/*
* 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.
+ * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_scan.h>
+#include <sys/zil.h>
#include <sys/callb.h>
+#include <sys/trace_txg.h>
/*
* ZFS Transaction Groups
* 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
+ * 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
+ * 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
* now transition to the syncing state.
*/
-static void txg_sync_thread(dsl_pool_t *dp);
-static void txg_quiesce_thread(dsl_pool_t *dp);
+static void txg_sync_thread(void *arg);
+static void txg_quiesce_thread(void *arg);
int zfs_txg_timeout = 5; /* max seconds worth of delta per txg */
int i;
mutex_init(&tx->tx_cpu[c].tc_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&tx->tx_cpu[c].tc_open_lock, NULL, MUTEX_NOLOCKDEP,
+ NULL);
for (i = 0; i < TXG_SIZE; i++) {
cv_init(&tx->tx_cpu[c].tc_cv[i], NULL, CV_DEFAULT,
NULL);
for (c = 0; c < max_ncpus; c++) {
int i;
+ mutex_destroy(&tx->tx_cpu[c].tc_open_lock);
mutex_destroy(&tx->tx_cpu[c].tc_lock);
for (i = 0; i < TXG_SIZE; i++) {
cv_destroy(&tx->tx_cpu[c].tc_cv[i]);
tx->tx_threads = 2;
tx->tx_quiesce_thread = thread_create(NULL, 0, txg_quiesce_thread,
- dp, 0, &p0, TS_RUN, minclsyspri);
+ dp, 0, &p0, TS_RUN, defclsyspri);
/*
* The sync thread can need a larger-than-default stack size on
* 32-bit x86. This is due in part to nested pools and
* scrub_visitbp() recursion.
*/
- tx->tx_sync_thread = thread_create(NULL, 32<<10, txg_sync_thread,
- dp, 0, &p0, TS_RUN, minclsyspri);
+ tx->tx_sync_thread = thread_create(NULL, 0, txg_sync_thread,
+ dp, 0, &p0, TS_RUN, defclsyspri);
mutex_exit(&tx->tx_sync_lock);
}
}
static void
-txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
+txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, clock_t time)
{
CALLB_CPR_SAFE_BEGIN(cpr);
if (time)
- (void) cv_timedwait_interruptible(cv, &tx->tx_sync_lock,
+ (void) cv_timedwait_sig(cv, &tx->tx_sync_lock,
ddi_get_lbolt() + time);
else
- cv_wait_interruptible(cv, &tx->tx_sync_lock);
+ cv_wait_sig(cv, &tx->tx_sync_lock);
CALLB_CPR_SAFE_END(cpr, &tx->tx_sync_lock);
}
tc = &tx->tx_cpu[CPU_SEQID];
kpreempt_enable();
- mutex_enter(&tc->tc_lock);
-
+ mutex_enter(&tc->tc_open_lock);
txg = tx->tx_open_txg;
+
+ mutex_enter(&tc->tc_lock);
tc->tc_count[txg & TXG_MASK]++;
+ mutex_exit(&tc->tc_lock);
th->th_cpu = tc;
th->th_txg = txg;
{
tx_cpu_t *tc = th->th_cpu;
- mutex_exit(&tc->tc_lock);
+ ASSERT(!MUTEX_HELD(&tc->tc_lock));
+ mutex_exit(&tc->tc_open_lock);
}
void
th->th_cpu = NULL; /* defensive */
}
+/*
+ * Blocks until all transactions in the group are committed.
+ *
+ * On return, the transaction group has reached a stable state in which it can
+ * then be passed off to the syncing context.
+ */
static void
txg_quiesce(dsl_pool_t *dp, uint64_t txg)
{
tx_state_t *tx = &dp->dp_tx;
+ uint64_t tx_open_time;
int g = txg & TXG_MASK;
int c;
/*
- * Grab all tx_cpu locks so nobody else can get into this txg.
+ * Grab all tc_open_locks so nobody else can get into this txg.
*/
for (c = 0; c < max_ncpus; c++)
- mutex_enter(&tx->tx_cpu[c].tc_lock);
+ mutex_enter(&tx->tx_cpu[c].tc_open_lock);
ASSERT(txg == tx->tx_open_txg);
tx->tx_open_txg++;
+ tx->tx_open_time = tx_open_time = gethrtime();
- spa_txg_history_set(dp->dp_spa, txg, TXG_STATE_OPEN, gethrtime());
- spa_txg_history_add(dp->dp_spa, tx->tx_open_txg);
+ DTRACE_PROBE2(txg__quiescing, dsl_pool_t *, dp, uint64_t, txg);
+ DTRACE_PROBE2(txg__opened, dsl_pool_t *, dp, uint64_t, tx->tx_open_txg);
/*
* Now that we've incremented tx_open_txg, we can let threads
* enter the next transaction group.
*/
for (c = 0; c < max_ncpus; c++)
- mutex_exit(&tx->tx_cpu[c].tc_lock);
+ mutex_exit(&tx->tx_cpu[c].tc_open_lock);
+
+ spa_txg_history_set(dp->dp_spa, txg, TXG_STATE_OPEN, tx_open_time);
+ spa_txg_history_add(dp->dp_spa, txg + 1, tx_open_time);
/*
* Quiesce the transaction group by waiting for everyone to txg_exit().
/*
* Dispatch the commit callbacks registered on this txg to worker threads.
+ *
+ * If no callbacks are registered for a given TXG, nothing happens.
+ * This function creates a taskq for the associated pool, if needed.
*/
static void
txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
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 */
+ /*
+ * No need to lock tx_cpu_t at this point, since this can
+ * only be called once a txg has been synced.
+ */
int g = txg & TXG_MASK;
* 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);
+ max_ncpus, defclsyspri, max_ncpus, max_ncpus * 2,
+ TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
}
- cb_list = kmem_alloc(sizeof (list_t), KM_PUSHPAGE);
+ cb_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
list_create(cb_list, sizeof (dmu_tx_callback_t),
offsetof(dmu_tx_callback_t, dcb_node));
tx_state_t *tx = &dp->dp_tx;
if (tx->tx_commit_cb_taskq != NULL)
- taskq_wait(tx->tx_commit_cb_taskq);
+ taskq_wait_outstanding(tx->tx_commit_cb_taskq, 0);
}
static void
-txg_sync_thread(dsl_pool_t *dp)
+txg_sync_thread(void *arg)
{
+ dsl_pool_t *dp = arg;
spa_t *spa = dp->dp_spa;
tx_state_t *tx = &dp->dp_tx;
callb_cpr_t cpr;
- vdev_stat_t *vs1, *vs2;
- 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 */
+ clock_t start, delta;
+ (void) spl_fstrans_mark();
txg_thread_enter(tx, &cpr);
- vs1 = kmem_alloc(sizeof(vdev_stat_t), KM_PUSHPAGE);
- vs2 = kmem_alloc(sizeof(vdev_stat_t), KM_PUSHPAGE);
-
start = delta = 0;
for (;;) {
- uint64_t timer, timeout;
+ clock_t timeout = zfs_txg_timeout * hz;
+ clock_t timer;
uint64_t txg;
-
- timeout = zfs_txg_timeout * hz;
+ txg_stat_t *ts;
/*
* We sync when we're scanning, there's someone waiting
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) {
+ tx->tx_quiesced_txg == 0 &&
+ dp->dp_dirty_total < zfs_dirty_data_sync) {
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);
txg_thread_wait(tx, &cpr, &tx->tx_quiesce_done_cv, 0);
}
- if (tx->tx_exiting) {
- kmem_free(vs2, sizeof(vdev_stat_t));
- kmem_free(vs1, sizeof(vdev_stat_t));
+ if (tx->tx_exiting)
txg_thread_exit(tx, &cpr, &tx->tx_sync_thread);
- }
-
- vdev_get_stats(spa->spa_root_vdev, vs1);
/*
* Consume the quiesced txg which has been handed off to
txg = tx->tx_quiesced_txg;
tx->tx_quiesced_txg = 0;
tx->tx_syncing_txg = txg;
+ DTRACE_PROBE2(txg__syncing, dsl_pool_t *, dp, uint64_t, txg);
+ ts = spa_txg_history_init_io(spa, txg, dp);
cv_broadcast(&tx->tx_quiesce_more_cv);
dprintf("txg=%llu quiesce_txg=%llu sync_txg=%llu\n",
mutex_enter(&tx->tx_sync_lock);
tx->tx_synced_txg = txg;
tx->tx_syncing_txg = 0;
+ DTRACE_PROBE2(txg__synced, dsl_pool_t *, dp, uint64_t, txg);
+ spa_txg_history_fini_io(spa, ts);
cv_broadcast(&tx->tx_sync_done_cv);
/*
* Dispatch commit callbacks to worker threads.
*/
txg_dispatch_callbacks(dp, txg);
-
- vdev_get_stats(spa->spa_root_vdev, vs2);
- spa_txg_history_set_io(spa, txg,
- vs2->vs_bytes[ZIO_TYPE_READ]-vs1->vs_bytes[ZIO_TYPE_READ],
- vs2->vs_bytes[ZIO_TYPE_WRITE]-vs1->vs_bytes[ZIO_TYPE_WRITE],
- vs2->vs_ops[ZIO_TYPE_READ]-vs1->vs_ops[ZIO_TYPE_READ],
- vs2->vs_ops[ZIO_TYPE_WRITE]-vs1->vs_ops[ZIO_TYPE_WRITE],
- dp->dp_space_towrite[txg & TXG_MASK] +
- dp->dp_tempreserved[txg & TXG_MASK] / 2);
- spa_txg_history_set(spa, txg, TXG_STATE_SYNCED, gethrtime());
}
}
static void
-txg_quiesce_thread(dsl_pool_t *dp)
+txg_quiesce_thread(void *arg)
{
+ dsl_pool_t *dp = arg;
tx_state_t *tx = &dp->dp_tx;
callb_cpr_t cpr;
*/
dprintf("quiesce done, handing off txg %llu\n", txg);
tx->tx_quiesced_txg = txg;
+ DTRACE_PROBE2(txg__quiesced, dsl_pool_t *, dp, uint64_t, txg);
cv_broadcast(&tx->tx_sync_more_cv);
cv_broadcast(&tx->tx_quiesce_done_cv);
}
}
/*
- * Delay this thread by 'ticks' if we are still in the open transaction
- * group and there is already a waiting txg quiesing or quiesced. Abort
- * the delay if this txg stalls or enters the quiesing state.
+ * Delay this thread by delay nanoseconds if we are still in the open
+ * transaction group and there is already a waiting txg quiesing or quiesced.
+ * Abort the delay if this txg stalls or enters the quiesing state.
*/
void
-txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
+txg_delay(dsl_pool_t *dp, uint64_t txg, hrtime_t delay, hrtime_t resolution)
{
tx_state_t *tx = &dp->dp_tx;
- clock_t timeout = ddi_get_lbolt() + ticks;
+ hrtime_t start = gethrtime();
- /* don't delay if this txg could transition to quiesing immediately */
+ /* don't delay if this txg could transition to quiescing immediately */
if (tx->tx_open_txg > txg ||
tx->tx_syncing_txg == txg-1 || tx->tx_synced_txg == txg-1)
return;
return;
}
- 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);
+ while (gethrtime() - start < delay &&
+ tx->tx_syncing_txg < txg-1 && !txg_stalled(dp)) {
+ (void) cv_timedwait_hires(&tx->tx_quiesce_more_cv,
+ &tx->tx_sync_lock, delay, resolution, 0);
+ }
DMU_TX_STAT_BUMP(dmu_tx_delay);
mutex_exit(&tx->tx_sync_lock);
}
+/*
+ * If there isn't a txg syncing or in the pipeline, push another txg through
+ * the pipeline by queiscing the open txg.
+ */
+void
+txg_kick(dsl_pool_t *dp)
+{
+ tx_state_t *tx = &dp->dp_tx;
+
+ ASSERT(!dsl_pool_config_held(dp));
+
+ mutex_enter(&tx->tx_sync_lock);
+ if (tx->tx_syncing_txg == 0 &&
+ tx->tx_quiesce_txg_waiting <= tx->tx_open_txg &&
+ tx->tx_sync_txg_waiting <= tx->tx_synced_txg &&
+ tx->tx_quiesced_txg <= tx->tx_synced_txg) {
+ tx->tx_quiesce_txg_waiting = tx->tx_open_txg + 1;
+ cv_broadcast(&tx->tx_quiesce_more_cv);
+ }
+ mutex_exit(&tx->tx_sync_lock);
+}
+
boolean_t
txg_stalled(dsl_pool_t *dp)
{
tx->tx_quiesced_txg != 0);
}
+/*
+ * Verify that this txg is active (open, quiescing, syncing). Non-active
+ * txg's should not be manipulated.
+ */
+void
+txg_verify(spa_t *spa, uint64_t txg)
+{
+ ASSERTV(dsl_pool_t *dp = spa_get_dsl(spa));
+ if (txg <= TXG_INITIAL || txg == ZILTEST_TXG)
+ return;
+ ASSERT3U(txg, <=, dp->dp_tx.tx_open_txg);
+ ASSERT3U(txg, >=, dp->dp_tx.tx_synced_txg);
+ ASSERT3U(txg, >=, dp->dp_tx.tx_open_txg - TXG_CONCURRENT_STATES);
+}
+
/*
* Per-txg object lists.
*/
void
-txg_list_create(txg_list_t *tl, size_t offset)
+txg_list_create(txg_list_t *tl, spa_t *spa, size_t offset)
{
int t;
mutex_init(&tl->tl_lock, NULL, MUTEX_DEFAULT, NULL);
tl->tl_offset = offset;
+ tl->tl_spa = spa;
for (t = 0; t < TXG_SIZE; t++)
tl->tl_head[t] = NULL;
boolean_t
txg_list_empty(txg_list_t *tl, uint64_t txg)
{
+ txg_verify(tl->tl_spa, txg);
return (tl->tl_head[txg & TXG_MASK] == NULL);
}
+/*
+ * Returns true if all txg lists are empty.
+ *
+ * Warning: this is inherently racy (an item could be added immediately
+ * after this function returns). We don't bother with the lock because
+ * it wouldn't change the semantics.
+ */
+boolean_t
+txg_all_lists_empty(txg_list_t *tl)
+{
+ for (int i = 0; i < TXG_SIZE; i++) {
+ if (!txg_list_empty(tl, i)) {
+ return (B_FALSE);
+ }
+ }
+ return (B_TRUE);
+}
+
/*
* Add an entry to the list (unless it's already on the list).
* Returns B_TRUE if it was actually added.
txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
boolean_t add;
+ txg_verify(tl->tl_spa, txg);
mutex_enter(&tl->tl_lock);
add = (tn->tn_member[t] == 0);
if (add) {
txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
boolean_t add;
+ txg_verify(tl->tl_spa, txg);
mutex_enter(&tl->tl_lock);
add = (tn->tn_member[t] == 0);
if (add) {
txg_node_t *tn;
void *p = NULL;
+ txg_verify(tl->tl_spa, txg);
mutex_enter(&tl->tl_lock);
if ((tn = tl->tl_head[t]) != NULL) {
p = (char *)tn - tl->tl_offset;
int t = txg & TXG_MASK;
txg_node_t *tn, **tp;
+ txg_verify(tl->tl_spa, txg);
mutex_enter(&tl->tl_lock);
for (tp = &tl->tl_head[t]; (tn = *tp) != NULL; tp = &tn->tn_next[t]) {
int t = txg & TXG_MASK;
txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+ txg_verify(tl->tl_spa, txg);
return (tn->tn_member[t] != 0);
}
int t = txg & TXG_MASK;
txg_node_t *tn = tl->tl_head[t];
+ txg_verify(tl->tl_spa, txg);
return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
}
int t = txg & TXG_MASK;
txg_node_t *tn = (txg_node_t *)((char *)p + tl->tl_offset);
+ txg_verify(tl->tl_spa, txg);
tn = tn->tn_next[t];
return (tn == NULL ? NULL : (char *)tn - tl->tl_offset);
projects / mirror_zfs.git / blobdiff