bcma: add PCI ID for BCM43224

This ID was found on the PCIe wireless card on the board of a Netgear
WNDR3400 using a bcm4716. The device with this ID is identified by b43
as "Broadcom 43224 WLAN".

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: remove fix for 4329b0 bad LPOM is detection

There is not core id with 0x4329, but at the same place in the open
source part of the Broadcom SDK is a check for some device with the
chip id of 0x4329. The device with a chip id of 0x4329 is a full mac
device, so it will never be supported by bcma, this part is running in
the firmware of the device and not on the host CPU.
This code is wrong and will never be used, so just remove it.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: add constants for chip ids

The chip IDs are used all over bcma and no constants where defined.
This patch adds the constants and makes bcma use them.

Acked-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: extend workaround for bcm4331

This patch is based on a recent version of the Broadcom SDK.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fixing register bit shift values of control packets to support TPC

Some register values of bit shifts are corrected in order to support the upcoming
transmission power control (tpc) for control packets as well.

Signed-off-by: Thomas Huehn <thomas@net.t-labs.tu-berlin.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Fix clearing of BTCOEX flags

BTCOEX flags are set/cleared by atomic operations.
We got to do the same in ath9k_btcoex_timer_resume,
while clearing those BTCOEX flags.

Acked-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: parse WPS IEs from beacon_data

Parse WPS IEs from start_ap as well as change_beacon handlers
and set them to FW. Beacon IEs, Probe Response IEs and Assoc
Response IEs are parsed from beacon_data and set to FW with
related masks.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: advertise WPS probe response offload support to cfg80211

Being a fullmac driver, mwifiex takes care of populating
beacon, probe response, association response WPS IEs to firmware.
And firmware is responsible for constructing these frames.
Advertise this to cfg80211.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: add change_beacon cfg80211 handler

This patch adds support for change_beacon handler which is needed
for setting modified management IEs to driver and FW.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: overwrite earlier IE buffers for new set IE request

All IE buffers are freshly supplied. In such case, there is no need
to preserve earlier buffers with same management mask and index.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: separate IE parsing for Head/Tail IEs and beacon_ies etc

While Head/Tail IEs may contain RSN/WPA IEs which needs to be set
for all mgmt subtypes, beacon_ies, probe_resp_ies, assoc_resp_ies
are for specific mgmt subtypes. Move them to separate function.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: pass cfg80211_beacon_data to mwifiex_set_mgmt_ie()

Pass cfg80211_beacon_data pointer instead of cfg80211_ap_settings.
While setting management IEs we only need cfg80211_beacon_data of
cfg80211_ap_settings which has Tail IE, Head IE and other IE elements.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Merge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-next

Merge branch 'for-john' of git://git.sipsolutions.net/mac80211-next

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless

Conflicts:
net/mac80211/mlme.c

iwlegacy: don't mess up the SCD when removing a key

When we remove a key, we put a key index which was supposed
to tell the fw that we are actually removing the key. But
instead the fw took that index as a valid index and messed
up the SRAM of the device.

This memory corruption on the device mangled the data of
the SCD. The impact on the user is that SCD queue 2 got
stuck after having removed keys.

Reported-by: Paul Bolle <pebolle@tiscali.nl>
Cc: stable@vger.kernel.org
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlegacy: always monitor for stuck queues

This is iwlegacy version of:

commit 342bbf3fee2fa9a18147e74b2e3c4229a4564912
Author: Johannes Berg <johannes.berg@intel.com>
Date:   Sun Mar 4 08:50:46 2012 -0800

    iwlwifi: always monitor for stuck queues

    If we only monitor while associated, the following
    can happen:
     - we're associated, and the queue stuck check
       runs, setting the queue "touch" time to X
     - we disassociate, stopping the monitoring,
       which leaves the time set to X
     - almost 2s later, we associate, and enqueue
       a frame
     - before the frame is transmitted, we monitor
       for stuck queues, and find the time set to
       X, although it is now later than X + 2000ms,
       so we decide that the queue is stuck and
       erroneously restart the device

Cc: stable@vger.kernel.org
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00usb: fix indexes ordering on RX queue kick

On rt2x00_dmastart() we increase index specified by Q_INDEX and on
rt2x00_dmadone() we increase index specified by Q_INDEX_DONE. So entries
between Q_INDEX_DONE and Q_INDEX are those we currently process in the
hardware. Entries between Q_INDEX and Q_INDEX_DONE are those we can
submit to the hardware.

According to that fix rt2x00usb_kick_queue(), as we need to submit RX
entries that are not processed by the hardware. It worked before only
for empty queue, otherwise was broken.

Note that for TX queues indexes ordering are ok. We need to kick entries
that have filled skb, but was not submitted to the hardware, i.e.
started from Q_INDEX_DONE and have ENTRY_DATA_PENDING bit set.

From practical standpoint this fixes RX queue stall, usually reproducible
in AP mode, like for example reported here:
https://bugzilla.redhat.com/show_bug.cgi?id=828824

Reported-and-tested-by: Franco Miceli <fmiceli@plan.ceibal.edu.uy>
Reported-and-tested-by: Tom Horsley <horsley1953@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: fix Coverity SCAN CID 709078: Resource leak (RESOURCE_LEAK)

> *. CID 709078: Resource leak (RESOURCE_LEAK)
>  - drivers/net/wireless/mwifiex/cfg80211.c, line: 935
> Assigning: "bss_cfg" = storage returned from "kzalloc(132UL, 208U)"
>  - but was not free
> drivers/net/wireless/mwifiex/cfg80211.c:935

Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: destroy assoc_data correctly if assoc fails

If association failed due to internal error (e.g. no
supported rates IE), we call ieee80211_destroy_assoc_data()
with assoc=true, while we actually reject the association.

This results in the BSSID not being zeroed out.

After passing assoc=false, we no longer have to call
sta_info_destroy_addr() explicitly. While on it, move
the "associated" message after the assoc_success check.

Cc: stable@vger.kernel.org [3.4+]
Signed-off-by: Eliad Peller <eliad@wizery.com>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

NFC: Prevent NULL deref when getting socket name

llcp_sock_getname can be called without a device attached to the nfc_llcp_sock.

This would lead to the following BUG:

[  362.341807] BUG: unable to handle kernel NULL pointer dereference at           (null)
[  362.341815] IP: [<ffffffff836258e5>] llcp_sock_getname+0x75/0xc0
[  362.341818] PGD 31b35067 PUD 30631067 PMD 0
[  362.341821] Oops: 0000 [#627] PREEMPT SMP DEBUG_PAGEALLOC
[  362.341826] CPU 3
[  362.341827] Pid: 7816, comm: trinity-child55 Tainted: G      D W    3.5.0-rc4-next-20120628-sasha-00005-g9f23eb7 #479
[  362.341831] RIP: 0010:[<ffffffff836258e5>]  [<ffffffff836258e5>] llcp_sock_getname+0x75/0xc0
[  362.341832] RSP: 0018:ffff8800304fde88  EFLAGS: 00010286
[  362.341834] RAX: 0000000000000000 RBX: ffff880033cb8000 RCX: 0000000000000001
[  362.341835] RDX: ffff8800304fdec4 RSI: ffff8800304fdec8 RDI: ffff8800304fdeda
[  362.341836] RBP: ffff8800304fdea8 R08: 7ebcebcb772b7ffb R09: 5fbfcb9c35bdfd53
[  362.341838] R10: 4220020c54326244 R11: 0000000000000246 R12: ffff8800304fdec8
[  362.341839] R13: ffff8800304fdec4 R14: ffff8800304fdec8 R15: 0000000000000044
[  362.341841] FS:  00007effa376e700(0000) GS:ffff880035a00000(0000) knlGS:0000000000000000
[  362.341843] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  362.341844] CR2: 0000000000000000 CR3: 0000000030438000 CR4: 00000000000406e0
[  362.341851] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  362.341856] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  362.341858] Process trinity-child55 (pid: 7816, threadinfo ffff8800304fc000, task ffff880031270000)
[  362.341858] Stack:
[  362.341862]  ffff8800304fdea8 ffff880035156780 0000000000000000 0000000000001000
[  362.341865]  ffff8800304fdf78 ffffffff83183b40 00000000304fdec8 0000006000000000
[  362.341868]  ffff8800304f0027 ffffffff83729649 ffff8800304fdee8 ffff8800304fdf48
[  362.341869] Call Trace:
[  362.341874]  [<ffffffff83183b40>] sys_getpeername+0xa0/0x110
[  362.341877]  [<ffffffff83729649>] ? _raw_spin_unlock_irq+0x59/0x80
[  362.341882]  [<ffffffff810f342b>] ? do_setitimer+0x23b/0x290
[  362.341886]  [<ffffffff81985ede>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[  362.341889]  [<ffffffff8372a539>] system_call_fastpath+0x16/0x1b
[  362.341921] Code: 84 00 00 00 00 00 b8 b3 ff ff ff 48 85 db 74 54 66 41 c7 04 24 27 00 49 8d 7c 24 12 41 c7 45 00 60 00 00 00 48 8b 83 28 05 00 00 <8b> 00 41 89 44 24 04 0f b6 83 41 05 00 00 41 88 44 24 10 0f b6
[  362.341924] RIP  [<ffffffff836258e5>] llcp_sock_getname+0x75/0xc0
[  362.341925]  RSP <ffff8800304fde88>
[  362.341926] CR2: 0000000000000000
[  362.341928] ---[ end trace 6d450e935ee18bf3 ]---

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: correct size the argument to kzalloc in minstrel_ht

msp has type struct minstrel_ht_sta_priv not struct minstrel_ht_sta.

(This incorporates the fixup originally posted as "mac80211: fix kzalloc
memory corruption introduced in minstrel_ht". -- JWL)

Reported-by: Fengguang Wu <wfg@linux.intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thomas Huehn <thomas@net.t-labs.tu-berlin.de>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211_hwsim: add testmode code to stop/wake queues

This was useful for debugging the queue stop/wake
issues and is pretty small so let's just put it in.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: fix crash with single-queue drivers

Larry (and some others I think) reported that with
single-queue drivers mac80211 crashes when waking
the queues. This happens because we allocate just
a single queue for each virtual interface in case
the driver doesn't have at least 4 queues, but the
code stopping/waking the virtual interface queues
wasn't taking this into account.

Reported-by: Larry Finger <Larry.Finger@lwfinger.net>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: always set in_reconfig=false on wakeup

If the interfaces were removed just before a restart
work was started, open_count will be 0, and most of
the reconfig work will be skipped, including the
resetting of local->in_reconfig to false.

Leaving local->inconfig = true will result in
dropping any incoming packet.

Fix it by always setting local->in_reconfig = false
(even if there are no active interfaces).

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: fix debugfs default key links

Due to the way the default key links are created,
it happens that a link is left dangling:
 * both unicast/multicast links are created
 * unicast link is destroyed, and the links
   are updated
 * during this update, adding the multicast
   link again fails because it is present,
   destroying the debugfs pointer
 * removing the multicast link won't work as
   the pointer has been destroyed

Fix this by always removing the links and then
re-creating them if needed.

Reported-by: Marek Lindner <lindner_marek@yahoo.de>
Reported-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: bitrate calculation for 60g

60g band uses different from .11n MCS scheme, so bitrate
should be calculated differently

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

{nl,cfg}80211: support high bitrates

Until now, a u16 value was used to represent bitrate value.
With VHT bitrates this becomes too small.

Introduce a new 32-bit bitrate attribute. nl80211 will report
both the new and the old attribute, unless the bitrate doesn't
fit into the old u16 attribute in which case only the new one
will be reported.

User space tools encouraged to prefer the 32-bit attribute, if
available (since it won't be available on older kernels.)

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
[reword commit message and comments a bit]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: fix locking regression in monitor channel tracking

Michal's monitor channel tracking introduce a locking problem
as it locked the rdev lock inside the netdev notifier which
isn't allowed as we might already hold it if we get there by
removing an interface that is up.

Fix this by relying only on the RTNL to protect the interface
counters, the RTNL is always held in these code paths anyway.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>

iwlwifi: disallow log_event access if interface down

'echo 1 > log_event' generates the bogus "MAC is in deep sleep"
or "Timeout waiting for hardware access" log messages when
the interface is down, we should just disallow accessing the
device through debugfs when it is down.

Signed-off-by: Richard A. Griffiths <richardx.a.griffiths@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

iwlwifi: remove unneeded NULL check

There's no need to check trans for non-null
here as it has already been checked in the
caller. This fixes an smatch warning that we
check after having dereferenced it.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: add TX prepare API

Some drivers require setup before being able to send
management frames in managed mode, in particular in
multi-channel cases.

Introduce API to allow the drivers to do such setup
while being able to sleep waiting for the setup to
finish in the device. This isn't possible inside the
TX call since that can't sleep.

A future patch may also restructure the TX retry to
wait for the driver to report the frame status, as
suggested by Arik in
http://mid.gmane.org/CA+XVXffKSEL6ZQPQ98x-zO-NL2=TNF1uN==mprRyUmAaRn254g@mail.gmail.com

Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211_hwsim: fix NUM_BANDS usage

Due to the recent change of NUM_BANDS from 2 to 3 hwsim
broke. Fix the code by using the right constant but don't
support bands other than 2.4 and 5 GHz.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: reduce IEEE80211_TX_MAX_RATES

IEEE80211_TX_MAX_RATES can be reduced from 5 to 4 as there
is no current hardware supporting a rate chain with 5 multi
rate stages (mrr), so 4 mrr stages are sufficient.

The memory that is freed within the ieee80211_tx_info struct
will be used in the upcoming Transmission Power Control (TPC)
implementation.

Suggested-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Thomas Huehn <thomas@net.t-labs.tu-berlin.de>
[reword commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: remove tx_frags driver callback

The implementation of tx_frags is buggy due to
not handling queue stop, and there's no driver
implementing it so remove it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: include VHT capability IE in probe requests

Insert the VHT capability IE into probe requests.

Signed-off-by: Mahesh Palivela <maheshp@posedge.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wireless: 60g protocol constants

Provide various constants as defined by the 802.11ad:
frame types, IE's, capability bits, action categories

Introduce GCMP cipher, mandatory by 802.11ad

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wireless: regulatory for 60g

Add regulatory rule for the 60g band

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add 802.11ad (60gHz band) support

Add enumerations for both cfg80211 and nl80211.
This expands wiphy.bands etc. arrays.

Extend channel <-> frequency translation to cover 60g band
and modify the rate check logic since there are no legacy
mandatory rates (only MCS is used.)

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: allow calling ieee80211_ap_probereq_get() during auth/assoc

Drivers might need getting the probe request
(e.g. in order to extract the ssid) even during
auth/assoc.

Make ieee80211_ap_probereq_get() support it
by considering auth_data/assoc_data as well.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

iwlwifi: fix debug message level

Debug messages should be printed using dev_dbg() not
dev_err() which requires DEBUG to be defined.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

iwlwifi: add trailing newline to some messages

Some messages were missing a trailing newline, add it.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: respect iface combinations when starting operation

devlist_mtx locking is changed to accomodate changes.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add channel checking for iface combinations

.connect cannot be handled since the driver scans
and connects on its own. It is up to the driver
then to refuse a connection (with -EBUSY for
example).

Non-fixed channel IBSSes always take a single
channel resource. For example two non-fixed
channel IBSSes always take up 2
num_different_channels, even if they operate on
the same channel at a given point of time.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211/mac80211: remove .get_channel

We do not need it anymore since cfg80211 tracks
monitor channel and monitor channel type.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: set initial monitor channel

Implements behaviour seen in mac80211. A running
monitor always has a channel - even before
.set_channel. This way we won't break current
behaviour.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: track monitor channel

Make it even more obvious we support single
monitor channel. This will allow us to remove
.get_channel.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: refuse to .set_monitor_channel when non-monitors are present

Having .set_monitor_channel work with non-monitor
interfaces running would make interface
combinations accounting ambiguous.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: refactor virtual monitor code

Use cfg80211 the new .set_monitor_enabled instead
of tracking it inside mac80211.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: track monitor interfaces count

Implements .set_monitor_enabled(wiphy, enabled).

Notifies driver upon change of interface layout.

If only monitor interfaces become present it is
called with 2nd argument being true. If
non-monitor interface appears then 2nd argument
is false. Driver is notified only upon change.

This makes it more obvious about the fact that
cfg80211 supports single monitor channel. Once we
implement multi-channel we don't want to allow
setting monitor channel while other interface
types are running. Otherwise it would be ambiguous
once we start considering num_different_channels.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: introduce cfg80211_get_chan_state

Helper function for finding out which channel is
used by a given interface.

An exclusive channel can be used only by a single
interface. This is mainly for non-fixed channel
IBSS handling.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: track ibss fixed channel

IBSS may hop between channels. It is necessary to
account this special case when considering
interface combinations.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: add channel tracking for AP and mesh

We need to know which channel is used by a running
AP and mesh for channel context accounting and
finding matching/active interface combination.

STA/IBSS have current_bss already which allows us
to check which channel a vif is tuned to.
Non-fixed channel IBSS can be handled with
additional changes.

Monitor mode is going to be handled differently.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: .stop_ap when interface is going down

We'll need this for proper channel tracking (which
is going to be needed for channel context
accounting and finding matching/active interface
combination).

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

cfg80211: introduce cfg80211_stop_ap

This functionality will be reused when interface
is going down. Avoids code duplication. Also adds
missing wdev locking.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mwifiex: retrieve correct max_power information in reg_notifier handler

As we don't provide custom regulatory rules to cfg80211,
"chan->max_power" remains uninitialized (0dbm) and
"chan->max_reg_power" will contain maximum power for a channel
extracted from regulatory rules provided by CRDA; hence use
"chan->max_reg_power" in reg_notifier handler instead of
"chan->max_power" to set max_power in firmware.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: do not advertise custom regulatory domain capability

Since we don't support custom regulatory domains,
WIPHY_FLAG_CUSTOM_REGULATORY should not be enabled during wiphy
registration.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: use correct firmware command to get power limits

"priv->max_tx_power_level" and "priv->min_tx_power_level" variables
are initialized to maximum and minimum power levels supported by
hardware by sending correct firmware command.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: wakeup main thread to handle command queued

We miss to wakeup main thread after adding command to cmd pending
queue at follwing places. These commands are handled later when
main thread is woken up for handling an interrupt for sleep event
from firmware. This adds worst case delay of 50msec.

1) We don't wakeup main thread when asynchronous command is added
to cmd pending queue. Move queue_work() call from
mwifiex_wait_queue_complete() to mwifiex_send_cmd_async() to wakeup
main thread for sync as well as async commands.

2) Scan operation is triggered due to following reasons
   a) request from user (ex. "iw scan" command)
   b) Scan performed by driver internally.
   In first case main thread is woken up when first scan command is
queued in cmd pending queue (we don't need to wakeup main thread for
subsequent scan commands, because they are queued in scan command
response handler), but it is not done for second case. queue_work()
is moved inside mwifiex_scan_networks() to handle both the cases.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: de-duplicate initvals

The initvals tool from https://github.com/mcgrof/qca-swiss-army-knife has
been modified to detect identical initval tables and replace them with
macros. This patch contains the generated changes.

On MIPS this reduces the binary size by 24 KB with no runtime changes.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: update AR934x initvals to latest version

Generated using the initvals tool from the qca-swiss-army-knife repository
from https://github.com/mcgrof/qca-swiss-army-knife

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Fix signedness in a MCI debug message

seems i got a message like this
ath: phy0: BT_Status_Update: is_link=0, linkId=2,
state=1, SEQ=-2085766476 initially.

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_hw: make use of the wrapper to check for MCI init

ath9k_hw_mci_is_enabled wrapper also takes care of
ATH9K_HW_CAP_MCI being set for the AR9462 under test.

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fix ANI operation in AP mode

ath9k_ani_reset (which is called at reset time) uses a state variable
ani->update_ani to prevent the ANI noise immunity state on the operating
channel from being overwritten by background scans. Unfortunately this
is also being set for AP mode, since it's mixed with code that is only
supposed to change the default settings after a reset.

In AP mode this has the side effect of having ANI run, but being unable to
change its runtime noise immunity level, making it effectively useless.

Fix this by getting rid of ani->update_ani and passing a parameter to
ath9k_hw_set_ofdm_nil and ath9k_hw_set_cck_nil instead.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlegacy: print how long queue was actually stuck

Every now and then, after resuming from suspend, the iwlegacy driver
prints
    iwl4965 0000:03:00.0: Queue 2 stuck for 2000 ms.
    iwl4965 0000:03:00.0: On demand firmware reload

I have no idea what causes these errors. But the code currently uses
wd_timeout in the first error. wd_timeout will generally be set at
IL_DEF_WD_TIMEOUT (ie, 2000). Perhaps printing for how long the queue
was actually stuck can clarify the cause of these errors.

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Merge remote-tracking branch 'wireless-next/master' into mac80211-next

cfg80211: allow advertising VHT capabilities

Allow drivers to advertise their VHT capabilities
and export them to userspace via nl80211.

Signed-off-by: Mahesh Palivela <maheshp@posedge.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

wireless: add VHT (802.11ac) definitions

Add the VHT definitions to be used by drivers supporting it.

Signed-off-by: Mahesh Palivela <maheshp@posedge.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: don't require associated->beacon_ies for ps

beacon_ies is needed only in order to extract the dtim
period. However, even if it's missing we can still enter
ps with dtim=1 (which also happens if the TIM ie is invalid).

Most drivers don't use conf.max_sleep_period/ps_dtim_period
anyway, and this check prevents them from entering ps if
they don't have beacon (but only probe response), even though
the beacon is not needed at all.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: flush queues before deauth/disassoc

On deauth/disassoc we tear down all BA sessions. These
DELBA packets are sent on the appropriate TID, while
deauth/disassoc is always sent on VO. This sometimes
ends with the DELBA being sent after the deauth was
already sent.

Fix it by flushing all the pending frames before
sending deauth/disassoc.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mac80211: don't expose ieee80211_add_srates_ie()

This and ieee80211_add_ext_srates_ie() aren't
exported, so can't be used by drivers anyway,
but there's also no reason that they should be
so make them private to mac80211 and use sdata
instead of vif arguments.

Acked-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

mwifiex: fix memory leak associated with IE manamgement

Free ap_custom_ie before return from function.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Fix compilation breakage

Wrap the MCI-work canceling with CONFIG_ATH9K_BTCOEX_SUPPORT.

Reported-by: Emmanuel Benisty <benisty.e@gmail.com>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: fix sparse warning introduced with checkdied patch

The commit "brcmfmac: introduce checkdied debugfs functionality"
also introduced a sparse warning:

..../brcmfmac/dhd_sdio.c:3147:45: sparse: cast to restricted __le32

This patch fixes this sparse warning.

Reported-by: Fengguang Wu <wfg@linux.intel.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: add BCM4334 support

BCM4334 is a dualband a/b/g/n WiFi chip support 20MHz/40MHz
channels. This patch adds support for its SDIO interface.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: reduce allocations needed during nvram data download

The nvram data is preprocessed before being sent to the device
and just before sending an additional allocation was done that
assured word alignment of the data. This has moved to the
preprocessing step to reduce allocations and subsequent copying
of the nvram data.

Reviewed-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: use firmware data buffer directly for nvram

The nvram file could be parsed directly in the data buffer in the
firmware structure passed by request_firmware function. This patch
gets rid of the redundant memcpy.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: move glom alignment setting to SDIO bus layer

txglomming alignment is a SDIO bus specific feature. It is more
appropriate to place it in SDIO bus layer instead of common layer.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: restrict dongle txglom disable to old SDIO core

txglomming is a firmware feature for sdio bus interface. For SDIO
device cores newer than revision 11, the default setting of
firmware should be used instead of disabling it from the host side.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

brcmfmac: add support for bus specific data command

brcmfmac need to support data command setting for dongle's bus
core. A list must be placed at brcmf_bus structure before calling
brcmf_bus_start in order to be sent by brcmf_c_preinit_dcmds.

Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: define some additional cores IDs

Some of them are BCM4706 specific AFAWK. Most of them was confirmed on
Netgear WNDR450.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fix panic caused by returning a descriptor we have queued for reuse

Commit 3a2923e83c introduced a bug when a corrupt descriptor
is encountered - although the following descriptor is discarded
and returned to the queue for reuse the associated frame is
also returned for processing. This leads to a panic:

BUG: unable to handle kernel NULL pointer dereference at 000000000000003a
IP: [<ffffffffa02599a5>] ath_rx_tasklet+0x165/0x1b00 [ath9k]
Call Trace:
<IRQ>
[<ffffffff812d7fa0>] ? map_single+0x60/0x60
[<ffffffffa028f044>] ? ath9k_ioread32+0x34/0x90 [ath9k]
[<ffffffffa0292eec>] athk9k_tasklet+0xdc/0x160 [ath9k]
[<ffffffff8105e133>] tasklet_action+0x63/0xd0
[<ffffffff8105dbc0>] __do_softirq+0xc0/0x1e0
[<ffffffff8101a873>] ? native_sched_clock+0x13/0x80
[<ffffffff815f9d5c>] call_softirq+0x1c/0x30
[<ffffffff810151f5>] do_softirq+0x75/0xb0
[<ffffffff8105df95>] irq_exit+0xb5/0xc0
[<ffffffff815fa5b3>] do_IRQ+0x63/0xe0
[<ffffffff815f0cea>] common_interrupt+0x6a/0x6a
<EOI>
[<ffffffff8131840a>] ? intel_idle+0xea/0x150
[<ffffffff813183eb>] ? intel_idle+0xcb/0x150
[<ffffffff814a1db9>] cpuidle_enter+0x19/0x20
[<ffffffff814a23d9>] cpuidle_idle_call+0xa9/0x240
[<ffffffff8101c4bf>] cpu_idle+0xaf/0x120
[<ffffffff815cda8e>] rest_init+0x72/0x74
[<ffffffff81cf4c1a>] start_kernel+0x3b7/0x3c4
[<ffffffff81cf4662>] ? repair_env_string+0x5e/0x5e
[<ffffffff81cf4346>] x86_64_start_reservations+0x131/0x135
[<ffffffff81cf444a>] x86_64_start_kernel+0x100/0x10f

Making sure bf is cleared to NULL in this case restores the
old behaviour.

Signed-off-by: Tom Hughes <tom@compton.nu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: correct behaviour on unrecognised action frames

When receiving an "individually addressed" action frame, the
receiver is required to return it to the sender. mac80211
gets this wrong as it also returns group addressed (mcast)
frames to the sender. Fix this and update the reference to
the new 802.11 standards version since things were shuffled
around significantly.

Cc: stable@kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: enable serialize_regmode for non-PCIE AR9287

https://bugzilla.kernel.org/show_bug.cgi?id=42903

Based on the work of <fynivx@gmail.com>

Signed-off-by: Panayiotis Karabassis <panayk@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: rtl8192cu: New USB IDs

The latest Realtek driver for the RTL8188CU and RTL8192CU chips adds three
new USB IDs.

Reported-by: Xose Vazquez Perez <xose.vazquez@gmail.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: don't allow WoWLAN support without CONFIG_PM

When CONFIG_PM is disabled, no device can possibly
support WoWLAN since it can't go to sleep to start
with. Due to this, mac80211 had even rejected the
hardware registration. By making all the code and
data for WoWLAN depend on CONFIG_PM we can promote
this runtime error to a compile-time error.

Add #ifdef around all WoWLAN code to remove it in
systems that don't need it as they never suspend.

Cc: Kalle Valo <kvalo@qca.qualcomm.com>
Acked-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>

Merge branch 'wl12xx-next' into for-linville

ath9k_htc: Fix IDLE power save

Remove the radio enable/disable stuff and fix the
transition to FULL_SLEEP mode when the device is idle.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_htc: Use atomic operations for op_flags

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_htc: Change default listen interval to 1

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Fix lockdep splat

Cancel the MCI work only when MCI is actually enabled.
Fixes this:

[96833.124051] Call Trace:
[96833.124060]  [<ffffffff810afaf8>] __lock_acquire+0x1518/0x1e40
[96833.124065]  [<ffffffff810ad126>] ? mark_held_locks+0x86/0x110
[96833.124069]  [<ffffffff810ad3ad>] ? trace_hardirqs_on+0xd/0x10
[96833.124073]  [<ffffffff814464f0>] ? _raw_spin_unlock_irq+0x30/0x70
[96833.124078]  [<ffffffff81072968>] ? wait_on_cpu_work+0x98/0xc0
[96833.124082]  [<ffffffff810b0a11>] lock_acquire+0xa1/0x150
[96833.124085]  [<ffffffff81072990>] ? wait_on_cpu_work+0xc0/0xc0
[96833.124088]  [<ffffffff81072990>] ? wait_on_cpu_work+0xc0/0xc0
[96833.124092]  [<ffffffff810729e2>] wait_on_work+0x52/0x120
[96833.124095]  [<ffffffff81072990>] ? wait_on_cpu_work+0xc0/0xc0
[96833.124099]  [<ffffffff81063b3f>] ? del_timer+0x7f/0x110
[96833.124102]  [<ffffffff81072c13>] __cancel_work_timer+0x83/0x130
[96833.124106]  [<ffffffff81072cf0>] cancel_work_sync+0x10/0x20
[96833.124113]  [<ffffffffa065b5cd>] __ath_cancel_work+0x4d/0x60 [ath9k]
[96833.124119]  [<ffffffffa065cf28>] ath9k_config+0x458/0x680 [ath9k]
[96833.124125]  [<ffffffffa065dd1e>] ? ath9k_flush+0x6e/0x1d0 [ath9k]
[96833.124129]  [<ffffffff8144394d>] ? __mutex_unlock_slowpath+0x10d/0x190
[96833.124146]  [<ffffffffa056c7b5>] ieee80211_hw_config+0x135/0x2a0 [mac80211]
[96833.124163]  [<ffffffffa057ebbb>] ieee80211_do_open+0x67b/0xc50 [mac80211]
[96833.124178]  [<ffffffffa057f1fd>] ieee80211_open+0x6d/0x80 [mac80211]
[96833.124183]  [<ffffffff8137a44f>] __dev_open+0x9f/0xf0
[96833.124187]  [<ffffffff8137a701>] __dev_change_flags+0xa1/0x180
[96833.124190]  [<ffffffff8137a898>] dev_change_flags+0x28/0x70
[96833.124195]  [<ffffffff813e1179>] devinet_ioctl+0x659/0x780
[96833.124199]  [<ffffffff8137aea0>] ? dev_ioctl+0x210/0x6d0
[96833.124203]  [<ffffffff813e1db5>] inet_ioctl+0x75/0x90
[96833.124208]  [<ffffffff8135e0e0>] sock_do_ioctl+0x30/0x70
[96833.124211]  [<ffffffff8135e3dd>] sock_ioctl+0x7d/0x2c0
[96833.124218]  [<ffffffff81193c39>] do_vfs_ioctl+0x99/0x580
[96833.124222]  [<ffffffff81447415>] ? sysret_check+0x22/0x5d
[96833.124226]  [<ffffffff811941b9>] sys_ioctl+0x99/0xa0
[96833.124230]  [<ffffffff814473e9>] system_call_fastpath+0x16/0x1b

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: raise aggregation limit to 64k for HT IBSS

mac80211 adds stations in HT IBSS as soon as a frame comes by,
even if the HT capabilities are not known yet (they are often
received later, e.g. in beacons). So far, ampdu factor/density
are only calculated when the station is initially added.

This patch changes this to update ampdu factor/density settings
when starting a blockack session.

Using this patch, we had performance boosts from 60 to 150 MBit/s
between two 2x2 Atheros devices in 5 GHz HT IBSS mode.

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: rtl8192se: Fix double inclusion of header pci.h

The command "make includecheck" yields the following for the rtlwifi tree:

/home/finger/linux-2.6/drivers/net/wireless/rtlwifi/rtl8192se/sw.c: ../pci.h is included more than once.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Fix IRQ disabled warning

The PCI-based drivers can generate the following warning:

[ 9497.776350] ------------[ cut here ]------------
[ 9497.776366] WARNING: at kernel/softirq.c:159 local_bh_enable_ip+0x7a/0xa0()
[ 9497.776370] Hardware name: 05794NC
[ 9497.776597] Pid: 6413, comm: hostapd Not tainted 3.3.0-4.fc16.x86_64 #1
[ 9497.776601] Call Trace:
[ 9497.776612]  [<ffffffff81057b1f>] warn_slowpath_common+0x7f/0xc0
[ 9497.776633]  [<ffffffffa034a099>] ? rtl_pci_reset_trx_ring+0x199/0x230
[rtlwifi]
[ 9497.776640]  [<ffffffff81057b7a>] warn_slowpath_null+0x1a/0x20
[ 9497.776646]  [<ffffffff8105f06a>] local_bh_enable_ip+0x7a/0xa0
[ 9497.776654]  [<ffffffff815f3ef6>] _raw_spin_unlock_bh+0x16/0x20
[ 9497.776671]  [<ffffffffa03e50de>] destroy_conntrack+0x9e/0x120
[nf_conntrack]
[ 9497.776681]  [<ffffffff81511847>] nf_conntrack_destroy+0x17/0x20
[ 9497.776689]  [<ffffffff814d9c85>] skb_release_head_state+0xe5/0x120
[ 9497.776695]  [<ffffffff814d98b6>] __kfree_skb+0x16/0xa0
[ 9497.776700]  [<ffffffff814d9a35>] kfree_skb+0x45/0xc0
[ 9497.776717]  [<ffffffffa034a099>] rtl_pci_reset_trx_ring+0x199/0x230
[rtlwifi]
[ 9497.776734]  [<ffffffffa034a155>] rtl_pci_start+0x25/0x1d0 [rtlwifi]
[ 9497.776750]  [<ffffffffa03440b5>] rtl_op_start+0x55/0x90 [rtlwifi]
[ 9497.776785]  [<ffffffffa02c4956>] ieee80211_do_open+0x296/0xa10 [mac80211]
[ 9497.776794]  [<ffffffff815f7ddd>] ? notifier_call_chain+0x4d/0x70
[ 9497.776828]  [<ffffffffa02c513d>] ieee80211_open+0x6d/0x80 [mac80211]
[ 9497.776836]  [<ffffffff814e8b3f>] __dev_open+0x8f/0xe0
[ 9497.776842]  [<ffffffff814e8de1>] __dev_change_flags+0xa1/0x180
[ 9497.776847]  [<ffffffff814e8f78>] dev_change_flags+0x28/0x70
[ 9497.776856]  [<ffffffff8154e99d>] devinet_ioctl+0x61d/0x7b0
[ 9497.776863]  [<ffffffff8154ef55>] inet_ioctl+0x75/0x90
[ 9497.776870]  [<ffffffff814cdd50>] sock_do_ioctl+0x30/0x70
[ 9497.776876]  [<ffffffff814cee09>] sock_ioctl+0x79/0x2f0
[ 9497.776885]  [<ffffffff81193498>] do_vfs_ioctl+0x98/0x550
[ 9497.776891]  [<ffffffff811939e1>] sys_ioctl+0x91/0xa0
[ 9497.776897]  [<ffffffff815fc029>] system_call_fastpath+0x16/0x1b
[ 9497.776902] ---[ end trace 22886c442489082d ]---

The cause is due to calling kfree_skb() with interrupts disabled.

This bug is discussed in https://bugzilla.redhat.com/show_bug.cgi?id=797709.

Reported-and-Tested by: Ivan Ivanovich <iivanich@gmail.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Merge branch 'for-john' of git://git.sipsolutions.net/mac80211-next

Merge branch 'for-wireless' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/nfc-3.0

wlcore: print stack trace in every recovery

As recovery queuing can now occur from multiple code paths
it's convenient to know what triggered it in all cases
other than an intended recovery which is part of the
switch between single role to multi role.

Signed-off-by: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wlcore: queue recovery in case of bus errors during cmd_remove_peer

Following the addition of propagating errors from the bus ops
there's a need to distinguish between bus errors (including timeout)
and a legitimate timeout occuring in cmd_wait_for_event_or_timeout.
In case of real bus errors we need to queue recovery even in cases
where a timeout on a response from the FW to a command is acceptable.

Reported-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wlcore: fix broken TX due to wrong queuing of recovery

commit 14bba17b "wl12xx: Propagate errors from wl1271_raw_write32"
breaks down TX in certain scenarios. wl1271_irq_locked() propagates
errors from wl1271_tx_work_locked however it may return -EBUSY
when the FW queues are full which is a legitimate case and not a
a real error. In this case a recovery is triggered by wl1271_irq
and this keeps repeating itself so TX is completely broken.
Fix it by avoiding propagating return values as errors even if they
aren't. Only bus (SDIO or SPI) ops failures would be progagated
as only these should trigger recovery.

Signed-off-by: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wlcore: fix some failure cases in wlcore_probe()

We need to release the IRQ if hw_info() or identify_chip() fails.  And
we need unregister the HW with mac80211 if there are any failures
after it's registered.

Signed-off-by: Luciano Coelho <coelho@ti.com>