[PATCH net 6/6] net: mvneta: Fix race condition during stopping
From: Gregory CLEMENT <hidden>
Date: 2016-01-29 16:27:22
Also in:
lkml, netdev
Subsystem:
marvell mvneta ethernet driver, networking drivers, the rest · Maintainers:
Marcin Wojtas, Andrew Lunn, "David S. Miller", Eric Dumazet, Jakub Kicinski, Paolo Abeni, Linus Torvalds
When stopping the port, the CPU notifier are still there whereas the mvneta_stop_dev function calls mvneta_percpu_disable() on each CPUs. It was possible to have a new CPU coming at this point which could be racy. This patch adds a flag preventing executing the code notifier for a new CPU when the port is stopping. Signed-off-by: Gregory CLEMENT <redacted> --- drivers/net/ethernet/marvell/mvneta.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 3358c9a70467..22962fac47c2 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c@@ -372,6 +372,7 @@ struct mvneta_port { int rxq_def; /* protect */ spinlock_t lock; + bool is_stopping; /* Core clock */ struct clk *clk;
@@ -2914,6 +2915,11 @@ static int mvneta_percpu_notifier(struct notifier_block *nfb, switch (action) { case CPU_ONLINE: case CPU_ONLINE_FROZEN: + /* Configuring the driver for a new CPU while the + * driver is stopping is racy, so just avoid it. + */ + if (pp->is_stopping) + break; netif_tx_stop_all_queues(pp->dev); /* We have to synchronise on tha napi of each CPU
@@ -3052,9 +3058,17 @@ static int mvneta_stop(struct net_device *dev) { struct mvneta_port *pp = netdev_priv(dev); + /* Inform that we are stopping so we don't want to setup the + * driver for new CPUs in the notifiers + */ + pp->is_stopping = true; mvneta_stop_dev(pp); mvneta_mdio_remove(pp); unregister_cpu_notifier(&pp->cpu_notifier); + /* Now that the notifier are unregistered, we can clear the + * flag + */ + pp->is_stopping = false; on_each_cpu(mvneta_percpu_disable, pp, true); free_percpu_irq(dev->irq, pp->ports); mvneta_cleanup_rxqs(pp);
--
2.5.0