/*

 * 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 2006 Sun Microsystems, Inc.  All rights reserved.

 * Use is subject to license terms.

 */

#pragma ident	"%Z%%M%	%I%	%E% SMI"

#include <sys/dmu.h>

#include <sys/dmu_impl.h>

#include <sys/dbuf.h>

#include <sys/dmu_tx.h>

#include <sys/dmu_objset.h>

#include <sys/dsl_dataset.h> /* for dsl_dataset_block_freeable() */

#include <sys/dsl_dir.h> /* for dsl_dir_tempreserve_*() */

#include <sys/dsl_pool.h>

#include <sys/zap_impl.h>	/* for ZAP_BLOCK_SHIFT */

#include <sys/spa.h>

#include <sys/zfs_context.h>

#ifdef ZFS_DEBUG

int dmu_use_tx_debug_bufs = 1;

#endif

dmu_tx_t *

dmu_tx_create_ds(dsl_dir_t *dd)

{

	dmu_tx_t *tx = kmem_zalloc(sizeof (dmu_tx_t), KM_SLEEP);

	tx->tx_dir = dd;

	if (dd)

		tx->tx_pool = dd->dd_pool;

	list_create(&tx->tx_holds, sizeof (dmu_tx_hold_t),

	    offsetof(dmu_tx_hold_t, dth_node));

	refcount_create(&tx->tx_space_written);

	refcount_create(&tx->tx_space_freed);

	return (tx);

}

dmu_tx_t *

dmu_tx_create(objset_t *os)

{

	dmu_tx_t *tx = dmu_tx_create_ds(os->os->os_dsl_dataset->ds_dir);

	tx->tx_objset = os;

	return (tx);

}

dmu_tx_t *

dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg)

{

	dmu_tx_t *tx = dmu_tx_create_ds(NULL);

	ASSERT3U(txg, <=, dp->dp_tx.tx_open_txg);

	tx->tx_pool = dp;

	tx->tx_txg = txg;

	tx->tx_anyobj = TRUE;

	return (tx);

}

int

dmu_tx_is_syncing(dmu_tx_t *tx)

{

	return (tx->tx_anyobj);

}

int

dmu_tx_private_ok(dmu_tx_t *tx)

{

}

static void

dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object,

    enum dmu_tx_hold_type type, dmu_tx_hold_func_t func,

    uint64_t arg1, uint64_t arg2)

{

	dmu_tx_hold_t *dth;

	dnode_t *dn = NULL;

	if (object != DMU_NEW_OBJECT) {

			mutex_enter(&dn->dn_mtx);

			/*

			 * dn->dn_assigned_txg == tx->tx_txg doesn't pose a

			 * problem, but there's no way for it to happen (for

			 * now, at least).

			 */

			ASSERT(dn->dn_assigned_txg == 0);

			ASSERT(dn->dn_assigned_tx == NULL);

			dn->dn_assigned_txg = tx->tx_txg;

			dn->dn_assigned_tx = tx;

			(void) refcount_add(&dn->dn_tx_holds, tx);

			mutex_exit(&dn->dn_mtx);

		}

	}

	dth = kmem_zalloc(sizeof (dmu_tx_hold_t), KM_SLEEP);

	dth->dth_dnode = dn;

	dth->dth_type = type;

	dth->dth_arg1 = arg1;

	dth->dth_arg2 = arg2;

	list_insert_tail(&tx->tx_holds, dth);

}

void

dmu_tx_add_new_object(dmu_tx_t *tx, objset_t *os, uint64_t object)

{

	/*

	 * If we're syncing, they can manipulate any object anyhow, and

	 * the hold on the dnode_t can cause problems.

	 */

	if (!dmu_tx_is_syncing(tx)) {

		dmu_tx_hold_object_impl(tx, os, object, THT_NEWOBJECT,

		    NULL, 0, 0);

	}

}

/* ARGSUSED */

static void

dmu_tx_count_write(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len)

{

	int min_bs, max_bs, min_ibs, max_ibs, epbs, bits;

	if (len == 0)

		return;

	min_bs = SPA_MINBLOCKSHIFT;

	max_bs = SPA_MAXBLOCKSHIFT;

	min_ibs = DN_MIN_INDBLKSHIFT;

	max_ibs = DN_MAX_INDBLKSHIFT;

	/*

	 * If there's more than one block, the blocksize can't change,

	 * so we can make a more precise estimate.  Alternatively,

	 * if the dnode's ibs is larger than max_ibs, always use that.

	 * This ensures that if we reduce DN_MAX_INDBLKSHIFT,

	 * the code will still work correctly on existing pools.

	 */

	if (dn && (dn->dn_maxblkid != 0 || dn->dn_indblkshift > max_ibs)) {

		min_ibs = max_ibs = dn->dn_indblkshift;

		if (dn->dn_datablkshift != 0)

			min_bs = max_bs = dn->dn_datablkshift;

	}

	/*

	 * 'end' is the last thing we will access, not one past.

	 * This way we won't overflow when accessing the last byte.

	 */

	start = P2ALIGN(off, 1ULL << max_bs);

	end = P2ROUNDUP(off + len, 1ULL << max_bs) - 1;

	space = end - start + 1;

	start >>= min_bs;

	end >>= min_bs;

	epbs = min_ibs - SPA_BLKPTRSHIFT;

	/*

	 * The object contains at most 2^(64 - min_bs) blocks,

	 * and each indirect level maps 2^epbs.

	 */

	for (bits = 64 - min_bs; bits >= 0; bits -= epbs) {

		start >>= epbs;

		end >>= epbs;

		/*

		 * If we increase the number of levels of indirection,

		 * we'll need new blkid=0 indirect blocks.  If start == 0,

		 * we're already accounting for that blocks; and if end == 0,

		 * we can't increase the number of levels beyond that.

		 */

		if (start != 0 && end != 0)

			space += 1ULL << max_ibs;

		space += (end - start + 1) << max_ibs;

	}

	ASSERT(space < 2 * DMU_MAX_ACCESS);

	tx->tx_space_towrite += space;

}

static void

dmu_tx_count_dnode(dmu_tx_t *tx, dnode_t *dn)

{

	dnode_t *mdn = tx->tx_objset->os->os_meta_dnode;

	uint64_t object = dn ? dn->dn_object : DN_MAX_OBJECT - 1;

	uint64_t pre_write_space;

	ASSERT(object < DN_MAX_OBJECT);

	pre_write_space = tx->tx_space_towrite;

	dmu_tx_count_write(tx, mdn, object << DNODE_SHIFT, 1 << DNODE_SHIFT);

	if (dn && dn->dn_dbuf->db_blkptr &&

	    dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,

		tx->tx_space_tooverwrite +=

			tx->tx_space_towrite - pre_write_space;

		tx->tx_space_towrite = pre_write_space;

	}

}

/* ARGSUSED */

static void

dmu_tx_hold_write_impl(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len)

{

	dmu_tx_count_write(tx, dn, off, len);

	dmu_tx_count_dnode(tx, dn);

}

void

dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len)

{

	ASSERT(tx->tx_txg == 0);

	ASSERT(UINT64_MAX - off >= len - 1);

	dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_WRITE,

	    dmu_tx_hold_write_impl, off, len);

}

static void

dmu_tx_count_free(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len)

{

	uint64_t blkid, nblks;

	uint64_t space = 0;

	dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;

	if (dn->dn_datablkshift == 0)

		return;

	/*

	 * not that the dnode can change, since it isn't dirty, but

	 * dbuf_hold_impl() wants us to have the struct_rwlock.

	 * also need it to protect dn_maxblkid.

	 */

	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	blkid = off >> dn->dn_datablkshift;

	nblks = (off + len) >> dn->dn_datablkshift;

	if (blkid + nblks > dn->dn_maxblkid)

		nblks = dn->dn_maxblkid - blkid;

	/* don't bother after the 100,000 blocks */

	nblks = MIN(nblks, 128*1024);

	if (dn->dn_phys->dn_nlevels == 1) {

		int i;

		for (i = 0; i < nblks; i++) {

			blkptr_t *bp = dn->dn_phys->dn_blkptr;

			ASSERT3U(blkid + i, <, dn->dn_phys->dn_nblkptr);

			bp += blkid + i;

				dprintf_bp(bp, "can free old%s", "");

				space += BP_GET_ASIZE(bp);

			}

		}

	}

	while (nblks) {

		dmu_buf_impl_t *dbuf;

		int err, epbs, blkoff, tochk;

		epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;

		blkoff = P2PHASE(blkid, 1<<epbs);

		tochk = MIN((1<<epbs) - blkoff, nblks);

		err = dbuf_hold_impl(dn, 1, blkid >> epbs, TRUE, FTAG, &dbuf);

		if (err == 0) {

			int i;

			blkptr_t *bp;

			bp = dbuf->db.db_data;

			bp += blkoff;

			for (i = 0; i < tochk; i++) {

				if (dsl_dataset_block_freeable(ds,

					dprintf_bp(&bp[i],

					    "can free old%s", "");

					space += BP_GET_ASIZE(&bp[i]);

				}

			}

		} else {

			/* the indirect block is sparse */

			ASSERT(err == ENOENT);

		}

		blkid += tochk;

		nblks -= tochk;

	}

	rw_exit(&dn->dn_struct_rwlock);

	tx->tx_space_tofree += space;

}

static void

dmu_tx_hold_free_impl(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len)

{

	/* first block */

	if (off != 0 /* || dn->dn_maxblkid == 0 */)

		dmu_tx_count_write(tx, dn, off, 1);

	/* last block */

	if (len != DMU_OBJECT_END)

		dmu_tx_count_write(tx, dn, off+len, 1);

	if (off >= (dn->dn_maxblkid+1) * dn->dn_datablksz)

		return;

	if (len == DMU_OBJECT_END)

		len = (dn->dn_maxblkid+1) * dn->dn_datablksz - off;

	/* XXX locking */

	dirty = dn->dn_dirtyblksz[0] | dn->dn_dirtyblksz[1] |

	    dn->dn_dirtyblksz[2] | dn->dn_dirtyblksz[3];

	if (dn->dn_assigned_tx != NULL && !dirty)

		dmu_tx_count_free(tx, dn, off, len);

}

void

dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len)

{

	ASSERT(tx->tx_txg == 0);

	dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_FREE,

	    dmu_tx_hold_free_impl, off, len);

}

/* ARGSUSED */

static void

{

	uint64_t nblocks;

	dmu_tx_count_dnode(tx, dn);

	if (dn == NULL) {

		/*

		 * block.  So there will be at most 2 blocks total,

		 * including the header block.

		 */

		dmu_tx_count_write(tx, dn, 0, 2 << fzap_default_block_shift);

		return;

	}

	ASSERT3P(dmu_ot[dn->dn_type].ot_byteswap, ==, zap_byteswap);

		/*

		 * If there is only one block  (i.e. this is a micro-zap)

		 */