RE: BIST

["David O'Shea" <daveo@corollary.com>]

I a target issues RETRY's to the PCI configuration requests, then the host
bridge will retry the operation.   *ASUMMING* that the transaction eventually
completes (is not retried and completes), then the entire RETRY sequence is
invisible to the BIOS.  (It just looks like a really slow PCI configuration 
cycle).   The problem is the ASSUMPTION here.   Host bus bridges have varied
responses to this condition.   Pretty much all of them will take a few retries
and until a transaction is finally completed.  But some will eventually give
up on PCI and potentially RETRY the host bus if it has such a feature.  If not,
the bridge might time out and fail the transaction, either on the PCI bus, or
on the host bus.  That usually results in some sort of fatal error.   On the
other hand, some host bridges are programmed to retry forever in this cicumstance.

The unfortunate answer is that the #of times a retry will be tolerated varies
by the host bus bridge design, and sometimes by the host bus bridge programming.
So again, there is not derivable minimum time in which the target board has to
get ready.  Often they will have time, but sometimes they will not.  (Usually 
fatal timeouts are rigged up in the 250-500 millisecond range in most programming
that I have seen, so usually target cards will have lots of time if they RETRY
trasactions until they are ready.   The problem again is my use of usually....)

David O'Shea

At 08:49 AM 3/18/97 -0800, Eric Rehm wrote:
>> 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.]
>
>Dave,
>
>With regard to this discussion, I did look in the archives.
>What was the resolution/answer to the final question below?  I can't
>find a response in the PCI-SIG archives.
>
>/eric rehm
>Equator Technologies, Inc.
>eric@equator.com


>> Date: Fri, 5 Jan 96 10:30:58 PST                                               ¦
>> From: scf@C-Cube.COM (Stephen Fenwick)                                         ¦
>> To: james@raid.us.dell.com                                                     ¦
>> Subject: Re: Latency guarantee for autoconfig?                                 ¦
>> Cc: pci-sig@znyx.com                                                           ¦
>>                                                                                ¦
>> > From: james@raid.us.dell.com (James Van Artsdalen)                           ¦
>> > > Date: Sat, 23 Dec 1995 02:48:05 -0500                                      ¦
>> > > From: Mitch Kahn <mitch@terraport.net>                                     ¦
>> >                                                                              ¦
>> > > The reason I'm asking is I'm wondering how long a PCI device that is       ¦
>> > > initialized after reset (i.e. sets it's device type, etc.) has to          ¦
>> > > do it.                                                                     ¦
>> >                                                                              ¦
>> > The Dell BIOS will do the first write to a PCI card (a write to COMMAND of 0)¦
>> > within a few hundred microseconds of reset.                                  ¦
>> >                                                                              ¦
>> > We will start scanning device ID's several dozen milliseconds later in       ¦
>> > order to initialize video.  A video device needs to be fully                 ¦
>> > functional at this point.                                                    ¦
>>                                                                                ¦
>> What happens if the device issues a Retry against one or both                  ¦
>> of these Configuration Cycles? Does the BIOS have some way to detect           ¦
>> that it's being stalled by a PCI device, or is this transparent to it?         ¦
>>                                                                                ¦
>> Steve                                                                          ¦
>
>
è´¡