Upstream: SHA2 reworking and API for iterating over multiple implemen…

…tations

The changes in the shared files to enable macOS support to PR

lib/libspl/include/sys/simd.h

@@ -49,6 +49,27 @@ static inline unsigned long getauxval(unsigned long key)
 #define	AT_HWCAP	16
 #define	AT_HWCAP2	26
 extern unsigned long getauxval(unsigned long type);
+#elif defined(__APPLE__)
+#include <sys/sysctl.h>
+#define	AT_HWCAP	0
+static inline unsigned long getauxval(unsigned long key)
+{
+	(void) key;
+	/* HWCAP_ are all defined halfway down this file */
+	unsigned long val = 1 /* HWCAP_FP */;
+	int intval;
+	size_t intvallen = sizeof (intval);
+	int err;
+	err = sysctlbyname("hw.optional.arm.FEAT_SHA256",
+	    &intval, &intvallen, NULL, 0);
+	if (err == 0 && intval != 0)
+		val |= 0x00000040; /* SHA256 */
+	err = sysctlbyname("hw.optional.arm.FEAT_SHA512",
+	    &intval, &intvallen, NULL, 0);
+	if (err == 0 && intval != 0)
+		val |= 0x00200000; /* SHA512 */
+	return (val);
+}
 #endif /* __linux__ */
 #endif /* arm || aarch64 || powerpc */
 

module/icp/algs/sha2/sha256_impl.c

@@ -240,7 +240,7 @@ sha256_param_set(const char *val, zfs_kernel_param_t *unused)
 	return (generic_impl_setname(val));
 }
 
-#elif defined(__FreeBSD__)
+#elif defined(__FreeBSD__) || defined(__APPLE__)
 
 #include <sys/sbuf.h>
 
@@ -272,7 +272,12 @@ sha256_param(ZFS_MODULE_PARAM_ARGS)
 			(void) sbuf_printf(s, fmt, generic_supp_impls[i]->name);
 		}
 
+#ifdef __APPLE__
+		err = SYSCTL_OUT(req, s->s_buf, s->s_len);
+		sbuf_finish(s);
+#else
 		err = sbuf_finish(s);
+#endif
 		sbuf_delete(s);
 
 		return (err);

module/icp/algs/sha2/sha512_impl.c

@@ -216,7 +216,7 @@ sha512_param_set(const char *val, zfs_kernel_param_t *unused)
 	return (generic_impl_setname(val));
 }
 
-#elif defined(__FreeBSD__)
+#elif defined(__FreeBSD__) || defined(__APPLE__)
 
 #include <sys/sbuf.h>
 
@@ -248,7 +248,12 @@ sha512_param(ZFS_MODULE_PARAM_ARGS)
 			(void) sbuf_printf(s, fmt, generic_supp_impls[i]->name);
 		}
 
+#ifdef __APPLE__
+		err = SYSCTL_OUT(req, s->s_buf, s->s_len);
+		sbuf_finish(s);
+#else
 		err = sbuf_finish(s);
+#endif
 		sbuf_delete(s);
 
 		return (err);

module/icp/asm-aarch64/sha2/sha256-armv8.S

@@ -19,12 +19,14 @@
  * - modified assembly to fit into OpenZFS
  */
 
+#include <sys/asm_linkage.h>
+
 #if defined(__aarch64__)
 
-.text
+SECTION_TEXT
 
-.align	6
-.type	.LK256,%object
+.balign	64
+SET_OBJ(.LK256)
 .LK256:
 	.long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
 	.long	0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
@@ -43,12 +45,9 @@
 	.long	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
 	.long	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
 	.long	0	//terminator
-.size	.LK256,.-.LK256
+SET_SIZE(.LK256)
 
-.globl	zfs_sha256_block_armv7
-.type	zfs_sha256_block_armv7,%function
-.align	6
-zfs_sha256_block_armv7:
+ENTRY_ALIGN(zfs_sha256_block_armv7, 64)
 	stp	x29,x30,[sp,#-128]!
 	add	x29,sp,#0
 
@@ -1009,12 +1008,9 @@ zfs_sha256_block_armv7:
 	ldp	x27,x28,[x29,#80]
 	ldp	x29,x30,[sp],#128
 	ret
-.size	zfs_sha256_block_armv7,.-zfs_sha256_block_armv7
+SET_SIZE(zfs_sha256_block_armv7)
 
-.globl	zfs_sha256_block_armv8
-.type	zfs_sha256_block_armv8,%function
-.align	6
-zfs_sha256_block_armv8:
+ENTRY_ALIGN(zfs_sha256_block_armv8, 64)
 .Lv8_entry:
 	stp		x29,x30,[sp,#-16]!
 	add		x29,sp,#0
@@ -1149,12 +1145,9 @@ zfs_sha256_block_armv8:
 
 	ldr		x29,[sp],#16
 	ret
-.size	zfs_sha256_block_armv8,.-zfs_sha256_block_armv8
+SET_SIZE(zfs_sha256_block_armv8)
 
-.globl	zfs_sha256_block_neon
-.type	zfs_sha256_block_neon,%function
-.align	4
-zfs_sha256_block_neon:
+ENTRY_ALIGN(zfs_sha256_block_neon, 16)
 .Lneon_entry:
 	stp	x29, x30, [sp, #-16]!
 	mov	x29, sp
@@ -1994,6 +1987,6 @@ zfs_sha256_block_neon:
 	ldr	x29,[x29]
 	add	sp,sp,#16*4+16
 	ret
-.size	zfs_sha256_block_neon,.-zfs_sha256_block_neon
+SET_SIZE(zfs_sha256_block_neon)
 
 #endif

module/icp/asm-aarch64/sha2/sha512-armv8.S

@@ -18,13 +18,14 @@
  * Portions Copyright (c) 2022 Tino Reichardt <milky-zfs@mcmilk.de>
  * - modified assembly to fit into OpenZFS
  */
+#include <sys/asm_linkage.h>
 
 #if defined(__aarch64__)
 
-.text
+SECTION_TEXT
 
-.align	6
-.type	.LK512,%object
+.balign	64
+SET_OBJ(.LK512)
 .LK512:
 	.quad	0x428a2f98d728ae22,0x7137449123ef65cd
 	.quad	0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
@@ -67,12 +68,9 @@
 	.quad	0x4cc5d4becb3e42b6,0x597f299cfc657e2a
 	.quad	0x5fcb6fab3ad6faec,0x6c44198c4a475817
 	.quad	0	// terminator
-.size	.LK512,.-.LK512
+SET_SIZE(.LK512)
 
-.globl	zfs_sha512_block_armv7
-.type	zfs_sha512_block_armv7,%function
-.align	6
-zfs_sha512_block_armv7:
+ENTRY_ALIGN(zfs_sha512_block_armv7, 64)
 	stp	x29,x30,[sp,#-128]!
 	add	x29,sp,#0
 
@@ -1033,13 +1031,10 @@ zfs_sha512_block_armv7:
 	ldp	x27,x28,[x29,#80]
 	ldp	x29,x30,[sp],#128
 	ret
-.size	zfs_sha512_block_armv7,.-zfs_sha512_block_armv7
+SET_SIZE(zfs_sha512_block_armv7)
 
 
-.globl	zfs_sha512_block_armv8
-.type	zfs_sha512_block_armv8,%function
-.align	6
-zfs_sha512_block_armv8:
+ENTRY_ALIGN(zfs_sha512_block_armv8, 64)
 .Lv8_entry:
 	// Armv8.3-A PAuth: even though x30 is pushed to stack it is not popped later
 	stp		x29,x30,[sp,#-16]!
@@ -1061,7 +1056,7 @@ zfs_sha512_block_armv8:
 	rev64		v23.16b,v23.16b
 	b		.Loop_hw
 
-.align	4
+.balign	16
 .Loop_hw:
 	ld1		{v24.2d},[x3],#16
 	subs		x2,x2,#1
@@ -1554,5 +1549,5 @@ zfs_sha512_block_armv8:
 
 	ldr		x29,[sp],#16
 	ret
-.size	zfs_sha512_block_armv8,.-zfs_sha512_block_armv8
+SET_SIZE(zfs_sha512_block_armv8)
 #endif

module/icp/asm-x86_64/sha2/sha256-x86_64.S

@@ -26,8 +26,8 @@
 
 SECTION_STATIC
 
-.align	64
-.type	K256,@object
+.balign	64
+SET_OBJ(K256)
 K256:
 .long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
 .long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
@@ -105,7 +105,7 @@ ENTRY_ALIGN(zfs_sha256_transform_x64, 16)
 	movl	24(%rdi),%r10d
 	movl	28(%rdi),%r11d
 	jmp	.Lloop
-.align	16
+.balign	16
 .Lloop:
 	movl	%ebx,%edi
 	leaq	K256(%rip),%rbp
@@ -622,7 +622,7 @@ ENTRY_ALIGN(zfs_sha256_transform_x64, 16)
 	addl	%r12d,%eax
 	leaq	20(%rbp),%rbp
 	jmp	.Lrounds_16_xx
-.align	16
+.balign	16
 .Lrounds_16_xx:
 	movl	4(%rsp),%r13d
 	movl	56(%rsp),%r15d
@@ -1436,7 +1436,7 @@ ENTRY_ALIGN(zfs_sha256_transform_shani, 64)
 	punpcklqdq	%xmm0,%xmm2
 	jmp	.Loop_shani
 
-.align	16
+.balign	16
 .Loop_shani:
 	movdqu	(%rsi),%xmm3
 	movdqu	16(%rsi),%xmm4
@@ -1666,7 +1666,7 @@ ENTRY_ALIGN(zfs_sha256_transform_ssse3, 64)
 	movl	28(%rdi),%r11d
 
 	jmp	.Lloop_ssse3
-.align	16
+.balign	16
 .Lloop_ssse3:
 	movdqa	K256+512(%rip),%xmm7
 	movdqu	0(%rsi),%xmm0
@@ -1696,7 +1696,7 @@ ENTRY_ALIGN(zfs_sha256_transform_ssse3, 64)
 	movl	%r8d,%r13d
 	jmp	.Lssse3_00_47
 
-.align	16
+.balign	16
 .Lssse3_00_47:
 	subq	$-128,%rbp
 	rorl	$14,%r13d
@@ -2779,7 +2779,7 @@ ENTRY_ALIGN(zfs_sha256_transform_avx, 64)
 	vmovdqa	K256+512+32(%rip),%xmm8
 	vmovdqa	K256+512+64(%rip),%xmm9
 	jmp	.Lloop_avx
-.align	16
+.balign	16
 .Lloop_avx:
 	vmovdqa	K256+512(%rip),%xmm7
 	vmovdqu	0(%rsi),%xmm0
@@ -2805,7 +2805,7 @@ ENTRY_ALIGN(zfs_sha256_transform_avx, 64)
 	movl	%r8d,%r13d
 	jmp	.Lavx_00_47
 
-.align	16
+.balign	16
 .Lavx_00_47:
 	subq	$-128,%rbp
 	vpalignr	$4,%xmm0,%xmm1,%xmm4
@@ -3858,7 +3858,7 @@ ENTRY_ALIGN(zfs_sha256_transform_avx2, 64)
 	vmovdqa	K256+512+32(%rip),%ymm8
 	vmovdqa	K256+512+64(%rip),%ymm9
 	jmp	.Loop_avx2
-.align	16
+.balign	16
 .Loop_avx2:
 	vmovdqa	K256+512(%rip),%ymm7
 	vmovdqu	-64+0(%rsi),%xmm0
@@ -3900,7 +3900,7 @@ ENTRY_ALIGN(zfs_sha256_transform_avx2, 64)
 	subq	$-32*4,%rbp
 	jmp	.Lavx2_00_47
 
-.align	16
+.balign	16
 .Lavx2_00_47:
 	leaq	-64(%rsp),%rsp
 .cfi_escape	0x0f,0x05,0x77,0x38,0x06,0x23,0x08
@@ -4842,7 +4842,7 @@ ENTRY_ALIGN(zfs_sha256_transform_avx2, 64)
 	xorl	%ecx,%edi
 	movl	%r9d,%r12d
 	jmp	.Lower_avx2
-.align	16
+.balign	16
 .Lower_avx2:
 	addl	0+16(%rbp),%r11d
 	andl	%r8d,%r12d

module/icp/asm-x86_64/sha2/sha512-x86_64.S

@@ -26,8 +26,8 @@
 
 SECTION_STATIC
 
-.align	64
-.type	K512,@object
+.balign	64
+SET_OBJ(K512)
 K512:
 .quad	0x428a2f98d728ae22,0x7137449123ef65cd
 .quad	0x428a2f98d728ae22,0x7137449123ef65cd
@@ -148,7 +148,7 @@ ENTRY_ALIGN(zfs_sha512_transform_x64, 16)
 	movq	48(%rdi),%r10
 	movq	56(%rdi),%r11
 	jmp	.Lloop
-.align	16
+.balign	16
 .Lloop:
 	movq	%rbx,%rdi
 	leaq	K512(%rip),%rbp
@@ -665,7 +665,7 @@ ENTRY_ALIGN(zfs_sha512_transform_x64, 16)
 	addq	%r12,%rax
 	leaq	24(%rbp),%rbp
 	jmp	.Lrounds_16_xx
-.align	16
+.balign	16
 .Lrounds_16_xx:
 	movq	8(%rsp),%r13
 	movq	112(%rsp),%r15
@@ -1501,7 +1501,7 @@ ENTRY_ALIGN(zfs_sha512_transform_avx, 64)
 	movq	48(%rdi),%r10
 	movq	56(%rdi),%r11
 	jmp	.Lloop_avx
-.align	16
+.balign	16
 .Lloop_avx:
 	vmovdqa	K512+1280(%rip),%xmm11
 	vmovdqu	0(%rsi),%xmm0
@@ -1543,7 +1543,7 @@ ENTRY_ALIGN(zfs_sha512_transform_avx, 64)
 	movq	%r8,%r13
 	jmp	.Lavx_00_47
 
-.align	16
+.balign	16
 .Lavx_00_47:
 	addq	$256,%rbp
 	vpalignr	$8,%xmm0,%xmm1,%xmm8
@@ -2670,7 +2670,7 @@ ENTRY_ALIGN(zfs_sha512_transform_avx2, 64)
 	movq	48(%rdi),%r10
 	movq	56(%rdi),%r11
 	jmp	.Loop_avx2
-.align	16
+.balign	16
 .Loop_avx2:
 	vmovdqu	-128(%rsi),%xmm0
 	vmovdqu	-128+16(%rsi),%xmm1
@@ -2732,7 +2732,7 @@ ENTRY_ALIGN(zfs_sha512_transform_avx2, 64)
 	addq	$32*8,%rbp
 	jmp	.Lavx2_00_47
 
-.align	16
+.balign	16
 .Lavx2_00_47:
 	leaq	-128(%rsp),%rsp
 .cfi_escape	0x0f,0x06,0x77,0xf8,0x00,0x06,0x23,0x08
@@ -3750,7 +3750,7 @@ ENTRY_ALIGN(zfs_sha512_transform_avx2, 64)
 	xorq	%rcx,%rdi
 	movq	%r9,%r12
 	jmp	.Lower_avx2
-.align	16
+.balign	16
 .Lower_avx2:
 	addq	0+16(%rbp),%r11
 	andq	%r8,%r12