[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Fwd: Message Signaled Interrupts]



I am forwarding this message as a request for help from anyone who can.
					Thanks
					Jim Freeman

-- BEGIN included message

Hi,
	I have a copy of your presentation "MSI changes everything" given at
the PC developers conference in May of this year. I have read the PCI
2.2 spec for the Vital Product Data that is initialized at runtime and
the operation of MSI. The presentation outlines a write-write protocol
for MSI. The way I understand it, the first write is for a DMA or other
transaction being performed and the second write is to some MSI receiver
in the system that starts the ISR for that devices particular message.
The PCI spec also outlines the issues of presenting the same message
while that message is also active and the possibility that the second
message of the same type can be ignored. If this would be a problem for
a device, the device driver is required to perform a handshake to the
interrupt source to prevent a second message from being generated. At
present, I see no mechanism for the acknowledge or a configuration
register on the device that ccan be set to indicate to the system that
the handshake is required. 
	My second question is how does a device receive the interrupt
acknowledge.
	Granted. this is not a problem for two different message types from the
same device.
	Also, Your presentation references a presentation by Jill  Hubbard on a
PCI sig proposal for MSI. Do you have her email address or can I get a
copy of the presentation from one of you. 
	BTW, I do have the PCI 2.2 spec and have read every word about MSI.

				Thanks
				Jim Freeman
				PLX Technology

-- END included message