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