Re: TRDY# Hang...

[Tom Keaveny <tak@core.rose.hp.com>]

> 
> >Gents,
> >
> >How should a system behave if ????
> >-Frame# was asserted for a valid address
> >-IRDY# asserted
> >-DEVSEL# asserted
> >-TRDY# never came down and the system hung.....
> >Abnormalities in PCI targets are related to DEVSEL not responding but not 
> >TRDY#.
> 
> The target is supposed to complete the initial data phase of the transaction
> within 16 clocks of assertion of FRAME# (PCI spec rev 2.1, page 248). This
> would mean that the target should either perform a data transfer, or assert
> stop (to retry, disconnect or abort the transaction). If your target does
> not do so, it is not a 2.1 compliant device.
> 
> 

Note, that it is quite possible that the above device could be "defective".
I don't think that the PCI spec explicitly deals with aberrant device
behaviour, and its associated recovery.  If the target system has 
some level of error resilliency requirements, then it is conceivable that
the PCI bridge in concert with the system processor implement some sort
of recovery protocol.  The situation implies some sort of bus reset, to
force the offending device to release its bus drivers.
­|k