*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2014 Integros [integros.com]
*/
/* Portions Copyright 2010 Robert Milkowski */
*/
int zfs_nocacheflush = 0;
+/*
+ * Limit SLOG write size per commit executed with synchronous priority.
+ * Any writes above that will be executed with lower (asynchronous) priority
+ * to limit potential SLOG device abuse by single active ZIL writer.
+ */
+unsigned long zil_slog_bulk = 768 * 1024;
+
static kmem_cache_t *zil_lwb_cache;
static void zil_async_to_sync(zilog_t *zilog, uint64_t foid);
#define LWB_EMPTY(lwb) ((BP_GET_LSIZE(&lwb->lwb_blk) - \
sizeof (zil_chain_t)) == (lwb->lwb_sz - lwb->lwb_nused))
-
-/*
- * ziltest is by and large an ugly hack, but very useful in
- * checking replay without tedious work.
- * When running ziltest we want to keep all itx's and so maintain
- * a single list in the zl_itxg[] that uses a high txg: ZILTEST_TXG
- * We subtract TXG_CONCURRENT_STATES to allow for common code.
- */
-#define ZILTEST_TXG (UINT64_MAX - TXG_CONCURRENT_STATES)
-
static int
zil_bp_compare(const void *x1, const void *x2)
{
}
static lwb_t *
-zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, uint64_t txg, boolean_t fastwrite)
+zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, boolean_t slog, uint64_t txg,
+ boolean_t fastwrite)
{
lwb_t *lwb;
lwb->lwb_zilog = zilog;
lwb->lwb_blk = *bp;
lwb->lwb_fastwrite = fastwrite;
+ lwb->lwb_slog = slog;
lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp));
lwb->lwb_max_txg = txg;
lwb->lwb_zio = NULL;
}
}
+/*
+ * Determine if the zil is dirty in the specified txg. Callers wanting to
+ * ensure that the dirty state does not change must hold the itxg_lock for
+ * the specified txg. Holding the lock will ensure that the zil cannot be
+ * dirtied (zil_itx_assign) or cleaned (zil_clean) while we check its current
+ * state.
+ */
+boolean_t
+zilog_is_dirty_in_txg(zilog_t *zilog, uint64_t txg)
+{
+ dsl_pool_t *dp = zilog->zl_dmu_pool;
+
+ if (txg_list_member(&dp->dp_dirty_zilogs, zilog, txg & TXG_MASK))
+ return (B_TRUE);
+ return (B_FALSE);
+}
+
+/*
+ * Determine if the zil is dirty. The zil is considered dirty if it has
+ * any pending itx records that have not been cleaned by zil_clean().
+ */
boolean_t
zilog_is_dirty(zilog_t *zilog)
{
blkptr_t blk;
int error = 0;
boolean_t fastwrite = FALSE;
+ boolean_t slog = FALSE;
/*
* Wait for any previous destroy to complete.
/*
* Allocate an initial log block if:
* - there isn't one already
- * - the existing block is the wrong endianess
+ * - the existing block is the wrong endianness
*/
if (BP_IS_HOLE(&blk) || BP_SHOULD_BYTESWAP(&blk)) {
tx = dmu_tx_create(zilog->zl_os);
}
error = zio_alloc_zil(zilog->zl_spa, txg, &blk,
- ZIL_MIN_BLKSZ, B_TRUE);
+ ZIL_MIN_BLKSZ, &slog);
fastwrite = TRUE;
if (error == 0)
* Allocate a log write buffer (lwb) for the first log block.
*/
if (error == 0)
- lwb = zil_alloc_lwb(zilog, &blk, txg, fastwrite);
+ lwb = zil_alloc_lwb(zilog, &blk, slog, txg, fastwrite);
/*
* If we just allocated the first log block, commit our transaction
zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb)
{
zbookmark_phys_t zb;
+ zio_priority_t prio;
SET_BOOKMARK(&zb, lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET],
ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
metaslab_fastwrite_mark(zilog->zl_spa, &lwb->lwb_blk);
lwb->lwb_fastwrite = 1;
}
+ if (!lwb->lwb_slog || zilog->zl_cur_used <= zil_slog_bulk)
+ prio = ZIO_PRIORITY_SYNC_WRITE;
+ else
+ prio = ZIO_PRIORITY_ASYNC_WRITE;
lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa,
0, &lwb->lwb_blk, lwb_abd, BP_GET_LSIZE(&lwb->lwb_blk),
- zil_lwb_write_done, lwb, ZIO_PRIORITY_SYNC_WRITE,
+ zil_lwb_write_done, lwb, prio,
ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE |
ZIO_FLAG_FASTWRITE, &zb);
}
UINT64_MAX
};
-/*
- * Use the slog as long as the current commit size is less than the
- * limit or the total list size is less than 2X the limit. Limit
- * checking is disabled by setting zil_slog_limit to UINT64_MAX.
- */
-unsigned long zil_slog_limit = 1024 * 1024;
-#define USE_SLOG(zilog) (((zilog)->zl_cur_used < zil_slog_limit) || \
- ((zilog)->zl_itx_list_sz < (zil_slog_limit << 1)))
-
/*
* Start a log block write and advance to the next log block.
* Calls are serialized.
uint64_t txg;
uint64_t zil_blksz, wsz;
int i, error;
- boolean_t use_slog;
+ boolean_t slog;
if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) {
zilc = (zil_chain_t *)lwb->lwb_buf;
* to clean up in the event of allocation failure or I/O failure.
*/
tx = dmu_tx_create(zilog->zl_os);
- VERIFY(dmu_tx_assign(tx, TXG_WAIT) == 0);
+
+ /*
+ * Since we are not going to create any new dirty data, and we
+ * can even help with clearing the existing dirty data, we
+ * should not be subject to the dirty data based delays. We
+ * use TXG_NOTHROTTLE to bypass the delay mechanism.
+ */
+ VERIFY0(dmu_tx_assign(tx, TXG_WAIT | TXG_NOTHROTTLE));
+
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
zilog->zl_prev_rotor = (zilog->zl_prev_rotor + 1) & (ZIL_PREV_BLKS - 1);
BP_ZERO(bp);
- use_slog = USE_SLOG(zilog);
- error = zio_alloc_zil(spa, txg, bp, zil_blksz,
- USE_SLOG(zilog));
- if (use_slog) {
+ error = zio_alloc_zil(spa, txg, bp, zil_blksz, &slog);
+ if (slog) {
ZIL_STAT_BUMP(zil_itx_metaslab_slog_count);
ZIL_STAT_INCR(zil_itx_metaslab_slog_bytes, lwb->lwb_nused);
} else {
/*
* Allocate a new log write buffer (lwb).
*/
- nlwb = zil_alloc_lwb(zilog, bp, txg, TRUE);
+ nlwb = zil_alloc_lwb(zilog, bp, slog, txg, TRUE);
/* Record the block for later vdev flushing */
zil_add_block(zilog, &lwb->lwb_blk);
static lwb_t *
zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
{
- lr_t *lrc = &itx->itx_lr; /* common log record */
- lr_write_t *lrw = (lr_write_t *)lrc;
+ lr_t *lrcb, *lrc;
+ lr_write_t *lrwb, *lrw;
char *lr_buf;
- uint64_t txg = lrc->lrc_txg;
- uint64_t reclen = lrc->lrc_reclen;
- uint64_t dlen = 0;
+ uint64_t dlen, dnow, lwb_sp, reclen, txg;
if (lwb == NULL)
return (NULL);
ASSERT(lwb->lwb_buf != NULL);
- ASSERT(zilog_is_dirty(zilog) ||
- spa_freeze_txg(zilog->zl_spa) != UINT64_MAX);
- if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY)
+ lrc = &itx->itx_lr; /* Common log record inside itx. */
+ lrw = (lr_write_t *)lrc; /* Write log record inside itx. */
+ if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) {
dlen = P2ROUNDUP_TYPED(
lrw->lr_length, sizeof (uint64_t), uint64_t);
-
+ } else {
+ dlen = 0;
+ }
+ reclen = lrc->lrc_reclen;
zilog->zl_cur_used += (reclen + dlen);
+ txg = lrc->lrc_txg;
zil_lwb_write_init(zilog, lwb);
+cont:
/*
* If this record won't fit in the current log block, start a new one.
+ * For WR_NEED_COPY optimize layout for minimal number of chunks.
*/
- if (lwb->lwb_nused + reclen + dlen > lwb->lwb_sz) {
+ lwb_sp = lwb->lwb_sz - lwb->lwb_nused;
+ if (reclen > lwb_sp || (reclen + dlen > lwb_sp &&
+ lwb_sp < ZIL_MAX_WASTE_SPACE && (dlen % ZIL_MAX_LOG_DATA == 0 ||
+ lwb_sp < reclen + dlen % ZIL_MAX_LOG_DATA))) {
lwb = zil_lwb_write_start(zilog, lwb);
if (lwb == NULL)
return (NULL);
zil_lwb_write_init(zilog, lwb);
ASSERT(LWB_EMPTY(lwb));
- if (lwb->lwb_nused + reclen + dlen > lwb->lwb_sz) {
- txg_wait_synced(zilog->zl_dmu_pool, txg);
- return (lwb);
- }
+ lwb_sp = lwb->lwb_sz - lwb->lwb_nused;
+ ASSERT3U(reclen + MIN(dlen, sizeof (uint64_t)), <=, lwb_sp);
}
+ dnow = MIN(dlen, lwb_sp - reclen);
lr_buf = lwb->lwb_buf + lwb->lwb_nused;
bcopy(lrc, lr_buf, reclen);
- lrc = (lr_t *)lr_buf;
- lrw = (lr_write_t *)lrc;
+ lrcb = (lr_t *)lr_buf; /* Like lrc, but inside lwb. */
+ lrwb = (lr_write_t *)lrcb; /* Like lrw, but inside lwb. */
ZIL_STAT_BUMP(zil_itx_count);
char *dbuf;
int error;
- if (dlen) {
- ASSERT(itx->itx_wr_state == WR_NEED_COPY);
+ if (itx->itx_wr_state == WR_NEED_COPY) {
dbuf = lr_buf + reclen;
- lrw->lr_common.lrc_reclen += dlen;
+ lrcb->lrc_reclen += dnow;
+ if (lrwb->lr_length > dnow)
+ lrwb->lr_length = dnow;
+ lrw->lr_offset += dnow;
+ lrw->lr_length -= dnow;
ZIL_STAT_BUMP(zil_itx_needcopy_count);
- ZIL_STAT_INCR(zil_itx_needcopy_bytes,
- lrw->lr_length);
+ ZIL_STAT_INCR(zil_itx_needcopy_bytes, dnow);
} else {
ASSERT(itx->itx_wr_state == WR_INDIRECT);
dbuf = NULL;
lrw->lr_length);
}
error = zilog->zl_get_data(
- itx->itx_private, lrw, dbuf, lwb->lwb_zio);
+ itx->itx_private, lrwb, dbuf, lwb->lwb_zio);
if (error == EIO) {
txg_wait_synced(zilog->zl_dmu_pool, txg);
return (lwb);
* equal to the itx sequence number because not all transactions
* are synchronous, and sometimes spa_sync() gets there first.
*/
- lrc->lrc_seq = ++zilog->zl_lr_seq; /* we are single threaded */
- lwb->lwb_nused += reclen + dlen;
+ lrcb->lrc_seq = ++zilog->zl_lr_seq; /* we are single threaded */
+ lwb->lwb_nused += reclen + dnow;
lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg);
ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_sz);
ASSERT0(P2PHASE(lwb->lwb_nused, sizeof (uint64_t)));
+ dlen -= dnow;
+ if (dlen > 0) {
+ zilog->zl_cur_used += reclen;
+ goto cont;
+ }
+
return (lwb);
}
itx_t *
zil_itx_create(uint64_t txtype, size_t lrsize)
{
+ size_t itxsize;
itx_t *itx;
lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t);
+ itxsize = offsetof(itx_t, itx_lr) + lrsize;
- itx = zio_data_buf_alloc(offsetof(itx_t, itx_lr) + lrsize);
+ itx = zio_data_buf_alloc(itxsize);
itx->itx_lr.lrc_txtype = txtype;
itx->itx_lr.lrc_reclen = lrsize;
- itx->itx_sod = lrsize; /* if write & WR_NEED_COPY will be increased */
itx->itx_lr.lrc_seq = 0; /* defensive */
itx->itx_sync = B_TRUE; /* default is synchronous */
itx->itx_callback = NULL;
itx->itx_callback_data = NULL;
+ itx->itx_size = itxsize;
return (itx);
}
void
zil_itx_destroy(itx_t *itx)
{
- zio_data_buf_free(itx, offsetof(itx_t, itx_lr)+itx->itx_lr.lrc_reclen);
+ zio_data_buf_free(itx, itx->itx_size);
}
/*
* this itxg. Save the itxs for release below.
* This should be rare.
*/
- atomic_add_64(&zilog->zl_itx_list_sz, -itxg->itxg_sod);
- itxg->itxg_sod = 0;
+ zfs_dbgmsg("zil_itx_assign: missed itx cleanup for "
+ "txg %llu", itxg->itxg_txg);
clean = itxg->itxg_itxs;
}
- ASSERT(itxg->itxg_sod == 0);
itxg->itxg_txg = txg;
itxs = itxg->itxg_itxs = kmem_zalloc(sizeof (itxs_t),
KM_SLEEP);
}
if (itx->itx_sync) {
list_insert_tail(&itxs->i_sync_list, itx);
- atomic_add_64(&zilog->zl_itx_list_sz, itx->itx_sod);
- itxg->itxg_sod += itx->itx_sod;
} else {
avl_tree_t *t = &itxs->i_async_tree;
uint64_t foid =
ASSERT3U(itxg->itxg_txg, <=, synced_txg);
ASSERT(itxg->itxg_txg != 0);
ASSERT(zilog->zl_clean_taskq != NULL);
- atomic_add_64(&zilog->zl_itx_list_sz, -itxg->itxg_sod);
- itxg->itxg_sod = 0;
clean_me = itxg->itxg_itxs;
itxg->itxg_itxs = NULL;
itxg->itxg_txg = 0;
{
uint64_t otxg, txg;
list_t *commit_list = &zilog->zl_itx_commit_list;
- uint64_t push_sod = 0;
if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */
otxg = ZILTEST_TXG;
else
otxg = spa_last_synced_txg(zilog->zl_spa) + 1;
+ /*
+ * This is inherently racy, since there is nothing to prevent
+ * the last synced txg from changing. That's okay since we'll
+ * only commit things in the future.
+ */
for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];
continue;
}
+ /*
+ * If we're adding itx records to the zl_itx_commit_list,
+ * then the zil better be dirty in this "txg". We can assert
+ * that here since we're holding the itxg_lock which will
+ * prevent spa_sync from cleaning it. Once we add the itxs
+ * to the zl_itx_commit_list we must commit it to disk even
+ * if it's unnecessary (i.e. the txg was synced).
+ */
+ ASSERT(zilog_is_dirty_in_txg(zilog, txg) ||
+ spa_freeze_txg(zilog->zl_spa) != UINT64_MAX);
list_move_tail(commit_list, &itxg->itxg_itxs->i_sync_list);
- push_sod += itxg->itxg_sod;
- itxg->itxg_sod = 0;
mutex_exit(&itxg->itxg_lock);
}
- atomic_add_64(&zilog->zl_itx_list_sz, -push_sod);
}
/*
else
otxg = spa_last_synced_txg(zilog->zl_spa) + 1;
+ /*
+ * This is inherently racy, since there is nothing to prevent
+ * the last synced txg from changing.
+ */
for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) {
itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK];
for (itx = list_head(&zilog->zl_itx_commit_list); itx != NULL;
itx = list_next(&zilog->zl_itx_commit_list, itx)) {
txg = itx->itx_lr.lrc_txg;
- ASSERT(txg);
+ ASSERT3U(txg, !=, 0);
+ /*
+ * This is inherently racy and may result in us writing
+ * out a log block for a txg that was just synced. This is
+ * ok since we'll end cleaning up that log block the next
+ * time we call zil_sync().
+ */
if (txg > spa_last_synced_txg(spa) || txg > spa_freeze_txg(spa))
lwb = zil_lwb_commit(zilog, itx, lwb);
}
mutex_exit(&zilog->zl_lock);
if (txg)
txg_wait_synced(zilog->zl_dmu_pool, txg);
+
+ if (zilog_is_dirty(zilog))
+ zfs_dbgmsg("zil (%p) is dirty, txg %llu", zilog, txg);
if (txg < spa_freeze_txg(zilog->zl_spa))
- ASSERT(!zilog_is_dirty(zilog));
+ VERIFY(!zilog_is_dirty(zilog));
taskq_destroy(zilog->zl_clean_taskq);
zilog->zl_clean_taskq = NULL;
EXPORT_SYMBOL(zil_set_sync);
EXPORT_SYMBOL(zil_set_logbias);
+/* BEGIN CSTYLED */
module_param(zil_replay_disable, int, 0644);
MODULE_PARM_DESC(zil_replay_disable, "Disable intent logging replay");
module_param(zfs_nocacheflush, int, 0644);
MODULE_PARM_DESC(zfs_nocacheflush, "Disable cache flushes");
-/* CSTYLED */
-module_param(zil_slog_limit, ulong, 0644);
-MODULE_PARM_DESC(zil_slog_limit, "Max commit bytes to separate log device");
+module_param(zil_slog_bulk, ulong, 0644);
+MODULE_PARM_DESC(zil_slog_bulk, "Limit in bytes slog sync writes per commit");
+/* END CSTYLED */
#endif