GitHub/MotorolaMobilityLLC/kernel-slsi.git
11 years agocfg80211: move locking into cfg80211_bss_age
Johannes Berg [Fri, 1 Feb 2013 00:04:14 +0000 (01:04 +0100)]
cfg80211: move locking into cfg80211_bss_age

There's no reason for it to require external
locking, move it into the function.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: fix BSS IE allocation comment
Johannes Berg [Wed, 30 Jan 2013 20:33:19 +0000 (21:33 +0100)]
cfg80211: fix BSS IE allocation comment

The comment about allocating the IEs together with
the BSS struct is no longer true, remove it. Also
fix a typo in the same area.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove unused SSID from BSS
Johannes Berg [Wed, 30 Jan 2013 08:03:08 +0000 (09:03 +0100)]
mac80211: remove unused SSID from BSS

The ssid/ssid_len fields in the private BSS
struct are unused, contrary to the comment
we do look up the SSID in the few cases we
need it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: remove unused cfg80211_get_mesh
Johannes Berg [Wed, 30 Jan 2013 07:50:37 +0000 (08:50 +0100)]
cfg80211: remove unused cfg80211_get_mesh

As Thomas pointed out, cfg80211_get_mesh() is
unused and can be removed.

Cc: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: simplify mesh BSS comparison
Johannes Berg [Tue, 29 Jan 2013 23:40:20 +0000 (00:40 +0100)]
cfg80211: simplify mesh BSS comparison

Instead of first checking if a BSS is an MBSS
and then doing the comparisons, inline it all
into the BSS comparison function. This avoids
doing the IE searches twice and is also a lot
less code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: fix BSS list hidden SSID lookup
Johannes Berg [Fri, 1 Feb 2013 18:20:03 +0000 (19:20 +0100)]
cfg80211: fix BSS list hidden SSID lookup

When trying to find a hidden SSID, the lookup function
is done wrong; the code is trying to combine the two
lookups into one, and as a consequence doesn't always
find the entry at all. To understand this, consider a
case where multiple BSS entries with the same channel
and BSSID exist but have different SSID length. Then
comparing against the probe response SSID length is
bound to cause problems since the hidden one might be
either zeroed out or zero-length.

To fix this we need to do two lookups for the two ways
to hide SSIDs.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: refactor hidden SSID finding
Johannes Berg [Tue, 29 Jan 2013 23:26:45 +0000 (00:26 +0100)]
cfg80211: refactor hidden SSID finding

Instead of duplicating the rbtree functions, pass
an argument to the compare function. This removes
the code duplication for the two searches.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: expand per-station byte counters to 64bit
Vladimir Kondratiev [Mon, 4 Feb 2013 11:53:11 +0000 (13:53 +0200)]
cfg80211: expand per-station byte counters to 64bit

In per-station statistics, present 32bit counters are too small
for practical purposes - with gigabit speeds, it get overlapped
every few seconds.

Expand counters in the struct station_info to be 64-bit.
Driver can still fill only 32-bit and indicate in @filled
only bits like STATION_INFO_[TR]X_BYTES; in case driver provides
full 64-bit counter, it should also set in @filled
bit STATION_INFO_[TR]RX_BYTES64

Netlink sends both 32-bit and 64-bit counters, if present, to not
break userspace.

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
[change to also have 32-bit counters if driver advertises 64-bit]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: always allow calling ieee80211_connection_loss()
Johannes Berg [Tue, 29 Jan 2013 12:13:50 +0000 (13:13 +0100)]
mac80211: always allow calling ieee80211_connection_loss()

With multi-channel, there's a corner case where a driver
doesn't receive a beacon soon enough to be able to sync
its timers with the AP. In this case, the only recovery
(after trying again) is to disconnect from the AP. Allow
calling ieee80211_connection_loss() for such cases. To
make that possible, modify the work function to not rely
on the IEEE80211_HW_CONNECTION_MONITOR flag but use new
state kept in the interface instead.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: send deauth when connection is lost
Johannes Berg [Tue, 29 Jan 2013 12:09:34 +0000 (13:09 +0100)]
mac80211: send deauth when connection is lost

If the driver determines the connection is lost,
send a deauth frame to the AP anyway just in case
it still considers the connection alive. The frame
might not go through, but at least we've tried.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: improve latency and throughput while software scanning
Stanislaw Gruszka [Wed, 23 Jan 2013 11:32:45 +0000 (12:32 +0100)]
mac80211: improve latency and throughput while software scanning

Patch vastly improve latency while scanning. Slight throughput
improvements were observed as well. Is intended for improve performance
of voice and video applications, when scan is periodically requested by
user space (i.e. default NetworkManager behaviour).

Patch remove latency requirement based on PM_QOS_NETWORK_LATENCY,
this value is 2000 seconds by default (i.e. approximately 0.5 hour !?!).

Also remove listen interval requirement, which based on beaconing and
depending on BSS parameters. It can make we stay off-channel for a
second or more.

Instead try to offer the best latency that we could, i.e. be off-channel
no longer than PASSIVE channel scan time: 125 ms. That mean we will
scan two ACTIVE channels and go back to on-channel, and one PASSIVE
channel, and go back to on-channel.

Patch also decrease PASSIVE channel scan time to about 110 ms.

As drawback patch increase overall scan time. On my tests, when scanning
both 2GHz and 5GHz bands, scanning time increase from 5 seconds up to 10
seconds. Since that increase happen only when we are associated, I think
it can be acceptable. If eventually better scan time is needed for
situations when we lose signal and quickly need to decide to which AP
roam, additional scan flag or parameter can be introduced.

I tested patch by doing:

while true; do iw dev wlan0 scan; sleep 3; done > /dev/null

and

ping -i0.2 -c 1000 HOST

on remote and local machine, results are as below:

* Ping from local periodically scanning machine to AP:
Unpatched: rtt min/avg/max/mdev = 0.928/24.946/182.135/36.873 ms
Patched:   rtt min/avg/max/mdev = 0.928/19.678/150.845/33.130 ms

* Ping from remote machine to periodically scanning machine:
Unpatched: rtt min/avg/max/mdev = 1.637/120.683/709.139/164.337 ms
Patched:   rtt min/avg/max/mdev = 1.807/26.893/201.435/40.284 ms

Throughput measured by scp show following results.

* Upload to periodically scanning machine:
Unpatched: 3.9MB/s   03:15
Patched:   4.3MB/s   02:58

* Download from periodically scanning machine:
Unpatched: 5.5MB/s   02:17
Patched:   6.2MB/s   02:02

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: start auth/assoc timeout on frame status
Johannes Berg [Tue, 29 Jan 2013 14:02:27 +0000 (15:02 +0100)]
mac80211: start auth/assoc timeout on frame status

When sending authentication/association frames they
might take a bit of time to go out because we may
have to synchronise with the AP, in particular in
the case where it's really a P2P GO. In this case
the 200ms fixed timeout could potentially be too
short if the beacon interval is relatively large.

For drivers that report TX status we can do better.
Instead of starting the timeout directly, start it
only when the frame status arrives. Since then the
frame was out on the air, we can wait shorter (the
typical response time is supposed to be 30ms, wait
100ms.) Also, if the frame failed to be transmitted
try again right away instead of waiting.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: remove free_priv BSS API
Johannes Berg [Tue, 29 Jan 2013 21:37:48 +0000 (22:37 +0100)]
cfg80211: remove free_priv BSS API

Now that mac80211 no longer uses this API, remove
it completely. If anyone needs it again, we can
revert this patch of course, but mac80211 was the
only user right now.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove unused mesh data from bss
Johannes Berg [Tue, 29 Jan 2013 21:37:04 +0000 (22:37 +0100)]
mac80211: remove unused mesh data from bss

These pointers/values are never used, remove them.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove last_probe_resp from bss
Johannes Berg [Tue, 29 Jan 2013 21:33:30 +0000 (22:33 +0100)]
mac80211: remove last_probe_resp from bss

We track this, but never use it, so we can
just remove it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: inform the driver about update of dtim_period
Emmanuel Grumbach [Wed, 12 Dec 2012 08:12:24 +0000 (10:12 +0200)]
mac80211: inform the driver about update of dtim_period

Currently, when the driver requires the DTIM period,
mac80211 will wait to hear a beacon before association.
This behavior is suboptimal since some drivers may be
able to deal with knowing the DTIM period after the
association, if they get it at all.

To address this, notify the drivers with bss_info_changed
with the new BSS_CHANGED_DTIM_PERIOD flag when the DTIM
becomes known. This might be when changing to associated,
or later when the entire association was done with only
probe response information.

Rename the hardware flag for the current behaviour to
IEEE80211_HW_NEED_DTIM_BEFORE_ASSOC to more accurately
reflect its behaviour. IEEE80211_HW_NEED_DTIM_PERIOD is
no longer accurate as all drivers get the DTIM period
now, just not before association.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove assoc data "sent_assoc"
Johannes Berg [Thu, 24 Jan 2013 23:44:12 +0000 (00:44 +0100)]
mac80211: remove assoc data "sent_assoc"

The field is never used, so remove it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211/mac80211: support reporting wakeup reason
Johannes Berg [Tue, 22 Jan 2013 11:34:29 +0000 (12:34 +0100)]
cfg80211/mac80211: support reporting wakeup reason

When waking up from WoWLAN, it is useful to know
what triggered the wakeup. Support reporting the
wakeup reason(s) in cfg80211 (and a pass-through
in mac80211) to allow userspace to know.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: dynamic short slot time for MBSSs
Thomas Pedersen [Sun, 27 Jan 2013 00:39:55 +0000 (16:39 -0800)]
mac80211: dynamic short slot time for MBSSs

The standard mandates mesh STAs to set the ERP Short Slot
Time capability info bit in beacons to 0. Even though this
is their way of disallowing short slot time for mesh STAs,
there should be no harm in enabling it if we determine all
STAs in the current MBSS support ERP rates.

Increases throughput about 20% for legacy rates when
enabled.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: add SME state to warning in __cfg80211_mlme_disassoc
Johannes Berg [Sat, 26 Jan 2013 21:16:08 +0000 (22:16 +0100)]
cfg80211: add SME state to warning in __cfg80211_mlme_disassoc

The warning here occasionally triggers but we haven't
found the cause yet. It's a valid warning since if it
triggers the SME state got confused, so add the SME
state to it to help narrow it down in the future.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove IEEE80211_HW_TEARDOWN_AGGR_ON_BAR_FAIL
Stanislaw Gruszka [Mon, 28 Jan 2013 13:42:30 +0000 (14:42 +0100)]
mac80211: remove IEEE80211_HW_TEARDOWN_AGGR_ON_BAR_FAIL

This is basically a revert of:

commit 5b632fe85ec82e5c43740b52e74c66df50a37db3
Author: Stanislaw Gruszka <sgruszka@redhat.com>
Date:   Mon Dec 3 12:56:33 2012 +0100

    mac80211: introduce IEEE80211_HW_TEARDOWN_AGGR_ON_BAR_FAIL

We do not need this flag any longer, rt2x00 BAR/BA problem was fixed
correctly by wireless-testing commit:

commit 84e9e8ebd369679a958200a8baca96aafb2393bb
Author: Helmut Schaa <helmut.schaa@googlemail.com>
Date:   Thu Jan 17 17:34:32 2013 +0100

    rt2x00: Improve TX status handling for BlockAckReq frames

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoMerge remote-tracking branch 'wireless-next/master' into HEAD
Johannes Berg [Tue, 29 Jan 2013 11:16:22 +0000 (12:16 +0100)]
Merge remote-tracking branch 'wireless-next/master' into HEAD

11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
John W. Linville [Mon, 28 Jan 2013 18:54:03 +0000 (13:54 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless

Conflicts:
drivers/net/wireless/ath/ath9k/main.c
drivers/net/wireless/iwlwifi/dvm/tx.c

11 years agomac80211: fix mesh_sta_info_get() reshuffle damage
Thomas Pedersen [Sun, 27 Jan 2013 01:00:02 +0000 (17:00 -0800)]
mac80211: fix mesh_sta_info_get() reshuffle damage

Before "mac80211: clean up mesh sta allocation warning"
was applied, mesh_sta_info_get() was reshuffled to please
sparse. As a result we neglect to initialize newly created
STAs. Fix this.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211/nl80211: add API for MAC address ACLs
Vasanthakumar Thiagarajan [Fri, 18 Jan 2013 05:48:45 +0000 (11:18 +0530)]
cfg80211/nl80211: add API for MAC address ACLs

Add API to enable drivers to implement MAC address based
access control in AP/P2P GO mode. Capable drivers advertise
this capability by setting the maximum number of MAC
addresses in such a list in wiphy->max_acl_mac_addrs.

An initial ACL may be given to the NL80211_CMD_START_AP
command and/or changed later with NL80211_CMD_SET_MAC_ACL.

Black- and whitelists are supported, but not simultaneously.

Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
[rewrite commit log, many cleanups]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: Move the definition of struct mac_address up
Vasanthakumar Thiagarajan [Fri, 18 Jan 2013 05:48:44 +0000 (11:18 +0530)]
cfg80211: Move the definition of struct mac_address up

struct mac_address will be used by ACL related configuration ops.

Signed-off-by: Vasanthakumar Thiagarajan <vthiagar@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: avoid a build warning
Arnd Bergmann [Fri, 25 Jan 2013 14:14:33 +0000 (14:14 +0000)]
mac80211: avoid a build warning

gcc cannot prove that the value of sdata->vif.type does not
change between the switch() statement and the second
comparison to NL80211_IFTYPE_AP, causing a harmless
warning.
Slightly reordering the code makes the warning go away
with no functional change.

Without this patch, building ARM at91sam9g45_defconfig with
gcc-4.6 results in:

net/mac80211/tx.c: In function 'ieee80211_subif_start_xmit':
net/mac80211/tx.c:1797:22: warning: 'chanctx_conf' may be used uninitialized in this function [-Wuninitialized]

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: "John W. Linville" <linville@tuxdriver.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: linux-wireless@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove redundant check
Johannes Berg [Fri, 18 Jan 2013 20:21:31 +0000 (21:21 +0100)]
mac80211: remove redundant check

There's no need to have two checks for "associated"
in ieee80211_sta_restart(), make the first one locked
to not race (unlikely at this point during resume)
and remove the second check.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: fix channel check in cfg80211_can_use_iftype_chan
Simon Wunderlich [Wed, 23 Jan 2013 14:15:57 +0000 (15:15 +0100)]
cfg80211: fix channel check in cfg80211_can_use_iftype_chan

In commit "cfg80211: check radar interface combinations" a regression
was introduced which might lead to NULL dereference if the argument
chan = NULL, which might happen in IBSS/wext case (and probably
others).

Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: support mesh rate updates
Thomas Pedersen [Wed, 23 Jan 2013 20:18:13 +0000 (12:18 -0800)]
mac80211: support mesh rate updates

An existing mesh station entry may change its rate
capabilities, so call rate_control_rate_update() to notify
the rate control.

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
[fix compilation]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: clean up mesh sta allocation warning
Thomas Pedersen [Wed, 23 Jan 2013 20:18:12 +0000 (12:18 -0800)]
mac80211: clean up mesh sta allocation warning

This refactoring fixes a "scheduling while atomic" warning
when allocating a mesh station entry while holding the RCU
read lock. Fix this by creating a new function
mesh_sta_info_get(), which correctly handles the locking
and returns under RCU.

Also move some unnecessarily #ifdefed mesh station init
code from sta_info_alloc() to __mesh_sta_info_alloc().

Signed-off-by: Thomas Pedersen <thomas@cozybit.com>
[change code flow to make sparse happy]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoieee80211: define AKM suite selectors type 5, 6 and 7
Bing Zhao [Thu, 24 Jan 2013 04:33:58 +0000 (20:33 -0800)]
ieee80211: define AKM suite selectors type 5, 6 and 7

Reference: IEEE 802.11-2012 8.4.2.27.3 "AKM suites"

Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: fix aggregation state with current drivers
Johannes Berg [Fri, 18 Jan 2013 22:47:05 +0000 (23:47 +0100)]
mac80211: fix aggregation state with current drivers

For drivers that don't actually flush their queues when
aggregation stop with the IEEE80211_AMPDU_TX_STOP_FLUSH
or IEEE80211_AMPDU_TX_STOP_FLUSH_CONT reasons is done,
like iwlwifi or iwlegacy, mac80211 can then transmit on
a TID that the driver still considers busy. This happens
in the following way:

 - IEEE80211_AMPDU_TX_STOP_FLUSH requested
 - driver marks TID as emptying
 - mac80211 removes tid_tx data, this can copy packets
   to the TX pending queues and also let new packets
   through to the driver
 - driver gets unexpected TX as it wasn't completely
   converted to the new API

In iwlwifi, this lead to the following warning:

WARNING: at drivers/net/wireless/iwlwifi/dvm/tx.c:442 iwlagn_tx_skb+0xc47/0xce0
Tx while agg.state = 4
Modules linked in: [...]
Pid: 0, comm: kworker/0:0 Tainted: G        W   3.1.0 #1
Call Trace:
 [<c1046e42>] warn_slowpath_common+0x72/0xa0
 [<c1046f13>] warn_slowpath_fmt+0x33/0x40
 [<fddffa17>] iwlagn_tx_skb+0xc47/0xce0 [iwldvm]
 [<fddfcaa3>] iwlagn_mac_tx+0x23/0x40 [iwldvm]
 [<fd8c98b6>] __ieee80211_tx+0xf6/0x3c0 [mac80211]
 [<fd8cbe00>] ieee80211_tx+0xd0/0x100 [mac80211]
 [<fd8cc176>] ieee80211_xmit+0x96/0xe0 [mac80211]
 [<fd8cc578>] ieee80211_subif_start_xmit+0x348/0xc80 [mac80211]
 [<c1445207>] dev_hard_start_xmit+0x337/0x6d0
 [<c145eee9>] sch_direct_xmit+0xa9/0x210
 [<c14462c0>] dev_queue_xmit+0x1b0/0x8e0

Fortunately, solving this problem is easy as the station
is being destroyed, so such transmit packets can only
happen due to races. Instead of trying to close the race
just let the race not reach the drivers by making two
changes:
 1) remove the explicit aggregation session teardown in
    the managed mode code, the same thing will be done
    when the station is removed, in __sta_info_destroy.
 2) When aggregation stop with AGG_STOP_DESTROY_STA is
    requested, leave the tid_tx data around as stopped.
    It will be cleared and freed in cleanup_single_sta
    later, but until then any racy packets will be put
    onto the tid_tx pending queue instead of transmitted
    which is fine since the station is being removed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: provide the vif in rssi_callback
Emmanuel Grumbach [Sun, 20 Jan 2013 15:32:41 +0000 (17:32 +0200)]
mac80211: provide the vif in rssi_callback

Since drivers can support several BSS / P2P Client
interfaces, the rssi callback needs to inform the driver
about the interface teh rssi event relates to.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoath9k_hw: remove a useless WARN_ON
Felix Fietkau [Sun, 20 Jan 2013 17:51:59 +0000 (18:51 +0100)]
ath9k_hw: remove a useless WARN_ON

&ah->curchan->ani can never be NULL

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: reduce struct ar5416AniState size
Felix Fietkau [Sun, 20 Jan 2013 17:51:58 +0000 (18:51 +0100)]
ath9k_hw: reduce struct ar5416AniState size

It is kept per-channel, so removing unnecessary (or constant) fields from
it can save quite a bit of memory.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: remove ath9k_hw_ani_setup and its variables
Felix Fietkau [Sun, 20 Jan 2013 17:51:57 +0000 (18:51 +0100)]
ath9k_hw: remove ath9k_hw_ani_setup and its variables

They are no longer needed for ANI functionality

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: make the initval parameter to ath9k_hw_write_array const
Felix Fietkau [Sun, 20 Jan 2013 17:51:55 +0000 (18:51 +0100)]
ath9k_hw: make the initval parameter to ath9k_hw_write_array const

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: use ath_tx_process_buffer instead of open-coding similar code
Felix Fietkau [Sun, 20 Jan 2013 17:51:54 +0000 (18:51 +0100)]
ath9k: use ath_tx_process_buffer instead of open-coding similar code

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: clean up processing of pending tx frames on reset
Felix Fietkau [Sun, 20 Jan 2013 17:51:53 +0000 (18:51 +0100)]
ath9k: clean up processing of pending tx frames on reset

Dropping packets from aggregation sessions is usually not a good idea, as
it might upset the synchronization of the BlockAck receive window of the
remote node. The use of the retry_tx parameter to reset/tx-drain functions
also seemed a bit arbitrary.
This patch removes this parameter altogether and ensures that pending tx
frames are not dropped for no good reason.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: stop rx after tx
Felix Fietkau [Sun, 20 Jan 2013 17:51:52 +0000 (18:51 +0100)]
ath9k: stop rx after tx

Completing frame transmission can fail if the rx engine is stopped
prematurely, as the hw might be waiting for an ACK from the other side.
Shutting down tx before rx might make the DMA shutdown more reliable.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agonet: wireless/rtlwifi: fix uninitialized variable issue
Cong Ding [Sat, 19 Jan 2013 12:56:36 +0000 (13:56 +0100)]
net: wireless/rtlwifi: fix uninitialized variable issue

The use of variable packet_beacon might be uninitialized in the three files.

Signed-off-by: Cong Ding <dinggnu@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoiwlegacy: don't return zero on failure paths in il4965_pci_probe()
Alexey Khoroshilov [Sat, 19 Jan 2013 12:56:34 +0000 (16:56 +0400)]
iwlegacy: don't return zero on failure paths in il4965_pci_probe()

If hardware is not ready, il4965_pci_probe() breaks off initialization,
deallocates all resources, but returns zero.
The patch adds -EIO as return value in this case.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agomwl8k: don't return zero on failure paths in mwl8k_probe[_hw]()
Alexey Khoroshilov [Fri, 18 Jan 2013 21:24:01 +0000 (01:24 +0400)]
mwl8k: don't return zero on failure paths in mwl8k_probe[_hw]()

If pci_iomap() fails in mwl8k_probe(), it breaks off initialization,
deallocates all resources, but returns zero.
There is a similar issue when priv->rxd_ops is NULL in mwl8k_probe_hw().

The patch adds proper error code return values.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agomwifiex: correction in status codes used for association failure
Amitkumar Karwar [Fri, 18 Jan 2013 01:41:57 +0000 (17:41 -0800)]
mwifiex: correction in status codes used for association failure

When AP responds with appropriate status code, we forward that
code correctly to cfg80211. But sometimes when there is no
response from AP, our firmware uses proprietary status codes.
We will map authentication timeout to WLAN_STATUS_AUTH_TIMEOUT
and other proprietary codes to WLAN_STATUS_UNSPECIFIED_FAILURE.

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>
11 years agort2x00: Improve TX status handling for BlockAckReq frames
Helmut Schaa [Thu, 17 Jan 2013 16:34:32 +0000 (17:34 +0100)]
rt2x00: Improve TX status handling for BlockAckReq frames

Since rt2800 hardware isn't capable of reporting the TX status of
BlockAckReq frames implement the TX status handling of BARs in
rt2x00lib. We keep track of all BARs that are send out and try to
match incoming BAs to the appropriate BARs. This allows us to report a
more or less accurate TX status for BAR frames which in turn improves
BA session stability.

This is loosley based on Christian Lamparter's patch for carl9170
"carl9170: fix HT peer BA session corruption".

We have to walk the list of pending BARs for every rx'red BA even
though most BAs don't belong to any of these BARs as they are just
acknowledging an AMPDU. To keep that overhead low use RCU which allows
us to walk the list of pending BARs without the need to acquire a lock.
This however requires us to _copy_ relevant information from the BAR
(RA, TA, control field, start sequence number) into our BAR list entry.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Tested-by: Andreas Hartmann <andihartmann@01019freenet.de>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoprism54: bug in getting auth type
Dan Carpenter [Sun, 13 Jan 2013 20:03:42 +0000 (23:03 +0300)]
prism54: bug in getting auth type

There is a missing break statement so SHARED_KEY authentication doesn't
work.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agob43: N-PHY: fix gain in b43_nphy_get_gain_ctl_workaround_ent()
Dan Carpenter [Sun, 20 Jan 2013 16:31:30 +0000 (19:31 +0300)]
b43: N-PHY: fix gain in b43_nphy_get_gain_ctl_workaround_ent()

There were no break statements in this switch statement so everything
used the default settings.  Per Walter Harms's suggestion, I've replaced
the switch statement and done a little cleanup.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agodrivers/net/wireless: need consider the not '\0' terminated string.
Chen Gang [Tue, 8 Jan 2013 05:33:03 +0000 (13:33 +0800)]
drivers/net/wireless: need consider the not '\0' terminated string.

  in ray_cs.c:
    the a_current_ess_id is "Null terminated unless ESSID_SIZE long"
    so we need buffer it with '\0' firstly, before using strlen or %s.

  additional information:
    in drivers/net/wireless/rayctl.h:
      "NULL terminated unless 32 long" is a comment at line 616, 664
      ESSID_SIZE is 32, at line 190
    in include/uapi/linux/wireless.h:
      IW_ESSID_MAX_SIZE is also 32
    in drivers/net/wireless/ray_cs.c:
      use strncpy for it, without '\0' terminated, at line 639
      use memcpy for it, assume not '\0' terminated in line 1092..1097
      buffer it with '\0' firstly, before using %s, in line 2576, 2598..2600

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: fix RF bank initialization
Felix Fietkau [Sun, 13 Jan 2013 18:54:58 +0000 (19:54 +0100)]
ath9k_hw: fix RF bank initialization

ar900*_init_mode_regs needs to be called before RF banks are allocated,
otherwise the storage size of RF banks isn't known. This patch fixes
a memory overrun that can show up as a crash on unloading the module.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac802...
John W. Linville [Tue, 22 Jan 2013 20:40:56 +0000 (15:40 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/jberg/mac80211-next

11 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi...
John W. Linville [Tue, 22 Jan 2013 20:39:51 +0000 (15:39 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/iwlwifi/iwlwifi-next

11 years agomwifiex: fix typo in PCIe adapter NULL check
Avinash Patil [Tue, 22 Jan 2013 05:04:10 +0000 (21:04 -0800)]
mwifiex: fix typo in PCIe adapter NULL check

Add missing "!" as we are supposed to check "!card->adapter"
in PCIe suspend handler.

Cc: "3.2+" <stable@vger.kernel.org>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Reviewed-by: Sergey V. <sftp.mtuci@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: allow setting arbitrary antenna masks on AR9003+
Felix Fietkau [Sun, 20 Jan 2013 20:55:22 +0000 (21:55 +0100)]
ath9k: allow setting arbitrary antenna masks on AR9003+

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: fix chain swap setting when setting rx chainmask to 5
Felix Fietkau [Sun, 20 Jan 2013 20:55:21 +0000 (21:55 +0100)]
ath9k_hw: fix chain swap setting when setting rx chainmask to 5

Chain swapping should only be enabled when the EEPROM chainmask is set to 5,
regardless of what the runtime chainmask is.

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: fix calibration issues on chainmask that don't include chain 0
Felix Fietkau [Sun, 20 Jan 2013 20:55:20 +0000 (21:55 +0100)]
ath9k_hw: fix calibration issues on chainmask that don't include chain 0

Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agortlwifi: Fix build warning introduced by commit a290593
Larry Finger [Sat, 19 Jan 2013 18:36:49 +0000 (12:36 -0600)]
rtlwifi: Fix build warning introduced by commit a290593

The kbuild test robot reports the following warning with x86_64-randconfig-x955:

warning: (RTL8192CE && RTL8192SE && RTL8192DE && RTL8723AE && RTL8192CU) selects
 RTLWIFI which has unmet direct dependencies (NETDEVICES && WLAN &&
 (RTL8192CE || RTL8192CU || RTL8192SE || RTL8192DE))

This warning was introduced in commit a290593, "rtlwifi: Modify files for addition
of rtl8723ae", and is d ue to a missing dependence of RTLWIFI on RTL8723AE.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: kbuild test robot <fengguang.wu@intel.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoiwlegacy: fix IBSS cleanup
Stanislaw Gruszka [Wed, 16 Jan 2013 10:45:15 +0000 (11:45 +0100)]
iwlegacy: fix IBSS cleanup

We do not correctly change interface type when switching from
IBSS mode to STA mode, that results in microcode errors.

Resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=886946

Reported-by: Jaroslav Skarvada <jskarvad@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211
John W. Linville [Tue, 22 Jan 2013 19:23:13 +0000 (14:23 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/jberg/mac80211

11 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi...
John W. Linville [Tue, 22 Jan 2013 19:22:01 +0000 (14:22 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/iwlwifi/iwlwifi-fixes

11 years agomac80211: allow drivers to access IPv6 information
Johannes Berg [Mon, 14 Jan 2013 14:14:34 +0000 (15:14 +0100)]
mac80211: allow drivers to access IPv6 information

To be able to implement NS response offloading (in
regular operation or while in WoWLAN) drivers need
to know the IPv6 addresses assigned to interfaces.
Implement an IPv6 notifier in mac80211 to call the
driver when addresses change.

Unlike for IPv4, implement it as a callback rather
than as a list in the BSS configuration, that is
more flexible.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: improve aggregation debug messages
Johannes Berg [Wed, 16 Jan 2013 09:39:44 +0000 (10:39 +0100)]
mac80211: improve aggregation debug messages

A lot of the aggregation messages don't indicate the
station so they're hard to understand if there are
multiple sessions in progress. Make that easier by
adding the MAC address to most messages. Also add
the TID if it wasn't already there.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: remove ARP filter enable/disable logic
Johannes Berg [Mon, 14 Jan 2013 15:39:07 +0000 (16:39 +0100)]
mac80211: remove ARP filter enable/disable logic

Depending on the driver, having ARP filtering for
some addresses may be possible. Remove the logic
that tracks whether ARP filter is enabled or not
and give the driver the total number of addresses
instead of the length of the list so it can make
its own decision.

Reviewed-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: add op to configure default key id
Yoni Divinsky [Wed, 30 May 2012 08:36:39 +0000 (11:36 +0300)]
mac80211: add op to configure default key id

There are hardwares which support offload of data packets
for example when auto ARP is enabled the hw will send
the ARP response. In such cases if WEP encryption is
configured the hw must know the default WEP key in order
to encrypt the packets correctly.

When hw_accel is enabled and encryption type is set to WEP,
the driver should get the default key index from mac80211.

Signed-off-by: Yoni Divinsky <yoni.divinsky@ti.com>
[cleanups, fixes, documentation]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: split out chandef tracing macros
Johannes Berg [Fri, 21 Dec 2012 11:36:33 +0000 (12:36 +0100)]
mac80211: split out chandef tracing macros

Some new APIs will require tracing a chandef without
it being part of a channel context, so separate out
the tracing macros for that.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: check radar interface combinations
Simon Wunderlich [Tue, 8 Jan 2013 13:04:07 +0000 (14:04 +0100)]
cfg80211: check radar interface combinations

To ease further DFS development regarding interface combinations, use
the interface combinations structure to test for radar capabilities.
Drivers can specify which channel widths they support, and in which
modes. Right now only a single AP interface is allowed, but as the
DFS code evolves other combinations can be enabled.

Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agocfg80211: Allow use_mfp to be specified with the connect command
Jouni Malinen [Tue, 15 Jan 2013 15:15:57 +0000 (17:15 +0200)]
cfg80211: Allow use_mfp to be specified with the connect command

The NL80211_ATTR_USE_MFP attribute was originally added for
NL80211_CMD_ASSOCIATE, but it is actually as useful (if not even more
useful) with NL80211_CMD_CONNECT, so process that attribute with the
connect command, too.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: add encrypt headroom to PERR frames
Bob Copeland [Wed, 9 Jan 2013 17:34:56 +0000 (12:34 -0500)]
mac80211: add encrypt headroom to PERR frames

Mesh PERR action frames are robust and thus may be encrypted, so add
proper head/tailroom to allow this.  Fixes this warning when operating
a Mesh STA on ath5k:

WARNING: at net/mac80211/wpa.c:427 ccmp_encrypt_skb.isra.5+0x7b/0x1a0 [mac80211]()
Call Trace:
 [<c011c5e7>] warn_slowpath_common+0x63/0x78
 [<c011c60b>] warn_slowpath_null+0xf/0x13
 [<e090621d>] ccmp_encrypt_skb.isra.5+0x7b/0x1a0 [mac80211]
 [<e090685c>] ieee80211_crypto_ccmp_encrypt+0x1f/0x37 [mac80211]
 [<e0917113>] invoke_tx_handlers+0xcad/0x10bd [mac80211]
 [<e0917665>] ieee80211_tx+0x87/0xb3 [mac80211]
 [<e0918932>] ieee80211_tx_pending+0xcc/0x170 [mac80211]
 [<c0121c43>] tasklet_action+0x3e/0x65

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: set NEED_TXPROCESSING for PERR frames
Bob Copeland [Wed, 9 Jan 2013 17:34:55 +0000 (12:34 -0500)]
mac80211: set NEED_TXPROCESSING for PERR frames

A user reported warnings in ath5k due to transmitting frames with no
rates set up.  The frames were Mesh PERR frames, and some debugging
showed an empty control block with just the vif pointer:

>  [  562.522682] XXX txinfo: 00000000: 00 00 00 00 00 00 00 00 00 00 00
>  00 00 00 00 00  ................
>  [  562.522688] XXX txinfo: 00000010: 00 00 00 00 00 00 00 00 54 b8 f2
>  db 00 00 00 00  ........T.......
>  [  562.522693] XXX txinfo: 00000020: 00 00 00 00 00 00 00 00 00 00 00
>  00 00 00 00 00  ................

Set the IEEE80211_TX_INTFL_NEED_TXPROCESSING flag to ensure that
rate control gets run before the frame is sent.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agonl80211: allow user-space to set address for P2P_DEVICE
Arend van Spriel [Tue, 8 Jan 2013 09:17:27 +0000 (10:17 +0100)]
nl80211: allow user-space to set address for P2P_DEVICE

As per email discussion Jouni Malinen pointed out that:

"P2P message exchanges can be executed on the current operating channel
of any operation (both P2P and non-P2P station). These can be on 5 GHz
and even on 60 GHz (so yes, you _can_ do GO Negotiation on 60 GHz).

As an example, it would be possible to receive a GO Negotiation Request
frame on a 5 GHz only radio and then to complete GO Negotiation on that
band. This can happen both when connected to a P2P group (through client
discoverability mechanism) and when connected to a legacy AP (assuming
the station receive Probe Request frame from full scan in the beginning
of P2P device discovery)."

This means that P2P messages can be sent over different radio devices.
However, these should use the same P2P device address so it should be
able to provision this from user-space. This patch adds a parameter for
this to struct vif_params which should only be used during creation of
the P2P device interface.

Cc: Jouni Malinen <j@w1.fi>
Cc: Greg Goldman <ggoldman@broadcom.com>
Cc: Jithu Jance <jithu@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
[add error checking]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years ago{cfg,nl}80211: mesh power mode primitives and userspace access
Marco Porsch [Mon, 7 Jan 2013 15:04:52 +0000 (16:04 +0100)]
{cfg,nl}80211: mesh power mode primitives and userspace access

Add the nl80211_mesh_power_mode enumeration which holds possible
values for the mesh power mode. These modes are unknown, active,
light sleep and deep sleep.

Add power_mode entry to the mesh config structure to hold the
user-configured default mesh power mode. This value will be used
for new peer links.

Add the dot11MeshAwakeWindowDuration value to the mesh config.
The awake window is a duration in TU describing how long the STA
will stay awake after transmitting its beacon in PS mode.

Add access routines to:
 - get/set local link-specific power mode (STA)
 - get remote STA's link-specific power mode (STA)
 - get remote STA's non-peer power mode (STA)
 - get/set default mesh power mode (mesh config)
 - get/set mesh awake window duration (mesh config)

All config changes may be done at mesh runtime and take effect
immediately.

Signed-off-by: Marco Porsch <marco@cozybit.com>
Signed-off-by: Ivan Bezyazychnyy <ivan.bezyazychnyy@gmail.com>
Signed-off-by: Mike Krinkin <krinkin.m.u@gmail.com>
[fix commit message line length, error handling in set station]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years ago{cfg,nl,mac}80211: set beacon interval and DTIM period on mesh join
Marco Porsch [Mon, 7 Jan 2013 15:04:51 +0000 (16:04 +0100)]
{cfg,nl,mac}80211: set beacon interval and DTIM period on mesh join

Move the default mesh beacon interval and DTIM period to cfg80211
and make them accessible to nl80211. This enables setting both
values when joining an MBSS.

Previously the DTIM parameter was not set by mac80211 so the
driver's default value was used.

Signed-off-by: Marco Porsch <marco@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: move add_tim to subfunction
Marco Porsch [Mon, 7 Jan 2013 15:04:50 +0000 (16:04 +0100)]
mac80211: move add_tim to subfunction

This functions will be used for mesh beacons, too.

Signed-off-by: Marco Porsch <marco@cozybit.com>
[some formatting fixes]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: update mesh peer link counter during userspace peering
Marco Porsch [Mon, 7 Jan 2013 15:04:49 +0000 (16:04 +0100)]
mac80211: update mesh peer link counter during userspace peering

The established peer link count is indicated in mesh beacons and
used for other internal tasks. Previously it was not updated when
authenticated peering is performed in userspace.

Signed-off-by: Marco Porsch <marco@cozybit.com>
Acked-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agonl80211: add range checks to mesh parameters
Marco Porsch [Mon, 7 Jan 2013 15:04:48 +0000 (16:04 +0100)]
nl80211: add range checks to mesh parameters

Ranges are taken from IEEE 802.11-2012, common sense or current
implementation requirements.

Signed-off-by: Marco Porsch <marco@cozybit.com>
Acked-by: Thomas Pedersen <thomas@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: allow disable power save in mesh
Chun-Yeow Yeoh [Thu, 10 Jan 2013 15:31:54 +0000 (23:31 +0800)]
mac80211: allow disable power save in mesh

Allow mesh interface to disable the power save which is by default
turn on in certain chipset. Testing with 2 units of ZCN-1523H-5-16
featuring AR9280 chipset which have power save enabled by default.
Constant reset if the average signal of the peer mesh STA is below
-80 dBm and power save is enabled.

Signed-off-by: Chun-Yeow Yeoh <yeohchunyeow@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: call restart complete at wowlan resume time
Johannes Berg [Thu, 10 Jan 2013 23:28:01 +0000 (00:28 +0100)]
mac80211: call restart complete at wowlan resume time

When the driver's resume function can't completely
restore the configuration in the device, it returns
1 from the callback which will be treated like a HW
restart request, but done directly.

In this case, also call the driver's restart_complete()
function so it can finish the reconfiguration there.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: fix resume from WoWLAN
Johannes Berg [Thu, 10 Jan 2013 22:55:33 +0000 (23:55 +0100)]
mac80211: fix resume from WoWLAN

My commit 529ba6e9313dbe60dab7e72c6fdf647a012e9f5b
("mac80211: clean up association better in suspend")
introduced a bug when resuming from WoWLAN when a
device reset is desired. This case must not use the
suspend_bss_conf as it hasn't been stored.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years ago{cfg,mac}80211.h: fix some kernel-doc warnings
Yacine Belkadi [Sat, 12 Jan 2013 12:54:14 +0000 (13:54 +0100)]
{cfg,mac}80211.h: fix some kernel-doc warnings

When building the 80211 DocBook, scripts/kernel-doc reports
the following type of warnings:

Warning(include/net/cfg80211.h:334): No description found for return value of 'cfg80211_get_chandef_type'

These warnings are only reported when scripts/kernel-doc
runs in verbose mode.

To fix these use "Return:" to describe function return values.

Signed-off-by: Yacine Belkadi <yacine.belkadi.1@gmail.com>
[adjust for freq_reg_info() change]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: fix monitor mode injection
Felix Fietkau [Sun, 13 Jan 2013 22:10:26 +0000 (23:10 +0100)]
mac80211: fix monitor mode injection

Channel contexts are not always used with monitor interfaces. If no channel
context is set, use the oper channel, otherwise tx fails.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
[check local->use_chanctx]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: synchronize scan off/on-channel and PS states
Stanislaw Gruszka [Thu, 20 Dec 2012 13:41:18 +0000 (14:41 +0100)]
mac80211: synchronize scan off/on-channel and PS states

Since:

commit b23b025fe246f3acc2988eb6d400df34c27cb8ae
Author: Ben Greear <greearb@candelatech.com>
Date:   Fri Feb 4 11:54:17 2011 -0800

    mac80211: Optimize scans on current operating channel.

we do not disable PS while going back to operational channel (on
ieee80211_scan_state_suspend) and deffer that until scan finish.
But since we are allowed to send frames, we can send a frame to AP
without PM bit set, so disable PS on AP side. Then when we switch
to off-channel (in ieee80211_scan_state_resume) we do not enable PS.
Hence we are off-channel with PS disabled, frames are not buffered
by AP.

To fix remove offchannel_ps_disable argument and always enable PS when
going off-channel and disable it when going on-channel, like it was
before.

Cc: stable@vger.kernel.org # 2.6.39+
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Tested-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agomac80211: fix FT roaming
Johannes Berg [Fri, 11 Jan 2013 13:34:25 +0000 (14:34 +0100)]
mac80211: fix FT roaming

During FT roaming, wpa_supplicant attempts to set the
key before association. This used to be rejected, but
as a side effect of my commit 66e67e418908442389d3a9e
("mac80211: redesign auth/assoc") the key was accepted
causing hardware crypto to not be used for it as the
station isn't added to the driver yet.

It would be possible to accept the key and then add it
to the driver when the station has been added. However,
this may run into issues with drivers using the state-
based station adding if they accept the key only after
association like it used to be.

For now, revert to the behaviour from before the auth
and assoc change.

Cc: stable@vger.kernel.org
Reported-by: Cédric Debarge <cedric.debarge@acksys.fr>
Tested-by: Cédric Debarge <cedric.debarge@acksys.fr>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: improve the reports in TX path
Emmanuel Grumbach [Wed, 2 Jan 2013 10:12:25 +0000 (12:12 +0200)]
iwlwifi: improve the reports in TX path

Also when things go wrong (queues don't get emtpy), try to
get some data from the HW.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: audit single frames from AGG queue in RS
Emmanuel Grumbach [Wed, 9 Jan 2013 08:20:36 +0000 (10:20 +0200)]
iwlwifi: audit single frames from AGG queue in RS

The rate scaling won't treat the information in a frame
with IEEE80211_TX_CTL_AMPDU set if IEEE80211_TX_STAT_AMPDU
is cleared. But all the frames coming from an AGG tx queue
have IEEE80211_TX_CTL_AMPDU set, and IEEE80211_TX_STAT_AMPDU
is set only if the frame was sent in an AMPDU.
This means that all the data in frames in AGG tx queues that
aren't sent as an AMPDU is thrown away.
This is even more harmful when in bad link conditions, the
frames are sent in an AMPDU and then finally sent as single
frame. So a lot of failures weren't reported and the rate
scaling got stuck in high rates leading to very poor
connectivity.

Fix that by clearing IEEE80211_TX_CTL_AMPDU when the frame
isn't part of an AMPDU.

This bug was introduced by

2eb81a40aa521035ff9c8c8309e482dff523f8c9
iwlwifi: don't clear CTL_AMPDU on frame status

This fix basically reverts the aforementioned commit.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: enable communication with WoWLAN firmware
Johannes Berg [Tue, 8 Jan 2013 10:25:44 +0000 (11:25 +0100)]
iwlwifi: enable communication with WoWLAN firmware

On resuming, the opmode may have to be able to talk
to the WoWLAN/D3 firmware in order to query it about
its status and wakeup reasons. To do that, the opmode
has to call the new d3_resume() transport API which
will set up the device for command communcation.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: simplify scheduler memory clearing
Johannes Berg [Tue, 8 Jan 2013 23:47:07 +0000 (00:47 +0100)]
iwlwifi: simplify scheduler memory clearing

Writing 130 dwords into the device one by one is
rather inefficient, every one needs to lock, grab
NIC access (a few register reads/writes) and then
write the address and data registers.

Use the new memory clearing function to make this
easier and faster.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: allow to zero SRAM
Emmanuel Grumbach [Wed, 9 Jan 2013 09:37:59 +0000 (11:37 +0200)]
iwlwifi: allow to zero SRAM

Sending a NULL pointer to iwl_trans_write_mem allows now
to zero SRAM.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: don't enable all interrupts on resuming
Johannes Berg [Tue, 8 Jan 2013 09:26:17 +0000 (10:26 +0100)]
iwlwifi: don't enable all interrupts on resuming

Enabling the RF-kill interrupt is sufficient for getting
RF-kill notifications, and no other interrupt is needed
as the device isn't functional when suspended and will be
restarted/reconfigured when mac80211 resumes it later.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agoiwlwifi: don't memset scalar values
Johannes Berg [Mon, 7 Jan 2013 23:25:21 +0000 (00:25 +0100)]
iwlwifi: don't memset scalar values

The dma_addr_t type is a scalar value, so it should
just be assigned, not memset.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
11 years agobcma: use consistent case for 'hex' constants
Nathan Hintz [Sat, 12 Jan 2013 10:46:17 +0000 (02:46 -0800)]
bcma: use consistent case for 'hex' constants

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: add support for 1 and 2 byte extended config space access
Nathan Hintz [Sat, 12 Jan 2013 10:46:16 +0000 (02:46 -0800)]
bcma: add support for 1 and 2 byte extended config space access

The sanity checks allow 1 and 2 byte reads/writes of the extended
PCI config space to proceed; however, the code only supports 4
byte reads/writes.  This patch adds support for 1 and 2 byte
reads/writes of the extended PCI config space.

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: don't map/unmap a subset of the PCI config space
Nathan Hintz [Sat, 12 Jan 2013 10:46:15 +0000 (02:46 -0800)]
bcma: don't map/unmap a subset of the PCI config space

For PCI config space access offsets < 256 for device '0',
bcma_extpci_write_config performs an 'ioremap_nocache' on a 4 byte
section of the PCI config space (an area that has already
previously been mapped), and then subsequently unmaps that 4 byte
section.  This can't be a good thing for future read access from
that now unmapped location.  Modify the config space writes to use
the existing access functions (similar to how it is done for the reads).

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: jump to 'out' label for invalid 'func' value
Nathan Hintz [Sat, 12 Jan 2013 10:46:14 +0000 (02:46 -0800)]
bcma: jump to 'out' label for invalid 'func' value

Consistently jump to the 'out' label for error conditions (adds
missing check for 'func' validity in bcma_extpci_write_config).

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: delete duplicate readl
Nathan Hintz [Sat, 12 Jan 2013 10:46:13 +0000 (02:46 -0800)]
bcma: delete duplicate readl

The 'val' has already been read by the prior call to 'mips_busprobe32';
this 'readl' is unnecessary.

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agobcma: fix bcm4716/bcm4748 i2s irqflag
Nathan Hintz [Sat, 12 Jan 2013 06:07:22 +0000 (22:07 -0800)]
bcma: fix bcm4716/bcm4748 i2s irqflag

The default irqflag assignment for the I2S core on some Broadcom
4716/4748 devices is invalid and needs to be corrected (from the
Broadcom SDK).

Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: ar9002_hw_spectral_scan_config() can be static
Fengguang Wu [Sat, 12 Jan 2013 02:28:12 +0000 (10:28 +0800)]
ath9k_hw: ar9002_hw_spectral_scan_config() can be static

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoMerge tag 'nfc-next-3.9-1' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo...
John W. Linville [Mon, 14 Jan 2013 20:08:52 +0000 (15:08 -0500)]
Merge tag 'nfc-next-3.9-1' of git://git./linux/kernel/git/sameo/nfc-next

Samuel Ortiz <sameo@linux.intel.com> says:

"This is the first NFC patchset targeted at the 3.9 merge window.

It brings the following goodies:

- LLCP socket timestamping (To be used e.g with the recently released nfctool
  application for a more efficient skb timestamping when sniffing).
- A pretty big pn533 rework from Waldemar, preparing the driver to support
  more flavours of pn533 based devices.
- HCI changes from Eric in preparation for the microread driver support.
- Some LLCP memory leak fixes, cleanups and slight improvements.
- pn544 and nfcwilink move to the devm_kzalloc API.
- An initial Secure Element (SE) API.
- An nfc.h license change from the original author, allowing non GPL
  application code to safely include it."

Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: disable the tasklet before taking the PCU lock
Felix Fietkau [Mon, 14 Jan 2013 15:56:46 +0000 (16:56 +0100)]
ath9k: disable the tasklet before taking the PCU lock

Fixes a reported CPU soft lockup where the tasklet tries to acquire the
lock and blocks while ath_prepare_reset (holding the lock) waits for it
to complete.

Cc: stable@vger.kernel.org
Reported-by: Robert Shade <robert.shade@gmail.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: remove sc->rx.rxbuflock to fix a deadlock
Felix Fietkau [Mon, 14 Jan 2013 09:50:15 +0000 (10:50 +0100)]
ath9k: remove sc->rx.rxbuflock to fix a deadlock

The commit "ath9k: fix rx flush handling" added a deadlock that happens
because ath_rx_tasklet is called in a section that has already taken the
rx buffer lock.

It seems that the only purpose of the rxbuflock was a band-aid fix to the
reset vs rx tasklet race, which has been properly fixed in the commit
"ath9k: add a better fix for the rx tasklet vs rx flush race".

Now that the fix is in, we can safely remove the lock to avoid such issues.

Cc: stable@vger.kernel.org
Reported-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoNFC: Change nfc.h license
Lauro Ramos Venancio [Thu, 6 Dec 2012 00:12:25 +0000 (21:12 -0300)]
NFC: Change nfc.h license

nfc.h being GPL makes it quite controversial for non GPL applications to
include it.
Moreover, nfc.h only includes structures and API definitions that are hardly
copyrightable.

Signed-off-by: Lauro Ramos Venancio <lauro.venancio@openbossa.org>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>