RE: BIST

["Kimmery, Clifford (FL51)" <kimmery@space.honeywell.com>]

Thanks for the comments.  The BIST functionality defined by PCI 2.1 is
not automatically activated by reset.  It has to be initiated under
software control by setting bit 6 in the BIST register.  Bit 6 must then
stay set until BIST is complete as a status indicator (this seems to
imply that the configuration target must remain functional).  The two
second restriction is on the time from initiation of BIST to completion.

Because BIST must be initiated by software, that software will be aware
of the circumstances and can make sure it doesn't configure the device
until BIST is complete.  On the other hand, if a device has been
configured prior to the initiation of BIST, either the software must
save the configuration for restoration after BIST is complete or BIST
must not permanently affect the device state.

Cliff Kimmery
Honeywell Inc.
kimmery@space.honeywell.com

>----------
>From: 	David O'Shea[SMTP:daveo@corollary.com]
>Sent: 	Wednesday, March 12, 1997 5:47 PM
>To: 	Clifford Kimmery; pci-sig@znyx.com
>Subject: 	Re: BIST
>
>What is really at issue here is when will the system BIOS scan your
>device, and what will it do if you don't respond because you are 
>doing BIST.   This is a system BIOS implementation specific question
>which requires a general answer for board implementors.
>
>Currently, it is not an issue for other configuration software (like
>that
>in an Operating System), because the actions of the BIOS usually take
>longer than a few seconds, which is more than enough time for PCI
>boards
>to become totally ready.
>
>Most BIOS will not rescan looking for non-ready PCI devices after their
>initial scan.  (In fact I don't know of any that rescan).  So, if your
>device is not ready when the system BIOS is ready to look for it, your
>device will be uninitialized by the system BIOS.  (No resource
>mappings,
>no BAR's programmed, no option ROMS mapped).
>
>So, the question is: when does a system BIOS do the device scan, i.e.
>how
>long after reset is removed?   Most BIOS scan in a few passes, handling
>certain devices earlier than others.  For instance, many BIOS like to
>find
>the video device and turn it on as soon as possible.
>
>Current BIOS are definitely approaching (and some have surpassed) the
>border
>of scanning sooner than 2 seconds after reset has been removed.  Some
>BIOS
>do their initial scan in fewer than 100 milliseconds after reset is
>removed.
>[NOTE: This was covered in past discussions under the subject of how
>quickly
>FPGA devices needed to download vendor ID's and subsystem ID's from
>serial
>non-volatile stores for use in configuration cycle response.  You might
>
>try looking in the archives for FPGA, SUBSSYSTEM, or VENDOR ID, to hit 
>uppon that material.]
>
>Your question is essentially the same question as was raised during
>that
>discussion, only you are talking about seconds.   The other issues was
>in
>the millisecond range.   
>
>Bottom line:  Your device will be skipped on some systems if you wait
>for
>2 seconds to have your configuration registers come ready.  I would not
>do 
>this.
>
>At 03:56 PM 3/12/97 -0500, Kimmery, Clifford (FL51) wrote:
>>The PCI 2.1 specification states (page 186) "A device's Configuration
>>Space must be accessible at all times, not just during system boot."  On
>>page 194, it says that BIST must be complete within 2 seconds (a
>>reasonable period).
>>
>>The ASIC BIST cell that I am using disables functional operation of the
>>device during execution of the BIST.  For system reasons, I would like
>>to tristate all device outputs during BIST as well.
>>
>>Is it a requirement that a PCI device implementing BIST remain
>>functional (at least to the extent of behaving as a configuration
>>target) while performing BIST?  Can it be non-functional for the two
>>second window?
>
>
>
>
ÜXG