In some cases we may want to stop the queues of a single vif (for
instance during a channel-switch).  Add a function that stops all the
queues that are assigned to a vif.  If a queue is assigned to more
than one vif, the corresponding netdev subqueue of the other vif(s)
will also be stopped.  If the HW doesn't set the
IEEE80211_HW_QUEUE_CONTROL flag, then all queues are stopped.

Also add a corresponding function to wake the queues of a vif back.
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Sometimes different vifs may be stopping the queues for the same
reason (e.g. when several interfaces are performing a channel switch).
Instead of using a bitmask for the reasons, use an integer that holds
a refcount instead.  In order to keep it backwards compatible,
introduce a boolean in some functions that tell us whether the queue
stopping should be refcounted or not.  For now, use not refcounted for
all calls to keep it functionally the same as before.
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

There is no need to stop all queues when we want to flush specific
queues, so stop only the queues that will be flushed.
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Converting time from one format to another seems to give coders a warm
and fuzzy feeling.

Use the proper interfaces.
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: John W. Linville <linville@tuxdriver.com>
[fix compile error]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

do_posix_clock_monotonic_gettime() is a leftover from the initial
posix timer implementation which maps to ktime_get_ts().
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

vif->csa_active is protected by mutexes only. This
means it is unreliable to depend on it on codeflow
in non-sleepable beacon and CSA code. There was no
guarantee to have vif->csa_active update be
visible before beacons are updated on SMP systems.

Using csa counter offsets which are embedded in
beacon struct (and thus are protected with single
RCU assignment) is much safer.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Having csa counters part of beacon and probe_resp
structures makes it easier to get rid of possible
races between setting a beacon and updating
counters on SMP systems by guaranteeing counters
are always consistent against given beacon struct.

While at it relax WARN_ON into WARN_ON_ONCE to
prevent spamming logs and racing.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
[remove pointless array check]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Allow send frames using monitor interface
when DFS chandef and we pass CAC (beaconing
allowed).

This fix problem when old kernel and new backports used,
in such case hostapd create/use also monitor interface.
Before this patch all frames hostapd send using monitor
iface were dropped when AP was configured on DFS channel.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Currently, cfg80211 tries to implement ethtool, but that doesn't
really scale well, with all the different operations. Make the
lower-level driver responsible for it, which currently only has
an effect on mac80211. It will similarly not scale well at that
level though, since mac80211 also has many drivers.

To cleanly implement this in mac80211, introduce a new file and
move some code to appropriate places.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Since WEP is practically dead, there seems very little
point in keeping WEP weak IV accounting.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Signed-off-by: NAntonio Ospite <ao2@ao2.it>
Cc: Luis Carlos Cobo <luisca@cozybit.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: linux-wireless@vger.kernel.org
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

The mesh_plink code is doing some interesting things with the
ignore_plink_timer flag.  It seems the original intent was to
handle this race:

cpu 0                           cpu 1
-----                           -----
                                start timer handler for state X
acquire sta_lock
change state from X to Y
mod_timer() / del_timer()
release sta_lock
                                acquire sta_lock
                                execute state Y timer too soon

However, using the mod_timer()/del_timer() return values to
detect these cases is broken.  As a result, timers get ignored
unnecessarily, and stations can get stuck in the peering state
machine.

Instead, we can detect the case by looking at the timer expiration.
In the case of del_timer, just ignore the timers in the following
(LISTEN/ESTAB) states since they won't have timers anyway.
Signed-off-by: NBob Copeland <me@bobcopeland.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It is currently possible to have a race due to the station PS
unblock work like this:
 * station goes to sleep with frames buffered in the driver
 * driver blocks wakeup
 * station wakes up again
 * driver flushes/returns frames, and unblocks, which schedules
   the unblock work
 * unblock work starts to run, and checks that the station is
   awake (i.e. that the WLAN_STA_PS_STA flag isn't set)
 * we process a received frame with PM=1, setting the flag again
 * ieee80211_sta_ps_deliver_wakeup() runs, delivering all frames
   to the driver, and then clearing the WLAN_STA_PS_DRIVER and
   WLAN_STA_PS_STA flags

In this scenario, mac80211 will think that the station is awake,
while it really is asleep, and any TX'ed frames should be filtered
by the device (it will know that the station is sleeping) but then
passed to mac80211 again, which will not buffer it either as it
thinks the station is awake, and eventually the packets will be
dropped.

Fix this by moving the clearing of the flags to exactly where we
learn about the situation. This creates a problem of reordering,
so introduce another flag indicating that delivery is being done,
this new flag also queues frames and is cleared only while the
spinlock is held (which the queuing code also holds) so that any
concurrent delivery/TX is handled correctly.
Reported-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Minstrel has long since proven its worth.
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

unregister_netdevice_many() API is error prone and we had too
many bugs because of dangling LIST_HEAD on stacks.

See commit f87e6f47 ("net: dont leave active on stack LIST_HEAD")

In fact, instead of making sure no caller leaves an active list_head,
just force a list_del() in the callee. No one seems to need to access
the list after unregister_netdevice_many()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Mark new interfaces as idle to allow operations that require that
interfaces are idle to take place. Interface types that are always
not idle (like AP interfaces) will be set as not idle when they are
assigned a channel context.
Signed-off-by: NAvraham Stern <avraham.stern@intel.com>
Signed-off-by: Emmanuel Grumbach<emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

During strong signal fluctuations under high throughput, few consecutive
failed A-MPDU transmissions can easily trigger packet loss notification,
and thus (in AP mode) client disconnection.

Reduce the number of false positives by checking the A-MPDU status flag
and treating a failed A-MPDU as a single packet.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Doing so will lead to an oops for a p2p-dev interface, since it has
no netdev.

Cc: stable@vger.kernel.org
Signed-off-by: NArik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

If the rate control algorithm uses a selection table, it
is leaked when the station is destroyed - fix that.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Reported-by: NChristophe Prévotaux <cprevotaux@nltinc.com>
Fixes: 0d528d85 ("mac80211: improve the rate control API")
Cc: stable@vger.kernel.org # v3.10+
[add commit log entry, remove pointless NULL check]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Without this fix, freshly rebooted Linux creates a new IBSS
instead of joining an existing one. Only when jiffies counter
overflows after 5 minutes the IBSS can be successfully joined.
Signed-off-by: NKrzysztof Hałasa <khalasa@piap.pl>
[edit commit message slightly]
Cc: stable@vger.kernel.org
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Since the commit below, cfg80211_chandef_dfs_required()
will warn if it gets a an NL80211_IFTYPE_UNSPECIFIED iftype
as explicitely written in the commit log.
When an virtual monitor interface is added, its type is set
in ieee80211_sub_if_data.vif.type, but not in
ieee80211_sub_if_data.wdev.iftype which is passed to
cfg80211_chandef_dfs_required() hence resulting in the
following warning:

WARNING: CPU: 1 PID: 21265 at net/wireless/chan.c:376 cfg80211_chandef_dfs_required+0xbc/0x130 [cfg80211]()
Modules linked in: [...]
CPU: 1 PID: 21265 Comm: ifconfig Tainted: G        W  O 3.13.11+ #12
Hardware name: Dell Inc. Latitude E6410/0667CC, BIOS A01 03/05/2010
 0000000000000009 ffff88008f5fdb08 ffffffff817d4219 ffff88008f5fdb50
 ffff88008f5fdb40 ffffffff8106f57d 0000000000000000 0000000000000000
 ffff880081062fb8 ffff8800810604e0 0000000000000001 ffff88008f5fdba0
Call Trace:
 [<ffffffff817d4219>] dump_stack+0x4d/0x66
 [<ffffffff8106f57d>] warn_slowpath_common+0x7d/0xa0
 [<ffffffff8106f5ec>] warn_slowpath_fmt+0x4c/0x50
 [<ffffffffa04ea4ec>] cfg80211_chandef_dfs_required+0xbc/0x130 [cfg80211]
 [<ffffffffa06b1024>] ieee80211_vif_use_channel+0x94/0x500 [mac80211]
 [<ffffffffa0684e6b>] ieee80211_add_virtual_monitor+0x1ab/0x5c0 [mac80211]
 [<ffffffffa0686ae5>] ieee80211_do_open+0xe75/0x1580 [mac80211]
 [<ffffffffa0687259>] ieee80211_open+0x69/0x70 [mac80211]
[snip]

Fixes: 00ec75fc ("cfg80211: pass the actual iftype when calling cfg80211_chandef_dfs_required()")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Acked-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In some cases, when the driver is already using all the channel
contexts it can handle at once, we have to do an in-place switch
(ie. we cannot afford using an extra context temporarily for the
transaction).  But some drivers may not support switching the channel
context assigned to a vif on the fly (ie. without unassigning and
assigning it) while others may only work if the context is changed on
the fly, without unassigning it first.

To allow these different scenarios, add a new driver operation that
let's the driver decide how to handle an in-place switch.
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Add get_expected_throughput() API to mac80211 so that each
driver can implement its own version based on the RC
algorithm they are using (might be using an HW RC algo).
The API returns a value expressed in Kbps.

Also, add the new get_expected_throughput() member
to the rate_control_ops structure in order to be
able to query the RC algorithm (this patch provides an
implementation of this API for both minstrel and
minstrel_ht).

The related member in the station_info object is now
filled accordingly when dumping a station.

Cc: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: NAntonio Quartulli <antonio@open-mesh.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

This propagates through all the drivers and mac80211.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Antonio and I were looking over this code and some things
didn't immediately make sense, so we came up with two small
clarifications.
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Make the beacon CSA counters part of ieee80211_mutable_offsets and don't
decrement CSA counters when generating a beacon template. This permits the
driver to offload the CSA counters handling. Since mac80211 updates the probe
responses with the correct counter, the driver should sync the counter's value
with mac80211 using ieee80211_csa_update_counter function.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Add a new API ieee80211_beacon_get_template, which doesn't
affect DTIM counter and should be used if the device generates beacon
frames, and new beacon template is needed. In addition set the offsets
to TIM IE for MESH interface.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Support up to IEEE80211_MAX_CSA_COUNTERS_NUM csa counters.
This is defined to be 2 now, to support both CSA and eCSA
counters.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Change the type of NL80211_ATTR_CSA_C_OFF_BEACON and
NL80211_ATTR_CSA_C_OFF_PRESP to be NLA_BINARY which allows
userspace to use beacons and probe responses with
multiple CSA counters.
This isn't breaking the API since userspace can
continue to use nla_put_u16 for this attributes, which
is equivalent to a single element u16 array.
In addition advertise max number of supported CSA counters.
This is needed when using CSA and eCSA IEs together.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Track current csa counter value and use it
to update mgmt frames at the provided offsets.
Signed-off-by: NAndrei Otcheretianski <andrei.otcheretianski@intel.com>
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Jouni reported that if a remain-on-channel was active on the
same channel as the current operating channel, then the ROC
would start, but any frames transmitted using mgmt-tx on the
same channel would get delayed until after the ROC.

The reason for this is that the ROC starts, but doesn't have
any handling for "remain on the same channel", so it stops
the interface queues. The later mgmt-tx then puts the frame
on the interface queues (since it's on the current operating
channel) and thus they get delayed until after the ROC.

To fix this, add some logic to handle remaining on the same
channel specially and not stop the queues etc. in this case.
This not only fixes the bug but also improves behaviour in
this case as data frames etc. can continue to flow.

Cc: stable@vger.kernel.org
Reported-by: NJouni Malinen <j@w1.fi>
Tested-by: NJouni Malinen <j@w1.fi>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

If the association is in progress while we suspend, the
stack will be in a messed up state. Clean it before we
suspend.

This patch completes Johannes's patch:

1a1cb744
Author: Johannes Berg <johannes.berg@intel.com>

    mac80211: fix suspend vs. authentication race

Cc: <stable@vger.kernel.org>
Fixes: 12e7f517 ("mac80211: cleanup generic suspend/resume procedures")
Signed-off-by: NEmmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

If sdata doesn't have a valid dev (e.g. in case of monitor
vif), the vif_name field was initialized with (a length of)
some short string, but later was set to a different,
potentially larger one.

This resulted in out-of-bounds write, which usually
appeared as garbage in the trace log.

Simply trace sdata->name, as it should always have the
correct name for both cases.
Signed-off-by: NEliad Peller <eliadx.peller@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

When the driver fails during HW restart or resume, the whole
stack goes into a very confused state with interfaces being
up while the hardware is down etc.

Address this by shutting down everything; we'll run into a
lot of warnings in the process but that's better than having
the whole stack get messed up.
Reviewed-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

There are two (related) issues with this.

One case, reported by Michal, is related to hostap: it unsets the
20/40 capability bit for stations that associate when it's in 20
MHz mode.

The other case, reported by Eyal, is that some APs like Netgear
R6300v2 and probably others based on the BCM4360 chipset can be
configured for doing VHT at 20Mhz. In this case the beacon has
a VHT IE but the HT cap indicates transmitter only support 20Mhz.

In both of these cases, we currently avoid VHT and use only HT
this means we can't use the highest rates (MCS8), so fixing this
leads to throughput improvements.
Reported-by: NMichal Kazior <michal.kazior@tieto.com>
Reported-by: NEyal Shapira <eyal@wizery.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Commit 59af6928 (mac80211: fix CSA tx queue stopping) introduced a
sparse warning:

net/mac80211/cfg.c:3274:5: warning: symbol '__ieee80211_channel_switch' was not declared. Should it be static?

Fix it by declaring the function static.
Signed-off-by: NLuciano Coelho <luciano.coelho@intel.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It doesn't make much sense to leave a crippled
interface running.

As a side effect this will unblock tx queues with
CSA reason immediately after failure instead of
until after userspace requests interface to stop.

This also gives userspace an opportunity to
indirectly see CSA failure.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
[small code cleanup]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It is not guaranteed that multi-vif channel
switching is tightly synchronized. It makes sense
to ignore cqm (missing beacons, et al) while csa
is progressing and re-check it after it completes.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

Improves readability and modularity.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

It was possible for tx queues to be stuck stopped
if AP CSA finalization failed. In that case
neither stop_ap nor do_stop woke the queues up.
This means it was impossible to perform tx at all
until driver was reloaded or a successful CSA was
performed later.

It was possible to solve this in a simpler manner
however this is more robust and future proof
(having multi-vif CSA in mind).

New sdata->csa_block_tx is introduced to keep
track of which interfaces requested tx to be
blocked for CSA. This is required because mac80211
stops all tx queues for that purpose. This means
queues must be awoken only when last tx-blocking
CSA interface is finished.

It is still possible to have tx queues stopped
after CSA failure but as soon as offending
interfaces are stopped from userspace (stop_ap or
ifdown) tx queues are woken up properly.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>