Re: [PARPORT] ECP mode transfers in 2.4.x kernels

From: rjh@world.std.com
Date: Mon Aug 27 2001 - 13:52:12 EDT

  • Next message: Wolfgang Rohdewald: "Re: [PARPORT] MP9805P PCI card 1283:8872"

    On 27 Aug, Dave Strauss wrote:
    > the fact that the parport ECP mode doesn't work unless DMA is used. I
    > don't have any problem removing the test as long as it's coupled with
    > appropriate changes to parport. Perhaps parport should just fail the
    > negotiate-to-ECP request unless FIFO/DMA support is turned on (and the
    > hardware supports it!). I'd be happy to submit a set of coupled
    > patches if that's the right thing to do.
    >

    That may be a correct description of a bug, but back in 2.2.x days, and
    in the last backport that I checked, ECP mode worked without DMA and
    both with and without FIFO. (I haven't tried the more recent 2.4.x
    variations, so a bug could be present.) However, it should be clear
    that this is a software bug and not a hardware restriction. I had ECP
    working reliably (and slowly) with software emulation, and I had ECP
    working reliably with FIFO and no DMA. Reliably means in excess of
    1,000,000 read/write operations without an error. I did have to write
    my own version of FIFO without DMA which I still have not had time to
    port to 2.4.x. FYI, you get about the same transfer speed using FIFO
    without DMA as you do for FIFO with DMA. The difference is CPU
    utilization.

    > BTW, Windows machines won't even attempt to do ECP mode transfers unless
    > the hardware supports DMA, no matter what your BIOS settings may be.
    > Perhaps we should learn from them :-)
    >

    Depends on whose driver you use. There are drivers that will do ECP
    without DMA capability.

    R Horn

    -- To unsubscribe, send mail to: linux-parport-request@torque.net --
    -- with the single word "unsubscribe" in the body of the message. --



    This archive was generated by hypermail 2b29 : Mon Aug 27 2001 - 15:33:57 EDT