File Manager

003 File Manager

Current Path: /usr/src/sys/contrib/openzfs/module/zfs

usr / src / sys / contrib / openzfs / module / zfs /

📁 ..
📄 Makefile.in(4.57 KB)
📄 THIRDPARTYLICENSE.cityhash(1.03 KB)
📄 THIRDPARTYLICENSE.cityhash.descrip(39 B)
📄 abd.c(31.72 KB)
📄 aggsum.c(8.46 KB)
📄 arc.c(330.91 KB)
📄 blkptr.c(4.26 KB)
📄 bplist.c(2.38 KB)
📄 bpobj.c(27.82 KB)
📄 bptree.c(8.2 KB)
📄 bqueue.c(4.7 KB)
📄 btree.c(65.19 KB)
📄 dataset_kstats.c(6.32 KB)
📄 dbuf.c(141.84 KB)
📄 dbuf_stats.c(6.05 KB)
📄 ddt.c(28.01 KB)
📄 ddt_zap.c(4.38 KB)
📄 dmu.c(60.44 KB)
📄 dmu_diff.c(6.25 KB)
📄 dmu_object.c(14.73 KB)
📄 dmu_objset.c(79.38 KB)
📄 dmu_recv.c(98.69 KB)
📄 dmu_redact.c(37.09 KB)
📄 dmu_send.c(95.51 KB)
📄 dmu_traverse.c(21.98 KB)
📄 dmu_tx.c(40.13 KB)
📄 dmu_zfetch.c(13.56 KB)
📄 dnode.c(71.32 KB)
📄 dnode_sync.c(24.93 KB)
📄 dsl_bookmark.c(51.22 KB)
📄 dsl_crypt.c(78.06 KB)
📄 dsl_dataset.c(141.61 KB)
📄 dsl_deadlist.c(27.94 KB)
📄 dsl_deleg.c(19.92 KB)
📄 dsl_destroy.c(36.73 KB)
📄 dsl_dir.c(65.45 KB)
📄 dsl_pool.c(42.92 KB)
📄 dsl_prop.c(32.88 KB)
📄 dsl_scan.c(134.47 KB)
📄 dsl_synctask.c(7.94 KB)
📄 dsl_userhold.c(18.26 KB)
📄 edonr_zfs.c(3.21 KB)
📄 fm.c(39.75 KB)
📄 gzip.c(2.57 KB)
📄 hkdf.c(4.5 KB)
📄 lz4.c(26.46 KB)
📄 lzjb.c(3.85 KB)
📄 metaslab.c(190.62 KB)
📄 mmp.c(24.93 KB)
📄 multilist.c(12.21 KB)
📄 objlist.c(2.58 KB)
📄 pathname.c(2.52 KB)
📄 range_tree.c(24.9 KB)
📄 refcount.c(7.82 KB)
📄 rrwlock.c(10.82 KB)
📄 sa.c(59.23 KB)
📄 sha256.c(2.89 KB)
📄 skein_zfs.c(2.92 KB)
📄 spa.c(272.81 KB)
📄 spa_boot.c(1.28 KB)
📄 spa_checkpoint.c(21.54 KB)
📄 spa_config.c(16.97 KB)
📄 spa_errlog.c(11.35 KB)
📄 spa_history.c(17.75 KB)
📄 spa_log_spacemap.c(47.13 KB)
📄 spa_misc.c(75.4 KB)
📄 spa_stats.c(27.08 KB)
📄 space_map.c(31.45 KB)
📄 space_reftree.c(4.21 KB)
📄 txg.c(27.49 KB)
📄 uberblock.c(2.17 KB)
📄 unique.c(2.53 KB)
📄 vdev.c(143.91 KB)
📄 vdev_cache.c(11.71 KB)
📄 vdev_draid.c(96.86 KB)
📄 vdev_draid_rand.c(1.16 KB)
📄 vdev_indirect.c(60.83 KB)
📄 vdev_indirect_births.c(5.88 KB)
📄 vdev_indirect_mapping.c(18.06 KB)
📄 vdev_initialize.c(22.38 KB)
📄 vdev_label.c(57.52 KB)
📄 vdev_mirror.c(26.76 KB)
📄 vdev_missing.c(3.74 KB)
📄 vdev_queue.c(37.61 KB)
📄 vdev_raidz.c(76.09 KB)
📄 vdev_raidz_math.c(17.18 KB)
📄 vdev_raidz_math_aarch64_neon.c(112.13 KB)
📄 vdev_raidz_math_aarch64_neon_common.h(23.64 KB)
📄 vdev_raidz_math_aarch64_neonx2.c(5.39 KB)
📄 vdev_raidz_math_avx2.c(10.64 KB)
📄 vdev_raidz_math_avx512bw.c(10.93 KB)
📄 vdev_raidz_math_avx512f.c(17.9 KB)
📄 vdev_raidz_math_impl.h(34.78 KB)
📄 vdev_raidz_math_powerpc_altivec.c(213.46 KB)
📄 vdev_raidz_math_powerpc_altivec_common.h(22.62 KB)
📄 vdev_raidz_math_scalar.c(9.54 KB)
📄 vdev_raidz_math_sse2.c(25.84 KB)
📄 vdev_raidz_math_ssse3.c(118 KB)
📄 vdev_rebuild.c(34.65 KB)
📄 vdev_removal.c(72.08 KB)
📄 vdev_root.c(4.28 KB)
📄 vdev_trim.c(50.92 KB)
📄 zap.c(35.35 KB)
📄 zap_leaf.c(22.75 KB)
📄 zap_micro.c(42.18 KB)
📄 zcp.c(42.29 KB)
📄 zcp_get.c(21.01 KB)
📄 zcp_global.c(1.9 KB)
📄 zcp_iter.c(18.21 KB)
📄 zcp_set.c(2.37 KB)
📄 zcp_synctask.c(13.69 KB)
📄 zfeature.c(17.6 KB)
📄 zfs_byteswap.c(5.64 KB)
📄 zfs_fm.c(40.58 KB)
📄 zfs_fuid.c(19.69 KB)
📄 zfs_ioctl.c(195.48 KB)
📄 zfs_log.c(20.6 KB)
📄 zfs_onexit.c(5.09 KB)
📄 zfs_quota.c(12.93 KB)
📄 zfs_ratelimit.c(2.35 KB)
📄 zfs_replay.c(26.11 KB)
📄 zfs_rlock.c(20.38 KB)
📄 zfs_sa.c(13.09 KB)
📄 zfs_vnops.c(22 KB)
📄 zil.c(111.32 KB)
📄 zio.c(144.72 KB)
📄 zio_checksum.c(17.28 KB)
📄 zio_compress.c(5.92 KB)
📄 zio_inject.c(25.79 KB)
📄 zle.c(2.52 KB)
📄 zrlock.c(4.45 KB)
📄 zthr.c(16.41 KB)
📄 zvol.c(42.96 KB)

Editing: spa_history.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
 * Copyright (c) 2014 Integros [integros.com]
 * Copyright 2017 Joyent, Inc.
 */

#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/zap.h>
#include <sys/dsl_synctask.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/cmn_err.h>
#include <sys/sunddi.h>
#include <sys/cred.h>
#include "zfs_comutil.h"
#include "zfs_gitrev.h"
#ifdef _KERNEL
#include <sys/zone.h>
#endif

/*
 * Routines to manage the on-disk history log.
 *
 * The history log is stored as a dmu object containing
 * <packed record length, record nvlist> tuples.
 *
 * Where "record nvlist" is an nvlist containing uint64_ts and strings, and
 * "packed record length" is the packed length of the "record nvlist" stored
 * as a little endian uint64_t.
 *
 * The log is implemented as a ring buffer, though the original creation
 * of the pool ('zpool create') is never overwritten.
 *
 * The history log is tracked as object 'spa_t::spa_history'.  The bonus buffer
 * of 'spa_history' stores the offsets for logging/retrieving history as
 * 'spa_history_phys_t'.  'sh_pool_create_len' is the ending offset in bytes of
 * where the 'zpool create' record is stored.  This allows us to never
 * overwrite the original creation of the pool.  'sh_phys_max_off' is the
 * physical ending offset in bytes of the log.  This tells you the length of
 * the buffer. 'sh_eof' is the logical EOF (in bytes).  Whenever a record
 * is added, 'sh_eof' is incremented by the size of the record.
 * 'sh_eof' is never decremented.  'sh_bof' is the logical BOF (in bytes).
 * This is where the consumer should start reading from after reading in
 * the 'zpool create' portion of the log.
 *
 * 'sh_records_lost' keeps track of how many records have been overwritten
 * and permanently lost.
 */

/* convert a logical offset to physical */
static uint64_t
spa_history_log_to_phys(uint64_t log_off, spa_history_phys_t *shpp)
{
	uint64_t phys_len;

phys_len = shpp->sh_phys_max_off - shpp->sh_pool_create_len;
	return ((log_off - shpp->sh_pool_create_len) % phys_len
	    + shpp->sh_pool_create_len);
}

void
spa_history_create_obj(spa_t *spa, dmu_tx_t *tx)
{
	dmu_buf_t *dbp;
	spa_history_phys_t *shpp;
	objset_t *mos = spa->spa_meta_objset;

ASSERT0(spa->spa_history);
	spa->spa_history = dmu_object_alloc(mos, DMU_OT_SPA_HISTORY,
	    SPA_OLD_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS,
	    sizeof (spa_history_phys_t), tx);

VERIFY0(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
	    DMU_POOL_HISTORY, sizeof (uint64_t), 1,
	    &spa->spa_history, tx));

VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
	ASSERT3U(dbp->db_size, >=, sizeof (spa_history_phys_t));

shpp = dbp->db_data;
	dmu_buf_will_dirty(dbp, tx);

/*
	 * Figure out maximum size of history log.  We set it at
	 * 0.1% of pool size, with a max of 1G and min of 128KB.
	 */
	shpp->sh_phys_max_off =
	    metaslab_class_get_dspace(spa_normal_class(spa)) / 1000;
	shpp->sh_phys_max_off = MIN(shpp->sh_phys_max_off, 1<<30);
	shpp->sh_phys_max_off = MAX(shpp->sh_phys_max_off, 128<<10);

dmu_buf_rele(dbp, FTAG);
}

/*
 * Change 'sh_bof' to the beginning of the next record.
 */
static int
spa_history_advance_bof(spa_t *spa, spa_history_phys_t *shpp)
{
	objset_t *mos = spa->spa_meta_objset;
	uint64_t firstread, reclen, phys_bof;
	char buf[sizeof (reclen)];
	int err;

phys_bof = spa_history_log_to_phys(shpp->sh_bof, shpp);
	firstread = MIN(sizeof (reclen), shpp->sh_phys_max_off - phys_bof);

if ((err = dmu_read(mos, spa->spa_history, phys_bof, firstread,
	    buf, DMU_READ_PREFETCH)) != 0)
		return (err);
	if (firstread != sizeof (reclen)) {
		if ((err = dmu_read(mos, spa->spa_history,
		    shpp->sh_pool_create_len, sizeof (reclen) - firstread,
		    buf + firstread, DMU_READ_PREFETCH)) != 0)
			return (err);
	}

reclen = LE_64(*((uint64_t *)buf));
	shpp->sh_bof += reclen + sizeof (reclen);
	shpp->sh_records_lost++;
	return (0);
}

static int
spa_history_write(spa_t *spa, void *buf, uint64_t len, spa_history_phys_t *shpp,
    dmu_tx_t *tx)
{
	uint64_t firstwrite, phys_eof;
	objset_t *mos = spa->spa_meta_objset;
	int err;

ASSERT(MUTEX_HELD(&spa->spa_history_lock));

/* see if we need to reset logical BOF */
	while (shpp->sh_phys_max_off - shpp->sh_pool_create_len -
	    (shpp->sh_eof - shpp->sh_bof) <= len) {
		if ((err = spa_history_advance_bof(spa, shpp)) != 0) {
			return (err);
		}
	}

phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
	firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
	shpp->sh_eof += len;
	dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);

len -= firstwrite;
	if (len > 0) {
		/* write out the rest at the beginning of physical file */
		dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
		    len, (char *)buf + firstwrite, tx);
	}

return (0);
}

/*
 * Post a history sysevent.
 *
 * The nvlist_t* passed into this function will be transformed into a new
 * nvlist where:
 *
 * 1. Nested nvlists will be flattened to a single level
 * 2. Keys will have their names normalized (to remove any problematic
 * characters, such as whitespace)
 *
 * The nvlist_t passed into this function will duplicated and should be freed
 * by caller.
 *
 */
static void
spa_history_log_notify(spa_t *spa, nvlist_t *nvl)
{
	nvlist_t *hist_nvl = fnvlist_alloc();
	uint64_t uint64;
	char *string;

if (nvlist_lookup_string(nvl, ZPOOL_HIST_CMD, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_CMD, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_NAME, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_ZONE, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_ZONE, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_HOST, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_HOST, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_DSNAME, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_STR, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_IOCTL, string);

if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME, &string) == 0)
		fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_NAME, string);

if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID, &uint64) == 0)
		fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_DSID, uint64);

if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG, &uint64) == 0)
		fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_TXG, uint64);

if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_TIME, &uint64) == 0)
		fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_TIME, uint64);

if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_WHO, &uint64) == 0)
		fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_WHO, uint64);

if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_INT_EVENT, &uint64) == 0)
		fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_INT_EVENT, uint64);

spa_event_notify(spa, NULL, hist_nvl, ESC_ZFS_HISTORY_EVENT);

nvlist_free(hist_nvl);
}

/*
 * Write out a history event.
 */
/*ARGSUSED*/
static void
spa_history_log_sync(void *arg, dmu_tx_t *tx)
{
	nvlist_t	*nvl = arg;
	spa_t		*spa = dmu_tx_pool(tx)->dp_spa;
	objset_t	*mos = spa->spa_meta_objset;
	dmu_buf_t	*dbp;
	spa_history_phys_t *shpp;
	size_t		reclen;
	uint64_t	le_len;
	char		*record_packed = NULL;
	int		ret;

/*
	 * If we have an older pool that doesn't have a command
	 * history object, create it now.
	 */
	mutex_enter(&spa->spa_history_lock);
	if (!spa->spa_history)
		spa_history_create_obj(spa, tx);
	mutex_exit(&spa->spa_history_lock);

/*
	 * Get the offset of where we need to write via the bonus buffer.
	 * Update the offset when the write completes.
	 */
	VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
	shpp = dbp->db_data;

dmu_buf_will_dirty(dbp, tx);

#ifdef ZFS_DEBUG
	{
		dmu_object_info_t doi;
		dmu_object_info_from_db(dbp, &doi);
		ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
	}
#endif

fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec());
	fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname()->nodename);

if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
		zfs_dbgmsg("command: %s",
		    fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD));
	} else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) {
		if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) {
			zfs_dbgmsg("txg %lld %s %s (id %llu) %s",
			    fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
			    fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
			    fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME),
			    fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID),
			    fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
		} else {
			zfs_dbgmsg("txg %lld %s %s",
			    fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
			    fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
			    fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
		}
		/*
		 * The history sysevent is posted only for internal history
		 * messages to show what has happened, not how it happened. For
		 * example, the following command:
		 *
		 * # zfs destroy -r tank/foo
		 *
		 * will result in one sysevent posted per dataset that is
		 * destroyed as a result of the command - which could be more
		 * than one event in total.  By contrast, if the sysevent was
		 * posted as a result of the ZPOOL_HIST_CMD key being present
		 * it would result in only one sysevent being posted with the
		 * full command line arguments, requiring the consumer to know
		 * how to parse and understand zfs(8) command invocations.
		 */
		spa_history_log_notify(spa, nvl);
	} else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) {
		zfs_dbgmsg("ioctl %s",
		    fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL));
	}

VERIFY3U(nvlist_pack(nvl, &record_packed, &reclen, NV_ENCODE_NATIVE,
	    KM_SLEEP), ==, 0);

mutex_enter(&spa->spa_history_lock);

/* write out the packed length as little endian */
	le_len = LE_64((uint64_t)reclen);
	ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
	if (!ret)
		ret = spa_history_write(spa, record_packed, reclen, shpp, tx);

/* The first command is the create, which we keep forever */
	if (ret == 0 && shpp->sh_pool_create_len == 0 &&
	    nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
		shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof;
	}

mutex_exit(&spa->spa_history_lock);
	fnvlist_pack_free(record_packed, reclen);
	dmu_buf_rele(dbp, FTAG);
	fnvlist_free(nvl);
}

/*
 * Write out a history event.
 */
int
spa_history_log(spa_t *spa, const char *msg)
{
	int err;
	nvlist_t *nvl = fnvlist_alloc();

fnvlist_add_string(nvl, ZPOOL_HIST_CMD, msg);
	err = spa_history_log_nvl(spa, nvl);
	fnvlist_free(nvl);
	return (err);
}

int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
	int err = 0;
	dmu_tx_t *tx;
	nvlist_t *nvarg, *in_nvl = NULL;

if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
		return (SET_ERROR(EINVAL));

err = nvlist_lookup_nvlist(nvl, ZPOOL_HIST_INPUT_NVL, &in_nvl);
	if (err == 0) {
		(void) nvlist_remove_all(in_nvl, ZPOOL_HIDDEN_ARGS);
	}

tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err) {
		dmu_tx_abort(tx);
		return (err);
	}

VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP));
	if (spa_history_zone() != NULL) {
		fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
		    spa_history_zone());
	}
	fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));

/* Kick this off asynchronously; errors are ignored. */
	dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync, nvarg, tx);
	dmu_tx_commit(tx);

/* spa_history_log_sync will free nvl */
	return (err);
}

/*
 * Read out the command history.
 */
int
spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
{
	objset_t *mos = spa->spa_meta_objset;
	dmu_buf_t *dbp;
	uint64_t read_len, phys_read_off, phys_eof;
	uint64_t leftover = 0;
	spa_history_phys_t *shpp;
	int err;

/*
	 * If the command history doesn't exist (older pool),
	 * that's ok, just return ENOENT.
	 */
	if (!spa->spa_history)
		return (SET_ERROR(ENOENT));

/*
	 * The history is logged asynchronously, so when they request
	 * the first chunk of history, make sure everything has been
	 * synced to disk so that we get it.
	 */
	if (*offp == 0 && spa_writeable(spa))
		txg_wait_synced(spa_get_dsl(spa), 0);

if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0)
		return (err);
	shpp = dbp->db_data;

#ifdef ZFS_DEBUG
	{
		dmu_object_info_t doi;
		dmu_object_info_from_db(dbp, &doi);
		ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
	}
#endif

mutex_enter(&spa->spa_history_lock);
	phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);

if (*offp < shpp->sh_pool_create_len) {
		/* read in just the zpool create history */
		phys_read_off = *offp;
		read_len = MIN(*len, shpp->sh_pool_create_len -
		    phys_read_off);
	} else {
		/*
		 * Need to reset passed in offset to BOF if the passed in
		 * offset has since been overwritten.
		 */
		*offp = MAX(*offp, shpp->sh_bof);
		phys_read_off = spa_history_log_to_phys(*offp, shpp);

/*
		 * Read up to the minimum of what the user passed down or
		 * the EOF (physical or logical).  If we hit physical EOF,
		 * use 'leftover' to read from the physical BOF.
		 */
		if (phys_read_off <= phys_eof) {
			read_len = MIN(*len, phys_eof - phys_read_off);
		} else {
			read_len = MIN(*len,
			    shpp->sh_phys_max_off - phys_read_off);
			if (phys_read_off + *len > shpp->sh_phys_max_off) {
				leftover = MIN(*len - read_len,
				    phys_eof - shpp->sh_pool_create_len);
			}
		}
	}

/* offset for consumer to use next */
	*offp += read_len + leftover;

/* tell the consumer how much you actually read */
	*len = read_len + leftover;

if (read_len == 0) {
		mutex_exit(&spa->spa_history_lock);
		dmu_buf_rele(dbp, FTAG);
		return (0);
	}

err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
	    DMU_READ_PREFETCH);
	if (leftover && err == 0) {
		err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
		    leftover, buf + read_len, DMU_READ_PREFETCH);
	}
	mutex_exit(&spa->spa_history_lock);

dmu_buf_rele(dbp, FTAG);
	return (err);
}

/*
 * The nvlist will be consumed by this call.
 */
static void
log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
    dmu_tx_t *tx, const char *fmt, va_list adx)
{
	char *msg;

/*
	 * If this is part of creating a pool, not everything is
	 * initialized yet, so don't bother logging the internal events.
	 * Likewise if the pool is not writeable.
	 */
	if (spa_is_initializing(spa) || !spa_writeable(spa)) {
		fnvlist_free(nvl);
		return;
	}

msg = kmem_vasprintf(fmt, adx);
	fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
	kmem_strfree(msg);

fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
	fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);

if (dmu_tx_is_syncing(tx)) {
		spa_history_log_sync(nvl, tx);
	} else {
		dsl_sync_task_nowait(spa_get_dsl(spa),
		    spa_history_log_sync, nvl, tx);
	}
	/* spa_history_log_sync() will free nvl */
}

void
spa_history_log_internal(spa_t *spa, const char *operation,
    dmu_tx_t *tx, const char *fmt, ...)
{
	dmu_tx_t *htx = tx;
	va_list adx;

/* create a tx if we didn't get one */
	if (tx == NULL) {
		htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
		if (dmu_tx_assign(htx, TXG_WAIT) != 0) {
			dmu_tx_abort(htx);
			return;
		}
	}

va_start(adx, fmt);
	log_internal(fnvlist_alloc(), operation, spa, htx, fmt, adx);
	va_end(adx);

/* if we didn't get a tx from the caller, commit the one we made */
	if (tx == NULL)
		dmu_tx_commit(htx);
}

void
spa_history_log_internal_ds(dsl_dataset_t *ds, const char *operation,
    dmu_tx_t *tx, const char *fmt, ...)
{
	va_list adx;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *nvl = fnvlist_alloc();

ASSERT(tx != NULL);

dsl_dataset_name(ds, namebuf);
	fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
	fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID, ds->ds_object);

va_start(adx, fmt);
	log_internal(nvl, operation, dsl_dataset_get_spa(ds), tx, fmt, adx);
	va_end(adx);
}

void
spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
    dmu_tx_t *tx, const char *fmt, ...)
{
	va_list adx;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *nvl = fnvlist_alloc();

ASSERT(tx != NULL);

dsl_dir_name(dd, namebuf);
	fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
	fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID,
	    dsl_dir_phys(dd)->dd_head_dataset_obj);

va_start(adx, fmt);
	log_internal(nvl, operation, dd->dd_pool->dp_spa, tx, fmt, adx);
	va_end(adx);
}

void
spa_history_log_version(spa_t *spa, const char *operation, dmu_tx_t *tx)
{
	utsname_t *u = utsname();

spa_history_log_internal(spa, operation, tx,
	    "pool version %llu; software version %s; uts %s %s %s %s",
	    (u_longlong_t)spa_version(spa), ZFS_META_GITREV,
	    u->nodename, u->release, u->version, u->machine);
}

#ifndef _KERNEL
const char *
spa_history_zone(void)
{
	return (NULL);
}
#endif

#if defined(_KERNEL)
EXPORT_SYMBOL(spa_history_create_obj);
EXPORT_SYMBOL(spa_history_get);
EXPORT_SYMBOL(spa_history_log);
EXPORT_SYMBOL(spa_history_log_internal);
EXPORT_SYMBOL(spa_history_log_version);
#endif

003 File Manager

Editing: spa_history.c

Upload File

Create Folder