diff --git a/include/sys/zstd/zstd.h b/include/sys/zstd/zstd.h
index f965df3198..e42e44c236 100644
--- a/include/sys/zstd/zstd.h
+++ b/include/sys/zstd/zstd.h
@@ -90,6 +90,7 @@ int zfs_zstd_decompress_level(void *s_start, void *d_start, size_t s_len,
     size_t d_len, uint8_t *level);
 int zfs_zstd_decompress(void *s_start, void *d_start, size_t s_len,
     size_t d_len, int n);
+void zfs_zstd_cache_reap_now(void);
 
 #ifdef	__cplusplus
 }
diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 1cce068e65..c54d53908e 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -308,6 +308,7 @@
 #include <sys/aggsum.h>
 #include <cityhash.h>
 #include <sys/vdev_trim.h>
+#include <sys/zstd/zstd.h>
 
 #ifndef _KERNEL
 /* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
@@ -4972,6 +4973,7 @@ static boolean_t
 arc_reap_cb_check(void *arg, zthr_t *zthr)
 {
 	int64_t free_memory = arc_available_memory();
+	static int reap_cb_check_counter = 0;
 
 	/*
 	 * If a kmem reap is already active, don't schedule more.  We must
@@ -4996,6 +4998,14 @@ arc_reap_cb_check(void *arg, zthr_t *zthr)
 		arc_no_grow = B_FALSE;
 	}
 
+	/*
+	 * Called unconditionally every 60 seconds to reclaim unused
+	 * zstd compression and decompression context. This is done
+	 * here to avoid the need for an independent thread.
+	 */
+	if (!((reap_cb_check_counter++) % 60))
+		zfs_zstd_cache_reap_now();
+
 	return (B_FALSE);
 }
 
diff --git a/module/zstd/zfs_zstd.c b/module/zstd/zfs_zstd.c
index c10b7a84d6..34c56b7a7f 100644
--- a/module/zstd/zfs_zstd.c
+++ b/module/zstd/zfs_zstd.c
@@ -238,7 +238,7 @@ zstd_mempool_alloc(struct zstd_pool *zstd_mempool, size_t size)
 			 * Check if objects fits the size, if so we take it and
 			 * update the timestamp.
 			 */
-			if (!mem && pool->mem && size <= pool->size) {
+			if (size && !mem && pool->mem && size <= pool->size) {
 				pool->timeout = gethrestime_sec() +
 				    ZSTD_POOL_TIMEOUT;
 				mem = pool->mem;
@@ -257,7 +257,7 @@ zstd_mempool_alloc(struct zstd_pool *zstd_mempool, size_t size)
 		}
 	}
 
-	if (mem) {
+	if (!size || mem) {
 		return (mem);
 	}
 
@@ -688,6 +688,19 @@ zstd_mempool_deinit(void)
 	zstd_mempool_cctx = NULL;
 }
 
+/* release unused memory from pool */
+
+void
+zfs_zstd_cache_reap_now(void)
+{
+	/*
+	 * calling alloc with zero size seeks
+	 * and releases old unused objects
+	 */
+	zstd_mempool_alloc(zstd_mempool_cctx, 0);
+	zstd_mempool_alloc(zstd_mempool_dctx, 0);
+}
+
 extern int __init
 zstd_init(void)
 {
@@ -735,4 +748,5 @@ ZFS_MODULE_VERSION(ZSTD_VERSION_STRING);
 EXPORT_SYMBOL(zfs_zstd_compress);
 EXPORT_SYMBOL(zfs_zstd_decompress_level);
 EXPORT_SYMBOL(zfs_zstd_decompress);
+EXPORT_SYMBOL(zfs_zstd_cache_reap_now);
 #endif