Print this page
2882 implement libzfs_core
2883 changing "canmount" property to "on" should not always remount dataset
2900 "zfs snapshot" should be able to create multiple, arbitrary snapshots at once
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Chris Siden <christopher.siden@delphix.com>
Reviewed by: Garrett D'Amore <garrett@damore.org>
Reviewed by: Bill Pijewski <wdp@joyent.com>
Reviewed by: Dan Kruchinin <dan.kruchinin@gmail.com>


   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 /*
  23  * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Copyright (c) 2011 by Delphix. All rights reserved.
  25  */
  26 
  27 #include <sys/spa.h>
  28 #include <sys/spa_impl.h>
  29 #include <sys/zap.h>
  30 #include <sys/dsl_synctask.h>
  31 #include <sys/dmu_tx.h>
  32 #include <sys/dmu_objset.h>


  33 #include <sys/utsname.h>
  34 #include <sys/cmn_err.h>
  35 #include <sys/sunddi.h>

  36 #include "zfs_comutil.h"
  37 #ifdef _KERNEL
  38 #include <sys/zone.h>
  39 #endif
  40 
  41 /*
  42  * Routines to manage the on-disk history log.
  43  *
  44  * The history log is stored as a dmu object containing
  45  * <packed record length, record nvlist> tuples.
  46  *
  47  * Where "record nvlist" is a nvlist containing uint64_ts and strings, and
  48  * "packed record length" is the packed length of the "record nvlist" stored
  49  * as a little endian uint64_t.
  50  *
  51  * The log is implemented as a ring buffer, though the original creation
  52  * of the pool ('zpool create') is never overwritten.
  53  *
  54  * The history log is tracked as object 'spa_t::spa_history'.  The bonus buffer
  55  * of 'spa_history' stores the offsets for logging/retrieving history as


 159                         return (err);
 160                 }
 161         }
 162 
 163         phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
 164         firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
 165         shpp->sh_eof += len;
 166         dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);
 167 
 168         len -= firstwrite;
 169         if (len > 0) {
 170                 /* write out the rest at the beginning of physical file */
 171                 dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
 172                     len, (char *)buf + firstwrite, tx);
 173         }
 174 
 175         return (0);
 176 }
 177 
 178 static char *
 179 spa_history_zone()
 180 {
 181 #ifdef _KERNEL


 182         return (curproc->p_zone->zone_name);
 183 #else
 184         return ("global");
 185 #endif
 186 }
 187 
 188 /*
 189  * Write out a history event.
 190  */
 191 /*ARGSUSED*/
 192 static void
 193 spa_history_log_sync(void *arg1, void *arg2, dmu_tx_t *tx)
 194 {
 195         spa_t           *spa = arg1;
 196         history_arg_t   *hap = arg2;
 197         const char      *history_str = hap->ha_history_str;
 198         objset_t        *mos = spa->spa_meta_objset;
 199         dmu_buf_t       *dbp;
 200         spa_history_phys_t *shpp;
 201         size_t          reclen;
 202         uint64_t        le_len;
 203         nvlist_t        *nvrecord;
 204         char            *record_packed = NULL;
 205         int             ret;
 206 
 207         /*
 208          * If we have an older pool that doesn't have a command
 209          * history object, create it now.
 210          */
 211         mutex_enter(&spa->spa_history_lock);
 212         if (!spa->spa_history)
 213                 spa_history_create_obj(spa, tx);
 214         mutex_exit(&spa->spa_history_lock);
 215 
 216         /*
 217          * Get the offset of where we need to write via the bonus buffer.
 218          * Update the offset when the write completes.
 219          */
 220         VERIFY(0 == dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
 221         shpp = dbp->db_data;
 222 
 223         dmu_buf_will_dirty(dbp, tx);
 224 
 225 #ifdef ZFS_DEBUG
 226         {
 227                 dmu_object_info_t doi;
 228                 dmu_object_info_from_db(dbp, &doi);
 229                 ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
 230         }
 231 #endif
 232 
 233         VERIFY(nvlist_alloc(&nvrecord, NV_UNIQUE_NAME, KM_SLEEP) == 0);
 234         VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_TIME,
 235             gethrestime_sec()) == 0);
 236         VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_WHO, hap->ha_uid) == 0);
 237         if (hap->ha_zone != NULL)
 238                 VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_ZONE,
 239                     hap->ha_zone) == 0);
 240 #ifdef _KERNEL
 241         VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_HOST,
 242             utsname.nodename) == 0);
 243 #endif
 244         if (hap->ha_log_type == LOG_CMD_POOL_CREATE ||
 245             hap->ha_log_type == LOG_CMD_NORMAL) {
 246                 VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_CMD,
 247                     history_str) == 0);
 248 
 249                 zfs_dbgmsg("command: %s", history_str);





 250         } else {
 251                 VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_INT_EVENT,
 252                     hap->ha_event) == 0);
 253                 VERIFY(nvlist_add_uint64(nvrecord, ZPOOL_HIST_TXG,
 254                     tx->tx_txg) == 0);
 255                 VERIFY(nvlist_add_string(nvrecord, ZPOOL_HIST_INT_STR,
 256                     history_str) == 0);
 257 
 258                 zfs_dbgmsg("internal %s pool:%s txg:%llu %s",
 259                     zfs_history_event_names[hap->ha_event], spa_name(spa),
 260                     (longlong_t)tx->tx_txg, history_str);
 261 
 262         }
 263 
 264         VERIFY(nvlist_size(nvrecord, &reclen, NV_ENCODE_XDR) == 0);
 265         record_packed = kmem_alloc(reclen, KM_SLEEP);
 266 
 267         VERIFY(nvlist_pack(nvrecord, &record_packed, &reclen,
 268             NV_ENCODE_XDR, KM_SLEEP) == 0);
 269 
 270         mutex_enter(&spa->spa_history_lock);
 271         if (hap->ha_log_type == LOG_CMD_POOL_CREATE)
 272                 VERIFY(shpp->sh_eof == shpp->sh_pool_create_len);
 273 
 274         /* write out the packed length as little endian */
 275         le_len = LE_64((uint64_t)reclen);
 276         ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
 277         if (!ret)
 278                 ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
 279 
 280         if (!ret && hap->ha_log_type == LOG_CMD_POOL_CREATE) {
 281                 shpp->sh_pool_create_len += sizeof (le_len) + reclen;
 282                 shpp->sh_bof = shpp->sh_pool_create_len;

 283         }
 284 
 285         mutex_exit(&spa->spa_history_lock);
 286         nvlist_free(nvrecord);
 287         kmem_free(record_packed, reclen);
 288         dmu_buf_rele(dbp, FTAG);
 289 
 290         strfree(hap->ha_history_str);
 291         if (hap->ha_zone != NULL)
 292                 strfree(hap->ha_zone);
 293         kmem_free(hap, sizeof (history_arg_t));
 294 }
 295 
 296 /*
 297  * Write out a history event.
 298  */
 299 int
 300 spa_history_log(spa_t *spa, const char *history_str, history_log_type_t what)












 301 {
 302         history_arg_t *ha;
 303         int err = 0;
 304         dmu_tx_t *tx;

 305 
 306         ASSERT(what != LOG_INTERNAL);

 307 
 308         tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
 309         err = dmu_tx_assign(tx, TXG_WAIT);
 310         if (err) {
 311                 dmu_tx_abort(tx);
 312                 return (err);
 313         }
 314 
 315         ha = kmem_alloc(sizeof (history_arg_t), KM_SLEEP);
 316         ha->ha_history_str = strdup(history_str);
 317         ha->ha_zone = strdup(spa_history_zone());
 318         ha->ha_log_type = what;
 319         ha->ha_uid = crgetuid(CRED());

 320 
 321         /* Kick this off asynchronously; errors are ignored. */
 322         dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
 323             spa_history_log_sync, spa, ha, 0, tx);
 324         dmu_tx_commit(tx);
 325 
 326         /* spa_history_log_sync will free ha and strings */
 327         return (err);

 328 }
 329 
 330 /*
 331  * Read out the command history.
 332  */
 333 int
 334 spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
 335 {
 336         objset_t *mos = spa->spa_meta_objset;
 337         dmu_buf_t *dbp;
 338         uint64_t read_len, phys_read_off, phys_eof;
 339         uint64_t leftover = 0;
 340         spa_history_phys_t *shpp;
 341         int err;
 342 
 343         /*
 344          * If the command history  doesn't exist (older pool),
 345          * that's ok, just return ENOENT.
 346          */
 347         if (!spa->spa_history)


 407         *len = read_len + leftover;
 408 
 409         if (read_len == 0) {
 410                 mutex_exit(&spa->spa_history_lock);
 411                 dmu_buf_rele(dbp, FTAG);
 412                 return (0);
 413         }
 414 
 415         err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
 416             DMU_READ_PREFETCH);
 417         if (leftover && err == 0) {
 418                 err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
 419                     leftover, buf + read_len, DMU_READ_PREFETCH);
 420         }
 421         mutex_exit(&spa->spa_history_lock);
 422 
 423         dmu_buf_rele(dbp, FTAG);
 424         return (err);
 425 }
 426 



 427 static void
 428 log_internal(history_internal_events_t event, spa_t *spa,
 429     dmu_tx_t *tx, const char *fmt, va_list adx)
 430 {
 431         history_arg_t *ha;
 432 
 433         /*
 434          * If this is part of creating a pool, not everything is
 435          * initialized yet, so don't bother logging the internal events.
 436          */
 437         if (tx->tx_txg == TXG_INITIAL)
 438                 return;
 439 
 440         ha = kmem_alloc(sizeof (history_arg_t), KM_SLEEP);
 441         ha->ha_history_str = kmem_alloc(vsnprintf(NULL, 0, fmt, adx) + 1,
 442             KM_SLEEP);
 443 
 444         (void) vsprintf(ha->ha_history_str, fmt, adx);
 445 
 446         ha->ha_log_type = LOG_INTERNAL;
 447         ha->ha_event = event;
 448         ha->ha_zone = NULL;
 449         ha->ha_uid = 0;
 450 
 451         if (dmu_tx_is_syncing(tx)) {
 452                 spa_history_log_sync(spa, ha, tx);
 453         } else {
 454                 dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
 455                     spa_history_log_sync, spa, ha, 0, tx);
 456         }
 457         /* spa_history_log_sync() will free ha and strings */
 458 }
 459 
 460 void
 461 spa_history_log_internal(history_internal_events_t event, spa_t *spa,
 462     dmu_tx_t *tx, const char *fmt, ...)
 463 {
 464         dmu_tx_t *htx = tx;
 465         va_list adx;
 466 
 467         /* create a tx if we didn't get one */
 468         if (tx == NULL) {
 469                 htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
 470                 if (dmu_tx_assign(htx, TXG_WAIT) != 0) {
 471                         dmu_tx_abort(htx);
 472                         return;
 473                 }
 474         }
 475 
 476         va_start(adx, fmt);
 477         log_internal(event, spa, htx, fmt, adx);
 478         va_end(adx);
 479 
 480         /* if we didn't get a tx from the caller, commit the one we made */
 481         if (tx == NULL)
 482                 dmu_tx_commit(htx);
 483 }
 484 
 485 void
 486 spa_history_log_version(spa_t *spa, history_internal_events_t event)







































 487 {
 488 #ifdef _KERNEL
 489         uint64_t current_vers = spa_version(spa);
 490 
 491         if (current_vers >= SPA_VERSION_ZPOOL_HISTORY) {
 492                 spa_history_log_internal(event, spa, NULL,
 493                     "pool spa %llu; zfs spa %llu; zpl %d; uts %s %s %s %s",
 494                     (u_longlong_t)current_vers, SPA_VERSION, ZPL_VERSION,
 495                     utsname.nodename, utsname.release, utsname.version,
 496                     utsname.machine);
 497         }
 498         cmn_err(CE_CONT, "!%s version %llu pool %s using %llu",
 499             event == LOG_POOL_IMPORT ? "imported" :
 500             event == LOG_POOL_CREATE ? "created" : "accessed",
 501             (u_longlong_t)current_vers, spa_name(spa), SPA_VERSION);
 502 #endif
 503 }


   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 /*
  23  * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Copyright (c) 2012 by Delphix. All rights reserved.
  25  */
  26 
  27 #include <sys/spa.h>
  28 #include <sys/spa_impl.h>
  29 #include <sys/zap.h>
  30 #include <sys/dsl_synctask.h>
  31 #include <sys/dmu_tx.h>
  32 #include <sys/dmu_objset.h>
  33 #include <sys/dsl_dataset.h>
  34 #include <sys/dsl_dir.h>
  35 #include <sys/utsname.h>
  36 #include <sys/cmn_err.h>
  37 #include <sys/sunddi.h>
  38 #include <sys/cred.h>
  39 #include "zfs_comutil.h"
  40 #ifdef _KERNEL
  41 #include <sys/zone.h>
  42 #endif
  43 
  44 /*
  45  * Routines to manage the on-disk history log.
  46  *
  47  * The history log is stored as a dmu object containing
  48  * <packed record length, record nvlist> tuples.
  49  *
  50  * Where "record nvlist" is a nvlist containing uint64_ts and strings, and
  51  * "packed record length" is the packed length of the "record nvlist" stored
  52  * as a little endian uint64_t.
  53  *
  54  * The log is implemented as a ring buffer, though the original creation
  55  * of the pool ('zpool create') is never overwritten.
  56  *
  57  * The history log is tracked as object 'spa_t::spa_history'.  The bonus buffer
  58  * of 'spa_history' stores the offsets for logging/retrieving history as


 162                         return (err);
 163                 }
 164         }
 165 
 166         phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
 167         firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
 168         shpp->sh_eof += len;
 169         dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);
 170 
 171         len -= firstwrite;
 172         if (len > 0) {
 173                 /* write out the rest at the beginning of physical file */
 174                 dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
 175                     len, (char *)buf + firstwrite, tx);
 176         }
 177 
 178         return (0);
 179 }
 180 
 181 static char *
 182 spa_history_zone(void)
 183 {
 184 #ifdef _KERNEL
 185         if (INGLOBALZONE(curproc))
 186                 return (NULL);
 187         return (curproc->p_zone->zone_name);
 188 #else
 189         return (NULL);
 190 #endif
 191 }
 192 
 193 /*
 194  * Write out a history event.
 195  */
 196 /*ARGSUSED*/
 197 static void
 198 spa_history_log_sync(void *arg1, void *arg2, dmu_tx_t *tx)
 199 {
 200         spa_t           *spa = arg1;
 201         nvlist_t        *nvl = arg2;

 202         objset_t        *mos = spa->spa_meta_objset;
 203         dmu_buf_t       *dbp;
 204         spa_history_phys_t *shpp;
 205         size_t          reclen;
 206         uint64_t        le_len;

 207         char            *record_packed = NULL;
 208         int             ret;
 209 
 210         /*
 211          * If we have an older pool that doesn't have a command
 212          * history object, create it now.
 213          */
 214         mutex_enter(&spa->spa_history_lock);
 215         if (!spa->spa_history)
 216                 spa_history_create_obj(spa, tx);
 217         mutex_exit(&spa->spa_history_lock);
 218 
 219         /*
 220          * Get the offset of where we need to write via the bonus buffer.
 221          * Update the offset when the write completes.
 222          */
 223         VERIFY(0 == dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
 224         shpp = dbp->db_data;
 225 
 226         dmu_buf_will_dirty(dbp, tx);
 227 
 228 #ifdef ZFS_DEBUG
 229         {
 230                 dmu_object_info_t doi;
 231                 dmu_object_info_from_db(dbp, &doi);
 232                 ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
 233         }
 234 #endif
 235 
 236         fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec());






 237 #ifdef _KERNEL
 238         fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname.nodename);

 239 #endif
 240         if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
 241                 zfs_dbgmsg("command: %s",
 242                     fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD));
 243         } else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) {
 244                 if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) {
 245                         zfs_dbgmsg("txg %lld %s %s (id %llu) %s",
 246                             fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
 247                             fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
 248                             fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME),
 249                             fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID),
 250                             fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
 251                 } else {
 252                         zfs_dbgmsg("txg %lld %s %s",
 253                             fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
 254                             fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
 255                             fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
 256                 }
 257         } else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) {
 258                 zfs_dbgmsg("ioctl %s",
 259                     fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL));



 260         }
 261 
 262         record_packed = fnvlist_pack(nvl, &reclen);




 263 
 264         mutex_enter(&spa->spa_history_lock);


 265 
 266         /* write out the packed length as little endian */
 267         le_len = LE_64((uint64_t)reclen);
 268         ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
 269         if (!ret)
 270                 ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
 271 
 272         /* The first command is the create, which we keep forever */
 273         if (ret == 0 && shpp->sh_pool_create_len == 0 &&
 274             nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
 275                 shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof;
 276         }
 277 
 278         mutex_exit(&spa->spa_history_lock);
 279         fnvlist_pack_free(record_packed, reclen);

 280         dmu_buf_rele(dbp, FTAG);
 281         fnvlist_free(nvl);




 282 }
 283 
 284 /*
 285  * Write out a history event.
 286  */
 287 int
 288 spa_history_log(spa_t *spa, const char *msg)
 289 {
 290         int err;
 291         nvlist_t *nvl = fnvlist_alloc();
 292 
 293         fnvlist_add_string(nvl, ZPOOL_HIST_CMD, msg);
 294         err = spa_history_log_nvl(spa, nvl);
 295         fnvlist_free(nvl);
 296         return (err);
 297 }
 298 
 299 int
 300 spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
 301 {

 302         int err = 0;
 303         dmu_tx_t *tx;
 304         nvlist_t *nvarg;
 305 
 306         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY)
 307                 return (EINVAL);
 308 
 309         tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
 310         err = dmu_tx_assign(tx, TXG_WAIT);
 311         if (err) {
 312                 dmu_tx_abort(tx);
 313                 return (err);
 314         }
 315 
 316         nvarg = fnvlist_dup(nvl);
 317         if (spa_history_zone() != NULL) {
 318                 fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
 319                     spa_history_zone());
 320         }
 321         fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
 322 
 323         /* Kick this off asynchronously; errors are ignored. */
 324         dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
 325             spa_history_log_sync, spa, nvarg, 0, tx);
 326         dmu_tx_commit(tx);
 327 
 328         /* spa_history_log_sync will free nvl */
 329         return (err);
 330 
 331 }
 332 
 333 /*
 334  * Read out the command history.
 335  */
 336 int
 337 spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
 338 {
 339         objset_t *mos = spa->spa_meta_objset;
 340         dmu_buf_t *dbp;
 341         uint64_t read_len, phys_read_off, phys_eof;
 342         uint64_t leftover = 0;
 343         spa_history_phys_t *shpp;
 344         int err;
 345 
 346         /*
 347          * If the command history doesn't exist (older pool),
 348          * that's ok, just return ENOENT.
 349          */
 350         if (!spa->spa_history)


 410         *len = read_len + leftover;
 411 
 412         if (read_len == 0) {
 413                 mutex_exit(&spa->spa_history_lock);
 414                 dmu_buf_rele(dbp, FTAG);
 415                 return (0);
 416         }
 417 
 418         err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
 419             DMU_READ_PREFETCH);
 420         if (leftover && err == 0) {
 421                 err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
 422                     leftover, buf + read_len, DMU_READ_PREFETCH);
 423         }
 424         mutex_exit(&spa->spa_history_lock);
 425 
 426         dmu_buf_rele(dbp, FTAG);
 427         return (err);
 428 }
 429 
 430 /*
 431  * The nvlist will be consumed by this call.
 432  */
 433 static void
 434 log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
 435     dmu_tx_t *tx, const char *fmt, va_list adx)
 436 {
 437         char *msg;
 438 
 439         /*
 440          * If this is part of creating a pool, not everything is
 441          * initialized yet, so don't bother logging the internal events.
 442          */
 443         if (tx->tx_txg == TXG_INITIAL)
 444                 return;
 445 
 446         msg = kmem_alloc(vsnprintf(NULL, 0, fmt, adx) + 1, KM_SLEEP);
 447         (void) vsprintf(msg, fmt, adx);
 448         fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
 449         strfree(msg);
 450 
 451         fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
 452         fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);



 453 
 454         if (dmu_tx_is_syncing(tx)) {
 455                 spa_history_log_sync(spa, nvl, tx);
 456         } else {
 457                 dsl_sync_task_do_nowait(spa_get_dsl(spa), NULL,
 458                     spa_history_log_sync, spa, nvl, 0, tx);
 459         }
 460         /* spa_history_log_sync() will free nvl */
 461 }
 462 
 463 void
 464 spa_history_log_internal(spa_t *spa, const char *operation,
 465     dmu_tx_t *tx, const char *fmt, ...)
 466 {
 467         dmu_tx_t *htx = tx;
 468         va_list adx;
 469 
 470         /* create a tx if we didn't get one */
 471         if (tx == NULL) {
 472                 htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
 473                 if (dmu_tx_assign(htx, TXG_WAIT) != 0) {
 474                         dmu_tx_abort(htx);
 475                         return;
 476                 }
 477         }
 478 
 479         va_start(adx, fmt);
 480         log_internal(fnvlist_alloc(), operation, spa, htx, fmt, adx);
 481         va_end(adx);
 482 
 483         /* if we didn't get a tx from the caller, commit the one we made */
 484         if (tx == NULL)
 485                 dmu_tx_commit(htx);
 486 }
 487 
 488 void
 489 spa_history_log_internal_ds(dsl_dataset_t *ds, const char *operation,
 490     dmu_tx_t *tx, const char *fmt, ...)
 491 {
 492         va_list adx;
 493         char namebuf[MAXNAMELEN];
 494         nvlist_t *nvl = fnvlist_alloc();
 495 
 496         ASSERT(tx != NULL);
 497 
 498         dsl_dataset_name(ds, namebuf);
 499         fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
 500         fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID, ds->ds_object);
 501 
 502         va_start(adx, fmt);
 503         log_internal(nvl, operation, dsl_dataset_get_spa(ds), tx, fmt, adx);
 504         va_end(adx);
 505 }
 506 
 507 void
 508 spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
 509     dmu_tx_t *tx, const char *fmt, ...)
 510 {
 511         va_list adx;
 512         char namebuf[MAXNAMELEN];
 513         nvlist_t *nvl = fnvlist_alloc();
 514 
 515         ASSERT(tx != NULL);
 516 
 517         dsl_dir_name(dd, namebuf);
 518         fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
 519         fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID,
 520             dd->dd_phys->dd_head_dataset_obj);
 521 
 522         va_start(adx, fmt);
 523         log_internal(nvl, operation, dd->dd_pool->dp_spa, tx, fmt, adx);
 524         va_end(adx);
 525 }
 526 
 527 void
 528 spa_history_log_version(spa_t *spa, const char *operation)
 529 {
 530 #ifdef _KERNEL
 531         uint64_t current_vers = spa_version(spa);
 532 
 533         spa_history_log_internal(spa, operation, NULL,
 534             "pool version %llu; software version %llu/%d; uts %s %s %s %s",

 535             (u_longlong_t)current_vers, SPA_VERSION, ZPL_VERSION,
 536             utsname.nodename, utsname.release, utsname.version,
 537             utsname.machine);
 538         cmn_err(CE_CONT, "!%s version %llu pool %s using %llu", operation,



 539             (u_longlong_t)current_vers, spa_name(spa), SPA_VERSION);
 540 #endif
 541 }