Wakeup from a PCI device

[Bruce Young <Bruce_Young@ccm.jf.intel.com>]

     Many of you may recall a thread back in April talking about how a PCI 
     device could initiate a wakeup for a powered down system. The proposal 
     that I had made as a part of the initial work I did on PCI Power 
     Management defined the standard INTx# signals as the mechanism to 
     accomplish this. After much discussion and working through as many of 
     the technical issues as we could, the PCI Power Management Working 
     Group has decided to request that a reserved pin be assigned for this 
     purpose. While the technical details are quite involved, the overall 
     reason is that we decided that we did not want to eliminate the 
     possibility of a device in a powered-down slot initiating a wakeup and 
     the only electrically feasible way to do this while maintaining 
     compatibility with older PCI cards is to use a new pin. 
     
     The impact of this should be minimal. Most PCI peripherals don't need 
     to initiate a wakeup and therefore have no need to add the pin. 
     Chipsets that want to support wakeup will, in general, have dedicated 
     inputs for wakeup from other devices (e.g. Lid open or suspend switch) 
     that shared for this new signal as well. The only devices which will 
     need to add a pin are those that have a need to wake a "sleeping" 
     system which will typically be comm and network cards. I realize that 
     this can be a significant impact on those devices but we could not 
     solve all the technical issues related to using the INTx# pins. 
     
     I want to emphasize that this is not final and no-one should start 
     designing Si with this new pin yet. The ECR will go through the 
     standard PCI SIG process of approval which consists of Working Group 
     approval (in this case the protocol working group), Steering Committee 
     approval and then a 30 day review period for the entire SIG. I am 
     posting this to highlight the issue and try to understand how this 
     will be received by the entire PCI design community. I believe that a 
     dialog early can help minimize the problems later in the process.
     
     So what do you think? Is this a non-issue that everyone agrees a 
     separate pin for wakeup is a good idea? Or is everyone dead-set 
     against adding a pin on principle!?
     
     -Bruce Young, Intel Corporation 
     PCI SIG Power Management Working Group Chairman 
'
̺