[tulip] Careful with Conexant LANfinity and 2.4

Donald Becker becker@scyld.com
Sat, 28 Jul 2001 16:18:15 -0400 (EDT)


On Fri, 27 Jul 2001, Kent Hunt wrote:

> 	No lockups so far. But there is something
> wrong with the TX_RING. I can easily get an
> Out-of-sync drity pointer by doing an outbound scp
> of a large file. The transfer then quickly gets
> staled.
> This doesn't happen with 2.2 kernels.

This is caused by a locking change in the 2.4 kernel.

Try adding these two lines in tulip.c line 2668

	if ( ! tp->tx_full)
		netif_unpause_tx_queue(dev);
+	else
+		netif_stop_tx_queue(dev);

Please send a report.

> 	Another thing that I noticed that is when I try to
> remove tulip.o out of the kernel I get:
> Trying to free nonexistent resource <d09dc000-d09dc07f>

I'm still trying to work out cross-compatible semantics for this.  The
2.4 kernel doesn't seem to match up request_region/release_region calls.

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993