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

From: Mirek Kwasniak (mirek@zind.ikem.pwr.wroc.pl)
Date: Sun Sep 02 2001 - 06:37:01 EDT

  • Next message: Ing. Giorgio Alboni: "[PARPORT] Red Hat 6.5 and Mandrake 8.0 ECP problem"

    On Thu, Aug 30, 2001 at 10:45:51AM -0400, Dave Strauss wrote:
    > On Thu, 30 Aug 2001 15:43:23 +0100, Philip Blundell <pb@nexus.co.uk> wrote:
    > >
    > > >no longer does the "keep-alive" thing. It downloads a bunch of data,
    > > >writes the flash, downloads a bunch of data, write the flash, and so
    > > >on. It can be almost a minute between downloads in some cases.
    > >
    > > Ah, I see.
    > >
    > > Well, in theory these timeouts should be non-fatal. The code in
    > > parport_pc goes to some trouble to make sure it knows how much data
    > > was left in the FIFO when the timeout occurred, in order to make
    > > sure that none is lost. But it's entirely possible that bugs are
    > > lurking, for example if it's misdetecting the depth of the FIFO at
    > > initialisation.
    >
    > Yes, I had a quick look at the timeout recovery code, but didn't see
    > anything obvious. :-(

    I think, that parport code isn't to bad :)

    Recently I did some graphics 1200dpi printings on my HP LJ2100. They were
    horrible long ( 1-2 minutes per page). In longs were some messages :)

      Aug 24 15:40:31 mirek kernel: DMA write timed out
      Aug 24 15:42:00 mirek kernel: DMA write timed out
      Aug 24 15:43:29 mirek kernel: DMA write timed out
      Aug 24 15:45:53 mirek kernel: parport0: FIFO is stuck
      Aug 24 15:45:53 mirek kernel: parport0: BUSY timeout (1) in compat_write_block_pio
      Aug 24 15:46:42 mirek kernel: DMA write timed out
      Aug 24 15:47:13 mirek kernel: parport0: BUSY timeout (1) in compat_write_block_pio
      Aug 24 15:47:31 mirek kernel: DMA write timed out
      Aug 24 15:48:17 mirek kernel: parport0: FIFO is stuck

    The PCL files have size only about 200-800 KB/page. Mesages like above
    doesn't appear with files size about 200KB, but files 600-800KB gave mostly
    one "DMA write timed out" message.

    I tried to find solution. But I thing it is only printer problem. In tests

      dd if=/dev/zero of=/dev/lp0

    I got troughput about 370 KB/s (Pentium 200, motherboard Asus T2P4 with SMC
    FDC37C669QF i/o chip). Not too bad :)

    In pure DOS (6.22) graphics printing (copy /b x.lj prn) speed is about 10%
    higher then in linux.

    Other side my parport tests, I checked again my old ZIP100 (ppa). Currently
    (kernel 2.4.8-ac4) hdparm gives only 140 kB/s :( It was once an experimental
    ppa (as I remeber I used it with kernel 2.0.x) with some tuning parameters
    that gave ZIP100 usable with about 650 kB/s (same computer) :(

    Mirek

    -- 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 : Sun Sep 02 2001 - 06:44:21 EDT