I then went into the driver and increased the queue-length
from
dev->tx_queue_len = 40;
to
dev->tx_queue_len = 60;
and then it worked.
It seems to me the IP-layer is not notified when the device queue is
emptied by the device and thus doesn't send out the missing fragments.
I tried to look into the net core code to find out: Indeed I could not
find any notification to the higher level protocols.
In vxWorks I am used to waking the higher-level protocols directly
(there is no generic device-specific queue). Thus IP stack can continue
with the next IP-fragment right away. Apparently the Linux IP-stack
can't as it is not notified that the device is ready to send again.
I am really understanding this correctly???
Also, I want to run a raw packet protocol along with IP. That protocol
works best with a small queue - maximum 10 packets!
Esben Nielsen
Vestas Wind Systems
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/