* CDDL HEADER END
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
- * Use is subject to license terms.
+ * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Portions Copyright 2011 Martin Matuska
+ * Copyright (c) 2013 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/txg_impl.h>
#include <sys/dmu_impl.h>
+#include <sys/dmu_tx.h>
#include <sys/dsl_pool.h>
+#include <sys/dsl_scan.h>
#include <sys/callb.h>
+#include <sys/spa_impl.h>
/*
- * Pool-wide transaction groups.
+ * ZFS Transaction Groups
+ * ----------------------
+ *
+ * ZFS transaction groups are, as the name implies, groups of transactions
+ * that act on persistent state. ZFS asserts consistency at the granularity of
+ * these transaction groups. Each successive transaction group (txg) is
+ * assigned a 64-bit consecutive identifier. There are three active
+ * transaction group states: open, quiescing, or syncing. At any given time,
+ * there may be an active txg associated with each state; each active txg may
+ * either be processing, or blocked waiting to enter the next state. There may
+ * be up to three active txgs, and there is always a txg in the open state
+ * (though it may be blocked waiting to enter the quiescing state). In broad
+ * strokes, transactions — operations that change in-memory structures — are
+ * accepted into the txg in the open state, and are completed while the txg is
+ * in the open or quiescing states. The accumulated changes are written to
+ * disk in the syncing state.
+ *
+ * Open
+ *
+ * When a new txg becomes active, it first enters the open state. New
+ * transactions — updates to in-memory structures — are assigned to the
+ * currently open txg. There is always a txg in the open state so that ZFS can
+ * accept new changes (though the txg may refuse new changes if it has hit
+ * some limit). ZFS advances the open txg to the next state for a variety of
+ * reasons such as it hitting a time or size threshold, or the execution of an
+ * administrative action that must be completed in the syncing state.
+ *
+ * Quiescing
+ *
+ * After a txg exits the open state, it enters the quiescing state. The
+ * quiescing state is intended to provide a buffer between accepting new
+ * transactions in the open state and writing them out to stable storage in
+ * the syncing state. While quiescing, transactions can continue their
+ * operation without delaying either of the other states. Typically, a txg is
+ * in the quiescing state very briefly since the operations are bounded by
+ * software latencies rather than, say, slower I/O latencies. After all
+ * transactions complete, the txg is ready to enter the next state.
+ *
+ * Syncing
+ *
+ * In the syncing state, the in-memory state built up during the open and (to
+ * a lesser degree) the quiescing states is written to stable storage. The
+ * process of writing out modified data can, in turn modify more data. For
+ * example when we write new blocks, we need to allocate space for them; those
+ * allocations modify metadata (space maps)... which themselves must be
+ * written to stable storage. During the sync state, ZFS iterates, writing out
+ * data until it converges and all in-memory changes have been written out.
+ * The first such pass is the largest as it encompasses all the modified user
+ * data (as opposed to filesystem metadata). Subsequent passes typically have
+ * far less data to write as they consist exclusively of filesystem metadata.
+ *
+ * To ensure convergence, after a certain number of passes ZFS begins
+ * overwriting locations on stable storage that had been allocated earlier in
+ * the syncing state (and subsequently freed). ZFS usually allocates new
+ * blocks to optimize for large, continuous, writes. For the syncing state to
+ * converge however it must complete a pass where no new blocks are allocated
+ * since each allocation requires a modification of persistent metadata.
+ * Further, to hasten convergence, after a prescribed number of passes, ZFS
+ * also defers frees, and stops compressing.
+ *
+ * In addition to writing out user data, we must also execute synctasks during
+ * the syncing context. A synctask is the mechanism by which some
+ * administrative activities work such as creating and destroying snapshots or
+ * datasets. Note that when a synctask is initiated it enters the open txg,
+ * and ZFS then pushes that txg as quickly as possible to completion of the
+ * syncing state in order to reduce the latency of the administrative
+ * activity. To complete the syncing state, ZFS writes out a new uberblock,
+ * the root of the tree of blocks that comprise all state stored on the ZFS
+ * pool. Finally, if there is a quiesced txg waiting, we signal that it can
+ * now transition to the syncing state.
*/
static void txg_sync_thread(dsl_pool_t *dp);
static void txg_quiesce_thread(dsl_pool_t *dp);
-int zfs_txg_timeout = 30; /* max seconds worth of delta per txg */
+int zfs_txg_timeout = 5; /* max seconds worth of delta per txg */
/*
* Prepare the txg subsystem.
int c;
bzero(tx, sizeof (tx_state_t));
- tx->tx_cpu = kmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
+ tx->tx_cpu = vmem_zalloc(max_ncpus * sizeof (tx_cpu_t), KM_SLEEP);
for (c = 0; c < max_ncpus; c++) {
int i;
for (i = 0; i < TXG_SIZE; i++) {
cv_init(&tx->tx_cpu[c].tc_cv[i], NULL, CV_DEFAULT,
NULL);
+ list_create(&tx->tx_cpu[c].tc_callbacks[i],
+ sizeof (dmu_tx_callback_t),
+ offsetof(dmu_tx_callback_t, dcb_node));
}
}
- rw_init(&tx->tx_suspend, NULL, RW_DEFAULT, NULL);
mutex_init(&tx->tx_sync_lock, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&tx->tx_sync_more_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&tx->tx_sync_done_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&tx->tx_quiesce_more_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&tx->tx_quiesce_done_cv, NULL, CV_DEFAULT, NULL);
+ cv_init(&tx->tx_exit_cv, NULL, CV_DEFAULT, NULL);
+
tx->tx_open_txg = txg;
}
ASSERT(tx->tx_threads == 0);
- rw_destroy(&tx->tx_suspend);
mutex_destroy(&tx->tx_sync_lock);
+ cv_destroy(&tx->tx_sync_more_cv);
+ cv_destroy(&tx->tx_sync_done_cv);
+ cv_destroy(&tx->tx_quiesce_more_cv);
+ cv_destroy(&tx->tx_quiesce_done_cv);
+ cv_destroy(&tx->tx_exit_cv);
+
for (c = 0; c < max_ncpus; c++) {
int i;
mutex_destroy(&tx->tx_cpu[c].tc_lock);
- for (i = 0; i < TXG_SIZE; i++)
+ for (i = 0; i < TXG_SIZE; i++) {
cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
+ list_destroy(&tx->tx_cpu[c].tc_callbacks[i]);
+ }
}
- kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
+ if (tx->tx_commit_cb_taskq != NULL)
+ taskq_destroy(tx->tx_commit_cb_taskq);
+
+ vmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
bzero(tx, sizeof (tx_state_t));
}
* 32-bit x86. This is due in part to nested pools and
* scrub_visitbp() recursion.
*/
- tx->tx_sync_thread = thread_create(NULL, 12<<10, txg_sync_thread,
+ tx->tx_sync_thread = thread_create(NULL, 32<<10, txg_sync_thread,
dp, 0, &p0, TS_RUN, minclsyspri);
mutex_exit(&tx->tx_sync_lock);
CALLB_CPR_SAFE_BEGIN(cpr);
if (time)
- (void) cv_timedwait(cv, &tx->tx_sync_lock, lbolt + time);
+ (void) cv_timedwait_interruptible(cv, &tx->tx_sync_lock,
+ ddi_get_lbolt() + time);
else
- cv_wait(cv, &tx->tx_sync_lock);
+ cv_wait_interruptible(cv, &tx->tx_sync_lock);
CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
}
* Finish off any work in progress.
*/
ASSERT(tx->tx_threads == 2);
- txg_wait_synced(dp, 0);
+
+ /*
+ * We need to ensure that we've vacated the deferred space_maps.
+ */
+ txg_wait_synced(dp, tx->tx_open_txg + TXG_DEFER_SIZE);
/*
* Wake all sync threads and wait for them to die.
txg_hold_open(dsl_pool_t *dp, txg_handle_t *th)
{
tx_state_t *tx = &dp->dp_tx;
- tx_cpu_t *tc = &tx->tx_cpu[CPU_SEQID];
+ tx_cpu_t *tc;
uint64_t txg;
+ /*
+ * It appears the processor id is simply used as a "random"
+ * number to index into the array, and there isn't any other
+ * significance to the chosen tx_cpu. Because.. Why not use
+ * the current cpu to index into the array?
+ */
+ kpreempt_disable();
+ tc = &tx->tx_cpu[CPU_SEQID];
+ kpreempt_enable();
+
mutex_enter(&tc->tc_lock);
txg = tx->tx_open_txg;
}
void
+txg_register_callbacks(txg_handle_t *th, list_t *tx_callbacks)
+{
+ tx_cpu_t *tc = th->th_cpu;
+ int g = th->th_txg & TXG_MASK;
+
+ mutex_enter(&tc->tc_lock);
+ list_move_tail(&tc->tc_callbacks[g], tx_callbacks);
+ mutex_exit(&tc->tc_lock);
+}
+
+void
txg_rele_to_sync(txg_handle_t *th)
{
tx_cpu_t *tc = th->th_cpu;
static void
txg_quiesce(dsl_pool_t *dp, uint64_t txg)
{
+ hrtime_t start;
+ txg_history_t *th;
tx_state_t *tx = &dp->dp_tx;
int g = txg & TXG_MASK;
int c;
mutex_exit(&tx->tx_cpu[c].tc_lock);
/*
+ * Measure how long the txg was open and replace the kstat.
+ */
+ th = dsl_pool_txg_history_get(dp, txg);
+ th->th_kstat.open_time = gethrtime() - th->th_kstat.birth;
+ th->th_kstat.state = TXG_STATE_QUIESCING;
+ dsl_pool_txg_history_put(th);
+ dsl_pool_txg_history_add(dp, tx->tx_open_txg);
+
+ /*
* Quiesce the transaction group by waiting for everyone to txg_exit().
*/
+ start = gethrtime();
+
for (c = 0; c < max_ncpus; c++) {
tx_cpu_t *tc = &tx->tx_cpu[c];
mutex_enter(&tc->tc_lock);
cv_wait(&tc->tc_cv[g], &tc->tc_lock);
mutex_exit(&tc->tc_lock);
}
+
+ /*
+ * Measure how long the txg took to quiesce.
+ */
+ th = dsl_pool_txg_history_get(dp, txg);
+ th->th_kstat.quiesce_time = gethrtime() - start;
+ dsl_pool_txg_history_put(th);
+}
+
+static void
+txg_do_callbacks(list_t *cb_list)
+{
+ dmu_tx_do_callbacks(cb_list, 0);
+
+ list_destroy(cb_list);
+
+ kmem_free(cb_list, sizeof (list_t));
+}
+
+/*
+ * Dispatch the commit callbacks registered on this txg to worker threads.
+ */
+static void
+txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
+{
+ int c;
+ tx_state_t *tx = &dp->dp_tx;
+ list_t *cb_list;
+
+ for (c = 0; c < max_ncpus; c++) {
+ tx_cpu_t *tc = &tx->tx_cpu[c];
+ /* No need to lock tx_cpu_t at this point */
+
+ int g = txg & TXG_MASK;
+
+ if (list_is_empty(&tc->tc_callbacks[g]))
+ continue;
+
+ if (tx->tx_commit_cb_taskq == NULL) {
+ /*
+ * Commit callback taskq hasn't been created yet.
+ */
+ tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
+ 100, minclsyspri, max_ncpus, INT_MAX,
+ TASKQ_THREADS_CPU_PCT | TASKQ_PREPOPULATE);
+ }
+
+ cb_list = kmem_alloc(sizeof (list_t), KM_PUSHPAGE);
+ list_create(cb_list, sizeof (dmu_tx_callback_t),
+ offsetof(dmu_tx_callback_t, dcb_node));
+
+ list_move_tail(cb_list, &tc->tc_callbacks[g]);
+
+ (void) taskq_dispatch(tx->tx_commit_cb_taskq, (task_func_t *)
+ txg_do_callbacks, cb_list, TQ_SLEEP);
+ }
+}
+
+/*
+ * Wait for pending commit callbacks of already-synced transactions to finish
+ * processing.
+ * Calling this function from within a commit callback will deadlock.
+ */
+void
+txg_wait_callbacks(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ if (tx->tx_commit_cb_taskq != NULL)
+ taskq_wait(tx->tx_commit_cb_taskq);
}
static void
txg_sync_thread(dsl_pool_t *dp)
{
+ spa_t *spa = dp->dp_spa;
tx_state_t *tx = &dp->dp_tx;
callb_cpr_t cpr;
uint64_t start, delta;
+#ifdef _KERNEL
+ /*
+ * Annotate this process with a flag that indicates that it is
+ * unsafe to use KM_SLEEP during memory allocations due to the
+ * potential for a deadlock. KM_PUSHPAGE should be used instead.
+ */
+ current->flags |= PF_NOFS;
+#endif /* _KERNEL */
+
txg_thread_enter(tx, &cpr);
start = delta = 0;
for (;;) {
- uint64_t timer, timeout = zfs_txg_timeout * hz;
+ hrtime_t hrstart;
+ txg_history_t *th;
+ uint64_t timer, timeout;
uint64_t txg;
+ timeout = zfs_txg_timeout * hz;
+
/*
- * We sync when we're scrubbing, there's someone waiting
+ * We sync when we're scanning, there's someone waiting
* on us, or the quiesce thread has handed off a txg to
* us, or we have reached our timeout.
*/
timer = (delta >= timeout ? 0 : timeout - delta);
- while ((dp->dp_scrub_func == SCRUB_FUNC_NONE ||
- spa_shutting_down(dp->dp_spa)) &&
+ while (!dsl_scan_active(dp->dp_scan) &&
!tx->tx_exiting && timer > 0 &&
tx->tx_synced_txg >= tx->tx_sync_txg_waiting &&
tx->tx_quiesced_txg == 0) {
dprintf("waiting; tx_synced=%llu waiting=%llu dp=%p\n",
tx->tx_synced_txg, tx->tx_sync_txg_waiting, dp);
txg_thread_wait(tx, &cpr, &tx->tx_sync_more_cv, timer);
- delta = lbolt - start;
+ delta = ddi_get_lbolt() - start;
timer = (delta > timeout ? 0 : timeout - delta);
}
if (tx->tx_exiting)
txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
- rw_enter(&tx->tx_suspend, RW_WRITER);
-
/*
* Consume the quiesced txg which has been handed off to
* us. This may cause the quiescing thread to now be
tx->tx_quiesced_txg = 0;
tx->tx_syncing_txg = txg;
cv_broadcast(&tx->tx_quiesce_more_cv);
- rw_exit(&tx->tx_suspend);
+
+ th = dsl_pool_txg_history_get(dp, txg);
+ th->th_kstat.state = TXG_STATE_SYNCING;
+ vdev_get_stats(spa->spa_root_vdev, &th->th_vs1);
+ dsl_pool_txg_history_put(th);
dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
txg, tx->tx_quiesce_txg_waiting, tx->tx_sync_txg_waiting);
mutex_exit(&tx->tx_sync_lock);
- start = lbolt;
- spa_sync(dp->dp_spa, txg);
- delta = lbolt - start;
+ start = ddi_get_lbolt();
+ hrstart = gethrtime();
+ spa_sync(spa, txg);
+ delta = ddi_get_lbolt() - start;
mutex_enter(&tx->tx_sync_lock);
- rw_enter(&tx->tx_suspend, RW_WRITER);
tx->tx_synced_txg = txg;
tx->tx_syncing_txg = 0;
- rw_exit(&tx->tx_suspend);
cv_broadcast(&tx->tx_sync_done_cv);
+
+ /*
+ * Dispatch commit callbacks to worker threads.
+ */
+ txg_dispatch_callbacks(dp, txg);
+
+ /*
+ * Measure the txg sync time determine the amount of I/O done.
+ */
+ th = dsl_pool_txg_history_get(dp, txg);
+ vdev_get_stats(spa->spa_root_vdev, &th->th_vs2);
+ th->th_kstat.sync_time = gethrtime() - hrstart;
+ th->th_kstat.nread = th->th_vs2.vs_bytes[ZIO_TYPE_READ] -
+ th->th_vs1.vs_bytes[ZIO_TYPE_READ];
+ th->th_kstat.nwritten = th->th_vs2.vs_bytes[ZIO_TYPE_WRITE] -
+ th->th_vs1.vs_bytes[ZIO_TYPE_WRITE];
+ th->th_kstat.reads = th->th_vs2.vs_ops[ZIO_TYPE_READ] -
+ th->th_vs1.vs_ops[ZIO_TYPE_READ];
+ th->th_kstat.writes = th->th_vs2.vs_ops[ZIO_TYPE_WRITE] -
+ th->th_vs1.vs_ops[ZIO_TYPE_WRITE];
+ th->th_kstat.state = TXG_STATE_COMMITTED;
+ dsl_pool_txg_history_put(th);
}
}
txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
{
tx_state_t *tx = &dp->dp_tx;
- int timeout = lbolt + ticks;
+ clock_t timeout = ddi_get_lbolt() + ticks;
/* don't delay if this txg could transition to quiesing immediately */
if (tx->tx_open_txg > txg ||
return;
}
- while (lbolt < timeout &&
+ while (ddi_get_lbolt() < timeout &&
tx->tx_syncing_txg < txg-1 && !txg_stalled(dp))
(void) cv_timedwait(&tx->tx_quiesce_more_cv, &tx->tx_sync_lock,
timeout);
+ DMU_TX_STAT_BUMP(dmu_tx_delay);
+
mutex_exit(&tx->tx_sync_lock);
}
mutex_enter(&tx->tx_sync_lock);
ASSERT(tx->tx_threads == 2);
if (txg == 0)
- txg = tx->tx_open_txg;
+ txg = tx->tx_open_txg + TXG_DEFER_SIZE;
if (tx->tx_sync_txg_waiting < txg)
tx->tx_sync_txg_waiting = txg;
dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
tx->tx_quiesced_txg != 0);
}
-void
-txg_suspend(dsl_pool_t *dp)
-{
- tx_state_t *tx = &dp->dp_tx;
- /* XXX some code paths suspend when they are already suspended! */
- rw_enter(&tx->tx_suspend, RW_READER);
-}
-
-void
-txg_resume(dsl_pool_t *dp)
-{
- tx_state_t *tx = &dp->dp_tx;
- rw_exit(&tx->tx_suspend);
-}
-
/*
* Per-txg object lists.
*/
mutex_destroy(&tl->tl_lock);
}
-int
+boolean_t
txg_list_empty(txg_list_t *tl, uint64_t txg)
{
return (tl->tl_head[txg & TXG_MASK] == NULL);
}
/*
+ * Add an entry to the end of the list (walks list to find end).
+ * Returns 0 if it's a new entry, 1 if it's already there.
+ */
+int
+txg_list_add_tail(txg_list_t *tl, void *p, uint64_t txg)
+{
+ int t = txg & TXG_MASK;
+ txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+ int already_on_list;
+
+ mutex_enter(&tl->tl_lock);
+ already_on_list = tn->tn_member[t];
+ if (!already_on_list) {
+ txg_node_t **tp;
+
+ for (tp = &tl->tl_head[t]; *tp != NULL; tp = &(*tp)->tn_next[t])
+ continue;
+
+ tn->tn_member[t] = 1;
+ tn->tn_next[t] = NULL;
+ *tp = tn;
+ }
+ mutex_exit(&tl->tl_lock);
+
+ return (already_on_list);
+}
+
+/*
* Remove the head of the list and return it.
*/
void *
return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
}
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+EXPORT_SYMBOL(txg_init);
+EXPORT_SYMBOL(txg_fini);
+EXPORT_SYMBOL(txg_sync_start);
+EXPORT_SYMBOL(txg_sync_stop);
+EXPORT_SYMBOL(txg_hold_open);
+EXPORT_SYMBOL(txg_rele_to_quiesce);
+EXPORT_SYMBOL(txg_rele_to_sync);
+EXPORT_SYMBOL(txg_register_callbacks);
+EXPORT_SYMBOL(txg_delay);
+EXPORT_SYMBOL(txg_wait_synced);
+EXPORT_SYMBOL(txg_wait_open);
+EXPORT_SYMBOL(txg_wait_callbacks);
+EXPORT_SYMBOL(txg_stalled);
+EXPORT_SYMBOL(txg_sync_waiting);
+
+module_param(zfs_txg_timeout, int, 0644);
+MODULE_PARM_DESC(zfs_txg_timeout, "Max seconds worth of delta per txg");
+#endif