Upstream: SHA2 reworking and API for iterating

over multiple implementations The changes in the shared files to enable macOS support to PR Signed-off-by: Jorgen Lundman <[email protected]>
openzfs · Oct 19, 2023 · 802749c · 802749c
1 parent 8812f26
commit 802749c
Show file tree

Hide file tree

Showing 11 changed files with 89 additions and 26 deletions.
diff --git a/include/os/macos/spl/sys/aarch64/asm_linkage.h b/include/os/macos/spl/sys/aarch64/asm_linkage.h
@@ -51,14 +51,16 @@
     .balign ASM_ENTRY_ALIGN %% \
     .globl _##x %% \
 _##x: %% \
-x:
+x: \
+	bti c // hint	#34
 
 #define	ENTRY_ALIGN(x, a) \
 	.text %% \
 	.balign	a %% \
 	.globl _##x %% \
 _##x: %% \
-x:
+x: \
+	bti c // hint	#34
 
 #define	FUNCTION(x) \
 x:

diff --git a/lib/libspl/include/sys/simd.h b/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 */
 

diff --git a/module/icp/algs/aes/aes_impl.c b/module/icp/algs/aes/aes_impl.c
@@ -418,14 +418,16 @@ aes_impl_set(const char *val)
 }
 
 #if defined(_KERNEL)
-#if defined(__linux__) || defined(__APPLE__)
 
+#if defined(__linux__)
 static int
 icp_aes_impl_set(const char *val, zfs_kernel_param_t *kp)
 {
 	return (aes_impl_set(val));
 }
+#endif
 
+#if defined(__linux__) || defined(__APPLE__)
 static int
 icp_aes_impl_get(char *buffer, zfs_kernel_param_t *kp)
 {

diff --git a/module/icp/algs/blake3/blake3_impl.c b/module/icp/algs/blake3/blake3_impl.c
@@ -364,7 +364,7 @@ blake3_param(ZFS_MODULE_PARAM_ARGS)
 	int err;
 
 	generic_impl_init();
-	if (req->newptr == NULL) {
+	if ((void *)req->newptr == NULL) {
 		const uint32_t impl = IMPL_READ(generic_impl_chosen);
 		const int init_buflen = 64;
 		const char *fmt;

diff --git a/module/icp/algs/modes/gcm.c b/module/icp/algs/modes/gcm.c
@@ -889,7 +889,13 @@ gcm_impl_init(void)
 	if (gcm_avx_will_work()) {
 #ifdef HAVE_MOVBE
 		if (zfs_movbe_available() == B_TRUE) {
+#ifdef __APPLE__
+			atomic_swap_32(
+			    (volatile unsigned int *)&gcm_avx_can_use_movbe,
+			    B_TRUE);
+#else
 			atomic_swap_32(&gcm_avx_can_use_movbe, B_TRUE);
+#endif
 		}
 #endif
 		if (GCM_IMPL_READ(user_sel_impl) == IMPL_FASTEST) {
@@ -992,13 +998,16 @@ gcm_impl_set(const char *val)
 }
 
 #if defined(_KERNEL)
-#if defined(__linux__) || defined(__APPLE__)
+
+#if defined(__linux__)
 static int
 icp_gcm_impl_set(const char *val, zfs_kernel_param_t *kp)
 {
 	return (gcm_impl_set(val));
 }
+#endif
 
+#if defined(__linux__) || defined(__APPLE__)
 static int
 icp_gcm_impl_get(char *buffer, zfs_kernel_param_t *kp)
 {
@@ -1120,7 +1129,11 @@ static inline void
 gcm_set_avx(boolean_t val)
 {
 	if (gcm_avx_will_work() == B_TRUE) {
+#ifdef __APPLE__
+		atomic_swap_32((volatile unsigned int *)&gcm_use_avx, val);
+#else
 		atomic_swap_32(&gcm_use_avx, val);
+#endif
 	}
 }
 
@@ -1571,6 +1584,8 @@ gcm_init_avx(gcm_ctx_t *ctx, const uint8_t *iv, size_t iv_len,
 }
 
 #if defined(_KERNEL)
+
+#if defined(__linux__)
 static int
 icp_gcm_avx_set_chunk_size(const char *buf, zfs_kernel_param_t *kp)
 {
@@ -1591,9 +1606,9 @@ icp_gcm_avx_set_chunk_size(const char *buf, zfs_kernel_param_t *kp)
 	error = param_set_uint(val_rounded, kp);
 	return (error);
 }
+#endif
 
 #ifdef __APPLE__
-
 /* Lives in here to have access to GCM macros */
 int
 param_icp_gcm_avx_set_chunk_size(ZFS_MODULE_PARAM_ARGS)
@@ -1622,10 +1637,8 @@ param_icp_gcm_avx_set_chunk_size(ZFS_MODULE_PARAM_ARGS)
 	gcm_avx_chunk_size = val;
 	return (rc);
 }
-
 #endif
 
-
 module_param_call(icp_gcm_avx_chunk_size, icp_gcm_avx_set_chunk_size,
     param_get_uint, &gcm_avx_chunk_size, 0644);
 

diff --git a/module/icp/algs/sha2/sha256_impl.c b/module/icp/algs/sha2/sha256_impl.c
@@ -260,7 +260,7 @@ sha256_param(ZFS_MODULE_PARAM_ARGS)
 	int err;
 
 	generic_impl_init();
-	if (req->newptr == NULL) {
+	if ((void *)req->newptr == NULL) {
 		const uint32_t impl = IMPL_READ(generic_impl_chosen);
 		const int init_buflen = 64;
 		const char *fmt;

diff --git a/module/icp/algs/sha2/sha512_impl.c b/module/icp/algs/sha2/sha512_impl.c
@@ -235,7 +235,7 @@ sha512_param(ZFS_MODULE_PARAM_ARGS)
 	int err;
 
 	generic_impl_init();
-	if (req->newptr == NULL) {
+	if ((void *)req->newptr == NULL) {
 		const uint32_t impl = IMPL_READ(generic_impl_chosen);
 		const int init_buflen = 64;
 		const char *fmt;

diff --git a/module/icp/asm-aarch64/sha2/sha256-armv8.S b/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
@@ -43,13 +45,17 @@
 	.long	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
 	.long	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
 	.long	0	//terminator
-.size	.LK256,.-.LK256
+SET_SIZE(.LK256)
 
+#ifdef __APPLE__
+ENTRY_ALIGN(zfs_sha256_block_armv7, 64)
+#else
 .globl	zfs_sha256_block_armv7
 .type	zfs_sha256_block_armv7,%function
 .align	6
 zfs_sha256_block_armv7:
 	hint	#34					// bti c
+#endif
 	stp	x29,x30,[sp,#-128]!
 	add	x29,sp,#0
 
@@ -1010,13 +1016,17 @@ 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)
 
+#ifdef __APPLE__
+ENTRY_ALIGN(zfs_sha256_block_armv8, 64)
+#else
 .globl	zfs_sha256_block_armv8
 .type	zfs_sha256_block_armv8,%function
 .align	6
 zfs_sha256_block_armv8:
 	hint		#34				// bti c
+#endif
 .Lv8_entry:
 	stp		x29,x30,[sp,#-16]!
 	add		x29,sp,#0
@@ -1151,13 +1161,17 @@ zfs_sha256_block_armv8:
 
 	ldr		x29,[sp],#16
 	ret
-.size	zfs_sha256_block_armv8,.-zfs_sha256_block_armv8
+SET_SIZE(zfs_sha256_block_armv8)
 
+#ifdef __APPLE__
+ENTRY_ALIGN(zfs_sha256_block_neon, 16)
+#else
 .globl	zfs_sha256_block_neon
 .type	zfs_sha256_block_neon,%function
 .align	4
 zfs_sha256_block_neon:
 	hint	#34					// bti c
+#endif
 .Lneon_entry:
 	stp	x29, x30, [sp, #-16]!
 	mov	x29, sp
@@ -1997,6 +2011,6 @@ 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
diff --git a/module/icp/asm-aarch64/sha2/sha512-armv8.S b/module/icp/asm-aarch64/sha2/sha512-armv8.S
@@ -18,13 +18,14 @@
  * Portions Copyright (c) 2022 Tino Reichardt <[email protected]>
  * - 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
@@ -67,13 +68,17 @@
 	.quad	0x4cc5d4becb3e42b6,0x597f299cfc657e2a
 	.quad	0x5fcb6fab3ad6faec,0x6c44198c4a475817
 	.quad	0	// terminator
-.size	.LK512,.-.LK512
+SET_SIZE(.LK512)
 
+#ifdef __APPLE__
+ENTRY_ALIGN(zfs_sha512_block_armv7, 64)
+#else
 .globl	zfs_sha512_block_armv7
 .type	zfs_sha512_block_armv7,%function
 .align	6
 zfs_sha512_block_armv7:
 	hint	#34					// bti c
+#endif
 	stp	x29,x30,[sp,#-128]!
 	add	x29,sp,#0
 
@@ -1034,14 +1039,18 @@ 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)
 
 
+#ifdef __APPLE__
+ENTRY_ALIGN(zfs_sha512_block_armv8, 64)
+#else
 .globl	zfs_sha512_block_armv8
 .type	zfs_sha512_block_armv8,%function
 .align	6
 zfs_sha512_block_armv8:
 	hint		#34				// bti c
+#endif
 .Lv8_entry:
 	// Armv8.3-A PAuth: even though x30 is pushed to stack it is not popped later
 	stp		x29,x30,[sp,#-16]!
@@ -1063,7 +1072,7 @@ 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
@@ -1556,5 +1565,5 @@ 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
diff --git a/module/zcommon/zfs_fletcher.c b/module/zcommon/zfs_fletcher.c
@@ -949,7 +949,7 @@ fletcher_4_param(ZFS_MODULE_PARAM_ARGS)
 {
 	int err = 0;
 
-	if (req->newptr == NULL) {
+	if ((void *)req->newptr == NULL) {
 		const uint32_t impl = IMPL_READ(fletcher_4_impl_chosen);
 		const int init_buflen = 64;
 		const char *fmt;

diff --git a/module/zfs/vdev_raidz_math.c b/module/zfs/vdev_raidz_math.c
@@ -634,14 +634,16 @@ vdev_raidz_impl_set(const char *val)
 }
 
 #if defined(_KERNEL)
-#if defined(__linux__) || defined(__APPLE__)
 
+#if defined(__linux__)
 static int
 zfs_vdev_raidz_impl_set(const char *val, zfs_kernel_param_t *kp)
 {
 	return (vdev_raidz_impl_set(val));
 }
+#endif
 
+#if defined(__linux__) || defined(__APPLE__)
 static int
 zfs_vdev_raidz_impl_get(char *buffer, zfs_kernel_param_t *kp)
 {