module/zfs/zfs_log.c

   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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2015, 2018 by Delphix. All rights reserved.
  24  */
  25
  26
  27 #include <sys/types.h>
  28 #include <sys/param.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/dbuf.h>
  46 #include <sys/spa.h>
  47 #include <sys/zfs_fuid.h>
  48 #include <sys/dsl_dataset.h>
  49
  50 /*
  51  * These zfs_log_* functions must be called within a dmu tx, in one
  52  * of 2 contexts depending on zilog->z_replay:
  53  *
  54  * Non replay mode
  55  * ---------------
  56  * We need to record the transaction so that if it is committed to
  57  * the Intent Log then it can be replayed.  An intent log transaction
  58  * structure (itx_t) is allocated and all the information necessary to
  59  * possibly replay the transaction is saved in it. The itx is then assigned
  60  * a sequence number and inserted in the in-memory list anchored in the zilog.
  61  *
  62  * Replay mode
  63  * -----------
  64  * We need to mark the intent log record as replayed in the log header.
  65  * This is done in the same transaction as the replay so that they
  66  * commit atomically.
  67  */
  68
  69 int
  70 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
  71 {
  72         int isxvattr = (vap->va_mask & ATTR_XVATTR);
  73         switch (type) {
  74         case Z_FILE:
  75                 if (vsecp == NULL && !isxvattr)
  76                         return (TX_CREATE);
  77                 if (vsecp && isxvattr)
  78                         return (TX_CREATE_ACL_ATTR);
  79                 if (vsecp)
  80                         return (TX_CREATE_ACL);
  81                 else
  82                         return (TX_CREATE_ATTR);
  83                 /*NOTREACHED*/
  84         case Z_DIR:
  85                 if (vsecp == NULL && !isxvattr)
  86                         return (TX_MKDIR);
  87                 if (vsecp && isxvattr)
  88                         return (TX_MKDIR_ACL_ATTR);
  89                 if (vsecp)
  90                         return (TX_MKDIR_ACL);
  91                 else
  92                         return (TX_MKDIR_ATTR);
  93         case Z_XATTRDIR:
  94                 return (TX_MKXATTR);
  95         }
  96         ASSERT(0);
  97         return (TX_MAX_TYPE);
  98 }
  99
 100 /*
 101  * build up the log data necessary for logging xvattr_t
 102  * First lr_attr_t is initialized.  following the lr_attr_t
 103  * is the mapsize and attribute bitmap copied from the xvattr_t.
 104  * Following the bitmap and bitmapsize two 64 bit words are reserved
 105  * for the create time which may be set.  Following the create time
 106  * records a single 64 bit integer which has the bits to set on
 107  * replay for the xvattr.
 108  */
 109 static void
 110 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
 111 {
 112         uint32_t        *bitmap;
 113         uint64_t        *attrs;
 114         uint64_t        *crtime;
 115         xoptattr_t      *xoap;
 116         void            *scanstamp;
 117         int             i;
 118
 119         xoap = xva_getxoptattr(xvap);
 120         ASSERT(xoap);
 121
 122         lrattr->lr_attr_masksize = xvap->xva_mapsize;
 123         bitmap = &lrattr->lr_attr_bitmap;
 124         for (i = 0; i != xvap->xva_mapsize; i++, bitmap++) {
 125                 *bitmap = xvap->xva_reqattrmap[i];
 126         }
 127
 128         /* Now pack the attributes up in a single uint64_t */
 129         attrs = (uint64_t *)bitmap;
 130         crtime = attrs + 1;
 131         scanstamp = (caddr_t)(crtime + 2);
 132         *attrs = 0;
 133         if (XVA_ISSET_REQ(xvap, XAT_READONLY))
 134                 *attrs |= (xoap->xoa_readonly == 0) ? 0 :
 135                     XAT0_READONLY;
 136         if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
 137                 *attrs |= (xoap->xoa_hidden == 0) ? 0 :
 138                     XAT0_HIDDEN;
 139         if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
 140                 *attrs |= (xoap->xoa_system == 0) ? 0 :
 141                     XAT0_SYSTEM;
 142         if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
 143                 *attrs |= (xoap->xoa_archive == 0) ? 0 :
 144                     XAT0_ARCHIVE;
 145         if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
 146                 *attrs |= (xoap->xoa_immutable == 0) ? 0 :
 147                     XAT0_IMMUTABLE;
 148         if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
 149                 *attrs |= (xoap->xoa_nounlink == 0) ? 0 :
 150                     XAT0_NOUNLINK;
 151         if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
 152                 *attrs |= (xoap->xoa_appendonly == 0) ? 0 :
 153                     XAT0_APPENDONLY;
 154         if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
 155                 *attrs |= (xoap->xoa_opaque == 0) ? 0 :
 156                     XAT0_APPENDONLY;
 157         if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
 158                 *attrs |= (xoap->xoa_nodump == 0) ? 0 :
 159                     XAT0_NODUMP;
 160         if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
 161                 *attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
 162                     XAT0_AV_QUARANTINED;
 163         if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
 164                 *attrs |= (xoap->xoa_av_modified == 0) ? 0 :
 165                     XAT0_AV_MODIFIED;
 166         if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
 167                 ZFS_TIME_ENCODE(&xoap->xoa_createtime, crtime);
 168         if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
 169                 ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));
 170
 171                 bcopy(xoap->xoa_av_scanstamp, scanstamp, AV_SCANSTAMP_SZ);
 172         } else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
 173                 /*
 174                  * XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
 175                  * at the same time, so we can share the same space.
 176                  */
 177                 bcopy(&xoap->xoa_projid, scanstamp, sizeof (uint64_t));
 178         }
 179         if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
 180                 *attrs |= (xoap->xoa_reparse == 0) ? 0 :
 181                     XAT0_REPARSE;
 182         if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
 183                 *attrs |= (xoap->xoa_offline == 0) ? 0 :
 184                     XAT0_OFFLINE;
 185         if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
 186                 *attrs |= (xoap->xoa_sparse == 0) ? 0 :
 187                     XAT0_SPARSE;
 188         if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
 189                 *attrs |= (xoap->xoa_projinherit == 0) ? 0 :
 190                     XAT0_PROJINHERIT;
 191 }
 192
 193 static void *
 194 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
 195 {
 196         zfs_fuid_t *zfuid;
 197         uint64_t *fuidloc = start;
 198
 199         /* First copy in the ACE FUIDs */
 200         for (zfuid = list_head(&fuidp->z_fuids); zfuid;
 201             zfuid = list_next(&fuidp->z_fuids, zfuid)) {
 202                 *fuidloc++ = zfuid->z_logfuid;
 203         }
 204         return (fuidloc);
 205 }
 206
 207
 208 static void *
 209 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
 210 {
 211         zfs_fuid_domain_t *zdomain;
 212
 213         /* now copy in the domain info, if any */
 214         if (fuidp->z_domain_str_sz != 0) {
 215                 for (zdomain = list_head(&fuidp->z_domains); zdomain;
 216                     zdomain = list_next(&fuidp->z_domains, zdomain)) {
 217                         bcopy((void *)zdomain->z_domain, start,
 218                             strlen(zdomain->z_domain) + 1);
 219                         start = (caddr_t)start +
 220                             strlen(zdomain->z_domain) + 1;
 221                 }
 222         }
 223         return (start);
 224 }
 225
 226 /*
 227  * If zp is an xattr node, check whether the xattr owner is unlinked.
 228  * We don't want to log anything if the owner is unlinked.
 229  */
 230 static int
 231 zfs_xattr_owner_unlinked(znode_t *zp)
 232 {
 233         int unlinked = 0;
 234         znode_t *dzp;
 235         igrab(ZTOI(zp));
 236         /*
 237          * if zp is XATTR node, keep walking up via z_xattr_parent until we
 238          * get the owner
 239          */
 240         while (zp->z_pflags & ZFS_XATTR) {
 241                 ASSERT3U(zp->z_xattr_parent, !=, 0);
 242                 if (zfs_zget(ZTOZSB(zp), zp->z_xattr_parent, &dzp) != 0) {
 243                         unlinked = 1;
 244                         break;
 245                 }
 246                 iput(ZTOI(zp));
 247                 zp = dzp;
 248                 unlinked = zp->z_unlinked;
 249         }
 250         iput(ZTOI(zp));
 251         return (unlinked);
 252 }
 253
 254 /*
 255  * Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
 256  * TK_MKXATTR transactions.
 257  *
 258  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
 259  * domain information appended prior to the name.  In this case the
 260  * uid/gid in the log record will be a log centric FUID.
 261  *
 262  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
 263  * may contain attributes, ACL and optional fuid information.
 264  *
 265  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
 266  * and ACL and normal users/groups in the ACEs.
 267  *
 268  * There may be an optional xvattr attribute information similar
 269  * to zfs_log_setattr.
 270  *
 271  * Also, after the file name "domain" strings may be appended.
 272  */
 273 void
 274 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 275     znode_t *dzp, znode_t *zp, char *name, vsecattr_t *vsecp,
 276     zfs_fuid_info_t *fuidp, vattr_t *vap)
 277 {
 278         itx_t *itx;
 279         lr_create_t *lr;
 280         lr_acl_create_t *lracl;
 281         size_t aclsize = 0;
 282         size_t xvatsize = 0;
 283         size_t txsize;
 284         xvattr_t *xvap = (xvattr_t *)vap;
 285         void *end;
 286         size_t lrsize;
 287         size_t namesize = strlen(name) + 1;
 288         size_t fuidsz = 0;
 289
 290         if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
 291                 return;
 292
 293         /*
 294          * If we have FUIDs present then add in space for
 295          * domains and ACE fuid's if any.
 296          */
 297         if (fuidp) {
 298                 fuidsz += fuidp->z_domain_str_sz;
 299                 fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
 300         }
 301
 302         if (vap->va_mask & ATTR_XVATTR)
 303                 xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
 304
 305         if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
 306             (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
 307             (int)txtype == TX_MKXATTR) {
 308                 txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
 309                 lrsize = sizeof (*lr);
 310         } else {
 311                 txsize =
 312                     sizeof (lr_acl_create_t) + namesize + fuidsz +
 313                     ZIL_ACE_LENGTH(aclsize) + xvatsize;
 314                 lrsize = sizeof (lr_acl_create_t);
 315         }
 316
 317         itx = zil_itx_create(txtype, txsize);
 318
 319         lr = (lr_create_t *)&itx->itx_lr;
 320         lr->lr_doid = dzp->z_id;
 321         lr->lr_foid = zp->z_id;
 322         /* Store dnode slot count in 8 bits above object id. */
 323         LR_FOID_SET_SLOTS(lr->lr_foid, zp->z_dnodesize >> DNODE_SHIFT);
 324         lr->lr_mode = zp->z_mode;
 325         if (!IS_EPHEMERAL(KUID_TO_SUID(ZTOI(zp)->i_uid))) {
 326                 lr->lr_uid = (uint64_t)KUID_TO_SUID(ZTOI(zp)->i_uid);
 327         } else {
 328                 lr->lr_uid = fuidp->z_fuid_owner;
 329         }
 330         if (!IS_EPHEMERAL(KGID_TO_SGID(ZTOI(zp)->i_gid))) {
 331                 lr->lr_gid = (uint64_t)KGID_TO_SGID(ZTOI(zp)->i_gid);
 332         } else {
 333                 lr->lr_gid = fuidp->z_fuid_group;
 334         }
 335         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
 336             sizeof (uint64_t));
 337         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
 338             lr->lr_crtime, sizeof (uint64_t) * 2);
 339
 340         if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(zp)), &lr->lr_rdev,
 341             sizeof (lr->lr_rdev)) != 0)
 342                 lr->lr_rdev = 0;
 343
 344         /*
 345          * Fill in xvattr info if any
 346          */
 347         if (vap->va_mask & ATTR_XVATTR) {
 348                 zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
 349                 end = (caddr_t)lr + lrsize + xvatsize;
 350         } else {
 351                 end = (caddr_t)lr + lrsize;
 352         }
 353
 354         /* Now fill in any ACL info */
 355
 356         if (vsecp) {
 357                 lracl = (lr_acl_create_t *)&itx->itx_lr;
 358                 lracl->lr_aclcnt = vsecp->vsa_aclcnt;
 359                 lracl->lr_acl_bytes = aclsize;
 360                 lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
 361                 lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
 362                 if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
 363                         lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
 364                 else
 365                         lracl->lr_acl_flags = 0;
 366
 367                 bcopy(vsecp->vsa_aclentp, end, aclsize);
 368                 end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
 369         }
 370
 371         /* drop in FUID info */
 372         if (fuidp) {
 373                 end = zfs_log_fuid_ids(fuidp, end);
 374                 end = zfs_log_fuid_domains(fuidp, end);
 375         }
 376         /*
 377          * Now place file name in log record
 378          */
 379         bcopy(name, end, namesize);
 380
 381         zil_itx_assign(zilog, itx, tx);
 382 }
 383
 384 void zil_remove_async(zilog_t *zilog, uint64_t oid);
 385
 386 /*
 387  * Handles both TX_REMOVE and TX_RMDIR transactions.
 388  */
 389 void
 390 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 391     znode_t *dzp, char *name, uint64_t foid, boolean_t unlinked)
 392 {
 393         itx_t *itx;
 394         lr_remove_t *lr;
 395         size_t namesize = strlen(name) + 1;
 396
 397         if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
 398                 return;
 399
 400         itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
 401         lr = (lr_remove_t *)&itx->itx_lr;
 402         lr->lr_doid = dzp->z_id;
 403         bcopy(name, (char *)(lr + 1), namesize);
 404
 405         itx->itx_oid = foid;
 406
 407         /*
 408          * Object ids can be re-instantiated in the next txg so
 409          * remove any async transactions to avoid future leaks.
 410          * This can happen if a fsync occurs on the re-instantiated
 411          * object for a WR_INDIRECT or WR_NEED_COPY write, which gets
 412          * the new file data and flushes a write record for the old object.
 413          */
 414         if (unlinked) {
 415                 ASSERT((txtype & ~TX_CI) == TX_REMOVE);
 416                 zil_remove_async(zilog, foid);
 417         }
 418         zil_itx_assign(zilog, itx, tx);
 419 }
 420
 421 /*
 422  * Handles TX_LINK transactions.
 423  */
 424 void
 425 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 426     znode_t *dzp, znode_t *zp, char *name)
 427 {
 428         itx_t *itx;
 429         lr_link_t *lr;
 430         size_t namesize = strlen(name) + 1;
 431
 432         if (zil_replaying(zilog, tx))
 433                 return;
 434
 435         itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
 436         lr = (lr_link_t *)&itx->itx_lr;
 437         lr->lr_doid = dzp->z_id;
 438         lr->lr_link_obj = zp->z_id;
 439         bcopy(name, (char *)(lr + 1), namesize);
 440
 441         zil_itx_assign(zilog, itx, tx);
 442 }
 443
 444 /*
 445  * Handles TX_SYMLINK transactions.
 446  */
 447 void
 448 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 449     znode_t *dzp, znode_t *zp, char *name, char *link)
 450 {
 451         itx_t *itx;
 452         lr_create_t *lr;
 453         size_t namesize = strlen(name) + 1;
 454         size_t linksize = strlen(link) + 1;
 455
 456         if (zil_replaying(zilog, tx))
 457                 return;
 458
 459         itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
 460         lr = (lr_create_t *)&itx->itx_lr;
 461         lr->lr_doid = dzp->z_id;
 462         lr->lr_foid = zp->z_id;
 463         lr->lr_uid = KUID_TO_SUID(ZTOI(zp)->i_uid);
 464         lr->lr_gid = KGID_TO_SGID(ZTOI(zp)->i_gid);
 465         lr->lr_mode = zp->z_mode;
 466         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
 467             sizeof (uint64_t));
 468         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
 469             lr->lr_crtime, sizeof (uint64_t) * 2);
 470         bcopy(name, (char *)(lr + 1), namesize);
 471         bcopy(link, (char *)(lr + 1) + namesize, linksize);
 472
 473         zil_itx_assign(zilog, itx, tx);
 474 }
 475
 476 /*
 477  * Handles TX_RENAME transactions.
 478  */
 479 void
 480 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 481     znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp)
 482 {
 483         itx_t *itx;
 484         lr_rename_t *lr;
 485         size_t snamesize = strlen(sname) + 1;
 486         size_t dnamesize = strlen(dname) + 1;
 487
 488         if (zil_replaying(zilog, tx))
 489                 return;
 490
 491         itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
 492         lr = (lr_rename_t *)&itx->itx_lr;
 493         lr->lr_sdoid = sdzp->z_id;
 494         lr->lr_tdoid = tdzp->z_id;
 495         bcopy(sname, (char *)(lr + 1), snamesize);
 496         bcopy(dname, (char *)(lr + 1) + snamesize, dnamesize);
 497         itx->itx_oid = szp->z_id;
 498
 499         zil_itx_assign(zilog, itx, tx);
 500 }
 501
 502 /*
 503  * zfs_log_write() handles TX_WRITE transactions. The specified callback is
 504  * called as soon as the write is on stable storage (be it via a DMU sync or a
 505  * ZIL commit).
 506  */
 507 long zfs_immediate_write_sz = 32768;
 508
 509 void
 510 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 511     znode_t *zp, offset_t off, ssize_t resid, int ioflag,
 512     zil_callback_t callback, void *callback_data)
 513 {
 514         dmu_buf_impl_t *db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl);
 515         uint32_t blocksize = zp->z_blksz;
 516         itx_wr_state_t write_state;
 517         uintptr_t fsync_cnt;
 518
 519         if (zil_replaying(zilog, tx) || zp->z_unlinked ||
 520             zfs_xattr_owner_unlinked(zp)) {
 521                 if (callback != NULL)
 522                         callback(callback_data);
 523                 return;
 524         }
 525
 526         if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
 527                 write_state = WR_INDIRECT;
 528         else if (!spa_has_slogs(zilog->zl_spa) &&
 529             resid >= zfs_immediate_write_sz)
 530                 write_state = WR_INDIRECT;
 531         else if (ioflag & (FSYNC | FDSYNC))
 532                 write_state = WR_COPIED;
 533         else
 534                 write_state = WR_NEED_COPY;
 535
 536         if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
 537                 (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
 538         }
 539
 540         while (resid) {
 541                 itx_t *itx;
 542                 lr_write_t *lr;
 543                 itx_wr_state_t wr_state = write_state;
 544                 ssize_t len = resid;
 545
 546                 /*
 547                  * A WR_COPIED record must fit entirely in one log block.
 548                  * Large writes can use WR_NEED_COPY, which the ZIL will
 549                  * split into multiple records across several log blocks
 550                  * if necessary.
 551                  */
 552                 if (wr_state == WR_COPIED &&
 553                     resid > zil_max_copied_data(zilog))
 554                         wr_state = WR_NEED_COPY;
 555                 else if (wr_state == WR_INDIRECT)
 556                         len = MIN(blocksize - P2PHASE(off, blocksize), resid);
 557
 558                 itx = zil_itx_create(txtype, sizeof (*lr) +
 559                     (wr_state == WR_COPIED ? len : 0));
 560                 lr = (lr_write_t *)&itx->itx_lr;
 561
 562                 DB_DNODE_ENTER(db);
 563                 if (wr_state == WR_COPIED && dmu_read_by_dnode(DB_DNODE(db),
 564                     off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) {
 565                         zil_itx_destroy(itx);
 566                         itx = zil_itx_create(txtype, sizeof (*lr));
 567                         lr = (lr_write_t *)&itx->itx_lr;
 568                         wr_state = WR_NEED_COPY;
 569                 }
 570                 DB_DNODE_EXIT(db);
 571
 572                 itx->itx_wr_state = wr_state;
 573                 lr->lr_foid = zp->z_id;
 574                 lr->lr_offset = off;
 575                 lr->lr_length = len;
 576                 lr->lr_blkoff = 0;
 577                 BP_ZERO(&lr->lr_blkptr);
 578
 579                 itx->itx_private = ZTOZSB(zp);
 580
 581                 if (!(ioflag & (FSYNC | FDSYNC)) && (zp->z_sync_cnt == 0) &&
 582                     (fsync_cnt == 0))
 583                         itx->itx_sync = B_FALSE;
 584
 585                 itx->itx_callback = callback;
 586                 itx->itx_callback_data = callback_data;
 587                 zil_itx_assign(zilog, itx, tx);
 588
 589                 off += len;
 590                 resid -= len;
 591         }
 592 }
 593
 594 /*
 595  * Handles TX_TRUNCATE transactions.
 596  */
 597 void
 598 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 599     znode_t *zp, uint64_t off, uint64_t len)
 600 {
 601         itx_t *itx;
 602         lr_truncate_t *lr;
 603
 604         if (zil_replaying(zilog, tx) || zp->z_unlinked ||
 605             zfs_xattr_owner_unlinked(zp))
 606                 return;
 607
 608         itx = zil_itx_create(txtype, sizeof (*lr));
 609         lr = (lr_truncate_t *)&itx->itx_lr;
 610         lr->lr_foid = zp->z_id;
 611         lr->lr_offset = off;
 612         lr->lr_length = len;
 613
 614         itx->itx_sync = (zp->z_sync_cnt != 0);
 615         zil_itx_assign(zilog, itx, tx);
 616 }
 617
 618 /*
 619  * Handles TX_SETATTR transactions.
 620  */
 621 void
 622 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 623     znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
 624 {
 625         itx_t           *itx;
 626         lr_setattr_t    *lr;
 627         xvattr_t        *xvap = (xvattr_t *)vap;
 628         size_t          recsize = sizeof (lr_setattr_t);
 629         void            *start;
 630
 631         if (zil_replaying(zilog, tx) || zp->z_unlinked)
 632                 return;
 633
 634         /*
 635          * If XVATTR set, then log record size needs to allow
 636          * for lr_attr_t + xvattr mask, mapsize and create time
 637          * plus actual attribute values
 638          */
 639         if (vap->va_mask & ATTR_XVATTR)
 640                 recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
 641
 642         if (fuidp)
 643                 recsize += fuidp->z_domain_str_sz;
 644
 645         itx = zil_itx_create(txtype, recsize);
 646         lr = (lr_setattr_t *)&itx->itx_lr;
 647         lr->lr_foid = zp->z_id;
 648         lr->lr_mask = (uint64_t)mask_applied;
 649         lr->lr_mode = (uint64_t)vap->va_mode;
 650         if ((mask_applied & ATTR_UID) && IS_EPHEMERAL(vap->va_uid))
 651                 lr->lr_uid = fuidp->z_fuid_owner;
 652         else
 653                 lr->lr_uid = (uint64_t)vap->va_uid;
 654
 655         if ((mask_applied & ATTR_GID) && IS_EPHEMERAL(vap->va_gid))
 656                 lr->lr_gid = fuidp->z_fuid_group;
 657         else
 658                 lr->lr_gid = (uint64_t)vap->va_gid;
 659
 660         lr->lr_size = (uint64_t)vap->va_size;
 661         ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
 662         ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
 663         start = (lr_setattr_t *)(lr + 1);
 664         if (vap->va_mask & ATTR_XVATTR) {
 665                 zfs_log_xvattr((lr_attr_t *)start, xvap);
 666                 start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
 667         }
 668
 669         /*
 670          * Now stick on domain information if any on end
 671          */
 672
 673         if (fuidp)
 674                 (void) zfs_log_fuid_domains(fuidp, start);
 675
 676         itx->itx_sync = (zp->z_sync_cnt != 0);
 677         zil_itx_assign(zilog, itx, tx);
 678 }
 679
 680 /*
 681  * Handles TX_ACL transactions.
 682  */
 683 void
 684 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
 685     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
 686 {
 687         itx_t *itx;
 688         lr_acl_v0_t *lrv0;
 689         lr_acl_t *lr;
 690         int txtype;
 691         int lrsize;
 692         size_t txsize;
 693         size_t aclbytes = vsecp->vsa_aclentsz;
 694
 695         if (zil_replaying(zilog, tx) || zp->z_unlinked)
 696                 return;
 697
 698         txtype = (ZTOZSB(zp)->z_version < ZPL_VERSION_FUID) ?
 699             TX_ACL_V0 : TX_ACL;
 700
 701         if (txtype == TX_ACL)
 702                 lrsize = sizeof (*lr);
 703         else
 704                 lrsize = sizeof (*lrv0);
 705
 706         txsize = lrsize +
 707             ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
 708             (fuidp ? fuidp->z_domain_str_sz : 0) +
 709             sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
 710
 711         itx = zil_itx_create(txtype, txsize);
 712
 713         lr = (lr_acl_t *)&itx->itx_lr;
 714         lr->lr_foid = zp->z_id;
 715         if (txtype == TX_ACL) {
 716                 lr->lr_acl_bytes = aclbytes;
 717                 lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
 718                 lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
 719                 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
 720                         lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
 721                 else
 722                         lr->lr_acl_flags = 0;
 723         }
 724         lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
 725
 726         if (txtype == TX_ACL_V0) {
 727                 lrv0 = (lr_acl_v0_t *)lr;
 728                 bcopy(vsecp->vsa_aclentp, (ace_t *)(lrv0 + 1), aclbytes);
 729         } else {
 730                 void *start = (ace_t *)(lr + 1);
 731
 732                 bcopy(vsecp->vsa_aclentp, start, aclbytes);
 733
 734                 start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
 735
 736                 if (fuidp) {
 737                         start = zfs_log_fuid_ids(fuidp, start);
 738                         (void) zfs_log_fuid_domains(fuidp, start);
 739                 }
 740         }
 741
 742         itx->itx_sync = (zp->z_sync_cnt != 0);
 743         zil_itx_assign(zilog, itx, tx);
 744 }
 745
 746 #if defined(_KERNEL)
 747 module_param(zfs_immediate_write_sz, long, 0644);
 748 MODULE_PARM_DESC(zfs_immediate_write_sz, "Largest data block to write to zil");
 749 #endif