s390/qeth: invoke softirqs after napi_schedule()
authorJulian Wiedmann <jwi@linux.ibm.com>
Mon, 17 Sep 2018 15:36:06 +0000 (17:36 +0200)
committerDavid S. Miller <davem@davemloft.net>
Mon, 17 Sep 2018 16:10:25 +0000 (09:10 -0700)
Calling napi_schedule() from process context does not ensure that the
NET_RX softirq is run in a timely fashion. So trigger it manually.

This is no big issue with current code. A call to ndo_open() is usually
followed by a ndo_set_rx_mode() call, and for qeth this contains a
spin_unlock_bh(). Except for OSN, where qeth_l2_set_rx_mode() bails out
early.
Nevertheless it's best to not depend on this behaviour, and just fix
the issue at its source like all other drivers do. For instance see
commit 83a0c6e58901 ("i40e: Invoke softirqs after napi_reschedule").

Fixes: a1c3ed4c9ca0 ("qeth: NAPI support for l2 and l3 discipline")
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/s390/net/qeth_l2_main.c
drivers/s390/net/qeth_l3_main.c

index 33b65471a68acab057a7e0caf138c85cb074e1a4..6285af373bdf668ee44a1ebc010c12d7341b9a42 100644 (file)
@@ -755,7 +755,10 @@ static int __qeth_l2_open(struct net_device *dev)
 
        if (qdio_stop_irq(card->data.ccwdev, 0) >= 0) {
                napi_enable(&card->napi);
+               local_bh_disable();
                napi_schedule(&card->napi);
+               /* kick-start the NAPI softirq: */
+               local_bh_enable();
        } else
                rc = -EIO;
        return rc;
index 5d7e2921ab367fdcbc4cfece432ebfc461443f62..8930d2a9fcad554e26b39960780afd5fc2929375 100644 (file)
@@ -2293,7 +2293,10 @@ static int __qeth_l3_open(struct net_device *dev)
 
        if (qdio_stop_irq(card->data.ccwdev, 0) >= 0) {
                napi_enable(&card->napi);
+               local_bh_disable();
                napi_schedule(&card->napi);
+               /* kick-start the NAPI softirq: */
+               local_bh_enable();
        } else
                rc = -EIO;
        return rc;