The device had an undocumented "feature": it can provide a sequence of
spurious link-down status data even if the link is up all the time.
A sequence of 10 was seen so update the link state only after the device
reports the same link state 20 times.
Signed-off-by: Ondrej Zary <li...@rainbow-software.org>
Reported-by: Michael Leun <lkml20120...@newton.leun.net>
Tested-by: Michael Leun <lkml20120...@newton.leun.net>
---
drivers/net/usb/mcs7830.c | 30 +++++++++++++++++++++---------
1 files changed, 21 insertions(+), 9 deletions(-)
diff --git a/drivers/net/usb/mcs7830.c b/drivers/net/usb/mcs7830.c
index 03c2d8d..cc7e720 100644
--- a/drivers/net/usb/mcs7830.c
+++ b/drivers/net/usb/mcs7830.c
@@ -117,6 +117,7 @@ enum {
struct mcs7830_data {
u8 multi_filter[8];
u8 config;
+ u8 link_counter;
};
static const char driver_name[] = "MOSCHIP usb-ethernet driver";
@@ -632,20 +633,31 @@ static int mcs7830_rx_fixup(struct usbnet *dev, struct
sk_buff *skb)
static void mcs7830_status(struct usbnet *dev, struct urb *urb)
{
u8 *buf = urb->transfer_buffer;
- bool link;
+ bool link, link_changed;
+ struct mcs7830_data *data = mcs7830_get_data(dev);
if (urb->actual_length < 16)
return;
link = !(buf[1] & 0x20);
- if (netif_carrier_ok(dev->net) != link) {
- if (link) {
- netif_carrier_on(dev->net);
- usbnet_defer_kevent(dev, EVENT_LINK_RESET);
- } else
- netif_carrier_off(dev->net);
- netdev_dbg(dev->net, "Link Status is: %d\n", link);
- }
+ link_changed = netif_carrier_ok(dev->net) != link;
+ if (link_changed) {
+ data->link_counter++;
+ /*
+ track link state 20 times to guard against erroneous
+ link state changes reported sometimes by the chip
+ */
+ if (data->link_counter > 20) {
+ data->link_counter = 0;
+ if (link) {
+ netif_carrier_on(dev->net);
+ usbnet_defer_kevent(dev, EVENT_LINK_RESET);
+ } else
+ netif_carrier_off(dev->net);
+ netdev_dbg(dev->net, "Link Status is: %d\n", link);
+ }
+ } else
+ data->link_counter = 0;
}
static const struct driver_info moschip_info = {
--
Ondrej Zary
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
