FW: Wakeup from a PCI device

[Ken Reneris <kenr@MICROSOFT.com>]

>I believe that devices would need to wake signal during runtime as well.
> 
>E.g.,  there are 2 scenarios in which WAKE is being designed for use.
>One is to wake a "sleeping" system, and the other scenario is for a sleeping
>device in a running system to signal that it needs to wake.
>
>Once a device is off, in a running system, the driver can no longer 
>access the device at interrupt time (since it's "off" and therefore
>the bus itself may be off if all devices on it are off) and therefore a 
>different signal which can be masked needs to be used.   When a wake 
>occurs on the bus, the OS will mask the signal and can then perform 
>the "long term"  operation of re-applying power/clocks to the bus and 
>determining which device has signaled the wake.   The same thing can't 
>be done with the device's IRQ signal as the IRQ may be shared with 
>devices on other buses and therefore can't be masked.
>
>
>However, the signal could still potentially be used to indicate 
>other events have occured.  (hot plug)
>
>	- Ken
>
>
>
>
>
>>>----------
>>>From: 	sigma@brentwood.bc.ca[SMTP:sigma@brentwood.bc.ca]
>>>Sent: 	Friday, August 16, 1996 3:34 PM
>>>To: 	Mailing List Recipients
>>>Subject: 	Re: Wakeup from a PCI device
>>>
>>>
>>>>I'll chime in here on this thread.  What relationship or involvement
>>>>will the wake-up pin have with the hot-swap (live insertion)  PCI
>>>>efforts underway in the SIG?  It seems that the wake-up pin could
>>>>serve the purpose of notifying an OS that a new board has been
>>>>inserted into the system for hot swap boards.
>>>
>>>>>I would think desktops and servers might use it also, so there is no
>>>>>reason to restrict the wakeup pin to laptops.
>>>>>
>>>>>Bruce Young's message suggested that only comm devices (modems, NICs)
>>>>>would use this. I'm not sure one should limit one's thinking. For
>>>>>example,
>>>>>an external data collection device might collect a sample only every 15
>>>>>minutes. A data collection box running on batteries or solar power could
>>>>>benefit also.
>>>
>>>Due to the limited number of reserved pins I would suggest defining the new
>>>wakeup pin as having a (possibly reserved) secondary purpose when the
>>>system is operating.
>>>
>>>The reserved purpose should be a potential synchronous signal as most other
>>>PCI signals.
>>>
>>>Otherwise, two high speed capable reserved pins will be consumed for low
>>>speed asynchronous signals that could have easily been implemented with a
>>>comparatively cheap separate connector.
>>>
>>>I speculate two pins as it seems to me that once you add a wake-up pin, you
>>>also ought to provide a power during sleep pin, although one might elect to
>>>use the wake-up pin for power too.
>>>
>>>James
>>>
>>>James MacPhail
>>>uo957@freenet.victoria.bc.ca    "Think not of engineering as art,
>>>Sigma Seven Systems Ltd.            but of art as engineering"
>>>sigma@brentwood.bc.ca
>>>
>>>
>>>
>>
>>
>
>--------------------------------
>  Alan Deikman, ZNYX Corporation
>  alan@znyx.com
>
>
_
ôá