PATCH [1/1] GIANFAR infiinate loop on reception of PAUSE FRAME

[Staale.Aakermann]

There is a undesired behavior in the GIANFAR driver that causes a infinite loop 
stalling the CPU when reception of PAUSE FRAMES.
I found this error during testing with a DSL modem (Westermo 
http://www.westermo.com). This equipment spawns PAUSE FRAMES continuously on 
its LAN side as long as the link is not up on the WAN side. The GIANFAR driver 
does not handle that it get an timeout on transmission on the first packet 
sent. It will try to gracefully stop all ongoing transactions, but this will 
fail as there aren't any, and it will loop forever.

--- drivers/net/ethernet/freescale/gianfar.c    2012-12-11 04:30:57.000000000 
+0100
+++ drivers/net/ethernet/freescale/gianfar.c    2013-02-20 20:28:13.201931602 
+0100
@@ -1600,8 +1600,10 @@
        struct gfar_private *priv = netdev_priv(dev);
        struct gfar __iomem *regs = NULL;
        u32 tempval;
-       int i;
-
+       u32 reset_value = DMACTRL_GRS;
+       u32 event_value = IEVENT_GRSC;
+       int i = 0;
+
        for (i = 0; i < priv->num_grps; i++) {
                regs = priv->gfargrp[i].regs;
                /* Mask all interrupts */
@@ -1612,22 +1614,32 @@
        }

        regs = priv->gfargrp[0].regs;
+
+       if (!((gfar_read(&regs->tctrl) & TCTRL_RFCPAUSE) == TCTRL_RFCPAUSE))
+       {
+               reset_value |= DMACTRL_GTS;
+               event_value |= IEVENT_GTSC;
+
+       }
+
        /* Stop the DMA, and wait for it to stop */
        tempval = gfar_read(&regs->dmactrl);
-       if ((tempval & (DMACTRL_GRS | DMACTRL_GTS)) !=
-           (DMACTRL_GRS | DMACTRL_GTS)) {
+       if ((tempval & reset_value) != reset_value) {
                int ret;

-               tempval |= (DMACTRL_GRS | DMACTRL_GTS);
+               tempval |= reset_value;
                gfar_write(&regs->dmactrl, tempval);

                do {
-                       ret = spin_event_timeout(((gfar_read(&regs->ievent) &
-                                (IEVENT_GRSC | IEVENT_GTSC)) ==
-                                (IEVENT_GRSC | IEVENT_GTSC)), 1000000, 0);
+
+                       ret = spin_event_timeout(((gfar_read(&regs->ievent) & 
event_value) == event_value), 1000000, 0);
+
                        if (!ret && !(gfar_read(&regs->ievent) & IEVENT_GRSC))
+                       {
                                ret = __gfar_is_rx_idle(priv);
+                       }
                } while (!ret);
+
        }
 }

@@ -1668,9 +1682,10 @@
        lock_rx_qs(priv);

        gfar_halt(dev);
-
+
        unlock_rx_qs(priv);
        unlock_tx_qs(priv);
+
        local_irq_restore(flags);

        /* Free the IRQs */
@@ -2424,22 +2439,32 @@
        struct gfar_private *priv = container_of(work, struct gfar_private,
                                                 reset_task);
        struct net_device *dev = priv->ndev;
-
        if (dev->flags & IFF_UP) {
                netif_tx_stop_all_queues(dev);
                stop_gfar(dev);
                startup_gfar(dev);
                netif_tx_start_all_queues(dev);
        }
-
        netif_tx_schedule_all(dev);
 }

 static void gfar_timeout(struct net_device *dev)
 {
        struct gfar_private *priv = netdev_priv(dev);
+       struct gfar __iomem *regs = NULL;

-       dev->stats.tx_errors++;
+       regs = priv->gfargrp[0].regs;
+
+       if ((gfar_read(&regs->tctrl) & TCTRL_RFCPAUSE) == TCTRL_RFCPAUSE)
+       {
+               printk(KERN_DEBUG "PAUSE FRAME RECEIVED\n");
+               dev->stats.tx_dropped++;
+       }
+       else
+       {
+               dev->stats.tx_errors++;
+       }
+
        schedule_work(&priv->reset_task);
 }


________________________________

CONFIDENTIALITY
This e-mail and any attachment contain KONGSBERG information which may be 
proprietary, confidential or subject to export regulations, and is only meant 
for the intended recipient(s). Any disclosure, copying, distribution or use is 
prohibited, if not otherwise explicitly agreed with KONGSBERG. If received in 
error, please delete it immediately from your system and notify the sender 
properly.
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev