[icp] fpu and asm cleanup for linux

Properly annotate functions and data section so that objtool does not complain
when CONFIG_STACK_VALIDATION and CONFIG_FRAME_POINTER are enabled.

Pass KERNELCPPFLAGS to assembler.

Use kfpu_begin()/kfpu_end() to protect SIMD regions in Linux kernel.

Reviewed-by: Tom Caputi <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Gvozden Neskovic <[email protected]>
Closes #5872 
Closes #5041 

10 files changed, 135 insertions, 307 deletions

diff --git a/module/icp/Makefile.in b/module/icp/Makefile.in
index b822635b7..77b2ec1b5 100644
--- a/module/icp/Makefile.in
+++ b/module/icp/Makefile.in
@@ -18,12 +18,13 @@ endif
 ifeq ($(TARGET_ASM_DIR), asm-i386)
 ASM_SOURCES :=
 endif
-	
+
 ifeq ($(TARGET_ASM_DIR), asm-generic)
 ASM_SOURCES :=
 endif
 
 EXTRA_CFLAGS = $(ZFS_MODULE_CFLAGS) @KERNELCPPFLAGS@
+EXTRA_AFLAGS = $(ZFS_MODULE_CFLAGS) @KERNELCPPFLAGS@
 
 obj-$(CONFIG_ZFS) := $(MODULE).o
 
diff --git a/module/icp/algs/aes/aes_impl.c b/module/icp/algs/aes/aes_impl.c
index 8592386dd..4c17e2a30 100644
--- a/module/icp/algs/aes/aes_impl.c
+++ b/module/icp/algs/aes/aes_impl.c
@@ -22,23 +22,22 @@
  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  */
 
-#include <sys/zfs_context.h>
-#include <sys/crypto/spi.h>
-#include <modes/modes.h>
-#include <aes/aes_impl.h>
-
-#ifdef __amd64
+#if defined(_KERNEL) && defined(__amd64)
+#include <linux/simd_x86.h>
 
-#ifdef _KERNEL
-/* Workaround for no XMM kernel thread save/restore */
-#define	KPREEMPT_DISABLE	kpreempt_disable()
-#define	KPREEMPT_ENABLE		kpreempt_enable()
+#define	KPREEMPT_DISABLE	kfpu_begin()
+#define	KPREEMPT_ENABLE		kfpu_end()
 
 #else
 #define	KPREEMPT_DISABLE
 #define	KPREEMPT_ENABLE
 #endif	/* _KERNEL */
-#endif  /* __amd64 */
+
+#include <sys/zfs_context.h>
+#include <sys/crypto/spi.h>
+#include <modes/modes.h>
+#include <aes/aes_impl.h>
+
 
 
 /*
diff --git a/module/icp/algs/modes/gcm.c b/module/icp/algs/modes/gcm.c
index 80acb6636..c0a26f524 100644
--- a/module/icp/algs/modes/gcm.c
+++ b/module/icp/algs/modes/gcm.c
@@ -22,6 +22,17 @@
  * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
  */
 
+#if defined(_KERNEL) && defined(__amd64)
+#include <linux/simd_x86.h>
+
+#define	KPREEMPT_DISABLE	kfpu_begin()
+#define	KPREEMPT_ENABLE		kfpu_end()
+
+#else
+#define	KPREEMPT_DISABLE
+#define	KPREEMPT_ENABLE
+#endif	/* _KERNEL */
+
 #include <sys/zfs_context.h>
 #include <modes/modes.h>
 #include <sys/crypto/common.h>
@@ -30,16 +41,6 @@
 
 #ifdef __amd64
 
-#ifdef _KERNEL
-/* Workaround for no XMM kernel thread save/restore */
-#define	KPREEMPT_DISABLE	kpreempt_disable()
-#define	KPREEMPT_ENABLE		kpreempt_enable()
-
-#else
-#define	KPREEMPT_DISABLE
-#define	KPREEMPT_ENABLE
-#endif	/* _KERNEL */
-
 extern void gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res);
 static int intel_pclmulqdq_instruction_present(void);
 #endif	/* __amd64 */
diff --git a/module/icp/asm-x86_64/aes/aes_amd64.S b/module/icp/asm-x86_64/aes/aes_amd64.S
index fa66dc321..9db3a3179 100644
--- a/module/icp/asm-x86_64/aes/aes_amd64.S
+++ b/module/icp/asm-x86_64/aes/aes_amd64.S
@@ -693,7 +693,8 @@ aes_decrypt_amd64(const uint32_t rk[], int Nr, const uint32_t ct[4],
  * int aes_encrypt(const unsigned char *in,
  *	unsigned char *out, const aes_encrypt_ctx cx[1])/
  */
-	.align	64
+.data
+.align	64
 enc_tab:
 	enc_vals(u8)
 #ifdef	LAST_ROUND_TABLES
@@ -702,7 +703,7 @@ enc_tab:
 #endif
 
 
-	ENTRY_NP(aes_encrypt_amd64)
+ENTRY_NP(aes_encrypt_amd64)
 #ifdef	GLADMAN_INTERFACE
 	// Original interface
 	sub	$[4*8], %rsp	// gnu/linux/opensolaris binary interface
@@ -797,7 +798,8 @@ enc_tab:
  * int aes_decrypt(const unsigned char *in,
  *	unsigned char *out, const aes_encrypt_ctx cx[1])/
  */
-	.align	64
+.data
+.align	64
 dec_tab:
 	dec_vals(v8)
 #ifdef	LAST_ROUND_TABLES
@@ -806,7 +808,7 @@ dec_tab:
 #endif
 
 
-	ENTRY_NP(aes_decrypt_amd64)
+ENTRY_NP(aes_decrypt_amd64)
 #ifdef	GLADMAN_INTERFACE
 	// Original interface
 	sub	$[4*8], %rsp	// gnu/linux/opensolaris binary interface
diff --git a/module/icp/asm-x86_64/aes/aes_intel.S b/module/icp/asm-x86_64/aes/aes_intel.S
index 6c5c0f919..ed0df75c5 100644
--- a/module/icp/asm-x86_64/aes/aes_intel.S
+++ b/module/icp/asm-x86_64/aes/aes_intel.S
@@ -149,6 +149,7 @@
  * ====================================================================
  */
 
+
 #if defined(lint) || defined(__lint)
 
 #include <sys/types.h>
@@ -182,111 +183,6 @@ rijndael_key_setup_dec_intel(uint32_t rk[], const uint32_t cipherKey[],
 #define _ASM
 #include <sys/asm_linkage.h>
 
-#ifdef _KERNEL
-	/*
-	 * Note: the CLTS macro clobbers P2 (%rsi) under i86xpv.  That is,
-	 * it calls HYPERVISOR_fpu_taskswitch() which modifies %rsi when it
-	 * uses it to pass P2 to syscall.
-	 * This also occurs with the STTS macro, but we dont care if
-	 * P2 (%rsi) is modified just before function exit.
-	 * The CLTS and STTS macros push and pop P1 (%rdi) already.
-	 */
-#ifdef __xpv
-#define	PROTECTED_CLTS \
-	push	%rsi; \
-	CLTS; \
-	pop	%rsi
-#else
-#define	PROTECTED_CLTS \
-	CLTS
-#endif	/* __xpv */
-
-#define	CLEAR_TS_OR_PUSH_XMM0_XMM1(tmpreg) \
-	push	%rbp; \
-	mov	%rsp, %rbp; \
-	movq	%cr0, tmpreg; \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	and	$-XMM_ALIGN, %rsp; \
-	sub	$[XMM_SIZE * 2], %rsp; \
-	movaps	%xmm0, 16(%rsp); \
-	movaps	%xmm1, (%rsp); \
-	jmp	2f; \
-1: \
-	PROTECTED_CLTS; \
-2:
-
-	/*
-	 * If CR0_TS was not set above, pop %xmm0 and %xmm1 off stack,
-	 * otherwise set CR0_TS.
-	 */
-#define	SET_TS_OR_POP_XMM0_XMM1(tmpreg) \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	movaps	(%rsp), %xmm1; \
-	movaps	16(%rsp), %xmm0; \
-	jmp	2f; \
-1: \
-	STTS(tmpreg); \
-2: \
-	mov	%rbp, %rsp; \
-	pop	%rbp
-
-	/*
-	 * If CR0_TS is not set, align stack (with push %rbp) and push
-	 * %xmm0 - %xmm6 on stack, otherwise clear CR0_TS
-	 */
-#define	CLEAR_TS_OR_PUSH_XMM0_TO_XMM6(tmpreg) \
-	push	%rbp; \
-	mov	%rsp, %rbp; \
-	movq	%cr0, tmpreg; \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	and	$-XMM_ALIGN, %rsp; \
-	sub	$[XMM_SIZE * 7], %rsp; \
-	movaps	%xmm0, 96(%rsp); \
-	movaps	%xmm1, 80(%rsp); \
-	movaps	%xmm2, 64(%rsp); \
-	movaps	%xmm3, 48(%rsp); \
-	movaps	%xmm4, 32(%rsp); \
-	movaps	%xmm5, 16(%rsp); \
-	movaps	%xmm6, (%rsp); \
-	jmp	2f; \
-1: \
-	PROTECTED_CLTS; \
-2:
-
-
-	/*
-	 * If CR0_TS was not set above, pop %xmm0 - %xmm6 off stack,
-	 * otherwise set CR0_TS.
-	 */
-#define	SET_TS_OR_POP_XMM0_TO_XMM6(tmpreg) \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	movaps	(%rsp), %xmm6; \
-	movaps	16(%rsp), %xmm5; \
-	movaps	32(%rsp), %xmm4; \
-	movaps	48(%rsp), %xmm3; \
-	movaps	64(%rsp), %xmm2; \
-	movaps	80(%rsp), %xmm1; \
-	movaps	96(%rsp), %xmm0; \
-	jmp	2f; \
-1: \
-	STTS(tmpreg); \
-2: \
-	mov	%rbp, %rsp; \
-	pop	%rbp
-
-
-#else
-#define	PROTECTED_CLTS
-#define	CLEAR_TS_OR_PUSH_XMM0_XMM1(tmpreg)
-#define	SET_TS_OR_POP_XMM0_XMM1(tmpreg)
-#define	CLEAR_TS_OR_PUSH_XMM0_TO_XMM6(tmpreg)
-#define	SET_TS_OR_POP_XMM0_TO_XMM6(tmpreg)
-#endif	/* _KERNEL */
-
 
 /*
  * _key_expansion_128(), * _key_expansion_192a(), _key_expansion_192b(),
@@ -302,9 +198,9 @@ rijndael_key_setup_dec_intel(uint32_t rk[], const uint32_t cipherKey[],
  * (%rcx)	AES key
  */
 
-.align	16
-_key_expansion_128:
-_key_expansion_256a:
+ENTRY_NP2(_key_expansion_128, _key_expansion_256a)
+_key_expansion_128_local:
+_key_expansion_256a_local:
 	pshufd	$0b11111111, %xmm1, %xmm1
 	shufps	$0b00010000, %xmm0, %xmm4
 	pxor	%xmm4, %xmm0
@@ -314,11 +210,13 @@ _key_expansion_256a:
 	movaps	%xmm0, (%rcx)
 	add	$0x10, %rcx
 	ret
-	SET_SIZE(_key_expansion_128)
-	SET_SIZE(_key_expansion_256a)
+	nop
+SET_SIZE(_key_expansion_128)
+SET_SIZE(_key_expansion_256a)
 
-.align 16
-_key_expansion_192a:
+
+ENTRY_NP(_key_expansion_192a)
+_key_expansion_192a_local:
 	pshufd	$0b01010101, %xmm1, %xmm1
 	shufps	$0b00010000, %xmm0, %xmm4
 	pxor	%xmm4, %xmm0
@@ -340,10 +238,11 @@ _key_expansion_192a:
 	movaps	%xmm1, 0x10(%rcx)
 	add	$0x20, %rcx
 	ret
-	SET_SIZE(_key_expansion_192a)
+SET_SIZE(_key_expansion_192a)
+
 
-.align 16
-_key_expansion_192b:
+ENTRY_NP(_key_expansion_192b)
+_key_expansion_192b_local:
 	pshufd	$0b01010101, %xmm1, %xmm1
 	shufps	$0b00010000, %xmm0, %xmm4
 	pxor	%xmm4, %xmm0
@@ -360,10 +259,11 @@ _key_expansion_192b:
 	movaps	%xmm0, (%rcx)
 	add	$0x10, %rcx
 	ret
-	SET_SIZE(_key_expansion_192b)
+SET_SIZE(_key_expansion_192b)
+
 
-.align 16
-_key_expansion_256b:
+ENTRY_NP(_key_expansion_256b)
+_key_expansion_256b_local:
 	pshufd	$0b10101010, %xmm1, %xmm1
 	shufps	$0b00010000, %xmm2, %xmm4
 	pxor	%xmm4, %xmm2
@@ -373,7 +273,7 @@ _key_expansion_256b:
 	movaps	%xmm2, (%rcx)
 	add	$0x10, %rcx
 	ret
-	SET_SIZE(_key_expansion_256b)
+SET_SIZE(_key_expansion_256b)
 
 
 /*
@@ -419,8 +319,7 @@ _key_expansion_256b:
 
 ENTRY_NP(rijndael_key_setup_enc_intel)
 rijndael_key_setup_enc_intel_local:
-	CLEAR_TS_OR_PUSH_XMM0_TO_XMM6(%r10)
-
+	FRAME_BEGIN
 	// NULL pointer sanity check
 	test	%USERCIPHERKEY, %USERCIPHERKEY
 	jz	.Lenc_key_invalid_param
@@ -446,38 +345,38 @@ rijndael_key_setup_enc_intel_local:
 	add	$0x10, %rcx
 
 	aeskeygenassist $0x1, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x1, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x2, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x2, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x4, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x4, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x8, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x8, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x10, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x10, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x20, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 	aeskeygenassist $0x20, %xmm0, %xmm1
-	call	_key_expansion_256b
+	call	_key_expansion_256b_local
 	aeskeygenassist $0x40, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_256a
+	call	_key_expansion_256a_local
 
-	SET_TS_OR_POP_XMM0_TO_XMM6(%r10)
 #ifdef	OPENSSL_INTERFACE
 	xor	%rax, %rax			// return 0 (OK)
 #else	/* Open Solaris Interface */
 	mov	$14, %rax			// return # rounds = 14
 #endif
+	FRAME_END
 	ret
 
 .align 4
@@ -493,28 +392,28 @@ rijndael_key_setup_enc_intel_local:
 
 	movq	0x10(%USERCIPHERKEY), %xmm2	// other user key
 	aeskeygenassist $0x1, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192a
+	call	_key_expansion_192a_local
 	aeskeygenassist $0x2, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192b
+	call	_key_expansion_192b_local
 	aeskeygenassist $0x4, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192a
+	call	_key_expansion_192a_local
 	aeskeygenassist $0x8, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192b
+	call	_key_expansion_192b_local
 	aeskeygenassist $0x10, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192a
+	call	_key_expansion_192a_local
 	aeskeygenassist $0x20, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192b
+	call	_key_expansion_192b_local
 	aeskeygenassist $0x40, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192a
+	call	_key_expansion_192a_local
 	aeskeygenassist $0x80, %xmm2, %xmm1	// expand the key
-	call	_key_expansion_192b
+	call	_key_expansion_192b_local
 
-	SET_TS_OR_POP_XMM0_TO_XMM6(%r10)
 #ifdef	OPENSSL_INTERFACE
 	xor	%rax, %rax			// return 0 (OK)
 #else	/* OpenSolaris Interface */
 	mov	$12, %rax			// return # rounds = 12
 #endif
+	FRAME_END
 	ret
 
 .align 4
@@ -529,51 +428,50 @@ rijndael_key_setup_enc_intel_local:
 #endif	/* OPENSSL_INTERFACE */
 
 	aeskeygenassist $0x1, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x2, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x4, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x8, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x10, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x20, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x40, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x80, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x1b, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 	aeskeygenassist $0x36, %xmm0, %xmm1	// expand the key
-	call	_key_expansion_128
+	call	_key_expansion_128_local
 
-	SET_TS_OR_POP_XMM0_TO_XMM6(%r10)
 #ifdef	OPENSSL_INTERFACE
 	xor	%rax, %rax			// return 0 (OK)
 #else	/* OpenSolaris Interface */
 	mov	$10, %rax			// return # rounds = 10
 #endif
+	FRAME_END
 	ret
 
 .Lenc_key_invalid_param:
 #ifdef	OPENSSL_INTERFACE
-	SET_TS_OR_POP_XMM0_TO_XMM6(%r10)
 	mov	$-1, %rax	// user key or AES key pointer is NULL
+	FRAME_END
 	ret
 #else
 	/* FALLTHROUGH */
 #endif	/* OPENSSL_INTERFACE */
 
 .Lenc_key_invalid_key_bits:
-	SET_TS_OR_POP_XMM0_TO_XMM6(%r10)
 #ifdef	OPENSSL_INTERFACE
 	mov	$-2, %rax	// keysize is invalid
 #else	/* Open Solaris Interface */
 	xor	%rax, %rax	// a key pointer is NULL or invalid keysize
 #endif	/* OPENSSL_INTERFACE */
-
+	FRAME_END
 	ret
 	SET_SIZE(rijndael_key_setup_enc_intel)
 
@@ -599,7 +497,9 @@ rijndael_key_setup_enc_intel_local:
  *	const int bits, AES_KEY *key);
  * Return value is non-zero on error, 0 on success.
  */
+
 ENTRY_NP(rijndael_key_setup_dec_intel)
+FRAME_BEGIN
 	// Generate round keys used for encryption
 	call	rijndael_key_setup_enc_intel_local
 	test	%rax, %rax
@@ -609,8 +509,6 @@ ENTRY_NP(rijndael_key_setup_dec_intel)
 	jz	.Ldec_key_exit	// Failed if returned 0
 #endif	/* OPENSSL_INTERFACE */
 
-	CLEAR_TS_OR_PUSH_XMM0_XMM1(%r10)
-
 	/*
 	 * Convert round keys used for encryption
 	 * to a form usable for decryption
@@ -647,11 +545,10 @@ ENTRY_NP(rijndael_key_setup_dec_intel)
 	cmp	%ENDAESKEY, %rcx
 	jnz	.Ldec_key_inv_loop
 
-	SET_TS_OR_POP_XMM0_XMM1(%r10)
-
 .Ldec_key_exit:
 	// OpenSolaris: rax = # rounds (10, 12, or 14) or 0 for error
 	// OpenSSL: rax = 0 for OK, or non-zero for error
+	FRAME_END
 	ret
 	SET_SIZE(rijndael_key_setup_dec_intel)
 
@@ -701,8 +598,8 @@ ENTRY_NP(rijndael_key_setup_dec_intel)
 #define	STATE		xmm0	/* temporary, 128 bits */
 #define	KEY		xmm1	/* temporary, 128 bits */
 
+
 ENTRY_NP(aes_encrypt_intel)
-	CLEAR_TS_OR_PUSH_XMM0_XMM1(%r10)
 
 	movups	(%INP), %STATE			// input
 	movaps	(%KEYP), %KEY			// key
@@ -759,7 +656,6 @@ ENTRY_NP(aes_encrypt_intel)
 	aesenclast	 %KEY, %STATE		// last round
 	movups	%STATE, (%OUTP)			// output
 
-	SET_TS_OR_POP_XMM0_XMM1(%r10)
 	ret
 	SET_SIZE(aes_encrypt_intel)
 
@@ -787,7 +683,6 @@ ENTRY_NP(aes_encrypt_intel)
  *	const AES_KEY *key);
  */
 ENTRY_NP(aes_decrypt_intel)
-	CLEAR_TS_OR_PUSH_XMM0_XMM1(%r10)
 
 	movups	(%INP), %STATE			// input
 	movaps	(%KEYP), %KEY			// key
@@ -844,7 +739,6 @@ ENTRY_NP(aes_decrypt_intel)
 	aesdeclast	%KEY, %STATE		// last round
 	movups	%STATE, (%OUTP)			// output
 
-	SET_TS_OR_POP_XMM0_XMM1(%r10)
 	ret
 	SET_SIZE(aes_decrypt_intel)
 
diff --git a/module/icp/asm-x86_64/modes/gcm_intel.S b/module/icp/asm-x86_64/modes/gcm_intel.S
index 109f9b47b..a43b5ebcb 100644
--- a/module/icp/asm-x86_64/modes/gcm_intel.S
+++ b/module/icp/asm-x86_64/modes/gcm_intel.S
@@ -95,99 +95,18 @@ gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res) {
 #define _ASM
 #include <sys/asm_linkage.h>
 
-#ifdef _KERNEL
-	/*
-	 * Note: the CLTS macro clobbers P2 (%rsi) under i86xpv.  That is,
-	 * it calls HYPERVISOR_fpu_taskswitch() which modifies %rsi when it
-	 * uses it to pass P2 to syscall.
-	 * This also occurs with the STTS macro, but we dont care if
-	 * P2 (%rsi) is modified just before function exit.
-	 * The CLTS and STTS macros push and pop P1 (%rdi) already.
-	 */
-#ifdef __xpv
-#define	PROTECTED_CLTS \
-	push	%rsi; \
-	CLTS; \
-	pop	%rsi
-#else
-#define	PROTECTED_CLTS \
-	CLTS
-#endif	/* __xpv */
-
-	/*
-	 * If CR0_TS is not set, align stack (with push %rbp) and push
-	 * %xmm0 - %xmm10 on stack, otherwise clear CR0_TS
-	 */
-#define	CLEAR_TS_OR_PUSH_XMM_REGISTERS(tmpreg) \
-	push	%rbp; \
-	mov	%rsp, %rbp; \
-	movq	%cr0, tmpreg; \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	and	$-XMM_ALIGN, %rsp; \
-	sub	$[XMM_SIZE * 11], %rsp; \
-	movaps	%xmm0, 160(%rsp); \
-	movaps	%xmm1, 144(%rsp); \
-	movaps	%xmm2, 128(%rsp); \
-	movaps	%xmm3, 112(%rsp); \
-	movaps	%xmm4, 96(%rsp); \
-	movaps	%xmm5, 80(%rsp); \
-	movaps	%xmm6, 64(%rsp); \
-	movaps	%xmm7, 48(%rsp); \
-	movaps	%xmm8, 32(%rsp); \
-	movaps	%xmm9, 16(%rsp); \
-	movaps	%xmm10, (%rsp); \
-	jmp	2f; \
-1: \
-	PROTECTED_CLTS; \
-2:
-
-
-	/*
-	 * If CR0_TS was not set above, pop %xmm0 - %xmm10 off stack,
-	 * otherwise set CR0_TS.
-	 */
-#define	SET_TS_OR_POP_XMM_REGISTERS(tmpreg) \
-	testq	$CR0_TS, tmpreg; \
-	jnz	1f; \
-	movaps	(%rsp), %xmm10; \
-	movaps	16(%rsp), %xmm9; \
-	movaps	32(%rsp), %xmm8; \
-	movaps	48(%rsp), %xmm7; \
-	movaps	64(%rsp), %xmm6; \
-	movaps	80(%rsp), %xmm5; \
-	movaps	96(%rsp), %xmm4; \
-	movaps	112(%rsp), %xmm3; \
-	movaps	128(%rsp), %xmm2; \
-	movaps	144(%rsp), %xmm1; \
-	movaps	160(%rsp), %xmm0; \
-	jmp	2f; \
-1: \
-	STTS(tmpreg); \
-2: \
-	mov	%rbp, %rsp; \
-	pop	%rbp
-
-
-#else
-#define	PROTECTED_CLTS
-#define	CLEAR_TS_OR_PUSH_XMM_REGISTERS(tmpreg)
-#define	SET_TS_OR_POP_XMM_REGISTERS(tmpreg)
-#endif	/* _KERNEL */
-
 /*
  * Use this mask to byte-swap a 16-byte integer with the pshufb instruction
  */
 
 // static uint8_t byte_swap16_mask[] = {
 //	 15, 14, 13, 12, 11, 10, 9, 8, 7, 6 ,5, 4, 3, 2, 1, 0 };
-.text
+.data
 .align XMM_ALIGN
 .Lbyte_swap16_mask:
 	.byte	15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
 
 
-
 /*
  * void gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res);
  *
@@ -221,8 +140,6 @@ gcm_mul_pclmulqdq(uint64_t *x_in, uint64_t *y, uint64_t *res) {
  */
 
 ENTRY_NP(gcm_mul_pclmulqdq)
-	CLEAR_TS_OR_PUSH_XMM_REGISTERS(%r10)
-
 	//
 	// Copy Parameters
 	//
@@ -325,9 +242,8 @@ ENTRY_NP(gcm_mul_pclmulqdq)
 
 
 	//
-	// Cleanup and Return
+	// Return
 	//
-	SET_TS_OR_POP_XMM_REGISTERS(%r10)
 	ret
 	SET_SIZE(gcm_mul_pclmulqdq)
 
diff --git a/module/icp/asm-x86_64/sha1/sha1-x86_64.S b/module/icp/asm-x86_64/sha1/sha1-x86_64.S
index 6fb4ac5da..cb923784a 100644
--- a/module/icp/asm-x86_64/sha1/sha1-x86_64.S
+++ b/module/icp/asm-x86_64/sha1/sha1-x86_64.S
@@ -1,15 +1,15 @@
 /*
  * !/usr/bin/env perl
- * 
+ *
  *  ====================================================================
  *  Written by Andy Polyakov <[email protected]> for the OpenSSL
  *  project. The module is, however, dual licensed under OpenSSL and
  *  CRYPTOGAMS licenses depending on where you obtain it. For further
  *  details see http://www.openssl.org/~appro/cryptogams/.
  *  ====================================================================
- * 
+ *
  *  sha1_block procedure for x86_64.
- * 
+ *
  *  It was brought to my attention that on EM64T compiler-generated code
  *  was far behind 32-bit assembler implementation. This is unlike on
  *  Opteron where compiler-generated code was only 15% behind 32-bit
@@ -24,29 +24,29 @@
  *  core must have managed to run-time optimize even 32-bit code just as
  *  good as 64-bit one. Performance improvement is summarized in the
  *  following table:
- * 
+ *
  * 		gcc 3.4		32-bit asm	cycles/byte
  *  Opteron	+45%		+20%		6.8
  *  Xeon P4	+65%		+0%		9.9
  *  Core2		+60%		+10%		7.0
- * 
- * 
+ *
+ *
  *  OpenSolaris OS modifications
- * 
+ *
  *  Sun elects to use this software under the BSD license.
- * 
+ *
  *  This source originates from OpenSSL file sha1-x86_64.pl at
  *  ftp://ftp.openssl.org/snapshot/openssl-0.9.8-stable-SNAP-20080131.tar.gz
  *  (presumably for future OpenSSL release 0.9.8h), with these changes:
- * 
+ *
  *  1. Added perl "use strict" and declared variables.
- * 
+ *
  *  2. Added OpenSolaris ENTRY_NP/SET_SIZE macros from
  *  /usr/include/sys/asm_linkage.h, .ident keywords, and lint(1B) guards.
- * 
+ *
  *  3. Removed x86_64-xlate.pl script (not needed for as(1) or gas(1)
  *  assemblers).
- * 
+ *
  */
 
 /*
@@ -58,6 +58,7 @@
 #include <sys/stdint.h>
 #include <sys/sha1.h>
 
+
 /* ARGSUSED */
 void
 sha1_block_data_order(SHA1_CTX *ctx, const void *inpp, size_t blocks)
@@ -86,7 +87,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	16(%r8),%r11d
 .align	4
 .Lloop:
-	mov	0(%r9),%eax	
+	mov	0(%r9),%eax
 	bswap	%eax
 	mov	%eax,0(%rsp)
 	lea	0x5a827999(%eax,%r11d),%r12d
@@ -94,7 +95,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	4(%r9),%eax
 	mov	%edx,%r11d
 	xor	%ebp,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%r11d
 	and	%esi,%ebx
 	mov	%eax,4(%rsp)
@@ -107,7 +108,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	8(%r9),%eax
 	mov	%r12d,%ebp
 	xor	%edi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%ebp
 	and	%edx,%ebx
 	mov	%eax,8(%rsp)
@@ -120,7 +121,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	12(%r9),%eax
 	mov	%r11d,%edi
 	xor	%esi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%edi
 	and	%r12d,%ebx
 	mov	%eax,12(%rsp)
@@ -133,7 +134,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	16(%r9),%eax
 	mov	%ebp,%esi
 	xor	%edx,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%esi
 	and	%r11d,%ebx
 	mov	%eax,16(%rsp)
@@ -146,7 +147,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	20(%r9),%eax
 	mov	%edi,%edx
 	xor	%r12d,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%edx
 	and	%ebp,%ebx
 	mov	%eax,20(%rsp)
@@ -159,7 +160,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	24(%r9),%eax
 	mov	%esi,%r12d
 	xor	%r11d,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%r12d
 	and	%edi,%ebx
 	mov	%eax,24(%rsp)
@@ -172,7 +173,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	28(%r9),%eax
 	mov	%edx,%r11d
 	xor	%ebp,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%r11d
 	and	%esi,%ebx
 	mov	%eax,28(%rsp)
@@ -185,7 +186,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	32(%r9),%eax
 	mov	%r12d,%ebp
 	xor	%edi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%ebp
 	and	%edx,%ebx
 	mov	%eax,32(%rsp)
@@ -198,7 +199,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	36(%r9),%eax
 	mov	%r11d,%edi
 	xor	%esi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%edi
 	and	%r12d,%ebx
 	mov	%eax,36(%rsp)
@@ -211,7 +212,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	40(%r9),%eax
 	mov	%ebp,%esi
 	xor	%edx,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%esi
 	and	%r11d,%ebx
 	mov	%eax,40(%rsp)
@@ -224,7 +225,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	44(%r9),%eax
 	mov	%edi,%edx
 	xor	%r12d,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%edx
 	and	%ebp,%ebx
 	mov	%eax,44(%rsp)
@@ -237,7 +238,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	48(%r9),%eax
 	mov	%esi,%r12d
 	xor	%r11d,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%r12d
 	and	%edi,%ebx
 	mov	%eax,48(%rsp)
@@ -250,7 +251,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	52(%r9),%eax
 	mov	%edx,%r11d
 	xor	%ebp,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%r11d
 	and	%esi,%ebx
 	mov	%eax,52(%rsp)
@@ -263,7 +264,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	56(%r9),%eax
 	mov	%r12d,%ebp
 	xor	%edi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%ebp
 	and	%edx,%ebx
 	mov	%eax,56(%rsp)
@@ -276,7 +277,7 @@ ENTRY_NP(sha1_block_data_order)
 	mov	60(%r9),%eax
 	mov	%r11d,%edi
 	xor	%esi,%ebx
-	bswap	%eax	
+	bswap	%eax
 	rol	$5,%edi
 	and	%r12d,%ebx
 	mov	%eax,60(%rsp)
@@ -1341,6 +1342,8 @@ ENTRY_NP(sha1_block_data_order)
 	pop	%rbx
 	ret
 SET_SIZE(sha1_block_data_order)
+
+.data
 .asciz	"SHA1 block transform for x86_64, CRYPTOGAMS by <[email protected]>"
 
 #endif /* lint || __lint */
diff --git a/module/icp/asm-x86_64/sha2/sha256_impl.S b/module/icp/asm-x86_64/sha2/sha256_impl.S
index 367795668..867dbb0f2 100644
--- a/module/icp/asm-x86_64/sha2/sha256_impl.S
+++ b/module/icp/asm-x86_64/sha2/sha256_impl.S
@@ -62,7 +62,7 @@
  */
 
 /*
- * This file was generated by a perl script (sha512-x86_64.pl) that were 
+ * This file was generated by a perl script (sha512-x86_64.pl) that were
  * used to generate sha256 and sha512 variants from the same code base.
  * The comments from the original file have been pasted above.
  */
@@ -2036,6 +2036,7 @@ ENTRY_NP(SHA256TransformBlocks)
 	ret
 SET_SIZE(SHA256TransformBlocks)
 
+.data
 .align	64
 .type	K256,@object
 K256:
diff --git a/module/icp/asm-x86_64/sha2/sha512_impl.S b/module/icp/asm-x86_64/sha2/sha512_impl.S
index 5a49cff4b..4acc72cbc 100644
--- a/module/icp/asm-x86_64/sha2/sha512_impl.S
+++ b/module/icp/asm-x86_64/sha2/sha512_impl.S
@@ -2037,6 +2037,7 @@ ENTRY_NP(SHA512TransformBlocks)
 	ret
 SET_SIZE(SHA512TransformBlocks)
 
+.data
 .align	64
 .type	K512,@object
 K512:
diff --git a/module/icp/include/sys/asm_linkage.h b/module/icp/include/sys/asm_linkage.h
index 380597857..49a494b46 100644
--- a/module/icp/include/sys/asm_linkage.h
+++ b/module/icp/include/sys/asm_linkage.h
@@ -33,4 +33,14 @@
 
 #endif
 
+#if defined(_KERNEL) && defined(HAVE_KERNEL_OBJTOOL)
+
+#include <asm/frame.h>
+
+#else /* userspace */
+#define	FRAME_BEGIN
+#define	FRAME_END
+#endif
+
+
 #endif	/* _SYS_ASM_LINKAGE_H */