/*

 * CDDL HEADER START

 *

 * The contents of this file are subject to the terms of the

 * Common Development and Distribution License (the "License").

 * You may not use this file except in compliance with the License.

 *

 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE

 * or http://www.opensolaris.org/os/licensing.

 * See the License for the specific language governing permissions

 * and limitations under the License.

 *

 * When distributing Covered Code, include this CDDL HEADER in each

 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.

 * If applicable, add the following below this CDDL HEADER, with the

 * fields enclosed by brackets "[]" replaced with your own identifying

 * information: Portions Copyright [yyyy] [name of copyright owner]

 *

 * CDDL HEADER END

 */

/*

 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.

 * Copyright (c) 2013 by Delphix. All rights reserved.

 * Copyright (c) 2013 Steven Hartland. All rights reserved.

 */

#include <sys/zfs_context.h>

#include <sys/dsl_userhold.h>

#include <sys/dsl_dataset.h>

#include <sys/dsl_synctask.h>

#include <sys/dmu_tx.h>

#include <sys/dsl_pool.h>

#include <sys/dsl_dir.h>

#include <sys/dmu_traverse.h>

#include <sys/dsl_scan.h>

#include <sys/dmu_objset.h>

#include <sys/zap.h>

#include <sys/zfeature.h>

#include <sys/zfs_ioctl.h>

#include <sys/dsl_deleg.h>

typedef struct dmu_snapshots_destroy_arg {

	nvlist_t *dsda_snaps;

	nvlist_t *dsda_successful_snaps;

	boolean_t dsda_defer;

	nvlist_t *dsda_errlist;

} dmu_snapshots_destroy_arg_t;

int

dsl_destroy_snapshot_check_impl(dsl_dataset_t *ds, boolean_t defer)

{

	if (!dsl_dataset_is_snapshot(ds))

		return (SET_ERROR(EINVAL));

	if (dsl_dataset_long_held(ds))

		return (SET_ERROR(EBUSY));

	/*

	 * Only allow deferred destroy on pools that support it.

	 * NOTE: deferred destroy is only supported on snapshots.

	 */

	if (defer) {

		if (spa_version(ds->ds_dir->dd_pool->dp_spa) <

		    SPA_VERSION_USERREFS)

			return (SET_ERROR(ENOTSUP));

		return (0);

	}

	/*

	 * If this snapshot has an elevated user reference count,

	 * we can't destroy it yet.

	 */

	if (ds->ds_userrefs > 0)

		return (SET_ERROR(EBUSY));

	/*

	 * Can't delete a branch point.

	 */

	if (ds->ds_phys->ds_num_children > 1)

		return (SET_ERROR(EEXIST));

	return (0);

}

static int

dsl_destroy_snapshot_check(void *arg, dmu_tx_t *tx)

{

	dmu_snapshots_destroy_arg_t *dsda = arg;

	dsl_pool_t *dp = dmu_tx_pool(tx);

	nvpair_t *pair;

	int error = 0;

	if (!dmu_tx_is_syncing(tx))

		return (0);

	for (pair = nvlist_next_nvpair(dsda->dsda_snaps, NULL);

	    pair != NULL; pair = nvlist_next_nvpair(dsda->dsda_snaps, pair)) {

		dsl_dataset_t *ds;

		error = dsl_dataset_hold(dp, nvpair_name(pair),

		    FTAG, &ds);

		/*

		 * If the snapshot does not exist, silently ignore it

		 * (it's "already destroyed").

		 */

		if (error == ENOENT)

			continue;

		if (error == 0) {

			error = dsl_destroy_snapshot_check_impl(ds,

			    dsda->dsda_defer);

			dsl_dataset_rele(ds, FTAG);

		}

		if (error == 0) {

			fnvlist_add_boolean(dsda->dsda_successful_snaps,

			    nvpair_name(pair));

		} else {

			fnvlist_add_int32(dsda->dsda_errlist,

			    nvpair_name(pair), error);

		}

	}

	pair = nvlist_next_nvpair(dsda->dsda_errlist, NULL);

	if (pair != NULL)

		return (fnvpair_value_int32(pair));

	return (0);

}

struct process_old_arg {

	dsl_dataset_t *ds;

	dsl_dataset_t *ds_prev;

	boolean_t after_branch_point;

	zio_t *pio;

	uint64_t used, comp, uncomp;

};

static int

process_old_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)

{

	struct process_old_arg *poa = arg;

	dsl_pool_t *dp = poa->ds->ds_dir->dd_pool;

	if (bp->blk_birth <= poa->ds->ds_phys->ds_prev_snap_txg) {

		dsl_deadlist_insert(&poa->ds->ds_deadlist, bp, tx);

		if (poa->ds_prev && !poa->after_branch_point &&

		    bp->blk_birth >

		    poa->ds_prev->ds_phys->ds_prev_snap_txg) {

			poa->ds_prev->ds_phys->ds_unique_bytes +=

			    bp_get_dsize_sync(dp->dp_spa, bp);

		}

	} else {

		poa->used += bp_get_dsize_sync(dp->dp_spa, bp);

		poa->comp += BP_GET_PSIZE(bp);

		poa->uncomp += BP_GET_UCSIZE(bp);

		dsl_free_sync(poa->pio, dp, tx->tx_txg, bp);

	}

	return (0);

}

static void

process_old_deadlist(dsl_dataset_t *ds, dsl_dataset_t *ds_prev,

    dsl_dataset_t *ds_next, boolean_t after_branch_point, dmu_tx_t *tx)

{

	struct process_old_arg poa = { 0 };

	dsl_pool_t *dp = ds->ds_dir->dd_pool;

	objset_t *mos = dp->dp_meta_objset;

	uint64_t deadlist_obj;

	ASSERT(ds->ds_deadlist.dl_oldfmt);

	ASSERT(ds_next->ds_deadlist.dl_oldfmt);

	poa.ds = ds;

	poa.ds_prev = ds_prev;

	poa.after_branch_point = after_branch_point;

	poa.pio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);

	VERIFY0(bpobj_iterate(&ds_next->ds_deadlist.dl_bpobj,

	    process_old_cb, &poa, tx));

	VERIFY0(zio_wait(poa.pio));

	ASSERT3U(poa.used, ==, ds->ds_phys->ds_unique_bytes);

	/* change snapused */

	dsl_dir_diduse_space(ds->ds_dir, DD_USED_SNAP,

	    -poa.used, -poa.comp, -poa.uncomp, tx);

	/* swap next's deadlist to our deadlist */

	dsl_deadlist_close(&ds->ds_deadlist);

	dsl_deadlist_close(&ds_next->ds_deadlist);

	deadlist_obj = ds->ds_phys->ds_deadlist_obj;

	ds->ds_phys->ds_deadlist_obj = ds_next->ds_phys->ds_deadlist_obj;

	ds_next->ds_phys->ds_deadlist_obj = deadlist_obj;

	dsl_deadlist_open(&ds->ds_deadlist, mos, ds->ds_phys->ds_deadlist_obj);

	dsl_deadlist_open(&ds_next->ds_deadlist, mos,

	    ds_next->ds_phys->ds_deadlist_obj);

}

static void

dsl_dataset_remove_clones_key(dsl_dataset_t *ds, uint64_t mintxg, dmu_tx_t *tx)

{

	objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;

	zap_cursor_t zc;

	zap_attribute_t za;

	/*

	 * If it is the old version, dd_clones doesn't exist so we can't

	 * find the clones, but dsl_deadlist_remove_key() is a no-op so it

	 * doesn't matter.

	 */

	if (ds->ds_dir->dd_phys->dd_clones == 0)

		return;

	for (zap_cursor_init(&zc, mos, ds->ds_dir->dd_phys->dd_clones);

	    zap_cursor_retrieve(&zc, &za) == 0;

	    zap_cursor_advance(&zc)) {

		dsl_dataset_t *clone;

		VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,

		    za.za_first_integer, FTAG, &clone));

		if (clone->ds_dir->dd_origin_txg > mintxg) {

			dsl_deadlist_remove_key(&clone->ds_deadlist,

			    mintxg, tx);

			dsl_dataset_remove_clones_key(clone, mintxg, tx);

		}

		dsl_dataset_rele(clone, FTAG);

	}

	zap_cursor_fini(&zc);

}

void

dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx)

{

	int err;

	int after_branch_point = FALSE;

	dsl_pool_t *dp = ds->ds_dir->dd_pool;

	objset_t *mos = dp->dp_meta_objset;

	dsl_dataset_t *ds_prev = NULL;

	uint64_t obj;

	ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));

	ASSERT3U(ds->ds_phys->ds_bp.blk_birth, <=, tx->tx_txg);

	ASSERT(refcount_is_zero(&ds->ds_longholds));

	if (defer &&

	    (ds->ds_userrefs > 0 || ds->ds_phys->ds_num_children > 1)) {

		ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);

		dmu_buf_will_dirty(ds->ds_dbuf, tx);

		ds->ds_phys->ds_flags |= DS_FLAG_DEFER_DESTROY;

		spa_history_log_internal_ds(ds, "defer_destroy", tx, "");

		return;

	}

	ASSERT3U(ds->ds_phys->ds_num_children, <=, 1);

	/* We need to log before removing it from the namespace. */

	spa_history_log_internal_ds(ds, "destroy", tx, "");

	dsl_scan_ds_destroyed(ds, tx);

	obj = ds->ds_object;

	if (ds->ds_phys->ds_prev_snap_obj != 0) {

		ASSERT3P(ds->ds_prev, ==, NULL);

		VERIFY0(dsl_dataset_hold_obj(dp,

		    ds->ds_phys->ds_prev_snap_obj, FTAG, &ds_prev));

		after_branch_point =

		    (ds_prev->ds_phys->ds_next_snap_obj != obj);

		dmu_buf_will_dirty(ds_prev->ds_dbuf, tx);

		if (after_branch_point &&

		    ds_prev->ds_phys->ds_next_clones_obj != 0) {

			dsl_dataset_remove_from_next_clones(ds_prev, obj, tx);

			if (ds->ds_phys->ds_next_snap_obj != 0) {

				VERIFY0(zap_add_int(mos,

				    ds_prev->ds_phys->ds_next_clones_obj,

				    ds->ds_phys->ds_next_snap_obj, tx));

			}

		}

		if (!after_branch_point) {

			ds_prev->ds_phys->ds_next_snap_obj =

			    ds->ds_phys->ds_next_snap_obj;

		}

	}

	dsl_dataset_t *ds_next;

	uint64_t old_unique;

	uint64_t used = 0, comp = 0, uncomp = 0;

	VERIFY0(dsl_dataset_hold_obj(dp,

	    ds->ds_phys->ds_next_snap_obj, FTAG, &ds_next));

	ASSERT3U(ds_next->ds_phys->ds_prev_snap_obj, ==, obj);

	old_unique = ds_next->ds_phys->ds_unique_bytes;

	dmu_buf_will_dirty(ds_next->ds_dbuf, tx);

	ds_next->ds_phys->ds_prev_snap_obj =

	    ds->ds_phys->ds_prev_snap_obj;

	ds_next->ds_phys->ds_prev_snap_txg =

	    ds->ds_phys->ds_prev_snap_txg;

	ASSERT3U(ds->ds_phys->ds_prev_snap_txg, ==,

	    ds_prev ? ds_prev->ds_phys->ds_creation_txg : 0);

	if (ds_next->ds_deadlist.dl_oldfmt) {

		process_old_deadlist(ds, ds_prev, ds_next,

		    after_branch_point, tx);

	} else {

		/* Adjust prev's unique space. */

		if (ds_prev && !after_branch_point) {

			dsl_deadlist_space_range(&ds_next->ds_deadlist,

			    ds_prev->ds_phys->ds_prev_snap_txg,

			    ds->ds_phys->ds_prev_snap_txg,

			    &used, &comp, &uncomp);

			ds_prev->ds_phys->ds_unique_bytes += used;

		}

		/* Adjust snapused. */

		dsl_deadlist_space_range(&ds_next->ds_deadlist,

		    ds->ds_phys->ds_prev_snap_txg, UINT64_MAX,

		    &used, &comp, &uncomp);

		dsl_dir_diduse_space(ds->ds_dir, DD_USED_SNAP,

		    -used, -comp, -uncomp, tx);

		/* Move blocks to be freed to pool's free list. */

		dsl_deadlist_move_bpobj(&ds_next->ds_deadlist,

		    &dp->dp_free_bpobj, ds->ds_phys->ds_prev_snap_txg,

		    tx);

		dsl_dir_diduse_space(tx->tx_pool->dp_free_dir,

		    DD_USED_HEAD, used, comp, uncomp, tx);

		/* Merge our deadlist into next's and free it. */

		dsl_deadlist_merge(&ds_next->ds_deadlist,

		    ds->ds_phys->ds_deadlist_obj, tx);

	}

	dsl_deadlist_close(&ds->ds_deadlist);

	dsl_deadlist_free(mos, ds->ds_phys->ds_deadlist_obj, tx);

	dmu_buf_will_dirty(ds->ds_dbuf, tx);

	ds->ds_phys->ds_deadlist_obj = 0;

	/* Collapse range in clone heads */

	dsl_dataset_remove_clones_key(ds,

	    ds->ds_phys->ds_creation_txg, tx);

	if (dsl_dataset_is_snapshot(ds_next)) {

		dsl_dataset_t *ds_nextnext;

		/*

		 * Update next's unique to include blocks which

		 * were previously shared by only this snapshot

		 * and it.  Those blocks will be born after the

		 * prev snap and before this snap, and will have

		 * died after the next snap and before the one

		 * after that (ie. be on the snap after next's

		 * deadlist).

		 */

		VERIFY0(dsl_dataset_hold_obj(dp,

		    ds_next->ds_phys->ds_next_snap_obj, FTAG, &ds_nextnext));

		dsl_deadlist_space_range(&ds_nextnext->ds_deadlist,

		    ds->ds_phys->ds_prev_snap_txg,

		    ds->ds_phys->ds_creation_txg,

		    &used, &comp, &uncomp);

		ds_next->ds_phys->ds_unique_bytes += used;

		dsl_dataset_rele(ds_nextnext, FTAG);

		ASSERT3P(ds_next->ds_prev, ==, NULL);

		/* Collapse range in this head. */

		dsl_dataset_t *hds;

		VERIFY0(dsl_dataset_hold_obj(dp,

		    ds->ds_dir->dd_phys->dd_head_dataset_obj, FTAG, &hds));

		dsl_deadlist_remove_key(&hds->ds_deadlist,

		    ds->ds_phys->ds_creation_txg, tx);

		dsl_dataset_rele(hds, FTAG);

	} else {

		ASSERT3P(ds_next->ds_prev, ==, ds);

		dsl_dataset_rele(ds_next->ds_prev, ds_next);

		ds_next->ds_prev = NULL;

		if (ds_prev) {

			VERIFY0(dsl_dataset_hold_obj(dp,

			    ds->ds_phys->ds_prev_snap_obj,

			    ds_next, &ds_next->ds_prev));

		}

		dsl_dataset_recalc_head_uniq(ds_next);

		/*

		 * Reduce the amount of our unconsumed refreservation

		 * being charged to our parent by the amount of

		 * new unique data we have gained.

		 */

		if (old_unique < ds_next->ds_reserved) {

			int64_t mrsdelta;

			uint64_t new_unique =

			    ds_next->ds_phys->ds_unique_bytes;

			ASSERT(old_unique <= new_unique);

			mrsdelta = MIN(new_unique - old_unique,

			    ds_next->ds_reserved - old_unique);

			dsl_dir_diduse_space(ds->ds_dir,

			    DD_USED_REFRSRV, -mrsdelta, 0, 0, tx);

		}

	}

	dsl_dataset_rele(ds_next, FTAG);

	/*

	 * This must be done after the dsl_traverse(), because it will

	 * re-open the objset.

	 */

	if (ds->ds_objset) {

		dmu_objset_evict(ds->ds_objset);

		ds->ds_objset = NULL;

	}

	/* remove from snapshot namespace */

	dsl_dataset_t *ds_head;

	ASSERT(ds->ds_phys->ds_snapnames_zapobj == 0);

	VERIFY0(dsl_dataset_hold_obj(dp,

	    ds->ds_dir->dd_phys->dd_head_dataset_obj, FTAG, &ds_head));

	VERIFY0(dsl_dataset_get_snapname(ds));

#ifdef ZFS_DEBUG

	{

		uint64_t val;

		err = dsl_dataset_snap_lookup(ds_head,

		    ds->ds_snapname, &val);

		ASSERT0(err);

		ASSERT3U(val, ==, obj);

	}

#endif

	VERIFY0(dsl_dataset_snap_remove(ds_head, ds->ds_snapname, tx));

	dsl_dataset_rele(ds_head, FTAG);

	if (ds_prev != NULL)

		dsl_dataset_rele(ds_prev, FTAG);

	spa_prop_clear_bootfs(dp->dp_spa, ds->ds_object, tx);

	if (ds->ds_phys->ds_next_clones_obj != 0) {

		uint64_t count;

		ASSERT0(zap_count(mos,

		    ds->ds_phys->ds_next_clones_obj, &count) && count == 0);

		VERIFY0(dmu_object_free(mos,

		    ds->ds_phys->ds_next_clones_obj, tx));

	}

	if (ds->ds_phys->ds_props_obj != 0)

		VERIFY0(zap_destroy(mos, ds->ds_phys->ds_props_obj, tx));

	if (ds->ds_phys->ds_userrefs_obj != 0)

		VERIFY0(zap_destroy(mos, ds->ds_phys->ds_userrefs_obj, tx));

	dsl_dir_rele(ds->ds_dir, ds);

	ds->ds_dir = NULL;

	VERIFY0(dmu_object_free(mos, obj, tx));

}

static void

dsl_destroy_snapshot_sync(void *arg, dmu_tx_t *tx)

{

	dmu_snapshots_destroy_arg_t *dsda = arg;

	dsl_pool_t *dp = dmu_tx_pool(tx);

	nvpair_t *pair;

	for (pair = nvlist_next_nvpair(dsda->dsda_successful_snaps, NULL);

	    pair != NULL;

	    pair = nvlist_next_nvpair(dsda->dsda_successful_snaps, pair)) {

		dsl_dataset_t *ds;

		VERIFY0(dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds));

		dsl_destroy_snapshot_sync_impl(ds, dsda->dsda_defer, tx);

		dsl_dataset_rele(ds, FTAG);

	}

}

/*

 * The semantics of this function are described in the comment above

 * lzc_destroy_snaps().  To summarize:

 *

 * The snapshots must all be in the same pool.

 *

 * Snapshots that don't exist will be silently ignored (considered to be

 * "already deleted").

 *

 * On success, all snaps will be destroyed and this will return 0.

 * On failure, no snaps will be destroyed, the errlist will be filled in,

 * and this will return an errno.

 */

int

dsl_destroy_snapshots_nvl(nvlist_t *snaps, boolean_t defer,

    nvlist_t *errlist)

{

	dmu_snapshots_destroy_arg_t dsda;

	int error;

	nvpair_t *pair;

	pair = nvlist_next_nvpair(snaps, NULL);

	if (pair == NULL)

		return (0);

	dsda.dsda_snaps = snaps;

	dsda.dsda_successful_snaps = fnvlist_alloc();

	dsda.dsda_defer = defer;

	dsda.dsda_errlist = errlist;

	error = dsl_sync_task(nvpair_name(pair),

	    dsl_destroy_snapshot_check, dsl_destroy_snapshot_sync,

	    &dsda, 0);

	fnvlist_free(dsda.dsda_successful_snaps);

	return (error);

}

int

dsl_destroy_snapshot(const char *name, boolean_t defer)

{

	int error;

	nvlist_t *nvl = fnvlist_alloc();

	nvlist_t *errlist = fnvlist_alloc();

	fnvlist_add_boolean(nvl, name);

	error = dsl_destroy_snapshots_nvl(nvl, defer, errlist);

	fnvlist_free(errlist);

	fnvlist_free(nvl);

	return (error);

}

struct killarg {

	dsl_dataset_t *ds;

	dmu_tx_t *tx;