linux-wlan-ng/CHANGES
Solomon Peachy 3d7a066a27 Fix oops in wlan_setup on 2.6.27-rc3
oops caused by not setting up network device queues.

Change wlan_setup to use the kernel provided alloc_netdev() rather than
hand coding it. This should help avoid any future problems.

Also change register_wlandev to not call dev_alloc_name() as
register_netdev() will do that if the device name has been setup.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
2008-09-30 11:01:09 -04:00

1269 lines
66 KiB
Plaintext

* 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.
*
* --------------------------------------------------------------------
- Fix oops due to 2.6.27 API change (Richard Kennedy)
- Fix compiles with 2.6.26 (Pavel Roskin)
- Add compatibility with 2.6.27+ WEXT API (Pavel Roskin)
- Fix misplaced variable in PCMCIA code (Pavel Roskin)
- Add compatibility for gfp_t and mutex APIs (Pavel Roskin)
- Fix non-posix find arguments in script (Mark Mathews)
0.2.9
- Reduce stack usage in prism2mib_priv (Richard Kennedy)
- WEXT functions should return -EBUSY when the hardware is still
being initialized (Richard Kennedy)
- Use round_jiffies wherever possible (Richard Kennedy)
- Disallow ioctls from running until the hardware probe function has
finished. (Richard Kennedy)
- Fix userspace scripts to work on Fedora 8 (Chris Rankin)
- Perform the hardware reset/init sequence *before* we register
the netdevice. (Karl Renton)
- Always do an ifstate_disable before a firmware load cycle.
- Default to wireless extension mode.
- Make the SIWENCODE behaivor more closely match the spec. (Tormod Volden)
- Apply a patch from Karl Relton that improves USB reliability.
- When in wlan_wext_write=1 mode, auto-transition to ifstate_enable
and bypass userspace firmware load altogether.
- Fix bug in the GIWENCODE handler (Richard Kennedy & Tormod Volden)
- Support 2.6.24 kernels (modified patch from Tormod Volden)
- Support 2.6.23 kernels. (With apologies to Martin Dauskardt for a
misapplied patch)
- Support "Intersil Americas USB 802.11b WLAN DEVICE" (Peter Levart)
- Tweaks to support Fedora 7 (Dan Williams)
- Support 2.6.22 kernels
0.2.8
- Fix up a pile of sparse warnings (Pavel Roskin)
- Support 2.6.20 kernels (Pavel Roskin)
- Tweaks to the udev rules (Richard Kennedy)
- Silence a large pile of warnings with GCC4.1
- pt_regs is gone for USB as well on 2.6.19+ (Chris Rankin)
- Large WEXT patch that allows use of NetworkManager. (Richard Kennedy)
0.2.7
- Fix license text in prism2dl.c (it is in fact MPL/GPL)
- Fix oops in wext_autojoin (Richard Kennedy)
- Report use of dBm for WEXT requests (Richard Kennedy)
- Configure script better handles PREEMPT kernels (Chris Rankin)
- 2.6.19+ uses a different IRQ handler function prototype.
- linux/config.h is no more on 2.6.19+ kernels.
0.2.6
- Fix another error that affected <2.6.15 builds on some compilers
- Fix a cut-n-paste error that broke 2.6.18 builds.
0.2.5
- 2.6.18-rc fixes.
- A few tweaks for Wireless-Extensions-21
- Fix an interrupt-related race that could lead to a crash on shutdown.
- Tweaks to the prism2_usb code to improve suspendability on 2.6.15+
- Silence many warnings under GCC 4.1
- prism2_cs support for 2.6.16+ (untested)
- Makefile fixes for 2.6.17+
- udev support for USB targets (Richard Kennedy)
- semaphores are deprecated as of 2.6.16 (Chris Rankin)
- prism2_usb build fix for 2.6.16
- Patches from Debian (via Victor Seva)
- Fix bugs in /etc/wlan/shared, including a security hole
- Fix build on 2.4.27 kernel
- Added man page for prism2dl
- Generate a WEXT Event when we gain/lose association.
- Make prism2_defer query ssid and bssid, as IBSS mode has this change.
- Clean up prism2_defer_timer a bit.
- Propogate linkstatus into p80211, and make ethtool aware of it.
- Eliminate wland and entire indication infrastructure. If you want to
play with it, use 0.2.4. We can put it back in if necessary.
0.2.4
- USB Build fix from Chris Rankin.
- Add ability to install prism2 firmware seperately. (Petteri Raty)
- Theoretically the per-frame signal/noise figures are now in dBm.
0.2.3
- WEXT 16+ fixes, which came to a head with v2.6.14+ kernels.
- Fix a circular symbol dependency glitch
- Fix a potential panic when receiving unknown CONTROL frames.
- get rid of p80211frm.c entirely; it's all unused code.
- Fix prototype for pci_suspend functions to match 2.6.11+ APIs.
- Build fixes for 2.6.14+
- Change the type of 'request_pending' to take care of warnings.
- Fixes for warnings generated by the 'sparse' tool (Richard Kennedy)
- Add an explicit README.firmware containing copyright information
on the firmware images.
- Compile fixes for 2.6.8 (Victor Seva Lopez)
0.2.2
- Fix an array overflow in the hotplug helper function.
0.2.1-final
- Added in-kernel devicetable for 2.6.13+ pcmcia
- Updates to bring prism2_cs into the newer 2.6 kernel APIs.
- Fixes in the hotplug invocation with WEXT (Pavel Roskin)
- Experimentally try using usb_reset_device on 2.4 kernels too
- use usb_reset_device() on 2.6 kernels (Andreas Schulte)
- Included kludged 1.8.4 PCI/PCMCIA firmware (Pavel Roskin)
- PCMCIA card Ident fixes to coexist with orinioco driver (Pavel Roskin)
- Ignore unexpected ctlx responses. (Chris Rankin)
- Fix an off-by-one in scan results.
- Change some log priorities
- [usb] Fix a memory leak (Chris Rankin * Eroc Koenders)
- [usb] clean up the PDA Read function a bit. (Chris Rankin)
- [usb] Don't overwrite current CTLX URB before we're done with it.
(Eric Koenders & Chris Rankin)
- Fix a byte-order problem when parsing scan results (Eroc Koenders)
- Added a mib item (lnxRSNAIE) to get/set the WPA/RSNA IE.
- Applied a patch from Chris Rankin which should solve the hangs with
kernel preemption turned on.
- Cleaned up some compile warnings with GCC 3.4.x
- Add a background thread to fetch comms quality from hardware.
- Fixed a series of bad memsets in prism2dl
- Added id for Viewsonic USB widget (Ekin Meroglu)
- Eliminate the extra frame copy in the usb rx path.
- When issuing a start/join, set MACMODE to none as to disable
transmits through the hardware.
- Get rid of the prism2_pci/plx 2.2.x compatibility code (Pavel Roskin)
- Change default behaivor of top-level Makefile (Pavel Roskin)
- Get rid of the #ifndef DECLARE_TASKLET cases; 2.4+ supports it.
- Fetch the BSSID when we get an AP_CHANGE event
- netif_carrier_off/netif_carrier_on at appropriate times
- Hook up the dot11CurrentTxPowerLevel MIB in the prism2 code.
- Add a default TMPDIR in case it isn't set.
- Removed some 2.2.x cruft
- CommsQuality now does the right thing with dBm values.
- More makefile work from Pavel Roskin
- A major USB patch from Chris Rankin. This rewrites the
command-queueing code to something considerably more sane.
- A few more Makefile tweaks
- Further compile warning fixes on 64-bit targets (Pavel Roskin)
- Fix prism2_cs on 64-bit targets (Pavel Roskin)
- Fix the 2.4 build (whoops)
-pre26
- Fix the "no hardware found" bug in 2.6.10+ kernels
- Cleanups in the Makefiles and build system
- Fix a typo in the shared.prism2 script (Chris Rankin)
- Clean up build warnings with gcc 3.3 (and possibly older too)
- Fix a nice race condition on device registration.
- Don't stop the netdev when we hit ifstate_disable.
- hw->state wasn't being set properly on drvr_stop, nor were we
clearing the port status in all cases.
- Fix the "Debug: sleeping function called from invalid context"
under 2.6 with prism2_usb (Chris Rankin)
- Wireless extension SCAN support (Dan Williams)
- 2.6 sysfs support in the form of SET_NETDEV_DEV (Dan Willimas)
-pre25
- Fix a brown-paper bag bug in the register_wlandev code. (Giacomo Lozito)
- Patch for the Sitecom WL-022 widget (Armijn Hemel)
- Build fix for ancient ETHTOOL versions
- build fix for old 2.4.x kernels
-pre24
- Fix build on 2.6.10-rc
- Fix USB build on 2.6.x (x < 7)
- Suspend the USB driver cleanly (Colin Leroy)
- Fix more USB badness (Colin Leroy)
- scan_timeout WAY too long. (Wolfram Gloger)
- dev_get is no more, as of 2.6.10-rc2 -- switch to dev_alloc_name
instead, which is present in older kernels. (reported by Colin Leroy)
-pre23
- Fix badness in 2.6.10-rc1 (Colin Leroy)
- Compile fix for 2.6.9-rc4+ (Jeff Chua)
-pre22
- Add support for x86-64 (and possibly other 64-bit targets) (J.D. Cole)
- Fix linkstate if driver reset without unloading. (Shiro Ninomiya)
- Rename the man pages to have a .1 extension
- Tweak the Configure script; change the default kernel build directory
- put a 'sleep 1' in the firmware load sequence to improve reliability.
(Arnold Liu)
- Add Linksys WUSB11 card ident
-pre21
- Add a per frame rx structure for OOB data to be passed to p80211
This is a step in the right direction for a major overhaul of
internal frame representation. The prism2 driver shouldn't care
about sniff headers or any of that crap; it should always pass
everything to p80211 (via this metadata) and the p80211conv_*
functions should do that work. It's also needed to eventually
support different encryption types.
- Added in beginnings of IWSPY support via patch from Josef Kriegl
- Install modules to $MOD_DIR/linux-wlan-ng instead of $MOD_DIR/kernel
- Primary/Secondary firmware (finally) bundled with the driver.
- Convert to newer 2.6 module parameter code.
- Fix an inverted test in GIWENCODE wireless extension (Shiro Ninomiya)
- Minor script fix for SSF cards (Pavel Roskin)
- Card idents.
- Fixes for SSF PCI cards; default to x16 SRAM as that's more common.
(download script now automatically detects hardware type)
- Fix script problems that result in a failed init returning success.
- Fix an endian-ness buglet in prism2sta.c (Federico Pellegrin)
- Compile fixes for 2.2.x kernels. (no tasklets or spin_lock_bh)
- Fix up pci resources on card init failure on plx card.
-pre20
- Fix up pci resources on card init failure
- Ident for "ROPEX FreeLan 802.11b USB Adapter" (Francesco Bochicchio)
- ETHTOOL support now optional, fixes build on older kernels.
- Fix prism2dl to not use kernel headers to fix 2.6.3 build (Pavel Roskin)
-pre19
- Add a 'lnxind_roam' indication.
- minor typo in the nwepgen makefile (Pavel Roskin)
- More makefile fixes (Pavel Roskin)
-pre18a
- brown-paper-bag fix for the 2.4 build. And it no longer builds on
2.2 thanks to an accidental makefile commit.
-pre18
- Fix broken makefiles so that prism2dl and the rc script install
- Get rid of the txfid lock, it's not necessary.
- Tasklet-driven RX handlers.
- Further work towards a Wireless Extension-enabled future.
- SIWFRAG, SIWRETRY, SIWFREQ, SIWRTS implemented
- SIWENCODE, SIWESSID taken from zdsta-1.0.3 (needs work)
- SIWCOMMIT Implemented ala autojoin.
- write support defaults to disabled, override via module parameter
- fix a warning in the wireless extension code.
- hfa384x.h header updates
- Fix prism2dl to handle the S7 record being elsewhere in the image.
(Pavel Roskin)
- Fix up brown paper bag script error for SSF hardware.
-pre17
- Spruce up the GIWNAME wireless extension ioctl. (Natsakis Konstantinos)
- Added in basic ethtool support; for link status reflection.
- Fixed a double-locked spinlock which deadlocked on SMP boxes.
- /proc/net/p80211/wlanX/nsd now has useful information about the
prism2 hardware/firmware revisions.
- eliminate the unused 'collptr' from the mib structures.
- Fix a bug with the pcmcia script and scanning (Dan Noe)
- Fix a crash on a VCC mismatch.
- Split out most pci/plx/pcmcia/usb specific code into their own .c files.
- Make prism2_cs build on 2.6.2-rc
- Changes to support SSF hardware.
- Card ident for D-Link DWL-650P1
- Move the prism2 link notification handler operate in scheduler
context to avoid a hang in softirq context.
- Melco WLI-USB-KB11 USB widget added (Bela Fenyvessy)
- Add in x86-64 support, plus other cleanups in wlan_compat.h
- Support the Netgear MA311 PCI card. (Steve Berry)
-pre16
- Implement wireless extension SETCHANNEL. (SIOCSIWFREQ)
- Currently broken. :)
- Fix several sniffing-related bugs:
- disable when we're already disabled
- no longer perform a full h/w reset when we're already enabled
- Vastly reduce CPU usage on channel hop.
- Fix a crash on unload for pci/plx widgets (Pavel Roskin)
-pre15
- Minor fix for the shared script (Jeff Chua)
- fix "bad: schedule while atomic" bug introduced by recent 2.6.0-test
kernels and wireless extensions.
- Switch to using spin_[un]lock_bh for the cmdlock; this way we
disable the transmit bh from running while issuing a card command.
Also have the txframe function obtain the lock.
- Hawking HighDB USB Widget (Allan Claghorn)
- Fix a double-lock in hfa384x_cmd_notify
- Identity for "T-Sinus 111 USB WlAN Adapter" (Roman Koutny)
- Fix up alignment problems on frame rx. (Ranjit Deshpande)
-pre14
- Update the HAS_HOTPLUG test to something considerably more foolproof
- Add ids for the Airvast Prism3 USB widget
- bap locking was broken when illegal parameters were passed. (Andre McCurdy)
- Don't use jiffies for timeouts when interrupts are disabled. (Andre McCurdy)
- Make the Configure script a little saner.
- Remove the Zydas 1201 USB widget. Use the Zydas modified
linux-wlan-ng release instead (we hope to merge it in eventually)
-pre13
- Fix two big-endian bugs in the scan code.
- Hotplug agent executes '/sbin/ifup INTERFACE' if possible.
- Planex GW-US11H USB adaptor (Henry Ip)
- Averatec USB Wlan Adapter (Stephan Miller)
- D-Link DWL-122 USB widget (Andrew Beresford)
- ASUS-WL140 Wireless USB widget
- Fix the WLAN_LOG_NOTICE macro for gcc 2.95.x
-pre12
- Fix up endian bugs in lnxerq_commsquality.
- Get rid of the pcmcia release timer for >2.6.0-test5
- Card ident for 'ZyDAS ZD1201 Wireless USB Adapter' (Albert Pauw)
- More tweaks for 2.6.0-test4
- fix some bash-isms in /etc/wlan/shared (Daniel Song)
- Collapsed the BOUNDEDINT and INT types into one.
- Eliminated the WLAN_LOG_***0() macros at long last thanks to CPP
trickery. Also eliminated more cruft from wlan_compat.h
- Card ident for the Samsung SWL-2210P PCI card
- Try not to schedule a USB endpoint reset if there's one already
pending. (Chris Rankin)
- Fix pci/plx operation on architechures with >32bit physical address
sizes. (namely mips64 & some mips32)
- get rid of save_flags/cli/restore_flags in prism2_cs on >2.2 builds
(tweaked from patch by Pavel Roskin)
- pci_device_id tables should not be marked __devinitdata
-pre11
- support genesis mode for RAM download of primary fimware
- prism2_usb now builds on 2.5/2.6.
- endpoint numeration removed.
- hfa384x_usb_corereset unimplemented for 2.5/2.6
- hfa384x.h updates.
- p80211conv_* functions should work with netdev->mtu, not MAX_ETHFRM.
- Simply erase the bogus '3rdparty' drivers on Mandrake systems.
- Card idents for Adaptec AWN-8030/AWN-8020, and Netgear MA111
- Strip out all 2.2.x stuff from prism2_usb, it doesn't work.
- add proper __devinit/__devexit/etc tags to prism2sta.c
- Minor USB janitorial patch from Chris Rankin.
- Make most card commands require superuser access (exception is
mibget) (Original patch by Pavel Roskin)
-pre10
- Make the wakeup flag in the usbctlx structure volatile, as well as
other wait condition variables.
- Make the rx_urb buffer dynamically allocated so we can re-post the
URB as soon as possible.
- Get rid of completion.h
- More wlan_compat updates.
- Disable TXEX reporting on USB as well as TXOK. should improve
stability a tad.
- Further 2.5 USB work. Not quite there yet.
-pre9
- Properly initialize the prism2_cs release timer (Pavel Roskin)
- Eliminate the CONFIG_ISA requirement for prism2_cs (Pavel Roskin)
- netlink build fixes for 2.5.71
- Hopefully fix the 2.2.x build.
- More USB fixes (Chris Rankin)
- Compile fixes for < 2.4.20
-pre8
- Fix the pcmcia breakage introduced in -pre2.
- On USB widgets, only signal txcomplete if there was a tx error.
- Minor cleanups in the PCMCIA code.
-pre7
- Add in the 'JVC MP-XP7250" USB Widget.
- Don't call ev_alloc on tx_complete event; only the tx urb completion.
- Minor tweaks to tx_timeout code
- More USB updates; primarily with error-handling. (Chris Rankin)
-pre6
- On USB transmits, don't return 'success' if the tx_urb is still
pending. Also don't wake the netdev queue if the tx fails.
- Hook up the tx watchdog timer and add a p80211-level handler.
- Fix compilation with older (<= 2.4.18) kernels.
- Another massive USB+other patch from Chris Rankin:
- a separate list for "finished" CTLX objects.
- correct (hopefully!) handling of -EPIPE errors, shamelessly copied
from^W^Winspired by David Brownell's usbnet driver.
- finished support for asynchronous commands, with proper tracking
through the CTLX lifecycle, clean-up etc. Each asynchronous command now
completes in its own tasklet.
- CTLX cancellation and error handling.
- further work on the timers and spin locks.
- clean up setconfig16() and setconfig32() functions so that they
take appropriate value parameters, even if the underlying
setconfig() commands don't.
- use an asynchronous command in setmulticast(). (CS, PCI and PLX
are still synchronous; they just pretend not to be.)
- further clean up in the disconnect() handler. However, it's still
not a good idea to remove the adapter while the
prism2sta_inf_hostscanresults() function is sleeping, and the wlandevice
struct could easily have been freed before the p80211req_dorequest()
function manages to clear the request_pending bit. It would be
nice if we could cancel both of these operations when we call
prism2sta_ifstate(P80211ENUM_ifstate_disable).
- extra support for Linux 2.5 (task queue -> work queue).
- reset the link status when we stop the driver.
- Some work on module locking. Not there yet, but getting better.
- Remove magic numbers from USB endpoint probe (Chris Rankin)
- Eliminate useless timer in p80211do_request, as all calls are
synchronous. Also clena up more function pointers, and a few other
doodads. (Modified from patch by Chris Rankin)
- Track the current SSID in the wlandev; Make the WEXT stuff key from
this. Update the SSID when we get a linkstatus_connect notification.
- A couple of misc cleanups from Chris Wilson
- Cleanups in the 2.5 makefile code. (Pavel Roskin)
- Add in another PCI device ID, to handle Sony VAIO PCG-V505AP
- Make sure CONFIG_ISA is enabled for the prism2_cs build. (x86 only)
-pre5
- Get rid of generic manfids in /etc/pcmcia/wlan.conf (Pavel Roskin)
- Fix the 'make install' target.
- only call flush_scheduled_tasks() if DECLARE_TASKLET not defined.
(fixes 2.5.x compile)
-pre4
- More script work; mostly cleanups and other enhancements.
- Substantial prism2_usb patch from Chris Rankin.
- using 2 Linux lists for CTLXs
- remove race condition and double-free with CTLXs in disconnect()
- initialise URBs correctly (vital for Linux 2.5+)
- separate callbacks for data/CTLX OUT URBs
- flush task queue on shutdown
- more thorough use of spinlocks with CTLXs
- set the ASYNC_UNLINK flag when it is time to unlink an URB
asynchronously, and not before.
- ensure we stop submitting URBs once the disconnect function
is called.
- More USB idents (Jeff Chua)
- Scan now uses active scans only; should speed things up a bit.
- Proper module init/cleanup in prism2sta.c, plus more work for 2.5.69
- Added in descriptions for module parameters.
- Major overhaul of the hotplug stuff. Now we use our own hotplug
event class, and use that to kick off configuration and whatnot.
All device types (pci/plx/pcmcia/usb) now use this mechanism.
Also brings us things like unified power management (suspend/resume)
- Supress spurious output when setting WEP parameters.
- Only query PrivacyOptionImplemented if wep was enabled in the config
Works around an apparent firmware bug, see http://bugs.debian.org/190851
- Eliminate hw->name.
- Card ident for NL-2511CD Plus pcmcia card.
- Minor tweak on pcmcia removal.
- Fix the USB compile on <2.4.20
-pre3
- get rid of wlandev_get_index and whatnot.
- URB cleanup in the USB code. (Chris Rankin)
- Moved host auth sequence out of interrupt context.
- USB paranoia checks (Chris Rankin)
- Partial rewrite of the USB shutdown sequence
- eliminate wlandev->hwremovedfn().
- Minor tweaks to the mib.
- Get rid of the horridly outdated skeleton driver.
-pre2
- Fix a crash-on-unload affecting pci and plx adapters
- have the init script load the module if needed.
- Handle the new 2.5.69+ irq handler semantics
- Fix 2.5 with pcmcia (Pavel Roskin)
- Massive cleanup of the prism2/driver/Makefile (Pavel Roskin)
- Don't wrap the [un]register_netdevice() calls in rtnl_[un]lock()
And use the [un]register_netdev() calls instead.
- Call the hwremoved() functions on pci/plx_remove
- Eliminate the prism2sta_priv_t structure.
- Fix a memory leak on ifstate_disable
- Minor changes to the USB driver
- Widget id for "Melco WLI-USB-KS11G" (Ichiro Doi)
- Function pointer syntax cleanup (Chris Rankin)
- Push the multicast/promisc set into the prism2 driver.
- Move type/multicast filtering into the p80211 layer.
- Format fix for p802addr_to_str (Pavel Roskin)
- Add a check for non-Intersil firmware.
- Fix a infinite recursion in the usb lowlevel code. (Chris Rankin)
-pre1
- Header updates.
- Add a card ident for the USR 1120 USB widget (Juan Conde)
- Cleaner shutdown; fixed a possible leak and a race condition.
- Another rearrangement of the command locks, to eliminate use of the
spin_is_locked() call.
- Fix a rare deadlock with cmd_initialize (Pavel Roskin)
- Add a Lucent-compatible key-generator in add-ons/lwepgen
- Fix a stack overflow problem. (Pavel Roskin)
- Further monitor mode fix for control frames (Pavel Roskin)
- Cleanups in the dependency generation code
- Initial 2.5 support. USB currently broken.
- Minor build system cleanups.
0.2.0
- Use PSUEDOIBSS mode for monitor mode; this should eliminate spurious
beacon generation. (Ryan Veety)
- Don't drop "unrecognized" frame types in monitor mode.
- Added ident for the Linksys WUSB12 (Brad Davidson)
- Eliminated most of the WLAN_LOG_???n macros.
- Minor fixups in /etc/wlan/shared
- Cleaned up some of the error messages in the conversion code.
- changed the "Host de-WEP failed" error to a debug message, as the
/proc/net/wireless stats are updated to reflect undecryptable frames
-pre10
- Added the p2TxPowerMax MIB item to specify the TxPower level.
0-30 dBm. Needs STA>=1.7.0 or AP>=1.4.0.
- Clean up some spurious warnings in prism2sta.c (Pavel Roskin)
- A series of patches from Pavel Kankovsky, somewhat tweaked. :)
- Properly set skb->mac.raw in non-monitor mode
- Enhancments to the p80211 frame conversion code
- Handle A4 frames.
- Don't issue linkstatus notifications in monitor mode
- Supress Linkstatus messages in monitor mode
- hfa384x.h updates for latest firmware.
-pre9
- ZyXEL ZyAir B200 Wireless USB widget ID added (Paul Lacatus)
- Only enable interrupts in one place (hfa384x_drvr_start)
- Fix the startup scripts to disable WEP completely if not enabled.
- ALLNET 0193 USB widget ID (Kurt Huwig)
- Disable the hfa384x port before issuing the autojoin stuff.
- Fixed a subtle flaw in the 802.11->802.3 conversion code.
- Updates to hfa384x.h
- Add productinfo for dlink DRC650 from Petr Slansky.
- Minor changes to the rfmon code.
- Only notify us on linkstatus changes if it actually changes.
- Fixed the path the manpages were installed to.
-pre8
- Properly reset the hardware state after a flash/ram download.
- Patch to allow escaped characters in the scan list (Derek Atkins)
- Re-added the txfid queue locking which somehow got dropped.
- Added DellTrueMobile 1180 USB ident. (Michael Hackett)
- Changed linkstatus messages to INFO level reporting.
- Ident info for the Intel Anypoint II 802.11b PCMCIA card.
- 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.