On Sun, 28 May 2017, allison via cctalk wrote: > > I have a VAX-11/750, an Able QNiverter (UNIBUS->Qbus adapter), a > > 22-bit Qbus backplane, and a VCB02 (QDSS) 4-bit graphics boardset. > > > > Theory: With an appropriately modified NetBSD driver for the VCB02, > > such that it provides only 18-bit addresses to the VCB02's DMA engine, > > I can get X running on the VCB02 on the 11/750. > > > > That is, the configuration is: 11/750->Unibus->QNiverter->Qbus->VCB02. > > > > Should work with one thing I can think of the VCB02 is Q22 not Q18. Its > 22bit. > > The VAX hardware is for memory 32bit and for IO only 16bit. ( A 16 bit > segment mapped to IO.) > The bit map is mapped as Q22 memory.
Depending on circumstances the DMA address space limitation can be tackled either by avoiding allocating memory for DMA outside the range supported by the device doing DMA or by using memory remapping, either with an IOMMU if available, or via software means aka bounce buffers. Support for the chosen logic will likely have to be placed then in a device driver for the QNiverter bus bridge. I suppose NetBSD will have some generic means of support for these techniques as this has been a common issue with many computer buses, which have run out of address lines needed to access all the system memory as DRAM capacities have grown. The bus bridge device driver might then be able to hook into it appropriately, serving any DMA-capable devices behind the bridge in a uniform way, without individual device drivers having to know their wiring. > > But this is something that randomly popped into my head (I was > > inspired by a research paper, "A VAX Based Data Acquisition Computer > > System at the Nuclear Structure Research Laboratory" wherein a Matrox > > QRGB Qbus graphics board was lashed to an 11/750 in a similar manner... > > > > What say you all? By all means do try it! Although not nearly as complex I have once tried a 24-plane TURBOchannel SFB+ graphics adapter in a VAXstation 4000/90 with the Linux TGA/SFB+ driver and it worked just fine as the OS console device, despite that I suppose nobody has actually envisaged such use, and there's nothing specific to the VAXstation in the TGA/SFB+ driver (originally written for the Alpha port and the PCI bus). I haven't tried X11, because the port was too immature/unstable to attempt compiling it, but I suppose it would just work given that other communication with the graphics hardware did. FWIW, Maciej
