[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: BURST Xactions CROSSING BAR BOUNDARIES
Stuart Milton said:
>
> For what it's worth I don't think that this situation calls for a hardware
> solution. Applications should never issue transfer requests which will
> cross a BAR boundary. The proposed response may not work too well since
> the transfer was continued across the boundary by the time the
> application/device driver terminates the transaction.
My $.02:
1) Clearly, the spec says to only assert devsel if the address is within
your range. During a burst, the address increments in a known, precise
fashion. So a target always knows the current address, and should, and
must, always know if that address belongs to it.
2) This requires a hardware solution within the target.
3) I've actually seen this happen when the core logic chip set started to
burst ahead of the cpu. A legal transaction, but had the chip not broken
off the burst, it would have grabbed someone else's address. Not good.
One can easily conceive of a pair of memory cards (ok, it's all on the
motherboard these days) and the core logic chip set bursting ahead as
fast as it can and expecting to get correct data for EVERY access. This
would only work as long as devices only accept transfers within their
own address space.
4) I only regret that I didn't come up with the "I'm not sure why anyone
would think that a computer system with 'fuzzy' imprecise address decoding
would ever work." My posting can't possibly be as good as that.
--------------------------------------------------------
| Steven Larky |
| Principal Engineer |
| larky@anchorchips.com Anchor Chips Inc. |
| Phone: 619-676-6815 12396 World Trade Dr., Ste 212 |
| Fax: 619-676-6896 San Diego, CA 92128 |
--------------------------------------------------------
Πz