Re: [Qemu-devel] [PATCH] hw/pxa2xx_dma.c

[andrzej zaborowski]

On 17/11/2007, andrzej zaborowski <address@hidden> wrote:
> Hi,
>
> On 15/11/2007, Thorsten Zitterell <address@hidden> wrote:
> > The following patch fixes the problem that DMA transfers are not performed 
> > when
> > the DCSR_STOPINTR bit is set.
> > --
> > Thorsten
> >
> > Index: hw/pxa2xx_dma.c
> > ===================================================================
> > RCS file: /sources/qemu/qemu/hw/pxa2xx_dma.c,v
> > retrieving revision 1.5
> > diff -u -r1.5 pxa2xx_dma.c
> > --- hw/pxa2xx_dma.c     11 Nov 2007 19:47:58 -0000      1.5
> > +++ hw/pxa2xx_dma.c     15 Nov 2007 09:28:22 -0000
> > @@ -186,7 +186,8 @@
> >          s->running = 1;
> >          for (c = 0; c < s->channels; c ++) {
> >              ch = &s->chan[c];
> > -
> > +
> > +            ch->state &= ~DCSR_STOPINTR;
> >              while ((ch->state & DCSR_RUN) && !(ch->state & DCSR_STOPINTR)) 
> > {
> >                  /* Test for pending requests */
> >                  if ((ch->cmd & (DCMD_FLOWSRC | DCMD_FLOWTRG)) && 
> > !ch->request)
> >
> >
>
> Can you check if the following change would make the gumstix NIC work
> too?  In my understanding of the specs it's more correct, but I'm not
> sure.

Sorry, rather this is what I meant:
--- a/hw/pxa2xx_dma.c
+++ b/hw/pxa2xx_dma.c
@@ -347,8 +348,10 @@ static void pxa2xx_dma_write(void *opaque,

         if (value & DCSR_NODESCFETCH) {
             /* No-descriptor-fetch mode */
-            if (value & DCSR_RUN)
+            if (value & DCSR_RUN) {
+                s->chan[channel].state &= ~DCSR_STOPINTR;
                 pxa2xx_dma_run(s);
+            }
         } else {
             /* Descriptor-fetch mode */
             if (value & DCSR_RUN) {

pxa-dma-stopintr-clear.patch
Description: Text Data