drivers/net/isa-skeleton.c queue management bug?

drivers/net/isa-skeleton.c queue management bug?

[Trammell Hudson]

I realize that the code is going on ten years old and isn't supposed
to be run, but is there a bug in the interface queue handling for
packet transmissions in isa-skeleton.c?

If TX_RING is defined, net_interrupt() will check to see if it was
called due to a completed packet, and if so, then it executes:

        if (status & TX_INTR) {
                /* Transmit complete. */                  
                net_tx(dev);
                np->stats.tx_packets++;                  
                netif_wake_queue(dev);
        }

However, the net_tx() function processes any completed entries and
then executes:

        if (netif_queue_stopped(dev) && ! tx_full(dev))
                netif_wake_queue(dev);     


It seems that the queue is going to be awoken in net_interrupt,
regardless of the state of the ring buffer.  I'm not exactly sure
under what circumstances this might occur, but I'm trying to track
down random hangs in a driver that is based on isa-skeleton.

Trammell      
--           
  -----|----- hudson@osresearch.net                   W 240-283-1700
*>=====[]L\   hudson@rotomotion.com                   M 505-463-1896
'     -'-`-   http://www.swcp.com/~hudson/                    KC5RNF