linux-wlan-ng/CHANGES

* CHANGES
*
* Copyright (C) 1999 AbsoluteValue Systems, Inc. All Rights Reserved.
* --------------------------------------------------------------------
*
* linux-wlan
*
*   The contents of this file are subject to the Mozilla Public
*   License Version 1.1 (the "License"); you may not use this file
*   except in compliance with the License. You may obtain a copy of
*   the License at http://www.mozilla.org/MPL/
*
*   Software distributed under the License is distributed on an "AS
*   IS" basis, WITHOUT WARRANTY OF ANY KIND, either express or
*   implied. See the License for the specific language governing
*   rights and limitations under the License.
*
*   Alternatively, the contents of this file may be used under the
*   terms of the GNU Public License version 2 (the "GPL"), in which
*   case the provisions of the GPL are applicable instead of the
*   above.  If you wish to allow the use of your version of this file
*   only under the terms of the GPL and not to allow others to use
*   your version of this file under the MPL, indicate your decision
*   by deleting the provisions above and replace them with the notice
*   and other provisions required by the GPL.  If you do not delete
*   the provisions above, a recipient may use your version of this
*   file under either the MPL or the GPL.
*
* --------------------------------------------------------------------
*
* Inquiries regarding the linux-wlan Open Source project can be
* made directly to:
*
* AbsoluteValue Systems Inc.
* info@linux-wlan.com
* http://www.linux-wlan.com
*
* --------------------------------------------------------------------
*
* Portions of the development of this software were funded by 
* Intersil Corporation as part of PRISM(R) chipset product development.
*
* --------------------------------------------------------------------
  - Added support for the SH architechure and many changes necessary for
    2.5 support.  Modified from the patches by Ryan Veety.
  - Compile fixes for 2.2.x introduced as part of the new hotplug support.
  - Added man pages contributed by Joey Hess
  - Add ident info for the Z-Com 725/726 USB Widgets.
  - Fixes to silence GCC 3.2 warnings. (Pavel Roskin)
-pre7
  - Patch to fix int array mib manipulation.  (Pavel Roskin)
  - Minor improvements to the build system.
  - Moved RX processing to a bottom half instead of hard irq context.
  - Further script improvements:  signal strength filtering, and some 
    space handling stuff.  The latter is not complete yet.
  - lnxreq_commsquality wasn't setting the status field properly on the
    response messages.  
  - Workaround for spurious interrupts generated before initialization
    is complete. (Clay Jones)
  - hotplug script now invokes 'ifup $DEVICE' if present.
  - Minor cleanups (Pavel Roskin)
  - Tenative SPARC support. (Olivier Bornet)
  - After leaving monitor mode, disable the port if we had to enable it.
  - Updates to hfa384x.h
-pre6
  - Further script enhancements; automatically use scan mode on hardware
    that is known to support it properly.
  - Fixed errors in the hotplug script, and other misc fixes.
  - 2.2.x fixes for the new proc stuff and deferred processing stuff.
  - Add support for the new sniff frame capture. (doc/capturefrm.txt)
  - Get rid of some bitrot with the sniffing code.
  - Populate the tx dropped frames statistics.
-pre5
  - Fixes for host scan when not currently joined.
  - /proc support; some informational stuff is kept in /proc/wlandev/wlanX
  - Complete rewrite of the configuration files and startup scripts.
     hotplug(pci+usb), pcmcia, rc, everything uses same config file(s).
     per-network configuration
     supports scanning and automatically choosing a network from a set.
  - Fix the TMD7160/ncp130 support so it acually works.  :)
  - Make the BUG() call in cmd_access enabled only in debug builds.
  - Minor makefile fixes.
  - Microsoft MN520 PCMCIA ident info added.
  - p2cnfShortPreamble is not a boolean value.  (Clay Jones)
-pre4
  - BAP access fixes for PCI platforms.  This mostly affects SMP 
    arrangements.
  - Back out W200 USB ident info; it's not prism-based.
  - Patch from Tom Prado to make 'keygen' more user-friendly.
  - bap_timeout parameter only valid on non-usb platforms
-pre3
  - Added Ident info for the USB Compaq/Intel W200 widget
  - dot11req_scan requires firmware >=1.3.2.
  - Added the "Acer Warplink USB Adapter" id.
  - Rework the prism2sta_inf_linkstatus call to defer processing of
    results until outside interrupt context.  Should fix the SMP
    deadlocks seen on some machines.
  - Rework the usb_disconnect code to hopefully alleviate the occasional
    crash-on-disconnect some people see.  Solution is not SMP-safe yet.
  - Add the D-Link DCF-660W ident info.
-pre2
  - New parameter, 'prism2_bap_timeout' to specify the timeout on
    bap setup.  timeout*10 == timeout in microseconds.  Note that this
    timeout occurs in spinlock context, so increasing it arbitrarily is
    BAD!  Defaults to 1000, ie 10000us.  
  - compile fix for prism2sta in debug mode.
  - Only prompt for pcmcia sources if we're using non-kernel pcmcia.
    (original patch by Beat Bolli, rewritten a bit..)
  - Add the Linksys WCF12 CF ident info.
  - A bunch of compile fixes for older versions of wireless extensions.
  - Added implementations of GIWTXPOWER and GIWRETRY
  - Some fixes to the wireless extension code (thanks to Jean Tourrilhes)
-pre1
  - wland was incorrectly using kernel headers in its compile. 
  - Adhoc mode was auto-roaming onto "any" SSID.  Fixed.
  - Implemented the dot11req_scan & dot11req_scan_results functions.
      This, combined with more intelligent startup scripts, will allow
      you per-network profiles.   Scripts are unwritten as yet.  :)
  - hfa384x.h and MIB updates.
  - Rewrote the wireless extension support code; moved to p80211 layer.
      Now we're compatible with at least v6-v14, and everything
      inbetween.   We hook up to the new iw_request structure, which
      makes further wireless extension support much easier to add.
  - Add the Microsoft MN510 USB device ID.
  - Finally fix the 'crash on unload' problem for 2.2.x kernels with PCI.
    Turns out the kcompat24 code wasn't quite complete.  (Matthew Rush)
  - Cleaned up our usage of __FUNCTION__ to make gcc 3.x happy.
  - Fix a typo in the prism2 makefile.  
0.1.15
  - Fix an obscure PCMCIA build problem. (modversions disabled but
    header present, and using non-kernel pcmcia)
-pre8
  - Added code to optionally truncate packets in monitor mode. (Clay Jones)
  - Fix a polarity problem with the stripfcs argument to monitor mode.
  - Add a test to make sure wireless.h is included.
  - Netgear MA401A card ident.
  - Further cleanups of the build system (Makefiles)
-pre7
  - Include wireless.h so wireless extensions work again.  :)
  - Work around the braindead RedHat kernel build system.
  - Fix a class of unaligned accesses in the message structures 
    (kernel driver) and user space parsers (wlanctl-ng, etc)
  - Fix an unpacked struct problem that affected ARM platforms
    (thanks to Clay Jones for spotting this one)
-pre6
  - Support the new auto-unknown mode present in firmware >1.3.3
    When issuing an autojoin, we try to join a BSS, then an IBSS, and if
    all fails, we create a new IBSS with the given parameters. 
 	*note*  currently disabled.
  - Updates to hfa384x.h to reflect latest documentation
  - Build system mini-enema.  Automagically detect kernel version,
    pcmcia status, and modversions.  Pick up the kernel compile flags.
  - Hopefully fixed the 'crash on unload' for 2.2.x kernels and PCI.
  - 'p2req_join' command, see doc/wlanctl-ng.p2req_join.txt 
    This lets you join a specific SSID.  Thanks to Clay Jones.
  - A couple more pcmcia/cf card idents
  - More deletions/fixes in wlan_compat.h
-pre5
  - Due to new hotplug code, atmel device detection removed
  - Further cleanups in wlan_compat.h and non-kernel makefiles
  - Debian package stuff yanked in favor of the "official" deb packages
  - Rearranged the wireless extension support code a bit
  - A fix for the promisc/allmulti stuff on USB devices
-pre4
  - Hotplug support for PCI/PLX/USB widgets
  - A compile fix for 2.2.x (x < 18) kernels and PCI adapters
  - Support for TMD7160-based "almost but not quite PLX" boards
  - HP-PA support.  In theory.  Anyone care to test?
  - pcmcia script cleanups
  - Add ident info for the "Intel PRO/Wireless 2011B LAN USB Device"
  - Include an almost-sane fix to the modversion madness
  - Tweak the init order of the PCI driver structs
-pre3
  - Added ident info for ASUS WL-110/WL-100 cards
  - Support for PCI power managment
  - Fixed a wep hang on the transmit side
  - Compile fixes; mainly for 2.2 kernels
  - Add support for the SIOCGIWRATE wireless extension
-pre2
  - Restructured the driver locking.  SMP should be MUCH better now.
  - Stripped out more dead code. (mainly FID_STACK)
  - Fix an skb access after we call netif_rx
  - Fixes for XScale CPUs.
  - Added added a MiniUSB ident info
  - Restructure of the docmd_* functions
-pre1
  - Host-based WEP decryption and encryption
  - Partial rewrite of copy_to_bap.
  - Debian package metadata updated to 3.0-STABLE. (David Everley)
  - Stripped out a lot of crap from wlan_compat.h
  - More atmel adapters added to the USB exclude list.
  - James Goodwin's patch to fix a bug in the 80211conv code.
  - The 3COM AirConnect board is a PLX adapter. 
0.1.14
  - Added PCI ident info for the 3Com AirConnect PCI.
  - FCS toggle now (really) defaults to false.
  - Added PCMCIA ident info for the 3Com AirConnect 3CRW737A/3CRW777A
  - David Everly's patch to kernel detection in the Configure script
  - David Everly's patch to add better debian packaging bits.
-pre7
  - Added kernel tags stuff.
  - Added PCMCIA ident info for the Actiontec 802CI2 card.
  - Added PCMCIA ident info for the Proxim RangeLAN-DS/LAN card.
  - Endian "fixes" for PCI PowerMacs.  This fix may be relevant for other
    BigEndian/PCI (and PLX) platforms.
  - FCS in monitor mode now optional; defaults to off.
  - Populate the monitor channel in the prismheader.
-pre6
  - monitor mode now works on USB widgets.
  - Fix an alignment bug in the rx path.
  - Fix a brown-paper-bag bug in the usb flash code.
  - Explicitly enumerate the USB widget endpoints, and use 'em.
  - REALLY fix the wlan.conf overwriting.
  - More minor cleanups in the PCI/PLX init code.
  - endian-ness fixes in the monitor code
  - Applied a patch from David Everly to fix another 2.2.x compile problem.
-pre5
  - Add USB entries for the "Melco WLI-USB-S11 11Mbps WLAN Adapter"
  - Don't overwrite /etc/wlan.conf if it already exists.
  - Compile sniffing code if packet socket is modularized.
  - Fix a typo with the init script.
  - Fix a compile problem with 2.2.x kernels
-pre4
  - First bits of the SMP/locking fixes.
  - Partial rewrite of the PCI/PLX/USB driver internals.
  - Added a makefile for the 'addons/keygen' tool.
  - Fixed a race condition in the USB code.
  - Applied Godmar Back's "SSIDs with spaces" patch.
-pre3
  - Applied pizza's patch to remove the pb_t structure and make all 
    802.11<->ether conversions happen in-place.  Noticeably reduces
    cpu load.  Also includes a few misc bugfixes to the conversion code.
  - Workaround for the kernel panic on bringing up a USB interface
  - Applied Mike Klar's USB patch that adds support for the MIPS
    target and should, among other things, fix WEP under USB.
  - Added Remy Cool's patch to add the Dynalink prism2 card ids.
  - Tim Miller's pcmcia script install path patch
  - Applied Tim Fletcher's patch to allow compilation against
    the newest versions of the wireless extensions. (V13)
  - Added jwyatt's patch for the Compaq Evo N600C USB adapter thing.
  - Per mwelk's suggestion, changed the displaystring totext to single
    quote the strings.  Hopefully this will take care of some of the 
    "SSID w/ space" problems.
-pre2
  - tseward's ID code patch for Siemens SpeedStream USB adapters.
  - pizza's patch for dropping/ignoring frames in promisc and allmulti
    modes.
  - crankin's new Actiontec USB IDs.
  - crankin's fix for wext /proc zero vs. garbage problem.
-pre1
  - pizza's chkconfig comment in rc.wlan.
  - pizza's patch to add true promiscious support to 0.1.13. Also includes
    Experimental ALLMULTI support.
  - pizza's patch to add PF_PACKET capture and raw transmit to 0.1.13-pre2, 
    (with toggles for WEP in sniffing and prism header)
    NOTE: this patch _removes_ the old netlink monitor mode (sniff) interface.
    If you're sniffing, you need to use the latest libpcap (v7.2-pre..from CVS),
    you'll also need to build your own ethereal (0.9.0 or greater) using the
    latest libpcap.
  - bhuang's fix for the hfa384x_test_command() function.
0.1.13
-pre4
  - crankin's revised USB ID patch.
  - Fixed a little buggage in the p80211knetdev_set_mac_address() function.  
-pre3
  - crankin's patch for a bug in usb rrid, DBFENTER/EXIT, and Actiontec
    USB IDs.
  - hmuurimaa's ID patch for the Omnibook500 integrated USB adapter
  - proskin's patch for MODULE_LICENSE conditional
  - dstates (JediElite) submission (others submitted it too) for the
    NDC PLX card.
  - crankin's patch for src/prism2/driver/Makefile. PCI and USB were
    linking with PLX's prism2wext.o file.  DOH!
-pre2
  - Synch'd the etc/pcmcia/wlan-ng scripts with the etc/wlan* scripts.
    This should help PCI/PLX/USB Adhoc users.
  - Added cmulliner's patch for set_mac_address(). Also changed it to
    use dot11StationID instead of the prism2 specific mib item...Oh
    Yeah, I also changed to comments to /*...*/  ;-)
  - Fixed PDA reads and PDA validation.
  - KNOWNBUG: prism2_usb + linux-2.4.17 + uhci.o : Very slow performance.
    We need to look into having multiple posted tx and rx urbs (this is
    also a TODO).
  - KNOWNBUG: prism2_usb + linux-2.2.20 + uhci.o : Physically 
    disconnecting the device causes a Panic.  The logs look like the
    usbdev is still holding some buffers (urbs?) associated with the
    device despite the fact that I've explicitly unlinked them.  This
    problem does not occur on 2.4.17 and, unfortunately, I don't have
    time to debug 2.2.20 this time around.
  - Made all corereset() calls dependent on the prism2_doreset module
    argument.
  - Added corereset() support for all device types.
  - Added the lnxreq_ifstate request message.  This request is used
    to bring the device up to an operable state _after_ the
    [hostbus]_probe() function is finished.  It can also be used to
    shutdown and/or restart the driver+device.  This takes the place
    of the "activate hardware on ifconfig up" code that was introduced
    in 0.1.11.  This is the change that should fix most of our dhcp
    client problems.
  - Massive changes/cleanups to the driver init/shutdown code for all
    device types (pcmcia|pci|plx|usb).
  - Removed the add-ons/mibedit application from the linux-wlan-ng package.
    We'll be re-releasing it as a separate tarball.
  - Modified hfa384x_usb.c per dbrownell's suggestion to mark the request
    urb's with the USB_ASYNC_UNLINK flag.  This because we're calling
    unlink_urb() from the timer callback (intterupt context).
  - Added arichter's patch for the RRID framelength.  It's a good catch
    to make things truly correct.  Unfortunately, I don't think it will
    have any effect.  Because of the 64-byte minimum, the actual RID value
    is being transferred regardless.  Also, I don't think the RRID handler
    in the MAC even looks at the framelen field of RRID frames.
  - Added mwelwarsky's patch for the tx path that may fix some of the
    "queue empty" errors.  Had to update it for 0.1.13.
  - Added new PLX device IDs from cyokoyama's patch, but deferred the rest 
    until we can get a clean patch against the latest release.
  - Added pizza's v2.1 NETLINK enable/disable patch.  Note that this whole
    NETLINK enable/disable thing is only barely tested by me.  I _always_
    have netlink available so maintaining a separate kernel setup just to
    test this feature is a pain.
  - Added jdiedrich's pcmcia card ids for the Compaq ipaq PCMCIA card.
  - Added crankin's patch fixing some more wext induced problems and adding
    the Actiontec USB ids.
  - Added abridgett's patch for PRIV_GENSTRING handling.
  - Added emckee's malloc.h -> slab.h patch and made the mods to the usb
    and wext files he missed. Some kind list soul told me
    that the change to slab.h goes way back in the 2.2.x history, so we
    should be OK.  If this change proves to be a problem for someone with
    an old kernel, please submit a patch with ifdefs for the version where
    things changed (not just the version you're using....thanks).
  - Added jtourrhiles patch for wext reporting level and noise in dbm.
    To enable the reporting in dbm, you must add "#define WEXT_IN_DB"
    to the top of src/prism2/driver/prism2wext.c.
  - Added support for the dbm-normalized RID for CommsQuality in prep
    for working with Jean's patch.  The mibs are called p2dbmCommsQuality*.  
    Note that this will only work for secondary firmware versions 1.2.0 
    and higher. Earlier versions will return with a 
    'resultcode=implementation_failure'.  Note also: The values returned
    are signed 16-bit integers that have been stuffed into unsigned 32-bit
    integers for the trip up to userland.  The following shell expression
    will convert them back to the signed quantity you want:
       if [ $val -lt 32767 ]; then
           echo $1
       else
           echo $[ $1 | 0xffff0000 ]
       fi
  - Added mkershaw's suggestion to lengthen the wait interval for 
    commands in hfa384x.c.
  - Added a note to the README about dleffler's observation regarding
    netlink troubles and using 'make mrproper'.
  - Added gback's patch to initialize wireless extensions for non-pcmcia
    targets.
-pre1
  - Fixed a patch induced (not sure which patch) Makefile syntax error.
  - Added rfloeter's cool Wireless Extensions patch which was kindly 
    updated to 0.1.12 by pizza.  Note that my own testing of this patch has
    been minimal but I've had reports of folks who are using it successfully.
    NOTE: The addition of this patch adds the requirement that Wireless
    Extensions (version >= 10) be compiled into one's kernel.
  - DOH! some somewhat kind soul on the lists pointed out that the keygen
    program wasn't in the distributed tarball.  Basically, I forgot to 
    check it into CVS, therefore the distribution build script couldn't
    find it.  Ahem...it's now in CVS and should be included in the tarballs.
  - Added pizza's patch fixing the "netlink disable" patch.  The 
    p80211ind_* functions need to be handled individually instead of two
    giant #if's, but that's a task for another day.
  - Added theBS's patch for making the /etc stuff installation relative and
    installing the FIXINST scripts for USB.
  - Added crankin's SMP deadlock fix patch.
0.1.12
  - Added dlyddy's string.h includes and comma fix.  Changing all malloc.h's
    to slab.h's will have to wait until I figure which kernel version marks
    the switch so we can ifdef it.
  - Added bgertfield's Corega support patch.
  - I added a check to Configure so we add the CONFIG_NETLINK symbol
    on the compile command line if necessary.
  - proskin's patches for netlink disable when not configured in kernel,
    mkprintstr fix, and turn off skeleton install.  The netlink patch
    pointed out something that might be the root cause of some problems
    with the _cs driver.  pcmcia_cs now strips the kernel CONFIG items
    it cares about and puts them in it's own include/linux/config.h 
    (which is symlinked to include/pcmcia/config.h).
  - Added mtaylor's suggested close() to do_ioctl() in wlanctl.
  - Added mkarmak's suggested change for the GL24110P device ID as
    PCI_DEVICE_GL24110P_ALT and an additional item in the device table
    for PLX devices.
  - Added crankin's suggested MODULE_LICENSE statements.  Had to figure
    out the version ifdefs myself though.  ;-)
  - Added vragor's patch for arm updates to the Makefiles and Configure.
  - Added jsuhr's patch of the PLX9052 irq enable code from the orinoco
    driver.  Note: PLX contributions are highly valued here because we rarely
    use these devices ourselves.
  - Added cfang's PCMCIA ident info for AirEZY card.
  - Made USB port reset in open() an option, DEFAULT IS OFF.
  - Cleaned up the top edge of hfa384x*.c (needs some more).
  - Added the notion of CMD vs. AUX format MAC chip addresses.
  - Modified isgoodpdr() to accept anything less than 0x1000.
  - Updated the wlan-ng scripts for the new open/close behavior.
  - Fixed a bunch of issues that caused bad breakage for hot-plugging
    the USB module.
  - Added a module argument to enable/disable the MAC (or usbport) reset
    performed in prism2sta_open().
  - Moved the port reset corereset() and the call to the open(). This
    gets the port reset out of the probe_usb() context and appears to
    solve the recursive probe problem in some cases.
  - Cleaned up the overlength usb packet problems
  - Added rmemreq/resp and wmemreq/resp support
  - Fixed some of the 64-byte packet problems in a few commands.
0.1.11-usbonly
  - Note: hfa384x*.c are currently undergoing some changes.  I believe
    these changes have broken some things for pcmcia/pci/plx support.  I'm
    not certain since I haven't tested them.  As such, this is a USB ONLY
    release.
  - Cleaned up the use of hfa384x_drvr_*() vs. hfa384x_cmd_*() a little.
    This is a work in progress.
  - Rewrote the USB request/response machine again.
  - Added tnewsham's keygen program to the add-ons directory.  I don't
    include it in the overall build because I'm not comfortable with making
    this package dependent on openssl right now.  It is supposed to generate
    keys in a manner identical to the windows software.  I have not personally
    verified this.  If someone wants to contribute more complete build
    instructions for this program, please do.
  - IMPORTANT! Note that we've changed the up/down behavior.  You now
    must 'up' the linux netdevice interface (e.g. ifconfig wlan0 up) prior
    to using wlanctl-ng.  Also, wlanctl commands aren't available after
    "ifconfig down".  From a low-level perspective, this is how most
    linux netdevices work.  We originally started trying to create a 
    setup where "ifconfig up" on an interface that wasn't associated would
    fail.  With PCI and even more so with USB, that has proven unworkable.
    The idea now is:
      Step 1: ifconfig wlan0 up 
               Enables the hardware so it's ready to handle wlanctl-ng commands.
      Step 2: wlanctl-ng <various commands>, repeat 
               Set up the dot11 configuration and get yourself associated or
               whatever.
      Step 3: ifconfig wlan0 <address>  
               After getting a "real" 802.11 connection, assign the address
               to the interface (via static numbers, dhcp or whatever).

      Note that it is very important that you do _not_ down the interface
      between the initial 'ifconfig up' and the later ifconfig that assigns
      the address.  'ifconfig down' will shut down and reset the hardware.
      If you down the interface, all that good work in step 2 will be lost.
  - Rewrote the USB request/response machine
  - Moved all memmapped/irq interface elements from prism2*.c to hfa384x.c
    in preparation for adding USB support.
  - Fixed 'make clean' so it doesn't generate .depend files.
  - Removed some leftover pcmcia-cs dependencies for non-pcmcia targets.
0.1.10
  - Added module param "prism2_ignorevcc".  If set to non-zero, we just
    accept the Voltage we get from pcmcia-cs without checking.  USE WITH
    CAUTION!!!!
  - Changed the voltage setting code such that we never try to change it.
  - Added NETGEAR PCI info.
  - Fixed a problem with the interrrupt enable in mlme_start().
  - Fixed the netdevice_t problem.
  - Added some new CIS entries.
  - Fixed some byte order things re: dblythe's patch.  The problem went
    a little farther than the patch so I had to change some things.
  - Added lye's fix for the netdevice_t problem on various targets.
  - Added an Ad-Hoc support patch from ????? (please identify yourself,
    it's a nice bit of code and stays within the spirit of things).  Adhoc
    still needs work though.  Remember, only the _first_ STA in an IBSS
    should issue an MLME-Start.request.  All the rest should scan/join
  - Re-added the monitor mode stuff.  At this point I guess Intersil won't 
    be taking it out.  Many thanks to those who maintained the patches and
    my apologies for the hassle.
0.1.9
  - As part of the PCI stuff, rewrote the scripts for 'fixed' (i.e. non-
    hotswappable) cards like the ISL3874 and PLX based PCI adapters.
  - Added support for the dot11req_reset command.  It should perform a
    complete reset of the MAC and driver.  Note that the MAC needs to be
    completely reconfigured from userland after the reset has occurred.
    Also added an internal call to the reset function when unloading a
    PCI or PLX driver.  We need this reset to make sure that the card is
    no longer generating interrupts.  Fixed this problem:  If the driver
    is unloaded, the card is still generating interrupts, and another active
    device+driver is using the same interrupt....system locks up.  DDT.
  - Rewrote the PCI support for PLX and ISL3874.  The rewrite was targeted
    at 2.4.7 with additional code to provide backward compatibility for
    2.2.x.  NOTE:  the PCI code still doesn't support the PCI power
    management stuff.  It will probably need more work when laptops w/
    the ISL3874 based mini-PCI cards show up.
  - Whoops, forgot to put this in the original 0.1.9 release
0.1.9-pre1
  - Added support for the ISL3874 based mini-pci card.  Also cleaned 
    up some of the PLX stuff (needs more work though).
0.1.8
  - Fixed some script bugs relating to WEP and generated keys.
  - Fixed the autojoin auth type control
  - Added/Changed some PDR numbers (thanks to BroMax for supplying hardware)
  - Received a large patch from bjames of Rebel Computing.  Here's a 
    summary in his words:

     > Minor bugs (not already found in 0.1.8-pre10):
     > 
     > 1. p80211metamsg.c:
     > - Wrong conversion functions for "p2req_mmi_read" (value).  enumint ==>
     > int
     > 
     > 2. prism2sta.c:
     > - "dot11_desired_bss_type" in "priv" structure left at 0 (i.e. an
     > invalid value) after initialization so mibget could not retrieve a value
     > unless the network was started.  Initialized to 1.
     > 
     > 3. prism2mgmt.c:
     > - "prism2mgmt_channel_info()" uses HFA384x_RID_SCANREQUEST_LEN instead
     > of ...CHANNELINFOREQUEST...  (However, they're the same size.)
     > 
     > 4. prism2mgmt.c:
     > - "prism2mgmt_auxport_read()" did not set the resultcode to "success" if
     > the read succeeded.
     > 
     > 5. p80211types.c:
     > - "p80211_fromtext_displaystr()" tests for length < maxlen.  This
     > probably should have been <= maxlen (i.e. excluding the \0), based on
     > the maxlen values in "p80211metamib.c" and based in what
     > "p80211_isvalid_displaystr()" checks.  Also,
     > "p80211_isvalid_displaystr()" assumes that "pstr->len" includes the \0
     > when it does the minlen check.  It probably shouldn't do this.
     > 
     > 6. p80211metamsg.c:
     > - dot11_reassociate command has a "dot11req_reassociatefailuretimeout"
     > argument which should probably be "reassociatefailuretimeout".
     > 
     > 7. prism2sta.c:
     > - WLAN_PCI ==> WLAN_PLX in "init_module()".
     > 
     > Enhancements:
     > 
     > 1. All Prism2 RIDs are now accessible.  Unfortunately, rather a lot was
     > changed to do this:
     > - Rather than cut and paste many more cases into the monster
     > "prism2mib.c" switch statement, "prism2mib.c" was completely reworked to
     > be table-driven with a bunch of semi-generic processing functions.  That
     > made it a lot easier to add new DIDs.  Several RID lengths in hfa384x.h
     > had to be changed from 0 to their actual value since more of the lengths
     > get used, now.  (Changed: prism2mib.c, hfa384x.h)
     > - 2 new data types were added to "p80211types.c".  One to handle integer
     > arrays (i.e. multi-word RID records...so that a DID didn't need to be
     > added for each element) and bit arrays (just to make them easier to deal
     > with).  However, there wasn't enough room in the packed DID to add new
     > types, so the type was removed from the DID (and from P80211DID_MKID)
     > and a function written to determine the type based on the conversion
     > functions.  (Changed: p80211meta.h, p80211types.h, p80211meta.c,
     > p80211types.c, mkmetastruct.c, p80211metamib.c p80211metamsg.c,
     > wlanctl/help.c)
     > Integer arrays look like: p2PRIIdentity=21,0,0,3
     > "meta->maxlen" specifies the number of integers.
     > Bit arrays look like: p2ChannelList=0,1,2,3,4,5,6,7,8,9,10
     > for bits 0-10.  "meta->min" and "meta->max" give the range of bits
     > allowed.
     > - The "p2Table" DIDs were split into p2Static, p2Dynamic, p2Behavior,
     > etc., in order to correspond with the Prism Driver Programmer's Manual.
     > A few of the names were also changed in order to correspond to the
     > manual.  (Changed: p80211metamib.c, prism2mib.c)  The changed names (of
     > the supported DIDs) are:
     > p2SystemScale ==> p2CnfSystemScale
     > p2MaxDataLen ==> p2CnfMaxDataLength
     > p2EnhancePowerSaveMode ==> p2CnfPMEPS
     > p2MaxSleepDurection ==> p2CnfMaxSleepDuration
     > p2PMHoldoverDuration ==> p2CnfPMHoldoverDuration
     > p2OwnName ==> p2CnfOwnName
     > p2McastPMBuffer ==> p2CnfMulticastPMBuffering
     > p2TxControlDepth ==>p2CnfTxControl
     > p2RoamingMode ==> p2CnfRoamingMode
     > p2MMLife ==> p2CnfMMLife
     > p2AltRetryCount ==> p2CnfAltRetryCount
     > p2ReserveDuration ==> p2CnfAPPCFInfo (bit 1)
     > p2ExcludeLongPreamble ==> p2CnfExcludeLongPreamble
     > - Added new RIDS from the Prism Driver Programmer's Manual 1.70
     > (2000-11-29).  Added p2CnfPriorityQUsage, p2CnfTimCtrl,
     > p2CnfThirty2Tally, p2CnfEnhSecurity.  Added "algorithm" to AuthRequest.
     > 
     > 2. A user utility program was written to do such things as query all
     > read-able MIBs, query all write-able MIBs, set all write-able MIBs,
     > etc.  A "p2req_enable" command was added (similar to the
     > "dot11req_start" command but without requiring all the 802.11
     > arguments).  This allows the driver to be started in a straight forward
     > manner, simply by setting all write-able MIBs and then enabling it.
     > (Changed: p80211metamsg.c, prism2sta.c, prism2mgmt.c, prism2mgmt.h)
     > 
     > 3. Support added for Prism2 CommTallies and a DID added to query the
     > values.  The values will come back as an array of integers.  Support
     > also added for 32-bit tallies (i.e. p2CnfThirty2Tally).  (Changed:
     > prism2mib.c, prism2sta.c, prism2mgmt.h, p80211metamib.c, hfa384x.h)
     > 
     > 4. Access Point support:
     > - Added support for maintaining a list of authenticated/associated
     > stations and a DID added to query the list.  (Changed: prism2mib.c,
     > prism2sta.c, prism2mgmt.h, p80211metamib.c)
     > - Added support for host control of authentications (allowed and denied
     > lists) and a DID added to set and query the lists.  Support for Prism2
     > AuthenticationRequest information frames.  Not tested yet.  (Changed:
     > prism2mib.c, prism2sta.c, prism2mgmt.h, p80211metamib.c)
     > - Added support for Prism2 PowerSaveUserCount information frames and a
     > DID added to query the value.  (Changed: prism2mib.c, prism2sta.c,
     > prism2mgmt.h, p80211metamib.c)
     > - Added a new data type to "p80211types.c" to handle arrays of MAC
     > address (used to set/query the above lists).  (Changed: p80211types.h,
     > p80211types.c, mkmetastruct.c, wlanctl/help.c)
     > Address arrays look like:
     > p2Authenticated=00:50:51:52:53:54,00:60:61:62:63:64,00:70:71:72:73:74
     > "meta->maxlen" specifies the maximum number of addresses.
     > - Unforetunately, all these lists are limited in length by the length of
     > the mibattribute parameter in mibget/mibset.  This is not very nice but
     > it was the quickest/easiest way to implement it.
     > 
     > 5. Changed maximum mibattribute length from 256 to 384 to support
     > querying commtallies and MAC address arrays.  (Changed: p80211types.h)
     > 
     > 6. 128-bit WEP key support:
     > - Rather than creating different MIB's for 64-bit and 128-bit WEP keys,
     > the "octetstr" definition for dot11WEPDefaultKey0, etc., was changed to
     > allow for a range of lengths (e.g. from 5 to 13).  The
     > dot11WEP128DefaultKey0, etc., MIBs were removed.  (Changed:
     > p80211meta.c, p80211types.c, p80211metamib.c, p80211metamsg.c,
     > mkmetastruct.c, wlanctl/help.c)
     > - WLAN_WEP_KEYLEN changed to WLAN_WEP_MAXKEYLEN and set to 13.
     > (Changed: p80211hdr.h)
     > - Optional "length" argument added to "nwepgen.c" so that either 5 or 13
     > byte keys can be handled.  (Changed: nwepgen.c)
     > - 128-bit-specific parts of "etc/pcmcia/wlan-ng" are no longer
     > necessary.  Genstr changes to handle generation of 13-byte keys using
     > nwepgen.  (Changed: etc/pcmcia/wlan-ng)
     > - WEPDefaultKey's were made write-only.  A new
     > "P80211ENUM_resultcode_cant_get_writeonly_mib" was added.  (Changed:
     > p80211types.h, p80211types.c, p80211metamib.c, prism2mib.c)
     > 
     > 7. Access code in DID:
     > - This was changed from a single bit flag (P80211DID_ACCESS_READONLY and
     > P80211DID_ACCESS_READWRITE) to a pair of flags (P80211DID_ACCESS_READ
     > and P80211DID_ACCESS_WRITE) since some MIBs are read/write, some are
     > read-only, and some are write-only (e.g. WEP keys).  This makes it
     > easier to determine which MIBs are read-able and write-able without
     > having to go all the way to the driver and then interpret a failure
     > error code.  (Changed: p80211types.h, p80211metamib.c)
     > 
     > 8. It was a bit of a hassle switching between the station driver and the
     > AP driver...which we seemed to be doing a lot.  The additional cost was
     > not particulary high to simply compile in both station functionality and
     > AP functionality.    A change was made so that "prism2sta_initmac()" set
     > an "ap" flag when it detected AP firmware.  "prism2mgmt.c" then uses the
     > flag to determine whether or not to allow station-only or AP-only
     > commands to be executed.  "prism2mib.c" also uses this to handle the
     > mode-specific MIBs.  This made it possible for the same module to
     > support either stations or APs, thereby simplifying the "make", the
     > install, and switching between stations and APs.  The only thing that
     > needs to change, is the value of "IS_AP" in "wlan-ng.opts".
     > 
     > 9. Miscellaneous small additions:
     > - Network state (stopped, started) added and a DID to query it.
     > - User-settable comment string added and a DID to set/query it.
     > - Some event logging (authentications, associations, etc.) can be
     > enabled/disabled at run time rather than only at compile time.  A DID
     > was added to set/query it.
     > - A "p2req_readcis" command (similar to "p2req_readpda") was added.
     > 
     > 10. Small changes:
     > - Added bounds checking to "p80211item_fromtext_boundedint".  This was
     > the only "fromtext" function that did not already do complete
     > validation.  It should no longer be necessary to call the "isvalid"
     > function if "fromtext" has just been called.  The call to "isvalid" was
     > removed from "wlanctl".  (Changed: p80211types.c, wlanctl.c)
     > - "p80211item_maxitemlen()" wasn't really adding much of value so it was
     > amalgamated into "p80211item_getoffset()".  The DID type is now only
     > checked once instead of three times.  (Changed: p80211meta.h,
     > p80211meta.c)
     > - "hfa384x_drvr_setconfig16()" and "hfa384x_drvr_setconfig32()" changed
     > their input parameters when converting to hf384x byte order.  Potential
     > source of bugs.  Changed to use a local variable.  (Changed: hfa384x.c)
     > - ScanRequest, JoinRequest, etc., moved from the "Information Record"
     > section to the "Configuration Record" section.  (Changed: hfa384x.h)
     > - Re-worked "p80211_fromtext_setmibattribute()" to make it a little
     > easier to follow the logic.  (Changed: p80211types.c)
     > - Type of "channellist" argument in "p2req_channel_info" command changed
     > from "octetstr" to "bitarray" to make it easier to deal with (can
     > specify a variable number of channels and don't need to use hex
     > values).  (Changed: p80211metamsg.c, prism2mgmt.c)
     > - "auxctl" parameter set to HFA384x_AUX_CTL_EXTDS in
     > "prism2mgmt_auxport_read()" and "prism2mgmt_auxport_write()" so that
     > they are still compatible with 0.1.7.  (Changed: prism2mgmt.c)
     > - Re-worked "prism2mgmt_set_grpaddr()" and added comments so that it's
     > more obvious what is going on.  Also, deleted the check which produced
     > an error if an attempt was made to delete an address from an empty
     > list.  This makes things more consistent and predictable (i.e. the same
     > as attempting to delete an address which is above the last group
     > address...the attempt is simply ignored).  Replaced the "memcpy()" call
     > with "memmove()" since "memcpy()" is not guaranteed to work with
     > overlapping ranges.  (Changed: prism2mib.c)
     > - Minimum length of "dot11OperationRateSet" changed from 126 to 1 to
     > match the 802.11 standard.  (Changed: p80211metamib.c)

  - Fixed some ci/co related problems pointed out by wbinjie.
  - Added ygalayda's startup scripts for PCI in the directory 
    ./etc/wlan_pci.  They aren't installed, that will have to be done
    by hand.  I have not tested these scripts myself.
  - Added sbaer's 'NULL check' fix in p80211_indicate_shutdown().
  - Added ctedrow's 'alloc not checked' fix in prism2sta_int_rx().
  - Added pkundrat's auth mib changes.
  - fgrau's conf.modules/modules.conf fix.
  - Added ygalayda's submission for the Global Sun GL24110P PLX based
    adapter.  Also included a Makefile fix.
  - Fixed some script bugs relating to 128-bit WEP based on smmclusky's
    observations.  These changes have not yet been tested by me.
  - Added rlazarev's additions to the pcmcia startup scripts
  - Added pkundrat's wlandev-name fix for 2.3.99+ kernels
  - Fixed some type/(to|from)text mismatches per note from pkundrat
  - Fixed some missing MKITEMNAMEs per note from pkundrat
  - Added D-LINK and SMC to the linux-wlan-ng.conf file
  - Added Zoom, Linksys and Addtron to the linux-wlan-ng.conf file
  - Added pkundrat's meta-item minlen stuff.
  - Added conditionals for PLX vs. PCMCIA build
  - Changed _pci driver filename to _plx
  - Fixed undconditional MOD_INC in prismt2sta_open()
  - Removed identification of old PCF firmware.
  - Fixed version.h generation bug from pre5
  - Added dhsu's (Eumitcom) code for supporting the PLX based PCI cards.
  - pkundrat's (whoops, sorry peter) PDA display program in ./scripts.  
    I haven't used it because my perl setup is messed up and I'm perl 
    challenged.
  - bcarr's patches for alpha boxes.
  - wwoods's patches for 2.2.18 and adding bytes received to the 
    statistics.
  - proskin's patch adding 'network stop'
  - Most of proskin's patch containing Configure fixes, adding file
    existence checks, and making sure directories are present.
  - tgriffin's patch for mrproper/memleak/nullcheck/freeonsuccess.  
    conf.modules change is already there.
  - Added most of jhicks ARM related patches.  I had to leave out the 
    HOST vs. TARGET compile variables in the Makefiles though.  Some folks
    are using that stuff.
  - Added script changes and new mib items to support 128-bit WEP per hong's
    suggestion.
  - Added conf.modules alias instead of symlink
  - Added a compat macro for kfree_s to wlan_compat.h per the suggestion
    from rbraun.
  - Added pkundrat's 'chinfo' patch
  - Added pkundrat's 'small fixes' patch
  - Added some contributed patches supporting a fix to our 'reassoc
    under load' problem
  - Added wlannoenable pcmcia scheme support to prevent card initialization.
  - Removed a PDR patch requirement.
0.1.7
  - Fixed the "converting eth in unknown mode" bug
  - Added APM/ACPI suspend/resume handling
  - Added PPC support
  - Added Prism test command support
  - Integrated patch to fix some structure alignment problems for ARM
  - Added multi-transmit buffer support
  - And much, much more.....
0.1.6
  - Fixed usage of pcmcia scheme file so we're not stomping on the user's
    scheme setting.
  - Updated the wlan-ng script for the new config variable names and added
    configurable download utility and a set of user MIB settings.
  - Added support for the Samsung and Z-Com cards.
  - Changed the name of the module (from pcmcia's point of view).
  - Fixed the "echo \c" problem in Configure
  - Added new and rearranged existing wlan-ng.opts configuration variables.
  - Created a scheme for "addon software" in the ./src/Makefile
  - Made some changes to the target detection conditionals in wlan_compat.h
    to better support LinuxPPC (port isn't finished yet).
  - Finally fixed the byteorder stuff to use (duh) byteorder.h
  - Fixed a bad free of the TX skb when we signal an error to higher layers.
  - Moved some code around and added rules to the prism2/driver (new directory
    too) Makefile.  This is so we build two separate drivers; one for STA,
    one for AP.  Both drivers are built from the same source files with
    the code conditional on the WLAN_AP and WLAN_STA defines.
  - Fixed the plugging in prism2dl so it reports ALL missing PDRs.
  - Added an anonymously contributed patch that adds support for the -m
    and -s cmdline options for prism2dl.  Thanks! You know who you are.@-)
  - Added some fixes to the dependency file generation.
  - Fixed a flash programming timeout miscalculation...and fixed it again 
    when I introduced an integer promotion problem.
  - Added the STA vs. AP conditionals to prism2mgmt.  Note that this causes
    a few compile time warnings.  They're harmless and will go away in the
    next release.
  - Created the file prism2mib.c to hold all the MIB get/set stuff which is
    getting a little out of hand.
  - Added pcmcia config support for the Z-Com (supposedly) dual voltage card.
    Unfortunately the engineering sample I have doesn't seem to work at 5v,
    or it might be the code.....not sure yet.  We're still a little
    bleeding edge with that card.
  - Cleaned up the version, compatibility range, and serial number 
    log messages.
  - Added a loop limit to the interrupt handler. (duh)
  - Changed the names of the driver modules for STA and AP.  make install
    then creates a softlink depending on the make config selection.  One
    note, both drivers report the same dev_info string to pcmcia-cs.  Trying
    to load both of them simultaneously would be bad (I have no idea what 
    would happen, I just thought of it).
  - Changed a bunch of types and constants for hfa384x.
  - Added functions for a couple of new commands in the hfa384x.
  - Began adding support for the Prism2 unique MIB items.