */
-		powercap_cpu_data[related_cpu].state = SUNWAY_POWERCAP_STATE_FREE;
+		powercap_cpu_data[cpu].state = SUNWAY_POWERCAP_STATE_FREE;
 
-		powercap_cpu_data[related_cpu].core = rcid_to_core_id(rcid);
-		powercap_cpu_data[related_cpu].node = rcid_to_domain_id(rcid);
+		powercap_cpu_data[cpu].core = rcid_to_core_id(rcid);
+		powercap_cpu_data[cpu].node = rcid_to_domain_id(rcid);
 
 		if (powercap_cpu_data[cpu].policy)
 			continue;