Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 25 Nov 2009 17:03:42 -0600

Mark, Matt,

On Wednesday 25 November 2009 15:23:15 Mark Rivers wrote:
> I don't know upgrading to 3.14.x will fix that, I have had problems that
> sound similar on MVME2700 on 3.14.8.2 and 3.14.10.  But I don't recall
> seeing the DMA underflow message. The network goes belly up and cannot
> recover, though everything else looks OK.  Andrew Johnson made a change
> to the BSP recently that he though might address the issue, but I have
> not tested it.  On the other hand this has not happened for many months
> now.  Sometimes it happened every week.

That was exactly the issue that caused me to release the new BSP.  Upgrading 
to R3.14.x will not help, because the dec driver basically locks up the 
network when this happens.

The fix is actually to simplify the code that Matt posted, replacing 
everything after the LOG_MSG as follows:

> /* restart tx if DMA underflow is detected */
>
> if (pTxD->tDesc0 & PCISWAP(TDESC0_UF))
> {
>   LOG_MSG ("%s%d - DMA underflow\n",
>   (int) DRV_NAME, pDrvCtrl->unit, 0, 0, 0, 0);
>
>    if (DRV_FLAGS_ISSET(DEC_TX_KICKSTART))
>        DEC_CSR_WRITE (CSR1, CSR1_TPD);
> }

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System