x86,kcfi: Fix EXPORT_SYMBOL vs kCFI

The expectation is that all EXPORT'ed symbols are free to have their
address taken and called indirectly. The majority of the assembly
defined functions currently violate this expectation.

Make then all use SYM_TYPED_FUNC_START() in order to emit the proper
kCFI preamble.

Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Reviewed-by: Sami Tolvanen <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Author: Peter Zijlstra

arch/x86/crypto/camellia-aesni-avx-asm_64.S

@@ -16,6 +16,7 @@
  */
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/frame.h>
 
 #define CAMELLIA_TABLE_BYTE_LEN 272
@@ -882,7 +883,7 @@ SYM_FUNC_START_LOCAL(__camellia_dec_blk16)
 	jmp .Ldec_max24;
 SYM_FUNC_END(__camellia_dec_blk16)
 
-SYM_FUNC_START(camellia_ecb_enc_16way)
+SYM_TYPED_FUNC_START(camellia_ecb_enc_16way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst (16 blocks)
@@ -907,7 +908,7 @@ SYM_FUNC_START(camellia_ecb_enc_16way)
 	RET;
 SYM_FUNC_END(camellia_ecb_enc_16way)
 
-SYM_FUNC_START(camellia_ecb_dec_16way)
+SYM_TYPED_FUNC_START(camellia_ecb_dec_16way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst (16 blocks)
@@ -937,7 +938,7 @@ SYM_FUNC_START(camellia_ecb_dec_16way)
 	RET;
 SYM_FUNC_END(camellia_ecb_dec_16way)
 
-SYM_FUNC_START(camellia_cbc_dec_16way)
+SYM_TYPED_FUNC_START(camellia_cbc_dec_16way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst (16 blocks)

arch/x86/crypto/camellia-aesni-avx2-asm_64.S

@@ -6,6 +6,7 @@
  */
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/frame.h>
 
 #define CAMELLIA_TABLE_BYTE_LEN 272

arch/x86/crypto/camellia-x86_64-asm_64.S

@@ -6,6 +6,7 @@
  */
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 
 .file "camellia-x86_64-asm_64.S"
 .text
@@ -177,7 +178,7 @@
 	bswapq				RAB0; \
 	movq RAB0,			4*2(RIO);
 
-SYM_FUNC_START(__camellia_enc_blk)
+SYM_TYPED_FUNC_START(__camellia_enc_blk)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -224,7 +225,7 @@ SYM_FUNC_START(__camellia_enc_blk)
 	RET;
 SYM_FUNC_END(__camellia_enc_blk)
 
-SYM_FUNC_START(camellia_dec_blk)
+SYM_TYPED_FUNC_START(camellia_dec_blk)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -411,7 +412,7 @@ SYM_FUNC_END(camellia_dec_blk)
 		bswapq				RAB1; \
 		movq RAB1,			12*2(RIO);
 
-SYM_FUNC_START(__camellia_enc_blk_2way)
+SYM_TYPED_FUNC_START(__camellia_enc_blk_2way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -460,7 +461,7 @@ SYM_FUNC_START(__camellia_enc_blk_2way)
 	RET;
 SYM_FUNC_END(__camellia_enc_blk_2way)
 
-SYM_FUNC_START(camellia_dec_blk_2way)
+SYM_TYPED_FUNC_START(camellia_dec_blk_2way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst

arch/x86/crypto/serpent-avx-x86_64-asm_64.S

@@ -9,6 +9,7 @@
  */
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/frame.h>
 #include "glue_helper-asm-avx.S"
 
@@ -656,7 +657,7 @@ SYM_FUNC_START_LOCAL(__serpent_dec_blk8_avx)
 	RET;
 SYM_FUNC_END(__serpent_dec_blk8_avx)
 
-SYM_FUNC_START(serpent_ecb_enc_8way_avx)
+SYM_TYPED_FUNC_START(serpent_ecb_enc_8way_avx)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -674,7 +675,7 @@ SYM_FUNC_START(serpent_ecb_enc_8way_avx)
 	RET;
 SYM_FUNC_END(serpent_ecb_enc_8way_avx)
 
-SYM_FUNC_START(serpent_ecb_dec_8way_avx)
+SYM_TYPED_FUNC_START(serpent_ecb_dec_8way_avx)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -692,7 +693,7 @@ SYM_FUNC_START(serpent_ecb_dec_8way_avx)
 	RET;
 SYM_FUNC_END(serpent_ecb_dec_8way_avx)
 
-SYM_FUNC_START(serpent_cbc_dec_8way_avx)
+SYM_TYPED_FUNC_START(serpent_cbc_dec_8way_avx)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst

arch/x86/crypto/twofish-x86_64-asm_64-3way.S

@@ -6,6 +6,7 @@
  */
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 
 .file "twofish-x86_64-asm-3way.S"
 .text
@@ -220,7 +221,7 @@
 	rorq $32,			RAB2; \
 	outunpack3(mov, RIO, 2, RAB, 2);
 
-SYM_FUNC_START(__twofish_enc_blk_3way)
+SYM_TYPED_FUNC_START(__twofish_enc_blk_3way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst
@@ -269,7 +270,7 @@ SYM_FUNC_START(__twofish_enc_blk_3way)
 	RET;
 SYM_FUNC_END(__twofish_enc_blk_3way)
 
-SYM_FUNC_START(twofish_dec_blk_3way)
+SYM_TYPED_FUNC_START(twofish_dec_blk_3way)
 	/* input:
 	 *	%rdi: ctx, CTX
 	 *	%rsi: dst

arch/x86/crypto/twofish-x86_64-asm_64.S

@@ -8,6 +8,7 @@
 .text
 
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/asm-offsets.h>
 
 #define a_offset	0
@@ -202,7 +203,7 @@
 	xor	%r8d,		d ## D;\
 	ror	$1,		d ## D;
 
-SYM_FUNC_START(twofish_enc_blk)
+SYM_TYPED_FUNC_START(twofish_enc_blk)
 	pushq    R1
 
 	/* %rdi contains the ctx address */
@@ -255,7 +256,7 @@ SYM_FUNC_START(twofish_enc_blk)
 	RET
 SYM_FUNC_END(twofish_enc_blk)
 
-SYM_FUNC_START(twofish_dec_blk)
+SYM_TYPED_FUNC_START(twofish_dec_blk)
 	pushq    R1
 
 	/* %rdi contains the ctx address */

arch/x86/lib/clear_page_64.S

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 #include <linux/export.h>
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/asm.h>
 
 /*
@@ -14,15 +15,15 @@
  * Zero a page.
  * %rdi	- page
  */
-SYM_FUNC_START(clear_page_rep)
+SYM_TYPED_FUNC_START(clear_page_rep)
 	movl $4096/8,%ecx
 	xorl %eax,%eax
 	rep stosq
 	RET
 SYM_FUNC_END(clear_page_rep)
 EXPORT_SYMBOL_GPL(clear_page_rep)
 
-SYM_FUNC_START(clear_page_orig)
+SYM_TYPED_FUNC_START(clear_page_orig)
 	xorl   %eax,%eax
 	movl   $4096/64,%ecx
 	.p2align 4
@@ -44,7 +45,7 @@ SYM_FUNC_START(clear_page_orig)
 SYM_FUNC_END(clear_page_orig)
 EXPORT_SYMBOL_GPL(clear_page_orig)
 
-SYM_FUNC_START(clear_page_erms)
+SYM_TYPED_FUNC_START(clear_page_erms)
 	movl $4096,%ecx
 	xorl %eax,%eax
 	rep stosb

arch/x86/lib/copy_page_64.S

@@ -3,6 +3,7 @@
 
 #include <linux/export.h>
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/cpufeatures.h>
 #include <asm/alternative.h>
 
@@ -13,7 +14,7 @@
  * prefetch distance based on SMP/UP.
  */
 	ALIGN
-SYM_FUNC_START(copy_page)
+SYM_TYPED_FUNC_START(copy_page)
 	ALTERNATIVE "jmp copy_page_regs", "", X86_FEATURE_REP_GOOD
 	movl	$4096/8, %ecx
 	rep	movsq

arch/x86/lib/memmove_64.S

@@ -8,6 +8,7 @@
  */
 #include <linux/export.h>
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/cpufeatures.h>
 #include <asm/alternative.h>
 
@@ -26,7 +27,7 @@
  * Output:
  * rax: dest
  */
-SYM_FUNC_START(__memmove)
+SYM_TYPED_FUNC_START(__memmove)
 
 	mov %rdi, %rax
 

arch/x86/lib/memset_64.S

@@ -3,6 +3,7 @@
 
 #include <linux/export.h>
 #include <linux/linkage.h>
+#include <linux/cfi_types.h>
 #include <asm/cpufeatures.h>
 #include <asm/alternative.h>
 
@@ -28,7 +29,7 @@
  * only for the return value that is the same as the source input,
  * which the compiler could/should do much better anyway.
  */
-SYM_FUNC_START(__memset)
+SYM_TYPED_FUNC_START(__memset)
 	ALTERNATIVE "jmp memset_orig", "", X86_FEATURE_FSRS
 
 	movq %rdi,%r9

arch/x86/lib/msr-reg.S

@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #include <linux/linkage.h>
 #include <linux/errno.h>
+#include <linux/cfi_types.h>
 #include <asm/asm.h>
 #include <asm/msr.h>
 
@@ -12,7 +13,7 @@
  *
  */
 .macro op_safe_regs op
-SYM_FUNC_START(\op\()_safe_regs)
+SYM_TYPED_FUNC_START(\op\()_safe_regs)
 	pushq %rbx
 	pushq %r12
 	movq	%rdi, %r10	/* Save pointer */