changes

dos下的网卡驱动程序。支持一般通用网卡

Changes from version 11.10 to 11.11:

  o It wasn't right to chain to the existing interrupt controller
    if we had serviced our own interrupt.  That interrupt controller
    would do the wrong thing if it had already serviced its interrupt.
  o Added a table of PCI ids.
  o Added a number of PCI ids for ICH3, ICH4, and ICH5.

Changes from version 11.9 to 11.10:

  o Wasn't properly checking to see if the PHY device number in the
    EEPROM was in range.  This resulted in a garbage message being printed.
  o A change in manufacturing caused the unset values in the EEPROM to
    default to 0xFF instead of 0.  This made it appear as if a second
    PHY chip was installed, which isn't supported.
  o Manual speed/duplex configuration was broken on all 82558/82559/82550
    boards.  Auto-negotiation happened, but wasn't being reported.
    Added a subroutine for 82555 setup.  Why 82555?  Because that's
    the Intel's original implementation of a PHY chip.  It got moved
    onto the same die for the 82558 && friends.
  o Moved the phy I/O routines from macros into subroutines.  No need
    for that much in-line code.
  o Helge Wagner added support for systems where the eeprom only has a
    MAC address and checksum.  If it reads back as all ones, we
    initialize to reasonable values and life goes on.

Changes from version 11.8 to 11.9:

  o Pro/100 VA support added.  Actually, all it does is recognize Yet
    Another PCI device id.
  o The ADL MSMP5S PC/104 board claims that it has neither primary
    nor secondary PHY chips.  Removed code that aborted out if
    neither PHY was defined in the EEPROM.

Changes from version 11.7 to 11.8:

  o Made interrupt sharing actually work properly.

Changes from version 11.6 to 11.7:

  o A few changes for MASM compatibility.
  o Changed the TX_THRESHOLD to 200 to delay transmission until the
    packet has been completely copied.  This was necessary for some
    machines to avoid sending runt packets.  If this causes you to
    have a cow, I can add a switch to set it back to 4.
  o newer boards have bigger eeproms, so the eeprom code is nearly all new.
  o Intel is now using multiple device ids, so the code needs to find
    all possible id's that might be used.
  o Default to half duplex.

Changes from version 11.5 to 11.6:

  o Changed the method used to locate adapters subsequent to the first.
    A user needed to specify multiple adapters in a consistent manner
    from machine to machine regardless of the assigned I/O address.
  o Enabled IRQ sharing.  Too much of a hassle to tweak BIOS
    parameters to ensure that each board has its own IRQ.

Changes from version 11.4 to 11.5:

  o Need to turn on bus_master in PCI command register.
  o We were printing the self test FAILED msg twice.

Changes from version 11.1 to 11.2:

  o Removed int_no from the command line parsing.  No need for it.
  o On some fast Pentiums, there was insufficient delay between
    EEPROM accesses