Fix memory allocation issue for BLAKE3 context

The kmem_alloc(sizeof (*ctx), KM_NOSLEEP) call on FreeBSD can't be
used in this code segment. Work around this by pre-allocating a percpu
context array for later use.

Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tino Reichardt <milky-zfs@mcmilk.de>
Closes #13568

module/zfs/blake3_zfs.c

@@ -47,18 +47,22 @@ void
 abd_checksum_blake3_native(abd_t *abd, uint64_t size, const void *ctx_template,
     zio_cksum_t *zcp)
 {
-	BLAKE3_CTX *ctx;
-
-	ctx = kmem_alloc(sizeof (*ctx), KM_NOSLEEP);
-	ASSERT(ctx != 0);
 	ASSERT(ctx_template != 0);
 
+#if defined(_KERNEL)
+	BLAKE3_CTX *ctx = blake3_per_cpu_ctx[CPU_SEQID_UNSTABLE];
+#else
+	BLAKE3_CTX *ctx = kmem_alloc(sizeof (*ctx), KM_SLEEP);
+#endif
+
 	memcpy(ctx, ctx_template, sizeof (*ctx));
 	(void) abd_iterate_func(abd, 0, size, blake3_incremental, ctx);
 	Blake3_Final(ctx, (uint8_t *)zcp);
 
+#if !defined(_KERNEL)
 	memset(ctx, 0, sizeof (*ctx));
 	kmem_free(ctx, sizeof (*ctx));
+#endif
 }
 
 /*

module/zfs/zfs_chksum.c

@@ -277,6 +277,9 @@ chksum_benchmark(void)
 void
 chksum_init(void)
 {
+#ifdef _KERNEL
+	blake3_per_cpu_ctx_init();
+#endif
 
 	/* Benchmark supported implementations */
 	chksum_benchmark();
@@ -313,4 +316,8 @@ chksum_fini(void)
 		chksum_stat_cnt = 0;
 		chksum_stat_data = 0;
 	}
+
+#ifdef _KERNEL
+	blake3_per_cpu_ctx_fini();
+#endif
 }