p54: fix queue stall due to underrun

Larry Finger discovered a weird behavior under load.
In essence, the queue's length count under runs,
which in turn renders the associated ac queue unusable.

Reported-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54usb: fix stalls caused by urb allocation failures

This patch squashes a few old bugs, which have been around since
the initial version of p54usb in one form or another.

we never freed a orphaned frame, when were denied the resources,
which are necessary to pass the data into the usb subsystem.
As a result we could end up with a full queue that wasn't emptied,
until the device was brought down.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rfkill: prep for rfkill API changes

We've designed the /dev/rfkill API in a way that we
can increase the event struct by adding members at
the end, should it become necessary. To validate the
events, userspace and the kernel need to have the
proper event size to check for -- when reading from
the other end they need to verify that it's at least
version 1 of the event API, with the current struct
size, so define a constant for that and make the
code a little more 'future proof'.

Not that I expect that we'll have to change the event
size any time soon, but it's better to write the code
in a way that lends itself to extending.

Due to the current size of the event struct, the code
is currently equivalent, but should the event struct
ever need to be increased the new code might not need
changing.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: write PCU registers on initial reset

"Ath5k: unify resets"
introduced a regression into 2.6.28 where the PCU registers are never
initialized, due to ath5k_reset() always passing true for change_channel.
We subsequently program a lot of these registers but several may start
in an unknown state.

Cc: stable@kernel.org
Reported-by: Forrest Zhang <forrest@hifulltech.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: do not release irq across suspend/resume

Paraphrasing Rafael J. Wysocki: "drivers should not release PCI IRQs
in suspend."  Doing so causes a warning during suspend/resume on some
platforms.

Cc: Rafael J. Wysocki <rjw@sisk.pl>
Reported-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath: remove unnecessary return in ath_regd_get_band_ctl

'default' case already returns NO_CTL

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: rework beacon configuration

Using the enable_beacon flag allows some simplifications and fixes
some corner cases in beacon handling.  This change adds a state
variable for beaconing in ath5k_beacon_config and handles both
enabling and disabling, thus eliminating the need for
ath5k_beacon_disable.  We also now configure the beacon when any
of the beacon parameters change, so ath5k_beacon_reconfig is no
longer needed (its mmiowb gets moved to ath5k_beacon_config).
Finally, by locking around the whole config function, we don't
need to worry about clearing the interrupt mask register before
installing the new mask.

The upshot is this correctly disables beaconing when the interfaces
are taken down, it fixes a potential restarting of beaconing
when ath5k_reset() is called, and ensures that updates to the
beacon interval take effect immediately.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: send buffered frames after the beacon

Enable the "Content" After Beacon queue and utilize it to send
any buffered frames for power-saving clients.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54spi: remove dead code and definitions

This patch removes some dead code:
p54spi.c:115: warning: ‘p54spi_read16’ defined but not used

and while we're at it, p54spi_registers_array is purged as well.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: two endian fixes

This patch fixes all CHECK_ENDIAN complains:

1. p54/fwio.c:296:6: warning: restricted __le32 degrades to integer
    p54/fwio.c:296:6: warning: restricted __le32 degrades to integer

2. p54/p54spi.c:172:32: warning: incorrect type in initializer
   p54spi.c:172:32:    expected restricted __le32 [usertype] buffer
   p54/p54spi.c:172:32:    got unsigned int

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: check for current_bss from giwrate

When connecting to an ESSID manually, we may not set the BSSID, and thus
wdev->wext.connect.bssid will be NULL.
wdev->current_bss is always updated when a connection is established so we
should check it first.

Signed-off-by: Samuel Ortiz <samuel.ortiz@intel.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: use wiphy rfkill interface

Remove the input_polldev from rt2x00 and replace it with
the rfkill interface offered by the wiphy structure. This
simplifies the entire rfkill handling in rt2x00 and allows
us to remove the CONFIG_RT2X00_LIB_RFKILL option and always
enables rfkill capabilities.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: fix aggregation limit

According to the documentation, the limit is 0x3f == 63, not 64.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: shorten the passive dwell time for sw scans

mac80211's software scan implementation uses a passive dwell time of
(HZ / 5) which means we stay 200ms on each passive channel. Compared
to iwlwifi's hw scan and the old ipw* drivers which use values around
120ms this is quite long.

Reducing the passive dwell time from 200ms to 125ms should save us
something around a second on cards capable of 11a and we should still be
able to catch beacons from most access points (assuming a ~100ms beacon
interval).

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove permissions from debugfs files

Don't allow users to open debugfs files, because it can cause oopses.

When a user opens some file, driver unlinks it and frees the
corresponding structure, we will dereference freed memory.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: remove permissions from debugfs files

Don't allow users to open debugfs files, because it can cause oopses.

When a user opens some file, driver unlinks it and frees the
corresponding structure, we will dereference freed memory.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

drivers/net/wireless/ath/ath9k: Remove unnecessary semicolons

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: always print buffer when error condition occurs

We want to see the buffer contents when the error occurs without
needing to set any debug flags.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: add utility to print buffer when error occurs

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: re-enable PS support for iwlagn

The register locking rework addressed the problem where nic
access was obtained incorrectly when PS is enabled.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwmc3200wifi: remove B0 hardware support

The patch removes B0 hardware support. Nobody is using it anyway.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwmc3200wifi: simplify calibration map

The patch simplifies calibration map by combining the init_calib_map
and periodic_calib_map into one calib_map in struct iwm_conf. Now the
initial calibration map is stored in the lower 16 bits of calib_map
and the periodic calibration map is stored in the higher 16 bits.

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwmc3200wifi: replace netif_rx with netif_rx_ni

The patch uses netif_rx_ni() over netif_rx() to post buffers to
upper network code because it is always scheduled in a workqueue.
The problem was first observed from a dynamic ticks warning:

    "NOHZ: local_softirq_pending ..."

Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

drivers/net/wireless: Use PCI_VDEVICE

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Handle different TX and RX streams properly.

This patch fixes an issue when the TX and RX streams are different.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: race condition in SCANNING state check during ANI calibration

ANI calibration shouldn't be done when we are not on our home channel.
This is already verified. However, it is racy. Fix this by proper
spin locks.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: stop ani when the STA gets disconnected.

ANI is not required when the STA is disconnected. So stop it and enable
ANI for adhoc and monitor mode.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove unnecessary STATION mode check.

Remove unncessary STATION mode check in ath9k_bss_assoc_info() as
it is called only for STATION mode.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Modify p54 files for new organization

Modify the remaining p54 files to account for the new file organization.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move TX/RX code

Copy the TX/RX code from p54common.c into a new file txrx.c

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move mac80211 glue code

Copy the mac80211 glue code from p54common.c into a new file main.c

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move LMAC interface definitions

Copy the LMAC Interface specific definitions from p54common.h into a new file lmac.h

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move LED code

Copy the LED code from p54common.c into a new file led.c

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move firmware code

Copy the firmware i/o code from p54common.c into a new file fwio.c

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move eeprom header

Copy the eeprom code from p54common.h into a new file eeprom.h

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: Move eeprom code

Copy the eeprom code from p54common.c into a new file eeprom.c

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: remove redundant tests on unsigned

bufsize and remainder are unsigned. When negative they are wrapped and caught by
the other test.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: redo rx_status into skb->cb

This patch slightly optimizes p54_rx_data's stack and code size.

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ar9170usb: module link in sysfs

Andrey Yurovsky reported that the driver forwarded erroneously the
parent device structure instead of the real thing, which of course
led to some dodgy sysfs links (at least?).

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: unify iwl_setup_rxon_timing

This patch unifies setup_rxon_timing funcions
of AGN and 3945. HWs differ only in supported maximal
beacon interval. This is reflected in hw_paras.max_beacon_itrvl

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: scan requested channels only

When userspace requests only certain channels to be scanned,
we currently ignore that request entirely.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: make software queue assignment more efficient

There really is no reason to be assigning txq->swq_id all the
time, once at aggregation setup is sufficient.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: drop sw_crypto from hw_params.

Each HW supported by iwlwifi is capable of hardware crypto
so drop this flag from hw_params structure.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: Check HW ready before prepare card.

Hardware may be ready for us to manage it without us trying to prepare
it first. Check if this is the case.

Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: differentiate quality reporting between legacy and HT configurations

We were not differentiating quality between legacy and HT
configurations. We change this to consider the differences.

New theory for reporting quality:

At a hardware RSSI of 45 you will be able to use MCS 7  reliably.
At a hardware RSSI of 45 you will be able to use MCS 15 reliably.
At a hardware RSSI of 35 you should be able use 54 Mbps reliably.

MCS 7  is the highets MCS index usable by a 1-stream device.
MCS 15 is the highest MCS index usable by a 2-stream device.

All ath9k devices are either 1-stream or 2-stream.

How many bars you see is derived from the qual reporting.

A more elaborate scheme can be used here but it requires tables
of SNR/throughput for each possible mode used. For the MCS table
you can refer to the wireless wiki:

http://wireless.kernel.org/en/developers/Documentation/ieee80211/802.11n

This should fix this bug report:

http://bugzilla.kernel.org/show_bug.cgi?id=13537

Cc: Janath.Peiris@atheros.com
Cc: Matt.Smith@atheros.com
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: fix card cleanup order in SDIO driver

The SDIO driver sets the surpriseremoved flag before calling
lbs_remove_card.  With IEEE PS enabled, lbs_remove_card must issue a
command to exit IEEE PS mode, however with that flag set the command
path is blocked and the card is never taken out of IEEE PS mode.  This
step is required to ensure that the driver can be reloaded.  This patch
moves the setting of surpriseremoved after lbs_remove_card is called.

Tested with V9 firmware by ensuring that IEEE PS is disabled when the
driver is removed.  Reloading the driver is not fully tested due to a
separate issue with module reload in the SDIO driver, however this
patch at least leaves the card in a better state when we bring the
driver down.

Signed-off-by: Andrey Yurovsky <andrey@cozybit.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: correct card cleanup order in SPI driver

The SPI driver does a couple of card cleanup steps in the wrong order on
module removal.  If IEEE PS is enabled, this results in the card being
left in IEEE PS mode and subsequent failures to reload the module.  The
problem is that the surpriseremoved flag is set before calling
lbs_remove_card, but that function needs to issue a command to exit IEEE
PS mode (the flag blocks the command path).  In addition, lbs_stop_card
should be called first because it clears out any pending commands.

Tested on a GSPI device with V9 firmware by confirming that we can
reload the module with or without IEEE PS enabled.

Also fix a warning from the wrong uint format in a printk.

V2: use z modifier, thanks Sebastian.

Signed-off-by: Andrey Yurovsky <andrey@cozybit.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: copy WPA keys to priv when associating

Libertas currently maintains a copy of the WPA unicast and group keys
when using WPA or WPA2.  This copy is checked when deciding whether or
not to return to sleep in IEEE PS mode but the actual copying back to
priv was omitted, which breaks IEEE PS mode with WPA/WPA2 when one
issues commands that require temporarily keeping the device awake.

This patch introduces the omitted copy-back of the keys so that IEEE PS
functions correctly in WPA/WPA2 mode.  Thanks to Dan Williams for
clearing up the issue.

V2: fix typo.  Also, this has been tested on GSPI and SDIO with V9 firmware.

Signed-off-by: Andrey Yurovsky <andrey@cozybit.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: fix giwrange

"cfg80211: Advertise ciphers via WE according to driver capability"
unfortunately broke iwrange -- it used the variable c
that needs to be 0 for the channel list.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: limit to one pairwise cipher for associate()

In this case, only one cipher makes sense, unlike for
connect() where it may be possible to have the card or
driver select.

No changes to mac80211 due to the way the structs are
laid out -- but the loop in net/mac80211/cfg.c will
degrade to just zero or one passes.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: fix printk format specifier for size_t arguments

This addresses the following compile warnings on 64-bit platforms.

drivers/net/wireless/orinoco/scan.c: In function 'orinoco_add_hostscan_results':
drivers/net/wireless/orinoco/scan.c:194: warning: format '%d' expects type 'int', but argument 3 has type 'size_t'
drivers/net/wireless/orinoco/scan.c:211: warning: format '%d' expects type 'int', but argument 3 has type 'size_t'
drivers/net/wireless/orinoco/scan.c:211: warning: format '%d' expects type 'int', but argument 4 has type 'size_t'

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: refuse authenticating to same BSSID twice

It is possible that there are different BSS structs with
the same BSSID, but we cannot authenticate with multiple
of them them because we need the BSSID to be unique for
deauthenticating/disassociating.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: keep track of BSSes

In order to avoid problems with BSS structs going away
while they're in use, I've long wanted to make cfg80211
keep track of them. Without the SME, that wasn't doable
but now that we have the SME we can do this too. It can
keep track of up to four separate authentications and
one association, regardless of whether it's controlled
by the cfg80211 SME or the userspace SME.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: assimilate and export ieee80211_bss_get_ie

This function from mac80211 seems generally useful, and
I will need it in cfg80211 soon.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: reset auth algorithm

When the interface is brought down, we need to
reset the auth algorithm because wpa_supplicant
doesn't reset it, and then we fail to use shared
key auth when required later.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: send events for userspace SME

When the userspace SME is in control, we are currently not sending
events, but this means that any userspace applications using wext
or nl80211 to receive events will not know what's going on unless
they can also interpret the nl80211 assoc event. Since we have all
the required code, let the SME follow events from the userspace
SME, this even means that you will be refused to connect() while
the userspace SME is in control and connected.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: remove dead code, clean up

With mac80211 now always controlled by an external SME,
a lot of code is dead -- SSID, BSSID, channel selection
is always done externally, etc. Additionally, rename
IEEE80211_STA_TKIP_WEP_USED to IEEE80211_STA_DISABLE_11N
and clean up the code a bit.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: remove auth algorithm retry

The automatic auth algorithm issue is now solved in
cfg80211, so mac80211 no longer needs code to try
different algorithms -- just using whatever cfg80211
asked for is good.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: re-add HT disabling

The IEEE80211_STA_TKIP_WEP_USED flag is used internally to
disable HT when WEP or TKIP are used. Now that cfg80211 is
giving us the required information, we can set the flag
appropriately again.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: implement get_wireless_stats

By dropping the noise reporting, we can implement
wireless stats in cfg80211. We also make the
handler return NULL if we have no information,
which is possible thanks to the recent wext change.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: implement IWRATE

For now, let's implement that using a very hackish way:
simply mirror the wext API in the cfg80211 API. This
will have to be changed later when we implement proper
bitrate API.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: implement IWAP for WDS

This implements siocsiwap/giwap for WDS mode.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: implement iwpower

Just on/off and timeout, and with a hacky cfg80211 method
until we figure out what we want, though this is probably
sufficient as we want to use pm_qos for wifi everywhere.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: managed mode wext compatibility

This adds code to make it possible to use the cfg80211
connect() API with wireless extensions, and because the
previous patch added emulation of that API with auth()
and assoc(), by extension also supports wext on that.
At the same time, removes code from mac80211 for wext,
but doesn't yet clean up mac80211's mlme code more.

Signed-off-by: Samuel Ortiz <samuel.ortiz@intel.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: emulate connect with auth/assoc

This adds code to cfg80211 so that drivers (mac80211 right
now) that don't implement connect but rather auth/assoc can
still be used with the nl80211 connect command. This will
also be necessary for the wext compat code.

Signed-off-by: Samuel Ortiz <samuel.ortiz@intel.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: connect/disconnect API

This patch introduces the cfg80211 connect/disconnect API.
The goal here is to run the AUTH and ASSOC steps in one call.
This is needed for some fullmac cards that run both steps
directly from the target, after the host driver sends a
connect command.

Additionally, all the new crypto parameters for connect()
are now also valid for associate() -- although associate
requires the IEs to be used, the information can be useful
for drivers and should be given.

Signed-off-by: Samuel Ortiz <samuel.ortiz@intel.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: define AKM suites

We'll need these values for some drivers using connect API
and for wext compat code, so let's define them.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: remove an unused function declaration

The ieee80211_scan_results function hasn't existed for a
long time now, so its declaration should be removed as
well.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: introduce nl80211 testmode command

This introduces a new NL80211_CMD_TESTMODE for testing
and calibration use with nl80211. There's no multiplexing
like like iwpriv had, and the command is not available by
default, it needs to be explicitly enabled in Kconfig and
shouldn't be enabled in most kernels.

The command requires a wiphy index or interface index to
identify the device to operate on, and the new TESTDATA
attribute. There also is API for sending replies to the
command, and testmode multicast messages (on a testmode
multicast group).

I've also updated mac80211 to be able to pass through the
command to the driver, since it itself doesn't implement
the testmode command.

Additionally, to give people an idea of how to use the
command, I've added a little code to hwsim that makes use
of the new command to set the powersave mode, this is
currently done via debugfs and should remain there, and
the testmode command only serves as an example of how to
use this best -- with nested netlink attributes in the
TESTDATA attribute. A hwsim testmode tool can be found at
http://git.sipsolutions.net/hwsim.git/. This tool is BSD
licensed so people can easily use it as a basis for their
own internal fabrication and validation tools.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wext: constify extra argument to wireless_send_event

This is never changed by the function, so can be marked const.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: tell SME about real auth state

When the auth algorithm is rejected, but we don't have
another one to try, we will eventually retry but that
isn't useful -- we'll then do it again and again until
we eventually give up. Instead, we should let the SME
know and go into disabled state. The same applies for
situations where the AP rejects with any other status
code.

Additionally, when trying the next auth algorithm, we
should reset the auth_tries so that just a single lost
frame doesn't lead to us giving up on the third auth
algorithm.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: remove wireless_dev->bssid

This variable isn't necessary -- the wext code keeps
track of the BSSID itself, and otherwise we have
current_bss.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: use proper allocation flags

Instead of hardcoding GFP_ATOMIC everywhere, add a
new function parameter that gets the flags from the
caller. Obviously then I need to update all callers
(all of them in mac80211), and it turns out that now
it's ok to use GFP_KERNEL in almost all places.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: clean up function definitions

I don't like the 'extern' keyword much when it's not
necessary, it makes lines rather long.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: move break statement to correct place

Move a break statement to the correct place _after_ the
#endif, otherwise w/o WIRELESS_EXT things break badly.
Also, while touching this code, do a cleanup and assign
dev->ieee80211_ptr to a new variable.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wext: default to y

The way I initially thought we could do wireless extensions
is by making all the compat code in cfg80211 be independent
of CONFIG_WIRELESS_EXT, but this is turning out to not be
feasible. Therefore, fix the Kconfig help text and make the
option default to yes, so people won't get a nasty surprise
when mac80211 will get rid of its 'select WIRELESS_EXT' any
time now.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix todo lock

The key todo lock can be taken from different locks
that require it to be _bh to avoid lock inversion
due to (soft)irqs.

This should fix the two problems reported by Bob and
Gabor:
http://mid.gmane.org/20090619113049.GB18956@hash.localnet
http://mid.gmane.org/4A3FA376.8020307@openwrt.org

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Bob Copeland <me@bobcopeland.com>
Cc: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wext: allow returning NULL stats

Currently, wext drivers cannot return NULL for stats even though
that would make the ioctl return -EOPNOTSUPP because that would
mean they are no longer listed in /proc/net/wireless. This patch
changes the wext core's behaviour to list them if they have any
wireless_handlers, but only show their stats when available, so
that drivers can start returning NULL if stats are currently not
available, reducing confusion for e.g. IBSS.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211_hwsim: fix-up build damage from removal of skb->dst

Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove ath_rx_ps_back_to_sleep helper

This helper only clears the SC_OP_WAIT_FOR_{BEACON,CAB} flags.
Remove it and clear these flags directly in the approptiate
places instead.

Changes-licensed-under: ISC
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove unnecessary clearing of SC_OP_WAIT_{BEACON,CAB} flags

All SC_OP_WAIT_* flags will be cleared in 'ath9k_conf' when PS mode
is disabled, so we don't have to clear it here.

Changes-licensed-under: ISC
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: send wext MLME-MICHAELMICFAILURE.indication

Instead of having mac80211 do it itself.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: remove WE nickname support

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: convert giwrange to cfg80211

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: convert scanning to cfg80211

This removes the custom scan cache used by orinoco.

We also have to avoid calling cfg80211_scan_done from the hard
interrupt, so we offload the entirety of scan processing to a workqueue.

This may behave strangely if you start scanning just prior to
suspending...

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: convert mode setting to cfg80211

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: provide generic commit function

This allows changes to be commited from cfg80211 functions.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Handle suspend/restore in core driver

Each device does almost exactly the same things on suspend and resume
when upping and downing the interface. So move this logic into a common
routine.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

airport: store irq in card private structure

... instead of relying on the net_device fields.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: move netdev interface creation to main driver

With the move to cfg80211 it's nice to keep the hardware operations
distinct from the interface, even though we can only support a single
interface.

This also means the driver resembles other cfg80211 drivers.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: make firmware download less verbose

The firmware download code has been in a couple of releases, without any
significant issues reported in this code.

Convert to use pr_debug, so the messages can be recoverred by defining
DEBUG.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: initiate cfg80211 conversion

Initialise and register a wiphy.

Store the orinoco_private structure in the new wiphy, and use the
net_device private area to store the wireless_dev. This results in a
change to the way we navigate from a net_device to the driver private
orinoco_private, which we encapsulate in the inline function ndev_priv.
Most of the remaining calls to netdev_priv are thus replaced by
ndev_priv.

We can immediately rely on cfg80211 to handle SIOCGIWNAME, so
orinoco_ioctl_getname is removed.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Change set_tkip to use orinoco_private instead of hermes_t

hw.h does not include hermes.h, and none of the other functions
requires types from that file. Also hermes_t is a (discouraged) typedef
so we can't add a forward declaration. Therefore change this function to
use orinoco_private.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: initialise independently of netdev

Initialise the orinoco driver before registerring with netdev, which
will help when we get to cfg80211...

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Replace net_device with orinoco_private in driver interfaces

Move away from using net_device as the main structure in orinoco
function calls. Use orinoco_private instead.

This makes more sense when we move to cfg80211, and we get wiphys as
well.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: firmware helpers should use dev_err and friends

We should be able to call these routines before we register with
netdev, so avoid printks using the netdev name.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: use dev_err in early initialisation routines

This allows us to use determine_fw_capabilities,
orinoco_hw_read_card_setting and orinoco_hw_allocate_fid prior to
netdev registration.

Since dev_dbg only prints if DEBUG is defined (or dynamic debug is
enabled), move a couple of the more useful prints up to info.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Move FID allocation to hw.c

This is part of refactorring the initialisation code so that we can
load the firmware before registerring with netdev.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Move card reading code into hw.c

This is part of refactorring the initialisation code so that we can
load the firmware before registerring with netdev.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Move firmware capability determination into hw.c

This is part of refactorring the initialisation code so that we can load
the firmware before registerring with netdev.

Signed-off-by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>