openwrt/staging/blogic.git
8 years agortl8xxxu: Use 1024 byte block loads for 8192eu firmware
Jes Sorensen [Mon, 29 Feb 2016 22:04:01 +0000 (17:04 -0500)]
rtl8xxxu: Use 1024 byte block loads for 8192eu firmware

The rtl8192eu can handle 1024 byte block writes, unlike it's
predecessors (8192cu/8188cu).

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortl8xxxu: Identify chip vendors correctly
Jes Sorensen [Mon, 29 Feb 2016 22:04:00 +0000 (17:04 -0500)]
rtl8xxxu: Identify chip vendors correctly

This identifies the chip vendors correctly and also picks the correct
firmware for rtl8192eu.

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortl8xxxu: Add initial code to parse rtl8192eu efuse
Jes Sorensen [Mon, 29 Feb 2016 22:03:59 +0000 (17:03 -0500)]
rtl8xxxu: Add initial code to parse rtl8192eu efuse

This is the start of 8192eu support. For now just detect the device
and parse the efuse.

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agowlcore/wl18xx: add radar_debug_mode handling
Eliad Peller [Sun, 6 Mar 2016 22:28:09 +0000 (00:28 +0200)]
wlcore/wl18xx: add radar_debug_mode handling

Add debugfs key (under CFG80211_CERTIFICATION_ONUS
configuration) to set/clear radar_debug_mode.
In this mode, the driver simply ignores radar
events (but prints them).

The fw is notified about this mode through
a special generic_cfg_feature command.

This mode is relevant only for ap mode. look for
it when initializing ap vif.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agowlcore: don't WARN_ON in case of existing ROC
Eliad Peller [Sun, 6 Mar 2016 22:28:08 +0000 (00:28 +0200)]
wlcore: don't WARN_ON in case of existing ROC

When working with AP + P2P, it's possible to get into
a state when the AP is in ROC (due to assiciating station)
while trying to ROC on the P2P interface.

Replace the WARN_ON with wl1271_error to avoid warnings
in this case.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: Empty Tx queue during suspend
Amitkumar Karwar [Wed, 9 Mar 2016 12:21:11 +0000 (04:21 -0800)]
mwifiex: Empty Tx queue during suspend

In cfg80211 suspend handler, stop the netif queue and
wait until all the Tx queues become empty. Start the
queues in resume handler.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: Remove waitqueue_active check
Hui Wang [Wed, 9 Mar 2016 07:25:26 +0000 (15:25 +0800)]
brcmfmac: Remove waitqueue_active check

We met a problem of pm_suspend  when repeated closing/opening the lid
on a Lenovo laptop (1/20 reproduce rate), below is the log:

[ 199.735876] PM: Entering mem sleep
[ 199.750516] e1000e: EEE TX LPI TIMER: 00000011
[ 199.856638] Trying to free nonexistent resource <000000000000d000-000000000000d0ff>
[ 201.753566] brcmfmac: brcmf_pcie_suspend: Timeout on response for entering D3 substate
[ 201.753581] pci_legacy_suspend(): brcmf_pcie_suspend+0x0/0x1f0 [brcmfmac] returns -5
[ 201.753585] dpm_run_callback(): pci_pm_suspend+0x0/0x160 returns -5
[ 201.753589] PM: Device 0000:04:00.0 failed to suspend async: error -5

Through debugging, we found when problem happens, it is not the device
fails to enter D3, but the signal D3_ACK comes too early to pass the
waitqueue_active() check.

Just like this:
brcmf_pcie_send_mb_data(devinfo, BRCMF_H2D_HOST_D3_INFORM);
// signal is triggered here
wait_event_timeout(devinfo->mbdata_resp_wait, devinfo->mbdata_completed,
   BRCMF_PCIE_MBDATA_TIMEOUT);

So far I think it is safe to remove waitqueue_active check since there
is only one place to trigger this signal (sending
BRCMF_H2D_HOST_D3_INFORM). And it is not a problem calling wake_up
event earlier than calling wait_event.

Cc: Brett Rudley <brudley@broadcom.com>
Cc: Hante Meuleman <meuleman@broadcom.com>
Cc: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Cc: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Cc: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agolibertas: fix an error code in probe
Dan Carpenter [Tue, 8 Mar 2016 12:04:08 +0000 (15:04 +0300)]
libertas: fix an error code in probe

We accidentally return success instead of a negative error code.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agoMerge tag 'iwlwifi-next-for-kalle-2016-03-09_2' of https://git.kernel.org/pub/scm...
Kalle Valo [Thu, 10 Mar 2016 12:53:35 +0000 (14:53 +0200)]
Merge tag 'iwlwifi-next-for-kalle-2016-03-09_2' of https://git./linux/kernel/git/iwlwifi/iwlwifi-next

* update GSCAN capabilities (Ayala)
* fix AES-CMAC in AP mode (Johannes)
* adapt prints to new firmware API
* rx path improvements (Sara and Gregory)
* fixes for the thermal / cooling device code (Chaya Rachel)
* fixes for GO uAPSD handling
* more code for the 9000 device family (Sara)
* infrastructure work for firmware notification (Chaya Rachel)
* improve association reliablity (Sara)
* runtime PM fixes
* fixes for ROC (HS2.0)

8 years agoiwlwifi: mvm: update GSCAN capabilities
Ayala Beker [Wed, 3 Feb 2016 13:36:52 +0000 (15:36 +0200)]
iwlwifi: mvm: update GSCAN capabilities

Gscan capabilities were updated with new capabilities supported
by the device. While at it, simplify the firmware support
conditional and move both conditions into the WARN() to make it
easier to undertand and use the unlikely() for both.

Signed-off-by: Ayala Beker <ayala.beker@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: don't try to offload AES-CMAC in AP/IBSS modes
Johannes Berg [Wed, 9 Mar 2016 13:58:47 +0000 (14:58 +0100)]
iwlwifi: mvm: don't try to offload AES-CMAC in AP/IBSS modes

The firmware/hardware only supports checking AES-CMAC on RX, not
using it on TX. For station mode this is fine, since it's the only
thing it will ever do. For AP mode, it never receives such frames,
but must be able to transmit them. This is currently broken since
we try to enable them for hardware crypto (for RX only) and then
treat them as TX_CMD_SEC_EXT, leading to FIFO underruns during TX
so the frames never go out to the air.

To fix this, simply use software on TX in AP (and IBSS) mode.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: adapt the firmware assert log to new firmware
Emmanuel Grumbach [Mon, 7 Mar 2016 08:16:38 +0000 (10:16 +0200)]
iwlwifi: mvm: adapt the firmware assert log to new firmware

Newer firmware versions put different data in the memory
which is read by the driver upon firmware crash. Just
change the variable names in the code and the name of the
data in the log that we print withouth any functional
change.
On older firmware, there will be a mismatch between the
names that are printed and the content itself, but that's
harmless.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: pcie: avoid restocks inside rx loop if not emergency
Gregory Greenman [Mon, 29 Feb 2016 13:34:25 +0000 (15:34 +0200)]
iwlwifi: pcie: avoid restocks inside rx loop if not emergency

When trying to reach high Rx throughput of more than 500Mbps on
a device with a relatively weak CPU (Atom x5-Z8500), CPU utilization
may become a bottleneck. Analysis showed that we are looping in
iwl_pcie_rx_handle for very long periods which led to starvation
of other threads (iwl_pcie_rx_handle runs with _bh disabled).
We were handling Rx and allocating new buffers and the new buffers
were ready quickly enough to be available before we had finished
handling all the buffers available in the hardware. As a
consequence, we called iwl_pcie_rxq_restock to refill the hardware
with the new buffers, and start again handling new buffers without
exiting the function. Since we read the hardware pointer again when
we goto restart, new buffers were handled immediately instead of
exiting the function.

This patch avoids refilling RBs inside rx handling loop, unless an
emergency situation is reached. It also doesn't read the hardware
pointer again unless we are in an emergency (unlikely) case.
This significantly reduce the maximal time we spend in
iwl_pcie_rx_handle with _bh disabled.

Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: return the cooling state index instead of the budget
Chaya Rachel Ivgi [Sun, 14 Feb 2016 12:03:10 +0000 (14:03 +0200)]
iwlwifi: mvm: return the cooling state index instead of the budget

iwl_mvm_tcool_get_cur_state is the function that returns the
cooling state index to the sysfs handler. This function returns
mvm->cooling_dev.cur_state but that variable was set to the
budget and not the cooling state index. Fix that.
Add a missing blank line while at it.

Signed-off-by: Chaya Rachel Ivgi <chaya.rachel.ivgi@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: remove RRM advertisement
Emmanuel Grumbach [Tue, 8 Mar 2016 11:53:05 +0000 (13:53 +0200)]
iwlwifi: mvm: remove RRM advertisement

mac80211 advertises this feature for all its drivers.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: don't let NDPs mess the packet tracking
Emmanuel Grumbach [Mon, 7 Mar 2016 20:23:52 +0000 (22:23 +0200)]
iwlwifi: mvm: don't let NDPs mess the packet tracking

We need to track the next packet that we will reclaim in
order to know when the Tx queues are empty. This is useful
when we open or tear down an A-MPDU session which requires
to switch queue.
The next packet being reclaimed is identified by its WiFi
sequence number and this is relevant only when we use QoS.
QoS NDPs do have a TID but have a meaningless sequence
number. The spec mandates the receiver to ignore the
sequence number in this case, allowing the transmitter to
put any sequence number. Our implementation leaves it 0.
When we reclaim a QoS NDP, we can't update the next_relcaim
counter since the sequence number of the QoS NDP itself is
invalid.
We used to update the next_reclaim based on the sequence
number of the QoS NDP which reset it to 1 (0 + 1) and
because of this, we never knew when the queue got empty.
This had to sad consequence to stuck the A-MPDU state
machine in a transient state.
To fix this, don't update next_reclaim when we reclaim
a QoS NDP.

Alesya saw this bug when testing u-APSD. Because the
A-MPDU state machine was stuck in EMPTYING_DELBA, we
updated mac80211 that we still have frames for that
station when it got back to sleep. mac80211 then wrongly
set the TIM bit in the beacon and requested to release
non-existent frames from the A-MPDU queue. This led to
a situation where the client was trying to poll frames
but we had no frames to send.

Reported-by: Alesya Shapira <alesya.shapira@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: add support for getting HW address from CSR
Sara Sharon [Mon, 7 Mar 2016 12:18:29 +0000 (14:18 +0200)]
iwlwifi: add support for getting HW address from CSR

From 9000 family on, we need to get HW address from host
CSR registers.
OEM can override it by fusing the override registers - read
those first, and if those are 0 - read the OTP registers instead.

In addition - bail out if no valid mac address is present. Make
it shared for all NICs.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: pcie: fine tune number of rxbs
Sara Sharon [Mon, 1 Feb 2016 11:46:06 +0000 (13:46 +0200)]
iwlwifi: pcie: fine tune number of rxbs

We kick the allocator when we have 2 RBDs that don't have
attached RBs, and the allocator allocates 8 RBs meaning
that it needs another 6 RBDs to attach the RBs to.
The design is that allocator should always have enough RBDs
to fulfill requests, so we give in advance 6 RBDs to the
allocator so that when it is kicked, it gets additional 2 RBDs
and has enough RBDs.
These RBDs were taken from the Rx queue itself, meaning
that each Rx queue didn't have the maximal number of
RBDs, but MAX - 6.
Change initial number of RBDs in the system to include both
queue size and allocator reserves.
Note the multi-queue is always 511 instead of 512 to avoid a
full queue since we cannot detect this state easily enough in
the 9000 arch.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: add support for async rx handler without hold the mutex
Chaya Rachel Ivgi [Thu, 3 Mar 2016 13:35:34 +0000 (15:35 +0200)]
iwlwifi: mvm: add support for async rx handler without hold the mutex

When running async rx handler the framework holds the mvm->mutex
before starting the async handler, that might cause a deadlock in case
the handler calls to ops that lock the mutex as well.
Add support for running async rx handler without hold the mutex before
activating the handler.

Signed-off-by: Chaya Rachel Ivgi <chaya.rachel.ivgi@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: ROC: cleanup time event info on FW failure
Matti Gottlieb [Wed, 9 Mar 2016 12:46:28 +0000 (14:46 +0200)]
iwlwifi: mvm: ROC: cleanup time event info on FW failure

Currently when the FW sends start/stop aux roc time event
notification with an error status, the driver returns an
error value, but does not remove the time event, and does
not notify the stack above that the time event is over.

This causes problems that the stack above assumes we are still
in the middle of a time event, and therefore can block different
events, such as scanning.

On FW failure notification, cleanup the time event parameters and
notify the stack above that the time event is over.

Signed-off-by: Matti Gottlieb <matti.gottlieb@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: turn off AMSDU bit in QoS control for de-aggregated AMSDUs
Sara Sharon [Sun, 6 Mar 2016 07:51:29 +0000 (09:51 +0200)]
iwlwifi: mvm: turn off AMSDU bit in QoS control for de-aggregated AMSDUs

Our hardware de-aggregates AMSDUs but copies the mac header
as it to the de-aggregated MPDUs. We need to turn off the AMSDU
bit in the QoS control ourselves.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: add ctdp operations to debugfs
Chaya Rachel Ivgi [Wed, 24 Feb 2016 10:19:22 +0000 (12:19 +0200)]
iwlwifi: mvm: add ctdp operations to debugfs

Add debugfs entries to get the ctdp budget average
and to stop ctdp.

Signed-off-by: Chaya Rachel Ivgi <chaya.rachel.ivgi@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agowireless: cw1200: use __maybe_unused to hide pm functions_
Arnd Bergmann [Wed, 2 Mar 2016 15:59:02 +0000 (16:59 +0100)]
wireless: cw1200: use __maybe_unused to hide pm functions_

The cw1200 uses #ifdef to check for CONFIG_PM, but then
uses SIMPLE_DEV_PM_OPS, which leaves the references out when
CONFIG_PM_SLEEP is not defined, so we get a warning with
PM=y && PM_SLEEP=n:

drivers/net/wireless/st/cw1200/cw1200_spi.c:450:12: error: 'cw1200_spi_suspend' defined but not used [-Werror=unused-function]

This removes the incorrect #ifdef and instead uses a __maybe_unused
annotation to let the compiler know it can silently drop
the function definition.

For the DEV_PM_OPS definition, we can use an IS_ENABLED() check
to avoid defining the structure when CONFIG_PM is not set without
the #ifdef.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortlwifi: Fix size of wireless mode variable
Larry Finger [Thu, 25 Feb 2016 17:03:01 +0000 (11:03 -0600)]
rtlwifi: Fix size of wireless mode variable

Smatch reports the following warning:

  CHECK   drivers/net/wireless/realtek/rtlwifi/rc.c
drivers/net/wireless/realtek/rtlwifi/rc.c:144 _rtl_rc_rate_set_series() warn: impossible condition '(wireless_mode == 256) => (0-255 == 256)'

This warning arises because commit acc6907b87a9 ("rtlwifi: Fix warning
from ieee80211_get_tx_rates() when using 5G") now checks the wireless
mode for WIRELESS_MODE_AC_ONLY (BIT(8)) in _rtl_rc_rate_set_series().
As a result, all quantities used to store the wireless mode must be u16.

This patch also reorders struct rtl_sta_info to save a little space.

Fixes: d76d65fd2695 ("rtlwifi: fix broken VHT support")
Reported-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agoMerge tag 'iwlwifi-next-for-kalle-2016-03-02' of https://git.kernel.org/pub/scm/linux...
Kalle Valo [Mon, 7 Mar 2016 13:48:56 +0000 (15:48 +0200)]
Merge tag 'iwlwifi-next-for-kalle-2016-03-02' of https://git./linux/kernel/git/iwlwifi/iwlwifi-next

* add support for thermal device / cooling device (Chaya Rachel)
* fixes for 9000 devices data path (Sara Sharon)
* improvements in scheduled scan w/o profiles (Luca)
* new firmware support (-21.ucode)
* add MSIX support for 9000 devices (Haim Dreyfuss)
* cleanup in PCIe initialization
* enable MU-MIMO and take care of firmware restart(Sara Sharon)
        ===> This needs mac80211-next
* add support for large SKBs in mvm to reach A-MSDU
        ===> This needs mac80211-next
* add support for filtering frames from a BA session (Sara Sharon)
        ===> This needs mac80211-next
* start implementing the new Rx path for 9000 devices (Sara Sharon)
* enable the new RRM feature flag (Beni Lev)
* fix U-APSD enablement on P2P Client (Avri Altman)
* fix beacon abort enablement (Avri Altman)
* forbid beacon storing with WoWLAN (Matti Gottlieb)
* support unified uSniffer / regular firmware image (Golan Ben-Ami)
* fix a race between debugfs hooks and iface up (Chaya Rachel Ivgi)
* fixes for runtime PM (Luca)
* add a new module paramater to disable VHT (Andrei Otcheretianski)
* build infrastructure for Dynamic Queue Allocation (Liad Kaufman)

8 years agobcma: move flash detection code to ChipCommon core driver
Rafał Miłecki [Fri, 12 Feb 2016 09:15:45 +0000 (10:15 +0100)]
bcma: move flash detection code to ChipCommon core driver

Flashes are attached and accessible through ChipCommon core, they aren't
connected to MIPS one. Moreover some new ARM devices (e.g. BCM47189) may
have serial flash accessibility using ChipCommon registers as well. To
support them we can't keep this code in MIPS core driver.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobcma: move parallel flash support to separated file
Rafał Miłecki [Fri, 12 Feb 2016 09:15:44 +0000 (10:15 +0100)]
bcma: move parallel flash support to separated file

This follows the way of handling other flashes and cleans code a bit. As
next task we will want to move flash code to ChipCommon driver as:
1) Flash controllers are accesible using ChipCommon registers
2) This code isn't MIPS specific
This change prepares bcma for that.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobcma: drop unneeded fields from bcma_pflash struct
Rafał Miłecki [Fri, 12 Feb 2016 09:15:43 +0000 (10:15 +0100)]
bcma: drop unneeded fields from bcma_pflash struct

Most of info stored in this struct wasn't really used anywhere as we put
all that data in platform data & resource as well.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomt7601u: do not free dma_buf when ivp allocation fails
Colin Ian King [Thu, 25 Feb 2016 23:24:27 +0000 (23:24 +0000)]
mt7601u: do not free dma_buf when ivp allocation fails

If the allocation of ivp fails the error handling attempts to
free an uninitialized dma_buf; this data structure just contains
garbage on the stack, so the freeing will cause issues when the
urb, buf and dma fields are free'd. Fix this by not free'ing the
dma_buf if the ivp allocation fails.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: Added missing spaces around brackets
Ujjal Roy [Wed, 24 Feb 2016 19:19:56 +0000 (00:49 +0530)]
mwifiex: Added missing spaces around brackets

This patch fixes the missing spaces issue in coding style.

Signed-off-by: Ujjal Roy <royujjal@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: Removed extra spaces before commas
Ujjal Roy [Wed, 24 Feb 2016 19:19:43 +0000 (00:49 +0530)]
mwifiex: Removed extra spaces before commas

This patch fixes spaces before commas issue in coding style.

Signed-off-by: Ujjal Roy <royujjal@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: Fixed incorrect indentation issue
Ujjal Roy [Wed, 24 Feb 2016 19:19:23 +0000 (00:49 +0530)]
mwifiex: Fixed incorrect indentation issue

This patch fixes the incorrect indentation of the case label.

Signed-off-by: Ujjal Roy <royujjal@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agort2x00: add new rt2800usb device Buffalo WLI-UC-G450
Anthony Wong [Tue, 23 Feb 2016 15:09:22 +0000 (23:09 +0800)]
rt2x00: add new rt2800usb device Buffalo WLI-UC-G450

Add USB ID 0411:01fd for Buffalo WLI-UC-G450 wireless adapter,
RT chipset 3593

Signed-off-by: Anthony Wong <anthony.wong@ubuntu.com>
Cc: stable@vger.kernel.org
Acked-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: do not set multiport flag for tx/rx single packet
Xinming Hu [Tue, 23 Feb 2016 13:16:19 +0000 (05:16 -0800)]
mwifiex: do not set multiport flag for tx/rx single packet

multiport address flag(0x1000) should not be set during sdio cmd53,
if we have only one packet to read/write.

Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Xinming Hu <huxm@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: add sdio multiport aggregation debug information
Xinming Hu [Tue, 23 Feb 2016 13:16:18 +0000 (05:16 -0800)]
mwifiex: add sdio multiport aggregation debug information

This patch sdio multi port aggregation statistics which can be
used for debugging. This debug data is collected in
/sys/kernel/debug/mwifiex/mlan0/debug.

Signed-off-by: Xinming Hu <huxm@marvell.com>
Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: fix corner case association failure
Amitkumar Karwar [Tue, 23 Feb 2016 13:16:17 +0000 (05:16 -0800)]
mwifiex: fix corner case association failure

This patch corrects the error case in association path by returning
-1. Earlier "media_connected" used to remain on in this error case
causing failure for further association attempts.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Fixes: b887664d882ee4 ('mwifiex: channel switch handling for station')
Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: add delay when tdls confirm frame is queued
Xinming Hu [Tue, 23 Feb 2016 13:16:16 +0000 (05:16 -0800)]
mwifiex: add delay when tdls confirm frame is queued

It is observed that driver may send the data packet to tdls peer
before tdls peer receives tdls setup confirm frame.
Similar race condition exists during tdls teardown procedure also.
This patch adds 10 milliseconds delay to resolve the race.

Signed-off-by: Xinming Hu <huxm@marvell.com>
Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortlwifi: pass struct rtl_stats by reference as it is more efficient
Colin Ian King [Mon, 22 Feb 2016 11:35:46 +0000 (11:35 +0000)]
rtlwifi: pass struct rtl_stats by reference as it is more efficient

passing rtl_stats by value is inefficient; the structure is over 300
bytes in size and generally just one field (packet_report_type)
is being accessed, so the pass by value is a relatively large overhead.
This change just affects just the rx_command_packet calls.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agoat76c50x-usb: avoid double usb_put_dev() after downloading internal firmware in at76_...
Alexey Khoroshilov [Sat, 20 Feb 2016 22:30:29 +0000 (01:30 +0300)]
at76c50x-usb: avoid double usb_put_dev() after downloading internal firmware in at76_probe()

There is no need in usb_put_dev() if at76_load_internal_fw() succeed.

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

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agomwifiex: Use to_delayed_work()
Amitoj Kaur Chawla [Wed, 17 Feb 2016 12:33:37 +0000 (18:03 +0530)]
mwifiex: Use to_delayed_work()

Introduce the use of to_delayed_work() helper function instead of open
coding it with container_of()

A simplified version of the Coccinelle semantic patch used to make
this change is:

//<smpl>
@@
expression a;
symbol work;
@@
- container_of(a, struct delayed_work, work)
+ to_delayed_work(a)
//</smpl>

Signed-off-by: Amitoj Kaur Chawla <amitoj1606@gmail.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: add 802.11w management frame protection support
Hante Meuleman [Wed, 17 Feb 2016 10:27:10 +0000 (11:27 +0100)]
brcmfmac: add 802.11w management frame protection support

Add full support for both AP and STA for management frame protection.

Reviewed-by: Arend Van Spriel <arend.van@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <franky.lin@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Signed-off-by: Hante Meuleman <hante.meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: integrate add_keyext in add_key
Hante Meuleman [Wed, 17 Feb 2016 10:27:09 +0000 (11:27 +0100)]
brcmfmac: integrate add_keyext in add_key

brcmf_add_keyext is called when a key is configured for a specific
mac address. This function is very similar to the calling function
brcmf_add_key. Integrate this function and also use existing del_key
function in case key is to be cleared.

Reviewed-by: Arend Van Spriel <arend.van@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <franky.lin@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieter-paul.giesberts@broadcom.com>
Signed-off-by: Hante Meuleman <hante.meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: merge platform data and module paramaters
Hante Meuleman [Wed, 17 Feb 2016 10:27:08 +0000 (11:27 +0100)]
brcmfmac: merge platform data and module paramaters

Merge module parameters and platform data in one struct. This is the
last step to move to the new platform data per device. Now parameters
of platform data will be merged with module parameters per device.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: switch to new platform data
Hante Meuleman [Wed, 17 Feb 2016 10:27:07 +0000 (11:27 +0100)]
brcmfmac: switch to new platform data

Platform data is only available for sdio. With this patch a new
platform data structure is being used which allows for platform
data for any device and configurable per device. This patch only
switches to the new structure and adds support for SDIO devices.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: keep ARP and ND offload enabled during WOWL
Hante Meuleman [Wed, 17 Feb 2016 10:27:05 +0000 (11:27 +0100)]
brcmfmac: keep ARP and ND offload enabled during WOWL

Currently ARP and ND (IPv6 Neigbor Discovery) offload get disabled
on entering suspend. However when firmwares support the wowl_cap
iovar then these offload routines can be kept enabled as they
will work during WOWL as well.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: move platform data retrieval code to common
Hante Meuleman [Wed, 17 Feb 2016 10:27:04 +0000 (11:27 +0100)]
brcmfmac: move platform data retrieval code to common

In preparation of module parameters for all devices the module
platform data retrieval is moved from sdio to common. It is still
only used for sdio devices.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: add wowl gtk rekeying offload support
Hante Meuleman [Wed, 17 Feb 2016 10:27:03 +0000 (11:27 +0100)]
brcmfmac: add wowl gtk rekeying offload support

This patch adds support for gtk rekeying offload and for gtk
rekeying failure during wowl mode.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: move module init and exit to common
Hante Meuleman [Wed, 17 Feb 2016 10:27:02 +0000 (11:27 +0100)]
brcmfmac: move module init and exit to common

In preparation of module parameters for all devices the module init
and exit routines are moved to the common file.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: increase timeout for tx eapol
Hante Meuleman [Wed, 17 Feb 2016 10:27:01 +0000 (11:27 +0100)]
brcmfmac: increase timeout for tx eapol

When keys get set and updated this has to happen after eapol got
transmitted (without key or old key) before the key can be updated.
To make sure the order of sending eapol and configuring key is done
correctly a timeout for tx of eapol is applied. This timeout is set
to 50 msec, which is not always enough. Especially in AP mode and
key updates the timeout may need to be much longer because client(s)
can be in powersave. Increase the timeout from 50 to 950 msec.

Reviewed-by: Arend Van Spriel <arend@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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: remove pcie gen1 support
Hante Meuleman [Wed, 17 Feb 2016 10:27:00 +0000 (11:27 +0100)]
brcmfmac: remove pcie gen1 support

The PCIE bus driver supports older gen1 (v1) chips, but there is no
actual device which is using this older pcie core which is supported
by brcmfmac. Remove all gen1 related code.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: add support for the PCIE 4366c0 chip
Hante Meuleman [Wed, 17 Feb 2016 10:26:59 +0000 (11:26 +0100)]
brcmfmac: add support for the PCIE 4366c0 chip

A newer version of the 4366 PCIE chip has been released. Add
support for this version of the chip.

Reviewed-by: Arend Van Spriel <arend@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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: use bar1 window size as provided by pci subsystem
Hante Meuleman [Wed, 17 Feb 2016 10:26:58 +0000 (11:26 +0100)]
brcmfmac: use bar1 window size as provided by pci subsystem

The PCIE bar1 window size is specified by chip. Currently the
ioremap of bar1 was using a define which always matched the size
of bar1, but newer chips can have a different bar1 sizes. With
this patch the ioremap will be called with the by chip provided
window size.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: use device memsize config from fw if defined
Hante Meuleman [Wed, 17 Feb 2016 10:26:57 +0000 (11:26 +0100)]
brcmfmac: use device memsize config from fw if defined

Newer type pcie devices have memory which get shared between fw and
hw. The division of this memory is done firmware compile time. As a
result the ramsize as used by driver needs to be adjusted for this.
This is done by reading the memory size from the firmware.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: check return for ARP ip setting iovar
Franky Lin [Wed, 17 Feb 2016 10:26:56 +0000 (11:26 +0100)]
brcmfmac: check return for ARP ip setting iovar

The return value of iovar set function should be saved and checked.

Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: add neighbor discovery offload ip address table configuration
Franky Lin [Wed, 17 Feb 2016 10:26:55 +0000 (11:26 +0100)]
brcmfmac: add neighbor discovery offload ip address table configuration

Configure ipv6 address for neighbor discovery offload ip table in
firmware obtained through ipv6 address notification callback.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: Add length checks on firmware events
Hante Meuleman [Wed, 17 Feb 2016 10:26:54 +0000 (11:26 +0100)]
brcmfmac: Add length checks on firmware events

Add additional length checks on firmware events to create more
robust code.

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: Lei Zhang <leizh@broadcom.com>
Signed-off-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: Configure country code using device specific settings
Hante Meuleman [Wed, 17 Feb 2016 10:26:53 +0000 (11:26 +0100)]
brcmfmac: Configure country code using device specific settings

Country code configuration in a device is a device specific
operation. For this the country code as specified by reg notifier
(iso3166 alpha2) needs to be translated to a device specific
country locale and revision number. This patch adds this
translation and puts a placeholder in the device specific settings
where the translation table can be stored. Additional patches will
be needed to read these tables from for example device platform
data.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: check for wowl support before enumerating feature flag
Franky Lin [Wed, 17 Feb 2016 10:26:52 +0000 (11:26 +0100)]
brcmfmac: check for wowl support before enumerating feature flag

In some cases wiphy->wowlan could be NULL if firmware doesn't have the
support. Driver should check for support before walking down the feature
flags.

Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: Limit memory allocs to <64K
Hante Meuleman [Wed, 17 Feb 2016 10:26:51 +0000 (11:26 +0100)]
brcmfmac: Limit memory allocs to <64K

Some systems have problems with allocating memory allocation larger
then 64K. Often on unload/load or suspend/resume a failure is
reported: Could not allocate wiphy device. This patch makes the
escan intermediate storage buf dynamically allocated, and smaller
than 64K.

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: Kalle Valo <kvalo@codeaurora.org>
8 years agobrcmfmac: change function name for brcmf_cfg80211_wait_vif_event_timeout()
Arend van Spriel [Wed, 17 Feb 2016 10:26:50 +0000 (11:26 +0100)]
brcmfmac: change function name for brcmf_cfg80211_wait_vif_event_timeout()

Dropping the '_timeout' from the function name as the fact that a timeout
value is passed makes it obvious a timeout is used. Also helps to keep code
lines a bit shorter and easier to stick to 80 char boundary.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortlwifi: Change long delays to sleeps
Larry Finger [Mon, 15 Feb 2016 22:12:07 +0000 (16:12 -0600)]
rtlwifi: Change long delays to sleeps

Routine rtl_addr_delay() uses delay statements in code that can
sleep. To improve system responsiveness, the various delay statements
are changed.

In addition, routines rtl_rfreg_delay() and rtl_bb_delay() are
rewritten to use the code in rtl_addr_delay() for most of their
input values.

Suggested-by: Byeoungwook Kim <quddnr145@gmail.com>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agocarl9170: import 1.9.9 firmware headers
Christian Lamparter [Mon, 15 Feb 2016 20:16:06 +0000 (21:16 +0100)]
carl9170: import 1.9.9 firmware headers

Import new headers from my firmware branch:
<https://github.com/chunkeey/carl9170fw>

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agort2x00: unterminated strlen of user data
One Thousand Gnomes [Mon, 15 Feb 2016 19:04:56 +0000 (19:04 +0000)]
rt2x00: unterminated strlen of user data

The buffer needs to be zero terminated in case the user data is not.
Otherwise we run off the end of the buffer.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agortlwifi: rtl8821ae: Silence useless log output
Larry Finger [Thu, 11 Feb 2016 16:56:22 +0000 (10:56 -0600)]
rtlwifi: rtl8821ae: Silence useless log output

When driver rtl8821ae is loaded but not connected to any AP, it logs
a "firmware not ready to run" message roughly once a minute. To
eliminate logging this massage under normal debug conditions, the
degug level needed to print this message is increased.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
8 years agoiwlwifi: mvm: extend time event duration
Sara Sharon [Tue, 2 Feb 2016 09:55:53 +0000 (11:55 +0200)]
iwlwifi: mvm: extend time event duration

Before authentication, we start a time event during
which we wait for a beacon in order to sync our timers.
If we didn't hear the beacon during this time - we abandon
the connection. However, in congested environment, it was
observed we might not hear beacons in that time slot.
Extend the time event to give the connection a better chance.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: set the correct amsdu enum values
Sara Sharon [Thu, 3 Mar 2016 14:49:05 +0000 (16:49 +0200)]
iwlwifi: mvm: set the correct amsdu enum values

The amsdu enum values are off by 1 bit. Fix it.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: fix unregistration of thermal in some error flows
Chaya Rachel Ivgi [Thu, 3 Mar 2016 11:31:39 +0000 (13:31 +0200)]
iwlwifi: mvm: fix unregistration of thermal in some error flows

The call to iwl_mvm_thermal_initialize() was too early in the
function.
Unregister will be performed when goto out_unregister is called,
but as the code was - out_free may be called and leave without
unregistering from thermal.

Signed-off-by: Chaya Rachel Ivgi <chaya.rachel.ivgi@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: refactor the code that reads the MAC address from the NVM
Sara Sharon [Tue, 1 Mar 2016 10:18:22 +0000 (12:18 +0200)]
iwlwifi: refactor the code that reads the MAC address from the NVM

It makes it slightly easier to follow. Pass the pointer to
the transport which allows to read WFMP_MAC_ADDR_X register
only when needed and to use IWL_ERR instead of the less
commonly used IWL_ERR_DEV logger macro.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: mvm: avoid panics with thermal device usage
Emmanuel Grumbach [Sun, 28 Feb 2016 08:15:08 +0000 (10:15 +0200)]
iwlwifi: mvm: avoid panics with thermal device usage

Thermal zone device registration can fail, and in this case
we don't want to remove WiFi functionality. This is why the
thermal zone registration function is void, and the flows
continue even if the thermal zone device registration failed.
Same applies for the cooling device.

This means that we at least need to remember that the thermal
zone device didn't register properly and take the minimal
precautions to avoid panic'ing when we access it.

This was missing.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: pcie: set RB chunk size back to 64
Sara Sharon [Wed, 2 Mar 2016 13:17:28 +0000 (15:17 +0200)]
iwlwifi: pcie: set RB chunk size back to 64

128 byte chunk size is supported only on PCIe and not
on IOSF. For now, change it back to 64 byte.

Reported-by: Oren Givon <oren.givon@intel.com>
Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: pcie: refactor RXBs reclaiming code
Sara Sharon [Mon, 15 Feb 2016 17:30:49 +0000 (19:30 +0200)]
iwlwifi: pcie: refactor RXBs reclaiming code

Change the code to move rxbs directly from the allocator's
list to the queue's free list. This makes the code more
readable, saves the interim array and the double loop over
the free RBs.

Signed-off-by: Sara Sharon <sara.sharon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoiwlwifi: pcie: forbid RTPM on device removal
Luca Coelho [Mon, 8 Feb 2016 09:18:46 +0000 (11:18 +0200)]
iwlwifi: pcie: forbid RTPM on device removal

The pci driver keeps any unbound device in active state and forbids
runtime PM.  When our driver gets probed, we take control of the
state.  When the device is released (i.e. during unbind or module
removal), we should return the state to what it was before.  To do so,
we need to forbid RTPM in the driver remove op.

Additionally, remove an unnecessary pm_runtime_disable() call, move
the initial ref_count setting to a better place and add some comments
explaining what is going on.

Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
8 years agoMerge 'net-next/master'
Kalle Valo [Sun, 6 Mar 2016 12:57:44 +0000 (14:57 +0200)]
Merge 'net-next/master'

Needed by the upcoming merge of iwlwifi-next-for-kalle-2016-03-02 tag.

8 years agoMerge ath-next from ath.git
Kalle Valo [Sun, 6 Mar 2016 12:41:51 +0000 (14:41 +0200)]
Merge ath-next from ath.git

ath.git patches for 4.6. Major changes:

ath10k

* dt: add bindings for ipq4019 wifi block
* start adding support for qca4019 chip

ath9k

* add device ID for Toshiba WLM-20U2/GN-1080
* allow more than one interface on DFS channels

8 years agoMerge branch 'reset_mac_header'
David S. Miller [Sat, 5 Mar 2016 03:45:14 +0000 (22:45 -0500)]
Merge branch 'reset_mac_header'

Zhang Shengju says:

====================
use reset to set header pointers

This patch series replace set function with reset when offset is zero.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agowireless: use reset to set mac header
Zhang Shengju [Thu, 3 Mar 2016 01:16:57 +0000 (01:16 +0000)]
wireless: use reset to set mac header

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211: use reset to set header pointer
Zhang Shengju [Thu, 3 Mar 2016 01:16:56 +0000 (01:16 +0000)]
mac80211: use reset to set header pointer

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomac80211_hwsim: use reset to set mac header
Zhang Shengju [Thu, 3 Mar 2016 01:16:55 +0000 (01:16 +0000)]
mac80211_hwsim: use reset to set mac header

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add
operation in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovxlan: use reset to set header pointers
Zhang Shengju [Thu, 3 Mar 2016 01:16:54 +0000 (01:16 +0000)]
vxlan: use reset to set header pointers

Since offset is zero, it's not necessary to use set function. Reset
function is straightforward, and will remove the unnecessary add operation
in set function.

Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'rxrpc-rewrite-20160304' of git://git.kernel.org/pub/scm/linux/kernel/git...
David S. Miller [Fri, 4 Mar 2016 22:19:16 +0000 (17:19 -0500)]
Merge tag 'rxrpc-rewrite-20160304' of git://git./linux/kernel/git/dhowells/linux-fs

David Howells says:

====================
RxRPC: Rewrite part 1

Here's the first set of patches from my RxRPC rewrite, aimed at net-next.
These do some clean ups and bug fixes.  Most of the changes are small, but
there are a couple of bigger changes:

 (*) Convert call flag numbers and event numbers into enums.  Then rename
     the event numbers to all have _EV_ in their name to stop confusion.
     Fix one instance of an event bit being used instead of a flag bit.

 (*) A copy of the Rx protocol header is kept in the sk_buff private data.
     Keep this in host byte order rather than network byte order as it
     makes more sense.  A number of other fields then get converted into
     host byte order too.

     Conversion between host and network byte order is then done at the
     packet reception/generation stage.

This is based on net-next/master
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'DIV_ROUND_UP-uapi'
David S. Miller [Fri, 4 Mar 2016 21:10:37 +0000 (16:10 -0500)]
Merge branch 'DIV_ROUND_UP-uapi'

Nicolas Dichtel says:

====================
uapi: consolidate DIV_ROUND_UP definition

The inital goal was to consolidate ethtool.h uapi header. But I took the
opportunity to remove all duplicate definitions of DIV_ROUND_UP.

v3: add patch #2 and #3

v2: split the patch
    define DIV_ROUND_UP in uapi
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethtool.h: define INT_MAX for userland
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:19 +0000 (11:52 +0100)]
ethtool.h: define INT_MAX for userland

INT_MAX needs limits.h in userland.
When ethtool.h is included by a userland app, we got the following error:

.../usr/include/linux/ethtool.h: In function 'ethtool_validate_speed':
.../usr/include/linux/ethtool.h:1471:18: error: 'INT_MAX' undeclared (first use in this function)
  return speed <= INT_MAX || speed == SPEED_UNKNOWN
                  ^

Fixes: e02564ee334a ("ethtool: make validate_speed accept all speeds between 0 and INT_MAX")
CC: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrm/vmwgfx: remove userland definition of DIV_ROUND_UP
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:18 +0000 (11:52 +0100)]
drm/vmwgfx: remove userland definition of DIV_ROUND_UP

Let's use __KERNEL_DIV_ROUND_UP, which is defined in uapi/linux/kernel.h.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4i: don't redefine DIV_ROUND_UP
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:17 +0000 (11:52 +0100)]
cxgb4i: don't redefine DIV_ROUND_UP

let's use the common definition to avoid the following warning during the
compilation:

drivers/scsi/cxgbi/cxgb4i/cxgb4i.c:161:0: warning: "DIV_ROUND_UP" redefined
 #define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
 ^
In file included from include/linux/list.h:8:0,
                 from include/linux/module.h:9,
                 from drivers/scsi/cxgbi/cxgb4i/cxgb4i.c:16:
include/linux/kernel.h:67:0: note: this is the location of the previous definition
 #define DIV_ROUND_UP __KERNEL_DIV_ROUND_UP
 ^

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agouapi: define DIV_ROUND_UP for userland
Nicolas Dichtel [Fri, 4 Mar 2016 10:52:16 +0000 (11:52 +0100)]
uapi: define DIV_ROUND_UP for userland

DIV_ROUND_UP is defined in linux/kernel.h only for the kernel.
When ethtool.h is included by a userland app, we got the following error:

include/linux/ethtool.h:1218:8: error: variably modified 'queue_mask' at file scope
  __u32 queue_mask[DIV_ROUND_UP(MAX_NUM_QUEUE, 32)];
        ^

Let's add a common definition in uapi and use it everywhere.

Fixes: ac2c7ad0e5d6 ("net/ethtool: introduce a new ioctl for per queue setting")
CC: Kan Liang <kan.liang@intel.com>
Suggested-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonfp: remove Rolf Neugebauer as co-maintainer
Simon Horman [Fri, 4 Mar 2016 06:58:07 +0000 (15:58 +0900)]
nfp: remove Rolf Neugebauer as co-maintainer

Rolf is no longer in his previous role at Netronome and as far as I know no
longer working on the NFP driver. Thus it does not seem appropriate for him
to be a co-maintainer anymore.

Reviewed-by: Dinan Gunawardena <dinan.gunawardena@netronome.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agorxrpc: Don't try to map ICMP to error as the lower layer already did that
David Howells [Fri, 4 Mar 2016 16:02:03 +0000 (16:02 +0000)]
rxrpc: Don't try to map ICMP to error as the lower layer already did that

In the ICMP message processing code, don't try to map ICMP codes to UNIX
error codes as the caller (IPv4/IPv6) already did that for us (ee_errno).

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Clear the unused part of a sockaddr_rxrpc for memcmp() use
David Howells [Fri, 4 Mar 2016 15:59:49 +0000 (15:59 +0000)]
rxrpc: Clear the unused part of a sockaddr_rxrpc for memcmp() use

Clear the unused part of a sockaddr_rxrpc structs so that memcmp() can be
used to compare them.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: rxkad: Casts are needed when comparing be32 values
David Howells [Fri, 4 Mar 2016 15:59:13 +0000 (15:59 +0000)]
rxrpc: rxkad: Casts are needed when comparing be32 values

Forced casts are needed to avoid sparse warning when directly comparing
be32 values.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: rxkad: The version number in the response should be net byte order
David Howells [Fri, 4 Mar 2016 15:59:00 +0000 (15:59 +0000)]
rxrpc: rxkad: The version number in the response should be net byte order

The version number rxkad places in the response should be network byte
order.

Whilst we're at it, rearrange the code to be more readable.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Use ACCESS_ONCE() when accessing circular buffer pointers
David Howells [Fri, 4 Mar 2016 15:58:06 +0000 (15:58 +0000)]
rxrpc: Use ACCESS_ONCE() when accessing circular buffer pointers

Use ACCESS_ONCE() when accessing the other-end pointer into a circular
buffer as it's possible the other-end pointer might change whilst we're
doing this, and if we access it twice, we might get some weird things
happening.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Adjust some whitespace and comments
David Howells [Fri, 4 Mar 2016 15:56:19 +0000 (15:56 +0000)]
rxrpc: Adjust some whitespace and comments

Remove some excess whitespace, insert some missing spaces and adjust a
couple of comments.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Be more selective about the types of received packets we accept
David Howells [Fri, 4 Mar 2016 15:56:06 +0000 (15:56 +0000)]
rxrpc: Be more selective about the types of received packets we accept

Currently, received RxRPC packets outside the range 1-13 are rejected.
There are, however, holes in the range that should also be rejected - plus
at least one type we don't yet support - so reject these also.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Fix defined range for /proc/sys/net/rxrpc/rx_mtu
David Howells [Fri, 4 Mar 2016 15:55:32 +0000 (15:55 +0000)]
rxrpc: Fix defined range for /proc/sys/net/rxrpc/rx_mtu

The upper bound of the defined range for rx_mtu is being set in the same
member as the lower bound (extra1) rather than the correct place (extra2).
I'm not entirely sure why this compiles.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: The protocol family should be set to PF_RXRPC not PF_UNIX
David Howells [Fri, 4 Mar 2016 15:54:27 +0000 (15:54 +0000)]
rxrpc: The protocol family should be set to PF_RXRPC not PF_UNIX

Fix the protocol family set in the proto_ops for rxrpc to be PF_RXRPC not
PF_UNIX.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Keep the skb private record of the Rx header in host byte order
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Keep the skb private record of the Rx header in host byte order

Currently, a copy of the Rx packet header is copied into the the sk_buff
private data so that we can advance the pointer into the buffer,
potentially discarding the original.  At the moment, this copy is held in
network byte order, but this means we're doing a lot of unnecessary
translations.

The reasons it was done this way are that we need the values in network
byte order occasionally and we can use the copy, slightly modified, as part
of an iov array when sending an ack or an abort packet.

However, it seems more reasonable on review that it would be better kept in
host byte order and that we make up a new header when we want to send
another packet.

To this end, rename the original header struct to rxrpc_wire_header (with
BE fields) and institute a variant called rxrpc_host_header that has host
order fields.  Change the struct in the sk_buff private data into an
rxrpc_host_header and translate the values when filling it in.

This further allows us to keep values kept in various structures in host
byte order rather than network byte order and allows removal of some fields
that are byteswapped duplicates.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Rename call events to begin RXRPC_CALL_EV_
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Rename call events to begin RXRPC_CALL_EV_

Rename call event names to begin RXRPC_CALL_EV_ to distinguish them from the
flags.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Convert call flag and event numbers into enums
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Convert call flag and event numbers into enums

Convert call flag and event numbers into enums and move their definitions
outside of the struct.

Also move the call state enum outside of the struct and add an extra
element to count the number of states.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agorxrpc: Fix a case where a call event bit is being used as a flag bit
David Howells [Fri, 4 Mar 2016 15:53:46 +0000 (15:53 +0000)]
rxrpc: Fix a case where a call event bit is being used as a flag bit

Fix a case where RXRPC_CALL_RELEASE (an event) is being used to specify a
flag bit.  RXRPC_CALL_RELEASED should be used instead.

Signed-off-by: David Howells <dhowells@redhat.com>
8 years agonet: sched: use pfifo_fast for non real queues
Eric Dumazet [Wed, 2 Mar 2016 16:21:43 +0000 (08:21 -0800)]
net: sched: use pfifo_fast for non real queues

Some devices declare a high number of TX queues, then set a much
lower real_num_tx_queues

This cause setups using fq_codel, sfq or fq as the default qdisc to consume
more memory than really needed.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>