drivers/net: Use DIV_ROUND_CLOSEST

The kernel.h macro DIV_ROUND_CLOSEST performs the computation (x + d/2)/d
but is perhaps more readable.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@haskernel@
@@

#include <linux/kernel.h>

@depends on haskernel@
expression x,__divisor;
@@

- (((x) + ((__divisor) / 2)) / (__divisor))
+ DIV_ROUND_CLOSEST(x,__divisor)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb3: fix Gen2 pci default settings

Modify control register settings to accommodate the bridge's max read
requset size.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb3: fix EDC filename string

Fix a cut'n paste error in the AEL2020 twinax EDC file name

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxge: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for vxge.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s2io: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for s2io.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlge: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for qlge.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgb: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for ixgb.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igbvf: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for igbvf.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for bnx2x.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for bnx2.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atl1e: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for atl1e.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atl1c: return PCI_ERS_RESULT_DISCONNECT on permanent failure

PCI drivers that implement the struct pci_error_handlers' error_detected
callback should return PCI_ERS_RESULT_DISCONNECT if the state passed in is
pci_channel_io_perm_failure. This patch fixes the issue for atl1c.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: delete unnecessary code from be_cmd_POST()

When the driver loads the POST stage of the card is expected to be
POST_STAGE_ARMFW_RDY.

Signed-off-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: some name changes for consistency

be_mbox_db_ring() has been changed to be_mbox_notify() (to be consistent with
be_mcc_notify()) and struct be_mcc_cq_entry changed to be_mcc_compl
to be consistent with rx/tx_compl.

Signed-off-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: replace some printks with dev_err()/dev_warn()

And get rid of some unnecessary printks.

Signed-off-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: remove pci_func field from be_adapter struct

pci function is available as PCI_FUNC(pdev->devfn); no need for a
separate field.

Signed-off-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: get rid of be_ctrl_info struct/abstraction

Currently only the be_ctrl_info struct ptr (instead of adapter) is
 passed to all the routines in be_cmds.c. Instead pass be_adapter
ptr to allow access to all its fields. Merge the contents of struct be_ctrl_info
into be_adapter. The resulting code is simpler.

Signed-off-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next-2.6

Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
drivers/net/wireless/iwlwifi/iwl-3945.h
drivers/net/wireless/iwlwifi/iwl-tx.c
drivers/net/wireless/iwlwifi/iwl3945-base.c

ixgbe: Fix RSC completion delay causing Rx interrupts to stop

When a user disables interrupt throttling with ethtool on 82599 devices,
the interrupt timer may not be re-enabled if hardware RSC is running.  The
RSC completions in hardware don't complete before the next ITR event tries
to fire, so the ITR timer never gets re-armed.  This patch increases the
amount of time between interrupts when throttling is disabled (rx-usecs =
0) when the hardware RSC deature is enabled.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Fix usage of second flags bitmap when using LRO/RSC

A second set of feature flag bits was added, and the hardware RSC engine
flags were moved there.  However, the code itself didn't make the move
completely to use the new bitmap.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Acked-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Fix netpoll to be properly multiqueue aware

Our ndo_poll_controller callback is broken for anything but non-multiqueue
setups.  This fixes that issue.

Signed-off-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm: select sane defaults for xfrm[4|6] gc_thresh

Choose saner defaults for xfrm[4|6] gc_thresh values on init

Currently, the xfrm[4|6] code has hard-coded initial gc_thresh values
(set to 1024).  Given that the ipv4 and ipv6 routing caches are sized
dynamically at boot time, the static selections can be non-sensical.
This patch dynamically selects an appropriate gc threshold based on
the corresponding main routing table size, using the assumption that
we should in the worst case be able to handle as many connections as
the routing table can.

For ipv4, the maximum route cache size is 16 * the number of hash
buckets in the route cache.  Given that xfrm4 starts garbage
collection at the gc_thresh and prevents new allocations at 2 *
gc_thresh, we set gc_thresh to half the maximum route cache size.

For ipv6, its a bit trickier.  there is no maximum route cache size,
but the ipv6 dst_ops gc_thresh is statically set to 1024.  It seems
sane to select a simmilar gc_thresh for the xfrm6 code that is half
the number of hash buckets in the v6 route cache times 16 (like the v4
code does).

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPVS: use pr_fmt

While being at it cleanup whitespace.

Signed-off-by: Hannes Eder <heder@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: fix coalescing setup in ethtool support

Parameter order for using mk_ic_value(count, time) was reversed,
the patch fixes this.

Signed-off-by: Jiajun Wu <b06378@freescale.com>
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pppoe: fix /proc/net/pppoe

If a socket is hashed in last slot of pppoe hash table (PPPOE_HASH_SIZE-1)
we report it many times (up to filling seq buffer)
(Only the last socket of last slot)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

eexpress: Read buffer overflow

start_code is 69 words, but the code always writes a multiple of 16 words,
so the last 11 words written are outside the array.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: ARP neigh procfs buffer overflow

If arp_format_neigh_entry() can be called with n->dev->addr_len == 0, then a
write to hbuffer[-1] occurs.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3c515: Write outside array bounds

if dev_alloc_skb() fails on the first iteration, a write to
cp->rx_ring[-1] occurs.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vlan: allow creating vlan when real device is not up

There is no reason for the arbitrary restriction that device must be
up to create a vlan. This patch was added to Vyatta kernel to resolve startup
ordering issues where vlan's are created but real device was disabled.

Note: the vlan already correctly inherits the operstate from real device; so
if vlan is created and real device is marked down, the vlan is marked
down.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

docbook: fix printk of ip address

Use the %pI4 format string instead of %d.%d.%d.%d and NIPQUAD.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Version 1.48.114-1

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: WoL only with current MAC address

In some cases with 57711E, depending on the functions unload sequence, other
functions MAC address could have been used to wake the system as well. Make sure
to block all but the current function if WoL is required by changing the mode
to single function WoL.

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Self-test kills management traffic

Self test used to play with the management FIFO possibly while management was
running...

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Receive traffic that maches management filtering rules

Due to lack of configuration, if the BMC configures the chip to pass all
broadcast/multicast traffic to it, the host will not receive it. On top of
fixing it, also make sure that in promiscuous mode, the host will receive the
management traffic as well.

Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iscsi: Use GFP_ATOMIC in iscsi_offload_mesg().

Changing to GFP_ATOMIC because the only caller in cnic/bnx2i may
be calling this function while holding spin_lock.

This problem was discovered by Mike Christie.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Acked-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/netlabel: Add kmalloc NULL tests

The test on map4 should be a test on map6.

The semantic match that finds this problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
expression *x;
identifier f;
constant char *C;
@@

x = \(kmalloc\|kcalloc\|kzalloc\)(...);
... when != x == NULL
    when != x != NULL
    when != (x || ...)
(
kfree(x)
|
f(...,C,...,x,...)
|
*f(...,x,...)
|
*x->f
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Acked-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

netxen: fix CONFIG_INET=n build

Wrap dest IP hashing code with #ifdef CONFIG_INET,
this feature makes no sense without INET, but other
driver can still work.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iwlwifi: debugFs to enable/disable HT40 support

Add debugfs file to enable/disable HT40(40MHz) channel support.
By default, 40MHz is supported if AP can support the function.

By echo "1" to "disable_ht40" file, iwlwifi driver will disable the
40MHz support and only allow 20MHz channel.

Because the information exchange happen during association time,
so enable/disable ht40 channel only can be performed when it is not
associated with AP.

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

ath9k: remove unused ATH_PCI_VERSION

Our version goes by the kernel release.

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

cfg80211: self-contained wext handling where possible

Finally! This is what you've all been waiting for!

This patch makes cfg80211 take care of wext emulation
_completely_ by itself, drivers that don't need things
cfg80211 doesn't do yet don't even need to be aware of
wireless extensions.
This means we can also clean up mac80211's and iwm's
Kconfig and make it possible to build them w/o wext
now!

RIP wext.

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

cfg80211: combine IWESSID handlers

Since we now have handlers IWESSID for all modes, we can
combine them into one.

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

cfg80211: combine IWAP handlers

Since we now have IWAP handlers for all modes, we can
combine them into one.

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

cfg80211: combine iwfreq implementations

Until now we implemented iwfreq for managed mode, we
needed to keep the implementations separate, but now
that we have all versions implemented we can combine
them and export just one handler.

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

mac80211: verify info->control.vif is not NULL

When enqueuing packets on the internal packet queue, we
need to ensure that we have a valid vif pointer since
that is required since the net namespace work. Add some
assertions to verify this, but also don't crash is for
some reason we don't end up with a vif pointer -- warn
and drop the packet in all these cases.

Since this code touches a number of hotpaths, it is
intended to be temporary, or maybe configurable in the
future, at least the bit that is in the path that gets
hit for every packet, ieee80211_tx_pending().

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

cfg80211: fix disassoc while not associated

When trying to disassociate while not associated,
the kernel would crash rather than refusing the
operation, fix this;

Reported-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Add debug counters for TX

Location: ath9k/phy#/xmit

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

mac80211: fix an oops in ieee80211_scan_state_set_channel

Fix an oops in ieee80211_scan_state_set_channel which was triggered
if the last scanned channel was skipped (for example due to regulatory
restrictions) by returning to the decision state after each skipped
channel.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

nl80211: add missing parameter clearing

Jouni and Maxim reported an oops when using wpa_supplicant -Dnl80211,
which seems to be due to random data being contained in the crypto
settings for the assoc() command. This seems to be due to the missing
memset here, so add it -- it's certainly missing but I'm not 100%
certain that it will fix the problem.

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

ath9k: get rid of unnecessary setpower calls

We are using setpower routines regardless of the current power mode.
Don't bother the hardware, if it is not necessary.

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

ath9k: fix race with IEEE80211_CONF_PS checks

There is a small window where the mac80211 changes the IEEE80211_CONF_PS
flag, and then informs the driver about the change. We have a race
condition if we are checking the flag in the same time. Avoid it by
introducing a local variable, and using that instead of checking the
IEEE80211_CONF_PS flag directly.

This fix the problem reported by Luis:
http://article.gmane.org/gmane.linux.kernel.wireless.general/34363

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

ar9170: fix build error when !CONFIG_AR9170_LEDS

Fix this build error when CONFIG_AR9170_LEDS is not set
drivers/net/wireless/ath/ar9170/main.c:1296: error: 'struct ar9170' has
no member named 'led_work'

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ray_cs: remove bogus NULL check at head of ray_get_wireless_stats

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Martin Ettl <ettl.martin@gmx.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix receiving deauth

Marcel reported a warning, which quite obviously comes
from an oversight in the code handling deauth frames,
and which resulted in multiple follow-up warnings due
to this missing handling. This patch adds the missing
deauth handling (telling cfg80211 about it) and also
removes the follow-up warnings since they could happen
due to races even if nothing is wrong. I've explained
the races in the comments.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Reported-by: Marcel Holtmann <marcel@holtmann.org>
Tested-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: fix circular lock dependency (1)

Luis reported this lockdep complaint, that he had also
reported earlier but when trying to analyse I had been
locking at the wrong code, and never saw the problem:

(slightly abridged)
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.31-rc4-wl #6
-------------------------------------------------------
wpa_supplicant/3799 is trying to acquire lock:
 (cfg80211_mutex){+.+.+.}, at: [<ffffffffa009246a>] cfg80211_get_dev_from_ifindex+0x1a/0x90 [cfg80211]

but task is already holding lock:
 (rtnl_mutex){+.+.+.}, at: [<ffffffff81400ff2>] rtnl_lock+0x12/0x20

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (rtnl_mutex){+.+.+.}:
       [<ffffffff810857b6>] __lock_acquire+0xd76/0x12b0
       [<ffffffff81085dd3>] lock_acquire+0xe3/0x120
       [<ffffffff814ee7a4>] mutex_lock_nested+0x44/0x350
       [<ffffffff81400ff2>] rtnl_lock+0x12/0x20
       [<ffffffffa009f6a5>] nl80211_send_reg_change_event+0x1f5/0x2a0 [cfg80211]
       [<ffffffffa009529e>] set_regdom+0x28e/0x4c0 [cfg80211]

-> #0 (cfg80211_mutex){+.+.+.}:
       [<ffffffff8108587b>] __lock_acquire+0xe3b/0x12b0
       [<ffffffff81085dd3>] lock_acquire+0xe3/0x120
       [<ffffffff814ee7a4>] mutex_lock_nested+0x44/0x350
       [<ffffffffa009246a>] cfg80211_get_dev_from_ifindex+0x1a/0x90 [cfg80211]
       [<ffffffffa009813f>] get_rdev_dev_by_info_ifindex+0x6f/0xa0 [cfg80211]
       [<ffffffffa009b12b>] nl80211_set_interface+0x3b/0x260 [cfg80211]

When looking at the correct code, the problem is quite
obvious. I'm not entirely sure which code paths lead
here, so until I can analyse it better let's just use
RCU to avoid the problem.

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

iwlagn: fix sparse warning when compiling without debug

C [M]  drivers/net/wireless/iwlwifi/iwl-core.o
drivers/net/wireless/iwlwifi/iwl-core.c:1341: warning:
‘iwl_dump_nic_error_log’ defined but not used

Reported-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlagn: fix null pointer access during ucode load on 1000

Commit "iwlwifi: Handle new firmware file with ucode build number
in header" introduced new ucode header parsing routines, but
neglected to initialize these routines for 1000. The system thus goes
into infinite loop trying to load ucode, failing every time with a null
pointer exception as it tries to parse the header.

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

mac80211: fix suspend

Jan reported that his b43-based laptop hangs during suspend.
The problem turned out to be mac80211 asking the driver to
stop the hardware before removing interfaces, and interface
removal caused b43 to touch the hardware (while down, which
causes the hang).

This patch fixes mac80211 to do reorder these operations to
have them in the correct order -- first remove interfaces
and then stop the hardware. Some more code is necessary to
be able to do so in a race-free manner, in particular it is
necessary to not process frames received during quiescing.

Fixes http://bugzilla.kernel.org/show_bug.cgi?id=13337.

Reported-by: Jan Scholz <scholz@fias.uni-frankfurt.de>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: Read buffer overflow

Several arrays were read before checking whether the index was within
bounds. ARRAY_SIZE() should be used to determine the size of arrays.

rates->rates has an arraysize of 1, so calling get_common_rates()
with a rates_size of MAX_RATES (14) was causing reads out of bounds.

tmp_size can increment at most to (ARRAY_SIZE(lbs_bg_rates) - 1) *
(*rates_size - 1), so that should be the number of elements of tmp[].

A goto can be eliminated: ret was already set upon its declaration.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: Read outside array bounds

reads bss->rates[j] before checking bounds of index, and should use
ARRAY_SIZE to determine the size of the array.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Holger Schurig <hs4233@mail.mn-solutions.de>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ip: fix logic of reverse path filter sysctl

Even though reverse path filter was changed from simple boolean to
trinary control, the loose mode only works if both all and device are
configured because of this logic error.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iwlwifi: fix LED config option

IWLWIFI_LEDS option should certainly have help comment, and should
default to y.

Signed-off-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: Name fix for MPDU density for TX aggregation

Fix incorrect name for HT MPDU Density.
default set to 4 uSec

Reported-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: print packet contents in error case

This data is more useful to debugging that the receive
buffer contents.

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: remove command callback return value

No existing callbacks use anything other than the return
value 1, which means that the caller should free the
reply skb, so it seems safer in terms of not introducing
memory leaks to simply remove the return value and let
the caller always free the skb.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: fix up command sending

The current command sending in iwlwifi is a bit of a mess:
 1) there is a struct, iwl_cmd, that contains both driver
    and device data in a single packed structure -- this
    is very confusing
 2) the on-stack data and the command metadata share a
    structure by embedding the latter in the former, which
    is also rather confusing because it leads to weird
    unions and similarly odd constructs
 3) each txq always has enough space for 256 commands,
    even if only 32 end up being used

This patch fixes these things:
 1) rename iwl_cmd to iwl_device_cmd and keep track of
    command metadata and device command separately, in
    two arrays in each tx queue
 2) remove the 'meta' member from iwl_host_cmd and only
    put in the required members
 3) allocate the cmd/meta arrays separately instead of
    embedding them into the txq structure

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: Thermal Throttling debugfs function

Add debugfs function to display current thermal throttling status for
both Legacy and Advance Thermal Throttling Management

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

iwlwifi: Thermal Throttling Management - part 2

Part 2 of Thermal Throttling Management -

Thermal Throttling feature is used to put NIC into low power state when
driver detect the Radio temperature reach pre-defined threshold

Two Thermal Throttling Management Methods; this patch introduce the
Advance Thermal Throttling:
TI-0: system power index, no tx/rx restriction, HT enabled
TI-1: power index 5, 1 spatial stream Tx, multiple spatial stream Rx, HT
enabled
TI-2: power index 5: 1 spatial stream Tx, 1 spatial stream Rx, HT
disabled
TI-CT-KILL: power index 5, no Tx, no Rx, HT disabled

For advance Thermal Throttling, CT_KILL_ENTER threshold and CT_KILL_EXIT
threshold are different; uCode will not stay awake until reach
CT_KILL_EXIT threshold.

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

iwlwifi: Thermal Throttling Management - Part 1

Part 1 of Thermal Throttling Management -

Thermal Throttling feature is used to put NIC into low power state when
driver detect the Radio temperature reach pre-defined threshold

Two Thermal Throttling Management Methods; this patch introduce the
Legacy Thermal Management:
   IWL_TI_0: normal temperature, system power state
   IWL_TI_1: high temperature detect, low power state
   IWL_TI_2: higher temperature detected, lower power state
   IWL_TI_CT_KILL: critical temperature detected, lowest power state

Once get into CT_KILL state, uCode go into sleep, driver will stop all
the active queues, then move to IWL_TI_CT_KILL state; also set up 5
seconds timer to toggle CSR flag, uCode wake up upon CSR flag change,
then measure the temperature.
If temperature is above CT_KILL exit threshold, uCode go backto sleep;
if temperature is below CT_KILL exit threshold, uCode send Card State
Notification response with appropriate CT_KILL status flag, and uCode
remain awake, Driver receive Card State Notification Response and update
the card temperature to the CT_KILL exit threshold.

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

iwlwifi: critical temperature enter/exit condition

If advance thermal throttling is used the driver need to pass both
"enter" and "exit" temperature to uCode.

Using different critical temperature threshold for legacy and advance
thermal throttling management based on the type of thermal throttling
method is used except 1000.
For 1000, it use advance thermal throttling critical temperature
threshold, but with legacy thermal management implementation until ucode
has the necessary implementations in place.

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

iwlwifi: revert to active table when rate is not valid

When performing rate scaling, if detected that the new rate
index is invalid, clear the search_better_tbl flag
so it will not be stuck in the loop.

Since the search table is already set up in uCode,
we need to empty out the the search table;
revert back to the "active" rate and throughput info.
Also pass the "active" table setup to uCode to make
sure the rate scale is functioning correctly.

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

mac80211: fix oops due to missing private data

This was caused by patch:

"mac80211: cooperate more with network namespaces"

The version of the patch applied doesn't match Johannes' latest:

http://johannes.sipsolutions.net/patches/kernel/all/LATEST/NNN-mac80211-netns.patch

The skb->cb virtual interface data wasn't being reset for
reuse so ath9k pooped out when trying to dereference the
private rate control info from the skb.

BUG: unable to handle kernel NULL pointer dereference
RIP: 0010:[<ffffffffa0258173>] ath_tx_rc_status+0x33/0x150 [ath9k]
<-- snip etc -->

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

ath9k: Add support for AR9287 based chipsets.

Signed-off-by: Vivek Natarajan <vnatarajan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix PS-poll response, race

When a station queries us for a PS-poll response, we wrongly
queue the frame on the virtual interface's queue rather than
the pending queue.

Additionally, fix a race condition where we could potentially
send multiple frames to the sleeping station due to using a
station flag rather than a packet flag. When converting to a
packet flag, we can also convert p54 and remove the filter
clearing we added for it.

(Also remove a now dead function)

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Reported-by: Bob Copeland <me@bobcopeland.com>
Tested-by: Bob Copeland <me@bobcopeland.com>
Cc: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix MLME issuing of probe requests while scanning

We were issuing probe requests to the associated AP on the wrong
band by having our beacon timer loss trigger while we are scanning.
When we would scan the timer could hit and force us to send a
probe request to the AP but with a chance we'd be on the wrong band.

This leads to finding no usable bitrate but we should not get so
far on the xmit path. We should not be trying to send these probe
request frames so prevent ieee80211_mgd_probe_ap() from sending
these.

As it turns out all callers of ieee80211_mgd_probe_ap() need this
check so we just move the scan check there. This means we can remove
the recenlty added check during ieee80211_sta_monitor_work().

Additionally we now fix a race condition added by the patch
"mac80211: do not monitor the connection while scanning" which
had the same check in ieee80211_sta_conn_mon_timer(). The race
happens because the timer routine *does* a valid check for
scanning but after it queues work into the mac80211 workqueue
the work callback can kick off with scanning enabled and cause
the same issue we were trying to avoid.

The more appropriate solution would be to disable the respective
timers during scan and re-enable them after scan but requires more
complex code and testing.

Cc: Christian Lamparter <chunkeey@web.de>
Cc: Larry Finger <Larry.Finger@lwfinger.net>
Reported-by: Fabio Rossi <rossi.f@inwind.it>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix mlme timeouts

When a new MLME work is created, its timeout is initialised
to 0. This is wrong, it could then be thought of as having
an actual timeout in the future (time_is_after_jiffies() can
return true). Instead, it should be initialised to jiffies
so that it will run right away as soon as the mlme work is
executed.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Reported-by: Luciano Roth Coelho <luciano.coelho@nokia.com>
Reported-by: Alban Browaeys <prahal@yahoo.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: increase scan result expire time

Using background scanning in mac80211 the time a scan needs to
finish can exceed 10 seconds. Hence, increase the scan results
expire time to 15 seconds which should be sufficient.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: rename scan_state to next_scan_state

Rename scan_state to next_scan_state to better reflect
what it is used for.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: implement basic background scanning

Introduce a new scan flag "SCAN_OFF_CHANNEL" which basically tells us
that we are currently on a different channel for scanning and cannot
RX/TX. "SCAN_SW_SCANNING" tells us that we are currently running a
software scan but we might as well be on the operating channel to RX/TX.
While "SCAN_SW_SCANNING" is set during the whole scan "SCAN_OFF_CHANNEL"
is set when leaving the operating channel and unset when coming back.

Introduce two new scan states "SCAN_LEAVE_OPER_CHANNEL" and
"SCAN_ENTER_OPER_CHANNEL" which basically implement the functionality we
need to leave the operating channel (send a nullfunc to the AP and stop
the queues) and enter it again (send a nullfunc to the AP and start the
queues again).

Enhance the scan state "SCAN_DECISION" to switch back to the operating
channel after each scanned channel. In the future it sould be simple
to enhance the decision state to scan as much channels in a row as the
qos latency allows us.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: Replace {sw, hw}_scanning variables with a bitfield

Use a bitfield to store the current scan mode instead of two boolean
variables {sw,hw}_scanning. This patch does not introduce functional
changes but allows us to enhance the scan flags later (for example
for background scanning).

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: introduce a new scan state "decision"

Introduce a new scan state "decision" which is entered after
every completed scan operation and decides about the next steps.
At first the decision is in any case to scan the next channel.
This shouldn't introduce any functional changes.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: advance the state machine immediately if no delay is needed

Instead of queueing the scan work again without delay just process the
next state immediately.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: refactor the scan code

Move the processing of each scan state into its own functions for better
readability. This patch does not introduce functional changes.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Remove a few redundant variables/macros

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

ath9k: Cleanup return values

Cleanup aggregation start/stop function interfaces.

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

ath9k: Remove redundant HT macros

These can be obtained from mac80211.

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

ath9k: Fix a sparse warning

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

ath9k: Trivial fix in Kconfig

Update filename for debug information.

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

mac80211: Add a few 802.11n defines for AMPDU parameters

Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Add init values for AR9287 based chipsets.

Signed-off-by: Vivek Natarajan <vnatarajan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath: map TH to FCC3_WORLD

TH gets 5 GHz.

Cc: David Quan <david.quan@atheros.com>
Cc: Michael Green <michael.green@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: Don't alter rt2x00dev->default_ant

rt2x00dev->default_ant should be initialized once by the driver,
and should not be changed afterwards. Because rt2x00lib_config_antenna()
was using a reference to the struct antenna_setup it actually had the oppurtunity
to change the default antenna setting and it actually did that during the validation.

Instead of passing a pointer to antenna_setup the entire structure should be copied.

Signed-off-by: Lars Ericsson <Lars_Ericsson@telia.com>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: allow using network namespaces

This finally opens up the ability to put mac80211 devices
into different network namespaces. As long as you don't
have sysfs, that is.

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

cfg80211: make aware of net namespaces

In order to make cfg80211/nl80211 aware of network namespaces,
we have to do the following things:

 * del_virtual_intf method takes an interface index rather
   than a netdev pointer - simply change this

 * nl80211 uses init_net a lot, it changes to use the sender's
   network namespace

 * scan requests use the interface index, hold a netdev pointer
   and reference instead

 * we want a wiphy and its associated virtual interfaces to be
   in one netns together, so
    - we need to be able to change ns for a given interface, so
      export dev_change_net_namespace()
    - for each virtual interface set the NETIF_F_NETNS_LOCAL
      flag, and clear that flag only when the wiphy changes ns,
      to disallow breaking this invariant

 * when a network namespace goes away, we need to reparent the
   wiphy to init_net

 * cfg80211 users that support creating virtual interfaces must
   create them in the wiphy's namespace, currently this affects
   only mac80211

The end result is that you can now switch an entire wiphy into
a different network namespace with the new command
iw phy#<idx> set netns <pid>
and all virtual interfaces will follow (or the operation fails).

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

mac80211: cooperate more with network namespaces

There are still two places in mac80211 that hardcode
the initial net namespace (init_net). One of them is
mandated by cfg80211 and will be removed by a separate
patch, the other one is used for finding the network
device of a pending packet via its ifindex.

Remove the latter use by keeping track of the device
pointer itself, via the vif pointer, and avoid it
going stale by dropping pending frames for a given
interface when the interface is removed.

To keep track of the vif pointer for the correct
interface, change the info->control.vif pointer's
internal use to always be the correct vif, and only
move it to the vif the driver expects (or NULL for
monitor interfaces and injected packets) right before
giving the packet to the driver.

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

mac80211: do not queue work after suspend in the dynamic ps timer

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

iwlwifi: Read outside array bounds

tid is bounded (above) by the size of default_tid_to_tx_fifo (17 elements), but
the size of priv->stations[].tid[] is MAX_TID_COUNT (9) elements.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Read outside array bounds

Incorrect limits leads to reads outside array bounds.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

airo: Buffer overflow

SSID_rid has space for only 3 ssids.
txPowerLevels[i] is read before the bounds check for i

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>