platform/kernel/linux-starfive.git
9 years agomwifiex: add SDIO rx single port aggregation
Zhaoyang Liu [Fri, 13 Mar 2015 12:07:58 +0000 (17:37 +0530)]
mwifiex: add SDIO rx single port aggregation

This patch brings in support for SDIO single port rx aggregation
to mwifiex.
Maximum read size support by SDIO cmd53 is 64K.
Based on multi port aggregation which is already part of mwifiex, idea here
is multiple packets received in FW can be aggregated into single buffer.
A separate upload type is defined for such packet aggregated to single port.
Packets from this single buffer are later deaggregated into individual packets.
This way, driver can receive more packets each time through single SDIO cmd53;
thereby reducing no of times MMC bus is accessed.

SDIO SP aggregation support is advertised by FW during load time and driver
would get FW block size in command response of HostCmd_CMD_SDIO_SP_RX_AGGR_CFG.

Signed-off-by: Zhaoyang Liu <liuzy@marvell.com>
Signed-off-by: Marc Yang <yangyang@marvell.com>
Reviewed-by: Amitkumar Karwar <akarwar@marvell.com>
Reviewed-by: Cathy Luo <cluo@marvell.com>
Reviewed-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: preprocess packets from TX queue
Zhaoyang Liu [Fri, 13 Mar 2015 12:07:57 +0000 (17:37 +0530)]
mwifiex: preprocess packets from TX queue

During profiling, we discovered that driver remains idle for time
when pakcet is downloaded to FW but no TX_DONE has been received
i.e. while data_sent is true.

This patch adds enhancement to TX routine where we preprocess
packets from TX queue, make them ready for TX and add them to
separate TX queue.

Signed-off-by: Zhaoyang Liu <liuzy@marvell.com>
Signed-off-by: Marc Yang <yangyang@marvell.com>
Signed-off-by: Chin-ran Lo <crlo@marvell.com>
Reviewed-by: Cathy Luo <cluo@marvell.com>
Reviewed-by: Amitkumar Karwar <akarwar@marvell.com>
Reviewed-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: remove_bss_prio_lock
Zhaoyang Liu [Fri, 13 Mar 2015 12:07:56 +0000 (17:37 +0530)]
mwifiex: remove_bss_prio_lock

This patch does away with spinlock in
mwifiex_wmm_get_highest_priolist_ptr in order to improve TP.

Signed-off-by: Zhaoyang Liu <liuzy@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: get rid of BA setup helper functions
Zhaoyang Liu [Fri, 13 Mar 2015 12:07:55 +0000 (17:37 +0530)]
mwifiex: get rid of BA setup helper functions

This patch removes BA setup helper routines
mwifiex_is_bastream_setup and  mwifiex_is_amsdu_in_ampdu_allowed.

Current code will use two functions to check bastream setup and
amsdu in ampdu. This patch change these functions to flags, thus
avoiding redundant spin_lock check while dequeuing TX packets.

Signed-off-by: Zhaoyang Liu <liuzy@marvell.com>
Reviewed-by: Cathy Luo <cluo@marvell.com>
Reviewed-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: avoid queue_work while work is ongoing
Shengzhen Li [Fri, 13 Mar 2015 12:07:54 +0000 (17:37 +0530)]
mwifiex: avoid queue_work while work is ongoing

Current code does not check whether main_work_queue or
rx_work_queue is running when preparing to do queue_work,
this code fix add check before calling queue_work, reducing
unnecessary queue_work switch.

This change instead sets more_task flag to ensure we run main_process
superloop once again.

Signed-off-by: Shengzhen Li <szli@marvell.com>
Signed-off-by: Zhaoyang Liu <liuzy@marvell.com>
Reviewed-by: Cathy Luo <cluo@marvell.com>
Reviewed-by: Amitkumar Karwar <akarwar@marvell.com>
Reviewed-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: enhance SD8897 MP aggregation limits
Avinash Patil [Fri, 13 Mar 2015 12:07:53 +0000 (17:37 +0530)]
mwifiex: enhance SD8897 MP aggregation limits

SD8897 support buffers of 4K and 16 such ports can be accomodated.
So basically 64K buffer size in single aggregation is supported.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: rename alloc_rx_buf to alloc_dma_aligned_buf
Avinash Patil [Fri, 13 Mar 2015 12:07:52 +0000 (17:37 +0530)]
mwifiex: rename alloc_rx_buf to alloc_dma_aligned_buf

Rename this function to reflect its purpose correctly.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: lock main process till reinitialization of vif is over
Avinash Patil [Fri, 13 Mar 2015 12:07:51 +0000 (17:37 +0530)]
mwifiex: lock main process till reinitialization of vif is over

A crash was detected while changing virtual interface type is in
progress. This was tracked to race condition in accessing bss_priority
table while change is in progress. This patch ensures that main_process
and rx_process works are locked while we change virtual interface.

Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowl18xx: show rx_frames_per_rates as an array as it really is
Nicolas Iooss [Fri, 13 Mar 2015 07:17:14 +0000 (15:17 +0800)]
wl18xx: show rx_frames_per_rates as an array as it really is

In struct wl18xx_acx_rx_rate_stat, rx_frames_per_rates field is an
array, not a number.  This means WL18XX_DEBUGFS_FWSTATS_FILE can't be
used to display this field in debugfs (it would display a pointer, not
the actual data).  Use WL18XX_DEBUGFS_FWSTATS_FILE_ARRAY instead.

This bug has been found by adding a __printf attribute to
wl1271_format_buffer.  gcc complained about "format '%u' expects
argument of type 'unsigned int', but argument 5 has type 'u32 *'".

Fixes: c5d94169e818 ("wl18xx: use new fw stats structures")
Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agort2x00usb: drop rt2x00usb_disable_radio() from rt2800usb_disable_radio()
Sebastian Andrzej Siewior [Thu, 12 Mar 2015 14:35:46 +0000 (15:35 +0100)]
rt2x00usb: drop rt2x00usb_disable_radio() from rt2800usb_disable_radio()

I have here FRITZ!WLAN USB Stick N v2 / idVendor=057c, idProduct=8501
and every single of this requests ends up with:

|ieee80211 phy0: rt2x00usb_vendor_request: Error - Vendor Request 0x0c  failed for offset 0x0000 with error -32

I was browsing the the vedor code and I haven't seen such request so I
remove it with this patch. If I wasn't look enough or if this error is
expected then please excuse :)

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agort2x00usb: check USB's request error code in rt2800usb_autorun_detect()
Sebastian Andrzej Siewior [Thu, 12 Mar 2015 14:35:45 +0000 (15:35 +0100)]
rt2x00usb: check USB's request error code in rt2800usb_autorun_detect()

rt2800usb_autorun_detect() blindly assumes assumes that it worked while
it could have failed. Check the error code here.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agort2x00usb: initialize the read value in case of failure
Sebastian Andrzej Siewior [Thu, 12 Mar 2015 14:35:44 +0000 (15:35 +0100)]
rt2x00usb: initialize the read value in case of failure

My understanding ist that rt2x00usb_register_read() is void and so the
reader is unaware of read errors and assumes that whatever was on the
stack as it was about to read.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: use del_timer variant in interrupt context
Amitkumar Karwar [Thu, 12 Mar 2015 07:38:40 +0000 (00:38 -0700)]
mwifiex: use del_timer variant in interrupt context

We might be in interrupt context at few places. So replace
del_timer_sync() with del_timer(). This patch fixes a kernel
trace problem seen occasionally during our testing.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: Add support for auto ARP in mwifiex.
Maithili Hinge [Thu, 12 Mar 2015 07:38:39 +0000 (00:38 -0700)]
mwifiex: Add support for auto ARP in mwifiex.

This patch adds support for auto ARP feature in mwifiex.
The device will respond to ARP requests from the network
with ARP response in suspended state without waking up the host.
This feature is enabled in the driver by default.

Signed-off-by: Maithili Hinge <maithili@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agomwifiex: fix a bug in Rx multiport aggregation logic
Cathy Luo [Thu, 12 Mar 2015 08:29:31 +0000 (01:29 -0700)]
mwifiex: fix a bug in Rx multiport aggregation logic

It's been observed Rx aggregated packets are always followed
by a single Rx packet. This patch improves our logic to add
that extra packet in next aggregation.

Signed-off-by: Cathy Luo <cluo@marvell.com>
Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: No need to export rtl_evm_dbm_jaguar anymore
Priit Laes [Wed, 11 Mar 2015 20:03:27 +0000 (22:03 +0200)]
rtlwifi: No need to export rtl_evm_dbm_jaguar anymore

This function is used only by rtl8821ae so move it there

Signed-off-by: Priit Laes <plaes@plaes.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: Clean rtl_evm_db_to_percentage a bit
Priit Laes [Wed, 11 Mar 2015 20:03:25 +0000 (22:03 +0200)]
rtlwifi: Clean rtl_evm_db_to_percentage a bit

Signed-off-by: Priit Laes <plaes@plaes.org>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: allow 40MHz radar detection width
Zefir Kurtisi [Tue, 10 Mar 2015 16:49:31 +0000 (17:49 +0100)]
ath9k: allow 40MHz radar detection width

Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: add DFS support for extension channel
Zefir Kurtisi [Tue, 10 Mar 2015 16:49:30 +0000 (17:49 +0100)]
ath9k: add DFS support for extension channel

In HT40 modes, pulse events on primary and extension
channel are processed individually. If valid, a pulse
event will be fed into the detector
* for primary frequency, or
* for extension frequency (+/-20MHz based on HT40-mode)
* or both

With that, a 40MHz radar will result in two individual
radar events.

Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: restart only triggering DFS detector line
Zefir Kurtisi [Tue, 10 Mar 2015 16:49:29 +0000 (17:49 +0100)]
ath9k: restart only triggering DFS detector line

To support HT40 DFS mode, a triggering detector must
reset only itself but not other detector lines.

Signed-off-by: Zefir Kurtisi <zefir.kurtisi@neratec.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoRevert "bcma: Kconfig: Let it depend on PCI"
Rafał Miłecki [Thu, 5 Mar 2015 17:25:12 +0000 (18:25 +0100)]
Revert "bcma: Kconfig: Let it depend on PCI"

This reverts commit b09f5ec18b16b82f4db8a735e453332db7514275.

Now that we have fully working BCMA_DRIVER_PCI symbol (in can be safely
disabled), there is no risk bcma will try to use PCI code without PCI
available.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agobcma: allow disabling (not building) PCI driver
Rafał Miłecki [Thu, 5 Mar 2015 17:25:11 +0000 (18:25 +0100)]
bcma: allow disabling (not building) PCI driver

It isn't required for bcma bus on SoCs, so provide some empty functions
and allow disabling it.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agobcma: move PCI IRQ control function to host specific code
Rafał Miłecki [Thu, 5 Mar 2015 17:25:10 +0000 (18:25 +0100)]
bcma: move PCI IRQ control function to host specific code

This function isn't really related to any bus core. It touches PCI
device config registers only, so move it to the (PCI) host file.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: rtl8192cu: remove unused arguments from _beacon_function_enable()
Taehee Yoo [Mon, 9 Mar 2015 15:07:08 +0000 (00:07 +0900)]
rtlwifi: rtl8192cu: remove unused arguments from _beacon_function_enable()

Remove unnecessary parameter in rtl8192cu/hw.c

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agolibertas_tf: if_usb.c: don't export static symbol
Julia Lawall [Wed, 11 Mar 2015 16:56:30 +0000 (17:56 +0100)]
libertas_tf: if_usb.c: don't export static symbol

The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
type T;
identifier f;
@@

static T f (...) { ... }

@@
identifier r.f;
declarer name EXPORT_SYMBOL_GPL;
@@

-EXPORT_SYMBOL_GPL(f);
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agobrcmfmac: fix watchdog timer regression
Arend van Spriel [Wed, 11 Mar 2015 15:11:33 +0000 (16:11 +0100)]
brcmfmac: fix watchdog timer regression

The watchdog timer is used to put the device in a low-power mode when
it is idle for some time. This timer is stopped during that mode and
should be restarted upon activity. This has been broken by commit
d4150fced0365 ("brcmfmac: Simplify watchdog sleep."). This patch
restores the behaviour as it was before that commit.

Reported-by: Pontus Fuchs <pontusf@broadcom.com>
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>
9 years agobrcmfmac: take save&restore memory into account for SDIO shared info
Arend van Spriel [Wed, 11 Mar 2015 15:11:32 +0000 (16:11 +0100)]
brcmfmac: take save&restore memory into account for SDIO shared info

The firmware provides pointer to SDIO shared information at end of
RAM during firmware initialization. End of RAM is obviously determined
by the actual ram size, but part of that may be used for save&restore
memory. In that case another location in RAM will hold the pointer.

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>
9 years agobrcmfmac: extract ram size info from internal memory registers
Arend van Spriel [Wed, 11 Mar 2015 15:11:31 +0000 (16:11 +0100)]
brcmfmac: extract ram size info from internal memory registers

Instead of hard-coded memory sizes it is possible to obtain that
information from the internal memory registers.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@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>
9 years agobrcmfmac: assure device is ready for download after brcmf_chip_attach()
Arend van Spriel [Wed, 11 Mar 2015 15:11:30 +0000 (16:11 +0100)]
brcmfmac: assure device is ready for download after brcmf_chip_attach()

Make the brcmf_chip_attach() function responsible for putting the
device in a state where it is accessible for firmware download.

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>
9 years agobrcmfmac: rename chip download functions
Arend van Spriel [Wed, 11 Mar 2015 15:11:29 +0000 (16:11 +0100)]
brcmfmac: rename chip download functions

The functions brcmf_chip_[enter/exit]_download() are not exclusively
used for firmware download so rename these more appropriate.

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>
9 years agobrcmfmac: always perform cores checks
Arend van Spriel [Wed, 11 Mar 2015 15:11:28 +0000 (16:11 +0100)]
brcmfmac: always perform cores checks

Instead of checking the cores in the chip only if CONFIG_BRCMDBG
is selected perform the check always and extend it with more sanity
checking.

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>
9 years agobrcmfmac: remove duplication of ramsize info
Arend van Spriel [Wed, 11 Mar 2015 15:11:27 +0000 (16:11 +0100)]
brcmfmac: remove duplication of ramsize info

Removing the ramsize from the brcmf_sdio structure to avoid
duplication. The information is available in brcmf_chip
structure.

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>
9 years agoath9k: make ath_frame_info fit into reduced-size rate_driver_data
Florian Westphal [Tue, 10 Mar 2015 14:44:22 +0000 (15:44 +0100)]
ath9k: make ath_frame_info fit into reduced-size rate_driver_data

pre-requisite to shrink size of ieee80211_tx_info which in turn is
needed to shrink skb->cb to 40 bytes again.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoar5523: use container_of
Florian Westphal [Tue, 10 Mar 2015 14:25:39 +0000 (15:25 +0100)]
ar5523: use container_of

If we want to shrink skb->cb then we'd have to see about
reducing struct ieee80211_tx_info, which gets embedded inside
skb->cb[].

It provides a scratch space to be used by wireless drivers.
ar5523 uses the maximum space available today (40 bytes), but it seems
we don't need this -- data->skb pointer seems to always point back to the
skb whose cb buffer the data structure resides, iow, given a pointer to the
embedded control buffer we can infer the skb address.

Tested-by: Pontus Fuchs <pontus.fuchs@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: Fix RTT chainmask usage
Sujith Manoharan [Mon, 9 Mar 2015 08:50:10 +0000 (14:20 +0530)]
ath9k: Fix RTT chainmask usage

Since the RTT registers need to be configured for all
valid chains irrespective of the runtime chainmask,
use the actual chainmask of the chip.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: Check allowed PCIE powersave configuration
Sujith Manoharan [Mon, 9 Mar 2015 08:50:09 +0000 (14:20 +0530)]
ath9k: Check allowed PCIE powersave configuration

When assigning the initvals for PCIE sleep/awake
registers, check the configuration that has been
assigned to pll_pwrsave during initialization. Also,
display a warning if we don't have valid arrays.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: Initialize pll_pwrsave for AR9462/AR9565
Sujith Manoharan [Mon, 9 Mar 2015 08:50:08 +0000 (14:20 +0530)]
ath9k: Initialize pll_pwrsave for AR9462/AR9565

Cards based on AR9462/AR9565 support more PCIE
power save mechanisms, so register them correctly.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: Fix PLL powersave for AR9485
Sujith Manoharan [Mon, 9 Mar 2015 08:50:07 +0000 (14:20 +0530)]
ath9k: Fix PLL powersave for AR9485

Use the value in ah->config.pll_pwrsave to determine
which array needs to be loaded. Also, initialize
pll_pwrsave to 1 by default.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath9k: Add PCIE powersave macros
Sujith Manoharan [Mon, 9 Mar 2015 08:50:06 +0000 (14:20 +0530)]
ath9k: Add PCIE powersave macros

These will be used to handle chip-specific
power save configuration.

Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowil6210: support AP isolation
Vladimir Kondratiev [Sun, 8 Mar 2015 13:42:03 +0000 (15:42 +0200)]
wil6210: support AP isolation

For the AP, configuration may say not to bridge traffic between
wireless clients. This is conveyed from user space (ex: hostapd has
ap_isolate parameter) with NL80211_CMD_SET_BSS, to the driver's
cfg80211 ops method change_bss

Add support for this setting.

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowil6210: re-submit Rx frames to the wireless media if appropriate
Vladimir Kondratiev [Sun, 8 Mar 2015 13:42:02 +0000 (15:42 +0200)]
wil6210: re-submit Rx frames to the wireless media if appropriate

This is for AP only. If Rx data frame targeted to one of associated clients,
transmit it back to the wireless media and don't deliver to the host.
For the multicast frames, deliver to both host and wireless media.

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agowil6210: NAPI completion refactor
Vladimir Kondratiev [Sun, 8 Mar 2015 13:42:01 +0000 (15:42 +0200)]
wil6210: NAPI completion refactor

It is expected that driver completes NAPI when less than
full budget is consumed.

Fulfill this requirement.

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agobrcmfmac: Add support for BCM4345 SDIO chipset.
Syed Asifful Dayyan [Fri, 6 Mar 2015 17:40:42 +0000 (18:40 +0100)]
brcmfmac: Add support for BCM4345 SDIO chipset.

These changes add support for BCM4345 SDIO chipset.

Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend Van Spriel <arend@broadcom.com>
Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Daniel (Deognyoun) Kim <dekim@broadcom.com>
Signed-off-by: Syed Asifful Dayyan <syedd@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agobrcmfmac: Fix race condition in msgbuf ioctl processing.
Hante Meuleman [Fri, 6 Mar 2015 17:40:41 +0000 (18:40 +0100)]
brcmfmac: Fix race condition in msgbuf ioctl processing.

Msgbuf is using a wait_event_timeout to wait for the response on
an ioctl. The wakeup routine uses waitqueue_active to see if
wait_event_timeout has been called. There is a chance that the
response arrives before wait_event_timeout is called, this
will result in situation that wait_event_timeout never gets
woken again and assumed result will be a timeout. This patch
removes that errornous situation by always setting the
ctl_completed var before checking for queue active.

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>
9 years agobrcmfmac: Fix possible race-condition.
Hante Meuleman [Fri, 6 Mar 2015 17:40:40 +0000 (18:40 +0100)]
brcmfmac: Fix possible race-condition.

SDIO is using a "shared" variable to handoff ctl frames to DPC
and to see when they are done. In a timeout situation this can
lead to erroneous situation where DPC started to handle the ctl
frame while the timeout expired. This patch will fix this by
adding locking around the shared variable.

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: Kalle Valo <kvalo@codeaurora.org>
9 years agobrcmfmac: Simplify watchdog sleep.
Hante Meuleman [Fri, 6 Mar 2015 17:40:39 +0000 (18:40 +0100)]
brcmfmac: Simplify watchdog sleep.

The watchdog thread is used to put the SDIO bus to sleep when the
system is idling. This patch simplifies the way it is determined
when sleep can be entered.

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: Kalle Valo <kvalo@codeaurora.org>
9 years agobrcmfmac: Fix oops when SDIO device is removed.
Hante Meuleman [Fri, 6 Mar 2015 17:40:38 +0000 (18:40 +0100)]
brcmfmac: Fix oops when SDIO device is removed.

On removal of SDIO card both functions of card will be getting
a remove call. When the first is hanging in ctrl frame xmit then
the second will cause oops. This patch fixes the xmit ctrl
handling in case of serious errors and also limits the handling
for remove to function 1 only.

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: Kalle Valo <kvalo@codeaurora.org>
9 years agortlwifi: Remove duplicated prototype
Taehee Yoo [Fri, 6 Mar 2015 15:47:17 +0000 (00:47 +0900)]
rtlwifi: Remove duplicated prototype

Remove duplicated prototype in base.h

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath5k: fix reset race
Sergey Ryazanov [Wed, 4 Mar 2015 02:12:11 +0000 (05:12 +0300)]
ath5k: fix reset race

To prepare for reset ath5k should finish all asynchronous tasks. At
first, it disables the interrupt generation, then it waits for the
interrupt handler and tasklets completion, and then proceeds to the HW
configuration update. But it does not consider that the interrupt
handler or tasklet re-enables the interrupt generation. And we fall in a
situation when ath5k assumes that interrupts are disabled, but it is
not.

This can lead to different consequences, such as reception of the frame,
when we do not expect it. Under certain circumstances, this can lead to
the following warning:

  WARNING: at ath5k/base.c:589 ath5k_tasklet_rx+0x318/0x6ec [ath5k]()
  invalid hw_rix: 1a
  [..]
  Call Trace:
  [<802656a8>] show_stack+0x48/0x70
  [<802dd92c>] warn_slowpath_common+0x88/0xbc
  [<802dd98c>] warn_slowpath_fmt+0x2c/0x38
  [<81b51be8>] ath5k_tasklet_rx+0x318/0x6ec [ath5k]
  [<8028ac64>] tasklet_action+0x8c/0xf0
  [<80075804>] __do_softirq+0x180/0x32c
  [<80196ce8>] irq_exit+0x54/0x70
  [<80041848>] ret_from_irq+0x0/0x4
  [<80182fdc>] ioread32+0x4/0xc
  [<81b4c42c>] ath5k_hw_set_sleep_clock+0x2ec/0x474 [ath5k]
  [<81b4cf28>] ath5k_hw_reset+0x50/0xeb8 [ath5k]
  [<81b50900>] ath5k_reset+0xd4/0x310 [ath5k]
  [<81b557e8>] ath5k_config+0x4c/0x104 [ath5k]
  [<80d01770>] ieee80211_hw_config+0x2f4/0x35c [mac80211]
  [<80d09aa8>] ieee80211_scan_work+0x2e4/0x414 [mac80211]
  [<8022c3f4>] process_one_work+0x28c/0x400
  [<802df8f8>] worker_thread+0x258/0x3c0
  [<801b5710>] kthread+0xe0/0xec
  [<800418a8>] ret_from_kernel_thread+0x14/0x1c

Fix this issue by adding a new status flag, which forbids to re-enable
the interrupt generation until the HW configuration is completed.

Note: previous patch, which reorders the Rx disable code helps to avoid
the above warning, but not fixes the root cause of unexpected frame
receiving.

CC: Jiri Slaby <jirislaby@gmail.com>
CC: Nick Kossifidis <mickflemm@gmail.com>
CC: Luis R. Rodriguez <mcgrof@do-not-panic.com>
Reported-by: Christophe Prevotaux <cprevotaux@nltinc.com>
Tested-by: Christophe Prevotaux <cprevotaux@nltinc.com>
Tested-by: Eric Bree <ebree@nltinc.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoath5k: channel change fix
Sergey Ryazanov [Wed, 4 Mar 2015 02:12:10 +0000 (05:12 +0300)]
ath5k: channel change fix

ath5k updates the channel pointer and after that it stops the Rx logic
and apply channel to HW. In case of channel switch, such sequence
creates a small window when a frame, which is received on the old
channel is considered as a frame received on the new one.

The most notable consequence of this situation occurs during the switch
from 2 GHz band (CCK+OFDM) to the 5GHz band (OFDM-only). Frame received
with CCK rate, e.g. beacon received at the 1mbps, causes the following
warning:

  WARNING: at ath5k/base.c:589 ath5k_tasklet_rx+0x318/0x6ec [ath5k]()
  invalid hw_rix: 1a
  [..]
  Call Trace:
  [<802656a8>] show_stack+0x48/0x70
  [<802dd92c>] warn_slowpath_common+0x88/0xbc
  [<802dd98c>] warn_slowpath_fmt+0x2c/0x38
  [<81b51be8>] ath5k_tasklet_rx+0x318/0x6ec [ath5k]
  [<8028ac64>] tasklet_action+0x8c/0xf0
  [<80075804>] __do_softirq+0x180/0x32c
  [<80196ce8>] irq_exit+0x54/0x70
  [<80041848>] ret_from_irq+0x0/0x4
  [<80182fdc>] ioread32+0x4/0xc
  [<81b4c42c>] ath5k_hw_set_sleep_clock+0x2ec/0x474 [ath5k]
  [<81b4cf28>] ath5k_hw_reset+0x50/0xeb8 [ath5k]
  [<81b50900>] ath5k_reset+0xd4/0x310 [ath5k]
  [<81b557e8>] ath5k_config+0x4c/0x104 [ath5k]
  [<80d01770>] ieee80211_hw_config+0x2f4/0x35c [mac80211]
  [<80d09aa8>] ieee80211_scan_work+0x2e4/0x414 [mac80211]
  [<8022c3f4>] process_one_work+0x28c/0x400
  [<802df8f8>] worker_thread+0x258/0x3c0
  [<801b5710>] kthread+0xe0/0xec
  [<800418a8>] ret_from_kernel_thread+0x14/0x1c

The easiest way to reproduce this warning is to run scan with dualband
NIC in noisy environments, when the channel 11 runs multiple APs. In my
tests if the APs num >= 12, the warning appears in the first few
seconds of scanning.

In order to fix this, the Rx disable code moved to a higher level and
placed before the channel pointer update. This is also makes the code a
bit more symmetrical, since we disable and enable the Rx in the same
function.

In fact, at the pointer update time new frames should not appear,
because interrupt generation at this point should already be disabled.
The next patch should address this issue.

CC: Jiri Slaby <jirislaby@gmail.com>
CC: Nick Kossifidis <mickflemm@gmail.com>
CC: Luis R. Rodriguez <mcgrof@do-not-panic.com>
Reported-by: Christophe Prevotaux <cprevotaux@nltinc.com>
Tested-by: Christophe Prevotaux <cprevotaux@nltinc.com>
Tested-by: Eric Bree <ebree@nltinc.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
9 years agoMerge tag 'iwlwifi-next-for-kalle-2015-03-12' of https://git.kernel.org/pub/scm/linux...
Kalle Valo [Fri, 13 Mar 2015 13:09:18 +0000 (15:09 +0200)]
Merge tag 'iwlwifi-next-for-kalle-2015-03-12' of https://git./linux/kernel/git/iwlwifi/iwlwifi-next

* Location Aware Regulatory was added by Arik
* 8000 device family work
* Update to the BT Coex firmware API

9 years agovxlan: Don't set s_addr in vxlan_create_sock
Simon Horman [Fri, 13 Mar 2015 02:00:58 +0000 (11:00 +0900)]
vxlan: Don't set s_addr in vxlan_create_sock

In the case of AF_INET s_addr was set to INADDR_ANY (0) which which both
symmetric with the AF_INET6 case, where s_addr is not set, and unnecessary
as udp_conf is zeroed out earlier in the same function.

I suspect this change does not have any run-time effect due to compiler
optimisations. But it does make the code a little easier on the/my eyes.

Cc: Tom Herbert <therbert@google.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agompls: In mpls_egress verify the packet length.
Eric W. Biederman [Thu, 12 Mar 2015 23:22:59 +0000 (18:22 -0500)]
mpls: In mpls_egress verify the packet length.

Reobert Shearman noticed that mpls_egress is failing to verify that
the bytes to be examined are in fact present in the packet before
mpls_egress reads those bytes.

As suggested by David Miller reduce this to a single pskb_may_pull
call so that we don't do unnecessary work in the fast path.

Reported-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: Only adjust tx_clk on link change
Jaeden Amero [Thu, 12 Mar 2015 23:07:54 +0000 (18:07 -0500)]
net/macb: Only adjust tx_clk on link change

The PHY state machine (in drivers/net/phy/phy.c) will unconditionally
call phydev->adjust_link (macb_handle_link_change) when polling in the
PHY_CHANGELINK state. As currently written, macb always ends up
requesting a new tx_clk frequency in macb_handle_link_change. It is a
waste of time to request a new tx_clk frequency if the link state hasn't
changed, as the tx_clk will already be configured properly.

Let's only request a new tx_clk clock frequency when necessary.

Signed-off-by: Jaeden Amero <jaeden.amero@ni.com>
Cc: Josh Cartwright <joshc@ni.com>
Cc: Soren Brinkmann <soren.brinkmann@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Fix read-side crash during rehash
Herbert Xu [Fri, 13 Mar 2015 01:54:10 +0000 (12:54 +1100)]
rhashtable: Fix read-side crash during rehash

This patch fixes a typo rhashtable_lookup_compare where we fail
to recompute the hash when looking up the new table.  This causes
elements to be missed and potentially a crash during a resize.

Reported-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: kill ht->shift atomic operations
Daniel Borkmann [Thu, 12 Mar 2015 14:28:40 +0000 (15:28 +0100)]
rhashtable: kill ht->shift atomic operations

Commit c0c09bfdc415 ("rhashtable: avoid unnecessary wakeup for worker
queue") changed ht->shift to be atomic, which is actually unnecessary.

Instead of leaving the current shift in the core rhashtable structure,
it can be cached inside the individual bucket tables.

There, it will only be initialized once during a new table allocation
in the shrink/expansion slow path, and from then onward it stays immutable
for the rest of the bucket table liftime.

That allows shift to be non-atomic. The patch also moves hash_rnd
management into the table setup. The rhashtable structure now consumes
3 instead of 4 cachelines.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Ying Xue <ying.xue@windriver.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Fix reader/rehash race
Herbert Xu [Thu, 12 Mar 2015 11:07:49 +0000 (22:07 +1100)]
rhashtable: Fix reader/rehash race

There is a potential race condition between readers and the rehasher.
In particular, the rehasher could have started a rehash while the
reader finishes a scan of the old table but fails to see the new
table pointer.

This patch closes this window by adding smp_wmb/smp_rmb.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'listener_refactor'
David S. Miller [Fri, 13 Mar 2015 02:58:27 +0000 (22:58 -0400)]
Merge branch 'listener_refactor'

Eric Dumazet says:

====================
inet: tcp listener refactoring, part 8

These patches prepare request socks being hashed into general ehash
table : We declare 3 aliases (ireq_state, ireq_refcnt, ireq_family)

Note that refcnt is not yet handled, this will be done later.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: introduce ireq_family
Eric Dumazet [Thu, 12 Mar 2015 23:44:10 +0000 (16:44 -0700)]
inet: introduce ireq_family

Before inserting request socks into general hash table,
fill their socket family.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: get_openreq4() & get_openreq6() do not need listener
Eric Dumazet [Thu, 12 Mar 2015 23:44:09 +0000 (16:44 -0700)]
inet: get_openreq4() & get_openreq6() do not need listener

ireq->ir_num contains local port, use it.

Also, get_openreq4() dumping listen_sk->refcnt makes litle sense.

inet_diag_fill_req() can also use ireq->ir_num

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: prepare sock_edemux() & sock_gen_put() for new SYN_RECV state
Eric Dumazet [Thu, 12 Mar 2015 23:44:08 +0000 (16:44 -0700)]
inet: prepare sock_edemux() & sock_gen_put() for new SYN_RECV state

sock_edemux() & sock_gen_put() should be ready to cope with request socks.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: add req_prot_cleanup() & req_prot_init() helpers
Eric Dumazet [Thu, 12 Mar 2015 23:44:07 +0000 (16:44 -0700)]
net: add req_prot_cleanup() & req_prot_init() helpers

Make proto_register() & proto_unregister() a bit nicer.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: add rsk_refcnt/ireq_refcnt to request socks
Eric Dumazet [Thu, 12 Mar 2015 23:44:06 +0000 (16:44 -0700)]
inet: add rsk_refcnt/ireq_refcnt to request socks

When request socks will be in ehash, they'll need to be refcounted.

This patch adds rsk_refcnt/ireq_refcnt macros, and adds
reqsk_put() function, but nothing yet use them.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: add ireq_state field to inet_request_sock
Eric Dumazet [Thu, 12 Mar 2015 23:44:05 +0000 (16:44 -0700)]
inet: add ireq_state field to inet_request_sock

We need to identify request sock when they'll be visible in
global ehash table.

ireq_state is an alias to req.__req_common.skc_state.

Its value is set to TCP_NEW_SYN_RECV

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinet: add TCP_NEW_SYN_RECV state
Eric Dumazet [Thu, 12 Mar 2015 23:44:04 +0000 (16:44 -0700)]
inet: add TCP_NEW_SYN_RECV state

TCP_SYN_RECV state is currently used by fast open sockets.

Initial TCP requests (the pseudo sockets created when a SYN is received)
are not yet associated to a state. They are attached to their parent,
and the parent is in TCP_LISTEN state.

This commit adds TCP_NEW_SYN_RECV state, so that we can convert
TCP stack to a different schem gradually.

This state is not exported to user space.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: add missing ireq_net & ir_cookie initializations
Eric Dumazet [Thu, 12 Mar 2015 23:44:03 +0000 (16:44 -0700)]
ipv6: add missing ireq_net & ir_cookie initializations

I forgot to update dccp_v6_conn_request() & cookie_v6_check().
They both need to set ireq->ireq_net and ireq->ir_cookie

Lets clear ireq->ir_cookie in inet_reqsk_alloc()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: 33cf7c90fe2f ("net: add real socket cookies")
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocls_bpf: do eBPF invocation under non-bh RCU lock variant for maps
Daniel Borkmann [Thu, 12 Mar 2015 19:03:12 +0000 (20:03 +0100)]
cls_bpf: do eBPF invocation under non-bh RCU lock variant for maps

Currently, it is possible in cls_bpf to access eBPF maps only under
rcu_read_lock_bh() variants: while on ingress side, that is, handle_ing(),
the classifier would be called from __netif_receive_skb_core() under
rcu_read_lock(); on egress side, however, it's rcu_read_lock_bh() via
__dev_queue_xmit().

This rcu/rcu_bh mix doesn't work together with eBPF maps as they require
soley to be called under rcu_read_lock(). eBPF maps could also be shared
among various other eBPF programs (possibly even with other eBPF program
types, f.e. tracing) and user space processes, so any context is assumed.

Therefore, a possible fix for cls_bpf is to wrap/nest eBPF program
invocation under non-bh RCU lock variant.

Fixes: e2e9b6541dd4 ("cls_bpf: add initial eBPF support for programmable classifiers")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'fib_trie_table_merge_fixes'
David S. Miller [Thu, 12 Mar 2015 22:26:58 +0000 (18:26 -0400)]
Merge branch 'fib_trie_table_merge_fixes'

Alexander Duyck says:

====================
fib_trie: Minor fixes for table merge

This patch set addresses two issues reported with the tables merged, the
first is a NULL pointer dereference, and the other is to remove a WARN_ON
and set the ordering for aliases from different tables with the same slen
values.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agofib_trie: Provide a deterministic order for fib_alias w/ tables merged
Alexander Duyck [Thu, 12 Mar 2015 21:46:29 +0000 (14:46 -0700)]
fib_trie: Provide a deterministic order for fib_alias w/ tables merged

This change makes it so that we should always have a deterministic ordering
for the main and local aliases within the merged table when two leaves
overlap.

So for example if we have a leaf with a key of 192.168.254.0.  If we
previously added two aliases with a prefix length of 24 from both local and
main the first entry would be first and the second would be second.  When I
was coding this I had added a WARN_ON should such a situation occur as I
wasn't sure how likely it would be.  However this WARN_ON has been
triggered so this is something that should be addressed.

With this patch the ordering of the aliases is as follows.  First they are
sorted on prefix length, then on their table ID, then tos, and finally
priority.  This way what we end up doing is essentially interleaving the
two tables on what used to be leaf_info structure boundaries.

Fixes: 0ddcf43d5 ("ipv4: FIB Local/MAIN table collapse")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agofib_trie: Avoid NULL pointer if local table is not allocated
Alexander Duyck [Thu, 12 Mar 2015 21:46:23 +0000 (14:46 -0700)]
fib_trie: Avoid NULL pointer if local table is not allocated

The function fib_unmerge assumed the local table had already been
allocated.  If that is not the case however when custom rules are applied
then this can result in a NULL pointer dereference.

In order to prevent this we must check the value of the local table pointer
and if it is NULL simply return 0 as there is no local table to separate
from the main.

Fixes: 0ddcf43d5 ("ipv4: FIB Local/MAIN table collapse")
Reported-by: Madhu Challa <challa@noironetworks.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoebpf: verifier: check that call reg with ARG_ANYTHING is initialized
Daniel Borkmann [Thu, 12 Mar 2015 16:21:42 +0000 (17:21 +0100)]
ebpf: verifier: check that call reg with ARG_ANYTHING is initialized

I noticed that a helper function with argument type ARG_ANYTHING does
not need to have an initialized value (register).

This can worst case lead to unintented stack memory leakage in future
helper functions if they are not carefully designed, or unintended
application behaviour in case the application developer was not careful
enough to match a correct helper function signature in the API.

The underlying issue is that ARG_ANYTHING should actually be split
into two different semantics:

  1) ARG_DONTCARE for function arguments that the helper function
     does not care about (in other words: the default for unused
     function arguments), and

  2) ARG_ANYTHING that is an argument actually being used by a
     helper function and *guaranteed* to be an initialized register.

The current risk is low: ARG_ANYTHING is only used for the 'flags'
argument (r4) in bpf_map_update_elem() that internally does strict
checking.

Fixes: 17a5267067f3 ("bpf: verifier (add verifier core)")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'possible_net_t'
David S. Miller [Thu, 12 Mar 2015 18:39:44 +0000 (14:39 -0400)]
Merge branch 'possible_net_t'

Eric W. Biederman says:

====================
Introduce possible_net_t

The current usage of write_pnet and read_pnet is a little laborious and
error prone as you only notice if you failed to include them if are
compiling with network namespaces enabled.

possible_net_t remedies that by using a type that is 0 bytes when
network namespaces are disabled and can only be read and written to with
read_pnet and write_pnet.

Aka less work and safer for the same effect.

I kill hold_net and release_net first as are they are haven't been used
since 2008 and are noise at the points where write_pnet and read_pnet
are used.

I have folded in Eric Dumazets suggestions to improve the killing of
hold_net and release net.  And respon.  I had to respin anyway as
there was enough changes elsewhere in the tree the previous version
of these patches did not quite apply cleanly.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Introduce possible_net_t
Eric W. Biederman [Thu, 12 Mar 2015 04:06:44 +0000 (23:06 -0500)]
net: Introduce possible_net_t

Having to say
> #ifdef CONFIG_NET_NS
>  struct net *net;
> #endif

in structures is a little bit wordy and a little bit error prone.

Instead it is possible to say:
> typedef struct {
> #ifdef CONFIG_NET_NS
>       struct net *net;
> #endif
> } possible_net_t;

And then in a header say:

>  possible_net_t net;

Which is cleaner and easier to use and easier to test, as the
possible_net_t is always there no matter what the compile options.

Further this allows read_pnet and write_pnet to be functions in all
cases which is better at catching typos.

This change adds possible_net_t, updates the definitions of read_pnet
and write_pnet, updates optional struct net * variables that
write_pnet uses on to have the type possible_net_t, and finally fixes
up the b0rked users of read_pnet and write_pnet.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: Kill hold_net release_net
Eric W. Biederman [Thu, 12 Mar 2015 04:04:08 +0000 (23:04 -0500)]
net: Kill hold_net release_net

hold_net and release_net were an idea that turned out to be useless.
The code has been disabled since 2008.  Kill the code it is long past due.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'rhashtable-cleanups'
David S. Miller [Thu, 12 Mar 2015 18:35:35 +0000 (14:35 -0400)]
Merge branch 'rhashtable-cleanups'

Herbert Xu says:

====================
rhashtable hash cleanups

This is a rebase on top of the nested lock annotation fix.

Nothing to see here, just a bunch of simple clean-ups before
I move onto something more substantial (hopefully).
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Remove obj_raw_hashfn
Herbert Xu [Thu, 12 Mar 2015 03:49:41 +0000 (14:49 +1100)]
rhashtable: Remove obj_raw_hashfn

Now that the only caller of obj_raw_hashfn is head_hashfn, we can
simply kill it and fold it into the latter.

This patch also moves the common shift from head_hashfn/key_hashfn
into rht_bucket_index.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Remove key length argument to key_hashfn
Herbert Xu [Thu, 12 Mar 2015 03:49:40 +0000 (14:49 +1100)]
rhashtable: Remove key length argument to key_hashfn

key_hashfn has only one caller and it doesn't really need to supply
the key length as an extra parameter.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Use head_hashfn instead of obj_raw_hashfn
Herbert Xu [Thu, 12 Mar 2015 03:49:39 +0000 (14:49 +1100)]
rhashtable: Use head_hashfn instead of obj_raw_hashfn

Now that we don't have cross-table hashes, we no longer need to
keep the entire hash value so all users of obj_raw_hashfn can
use head_hashfn instead.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Move masking back into key_hashfn
Herbert Xu [Thu, 12 Mar 2015 03:49:38 +0000 (14:49 +1100)]
rhashtable: Move masking back into key_hashfn

This patch reverts commit c88455ce50ae4224d84960ce2baa53e61580df27
("rhashtable: key_hashfn() must return full hash value") because
the only user of it always masks the hash value.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'iwlwifi-fixes' into iwlwifi-next
Emmanuel Grumbach [Thu, 12 Mar 2015 12:38:26 +0000 (14:38 +0200)]
Merge branch 'iwlwifi-fixes' into iwlwifi-next

9 years agoiwlwifi: mvm: simplify iwl_mvm_get_wakeup_status() return
Johannes Berg [Tue, 3 Mar 2015 19:23:47 +0000 (20:23 +0100)]
iwlwifi: mvm: simplify iwl_mvm_get_wakeup_status() return

The return value in iwl_mvm_get_wakeup_status() is a bit unclear in
that it's not obvious that we don't leak fw_status in some cases.
Use fw_status directly with ERR_PTR() and return only it, that way
the compiler has a chance of proving that it's uninitialized (if it
ever is due to new changes.)

Additionally, this removes a smatch warning since smatch couldn't
figure out that fw_status can't, in fact, leak here.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: don't double unlock the mutex in __iwl_mvm_resume()
Luciano Coelho [Thu, 12 Mar 2015 07:25:15 +0000 (09:25 +0200)]
iwlwifi: mvm: don't double unlock the mutex in __iwl_mvm_resume()

When IWLWIFI_DEBUGFS is not set, we should not unlock the mutex after
calling iwl_mvm_query_wakeup_reasons(), because this function unlocks
it already.  Move the goto out_iterate outside the #ifdef.

Change-Id: I13d86402aecf0eeec44b1abbe2b244fbc706a5eb
Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
9 years agoiwlwifi: mvm: clarify time event end handling
Johannes Berg [Wed, 11 Mar 2015 19:27:06 +0000 (20:27 +0100)]
iwlwifi: mvm: clarify time event end handling

The code here is a little confusing, the iwl_mvm_te_check_disconnect()
will check that the interface is a station, but going into it after
already having processed the time even end for P2P seems strange at
first look.

Put a switch statement there to distinguish the interface types and
make this more readable.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: Always enable the smart FIFO
Eran Harary [Sun, 8 Feb 2015 11:58:50 +0000 (13:58 +0200)]
iwlwifi: mvm: Always enable the smart FIFO

We previously enabled the smart FIFO (SF) in BSS only after
association.
This cause interrupt latency on P2P on certain devices.
Change the working model to enable the SF all the time and
play with the timeout values based on the association state.
This change was not tested on older firwmares, so make it
happen only on -13.ucode and up.

Signed-off-by: Eran Harary <eran.harary@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: update copyright to include 2015
Emmanuel Grumbach [Sun, 8 Mar 2015 12:18:17 +0000 (14:18 +0200)]
iwlwifi: update copyright to include 2015

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: add more new 8260 series PCI IDs
Oren Givon [Mon, 9 Mar 2015 09:25:59 +0000 (11:25 +0200)]
iwlwifi: add more new 8260 series PCI IDs

More sub system IDs were introduced for the 8260 series.
Add the new sub system IDs so the cards can be recognized.

Signed-off-by: Oren Givon <oren.givon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: BT Coex - update the new API
Emmanuel Grumbach [Thu, 19 Feb 2015 13:00:18 +0000 (15:00 +0200)]
iwlwifi: mvm: BT Coex - update the new API

The firmware was not using the new API, so we don't need to
differentiate between the different stages of this new API.
The main difference here is that most of the hard coded
values are not sent through the command anymore.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: fix force NMI for 8000
Emmanuel Grumbach [Mon, 9 Mar 2015 10:37:59 +0000 (12:37 +0200)]
iwlwifi: mvm: fix force NMI for 8000

The newer devices will enable a new register for this
(DEVICE_SET_NMI_8000B_REG), but the interrupt handler
isn't wired yet.
Keep the old register for now.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: freeze the non-shared queues when a station goes to sleep
Emmanuel Grumbach [Thu, 22 Jan 2015 11:15:15 +0000 (13:15 +0200)]
iwlwifi: mvm: freeze the non-shared queues when a station goes to sleep

When a station goes to sleep, we can't transmit any frame
to it. This means that until that station will wake up, a
queue that is dedicated to this station won't progress at
all. Take this into account when monitoring stuck queues
and don't account for the time the station was asleep.
This allows to mask false positives where the queues are
stuck not because of a bug, but because of the station
being asleep.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: pcie: allow the op_mode to freeze the stuck queue timer
Emmanuel Grumbach [Tue, 20 Jan 2015 15:02:40 +0000 (17:02 +0200)]
iwlwifi: pcie: allow the op_mode to freeze the stuck queue timer

This allows the op_mode to let the transport know that a
queue is currently frozen and that its timer should be
stopped.
When the queue is unfrozen, its timer should be set to
expire after the remainder of the timeout has elapsed.
This can be used when stations go to sleep. When a station
goes to sleep, the op_mode can freeze the timer so that the
queue will never be considered as stuck. When the station
wakes up, the queue will be unfrozen.
This is meant to avoid false positives that would happen if
a buggy station goes to sleep for a very long time. In case
we have a dedicated queue for this station (BA agreement)
and it goes to sleep for a very long time, the queue would
rightfully be stopped during all that time. In this case,
the stuck queue timer could fire and that would be a false
positive.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: rs: update Tx statistics when using fixed rate
Eyal Shapira [Sun, 8 Mar 2015 17:43:41 +0000 (19:43 +0200)]
iwlwifi: mvm: rs: update Tx statistics when using fixed rate

The Tx statistics weren't updated when using fixed rate for
debugging. Fix this as Tx statistics are useful in this use case.

Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: don't init MCC during CT-kill
Arik Nemtsov [Wed, 4 Mar 2015 13:08:00 +0000 (15:08 +0200)]
iwlwifi: mvm: don't init MCC during CT-kill

RTNL is not taken during CT-kill so regulatory APIs cannot be invoked.
That's fine, since the HW is only brought up to check the temperature
during CT-kill. We don't expect Tx or scanning.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: remove warning on station exhaustion
Johannes Berg [Tue, 10 Mar 2015 19:32:08 +0000 (20:32 +0100)]
iwlwifi: mvm: remove warning on station exhaustion

When using IBSS, it's easily possible to exhaust the number
of available stations in the driver, so don't warn on it.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: trans: Take ownership on secure machine before FW load
Eran Harary [Tue, 3 Mar 2015 11:53:28 +0000 (13:53 +0200)]
iwlwifi: trans: Take ownership on secure machine before FW load

When we load the firmware for the 8000 B step device, it'll
verify its signature. In the current version of the
hardware, there can be a race between the WiFi firmware
being loaded and the Bluetooth firmware being loaded.

Check that WiFi is authenticated, if not, take ownership
on the authentication machine to make sure that the WiFi
firmware will be authenticated.

Signed-off-by: Eran Harary <eran.harary@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: add new 8260 series PCI IDs
Oren Givon [Wed, 4 Mar 2015 14:26:45 +0000 (16:26 +0200)]
iwlwifi: add new 8260 series PCI IDs

New sub system IDs were introduced for the 8260 series.
This patch adds them so new 8260 cards can be recognized.

Signed-off-by: Oren Givon <oren.givon@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: rs: improve ss_params debug print
Eyal Shapira [Mon, 2 Mar 2015 08:35:19 +0000 (10:35 +0200)]
iwlwifi: mvm: rs: improve ss_params debug print

Make the print a bit more readable.

Reported-by: Joe Perches <joe@perches.com>
Signed-off-by: Eyal Shapira <eyalx.shapira@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: don't allow the FW to return invalid ch indices
Arik Nemtsov [Sun, 1 Mar 2015 16:24:57 +0000 (18:24 +0200)]
iwlwifi: don't allow the FW to return invalid ch indices

If the FW returns an invalid channels count in response to an MCC request,
make sure we don't reference invalid indices in the channels array.

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>
9 years agoiwlwifi: mvm: reflect TDLS pm state in mvmvif->pm_enabled
Arik Nemtsov [Sun, 22 Feb 2015 17:15:04 +0000 (19:15 +0200)]
iwlwifi: mvm: reflect TDLS pm state in mvmvif->pm_enabled

When entering D0i3, the MVM mutex cannot be grabbed. This interferes
with the calculation of the number of connected TDLS stations during
the setup of the power cmd.
The goal is to disable power saving for all vifs while any TDLS station
is connected. For this purpose it is enough to keep the pm_enabled
member of all mvmvifs as false. An update of the power state already
occurs when a TDLS station is added/removed, so the values are correctly
updated.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: fix identation
Emmanuel Grumbach [Mon, 2 Mar 2015 12:39:03 +0000 (14:39 +0200)]
iwlwifi: mvm: fix identation

mvm->fw->dbg_dest_tlv really needs to be under the right
parenthesis.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
9 years agoiwlwifi: mvm: remove unneeded include iwl-fw-error-dump.h
Emmanuel Grumbach [Mon, 2 Mar 2015 12:35:41 +0000 (14:35 +0200)]
iwlwifi: mvm: remove unneeded include iwl-fw-error-dump.h

The functions related to firmware error dump moved. No need
for this unclude anymore.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>