Legacy driver status

Ben Hutchings ben at decadent.org.uk
Sun Jun 3 03:10:37 UTC 2007


Here's my summary of changes made in CVS based on reading both commit
messages and diffs.  This should aid a decision on whether and where to
take new CVS snapshots or selected changes.

rt2400
2006-07-25: Readd proper module support - patch from Kel
2006-08-23: Fix register length defines.
2006-11-30: Fix compile isssue for recent kernels.
            Further kernel 2.6.19 compile fixes.
2006-12-27: Fix essid truncation on kernels >= 2.6.19
2007-02-20: Add debugfs support.
            [Replaces register and EEPROM access via ethtool.]
2007-03-21: RFE 1512276 - belated Cruft Removal at Bryan's suggestion.
2007-03-30: Bug 1689965 - Change all char definitions to signed char to resolve ARM compile issue
2007-04-25: "Back-porting" the rt61 patch against auth timeout kernel freezes on SMP systems
            Further backporting of rt61 patches against kernel freezes on SMP systems.
            This patch corrects wrong locking mechanisms around the MLME queue.
            [Also fixes authentication timer reset.]
2007-05-03: Remove fake MlmeEmptyQueue error message
2007-05-05: Received one too many RaConfig support request on the forum...
            Enforces what we decided earlier: promote RutilT and get rid of RaConfig altogether.
2007-05-05: [Improvements to TESTING documentation.]
2007-05-07: [Improvements to TESTING documentation.]
2007-05-12: Code cleaning (step 3), quick hacks to get rid of the compile warnings.
            [This seems like a really stupid change.]

rt2500
2006-07-13: 256bit key support in ADHOC patch by bbatten
2006-07-13: Documentation cleanups from bbatten
2006-08-15: [Fix definition of SHORT type.]
2006-08-15: Setting default key id patch by bbatten
2006-08-19: Missing MLME commands (bbatten)
2006-08-23: Fix register length defines.
2006-08-29: Fix to RSN IE WPA problem (patch by bbaatten)
2006-11-28: This fixes the "RSN IE mismatched" problem in message 3 of the EAPOL exchange.
            It handles the case where a beacon or probe response frame sends either a WPA1 IE followed by a WPA2 IE, or a WPA2 IE followed by a WPA1 IE.
            Patch by Bryan Batten
2006-11-30: Further compile fixes for 2.6.19.
2006-12-27: Fix essid truncation for kernel versions >= 2.6.19.
2007-01-09: Improve iwpriv interface.
            [Split RFMONTX ioctl into SET/GET pair.  This may be an incompatible change.]
2007-01-09: Patch #1579836 - iwpriv_usage cleanup (from bbatten)
2007-01-20: 2.6.20 kernel fixes for rt2500 (patch by Tobias)
2007-01-22: Compile fix.
2007-02-20: Add debugfs support.
            [Replaces register and EEPROM access via ethtool.]
2007-03-17: In-kernel compile support (Bug #1642144): Declare debug var as static.
            Implement as bit mask.
2007-03-30: Bug 1689965 - Change all char definitions to signed char to resolve ARM compile issue
2007-04-25: "Back-porting" the rt61 patch against auth timeout kernel freezes on SMP systems
2007-05-05: Received one too many RaConfig support request on the forum...
            Enforces what we decided earlier: promote RutilT and get rid of RaConfig altogether.
2007-05-05: Bypass MlmePeriodicExec processing when BSS_SCAN_IN_PROGRESS
2007-05-05: Don't allow BSS scans when interface is down (MLME not initialized)
2007-05-05: [Improvements to TESTING documentation.]
2007-05-06: Another timer soft_lockup identified: timer trying to del_timer_sync itself (this time it is a scan timer), so i generalize the timer_pending test.
2007-05-07: [Improvements to TESTING documentation.]
2007-05-12: Code cleaning (step 3), quick hacks to get rid of the compile warnings.
            [This seems like a really stupid change.]
2007-05-13: Error message on failed to resume
2007-05-15: Upgrade base code to the Ralink driver version 1.4.6.6.
            [Removed potential null pointer dereference.
             Added packed attribute to some structure definitions.
             Fixed ordering of descriptor writes on big-endian systems,
             though this is subject to reordering by the compiler.
             Added checks for interface down in scanning functions.
             Corrected offset of variable-length information elements in NDIS_WLAN_BSSID_EX.  Size is still calulcated incorrectly!
             Added error message on failure to read EEPROM.
             Changed signal quality calculations.
             Added check to ignore AP beacons when in ad-hoc mode.
             Implemented pairwise WPA keys properly.
             Corrected IV16 on big-endian systems.  Maybe.]
2007-05-31: SIOCGIWSCAN returns EAGAIN until all channels scanned.

I need to go through these to work out where everything broke between
2006-06-20 and 2007-02-19.  The RSN IE changes look like the most likely
culprits.  I can't see why they would cause complete breakage for our
users though without this being detected upstream though.  Maybe a
difference in authentication mode or kernel version?

rt2570
2006-06-20: Remove rt73usb USB ID. This had been accidently added in the past.
2006-07-13: MAC changing patch from ASPj
            [Avoid resetting MAC address on device reopen.]
2006-07-13: Wireless stats fix from bbatten
2006-08-23: Fix register length defines.
            Read register from correct offset.
2006-08-26: Fix register/eeprom reading for ethtool.
2006-10-06: [Added USB device id for Siemens 802.11g USB Wireless Network Adapter.]
2006-11-30: Fix compile isssue for recent kernels.
2006-12-27: Fix essid truncation on kernel versions >= 2.6.19.
2007-01-09: Improve iwpriv interface.
            [Split ADHOCMODE, RFMONTX and FORCEPRISMHEADER ioctls into
             SET/GET pair.  This may be an incompatible change.]
2007-01-09: [Prevented building in SMP configurations which will be broken.
             This resulted in reporting of bug #423127, but the bug was
             present already.]
2007-02-20: Add debugfs support.
            [Replaces register and EEPROM access via ethtool.]
2007-02-23: Romain's patch for Rutil
            [Move RTPRIV_IOCTL_GETRTPRIV around.  I don't know quite what the effect is.]
2007-03-28: In-kernel compile (Bug #1642144) - static debug mask.
2007-03-30: Bug 1689965 - Change all char definitions to signed char to resolve ARM compile issue
2007-04-25: "Back-porting" the rt61 patch against auth timeout kernel freezes on SMP systems
2007-04-25: Further backporting of rt61 patches against kernel freezes on SMP systems.
            This patch corrects wrong locking mechanisms around the MLME queue.
2007-05-05: [Improved TESTING documentation.]
2007-05-06: Generalize the timer_pending test to avoid soft lockups when a timer tries to del_timer_sync itself.
2007-05-07: [Improved TESTING documentation.]
2007-05-13: Update base code to the Ralink 2.0.8.0 driver version.
            [Fixed MD5 calculation on big-endian systems.
             Added support for WPA PSK and WPA2.
             Added support for RSN IE.  (Probably part of the above.)
             Added handling of Cisco WPA IE.
             Added checks for interface down in various functions.
             Changed signal quality calculations.
             Changed max USB packet size to follow endpoint settings.
             Added check to ignore AP beacons when in ad-hoc mode.]
2007-05-20: Fix hmac5 MIC length in wpa.c (patch by Adam Gashlin).
2007-05-31: Accept 64 hex character WPA encryption key.

Ben.

-- 
Ben Hutchings
The obvious mathematical breakthrough [to break modern encryption] would be
development of an easy way to factor large prime numbers. - Bill Gates
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.alioth.debian.org/pipermail/pkg-ralink-maintainers/attachments/20070603/94b6c019/attachment.pgp 


More information about the Pkg-ralink-maintainers mailing list