1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
23 */
24
25#include <sys/dmu.h>
26#include <sys/dmu_impl.h>
27#include <sys/dmu_tx.h>
28#include <sys/dbuf.h>
29#include <sys/dnode.h>
30#include <sys/zfs_context.h>
31#include <sys/dmu_objset.h>
32#include <sys/dmu_traverse.h>
33#include <sys/dsl_dataset.h>
34#include <sys/dsl_dir.h>
35#include <sys/dsl_pool.h>
36#include <sys/dsl_synctask.h>
37#include <sys/zfs_ioctl.h>
38#include <sys/zap.h>
39#include <sys/zio_checksum.h>
40#include <sys/zfs_znode.h>
41
42struct diffarg {
43	struct vnode *da_vp;		/* file to which we are reporting */
44	offset_t *da_offp;
45	int da_err;			/* error that stopped diff search */
46	dmu_diff_record_t da_ddr;
47};
48
49static int
50write_record(struct diffarg *da)
51{
52	ssize_t resid; /* have to get resid to get detailed errno */
53
54	if (da->da_ddr.ddr_type == DDR_NONE) {
55		da->da_err = 0;
56		return (0);
57	}
58
59	da->da_err = vn_rdwr(UIO_WRITE, da->da_vp, (caddr_t)&da->da_ddr,
60	    sizeof (da->da_ddr), 0, UIO_SYSSPACE, FAPPEND,
61	    RLIM64_INFINITY, CRED(), &resid);
62	*da->da_offp += sizeof (da->da_ddr);
63	return (da->da_err);
64}
65
66static int
67report_free_dnode_range(struct diffarg *da, uint64_t first, uint64_t last)
68{
69	ASSERT(first <= last);
70	if (da->da_ddr.ddr_type != DDR_FREE ||
71	    first != da->da_ddr.ddr_last + 1) {
72		if (write_record(da) != 0)
73			return (da->da_err);
74		da->da_ddr.ddr_type = DDR_FREE;
75		da->da_ddr.ddr_first = first;
76		da->da_ddr.ddr_last = last;
77		return (0);
78	}
79	da->da_ddr.ddr_last = last;
80	return (0);
81}
82
83static int
84report_dnode(struct diffarg *da, uint64_t object, dnode_phys_t *dnp)
85{
86	ASSERT(dnp != NULL);
87	if (dnp->dn_type == DMU_OT_NONE)
88		return (report_free_dnode_range(da, object, object));
89
90	if (da->da_ddr.ddr_type != DDR_INUSE ||
91	    object != da->da_ddr.ddr_last + 1) {
92		if (write_record(da) != 0)
93			return (da->da_err);
94		da->da_ddr.ddr_type = DDR_INUSE;
95		da->da_ddr.ddr_first = da->da_ddr.ddr_last = object;
96		return (0);
97	}
98	da->da_ddr.ddr_last = object;
99	return (0);
100}
101
102#define	DBP_SPAN(dnp, level)				  \
103	(((uint64_t)dnp->dn_datablkszsec) << (SPA_MINBLOCKSHIFT + \
104	(level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT)))
105
106/* ARGSUSED */
107static int
108diff_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, arc_buf_t *pbuf,
109    const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
110{
111	struct diffarg *da = arg;
112	int err = 0;
113
114	if (issig(JUSTLOOKING) && issig(FORREAL))
115		return (EINTR);
116
117	if (zb->zb_object != DMU_META_DNODE_OBJECT)
118		return (0);
119
120	if (bp == NULL) {
121		uint64_t span = DBP_SPAN(dnp, zb->zb_level);
122		uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT;
123
124		err = report_free_dnode_range(da, dnobj,
125		    dnobj + (span >> DNODE_SHIFT) - 1);
126		if (err)
127			return (err);
128	} else if (zb->zb_level == 0) {
129		dnode_phys_t *blk;
130		arc_buf_t *abuf;
131		uint32_t aflags = ARC_WAIT;
132		int blksz = BP_GET_LSIZE(bp);
133		int i;
134
135		if (dsl_read(NULL, spa, bp, pbuf,
136		    arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
137		    ZIO_FLAG_CANFAIL, &aflags, zb) != 0)
138			return (EIO);
139
140		blk = abuf->b_data;
141		for (i = 0; i < blksz >> DNODE_SHIFT; i++) {
142			uint64_t dnobj = (zb->zb_blkid <<
143			    (DNODE_BLOCK_SHIFT - DNODE_SHIFT)) + i;
144			err = report_dnode(da, dnobj, blk+i);
145			if (err)
146				break;
147		}
148		(void) arc_buf_remove_ref(abuf, &abuf);
149		if (err)
150			return (err);
151		/* Don't care about the data blocks */
152		return (TRAVERSE_VISIT_NO_CHILDREN);
153	}
154	return (0);
155}
156
157int
158dmu_diff(objset_t *tosnap, objset_t *fromsnap, struct vnode *vp, offset_t *offp)
159{
160	struct diffarg da;
161	dsl_dataset_t *ds = tosnap->os_dsl_dataset;
162	dsl_dataset_t *fromds = fromsnap->os_dsl_dataset;
163	dsl_dataset_t *findds;
164	dsl_dataset_t *relds;
165	int err = 0;
166
167	/* make certain we are looking at snapshots */
168	if (!dsl_dataset_is_snapshot(ds) || !dsl_dataset_is_snapshot(fromds))
169		return (EINVAL);
170
171	/* fromsnap must be earlier and from the same lineage as tosnap */
172	if (fromds->ds_phys->ds_creation_txg >= ds->ds_phys->ds_creation_txg)
173		return (EXDEV);
174
175	relds = NULL;
176	findds = ds;
177
178	while (fromds->ds_dir != findds->ds_dir) {
179		dsl_pool_t *dp = ds->ds_dir->dd_pool;
180
181		if (!dsl_dir_is_clone(findds->ds_dir)) {
182			if (relds)
183				dsl_dataset_rele(relds, FTAG);
184			return (EXDEV);
185		}
186
187		rw_enter(&dp->dp_config_rwlock, RW_READER);
188		err = dsl_dataset_hold_obj(dp,
189		    findds->ds_dir->dd_phys->dd_origin_obj, FTAG, &findds);
190		rw_exit(&dp->dp_config_rwlock);
191
192		if (relds)
193			dsl_dataset_rele(relds, FTAG);
194
195		if (err)
196			return (EXDEV);
197
198		relds = findds;
199	}
200
201	if (relds)
202		dsl_dataset_rele(relds, FTAG);
203
204	da.da_vp = vp;
205	da.da_offp = offp;
206	da.da_ddr.ddr_type = DDR_NONE;
207	da.da_ddr.ddr_first = da.da_ddr.ddr_last = 0;
208	da.da_err = 0;
209
210	err = traverse_dataset(ds, fromds->ds_phys->ds_creation_txg,
211	    TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, diff_cb, &da);
212
213	if (err) {
214		da.da_err = err;
215	} else {
216		/* we set the da.da_err we return as side-effect */
217		(void) write_record(&da);
218	}
219
220	return (da.da_err);
221}
222