On Wed, Apr 19, 2000 at 07:59:17AM +0200, Jean-Louis Debert wrote:
-> "D. R. Evans" wrote:
-> > So I'm wondering, is there any way to get a device-by-device listing of
-> > what's occupying the various IRQs? (One hypothesis being that something
-> > else was sitting on IRQ 5 and moved to IRQ 10 when the network card
-> > appeared and occupied IRQ 5. I know that the PCI bus is not supposed to
-> > allow two devices to occupy the same IRQ, but I'm grasping at straws.)
->
->
-> You have that under /proc
->
-> Note that PCI bus _IS_ supposed to allow to devices to occupy
-> the same IRQ (there is arbitration _if needed_, i.e. if either
-> of the devices indicates that it _CAN'T_ share a resource,
-> which would probably be the case for either modem or ethernet,
-> because of timing considerations).
The PCI 2.x specs require that all devices daisy chain interrupt drivers
and be able to share resources. This implies sufficient buffering to last
a while between interrupt services, and encourages the use of bus
mastering for delivering the data. If SCSI host adapters can share an
interrupt, modems and ethernet for dang sure should be able to.
->
-> Besides, did you consider that it might _NOT_ be the IRQ,
-> but also a possible I/O (including memory mapped I/O) port
-> or address, that could cause the conflict, and/or confuse
-> the driver because the wrong device would answer ?
-> The PCI arbitration is supposed to resolve this too, but
-> maybe the devices _CAN'T_ use any other address, in which
-> case you are screwed ...
And in which case the card is out of spec. The PCI spec requires that
software be able to assign a base address to the card in all address
spaces.
->
->
-> --
-> Jean-Louis Debert [EMAIL PROTECTED]
-> 74 Annemasse France
-> old Linux fan
--
-- C^2
No windows were crashed in the making of this email.
Looking for fine software and/or web pages?
http://w3.trib.com/~ccurley
