Add 'dmu_tx' kstats entry

[zfs.git] / module / zfs / txg.c

diff --git a/module/zfs/txg.c b/module/zfs/txg.c
index f478ad0..6e64adf 100644 (file)
--- a/module/zfs/txg.c
+++ b/module/zfs/txg.c
@@ -37,7 +37,7 @@
 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.
@@ -49,7 +49,7 @@ txg_init(dsl_pool_t *dp, uint64_t txg)
        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;
@@ -107,7 +107,7 @@ txg_fini(dsl_pool_t *dp)
        if (tx->tx_commit_cb_taskq != NULL)
                taskq_destroy(tx->tx_commit_cb_taskq);
 
-       kmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
+       vmem_free(tx->tx_cpu, max_ncpus * sizeof (tx_cpu_t));
 
        bzero(tx, sizeof (tx_state_t));
 }
@@ -166,10 +166,10 @@ txg_thread_wait(tx_state_t *tx, callb_cpr_t *cpr, kcondvar_t *cv, uint64_t time)
        CALLB_CPR_SAFE_BEGIN(cpr);
 
        if (time)
-               (void) cv_timedwait(cv, &tx->tx_sync_lock,
+               (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);
 }
@@ -335,21 +335,35 @@ txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
                         * Commit callback taskq hasn't been created yet.
                         */
                        tx->tx_commit_cb_taskq = taskq_create("tx_commit_cb",
-                           max_ncpus, minclsyspri, max_ncpus, max_ncpus * 2,
-                           TASKQ_PREPOPULATE);
+                           100, minclsyspri, max_ncpus, INT_MAX,
+                           TASKQ_THREADS_CPU_PCT | TASKQ_PREPOPULATE);
                }
 
                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));
 
-               list_move_tail(&tc->tc_callbacks[g], cb_list);
+               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)
 {
@@ -358,6 +372,19 @@ txg_sync_thread(dsl_pool_t *dp)
        callb_cpr_t cpr;
        uint64_t start, delta;
 
+#ifdef _KERNEL
+       /*
+        * Disable the normal reclaim path for the txg_sync thread.  This
+        * ensures the thread will never enter dmu_tx_assign() which can
+        * otherwise occur due to direct reclaim.  If this is allowed to
+        * happen the system can deadlock.  Direct reclaim call path:
+        *
+        *   ->shrink_icache_memory->prune_icache->dispose_list->
+        *   clear_inode->zpl_clear_inode->zfs_inactive->dmu_tx_assign
+        */
+       current->flags |= PF_MEMALLOC;
+#endif /* _KERNEL */
+
        txg_thread_enter(tx, &cpr);
 
        start = delta = 0;
@@ -479,7 +506,7 @@ void
 txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
 {
        tx_state_t *tx = &dp->dp_tx;
-       int timeout = ddi_get_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 ||
@@ -497,6 +524,8 @@ txg_delay(dsl_pool_t *dp, uint64_t txg, int ticks)
                (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);
 }
 
@@ -722,3 +751,20 @@ txg_list_next(txg_list_t *tl, void *p, uint64_t txg)
 
        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);
+#endif