diff --git a/net/wireguard/compat/Kbuild.include b/net/wireguard/compat/Kbuild.include
index 209c0ccd49..0192ecdcbc 100644
--- a/net/wireguard/compat/Kbuild.include
+++ b/net/wireguard/compat/Kbuild.include
@@ -12,6 +12,10 @@ ifeq ($(wildcard $(srctree)/include/linux/ptr_ring.h),)
 ccflags-y += -I$(kbuild-dir)/compat/ptr_ring/include
 endif
 
+ifeq ($(wildcard $(srctree)/include/linux/skb_array.h),)
+ccflags-y += -I$(kbuild-dir)/compat/skb_array/include
+endif
+
 ifeq ($(wildcard $(srctree)/include/linux/siphash.h),)
 ccflags-y += -I$(kbuild-dir)/compat/siphash/include
 wireguard-y += compat/siphash/siphash.o
@@ -65,6 +69,10 @@ ifeq ($(wildcard $(srctree)/arch/arm64/include/asm/neon.h)$(CONFIG_ARM64),y)
 ccflags-y += -I$(kbuild-dir)/compat/neon-arm/include
 endif
 
+ifeq ($(wildcard $(srctree)/include/net/dst_metadata.h),)
+ccflags-y += -I$(kbuild-dir)/compat/dstmetadata/include
+endif
+
 ifeq ($(CONFIG_X86_64),y)
 	ifeq ($(ssse3_instr),)
 		ssse3_instr := $(call as-instr,pshufb %xmm0$(comma)%xmm0,-DCONFIG_AS_SSSE3=1)
diff --git a/net/wireguard/compat/compat.h b/net/wireguard/compat/compat.h
index cd1894e904..36e902b062 100644
--- a/net/wireguard/compat/compat.h
+++ b/net/wireguard/compat/compat.h
@@ -22,9 +22,7 @@
 #endif
 #endif
 #ifdef UTS_UBUNTU_RELEASE_ABI
-#if LINUX_VERSION_CODE == KERNEL_VERSION(3, 13, 11)
-#define ISUBUNTU1404
-#elif LINUX_VERSION_CODE < KERNEL_VERSION(4, 5, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 4, 0)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 5, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 4, 0)
 #define ISUBUNTU1604
 #elif LINUX_VERSION_CODE < KERNEL_VERSION(4, 16, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(4, 15, 0)
 #define ISUBUNTU1804
@@ -219,7 +217,7 @@ static inline void skb_scrub_packet(struct sk_buff *skb, bool xnet)
 #define skb_scrub_packet(a, b) skb_scrub_packet(a)
 #endif
 
-#if ((LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 63) || defined(ISUBUNTU1404)) && !defined(ISRHEL7)
+#if ((LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 63)) && !defined(ISRHEL7)
 #include <linux/random.h>
 static inline u32 __compat_prandom_u32_max(u32 ep_ro)
 {
@@ -268,7 +266,7 @@ static inline u32 __compat_prandom_u32_max(u32 ep_ro)
 #endif
 #endif
 
-#if (LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 3) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 17, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 16, 35) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 15, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 24) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0) && !defined(ISUBUNTU1404)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 33) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 60) && !defined(ISRHEL7))
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(3, 17, 3) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 17, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 16, 35) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 15, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 14, 24) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 33) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)) || (LINUX_VERSION_CODE < KERNEL_VERSION(3, 10, 60) && !defined(ISRHEL7))
 static inline void memzero_explicit(void *s, size_t count)
 {
 	memset(s, 0, count);
@@ -502,7 +500,7 @@ static inline void *__compat_kvzalloc(size_t size, gfp_t flags)
 #define kvzalloc __compat_kvzalloc
 #endif
 
-#if ((LINUX_VERSION_CODE < KERNEL_VERSION(3, 15, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 41)) && !defined(ISUBUNTU1404)
+#if ((LINUX_VERSION_CODE < KERNEL_VERSION(3, 15, 0) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 13, 0)) || LINUX_VERSION_CODE < KERNEL_VERSION(3, 12, 41))
 #include <linux/vmalloc.h>
 #include <linux/mm.h>
 static inline void __compat_kvfree(const void *addr)
diff --git a/net/wireguard/compat/dstmetadata/include/net/dst_metadata.h b/net/wireguard/compat/dstmetadata/include/net/dst_metadata.h
new file mode 100644
index 0000000000..995094d4f0
--- /dev/null
+++ b/net/wireguard/compat/dstmetadata/include/net/dst_metadata.h
@@ -0,0 +1,3 @@
+#ifndef skb_valid_dst
+#define skb_valid_dst(skb) (!!skb_dst(skb))
+#endif
diff --git a/net/wireguard/compat/skb_array/include/linux/skb_array.h b/net/wireguard/compat/skb_array/include/linux/skb_array.h
new file mode 100644
index 0000000000..c91fedcdbf
--- /dev/null
+++ b/net/wireguard/compat/skb_array/include/linux/skb_array.h
@@ -0,0 +1,11 @@
+#ifndef _WG_SKB_ARRAY_H
+#define _WG_SKB_ARRAY_H
+
+#include <linux/skbuff.h>
+
+static void __skb_array_destroy_skb(void *ptr)
+{
+	kfree_skb(ptr);
+}
+
+#endif
diff --git a/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c
index f26ed5d897..8b6872a2f0 100644
--- a/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c
+++ b/net/wireguard/crypto/zinc/curve25519/curve25519-x86_64.c
@@ -34,11 +34,11 @@ static inline u64 add_scalar(u64 *out, const u64 *f1, u64 f2)
 
 	asm volatile(
 		/* Clear registers to propagate the carry bit */
-		"  xor %%r8, %%r8;"
-		"  xor %%r9, %%r9;"
-		"  xor %%r10, %%r10;"
-		"  xor %%r11, %%r11;"
-		"  xor %1, %1;"
+		"  xor %%r8d, %%r8d;"
+		"  xor %%r9d, %%r9d;"
+		"  xor %%r10d, %%r10d;"
+		"  xor %%r11d, %%r11d;"
+		"  xor %k1, %k1;"
 
 		/* Begin addition chain */
 		"  addq 0(%3), %0;"
@@ -52,10 +52,9 @@ static inline u64 add_scalar(u64 *out, const u64 *f1, u64 f2)
 
 		/* Return the carry bit in a register */
 		"  adcx %%r11, %1;"
-	: "+&r" (f2), "=&r" (carry_r)
-	: "r" (out), "r" (f1)
-	: "%r8", "%r9", "%r10", "%r11", "memory", "cc"
-	);
+		: "+&r"(f2), "=&r"(carry_r)
+		: "r"(out), "r"(f1)
+		: "%r8", "%r9", "%r10", "%r11", "memory", "cc");
 
 	return carry_r;
 }
@@ -82,7 +81,7 @@ static inline void fadd(u64 *out, const u64 *f1, const u64 *f2)
 		"  cmovc %0, %%rax;"
 
 		/* Step 2: Add carry*38 to the original sum */
-		"  xor %%rcx, %%rcx;"
+		"  xor %%ecx, %%ecx;"
 		"  add %%rax, %%r8;"
 		"  adcx %%rcx, %%r9;"
 		"  movq %%r9, 8(%1);"
@@ -96,17 +95,16 @@ static inline void fadd(u64 *out, const u64 *f1, const u64 *f2)
 		"  cmovc %0, %%rax;"
 		"  add %%rax, %%r8;"
 		"  movq %%r8, 0(%1);"
-	: "+&r" (f2)
-	: "r" (out), "r" (f1)
-	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc"
-	);
+		: "+&r"(f2)
+		: "r"(out), "r"(f1)
+		: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc");
 }
 
-/* Computes the field substraction of two field elements */
+/* Computes the field subtraction of two field elements */
 static inline void fsub(u64 *out, const u64 *f1, const u64 *f2)
 {
 	asm volatile(
-		/* Compute the raw substraction of f1-f2 */
+		/* Compute the raw subtraction of f1-f2 */
 		"  movq 0(%1), %%r8;"
 		"  subq 0(%2), %%r8;"
 		"  movq 8(%1), %%r9;"
@@ -123,7 +121,7 @@ static inline void fsub(u64 *out, const u64 *f1, const u64 *f2)
 		"  mov $38, %%rcx;"
 		"  cmovc %%rcx, %%rax;"
 
-		/* Step 2: Substract carry*38 from the original difference */
+		/* Step 2: Subtract carry*38 from the original difference */
 		"  sub %%rax, %%r8;"
 		"  sbb $0, %%r9;"
 		"  sbb $0, %%r10;"
@@ -139,10 +137,9 @@ static inline void fsub(u64 *out, const u64 *f1, const u64 *f2)
 		"  movq %%r9, 8(%0);"
 		"  movq %%r10, 16(%0);"
 		"  movq %%r11, 24(%0);"
-	:
-	: "r" (out), "r" (f1), "r" (f2)
-	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc"
-	);
+		:
+		: "r"(out), "r"(f1), "r"(f2)
+		: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "memory", "cc");
 }
 
 /* Computes a field multiplication: out <- f1 * f2
@@ -150,239 +147,400 @@ static inline void fsub(u64 *out, const u64 *f1, const u64 *f2)
 static inline void fmul(u64 *out, const u64 *f1, const u64 *f2, u64 *tmp)
 {
 	asm volatile(
+
 		/* Compute the raw multiplication: tmp <- src1 * src2 */
 
 		/* Compute src1[0] * src2 */
-		"  movq 0(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 0(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 8(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"
+		"  movq 0(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  movq %%r8, 0(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  movq %%r10, 8(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+
 		/* Compute src1[1] * src2 */
-		"  movq 8(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 8(%0), %%r8;"    "  movq %%r8, 8(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 16(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 8(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 8(%2), %%r8;"
+		"  movq %%r8, 8(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 16(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[2] * src2 */
-		"  movq 16(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 16(%0), %%r8;"    "  movq %%r8, 16(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 24(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 16(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 16(%2), %%r8;"
+		"  movq %%r8, 16(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 24(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[3] * src2 */
-		"  movq 24(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 24(%0), %%r8;"    "  movq %%r8, 24(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 32(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  movq %%rbx, 40(%0);"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 48(%0);"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 56(%0);"
+		"  movq 24(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 24(%2), %%r8;"
+		"  movq %%r8, 24(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 32(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  movq %%rbx, 40(%2);"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  movq %%r14, 48(%2);"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+		"  movq %%rax, 56(%2);"
+
 		/* Line up pointers */
-		"  mov %0, %1;"
 		"  mov %2, %0;"
+		"  mov %3, %2;"
 
 		/* Wrap the result back into the field */
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 32(%1), %%r8, %%r13;"
-		"  xor %3, %3;"
-		"  adoxq 0(%1), %%r8;"
-		"  mulxq 40(%1), %%r9, %%rbx;"
+		"  mulxq 32(%0), %%r8, %%r13;"
+		"  xor %k1, %k1;"
+		"  adoxq 0(%0), %%r8;"
+		"  mulxq 40(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 8(%1), %%r9;"
-		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adoxq 8(%0), %%r9;"
+		"  mulxq 48(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 16(%1), %%r10;"
-		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adoxq 16(%0), %%r10;"
+		"  mulxq 56(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 24(%1), %%r11;"
-		"  adcx %3, %%rax;"
-		"  adox %3, %%rax;"
+		"  adoxq 24(%0), %%r11;"
+		"  adcx %1, %%rax;"
+		"  adox %1, %%rax;"
 		"  imul %%rdx, %%rax;"
 
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
-		"  adcx %3, %%r9;"
-		"  movq %%r9, 8(%0);"
-		"  adcx %3, %%r10;"
-		"  movq %%r10, 16(%0);"
-		"  adcx %3, %%r11;"
-		"  movq %%r11, 24(%0);"
+		"  adcx %1, %%r9;"
+		"  movq %%r9, 8(%2);"
+		"  adcx %1, %%r10;"
+		"  movq %%r10, 16(%2);"
+		"  adcx %1, %%r11;"
+		"  movq %%r11, 24(%2);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 0(%0);"
-	: "+&r" (tmp), "+&r" (f1), "+&r" (out), "+&r" (f2)
-	:
-	: "%rax", "%rdx", "%r8", "%r9", "%r10", "%r11", "%rbx", "%r13", "%r14", "memory", "cc"
-	);
+		"  movq %%r8, 0(%2);"
+		: "+&r"(f1), "+&r"(f2), "+&r"(tmp)
+		: "r"(out)
+		: "%rax", "%rbx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r13",
+		  "%r14", "memory", "cc");
 }
 
 /* Computes two field multiplications:
- * out[0] <- f1[0] * f2[0]
- * out[1] <- f1[1] * f2[1]
- * Uses the 16-element buffer tmp for intermediate results. */
+ *   out[0] <- f1[0] * f2[0]
+ *   out[1] <- f1[1] * f2[1]
+ * Uses the 16-element buffer tmp for intermediate results: */
 static inline void fmul2(u64 *out, const u64 *f1, const u64 *f2, u64 *tmp)
 {
 	asm volatile(
+
 		/* Compute the raw multiplication tmp[0] <- f1[0] * f2[0] */
 
 		/* Compute src1[0] * src2 */
-		"  movq 0(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 0(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 8(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"
+		"  movq 0(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  movq %%r8, 0(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  movq %%r10, 8(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+
 		/* Compute src1[1] * src2 */
-		"  movq 8(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 8(%0), %%r8;"    "  movq %%r8, 8(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 16(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 8(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 8(%2), %%r8;"
+		"  movq %%r8, 8(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 16(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[2] * src2 */
-		"  movq 16(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 16(%0), %%r8;"    "  movq %%r8, 16(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 24(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 16(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 16(%2), %%r8;"
+		"  movq %%r8, 16(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 24(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[3] * src2 */
-		"  movq 24(%1), %%rdx;"
-		"  mulxq 0(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 24(%0), %%r8;"    "  movq %%r8, 24(%0);"
-		"  mulxq 8(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 32(%0);"
-		"  mulxq 16(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  movq %%rbx, 40(%0);"    "  mov $0, %%r8;"
-		"  mulxq 24(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 48(%0);"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 56(%0);"
+		"  movq 24(%0), %%rdx;"
+		"  mulxq 0(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 24(%2), %%r8;"
+		"  movq %%r8, 24(%2);"
+		"  mulxq 8(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 32(%2);"
+		"  mulxq 16(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  movq %%rbx, 40(%2);"
+		"  mov $0, %%r8;"
+		"  mulxq 24(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  movq %%r14, 48(%2);"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+		"  movq %%rax, 56(%2);"
 
 		/* Compute the raw multiplication tmp[1] <- f1[1] * f2[1] */
 
 		/* Compute src1[0] * src2 */
-		"  movq 32(%1), %%rdx;"
-		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  movq %%r8, 64(%0);"
-		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  movq %%r10, 72(%0);"
-		"  mulxq 48(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"
-		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"
+		"  movq 32(%0), %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  movq %%r8, 64(%2);"
+		"  mulxq 40(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  movq %%r10, 72(%2);"
+		"  mulxq 48(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  mulxq 56(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+
 		/* Compute src1[1] * src2 */
-		"  movq 40(%1), %%rdx;"
-		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"     "  adcxq 72(%0), %%r8;"    "  movq %%r8, 72(%0);"
-		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 80(%0);"
-		"  mulxq 48(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 40(%0), %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 72(%2), %%r8;"
+		"  movq %%r8, 72(%2);"
+		"  mulxq 40(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 80(%2);"
+		"  mulxq 48(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 56(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[2] * src2 */
-		"  movq 48(%1), %%rdx;"
-		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 80(%0), %%r8;"    "  movq %%r8, 80(%0);"
-		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 88(%0);"
-		"  mulxq 48(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  mov $0, %%r8;"
-		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"
+		"  movq 48(%0), %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 80(%2), %%r8;"
+		"  movq %%r8, 80(%2);"
+		"  mulxq 40(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 88(%2);"
+		"  mulxq 48(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  mov $0, %%r8;"
+		"  mulxq 56(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+
 		/* Compute src1[3] * src2 */
-		"  movq 56(%1), %%rdx;"
-		"  mulxq 32(%3), %%r8, %%r9;"       "  xor %%r10, %%r10;"    "  adcxq 88(%0), %%r8;"    "  movq %%r8, 88(%0);"
-		"  mulxq 40(%3), %%r10, %%r11;"     "  adox %%r9, %%r10;"     "  adcx %%rbx, %%r10;"    "  movq %%r10, 96(%0);"
-		"  mulxq 48(%3), %%rbx, %%r13;"    "  adox %%r11, %%rbx;"    "  adcx %%r14, %%rbx;"    "  movq %%rbx, 104(%0);"    "  mov $0, %%r8;"
-		"  mulxq 56(%3), %%r14, %%rdx;"    "  adox %%r13, %%r14;"    "  adcx %%rax, %%r14;"    "  movq %%r14, 112(%0);"    "  mov $0, %%rax;"
-		                                   "  adox %%rdx, %%rax;"    "  adcx %%r8, %%rax;"     "  movq %%rax, 120(%0);"
+		"  movq 56(%0), %%rdx;"
+		"  mulxq 32(%1), %%r8, %%r9;"
+		"  xor %%r10d, %%r10d;"
+		"  adcxq 88(%2), %%r8;"
+		"  movq %%r8, 88(%2);"
+		"  mulxq 40(%1), %%r10, %%r11;"
+		"  adox %%r9, %%r10;"
+		"  adcx %%rbx, %%r10;"
+		"  movq %%r10, 96(%2);"
+		"  mulxq 48(%1), %%rbx, %%r13;"
+		"  adox %%r11, %%rbx;"
+		"  adcx %%r14, %%rbx;"
+		"  movq %%rbx, 104(%2);"
+		"  mov $0, %%r8;"
+		"  mulxq 56(%1), %%r14, %%rdx;"
+		"  adox %%r13, %%r14;"
+		"  adcx %%rax, %%r14;"
+		"  movq %%r14, 112(%2);"
+		"  mov $0, %%rax;"
+		"  adox %%rdx, %%rax;"
+		"  adcx %%r8, %%rax;"
+		"  movq %%rax, 120(%2);"
+
 		/* Line up pointers */
-		"  mov %0, %1;"
 		"  mov %2, %0;"
+		"  mov %3, %2;"
 
 		/* Wrap the results back into the field */
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 32(%1), %%r8, %%r13;"
-		"  xor %3, %3;"
-		"  adoxq 0(%1), %%r8;"
-		"  mulxq 40(%1), %%r9, %%rbx;"
+		"  mulxq 32(%0), %%r8, %%r13;"
+		"  xor %k1, %k1;"
+		"  adoxq 0(%0), %%r8;"
+		"  mulxq 40(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 8(%1), %%r9;"
-		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adoxq 8(%0), %%r9;"
+		"  mulxq 48(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 16(%1), %%r10;"
-		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adoxq 16(%0), %%r10;"
+		"  mulxq 56(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 24(%1), %%r11;"
-		"  adcx %3, %%rax;"
-		"  adox %3, %%rax;"
+		"  adoxq 24(%0), %%r11;"
+		"  adcx %1, %%rax;"
+		"  adox %1, %%rax;"
 		"  imul %%rdx, %%rax;"
 
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
-		"  adcx %3, %%r9;"
-		"  movq %%r9, 8(%0);"
-		"  adcx %3, %%r10;"
-		"  movq %%r10, 16(%0);"
-		"  adcx %3, %%r11;"
-		"  movq %%r11, 24(%0);"
+		"  adcx %1, %%r9;"
+		"  movq %%r9, 8(%2);"
+		"  adcx %1, %%r10;"
+		"  movq %%r10, 16(%2);"
+		"  adcx %1, %%r11;"
+		"  movq %%r11, 24(%2);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 0(%0);"
+		"  movq %%r8, 0(%2);"
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 96(%1), %%r8, %%r13;"
-		"  xor %3, %3;"
-		"  adoxq 64(%1), %%r8;"
-		"  mulxq 104(%1), %%r9, %%rbx;"
+		"  mulxq 96(%0), %%r8, %%r13;"
+		"  xor %k1, %k1;"
+		"  adoxq 64(%0), %%r8;"
+		"  mulxq 104(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 72(%1), %%r9;"
-		"  mulxq 112(%1), %%r10, %%r13;"
+		"  adoxq 72(%0), %%r9;"
+		"  mulxq 112(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 80(%1), %%r10;"
-		"  mulxq 120(%1), %%r11, %%rax;"
+		"  adoxq 80(%0), %%r10;"
+		"  mulxq 120(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 88(%1), %%r11;"
-		"  adcx %3, %%rax;"
-		"  adox %3, %%rax;"
+		"  adoxq 88(%0), %%r11;"
+		"  adcx %1, %%rax;"
+		"  adox %1, %%rax;"
 		"  imul %%rdx, %%rax;"
 
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
-		"  adcx %3, %%r9;"
-		"  movq %%r9, 40(%0);"
-		"  adcx %3, %%r10;"
-		"  movq %%r10, 48(%0);"
-		"  adcx %3, %%r11;"
-		"  movq %%r11, 56(%0);"
+		"  adcx %1, %%r9;"
+		"  movq %%r9, 40(%2);"
+		"  adcx %1, %%r10;"
+		"  movq %%r10, 48(%2);"
+		"  adcx %1, %%r11;"
+		"  movq %%r11, 56(%2);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 32(%0);"
-	: "+&r" (tmp), "+&r" (f1), "+&r" (out), "+&r" (f2)
-	:
-	: "%rax", "%rdx", "%r8", "%r9", "%r10", "%r11", "%rbx", "%r13", "%r14", "memory", "cc"
-	);
+		"  movq %%r8, 32(%2);"
+		: "+&r"(f1), "+&r"(f2), "+&r"(tmp)
+		: "r"(out)
+		: "%rax", "%rbx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%r13",
+		  "%r14", "memory", "cc");
 }
 
-/* Computes the field multiplication of four-element f1 with value in f2 */
+/* Computes the field multiplication of four-element f1 with value in f2
+ * Requires f2 to be smaller than 2^17 */
 static inline void fmul_scalar(u64 *out, const u64 *f1, u64 f2)
 {
 	register u64 f2_r asm("rdx") = f2;
 
 	asm volatile(
 		/* Compute the raw multiplication of f1*f2 */
-		"  mulxq 0(%2), %%r8, %%rcx;"      /* f1[0]*f2 */
-		"  mulxq 8(%2), %%r9, %%rbx;"      /* f1[1]*f2 */
+		"  mulxq 0(%2), %%r8, %%rcx;" /* f1[0]*f2 */
+		"  mulxq 8(%2), %%r9, %%rbx;" /* f1[1]*f2 */
 		"  add %%rcx, %%r9;"
 		"  mov $0, %%rcx;"
-		"  mulxq 16(%2), %%r10, %%r13;"    /* f1[2]*f2 */
+		"  mulxq 16(%2), %%r10, %%r13;" /* f1[2]*f2 */
 		"  adcx %%rbx, %%r10;"
-		"  mulxq 24(%2), %%r11, %%rax;"    /* f1[3]*f2 */
+		"  mulxq 24(%2), %%r11, %%rax;" /* f1[3]*f2 */
 		"  adcx %%r13, %%r11;"
 		"  adcx %%rcx, %%rax;"
 
@@ -406,17 +564,17 @@ static inline void fmul_scalar(u64 *out, const u64 *f1, u64 f2)
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
 		"  movq %%r8, 0(%1);"
-	: "+&r" (f2_r)
-	: "r" (out), "r" (f1)
-	: "%rax", "%rcx", "%r8", "%r9", "%r10", "%r11", "%rbx", "%r13", "memory", "cc"
-	);
+		: "+&r"(f2_r)
+		: "r"(out), "r"(f1)
+		: "%rax", "%rbx", "%rcx", "%r8", "%r9", "%r10", "%r11", "%r13",
+		  "memory", "cc");
 }
 
 /* Computes p1 <- bit ? p2 : p1 in constant time */
 static inline void cswap2(u64 bit, const u64 *p1, const u64 *p2)
 {
 	asm volatile(
-		/* Invert the polarity of bit to match cmov expectations */
+		/* Transfer bit into CF flag */
 		"  add $18446744073709551615, %0;"
 
 		/* cswap p1[0], p2[0] */
@@ -490,10 +648,9 @@ static inline void cswap2(u64 bit, const u64 *p1, const u64 *p2)
 		"  cmovc %%r10, %%r9;"
 		"  movq %%r8, 56(%1);"
 		"  movq %%r9, 56(%2);"
-	: "+&r" (bit)
-	: "r" (p1), "r" (p2)
-	: "%r8", "%r9", "%r10", "memory", "cc"
-	);
+		: "+&r"(bit)
+		: "r"(p1), "r"(p2)
+		: "%r8", "%r9", "%r10", "memory", "cc");
 }
 
 /* Computes the square of a field element: out <- f * f
@@ -504,18 +661,25 @@ static inline void fsqr(u64 *out, const u64 *f, u64 *tmp)
 		/* Compute the raw multiplication: tmp <- f * f */
 
 		/* Step 1: Compute all partial products */
-		"  movq 0(%1), %%rdx;"                                       /* f[0] */
-		"  mulxq 8(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
-		"  mulxq 16(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
-		"  mulxq 24(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
-		"  movq 24(%1), %%rdx;"                                      /* f[3] */
-		"  mulxq 8(%1), %%r11, %%rbx;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
-		"  mulxq 16(%1), %%rax, %%r13;"    "  adcx %%rax, %%rbx;"    /* f[2]*f[3] */
-		"  movq 8(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
-		"  mulxq 16(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+		"  movq 0(%0), %%rdx;" /* f[0] */
+		"  mulxq 8(%0), %%r8, %%r14;"
+		"  xor %%r15d, %%r15d;" /* f[1]*f[0] */
+		"  mulxq 16(%0), %%r9, %%r10;"
+		"  adcx %%r14, %%r9;" /* f[2]*f[0] */
+		"  mulxq 24(%0), %%rax, %%rcx;"
+		"  adcx %%rax, %%r10;" /* f[3]*f[0] */
+		"  movq 24(%0), %%rdx;" /* f[3] */
+		"  mulxq 8(%0), %%r11, %%rbx;"
+		"  adcx %%rcx, %%r11;" /* f[1]*f[3] */
+		"  mulxq 16(%0), %%rax, %%r13;"
+		"  adcx %%rax, %%rbx;" /* f[2]*f[3] */
+		"  movq 8(%0), %%rdx;"
+		"  adcx %%r15, %%r13;" /* f1 */
+		"  mulxq 16(%0), %%rax, %%rcx;"
+		"  mov $0, %%r14;" /* f[2]*f[1] */
 
 		/* Step 2: Compute two parallel carry chains */
-		"  xor %%r15, %%r15;"
+		"  xor %%r15d, %%r15d;"
 		"  adox %%rax, %%r10;"
 		"  adcx %%r8, %%r8;"
 		"  adox %%rcx, %%r11;"
@@ -530,39 +694,50 @@ static inline void fsqr(u64 *out, const u64 *f, u64 *tmp)
 		"  adcx %%r14, %%r14;"
 
 		/* Step 3: Compute intermediate squares */
-		"  movq 0(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
-		                           "  movq %%rax, 0(%0);"
-		"  add %%rcx, %%r8;"       "  movq %%r8, 8(%0);"
-		"  movq 8(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
-		"  adcx %%rax, %%r9;"      "  movq %%r9, 16(%0);"
-		"  adcx %%rcx, %%r10;"     "  movq %%r10, 24(%0);"
-		"  movq 16(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
-		"  adcx %%rax, %%r11;"     "  movq %%r11, 32(%0);"
-		"  adcx %%rcx, %%rbx;"     "  movq %%rbx, 40(%0);"
-		"  movq 24(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
-		"  adcx %%rax, %%r13;"     "  movq %%r13, 48(%0);"
-		"  adcx %%rcx, %%r14;"     "  movq %%r14, 56(%0);"
+		"  movq 0(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[0]^2 */
+		"  movq %%rax, 0(%1);"
+		"  add %%rcx, %%r8;"
+		"  movq %%r8, 8(%1);"
+		"  movq 8(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[1]^2 */
+		"  adcx %%rax, %%r9;"
+		"  movq %%r9, 16(%1);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 24(%1);"
+		"  movq 16(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[2]^2 */
+		"  adcx %%rax, %%r11;"
+		"  movq %%r11, 32(%1);"
+		"  adcx %%rcx, %%rbx;"
+		"  movq %%rbx, 40(%1);"
+		"  movq 24(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[3]^2 */
+		"  adcx %%rax, %%r13;"
+		"  movq %%r13, 48(%1);"
+		"  adcx %%rcx, %%r14;"
+		"  movq %%r14, 56(%1);"
 
 		/* Line up pointers */
-		"  mov %0, %1;"
-		"  mov %2, %0;"
+		"  mov %1, %0;"
+		"  mov %2, %1;"
 
 		/* Wrap the result back into the field */
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 32(%1), %%r8, %%r13;"
-		"  xor %%rcx, %%rcx;"
-		"  adoxq 0(%1), %%r8;"
-		"  mulxq 40(%1), %%r9, %%rbx;"
+		"  mulxq 32(%0), %%r8, %%r13;"
+		"  xor %%ecx, %%ecx;"
+		"  adoxq 0(%0), %%r8;"
+		"  mulxq 40(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 8(%1), %%r9;"
-		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adoxq 8(%0), %%r9;"
+		"  mulxq 48(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 16(%1), %%r10;"
-		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adoxq 16(%0), %%r10;"
+		"  mulxq 56(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 24(%1), %%r11;"
+		"  adoxq 24(%0), %%r11;"
 		"  adcx %%rcx, %%rax;"
 		"  adox %%rcx, %%rax;"
 		"  imul %%rdx, %%rax;"
@@ -570,43 +745,50 @@ static inline void fsqr(u64 *out, const u64 *f, u64 *tmp)
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
 		"  adcx %%rcx, %%r9;"
-		"  movq %%r9, 8(%0);"
+		"  movq %%r9, 8(%1);"
 		"  adcx %%rcx, %%r10;"
-		"  movq %%r10, 16(%0);"
+		"  movq %%r10, 16(%1);"
 		"  adcx %%rcx, %%r11;"
-		"  movq %%r11, 24(%0);"
+		"  movq %%r11, 24(%1);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 0(%0);"
-	: "+&r,&r" (tmp), "+&r,&r" (f)
-	: "r,m" (out)
-	: "%rax", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%rbx", "%r13", "%r14", "%r15", "memory", "cc"
-	);
+		"  movq %%r8, 0(%1);"
+		: "+&r,&r"(f), "+&r,&r"(tmp)
+		: "r,m"(out)
+		: "%rax", "%rbx", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11",
+		  "%r13", "%r14", "%r15", "memory", "cc");
 }
 
 /* Computes two field squarings:
- * out[0] <- f[0] * f[0]
- * out[1] <- f[1] * f[1]
+ *   out[0] <- f[0] * f[0]
+ *   out[1] <- f[1] * f[1]
  * Uses the 16-element buffer tmp for intermediate results */
 static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
 {
 	asm volatile(
 		/* Step 1: Compute all partial products */
-		"  movq 0(%1), %%rdx;"                                       /* f[0] */
-		"  mulxq 8(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
-		"  mulxq 16(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
-		"  mulxq 24(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
-		"  movq 24(%1), %%rdx;"                                      /* f[3] */
-		"  mulxq 8(%1), %%r11, %%rbx;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
-		"  mulxq 16(%1), %%rax, %%r13;"    "  adcx %%rax, %%rbx;"    /* f[2]*f[3] */
-		"  movq 8(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
-		"  mulxq 16(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+		"  movq 0(%0), %%rdx;" /* f[0] */
+		"  mulxq 8(%0), %%r8, %%r14;"
+		"  xor %%r15d, %%r15d;" /* f[1]*f[0] */
+		"  mulxq 16(%0), %%r9, %%r10;"
+		"  adcx %%r14, %%r9;" /* f[2]*f[0] */
+		"  mulxq 24(%0), %%rax, %%rcx;"
+		"  adcx %%rax, %%r10;" /* f[3]*f[0] */
+		"  movq 24(%0), %%rdx;" /* f[3] */
+		"  mulxq 8(%0), %%r11, %%rbx;"
+		"  adcx %%rcx, %%r11;" /* f[1]*f[3] */
+		"  mulxq 16(%0), %%rax, %%r13;"
+		"  adcx %%rax, %%rbx;" /* f[2]*f[3] */
+		"  movq 8(%0), %%rdx;"
+		"  adcx %%r15, %%r13;" /* f1 */
+		"  mulxq 16(%0), %%rax, %%rcx;"
+		"  mov $0, %%r14;" /* f[2]*f[1] */
 
 		/* Step 2: Compute two parallel carry chains */
-		"  xor %%r15, %%r15;"
+		"  xor %%r15d, %%r15d;"
 		"  adox %%rax, %%r10;"
 		"  adcx %%r8, %%r8;"
 		"  adox %%rcx, %%r11;"
@@ -621,32 +803,50 @@ static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
 		"  adcx %%r14, %%r14;"
 
 		/* Step 3: Compute intermediate squares */
-		"  movq 0(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
-		                           "  movq %%rax, 0(%0);"
-		"  add %%rcx, %%r8;"       "  movq %%r8, 8(%0);"
-		"  movq 8(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
-		"  adcx %%rax, %%r9;"      "  movq %%r9, 16(%0);"
-		"  adcx %%rcx, %%r10;"     "  movq %%r10, 24(%0);"
-		"  movq 16(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
-		"  adcx %%rax, %%r11;"     "  movq %%r11, 32(%0);"
-		"  adcx %%rcx, %%rbx;"     "  movq %%rbx, 40(%0);"
-		"  movq 24(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
-		"  adcx %%rax, %%r13;"     "  movq %%r13, 48(%0);"
-		"  adcx %%rcx, %%r14;"     "  movq %%r14, 56(%0);"
+		"  movq 0(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[0]^2 */
+		"  movq %%rax, 0(%1);"
+		"  add %%rcx, %%r8;"
+		"  movq %%r8, 8(%1);"
+		"  movq 8(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[1]^2 */
+		"  adcx %%rax, %%r9;"
+		"  movq %%r9, 16(%1);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 24(%1);"
+		"  movq 16(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[2]^2 */
+		"  adcx %%rax, %%r11;"
+		"  movq %%r11, 32(%1);"
+		"  adcx %%rcx, %%rbx;"
+		"  movq %%rbx, 40(%1);"
+		"  movq 24(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[3]^2 */
+		"  adcx %%rax, %%r13;"
+		"  movq %%r13, 48(%1);"
+		"  adcx %%rcx, %%r14;"
+		"  movq %%r14, 56(%1);"
 
 		/* Step 1: Compute all partial products */
-		"  movq 32(%1), %%rdx;"                                       /* f[0] */
-		"  mulxq 40(%1), %%r8, %%r14;"      "  xor %%r15, %%r15;"     /* f[1]*f[0] */
-		"  mulxq 48(%1), %%r9, %%r10;"     "  adcx %%r14, %%r9;"     /* f[2]*f[0] */
-		"  mulxq 56(%1), %%rax, %%rcx;"    "  adcx %%rax, %%r10;"    /* f[3]*f[0] */
-		"  movq 56(%1), %%rdx;"                                      /* f[3] */
-		"  mulxq 40(%1), %%r11, %%rbx;"     "  adcx %%rcx, %%r11;"    /* f[1]*f[3] */
-		"  mulxq 48(%1), %%rax, %%r13;"    "  adcx %%rax, %%rbx;"    /* f[2]*f[3] */
-		"  movq 40(%1), %%rdx;"             "  adcx %%r15, %%r13;"    /* f1 */
-		"  mulxq 48(%1), %%rax, %%rcx;"    "  mov $0, %%r14;"        /* f[2]*f[1] */
+		"  movq 32(%0), %%rdx;" /* f[0] */
+		"  mulxq 40(%0), %%r8, %%r14;"
+		"  xor %%r15d, %%r15d;" /* f[1]*f[0] */
+		"  mulxq 48(%0), %%r9, %%r10;"
+		"  adcx %%r14, %%r9;" /* f[2]*f[0] */
+		"  mulxq 56(%0), %%rax, %%rcx;"
+		"  adcx %%rax, %%r10;" /* f[3]*f[0] */
+		"  movq 56(%0), %%rdx;" /* f[3] */
+		"  mulxq 40(%0), %%r11, %%rbx;"
+		"  adcx %%rcx, %%r11;" /* f[1]*f[3] */
+		"  mulxq 48(%0), %%rax, %%r13;"
+		"  adcx %%rax, %%rbx;" /* f[2]*f[3] */
+		"  movq 40(%0), %%rdx;"
+		"  adcx %%r15, %%r13;" /* f1 */
+		"  mulxq 48(%0), %%rax, %%rcx;"
+		"  mov $0, %%r14;" /* f[2]*f[1] */
 
 		/* Step 2: Compute two parallel carry chains */
-		"  xor %%r15, %%r15;"
+		"  xor %%r15d, %%r15d;"
 		"  adox %%rax, %%r10;"
 		"  adcx %%r8, %%r8;"
 		"  adox %%rcx, %%r11;"
@@ -661,37 +861,48 @@ static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
 		"  adcx %%r14, %%r14;"
 
 		/* Step 3: Compute intermediate squares */
-		"  movq 32(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[0]^2 */
-		                           "  movq %%rax, 64(%0);"
-		"  add %%rcx, %%r8;"       "  movq %%r8, 72(%0);"
-		"  movq 40(%1), %%rdx;"     "  mulx %%rdx, %%rax, %%rcx;"    /* f[1]^2 */
-		"  adcx %%rax, %%r9;"      "  movq %%r9, 80(%0);"
-		"  adcx %%rcx, %%r10;"     "  movq %%r10, 88(%0);"
-		"  movq 48(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[2]^2 */
-		"  adcx %%rax, %%r11;"     "  movq %%r11, 96(%0);"
-		"  adcx %%rcx, %%rbx;"     "  movq %%rbx, 104(%0);"
-		"  movq 56(%1), %%rdx;"    "  mulx %%rdx, %%rax, %%rcx;"    /* f[3]^2 */
-		"  adcx %%rax, %%r13;"     "  movq %%r13, 112(%0);"
-		"  adcx %%rcx, %%r14;"     "  movq %%r14, 120(%0);"
+		"  movq 32(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[0]^2 */
+		"  movq %%rax, 64(%1);"
+		"  add %%rcx, %%r8;"
+		"  movq %%r8, 72(%1);"
+		"  movq 40(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[1]^2 */
+		"  adcx %%rax, %%r9;"
+		"  movq %%r9, 80(%1);"
+		"  adcx %%rcx, %%r10;"
+		"  movq %%r10, 88(%1);"
+		"  movq 48(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[2]^2 */
+		"  adcx %%rax, %%r11;"
+		"  movq %%r11, 96(%1);"
+		"  adcx %%rcx, %%rbx;"
+		"  movq %%rbx, 104(%1);"
+		"  movq 56(%0), %%rdx;"
+		"  mulx %%rdx, %%rax, %%rcx;" /* f[3]^2 */
+		"  adcx %%rax, %%r13;"
+		"  movq %%r13, 112(%1);"
+		"  adcx %%rcx, %%r14;"
+		"  movq %%r14, 120(%1);"
 
 		/* Line up pointers */
-		"  mov %0, %1;"
-		"  mov %2, %0;"
+		"  mov %1, %0;"
+		"  mov %2, %1;"
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 32(%1), %%r8, %%r13;"
-		"  xor %%rcx, %%rcx;"
-		"  adoxq 0(%1), %%r8;"
-		"  mulxq 40(%1), %%r9, %%rbx;"
+		"  mulxq 32(%0), %%r8, %%r13;"
+		"  xor %%ecx, %%ecx;"
+		"  adoxq 0(%0), %%r8;"
+		"  mulxq 40(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 8(%1), %%r9;"
-		"  mulxq 48(%1), %%r10, %%r13;"
+		"  adoxq 8(%0), %%r9;"
+		"  mulxq 48(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 16(%1), %%r10;"
-		"  mulxq 56(%1), %%r11, %%rax;"
+		"  adoxq 16(%0), %%r10;"
+		"  mulxq 56(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 24(%1), %%r11;"
+		"  adoxq 24(%0), %%r11;"
 		"  adcx %%rcx, %%rax;"
 		"  adox %%rcx, %%rax;"
 		"  imul %%rdx, %%rax;"
@@ -699,32 +910,32 @@ static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
 		"  adcx %%rcx, %%r9;"
-		"  movq %%r9, 8(%0);"
+		"  movq %%r9, 8(%1);"
 		"  adcx %%rcx, %%r10;"
-		"  movq %%r10, 16(%0);"
+		"  movq %%r10, 16(%1);"
 		"  adcx %%rcx, %%r11;"
-		"  movq %%r11, 24(%0);"
+		"  movq %%r11, 24(%1);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 0(%0);"
+		"  movq %%r8, 0(%1);"
 
 		/* Step 1: Compute dst + carry == tmp_hi * 38 + tmp_lo */
 		"  mov $38, %%rdx;"
-		"  mulxq 96(%1), %%r8, %%r13;"
-		"  xor %%rcx, %%rcx;"
-		"  adoxq 64(%1), %%r8;"
-		"  mulxq 104(%1), %%r9, %%rbx;"
+		"  mulxq 96(%0), %%r8, %%r13;"
+		"  xor %%ecx, %%ecx;"
+		"  adoxq 64(%0), %%r8;"
+		"  mulxq 104(%0), %%r9, %%rbx;"
 		"  adcx %%r13, %%r9;"
-		"  adoxq 72(%1), %%r9;"
-		"  mulxq 112(%1), %%r10, %%r13;"
+		"  adoxq 72(%0), %%r9;"
+		"  mulxq 112(%0), %%r10, %%r13;"
 		"  adcx %%rbx, %%r10;"
-		"  adoxq 80(%1), %%r10;"
-		"  mulxq 120(%1), %%r11, %%rax;"
+		"  adoxq 80(%0), %%r10;"
+		"  mulxq 120(%0), %%r11, %%rax;"
 		"  adcx %%r13, %%r11;"
-		"  adoxq 88(%1), %%r11;"
+		"  adoxq 88(%0), %%r11;"
 		"  adcx %%rcx, %%rax;"
 		"  adox %%rcx, %%rax;"
 		"  imul %%rdx, %%rax;"
@@ -732,21 +943,21 @@ static inline void fsqr2(u64 *out, const u64 *f, u64 *tmp)
 		/* Step 2: Fold the carry back into dst */
 		"  add %%rax, %%r8;"
 		"  adcx %%rcx, %%r9;"
-		"  movq %%r9, 40(%0);"
+		"  movq %%r9, 40(%1);"
 		"  adcx %%rcx, %%r10;"
-		"  movq %%r10, 48(%0);"
+		"  movq %%r10, 48(%1);"
 		"  adcx %%rcx, %%r11;"
-		"  movq %%r11, 56(%0);"
+		"  movq %%r11, 56(%1);"
 
 		/* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
 		"  mov $0, %%rax;"
 		"  cmovc %%rdx, %%rax;"
 		"  add %%rax, %%r8;"
-		"  movq %%r8, 32(%0);"
-	: "+&r,&r" (tmp), "+&r,&r" (f)
-	: "r,m" (out)
-	: "%rax", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11", "%rbx", "%r13", "%r14", "%r15", "memory", "cc"
-	);
+		"  movq %%r8, 32(%1);"
+		: "+&r,&r"(f), "+&r,&r"(tmp)
+		: "r,m"(out)
+		: "%rax", "%rbx", "%rcx", "%rdx", "%r8", "%r9", "%r10", "%r11",
+		  "%r13", "%r14", "%r15", "memory", "cc");
 }
 
 static void point_add_and_double(u64 *q, u64 *p01_tmp1, u64 *tmp2)
diff --git a/net/wireguard/device.c b/net/wireguard/device.c
index ece4ad2db8..062490f1b8 100644
--- a/net/wireguard/device.c
+++ b/net/wireguard/device.c
@@ -19,6 +19,7 @@
 #include <linux/if_arp.h>
 #include <linux/icmp.h>
 #include <linux/suspend.h>
+#include <net/dst_metadata.h>
 #include <net/icmp.h>
 #include <net/rtnetlink.h>
 #include <net/ip_tunnels.h>
@@ -160,7 +161,7 @@ static netdev_tx_t wg_xmit(struct sk_buff *skb, struct net_device *dev)
 		goto err_peer;
 	}
 
-	mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
+	mtu = skb_valid_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 
 	__skb_queue_head_init(&packets);
 	if (!skb_is_gso(skb)) {
diff --git a/net/wireguard/queueing.c b/net/wireguard/queueing.c
index 1de413b19e..8084e7408c 100644
--- a/net/wireguard/queueing.c
+++ b/net/wireguard/queueing.c
@@ -4,6 +4,7 @@
  */
 
 #include "queueing.h"
+#include <linux/skb_array.h>
 
 struct multicore_worker __percpu *
 wg_packet_percpu_multicore_worker_alloc(work_func_t function, void *ptr)
@@ -42,7 +43,7 @@ void wg_packet_queue_free(struct crypt_queue *queue, bool purge)
 {
 	free_percpu(queue->worker);
 	WARN_ON(!purge && !__ptr_ring_empty(&queue->ring));
-	ptr_ring_cleanup(&queue->ring, purge ? (void(*)(void*))kfree_skb : NULL);
+	ptr_ring_cleanup(&queue->ring, purge ? __skb_array_destroy_skb : NULL);
 }
 
 #define NEXT(skb) ((skb)->prev)
diff --git a/net/wireguard/socket.c b/net/wireguard/socket.c
index bd887f33a3..9e0af9320c 100644
--- a/net/wireguard/socket.c
+++ b/net/wireguard/socket.c
@@ -160,6 +160,7 @@ static int send6(struct wg_device *wg, struct sk_buff *skb,
 	rcu_read_unlock_bh();
 	return ret;
 #else
+	kfree_skb(skb);
 	return -EAFNOSUPPORT;
 #endif
 }
@@ -241,7 +242,7 @@ int wg_socket_endpoint_from_skb(struct endpoint *endpoint,
 		endpoint->addr4.sin_addr.s_addr = ip_hdr(skb)->saddr;
 		endpoint->src4.s_addr = ip_hdr(skb)->daddr;
 		endpoint->src_if4 = skb->skb_iif;
-	} else if (skb->protocol == htons(ETH_P_IPV6)) {
+	} else if (IS_ENABLED(CONFIG_IPV6) && skb->protocol == htons(ETH_P_IPV6)) {
 		endpoint->addr6.sin6_family = AF_INET6;
 		endpoint->addr6.sin6_port = udp_hdr(skb)->source;
 		endpoint->addr6.sin6_addr = ipv6_hdr(skb)->saddr;
@@ -284,7 +285,7 @@ void wg_socket_set_peer_endpoint(struct wg_peer *peer,
 		peer->endpoint.addr4 = endpoint->addr4;
 		peer->endpoint.src4 = endpoint->src4;
 		peer->endpoint.src_if4 = endpoint->src_if4;
-	} else if (endpoint->addr.sa_family == AF_INET6) {
+	} else if (IS_ENABLED(CONFIG_IPV6) && endpoint->addr.sa_family == AF_INET6) {
 		peer->endpoint.addr6 = endpoint->addr6;
 		peer->endpoint.src6 = endpoint->src6;
 	} else {
diff --git a/net/wireguard/tests/netns.sh b/net/wireguard/tests/netns.sh
index eff6030ae1..f21d59924a 100755
--- a/net/wireguard/tests/netns.sh
+++ b/net/wireguard/tests/netns.sh
@@ -22,10 +22,12 @@
 # interfaces in $ns1 and $ns2. See https://www.wireguard.com/netns/ for further
 # details on how this is accomplished.
 set -e
+shopt -s extglob
 
 exec 3>&1
 export LANG=C
 export WG_HIDE_KEYS=never
+NPROC=( /sys/devices/system/cpu/cpu+([0-9]) ); NPROC=${#NPROC[@]}
 netns0="wg-test-$$-0"
 netns1="wg-test-$$-1"
 netns2="wg-test-$$-2"
@@ -147,17 +149,15 @@ tests() {
 	[[ $(< /proc/version) =~ ^Linux\ version\ 5\.4[.\ ] ]] || return 0
 
 	# TCP over IPv4, in parallel
-	for max in 4 5 50; do
-		local pids=( )
-		for ((i=0; i < max; ++i)) do
-			n2 iperf3 -p $(( 5200 + i )) -s -1 -B 192.168.241.2 &
-			pids+=( $! ); waitiperf $netns2 $! $(( 5200 + i ))
-		done
-		for ((i=0; i < max; ++i)) do
-			n1 iperf3 -Z -t 3 -p $(( 5200 + i )) -c 192.168.241.2 &
-		done
-		wait "${pids[@]}"
+	local pids=( ) i
+	for ((i=0; i < NPROC; ++i)) do
+		n2 iperf3 -p $(( 5200 + i )) -s -1 -B 192.168.241.2 &
+		pids+=( $! ); waitiperf $netns2 $! $(( 5200 + i ))
 	done
+	for ((i=0; i < NPROC; ++i)) do
+		n1 iperf3 -Z -t 3 -p $(( 5200 + i )) -c 192.168.241.2 &
+	done
+	wait "${pids[@]}"
 }
 
 [[ $(ip1 link show dev wg0) =~ mtu\ ([0-9]+) ]] && orig_mtu="${BASH_REMATCH[1]}"
@@ -284,7 +284,19 @@ read _ _ tx_bytes_before < <(n0 wg show wg1 transfer)
 ! n0 ping -W 1 -c 10 -f 192.168.241.2 || false
 sleep 1
 read _ _ tx_bytes_after < <(n0 wg show wg1 transfer)
-(( tx_bytes_after - tx_bytes_before < 70000 ))
+if ! (( tx_bytes_after - tx_bytes_before < 70000 )); then
+	errstart=$'\x1b[37m\x1b[41m\x1b[1m'
+	errend=$'\x1b[0m'
+	echo "${errstart}                                                ${errend}"
+	echo "${errstart}                   E  R  R  O  R                ${errend}"
+	echo "${errstart}                                                ${errend}"
+	echo "${errstart} This architecture does not do the right thing  ${errend}"
+	echo "${errstart} with cross-namespace routing loops. This test  ${errend}"
+	echo "${errstart} has thus technically failed but, as this issue ${errend}"
+	echo "${errstart} is as yet unsolved, these tests will continue  ${errend}"
+	echo "${errstart} onward. :(                                     ${errend}"
+	echo "${errstart}                                                ${errend}"
+fi
 
 ip0 link del wg1
 ip1 link del wg0
diff --git a/net/wireguard/tests/qemu/Makefile b/net/wireguard/tests/qemu/Makefile
index 2a92e0b0cb..2846b92923 100644
--- a/net/wireguard/tests/qemu/Makefile
+++ b/net/wireguard/tests/qemu/Makefile
@@ -86,8 +86,10 @@ CROSS_COMPILE_FLAG := --build=$(CBUILD) --host=$(CHOST)
 export CROSS_COMPILE=$(CBUILD)-
 STRIP := $(CBUILD)-strip
 endif
+QEMU_VPORT_RESULT :=
 ifeq ($(ARCH),aarch64)
 QEMU_ARCH := aarch64
+QEMU_VPORT_RESULT := virtio-serial-device
 KERNEL_ARCH := arm64
 KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm64/boot/Image
 ifeq ($(HOST_ARCH),$(ARCH))
@@ -98,6 +100,7 @@ CFLAGS += -march=armv8-a -mtune=cortex-a53
 endif
 else ifeq ($(ARCH),aarch64_be)
 QEMU_ARCH := aarch64
+QEMU_VPORT_RESULT := virtio-serial-device
 KERNEL_ARCH := arm64
 KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm64/boot/Image
 ifeq ($(HOST_ARCH),$(ARCH))
@@ -108,6 +111,7 @@ CFLAGS += -march=armv8-a -mtune=cortex-a53
 endif
 else ifeq ($(ARCH),arm)
 QEMU_ARCH := arm
+QEMU_VPORT_RESULT := virtio-serial-device
 KERNEL_ARCH := arm
 KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm/boot/zImage
 ifeq ($(HOST_ARCH),$(ARCH))
@@ -118,6 +122,7 @@ CFLAGS += -march=armv7-a -mtune=cortex-a15 -mabi=aapcs-linux
 endif
 else ifeq ($(ARCH),armeb)
 QEMU_ARCH := arm
+QEMU_VPORT_RESULT := virtio-serial-device
 KERNEL_ARCH := arm
 KERNEL_BZIMAGE := $(KERNEL_PATH)/arch/arm/boot/zImage
 ifeq ($(HOST_ARCH),$(ARCH))
@@ -217,7 +222,7 @@ KERNEL_ARCH := m68k
 KERNEL_BZIMAGE := $(KERNEL_PATH)/vmlinux
 KERNEL_CMDLINE := $(shell sed -n 's/CONFIG_CMDLINE=\(.*\)/\1/p' arch/m68k.config)
 ifeq ($(HOST_ARCH),$(ARCH))
-QEMU_MACHINE := -cpu host,accel=kvm -machine q800 -smp 1 -append $(KERNEL_CMDLINE)
+QEMU_MACHINE := -cpu host,accel=kvm -machine q800 -append $(KERNEL_CMDLINE)
 else
 QEMU_MACHINE := -machine q800 -smp 1 -append $(KERNEL_CMDLINE)
 endif
@@ -230,6 +235,7 @@ MUSL_CC := $(BUILD_PATH)/musl-gcc
 export CC := $(MUSL_CC)
 USERSPACE_DEPS := $(MUSL_CC) $(BUILD_PATH)/include/.installed $(BUILD_PATH)/include/linux/.installed
 
+comma := ,
 build: $(KERNEL_BZIMAGE)
 qemu: $(KERNEL_BZIMAGE)
 	rm -f $(BUILD_PATH)/result
@@ -240,7 +246,8 @@ qemu: $(KERNEL_BZIMAGE)
 		$(QEMU_MACHINE) \
 		-m $$(grep -q CONFIG_DEBUG_KMEMLEAK=y $(KERNEL_PATH)/.config && echo 1G || echo 256M) \
 		-serial stdio \
-		-serial file:$(BUILD_PATH)/result \
+		-chardev file,path=$(BUILD_PATH)/result,id=result \
+		$(if $(QEMU_VPORT_RESULT),-device $(QEMU_VPORT_RESULT) -device virtserialport$(comma)chardev=result,-serial chardev:result) \
 		-no-reboot \
 		-monitor none \
 		-kernel $<
@@ -277,12 +284,6 @@ $(KERNEL_PATH)/.installed: $(KERNEL_TAR)
 	printf 'ifdef CONFIG_X86_64\nLDFLAGS += $$(call ld-option, -z max-page-size=0x200000)\nendif\n' >> $(KERNEL_PATH)/arch/x86/Makefile
 	sed -i 's/^Elf_Addr per_cpu_load_addr;$$/static \0/' $(KERNEL_PATH)/arch/x86/tools/relocs.c || true
 	if grep -sqr UTS_UBUNTU_RELEASE_ABI $(KERNEL_PATH)/debian/rules.d; then echo 'KBUILD_CFLAGS += -DUTS_UBUNTU_RELEASE_ABI=0' >> $(KERNEL_PATH)/Makefile; fi
-	if grep -sq 'RHEL_MAJOR = 8' $(KERNEL_PATH)/Makefile.rhelver; then \
-		sed -i '/#include <asm\//a #include <asm\/acpi.h>/' $(KERNEL_PATH)/arch/x86/kernel/{apic/apic.c,irqinit.c,kvm.c,mpparse.c} && \
-		sed -i '/#include <asm\//a #include <asm\/numa.h>/' $(KERNEL_PATH)/arch/x86/kernel/setup.c && \
-		sed -i '/irq_hv_callback_count/d' $(KERNEL_PATH)/arch/x86/kernel/kvm.c && \
-		sed -i '/do_vmm_communication/d' $(KERNEL_PATH)/arch/x86/entry/entry_64.S; \
-	fi
 	sed -i "/^if INET\$$/a source \"net/wireguard/Kconfig\"" $(KERNEL_PATH)/net/Kconfig
 	sed -i "/^obj-\$$(CONFIG_NETFILTER).*+=/a obj-\$$(CONFIG_WIREGUARD) += wireguard/" $(KERNEL_PATH)/net/Makefile
 	ln -sfT $(shell readlink -f ../..) $(KERNEL_PATH)/net/wireguard
diff --git a/net/wireguard/tests/qemu/arch/aarch64.config b/net/wireguard/tests/qemu/arch/aarch64.config
index 3d063bb247..09016880ce 100644
--- a/net/wireguard/tests/qemu/arch/aarch64.config
+++ b/net/wireguard/tests/qemu/arch/aarch64.config
@@ -1,5 +1,8 @@
 CONFIG_SERIAL_AMBA_PL011=y
 CONFIG_SERIAL_AMBA_PL011_CONSOLE=y
+CONFIG_VIRTIO_MENU=y
+CONFIG_VIRTIO_MMIO=y
+CONFIG_VIRTIO_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyAMA0 wg.success=ttyAMA1"
+CONFIG_CMDLINE="console=ttyAMA0 wg.success=vport0p1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1280
diff --git a/net/wireguard/tests/qemu/arch/aarch64_be.config b/net/wireguard/tests/qemu/arch/aarch64_be.config
index dbdc7e406a..19ff66e4c6 100644
--- a/net/wireguard/tests/qemu/arch/aarch64_be.config
+++ b/net/wireguard/tests/qemu/arch/aarch64_be.config
@@ -1,6 +1,9 @@
 CONFIG_CPU_BIG_ENDIAN=y
 CONFIG_SERIAL_AMBA_PL011=y
 CONFIG_SERIAL_AMBA_PL011_CONSOLE=y
+CONFIG_VIRTIO_MENU=y
+CONFIG_VIRTIO_MMIO=y
+CONFIG_VIRTIO_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyAMA0 wg.success=ttyAMA1"
+CONFIG_CMDLINE="console=ttyAMA0 wg.success=vport0p1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1280
diff --git a/net/wireguard/tests/qemu/arch/arm.config b/net/wireguard/tests/qemu/arch/arm.config
index 148f499054..fc7959bef9 100644
--- a/net/wireguard/tests/qemu/arch/arm.config
+++ b/net/wireguard/tests/qemu/arch/arm.config
@@ -4,6 +4,9 @@ CONFIG_ARCH_VIRT=y
 CONFIG_THUMB2_KERNEL=n
 CONFIG_SERIAL_AMBA_PL011=y
 CONFIG_SERIAL_AMBA_PL011_CONSOLE=y
+CONFIG_VIRTIO_MENU=y
+CONFIG_VIRTIO_MMIO=y
+CONFIG_VIRTIO_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyAMA0 wg.success=ttyAMA1"
+CONFIG_CMDLINE="console=ttyAMA0 wg.success=vport0p1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/armeb.config b/net/wireguard/tests/qemu/arch/armeb.config
index bd76b07d00..f3066be81c 100644
--- a/net/wireguard/tests/qemu/arch/armeb.config
+++ b/net/wireguard/tests/qemu/arch/armeb.config
@@ -4,7 +4,10 @@ CONFIG_ARCH_VIRT=y
 CONFIG_THUMB2_KERNEL=n
 CONFIG_SERIAL_AMBA_PL011=y
 CONFIG_SERIAL_AMBA_PL011_CONSOLE=y
+CONFIG_VIRTIO_MENU=y
+CONFIG_VIRTIO_MMIO=y
+CONFIG_VIRTIO_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyAMA0 wg.success=ttyAMA1"
+CONFIG_CMDLINE="console=ttyAMA0 wg.success=vport0p1 panic_on_warn=1"
 CONFIG_CPU_BIG_ENDIAN=y
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/i686.config b/net/wireguard/tests/qemu/arch/i686.config
index a85025d720..6d90892a85 100644
--- a/net/wireguard/tests/qemu/arch/i686.config
+++ b/net/wireguard/tests/qemu/arch/i686.config
@@ -1,5 +1,6 @@
+CONFIG_ACPI=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/m68k.config b/net/wireguard/tests/qemu/arch/m68k.config
index 62a15bdb87..82c925e49b 100644
--- a/net/wireguard/tests/qemu/arch/m68k.config
+++ b/net/wireguard/tests/qemu/arch/m68k.config
@@ -5,5 +5,5 @@ CONFIG_MAC=y
 CONFIG_SERIAL_PMACZILOG=y
 CONFIG_SERIAL_PMACZILOG_TTYS=y
 CONFIG_SERIAL_PMACZILOG_CONSOLE=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/mips.config b/net/wireguard/tests/qemu/arch/mips.config
index df71d6b955..d7ec63c17b 100644
--- a/net/wireguard/tests/qemu/arch/mips.config
+++ b/net/wireguard/tests/qemu/arch/mips.config
@@ -7,5 +7,5 @@ CONFIG_POWER_RESET_SYSCON=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/mips64.config b/net/wireguard/tests/qemu/arch/mips64.config
index 90c783f725..0994947e33 100644
--- a/net/wireguard/tests/qemu/arch/mips64.config
+++ b/net/wireguard/tests/qemu/arch/mips64.config
@@ -10,5 +10,5 @@ CONFIG_POWER_RESET_SYSCON=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1280
diff --git a/net/wireguard/tests/qemu/arch/mips64el.config b/net/wireguard/tests/qemu/arch/mips64el.config
index 435b0b43e0..591184342f 100644
--- a/net/wireguard/tests/qemu/arch/mips64el.config
+++ b/net/wireguard/tests/qemu/arch/mips64el.config
@@ -11,5 +11,5 @@ CONFIG_POWER_RESET_SYSCON=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1280
diff --git a/net/wireguard/tests/qemu/arch/mipsel.config b/net/wireguard/tests/qemu/arch/mipsel.config
index 62bb50c4a8..18a4982937 100644
--- a/net/wireguard/tests/qemu/arch/mipsel.config
+++ b/net/wireguard/tests/qemu/arch/mipsel.config
@@ -8,5 +8,5 @@ CONFIG_POWER_RESET_SYSCON=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/powerpc.config b/net/wireguard/tests/qemu/arch/powerpc.config
index 57957093b7..5e04882e8e 100644
--- a/net/wireguard/tests/qemu/arch/powerpc.config
+++ b/net/wireguard/tests/qemu/arch/powerpc.config
@@ -6,5 +6,5 @@ CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_MATH_EMULATION=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1024
diff --git a/net/wireguard/tests/qemu/arch/powerpc64le.config b/net/wireguard/tests/qemu/arch/powerpc64le.config
index f52f1e2bc7..8148b9d122 100644
--- a/net/wireguard/tests/qemu/arch/powerpc64le.config
+++ b/net/wireguard/tests/qemu/arch/powerpc64le.config
@@ -7,7 +7,7 @@ CONFIG_PPC_RADIX_MMU=y
 CONFIG_HVC_CONSOLE=y
 CONFIG_CPU_LITTLE_ENDIAN=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=hvc0 wg.success=hvc1"
+CONFIG_CMDLINE="console=hvc0 wg.success=hvc1 panic_on_warn=1"
 CONFIG_SECTION_MISMATCH_WARN_ONLY=y
 CONFIG_FRAME_WARN=1280
 CONFIG_THREAD_SHIFT=14
diff --git a/net/wireguard/tests/qemu/arch/x86_64.config b/net/wireguard/tests/qemu/arch/x86_64.config
index 00a1ef4869..efa00693e0 100644
--- a/net/wireguard/tests/qemu/arch/x86_64.config
+++ b/net/wireguard/tests/qemu/arch/x86_64.config
@@ -1,5 +1,6 @@
+CONFIG_ACPI=y
 CONFIG_SERIAL_8250=y
 CONFIG_SERIAL_8250_CONSOLE=y
 CONFIG_CMDLINE_BOOL=y
-CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1"
+CONFIG_CMDLINE="console=ttyS0 wg.success=ttyS1 panic_on_warn=1"
 CONFIG_FRAME_WARN=1280
diff --git a/net/wireguard/tests/qemu/init.c b/net/wireguard/tests/qemu/init.c
index 3e2a237e6c..6be8064cf9 100644
--- a/net/wireguard/tests/qemu/init.c
+++ b/net/wireguard/tests/qemu/init.c
@@ -58,27 +58,17 @@ static void print_banner(void)
 
 static void seed_rng(void)
 {
-	int fd;
-	struct {
-		int entropy_count;
-		int buffer_size;
-		unsigned char buffer[256];
-	} entropy = {
-		.entropy_count = sizeof(entropy.buffer) * 8,
-		.buffer_size = sizeof(entropy.buffer),
-		.buffer = "Adding real entropy is not actually important for these tests. Don't try this at home, kids!"
-	};
+	int bits = 4096, fd;
 
-	if (mknod("/dev/urandom", S_IFCHR | 0644, makedev(1, 9)))
-		panic("mknod(/dev/urandom)");
-	fd = open("/dev/urandom", O_WRONLY);
+	pretty_message("[+] Fake seeding RNG...");
+	fd = open("/dev/random", O_WRONLY);
 	if (fd < 0)
-		panic("open(urandom)");
+		panic("open(random)");
 	for (;;) {
-		if (getrandom(entropy.buffer, sizeof(entropy.buffer), GRND_NONBLOCK) != -1 || errno != EAGAIN)
+		if (!getrandom(NULL, 0, GRND_NONBLOCK) || errno == ENOSYS)
 			break;
-		if (ioctl(fd, RNDADDENTROPY, &entropy) < 0)
-			panic("ioctl(urandom)");
+		if (ioctl(fd, RNDADDTOENTCNT, &bits) < 0)
+			panic("ioctl(RNDADDTOENTCNT)");
 	}
 	close(fd);
 }
@@ -126,12 +116,6 @@ static void enable_logging(void)
 			panic("write(exception-trace)");
 		close(fd);
 	}
-	fd = open("/proc/sys/kernel/panic_on_warn", O_WRONLY);
-	if (fd >= 0) {
-		if (write(fd, "1\n", 2) != 2)
-			panic("write(panic_on_warn)");
-		close(fd);
-	}
 }
 
 static void kmod_selftests(void)
@@ -274,10 +258,10 @@ static void check_leaks(void)
 
 int main(int argc, char *argv[])
 {
-	seed_rng();
 	ensure_console();
 	print_banner();
 	mount_filesystems();
+	seed_rng();
 	kmod_selftests();
 	enable_logging();
 	clear_leaks();