Hi,

following up to my mail from 08/09/2007...

The patch below (adapted from src/sys/dev/pci/if_hme.c) eliminates
the problem. Maybe it's not "The Rigth Thing" to do, but gets
things going... Worth a send-pr?

Kurt

>-1030: cvs diff -u viaide.c 
Index: viaide.c
===================================================================
RCS file: /cvsroot/src/sys/dev/pci/viaide.c,v
retrieving revision 1.41
diff -u -r1.41 viaide.c
--- viaide.c    27 Mar 2007 00:04:04 -0000      1.41
+++ viaide.c    10 Aug 2007 13:36:08 -0000
@@ -674,6 +674,23 @@
        pciide_mapreg_dma(sc, pa);
        aprint_verbose("\n");
 
+       /*
+        * enable memory-space access. sometimes memory space is
+        * _not_ enabled so "just do it" to get things going!
+        */
+       {
+       pcireg_t csr;
+       if (pa->pa_memt)
+           pa->pa_flags |= PCI_FLAGS_MEM_ENABLED;
+       csr = pci_conf_read(pa->pa_pc, pa->pa_tag, PCI_COMMAND_STATUS_REG);
+       if (pa->pa_memt) {
+               csr |= PCI_COMMAND_MEM_ENABLE;
+               sc->sc_ba5_st = pa->pa_memt;
+       }
+       pci_conf_write(pa->pa_pc, pa->pa_tag, PCI_COMMAND_STATUS_REG,
+           csr | PCI_COMMAND_MEM_ENABLE);
+       }
+
        if (sc->sc_dma_ok) {
                sc->sc_wdcdev.sc_atac.atac_cap |= ATAC_CAP_UDMA | ATAC_CAP_DMA;
                sc->sc_wdcdev.irqack = pciide_irqack;