Fix blake3 on macOS/arm64

    BLAKE3_CTX *ctx = blake3_per_cpu_ctx[CPU_SEQID_UNSTABLE];

We have macOS arm64 to call kmem_alloc() as the cpu_number()
changes quite frequently, and would reuse an already active
ctx.

If in future we want to avoid kmem_alloc, we can use the
blake3_per_cpu_ctx[CPU_SEQID_UNSTABLE] but check if it is
busy, and move to the next free slot. Easily implemented with
CAS.

Signed-off-by: Jorgen Lundman <[email protected]>

Author: lundman

module/icp/algs/blake3/blake3_impl.c

@@ -364,7 +364,7 @@ blake3_param(ZFS_MODULE_PARAM_ARGS)
 	int err;
 
 	generic_impl_init();
-	if ((void *)req->newptr == NULL) {
+	if ((const void *)req->newptr == NULL) {
 		const uint32_t impl = IMPL_READ(generic_impl_chosen);
 		const int init_buflen = 64;
 		const char *fmt;

module/zfs/blake3_zfs.c

@@ -49,7 +49,7 @@ abd_checksum_blake3_native(abd_t *abd, uint64_t size, const void *ctx_template,
 {
 	ASSERT(ctx_template != NULL);
 
-#if defined(_KERNEL)
+#if defined(_KERNEL) && !(defined(__APPLE__) && defined(__aarch64__))
 	kpreempt_disable();
 	BLAKE3_CTX *ctx = blake3_per_cpu_ctx[CPU_SEQID];
 #else
@@ -60,7 +60,8 @@ abd_checksum_blake3_native(abd_t *abd, uint64_t size, const void *ctx_template,
 	(void) abd_iterate_func(abd, 0, size, blake3_incremental, ctx);
 	Blake3_Final(ctx, (uint8_t *)zcp);
 
-#if defined(_KERNEL)
+#if defined(_KERNEL) && !(defined(__APPLE__) && defined(__aarch64__))
+	/* To keep conditionals the same */
 	kpreempt_enable();
 #else
 	memset(ctx, 0, sizeof (*ctx));