Hi,

ASIX AX88772B based USB 10/100 Ethernet adapter doesn't come
up ("carrier off"), despite the built-in 100BASE-FX PHY positive link
indication.

The problem appears to be in usbnet.c framework:

void usbnet_link_change(struct usbnet *dev, bool link, bool need_reset)
{
	/* update link after link is reseted */
	if (link && !need_reset)
		netif_carrier_on(dev->net);
	else
		netif_carrier_off(dev->net);

	if (need_reset && link)
		usbnet_defer_kevent(dev, EVENT_LINK_RESET);
	else
		usbnet_defer_kevent(dev, EVENT_LINK_CHANGE);
}

I think the author's idea was a bit different than what the code really
ended doing. Especially when called with link = 1 and need_reset = 1.

It seems it may have already caused problems - possible workarounds:
- commit 7be4cb7189f7 ("net: usb: ax88179_178a: improve reset check")
- commit ecf848eb934b ("net: usb: ax88179_178a: fix link status when
  link is set to down/up")
Can't check those due to -ENOHW but ax88179_link_reset() adds
a netif_carrier_on() call on link ups.

Not sure about the "reset" name, though (and the comment in
usbnet_link_change()). It seems it's when the link goes up.

Signed-off-by: Krzysztof Hałasa <khalasa@piap.pl>
Fixes: ac64995da872 ("usbnet: introduce usbnet_link_change API")
Fixes: 4b49f58fff00 ("usbnet: handle link change")

--- a/drivers/net/usb/usbnet.c
+++ b/drivers/net/usb/usbnet.c

@@ -1978,16 +1978,18 @@ EXPORT_SYMBOL(usbnet_manage_power);
 
 void usbnet_link_change(struct usbnet *dev, bool link, bool need_reset)
 {
-	/* update link after link is reseted */
-	if (link && !need_reset)
+	if (link)
 		netif_carrier_on(dev->net);
 	else
 		netif_carrier_off(dev->net);
 
-	if (need_reset && link)
-		usbnet_defer_kevent(dev, EVENT_LINK_RESET);
-	else
-		usbnet_defer_kevent(dev, EVENT_LINK_CHANGE);
+	if (need_reset) {
+		/* update link after link is reset */
+		if (link)
+			usbnet_defer_kevent(dev, EVENT_LINK_RESET);
+		else
+			usbnet_defer_kevent(dev, EVENT_LINK_CHANGE);
+	}
 }
 EXPORT_SYMBOL(usbnet_link_change);
 

The code has been introduced in 2013, by 4b49f58fff00 and a bunch of
related commits. Perhaps it visibly affects only AXIS and dm9601
adapters, though.
-- 
Krzysztof "Chris" Hałasa

Sieć Badawcza Łukasiewicz
Przemysłowy Instytut Automatyki i Pomiarów PIAP
Al. Jerozolimskie 202, 02-486 Warszawa