I've updated the configuration file in CVS to move the SCSI device to
a different PCI id, and add some more comments.
Basically, it comes down to this:
ali, ali_ide and ali_usb should always be at fixed locations (pci0.7,
pci0.15, pci0.19). The other devices can be on pci0.[1..4] or
pci1.[1..6]. Other values may work, but don't exist on a real ES40.
Camiel.
On Feb 13, 2008 5:29 PM, Camiel Vanderhoeven <iam...@gm...> wrote:
>
> Hello Eduardo,
>
> I think that pci slot 0.16 does not exist on a real ES40, and that SRM
> can't handle it for that reason.
>
> The USB interrupt is a different thing; the USB interrupt is tied to
> an ISA IRQ, at least, according to the NetBSD source code
> (sys/arch/alpha/pci/pci_6600.h) that is what the value 0xEE means:
>
> /*
> * Some Tsunami models have a PCI device (the USB controller) with interrupts
> * tied to ISA IRQ lines. The IRQ is encoded as:
> *
> * line = 0xe0 | isa_irq;
> */
> #define DEC_6600_LINE_IS_ISA(line) ((line) >= 0xe0 &&
> (line) <= 0xef)
> #define DEC_6600_LINE_ISA_IRQ(line) ((line) & 0x0f)
>
> I need to find out exactly what slots are usable, and put information
> about this in the config file (in comments)
>
> Camiel.
>
> P.S. I'm cc'ing this to the es40-developers mailing list, as it may be
> of interest to others.
>
> > On Feb 13, 2008 4:20 PM, Eduardo Marcelo Serrat wrote:
> > >
> > > Hi Camiel,
> > >
> > > Yesterday I tried SCSI and found the following.
> > >
> > >
> > > // SCSI
> > > a) When scsi adapter is configured to use high slots (i.e. 16 or above - pci0.16 = SYM53c895) SRM sets the CFIT PCI configuration register to Interrupt line = 0x44. This turns to be interrupt #68 and in consequence it cannot be delivered correctly in that DRIR only supports 48 PCI interrupts.
> > >
> > > I also found this is the case of ali.usb on slot 19 (CFIT interrupt line set to 0xEE). I guess that some interrupt routing logic is needed in those cases. (I recall that ES40s had 2 models available ... model 1 and model 2, with the only differences in PCI available slots.)
> > >
> > > This is the cause of pka0.... polling messages and any scsi device configured when issuing a SHOW DEV at srm console.
> > >
> > > b) At the disk polling sequence a SDTR is sent to setup synchronous transfer capabilities. We could safely ignore this message and move from MSG_OUT to COMMAND phase directly.
> > >
> > > Note: The SYM53C895 implementation in ES40 emulator is incredible precise!
> > >
> > > Regards
> > > Eduardo
>
|
