Illumos 4951 - ZFS administrative commands should use reserved space

4951 ZFS administrative commands should use reserved space, not with ENOSPC
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>
Approved by: Garrett D'Amore <garrett@damore.org>

References:
  https://www.illumos.org/issues/4373
  illumos/illumos-gate@7d46dc6

Ported by: Brian Behlendorf <behlendorf1@llnl.gov>

Author: ahrens

cmd/zfs/zfs_main.c

@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  * Copyright (c) 2013 Steven Hartland.  All rights reserved.
  * Copyright 2013 Nexenta Systems, Inc. All rights reserved.
@@ -6636,6 +6636,9 @@ zfs_do_bookmark(int argc, char **argv)
 		case ENOTSUP:
 			err_msg = "bookmark feature not enabled";
 			break;
+		case ENOSPC:
+			err_msg = "out of space";
+			break;
 		default:
 			err_msg = "unknown error";
 			break;
@@ -6644,7 +6647,7 @@ zfs_do_bookmark(int argc, char **argv)
 		    dgettext(TEXT_DOMAIN, err_msg));
 	}
 
-	return (ret);
+	return (ret != 0);
 
 usage:
 	usage(B_FALSE);

cmd/zhack/zhack.c

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
@@ -362,7 +362,7 @@ zhack_do_feature_enable(int argc, char **argv)
 		    feature.fi_guid);
 
 	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
-	    zhack_feature_enable_sync, &feature, 5));
+	    zhack_feature_enable_sync, &feature, 5, ZFS_SPACE_CHECK_NORMAL));
 
 	spa_close(spa, FTAG);
 
@@ -473,7 +473,8 @@ zhack_do_feature_ref(int argc, char **argv)
 	}
 
 	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
-	    decr ? feature_decr_sync : feature_incr_sync, &feature, 5));
+	    decr ? feature_decr_sync : feature_incr_sync, &feature,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 
 	spa_close(spa, FTAG);
 }

include/sys/dsl_synctask.h

@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef	_SYS_DSL_SYNCTASK_H
@@ -38,23 +38,53 @@ struct dsl_pool;
 typedef int (dsl_checkfunc_t)(void *, dmu_tx_t *);
 typedef void (dsl_syncfunc_t)(void *, dmu_tx_t *);
 
+typedef enum zfs_space_check {
+	/*
+	 * Normal space check: if there is less than 3.2% free space,
+	 * the operation will fail.  Operations which are logically
+	 * creating things should use this (e.g. "zfs create", "zfs snapshot").
+	 * User writes (via the ZPL / ZVOL) also fail at this point.
+	 */
+	ZFS_SPACE_CHECK_NORMAL,
+
+	/*
+	 * Space check allows use of half the slop space.  If there
+	 * is less than 1.6% free space, the operation will fail.  Most
+	 * operations should use this (e.g. "zfs set", "zfs rename"),
+	 * because we want them to succeed even after user writes are failing,
+	 * so that they can be used as part of the space recovery process.
+	 */
+	ZFS_SPACE_CHECK_RESERVED,
+
+	/*
+	 * No space check is performed.  Only operations which we expect to
+	 * result in a net reduction in space should use this
+	 * (e.g. "zfs destroy". Setting quotas & reservations also uses
+	 * this because it needs to circumvent the quota/reservation checks).
+	 *
+	 * See also the comments above spa_slop_shift.
+	 */
+	ZFS_SPACE_CHECK_NONE,
+} zfs_space_check_t;
+
 typedef struct dsl_sync_task {
 	txg_node_t dst_node;
 	struct dsl_pool *dst_pool;
 	uint64_t dst_txg;
 	int dst_space;
+	zfs_space_check_t dst_space_check;
 	dsl_checkfunc_t *dst_checkfunc;
 	dsl_syncfunc_t *dst_syncfunc;
 	void *dst_arg;
 	int dst_error;
 	boolean_t dst_nowaiter;
 } dsl_sync_task_t;
 
-void dsl_sync_task_sync(dsl_sync_task_t *dst, dmu_tx_t *tx);
-int dsl_sync_task(const char *pool, dsl_checkfunc_t *checkfunc,
-    dsl_syncfunc_t *syncfunc, void *arg, int blocks_modified);
-void dsl_sync_task_nowait(struct dsl_pool *dp, dsl_syncfunc_t *syncfunc,
-    void *arg, int blocks_modified, dmu_tx_t *tx);
+void dsl_sync_task_sync(dsl_sync_task_t *, dmu_tx_t *);
+int dsl_sync_task(const char *, dsl_checkfunc_t *,
+    dsl_syncfunc_t *, void *, int, zfs_space_check_t);
+void dsl_sync_task_nowait(struct dsl_pool *, dsl_syncfunc_t *,
+    void *, int, zfs_space_check_t, dmu_tx_t *);
 
 #ifdef	__cplusplus
 }

module/zfs/dmu_objset.c

@@ -865,7 +865,8 @@ dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
 	doca.doca_type = type;
 
 	return (dsl_sync_task(name,
-	    dmu_objset_create_check, dmu_objset_create_sync, &doca, 5));
+	    dmu_objset_create_check, dmu_objset_create_sync, &doca,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 typedef struct dmu_objset_clone_arg {
@@ -964,7 +965,8 @@ dmu_objset_clone(const char *clone, const char *origin)
 	doca.doca_cred = CRED();
 
 	return (dsl_sync_task(clone,
-	    dmu_objset_clone_check, dmu_objset_clone_sync, &doca, 5));
+	    dmu_objset_clone_check, dmu_objset_clone_sync, &doca,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 int

module/zfs/dmu_send.c

@@ -1191,7 +1191,7 @@ dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *drrb,
 	drba.drba_cred = CRED();
 
 	return (dsl_sync_task(tofs, dmu_recv_begin_check, dmu_recv_begin_sync,
-	    &drba, 5));
+	    &drba, 5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 struct restorearg {
@@ -2108,7 +2108,7 @@ dmu_recv_existing_end(dmu_recv_cookie_t *drc)
 
 	error = dsl_sync_task(drc->drc_tofs,
 	    dmu_recv_end_check, dmu_recv_end_sync, drc,
-	    dmu_recv_end_modified_blocks);
+	    dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL);
 
 	if (error != 0)
 		dmu_recv_cleanup_ds(drc);
@@ -2122,7 +2122,7 @@ dmu_recv_new_end(dmu_recv_cookie_t *drc)
 
 	error = dsl_sync_task(drc->drc_tofs,
 	    dmu_recv_end_check, dmu_recv_end_sync, drc,
-	    dmu_recv_end_modified_blocks);
+	    dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL);
 
 	if (error != 0) {
 		dmu_recv_cleanup_ds(drc);

module/zfs/dsl_bookmark.c

@@ -13,7 +13,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -249,7 +249,8 @@ dsl_bookmark_create(nvlist_t *bmarks, nvlist_t *errors)
 	dbca.dbca_errors = errors;
 
 	return (dsl_sync_task(nvpair_name(pair), dsl_bookmark_create_check,
-	    dsl_bookmark_create_sync, &dbca, fnvlist_num_pairs(bmarks)));
+	    dsl_bookmark_create_sync, &dbca,
+	    fnvlist_num_pairs(bmarks), ZFS_SPACE_CHECK_NORMAL));
 }
 
 int
@@ -454,7 +455,8 @@ dsl_bookmark_destroy(nvlist_t *bmarks, nvlist_t *errors)
 	dbda.dbda_success = fnvlist_alloc();
 
 	rv = dsl_sync_task(nvpair_name(pair), dsl_bookmark_destroy_check,
-	    dsl_bookmark_destroy_sync, &dbda, fnvlist_num_pairs(bmarks));
+	    dsl_bookmark_destroy_sync, &dbda, fnvlist_num_pairs(bmarks),
+	    ZFS_SPACE_CHECK_RESERVED);
 	fnvlist_free(dbda.dbda_success);
 	return (rv);
 }

module/zfs/dsl_dataset.c

@@ -1406,7 +1406,7 @@ dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors)
 	if (error == 0) {
 		error = dsl_sync_task(firstname, dsl_dataset_snapshot_check,
 		    dsl_dataset_snapshot_sync, &ddsa,
-		    fnvlist_num_pairs(snaps) * 3);
+		    fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL);
 	}
 
 	if (suspended != NULL) {
@@ -1518,7 +1518,7 @@ dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
 	}
 
 	error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check,
-	    dsl_dataset_snapshot_tmp_sync, &ddsta, 3);
+	    dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED);
 
 	if (needsuspend)
 		zil_resume(cookie);
@@ -1874,7 +1874,8 @@ dsl_dataset_rename_snapshot(const char *fsname,
 	ddrsa.ddrsa_recursive = recursive;
 
 	error = dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check,
-	    dsl_dataset_rename_snapshot_sync, &ddrsa, 1);
+	    dsl_dataset_rename_snapshot_sync, &ddrsa,
+	    1, ZFS_SPACE_CHECK_RESERVED);
 
 	if (error)
 	    return (SET_ERROR(error));
@@ -2064,7 +2065,8 @@ dsl_dataset_rollback(const char *fsname, void *owner, nvlist_t *result)
 	ddra.ddra_result = result;
 
 	return (dsl_sync_task(fsname, dsl_dataset_rollback_check,
-	    dsl_dataset_rollback_sync, &ddra, 1));
+	    dsl_dataset_rollback_sync, &ddra,
+	    1, ZFS_SPACE_CHECK_RESERVED));
 }
 
 struct promotenode {
@@ -2595,7 +2597,8 @@ dsl_dataset_promote(const char *name, char *conflsnap)
 	ddpa.cr = CRED();
 
 	return (dsl_sync_task(name, dsl_dataset_promote_check,
-	    dsl_dataset_promote_sync, &ddpa, 2 + numsnaps));
+	    dsl_dataset_promote_sync, &ddpa,
+	    2 + numsnaps, ZFS_SPACE_CHECK_RESERVED));
 }
 
 int
@@ -2949,7 +2952,7 @@ dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
 	ddsqra.ddsqra_value = refquota;
 
 	return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check,
-	    dsl_dataset_set_refquota_sync, &ddsqra, 0));
+	    dsl_dataset_set_refquota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 static int
@@ -3064,7 +3067,8 @@ dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
 	ddsqra.ddsqra_value = refreservation;
 
 	return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check,
-	    dsl_dataset_set_refreservation_sync, &ddsqra, 0));
+	    dsl_dataset_set_refreservation_sync, &ddsqra,
+	    0, ZFS_SPACE_CHECK_NONE));
 }
 
 /*

module/zfs/dsl_deleg.c

@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  */
 
 /*
@@ -282,7 +282,7 @@ dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset)
 
 	return (dsl_sync_task(ddname, dsl_deleg_check,
 	    unset ? dsl_deleg_unset_sync : dsl_deleg_set_sync,
-	    &dda, fnvlist_num_pairs(nvp)));
+	    &dda, fnvlist_num_pairs(nvp), ZFS_SPACE_CHECK_RESERVED));
 }
 
 /*

module/zfs/dsl_destroy.c

@@ -520,7 +520,7 @@ dsl_destroy_snapshots_nvl(nvlist_t *snaps, boolean_t defer,
 
 	error = dsl_sync_task(nvpair_name(pair),
 	    dsl_destroy_snapshot_check, dsl_destroy_snapshot_sync,
-	    &dsda, 0);
+	    &dsda, 0, ZFS_SPACE_CHECK_NONE);
 	fnvlist_free(dsda.dsda_successful_snaps);
 
 	return (error);
@@ -918,7 +918,8 @@ dsl_destroy_head(const char *name)
 		objset_t *os;
 
 		error = dsl_sync_task(name, dsl_destroy_head_check,
-		    dsl_destroy_head_begin_sync, &ddha, 0);
+		    dsl_destroy_head_begin_sync, &ddha,
+		    0, ZFS_SPACE_CHECK_NONE);
 		if (error != 0)
 			return (error);
 
@@ -944,7 +945,7 @@ dsl_destroy_head(const char *name)
 	}
 
 	return (dsl_sync_task(name, dsl_destroy_head_check,
-	    dsl_destroy_head_sync, &ddha, 0));
+	    dsl_destroy_head_sync, &ddha, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 /*

module/zfs/dsl_dir.c

@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Martin Matuska. All rights reserved.
  * Copyright (c) 2014 Joyent, Inc. All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
@@ -665,7 +665,8 @@ dsl_dir_activate_fs_ss_limit(const char *ddname)
 	int error;
 
 	error = dsl_sync_task(ddname, dsl_dir_actv_fs_ss_limit_check,
-	    dsl_dir_actv_fs_ss_limit_sync, (void *)ddname, 0);
+	    dsl_dir_actv_fs_ss_limit_sync, (void *)ddname, 0,
+	    ZFS_SPACE_CHECK_RESERVED);
 
 	if (error == EALREADY)
 		error = 0;
@@ -1529,7 +1530,7 @@ dsl_dir_set_quota(const char *ddname, zprop_source_t source, uint64_t quota)
 	ddsqra.ddsqra_value = quota;
 
 	return (dsl_sync_task(ddname, dsl_dir_set_quota_check,
-	    dsl_dir_set_quota_sync, &ddsqra, 0));
+	    dsl_dir_set_quota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 int
@@ -1650,7 +1651,7 @@ dsl_dir_set_reservation(const char *ddname, zprop_source_t source,
 	ddsqra.ddsqra_value = reservation;
 
 	return (dsl_sync_task(ddname, dsl_dir_set_reservation_check,
-	    dsl_dir_set_reservation_sync, &ddsqra, 0));
+	    dsl_dir_set_reservation_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 static dsl_dir_t *
@@ -1932,7 +1933,8 @@ dsl_dir_rename(const char *oldname, const char *newname)
 	ddra.ddra_cred = CRED();
 
 	return (dsl_sync_task(oldname,
-	    dsl_dir_rename_check, dsl_dir_rename_sync, &ddra, 3));
+	    dsl_dir_rename_check, dsl_dir_rename_sync, &ddra,
+	    3, ZFS_SPACE_CHECK_RESERVED));
 }
 
 int