fwochi.c and bus_space_barrier()

[Sean Bruno]

On Mon, 2009-04-20 at 21:38 +0200, Andreas Tobler wrote:
> Sean Bruno wrote:
> > On Mon, 2009-04-20 at 20:36 +0200, Andreas Tobler wrote:
> >> resetting OHCI...done (loop=0)
> > 
> > 
> > Can you recomplile with firewire_debug = 1 and resend the output?
> > 
> > I'm interested in:
> > device_printf(sc->fc.dev, "%s: OHCI_INT_REG_FAIL.\n", __func__);
> > 
> > If that doesn't get printed, then I need to debug a bit further.
> 
> I always use firewire_debug=1, in the last try even > 1. All the traces 
> I sent are with firewire_debug=1.
> 
> I didn't see the above, I suspect the early OWRITE/READ in rddata are 
> too early for the silicon. Unfortunately adding printf's there, cures 
> the issue.
> 
> Andreas

I *think* this section of fwphy_rddata() is suspect:
        /*
         * Setup command to PHY
         */
        fun = PHYDEV_RDCMD | (addr << PHYDEV_REGADDR);
        OWRITE(sc, OHCI_PHYACCESS, fun);
        bus_space_barrier(sc->bst, sc->bsh, OHCI_PHYACCESS,
                                4, BUS_SPACE_BARRIER_WRITE);


According to the specification, this access is illegal if SCLK has not
started.  So, there's no way out of this error without a pause() after
LPS is set in fwohci_probe_phy().  

Although this adventure did teach me a great deal regarding firewire.
Thank you for the challenging problem.

Find the final version of my update attached.  Let me know what you find
with it.

Sean
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fwohci.c.diff
Type: text/x-patch
Size: 5984 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-firewire/attachments/20090420/b9dd1599/fwohci.c.bin