Taketo Kabe [Sun, 13 May 2018 09:16:40 +0000 (18:16 +0900)]
b43: fix transmit failure when VT is switched
Setup:
Using BCM4306 rev.03 chip based CardBus wireless card.
IRQ is shared with yenta (cardbus bridge) and i915 (display) driver.
For firmware, installed latest but dated openfwwf 5.2
(http://netweb.ing.unibs.it/~openfwwf/)
How-to-reproduce:
Do "ssh <NetBSD-remotehost>", then "ls -lR /" to generate traffic, then
repeatedly switch VTs by Alt-F1<>Alt-F2.
Eventually (within a minute) the card stops working.
You can receive traffic but no transmission.
For unknown reason it doesn't occur when just generating traffic by
"ssh <remotehost> ls -lR /".
With CONFIG_B43_DEBUG=y kernel config, when it stops,
the debug message shows
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 148, but got 180
The slot offset I observed so far was always 32.
When err_out2 is not set to make error messages successive,
the debug output will be like this:
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 148
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 150
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 120
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 152
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 122
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 154
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 124
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 156
kernel: b43-phy1 debug: Out of order TX status report on DMA ring 1. Expected 116, but got 126
The TX ring alternates between 2 sequences; the ring seems
to be completely confused. Controller restart is needed.
Workaround(1):
This problem doesn't occur when using propriatory firmware
you will extract by b43-fwcutter, so it may be a bug in
openfwwf firmware, as the comment in the b43_dma_handle_txstatus() suggests.
I wasn't able to find a bug in the terse openfwwf code though.
Workaround(2):
Using "pio=1" option to not use DMA makes this problem to
not occur.
Description of the patch:
This patch will forcibly reset the controller to make it
work again. Very kludgy and doesn't look right, but
the traffic will continue to flow.
Siva Rebbagondla [Fri, 11 May 2018 14:57:53 +0000 (20:27 +0530)]
rsi: Set wowlan flag while writing wowlan config parameters
As wowlan enable flag did not set, while writing wowlan parameters to
card using rsi_send_vap_dynamic_update, which results firmware is unable to
set wowlan configurations. Hence, setting wowlan flag before sending
parameters.
Siva Rebbagondla [Fri, 11 May 2018 14:57:52 +0000 (20:27 +0530)]
rsi: reset hibernate_resume flag to work hibernate resume in coex mode.
In coex mode, observed hibernate resume is not working properly, as the
hibernate_resume flag is not getting reset in rsi_coex_recv_pkt(),
when common card ready indication received from firmware. Hence resetting
hibernate_resume flag in this function.
rsi: Add null check for virtual interfaces in wowlan config
When the "poweroff" command is executed after wowlan enabled, we have
observed a system crash. In the system "poweroff" sequence, network-manager
is sent to inactive state by cleaning up the network interfaces, using
rsi_mac80211_remove_interface() and when driver tries to access those
network interfaces in rsi_wowlan_config() which was invoked by SDIO
shutdown, results in a crash. Added a NULL check before accessing the
network interfaces in rsi_wowlan_config().
Rafał Miłecki [Thu, 10 May 2018 13:21:39 +0000 (15:21 +0200)]
brcmfmac: set WIPHY_FLAG_HAVE_AP_SME flag
brcmfmac is a FullMAC driver and it implements/uses cfg80211 interface
for stations management. At the same time it doesn't receive or pass up
management frames.
This flag indicates that authenticator doesn't have to subscribe to or
handle management frames. Some authenticators (e.g. hostapd) were
working with brcmfmac thanks to some extra assumptions. This commit
clears up the situation.
Since the addition of the TXQ stats to cfg80211, the station_info struct
has grown to be quite large, which results in warnings when allocated on
the stack. Fix the affected places to do dynamic allocations instead.
Fixes: 52539ca89f36 ("cfg80211: Expose TXQ stats and parameters to userspace") Reviewed-by: Sergey Matyukevich <[email protected]> Signed-off-by: Toke Høiland-Jørgensen <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
Ganapathi Bhat [Mon, 7 May 2018 18:40:58 +0000 (00:10 +0530)]
mwifiex: increase TX threashold to avoid TX timeout during ED MAC test
While carrying energy detection(ED) tests, the chip will stop
transmission upon detecting an energy in the connected channel.
As a feedback, driver will stop dequeuing TX packets and due to
which wmm_tx_pending keep incremeting. Once wmm_tx_pending
reaches 100, driver calls netif_tx_stop_queue(). If TX ques is
not restarted within 5(watchdog_timeo) seconds, it will result in
TX timeout.
The ED test is carried out for one minute and the current
threshold of 100 is easily overcome by the traffic, cuasing TX
timeouts. To fix this increase the threshold to 400.
Sean Lanigan [Fri, 4 May 2018 06:48:23 +0000 (16:48 +1000)]
brcmfmac: Add support for bcm43364 wireless chipset
Add support for the BCM43364 chipset via an SDIO interface, as used in
e.g. the Murata 1FX module.
The BCM43364 uses the same firmware as the BCM43430 (which is already
included), the only difference is the omission of Bluetooth.
However, the SDIO_ID for the BCM43364 is 02D0:A9A4, giving it a MODALIAS
of sdio:c00v02D0dA9A4, which doesn't get recognised and hence doesn't
load the brcmfmac module. Adding the 'A9A4' ID in the appropriate place
triggers the brcmfmac driver to load, and then correctly use the
firmware file 'brcmfmac43430-sdio.bin'.
We have check if queue is not empty when start flushing queues on
by mac80211 callback, but we also can start flushing queues by internal
driver calls. So move check into rt2x00queue_flush_queue() to assure
we do not flush empty queue anytime.
Additionally add warning if we start to kick empty queue as in such
situation we set wrong index in the HW queue, what can confuse the HW
and have various negative consequences.
Franky Lin [Thu, 26 Apr 2018 10:16:51 +0000 (12:16 +0200)]
brcmfmac: coarse support for PCIe shared structure rev7
Revision 7 of PCIe dongle interface increases the item size of tx and rx
complete rings to accommodate extra payload for new feature. This patch
simply bump up the size of these two rings without adding the support
for utilizing the new space. This makes brcmfmac compatible with rev7
firmware.
Franky Lin [Thu, 26 Apr 2018 10:16:48 +0000 (12:16 +0200)]
brcmfmac: use nl80211_band directly to get ieee80211 channel
The enum nl80211_band used to retrieve wiphy->bands is the same as
wiphy->bands->band which is checked by wiphy_register(). So it can be used
directly as parameter of ieee80211_channel_to_frequency().
Franky Lin [Thu, 26 Apr 2018 10:18:35 +0000 (12:18 +0200)]
brcmfmac: reports boottime_ns while informing bss
Provides a timestamp in bss information so user space can see when the
bss info was updated. Since tsf is not available from the dongle events
boottime is reported instead.
The mac address for p2pdev must be different from the primary interface
due to firmware requirement. Add an explicit check for this requirement
if user-space provides a mac address.
Function rsi_hal_key_config returns before reaching code at line
922 if (status), hence this code is structurally dead.
Fix this by storing the value returned by rsi_hal_load_key
into _status_ for its further evaluation and use.
Addresses-Coverity-ID: 1468409 ("Structurally dead code") Fixes: 4fd6c4762f37 ("rsi: roaming enhancements") Signed-off-by: Gustavo A. R. Silva <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
pm_runtime handles sdio power on and power off transitions.
An old workaround for trying to control the power explicitly from the
driver was in fact causing failures on suspend/resume as the mmc layer
already power the module on resume.
In case of resume pm_runtime_get sync returns a positive device's usage
count causing the driver to try an re-initialize an already initialized
device. This was causing sdio bus failure on resume.
Remove this manual power on/off sequence as it is in-fact not needed.
qtnfmac: fix qtnf_netdev_hard_start_xmit()'s return type
The method ndo_start_xmit() is defined as returning an 'netdev_tx_t',
which is a typedef for an enum type, but the implementation in this
driver returns an 'int'.
Fix this by returning 'netdev_tx_t' in this driver too.
Jia-Ju Bai [Tue, 10 Apr 2018 13:54:19 +0000 (21:54 +0800)]
net: wireless: b43legacy: Replace GFP_ATOMIC with GFP_KERNEL in dma_tx_fragment
dma_tx_fragment() is never called in atomic context.
dma_tx_fragment() is only called by b43legacy_dma_tx(), which is
only called by b43legacy_tx_work().
b43legacy_tx_work() is only set a parameter of INIT_WORK() in
b43legacy_wireless_init().
Despite never getting called from atomic context,
dma_tx_fragment() calls alloc_skb() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
This is found by a static analysis tool named DCNS written by myself.
And I also manually check it.
mwifiex: fix mwifiex_hard_start_xmit()'s return type
The method ndo_start_xmit() is defined as returning an 'netdev_tx_t',
which is a typedef for an enum type, but the implementation in this
driver returns an 'int'.
Fix this by returning 'netdev_tx_t' in this driver too.
Felix Fietkau [Wed, 25 Apr 2018 09:11:22 +0000 (11:11 +0200)]
mt76: add rcu locking in tid reorder function
Avoids having the tid or station entry disappear prematurely.
Also cancel the reorder work earlier to avoid further processing delayed
by waiting for the lock to be released
Felix Fietkau [Wed, 25 Apr 2018 09:11:21 +0000 (11:11 +0200)]
mt76: fix concurrent rx calls on A-MPDU release
Add a spinlock in mt76_rx_complete. Without this, multiple stats updates
could happen in parallel, which can lead to deadlocks. There are
probably more corner cases fixed by this change.
Lorenzo Bianconi [Tue, 17 Apr 2018 22:27:18 +0000 (00:27 +0200)]
mt76x2: fix is_mt7612 routine
Fix is_mt7612 routine since asic version is set in mt76_dev revision
field and not in mt76x2_dev one.
Moreover remove mt76x2_dev rev field since it is never used in the
driver
Kalle Valo [Mon, 30 Apr 2018 10:02:30 +0000 (13:02 +0300)]
Merge tag 'iwlwifi-next-for-kalle-2018-04-26' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-next
Second batch of iwlwifi patches for 4.18
* Some preparations for new hardware;
* A workaround to rescan the bus for the rare situation when older
devices become irresponsive,
* Hardening of the firmware loading code to avoid issues with
corrupted files;
* A few clean-ups and bugfixes.
Eliad Peller [Thu, 22 Feb 2018 16:45:49 +0000 (18:45 +0200)]
iwlwifi: mvm: set wakeup filters for wowlan "any" configuration
In case of "any" wowlan trigger is configured, no valid wakeup filter
was configured.
Moreover, the fw assumes there's no connection when there are no configured
wakeup filters.
This leads to the station info not being updated on D3 command, causing
rate_n_flags to be 0 when the offloading code sends tx frame (triggering
SYSASSERT_102C due to invalid antenna param)
Note: "any" trigger is currently assumed to only be used when entering
d0i3 (which has a different flow). However, we still reach this code
when using d3_test.
Luca Coelho [Thu, 22 Feb 2018 12:50:06 +0000 (14:50 +0200)]
iwlwifi: make bitfield a u32 instead of u16
The bitfield we use in struct iwl_cfg contains 17 bits, but we declare
it as u16. The compiler doesn't seem to have any problems with that
(it probably automatically makes it u32), but it's cleaner to use a
type that is long enough for all the flags.
Luca Coelho [Tue, 12 Dec 2017 06:58:41 +0000 (08:58 +0200)]
iwlwifi: pcie: remove non-responsive device
If we fail to to grab NIC access because the device is not responding
(i.e. CSR_GP_CNTRL returns 0xFFFFFFFF), remove the device from the PCI
bus, to avoid any further damage, and to let the user space rescan.
In order to inform the userspace that a rescan is needed, we send a
kobject uevent with "INACCESSIBLE".
This functionality is disabled by default, but can be enabled via a
new module parameter called "remove_when_gone". In the future we may
change this module parameter to include 3 modes instead: do nothing;
auto-rescan or; send uevent.
Luca Coelho [Wed, 14 Feb 2018 06:54:40 +0000 (08:54 +0200)]
iwlwifi: fw: combine loading of last page block into main copy loop
Now that we check and copy only the actual size of the page block,
there is no need to treat the last block separately. Remove the
mostly duplicate code and make the main copy loop handle also the last
block.
Luca Coelho [Tue, 13 Feb 2018 09:09:40 +0000 (11:09 +0200)]
iwlwifi: fw: harden page loading code
The page loading code trusts the data provided in the firmware images
a bit too much and may cause a buffer overflow or copy unknown data if
the block sizes don't match what we expect.
To prevent potential problems, harden the code by checking if the
sizes we are copying are what we expect.
Future changes will require moving the HE radiotap data into
the SKB head, but this means we need to have the alignment
reservation done before that. To prepare, move the alignment
reservation earlier here.
Luca Coelho [Tue, 13 Feb 2018 11:52:43 +0000 (13:52 +0200)]
iwlwifi: cfg: remove unnecessary cfg data in non-dvm devices
The max_data_size and max_inst_size values are only needed for DVM
devices. Remove the assignment to those fields in 7000 and newer
families so we can also remove the otherwise unnecessary inclusion of
iwl-agn.h headers.
iwlwifi: introduce Image Loader (IML) - new firmware image
In future devices a new image will be introduced - IML. The IML, image
loader, is loaded by the ROM, and as part of the new self-init flow,
loads the rest of the firmware images to the device.
Store the image, so the ROM can load it to the device.
Golan Ben Ami [Sun, 22 Oct 2017 12:58:26 +0000 (15:58 +0300)]
iwlwifi: allow different csr flags for different device families
Different device families may have different flag values
for passing a message to the fw (i.e. SW_RESET).
In order to keep the code readable, and avoid conditioning
upon the family, store a value for each flag, which indicates
the bit that needs to be enabled.
RT5370G has hardware RX antenna diversity like RT5390R.
Based on DPO_RT5572_LinuxSTA_2.6.1.3_20121022.tar.bz2 manufacturer driver:
https://d86o2zu8ugzlg.cloudfront.net/mediatek-craft/drivers/DPO_RT5572_LinuxSTA_2.6.1.3_20121022.tar.bz2
Remove and refactor some code in order to avoid having identical code
for different branches.
Notice that this piece of code hasn't been modified since 2011.
Addresses-Coverity-ID: 1226756 ("Identical code for different branches") Signed-off-by: Gustavo A. R. Silva <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
Dan Carpenter [Thu, 5 Apr 2018 11:17:19 +0000 (14:17 +0300)]
mwifiex: pcie: tighten a check in mwifiex_pcie_process_event_ready()
If "evt_len" is 1 then we try to memcpy() negative 3 bytes and it would
cause memory corruption.
Fixes: d930faee141b ("mwifiex: add support for Marvell pcie8766 chipset") Signed-off-by: Dan Carpenter <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
Lorenzo Bianconi [Mon, 16 Apr 2018 23:05:59 +0000 (01:05 +0200)]
mt76x2: fix TXD_INFO bitmask definition
Despite this issue is not harmful since it is not actually used in
mt76 driver, fix DMA txinfo bitmask definition
Fixes: 7bc04215a66b ('mt76: add driver code for MT76x2e') Signed-off-by: Lorenzo Bianconi <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
mt7601u: run calibration works after finishing scanning
When finishing scanning we switch to operational channel sill with
SCANNING flag. This mean that we never perform calibration works after
scanning. To fix the problem queue calibration works on
.sw_scan_complete() routine.
avr_rssi is not calculated correctly as we do not divide result
by 256 (mt76 sum avg_rssi1 and avg_rssi2 and divide by 512).
However dividing by 256 will make avg_rssi almost the same as
last rssi value - not really an average. So use EWMA to calculate
avg_rssi. I've chosen weight_rcp=4 to convergence quicker on signal
strength changes.
Fix mt76x2_temp_tx_alc_enabled routine since in order to enable tx_alc
temperature compensation it necessary to take into account BIT(15) of
MT_EE_TX_POWER_EXT_PA_5G eeprom info
Fixes: 7bc04215a66b ("mt76: add driver code for MT76x2e") Signed-off-by: Lorenzo Bianconi <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
Felix Fietkau [Tue, 3 Apr 2018 19:52:51 +0000 (21:52 +0200)]
mt76: rework tx power handling
There were a number of issues in the existing tx power handling code:
1. The EEPROM target power for chain 0 refers to the actual output power
after the channel and bandwidth delta have been added to the initial
channel gain. This means the delta values should not be added/subtracted
for the rate power calculation
2. When power is reduced significantly, the initial channel gain
underflows very quickly, while the per-rate power offsets are high.
This miscalculation causes effective tx power to be increased when it
should actually be lowered.
Fix this by trying to adjust the channel gain to the lowest power from
the rate table. In case of under- or overflow, compensate by adding an
appropriate delta to the rate power values.
This makes power configuration more accurate on a wider range of
configurable values
Felix Fietkau [Tue, 3 Apr 2018 19:52:48 +0000 (21:52 +0200)]
mt76: stop tx queues from the driver callback instead of common code
Allows the driver to control whether to send a BlockAckReq after waking
up. MT76x2 needs this set to true, for MT7603 (to be submitted later)
it needs to be false.
rtlwifi: btcoex: remove identical statements within if-else branches
Since the statements are identical, we can safely remove the statements.
The commit 42e74946f016 ("rtlwifi: btcoexist: Fix if == else warnings in
halbtc8821a1ant.c") had fixed the statements, but the commit c6821613e653
("rtlwifi: btcoex: follow linux coding style") that converted coding style
and upgraded btcoex didn't include the fix.
While performing cleanup, driver is messing with card->ocr
value by not masking rocr against ocr_avail. Below panic
is observed with some of the SDIO host controllers due to
this. Issue is resolved by reverting incorrect modifications
to vdd.
Following overflow kernel panic is observed on some platforms while
loading the driver. It is fixed if dynamically allocated memory is
passed to SDIO instead of static one
Currently TID is extracted by checking at specific offset in data frame.
This approach doesn't work for some of the frames. This patch uses mac80211
API and do it correctly
rsi: use appropriate interface for power save configuration
Power save request should be sent on station interface. Virtual
interface which is connected should be preferred. This patch
resolves device not entering power save problem in certain
situations
Dan Carpenter [Fri, 6 Apr 2018 08:37:17 +0000 (11:37 +0300)]
rsi: remove unecessary PTR_ALIGN()s
The issue here is that we allocate "data" and then set
"data = PTR_ALIGN(data, 8);" and then we free the aligned pointer
instead of the original pointer.
kmalloc() pointers are already ARCH_SLAB_MINALIGN aligned which is 8 or
more on everything except certain Xtensa variants. We decided that if
the Xtensa people ever notice a bug here then we'll tell them the bug is
on their side. ;)
To support roaming below changes are done:
* Station notify frame is send to firmware after sending assoc
request. This will avoid dropping of first EAPOL frame due to
delay in creation of station control block in firmware.
* Data queues are unblocked after sending station notify in open
mode, after configuring key in WEP mode, and after receiving
EAPOL4 confirm in WPA mode.
* Initial EAPOL frames priority is chaged to MGMT, rekey EAPOL
frames priority changed to VO.
rsi: move xtend_desc structure from rsi_main.h to rsi_mgmt.h
All descriptor structures are in rsi_mgmt.h except this
extended descriptor structure. Hence moving it to rsi_mgmt.h
and also renaming to rsi_xtend_desc.
Dmitry Lebed [Mon, 26 Mar 2018 13:36:33 +0000 (16:36 +0300)]
qtnfmac: add DFS offload support
DFS offload support implemented:
- DFS_OFFLOAD feature is advertised depending on HW capabilities
- CAC_STARTED event forwarding from HW implemented
- start_radar_detection() callback now returning -ENOTSUPP
if DFS_OFFLOAD is enabled
Some HP laptops have only a single wifi antenna. This would not be a
problem except that they were shipped with an incorrectly encoded
EFUSE. It should have been possible to open the computer and transfer
the antenna connection to the other terminal except that such action
might void the warranty, and moving the antenna broke the Windows
driver. The fix was to add a module option that would override the
EFUSE encoding. That was done with commit c18d8f509571 ("rtlwifi:
rtl8723be: Add antenna select module parameter"). There was still a
problem with Bluetooth coexistence, which was addressed with commit baa170229095 ("rtlwifi: btcoexist: Implement antenna selection").
There were still problems, thus there were commit 0ff78adeef11
("rtlwifi: rtl8723be: fix ant_sel code") and commit 6d6226928369
("rtlwifi: btcoexist: Fix antenna selection code"). Despite all these
attempts at fixing the problem, the code is not yet right. A proper
fix is important as there are now instances of laptops having
RTL8723DE chips with the same problem.
The module parameter ant_sel is used to control antenna number and path.
At present enum ANT_{X2,X1} is used to define the antenna number, but
this choice is not intuitive, thus change to a new enum ANT_{MAIN,AUX}
to make it more readable. This change showed examples where incorrect
values were used. It was also possible to remove a workaround in
halbtcoutsrc.c.
The experimental results with single antenna connected to specific path
are now as follows:
ant_sel ANT_MAIN(#1) ANT_AUX(#2)
0 -8 -62
1 -62 -10
2 -6 -60
When version 8 of the scan command API was introduced, only the size
of version 7 was updated, causing older versions of the firmware to
throw BAD_COMMAND errors.
Calculating the old version based on the size of the latest version
got too complicated and the size of the older versions will never
change anyway, so it's better to just hardcoded the sizes.
Fixes: 66fa2424df16 ("iwlwifi: fw api: support the new scan request FW API version") Reported-by: Scott Register <[email protected]> Signed-off-by: Luca Coelho <[email protected]> Signed-off-by: Kalle Valo <[email protected]>
Kalle Valo [Tue, 24 Apr 2018 10:00:03 +0000 (13:00 +0300)]
Merge tag 'iwlwifi-next-for-kalle-2018-04-20' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-next
iwlwifi patches for 4.18
* implement Traffic Condition Monitor;
* use TCM for scan and BT coex;
* use TCM to detect when the AP doesn't support UAPSD properly;
* some more work for the 22000 family of devices;
* introduce AMSDU rate control offload;
* a couple of clean-ups and bugfixes.
Daniel Mack [Tue, 17 Apr 2018 13:23:37 +0000 (15:23 +0200)]
wcn36xx: pass information elements in scan requests
When the wifi driver core passes IE elements in the scan request, append
them to the firmware message. The driver currently tells the core that
it is capable of attaching up to WCN36XX_MAX_SCAN_IE_LEN octets, but
doesn't actually pass them to the the hardware.
Note that this patch doesn't fix a bug that was observed. The change is
merely done for the sake of completeness as the hardware supports
appending IEs in scans. Tests show that network scans work fine with
this patch applied.
Some defines were moved around to avoid cyclic include dependencies.
projects / linux.git / log