[PATCH 3.16 130/204] can: gs_usb: fix busy loop if no more TX context is available

From: Ben Hutchings
Date: Thu Dec 28 2017 - 12:31:32 EST


3.16.52-rc1 review patch. If anyone has any objections, please let me know.

------------------

From: Wolfgang Grandegger <wg@xxxxxxxxxxxxxx>

commit 97819f943063b622eca44d3644067c190dc75039 upstream.

If sending messages with no cable connected, it quickly happens that
there is no more TX context available. Then "gs_can_start_xmit()"
returns with "NETDEV_TX_BUSY" and the upper layer does retry
immediately keeping the CPU busy. To fix that issue, I moved
"atomic_dec(&dev->active_tx_urbs)" from "gs_usb_xmit_callback()" to
the TX done handling in "gs_usb_receive_bulk_callback()". Renaming
"active_tx_urbs" to "active_tx_contexts" and moving it into
"gs_[alloc|free]_tx_context()" would also make sense.

Signed-off-by: Wolfgang Grandegger <wg@xxxxxxxxxxxxxx>
Signed-off-by: Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx>
Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx>
---
drivers/net/can/usb/gs_usb.c | 10 ++--------
1 file changed, 2 insertions(+), 8 deletions(-)

--- a/drivers/net/can/usb/gs_usb.c
+++ b/drivers/net/can/usb/gs_usb.c
@@ -356,6 +356,8 @@ static void gs_usb_recieve_bulk_callback

gs_free_tx_context(txc);

+ atomic_dec(&dev->active_tx_urbs);
+
netif_wake_queue(netdev);
}

@@ -444,14 +446,6 @@ static void gs_usb_xmit_callback(struct
urb->transfer_buffer_length,
urb->transfer_buffer,
urb->transfer_dma);
-
- atomic_dec(&dev->active_tx_urbs);
-
- if (!netif_device_present(netdev))
- return;
-
- if (netif_queue_stopped(netdev))
- netif_wake_queue(netdev);
}

static netdev_tx_t gs_can_start_xmit(struct sk_buff *skb, struct net_device *netdev)