Fix up new x86 string functions.

4 files changed, 151 insertions, 38 deletions

diff --git a/sysdeps/i386/i686/multiarch/memcmp-sse4.S b/sysdeps/i386/i686/multiarch/memcmp-sse4.S
index 06437e484c..71c4e1c337 100644
--- a/sysdeps/i386/i686/multiarch/memcmp-sse4.S
+++ b/sysdeps/i386/i686/multiarch/memcmp-sse4.S
@@ -105,43 +105,43 @@ L(less8bytes):
 	mov	1(%eax), %bl
 	cmpb	1(%edx), %bl
 	jne	L(nonzero)
-	
-	cmp	$2, %ecx		
+
+	cmp	$2, %ecx
 	jz	L(0bytes)
 
 	mov	2(%eax), %bl
 	cmpb	2(%edx), %bl
 	jne	L(nonzero)
-	
-	cmp	$3, %ecx		
+
+	cmp	$3, %ecx
 	jz	L(0bytes)
-	
+
 	mov	3(%eax), %bl
 	cmpb	3(%edx), %bl
 	jne	L(nonzero)
-	
-	cmp	$4, %ecx		
+
+	cmp	$4, %ecx
 	jz	L(0bytes)
-	
+
 	mov	4(%eax), %bl
 	cmpb	4(%edx), %bl
 	jne	L(nonzero)
 
-	cmp	$5, %ecx		
+	cmp	$5, %ecx
 	jz	L(0bytes)
-	
+
 	mov	5(%eax), %bl
 	cmpb	5(%edx), %bl
 	jne	L(nonzero)
 
-	cmp	$6, %ecx		
+	cmp	$6, %ecx
 	jz	L(0bytes)
-	
+
 	mov	6(%eax), %bl
 	cmpb	6(%edx), %bl
 	je	L(0bytes)
 L(nonzero):
-	POP (%ebx)	
+	POP (%ebx)
 	mov	$1, %eax
 	ja	L(above)
 	neg	%eax
@@ -151,11 +151,11 @@ L(above):
 
 	ALIGN (4)
 L(0bytes):
-	POP (%ebx)	
+	POP (%ebx)
 	xor	%eax, %eax
 	ret
 	CFI_PUSH (%ebx)
-	
+
 	ALIGN (4)
 L(less1bytes):
 	jb	L(0bytesend)
@@ -609,7 +609,7 @@ L(26bytes):
 	mov	-6(%edx), %ebx
 	cmp	%ebx, %ecx
 	jne	L(find_diff)
-	
+
 	movzwl	-2(%eax), %ecx
 	movzwl	-2(%edx), %ebx
 	cmp	%bl, %cl
@@ -873,7 +873,7 @@ L(32bytes):
 L(less16bytes):
 	add	%ebx, %eax
 	add	%ebx, %edx
-	
+
 	mov	(%eax), %ecx
 	mov	(%edx), %ebx
 	cmp	%ebx, %ecx
@@ -908,7 +908,7 @@ L(find_diff):
 	jne	L(end)
 	cmp	%bx, %cx
 L(end):
-	POP (%ebx)	
+	POP (%ebx)
 	mov	$1, %eax
 	ja	L(bigger)
 	neg	%eax
diff --git a/sysdeps/i386/i686/multiarch/memcmp-ssse3.S b/sysdeps/i386/i686/multiarch/memcmp-ssse3.S
index bfcf660729..869f37a912 100644
--- a/sysdeps/i386/i686/multiarch/memcmp-ssse3.S
+++ b/sysdeps/i386/i686/multiarch/memcmp-ssse3.S
@@ -43,8 +43,7 @@
 #define BLK2		BLK1+4
 #define LEN		BLK2+4
 #define RETURN_END	POP (%edi); POP (%esi); POP (%ebx); ret
-#define RETURN		RETURN_END; CFI_PUSH (%ebx); CFI_PUSH (%edi); \
-			CFI_PUSH (%esi)
+#define RETURN		RETURN_END; cfi_restore_state; cfi_remember_state
 
 	.section .text.ssse3,"ax",@progbits
 ENTRY (MEMCMP)
@@ -76,12 +75,13 @@ L(1bytesend):
 L(zero):
 	mov	$0, %eax
 	ret
-	
+
 	ALIGN (4)
 L(48bytesormore):
 	PUSH (%ebx)
 	PUSH (%esi)
 	PUSH (%edi)
+	cfi_remember_state
 	movdqu    (%eax), %xmm3
 	movdqu    (%edx), %xmm0
 	movl	%eax, %edi
@@ -155,7 +155,7 @@ L(shr_0):
 	add	$32, %esi
 	sub	$0xffff, %edx
 	jnz	L(exit)
-	
+
 
 	lea	(%ecx, %edi,1), %eax
 	lea	(%ecx, %esi,1), %edx
@@ -163,6 +163,8 @@ L(shr_0):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_0_gobble):
 	lea	-48(%ecx), %ecx
@@ -207,6 +209,8 @@ L(shr_0_gobble_loop_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_1):
 	cmp	$80, %ecx
@@ -235,6 +239,8 @@ L(shr_1):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_1_gobble):
 	sub	$32, %ecx
@@ -286,6 +292,8 @@ L(shr_1_gobble_next):
 	jmp	L(less48bytes)
 
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_2):
 	cmp	$80, %ecx
@@ -314,6 +322,8 @@ L(shr_2):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_2_gobble):
 	sub	$32, %ecx
@@ -364,6 +374,8 @@ L(shr_2_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_3):
 	cmp	$80, %ecx
@@ -392,6 +404,8 @@ L(shr_3):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_3_gobble):
 	sub	$32, %ecx
@@ -442,6 +456,8 @@ L(shr_3_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_4):
 	cmp	$80, %ecx
@@ -470,6 +486,8 @@ L(shr_4):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_4_gobble):
 	sub	$32, %ecx
@@ -520,6 +538,8 @@ L(shr_4_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_5):
 	cmp	$80, %ecx
@@ -548,6 +568,8 @@ L(shr_5):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_5_gobble):
 	sub	$32, %ecx
@@ -598,6 +620,8 @@ L(shr_5_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_6):
 	cmp	$80, %ecx
@@ -626,6 +650,8 @@ L(shr_6):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_6_gobble):
 	sub	$32, %ecx
@@ -676,6 +702,8 @@ L(shr_6_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_7):
 	cmp	$80, %ecx
@@ -704,6 +732,8 @@ L(shr_7):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_7_gobble):
 	sub	$32, %ecx
@@ -754,6 +784,8 @@ L(shr_7_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_8):
 	cmp	$80, %ecx
@@ -782,6 +814,8 @@ L(shr_8):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_8_gobble):
 	sub	$32, %ecx
@@ -832,6 +866,8 @@ L(shr_8_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_9):
 	cmp	$80, %ecx
@@ -860,6 +896,8 @@ L(shr_9):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_9_gobble):
 	sub	$32, %ecx
@@ -910,6 +948,8 @@ L(shr_9_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_10):
 	cmp	$80, %ecx
@@ -938,6 +978,8 @@ L(shr_10):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_10_gobble):
 	sub	$32, %ecx
@@ -988,6 +1030,8 @@ L(shr_10_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_11):
 	cmp	$80, %ecx
@@ -1016,6 +1060,8 @@ L(shr_11):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_11_gobble):
 	sub	$32, %ecx
@@ -1066,6 +1112,8 @@ L(shr_11_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_12):
 	cmp	$80, %ecx
@@ -1094,6 +1142,8 @@ L(shr_12):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_12_gobble):
 	sub	$32, %ecx
@@ -1144,6 +1194,8 @@ L(shr_12_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_13):
 	cmp	$80, %ecx
@@ -1172,6 +1224,8 @@ L(shr_13):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_13_gobble):
 	sub	$32, %ecx
@@ -1222,6 +1276,8 @@ L(shr_13_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_14):
 	cmp	$80, %ecx
@@ -1250,6 +1306,8 @@ L(shr_14):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_14_gobble):
 	sub	$32, %ecx
@@ -1300,6 +1358,8 @@ L(shr_14_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_15):
 	cmp	$80, %ecx
@@ -1328,6 +1388,8 @@ L(shr_15):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(shr_15_gobble):
 	sub	$32, %ecx
@@ -1378,6 +1440,8 @@ L(shr_15_gobble_next):
 	POP (%esi)
 	jmp	L(less48bytes)
 
+	cfi_restore_state
+	cfi_remember_state
 	ALIGN (4)
 L(exit):
 	pmovmskb %xmm1, %ebx
@@ -1497,8 +1561,9 @@ L(Byte31):
 	movzbl	 -9(%edi), %eax
 	movzbl	 -9(%esi), %edx
 	sub	%edx, %eax
-	RETURN
+	RETURN_END
 
+	CFI_PUSH (%ebx)
 	ALIGN (4)
 L(more8bytes):
 	cmp	$16, %ecx
diff --git a/sysdeps/i386/i686/multiarch/strcmp-sse4.S b/sysdeps/i386/i686/multiarch/strcmp-sse4.S
index 977647203f..4b47851ed4 100644
--- a/sysdeps/i386/i686/multiarch/strcmp-sse4.S
+++ b/sysdeps/i386/i686/multiarch/strcmp-sse4.S
@@ -176,6 +176,7 @@ L(first4bytes):
 	PUSH	(%ebx)
 	PUSH	(%edi)
 	PUSH	(%esi)
+	cfi_remember_state
 	mov	%edx, %edi
 	mov	%eax, %esi
 	xorl	%eax, %eax
@@ -241,6 +242,7 @@ L(ret):
 #endif
 	ret
 
+	cfi_restore_state
 #ifdef USE_AS_STRNCMP
 L(more16byteseq):
 	POP	(%esi)
@@ -253,6 +255,10 @@ L(eq):
 	POP	(%ebp)
 #endif
 	ret
+
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(neq):
 	mov	$1, %eax
 	ja	L(neq_bigger)
@@ -263,6 +269,9 @@ L(neq_bigger):
 #endif
 	ret
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(less16bytes):
 	add	$0xfefefeff, %ecx
 	jnc	L(less4bytes)
@@ -370,8 +379,13 @@ L(more4bytes):
 	movzbl	7(%eax), %ecx
 	cmpb	%cl, 7(%edx)
 	jne	L(neq)
+#if 0
+	// XXX bug in original code.  It had a fallthru without any code
 	cmpl	$0, %ecx
 	je	L(eq)
+#else
+	jmp	L(eq)
+#endif
 
 END (STRCMP)
 
diff --git a/sysdeps/i386/i686/multiarch/strcmp-ssse3.S b/sysdeps/i386/i686/multiarch/strcmp-ssse3.S
index 14caae29a1..338b00339d 100644
--- a/sysdeps/i386/i686/multiarch/strcmp-ssse3.S
+++ b/sysdeps/i386/i686/multiarch/strcmp-ssse3.S
@@ -160,6 +160,9 @@ L(crosspage):
 	PUSH	(%ebx)
 	PUSH	(%edi)
 	PUSH	(%esi)
+#ifdef USE_AS_STRNCMP
+	cfi_remember_state
+#endif
 
 	movl	%edx, %edi
 	movl	%eax, %ecx
@@ -254,7 +257,7 @@ L(loop_ashr_0):
 
 /*
  * The following cases will be handled by ashr_1
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(15)            n -15            0(15 +(n-15) - n)         ashr_1
  */
 	.p2align 4
@@ -360,7 +363,7 @@ L(ashr_1_exittail):
 
 /*
  * The following cases will be handled by ashr_2
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(14~15)            n -14            1(15 +(n-14) - n)         ashr_2
  */
 	.p2align 4
@@ -467,7 +470,7 @@ L(ashr_2_exittail):
 
 /*
  * The following cases will be handled by ashr_3
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(13~15)            n -13            2(15 +(n-13) - n)         ashr_3
  */
 	.p2align 4
@@ -573,7 +576,7 @@ L(ashr_3_exittail):
 
 /*
  * The following cases will be handled by ashr_4
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(12~15)            n -12            3(15 +(n-12) - n)         ashr_4
  */
 	.p2align 4
@@ -682,7 +685,7 @@ L(ashr_4_exittail):
 
 /*
  * The following cases will be handled by ashr_5
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(11~15)            n -11            4(15 +(n-11) - n)         ashr_5
  */
 	.p2align 4
@@ -788,7 +791,7 @@ L(ashr_5_exittail):
 
 /*
  * The following cases will be handled by ashr_6
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(10~15)            n -10            5(15 +(n-10) - n)         ashr_6
  */
 
@@ -896,7 +899,7 @@ L(ashr_6_exittail):
 
 /*
  * The following cases will be handled by ashr_7
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(9~15)            n - 9            6(15 +(n-9) - n)         ashr_7
  */
 
@@ -1006,7 +1009,7 @@ L(ashr_7_exittail):
 
 /*
  * The following cases will be handled by ashr_8
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(8~15)            n - 8            7(15 +(n-8) - n)         ashr_8
  */
 	.p2align 4
@@ -1113,7 +1116,7 @@ L(ashr_8_exittail):
 
 /*
  * The following cases will be handled by ashr_9
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(7~15)            n - 7            8(15 +(n-7) - n)         ashr_9
  */
 	.p2align 4
@@ -1219,7 +1222,7 @@ L(ashr_9_exittail):
 
 /*
  * The following cases will be handled by ashr_10
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(6~15)            n - 6            9(15 +(n-6) - n)         ashr_10
  */
 	.p2align 4
@@ -1325,7 +1328,7 @@ L(ashr_10_exittail):
 
 /*
  * The following cases will be handled by ashr_11
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(5~15)            n - 5            10(15 +(n-5) - n)         ashr_11
  */
 	.p2align 4
@@ -1431,7 +1434,7 @@ L(ashr_11_exittail):
 
 /*
  * The following cases will be handled by ashr_12
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(4~15)            n - 4            11(15 +(n-4) - n)         ashr_12
  */
 	.p2align 4
@@ -1537,7 +1540,7 @@ L(ashr_12_exittail):
 
 /*
  * The following cases will be handled by ashr_13
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(3~15)            n - 3            12(15 +(n-3) - n)         ashr_13
  */
 	.p2align 4
@@ -1643,7 +1646,7 @@ L(ashr_13_exittail):
 
 /*
  * The following cases will be handled by ashr_14
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(2~15)            n - 2            13(15 +(n-2) - n)         ashr_14
  */
 	.p2align 4
@@ -1749,7 +1752,7 @@ L(ashr_14_exittail):
 
 /*
  * The following cases will be handled by ashr_14
- * ecx(offset of esi)  eax(offset of edi)   relative offset   	corresponding case
+ * ecx(offset of esi)  eax(offset of edi)   relative offset	corresponding case
  *        n(1~15)            n - 1            14(15 +(n-1) - n)         ashr_15
  */
 
@@ -1916,6 +1919,9 @@ L(less16bytes):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte0):
 #ifdef USE_AS_STRNCMP
 	cmp	$0, %ebp
@@ -1931,6 +1937,9 @@ L(Byte0):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte1):
 #ifdef USE_AS_STRNCMP
 	cmp	$1, %ebp
@@ -1946,6 +1955,9 @@ L(Byte1):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte2):
 #ifdef USE_AS_STRNCMP
 	cmp	$2, %ebp
@@ -1961,6 +1973,9 @@ L(Byte2):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte3):
 #ifdef USE_AS_STRNCMP
 	cmp	$3, %ebp
@@ -1976,6 +1991,9 @@ L(Byte3):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte4):
 #ifdef USE_AS_STRNCMP
 	cmp	$4, %ebp
@@ -1989,7 +2007,11 @@ L(Byte4):
 	POP	(%ebp)
 #endif
 	ret
+
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte5):
 #ifdef USE_AS_STRNCMP
 	cmp	$5, %ebp
@@ -2005,6 +2027,9 @@ L(Byte5):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(Byte6):
 #ifdef USE_AS_STRNCMP
 	cmp	$6, %ebp
@@ -2020,6 +2045,9 @@ L(Byte6):
 	ret
 
 	.p2align 4
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(2next_8_bytes):
 	add	$8, %eax
 	add	$8, %edx
@@ -2063,6 +2091,9 @@ L(2next_8_bytes):
 #endif
 	ret
 
+#ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
+#endif
 L(neq):
 	mov	$1, %eax
 	ja	L(neq_bigger)
@@ -2074,6 +2105,7 @@ L(neq_bigger):
 	ret
 
 #ifdef USE_AS_STRNCMP
+	cfi_remember_state
 L(more8byteseq):
 	POP	(%esi)
 	POP	(%edi)
@@ -2087,7 +2119,9 @@ L(eq):
 #endif
 	xorl	%eax, %eax
 	ret
+
 #ifdef USE_AS_STRNCMP
+	CFI_PUSH (%ebp)
 L(less16bytes_sncmp):
 	test	%ebp, %ebp
 	jz	L(eq)