Re: Help, eth0: transmit timed out!

From: Denis Vlasenko (vda@port.imtp.ilyichevsk.odessa.ua)
Date: Wed Mar 12 2003 - 08:37:08 EST


On 12 March 2003 07:23, David Shirley wrote:
> Hi All,
>
> Just installed Redhat 7.3 on an Athlon 2000 system (2.4.20).
>
> We have a 3c905c network card, and when I try copying files from an
> nfs mount I get this in the logs:
>
Mar 12 13:15:11 ark kernel: NETDEV WATCHDOG: eth0: transmit timed out
Mar 12 13:15:11 ark kernel: eth0: transmit timed out, tx_status 00 status e000.
Mar 12 13:15:11 ark kernel: diagnostics: net 0cc6 media 8880 dma 000000a0.
Mar 12 13:15:11 ark kernel: Flags; bus-master 1, dirty 4048(0) current 4064(0)
Mar 12 13:15:11 ark kernel: Transmit list fffffff8 vs. f7ea9200.
Mar 12 13:15:11 ark kernel: 0: @f7ea9200 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 1: @f7ea9240 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 2: @f7ea9280 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 3: @f7ea92c0 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 4: @f7ea9300 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 5: @f7ea9340 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 6: @f7ea9380 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 7: @f7ea93c0 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 8: @f7ea9400 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 9: @f7ea9440 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 10: @f7ea9480 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 11: @f7ea94c0 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 12: @f7ea9500 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 13: @f7ea9540 length 800000b6 status 000000b6
Mar 12 13:15:11 ark kernel: 14: @f7ea9580 length 800000b6 status 800000b6
Mar 12 13:15:11 ark kernel: 15: @f7ea95c0 length 800000b6 status 800000b6
Mar 12 13:15:11 ark kernel: eth0: Resetting the Tx ring pointer.
>
> Usually the machine freezes up after this, and I can't ping it,

Drivers tend to be less stable that core kernel. Maybe there's
a bug in 3c59x.c network card driver. You may read the code
and start putting some printks there, especially in error
paths.

Say, "Resetting the Tx ring pointer" comes from
vortex_tx_timeout(struct net_device *dev), and looking at that code
you may notice that resetting is done without disabling interrupts.
I know nil about low-level network driver stuff, but

                printk(KERN_DEBUG "%s: Resetting the Tx ring pointer.\n", dev->name);
                if (vp->cur_tx - vp->dirty_tx > 0 && inl(ioaddr + DownListPtr) == 0)
                        outl(vp->tx_ring_dma + (vp->dirty_tx % TX_RING_SIZE) * sizeof(struct boom_tx_desc),
                                 ioaddr + DownListPtr);
                if (vp->cur_tx - vp->dirty_tx < TX_RING_SIZE)
===> netif_wake_queue (dev);
                if (vp->drv_flags & IS_BOOMERANG)
                        outb(PKT_BUF_SZ>>8, ioaddr + TxFreeThreshold);
                outw(DownUnstall, ioaddr + EL3_CMD);

looks a bit strange. I'd move netif_wake_queue() below the out()s
and encased out()s in IRQ disabled region.

Or some such.

Whee, this module can talk a lot, see:
MODULE_PARM_DESC(debug, "3c59x debug level (0-6)");

;)

--
vda
-
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/



This archive was generated by hypermail 2b29 : Sat Mar 15 2003 - 22:00:30 EST