[dpdk-stable] [PATCH v3 07/30] net/mlx5: fix reta update can segfault
Nelio Laranjeiro
nelio.laranjeiro at 6wind.com
Mon Oct 9 16:44:43 CEST 2017
Reta update needs to stop/start the port but stopping the port does not
disable the polling functions which may end in a segfault if a core is
polling the queue while the control thread is modifying it.
This patch changes the sequences to an order where such situation cannot
happen.
Fixes: aa13338faf5e ("net/mlx5: rebuild flows on updating RETA")
Cc: yskoh at mellanox.com
Cc: stable at dpdk.org
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro at 6wind.com>
Acked-by: Yongseok Koh <yskoh at mellanox.com>
---
drivers/net/mlx5/mlx5_rss.c | 9 +++++----
drivers/net/mlx5/mlx5_trigger.c | 7 +++++++
2 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_rss.c b/drivers/net/mlx5/mlx5_rss.c
index d3d2603..8942879 100644
--- a/drivers/net/mlx5/mlx5_rss.c
+++ b/drivers/net/mlx5/mlx5_rss.c
@@ -351,11 +351,12 @@ mlx5_dev_rss_reta_update(struct rte_eth_dev *dev,
struct priv *priv = dev->data->dev_private;
assert(!mlx5_is_secondary());
- mlx5_dev_stop(dev);
priv_lock(priv);
ret = priv_dev_rss_reta_update(priv, reta_conf, reta_size);
priv_unlock(priv);
- if (ret)
- return -ret;
- return mlx5_dev_start(dev);
+ if (dev->data->dev_started) {
+ mlx5_dev_stop(dev);
+ mlx5_dev_start(dev);
+ }
+ return -ret;
}
diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
index 212b4df..eeb9585 100644
--- a/drivers/net/mlx5/mlx5_trigger.c
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -30,6 +30,7 @@
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
+#include <unistd.h>
#include <rte_ether.h>
#include <rte_ethdev.h>
@@ -118,6 +119,12 @@ mlx5_dev_stop(struct rte_eth_dev *dev)
return;
priv_lock(priv);
+ dev->data->dev_started = 0;
+ /* Prevent crashes when queues are still in use. */
+ dev->rx_pkt_burst = removed_rx_burst;
+ dev->tx_pkt_burst = removed_tx_burst;
+ rte_wmb();
+ usleep(1000 * priv->rxqs_n);
DEBUG("%p: cleaning up and destroying hash RX queues", (void *)dev);
priv_special_flow_disable_all(priv);
priv_mac_addrs_disable(priv);
--
2.1.4
More information about the stable
mailing list