avoid retrieving unused snapshot props

This patch modifies the zfs_ioc_snapshot_list_next() ioctl to enable it
to take input parameters that alter the way looping through the list of
snapshots is performed. The idea here is to restrict functions that
throw away some of the snapshots returned by the ioctl to a range of
snapshots that these functions actually use. This improves efficiency
and execution speed for some rollback and send operations.

Signed-off-by: Alek Pinchuk <apinchuk@datto.com>

cmd/zfs/zfs_iter.c

@@ -140,7 +140,7 @@ zfs_callback(zfs_handle_t *zhp, void *data)
 		    ZFS_TYPE_BOOKMARK)) == 0) && include_snaps)
 			(void) zfs_iter_snapshots(zhp,
 			    (cb->cb_flags & ZFS_ITER_SIMPLE) != 0, zfs_callback,
-			    data);
+			    data, NULL);
 		if (((zfs_get_type(zhp) & (ZFS_TYPE_SNAPSHOT |
 		    ZFS_TYPE_BOOKMARK)) == 0) && include_bmarks)
 			(void) zfs_iter_bookmarks(zhp, zfs_callback, data);

cmd/zfs/zfs_main.c

@@ -1255,7 +1255,7 @@ destroy_print_snapshots(zfs_handle_t *fs_zhp, destroy_cbdata_t *cb)
 	int err;
 	assert(cb->cb_firstsnap == NULL);
 	assert(cb->cb_prevsnap == NULL);
-	err = zfs_iter_snapshots_sorted(fs_zhp, destroy_print_cb, cb);
+	err = zfs_iter_snapshots_sorted(fs_zhp, destroy_print_cb, cb, NULL);
 	if (cb->cb_firstsnap != NULL) {
 		uint64_t used = 0;
 		if (err == 0) {
@@ -3581,6 +3581,7 @@ zfs_do_promote(int argc, char **argv)
  */
 typedef struct rollback_cbdata {
 	uint64_t	cb_create;
+	uint8_t		cb_younger_ds_printed;
 	boolean_t	cb_first;
 	int		cb_doclones;
 	char		*cb_target;
@@ -3612,15 +3613,20 @@ rollback_check_dependent(zfs_handle_t *zhp, void *data)
 
 
 /*
- * Report any snapshots more recent than the one specified.  Used when '-r' is
- * not specified.  We reuse this same callback for the snapshot dependents - if
- * 'cb_dependent' is set, then this is a dependent and we should report it
- * without checking the transaction group.
+ * Report some snapshots/bookmarks more recent than the one specified.
+ * Used when '-r' is not specified. We reuse this same callback for the
+ * snapshot dependents - if 'cb_dependent' is set, then this is a
+ * dependent and we should report it without checking the transaction group.
  */
 static int
 rollback_check(zfs_handle_t *zhp, void *data)
 {
 	rollback_cbdata_t *cbp = data;
+	/*
+	 * Max number of younger snapshtos and/or bookmarks to display before
+	 * we stop the iteration
+	 */
+	const uint8_t max_younger = 32;
 
 	if (cbp->cb_doclones) {
 		zfs_close(zhp);
@@ -3649,9 +3655,25 @@ rollback_check(zfs_handle_t *zhp, void *data)
 		} else {
 			(void) fprintf(stderr, "%s\n",
 			    zfs_get_name(zhp));
+			cbp->cb_younger_ds_printed++;
 		}
 	}
 	zfs_close(zhp);
+
+	if (cbp->cb_younger_ds_printed == max_younger) {
+		/*
+		 * This non-recursive rollback is going to fail due to the
+		 * presence of snapshots and/or bookmarks that are younger than
+		 * the rollback target.
+		 * We printed some of the offending objects, now we stop
+		 * zfs_iter_snapshot/bookmark iteration so we can fail fast and
+		 * avoid iterating over the rest of the younger objects
+		 */
+		(void) fprintf(stderr, gettext("these are the first %d "
+		    "snapshots and/or bookmarks following %s\n"),
+		    max_younger, cbp->cb_target);
+		return (-1);
+	}
 	return (0);
 }
 
@@ -3665,6 +3687,7 @@ zfs_do_rollback(int argc, char **argv)
 	zfs_handle_t *zhp, *snap;
 	char parentname[ZFS_MAX_DATASET_NAME_LEN];
 	char *delim;
+	nvlist_t *range_nvl = NULL;
 
 	/* check options */
 	while ((c = getopt(argc, argv, "rRf")) != -1) {
@@ -3720,7 +3743,15 @@ zfs_do_rollback(int argc, char **argv)
 	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
 	cb.cb_first = B_TRUE;
 	cb.cb_error = 0;
-	if ((ret = zfs_iter_snapshots(zhp, B_FALSE, rollback_check, &cb)) != 0)
+
+	if (cb.cb_create > 0) {
+		range_nvl = fnvlist_alloc();
+		fnvlist_add_uint64(range_nvl, SNAP_ITER_SKIP_TO_TXG,
+		    cb.cb_create);
+	}
+
+	if ((ret = zfs_iter_snapshots(zhp, B_FALSE, rollback_check, &cb,
+	    range_nvl)) != 0)
 		goto out;
 	if ((ret = zfs_iter_bookmarks(zhp, rollback_check, &cb)) != 0)
 		goto out;
@@ -3734,6 +3765,7 @@ zfs_do_rollback(int argc, char **argv)
 	ret = zfs_rollback(zhp, snap, force);
 
 out:
+	nvlist_free(range_nvl);
 	zfs_close(snap);
 	zfs_close(zhp);
 

include/libzfs.h

@@ -570,8 +570,10 @@ extern int zfs_iter_root(libzfs_handle_t *, zfs_iter_f, void *);
 extern int zfs_iter_children(zfs_handle_t *, zfs_iter_f, void *);
 extern int zfs_iter_dependents(zfs_handle_t *, boolean_t, zfs_iter_f, void *);
 extern int zfs_iter_filesystems(zfs_handle_t *, zfs_iter_f, void *);
-extern int zfs_iter_snapshots(zfs_handle_t *, boolean_t, zfs_iter_f, void *);
-extern int zfs_iter_snapshots_sorted(zfs_handle_t *, zfs_iter_f, void *);
+extern int zfs_iter_snapshots(zfs_handle_t *, boolean_t, zfs_iter_f, void *,
+    nvlist_t *);
+extern int zfs_iter_snapshots_sorted(zfs_handle_t *, zfs_iter_f, void *,
+    nvlist_t *);
 extern int zfs_iter_snapspec(zfs_handle_t *, const char *, zfs_iter_f, void *);
 extern int zfs_iter_bookmarks(zfs_handle_t *, zfs_iter_f, void *);
 extern int zfs_iter_mounted(zfs_handle_t *, zfs_iter_f, void *);

include/sys/fs/zfs.h

@@ -27,6 +27,7 @@
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright (c) 2017 Datto Inc.
  * Copyright (c) 2017, Intel Corporation.
+ * Copyright (c) 2018 Datto Inc.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -1076,7 +1077,7 @@ typedef enum pool_initialize_func {
  * is passed between kernel and userland as an nvlist uint64 array.
  */
 typedef struct ddt_object {
-	uint64_t	ddo_count;	/* number of elements in ddt 	*/
+	uint64_t	ddo_count;	/* number of elements in ddt	*/
 	uint64_t	ddo_dspace;	/* size of ddt on disk		*/
 	uint64_t	ddo_mspace;	/* size of ddt in-core		*/
 } ddt_object_t;
@@ -1123,6 +1124,13 @@ typedef enum {
 	VDEV_INITIALIZE_COMPLETE
 } vdev_initializing_state_t;
 
+/*
+ * nvlist name constants. Facilitate restricting snapshot iteration range for
+ * the "list next snapshot" ioctl
+ */
+#define	SNAP_ITER_SKIP_AFTER_TXG	"snap_iter_skip_after_txg"
+#define	SNAP_ITER_SKIP_TO_TXG		"snap_iter_skip_to_txg"
+
 /*
  * /dev/zfs ioctl numbers.
  *

lib/libzfs/libzfs_dataset.c

@@ -4371,6 +4371,7 @@ zfs_rollback(zfs_handle_t *zhp, zfs_handle_t *snap, boolean_t force)
 	boolean_t restore_resv = 0;
 	uint64_t old_volsize = 0, new_volsize;
 	zfs_prop_t resv_prop = { 0 };
+	nvlist_t *range = NULL;
 
 	assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM ||
 	    zhp->zfs_type == ZFS_TYPE_VOLUME);
@@ -4381,7 +4382,15 @@ zfs_rollback(zfs_handle_t *zhp, zfs_handle_t *snap, boolean_t force)
 	cb.cb_force = force;
 	cb.cb_target = snap->zfs_name;
 	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
-	(void) zfs_iter_snapshots(zhp, B_FALSE, rollback_destroy, &cb);
+
+	if (cb.cb_create > 0) {
+		range = fnvlist_alloc();
+		fnvlist_add_uint64(range, SNAP_ITER_SKIP_TO_TXG, cb.cb_create);
+	}
+
+	(void) zfs_iter_snapshots(zhp, B_FALSE, rollback_destroy, &cb, range);
+	nvlist_free(range);
+
 	(void) zfs_iter_bookmarks(zhp, rollback_destroy, &cb);
 
 	if (cb.cb_error)

lib/libzfs/libzfs_iter.c

@@ -141,7 +141,7 @@ zfs_iter_filesystems(zfs_handle_t *zhp, zfs_iter_f func, void *data)
  */
 int
 zfs_iter_snapshots(zfs_handle_t *zhp, boolean_t simple, zfs_iter_f func,
-    void *data)
+    void *data, nvlist_t *range_nvl)
 {
 	zfs_cmd_t zc = {"\0"};
 	zfs_handle_t *nzhp;
@@ -155,6 +155,13 @@ zfs_iter_snapshots(zfs_handle_t *zhp, boolean_t simple, zfs_iter_f func,
 
 	if (zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0) != 0)
 		return (-1);
+
+	if (range_nvl != NULL &&
+	    zcmd_write_src_nvlist(zhp->zfs_hdl, &zc, range_nvl) != 0) {
+			zcmd_free_nvlists(&zc);
+			return (-1);
+	}
+
 	while ((ret = zfs_do_list_ioctl(zhp, ZFS_IOC_SNAPSHOT_LIST_NEXT,
 	    &zc)) == 0) {
 
@@ -282,7 +289,8 @@ zfs_snapshot_compare(const void *larg, const void *rarg)
 }
 
 int
-zfs_iter_snapshots_sorted(zfs_handle_t *zhp, zfs_iter_f callback, void *data)
+zfs_iter_snapshots_sorted(zfs_handle_t *zhp, zfs_iter_f callback, void *data,
+    nvlist_t *range_nvl)
 {
 	int ret = 0;
 	zfs_node_t *node;
@@ -292,7 +300,7 @@ zfs_iter_snapshots_sorted(zfs_handle_t *zhp, zfs_iter_f callback, void *data)
 	avl_create(&avl, zfs_snapshot_compare,
 	    sizeof (zfs_node_t), offsetof(zfs_node_t, zn_avlnode));
 
-	ret = zfs_iter_snapshots(zhp, B_FALSE, zfs_sort_snaps, &avl);
+	ret = zfs_iter_snapshots(zhp, B_FALSE, zfs_sort_snaps, &avl, range_nvl);
 
 	for (node = avl_first(&avl); node != NULL; node = AVL_NEXT(&avl, node))
 		ret |= callback(node->zn_handle, data);
@@ -395,7 +403,7 @@ zfs_iter_snapspec(zfs_handle_t *fs_zhp, const char *spec_orig,
 			}
 
 			err = zfs_iter_snapshots_sorted(fs_zhp,
-			    snapspec_cb, &ssa);
+			    snapspec_cb, &ssa, NULL);
 			if (ret == 0)
 				ret = err;
 			if (ret == 0 && (!ssa.ssa_seenfirst ||
@@ -435,7 +443,7 @@ zfs_iter_children(zfs_handle_t *zhp, zfs_iter_f func, void *data)
 {
 	int ret;
 
-	if ((ret = zfs_iter_snapshots(zhp, B_FALSE, func, data)) != 0)
+	if ((ret = zfs_iter_snapshots(zhp, B_FALSE, func, data, NULL)) != 0)
 		return (ret);
 
 	return (zfs_iter_filesystems(zhp, func, data));
@@ -501,7 +509,7 @@ iter_dependents_cb(zfs_handle_t *zhp, void *arg)
 		err = zfs_iter_filesystems(zhp, iter_dependents_cb, ida);
 		if (err == 0)
 			err = zfs_iter_snapshots(zhp, B_FALSE,
-			    iter_dependents_cb, ida);
+			    iter_dependents_cb, ida, NULL);
 		ida->stack = isf.next;
 	}