Shooting trouble on a PCI bus hang

[Ansar Mohammed]

So with a plethora of printfs in pci.c I managed to track it down to a call
to *pci_read_bar *against an AMD CS5536 PCI-ISA bridge. This somehow
intermittently hangs on bootup.
Any suggestions?


On Mon, Jun 20, 2011 at 12:07 PM, Warner Losh <imp at bsdimp.com> wrote:

>
> On Jun 20, 2011, at 6:49 AM, John Baldwin wrote:
>
> > On Sunday, June 19, 2011 12:35:49 pm Ansar Mohammed wrote:
> >> I appreciate that. The system works fine with NetBSD, LInux and Windows
> XP,
> >> so I doubt its hardware.
> >>
> >> Interesting though that OpenBSD has the same issue.
> >>
> >> A question about the debug kernel load process: as it hangs on *
> >> pci_print_verbose* in pci.c, can I deduce that this is the exact code
> >> segment that is the issue?
> >
> > Well, if that was the last device on the bus it might be in a device
> driver
> > probe routine.  You can try adding more printfs to device_probe(), etc.
> to
> > output each driver name as it probes each device perhaps.
>
> We've also had problems with interrupt storms after the drivers have all
> probed, and people blame the last printf :)
>
> It sounds, however, with the OpenBSD datapoint that something is being
> initialized bogusly, perhaps tripping over an erratum that the other systems
> avoid or have specific code to work around.
>
> Warner