Re: PCI MMIO flushing and stuff (was Re: 2.2.15 with eepro100: eth0: Too much work at interrupt)

From: David Wragg (dpw@doc.ic.ac.uk)
Date: Sun May 21 2000 - 15:29:32 EDT

  • Next message: Quintela Carreira Juan J.: "Re: Basic testing shows 2.3.99-pre9-3 bad, pre9-2 good"

    Andrew Morton <andrewm@uow.edu.au> writes:
    > We know that Pentium uses 'processor order'.

    Yes, but processor ordering only applies when the write-back memory
    type is in effect. MMIO is most often done through uncached memory, so
    (accoding to the Intel docs) all reads and writes appear in-order on
    the processor bus. (Even with this, I suppose it's possible for the
    chipset to coalesce transactions, but I haven't read that any Intel
    chipsets do so)

    The memory types become gradually more relaxed as you go from uncached
    to write combining to write through to write back. At least for P6
    processors. (Pentium is similar, I think, but I haven't read the
    Pentium-era Intel docs as thoroughly)

    David Wragg

    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.rutgers.edu
    Please read the FAQ at http://www.tux.org/lkml/



    This archive was generated by hypermail 2b29 : Sun May 21 2000 - 15:37:12 EDT