GitHub/LineageOS/G12/android_kernel_amlogic_linux-4.9.git
9 years agonet: rfkill: Add Broadcom BCM2E40 bluetooth ACPI ID
Mika Westerberg [Thu, 29 Jan 2015 16:26:00 +0000 (18:26 +0200)]
net: rfkill: Add Broadcom BCM2E40 bluetooth ACPI ID

This is yet another Broadcom bluetooth chip with ACPI ID BCM2E40.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: don't document per-wiphy interface dump
Johannes Berg [Thu, 29 Jan 2015 15:52:16 +0000 (16:52 +0100)]
nl80211: don't document per-wiphy interface dump

Such a feature doesn't exist and isn't really needed since you
probably won't have enough interfaces to make it worthwhile, so
just remove that from the documentation.

Reported-by: booto [on IRC]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agowireless: docs: fix 'make pdfdocs' failure
Sharon Dvir [Tue, 27 Jan 2015 21:44:20 +0000 (21:44 +0000)]
wireless: docs: fix 'make pdfdocs' failure

'make pdfdocs' failed for 80211.tmpl and reported errors of the form:
'document type does not allow element "refentry" here'
Fix this by moving </section> tags to a location that makes sense,
this fixes the errors and the pdf looks OK in relevant areas.

Signed-off-by: Sharon Dvir <sharon.dvir1@mail.huji.ac.il>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80111: Add BIP-GMAC-128 and BIP-GMAC-256 ciphers
Jouni Malinen [Sat, 24 Jan 2015 17:52:09 +0000 (19:52 +0200)]
mac80111: Add BIP-GMAC-128 and BIP-GMAC-256 ciphers

This allows mac80211 to configure BIP-GMAC-128 and BIP-GMAC-256 to the
driver and also use software-implementation within mac80211 when the
driver does not support this with hardware accelaration.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80111: Add BIP-CMAC-256 cipher
Jouni Malinen [Sat, 24 Jan 2015 17:52:08 +0000 (19:52 +0200)]
mac80111: Add BIP-CMAC-256 cipher

This allows mac80211 to configure BIP-CMAC-256 to the driver and also
use software-implementation within mac80211 when the driver does not
support this with hardware accelaration.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80111: Add CCMP-256 cipher
Jouni Malinen [Sat, 24 Jan 2015 17:52:07 +0000 (19:52 +0200)]
mac80111: Add CCMP-256 cipher

This allows mac80211 to configure CCMP-256 to the driver and also use
software-implementation within mac80211 when the driver does not support
this with hardware accelaration.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
[squash ccmp256 -> mic_len argument change]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80111: Add GCMP and GCMP-256 ciphers
Jouni Malinen [Sat, 24 Jan 2015 17:52:06 +0000 (19:52 +0200)]
mac80111: Add GCMP and GCMP-256 ciphers

This allows mac80211 to configure GCMP and GCMP-256 to the driver and
also use software-implementation within mac80211 when the driver does
not support this with hardware accelaration.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
[remove a spurious newline]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: Add new GCMP, CCMP-256, BIP-GMAC, BIP-CMAC-256 ciphers
Jouni Malinen [Sat, 24 Jan 2015 17:52:05 +0000 (19:52 +0200)]
cfg80211: Add new GCMP, CCMP-256, BIP-GMAC, BIP-CMAC-256 ciphers

This makes cfg80211 aware of the GCMP, GCMP-256, CCMP-256, BIP-GMAC-128,
BIP-GMAC-256, and BIP-CMAC-256 cipher suites. These new cipher suites
were defined in IEEE Std 802.11ac-2013.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: Fix BIP (AES-CMAC) cipher validation
Jouni Malinen [Sat, 24 Jan 2015 17:52:04 +0000 (19:52 +0200)]
cfg80211: Fix BIP (AES-CMAC) cipher validation

This cipher can be used only as a group management frame cipher and as
such, there is no point in validating that it is not used with non-zero
key-index. Instead, verify that it is not used as a pairwise cipher
regardless of the key index.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
[change code to use switch statement which is easier to extend]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoath10k: use IEEE80211_HW_SW_CRYPTO_CONTROL
Johannes Berg [Thu, 22 Jan 2015 20:38:45 +0000 (21:38 +0100)]
ath10k: use IEEE80211_HW_SW_CRYPTO_CONTROL

The ath10k driver only supports HW crypto (except for management
frames or so - CMAC needs to be done in software). Make it use
the new IEEE80211_HW_SW_CRYPTO_CONTROL flag to ensure mac80211
doesn't erroneously attempt to use software crypto.

Taking through my tree after Kalle's ACK to avoid breaking the
ath10k driver with the next crypto patches.

Acked-by: Kalle Valo <kvalo@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: handle potential race between suspend and scan completion
Luciano Coelho [Sat, 24 Jan 2015 08:30:02 +0000 (10:30 +0200)]
mac80211: handle potential race between suspend and scan completion

If suspend starts while ieee80211_scan_completed() is running, between
the point where SCAN_COMPLETED is set and the work is queued,
ieee80211_scan_cancel() will not catch the work and we may finish
suspending before the work is actually executed, leaving the scan
running while suspended.

To fix this race, queue the scan work during resume if the
SCAN_COMPLETED flag is set and flush it immediately.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: support beacon statistics
Johannes Berg [Wed, 21 Jan 2015 20:09:02 +0000 (21:09 +0100)]
mac80211: support beacon statistics

For drivers without beacon filtering, support beacon statistics
entirely, i.e. report the number of beacons and average signal.

For drivers with beacon filtering, give them the number of beacons
received by mac80211 -- in case the device reports only the number
of filtered beacons then driver doesn't have to count all beacons
again as mac80211 already does.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: fix per-TID RX-MSDU counter
Johannes Berg [Fri, 23 Jan 2015 12:23:48 +0000 (13:23 +0100)]
mac80211: fix per-TID RX-MSDU counter

In the case of non-QoS association, the counter was actually
wrong. The right index isn't security_idx but seqno_idx, as
security_idx will be 0 for data frames, while 16 is needed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: tdls: disentangle HT supported conditions
Johannes Berg [Fri, 23 Jan 2015 10:42:14 +0000 (11:42 +0100)]
mac80211: tdls: disentangle HT supported conditions

These conditions are rather difficult to follow, for example
because "!sta" only exists to not crash in the case that we
don't have a station pointer (WLAN_TDLS_SETUP_REQUEST) in
which the additional condition (peer supports HT) doesn't
actually matter anyway.

Cleaning this up only duplicates two lines of code but makes
the rest far easier to read, so do that.

As a side effect, smatch stops complaining about the lack of
a sta pointer test after the !sta (since the !sta goes away)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: tdls: remove shadowing variable
Johannes Berg [Fri, 23 Jan 2015 10:36:54 +0000 (11:36 +0100)]
mac80211: tdls: remove shadowing variable

There's no need to use another local 'sta' variable as the
original (outer scope) one isn't needed any more and has
become invalid anyway when exiting the RCU read section.

Remove the inner scope one and along with it the useless NULL
initialization.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: suppress smatch warnings
Johannes Berg [Fri, 23 Jan 2015 10:25:20 +0000 (11:25 +0100)]
nl80211: suppress smatch warnings

smatch warns that we once checked request->ssids in two functions
and then unconditionally used it later again.

This is actually fine, because the code has a relationship between
attrs[NL80211_ATTR_SCAN_SSIDS], n_ssids and request->ssids, but
smatch isn't smart enough to realize that.

Suppress the warnings by always checking just n_ssids - that way
smatch won't know that request->ssids could be NULL, and since it
is only NULL when n_ssids is 0 we still check everything correctly.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoRevert "mac80211: keep sending peer candidate events while in listen state"
Bob Copeland [Thu, 22 Jan 2015 21:49:53 +0000 (16:49 -0500)]
Revert "mac80211: keep sending peer candidate events while in listen state"

This reverts commit 2ae70efcea7a695a62bb47170d3fb16674b8dbea.

The new peer events that are generated by the change are causing problems
with wpa_supplicant in userspace: wpa_s tries to restart SAE authentication
with the peer when receiving the event, even though authentication may be in
progress already, and it gets very confused.

Revert back to the original operating mode, which is to only get events when
there is no corresponding station entry.

Cc: Nishikawa, Kenzoh <Kenzoh.Nishikawa@jp.sony.com>
Cc: Masashi Honma <masashi.honma@gmail.com>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: complete scan work immediately if quiesced or suspended
Luciano Coelho [Thu, 22 Jan 2015 21:34:10 +0000 (23:34 +0200)]
mac80211: complete scan work immediately if quiesced or suspended

It is possible that a deferred scan is queued after the queues are
flushed in __ieee80211_suspend().  The deferred scan work may be
scheduled by ROC or ieee80211_stop_poll().

To make sure don't start a new scan while suspending, check whether
we're quiescing or suspended and complete the scan immediately if
that's the case.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: avoid races related to suspend flow
Emmanuel Grumbach [Thu, 22 Jan 2015 21:32:46 +0000 (23:32 +0200)]
mac80211: avoid races related to suspend flow

When we go to suspend, there is complex set of states that
avoids races. The quiescing variable is set whlie
__ieee80211_suspend is running. Then suspended is set.
The code makes sure there is no window without any of these
flags.

The problem is that workers can still be enqueued while we
are quiescing. This leads to situations where the driver is
already suspending and other flows like disassociation are
handled by a worker.

To fix this, we need to check quiescing and suspended flags
in the worker itself and not only before enqueueing it.
I also add here extensive documentation to ease the
understanding of these complex issues.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: synchronize_net() before flushing the queues
Emmanuel Grumbach [Thu, 22 Jan 2015 21:30:19 +0000 (23:30 +0200)]
mac80211: synchronize_net() before flushing the queues

When mac80211 disconnects, it drops all the packets on the
queues. This happens after the net stack has been notified
that we have no link anymore (netif_carrier_off).
netif_carrier_off ensures that no new packets are sent to
xmit() callback, but we might have older packets in the
middle of the Tx path. These packets will land in the
driver's queues after the latter have been flushed.
Synchronize_net() between netif_carrier_off and drv_flush()
will fix this.

Note that we can't call synchronize_net inside
ieee80211_flush_queues since there are flows that call
ieee80211_flush_queues and don't need synchronize_net()
which is an expensive operation.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
[reword comment to be more accurate]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: add an attribute to allow delaying the first scheduled scan cycle
Luciano Coelho [Fri, 16 Jan 2015 14:04:09 +0000 (16:04 +0200)]
nl80211: add an attribute to allow delaying the first scheduled scan cycle

The userspace may want to delay the the first scheduled scan or
net-detect cycle.  Add an optional attribute to the scheduled scan
configuration to pass the delay to be (optionally) used by the driver.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
[add the attribute to the policy to validate it]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: enable TPC through mac80211 stack
Lorenzo Bianconi [Wed, 14 Jan 2015 11:55:08 +0000 (12:55 +0100)]
mac80211: enable TPC through mac80211 stack

Control per packet Transmit Power Control (TPC) in lower drivers
according to TX power settings configured by the user. In particular TPC is
enabled if value passed in enum nl80211_tx_power_setting is
NL80211_TX_POWER_LIMITED (allow using less than specified from userspace),
whereas TPC is disabled if nl80211_tx_power_setting is set to
NL80211_TX_POWER_FIXED (use value configured from userspace)

Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi83@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: Allow set network namespace by fd
Vadim Kochan [Mon, 12 Jan 2015 14:34:05 +0000 (16:34 +0200)]
nl80211: Allow set network namespace by fd

Added new NL80211_ATTR_NETNS_FD which allows to
set namespace via nl80211 by fd.

Signed-off-by: Vadim Kochan <vadim4j@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: allow drivers to control software crypto
Johannes Berg [Thu, 22 Jan 2015 17:44:19 +0000 (18:44 +0100)]
mac80211: allow drivers to control software crypto

Some drivers unfortunately cannot support software crypto, but
mac80211 currently assumes that they do.

This has the issue that if the hardware enabling fails for some
reason, the software fallback is used, which won't work. This
clearly isn't desirable, the error should be reported and the
key setting refused.

Support this in mac80211 by allowing drivers to set a new HW
flag IEEE80211_HW_SW_CRYPTO_CONTROL, in which case mac80211 will
only allow software fallback if the set_key() method returns 1.
The driver will also need to advertise supported cipher suites
so that mac80211 doesn't advertise any (future) software ciphers
that the driver can't actually do.

While at it, to make it easier to support this, refactor the
ieee80211_init_cipher_suites() code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: fix HW registration error paths
Johannes Berg [Wed, 21 Jan 2015 12:52:24 +0000 (13:52 +0100)]
mac80211: fix HW registration error paths

Station info state is started in allocation, so should be
destroyed on free (it's just a timer); rate control must
be freed if anything afterwards fails to initialize.

LED exit should be later, no need for locking there, but
it needs to be done also when rate init failed.

Also clean up the code by moving a label so the locking
doesn't have to be done separately.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: delete the assoc/auth timer upon suspend
Emmanuel Grumbach [Mon, 19 Jan 2015 14:53:06 +0000 (16:53 +0200)]
mac80211: delete the assoc/auth timer upon suspend

While suspending, we destroy the authentication /
association that might be taking place. While doing so, we
forgot to delete the timer which can be firing after
local->suspended is already set, producing the warning below.

Fix that by deleting the timer.

[66722.825487] WARNING: CPU: 2 PID: 5612 at net/mac80211/util.c:755 ieee80211_can_queue_work.isra.18+0x32/0x40 [mac80211]()
[66722.825487] queueing ieee80211 work while going to suspend
[66722.825529] CPU: 2 PID: 5612 Comm: kworker/u16:69 Tainted: G        W  O  3.16.1+ #24
[66722.825537] Workqueue: events_unbound async_run_entry_fn
[66722.825545] Call Trace:
[66722.825552]  <IRQ>  [<ffffffff817edbb2>] dump_stack+0x4d/0x66
[66722.825556]  [<ffffffff81075cad>] warn_slowpath_common+0x7d/0xa0
[66722.825572]  [<ffffffffa06b5b90>] ? ieee80211_sta_bcn_mon_timer+0x50/0x50 [mac80211]
[66722.825573]  [<ffffffff81075d1c>] warn_slowpath_fmt+0x4c/0x50
[66722.825586]  [<ffffffffa06977a2>] ieee80211_can_queue_work.isra.18+0x32/0x40 [mac80211]
[66722.825598]  [<ffffffffa06977d5>] ieee80211_queue_work+0x25/0x50 [mac80211]
[66722.825611]  [<ffffffffa06b5bac>] ieee80211_sta_timer+0x1c/0x20 [mac80211]
[66722.825614]  [<ffffffff8108655a>] call_timer_fn+0x8a/0x300

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoRevert "wireless: Support of IFLA_INFO_KIND rtnl attribute"
Johannes Berg [Mon, 19 Jan 2015 17:49:50 +0000 (18:49 +0100)]
Revert "wireless: Support of IFLA_INFO_KIND rtnl attribute"

This reverts commit ba1debdfed974f25aa598c283567878657b292ee.

Oliver reported that it breaks network-manager, for some reason with
this patch NM decides that the device isn't wireless but "generic"
(ethernet), sees no carrier (as expected with wifi) and fails to do
anything else with it.

Revert this to unbreak userspace.

Reported-by: Oliver Hartkopp <socketcan@hartkopp.net>
Tested-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: fix checking nl80211_send_station() return value
Johannes Berg [Fri, 16 Jan 2015 20:05:02 +0000 (21:05 +0100)]
cfg80211: fix checking nl80211_send_station() return value

The return value from nl80211_send_station() is the length of the
skb, or a negative error, so abort sending the message only when
the return value was negative.

This fixes the ibss_rsn wpa_supplicant test case.

Reported-by: Jouni Malinen <j@w1.fi>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove doubled semicolon
Johannes Berg [Fri, 16 Jan 2015 12:27:56 +0000 (13:27 +0100)]
mac80211: remove doubled semicolon

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: change bandwidth reporting to explicit field
Johannes Berg [Thu, 15 Jan 2015 15:14:02 +0000 (16:14 +0100)]
cfg80211: change bandwidth reporting to explicit field

For some reason, we made the bandwidth separate flags, which
is rather confusing - a single rate cannot have different
bandwidths at the same time.

Change this to no longer be flags but use a separate field
for the bandwidth ('bw') instead.

While at it, add support for 5 and 10 MHz rates - these are
reported as regular legacy rates with their real bitrate,
but tagged as 5/10 now to make it easier to distinguish them.

In the nl80211 API, the flags are preserved, but the code
now can also clearly only set a single one of the flags.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: remove 80+80 MHz rate reporting
Johannes Berg [Thu, 15 Jan 2015 15:05:21 +0000 (16:05 +0100)]
cfg80211: remove 80+80 MHz rate reporting

These rates are treated the same as 160 MHz in the spec, so
it makes no sense to distinguish them. As no driver uses them
yet, this is also not a problem, just remove them.

In the userspace API the field remains reserved to preserve
API and ABI.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove 80+80 MHz rate reporting
Johannes Berg [Thu, 15 Jan 2015 15:02:46 +0000 (16:02 +0100)]
mac80211: remove 80+80 MHz rate reporting

These rates are treated the same as 160 MHz in the spec,
so it makes no sense to distinguish them. As no driver
uses them yet, this is also not a problem, just remove
them.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: docs: remove station_info_flags
Johannes Berg [Wed, 14 Jan 2015 12:57:36 +0000 (13:57 +0100)]
cfg80211: docs: remove station_info_flags

I removed the enum earlier, but forgot to remove it from the
documentation - do that now.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: send netdetect configuration info in NL80211_CMD_GET_WOWLAN
Luciano Coelho [Fri, 9 Jan 2015 12:06:37 +0000 (14:06 +0200)]
nl80211: send netdetect configuration info in NL80211_CMD_GET_WOWLAN

Send the netdetect configuration information in the response to
NL8021_CMD_GET_WOWLAN commands.  This includes the scan interval,
SSIDs to match and frequencies to scan.

Additionally, add the NL80211_WOWLAN_TRIG_NET_DETECT with
NL80211_ATTR_WOWLAN_TRIGGERS_SUPPORTED.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: avoid reg-hints in self-managed only systems
Arik Nemtsov [Wed, 7 Jan 2015 14:47:20 +0000 (16:47 +0200)]
cfg80211: avoid reg-hints in self-managed only systems

When a system contains only self-managed regulatory devices all hints
from the regulatory core are ignored. Stop hint processing early in this
case. These systems usually don't have CRDA deployed, which results in
endless (irrelevent) logs of the form:
cfg80211: Calling CRDA to update world regulatory domain

Make sure there's at least one self-managed device before discarding a
hint, in order to prevent initial hints from disappearing on CRDA
managed systems.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: introduce sync regdom set API for self-managed
Arik Nemtsov [Wed, 7 Jan 2015 14:47:19 +0000 (16:47 +0200)]
cfg80211: introduce sync regdom set API for self-managed

A self-managed device will sometimes need to set its regdomain synchronously.
Notably it should be set before usermode has a chance to query it. Expose
a new API to accomplish this which requires the RTNL.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: don't defer scans in case of radar detection
Eliad Peller [Wed, 7 Jan 2015 15:50:11 +0000 (17:50 +0200)]
mac80211: don't defer scans in case of radar detection

Radar detection can last indefinite time. There is no
point in deferring a scan request in this case - simply
return -EBUSY.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: consider only relevant vifs for radar_required calculation
Eliad Peller [Wed, 7 Jan 2015 15:50:10 +0000 (17:50 +0200)]
mac80211: consider only relevant vifs for radar_required calculation

ctx->conf.radar_enabled should reflect whether radar
detection is enabled for the channel context.

When calculating it, make it consider only the vifs
that have this context assigned (instead of all the
vifs).

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove local->radar_detect_enabled
Eliad Peller [Wed, 7 Jan 2015 15:50:09 +0000 (17:50 +0200)]
mac80211: remove local->radar_detect_enabled

local->radar_detect_enabled should tell whether
radar_detect is enabled on any interface belonging
to local.

However, it's not getting updated correctly
in many cases (actually, when testing with hwsim
it's never been set, even when the dfs master
is beaconing).

Instead of handling all the corner cases
(e.g. channel switch), simply check whether
radar detection is enabled only when needed,
instead of caching the result.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: add TDLS supported channels correctly
Arik Nemtsov [Wed, 7 Jan 2015 14:45:07 +0000 (16:45 +0200)]
mac80211: add TDLS supported channels correctly

The function adding the supported channels IE during a TDLS connection had
several issues:
1. If the entire subband is usable, the function exitted the loop without
   adding it
2. The function only checked chandef_usable, ignoring flags like RADAR
   which would prevent TDLS off-channel communcation.
3. HT20 was explicitly required in the chandef, while not a requirement
   for TDLS off-channel.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: let flush() drop packets when possible
Emmanuel Grumbach [Wed, 7 Jan 2015 13:42:39 +0000 (15:42 +0200)]
mac80211: let flush() drop packets when possible

When roaming / suspending, it makes no sense to wait until
the transmit queues of the device are empty. In extreme
condition they can be starved (VO saturating the air), but
even in regular cases, it is pointless to delay the roaming
because the low level driver is trying to send packets to
an AP which is far away. We'd rather drop these packets and
let TCP retransmit if needed. This will allow to speed up
the roaming.

For suspend, the explanation is even more trivial.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agorfkill: document rfkill module parameters
Andrew Clausen [Fri, 9 Jan 2015 20:24:55 +0000 (20:24 +0000)]
rfkill: document rfkill module parameters

Document the rfkill module parameters default_state and
master_switch_mode.

Signed-off-by: Andrew Clausen <andrew.p.clausen@gmail.com>
[rewrite commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: fix handling TIM IE when stations disconnect
Johannes Berg [Fri, 9 Jan 2015 10:40:39 +0000 (11:40 +0100)]
mac80211: fix handling TIM IE when stations disconnect

When a station disconnects with frames still pending, we clear
the TIM bit, but too late - it's only cleared when the station
is already removed from the driver, and thus the driver can get
confused (and hwsim will loudly complain.)

Fix this by clearing the TIM bit earlier, when the station has
been unlinked but not removed from the driver yet. To do this,
refactor the TIM recalculation to in that case ignore traffic
and simply assume no pending traffic - this is correct for the
disconnected station even though the frames haven't been freed
yet at that point.

This patch isn't needed for current drivers though as they don't
check the station argument to the set_tim() operation and thus
don't really run into the possible confusion.

Reported-by: Jouni Malinen <j@w1.fi>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211_hwsim: fix PS debugfs file locking
Johannes Berg [Fri, 9 Jan 2015 10:29:50 +0000 (11:29 +0100)]
mac80211_hwsim: fix PS debugfs file locking

The functions called within the iterators must be called with
tasklets disabled, so use atomic iteration like the rest of
the code and disable tasklets around the whole operation.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: provide per-TID RX/TX MSDU counters
Johannes Berg [Fri, 21 Nov 2014 13:26:31 +0000 (14:26 +0100)]
mac80211: provide per-TID RX/TX MSDU counters

Implement the new counters cfg80211 can now advertise to userspace.
The TX code is in the sequence number handler, which is a bit odd,
but that place already knows the TID and frame type, so it was
easiest and least impact there.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: support per-TID station statistics
Johannes Berg [Fri, 19 Dec 2014 11:34:00 +0000 (12:34 +0100)]
nl80211: support per-TID station statistics

The base for the current statistics is pretty mixed up, support
exporting RX/TX statistics for MSDUs per TID. This (currently)
covers received MSDUs, transmitted MSDUs and retries/failures
thereof.

Doing it per TID for MSDUs makes more sense than say only per AC
because it's symmetric - we could export per-AC statistics for all
frames (which AC we used for transmission can be determined also
for management frames) but per TID is better and usually data
frames are really the ones we care about. Also, on RX we can't
determine the AC - but we do know the TID for any QoS MPDU we
received.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: clarify packet statistics descriptions
Johannes Berg [Fri, 21 Nov 2014 11:40:05 +0000 (12:40 +0100)]
nl80211: clarify packet statistics descriptions

The current statistics we keep aren't very clear, some are on
MPDUs and some on MSDUs/MMPDUs. Clarify the descriptions based
on the counters mac80211 keeps.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: add nl80211 beacon-only statistics
Johannes Berg [Mon, 17 Nov 2014 13:12:22 +0000 (14:12 +0100)]
cfg80211: add nl80211 beacon-only statistics

Add these two values:
 * BEACON_RX: number of beacons received from this peer
 * BEACON_SIGNAL_AVG: signal strength average for beacons only

These can then be used for Android Lollipop's statistics request.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: remove enum station_info_flags
Johannes Berg [Mon, 17 Nov 2014 13:08:11 +0000 (14:08 +0100)]
cfg80211: remove enum station_info_flags

This is really just duplicating the list of information that's
already available in the nl80211 attribute, so remove the list.
Two small changes are needed:
 * remove STATION_INFO_ASSOC_REQ_IES complete, but the length
   (assoc_req_ies_len) can be used instead
 * add NL80211_STA_INFO_RX_DROP_MISC which exists internally
   but not in nl80211 yet

This gets rid of the duplicate maintenance of the two lists.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: allow drivers to provide most station statistics
Johannes Berg [Mon, 17 Nov 2014 10:35:23 +0000 (11:35 +0100)]
mac80211: allow drivers to provide most station statistics

In many cases, drivers can filter things like beacons that will
skew statistics reported by mac80211. To get correct statistics
in these cases, call drivers to obtain statistics and let them
override all values, filling values from mac80211 if the driver
didn't provide them. Not all of them make sense for the driver
to fill, so some are still always done by mac80211.

Note that this doesn't currently allow a driver to say "I know
this value is wrong, don't report it at all", or to sum it up
with a mac80211 value (as could be useful for "dropped misc"),
that can be added if it turns out to be needed.

This also gets rid of the get_rssi() method as is can now be
implemented using sta_statistics().

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: send statistics with delete station event
Johannes Berg [Fri, 14 Nov 2014 19:32:57 +0000 (20:32 +0100)]
mac80211: send statistics with delete station event

Use the new cfg80211_del_sta_sinfo() function to send the
statistics about the deleted station with the delete event.
This lets userspace see how much traffic etc. the deleted
station used.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: allow including station info in delete event
Johannes Berg [Fri, 14 Nov 2014 16:14:00 +0000 (17:14 +0100)]
cfg80211: allow including station info in delete event

When a station is removed, its statistics may be interesting to
userspace, for example for further aggregation of statistics of
all stations that ever connected to an AP.

Introduce a new cfg80211_del_sta_sinfo() function (and make the
cfg80211_del_sta() a static inline calling it) to allow passing
a struct station_info along with this, and send the data in the
nl80211 event message.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: add scan time to survey data
Johannes Berg [Fri, 14 Nov 2014 15:44:11 +0000 (16:44 +0100)]
cfg80211: add scan time to survey data

Add the time spent scanning to the survey data so it can be
reported by drivers that collect such information.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: allow survey data to return global data
Johannes Berg [Fri, 14 Nov 2014 15:43:50 +0000 (16:43 +0100)]
cfg80211: allow survey data to return global data

Not all devices are able to report survey data (particularly
time spent for various operations) per channel. As all these
statistics already exist in survey data, allow such devices
to report them (if userspace requested it)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: remove "channel" from survey names
Johannes Berg [Fri, 14 Nov 2014 15:35:34 +0000 (16:35 +0100)]
cfg80211: remove "channel" from survey names

All of the survey data is (currently) per channel anyway,
so having the word "channel" in the name does nothing. In
the next patch I'll introduce global data to the survey,
where the word "channel" is actually confusing.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: Re-fix accounting of the tailroom-needed counter
Ido Yariv [Tue, 6 Jan 2015 13:39:02 +0000 (08:39 -0500)]
mac80211: Re-fix accounting of the tailroom-needed counter

When hw acceleration is enabled, the GENERATE_IV or PUT_IV_SPACE flags
only require headroom space. Therefore, the tailroom-needed counter can
safely be decremented for most drivers.

The older incarnation of this patch (ca34e3b5) assumed that the above
holds true for all drivers. As reported by Christopher Chavez and
researched by Christian Lamparter and Larry Finger, this isn't a valid
assumption for p54 and cw1200.

Drivers that still require tailroom for ICV/MIC even when HW encryption
is enabled can use IEEE80211_KEY_FLAG_RESERVE_TAILROOM to indicate it.

Signed-off-by: Ido Yariv <idox.yariv@intel.com>
Cc: Christopher Chavez <chrischavez@gmx.us>
Cc: Christian Lamparter <chunkeey@googlemail.com>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Solomon Peachy <pizza@shaftnet.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoMerge branch 'mac80211' into mac80211-next
Johannes Berg [Wed, 7 Jan 2015 13:39:02 +0000 (14:39 +0100)]
Merge branch 'mac80211' into mac80211-next

Merge mac80211.git to get some changes that would otherwise
cause conflicts with new changes coming here.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: skip disabled channels in VHT check
Arik Nemtsov [Thu, 1 Jan 2015 11:43:17 +0000 (13:43 +0200)]
mac80211: skip disabled channels in VHT check

The patch "40a11ca mac80211: check if channels allow 80 MHz for VHT
probe requests" considered disabled channels as VHT enabled, and
mistakenly sent out probe-requests with the VHT IE.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: define multicast group names in header
Johannes Berg [Tue, 23 Dec 2014 16:17:38 +0000 (17:17 +0100)]
nl80211: define multicast group names in header

Put the group names into the userspace API header file so that
userspace clients can use symbolic names from there instead of
hardcoding the actual names. This doesn't really change much,
but seems somewhat cleaner.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: add extensible feature flag attribute
Gautam Kumar Shukla [Tue, 23 Dec 2014 15:55:19 +0000 (16:55 +0100)]
cfg80211: add extensible feature flag attribute

With the wiphy::features flag being used up this patch adds a
new field wiphy::ext_features. Considering extensibility this
new field is declared as a byte array. This extensible flag is
exposed to user-space by NL80211_ATTR_EXT_FEATURES.

Cc: Avinash Patil <patila@marvell.com>
Signed-off-by: Gautam (Gautam Kumar) Shukla <gautams@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: document NL80211_BSS_STATUS_AUTHENTICATED isn't used
Johannes Berg [Mon, 5 Jan 2015 10:16:42 +0000 (11:16 +0100)]
nl80211: document NL80211_BSS_STATUS_AUTHENTICATED isn't used

The flag is no longer used (and hasn't been for a long time)
since trying to track authentication (and make decisions based
on state) was just causing issues all over - see commit
95de817b9034d50860319f6033ec85d25024694c.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoRevert "mac80211: Fix accounting of the tailroom-needed counter"
Johannes Berg [Mon, 5 Jan 2015 09:28:49 +0000 (10:28 +0100)]
Revert "mac80211: Fix accounting of the tailroom-needed counter"

This reverts commit ca34e3b5c808385b175650605faa29e71e91991b.

It turns out that the p54 and cw2100 drivers assume that there's
tailroom even when they don't say they really need it. However,
there's currently no way for them to explicitly say they do need
it, so for now revert this.

This fixes https://bugzilla.kernel.org/show_bug.cgi?id=90331.

Cc: stable@vger.kernel.org
Fixes: ca34e3b5c808 ("mac80211: Fix accounting of the tailroom-needed counter")
Reported-by: Christopher Chavez <chrischavez@gmx.us>
Bisected-by: Larry Finger <Larry.Finger@lwfinger.net>
Debugged-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: fix dot11MulticastTransmittedFrameCount tested address
Eliad Peller [Sun, 21 Dec 2014 13:25:28 +0000 (15:25 +0200)]
mac80211: fix dot11MulticastTransmittedFrameCount tested address

dot11MulticastTransmittedFrameCount should be updated according
to the DA, which might be different from A1. Checking A1 results
in the counter being 0 in case of station, as to-DS data frames
use A1 for the BSSID.

This behaviour is defined in state machines, specifically in the
sta_tx_dcf_3.1d(10) description of 802.11-2012.

Signed-off-by: Eliad Peller <eliad@wizery.com>
[rewrite commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211_hwsim: fix check for custom world regdom array size
Johannes Berg [Mon, 22 Dec 2014 11:51:25 +0000 (12:51 +0100)]
mac80211_hwsim: fix check for custom world regdom array size

David Binderman reports that the conditions in the first loop
are the wrong way around - checking the array contents before
the size.

Instead of leaving the empty loop there and reordering the two
checks unify it into a single loop that skips over non-matches
and exits after the first match.

Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: minstrel: reduce size of struct minstrel_rate_stats
Felix Fietkau [Wed, 17 Dec 2014 12:38:34 +0000 (13:38 +0100)]
mac80211: minstrel: reduce size of struct minstrel_rate_stats

On minstrel_ht, the size of the per-sta struct is almost 18k, making it
an order-3 allocation.
A few fields inside the per-rate statistics are bigger than they need to
be. This patch reduces the size enough to cut down the per-sta struct to
about 13k (order-2 allocation).

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonl80211: increase the max number of rules in regdomain
Arik Nemtsov [Wed, 17 Dec 2014 16:00:44 +0000 (18:00 +0200)]
nl80211: increase the max number of rules in regdomain

Some network cards (Intel) produce per-channel regdomains and rely on
cfg80211 to merge rules as needed. This hits the max rules limit and
fails.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonl80211: Stop scheduled scan if netlink client disappears
Jukka Rissanen [Mon, 15 Dec 2014 11:25:39 +0000 (13:25 +0200)]
nl80211: Stop scheduled scan if netlink client disappears

An attribute NL80211_ATTR_SOCKET_OWNER can be set by the scan initiator.
If present, the attribute will cause the scan to be stopped if the client
dies.

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonl80211: Convert sched_scan_req pointer to RCU pointer
Jukka Rissanen [Mon, 15 Dec 2014 11:25:38 +0000 (13:25 +0200)]
nl80211: Convert sched_scan_req pointer to RCU pointer

Because of possible races when accessing sched_scan_req pointer in
rdev, the sched_scan_req is converted to RCU pointer.

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: avoid reconfig if no interfaces are up
Eliad Peller [Sun, 14 Dec 2014 09:05:53 +0000 (11:05 +0200)]
mac80211: avoid reconfig if no interfaces are up

If there are no interfaces up, there is no reason
to continue the reconfig flow.

The current code might end up calling driver
callbacks (e.g. resume(), reconfig_complete())
while the driver is already stopped.

Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: remove unused variable in ieee80211_parse_ch_switch_ie()
Luciano Coelho [Sun, 14 Dec 2014 09:05:52 +0000 (11:05 +0200)]
mac80211: remove unused variable in ieee80211_parse_ch_switch_ie()

The ht_oper variable is assigned a value, but never used in
ieee80211_parse_ch_switch_ie().  Remove it.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: update sta bw on ht chanwidth action frame
Eliad Peller [Sun, 14 Dec 2014 09:05:51 +0000 (11:05 +0200)]
mac80211: update sta bw on ht chanwidth action frame

Commit e1a0c6b ("mac80211: stop toggling IEEE80211_HT_CAP_SUP_WIDTH_20_40")
mistakenly removed the actual update of sta->sta.bandwidth.

Refactor ieee80211_sta_cur_vht_bw() into multiple functions
(calculate caps-bw and chandef-bw separately, and min them
with cur_max_bandwidth).

On ht chanwidth action frame set only cur_max_bandwidth
(according to the sta capabilities) and recalc the sta bw.

Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: handle power constraint and country IEs in RRM
Moshe Benji [Sun, 14 Dec 2014 09:05:50 +0000 (11:05 +0200)]
mac80211: handle power constraint and country IEs in RRM

In beacons, handle the Country IE even if no Power Constraint IE
is present, and, capability wise, also in case that the Radio
Measurements capability is enabled.

In cases where the Country IE should be handled and that the
Power Constraint IE is not present, the Country IE alone will
set the power limit (and not both Country and Power Constraint
IEs).

Signed-off-by: Moshe Benji <moshe.benji@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: Fix ignored HT override configurations
Chaya Rachel Ivgi [Sun, 14 Dec 2014 08:38:59 +0000 (10:38 +0200)]
mac80211: Fix ignored HT override configurations

HT override configurations was ignored when choosing the channel
(until now, the override configuration affected only the
capabilities shown in the IEs).

The override configurations received only on association time,
so in this case we should determine the channel again.

Signed-off-by: Chaya Rachel Ivgi <chaya.rachel.ivgi@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: free management frame keys when removing station
Johannes Berg [Wed, 17 Dec 2014 12:55:49 +0000 (13:55 +0100)]
mac80211: free management frame keys when removing station

When writing the code to allow per-station GTKs, I neglected to
take into account the management frame keys (index 4 and 5) when
freeing the station and only added code to free the first four
data frame keys.

Fix this by iterating the array of keys over the right length.

Cc: stable@vger.kernel.org
Fixes: e31b82136d1a ("cfg80211/mac80211: allow per-station GTKs")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: avoid intersection when applying self-managed reg
Arik Nemtsov [Mon, 15 Dec 2014 17:26:02 +0000 (19:26 +0200)]
cfg80211: avoid intersection when applying self-managed reg

The custom-reg handling function can currently only add flags to a given
channel. This results in stale flags being left applied. In some cases
a channel was disabled and even the orig_flags were changed to reflect
this.

Previously the API was designed for a single invocation before wiphy
registration, so this didn't matter. The previous approach doesn't scale
well to self-managed regulatory devices, particularly when a more
permissive regdom is applied after a restrictive one.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: return private regdom for self-managed devices
Arik Nemtsov [Mon, 15 Dec 2014 17:26:01 +0000 (19:26 +0200)]
cfg80211: return private regdom for self-managed devices

If a device has self-managed regulatory, insist on returning the wiphy
specific regdomain if a wiphy-idx is specified. The global regdomain is
meaningless for such devices.

Also add an attribute for self-managed devices, so usermode can
distinguish them as such.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: allow wiphy specific regdomain management
Jonathan Doron [Mon, 15 Dec 2014 17:26:00 +0000 (19:26 +0200)]
cfg80211: allow wiphy specific regdomain management

Add a new regulatory flag that allows a driver to manage regdomain
changes/updates for its own wiphy.
A self-managed wiphys only employs regulatory information obtained from
the FW and driver and does not use other cfg80211 sources like
beacon-hints, country-code IEs and hints from other devices on the same
system. Conversely, a self-managed wiphy does not share its regulatory
hints with other devices in the system. If a system contains several
devices, one or more of which are self-managed, there might be
contradictory regulatory settings between them. Usage of flag is
generally discouraged. Only use it if the FW/driver is incompatible
with non-locally originated hints.

A new API lets the driver send a complete regdomain, to be applied on
its wiphy only.

After a wiphy-specific regdomain change takes place, usermode will get
a new type of change notification. The regulatory core also takes care
enforce regulatory restrictions, in case some interfaces are on
forbidden channels.

Signed-off-by: Jonathan Doron <jonathanx.doron@intel.com>
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: allow usermode to query wiphy specific regdom
Arik Nemtsov [Mon, 15 Dec 2014 17:25:59 +0000 (19:25 +0200)]
cfg80211: allow usermode to query wiphy specific regdom

If a wiphy-idx is specified, the kernel will return the wiphy specific
regdomain, if such exists. Otherwise return the global regdom.

When no wiphy-idx is specified, return the global regdomain as well as
all wiphy-specific regulatory domains in the system, via a new nested
list of attributes.

Add a new attribute for each wiphy-specific regdomain, for usermode to
identify it as such.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: keep sending peer candidate events while in listen state
Nishikawa, Kenzoh [Tue, 16 Dec 2014 01:41:06 +0000 (01:41 +0000)]
mac80211: keep sending peer candidate events while in listen state

Instead of sending peer candidate events just once, send them as long
as the peer remains in the LISTEN state in the peering state machine,
when userspace is implementing the peering manager.
Userspace may silence the events from a peer by progressing the state
machine or by setting the link state to BLOCKED.

Fixes the problem that a mesh peering process won't be fired again after
the previous first peering trial fails due to like air propagation error
if the peering is managed by user space such as wpa_supplicant.

This patch works with another patch for wpa_supplicant described here
which fires a peering process again triggered by the notice from kernel.
http://lists.shmoo.com/pipermail/hostap/2014-November/031235.html

Signed-off-by: Kenzoh Nishikawa <Kenzoh.Nishikawa@jp.sony.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: notify channel switch at the end of ieee80211_chswitch_post_beacon()
Luciano Coelho [Tue, 16 Dec 2014 14:01:39 +0000 (16:01 +0200)]
mac80211: notify channel switch at the end of ieee80211_chswitch_post_beacon()

The call to cfg80211_ch_switch_notify() should be at the end of the
ieee80211_chswitch_post_beacon() function, because it should only be
sent if everything succeeded.

Fixes: d04b5ac9e70b ("cfg80211/mac80211: allow any interface to send channel switch notifications")
Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: notify NSS changed when IBSS and HT
Janusz Dziedzic [Wed, 17 Dec 2014 09:00:52 +0000 (10:00 +0100)]
mac80211: notify NSS changed when IBSS and HT

When using IBSS in HT mode, we always get NSS=1
in rc_update callback. Force NSS recalculation when
rates updated and notify driver that NSS changed.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: move U-APSD enablement to vif flags
Johannes Berg [Tue, 11 Nov 2014 11:48:42 +0000 (12:48 +0100)]
mac80211: move U-APSD enablement to vif flags

In order to let drivers have more dynamic U-APSD support,
move the enablement flag to the virtual interface driver
flags. This lets drivers not only set it up differently
for different interfaces, but also enable/disable on the
fly if needed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: ask driver to look at power level when starting AP
Johannes Berg [Thu, 13 Nov 2014 10:23:53 +0000 (11:23 +0100)]
mac80211: ask driver to look at power level when starting AP

The power level might have been set, but as the interface was idle
it might not have taken effect yet. Ask the driver to check the
power level when starting up an AP so that in this case the correct
power level is used in case the device/driver can only set it when
the interface is actually active.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: Fix accounting of multicast frames
Sujith Manoharan [Wed, 10 Dec 2014 15:56:11 +0000 (21:26 +0530)]
mac80211: Fix accounting of multicast frames

Since multicast frames are marked as no-ack, using
IEEE80211_TX_STAT_ACK to check if they have been
successfully transmitted by the driver is incorrect
since a driver can choose to ignore transmission status
for no-ack frames. This results in incorrect accounting
for such frames.

To fix this issue, this patch introduces a new flag
that can be used by drivers to indicate error-free
transmission of no-ack frames.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
[add a note about not setting the flag for non-no-ack frames]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: Move IEEE80211_TX_CTL_PS_RESPONSE
Sujith Manoharan [Wed, 10 Dec 2014 15:56:10 +0000 (21:26 +0530)]
mac80211: Move IEEE80211_TX_CTL_PS_RESPONSE

Move IEEE80211_TX_CTL_PS_RESPONSE to info->control.flags since
this is used only in the TX path (by ath9k). This frees up
a bit which can be used for other purposes.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agowireless: Support of IFLA_INFO_KIND rtnl attribute
Vadim Kochan [Tue, 9 Dec 2014 22:21:06 +0000 (00:21 +0200)]
wireless: Support of IFLA_INFO_KIND rtnl attribute

It allows to identify the wlan kind of device for the user application,
e.g.:

    # ip -d link

    1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default
        link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 promiscuity 0
    2: enp0s25: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000
        link/ether XX:XX:XX:XX:XX:XX brd ff:ff:ff:ff:ff:ff promiscuity 0
    3: wlp3s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP mode DEFAULT group default qlen 1000
        link/ether XX:XX:XX:XX:XX:XX brd ff:ff:ff:ff:ff:ff promiscuity 0
        wlan

Signed-off-by: Vadim Kochan <vadim4j@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
[make wireless_link_ops const]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: correctly check ad-hoc channels
Arik Nemtsov [Wed, 3 Dec 2014 16:08:17 +0000 (18:08 +0200)]
cfg80211: correctly check ad-hoc channels

Ad-hoc requires beaconing for regulatory purposes. Validate that the
channel is valid for beaconing, and not only enabled.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: don't WARN about two consecutive Country IE hint
Emmanuel Grumbach [Tue, 2 Dec 2014 07:53:25 +0000 (09:53 +0200)]
cfg80211: don't WARN about two consecutive Country IE hint

This can happen and there is no point in added more
detection code lower in the stack. Catching these in one
single point (cfg80211) is enough. Stop WARNING about this
case.

This fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=89001

Cc: stable@vger.kernel.org
Fixes: 2f1c6c572d7b ("cfg80211: process non country IE conflicting first")
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: update the channel context after channel switch
Emmanuel Grumbach [Sun, 30 Nov 2014 15:17:26 +0000 (17:17 +0200)]
mac80211: update the channel context after channel switch

When the channel switch has been made, a vif is now using
the channel context which was reserved. When that happens,
we need to update the channel context since its parameters
may change.

I hit a case in which I switched to a 40Mhz channel but the
reserved channel context was still on 20Mhz. The rate control
would try to send 40Mhz packets on a 20Mhz channel context and
that made iwlwifi's firmware unhappy.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonl80211: check matches array length before acessing it
Luciano Coelho [Mon, 1 Dec 2014 09:32:09 +0000 (11:32 +0200)]
nl80211: check matches array length before acessing it

If the userspace passes a malformed sched scan request (or a net
detect wowlan configuration) by adding a NL80211_ATTR_SCHED_SCAN_MATCH
attribute without any nested matchsets, a NULL pointer dereference
will occur.  Fix this by checking that we do have matchsets in our
array before trying to access it.

BUG: unable to handle kernel NULL pointer dereference at 0000000000000024
IP: [<ffffffffa002fd69>] nl80211_parse_sched_scan.part.67+0x6e9/0x900 [cfg80211]
PGD 865c067 PUD 865b067 PMD 0
Oops: 0002 [#1] SMP
Modules linked in: iwlmvm(O) iwlwifi(O) mac80211(O) cfg80211(O) compat(O) [last unloaded: compat]
CPU: 2 PID: 2442 Comm: iw Tainted: G           O   3.17.2 #31
Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
task: ffff880013800790 ti: ffff880008d80000 task.ti: ffff880008d80000
RIP: 0010:[<ffffffffa002fd69>]  [<ffffffffa002fd69>] nl80211_parse_sched_scan.part.67+0x6e9/0x900 [cfg80211]
RSP: 0018:ffff880008d838d0  EFLAGS: 00010293
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 000000000000143c RSI: 0000000000000000 RDI: ffff880008ee8dd0
RBP: ffff880008d83948 R08: 0000000000000002 R09: 0000000000000019
R10: ffff88001d1b3c40 R11: 0000000000000002 R12: ffff880019e85e00
R13: 00000000fffffed4 R14: ffff880009757800 R15: 0000000000001388
FS:  00007fa3b6d13700(0000) GS:ffff88003e200000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000024 CR3: 0000000008670000 CR4: 00000000000006e0
Stack:
 ffff880009757800 ffff880000000001 0000000000000000 ffff880008ee84e0
 0000000000000000 ffff880009757800 00000000fffffed4 ffff880008d83948
 ffffffff814689c9 ffff880009757800 ffff880008ee8000 0000000000000000
Call Trace:
 [<ffffffff814689c9>] ? nla_parse+0xb9/0x120
 [<ffffffffa00306de>] nl80211_set_wowlan+0x75e/0x960 [cfg80211]
 [<ffffffff810bf3d5>] ? mark_held_locks+0x75/0xa0
 [<ffffffff8161a77b>] genl_family_rcv_msg+0x18b/0x360
 [<ffffffff810bf66d>] ? trace_hardirqs_on+0xd/0x10
 [<ffffffff8161a9d4>] genl_rcv_msg+0x84/0xc0
 [<ffffffff8161a950>] ? genl_family_rcv_msg+0x360/0x360
 [<ffffffff81618e79>] netlink_rcv_skb+0xa9/0xd0
 [<ffffffff81619458>] genl_rcv+0x28/0x40
 [<ffffffff816184a5>] netlink_unicast+0x105/0x180
 [<ffffffff8161886f>] netlink_sendmsg+0x34f/0x7a0
 [<ffffffff8105a097>] ? kvm_clock_read+0x27/0x40
 [<ffffffff815c644d>] sock_sendmsg+0x8d/0xc0
 [<ffffffff811a75c9>] ? might_fault+0xb9/0xc0
 [<ffffffff811a756e>] ? might_fault+0x5e/0xc0
 [<ffffffff815d5d26>] ? verify_iovec+0x56/0xe0
 [<ffffffff815c73e0>] ___sys_sendmsg+0x3d0/0x3e0
 [<ffffffff810a7be8>] ? sched_clock_cpu+0x98/0xd0
 [<ffffffff810611b4>] ? __do_page_fault+0x254/0x580
 [<ffffffff810bb39f>] ? up_read+0x1f/0x40
 [<ffffffff810611b4>] ? __do_page_fault+0x254/0x580
 [<ffffffff812146ed>] ? __fget_light+0x13d/0x160
 [<ffffffff815c7b02>] __sys_sendmsg+0x42/0x80
 [<ffffffff815c7b52>] SyS_sendmsg+0x12/0x20
 [<ffffffff81751f69>] system_call_fastpath+0x16/0x1b

Fixes: ea73cbce4e1f ("nl80211: fix scheduled scan RSSI matchset attribute confusion")
Cc: stable@vger.kernel.org [3.15+]
Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: use __force __rcu to suppress sparse warning
Johannes Berg [Fri, 12 Dec 2014 11:26:25 +0000 (12:26 +0100)]
cfg80211: use __force __rcu to suppress sparse warning

The code assigns a constant value (a pointer to a static variable)
to an RCU pointer, which results in a sparse warning:
  reg.c:112:10: warning: cast adds address space to expression (<asn:4>)

Suppress this warning by using __force.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: avoid mem leak on driver hint set
Arik Nemtsov [Thu, 4 Dec 2014 10:22:16 +0000 (12:22 +0200)]
cfg80211: avoid mem leak on driver hint set

In the already-set and intersect case of a driver-hint, the previous
wiphy regdomain was not freed before being reset with a copy of the
cfg80211 regdomain.

Cc: stable@vger.kernel.org
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Acked-by: Luis R. Rodriguez <mcgrof@suse.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: Fix 160 MHz channels with 80+80 and 160 MHz drivers
Jouni Malinen [Thu, 11 Dec 2014 21:48:55 +0000 (23:48 +0200)]
cfg80211: Fix 160 MHz channels with 80+80 and 160 MHz drivers

The VHT supported channel width field is a two bit integer, not a
bitfield. cfg80211_chandef_usable() was interpreting it incorrectly and
ended up rejecting 160 MHz channel width if the driver indicated support
for both 160 and 80+80 MHz channels.

Cc: stable@vger.kernel.org (3.16+)
Fixes: 3d9d1d6656a73 ("nl80211/cfg80211: support VHT channel configuration")
       (however, no real drivers had 160 MHz support it until 3.16)
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: fix multicast LED blinking and counter
Andreas Müller [Fri, 12 Dec 2014 11:11:11 +0000 (12:11 +0100)]
mac80211: fix multicast LED blinking and counter

As multicast-frames can't be fragmented, "dot11MulticastReceivedFrameCount"
stopped being incremented after the use-after-free fix. Furthermore, the
RX-LED will be triggered by every multicast frame (which wouldn't happen
before) which wouldn't allow the LED to rest at all.

Fixes https://bugzilla.kernel.org/show_bug.cgi?id=89431 which also had the
patch.

Cc: stable@vger.kernel.org
Fixes: b8fff407a180 ("mac80211: fix use-after-free in defragmentation")
Signed-off-by: Andreas Müller <goo@stapelspeicher.org>
[rewrite commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: avoid using uninitialized stack data
Jes Sorensen [Wed, 10 Dec 2014 19:14:07 +0000 (14:14 -0500)]
mac80211: avoid using uninitialized stack data

Avoid a case where we would access uninitialized stack data if the AP
advertises HT support without 40MHz channel support.

Cc: stable@vger.kernel.org
Fixes: f3000e1b43f1 ("mac80211: fix broken use of VHT/20Mhz with some APs")
Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agor8169:update rtl8168g pcie ephy parameter
Chun-Hao Lin [Wed, 10 Dec 2014 13:28:38 +0000 (21:28 +0800)]
r8169:update rtl8168g pcie ephy parameter

Add ephy parameter to rtl8168g.
Also change the common function of rtl8168g from "rtl_hw_start_8168g_1" to
 "rtl_hw_start_8168g". And function "rtl_hw_start_8168g_1" is used for
setting rtl8168g hardware parameters.

Following is the explanation of what hardware parameter change for.
rtl8168g may erroneous judge the PCIe signal quality and show the error bit
on PCI configuration space when in PCIe low power mode.
The following ephy parameters are for above issue.
{ 0x00, 0x0000, 0x0008 }
{ 0x0c, 0x37d0, 0x0820 }
{ 0x1e, 0x0000, 0x0001 }

rtl8168g may return to PCIe L0 from PCIe L0s low power mode too slow.
The following ephy parameter is for above issue.
{ 0x19, 0x8000, 0x0000 }

Signed-off-by: Chunhao Lin <hau@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: dsa: bcm_sf2: force link for all fixed PHY devices
Florian Fainelli [Fri, 12 Dec 2014 02:12:42 +0000 (18:12 -0800)]
net: dsa: bcm_sf2: force link for all fixed PHY devices

For ports of the switch that we define as "fixed PHYs" such as MoCA, we
would have our Port 7 special handling that would allow us to assert the
link status indication.

For other ports, such as e.g: RGMII_1 connected to a cable modem, we
would rely on whatever the bootloader has left configured, which is a
bad assumption to make, we really need to force the link status
indication here.

Fixes: 246d7f773c13 ("net: dsa: add Broadcom SF2 switch driver")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'dma_mb'
David S. Miller [Fri, 12 Dec 2014 02:15:37 +0000 (21:15 -0500)]
Merge branch 'dma_mb'

Alexander Duyck says:

====================
arch: Add lightweight memory barriers for coherent memory access

These patches introduce two new primitives for synchronizing cache coherent
memory writes and reads.  These two new primitives are:

dma_rmb()
dma_wmb()

The first patch cleans up some unnecessary overhead related to the
definition of read_barrier_depends, smp_read_barrier_depends, and comments
related to the barrier.

The second patch adds the primitives for the applicable architectures and
asm-generic.

The third patch adds the barriers to r8169 which turns out to be a good
example of where the new barriers might be useful as they have full
rmb()/wmb() barriers ordering accesses to the descriptors and the DescOwn
bit.

The fourth patch adds support for coherent_rmb() to the Intel fm10k, igb,
and ixgbe drivers.  Testing with the ixgbe driver has shown a processing
time reduction of at least 7ns per 64B frame on a Core i7-4930K.

This patch series is essentially the v7 for:
v4-7: Add lightweight memory barriers for coherent memory access
v3: Add lightweight memory barriers fast_rmb() and fast_wmb()
v2: Introduce load_acquire() and store_release()
v1: Introduce read_acquire()

The key changes in this patch series versus the earlier patches are:
v7 resubmit:
- Added Acked-by: Ben Herrenschmidt from v5 to dma_rmb/wmb patch
- No code changes from previous set, still applies cleanly and builds.
v7:
- Dropped test/debug patch that was accidentally slipped in
v6:
- Replaced "memory based device I/O" with "consistent memory" in
  docs
- Added reference to DMA-API.txt to explain consistent memory
v5:
- Renamed barriers dma_rmb and dma_wmb
- Undid smp_wmb changes in x86 and PowerPC
- Defined smp_rmb as __lwsync for SMP case on PowerPC
v4:
- Renamed barriers coherent_rmb and coherent_wmb
- Added smp_lwsync for use in smp_load_acquire/smp_store_release
v3:
- Moved away from acquire()/store() and instead focused on barriers
- Added cleanup of read_barrier_depends
- Added change in r8169 to fix cur_tx/DescOwn ordering
- Simplified changes to just replacing/moving barriers in r8169
- Added update to documentation with code example
v2:
- Renamed read_acquire() to be consistent with smp_load_acquire()
- Changed barrier used to be consistent with smp_load_acquire()
- Updated PowerPC code to use __lwsync based on IBM article
- Added store_release() as this is a viable use case for drivers
- Added r8169 patch which is able to fully use primitives
- Added fm10k/igb/ixgbe patch which is able to test performance
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofm10k/igb/ixgbe: Use dma_rmb on Rx descriptor reads
Alexander Duyck [Thu, 11 Dec 2014 23:02:28 +0000 (15:02 -0800)]
fm10k/igb/ixgbe: Use dma_rmb on Rx descriptor reads

This change makes it so that dma_rmb is used when reading the Rx
descriptor.  The advantage of dma_rmb is that it allows for a much
lower cost barrier on x86, powerpc, arm, and arm64 architectures than a
traditional memory barrier when dealing with reads that only have to
synchronize to coherent memory.

In addition I have updated the code so that it just checks to see if any
bits have been set instead of just the DD bit since the DD bit will always
be set as a part of a descriptor write-back so we just need to check for a
non-zero value being present at that memory location rather than just
checking for any specific bit.  This allows the code itself to appear much
cleaner and allows the compiler more room to optimize.

Cc: Matthew Vick <matthew.vick@intel.com>
Cc: Don Skidmore <donald.c.skidmore@intel.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8169: Use dma_rmb() and dma_wmb() for DescOwn checks
Alexander Duyck [Thu, 11 Dec 2014 23:02:17 +0000 (15:02 -0800)]
r8169: Use dma_rmb() and dma_wmb() for DescOwn checks

The r8169 use a pair of wmb() calls when setting up the descriptor rings.
The first is to synchronize the descriptor data with the descriptor status,
and the second is to synchronize the descriptor status with the use of the
MMIO doorbell to notify the device that descriptors are ready.  This can
come at a heavy price on some systems, and is not really necessary on
systems such as x86 as a simple barrier() would suffice to order store/store
accesses.  As such we can replace the first memory barrier with
dma_wmb() to reduce the cost for these accesses.

In addition the r8169 uses a rmb() to prevent compiler optimization in the
cleanup paths, however by moving the barrier down a few lines and replacing
it with a dma_rmb() we should be able to use it to guarantee
descriptor accesses do not occur until the device has updated the DescOwn
bit from its end.

One last change I made is to move the update of cur_tx in the xmit path to
after the wmb.  This way we can guarantee the device and all CPUs should
see the DescOwn update before they see the cur_tx value update.

Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Cc: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>