orinoco.c

pcmcia source code

/* orinoco.c 0.11b	- (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 loose 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 Mar 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 -> v0.11b - 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.
 *
 * TODO
 *	o New wireless extensions API
 *	o Handle de-encapsulation within network layer, provide 802.11
 *	  headers
 *	o Fix possible races in SPY handling.
 *	o Disconnect wireless extensions from fundamental configuration.
 *
 *	o Convert /proc debugging stuff to seqfile
 *	o Use multiple Tx buffers
 */
/* Notes on locking:
 *
 * The basic principle of operation is that everything except the
 * interrupt handler is serialized through a single spinlock in the
 * struct orinoco_private structure, using orinoco_lock() and
 * orinoco_unlock() (which in turn use spin_lock_bh() and
 * spin_unlock_bh()).
 *
 * The kernel's IRQ handling stuff ensures that the interrupt handler
 * does not re-enter itself. The interrupt handler is written such
 * that everything it does is safe without a lock: chiefly this means
 * that the Rx path uses one of the Hermes chipset's BAPs while
 * everything else uses the other.
 *
 * Actually, strictly speaking, the updating of the statistics from
 * the interrupt handler isn't safe without a lock.  However the worst
 * that can happen is that we perturb the packet/byte counts slightly.
 * We could fix this to use atomic types, but it's probably not worth
 * it.
 *
 * The big exception is that that we don't want the irq handler
 * running when we actually reset or shut down the card, because
 * strange things might happen (probably the worst would be one packet
 * of garbage, but you can't be too careful). For this we use
 * __orinoco_stop_irqs() which will set a flag to disable the interrupt
 * handler, and wait for any outstanding instances of the handler to
 * complete. THIS WILL LOSE INTERRUPTS! so it shouldn't be used except
 * for resets, where losing a few interrupts is acceptable. */

#include <linux/config.h>

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/ptrace.h>
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/timer.h>
#include <linux/ioport.h>
#include <asm/uaccess.h>
#include <asm/io.h>
#include <asm/system.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/netdevice.h>
#include <linux/if_arp.h>
#include <linux/etherdevice.h>
#include <linux/wireless.h>
#include <linux/list.h>

#include "hermes.h"
#include "hermes_rid.h"
#include "orinoco.h"
#include "ieee802_11.h"

/* Wireless extensions backwards compatibility */
#ifndef SIOCIWFIRSTPRIV
#define SIOCIWFIRSTPRIV		SIOCDEVPRIVATE
#endif /* SIOCIWFIRSTPRIV */

/* We do this this way to avoid ifdefs in the actual code */
#ifdef WIRELESS_SPY
#define SPY_NUMBER(priv)	(priv->spy_number)
#else
#define SPY_NUMBER(priv)	0
#endif /* WIRELESS_SPY */

static char version[] __initdata = "orinoco.c 0.11b (David Gibson <[e-mail]> and others)";
MODULE_AUTHOR("David Gibson <[e-mail]>");
MODULE_DESCRIPTION("Driver for Lucent Orinoco, Prism II based and similar wireless cards");
#ifdef MODULE_LICENSE
MODULE_LICENSE("Dual MPL/GPL");
#endif

/* Level of debugging. Used in the macros in orinoco.h */
#ifdef ORINOCO_DEBUG
int orinoco_debug = ORINOCO_DEBUG;
MODULE_PARM(orinoco_debug, "i");
EXPORT_SYMBOL(orinoco_debug);
#endif

#define ORINOCO_MIN_MTU		256
#define ORINOCO_MAX_MTU		(IEEE802_11_DATA_LEN - ENCAPS_OVERHEAD)

#define SYMBOL_MAX_VER_LEN	(14)
#define USER_BAP		0
#define IRQ_BAP			1
#define MAX_IRQLOOPS_PER_IRQ	10
#define MAX_IRQLOOPS_PER_JIFFY	(20000/HZ)	/* Based on a guestimate of how many events the
						   device can legitimately generate */
#define SMALL_KEY_SIZE		5
#define LARGE_KEY_SIZE		13
#define TX_NICBUF_SIZE_BUG	1585		/* Bug in Symbol firmware */

#define DUMMY_FID		0xFFFF

/*#define MAX_MULTICAST(priv)	(priv->firmware_type == FIRMWARE_TYPE_AGERE ? \
  HERMES_MAX_MULTICAST : 0)*/
#define MAX_MULTICAST(priv)	(HERMES_MAX_MULTICAST)

/********************************************************************/
/* Data tables                                                      */
/********************************************************************/

/* The frequency of each channel in MHz */
const long channel_frequency[] = {
	2412, 2417, 2422, 2427, 2432, 2437, 2442,
	2447, 2452, 2457, 2462, 2467, 2472, 2484
};
#define NUM_CHANNELS ( sizeof(channel_frequency) / sizeof(channel_frequency[0]) )

/* This tables gives the actual meanings of the bitrate IDs returned by the firmware. */
struct {
	int bitrate; /* in 100s of kilbits */
	int automatic;
	u16 agere_txratectrl;
	u16 intersil_txratectrl;
} bitrate_table[] = {
	{110, 1,  3, 15}, /* Entry 0 is the default */
	{10,  0,  1,  1},
	{10,  1,  1,  1},
	{20,  0,  2,  2},
	{20,  1,  6,  3},
	{55, 0,  4,  4},
	{55, 1,  7,  7},
	{110, 0,  5,  8},
};
#define BITRATE_TABLE_SIZE (sizeof(bitrate_table) / sizeof(bitrate_table[0]))

struct header_struct {
	/* 802.3 */