openwrt/staging/blogic.git
10 years agobrcmfmac: Add wowl patterns support.
Hante Meuleman [Tue, 28 Oct 2014 13:56:06 +0000 (14:56 +0100)]
brcmfmac: Add wowl patterns support.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Daniel (Deognyoun) Kim <dekim@broadcom.com>
Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agobrcmfmac: Add wowl support for SDIO devices.
Hante Meuleman [Tue, 28 Oct 2014 13:56:05 +0000 (14:56 +0100)]
brcmfmac: Add wowl support for SDIO devices.

This patch adds wowl support for SDIO bus devices. This feature
requires FW which has support for wowl built in.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agobrcmfmac: Add wowl support for USB devices.
Hante Meuleman [Tue, 28 Oct 2014 13:56:04 +0000 (14:56 +0100)]
brcmfmac: Add wowl support for USB devices.

This patch adds wowl support for USB bus devices. This feature
requires FW which has support for wowl built in.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Daniel (Deognyoun) Kim <dekim@broadcom.com>
Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: set pulse_rssi threshold to 15
Lorenzo Bianconi [Tue, 28 Oct 2014 11:37:16 +0000 (12:37 +0100)]
ath9k: set pulse_rssi threshold to 15

Reduce pulse_rssi threshold to 15 in order to improve radar pattern detection
probability on ext channel

Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi83@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoMerge tag 'iwlwifi-next-for-john-2014-10-29' of git://git.kernel.org/pub/scm/linux...
John W. Linville [Thu, 30 Oct 2014 19:22:06 +0000 (15:22 -0400)]
Merge tag 'iwlwifi-next-for-john-2014-10-29' of git://git./linux/kernel/git/iwlwifi/iwlwifi-next

Emmanuel Grumbach <egrumbach@gmail.com> says:

"The big new thing here is netdetect which allows the
firmware to wake up the platform when a specific network
is detected. Along with that I have fixes for d3 operation.
The usual amount of rate scaling stuff - we now support STBC.
The other commit that stands out is Johannes's work on
devcoredump. He basically starts to use the standard
infrastructure he built."

Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoiwlwifi: mvm: unref SCAN ref on scan completion
Eliad Peller [Sun, 22 Jun 2014 11:44:44 +0000 (14:44 +0300)]
iwlwifi: mvm: unref SCAN ref on scan completion

The scan reference wasn't released in case of offloaded scan,
causing the refcount to go wrong and prevent the device
from going into d0i3.

Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: rs: add Tx STBC support
Eyal Shapira [Thu, 23 Oct 2014 10:58:38 +0000 (13:58 +0300)]
iwlwifi: mvm: rs: add Tx STBC support

STBC is enabled on HT/VHT SISO rates in case we don't care
about power consumption and it won't hurt BT.
This is done only in case the peer and our chip support STBC
of course.
While at it fixed a bug which was causing bw and ldpc
flags to be set incorrectly in the rate scale table in case
we are switching to a legacy Tx column. This had no real impact.

Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: fix scan condition iterator
David Spinadel [Sun, 26 Oct 2014 13:53:27 +0000 (15:53 +0200)]
iwlwifi: mvm: fix scan condition iterator

Scan condition iterator assumes that an interface is associated if
phy_ctxt is assigned, but this isn't the sutuation in P2P device.
OTOH P2P device is never associated so we can simply ignore it.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: port to devcoredump framework
Johannes Berg [Thu, 9 Oct 2014 15:01:36 +0000 (17:01 +0200)]
iwlwifi: mvm: port to devcoredump framework

iwlwifi features a debug mechanism that allows to dump
binary data which is helpful to debug the firmware.
Until now, this data was made available for the userspace
through debugfs. For this exact purpose, devcoredump was
created. Move to the new infrastructure.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: enable STBC support for 8000 chips
Eyal Shapira [Thu, 23 Oct 2014 12:30:47 +0000 (15:30 +0300)]
iwlwifi: enable STBC support for 8000 chips

These chips support STBC in both Tx and Rx.
Given that we've added STBC Tx support enable it.

Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: dvm: Fix probable mask then right shift defect
Joe Perches [Mon, 27 Oct 2014 05:25:03 +0000 (22:25 -0700)]
iwlwifi: dvm: Fix probable mask then right shift defect

Precedence of & and >> is not the same and is not left to right.
shift has higher precedence and should be done after the mask.

Add parentheses around the mask.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: remove unnecessary includes in tt.c
Luciano Coelho [Fri, 10 Oct 2014 07:03:02 +0000 (10:03 +0300)]
iwlwifi: mvm: remove unnecessary includes in tt.c

We used to need to include some headers in order to read the registers
and do all the calculations by ourselves, but this is not done anymore
(i.e. we let the firmware do this for us), so we don't need to include
those headers anymore.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: flush queues without mutex held
Johannes Berg [Fri, 10 Oct 2014 06:59:27 +0000 (08:59 +0200)]
iwlwifi: mvm: flush queues without mutex held

For WMM-AC we may have to be changing QoS parameters while the
queues are being flushed. This is not relevant in the "drop"
case, but matters when we wait for the device to finish the
transmissions.

To allow this, do the actual waiting without holding the mutex.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: improve MCS rate warning
Johannes Berg [Wed, 8 Oct 2014 07:57:29 +0000 (09:57 +0200)]
iwlwifi: mvm: improve MCS rate warning

The warning gives no information about the frame, and presents
the flags so that one might think they're the frame. Clarify
and add more information.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: BT Coex - make the multiprio lut a constant
Emmanuel Grumbach [Sun, 19 Oct 2014 11:18:14 +0000 (14:18 +0300)]
iwlwifi: mvm: BT Coex - make the multiprio lut a constant

This makes it easier to tune the values during the testing.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: remove mvm argument from get_queues_mask
Arik Nemtsov [Sun, 21 Sep 2014 16:00:42 +0000 (19:00 +0300)]
iwlwifi: mvm: remove mvm argument from get_queues_mask

It is unused and won't be available in some future invocations of the
function.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: add MVM_FW_BCAST_FILTER_PASS_ALL option
Max Stepanov [Wed, 15 Oct 2014 08:27:16 +0000 (11:27 +0300)]
iwlwifi: mvm: add MVM_FW_BCAST_FILTER_PASS_ALL option

The MVM_FW_BCAST_FILTER_PASS_ALL option defined in iwl-dbg-cfg.ini
configuration file allows to enable/disable FW broadcast filtering.

If MVM_FW_BCAST_FILTER_PASS_ALL is not defined in iwl-dbg-cfg.ini or
its value is 0, the mvm broadcast filtering is enabled.

Setting MVM_FW_BCAST_FILTER_PASS_ALL=1 in iwl-dbg-cfg.ini blocks sending
of BCAST_FILTER_CMD to FW.

Signed-off-by: Max Stepanov <Max.Stepanov@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: BT Coex - move BT_ANTENNA_COUPLING_THRESHOLD to constants
Emmanuel Grumbach [Tue, 7 Oct 2014 11:44:18 +0000 (14:44 +0300)]
iwlwifi: mvm: BT Coex - move BT_ANTENNA_COUPLING_THRESHOLD to constants

This is were it really needs to be.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: always run the secured flow for family 8000
Eran Harary [Mon, 29 Sep 2014 06:27:56 +0000 (08:27 +0200)]
iwlwifi: always run the secured flow for family 8000

In the new format the "CSS section" has the same TLV type
as the "mem section". So we need to run the secured flow
for all the 8000 products.

Signed-off-by: Eran Harary <eran.harary@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: pcie: fix recovery from ARC reset in WoWLAN
Emmanuel Grumbach [Sun, 12 Oct 2014 10:23:40 +0000 (13:23 +0300)]
iwlwifi: pcie: fix recovery from ARC reset in WoWLAN

When the ARC is reset when we exit from Sx in case we had
WoWLAN running, we can't access the prph before we reset
the NIC.

Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: pcie: warn if extern fw_debug buffer failed
Emmanuel Grumbach [Thu, 18 Sep 2014 12:44:04 +0000 (15:44 +0300)]
iwlwifi: pcie: warn if extern fw_debug buffer failed

Otherwise we have no way to know that the buffer hasn't been
allocated.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: allow to restrict Tx aggregation size per-device
Gregory Greenman [Tue, 2 Sep 2014 14:04:58 +0000 (16:04 +0200)]
iwlwifi: allow to restrict Tx aggregation size per-device

A few slave bus devices show better performance when we
reduce the size of the Tx A-MPDU.

Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: change the iwl_mvm_d3_iface_iterator into a generic function
Luciano Coelho [Thu, 2 Oct 2014 12:38:04 +0000 (15:38 +0300)]
iwlwifi: mvm: change the iwl_mvm_d3_iface_iterator into a generic function

Getting the BSS station vif is something that may be needed by other
parts of the code.  So, instead of having an iterator specifically for
d3, change it into a generic one in utils.c.  Additionally, add a
iwl_mvm_get_bss_vif() function to make it easier to retrieving it.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: add debugfs entry to configure netdetect SSIDs
Luciano Coelho [Wed, 24 Sep 2014 05:29:11 +0000 (08:29 +0300)]
iwlwifi: mvm: add debugfs entry to configure netdetect SSIDs

Before we get all the chain (ie. mac80211, cfg80211, nl80211 and
userspace) changed to support net-detect, we can use this debugfs
entry for easy testing and as a proof of concept.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: spin off a function to start scan offload
Luciano Coelho [Wed, 1 Oct 2014 10:22:40 +0000 (13:22 +0300)]
iwlwifi: mvm: spin off a function to start scan offload

The net-detect feature will require a scan offload to be started in
the same way it is done now for scheduled scan.  Spin a new function
off of the sched_scan_start op code for reuse.  Additionally,
restructure the function a bit for more readability.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: move the check if associated outside of the iterator for wowlan
Luciano Coelho [Mon, 22 Sep 2014 10:01:59 +0000 (13:01 +0300)]
iwlwifi: mvm: move the check if associated outside of the iterator for wowlan

Instead of checking if we are associated when suspending with wowlan
enabled in the interface iterator, allow it to return an unassociated
vif and move the check to the main suspend function.  This will be
needed by netdetect, since it should also work when we are not
associated but the vif is active.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: split wowlan handling out of the main suspend function
Luciano Coelho [Mon, 22 Sep 2014 07:59:53 +0000 (10:59 +0300)]
iwlwifi: mvm: split wowlan handling out of the main suspend function

Take the WoWLAN handling code out of the main suspend function,
dividing it into three parts: get_config (which is used before the
firmware is switched), switch_to_d3 (which handles the rebooting of
the hardware with the D3 firmware) and config (which configures the D3
firmware for WoWLAN operation).  This is necessary to prepare for the
net-detect implementation, which will use only the switch_to_d3 part
of this flow.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: hold the rtnl when resuming from a d3 test
Luciano Coelho [Mon, 22 Sep 2014 10:22:51 +0000 (13:22 +0300)]
iwlwifi: mvm: hold the rtnl when resuming from a d3 test

We use ieee80211_iter_keys() which requires the rtnl to be held.  If
we don't lock the rtnl, like we do when we suspend during a d3 test,
we get the following splat:

 RTNL: assertion failed at net/mac80211/key.c (566)
 CPU: 1 PID: 26529 Comm: cat Tainted: G        W  O 3.10.29-dev #1
 Hardware name: Dell Inc. Latitude E6430/0CPWYR, BIOS A09 12/13/2012
  e7b15008 e7b15008 e68adc1c c168aa62 e68adc54 f91f20b6 f923700c f9236fd8
  00000236 00000000 ece23874 00000000 f94941e0 00000000 e43b8e48 e7b15008
  00000000 e8b69e78 e68adcc0 f9493ab9 e68adc68 00000000 e43b8e48 e7b15008
 Call Trace:
  [<c168aa62>] dump_stack+0x16/0x18
  [<f91f20b6>] ieee80211_iter_keys+0x166/0x170 [mac80211]
  [<f94941e0>] ? iwl_mvm_query_wakeup_reasons+0x5f0/0x5f0 [iwlmvm]
  [<f9493ab9>] iwl_mvm_setup_connection_keep.isra.5+0x99/0x1d0 [iwlmvm]
  [<f9165e28>] ? cfg80211_report_wowlan_wakeup+0x308/0x510 [cfg80211]
  [<f9493fe5>] iwl_mvm_query_wakeup_reasons+0x3f5/0x5f0 [iwlmvm]
  [<c116125a>] ? init_object+0x3a/0x70
  [<f8a5b8ee>] ? iwl_trans_pcie_d3_resume+0x1be/0x3a0 [iwlwifi]
  [<f94956da>] __iwl_mvm_resume+0x14a/0x180 [iwlmvm]
  [<f9495736>] iwl_mvm_d3_test_release+0x26/0xc0 [iwlmvm]
  [<c117a64d>] __fput+0xad/0x210
  [<c117a7bd>] ____fput+0xd/0x10
  [<c10601a1>] task_work_run+0x81/0xb0
  [<c1040fa5>] do_exit+0x255/0xac0
  [<c104e511>] ? dequeue_signal+0x31/0x1a0
  [<c1041888>] do_group_exit+0x38/0xa0
  [<c10a6cfb>] ? trace_hardirqs_on+0xb/0x10
  [<c1051c31>] get_signal_to_deliver+0x1e1/0x8e0
  [<c104bd52>] ? try_to_del_timer_sync+0x42/0x60
  [<c104bd70>] ? try_to_del_timer_sync+0x60/0x60
  [<c168dc82>] ? schedule_timeout+0x102/0x2a0
  [<c10011ba>] do_signal+0x3a/0x8e0
  [<c104b190>] ? __internal_add_timer+0xb0/0xb0
  [<c168de7a>] ? schedule_timeout_interruptible+0x1a/0x20
  [<c104ce39>] ? msleep_interruptible+0x39/0x40
  [<f94939a9>] ? iwl_mvm_d3_test_read+0x49/0x70 [iwlmvm]
  [<c11797fc>] ? vfs_read+0x8c/0x160
  [<c11243af>] ? SyS_fadvise64_64+0x15f/0x2b0
  [<f9493960>] ? iwl_mvm_wowlan_program_keys+0x4a0/0x4a0 [iwlmvm]
  [<c1179a57>] ? SyS_read+0x57/0xa0
  [<c1001acf>] do_notify_resume+0x6f/0xa0
  [<c1692500>] work_notifysig+0x29/0x31

Fix this by hold the rtnl lock when calling __iwl_mvm_resume() in the
d3 test wake up path.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: remove unneeded NULL pointer check
Emmanuel Grumbach [Sun, 21 Sep 2014 10:29:48 +0000 (13:29 +0300)]
iwlwifi: mvm: remove unneeded NULL pointer check

mac80211 will never call rate_control_tx_status with a NULL
pointer for sta. Remove the superfluous check. This check
misled smatch.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: rs: option to easily disable Tx MIMO
Eyal Shapira [Wed, 17 Sep 2014 10:56:29 +0000 (13:56 +0300)]
iwlwifi: rs: option to easily disable Tx MIMO

The current TLC isn't optimized for low latency and some devices
have issues with MIMO. This kind of combo creates latency issues.
Allow to temporarily disable MIMO for tests in order to solve
the latency issues without the added complexity of MIMO.

Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoMerge remote-tracking branch 'iwlwifi-fixes/master' into iwlwifi-next
Emmanuel Grumbach [Wed, 29 Oct 2014 10:48:37 +0000 (12:48 +0200)]
Merge remote-tracking branch 'iwlwifi-fixes/master' into iwlwifi-next

10 years agoath9k: use a random MAC address if the EEPROM address is invalid
Felix Fietkau [Sat, 25 Oct 2014 15:19:35 +0000 (17:19 +0200)]
ath9k: use a random MAC address if the EEPROM address is invalid

Based on OpenWrt patch by Gabor Juhos

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: allow disabling bands via platform data
Felix Fietkau [Sat, 25 Oct 2014 15:19:34 +0000 (17:19 +0200)]
ath9k: allow disabling bands via platform data

Some devices have multiple bands enables in the EEPROM data, even though
they are only calibrated for one. Allow platform data to disable
unsupported bands.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: add support for endian swap of eeprom from platform data
Felix Fietkau [Sat, 25 Oct 2014 15:19:33 +0000 (17:19 +0200)]
ath9k: add support for endian swap of eeprom from platform data

On some devices (especially little-endian ones), the flash EEPROM data
has a different endian, which needs to be detected.
Add a flag to the platform data to allow overriding that behavior

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: start initial NF calibration after PA calibration on <AR9003
Felix Fietkau [Sat, 25 Oct 2014 15:19:32 +0000 (17:19 +0200)]
ath9k_hw: start initial NF calibration after PA calibration on <AR9003

This makes the initial NF calibration less likely to fail.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: do not run NF and periodic calibration at the same time
Felix Fietkau [Sat, 25 Oct 2014 15:19:31 +0000 (17:19 +0200)]
ath9k_hw: do not run NF and periodic calibration at the same time

It can cause inconsistent calibration results or in some cases turn the
radio deaf.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: restart hardware after noise floor calibration failure
Felix Fietkau [Sat, 25 Oct 2014 15:19:30 +0000 (17:19 +0200)]
ath9k: restart hardware after noise floor calibration failure

When NF calibration fails, the radio often becomes deaf. The usual
hardware hang checks do not detect this, so it's better to issue a reset
when that happens.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: clean up debugfs print of reset causes
Felix Fietkau [Sat, 25 Oct 2014 15:19:29 +0000 (17:19 +0200)]
ath9k: clean up debugfs print of reset causes

Reduce code duplication

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: fix processing RXORN interrupts
Felix Fietkau [Sat, 25 Oct 2014 15:19:28 +0000 (17:19 +0200)]
ath9k: fix processing RXORN interrupts

The "goto chip_reset" is a bit misleading, because it does not actually
issue a chip reset. Instead it is bypassing processing of other
interrupts and assumes that the tasklet will issue a chip reset.

In the case of RXORN this does not happen, so bypassing processing of
other interrupts will simply allow them to fire again. Even if RXORN
was triggering a reset, it is not critical enough to need the bypass
here.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: remove support for UB124 tx gain table
Felix Fietkau [Sat, 25 Oct 2014 15:19:27 +0000 (17:19 +0200)]
ath9k_hw: remove support for UB124 tx gain table

UB124 is a USB based reference design not supported by ath9k or
ath9k_htc.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: make support for PC-OEM cards optional
Felix Fietkau [Sat, 25 Oct 2014 15:19:26 +0000 (17:19 +0200)]
ath9k_hw: make support for PC-OEM cards optional

The initvals use up quite a bit of space, and PC-OEM support is
typically not needed on embedded systems

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Update AR955x initvals
Sujith Manoharan [Wed, 22 Oct 2014 06:51:45 +0000 (12:21 +0530)]
ath9k: Update AR955x initvals

* Fix a 11b/EVM issue by adjusting
  FIR filter coefficients.

* Fix a problem with receiving probe request
  frames sent at 11b rate.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Update AR9580 initvals
Sujith Manoharan [Wed, 22 Oct 2014 06:51:44 +0000 (12:21 +0530)]
ath9k: Update AR9580 initvals

This fixes RX sensitivity issues with AR9580.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Do not start BA when scanning
Sujith Manoharan [Fri, 17 Oct 2014 02:10:30 +0000 (07:40 +0530)]
ath9k: Do not start BA when scanning

mac80211 currently has a race which can be hit
with this sequence:

* Start a scan operation.
* TX BA is initiated by ieee80211_start_tx_ba_session().
* Driver sets up internal state and calls
  ieee80211_start_tx_ba_cb_irqsafe().
* mac80211 adds a packet to sdata->skb_queue with
  type IEEE80211_SDATA_QUEUE_AGG_START.
* ieee80211_iface_work() doesn't process the
  packet because scan is in progress.
* ADDBA response timer expires and the sta/tid is
  torn down.
* Driver receives BA stop notification and calls
  ieee80211_stop_tx_ba_cb_irqsafe().
* This is also added to the queue by mac80211.
* Now, scan finishes.

At this point, the queued up packets might be processed
if some other operation schedules the sdata work. Since
the tids have been cleaned up already, warnings are hit.

If this doesn't happen, the packets are left in the queue
until the interface is torn down.

Since initiating a BA session when scan is in progress
leads to flaky connections, especially in MCC mode, we
can drop the TX BA request. This improves connectivity
with legacy clients in MCC mode.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Improve flush() in mcc mode
Sujith Manoharan [Fri, 17 Oct 2014 02:10:29 +0000 (07:40 +0530)]
ath9k: Improve flush() in mcc mode

The flush timeout in MCC mode is very small, since
we are constrained by the time slice for each
channel context, but since only the HW queues are
flushed when switching contexts, it is acceptable.

Since the SW queues are also emptied in the mac80211 flush()
callback, a larger duration is needed. Add an override
argument to __ath9k_flush() and set it when flush()
is called in MCC mode. This allows the driver to
drain both the SW and HW queues.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix HW scan abort
Sujith Manoharan [Fri, 17 Oct 2014 02:10:28 +0000 (07:40 +0530)]
ath9k: Fix HW scan abort

Instead of using ATH_CHANCTX_EVENT_ASSIGN to abort
a HW scan when a new interface becomes active, use the
mgd_prepare_tx() callback. This allows us to make
sure that the GO's channel becomes operational by
using flush_work().

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Send oneshot NoA
Sujith Manoharan [Fri, 17 Oct 2014 02:10:27 +0000 (07:40 +0530)]
ath9k: Send oneshot NoA

This patch makes sure that a GO interface
sends out a new NoA schedule with 200ms duration
when mgd_prepare_tx() is called.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix RoC expiration
Sujith Manoharan [Fri, 17 Oct 2014 02:10:26 +0000 (07:40 +0530)]
ath9k: Fix RoC expiration

mac80211 has to be notified when a RoC period
expires in the driver. In MCC mode, since the
offchannel/RoC timer is set with the requested
duration, ieee80211_remain_on_channel_expired() needs
to be called when the timer expires.

But, currently it is done after we move back to
the operating channel. This is incorrect - fix this
by calling ieee80211_remain_on_channel_expired() when
the RoC timer expires and in ath_roc_complete() when
the RoC request is aborted.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Use a helper function for bmiss
Sujith Manoharan [Fri, 17 Oct 2014 02:10:25 +0000 (07:40 +0530)]
ath9k: Use a helper function for bmiss

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Use a helper function to set NoA
Sujith Manoharan [Fri, 17 Oct 2014 02:10:24 +0000 (07:40 +0530)]
ath9k: Use a helper function to set NoA

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Use a helper function for offchannel NoA
Sujith Manoharan [Fri, 17 Oct 2014 02:10:23 +0000 (07:40 +0530)]
ath9k: Use a helper function for offchannel NoA

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Check for active GO in mgd_prepare_tx()
Sujith Manoharan [Fri, 17 Oct 2014 02:10:22 +0000 (07:40 +0530)]
ath9k: Check for active GO in mgd_prepare_tx()

If a GO interface is active when we receive a
mgd_prepare_tx() call, then we need to send
out a new NoA before switching to a new context.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Add a function to check for an active GO
Sujith Manoharan [Fri, 17 Oct 2014 02:10:21 +0000 (07:40 +0530)]
ath9k: Add a function to check for an active GO

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix address management
Sujith Manoharan [Fri, 17 Oct 2014 02:10:20 +0000 (07:40 +0530)]
ath9k: Fix address management

Since both the arguments need to satisfy
the alignment requirements of ether_addr_copy(),
use memcpy() in cases where there will be no
big performance benefit and make sure that
ether_addr_copy() calls use properly aligned
arguments.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Send AUTHORIZED event only for station mode
Sujith Manoharan [Fri, 17 Oct 2014 02:10:19 +0000 (07:40 +0530)]
ath9k: Send AUTHORIZED event only for station mode

ATH_CHANCTX_EVENT_AUTHORIZED is required to trigger
the MCC scheduler when a station interface becomes
authorized. But, since the driver gets station state
notifications when the current operating mode is AP
too, make sure that we send ATH_CHANCTX_EVENT_AUTHORIZED
only when the interface is in station mode.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Check for pending frames properly
Sujith Manoharan [Fri, 17 Oct 2014 02:10:18 +0000 (07:40 +0530)]
ath9k: Check for pending frames properly

Pending frames in the driver can be present
either in the HW queues or SW. ath9k_has_pending_frames()
currently checks for the HW queues first and then
checks if any ACs are queued in the driver.

In MCC mode, we need to check the HW queues alone, since
the SW queues are just marked as 'stopped' - they will
be processed in the next context switch. But since we
don't differentiate this now, mention whether we want
to check if there are frames in the SW queues.

* The flush() callback checks both HW and SW queues.
* The tx_frames_pending() callback does the same.
* The call to __ath9k_flush() in MCC mode checks HW queues alone.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix offchannel flush timeout
Sujith Manoharan [Fri, 17 Oct 2014 02:10:17 +0000 (07:40 +0530)]
ath9k: Fix offchannel flush timeout

An offchannel operation also needs to have
a flush timeout that doesn't exceed the NoA
absence duration of a GO context, so use
channel_switch_time. The first offchannel
operation is set a flush timeout of 10ms since
channel_switch_time will be zero.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix MCC flush timeout
Sujith Manoharan [Fri, 17 Oct 2014 02:10:16 +0000 (07:40 +0530)]
ath9k: Fix MCC flush timeout

In MCC mode, the duration for a channel context
is half the beacon interval and having a large
flush timeout will adversely affect GO operation,
since the default value of 200ms will overshoot
the advertised NoA absence duration.

The scheduler initiates a channel context switch
only when the slot duration for the current
context expires, so there is no possibility of
having a fixed timeout for flush.

Since the channel_switch_time is added to the
absence duration when the GO sets up the NoA
attribute, this is the maximum time that we
have to flush the TX queues. The duration is very
small, but we don't have a choice in MCC mode.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Use configurable timeout for flush
Sujith Manoharan [Fri, 17 Oct 2014 02:10:15 +0000 (07:40 +0530)]
ath9k: Use configurable timeout for flush

The timeout value for flushing the TX queues
is hardcoded at 200ms right now. Use a channel
context-specific value instead to allow adjustments
to the timeout in case MCC is enabled.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Clear NoA schedule properly
Sujith Manoharan [Fri, 17 Oct 2014 02:10:14 +0000 (07:40 +0530)]
ath9k: Clear NoA schedule properly

When an active context transitions to inactive
state, the NoA schedule needs to be removed
for the context that has beaconing enabled.
Not doing this will affect p2p clients.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Disable beacon tasklet during reset
Sujith Manoharan [Fri, 17 Oct 2014 02:10:13 +0000 (07:40 +0530)]
ath9k: Disable beacon tasklet during reset

When a chip reset is done, all running timers,
tasklets etc. are stopped but the beacon tasklet
is left running. Fix this.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Set ATH_OP_HW_RESET before HW reset
Sujith Manoharan [Fri, 17 Oct 2014 02:10:12 +0000 (07:40 +0530)]
ath9k: Set ATH_OP_HW_RESET before HW reset

When a HW reset is done, the interrupt tasklet is
disabled before ISRs are disabled in the HW. This
allows a small window where the HW can still generate
interrupts. Since the tasklet is disabled and not killed,
it is not scheduled but deferred for execution at a later
time.

This happens because ATH_OP_HW_RESET is not set when ath_reset()
is called. When the hw_reset_work workqueue is used, this
problem doesn't arise because ATH_OP_HW_RESET is set
and the ISR bails out.

Set ATH_OP_HW_RESET properly in ath_reset() to avoid
this race - all the ath_reset_internal() callers have
been converted to use ath_reset() in the previous patch.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Unify reset API
Sujith Manoharan [Fri, 17 Oct 2014 02:10:11 +0000 (07:40 +0530)]
ath9k: Unify reset API

Instead of having ath_reset_internal() and ath_reset()
as two separate calls to perform a HW reset, have
one function. This makes sure that the behavior will
be the same at all callsites.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Process beacons properly
Sujith Manoharan [Fri, 17 Oct 2014 02:10:10 +0000 (07:40 +0530)]
ath9k: Process beacons properly

When the current operating channel context has
been marked as ATH_CHANCTX_STATE_FORCE_ACTIVE,
do not process beacons that might be received,
since we have to wait for the station to become
authorized.

Also, since the cached TSF value will be zero
initially do not rearm the timer in this
case when a beacon is received, since it results
in spurious values.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Enable multi-channel properly
Sujith Manoharan [Fri, 17 Oct 2014 02:10:09 +0000 (07:40 +0530)]
ath9k: Enable multi-channel properly

In MCC mode, currently the decision to enable
the multi-channel state machine is done
based on the association status if one of
the interfaces assigned to a context is in
station mode.

This allows the driver to switch to the other
context before the current station is able to
complete the 4-way handshake in case it is
required and this causes problems.

Instead, enable multi-channel mode when the
station moves to the authorized state. This
disallows an early switch to the other channel.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Use sta_state() callback
Sujith Manoharan [Fri, 17 Oct 2014 02:10:08 +0000 (07:40 +0530)]
ath9k: Use sta_state() callback

Instead of using the sta_add()/sta_remove() callbacks,
use the sta_state() callback since this gives
more fine-grained control.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: do not overwrite AR_PHY_RADAR_1 MSB
Lorenzo Bianconi [Tue, 14 Oct 2014 20:21:00 +0000 (22:21 +0200)]
ath9k: do not overwrite AR_PHY_RADAR_1 MSB

Do not overwrite AR_PHY_RADAR_1 most significant byte default value

Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi83@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoipw2x00: clean up a condition
Dan Carpenter [Thu, 9 Oct 2014 09:59:39 +0000 (12:59 +0300)]
ipw2x00: clean up a condition

The original condition was "(PAGE_SIZE - len)" when "(len < PAGE_SIZE)"
is intended.

This condition is not really sufficient, but also not really needed...
If "len > PAGE_SIZE" then it we will print a warning message in dmesg
but there are no other effects.  Maybe we should just remove the
condition?

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoipw2x00: remove unused ->ibss_dfs pointer
Dan Carpenter [Thu, 9 Oct 2014 09:57:42 +0000 (12:57 +0300)]
ipw2x00: remove unused ->ibss_dfs pointer

The ->ibss_dfs pointer is always allocated with a user controlled
length.  This caused a static checker warning because what if the length
was zero?  In that case, any dereference of ->ibss_dfs would lead to an
Oops.

It turns out that this isn't a problem because the ->ibss_dfs pointer is
never used.  This patch deletes it along with all the related code.  In
particular the entire libipw_network_reset() function can be removed.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agort2x00: tune multi-registers I/O timeout
Stanislaw Gruszka [Wed, 8 Oct 2014 12:16:37 +0000 (14:16 +0200)]
rt2x00: tune multi-registers I/O timeout

We provide timeout value to rt2x00usb_vendor_request_buff() based on
number of registers to process. That value is passed down to
rt2x00usb_vendor_req_buff_lock() and ends in usb_control_msg(). But we
do not read/write all registers in rt2x00usb_vendor_req_buff_lock() at
once. We read/write them in chunks of 64 bytes in the loop, hence passed
timeout value to low level is too big.

Patch removes timeout argument from rt2x00usb_vendor_request_buff() and
use short REGISTER_TIMEOUT in rt2x00usb_vendor_req_buff_lock(). That
timeout value should be fine for 64 bytes and smaller requests. For
EEPROM read we introduced new timeout value equal to 2 seconds.

Patch fixes process uninterruptible sleep stalls for long period, when
USB bus has problem to satisfy a request and we wait very long time on
usb_start_wait_urb().

Reported-and-tested-by: Sakari Ailus <sakari.ailus@iki.fi>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agobcma: fill core details for every device
Rafał Miłecki [Fri, 3 Oct 2014 15:00:24 +0000 (17:00 +0200)]
bcma: fill core details for every device

We were setting things like dma_dev, IRQ, etc. during core registration
only. We need such info for cores handled internally (e.g. ChipCommon)
as well.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoMerge tag 'for-linville-20141024' of git://github.com/kvalo/ath
John W. Linville [Mon, 27 Oct 2014 18:09:36 +0000 (14:09 -0400)]
Merge tag 'for-linville-20141024' of git://github.com/kvalo/ath

Conflicts:
drivers/net/wireless/ath/wil6210/wil6210.h

10 years agoiwlwifi: pcie: fix polling in various places
Emmanuel Grumbach [Thu, 23 Oct 2014 05:53:21 +0000 (08:53 +0300)]
iwlwifi: pcie: fix polling in various places

iwl_poll_bit may return a strictly positive value when the
poll doesn't match on the first try.
This was caught when WoWLAN started failing upon resume
even if the poll_bit actually succeeded.

Also change a wrong print. If we reach the end of
iwl_pcie_prepare_card_hw, it means that we couldn't
get the devices.

Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoRevert "iwlwifi: mvm: treat EAPOLs like mgmt frames wrt rate"
Emmanuel Grumbach [Mon, 20 Oct 2014 05:29:55 +0000 (08:29 +0300)]
Revert "iwlwifi: mvm: treat EAPOLs like mgmt frames wrt rate"

This reverts commit aa11bbf3df026d6b1c6b528bef634fd9de7c2619.
This commit was causing connection issues and is not needed
if IWL_MVM_RS_RSSI_BASED_INIT_RATE is set to false by default.

Regardless of the issues mentioned above, this patch added the
following WARNING:

WARNING: CPU: 0 PID: 3946 at drivers/net/wireless/iwlwifi/mvm/tx.c:190 iwl_mvm_set_tx_params+0x60a/0x6f0 [iwlmvm]()
Got an HT rate for a non data frame 0x8
CPU: 0 PID: 3946 Comm: wpa_supplicant Tainted: G           O   3.17.0+ #6
Hardware name: LENOVO 20ANCTO1WW/20ANCTO1WW, BIOS GLET71WW (2.25 ) 07/02/2014
 0000000000000009 ffffffff814fa911 ffff8804288db8f8 ffffffff81064f52
 0000000000001808 ffff8804288db948 ffff88040add8660 ffff8804291b5600
 0000000000000000 ffffffff81064fb7 ffffffffa07b73d0 0000000000000020
Call Trace:
 [<ffffffff814fa911>] ? dump_stack+0x41/0x51
 [<ffffffff81064f52>] ? warn_slowpath_common+0x72/0x90
 [<ffffffff81064fb7>] ? warn_slowpath_fmt+0x47/0x50
 [<ffffffffa07a39ea>] ? iwl_mvm_set_tx_params+0x60a/0x6f0 [iwlmvm]
 [<ffffffffa07a3cf8>] ? iwl_mvm_tx_skb+0x48/0x3c0 [iwlmvm]
 [<ffffffffa079cb9b>] ? iwl_mvm_mac_tx+0x7b/0x180 [iwlmvm]
 [<ffffffffa0746ce9>] ? __ieee80211_tx+0x2b9/0x3c0 [mac80211]
 [<ffffffffa07492f3>] ? ieee80211_tx+0xb3/0x100 [mac80211]
 [<ffffffffa0749c49>] ? ieee80211_subif_start_xmit+0x459/0xca0 [mac80211]
 [<ffffffff814116e7>] ? dev_hard_start_xmit+0x337/0x5f0
 [<ffffffff81430d46>] ? sch_direct_xmit+0x96/0x1f0
 [<ffffffff81411ba3>] ? __dev_queue_xmit+0x203/0x4f0
 [<ffffffff8142f670>] ? ether_setup+0x70/0x70
 [<ffffffff814e96a1>] ? packet_sendmsg+0xf81/0x1110
 [<ffffffff8140625c>] ? skb_free_datagram+0xc/0x40
 [<ffffffff813f7538>] ? sock_sendmsg+0x88/0xc0
 [<ffffffff813f7274>] ? move_addr_to_kernel.part.20+0x14/0x60
 [<ffffffff811c47c2>] ? __inode_wait_for_writeback+0x62/0xb0
 [<ffffffff813f7a91>] ? SYSC_sendto+0xf1/0x180
 [<ffffffff813f88f9>] ? __sys_recvmsg+0x39/0x70
 [<ffffffff8150066d>] ? system_call_fastpath+0x1a/0x1f
---[ end trace cc19a150d311fc63 ]---

which was reported here: https://bugzilla.kernel.org/show_bug.cgi?id=85691

CC: <stable@vger.kernel.org> [3.13+]
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: dvm: drop non VO frames when flushing
Emmanuel Grumbach [Sun, 5 Oct 2014 06:11:14 +0000 (09:11 +0300)]
iwlwifi: dvm: drop non VO frames when flushing

When mac80211 wants to ensure that a frame is sent, it calls
the flush() callback. Until now, iwldvm implemented this by
waiting that all the frames are sent (ACKed or timeout).
In case of weak signal, this can take a significant amount
of time, delaying the next connection (in case of roaming).
Many users have reported that the flush would take too long
leading to the following error messages to be printed:

iwlwifi 0000:03:00.0: fail to flush all tx fifo queues Q 2
iwlwifi 0000:03:00.0: Current SW read_ptr 161 write_ptr 201
iwl data: 00000000: 00 00 00 00 00 00 00 00 fe ff 01 00 00 00 00 00
[snip]
iwlwifi 0000:03:00.0: FH TRBs(0) = 0x00000000
[snip]
iwlwifi 0000:03:00.0: Q 0 is active and mapped to fifo 3 ra_tid 0x0000 [9,9]
[snip]

Instead of waiting for these packets, simply drop them. This
significantly improves the responsiveness of the network.
Note that all the queues are flushed, but the VO one. This
is not typically used by the applications and it likely
contains management frames that are useful for connection
or roaming.

This bug is tracked here:
https://bugzilla.kernel.org/show_bug.cgi?id=56581

But it is duplicated in distributions' trackers.
A simple search in Ubuntu's database led to these bugs:

https://bugs.launchpad.net/ubuntu/+source/linux-firmware/+bug/1270808
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1305406
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1356236
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1360597
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1361809

Cc: <stable@vger.kernel.org>
Depends-on: 77be2c54c5bd ("mac80211: add vif to flush call")
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: ROC - bug fixes around time events and locking
Matti Gottlieb [Mon, 29 Sep 2014 08:46:04 +0000 (11:46 +0300)]
iwlwifi: mvm: ROC - bug fixes around time events and locking

Don't add the time event to the list. We added it several
times the same time event, which leads to an infinite loop
when walking the list.

Since we (currently) don't support more than one ROC for STA
vif at a time, enforce this and don't add the time event
to any list.

We were also missing the locking of the mutex which led to
a lockdep splat - fix that.

Signed-off-by: Matti Gottlieb <matti.gottlieb@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: Add tx power condition to bss_info_changed_ap_ibss
Haim Dreyfuss [Sun, 14 Sep 2014 09:40:00 +0000 (12:40 +0300)]
iwlwifi: mvm: Add tx power condition to bss_info_changed_ap_ibss

The tx power should be limited from many reasons.
currently, setting the tx power is available by the mvm only for
station interface. Adding the tx power condition to
bss_info_changed_ap_ibss make it available also for AP.

Signed-off-by: Haim Dreyfuss <haim.dreyfuss@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: BT coex - fix BT prio for probe requests
Emmanuel Grumbach [Mon, 22 Sep 2014 09:03:41 +0000 (12:03 +0300)]
iwlwifi: mvm: BT coex - fix BT prio for probe requests

The probe requests sent during scan must get BT prio 3.
Fix that.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: BT Coex - update the MPLUT Boost register value
Emmanuel Grumbach [Mon, 22 Sep 2014 13:12:24 +0000 (16:12 +0300)]
iwlwifi: mvm: BT Coex - update the MPLUT Boost register value

Cc: <stable@vger.kernel.org> [3.16+]
Fixes: 2adc8949efab ("iwlwifi: mvm: BT Coex - fix boost register / LUT values")
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: 8000: fix string given to MODULE_FIRMWARE
Liad Kaufman [Tue, 23 Sep 2014 12:15:17 +0000 (15:15 +0300)]
iwlwifi: 8000: fix string given to MODULE_FIRMWARE

I changed the string but forgot to update the fix also to
MODULE_FIRMWARE().

Signed-off-by: Liad Kaufman <liad.kaufman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: configure the LTR
Emmanuel Grumbach [Tue, 23 Sep 2014 20:02:41 +0000 (23:02 +0300)]
iwlwifi: configure the LTR

The LTR is the handshake between the device and the root
complex about the latency allowed when the bus exits power
save. This configuration was missing and this led to high
latency in the link power up. The end user could experience
high latency in the network because of this.

Cc: <stable@vger.kernel.org> [3.10+]
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoath10k: split ce pipe init/alloc further
Michal Kazior [Mon, 20 Oct 2014 12:14:39 +0000 (14:14 +0200)]
ath10k: split ce pipe init/alloc further

Calling init to reinit ce pipe state would also
re-set all static structure links and setting
(which don't change over driver lifecycle).

Make it so alloc links structures and initializes
static data and init part to setup state
variables and clear stuff.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: mask/unmask msi fw irq
Michal Kazior [Mon, 20 Oct 2014 12:14:38 +0000 (14:14 +0200)]
ath10k: mask/unmask msi fw irq

This was the final missing bit to making sure the
device doesn't assert interrupts to host.

This should fix possible race when target crashes
during driver teardown.

This also removes an early warm reset workaround
during pci probing.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: re-disable interrupts after target init
Michal Kazior [Mon, 20 Oct 2014 12:14:37 +0000 (14:14 +0200)]
ath10k: re-disable interrupts after target init

If MSI isn't configured device ROM program expects
legacy interrupts to be enabled before it can
fully boot. Don't forget to disable legacy
interrupts after that.

While at it re-use the legacy irq enabling helper
instead of calling ath10k_pci_write32().

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: use bss_info as txpower source
Michal Kazior [Tue, 21 Oct 2014 07:40:15 +0000 (10:40 +0300)]
ath10k: use bss_info as txpower source

This simply changes the source for txpower setup.
It does not make ath10k use different txpower
values for different vifs.

This will make it easier to implement chanctx in
ath10k in the future.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: simplify computation of mgmt rx band
Michal Kazior [Tue, 21 Oct 2014 07:10:30 +0000 (10:10 +0300)]
ath10k: simplify computation of mgmt rx band

Using global channel won't work with chanctx. Try
to determine the channel from the information
provided in the wmi event itself alone. This
should be sufficient.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: clean up sta auth/assoc code
Michal Kazior [Tue, 21 Oct 2014 07:10:29 +0000 (10:10 +0300)]
ath10k: clean up sta auth/assoc code

The code can be symmetrical so make it so. This
makes it easier to understand and work with.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: skip some commands on reassoc
Michal Kazior [Tue, 21 Oct 2014 07:10:29 +0000 (10:10 +0300)]
ath10k: skip some commands on reassoc

It doesn't make much sense to reconfigure peer
completely upon reassociation. This will make it
easier to have a more uniform association code
across different modes.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: clean up assoc code
Michal Kazior [Tue, 21 Oct 2014 07:10:29 +0000 (10:10 +0300)]
ath10k: clean up assoc code

There's no need to pass bss_conf explicitly as it
is accessible via vif pointer. This requires
slight changes in function prototypes. While at it
clean up listen interval workaround/command.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: retrieve calibration data from file
Kalle Valo [Mon, 13 Oct 2014 06:40:59 +0000 (09:40 +0300)]
ath10k: retrieve calibration data from file

A frequent request have been to be able to provide calibration data from a
file as some of the AP devices store the calibration data on an MTD partition.
This patchset adds support for that and also makes it easier to add Device Tree
support later on.

The calibration data is found by using the id string provided by dev_name()
using this format:

cal-<bus>-<id>.bin

With PCI the id string contains bus, slot and func values. For example for a
PCI device in bus 2 slot 0, ath10k will try to retrieve a calibration data from
a file:

/lib/firmware/ath10k/cal-pci-0000:02:00.0.bin

The calibration data sequence is:

1. Check with request_firmware() if there's a calibration file
   ("cal-<bus>-<id>.bin") on the filesystem for this device. If yes, use that. If
   not, goto 2

2. Check if otp.bin is able to successfully load the calibration data
   from OTP. If yes, use that. If not, goto 3.

4. Print an error message that no calibration data found and stop driver
   initialization for this device.

Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: refactor ath10k_init_download_firmware()
Kalle Valo [Mon, 13 Oct 2014 06:40:53 +0000 (09:40 +0300)]
ath10k: refactor ath10k_init_download_firmware()

This is preparation for being able to download calibration data from a file.

Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoath10k: add back enum ath10k_bus
Kalle Valo [Mon, 13 Oct 2014 06:40:47 +0000 (09:40 +0300)]
ath10k: add back enum ath10k_bus

Commit 3a0861fffd223 ("ath10k: remove ath10k_bus") removed enum ath10k_bus
because it was not used for anything at the time. But now it's needed for for
retrieving the right calibration data file so add it back. Only new addition is
ath10k_bus_str().

Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
10 years agoMerge tag 'ntb-3.18' of git://github.com/jonmason/ntb
Linus Torvalds [Sun, 19 Oct 2014 19:58:22 +0000 (12:58 -0700)]
Merge tag 'ntb-3.18' of git://github.com/jonmason/ntb

Pull ntb (non-transparent bridge) updates from Jon Mason:
 "Add support for Haswell NTB split BARs, a debugfs entry for basic
  debugging info, and some code clean-ups"

* tag 'ntb-3.18' of git://github.com/jonmason/ntb:
  ntb: Adding split BAR support for Haswell platforms
  ntb: use errata flag set via DID to implement workaround
  ntb: conslidate reading of PPD to move platform detection earlier
  ntb: move platform detection to separate function
  NTB: debugfs device entry

10 years agoMerge branch 'i2c/for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sun, 19 Oct 2014 19:50:44 +0000 (12:50 -0700)]
Merge branch 'i2c/for-next' of git://git./linux/kernel/git/wsa/linux

Pull i2c updates from Wolfram Sang:
 "Highlights from the I2C subsystem for 3.18:

   - new drivers for Axxia AM55xx, and Hisilicon hix5hd2 SoC.

   - designware driver gained AMD support, exynos gained exynos7 support

  The rest is usual driver stuff.  Hopefully no lowlights this time"

* 'i2c/for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: i801: Add Device IDs for Intel Sunrise Point PCH
  i2c: hix5hd2: add i2c controller driver
  i2c-imx: Disable the clock on probe failure
  i2c: designware: Add support for AMD I2C controller
  i2c: designware: Rework probe() to get clock a bit later
  i2c: designware: Default to fast mode in case of ACPI
  i2c: axxia: Add I2C driver for AXM55xx
  i2c: exynos: add support for HSI2C module on Exynos7
  i2c: mxs: detect No Slave Ack on SELECT in PIO mode
  i2c: cros_ec: Remove EC_I2C_FLAG_10BIT
  i2c: cros-ec-tunnel: Add of match table
  i2c: rcar: remove sign-compare flaw
  i2c: ismt: Use minimum descriptor size
  i2c: imx: Add arbitration lost check
  i2c: rk3x: Remove unlikely() annotations
  i2c: rcar: check for no IRQ in rcar_i2c_irq()
  i2c: rcar: make rcar_i2c_prepare_msg() *void*
  i2c: rcar: simplify check for last message
  i2c: designware: add support of platform data to set I2C mode
  i2c: designware: add support of I2C standard mode

10 years agoMerge tag 'sound-fix-3.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Sun, 19 Oct 2014 19:45:36 +0000 (12:45 -0700)]
Merge tag 'sound-fix-3.18-rc1' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Here are a collection of small fixes after 3.18 merge.

  The urgent one is the fix for kernel panics with linked PCM substream
  triggered by the recent nonatomic PCM ops support.  Other two fixes
  (emu10k1 and bebob) are stable fixes, and one easy PCI ID addition for
  a new Intel HD-audio controller"

* tag 'sound-fix-3.18-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda_intel: Add Device IDs for Intel Sunrise Point PCH
  ALSA: emu10k1: Fix deadlock in synth voice lookup
  ALSA: pcm: Fix referred substream in snd_pcm_action_group() unlock loop
  ALSA: bebob: Fix failure to detect source of clock for Terratec Phase 88

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 19 Oct 2014 19:40:24 +0000 (12:40 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull second round of input updates from Dmitry Torokhov:
 "Mostly simple bug fixes, although we do have one brand new driver for
  Microchip AR1021 i2c touchscreen.

  Also there is the change to stop trying to use i8042 active
  multiplexing by default (it is still possible to activate it via
  i8042.nomux=0 on boxes that implement it)"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: xpad - add Thrustmaster as Xbox 360 controller vendor
  Input: xpad - add USB ID for Thrustmaster Ferrari 458 Racing Wheel
  Input: max77693-haptic - fix state check in imax77693_haptic_disable()
  Input: xen-kbdfront - free grant table entry in xenkbd_disconnect_backend
  Input: alps - fix v4 button press recognition
  Input: i8042 - disable active multiplexing by default
  Input: i8042 - add noloop quirk for Asus X750LN
  Input: synaptics - gate forcepad support by DMI check
  Input: Add Microchip AR1021 i2c touchscreen
  Input: cros_ec_keyb - add of match table
  Input: serio - avoid negative serio device numbers
  Input: avoid negative input device numbers
  Input: automatically set EV_ABS bit in input_set_abs_params
  Input: adp5588-keys - cancel workqueue in failure path
  Input: opencores-kbd - switch to using managed resources
  Input: evdev - fix EVIOCG{type} ioctl

10 years agoMerge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Sun, 19 Oct 2014 19:29:23 +0000 (12:29 -0700)]
Merge tag 'rdma-for-linus' of git://git./linux/kernel/git/roland/infiniband

Pull infiniband/RDMA updates from Roland Dreier:
 - large set of iSER initiator improvements
 - hardware driver fixes for cxgb4, mlx5 and ocrdma
 - small fixes to core midlayer

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (47 commits)
  RDMA/cxgb4: Fix ntuple calculation for ipv6 and remove duplicate line
  RDMA/cxgb4: Add missing neigh_release in find_route
  RDMA/cxgb4: Take IPv6 into account for best_mtu and set_emss
  RDMA/cxgb4: Make c4iw_wr_log_size_order static
  IB/core: Fix XRC race condition in ib_uverbs_open_qp
  IB/core: Clear AH attr variable to prevent garbage data
  RDMA/ocrdma: Save the bit environment, spare unncessary parenthesis
  RDMA/ocrdma: The kernel has a perfectly good BIT() macro - use it
  RDMA/ocrdma: Don't memset() buffers we just allocated with kzalloc()
  RDMA/ocrdma: Remove a unused-label warning
  RDMA/ocrdma: Convert kernel VA to PA for mmap in user
  RDMA/ocrdma: Get vlan tag from ib_qp_attrs
  RDMA/ocrdma: Add default GID at index 0
  IB/mlx5, iser, isert: Add Signature API additions
  Target/iser: Centralize ib_sig_domain setting
  IB/iser: Centralize ib_sig_domain settings
  IB/mlx5: Use extended internal signature layout
  IB/iser: Set IP_CSUM as default guard type
  IB/iser: Remove redundant assignment
  IB/mlx5: Use enumerations for PI copy mask
  ...

10 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 19 Oct 2014 18:55:41 +0000 (11:55 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull more perf updates from Ingo Molnar:
 "A second (and last) round of late coming fixes and changes, almost all
  of them in perf tooling:

  User visible tooling changes:

   - Add period data column and make it default in 'perf script' (Jiri
     Olsa)

   - Add a visual cue for toggle zeroing of samples in 'perf top'
     (Taeung Song)

   - Improve callchains when using libunwind (Namhyung Kim)

  Tooling fixes and infrastructure changes:

   - Fix for double free in 'perf stat' when using some specific invalid
     command line combo (Yasser Shalabi)

   - Fix off-by-one bugs in map->end handling (Stephane Eranian)

   - Fix off-by-one bug in maps__find(), also related to map->end
     handling (Namhyung Kim)

   - Make struct symbol->end be the first addr after the symbol range,
     to make it match the convention used for struct map->end.  (Arnaldo
     Carvalho de Melo)

   - Fix perf_evlist__add_pollfd() error handling in 'perf kvm stat
     live' (Jiri Olsa)

   - Fix python test build by moving callchain_param to an object linked
     into the python binding (Jiri Olsa)

   - Document sysfs events/ interfaces (Cody P Schafer)

   - Fix typos in perf/Documentation (Masanari Iida)

   - Add missing 'struct option' forward declaration (Arnaldo Carvalho
     de Melo)

   - Add option to copy events when queuing for sorting across cpu
     buffers and enable it for 'perf kvm stat live', to avoid having
     events left in the queue pointing to the ring buffer be rewritten
     in high volume sessions.  (Alexander Yarygin, improving work done
     by David Ahern):

   - Do not include a struct hists per perf_evsel, untangling the
     histogram code from perf_evsel, to pave the way for exporting a
     minimalistic tools/lib/api/perf/ library usable by tools/perf and
     initially by the rasd daemon being developed by Borislav Petkov,
     Robert Richter and Jean Pihet.  (Arnaldo Carvalho de Melo)

   - Make perf_evlist__open(evlist, NULL, NULL), i.e. without cpu and
     thread maps mean syswide monitoring, reducing the boilerplate for
     tools that only want system wide mode.  (Arnaldo Carvalho de Melo)

   - Move exit stuff from perf_evsel__delete to perf_evsel__exit, delete
     should be just a front end for exit + free (Arnaldo Carvalho de
     Melo)

   - Add support to new style format of kernel PMU event.  (Kan Liang)

  and other misc fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (45 commits)
  perf script: Add period as a default output column
  perf script: Add period data column
  perf evsel: No need to drag util/cgroup.h
  perf evlist: Add missing 'struct option' forward declaration
  perf evsel: Move exit stuff from __delete to __exit
  kprobes/x86: Remove stale ARCH_SUPPORTS_KPROBES_ON_FTRACE define
  perf kvm stat live: Enable events copying
  perf session: Add option to copy events when queueing
  perf Documentation: Fix typos in perf/Documentation
  perf trace: Use thread_{,_set}_priv helpers
  perf kvm: Use thread_{,_set}_priv helpers
  perf callchain: Create an address space per thread
  perf report: Set callchain_param.record_mode for future use
  perf evlist: Fix for double free in tools/perf stat
  perf test: Add test case for pmu event new style format
  perf tools: Add support to new style format of kernel PMU event
  perf tools: Parse the pmu event prefix and suffix
  Revert "perf tools: Default to cpu// for events v5"
  perf Documentation: Remove Ruplicated docs for powerpc cpu specific events
  perf Documentation: sysfs events/ interfaces
  ...

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc
Linus Torvalds [Sun, 19 Oct 2014 18:46:09 +0000 (11:46 -0700)]
Merge git://git./linux/kernel/git/davem/sparc

Pull sparc fixes from David Miller:
 "Here we have two bug fixes:

  1) The current thread's fault_code is not setup properly upon entry to
     do_sparc64_fault() in some paths, leading to spurious SIGBUS.

  2) Don't use a zero length array at the end of thread_info on sparc64,
     otherwise end_of_stack() isn't right"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc64: Do not define thread fpregs save area as zero-length array.
  sparc64: Fix corrupted thread fault code.

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 19 Oct 2014 18:41:57 +0000 (11:41 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:
 "A quick batch of bug fixes:

  1) Fix build with IPV6 disabled, from Eric Dumazet.

  2) Several more cases of caching SKB data pointers across calls to
     pskb_may_pull(), thus referencing potentially free'd memory.  From
     Li RongQing.

  3) DSA phy code tests operation presence improperly, instead of going:

        if (x->ops->foo)
                r = x->ops->foo(args);

     it was going:

        if (x->ops->foo(args))
                r = x->ops->foo(args);

   Fix from Andew Lunn"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  Net: DSA: Fix checking for get_phy_flags function
  ipv6: fix a potential use after free in sit.c
  ipv6: fix a potential use after free in ip6_offload.c
  ipv4: fix a potential use after free in gre_offload.c
  tcp: fix build error if IPv6 is not enabled