Re: Invalid INTLN assignment (0x28) in bridged config

[Carter Buck <CBuck@PLXTech.com>]

Hello,

The PCI 9052 has an erratum in which a Configuration Write to the Subsystem 
ID register (offset 2Ch which is read-only) will cause the value to be 
written to the INTLN register at offset 3Ch.  Windows 98, for example, 
writes to the read-only Subsystem ID register to determine if the Subsystem 
ID register is implemented in the device.  This Configuration Write can be 
precluded if a particular value is written into the Windows Registry (we 
can provide a .reg file to enter the value into the Registry).  So perhaps 
your system software is attempting to write 28h to the Subsystem ID register.

Regards,

Carter Buck
PLX Technology, Inc.

At 10:50 AM 3/1/02, Steven Kan wrote:
>We develop CompactPCI-based instrumentation cards, and we deploy them in
>a bridged-chassis configuration with a single-board-computer and a
>CPCI-CPCI bridge from SBS Technologies (210-2 and 212-2) cards. Our
>instrument cards use the PLX9052 PCI bridge chip
>
>We have been having an intermittent problem with INTLN assignments. Most
>of the cards receive a valid INTLN assignment at boot (e.g. 05 hex or 09
>hex). But some of our cards are being assigned an invalid value (28
>hex).
>
>It seems to occur only in the slave chassis, and it is repeatable for a
>given card order. But it's not very logical. I can swap two otherwise
>identical cards and make the problem disappear, then swap them back in
>make it reappear. Or I can move a card from slot 3 to slot 4 and make
>the problem move from slot 5 to slot 6. Weird.
>
>The invalid interrupt number doesn't appear to follow any particular
>card or slot, but it _is_ repeatable.
>
>We're doing some investigation of timing, etc., but I was wondering
>whether anyone had seen this before and whether the 28 hex value had any
>special meaning.
>
>Thanks!