wl12xx: start/stop queues according to global per-AC counters

Split tx_queue_count to count per-AC skb's queued, instead of relying on
the skb-queue len. The skb queues used were only valid in STA-mode, as
AP-mode uses per-link queues.

This fixes a major regression in AP-mode, caused by the patch
"wl12xx: implement Tx watermarks per AC". With that patch applied, we
effectively had no regulation of Tx queues in AP-mode. Therefore a
sustained high rate of Tx could cause exhaustion of the skb memory pool.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: always initialize AP-mode max power level

Sometimes we only get the Tx power level via op->config when the FW is
off. Record the received power level when this happens and use it to
initialize the firmware during boot.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: initialize max_power levels for 11a channels

Set the maximum FW supported power levels for 11a channels.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: implement cancel_hw_scan() callback

After implementing cancel_hw_scan(), there is no more need to
flush scan_complete_work on suspend.

scan_complete_work enqueues a new scan_work, that might lead
to the following warning (if the device was already suspended):

WARNING: at drivers/mmc/core/core.c:169 mmc_wait_for_req+0x1ac/0x2c8()
Backtrace:
[<c004b50c>] (dump_backtrace+0x0/0x110)
[<c04fda50>] (dump_stack+0x0/0x1c)
[<c008f700>] (warn_slowpath_common+0x0/0x6c)
[<c008f76c>] (warn_slowpath_null+0x0/0x2c)
[<c034b058>] (mmc_wait_for_req+0x0/0x2c8)
[<c034f804>] (mmc_io_rw_extended+0x0/0x1e4)
[<c0350b5c>] (sdio_io_rw_ext_helper+0x0/0x1ac)
[<c0350e00>] (sdio_memcpy_fromio+0x0/0x2c)
[<bf194464>] (wl1271_sdio_raw_read+0x0/0x1bc [wl12xx_sdio])
[<bf15ac14>] (wl1271_cmd_wait_for_event_or_timeout+0x0/0x19c [wl12xx])
[<bf15adb0>] (wl1271_cmd_wait_for_event+0x0/0x30 [wl12xx])
[<bf15b7dc>] (wl1271_cmd_disconnect+0x0/0x14c [wl12xx])
[<bf158fd0>] (wl1271_unjoin+0x0/0x48 [wl12xx])
[<bf159070>] (wl1271_op_config+0x0/0x490 [wl12xx])
[<bf105084>] (ieee80211_hw_config+0x0/0x14c [mac80211])
[<bf113120>] (ieee80211_recalc_idle+0x0/0x4c [mac80211])
[<bf109a70>] (__ieee80211_scan_completed+0x0/0x1ac [mac80211])
[<bf109fd4>] (ieee80211_scan_work+0x0/0x484 [mac80211])
[<c00a197c>] (worker_thread+0x0/0x1e8)
[<c00a5874>] (kthread+0x0/0x90)

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: add stop_scan command

define wl1271_scan_stop() to cancel an ongoing hw scan.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: configure suspend/resume only if associated/started

Configure the device (to wowlan) only if the sta/ap
is associated/started.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: one SDIO initialization is enough

Skip manual device power on in case runtime PM is enabled for
our device.

This eliminates a secondary & redundant SDIO init sequence.

Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: implement Tx watermarks per AC

Each AC is stopped when its queue is filled up to the high watermark,
and restarted when its queue it lower than the low watermark. This
ensures congested ACs are not able to starve other ACs.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Increase low & high TX watermarks

The current TX watermark settings cause the driver to stop queues too
frequently. Among other things, this can have a negative impact on WMM
prioritization, since mac80211 sorts pending packets by their ACs.

Fix this by increasing the high watermark to 256 packets. Increase the
low watermark to 32 to minimize periods with queues being stopped.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: schedule TX packets according to FW occupancy

When selecting packets for transmission, prefer the ACs that are least
occupied in the FW. When packets for multiple ACs are present in the FW,
it decides which to transmit according to WMM QoS parameters.

With these changes, lower priority ACs should not be starved when higher
priority traffic is present.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: use 802.11 header location after relocation to frame start

When operating with TKIP encryption, the function wl1271_tx_fill_hdr()
relocates the 802.11 header to the start of the frame, and leaves room
for the security header.

Some functions in the Tx path rely on the location of the header,
namely, for purposes of roaming in STA mode and connecting new stations
in AP mode. Call these functions only after the header is relocated.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: AP mode - support FW TX inactivity triggers

In AP mode we register for the MAX_TX_RETRY and INACTIVE_STA events.
Both are reported to the upper layers as a TX failure in the offending
stations.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: fix Tx security sequence number handling

Do not reset the security sequence number when issuing a join command or
interface is removed. Instead, reset the counter only during the unjoin
command.

Added the notion of counter wrap-around to the LSB number in
wl1271_tx_complete_packet.

Added post recovery padding to adjust for potential security number
progress during the recovery process by the firmware and avoid
potential interop issues in encrypted networks.

Signed-off-by: Oz Krakowski <ozk@ti.com>
Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Support routing FW logs to the host

A recently added feature to the firmware enables the driver to retrieve
firmware logs via the host bus (SDIO or SPI).

There are two modes of operation:
1. On-demand: The FW collects its log in an internal ring buffer. This
   buffer can later be read, for example, upon recovery.
2. Continuous: The FW pushes the FW logs as special packets in the RX
   path.

Reading the internal ring buffer does not involve the FW. Thus, as long
as the HW is not in ELP, it should be possible to read the logs, even if
the FW crashes.

A sysfs binary file named "fwlog" was added to support this feature,
letting a monitor process read the FW messages. The log is transferred
from the FW only when available, so the reading process might block.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Avoid recovery while one is already in progress

During recovery work commands sent to the FW could fail and schedule
additional recovery work. Since the chip is going to be powered off,
avoid recursive recoveries.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Check for FW quirks as soon as the FW boots

The FW initialization might depend on the FW revision, so check for any
FW quirks right after booting it.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: use freezable workqueue for netstack_work

When resuming (after wowlan), we want the rx packets (which is
usually the wake-up packet itself) to be passed to mac80211 only
after the resume notifier was completed, and mac80211 is up and
running (otherwise, the packets will be dropped).

By enqueueing the netstack_work to a freezable workqueue, we can
guarantee the rx processing to occur only after mac80211 was resumed.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: use _ni version of ieee80211_tx_status

wl1271_flush_deferred_work(), which calls ieee80211_rx() and
ieee80211_tx_status(), is called from a process context.
hence, use ieee80211_tx_status_ni() instead of ieee80211_tx_status().

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: check the vif's operstate after join

When resuming while connected (without wowlan), the interface
is already IF_OPER_UP, so we won't get the notifier callback,
and hence never complete the association (from wl12xx perspective)

This situation, among other potential problems, prevents the
station from entering psm.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx_sdio: enable wowlan only if enable_irq_wake() succeeded

Some platforms don't support the wake_irq, so disable wowlan
in this case, and avoid the "Unbalanced IRQ wake disable"
warning on disable_irq_wake().

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: enable/disable beacon filtering on ap suspend/resume

Beacon filtering needs to be enabled so AP won't wake up by
by every received beacon.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: clear wl->wow_enabled on resume

We set wl->wow_enabled on every suspend(), so we need to clear it
on every resume().
(we can't rely on setting wl->wow_enabled=false in suspend(),
as it being called only when wowlan triggers are configured)

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: don't check wow param on suspend/resume

Since mac80211 calls suspend/resume only when wowlan triggers
exist, there is no need to check for triggers existance in the
callbacks as well.
Add a WARN_ON() to verify it.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Add Support for Low Power DRPw (LPD) Mode

The Low Power DRPw (LPD) mode contains several optimizations
that designed to reduce power consumption. The purpose
is to save current consumption in RX and Listen mode.

LPD setting apply only for wl127x AP mode (not wl128x)

Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: don't bail if mmc isn't MMC_CAP_POWER_OFF_CARD

If our SDIO function has its runtime PM disabled, don't try to
manipulate its runtime PM status at all. This way we can still
power on cards plugged to mmc hosts that are not MMC_CAP_POWER_OFF_CARD.

Reported-and-tested-by: Tim Yamin <tim@kangatronix.co.uk>
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: split channel array per band in sched_scan

The firmware, in practice, treats the channels in three separate
blocks, one for each band (bg, a and j).  Instead of using a single
array and doing some magic with indices, split the array in 3 to make
it more readable.

Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: fix erroneous commit (cb5ae0)

Due to rebase error, the patch for commit cb5ae0 ("wl12xx: configure
rates when working in ibss mode") was wrong - a blob was added
into the wrong function. fix it.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: remove unused crc7 references

crc7 is used only in wl12xx_spi.

Remove redundant crc7.h includes, and update Kconfig to select CRC7
only if WL12XX_SPI is being selected.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: Enable beacon early termination in 2.4GHz band only

Beacon early termination doesn't help much in the 5GHz band and masks
channel switch IE Beacons.  Thus, change the code to use BET only in
2.4GHz.

[Reworded the commit log slightly -- Luca.]

Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: add rx_streaming debugfs entry

Allow control over rx_streaming interval and operation mode
(always/only on coex) via debugfs.

e.g.
echo 100 > /debug/ieee80211/phy0/wl12xx/rx_streaming/interval
echo 1 > /debug/ieee80211/phy0/wl12xx/rx_streaming/always

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: add automatic rx streaming triggers

When rx_streaming.interval is non-zero, use automatic rx streaming.
Enable rx streaming on the each rx/tx packet, and disable it
rx_streaming.duration msecs later.

When rx_streaming.always=0 (default), rx streaming is enabled only
when there is a coex operation.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx: add support for rx streaming

wl12xx supports the "rx streaming" feature:

When in ps mode, and @timeout msecs have been passed since
the last rx/tx, it issues trigger packets (QoS-null/PS-Poll packets,
according to the ac type) in const intervals (in order to reduce
the rx time).

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

net: wl12xx: remove unnecessary prints

Those have little value. Remove those to
make the driver less noisy.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

net: wl12xx: remove the nops

Nops aren't needed. When we actually need
those calls, then we add them with meat
and barbecue sauce.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

net: wl12xx: care for optional operations

->init and ->reset are optional - at least
sdio.c doesn't implement them - so allow those
pointers to be NULL.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

net: wl12xx: remove some unnecessary prints

Those have little value. Remove those to make
the driver less noisy.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

net: wl12xx: sdio: id_tables should be __devinitconst

That's only needed during init anyway, let's free
some space after we're done probing.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

ssb: add __devinit to some functions

Two functions in ssb are using register_pci_controller() which is
__devinit. The functions ssb_pcicore_init_hostmode() and
ssb_gige_probe() should also be __devinit.

This fixes the following warning:
WARNING: vmlinux.o(.text+0x2727b8): Section mismatch in reference from the function ssb_pcicore_init_hostmode() to the function .devinit.text:register_pci_controller()
The function ssb_pcicore_init_hostmode() references
the function __devinit register_pci_controller().
This is often because ssb_pcicore_init_hostmode lacks a __devinit
annotation or the annotation of register_pci_controller is wrong.

WARNING: vmlinux.o(.text+0x273398): Section mismatch in reference from the function ssb_gige_probe() to the function .devinit.text:register_pci_controller()
The function ssb_gige_probe() references
the function __devinit register_pci_controller().
This is often because ssb_gige_probe lacks a __devinit
annotation or the annotation of register_pci_controller is wrong.

Reported-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: enable support for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add external_reset callback to ath9k_platfom_data for AR9330

The patch adds a callback to ath9k_platform_data. If the
callback is provided by the platform code, then it can be
used to hard reset the WMAC device.

The callback is required for doing a hard reset of the AR9330
chips to get them working again after a hang.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: configure internal regulator for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: configure diversity based on eeprom contents for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fix few register offsets for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: set diversity group for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add fast_div_bias configuration for diversity group 1

The patch adds the missing configuration code for diversity
group 1.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove superfluous parentheses

Also remove two empty lines.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fix antenna diversity configuration for non AR9485 chips

The current values must be used only on AR9485 chips. Other chips
requires different values.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add a missing case statement to ath_ant_div_comb_alt_check

The 'ar9003_hw_antdiv_comb_conf_get' function sets div_group to 2
however that value is not used in 'ath_ant_div_comb_alt_check'.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: use correct max_streams for AR9330

The AR9330 is a single stream chip.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: enable antenna diversity for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: set NF limits for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: configure xpa bias level for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: use correct EEPROM base for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: enable tuning capacitance configuration for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: skip internal regulator configuration for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: enable TX/RX data byte swap for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: configure RF channel frequency for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: fix channel frequency calculation for AR9340

A multiplication is missing from the current formula.

Cc: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: read spur frequency information from eeprom for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: don't enable LDPC for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: use different BB watchdog timeout value for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add AR9330 specific PLL initialization

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: initialize rx gain table for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: initialize tx gain table for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: initialize mode registers for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add AR9330 initvals

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add platform device id for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: add MAC revision detection for AR9330

The AR9330 1.0 and 1.1 are using the same revision,
thus it is not possible to distinguish the two chips.
The platform setup code can distinguish the chips based
on the SoC revision.

Add a callback function to ath9k_platform_data in order
to allow getting the revision number from the platform code.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: define device id for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: define mac version for AR9330

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_hw: Fix calculation of PAPRD training power at 5Ghz

higher the chainmask, lesser the power_delta to be added
to the paprd_training_power

Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: use netlink consistent dump feature for BSS dumps

Use the new consistent dump feature from (generic) netlink
to advertise when dumps are incomplete.

Readers may note that this does not initialize the
rdev->bss_generation counter to a non-zero value. This is
still OK since the value is modified only under spinlock
when the list is modified. Since the dump code holds the
spinlock, the value will either be > 0 already, or the
list will still be empty in which case a consistent dump
will actually be made (and be empty).

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: remove casts of void pointers

In some cases local pointers are used to cast void pointers passed to
the function. Those unnecessary local pointers are also removed.

This patch was inspired by Joe Perches' patch
[PATCH net-next 1/2] wireless: Remove casts of void *;
and the comments from Julian Calaby.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Cc: Joe Perches <joe@perches.com>
Cc: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: detect stalled beacon interrupt faster

As USB_INT_ID_RETRY_FAILED can override USB_INT_ID_REGS, beacon interrupt
(CR_INTERRUPT) might be lost. Problem is that when device trigger CR_INTERRUPT
it disables HW interrupt. Now if USB_INT_ID_REGS with CR_INTERRUPT gets lost,
beacon interrupt stays disabled until beacon watchdog notices the stall. This
happen very often on heavy TX. Improve watchdog to trigger earlier, after three
missing beacon interrupts.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: don't let zd_mac_config_beacon() run too long from beacon interrupt handler

zd_mac_config_beacon() has only limited time to set up beacon when called from
beacon interrupt handler/worker. So do not let it retry acquiring beacon fifo
semaphore in interrupt handler. Beacon fifo semaphore should not be locked by
firmware anyway at this time, it's only locked when device is using/txing
beacon.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: handle lost read-reg interrupts

Device losses read-reg interrupts. By looking at usbmon it appears that
USB_INT_ID_RETRY_FAILED can override USB_INT_ID_REGS. This causes read
command to timeout, usually under heavy TX.

Fix by retrying read registers again if USB_INT_ID_RETRY_FAILED is received
while waiting for USB_INT_ID_REGS.

However USB_INT_ID_REGS is not always lost but is received after
USB_INT_ID_RETRY_FAILED and is usually received by the retried read
command. USB_INT_ID_REGS of the retry is then left unhandled and might
be received by next read command. Handle this by ignoring previous
USB_INT_ID_REGS that doesn't match current read command request.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: only update HW beacon if new beacon differs from currect

Update HW beacon only when needed. This appears to make device work in AP-mode
(dtim_period=1) somewhat more stable.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: make 'handle_rx_packet: invalid, small RX packet' message debug-only

This message is should be debug-only as it tells almost nothing useful. It also
happens very often and just floods logs.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: fix invalid signal values from device

Driver uses IEEE80211_HW_SIGNAL_UNSPEC and so signal values reported to
mac80211 should be in range 0..100. Sometimes device return out of range
values. These out of range values can then trigger warning in
cfg80211_inform_bss_frame.

This patch adds checks to enforce range returned from driver to
mac80211 be in 0..100 range.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

netlink: advertise incomplete dumps

Consider the following situation:
 * a dump that would show 8 entries, four in the first
   round, and four in the second
 * between the first and second rounds, 6 entries are
   removed
 * now the second round will not show any entry, and
   even if there is a sequence/generation counter the
   application will not know

To solve this problem, add a new flag NLM_F_DUMP_INTR
to the netlink header that indicates the dump wasn't
consistent, this flag can also be set on the MSG_DONE
message that terminates the dump, and as such above
situation can be detected.

To achieve this, add a sequence counter to the netlink
callback struct. Of course, netlink code still needs
to use this new functionality. The correct way to do
that is to always set cb->seq when a dumpit callback
is invoked and call nl_dump_check_consistent() for
each new message. The core code will also call this
function for the final MSG_DONE message.

To make it usable with generic netlink, a new function
genlmsg_nlhdr() is needed to obtain the netlink header
from the genetlink user header.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: calibrate radio after switching channel

After uploading radio values calibration goes in. In MMIO dump it is:
 radio_read(0x002b) -> 0x0008
radio_write(0x002b) <- 0x0008
 radio_read(0x002e) -> 0x0004
radio_write(0x002e) <- 0x0000
 radio_read(0x002e) -> 0x0000
radio_write(0x002e) <- 0x0004
 radio_read(0x002b) -> 0x0008
radio_write(0x002b) <- 0x0009
To find masks and sets, MMIO hacks were used to fool closed driver.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: reset usb devices at probe

When module is reloaded, device may fail to work, with messages:

[  342.211926] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x0059, type=2.
[  342.314254] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x004a, type=2.
[  342.416458] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 3183 max 2432.
[  342.518605] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x00c9, type=2.
[  342.620836] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x00ae, type=1.
[  342.723201] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 0 max 2432.
[  342.825399] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 0 max 2432.
[  342.927624] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 0 max 2432.
[  343.029804] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 2491 max 2432.
[  343.132008] phy40 -> rt2x00lib_rxdone: Warning - Wrong frame size 2576 max 2432.
[  343.234326] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x004c, type=1.
[  343.438723] phy40 -> rt2x00lib_rxdone_read_signal: Warning - Frame received with unrecognized signal, mode=0x0000, signal=0x00e6, type=1.

Whereas replugging device make it functional. To solve that problem
force reset device during probe.

With patch messages are gone. Unfortunately device may sometimes
still does not operate correctly after module reload (fail to receive
data after associate), but such cases are rarer than without the patch.

Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: fix possible memory corruption in case of invalid rxdesc.size

Sometimes rxdesc descriptor provided by hardware contains invalid
(random) data. For example rxdesc.size can be bigger than actual
size of the buffer. When this happen rt2x00crypto_rx_insert_iv()
corrupt memory doing memmove outside of buffer boundaries.

Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: N-PHY: move declarations of chantables functions to correct files

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: N-PHY: update switching analog on/off

Specs were recently updated.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: switch radio to requested channel

Switching channel happens after specific SHM write to B43_SHM_SH_CHAN.
This is the way we found it in BCM4331 MMIO dumps. By comparing with
N-PHY code we noticed there is routing used for SYN and TX/RX.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: add place for implementing channel switching

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: define standard structs for channel switching

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: enable radio

The trick was to find 0x810 PHY reg ops close to analog enabling code.
To find out proper masks and sets, MMIO hacks were used.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: HT-PHY: implement killing radio

Closed drivers kill radio right after reading radio version and MACCTL,
so it was easy to find related PHY ops:
 phy_read(0x0810) -> 0x0000
phy_write(0x0810) <- 0x0000
To find out the mask of above OP, MMIO hack was used to fake read val:
 phy_read(0x0810) -> 0xffff
phy_write(0x0810) <- 0x0000

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Merge branch 'master' of /linux/kernel/git/padovan/bluetooth-next-2.6

Bluetooth: Fix not setting the chan state

When the connection is ready we should set the connection
to CONNECTED so userspace can use it.

Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@openbossa.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>

Merge /linux/kernel/git/padovan/bluetooth-2.6

Conflicts:
net/bluetooth/l2cap_core.c

rtlwifi: rtl8192{c,ce,cu,se}: Remove comparisons of booleans with true

These are a potential source of confusion, as most C code treats all
non-zero values as true.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: rtl8192cu: Remove unused parameter

rtl_ps_set_rf_state's protect_or_not parameter is not set to
true anywhere, except for commented out code.

It enables some legacy locking code, which is no longer used,
so delete the parameter and the old locking code.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Fix typo in variable name

offchan_deley should be offchan_delay

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: clean exports of functions

Function managing IRQs is needed for external drivers like b43.
On the other side we do not expect writing any hosts drivers outside of
bcma, so this is safe to do not export functions related to this.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: add missing ieee80211_free_hw() on unload in AHB code

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: free sc->ah in ath5k_deinit_softc

Nothing else is freeing it, and this mirrors the error handling path
already in ath5k_init_softc.

https://bugzilla.kernel.org/show_bug.cgi?id=37592

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Reported-by: Tsyvarev Andrey <tsyvarev@ispras.ru>

bcma: fix BCM43224 rev 0 workaround

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>