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 https://opensource.org/licenses/CDDL-1.0.
  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/acl.h>
  43 #include <sys/dmu.h>
  44 #include <sys/dbuf.h>
  45 #include <sys/spa.h>
  46 #include <sys/zfs_fuid.h>
  47 #include <sys/dsl_dataset.h>
  48
  49 /*
  50  * These zfs_log_* functions must be called within a dmu tx, in one
  51  * of 2 contexts depending on zilog->z_replay:
  52  *
  53  * Non replay mode
  54  * ---------------
  55  * We need to record the transaction so that if it is committed to
  56  * the Intent Log then it can be replayed.  An intent log transaction
  57  * structure (itx_t) is allocated and all the information necessary to
  58  * possibly replay the transaction is saved in it. The itx is then assigned
  59  * a sequence number and inserted in the in-memory list anchored in the zilog.
  60  *
  61  * Replay mode
  62  * -----------
  63  * We need to mark the intent log record as replayed in the log header.
  64  * This is done in the same transaction as the replay so that they
  65  * commit atomically.
  66  */
  67
  68 int
  69 zfs_log_create_txtype(zil_create_t type, vsecattr_t *vsecp, vattr_t *vap)
  70 {
  71         int isxvattr = (vap->va_mask & ATTR_XVATTR);
  72         switch (type) {
  73         case Z_FILE:
  74                 if (vsecp == NULL && !isxvattr)
  75                         return (TX_CREATE);
  76                 if (vsecp && isxvattr)
  77                         return (TX_CREATE_ACL_ATTR);
  78                 if (vsecp)
  79                         return (TX_CREATE_ACL);
  80                 else
  81                         return (TX_CREATE_ATTR);
  82         case Z_DIR:
  83                 if (vsecp == NULL && !isxvattr)
  84                         return (TX_MKDIR);
  85                 if (vsecp && isxvattr)
  86                         return (TX_MKDIR_ACL_ATTR);
  87                 if (vsecp)
  88                         return (TX_MKDIR_ACL);
  89                 else
  90                         return (TX_MKDIR_ATTR);
  91         case Z_XATTRDIR:
  92                 return (TX_MKXATTR);
  93         }
  94         ASSERT(0);
  95         return (TX_MAX_TYPE);
  96 }
  97
  98 /*
  99  * build up the log data necessary for logging xvattr_t
 100  * First lr_attr_t is initialized.  following the lr_attr_t
 101  * is the mapsize and attribute bitmap copied from the xvattr_t.
 102  * Following the bitmap and bitmapsize two 64 bit words are reserved
 103  * for the create time which may be set.  Following the create time
 104  * records a single 64 bit integer which has the bits to set on
 105  * replay for the xvattr.
 106  */
 107 static void
 108 zfs_log_xvattr(lr_attr_t *lrattr, xvattr_t *xvap)
 109 {
 110         xoptattr_t *xoap;
 111
 112         xoap = xva_getxoptattr(xvap);
 113         ASSERT(xoap);
 114
 115         lrattr->lr_attr_masksize = xvap->xva_mapsize;
 116         uint32_t *bitmap = &lrattr->lr_attr_bitmap;
 117         for (int i = 0; i != xvap->xva_mapsize; i++, bitmap++)
 118                 *bitmap = xvap->xva_reqattrmap[i];
 119
 120         lr_attr_end_t *end = (lr_attr_end_t *)bitmap;
 121         end->lr_attr_attrs = 0;
 122         end->lr_attr_crtime[0] = 0;
 123         end->lr_attr_crtime[1] = 0;
 124         memset(end->lr_attr_scanstamp, 0, AV_SCANSTAMP_SZ);
 125
 126         if (XVA_ISSET_REQ(xvap, XAT_READONLY))
 127                 end->lr_attr_attrs |= (xoap->xoa_readonly == 0) ? 0 :
 128                     XAT0_READONLY;
 129         if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
 130                 end->lr_attr_attrs |= (xoap->xoa_hidden == 0) ? 0 :
 131                     XAT0_HIDDEN;
 132         if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
 133                 end->lr_attr_attrs |= (xoap->xoa_system == 0) ? 0 :
 134                     XAT0_SYSTEM;
 135         if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
 136                 end->lr_attr_attrs |= (xoap->xoa_archive == 0) ? 0 :
 137                     XAT0_ARCHIVE;
 138         if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
 139                 end->lr_attr_attrs |= (xoap->xoa_immutable == 0) ? 0 :
 140                     XAT0_IMMUTABLE;
 141         if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
 142                 end->lr_attr_attrs |= (xoap->xoa_nounlink == 0) ? 0 :
 143                     XAT0_NOUNLINK;
 144         if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
 145                 end->lr_attr_attrs |= (xoap->xoa_appendonly == 0) ? 0 :
 146                     XAT0_APPENDONLY;
 147         if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
 148                 end->lr_attr_attrs |= (xoap->xoa_opaque == 0) ? 0 :
 149                     XAT0_APPENDONLY;
 150         if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
 151                 end->lr_attr_attrs |= (xoap->xoa_nodump == 0) ? 0 :
 152                     XAT0_NODUMP;
 153         if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
 154                 end->lr_attr_attrs |= (xoap->xoa_av_quarantined == 0) ? 0 :
 155                     XAT0_AV_QUARANTINED;
 156         if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
 157                 end->lr_attr_attrs |= (xoap->xoa_av_modified == 0) ? 0 :
 158                     XAT0_AV_MODIFIED;
 159         if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
 160                 ZFS_TIME_ENCODE(&xoap->xoa_createtime, end->lr_attr_crtime);
 161         if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
 162                 ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));
 163
 164                 memcpy(end->lr_attr_scanstamp, xoap->xoa_av_scanstamp,
 165                     AV_SCANSTAMP_SZ);
 166         } else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
 167                 /*
 168                  * XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
 169                  * at the same time, so we can share the same space.
 170                  */
 171                 memcpy(end->lr_attr_scanstamp, &xoap->xoa_projid,
 172                     sizeof (uint64_t));
 173         }
 174         if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
 175                 end->lr_attr_attrs |= (xoap->xoa_reparse == 0) ? 0 :
 176                     XAT0_REPARSE;
 177         if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
 178                 end->lr_attr_attrs |= (xoap->xoa_offline == 0) ? 0 :
 179                     XAT0_OFFLINE;
 180         if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
 181                 end->lr_attr_attrs |= (xoap->xoa_sparse == 0) ? 0 :
 182                     XAT0_SPARSE;
 183         if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
 184                 end->lr_attr_attrs |= (xoap->xoa_projinherit == 0) ? 0 :
 185                     XAT0_PROJINHERIT;
 186 }
 187
 188 static void *
 189 zfs_log_fuid_ids(zfs_fuid_info_t *fuidp, void *start)
 190 {
 191         zfs_fuid_t *zfuid;
 192         uint64_t *fuidloc = start;
 193
 194         /* First copy in the ACE FUIDs */
 195         for (zfuid = list_head(&fuidp->z_fuids); zfuid;
 196             zfuid = list_next(&fuidp->z_fuids, zfuid)) {
 197                 *fuidloc++ = zfuid->z_logfuid;
 198         }
 199         return (fuidloc);
 200 }
 201
 202
 203 static void *
 204 zfs_log_fuid_domains(zfs_fuid_info_t *fuidp, void *start)
 205 {
 206         zfs_fuid_domain_t *zdomain;
 207
 208         /* now copy in the domain info, if any */
 209         if (fuidp->z_domain_str_sz != 0) {
 210                 for (zdomain = list_head(&fuidp->z_domains); zdomain;
 211                     zdomain = list_next(&fuidp->z_domains, zdomain)) {
 212                         memcpy(start, zdomain->z_domain,
 213                             strlen(zdomain->z_domain) + 1);
 214                         start = (caddr_t)start +
 215                             strlen(zdomain->z_domain) + 1;
 216                 }
 217         }
 218         return (start);
 219 }
 220
 221 /*
 222  * If zp is an xattr node, check whether the xattr owner is unlinked.
 223  * We don't want to log anything if the owner is unlinked.
 224  */
 225 static int
 226 zfs_xattr_owner_unlinked(znode_t *zp)
 227 {
 228         int unlinked = 0;
 229         znode_t *dzp;
 230 #ifdef __FreeBSD__
 231         znode_t *tzp = zp;
 232
 233         /*
 234          * zrele drops the vnode lock which violates the VOP locking contract
 235          * on FreeBSD. See comment at the top of zfs_replay.c for more detail.
 236          */
 237         /*
 238          * if zp is XATTR node, keep walking up via z_xattr_parent until we
 239          * get the owner
 240          */
 241         while (tzp->z_pflags & ZFS_XATTR) {
 242                 ASSERT3U(zp->z_xattr_parent, !=, 0);
 243                 if (zfs_zget(ZTOZSB(tzp), tzp->z_xattr_parent, &dzp) != 0) {
 244                         unlinked = 1;
 245                         break;
 246                 }
 247
 248                 if (tzp != zp)
 249                         zrele(tzp);
 250                 tzp = dzp;
 251                 unlinked = tzp->z_unlinked;
 252         }
 253         if (tzp != zp)
 254                 zrele(tzp);
 255 #else
 256         zhold(zp);
 257         /*
 258          * if zp is XATTR node, keep walking up via z_xattr_parent until we
 259          * get the owner
 260          */
 261         while (zp->z_pflags & ZFS_XATTR) {
 262                 ASSERT3U(zp->z_xattr_parent, !=, 0);
 263                 if (zfs_zget(ZTOZSB(zp), zp->z_xattr_parent, &dzp) != 0) {
 264                         unlinked = 1;
 265                         break;
 266                 }
 267
 268                 zrele(zp);
 269                 zp = dzp;
 270                 unlinked = zp->z_unlinked;
 271         }
 272         zrele(zp);
 273 #endif
 274         return (unlinked);
 275 }
 276
 277 /*
 278  * Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
 279  * TK_MKXATTR transactions.
 280  *
 281  * TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
 282  * domain information appended prior to the name.  In this case the
 283  * uid/gid in the log record will be a log centric FUID.
 284  *
 285  * TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
 286  * may contain attributes, ACL and optional fuid information.
 287  *
 288  * TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
 289  * and ACL and normal users/groups in the ACEs.
 290  *
 291  * There may be an optional xvattr attribute information similar
 292  * to zfs_log_setattr.
 293  *
 294  * Also, after the file name "domain" strings may be appended.
 295  */
 296 void
 297 zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 298     znode_t *dzp, znode_t *zp, const char *name, vsecattr_t *vsecp,
 299     zfs_fuid_info_t *fuidp, vattr_t *vap)
 300 {
 301         itx_t *itx;
 302         lr_create_t *lr;
 303         lr_acl_create_t *lracl;
 304         size_t aclsize = 0;
 305         size_t xvatsize = 0;
 306         size_t txsize;
 307         xvattr_t *xvap = (xvattr_t *)vap;
 308         void *end;
 309         size_t lrsize;
 310         size_t namesize = strlen(name) + 1;
 311         size_t fuidsz = 0;
 312
 313         if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
 314                 return;
 315
 316         /*
 317          * If we have FUIDs present then add in space for
 318          * domains and ACE fuid's if any.
 319          */
 320         if (fuidp) {
 321                 fuidsz += fuidp->z_domain_str_sz;
 322                 fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
 323         }
 324
 325         if (vap->va_mask & ATTR_XVATTR)
 326                 xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);
 327
 328         if ((int)txtype == TX_CREATE_ATTR || (int)txtype == TX_MKDIR_ATTR ||
 329             (int)txtype == TX_CREATE || (int)txtype == TX_MKDIR ||
 330             (int)txtype == TX_MKXATTR) {
 331                 txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
 332                 lrsize = sizeof (*lr);
 333         } else {
 334                 txsize =
 335                     sizeof (lr_acl_create_t) + namesize + fuidsz +
 336                     ZIL_ACE_LENGTH(aclsize) + xvatsize;
 337                 lrsize = sizeof (lr_acl_create_t);
 338         }
 339
 340         itx = zil_itx_create(txtype, txsize);
 341
 342         lr = (lr_create_t *)&itx->itx_lr;
 343         lr->lr_doid = dzp->z_id;
 344         lr->lr_foid = zp->z_id;
 345         /* Store dnode slot count in 8 bits above object id. */
 346         LR_FOID_SET_SLOTS(lr->lr_foid, zp->z_dnodesize >> DNODE_SHIFT);
 347         lr->lr_mode = zp->z_mode;
 348         if (!IS_EPHEMERAL(KUID_TO_SUID(ZTOUID(zp)))) {
 349                 lr->lr_uid = (uint64_t)KUID_TO_SUID(ZTOUID(zp));
 350         } else {
 351                 lr->lr_uid = fuidp->z_fuid_owner;
 352         }
 353         if (!IS_EPHEMERAL(KGID_TO_SGID(ZTOGID(zp)))) {
 354                 lr->lr_gid = (uint64_t)KGID_TO_SGID(ZTOGID(zp));
 355         } else {
 356                 lr->lr_gid = fuidp->z_fuid_group;
 357         }
 358         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
 359             sizeof (uint64_t));
 360         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
 361             lr->lr_crtime, sizeof (uint64_t) * 2);
 362
 363         if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(zp)), &lr->lr_rdev,
 364             sizeof (lr->lr_rdev)) != 0)
 365                 lr->lr_rdev = 0;
 366
 367         /*
 368          * Fill in xvattr info if any
 369          */
 370         if (vap->va_mask & ATTR_XVATTR) {
 371                 zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
 372                 end = (caddr_t)lr + lrsize + xvatsize;
 373         } else {
 374                 end = (caddr_t)lr + lrsize;
 375         }
 376
 377         /* Now fill in any ACL info */
 378
 379         if (vsecp) {
 380                 lracl = (lr_acl_create_t *)&itx->itx_lr;
 381                 lracl->lr_aclcnt = vsecp->vsa_aclcnt;
 382                 lracl->lr_acl_bytes = aclsize;
 383                 lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
 384                 lracl->lr_fuidcnt  = fuidp ? fuidp->z_fuid_cnt : 0;
 385                 if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
 386                         lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
 387                 else
 388                         lracl->lr_acl_flags = 0;
 389
 390                 memcpy(end, vsecp->vsa_aclentp, aclsize);
 391                 end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
 392         }
 393
 394         /* drop in FUID info */
 395         if (fuidp) {
 396                 end = zfs_log_fuid_ids(fuidp, end);
 397                 end = zfs_log_fuid_domains(fuidp, end);
 398         }
 399         /*
 400          * Now place file name in log record
 401          */
 402         memcpy(end, name, namesize);
 403
 404         zil_itx_assign(zilog, itx, tx);
 405 }
 406
 407 /*
 408  * Handles both TX_REMOVE and TX_RMDIR transactions.
 409  */
 410 void
 411 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 412     znode_t *dzp, const char *name, uint64_t foid, boolean_t unlinked)
 413 {
 414         itx_t *itx;
 415         lr_remove_t *lr;
 416         size_t namesize = strlen(name) + 1;
 417
 418         if (zil_replaying(zilog, tx) || zfs_xattr_owner_unlinked(dzp))
 419                 return;
 420
 421         itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
 422         lr = (lr_remove_t *)&itx->itx_lr;
 423         lr->lr_doid = dzp->z_id;
 424         memcpy(lr + 1, name, namesize);
 425
 426         itx->itx_oid = foid;
 427
 428         /*
 429          * Object ids can be re-instantiated in the next txg so
 430          * remove any async transactions to avoid future leaks.
 431          * This can happen if a fsync occurs on the re-instantiated
 432          * object for a WR_INDIRECT or WR_NEED_COPY write, which gets
 433          * the new file data and flushes a write record for the old object.
 434          */
 435         if (unlinked) {
 436                 ASSERT((txtype & ~TX_CI) == TX_REMOVE);
 437                 zil_remove_async(zilog, foid);
 438         }
 439         zil_itx_assign(zilog, itx, tx);
 440 }
 441
 442 /*
 443  * Handles TX_LINK transactions.
 444  */
 445 void
 446 zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 447     znode_t *dzp, znode_t *zp, const char *name)
 448 {
 449         itx_t *itx;
 450         lr_link_t *lr;
 451         size_t namesize = strlen(name) + 1;
 452
 453         if (zil_replaying(zilog, tx))
 454                 return;
 455
 456         itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
 457         lr = (lr_link_t *)&itx->itx_lr;
 458         lr->lr_doid = dzp->z_id;
 459         lr->lr_link_obj = zp->z_id;
 460         memcpy(lr + 1, name, namesize);
 461
 462         zil_itx_assign(zilog, itx, tx);
 463 }
 464
 465 /*
 466  * Handles TX_SYMLINK transactions.
 467  */
 468 void
 469 zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 470     znode_t *dzp, znode_t *zp, const char *name, const char *link)
 471 {
 472         itx_t *itx;
 473         lr_create_t *lr;
 474         size_t namesize = strlen(name) + 1;
 475         size_t linksize = strlen(link) + 1;
 476
 477         if (zil_replaying(zilog, tx))
 478                 return;
 479
 480         itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
 481         lr = (lr_create_t *)&itx->itx_lr;
 482         lr->lr_doid = dzp->z_id;
 483         lr->lr_foid = zp->z_id;
 484         lr->lr_uid = KUID_TO_SUID(ZTOUID(zp));
 485         lr->lr_gid = KGID_TO_SGID(ZTOGID(zp));
 486         lr->lr_mode = zp->z_mode;
 487         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
 488             sizeof (uint64_t));
 489         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
 490             lr->lr_crtime, sizeof (uint64_t) * 2);
 491         memcpy((char *)(lr + 1), name, namesize);
 492         memcpy((char *)(lr + 1) + namesize, link, linksize);
 493
 494         zil_itx_assign(zilog, itx, tx);
 495 }
 496
 497 /*
 498  * Handles TX_RENAME transactions.
 499  */
 500 void
 501 zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype, znode_t *sdzp,
 502     const char *sname, znode_t *tdzp, const char *dname, znode_t *szp)
 503 {
 504         itx_t *itx;
 505         lr_rename_t *lr;
 506         size_t snamesize = strlen(sname) + 1;
 507         size_t dnamesize = strlen(dname) + 1;
 508
 509         if (zil_replaying(zilog, tx))
 510                 return;
 511
 512         itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
 513         lr = (lr_rename_t *)&itx->itx_lr;
 514         lr->lr_sdoid = sdzp->z_id;
 515         lr->lr_tdoid = tdzp->z_id;
 516         memcpy((char *)(lr + 1), sname, snamesize);
 517         memcpy((char *)(lr + 1) + snamesize, dname, dnamesize);
 518         itx->itx_oid = szp->z_id;
 519
 520         zil_itx_assign(zilog, itx, tx);
 521 }
 522
 523 /*
 524  * zfs_log_write() handles TX_WRITE transactions. The specified callback is
 525  * called as soon as the write is on stable storage (be it via a DMU sync or a
 526  * ZIL commit).
 527  */
 528 static long zfs_immediate_write_sz = 32768;
 529
 530 void
 531 zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 532     znode_t *zp, offset_t off, ssize_t resid, int ioflag,
 533     zil_callback_t callback, void *callback_data)
 534 {
 535         dmu_buf_impl_t *db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl);
 536         uint32_t blocksize = zp->z_blksz;
 537         itx_wr_state_t write_state;
 538         uintptr_t fsync_cnt;
 539         uint64_t gen = 0;
 540         ssize_t size = resid;
 541
 542         if (zil_replaying(zilog, tx) || zp->z_unlinked ||
 543             zfs_xattr_owner_unlinked(zp)) {
 544                 if (callback != NULL)
 545                         callback(callback_data);
 546                 return;
 547         }
 548
 549         if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
 550                 write_state = WR_INDIRECT;
 551         else if (!spa_has_slogs(zilog->zl_spa) &&
 552             resid >= zfs_immediate_write_sz)
 553                 write_state = WR_INDIRECT;
 554         else if (ioflag & (O_SYNC | O_DSYNC))
 555                 write_state = WR_COPIED;
 556         else
 557                 write_state = WR_NEED_COPY;
 558
 559         if ((fsync_cnt = (uintptr_t)tsd_get(zfs_fsyncer_key)) != 0) {
 560                 (void) tsd_set(zfs_fsyncer_key, (void *)(fsync_cnt - 1));
 561         }
 562
 563         (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &gen,
 564             sizeof (gen));
 565
 566         while (resid) {
 567                 itx_t *itx;
 568                 lr_write_t *lr;
 569                 itx_wr_state_t wr_state = write_state;
 570                 ssize_t len = resid;
 571
 572                 /*
 573                  * A WR_COPIED record must fit entirely in one log block.
 574                  * Large writes can use WR_NEED_COPY, which the ZIL will
 575                  * split into multiple records across several log blocks
 576                  * if necessary.
 577                  */
 578                 if (wr_state == WR_COPIED &&
 579                     resid > zil_max_copied_data(zilog))
 580                         wr_state = WR_NEED_COPY;
 581                 else if (wr_state == WR_INDIRECT)
 582                         len = MIN(blocksize - P2PHASE(off, blocksize), resid);
 583
 584                 itx = zil_itx_create(txtype, sizeof (*lr) +
 585                     (wr_state == WR_COPIED ? len : 0));
 586                 lr = (lr_write_t *)&itx->itx_lr;
 587
 588                 /*
 589                  * For WR_COPIED records, copy the data into the lr_write_t.
 590                  */
 591                 if (wr_state == WR_COPIED) {
 592                         int err;
 593                         DB_DNODE_ENTER(db);
 594                         err = dmu_read_by_dnode(DB_DNODE(db), off, len, lr + 1,
 595                             DMU_READ_NO_PREFETCH);
 596                         if (err != 0) {
 597                                 zil_itx_destroy(itx);
 598                                 itx = zil_itx_create(txtype, sizeof (*lr));
 599                                 lr = (lr_write_t *)&itx->itx_lr;
 600                                 wr_state = WR_NEED_COPY;
 601                         }
 602                         DB_DNODE_EXIT(db);
 603                 }
 604
 605                 itx->itx_wr_state = wr_state;
 606                 lr->lr_foid = zp->z_id;
 607                 lr->lr_offset = off;
 608                 lr->lr_length = len;
 609                 lr->lr_blkoff = 0;
 610                 BP_ZERO(&lr->lr_blkptr);
 611
 612                 itx->itx_private = ZTOZSB(zp);
 613                 itx->itx_gen = gen;
 614
 615                 if (!(ioflag & (O_SYNC | O_DSYNC)) && (zp->z_sync_cnt == 0) &&
 616                     (fsync_cnt == 0))
 617                         itx->itx_sync = B_FALSE;
 618
 619                 itx->itx_callback = callback;
 620                 itx->itx_callback_data = callback_data;
 621                 zil_itx_assign(zilog, itx, tx);
 622
 623                 off += len;
 624                 resid -= len;
 625         }
 626
 627         if (write_state == WR_COPIED || write_state == WR_NEED_COPY) {
 628                 dsl_pool_wrlog_count(zilog->zl_dmu_pool, size, tx->tx_txg);
 629         }
 630 }
 631
 632 /*
 633  * Handles TX_TRUNCATE transactions.
 634  */
 635 void
 636 zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 637     znode_t *zp, uint64_t off, uint64_t len)
 638 {
 639         itx_t *itx;
 640         lr_truncate_t *lr;
 641
 642         if (zil_replaying(zilog, tx) || zp->z_unlinked ||
 643             zfs_xattr_owner_unlinked(zp))
 644                 return;
 645
 646         itx = zil_itx_create(txtype, sizeof (*lr));
 647         lr = (lr_truncate_t *)&itx->itx_lr;
 648         lr->lr_foid = zp->z_id;
 649         lr->lr_offset = off;
 650         lr->lr_length = len;
 651
 652         itx->itx_sync = (zp->z_sync_cnt != 0);
 653         zil_itx_assign(zilog, itx, tx);
 654 }
 655
 656 /*
 657  * Handles TX_SETATTR transactions.
 658  */
 659 void
 660 zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 661     znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
 662 {
 663         itx_t           *itx;
 664         lr_setattr_t    *lr;
 665         xvattr_t        *xvap = (xvattr_t *)vap;
 666         size_t          recsize = sizeof (lr_setattr_t);
 667         void            *start;
 668
 669         if (zil_replaying(zilog, tx) || zp->z_unlinked)
 670                 return;
 671
 672         /*
 673          * If XVATTR set, then log record size needs to allow
 674          * for lr_attr_t + xvattr mask, mapsize and create time
 675          * plus actual attribute values
 676          */
 677         if (vap->va_mask & ATTR_XVATTR)
 678                 recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);
 679
 680         if (fuidp)
 681                 recsize += fuidp->z_domain_str_sz;
 682
 683         itx = zil_itx_create(txtype, recsize);
 684         lr = (lr_setattr_t *)&itx->itx_lr;
 685         lr->lr_foid = zp->z_id;
 686         lr->lr_mask = (uint64_t)mask_applied;
 687         lr->lr_mode = (uint64_t)vap->va_mode;
 688         if ((mask_applied & ATTR_UID) && IS_EPHEMERAL(vap->va_uid))
 689                 lr->lr_uid = fuidp->z_fuid_owner;
 690         else
 691                 lr->lr_uid = (uint64_t)vap->va_uid;
 692
 693         if ((mask_applied & ATTR_GID) && IS_EPHEMERAL(vap->va_gid))
 694                 lr->lr_gid = fuidp->z_fuid_group;
 695         else
 696                 lr->lr_gid = (uint64_t)vap->va_gid;
 697
 698         lr->lr_size = (uint64_t)vap->va_size;
 699         ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
 700         ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
 701         start = (lr_setattr_t *)(lr + 1);
 702         if (vap->va_mask & ATTR_XVATTR) {
 703                 zfs_log_xvattr((lr_attr_t *)start, xvap);
 704                 start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
 705         }
 706
 707         /*
 708          * Now stick on domain information if any on end
 709          */
 710
 711         if (fuidp)
 712                 (void) zfs_log_fuid_domains(fuidp, start);
 713
 714         itx->itx_sync = (zp->z_sync_cnt != 0);
 715         zil_itx_assign(zilog, itx, tx);
 716 }
 717
 718 /*
 719  * Handles TX_SETSAXATTR transactions.
 720  */
 721 void
 722 zfs_log_setsaxattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
 723     znode_t *zp, const char *name, const void *value, size_t size)
 724 {
 725         itx_t           *itx;
 726         lr_setsaxattr_t *lr;
 727         size_t          recsize = sizeof (lr_setsaxattr_t);
 728         void            *xattrstart;
 729         int             namelen;
 730
 731         if (zil_replaying(zilog, tx) || zp->z_unlinked)
 732                 return;
 733
 734         namelen = strlen(name) + 1;
 735         recsize += (namelen + size);
 736         itx = zil_itx_create(txtype, recsize);
 737         lr = (lr_setsaxattr_t *)&itx->itx_lr;
 738         lr->lr_foid = zp->z_id;
 739         xattrstart = (char *)(lr + 1);
 740         memcpy(xattrstart, name, namelen);
 741         if (value != NULL) {
 742                 memcpy((char *)xattrstart + namelen, value, size);
 743                 lr->lr_size = size;
 744         } else {
 745                 lr->lr_size = 0;
 746         }
 747
 748         itx->itx_sync = (zp->z_sync_cnt != 0);
 749         zil_itx_assign(zilog, itx, tx);
 750 }
 751
 752 /*
 753  * Handles TX_ACL transactions.
 754  */
 755 void
 756 zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
 757     vsecattr_t *vsecp, zfs_fuid_info_t *fuidp)
 758 {
 759         itx_t *itx;
 760         lr_acl_v0_t *lrv0;
 761         lr_acl_t *lr;
 762         int txtype;
 763         int lrsize;
 764         size_t txsize;
 765         size_t aclbytes = vsecp->vsa_aclentsz;
 766
 767         if (zil_replaying(zilog, tx) || zp->z_unlinked)
 768                 return;
 769
 770         txtype = (ZTOZSB(zp)->z_version < ZPL_VERSION_FUID) ?
 771             TX_ACL_V0 : TX_ACL;
 772
 773         if (txtype == TX_ACL)
 774                 lrsize = sizeof (*lr);
 775         else
 776                 lrsize = sizeof (*lrv0);
 777
 778         txsize = lrsize +
 779             ((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
 780             (fuidp ? fuidp->z_domain_str_sz : 0) +
 781             sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);
 782
 783         itx = zil_itx_create(txtype, txsize);
 784
 785         lr = (lr_acl_t *)&itx->itx_lr;
 786         lr->lr_foid = zp->z_id;
 787         if (txtype == TX_ACL) {
 788                 lr->lr_acl_bytes = aclbytes;
 789                 lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
 790                 lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
 791                 if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
 792                         lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
 793                 else
 794                         lr->lr_acl_flags = 0;
 795         }
 796         lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;
 797
 798         if (txtype == TX_ACL_V0) {
 799                 lrv0 = (lr_acl_v0_t *)lr;
 800                 memcpy(lrv0 + 1, vsecp->vsa_aclentp, aclbytes);
 801         } else {
 802                 void *start = (ace_t *)(lr + 1);
 803
 804                 memcpy(start, vsecp->vsa_aclentp, aclbytes);
 805
 806                 start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);
 807
 808                 if (fuidp) {
 809                         start = zfs_log_fuid_ids(fuidp, start);
 810                         (void) zfs_log_fuid_domains(fuidp, start);
 811                 }
 812         }
 813
 814         itx->itx_sync = (zp->z_sync_cnt != 0);
 815         zil_itx_assign(zilog, itx, tx);
 816 }
 817
 818 ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, LONG, ZMOD_RW,
 819         "Largest data block to write to zil");