]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 2008 Sun Microsystems, Inc. All rights reserved. | |
23 | * Use is subject to license terms. | |
24 | */ | |
25 | ||
34dc7c2f BB |
26 | #include <sys/types.h> |
27 | #include <sys/param.h> | |
28 | #include <sys/systm.h> | |
29 | #include <sys/sysmacros.h> | |
30 | #include <sys/cmn_err.h> | |
31 | #include <sys/kmem.h> | |
32 | #include <sys/thread.h> | |
33 | #include <sys/file.h> | |
34 | #include <sys/vfs.h> | |
35 | #include <sys/zfs_znode.h> | |
36 | #include <sys/zfs_dir.h> | |
37 | #include <sys/zil.h> | |
38 | #include <sys/zil_impl.h> | |
39 | #include <sys/byteorder.h> | |
40 | #include <sys/policy.h> | |
41 | #include <sys/stat.h> | |
42 | #include <sys/mode.h> | |
43 | #include <sys/acl.h> | |
44 | #include <sys/dmu.h> | |
45 | #include <sys/spa.h> | |
46 | #include <sys/zfs_fuid.h> | |
47 | #include <sys/ddi.h> | |
fb5f0bc8 BB |
48 | #include <sys/dsl_dataset.h> |
49 | ||
50 | #define ZFS_HANDLE_REPLAY(zilog, tx) \ | |
51 | if (zilog->zl_replay) { \ | |
52 | dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); \ | |
53 | zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] = \ | |
54 | zilog->zl_replaying_seq; \ | |
55 | return; \ | |
56 | } | |
34dc7c2f BB |
57 | |
58 | /* | |
fb5f0bc8 BB |
59 | * These zfs_log_* functions must be called within a dmu tx, in one |
60 | * of 2 contexts depending on zilog->z_replay: | |
61 | * | |
62 | * Non replay mode | |
63 | * --------------- | |
64 | * We need to record the transaction so that if it is committed to | |
65 | * the Intent Log then it can be replayed. An intent log transaction | |
66 | * structure (itx_t) is allocated and all the information necessary to | |
67 | * possibly replay the transaction is saved in it. The itx is then assigned | |
68 | * a sequence number and inserted in the in-memory list anchored in the zilog. | |
69 | * | |
70 | * Replay mode | |
71 | * ----------- | |
72 | * We need to mark the intent log record as replayed in the log header. | |
73 | * This is done in the same transaction as the replay so that they | |
74 | * commit atomically. | |
34dc7c2f BB |
75 | */ |
76 | ||
77 | int | |
78 | zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap) | |
79 | { | |
80 | int isxvattr = (vap->va_mask & AT_XVATTR); | |
81 | switch (type) { | |
82 | case Z_FILE: | |
83 | if (vsecp == NULL && !isxvattr) | |
84 | return (TX_CREATE); | |
85 | if (vsecp && isxvattr) | |
86 | return (TX_CREATE_ACL_ATTR); | |
87 | if (vsecp) | |
88 | return (TX_CREATE_ACL); | |
89 | else | |
90 | return (TX_CREATE_ATTR); | |
91 | /*NOTREACHED*/ | |
92 | case Z_DIR: | |
93 | if (vsecp == NULL && !isxvattr) | |
94 | return (TX_MKDIR); | |
95 | if (vsecp && isxvattr) | |
96 | return (TX_MKDIR_ACL_ATTR); | |
97 | if (vsecp) | |
98 | return (TX_MKDIR_ACL); | |
99 | else | |
100 | return (TX_MKDIR_ATTR); | |
101 | case Z_XATTRDIR: | |
102 | return (TX_MKXATTR); | |
103 | } | |
104 | ASSERT(0); | |
105 | return (TX_MAX_TYPE); | |
106 | } | |
107 | ||
108 | /* | |
109 | * build up the log data necessary for logging xvattr_t | |
110 | * First lr_attr_t is initialized. following the lr_attr_t | |
111 | * is the mapsize and attribute bitmap copied from the xvattr_t. | |
112 | * Following the bitmap and bitmapsize two 64 bit words are reserved | |
113 | * for the create time which may be set. Following the create time | |
114 | * records a single 64 bit integer which has the bits to set on | |
115 | * replay for the xvattr. | |
116 | */ | |
117 | static void | |
118 | zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap) | |
119 | { | |
120 | uint32_t *bitmap; | |
121 | uint64_t *attrs; | |
122 | uint64_t *crtime; | |
123 | xoptattr_t *xoap; | |
124 | void *scanstamp; | |
125 | int i; | |
126 | ||
127 | xoap = xva_getxoptattr(xvap); | |
128 | ASSERT(xoap); | |
129 | ||
130 | lrattr->lr_attr_masksize = xvap->xva_mapsize; | |
131 | bitmap = &lrattr->lr_attr_bitmap; | |
132 | for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) { | |
133 | *bitmap = xvap->xva_reqattrmap[i]; | |
134 | } | |
135 | ||
136 | /* Now pack the attributes up in a single uint64_t */ | |
137 | attrs = (uint64_t *)bitmap; | |
138 | crtime = attrs + 1; | |
139 | scanstamp = (caddr_t)(crtime + 2); | |
140 | *attrs = 0; | |
141 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) | |
142 | *attrs |= (xoap->xoa_readonly == 0) ? 0 : | |
143 | XAT0_READONLY; | |
144 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) | |
145 | *attrs |= (xoap->xoa_hidden == 0) ? 0 : | |
146 | XAT0_HIDDEN; | |
147 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) | |
148 | *attrs |= (xoap->xoa_system == 0) ? 0 : | |
149 | XAT0_SYSTEM; | |
150 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) | |
151 | *attrs |= (xoap->xoa_archive == 0) ? 0 : | |
152 | XAT0_ARCHIVE; | |
153 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) | |
154 | *attrs |= (xoap->xoa_immutable == 0) ? 0 : | |
155 | XAT0_IMMUTABLE; | |
156 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) | |
157 | *attrs |= (xoap->xoa_nounlink == 0) ? 0 : | |
158 | XAT0_NOUNLINK; | |
159 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) | |
160 | *attrs |= (xoap->xoa_appendonly == 0) ? 0 : | |
161 | XAT0_APPENDONLY; | |
162 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) | |
163 | *attrs |= (xoap->xoa_opaque == 0) ? 0 : | |
164 | XAT0_APPENDONLY; | |
165 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) | |
166 | *attrs |= (xoap->xoa_nodump == 0) ? 0 : | |
167 | XAT0_NODUMP; | |
168 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) | |
169 | *attrs |= (xoap->xoa_av_quarantined == 0) ? 0 : | |
170 | XAT0_AV_QUARANTINED; | |
171 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) | |
172 | *attrs |= (xoap->xoa_av_modified == 0) ? 0 : | |
173 | XAT0_AV_MODIFIED; | |
174 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) | |
175 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime); | |
176 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
177 | bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ); | |
178 | } | |
179 | ||
180 | static void * | |
181 | zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start) | |
182 | { | |
183 | zfs_fuid_t *zfuid; | |
184 | uint64_t *fuidloc = start; | |
185 | ||
186 | /* First copy in the ACE FUIDs */ | |
187 | for (zfuid = list_head(&fuidp->z_fuids); zfuid; | |
188 | zfuid = list_next(&fuidp->z_fuids, zfuid)) { | |
189 | *fuidloc++ = zfuid->z_logfuid; | |
190 | } | |
191 | return (fuidloc); | |
192 | } | |
193 | ||
194 | ||
195 | static void * | |
196 | zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start) | |
197 | { | |
198 | zfs_fuid_domain_t *zdomain; | |
199 | ||
200 | /* now copy in the domain info, if any */ | |
201 | if (fuidp->z_domain_str_sz != 0) { | |
202 | for (zdomain = list_head(&fuidp->z_domains); zdomain; | |
203 | zdomain = list_next(&fuidp->z_domains, zdomain)) { | |
204 | bcopy((void *)zdomain->z_domain, start, | |
205 | strlen(zdomain->z_domain) + 1); | |
206 | start = (caddr_t)start + | |
207 | strlen(zdomain->z_domain) + 1; | |
208 | } | |
209 | } | |
210 | return (start); | |
211 | } | |
212 | ||
213 | /* | |
214 | * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, | |
215 | * TX_MKDIR_ATTR and TX_MKXATTR | |
216 | * transactions. | |
217 | * | |
218 | * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID | |
219 | * domain information appended prior to the name. In this case the | |
220 | * uid/gid in the log record will be a log centric FUID. | |
221 | * | |
222 | * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that | |
223 | * may contain attributes, ACL and optional fuid information. | |
224 | * | |
225 | * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify | |
226 | * and ACL and normal users/groups in the ACEs. | |
227 | * | |
228 | * There may be an optional xvattr attribute information similar | |
229 | * to zfs_log_setattr. | |
230 | * | |
231 | * Also, after the file name "domain" strings may be appended. | |
232 | */ | |
233 | void | |
234 | zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, | |
235 | znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp, | |
236 | zfs_fuid_info_t *fuidp, vattr_t *vap) | |
237 | { | |
238 | itx_t *itx; | |
239 | uint64_t seq; | |
240 | lr_create_t *lr; | |
241 | lr_acl_create_t *lracl; | |
242 | size_t aclsize; | |
243 | size_t xvatsize = 0; | |
244 | size_t txsize; | |
245 | xvattr_t *xvap = (xvattr_t *)vap; | |
246 | void *end; | |
247 | size_t lrsize; | |
34dc7c2f BB |
248 | size_t namesize = strlen(name) + 1; |
249 | size_t fuidsz = 0; | |
250 | ||
251 | if (zilog == NULL) | |
252 | return; | |
253 | ||
fb5f0bc8 BB |
254 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
255 | ||
34dc7c2f BB |
256 | /* |
257 | * If we have FUIDs present then add in space for | |
258 | * domains and ACE fuid's if any. | |
259 | */ | |
260 | if (fuidp) { | |
261 | fuidsz += fuidp->z_domain_str_sz; | |
262 | fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t); | |
263 | } | |
264 | ||
265 | if (vap->va_mask & AT_XVATTR) | |
266 | xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize); | |
267 | ||
268 | if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR || | |
269 | (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR || | |
270 | (int)txtype == TX_MKXATTR) { | |
271 | txsize = sizeof (*lr) + namesize + fuidsz + xvatsize; | |
272 | lrsize = sizeof (*lr); | |
273 | } else { | |
274 | aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0; | |
275 | txsize = | |
276 | sizeof (lr_acl_create_t) + namesize + fuidsz + | |
277 | ZIL_ACE_LENGTH(aclsize) + xvatsize; | |
278 | lrsize = sizeof (lr_acl_create_t); | |
279 | } | |
280 | ||
281 | itx = zil_itx_create(txtype, txsize); | |
282 | ||
283 | lr = (lr_create_t *)&itx->itx_lr; | |
284 | lr->lr_doid = dzp->z_id; | |
285 | lr->lr_foid = zp->z_id; | |
286 | lr->lr_mode = zp->z_phys->zp_mode; | |
287 | if (!IS_EPHEMERAL(zp->z_phys->zp_uid)) { | |
288 | lr->lr_uid = (uint64_t)zp->z_phys->zp_uid; | |
289 | } else { | |
290 | lr->lr_uid = fuidp->z_fuid_owner; | |
291 | } | |
292 | if (!IS_EPHEMERAL(zp->z_phys->zp_gid)) { | |
293 | lr->lr_gid = (uint64_t)zp->z_phys->zp_gid; | |
294 | } else { | |
295 | lr->lr_gid = fuidp->z_fuid_group; | |
296 | } | |
297 | lr->lr_gen = zp->z_phys->zp_gen; | |
298 | lr->lr_crtime[0] = zp->z_phys->zp_crtime[0]; | |
299 | lr->lr_crtime[1] = zp->z_phys->zp_crtime[1]; | |
300 | lr->lr_rdev = zp->z_phys->zp_rdev; | |
301 | ||
302 | /* | |
303 | * Fill in xvattr info if any | |
304 | */ | |
305 | if (vap->va_mask & AT_XVATTR) { | |
306 | zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap); | |
307 | end = (caddr_t)lr + lrsize + xvatsize; | |
308 | } else { | |
309 | end = (caddr_t)lr + lrsize; | |
310 | } | |
311 | ||
312 | /* Now fill in any ACL info */ | |
313 | ||
314 | if (vsecp) { | |
315 | lracl = (lr_acl_create_t *)&itx->itx_lr; | |
316 | lracl->lr_aclcnt = vsecp->vsa_aclcnt; | |
317 | lracl->lr_acl_bytes = aclsize; | |
318 | lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0; | |
319 | lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0; | |
320 | if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS) | |
321 | lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags; | |
322 | else | |
323 | lracl->lr_acl_flags = 0; | |
324 | ||
325 | bcopy(vsecp->vsa_aclentp, end, aclsize); | |
326 | end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize); | |
327 | } | |
328 | ||
329 | /* drop in FUID info */ | |
330 | if (fuidp) { | |
331 | end = zfs_log_fuid_ids(fuidp, end); | |
332 | end = zfs_log_fuid_domains(fuidp, end); | |
333 | } | |
334 | /* | |
335 | * Now place file name in log record | |
336 | */ | |
337 | bcopy(name, end, namesize); | |
338 | ||
339 | seq = zil_itx_assign(zilog, itx, tx); | |
340 | dzp->z_last_itx = seq; | |
341 | zp->z_last_itx = seq; | |
342 | } | |
343 | ||
344 | /* | |
345 | * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions. | |
346 | */ | |
347 | void | |
348 | zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, | |
349 | znode_t *dzp, char *name) | |
350 | { | |
351 | itx_t *itx; | |
352 | uint64_t seq; | |
353 | lr_remove_t *lr; | |
354 | size_t namesize = strlen(name) + 1; | |
355 | ||
356 | if (zilog == NULL) | |
357 | return; | |
358 | ||
fb5f0bc8 BB |
359 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
360 | ||
34dc7c2f BB |
361 | itx = zil_itx_create(txtype, sizeof (*lr) + namesize); |
362 | lr = (lr_remove_t *)&itx->itx_lr; | |
363 | lr->lr_doid = dzp->z_id; | |
364 | bcopy(name, (char *)(lr + 1), namesize); | |
365 | ||
366 | seq = zil_itx_assign(zilog, itx, tx); | |
367 | dzp->z_last_itx = seq; | |
368 | } | |
369 | ||
370 | /* | |
371 | * zfs_log_link() handles TX_LINK transactions. | |
372 | */ | |
373 | void | |
374 | zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, | |
375 | znode_t *dzp, znode_t *zp, char *name) | |
376 | { | |
377 | itx_t *itx; | |
378 | uint64_t seq; | |
379 | lr_link_t *lr; | |
380 | size_t namesize = strlen(name) + 1; | |
381 | ||
382 | if (zilog == NULL) | |
383 | return; | |
384 | ||
fb5f0bc8 BB |
385 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
386 | ||
34dc7c2f BB |
387 | itx = zil_itx_create(txtype, sizeof (*lr) + namesize); |
388 | lr = (lr_link_t *)&itx->itx_lr; | |
389 | lr->lr_doid = dzp->z_id; | |
390 | lr->lr_link_obj = zp->z_id; | |
391 | bcopy(name, (char *)(lr + 1), namesize); | |
392 | ||
393 | seq = zil_itx_assign(zilog, itx, tx); | |
394 | dzp->z_last_itx = seq; | |
395 | zp->z_last_itx = seq; | |
396 | } | |
397 | ||
398 | /* | |
399 | * zfs_log_symlink() handles TX_SYMLINK transactions. | |
400 | */ | |
401 | void | |
402 | zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, | |
403 | znode_t *dzp, znode_t *zp, char *name, char *link) | |
404 | { | |
405 | itx_t *itx; | |
406 | uint64_t seq; | |
407 | lr_create_t *lr; | |
408 | size_t namesize = strlen(name) + 1; | |
409 | size_t linksize = strlen(link) + 1; | |
410 | ||
411 | if (zilog == NULL) | |
412 | return; | |
413 | ||
fb5f0bc8 BB |
414 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
415 | ||
34dc7c2f BB |
416 | itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize); |
417 | lr = (lr_create_t *)&itx->itx_lr; | |
418 | lr->lr_doid = dzp->z_id; | |
419 | lr->lr_foid = zp->z_id; | |
420 | lr->lr_mode = zp->z_phys->zp_mode; | |
421 | lr->lr_uid = zp->z_phys->zp_uid; | |
422 | lr->lr_gid = zp->z_phys->zp_gid; | |
423 | lr->lr_gen = zp->z_phys->zp_gen; | |
424 | lr->lr_crtime[0] = zp->z_phys->zp_crtime[0]; | |
425 | lr->lr_crtime[1] = zp->z_phys->zp_crtime[1]; | |
426 | bcopy(name, (char *)(lr + 1), namesize); | |
427 | bcopy(link, (char *)(lr + 1) + namesize, linksize); | |
428 | ||
429 | seq = zil_itx_assign(zilog, itx, tx); | |
430 | dzp->z_last_itx = seq; | |
431 | zp->z_last_itx = seq; | |
432 | } | |
433 | ||
434 | /* | |
435 | * zfs_log_rename() handles TX_RENAME transactions. | |
436 | */ | |
437 | void | |
438 | zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, | |
439 | znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp) | |
440 | { | |
441 | itx_t *itx; | |
442 | uint64_t seq; | |
443 | lr_rename_t *lr; | |
444 | size_t snamesize = strlen(sname) + 1; | |
445 | size_t dnamesize = strlen(dname) + 1; | |
446 | ||
447 | if (zilog == NULL) | |
448 | return; | |
449 | ||
fb5f0bc8 BB |
450 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
451 | ||
34dc7c2f BB |
452 | itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize); |
453 | lr = (lr_rename_t *)&itx->itx_lr; | |
454 | lr->lr_sdoid = sdzp->z_id; | |
455 | lr->lr_tdoid = tdzp->z_id; | |
456 | bcopy(sname, (char *)(lr + 1), snamesize); | |
457 | bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize); | |
458 | ||
459 | seq = zil_itx_assign(zilog, itx, tx); | |
460 | sdzp->z_last_itx = seq; | |
461 | tdzp->z_last_itx = seq; | |
462 | szp->z_last_itx = seq; | |
463 | } | |
464 | ||
465 | /* | |
466 | * zfs_log_write() handles TX_WRITE transactions. | |
467 | */ | |
468 | ssize_t zfs_immediate_write_sz = 32768; | |
469 | ||
470 | #define ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \ | |
471 | sizeof (lr_write_t)) | |
472 | ||
473 | void | |
474 | zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype, | |
475 | znode_t *zp, offset_t off, ssize_t resid, int ioflag) | |
476 | { | |
477 | itx_wr_state_t write_state; | |
478 | boolean_t slogging; | |
479 | uintptr_t fsync_cnt; | |
480 | ||
481 | if (zilog == NULL || zp->z_unlinked) | |
482 | return; | |
483 | ||
fb5f0bc8 BB |
484 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
485 | ||
34dc7c2f BB |
486 | /* |
487 | * Writes are handled in three different ways: | |
488 | * | |
489 | * WR_INDIRECT: | |
b128c09f BB |
490 | * In this mode, if we need to commit the write later, then the block |
491 | * is immediately written into the file system (using dmu_sync), | |
492 | * and a pointer to the block is put into the log record. | |
493 | * When the txg commits the block is linked in. | |
494 | * This saves additionally writing the data into the log record. | |
495 | * There are a few requirements for this to occur: | |
496 | * - write is greater than zfs_immediate_write_sz | |
497 | * - not using slogs (as slogs are assumed to always be faster | |
498 | * than writing into the main pool) | |
499 | * - the write occupies only one block | |
34dc7c2f BB |
500 | * WR_COPIED: |
501 | * If we know we'll immediately be committing the | |
502 | * transaction (FSYNC or FDSYNC), the we allocate a larger | |
503 | * log record here for the data and copy the data in. | |
504 | * WR_NEED_COPY: | |
505 | * Otherwise we don't allocate a buffer, and *if* we need to | |
506 | * flush the write later then a buffer is allocated and | |
507 | * we retrieve the data using the dmu. | |
508 | */ | |
509 | slogging = spa_has_slogs(zilog->zl_spa); | |
b128c09f | 510 | if (resid > zfs_immediate_write_sz && !slogging && resid <= zp->z_blksz) |
34dc7c2f BB |
511 | write_state = WR_INDIRECT; |
512 | else if (ioflag & (FSYNC | FDSYNC)) | |
513 | write_state = WR_COPIED; | |
514 | else | |
515 | write_state = WR_NEED_COPY; | |
516 | ||
517 | if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) { | |
518 | (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1)); | |
519 | } | |
520 | ||
521 | while (resid) { | |
522 | itx_t *itx; | |
523 | lr_write_t *lr; | |
524 | ssize_t len; | |
525 | ||
526 | /* | |
b128c09f | 527 | * If the write would overflow the largest block then split it. |
34dc7c2f | 528 | */ |
b128c09f | 529 | if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA) |
34dc7c2f BB |
530 | len = SPA_MAXBLOCKSIZE >> 1; |
531 | else | |
532 | len = resid; | |
533 | ||
534 | itx = zil_itx_create(txtype, sizeof (*lr) + | |
535 | (write_state == WR_COPIED ? len : 0)); | |
536 | lr = (lr_write_t *)&itx->itx_lr; | |
537 | if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os, | |
538 | zp->z_id, off, len, lr + 1) != 0) { | |
539 | kmem_free(itx, offsetof(itx_t, itx_lr) + | |
540 | itx->itx_lr.lrc_reclen); | |
541 | itx = zil_itx_create(txtype, sizeof (*lr)); | |
542 | lr = (lr_write_t *)&itx->itx_lr; | |
543 | write_state = WR_NEED_COPY; | |
544 | } | |
545 | ||
546 | itx->itx_wr_state = write_state; | |
547 | if (write_state == WR_NEED_COPY) | |
548 | itx->itx_sod += len; | |
549 | lr->lr_foid = zp->z_id; | |
550 | lr->lr_offset = off; | |
551 | lr->lr_length = len; | |
552 | lr->lr_blkoff = 0; | |
553 | BP_ZERO(&lr->lr_blkptr); | |
554 | ||
555 | itx->itx_private = zp->z_zfsvfs; | |
556 | ||
557 | if ((zp->z_sync_cnt != 0) || (fsync_cnt != 0) || | |
558 | (ioflag & (FSYNC | FDSYNC))) | |
559 | itx->itx_sync = B_TRUE; | |
560 | else | |
561 | itx->itx_sync = B_FALSE; | |
562 | ||
563 | zp->z_last_itx = zil_itx_assign(zilog, itx, tx); | |
564 | ||
565 | off += len; | |
566 | resid -= len; | |
567 | } | |
568 | } | |
569 | ||
570 | /* | |
571 | * zfs_log_truncate() handles TX_TRUNCATE transactions. | |
572 | */ | |
573 | void | |
574 | zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype, | |
575 | znode_t *zp, uint64_t off, uint64_t len) | |
576 | { | |
577 | itx_t *itx; | |
578 | uint64_t seq; | |
579 | lr_truncate_t *lr; | |
580 | ||
581 | if (zilog == NULL || zp->z_unlinked) | |
582 | return; | |
583 | ||
fb5f0bc8 BB |
584 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
585 | ||
34dc7c2f BB |
586 | itx = zil_itx_create(txtype, sizeof (*lr)); |
587 | lr = (lr_truncate_t *)&itx->itx_lr; | |
588 | lr->lr_foid = zp->z_id; | |
589 | lr->lr_offset = off; | |
590 | lr->lr_length = len; | |
591 | ||
592 | itx->itx_sync = (zp->z_sync_cnt != 0); | |
593 | seq = zil_itx_assign(zilog, itx, tx); | |
594 | zp->z_last_itx = seq; | |
595 | } | |
596 | ||
597 | /* | |
598 | * zfs_log_setattr() handles TX_SETATTR transactions. | |
599 | */ | |
600 | void | |
601 | zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype, | |
602 | znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp) | |
603 | { | |
604 | itx_t *itx; | |
605 | uint64_t seq; | |
606 | lr_setattr_t *lr; | |
607 | xvattr_t *xvap = (xvattr_t *)vap; | |
608 | size_t recsize = sizeof (lr_setattr_t); | |
609 | void *start; | |
610 | ||
611 | ||
612 | if (zilog == NULL || zp->z_unlinked) | |
613 | return; | |
614 | ||
fb5f0bc8 BB |
615 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
616 | ||
34dc7c2f BB |
617 | /* |
618 | * If XVATTR set, then log record size needs to allow | |
619 | * for lr_attr_t + xvattr mask, mapsize and create time | |
620 | * plus actual attribute values | |
621 | */ | |
622 | if (vap->va_mask & AT_XVATTR) | |
623 | recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize); | |
624 | ||
625 | if (fuidp) | |
626 | recsize += fuidp->z_domain_str_sz; | |
627 | ||
628 | itx = zil_itx_create(txtype, recsize); | |
629 | lr = (lr_setattr_t *)&itx->itx_lr; | |
630 | lr->lr_foid = zp->z_id; | |
631 | lr->lr_mask = (uint64_t)mask_applied; | |
632 | lr->lr_mode = (uint64_t)vap->va_mode; | |
633 | if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid)) | |
634 | lr->lr_uid = fuidp->z_fuid_owner; | |
635 | else | |
636 | lr->lr_uid = (uint64_t)vap->va_uid; | |
637 | ||
638 | if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid)) | |
639 | lr->lr_gid = fuidp->z_fuid_group; | |
640 | else | |
641 | lr->lr_gid = (uint64_t)vap->va_gid; | |
642 | ||
643 | lr->lr_size = (uint64_t)vap->va_size; | |
644 | ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime); | |
645 | ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime); | |
646 | start = (lr_setattr_t *)(lr + 1); | |
647 | if (vap->va_mask & AT_XVATTR) { | |
648 | zfs_log_xvattr((lr_attr_t *)start, xvap); | |
649 | start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize); | |
650 | } | |
651 | ||
652 | /* | |
653 | * Now stick on domain information if any on end | |
654 | */ | |
655 | ||
656 | if (fuidp) | |
657 | (void) zfs_log_fuid_domains(fuidp, start); | |
658 | ||
659 | itx->itx_sync = (zp->z_sync_cnt != 0); | |
660 | seq = zil_itx_assign(zilog, itx, tx); | |
661 | zp->z_last_itx = seq; | |
662 | } | |
663 | ||
664 | /* | |
665 | * zfs_log_acl() handles TX_ACL transactions. | |
666 | */ | |
667 | void | |
668 | zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp, | |
669 | vsecattr_t *vsecp, zfs_fuid_info_t *fuidp) | |
670 | { | |
671 | itx_t *itx; | |
672 | uint64_t seq; | |
673 | lr_acl_v0_t *lrv0; | |
674 | lr_acl_t *lr; | |
675 | int txtype; | |
676 | int lrsize; | |
677 | size_t txsize; | |
678 | size_t aclbytes = vsecp->vsa_aclentsz; | |
679 | ||
b128c09f BB |
680 | if (zilog == NULL || zp->z_unlinked) |
681 | return; | |
682 | ||
fb5f0bc8 BB |
683 | ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ |
684 | ||
b128c09f | 685 | txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ? |
34dc7c2f BB |
686 | TX_ACL_V0 : TX_ACL; |
687 | ||
688 | if (txtype == TX_ACL) | |
689 | lrsize = sizeof (*lr); | |
690 | else | |
691 | lrsize = sizeof (*lrv0); | |
692 | ||
34dc7c2f BB |
693 | txsize = lrsize + |
694 | ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) + | |
695 | (fuidp ? fuidp->z_domain_str_sz : 0) + | |
b128c09f | 696 | sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0); |
34dc7c2f BB |
697 | |
698 | itx = zil_itx_create(txtype, txsize); | |
699 | ||
700 | lr = (lr_acl_t *)&itx->itx_lr; | |
701 | lr->lr_foid = zp->z_id; | |
702 | if (txtype == TX_ACL) { | |
703 | lr->lr_acl_bytes = aclbytes; | |
704 | lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0; | |
705 | lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0; | |
706 | if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) | |
707 | lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags; | |
708 | else | |
709 | lr->lr_acl_flags = 0; | |
710 | } | |
711 | lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt; | |
712 | ||
713 | if (txtype == TX_ACL_V0) { | |
714 | lrv0 = (lr_acl_v0_t *)lr; | |
715 | bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes); | |
716 | } else { | |
717 | void *start = (ace_t *)(lr + 1); | |
718 | ||
719 | bcopy(vsecp->vsa_aclentp, start, aclbytes); | |
720 | ||
721 | start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes); | |
722 | ||
723 | if (fuidp) { | |
724 | start = zfs_log_fuid_ids(fuidp, start); | |
725 | (void) zfs_log_fuid_domains(fuidp, start); | |
726 | } | |
727 | } | |
728 | ||
729 | itx->itx_sync = (zp->z_sync_cnt != 0); | |
730 | seq = zil_itx_assign(zilog, itx, tx); | |
731 | zp->z_last_itx = seq; | |
732 | } |