rtlwifi: Fix logic in rx_interrupt

Should pass along packet if there's no CRC and no hardware error.

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

rtlwifi: Use write barrier when assigning ownership

Make sure all updates to a descriptor are flushed to memory
before assigning ownship to hardware.

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

rtlwifi: Assign rx buffer ownership to hardware last

Ownership of an rx buffer should only be given to the hardware
after all other changes are written, otherwise there's
a potential race.

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

rtlwifi: Don't block interrupts in spinlocks

Now power state transitions are not called from an
interrupt context, there's no need to block interrupts.

This code appears to block interrupts for too long,
causing my trackpad to lose sync occasionally.

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

rtlwifi: Run IPS leave work in a tasklet

This removes the need to use IRQ safe spinlocks in many places.

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

rtlwifi: Store loop index in local variable

_rtl_pci_rx_interrupt uses rtlpci->rx_ring[rx_queue_idx].idx a
few times, so store it in a separate variable.

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

rtlwifi: Remove set_rfpowerstate_inprogress

set_rfpowerstate_inprogress is only set and never read
so remove it.

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

rtlwifi: Synchronize IRQ after disabling it

This make sure any IRQ handlers running on other CPUs complete.

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

iwlagn: fix interface combinations

My patch to advertise interface combinations
worked by pure luck in the P2P case, but all
other cases are broken. This is due to a dumb
mistake in the code that checks what should
be advertised, fix that.

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

iwlagn: change log to better represent the state of aggregation process

Multiple A-MPDU actions will received from mac80211 while setting up the
aggregation queue, change the message log to better represent the states.

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: change the logging level for aggregation enable check

Aggregation will not enable if the traffic is lower than the threshold,
this is not an error condition, so change the logging level.

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: advertise interface combinations

With this, iwlwifi will advertise the limits on
concurrency of virtual interfaces.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: free the ICT ISR when the request_irq failed

Fix a memory leak in case request_irq fails.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: disambiguate invalid DMA index warnings

The exact same error message is used in three different functions in
iwlagn. Add the function name to the error string to disambiguate where
the error is coming from.

Signed-off-by: Daniel Halperin <dhalperi@cs.washington.edu>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: solve sparse warning

Remove local variable that was shadowing another one
sta_priv hasn't changed since the beginning of the function, so don't define
another pointer with the same name to the same variable

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: fix dual-mode RXON

 The PAN context has three states: disabled, disassociated
 and associated. It seems that wasn't quite as much of an
 issue in previous versions of the microcode, but now we
 really have to use all the three states properly. So add
 code to switch accordingly.

Additionally, PAN parameters need to be sent differently
and the timing for PAN RXON needs to be inbetween.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: remove un-necessary tx power ops

All agn devices use the same tx power function, remove the ops

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: send tx power command if defer cause by RXON not match

During channge channel, tx power will not send to uCode, the tx power command
should send after scan complete. but should also can send after RXON command.

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: skip disabled channels on channel survey

The channel survey information will be empy for
disabled channels so simply discard those entries.

Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: better position PAPRD debug message

the training power calculation indirectly depends on target power under
some scenarios, unless we have a valid training power, the PAPRD frames
won't be sent out. so when we get an invalid training power, its better
to display them before returning back.

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Add a debug entry to start/stop ANI

this helps the user to start/stop ANI dynamically.

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Further fix for mesh beaconing

This fix ensure the timers to be set at beacon interval boundaries.
Without this change timers can be set improperly resulting in the absence of beacons.

Signed-off-by: Fabrice Deyber <fabricedeyber@agilemesh.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: wake up chip before initializing PAPRD table

In PAPRD table initialization path we do some register read, so
make sure the chip is awake during that. Currently PAPRD is disabled.

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: rtl8192se: Synchronize IRQ after disabling it

As suggested by Mike McCormack <mikem@ring3k.org> for rtl8192ce, make sure any IRQ
handlers running on other CPUs complete.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: rtl8192se: Use write barrier when assigning ownership

As noted by Mike McCormack <mikem@ring3k.org> for rtl8192ce, make sure
all updates to a descriptor are flushed to memory before assigning ownship
to hardware.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: remove support for old chipsets W0/W1

For SD8787 A0/A1 chipsets we use the default firmware image
file 'mrvl/sd8787_uapsta.bin'. So the chip rev id variable
is removed.

The global variable fw_name is moved to adapter structure
so that we can support a different interface, such as PCIe,
in future.

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43legacy: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for b43legacy:

  CC [M]  drivers/net/wireless/b43legacy/main.o
drivers/net/wireless/b43legacy/main.c: In function ‘b43legacy_request_firmware’:
drivers/net/wireless/b43legacy/main.c:1567:6: warning: variable ‘tmshigh’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/main.c: In function ‘b43legacy_op_dev_config’:
drivers/net/wireless/b43legacy/main.c:2637:6: warning: variable ‘antenna_rx’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/main.c: In function ‘b43legacy_op_bss_info_changed’:
drivers/net/wireless/b43legacy/main.c:2778:24: warning: variable ‘phy’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/xmit.c: In function ‘generate_txhdr_fw3’:
drivers/net/wireless/b43legacy/xmit.c:324:7: warning: variable ‘rts_rate_ofdm’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/b43legacy/dma.o
drivers/net/wireless/b43legacy/dma.c: In function ‘free_all_descbuffers’:
drivers/net/wireless/b43legacy/dma.c:820:36: warning: variable ‘desc’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/dma.c: In function ‘b43legacy_dma_tx’:
drivers/net/wireless/b43legacy/dma.c:1377:28: warning: variable ‘info’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/dma.c:1374:24: warning: variable ‘hdr’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43legacy/dma.c: In function ‘b43legacy_dma_handle_txstatus’:
drivers/net/wireless/b43legacy/dma.c:1438:36: warning: variable ‘desc’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8192se: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for rtl8192se:

  CC [M]  drivers/net/wireless/rtlwifi/rtl8192se/dm.o
drivers/net/wireless/rtlwifi/rtl8192se/dm.c: In function ‘_rtl92s_dm_refresh_rateadaptive_mask’:
drivers/net/wireless/rtlwifi/rtl8192se/dm.c:225:5: warning: variable ‘rssi_level’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192se/fw.o
drivers/net/wireless/rtlwifi/rtl8192se/fw.c: In function ‘rtl92s_download_fw’:
drivers/net/wireless/rtlwifi/rtl8192se/fw.c:361:6: warning: variable ‘file_length’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192se/hw.o
drivers/net/wireless/rtlwifi/rtl8192se/hw.c: In function ‘_rtl92se_hw_configure’:
drivers/net/wireless/rtlwifi/rtl8192se/hw.c:887:6: warning: variable ‘reg_ratr’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/rtl8192se/hw.c: In function ‘_rtl92se_set_media_status’:
drivers/net/wireless/rtlwifi/rtl8192se/hw.c:1125:20: warning: variable ‘ledaction’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/rtl8192se/hw.c: In function ‘rtl92se_gpio_radio_on_off_checking’:
drivers/net/wireless/rtlwifi/rtl8192se/hw.c:2274:32: warning: variable ‘cur_rfstate’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192se/phy.o
drivers/net/wireless/rtlwifi/rtl8192se/phy.c: In function ‘rtl92s_phy_set_bw_mode’:
drivers/net/wireless/rtlwifi/rtl8192se/phy.c:266:5: warning: variable ‘reg_prsr_rsc’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192se/trx.o
drivers/net/wireless/rtlwifi/rtl8192se/trx.c: In function ‘_rtl92se_translate_rx_signal_stuff’:
drivers/net/wireless/rtlwifi/rtl8192se/trx.c:584:6: warning: variable ‘psaddr’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8192cu: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for rtl8192cu:

  CC [M]  drivers/net/wireless/rtlwifi/rtl8192cu/mac.o
drivers/net/wireless/rtlwifi/rtl8192cu/mac.c: In function ‘rtl92c_translate_rx_signal_stuff’:
drivers/net/wireless/rtlwifi/rtl8192cu/mac.c:1116:6: warning: variable ‘psaddr’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8192ce: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for rtl8192ce:

  CC [M]  drivers/net/wireless/rtlwifi/rtl8192ce/hw.o
drivers/net/wireless/rtlwifi/rtl8192ce/hw.c: In function ‘_rtl92ce_hw_configure’:
drivers/net/wireless/rtlwifi/rtl8192ce/hw.c:766:6: warning: variable ‘reg_ratr’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/rtl8192ce/hw.c: In function ‘rtl92ce_gpio_radio_on_off_checking’:
drivers/net/wireless/rtlwifi/rtl8192ce/hw.c:1972:41: warning: variable ‘cur_rfstate’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192ce/phy.o
drivers/net/wireless/rtlwifi/rtl8192ce/phy.c: In function ‘rtl92c_phy_config_rf_with_headerfile’:
drivers/net/wireless/rtlwifi/rtl8192ce/phy.c:284:7: warning: variable ‘rtstatus’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192ce/trx.o
drivers/net/wireless/rtlwifi/rtl8192ce/trx.c: In function ‘_rtl92ce_translate_rx_signal_stuff’:
drivers/net/wireless/rtlwifi/rtl8192ce/trx.c:595:6: warning: variable ‘psaddr’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8192c-common: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for rtl8192c-common:

  CC [M]  drivers/net/wireless/rtlwifi/rtl8192c/fw_common.o
drivers/net/wireless/rtlwifi/rtl8192c/fw_common.c: In function ‘_rtl92c_cmd_send_packet’:
drivers/net/wireless/rtlwifi/rtl8192c/fw_common.c:549:5: warning: variable ‘own’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/rtl8192c/phy_common.o
drivers/net/wireless/rtlwifi/rtl8192c/phy_common.c: In function ‘_rtl92c_phy_iq_calibrate’:
drivers/net/wireless/rtlwifi/rtl8192c/phy_common.c:1256:6: warning: variable ‘bbvalue’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/rtl8192c/phy_common.c: In function ‘rtl92c_phy_iq_calibrate’:
drivers/net/wireless/rtlwifi/rtl8192c/phy_common.c:1766:6: warning: variable ‘reg_ecc’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/rtl8192c/phy_common.c:1765:34: warning: variable ‘reg_eac’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for rtlwifi:

  CC [M]  drivers/net/wireless/rtlwifi/base.o
drivers/net/wireless/rtlwifi/base.c: In function ‘rtl_tx_agg_stop’:
drivers/net/wireless/rtlwifi/base.c:891:23: warning: variable ‘tid_data’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/base.c: In function ‘rtl_tx_agg_oper’:
drivers/net/wireless/rtlwifi/base.c:921:23: warning: variable ‘tid_data’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/efuse.o
drivers/net/wireless/rtlwifi/efuse.c: In function ‘efuse_pg_packet_write’:
drivers/net/wireless/rtlwifi/efuse.c:928:24: warning: variable ‘dataempty’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/efuse.c: In function ‘efuse_get_current_size’:
drivers/net/wireless/rtlwifi/efuse.c:1179:5: warning: variable ‘hoffset’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/ps.o
drivers/net/wireless/rtlwifi/ps.c: In function ‘rtl_ps_set_rf_state’:
drivers/net/wireless/rtlwifi/ps.c:85:19: warning: variable ‘rtstate’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/regd.o
drivers/net/wireless/rtlwifi/regd.c: In function ‘_rtl_dump_channel_map’:
drivers/net/wireless/rtlwifi/regd.c:310:28: warning: variable ‘ch’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/rtlwifi/usb.o
drivers/net/wireless/rtlwifi/usb.c: In function ‘_rtl_usb_transmit’:
drivers/net/wireless/rtlwifi/usb.c:826:21: warning: variable ‘urb_list’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/rtlwifi/usb.c:825:23: warning: variable ‘skb_list’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: Fix warnings from gcc 4.6.0

gcc 4.6.0 warnings for b43:

  CC [M]  drivers/net/wireless/b43/lo.o
drivers/net/wireless/b43/lo.c: In function ‘lo_measure_gain_values’:
drivers/net/wireless/b43/lo.c:304:7: warning: variable ‘trsw_rx’ set but not used [-Wunused-but-set-variable]
  CC [M]  drivers/net/wireless/b43/dma.o
drivers/net/wireless/b43/dma.c: In function ‘free_all_descbuffers’:
drivers/net/wireless/b43/dma.c:760:30: warning: variable ‘desc’ set but not used [-Wunused-but-set-variable]
drivers/net/wireless/b43/dma.c: In function ‘b43_dma_handle_txstatus’:
drivers/net/wireless/b43/dma.c:1391:30: warning: variable ‘desc’ set but not used [-Wunused-but-set-variable]

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

minstrel_ht: fixed rate mode through debugfs

Found several threads about fixed rate mode in minstrel_ht for test
environments, but no patches for it.

This patch provides such a mode through debugfs.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: add testmode dump support

This adds dump support to testmode. The testmode
dump support in nl80211 requires using two of the
six cb->args, the rest can be used by the driver
to figure out where the dump position is at or to
store other data across invocations.

Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Handle IBSS status changes in BSS_CHANGED_IBSS

Configure ibss node's bss_changes under BSS_CHANGED_IBSS.
And also start/stop ani timer only if the station join/leave
the group.

Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Remove ATH9K_BEACON_PERIOD mask

Earlier beacon_interval is used to hold interval value and
some flags (ATH9K_BEACON_ENA &ATH9K_BEACON_PERIOD). So to
extract interval ATH9K_BEACON_PERIOD is used. Those flags
were completely removed. So masking beacon_interval is
not required.

Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: host pci: implement block R/W operations

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

b43: bus: add helpers for checking host type

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

b43: use new pointers path to access ssb_device in SSB specific places

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

b43: SDIO: use new pointers path to access ssb_device

SDIO code is SSB specific, we can safely just use "sdev"

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

b43: LP-PHY: use new pointers path to access ssb_device

LP-PHY code is SSB specific, add check for bus type.

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

ath9k: mark few functions as static

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: make ath9k_setpower function as static

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: pci: implement interrupts control

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

ath9k_hw: remove ath9k_hw_get_desc_link

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: unify edma and non-edma tx code, improve tx fifo handling

EDMA based chips (AR9380+) have 8 Tx FIFO slots, which are used to fix the
tx queue start/stop race conditions which have to be worked around for
earlier chips by keeping the last descriptor in the queue. The current code
stores all frames that do not fit onto the 8 FIFO slots in a separate
list. Whenever a FIFO slot is freed up, the next frame (or A-MPDU) from the
pending queue gets moved to that slot.

This process is not only inefficient, but also unnecessary. The code can
be improved visibly by keeping the pending queue fully linked, and moving
the contents of the entire queue to a FIFO slot as it becomes available.

This patch makes the necessary changes for that and also merges some code
that was duplicated for EDMA vs non-EDMA. It changes txq->axq_link to point
to the last descriptor instead of the link pointer, so that
ath9k_hw_set_desc_link can be used, which works on all chips.

With this patch, a small performance increase for non-aggregated traffic
was observed on AR9380 based embedded hardware.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: remove redundant if check

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211_hwsim driver support userspace frame tx/rx

This patch adds to mac80211_hwsim the capability to send traffic via
userspace.

Frame exchange between kernel and user spaces is done through generic
netlink communication protocol. A new generic netlink family
MAC80211_HWSIM is proposed, this family contains three basic commands
HWSIM_CMD_REGISTER, which is the command used to register a new
traffic listener, HWSIM_CMD_FRAME, to exchange the frames from kernel
to user and vice-versa, and HWSIM_CMD_TX_INFO_FRAME which returns
from user all the information about retransmissions, rates, rx signal,
and so on.

How it works:

Once the driver is loaded the MAC80211_HWSIM family will be registered.
In the absence of userspace daemon, the driver itselfs implements a
perfect wireless medium as it did in the past. When a daemon sends a
HWSIM_CMD_REGISTER command, the module stores the application PID, and
from this moment all frames will be sent to the registered daemon.

The user space application will be in charge of process/forward all
frames broadcast by any mac80211_hwsim radio. If the user application
is stopped, the kernel module will detect the release of the socket
and it will switch back to in-kernel perfect channel simulation.

The userspace daemon must be waiting for incoming HWSIM_CMD_FRAME
commands sent from kernel, for each HWSIM_CMD_FRAME command the
application will try to broadcast this frame to all mac80211_hwsim
radios, however the application may decide to forward/drop this frame.
In the case of forwarding the frame, a new HWSIM_CMD_FRAME command will
be created, all necessary attributes will be populated and the frame
will be sent back to the kernel.

Also after the frame broadcast phase, a HWSIM_CMD_TX_INFO_FRAME
command will be sent from userspace to kernel, this command contains
all the information regarding the transmission, such as number of
tries, rates, ack signal, etc.

You can find the actual implementation of wireless mediumd daemon
(wmediumd) at:

* Last version tarball: https://github.com/jlopex/cozybit/tarball/master
* Or visiting my github tree: https://github.com/jlopex/cozybit/tree

Signed-off-by: Javier Lopez <jlopex@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Enabled rt35xx device support by default.

Now that support for these devices has been added we can enable them
by default and remove the Kconfig not on support for these devices to
be non-functional.

Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Move rt2800_txdone and rt2800_txdone_entry_check to rt2800usb.

These two functions are only used by rt2800usb so they don't have to be
in rt2800lib.

Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Interface sequence lock doesn't have to disable interrupts.

This lock is only used in the TX path and thus in process context. Therefore
we can use a much lighter spinlock variant.

Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Add support for RT3572/RT3592/RT3592+Bluetooth combo card

(based on an earlier patch submitted by Shiang)

Add support for RT3572/RT3592/RT3592+Bluetooth combo card

Signed-off-by: Shiang Tu <shiang_tu@ralinktech.com>
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Don't disable G0 PA_PE bit in case of BT coexistence.

(split off from the earlier RT35xx patch submitted by Shiang)

Signed-off-by: Shiang Tu <shiang_tu@ralinktech.com>
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Enable PA_PE bits in TX_PIN_CFG according to active band.

(split off from the earlier RT35xx patch submitted by Shiang)

There's no point in enabling the PA_PE bits for the bands that we are
not active on.

Signed-off-by: Shiang Tu <shiang_tu@ralinktech.com>
Signed-off-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bcma: add IRQ number and pointer to DMA dev

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

b43: bus: abstract bus and core operations

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

b43: bus: abstract board info

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

b43: bus: abstract chip info

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

b43: bus: abstract device structs and irq

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

b43: bus: abstract SPROM

SPROM is another frequently used struct. We decided to share SPROM
struct between ssb na bcma as long as we will not need any hacks.

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

b43: bus: abstract 80211 core info

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

b43: bus: abstract R/W operations

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

b43: add bus device abstraction layer

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

b43: rename ssb_device variable in ssb specific functions

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

Merge git://git./linux/kernel/git/linville/wireless-next-2.6 into for-davem

net: Kill ratelimit.h dependency in linux/net.h

Ingo Molnar noticed that we have this unnecessary ratelimit.h
dependency in linux/net.h, which hid compilation problems from
people doing builds only with CONFIG_NET enabled.

Move this stuff out to a seperate net/net_ratelimit.h file and
include that in the only two places where this thing is needed.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Ingo Molnar <mingo@elte.hu>

net: Add linux/sysctl.h includes where needed.

Several networking headers were depending upon the implicit
linux/sysctl.h include they get when including linux/net.h

Add explicit includes.

Signed-off-by: David S. Miller <davem@davemloft.net>

net: Kill ether_table[] declaration.

This got missed back in 2006 when Jes Sorensen deleted
net/ethernet/sysctl_net_ether.c

Signed-off-by: David S. Miller <davem@davemloft.net>

inetpeer: fix race in unused_list manipulations

Several crashes in cleanup_once() were reported in recent kernels.

Commit d6cc1d642de9 (inetpeer: various changes) added a race in
unlink_from_unused().

One way to avoid taking unused_peers.lock before doing the list_empty()
test is to catch 0->1 refcnt transitions, using full barrier atomic
operations variants (atomic_cmpxchg() and atomic_inc_return()) instead
of previous atomic_inc() and atomic_add_unless() variants.

We then call unlink_from_unused() only for the owner of the 0->1
transition.

Add a new atomic_add_unless_return() static helper

With help from Arun Sharma.

Refs: https://bugzilla.kernel.org/show_bug.cgi?id=32772

Reported-by: Arun Sharma <asharma@fb.com>
Reported-by: Maximilian Engelhardt <maxi@daemonizer.de>
Reported-by: Yann Dupont <Yann.Dupont@univ-nantes.fr>
Reported-by: Denys Fedoryshchenko <denys@visp.net.lb>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atm: expose ATM device index in sysfs

It's currently exposed only through /proc which, besides requiring
screen-scraping, doesn't allow userspace to distinguish between two
identical ATM adapters with different ATM indexes.  The ATM device index
is required when using PPPoATM on a system with multiple ATM adapters.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: David Woodhouse <dwmw2@infradead.org>
Cc: stable@kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'pablo/nf-2.6-updates' of git://1984.lsi.us.es/net-2.6

ath9k: Fix AR9287 calibration

The AR9287 calibration code was not being called because of an
incorrect MAC revision check.
This forced the AR9287 to use the AR9285 initial calibration code and
bypass the AR9287 code entirely.

Signed-off-by: Adrian Chadd <adrian@freebsd.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: Remove duplicate linux/slab.h include from net/mac80211/scan.c

Commit 79f460ca49d8d5700756ab7071c951311c7f29cc add a duplicate
linux/slab.h include to net/mac80211/scan.c - remove it.

Signed-off-by: Jesper Juhl <jj@chaosbits.net>
Acked-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: clear local->ps_data on disassoc

local->ps_data wasn't cleared on disassociation, which
(in some corner cases) caused reconnections to enter
psm before association completed.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: Default to 'n' for 2 new added devices in Kconfig.

We make oldconfig every time when a new kernel arrives, but
if we don't have such a device(I guess this is the most common
case for a new device), the default value should be 'n' so
that the kernel size we build doesn't grow up too much quickly.
For anyone who has the device, it is OK for them to turn it on
by themselves.

Cc: "John W. Linville" <linville@tuxdriver.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwl4965: fix 5GHz operation

rx_status.band is used uninitialized, what disallow to work on 5GHz .

Cc: stable@kernel.org # 2.6.39+
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

IPVS: bug in ip_vs_ftp, same list heaad used in all netns.

When ip_vs was adapted to netns the ftp application was not adapted
in a correct way.
However this is a fix to avoid kernel errors. In the long term another solution
might be chosen.  I.e the ports that the ftp appl, uses should be per netns.

Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

mac80211: stop queues before rate control updation

Stop tx queues before updating rate control to ensure
proper rate selection. Otherwise packets can be transmitted
in 40 Mhz whereas hw is configured in HT20.

Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: set 40 Mhz rate only if hw is configured in ht40

Whenever there is a channel width change from 40 Mhz to 20 Mhz,
the hardware is reconfigured to ht20. Meantime before doing
the rate control updation, the packets are being transmitted are
selected rate with IEEE80211_TX_RC_40_MHZ_WIDTH.

While transmitting ht40 rate packets in ht20 mode is causing
baseband panic with AR9003 based chips.

==== BB update: BB status=0x02001109 ====
ath: ** BB state: wd=1 det=1 rdar=0 rOFDM=1 rCCK=1 tOFDM=0 tCCK=0 agc=2
src=0 **
ath: ** BB WD cntl: cntl1=0xffff0085 cntl2=0x00000004 **
ath: ** BB mode: BB_gen_controls=0x000033c0 **
ath: ** BB busy times: rx_clear=99%, rx_frame=0%, tx_frame=0% **
ath: ==== BB update: done ====

Cc: stable@kernel.org
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k_hw: disable phy restart on baseband panic caused by RXSM

While receiving unsupported rate frame rx state machine
gets into a state 0xb and if phy_restart happens in that
state, BB would go hang. If RXSM is in 0xb state after
first bb panic, ensure to disable the phy_restart.

Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Reset chip on baseband hang

Resetting hardware helps to recover from baseband
hang/panic for AR9003 based chips.

Cc: stable@kernel.org
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: N-PHY: initialize last var in calibration function

Reported-by: Larry Finger <larry.finger@lwfinger.net>
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Use order 2 RX buffer allocation only if necessary

Although a previous fix handles the kernel panics that result from
failure to allocate a new RX buffer, memory fragmentation can be
reduced if the amsdu_8k capability is disabled as new buffers need only
be of O(0), not O(2).

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Fix kernel panic resulting from RX buffer allocation failure

To handle amsdu_8k capability, the PCI routine of this driver must
allocate receive buffers of order 2. Under heavy load, this causes
fragmentation of memory. The present code releases the current buffer
before checking to see if a new one is availble. Recovery from
allocation failures is not possible, which results in kernel panics.

The fix is to reorder the code to check that a new buffer can be
allocated before the old one is released. If not possible, the
received frame is dropped and the old one is reused. Without this
change, it is impossible to transfer a 2 GB file without a kernel panic.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org> [2.6.{37,38,39}]
Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: fix check for valid SSID size in scan operations

In both trigger_scan and sched_scan operations, we were checking for
the SSID length before assigning the value correctly.  Since the
memory was just kzalloc'ed, the check was always failing and SSID with
over 32 characters were allowed to go through.

This was causing a buffer overflow when copying the actual SSID to the
proper place.

This bug has been there since 2.6.29-rc4.

Cc: stable@kernel.org
Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwifiex: correct event header length

While decoding received event packet from firmware, 4 bytes
of interface header are already removed unconditionally.
So for handling event only 4 more bytes needs to be pulled.
This is achieved by changing event header length to 4.

Almost all the events, except BA stream related and AMSDU
aggregation control events, do not have the payload in their
event skb. Such events handling depends only on the event ID.
This event ID is the first four bytes of the event skb, which
is copied to a separate variable before pulling the skb header.
Hence event handling worked only for those events that didn't
have payload in event skb.

This patch fixes the broken event path of the events with
payload in their event skb without harming existing working
event path for the events without payload.

Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: Kiran Divekar <dkiran@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

bug.h: Move ratelimit warn interfaces to ratelimit.h

As reported by Ingo Molnar, we still have configuration combinations
where use of the WARN_RATELIMIT interfaces break the build because
dependencies don't get met.

Instead of going down the long road of trying to make it so that
ratelimit.h can get included by kernel.h or asm-generic/bug.h,
just move the interface into ratelimit.h and make users have
to include that.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>

bonding: cleanup module option descriptions

Weiping Pan noticed that the module option description for
xmit_hash_policy was incorrect and was nice enough to post a patch to
fix it.  The text was correct, but created a line over 80 characters and
I would rather not add those.  I realized I could take a few minutes and
clean up all the descriptions and things would look much better.  This
is the result.

Based on patch from Weiping Pan <panweiping3@gmail.com>.

Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
CC: Weiping Pan <panweiping3@gmail.com>
Reviewed-by: Weiping Pan <panweiping3@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net:8021q:vlan.c Fix pr_info to just give the vlan fullname and version.

The below patch removes vlan_buggyright and vlan_copyright from vlan_proto_init,
so that it prints out just the fullname of vlan and the version number.

before:

[   30.438203] 802.1Q VLAN Support v1.8 Ben Greear <greearb@candelatech.com>
[   30.441542] All bugs added by David S. Miller <davem@redhat.com>

after:

[   31.513910] 802.1Q VLAN Support v1.8

Signed-off-by: Justin P. Mattock <justinmattock@gmail.com>
CC: Joe Perches <joe@perches.com>
CC: David S. Miller <davem@davemloft.net>
CC: Ben Greear <greearb@candelatech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: davinci_emac: fix dev_err use at probe

Use platform device rather than net device in dev_err calls before net
device has been registered to avoid messages such as

(null): DaVinci EMAC: Failed to get EMAC clock

Also replace remaining printks in probe with dev_{err,warn}.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: convert to %pK for kptr_restrict support

As these pointers have been printed without using %p they were missed in the
big network kptr_restrict conversion patch %p -> %pK from Dan Rosenberg.

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix ETHTOOL_SFEATURES compatibility with old ethtool_ops.set_flags

Current code squashes flags to bool - this makes set_flags fail whenever
some ETH_FLAG_* equivalent features are set. Fix this.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: Fix several warnings in compat_mtw_from_user().

Kill set but not used 'entry_offset'.

Add a default case to the switch statement so the compiler
can see that we always initialize off and size_kern before
using them.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: ipset: fix ip_set_flush return code

ip_set_flush returned -EPROTO instead of -IPSET_ERR_PROTOCOL, fixed

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: ipset: remove unused variable from type_pf_tdel()

Variable 'ret' is set in type_pf_tdel() but not used, remove.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

netfilter: ipset: Use proper timeout value to jiffies conversion

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

bonding: documentation and code cleanup for resend_igmp

Improves the documentation about how IGMP resend parameter
works, fix two missing checks and coding style issues.

Signed-off-by: Flavio Leitner <fbl@redhat.com>
Acked-by: Rick Jones <rick.jones2@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: prevent deadlock on slave store with alb mode (v3)

This soft lockup was recently reported:

[root@dell-per715-01 ~]# echo +bond5 > /sys/class/net/bonding_masters
[root@dell-per715-01 ~]# echo +eth1 > /sys/class/net/bond5/bonding/slaves
bonding: bond5: doing slave updates when interface is down.
bonding bond5: master_dev is not up in bond_enslave
[root@dell-per715-01 ~]# echo -eth1 > /sys/class/net/bond5/bonding/slaves
bonding: bond5: doing slave updates when interface is down.

BUG: soft lockup - CPU#12 stuck for 60s! [bash:6444]
CPU 12:
Modules linked in: bonding autofs4 hidp rfcomm l2cap bluetooth lockd sunrpc
be2d
Pid: 6444, comm: bash Not tainted 2.6.18-262.el5 #1
RIP: 0010:[<ffffffff80064bf0>]  [<ffffffff80064bf0>]
.text.lock.spinlock+0x26/00
RSP: 0018:ffff810113167da8  EFLAGS: 00000286
RAX: ffff810113167fd8 RBX: ffff810123a47800 RCX: 0000000000ff1025
RDX: 0000000000000000 RSI: ffff810123a47800 RDI: ffff81021b57f6f8
RBP: ffff81021b57f500 R08: 0000000000000000 R09: 000000000000000c
R10: 00000000ffffffff R11: ffff81011d41c000 R12: ffff81021b57f000
R13: 0000000000000000 R14: 0000000000000282 R15: 0000000000000282
FS:  00002b3b41ef3f50(0000) GS:ffff810123b27940(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00002b3b456dd000 CR3: 000000031fc60000 CR4: 00000000000006e0

Call Trace:
 [<ffffffff80064af9>] _spin_lock_bh+0x9/0x14
 [<ffffffff886937d7>] :bonding:tlb_clear_slave+0x22/0xa1
 [<ffffffff8869423c>] :bonding:bond_alb_deinit_slave+0xba/0xf0
 [<ffffffff8868dda6>] :bonding:bond_release+0x1b4/0x450
 [<ffffffff8006457b>] __down_write_nested+0x12/0x92
 [<ffffffff88696ae4>] :bonding:bonding_store_slaves+0x25c/0x2f7
 [<ffffffff801106f7>] sysfs_write_file+0xb9/0xe8
 [<ffffffff80016b87>] vfs_write+0xce/0x174
 [<ffffffff80017450>] sys_write+0x45/0x6e
 [<ffffffff8005d28d>] tracesys+0xd5/0xe0

It occurs because we are able to change the slave configuarion of a bond while
the bond interface is down.  The bonding driver initializes some data structures
only after its ndo_open routine is called.  Among them is the initalization of
the alb tx and rx hash locks.  So if we add or remove a slave without first
opening the bond master device, we run the risk of trying to lock/unlock a
spinlock that has garbage for data in it, which results in our above softlock.

Note that sometimes this works, because in many cases an unlocked spinlock has
the raw_lock parameter initialized to zero (meaning that the kzalloc of the
net_device private data is equivalent to calling spin_lock_init), but thats not
true in all cases, and we aren't guaranteed that condition, so we need to pass
the relevant spinlocks through the spin_lock_init function.

Fix it by moving the spin_lock_init calls for the tx and rx hashtable locks to
the ndo_init path, so they are ready for use by the bond_store_slaves path.

Change notes:
v2) Based on conversation with Jay and Nicolas it seems that the ability to
enslave devices while the bond master is down should be safe to do.  As such
this is an outlier bug, and so instead we'll just initalize the errant spinlocks
in the init path rather than the open path, solving the problem.  We'll also
remove the warnings about the bond being down during enslave operations, since
it should be safe

v3) Fix spelling error

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: jtluka@redhat.com
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: nicolas.2p.debian@gmail.com
CC: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hold rtnl again in dump callbacks

Commit e67f88dd12f6 (dont hold rtnl mutex during netlink dump callbacks)
missed fact that rtnl_fill_ifinfo() must be called with rtnl held.

Because of possible deadlocks between two mutexes (cb_mutex and rtnl),
its not easy to solve this problem, so revert this part of the patch.

It also forgot one rcu_read_unlock() in FIB dump_rules()

Add one ASSERT_RTNL() in rtnl_fill_ifinfo() to remind us the rule.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Patrick McHardy <kaber@trash.net>
CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>