orinoco.c

优龙2410linux2.6.8内核源代码

/* orinoco.c 0.13e	- (formerly known as dldwd_cs.c and orinoco_cs.c)
 *
 * A driver for Hermes or Prism 2 chipset based PCMCIA wireless
 * adaptors, with Lucent/Agere, Intersil or Symbol firmware.
 *
 * Copyright (C) 2000 David Gibson, Linuxcare Australia <[e-mail]>
 *	With some help from :
 * Copyright (C) 2001 Jean Tourrilhes, HP Labs <[e-mail]>
 * Copyright (C) 2001 Benjamin Herrenschmidt <[e-mail]>
 *
 * Based on dummy_cs.c 1.27 2000/06/12 21:27:25
 *
 * Portions based on wvlan_cs.c 1.0.6, Copyright Andreas Neuhaus <[e-mail]>
 *      http://www.fasta.fh-dortmund.de/users/andy/wvlan/
 *
 * 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.
 *
 * The initial developer of the original code is David A. Hinds
 * <[e-mail]>.  Portions created by David
 * A. Hinds are Copyright (C) 1999 David A. Hinds.  All Rights
 * Reserved.
 *
 * Alternatively, the contents of this file may be used under the
 * terms of the GNU General 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.  */

/*
 * v0.01 -> v0.02 - 21/3/2001 - Jean II
 *	o Allow to use regular ethX device name instead of dldwdX
 *	o Warning on IBSS with ESSID=any for firmware 6.06
 *	o Put proper range.throughput values (optimistic)
 *	o IWSPY support (IOCTL and stat gather in Rx path)
 *	o Allow setting frequency in Ad-Hoc mode
 *	o Disable WEP setting if !has_wep to work on old firmware
 *	o Fix txpower range
 *	o Start adding support for Samsung/Compaq firmware
 *
 * v0.02 -> v0.03 - 23/3/2001 - Jean II
 *	o Start adding Symbol support - need to check all that
 *	o Fix Prism2/Symbol WEP to accept 128 bits keys
 *	o Add Symbol WEP (add authentication type)
 *	o Add Prism2/Symbol rate
 *	o Add PM timeout (holdover duration)
 *	o Enable "iwconfig eth0 key off" and friends (toggle flags)
 *	o Enable "iwconfig eth0 power unicast/all" (toggle flags)
 *	o Try with an intel card. It report firmware 1.01, behave like
 *	  an antiquated firmware, however on windows it says 2.00. Yuck !
 *	o Workaround firmware bug in allocate buffer (Intel 1.01)
 *	o Finish external renaming to orinoco...
 *	o Testing with various Wavelan firmwares
 *
 * v0.03 -> v0.04 - 30/3/2001 - Jean II
 *	o Update to Wireless 11 -> add retry limit/lifetime support
 *	o Tested with a D-Link DWL 650 card, fill in firmware support
 *	o Warning on Vcc mismatch (D-Link 3.3v card in Lucent 5v only slot)
 *	o Fixed the Prims2 WEP bugs that I introduced in v0.03 :-(
 *	  It work on D-Link *only* after a tcpdump. Weird...
 *	  And still doesn't work on Intel card. Grrrr...
 *	o Update the mode after a setport3
 *	o Add preamble setting for Symbol cards (not yet enabled)
 *	o Don't complain as much about Symbol cards...
 *
 * v0.04 -> v0.04b - 22/4/2001 - David Gibson
 *      o Removed the 'eth' parameter - always use ethXX as the
 *        interface name instead of dldwdXX.  The other was racy
 *        anyway.
 *	o Clean up RID definitions in hermes.h, other cleanups
 *
 * v0.04b -> v0.04c - 24/4/2001 - Jean II
 *	o Tim Hurley <[e-mail]> reported a D-Link card
 *	  with vendor 02 and firmware 0.08. Added in the capabilities...
 *	o Tested Lucent firmware 7.28, everything works...
 *
 * v0.04c -> v0.05 - 3/5/2001 - Benjamin Herrenschmidt
 *	o Spin-off Pcmcia code. This file is renamed orinoco.c,
 *	  and orinoco_cs.c now contains only the Pcmcia specific stuff
 *	o Add Airport driver support on top of orinoco.c (see airport.c)
 *
 * v0.05 -> v0.05a - 4/5/2001 - Jean II
 *	o Revert to old Pcmcia code to fix breakage of Ben's changes...
 *
 * v0.05a -> v0.05b - 4/5/2001 - Jean II
 *	o add module parameter 'ignore_cis_vcc' for D-Link @ 5V
 *	o D-Link firmware doesn't support multicast. We just print a few
 *	  error messages, but otherwise everything works...
 *	o For David : set/getport3 works fine, just upgrade iwpriv...
 *
 * v0.05b -> v0.05c - 5/5/2001 - Benjamin Herrenschmidt
 *	o Adapt airport.c to latest changes in orinoco.c
 *	o Remove deferred power enabling code
 *
 * v0.05c -> v0.05d - 5/5/2001 - Jean II
 *	o Workaround to SNAP decapsulate frame from LinkSys AP
 *	  original patch from : Dong Liu <[e-mail]>
 *	  (note : the memcmp bug was mine - fixed)
 *	o Remove set_retry stuff, no firmware support it (bloat--).
 *
 * v0.05d -> v0.06 - 25/5/2001 - Jean II
 *		Original patch from "Hong Lin" <[e-mail]>,
 *		"Ian Kinner" <[e-mail]>
 *		and "David Smith" <[e-mail]>
 *	o Init of priv->tx_rate_ctrl in firmware specific section.
 *	o Prism2/Symbol rate, upto should be 0xF and not 0x15. Doh !
 *	o Spectrum card always need cor_reset (for every reset)
 *	o Fix cor_reset to not lose bit 7 in the register
 *	o flush_stale_links to remove zombie Pcmcia instances
 *	o Ack previous hermes event before reset
 *		Me (with my little hands)
 *	o Allow orinoco.c to call cor_reset via priv->card_reset_handler
 *	o Add priv->need_card_reset to toggle this feature
 *	o Fix various buglets when setting WEP in Symbol firmware
 *	  Now, encryption is fully functional on Symbol cards. Youpi !
 *
 * v0.06 -> v0.06b - 25/5/2001 - Jean II
 *	o IBSS on Symbol use port_mode = 4. Please don't ask...
 *
 * v0.06b -> v0.06c - 29/5/2001 - Jean II
 *	o Show first spy address in /proc/net/wireless for IBSS mode as well
 *
 * v0.06c -> v0.06d - 6/7/2001 - David Gibson
 *      o Change a bunch of KERN_INFO messages to KERN_DEBUG, as per Linus'
 *        wishes to reduce the number of unecessary messages.
 *	o Removed bogus message on CRC error.
 *	o Merged fixeds for v0.08 Prism 2 firmware from William Waghorn
 *	  <[e-mail]>
 *	o Slight cleanup/re-arrangement of firmware detection code.
 *
 * v0.06d -> v0.06e - 1/8/2001 - David Gibson
 *	o Removed some redundant global initializers (orinoco_cs.c).
 *	o Added some module metadataa
 *
 * v0.06e -> v0.06f - 14/8/2001 - David Gibson
 *	o Wording fix to license
 *	o Added a 'use_alternate_encaps' module parameter for APs which need an
 *	  oui of 00:00:00.  We really need a better way of handling this, but
 *	  the module flag is better than nothing for now.
 *
 * v0.06f -> v0.07 - 20/8/2001 - David Gibson
 *	o Removed BAP error retries from hermes_bap_seek().  For Tx we now
 *	  let the upper layers handle the retry, we retry explicitly in the
 *	  Rx path, but don't make as much noise about it.
 *	o Firmware detection cleanups.
 *
 * v0.07 -> v0.07a - 1/10/3001 - Jean II
 *	o Add code to read Symbol firmware revision, inspired by latest code
 *	  in Spectrum24 by Lee John Keyser-Allen - Thanks Lee !
 *	o Thanks to Jared Valentine <[e-mail]> for "providing" me
 *	  a 3Com card with a recent firmware, fill out Symbol firmware
 *	  capabilities of latest rev (2.20), as well as older Symbol cards.
 *	o Disable Power Management in newer Symbol firmware, the API 
 *	  has changed (documentation needed).
 *
 * v0.07a -> v0.08 - 3/10/2001 - David Gibson
 *	o Fixed a possible buffer overrun found by the Stanford checker (in
 *	  dldwd_ioctl_setiwencode()).  Can only be called by root anyway, so not
 *	  a big problem.
 *	o Turned has_big_wep on for Intersil cards.  That's not true for all of
 *	  them but we should at least let the capable ones try.
 *	o Wait for BUSY to clear at the beginning of hermes_bap_seek().  I
 *	  realised that my assumption that the driver's serialization
 *	  would prevent the BAP being busy on entry was possibly false, because
 *	  things other than seeks may make the BAP busy.
 *	o Use "alternate" (oui 00:00:00) encapsulation by default.
 *	  Setting use_old_encaps will mimic the old behaviour, but I think we
 *	  will be able to eliminate this.
 *	o Don't try to make __initdata const (the version string).  This can't
 *	  work because of the way the __initdata sectioning works.
 *	o Added MODULE_LICENSE tags.
 *	o Support for PLX (transparent PCMCIA->PCI brdge) cards.
 *	o Changed to using the new type-facist min/max.
 *
 * v0.08 -> v0.08a - 9/10/2001 - David Gibson
 *	o Inserted some missing acknowledgements/info into the Changelog.
 *	o Fixed some bugs in the normalisation of signel level reporting.
 *	o Fixed bad bug in WEP key handling on Intersil and Symbol firmware,
 *	  which led to an instant crash on big-endian machines.
 *
 * v0.08a -> v0.08b - 20/11/2001 - David Gibson
 *	o Lots of cleanup and bugfixes in orinoco_plx.c
 *	o Cleanup to handling of Tx rate setting.
 *	o Removed support for old encapsulation method.
 *	o Removed old "dldwd" names.
 *	o Split RID constants into a new file hermes_rid.h
 *	o Renamed RID constants to match linux-wlan-ng and prism2.o
 *	o Bugfixes in hermes.c
 *	o Poke the PLX's INTCSR register, so it actually starts
 *	  generating interrupts.  These cards might actually work now.
 *	o Update to wireless extensions v12 (Jean II)
 *	o Support for tallies and inquire command (Jean II)
 *	o Airport updates for newer PPC kernels (BenH)
 *
 * v0.08b -> v0.09 - 21/12/2001 - David Gibson
 *	o Some new PCI IDs for PLX cards.
 *	o Removed broken attempt to do ALLMULTI reception.  Just use
 *	  promiscuous mode instead
 *	o Preliminary work for list-AP (Jean II)
 *	o Airport updates from (BenH)
 *	o Eliminated racy hw_ready stuff
 *	o Fixed generation of fake events in irq handler.  This should
 *	  finally kill the EIO problems (Jean II & dgibson)
 *	o Fixed breakage of bitrate set/get on Agere firmware (Jean II)
 *
 * v0.09 -> v0.09a - 2/1/2002 - David Gibson
 *	o Fixed stupid mistake in multicast list handling, triggering
 *	  a BUG()
 *
 * v0.09a -> v0.09b - 16/1/2002 - David Gibson
 *	o Fixed even stupider mistake in new interrupt handling, which
 *	  seriously broke things on big-endian machines.
 *	o Removed a bunch of redundant includes and exports.
 *	o Removed a redundant MOD_{INC,DEC}_USE_COUNT pair in airport.c
 *	o Don't attempt to do hardware level multicast reception on
 *	  Intersil firmware, just go promisc instead.
 *	o Typo fixed in hermes_issue_cmd()
 *	o Eliminated WIRELESS_SPY #ifdefs
 *	o Status code reported on Tx exceptions
 *	o Moved netif_wake_queue() from ALLOC interrupts to TX and TXEXC
 *	  interrupts, which should fix the timeouts we're seeing.
 *
 * v0.09b -> v0.10 - 25 Feb 2002 - David Gibson
 *	o Removed nested structures used for header parsing, so the
 *	  driver should now work without hackery on ARM
 *	o Fix for WEP handling on Intersil (Hawk Newton)
 *	o Eliminated the /proc/hermes/ethXX/regs debugging file.  It
 *	  was never very useful.
 *	o Make Rx errors less noisy.
 *
 * v0.10 -> v0.11 - 5 Apr 2002 - David Gibson
 *	o Laid the groundwork in hermes.[ch] for devices which map
 *	  into PCI memory space rather than IO space.
 *	o Fixed bug in multicast handling (cleared multicast list when
 *	  leaving promiscuous mode).
 *	o Relegated Tx error messages to debug.
 *	o Cleaned up / corrected handling of allocation lengths.
 *	o Set OWNSSID in IBSS mode for WinXP interoperability (jimc).
 *	o Change to using alloc_etherdev() for structure allocations. 
 *	o Check for and drop undersized packets.
 *	o Fixed a race in stopping/waking the queue.  This should fix
 *	  the timeout problems (Pavel Roskin)
 *	o Reverted to netif_wake_queue() on the ALLOC event.
 *	o Fixes for recent Symbol firmwares which lack AP density
 *	  (Pavel Roskin).
 *
 * v0.11 -> v0.11a - 29 Apr 2002 - David Gibson
 *	o Handle different register spacing, necessary for Prism 2.5
 *	  PCI adaptors (Steve Hill).
 *	o Cleaned up initialization of card structures in orinoco_cs
 *	  and airport.  Removed card->priv field.
 *	o Make response structure optional for hermes_docmd_wait()
 *	  Pavel Roskin)
 *	o Added PCI id for Nortel emobility to orinoco_plx.c.
 *	o Cleanup to handling of Symbol's allocation bug. (Pavel Roskin)
 *	o Cleanups to firmware capability detection.
 *	o Arrange for orinoco_pci.c to override firmware detection.
 *	  We should be able to support the PCI Intersil cards now.
 *	o Cleanup handling of reset_cor and hard_reset (Pavel Roskin).
 *	o Remove erroneous use of USER_BAP in the TxExc handler (Jouni
 *	  Malinen).
 *	o Makefile changes for better integration into David Hinds
 *	  pcmcia-cs package.
 *
 * v0.11a -> v0.11b - 1 May 2002 - David Gibson
 *	o Better error reporting in orinoco_plx_init_one()
 *	o Fixed multiple bad kfree() bugs introduced by the
 *	  alloc_orinocodev() changes.
 *
 * v0.11b -> v0.12 - 19 Jun 2002 - David Gibson
 *	o Support changing the MAC address.
 *	o Correct display of Intersil firmware revision numbers.
 *	o Entirely revised locking scheme.  Should be both simpler and
 *	   better.
 *	o Merged some common code in orinoco_plx, orinoco_pci and
 *	  airport by creating orinoco_default_{open,stop,reset}()
 *	  which are used as the dev->open, dev->stop, priv->reset
 *	  callbacks if none are specified when alloc_orinocodev() is
 *	  called.
 *	o Removed orinoco_plx_interrupt() and orinoco_pci_interrupt().
 *	  They didn't do anything.
 *
 * v0.12 -> v0.12a - 4 Jul 2002 - David Gibson
 *	o Some rearrangement of code.
 *	o Numerous fixups to locking and rest handling, particularly
 *	  for PCMCIA.
 *	o This allows open and stop net_device methods to be in
 *	  orinoco.c now, rather than in the init modules.
 *	o In orinoco_cs.c link->priv now points to the struct
 *	  net_device not to the struct orinoco_private.
 *	o Added a check for undersized SNAP frames, which could cause
 *	  crashes.
 *
 * v0.12a -> v0.12b - 11 Jul 2002 - David Gibson
 *	o Fix hw->num_init testing code, so num_init is actually
 *	  incremented.
 *	o Fix very stupid bug in orinoco_cs which broke compile with
 *	  CONFIG_SMP.
 *	o Squashed a warning.
 *
 * v0.12b -> v0.12c - 26 Jul 2002 - David Gibson
 *	o Change to C9X style designated initializers.
 *	o Add support for 3Com AirConnect PCI.
 *	o No longer ignore the hard_reset argument to
 *	  alloc_orinocodev().  Oops.
 *
 * v0.12c -> v0.13beta1 - 13 Sep 2002 - David Gibson
 *	o Revert the broken 0.12* locking scheme and go to a new yet
 *	  simpler scheme.
 *	o Do firmware resets only in orinoco_init() and when waking
 *	  the card from hard sleep.
 *
 * v0.13beta1 -> v0.13 - 27 Sep 2002 - David Gibson
 *	o Re-introduced full resets (via schedule_task()) on Tx
 *	  timeout.
 *
 * v0.13 -> v0.13a - 30 Sep 2002 - David Gibson
 *	o Minor cleanups to info frame handling.  Add basic support
 *	  for linkstatus info frames.
 *	o Include required kernel headers in orinoco.h, to avoid
 *	  compile problems.
 *
 * v0.13a -> v0.13b - 10 Feb 2003 - David Gibson
 *	o Implemented hard reset for Airport cards
 *	o Experimental suspend/resume implementation for orinoco_pci
 *	o Abolished /proc debugging support, replaced with a debugging
 *	  iwpriv.  Now it's ugly and simple instead of ugly and complex.
 *	o Bugfix in hermes.c if the firmware returned a record length
 *	  of 0, we could go clobbering memory.
 *	o Bugfix in orinoco_stop() - it used to fail if hw_unavailable
 *	  was set, which was usually true on PCMCIA hot removes.
 * 	o Track LINKSTATUS messages, silently drop Tx packets before
 * 	  we are connected (avoids cofusing the firmware), and only
 * 	  give LINKSTATUS printk()s if the status has changed.
 *
 * v0.13b -> v0.13c - 11 Mar 2003 - David Gibson
 *	o Cleanup: use dev instead of priv in various places.
 *	o Bug fix: Don't ReleaseConfiguration on RESET_PHYSICAL event
 *	  if we're in the middle of a (driver initiated) hard reset.
 *	o Bug fix: ETH_ZLEN is supposed to include the header
 *	  (Dionysus Blazakis & Manish Karir)
 *	o Convert to using workqueues instead of taskqueues (and
 *	  backwards compatibility macros for pre 2.5.41 kernels).
 *	o Drop redundant (I think...) MOD_{INC,DEC}_USE_COUNT in
 *	  airport.c
 *	o New orinoco_tmd.c init module from Joerg Dorchain for
 *	  TMD7160 based PCI to PCMCIA bridges (similar to
 *	  orinoco_plx.c).
 *
 * v0.13c -> v0.13d - 22 Apr 2003 - David Gibson
 *	o Make hw_unavailable a counter, rather than just a flag, this
 *	  is necessary to avoid some races (such as a card being
 *	  removed in the middle of orinoco_reset().
 *	o Restore Release/RequestConfiguration in the PCMCIA event handler
 *	  when dealing with a driver initiated hard reset.  This is
 *	  necessary to prevent hangs due to a spurious interrupt while
 *	  the reset is in progress.
 *	o Clear the 802.11 header when transmitting, even though we
 *	  don't use it.  This fixes a long standing bug on some
 *	  firmwares, which seem to get confused if that isn't done.
 *	o Be less eager to de-encapsulate SNAP frames, only do so if
 *	  the OUI is 00:00:00 or 00:00:f8, leave others alone.  The old
 *	  behaviour broke CDP (Cisco Discovery Protocol).
 *	o Use dev instead of priv for free_irq() as well as
 *	  request_irq() (oops).
 *	o Attempt to reset rather than giving up if we get too many
 *	  IRQs.
 *	o Changed semantics of __orinoco_down() so it can be called
 *	  safely with hw_unavailable set.  It also now clears the
 *	  linkstatus (since we're going to have to reassociate).
 *
 * v0.13d -> v0.13e - 12 May 2003 - David Gibson
 *	o Support for post-2.5.68 return values from irq handler.
 *	o Fixed bug where underlength packets would be double counted
 *	  in the rx_dropped statistics.
 *	o Provided a module parameter to suppress linkstatus messages.
 *
 * TODO
 *	o New wireless extensions API (patch from Moustafa
 *	  Youssef, updated by Jim Carter and Pavel Roskin).
 *	o Handle de-encapsulation within network layer, provide 802.11
 *	  headers (patch from Thomas 'Dent' Mirlacher)
 *	o RF monitor mode support
 *	o Fix possible races in SPY handling.
 *	o Disconnect wireless extensions from fundamental configuration.
 *	o (maybe) Software WEP support (patch from Stano Meduna).
 *	o (maybe) Use multiple Tx buffers - driver handling queue
 *	  rather than firmware. */

/* Locking and synchronization:
 *
 * The basic principle is that everything is serialized through a
 * single spinlock, priv->lock.  The lock is used in user, bh and irq
 * context, so when taken outside hardirq context it should always be
 * taken with interrupts disabled.  The lock protects both the
 * hardware and the struct orinoco_private.
 *
 * Another flag, priv->hw_unavailable indicates that the hardware is
 * unavailable for an extended period of time (e.g. suspended, or in
 * the middle of a hard reset).  This flag is protected by the
 * spinlock.  All code which touches the hardware should check the
 * flag after taking the lock, and if it is set, give up on whatever