xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_log.c (revision 74e20cfe817b82802b16fac8690dadcda76f54f5)
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 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysmacros.h>
32 #include <sys/cmn_err.h>
33 #include <sys/kmem.h>
34 #include <sys/thread.h>
35 #include <sys/file.h>
36 #include <sys/vfs.h>
37 #include <sys/zfs_znode.h>
38 #include <sys/zfs_dir.h>
39 #include <sys/zil.h>
40 #include <sys/byteorder.h>
41 #include <sys/policy.h>
42 #include <sys/stat.h>
43 #include <sys/mode.h>
44 #include <sys/acl.h>
45 #include <sys/dmu.h>
46 #include <sys/spa.h>
47 #include <sys/ddi.h>
48 
49 /*
50  * All the functions in this file are used to construct the log entries
51  * to record transactions. They allocate * a intent log transaction
52  * structure (itx_t) and save within it all the information necessary to
53  * possibly replay the transaction. The itx is then assigned a sequence
54  * number and inserted in the in-memory list anchored in the zilog.
55  */
56 
57 /*
58  * zfs_log_create() is used to handle TX_CREATE, TX_MKDIR and TX_MKXATTR
59  * transactions.
60  */
61 void
62 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, int txtype,
63 	znode_t *dzp, znode_t *zp, char *name)
64 {
65 	itx_t *itx;
66 	uint64_t seq;
67 	lr_create_t *lr;
68 	size_t namesize = strlen(name) + 1;
69 
70 	if (zilog == NULL)
71 		return;
72 
73 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
74 	lr = (lr_create_t *)&itx->itx_lr;
75 	lr->lr_doid = dzp->z_id;
76 	lr->lr_foid = zp->z_id;
77 	lr->lr_mode = zp->z_phys->zp_mode;
78 	lr->lr_uid = zp->z_phys->zp_uid;
79 	lr->lr_gid = zp->z_phys->zp_gid;
80 	lr->lr_gen = zp->z_phys->zp_gen;
81 	lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
82 	lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
83 	lr->lr_rdev = zp->z_phys->zp_rdev;
84 	bcopy(name, (char *)(lr + 1), namesize);
85 
86 	seq = zil_itx_assign(zilog, itx, tx);
87 	dzp->z_last_itx = seq;
88 	zp->z_last_itx = seq;
89 }
90 
91 /*
92  * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions.
93  */
94 void
95 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, int txtype,
96 	znode_t *dzp, char *name)
97 {
98 	itx_t *itx;
99 	uint64_t seq;
100 	lr_remove_t *lr;
101 	size_t namesize = strlen(name) + 1;
102 
103 	if (zilog == NULL)
104 		return;
105 
106 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
107 	lr = (lr_remove_t *)&itx->itx_lr;
108 	lr->lr_doid = dzp->z_id;
109 	bcopy(name, (char *)(lr + 1), namesize);
110 
111 	seq = zil_itx_assign(zilog, itx, tx);
112 	dzp->z_last_itx = seq;
113 }
114 
115 /*
116  * zfs_log_link() handles TX_LINK transactions.
117  */
118 void
119 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, int txtype,
120 	znode_t *dzp, znode_t *zp, char *name)
121 {
122 	itx_t *itx;
123 	uint64_t seq;
124 	lr_link_t *lr;
125 	size_t namesize = strlen(name) + 1;
126 
127 	if (zilog == NULL)
128 		return;
129 
130 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
131 	lr = (lr_link_t *)&itx->itx_lr;
132 	lr->lr_doid = dzp->z_id;
133 	lr->lr_link_obj = zp->z_id;
134 	bcopy(name, (char *)(lr + 1), namesize);
135 
136 	seq = zil_itx_assign(zilog, itx, tx);
137 	dzp->z_last_itx = seq;
138 	zp->z_last_itx = seq;
139 }
140 
141 /*
142  * zfs_log_symlink() handles TX_SYMLINK transactions.
143  */
144 void
145 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, int txtype,
146 	znode_t *dzp, znode_t *zp, char *name, char *link)
147 {
148 	itx_t *itx;
149 	uint64_t seq;
150 	lr_create_t *lr;
151 	size_t namesize = strlen(name) + 1;
152 	size_t linksize = strlen(link) + 1;
153 
154 	if (zilog == NULL)
155 		return;
156 
157 	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
158 	lr = (lr_create_t *)&itx->itx_lr;
159 	lr->lr_doid = dzp->z_id;
160 	lr->lr_foid = zp->z_id;
161 	lr->lr_mode = zp->z_phys->zp_mode;
162 	lr->lr_uid = zp->z_phys->zp_uid;
163 	lr->lr_gid = zp->z_phys->zp_gid;
164 	lr->lr_gen = zp->z_phys->zp_gen;
165 	lr->lr_crtime[0] = zp->z_phys->zp_crtime[0];
166 	lr->lr_crtime[1] = zp->z_phys->zp_crtime[1];
167 	bcopy(name, (char *)(lr + 1), namesize);
168 	bcopy(link, (char *)(lr + 1) + namesize, linksize);
169 
170 	seq = zil_itx_assign(zilog, itx, tx);
171 	dzp->z_last_itx = seq;
172 	zp->z_last_itx = seq;
173 }
174 
175 /*
176  * zfs_log_rename() handles TX_RENAME transactions.
177  */
178 void
179 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, int txtype,
180 	znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
181 {
182 	itx_t *itx;
183 	uint64_t seq;
184 	lr_rename_t *lr;
185 	size_t snamesize = strlen(sname) + 1;
186 	size_t dnamesize = strlen(dname) + 1;
187 
188 	if (zilog == NULL)
189 		return;
190 
191 	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
192 	lr = (lr_rename_t *)&itx->itx_lr;
193 	lr->lr_sdoid = sdzp->z_id;
194 	lr->lr_tdoid = tdzp->z_id;
195 	bcopy(sname, (char *)(lr + 1), snamesize);
196 	bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
197 
198 	seq = zil_itx_assign(zilog, itx, tx);
199 	sdzp->z_last_itx = seq;
200 	tdzp->z_last_itx = seq;
201 	szp->z_last_itx = seq;
202 }
203 
204 /*
205  * zfs_log_write() handles TX_WRITE transactions.
206  */
207 ssize_t zfs_immediate_write_sz = 32768;
208 
209 void
210 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
211 	znode_t *zp, offset_t off, ssize_t len, int ioflag, uio_t *uio)
212 {
213 	itx_t *itx;
214 	uint64_t seq;
215 	lr_write_t *lr;
216 	itx_wr_state_t write_state;
217 	size_t dlen;
218 	int err;
219 
220 	if (zilog == NULL || zp->z_reap)
221 		return;
222 
223 	/*
224 	 * Writes are handled in three different ways:
225 	 *
226 	 * WR_INDIRECT:
227 	 *    If the write is greater than zfs_immediate_write_sz then
228 	 *    later *if* we need to log the write then dmu_sync() is used
229 	 *    to immediately write the block and it's block pointer is put
230 	 *    in the log record.
231 	 * WR_COPIED:
232 	 *    If we know we'll immediately be committing the
233 	 *    transaction (FDSYNC (O_DSYNC)), the we allocate a larger
234 	 *    log record here for the data and copy the data in.
235 	 * WR_NEED_COPY:
236 	 *    Otherwise we don't allocate a buffer, and *if* we need to
237 	 *    flush the write later then a buffer is allocated and
238 	 *    we retrieve the data using the dmu.
239 	 */
240 	if (len > zfs_immediate_write_sz) {
241 		dlen = 0;
242 		write_state = WR_INDIRECT;
243 	} else if (ioflag & FDSYNC) {
244 		dlen = len;
245 		write_state = WR_COPIED;
246 	} else {
247 		dlen = 0;
248 		write_state = WR_NEED_COPY;
249 	}
250 	itx = zil_itx_create(txtype, sizeof (*lr) + dlen);
251 	if (write_state == WR_COPIED) {
252 		err = xcopyin(uio->uio_iov->iov_base - len,
253 		    (char *)itx + offsetof(itx_t, itx_lr) + sizeof (*lr), len);
254 		/*
255 		 * xcopyin shouldn't error as we've already successfully
256 		 * copied it to a dmu buffer. However if it does we'll get
257 		 * the data from the dmu later.
258 		 */
259 		if (err) {
260 			kmem_free(itx, offsetof(itx_t, itx_lr)
261 			    + itx->itx_lr.lrc_reclen);
262 			itx = zil_itx_create(txtype, sizeof (*lr));
263 			write_state = WR_NEED_COPY;
264 		}
265 	}
266 	itx->itx_wr_state = write_state;
267 	lr = (lr_write_t *)&itx->itx_lr;
268 	lr->lr_foid = zp->z_id;
269 	lr->lr_offset = off;
270 	lr->lr_length = len;
271 	lr->lr_blkoff = 0;
272 	BP_ZERO(&lr->lr_blkptr);
273 
274 	itx->itx_private = zp->z_zfsvfs;
275 
276 	itx->itx_sync = (zp->z_sync_cnt != 0);
277 	seq = zil_itx_assign(zilog, itx, tx);
278 	zp->z_last_itx = seq;
279 }
280 
281 /*
282  * zfs_log_truncate() handles TX_TRUNCATE transactions.
283  */
284 void
285 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
286 	znode_t *zp, uint64_t off, uint64_t len)
287 {
288 	itx_t *itx;
289 	uint64_t seq;
290 	lr_truncate_t *lr;
291 
292 	if (zilog == NULL || zp->z_reap)
293 		return;
294 
295 	itx = zil_itx_create(txtype, sizeof (*lr));
296 	lr = (lr_truncate_t *)&itx->itx_lr;
297 	lr->lr_foid = zp->z_id;
298 	lr->lr_offset = off;
299 	lr->lr_length = len;
300 
301 	itx->itx_sync = (zp->z_sync_cnt != 0);
302 	seq = zil_itx_assign(zilog, itx, tx);
303 	zp->z_last_itx = seq;
304 }
305 
306 /*
307  * zfs_log_setattr() handles TX_SETATTR transactions.
308  */
309 void
310 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
311 	znode_t *zp, vattr_t *vap, uint_t mask_applied)
312 {
313 	itx_t *itx;
314 	uint64_t seq;
315 	lr_setattr_t *lr;
316 
317 	if (zilog == NULL || zp->z_reap)
318 		return;
319 
320 	itx = zil_itx_create(txtype, sizeof (*lr));
321 	lr = (lr_setattr_t *)&itx->itx_lr;
322 	lr->lr_foid = zp->z_id;
323 	lr->lr_mask = (uint64_t)mask_applied;
324 	lr->lr_mode = (uint64_t)vap->va_mode;
325 	lr->lr_uid = (uint64_t)vap->va_uid;
326 	lr->lr_gid = (uint64_t)vap->va_gid;
327 	lr->lr_size = (uint64_t)vap->va_size;
328 	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
329 	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
330 
331 	itx->itx_sync = (zp->z_sync_cnt != 0);
332 	seq = zil_itx_assign(zilog, itx, tx);
333 	zp->z_last_itx = seq;
334 }
335 
336 /*
337  * zfs_log_acl() handles TX_ACL transactions.
338  */
339 void
340 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, int txtype,
341 	znode_t *zp, int aclcnt, ace_t *z_ace)
342 {
343 	itx_t *itx;
344 	uint64_t seq;
345 	lr_acl_t *lr;
346 
347 	if (zilog == NULL || zp->z_reap)
348 		return;
349 
350 	itx = zil_itx_create(txtype, sizeof (*lr) + aclcnt * sizeof (ace_t));
351 	lr = (lr_acl_t *)&itx->itx_lr;
352 	lr->lr_foid = zp->z_id;
353 	lr->lr_aclcnt = (uint64_t)aclcnt;
354 	bcopy(z_ace, (ace_t *)(lr + 1), aclcnt * sizeof (ace_t));
355 
356 	itx->itx_sync = (zp->z_sync_cnt != 0);
357 	seq = zil_itx_assign(zilog, itx, tx);
358 	zp->z_last_itx = seq;
359 }
360