Fix zfs_txg_timeout module parameter

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

diff --git a/module/zfs/txg.c b/module/zfs/txg.c
index 340c42a..5702a61 100644 (file)
--- a/module/zfs/txg.c
+++ b/module/zfs/txg.c
@@ -218,9 +218,19 @@ uint64_t
 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;
@@ -339,7 +349,7 @@ txg_dispatch_callbacks(dsl_pool_t *dp, uint64_t txg)
                            TASKQ_THREADS_CPU_PCT | TASKQ_PREPOPULATE);
                }
 
-               cb_list = kmem_alloc(sizeof (list_t), KM_SLEEP);
+               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));
 
@@ -374,24 +384,22 @@ txg_sync_thread(dsl_pool_t *dp)
 
 #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
+        * 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_MEMALLOC;
+       current->flags |= PF_NOFS;
 #endif /* _KERNEL */
 
        txg_thread_enter(tx, &cpr);
 
        start = delta = 0;
        for (;;) {
-               uint64_t timer, timeout = zfs_txg_timeout * hz;
+               uint64_t timer, timeout;
                uint64_t txg;
 
+               timeout = zfs_txg_timeout * hz;
+
                /*
                 * We sync when we're scanning, there's someone waiting
                 * on us, or the quiesce thread has handed off a txg to
@@ -506,7 +514,7 @@ 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 ||
@@ -524,6 +532,8 @@ 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);
 }
 
@@ -765,4 +775,7 @@ 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