diff --git a/man/man4/zfs.4 b/man/man4/zfs.4
index 7f1adaceb..60e17ddce 100644
--- a/man/man4/zfs.4
+++ b/man/man4/zfs.4
@@ -18,7 +18,7 @@
 .\" own identifying information:
 .\" Portions Copyright [yyyy] [name of copyright owner]
 .\"
-.Dd August 14, 2025
+.Dd September 15, 2025
 .Dt ZFS 4
 .Os
 .
@@ -2583,6 +2583,27 @@ the xattr so as to not accumulate duplicates.
 .It Sy zio_requeue_io_start_cut_in_line Ns = Ns Sy 0 Ns | Ns 1 Pq int
 Prioritize requeued I/O.
 .
+.It Sy zfs_delete_inode Ns = Ns Sy 0 Ns | Ns 1 Pq int
+Sets whether the kernel should free an inode structure when the last reference
+is released, or cache it in memory.
+Intended for testing/debugging.
+.Pp
+A live inode structure "pins" versious internal OpenZFS structures in memory,
+which can result in large amounts of "unusable" memory on systems with lots of
+infrequently-accessed files, until the kernel's memory pressure mechanism
+asks OpenZFS to release them.
+.Pp
+The default value of
+.Sy 0
+always caches inodes that appear to still exist on disk.
+Setting it to
+.Sy 1
+will immediately release unused inodes and their associated memory back to the
+dbuf cache or the ARC for reuse, but may reduce performance if inodes are
+frequently evicted and reloaded.
+.Pp
+This parameter is only available on Linux.
+.
 .It Sy zio_taskq_batch_pct Ns = Ns Sy 80 Ns % Pq uint
 Percentage of online CPUs which will run a worker thread for I/O.
 These workers are responsible for I/O work such as compression, encryption,
diff --git a/module/os/linux/zfs/zpl_super.c b/module/os/linux/zfs/zpl_super.c
index 538196286..591a4ee28 100644
--- a/module/os/linux/zfs/zpl_super.c
+++ b/module/os/linux/zfs/zpl_super.c
@@ -22,6 +22,7 @@
 /*
  * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
  * Copyright (c) 2023, Datto Inc. All rights reserved.
+ * Copyright (c) 2025, Klara, Inc.
  */
 
 
@@ -33,6 +34,12 @@
 #include <linux/iversion.h>
 #include <linux/version.h>
 
+/*
+ * What to do when the last reference to an inode is released. If 0, the kernel
+ * will cache it on the superblock. If 1, the inode will be freed immediately.
+ * See zpl_drop_inode().
+ */
+int zfs_delete_inode = 0;
 
 static struct inode *
 zpl_inode_alloc(struct super_block *sb)
@@ -77,11 +84,36 @@ zpl_dirty_inode(struct inode *ip, int flags)
 }
 
 /*
- * When ->drop_inode() is called its return value indicates if the
- * inode should be evicted from the inode cache.  If the inode is
- * unhashed and has no links the default policy is to evict it
- * immediately.
+ * ->drop_inode() is called when the last reference to an inode is released.
+ * Its return value indicates if the inode should be destroyed immediately, or
+ * cached on the superblock structure.
  *
+ * By default (zfs_delete_inode=0), we call generic_drop_inode(), which returns
+ * "destroy immediately" if the inode is unhashed and has no links (roughly: no
+ * longer exists on disk). On datasets with millions of rarely-accessed files,
+ * this can cause a large amount of memory to be "pinned" by cached inodes,
+ * which in turn pin their associated dnodes and dbufs, until the kernel starts
+ * reporting memory pressure and requests OpenZFS release some memory (see
+ * zfs_prune()).
+ *
+ * When set to 1, we call generic_delete_node(), which always returns "destroy
+ * immediately", resulting in inodes being destroyed immediately, releasing
+ * their associated dnodes and dbufs to the dbuf cached and the ARC to be
+ * evicted as normal.
+ *
+ * Note that the "last reference" doesn't always mean the last _userspace_
+ * reference; the dentry cache also holds a reference, so "busy" inodes will
+ * still be kept alive that way (subject to dcache tuning).
+ */
+static int
+zpl_drop_inode(struct inode *ip)
+{
+	if (zfs_delete_inode)
+		return (generic_delete_inode(ip));
+	return (generic_drop_inode(ip));
+}
+
+/*
  * The ->evict_inode() callback must minimally truncate the inode pages,
  * and call clear_inode().  For 2.6.35 and later kernels this will
  * simply update the inode state, with the sync occurring before the
@@ -470,6 +502,7 @@ const struct super_operations zpl_super_operations = {
 	.destroy_inode		= zpl_inode_destroy,
 	.dirty_inode		= zpl_dirty_inode,
 	.write_inode		= NULL,
+	.drop_inode		= zpl_drop_inode,
 	.evict_inode		= zpl_evict_inode,
 	.put_super		= zpl_put_super,
 	.sync_fs		= zpl_sync_fs,
@@ -491,3 +524,6 @@ struct file_system_type zpl_fs_type = {
 	.mount			= zpl_mount,
 	.kill_sb		= zpl_kill_sb,
 };
+
+ZFS_MODULE_PARAM(zfs, zfs_, delete_inode, INT, ZMOD_RW,
+	"Delete inodes as soon as the last reference is released.");