zfs_log.c revision 209962
1168404Spjd/* 2168404Spjd * CDDL HEADER START 3168404Spjd * 4168404Spjd * The contents of this file are subject to the terms of the 5168404Spjd * Common Development and Distribution License (the "License"). 6168404Spjd * You may not use this file except in compliance with the License. 7168404Spjd * 8168404Spjd * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9168404Spjd * or http://www.opensolaris.org/os/licensing. 10168404Spjd * See the License for the specific language governing permissions 11168404Spjd * and limitations under the License. 12168404Spjd * 13168404Spjd * When distributing Covered Code, include this CDDL HEADER in each 14168404Spjd * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15168404Spjd * If applicable, add the following below this CDDL HEADER, with the 16168404Spjd * fields enclosed by brackets "[]" replaced with your own identifying 17168404Spjd * information: Portions Copyright [yyyy] [name of copyright owner] 18168404Spjd * 19168404Spjd * CDDL HEADER END 20168404Spjd */ 21168404Spjd/* 22185029Spjd * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23168404Spjd * Use is subject to license terms. 24168404Spjd */ 25168404Spjd 26168404Spjd#include <sys/types.h> 27168404Spjd#include <sys/param.h> 28168404Spjd#include <sys/systm.h> 29168404Spjd#include <sys/sysmacros.h> 30168404Spjd#include <sys/cmn_err.h> 31168404Spjd#include <sys/kmem.h> 32168404Spjd#include <sys/file.h> 33168404Spjd#include <sys/vfs.h> 34168404Spjd#include <sys/zfs_znode.h> 35168404Spjd#include <sys/zfs_dir.h> 36168404Spjd#include <sys/zil.h> 37185029Spjd#include <sys/zil_impl.h> 38168404Spjd#include <sys/byteorder.h> 39168962Spjd#include <sys/policy.h> 40168404Spjd#include <sys/stat.h> 41168404Spjd#include <sys/acl.h> 42168404Spjd#include <sys/dmu.h> 43168404Spjd#include <sys/spa.h> 44185029Spjd#include <sys/zfs_fuid.h> 45209962Smm#include <sys/dsl_dataset.h> 46168404Spjd 47209962Smm#define ZFS_HANDLE_REPLAY(zilog, tx) \ 48209962Smm if (zilog->zl_replay) { \ 49209962Smm dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); \ 50209962Smm zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] = \ 51209962Smm zilog->zl_replaying_seq; \ 52209962Smm return; \ 53209962Smm } 54209962Smm 55168404Spjd/* 56209962Smm * These zfs_log_* functions must be called within a dmu tx, in one 57209962Smm * of 2 contexts depending on zilog->z_replay: 58209962Smm * 59209962Smm * Non replay mode 60209962Smm * --------------- 61209962Smm * We need to record the transaction so that if it is committed to 62209962Smm * the Intent Log then it can be replayed. An intent log transaction 63209962Smm * structure (itx_t) is allocated and all the information necessary to 64209962Smm * possibly replay the transaction is saved in it. The itx is then assigned 65209962Smm * a sequence number and inserted in the in-memory list anchored in the zilog. 66209962Smm * 67209962Smm * Replay mode 68209962Smm * ----------- 69209962Smm * We need to mark the intent log record as replayed in the log header. 70209962Smm * This is done in the same transaction as the replay so that they 71209962Smm * commit atomically. 72168404Spjd */ 73168404Spjd 74185029Spjdint 75185029Spjdzfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap) 76185029Spjd{ 77185029Spjd int isxvattr = (vap->va_mask & AT_XVATTR); 78185029Spjd switch (type) { 79185029Spjd case Z_FILE: 80185029Spjd if (vsecp == NULL && !isxvattr) 81185029Spjd return (TX_CREATE); 82185029Spjd if (vsecp && isxvattr) 83185321Strasz#ifdef TODO 84185029Spjd return (TX_CREATE_ACL_ATTR); 85185321Strasz#else 86185321Strasz panic("%s:%u: unsupported condition", __func__, __LINE__); 87185321Strasz#endif 88185029Spjd if (vsecp) 89185029Spjd return (TX_CREATE_ACL); 90185029Spjd else 91185029Spjd return (TX_CREATE_ATTR); 92185029Spjd /*NOTREACHED*/ 93185029Spjd case Z_DIR: 94185029Spjd if (vsecp == NULL && !isxvattr) 95185029Spjd return (TX_MKDIR); 96185029Spjd if (vsecp && isxvattr) 97185321Strasz#ifdef TODO 98185029Spjd return (TX_MKDIR_ACL_ATTR); 99185321Strasz#else 100185321Strasz panic("%s:%u: unsupported condition", __func__, __LINE__); 101185321Strasz#endif 102185029Spjd if (vsecp) 103185029Spjd return (TX_MKDIR_ACL); 104185029Spjd else 105185029Spjd return (TX_MKDIR_ATTR); 106185029Spjd case Z_XATTRDIR: 107185029Spjd return (TX_MKXATTR); 108185029Spjd } 109185029Spjd ASSERT(0); 110185029Spjd return (TX_MAX_TYPE); 111185029Spjd} 112185029Spjd 113168404Spjd/* 114185029Spjd * build up the log data necessary for logging xvattr_t 115185029Spjd * First lr_attr_t is initialized. following the lr_attr_t 116185029Spjd * is the mapsize and attribute bitmap copied from the xvattr_t. 117185029Spjd * Following the bitmap and bitmapsize two 64 bit words are reserved 118185029Spjd * for the create time which may be set. Following the create time 119185029Spjd * records a single 64 bit integer which has the bits to set on 120185029Spjd * replay for the xvattr. 121185029Spjd */ 122185029Spjdstatic void 123185029Spjdzfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap) 124185029Spjd{ 125185029Spjd uint32_t *bitmap; 126185029Spjd uint64_t *attrs; 127185029Spjd uint64_t *crtime; 128185029Spjd xoptattr_t *xoap; 129185029Spjd void *scanstamp; 130185029Spjd int i; 131185029Spjd 132185029Spjd xoap = xva_getxoptattr(xvap); 133185029Spjd ASSERT(xoap); 134185029Spjd 135185029Spjd lrattr->lr_attr_masksize = xvap->xva_mapsize; 136185029Spjd bitmap = &lrattr->lr_attr_bitmap; 137185029Spjd for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) { 138185029Spjd *bitmap = xvap->xva_reqattrmap[i]; 139185029Spjd } 140185029Spjd 141185029Spjd /* Now pack the attributes up in a single uint64_t */ 142185029Spjd attrs = (uint64_t *)bitmap; 143185029Spjd crtime = attrs + 1; 144185029Spjd scanstamp = (caddr_t)(crtime + 2); 145185029Spjd *attrs = 0; 146185029Spjd if (XVA_ISSET_REQ(xvap, XAT_READONLY)) 147185029Spjd *attrs |= (xoap->xoa_readonly == 0) ? 0 : 148185029Spjd XAT0_READONLY; 149185029Spjd if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) 150185029Spjd *attrs |= (xoap->xoa_hidden == 0) ? 0 : 151185029Spjd XAT0_HIDDEN; 152185029Spjd if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) 153185029Spjd *attrs |= (xoap->xoa_system == 0) ? 0 : 154185029Spjd XAT0_SYSTEM; 155185029Spjd if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) 156185029Spjd *attrs |= (xoap->xoa_archive == 0) ? 0 : 157185029Spjd XAT0_ARCHIVE; 158185029Spjd if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) 159185029Spjd *attrs |= (xoap->xoa_immutable == 0) ? 0 : 160185029Spjd XAT0_IMMUTABLE; 161185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) 162185029Spjd *attrs |= (xoap->xoa_nounlink == 0) ? 0 : 163185029Spjd XAT0_NOUNLINK; 164185029Spjd if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) 165185029Spjd *attrs |= (xoap->xoa_appendonly == 0) ? 0 : 166185029Spjd XAT0_APPENDONLY; 167185029Spjd if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) 168185029Spjd *attrs |= (xoap->xoa_opaque == 0) ? 0 : 169185029Spjd XAT0_APPENDONLY; 170185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) 171185029Spjd *attrs |= (xoap->xoa_nodump == 0) ? 0 : 172185029Spjd XAT0_NODUMP; 173185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) 174185029Spjd *attrs |= (xoap->xoa_av_quarantined == 0) ? 0 : 175185029Spjd XAT0_AV_QUARANTINED; 176185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) 177185029Spjd *attrs |= (xoap->xoa_av_modified == 0) ? 0 : 178185029Spjd XAT0_AV_MODIFIED; 179185029Spjd if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) 180185029Spjd ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime); 181185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) 182185029Spjd bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ); 183185029Spjd} 184185029Spjd 185185029Spjdstatic void * 186185029Spjdzfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start) 187185029Spjd{ 188185029Spjd zfs_fuid_t *zfuid; 189185029Spjd uint64_t *fuidloc = start; 190185029Spjd 191185029Spjd /* First copy in the ACE FUIDs */ 192185029Spjd for (zfuid = list_head(&fuidp->z_fuids); zfuid; 193185029Spjd zfuid = list_next(&fuidp->z_fuids, zfuid)) { 194185029Spjd *fuidloc++ = zfuid->z_logfuid; 195185029Spjd } 196185029Spjd return (fuidloc); 197185029Spjd} 198185029Spjd 199185029Spjd 200185029Spjdstatic void * 201185029Spjdzfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start) 202185029Spjd{ 203185029Spjd zfs_fuid_domain_t *zdomain; 204185029Spjd 205185029Spjd /* now copy in the domain info, if any */ 206185029Spjd if (fuidp->z_domain_str_sz != 0) { 207185029Spjd for (zdomain = list_head(&fuidp->z_domains); zdomain; 208185029Spjd zdomain = list_next(&fuidp->z_domains, zdomain)) { 209185029Spjd bcopy((void *)zdomain->z_domain, start, 210185029Spjd strlen(zdomain->z_domain) + 1); 211185029Spjd start = (caddr_t)start + 212185029Spjd strlen(zdomain->z_domain) + 1; 213185029Spjd } 214185029Spjd } 215185029Spjd return (start); 216185029Spjd} 217185029Spjd 218185029Spjd/* 219185029Spjd * zfs_log_create() is used to handle TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, 220185029Spjd * TX_MKDIR_ATTR and TX_MKXATTR 221168404Spjd * transactions. 222185029Spjd * 223185029Spjd * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID 224185029Spjd * domain information appended prior to the name. In this case the 225185029Spjd * uid/gid in the log record will be a log centric FUID. 226185029Spjd * 227185029Spjd * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that 228185029Spjd * may contain attributes, ACL and optional fuid information. 229185029Spjd * 230185029Spjd * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify 231185029Spjd * and ACL and normal users/groups in the ACEs. 232185029Spjd * 233185029Spjd * There may be an optional xvattr attribute information similar 234185029Spjd * to zfs_log_setattr. 235185029Spjd * 236185029Spjd * Also, after the file name "domain" strings may be appended. 237168404Spjd */ 238168404Spjdvoid 239185029Spjdzfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, 240185029Spjd znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp, 241185029Spjd zfs_fuid_info_t *fuidp, vattr_t *vap) 242168404Spjd{ 243168404Spjd itx_t *itx; 244168404Spjd uint64_t seq; 245168404Spjd lr_create_t *lr; 246185029Spjd lr_acl_create_t *lracl; 247185029Spjd size_t aclsize; 248185029Spjd size_t xvatsize = 0; 249185029Spjd size_t txsize; 250185029Spjd xvattr_t *xvap = (xvattr_t *)vap; 251185029Spjd void *end; 252185029Spjd size_t lrsize; 253168404Spjd size_t namesize = strlen(name) + 1; 254185029Spjd size_t fuidsz = 0; 255168404Spjd 256168404Spjd if (zilog == NULL) 257168404Spjd return; 258168404Spjd 259209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 260209962Smm 261185029Spjd /* 262185029Spjd * If we have FUIDs present then add in space for 263185029Spjd * domains and ACE fuid's if any. 264185029Spjd */ 265185029Spjd if (fuidp) { 266185029Spjd fuidsz += fuidp->z_domain_str_sz; 267185029Spjd fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t); 268185029Spjd } 269185029Spjd 270185029Spjd if (vap->va_mask & AT_XVATTR) 271185029Spjd xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize); 272185029Spjd 273185029Spjd if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR || 274185029Spjd (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR || 275185029Spjd (int)txtype == TX_MKXATTR) { 276185029Spjd txsize = sizeof (*lr) + namesize + fuidsz + xvatsize; 277185029Spjd lrsize = sizeof (*lr); 278185029Spjd } else { 279185029Spjd aclsize = (vsecp) ? vsecp->vsa_aclentsz : 0; 280185029Spjd txsize = 281185029Spjd sizeof (lr_acl_create_t) + namesize + fuidsz + 282185029Spjd ZIL_ACE_LENGTH(aclsize) + xvatsize; 283185029Spjd lrsize = sizeof (lr_acl_create_t); 284185029Spjd } 285185029Spjd 286185029Spjd itx = zil_itx_create(txtype, txsize); 287185029Spjd 288168404Spjd lr = (lr_create_t *)&itx->itx_lr; 289168404Spjd lr->lr_doid = dzp->z_id; 290168404Spjd lr->lr_foid = zp->z_id; 291168404Spjd lr->lr_mode = zp->z_phys->zp_mode; 292185029Spjd if (!IS_EPHEMERAL(zp->z_phys->zp_uid)) { 293185029Spjd lr->lr_uid = (uint64_t)zp->z_phys->zp_uid; 294185029Spjd } else { 295185029Spjd lr->lr_uid = fuidp->z_fuid_owner; 296185029Spjd } 297185029Spjd if (!IS_EPHEMERAL(zp->z_phys->zp_gid)) { 298185029Spjd lr->lr_gid = (uint64_t)zp->z_phys->zp_gid; 299185029Spjd } else { 300185029Spjd lr->lr_gid = fuidp->z_fuid_group; 301185029Spjd } 302168404Spjd lr->lr_gen = zp->z_phys->zp_gen; 303168404Spjd lr->lr_crtime[0] = zp->z_phys->zp_crtime[0]; 304168404Spjd lr->lr_crtime[1] = zp->z_phys->zp_crtime[1]; 305168404Spjd lr->lr_rdev = zp->z_phys->zp_rdev; 306168404Spjd 307185029Spjd /* 308185029Spjd * Fill in xvattr info if any 309185029Spjd */ 310185029Spjd if (vap->va_mask & AT_XVATTR) { 311185029Spjd zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap); 312185029Spjd end = (caddr_t)lr + lrsize + xvatsize; 313185029Spjd } else { 314185029Spjd end = (caddr_t)lr + lrsize; 315185029Spjd } 316185029Spjd 317185029Spjd /* Now fill in any ACL info */ 318185029Spjd 319185029Spjd if (vsecp) { 320185029Spjd lracl = (lr_acl_create_t *)&itx->itx_lr; 321185029Spjd lracl->lr_aclcnt = vsecp->vsa_aclcnt; 322185029Spjd lracl->lr_acl_bytes = aclsize; 323185029Spjd lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0; 324185029Spjd lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0; 325185029Spjd if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS) 326185029Spjd lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags; 327185029Spjd else 328185029Spjd lracl->lr_acl_flags = 0; 329185029Spjd 330185029Spjd bcopy(vsecp->vsa_aclentp, end, aclsize); 331185029Spjd end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize); 332185029Spjd } 333185029Spjd 334185029Spjd /* drop in FUID info */ 335185029Spjd if (fuidp) { 336185029Spjd end = zfs_log_fuid_ids(fuidp, end); 337185029Spjd end = zfs_log_fuid_domains(fuidp, end); 338185029Spjd } 339185029Spjd /* 340185029Spjd * Now place file name in log record 341185029Spjd */ 342185029Spjd bcopy(name, end, namesize); 343185029Spjd 344168404Spjd seq = zil_itx_assign(zilog, itx, tx); 345168404Spjd dzp->z_last_itx = seq; 346168404Spjd zp->z_last_itx = seq; 347168404Spjd} 348168404Spjd 349168404Spjd/* 350168404Spjd * zfs_log_remove() handles both TX_REMOVE and TX_RMDIR transactions. 351168404Spjd */ 352168404Spjdvoid 353185029Spjdzfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, 354168404Spjd znode_t *dzp, char *name) 355168404Spjd{ 356168404Spjd itx_t *itx; 357168404Spjd uint64_t seq; 358168404Spjd lr_remove_t *lr; 359168404Spjd size_t namesize = strlen(name) + 1; 360168404Spjd 361168404Spjd if (zilog == NULL) 362168404Spjd return; 363168404Spjd 364209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 365209962Smm 366168404Spjd itx = zil_itx_create(txtype, sizeof (*lr) + namesize); 367168404Spjd lr = (lr_remove_t *)&itx->itx_lr; 368168404Spjd lr->lr_doid = dzp->z_id; 369168404Spjd bcopy(name, (char *)(lr + 1), namesize); 370168404Spjd 371168404Spjd seq = zil_itx_assign(zilog, itx, tx); 372168404Spjd dzp->z_last_itx = seq; 373168404Spjd} 374168404Spjd 375168404Spjd/* 376168404Spjd * zfs_log_link() handles TX_LINK transactions. 377168404Spjd */ 378168404Spjdvoid 379185029Spjdzfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, 380168404Spjd znode_t *dzp, znode_t *zp, char *name) 381168404Spjd{ 382168404Spjd itx_t *itx; 383168404Spjd uint64_t seq; 384168404Spjd lr_link_t *lr; 385168404Spjd size_t namesize = strlen(name) + 1; 386168404Spjd 387168404Spjd if (zilog == NULL) 388168404Spjd return; 389168404Spjd 390209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 391209962Smm 392168404Spjd itx = zil_itx_create(txtype, sizeof (*lr) + namesize); 393168404Spjd lr = (lr_link_t *)&itx->itx_lr; 394168404Spjd lr->lr_doid = dzp->z_id; 395168404Spjd lr->lr_link_obj = zp->z_id; 396168404Spjd bcopy(name, (char *)(lr + 1), namesize); 397168404Spjd 398168404Spjd seq = zil_itx_assign(zilog, itx, tx); 399168404Spjd dzp->z_last_itx = seq; 400168404Spjd zp->z_last_itx = seq; 401168404Spjd} 402168404Spjd 403168404Spjd/* 404168404Spjd * zfs_log_symlink() handles TX_SYMLINK transactions. 405168404Spjd */ 406168404Spjdvoid 407185029Spjdzfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, 408185029Spjd znode_t *dzp, znode_t *zp, char *name, char *link) 409168404Spjd{ 410168404Spjd itx_t *itx; 411168404Spjd uint64_t seq; 412168404Spjd lr_create_t *lr; 413168404Spjd size_t namesize = strlen(name) + 1; 414168404Spjd size_t linksize = strlen(link) + 1; 415168404Spjd 416168404Spjd if (zilog == NULL) 417168404Spjd return; 418168404Spjd 419209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 420209962Smm 421168404Spjd itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize); 422168404Spjd lr = (lr_create_t *)&itx->itx_lr; 423168404Spjd lr->lr_doid = dzp->z_id; 424168404Spjd lr->lr_foid = zp->z_id; 425168404Spjd lr->lr_mode = zp->z_phys->zp_mode; 426168404Spjd lr->lr_uid = zp->z_phys->zp_uid; 427168404Spjd lr->lr_gid = zp->z_phys->zp_gid; 428168404Spjd lr->lr_gen = zp->z_phys->zp_gen; 429168404Spjd lr->lr_crtime[0] = zp->z_phys->zp_crtime[0]; 430168404Spjd lr->lr_crtime[1] = zp->z_phys->zp_crtime[1]; 431168404Spjd bcopy(name, (char *)(lr + 1), namesize); 432168404Spjd bcopy(link, (char *)(lr + 1) + namesize, linksize); 433168404Spjd 434168404Spjd seq = zil_itx_assign(zilog, itx, tx); 435168404Spjd dzp->z_last_itx = seq; 436168404Spjd zp->z_last_itx = seq; 437168404Spjd} 438168404Spjd 439168404Spjd/* 440168404Spjd * zfs_log_rename() handles TX_RENAME transactions. 441168404Spjd */ 442168404Spjdvoid 443185029Spjdzfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, 444168404Spjd znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp) 445168404Spjd{ 446168404Spjd itx_t *itx; 447168404Spjd uint64_t seq; 448168404Spjd lr_rename_t *lr; 449168404Spjd size_t snamesize = strlen(sname) + 1; 450168404Spjd size_t dnamesize = strlen(dname) + 1; 451168404Spjd 452168404Spjd if (zilog == NULL) 453168404Spjd return; 454168404Spjd 455209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 456209962Smm 457168404Spjd itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize); 458168404Spjd lr = (lr_rename_t *)&itx->itx_lr; 459168404Spjd lr->lr_sdoid = sdzp->z_id; 460168404Spjd lr->lr_tdoid = tdzp->z_id; 461168404Spjd bcopy(sname, (char *)(lr + 1), snamesize); 462168404Spjd bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize); 463168404Spjd 464168404Spjd seq = zil_itx_assign(zilog, itx, tx); 465168404Spjd sdzp->z_last_itx = seq; 466168404Spjd tdzp->z_last_itx = seq; 467168404Spjd szp->z_last_itx = seq; 468168404Spjd} 469168404Spjd 470168404Spjd/* 471168404Spjd * zfs_log_write() handles TX_WRITE transactions. 472168404Spjd */ 473168404Spjdssize_t zfs_immediate_write_sz = 32768; 474168404Spjd 475185029Spjd#define ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_trailer_t) - \ 476185029Spjd sizeof (lr_write_t)) 477185029Spjd 478168404Spjdvoid 479168404Spjdzfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype, 480185029Spjd znode_t *zp, offset_t off, ssize_t resid, int ioflag) 481168404Spjd{ 482168404Spjd itx_wr_state_t write_state; 483185029Spjd boolean_t slogging; 484185029Spjd uintptr_t fsync_cnt; 485168404Spjd 486168404Spjd if (zilog == NULL || zp->z_unlinked) 487168404Spjd return; 488168404Spjd 489209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 490209962Smm 491168404Spjd /* 492168404Spjd * Writes are handled in three different ways: 493168404Spjd * 494168404Spjd * WR_INDIRECT: 495185029Spjd * In this mode, if we need to commit the write later, then the block 496185029Spjd * is immediately written into the file system (using dmu_sync), 497185029Spjd * and a pointer to the block is put into the log record. 498185029Spjd * When the txg commits the block is linked in. 499185029Spjd * This saves additionally writing the data into the log record. 500185029Spjd * There are a few requirements for this to occur: 501185029Spjd * - write is greater than zfs_immediate_write_sz 502185029Spjd * - not using slogs (as slogs are assumed to always be faster 503185029Spjd * than writing into the main pool) 504185029Spjd * - the write occupies only one block 505168404Spjd * WR_COPIED: 506168404Spjd * If we know we'll immediately be committing the 507185029Spjd * transaction (FSYNC or FDSYNC), the we allocate a larger 508168404Spjd * log record here for the data and copy the data in. 509168404Spjd * WR_NEED_COPY: 510168404Spjd * Otherwise we don't allocate a buffer, and *if* we need to 511168404Spjd * flush the write later then a buffer is allocated and 512168404Spjd * we retrieve the data using the dmu. 513168404Spjd */ 514185029Spjd slogging = spa_has_slogs(zilog->zl_spa); 515185029Spjd if (resid > zfs_immediate_write_sz && !slogging && resid <= zp->z_blksz) 516168404Spjd write_state = WR_INDIRECT; 517185029Spjd else if (ioflag & (FSYNC | FDSYNC)) 518168404Spjd write_state = WR_COPIED; 519168404Spjd else 520168404Spjd write_state = WR_NEED_COPY; 521168404Spjd 522185029Spjd if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) { 523185029Spjd (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1)); 524185029Spjd } 525185029Spjd 526185029Spjd while (resid) { 527185029Spjd itx_t *itx; 528185029Spjd lr_write_t *lr; 529185029Spjd ssize_t len; 530185029Spjd 531185029Spjd /* 532185029Spjd * If the write would overflow the largest block then split it. 533185029Spjd */ 534185029Spjd if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA) 535185029Spjd len = SPA_MAXBLOCKSIZE >> 1; 536185029Spjd else 537185029Spjd len = resid; 538185029Spjd 539185029Spjd itx = zil_itx_create(txtype, sizeof (*lr) + 540185029Spjd (write_state == WR_COPIED ? len : 0)); 541185029Spjd lr = (lr_write_t *)&itx->itx_lr; 542185029Spjd if (write_state == WR_COPIED && dmu_read(zp->z_zfsvfs->z_os, 543209962Smm zp->z_id, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) { 544168404Spjd kmem_free(itx, offsetof(itx_t, itx_lr) + 545168404Spjd itx->itx_lr.lrc_reclen); 546168404Spjd itx = zil_itx_create(txtype, sizeof (*lr)); 547168404Spjd lr = (lr_write_t *)&itx->itx_lr; 548168404Spjd write_state = WR_NEED_COPY; 549168404Spjd } 550168404Spjd 551185029Spjd itx->itx_wr_state = write_state; 552185029Spjd if (write_state == WR_NEED_COPY) 553185029Spjd itx->itx_sod += len; 554185029Spjd lr->lr_foid = zp->z_id; 555185029Spjd lr->lr_offset = off; 556185029Spjd lr->lr_length = len; 557185029Spjd lr->lr_blkoff = 0; 558185029Spjd BP_ZERO(&lr->lr_blkptr); 559168404Spjd 560185029Spjd itx->itx_private = zp->z_zfsvfs; 561168404Spjd 562185029Spjd if ((zp->z_sync_cnt != 0) || (fsync_cnt != 0) || 563185029Spjd (ioflag & (FSYNC | FDSYNC))) 564185029Spjd itx->itx_sync = B_TRUE; 565185029Spjd else 566185029Spjd itx->itx_sync = B_FALSE; 567185029Spjd 568185029Spjd zp->z_last_itx = zil_itx_assign(zilog, itx, tx); 569185029Spjd 570185029Spjd off += len; 571185029Spjd resid -= len; 572185029Spjd } 573168404Spjd} 574168404Spjd 575168404Spjd/* 576168404Spjd * zfs_log_truncate() handles TX_TRUNCATE transactions. 577168404Spjd */ 578168404Spjdvoid 579168404Spjdzfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype, 580168404Spjd znode_t *zp, uint64_t off, uint64_t len) 581168404Spjd{ 582168404Spjd itx_t *itx; 583168404Spjd uint64_t seq; 584168404Spjd lr_truncate_t *lr; 585168404Spjd 586168404Spjd if (zilog == NULL || zp->z_unlinked) 587168404Spjd return; 588168404Spjd 589209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 590209962Smm 591168404Spjd itx = zil_itx_create(txtype, sizeof (*lr)); 592168404Spjd lr = (lr_truncate_t *)&itx->itx_lr; 593168404Spjd lr->lr_foid = zp->z_id; 594168404Spjd lr->lr_offset = off; 595168404Spjd lr->lr_length = len; 596168404Spjd 597168404Spjd itx->itx_sync = (zp->z_sync_cnt != 0); 598168404Spjd seq = zil_itx_assign(zilog, itx, tx); 599168404Spjd zp->z_last_itx = seq; 600168404Spjd} 601168404Spjd 602168404Spjd/* 603168404Spjd * zfs_log_setattr() handles TX_SETATTR transactions. 604168404Spjd */ 605168404Spjdvoid 606168404Spjdzfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype, 607185029Spjd znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp) 608168404Spjd{ 609185029Spjd itx_t *itx; 610185029Spjd uint64_t seq; 611185029Spjd lr_setattr_t *lr; 612185029Spjd xvattr_t *xvap = (xvattr_t *)vap; 613185029Spjd size_t recsize = sizeof (lr_setattr_t); 614185029Spjd void *start; 615168404Spjd 616185029Spjd 617168404Spjd if (zilog == NULL || zp->z_unlinked) 618168404Spjd return; 619168404Spjd 620209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 621209962Smm 622185029Spjd /* 623185029Spjd * If XVATTR set, then log record size needs to allow 624185029Spjd * for lr_attr_t + xvattr mask, mapsize and create time 625185029Spjd * plus actual attribute values 626185029Spjd */ 627185029Spjd if (vap->va_mask & AT_XVATTR) 628185029Spjd recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize); 629185029Spjd 630185029Spjd if (fuidp) 631185029Spjd recsize += fuidp->z_domain_str_sz; 632185029Spjd 633185029Spjd itx = zil_itx_create(txtype, recsize); 634168404Spjd lr = (lr_setattr_t *)&itx->itx_lr; 635168404Spjd lr->lr_foid = zp->z_id; 636168404Spjd lr->lr_mask = (uint64_t)mask_applied; 637168404Spjd lr->lr_mode = (uint64_t)vap->va_mode; 638185029Spjd if ((mask_applied & AT_UID) && IS_EPHEMERAL(vap->va_uid)) 639185029Spjd lr->lr_uid = fuidp->z_fuid_owner; 640185029Spjd else 641185029Spjd lr->lr_uid = (uint64_t)vap->va_uid; 642185029Spjd 643185029Spjd if ((mask_applied & AT_GID) && IS_EPHEMERAL(vap->va_gid)) 644185029Spjd lr->lr_gid = fuidp->z_fuid_group; 645185029Spjd else 646185029Spjd lr->lr_gid = (uint64_t)vap->va_gid; 647185029Spjd 648168404Spjd lr->lr_size = (uint64_t)vap->va_size; 649168404Spjd ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime); 650168404Spjd ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime); 651185029Spjd start = (lr_setattr_t *)(lr + 1); 652185029Spjd if (vap->va_mask & AT_XVATTR) { 653185029Spjd zfs_log_xvattr((lr_attr_t *)start, xvap); 654185029Spjd start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize); 655185029Spjd } 656168404Spjd 657185029Spjd /* 658185029Spjd * Now stick on domain information if any on end 659185029Spjd */ 660185029Spjd 661185029Spjd if (fuidp) 662185029Spjd (void) zfs_log_fuid_domains(fuidp, start); 663185029Spjd 664168404Spjd itx->itx_sync = (zp->z_sync_cnt != 0); 665168404Spjd seq = zil_itx_assign(zilog, itx, tx); 666168404Spjd zp->z_last_itx = seq; 667168404Spjd} 668168404Spjd 669168404Spjd/* 670168404Spjd * zfs_log_acl() handles TX_ACL transactions. 671168404Spjd */ 672168404Spjdvoid 673185029Spjdzfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp, 674185029Spjd vsecattr_t *vsecp, zfs_fuid_info_t *fuidp) 675168404Spjd{ 676168404Spjd itx_t *itx; 677168404Spjd uint64_t seq; 678185029Spjd lr_acl_v0_t *lrv0; 679168404Spjd lr_acl_t *lr; 680185029Spjd int txtype; 681185029Spjd int lrsize; 682185029Spjd size_t txsize; 683185029Spjd size_t aclbytes = vsecp->vsa_aclentsz; 684168404Spjd 685168404Spjd if (zilog == NULL || zp->z_unlinked) 686168404Spjd return; 687168404Spjd 688209962Smm ZFS_HANDLE_REPLAY(zilog, tx); /* exits if replay */ 689209962Smm 690185029Spjd txtype = (zp->z_zfsvfs->z_version < ZPL_VERSION_FUID) ? 691185029Spjd TX_ACL_V0 : TX_ACL; 692185029Spjd 693185029Spjd if (txtype == TX_ACL) 694185029Spjd lrsize = sizeof (*lr); 695185029Spjd else 696185029Spjd lrsize = sizeof (*lrv0); 697185029Spjd 698185029Spjd txsize = lrsize + 699185029Spjd ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) + 700185029Spjd (fuidp ? fuidp->z_domain_str_sz : 0) + 701185029Spjd sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0); 702185029Spjd 703185029Spjd itx = zil_itx_create(txtype, txsize); 704185029Spjd 705168404Spjd lr = (lr_acl_t *)&itx->itx_lr; 706168404Spjd lr->lr_foid = zp->z_id; 707185029Spjd if (txtype == TX_ACL) { 708185029Spjd lr->lr_acl_bytes = aclbytes; 709185029Spjd lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0; 710185029Spjd lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0; 711185029Spjd if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS) 712185029Spjd lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags; 713185029Spjd else 714185029Spjd lr->lr_acl_flags = 0; 715185029Spjd } 716185029Spjd lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt; 717168404Spjd 718185029Spjd if (txtype == TX_ACL_V0) { 719185029Spjd lrv0 = (lr_acl_v0_t *)lr; 720185029Spjd bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes); 721185029Spjd } else { 722185029Spjd void *start = (ace_t *)(lr + 1); 723185029Spjd 724185029Spjd bcopy(vsecp->vsa_aclentp, start, aclbytes); 725185029Spjd 726185029Spjd start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes); 727185029Spjd 728185029Spjd if (fuidp) { 729185029Spjd start = zfs_log_fuid_ids(fuidp, start); 730185029Spjd (void) zfs_log_fuid_domains(fuidp, start); 731185029Spjd } 732185029Spjd } 733185029Spjd 734168404Spjd itx->itx_sync = (zp->z_sync_cnt != 0); 735168404Spjd seq = zil_itx_assign(zilog, itx, tx); 736168404Spjd zp->z_last_itx = seq; 737168404Spjd} 738