Merge branch 'master' of ssh:///linux/kernel/git/rusty/linux-2.6-for-linus

mac8390: update to net_device_ops

Another related 8390 driver. Since this is for nubus, not sure if
anyone still has the hardware?

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

etherh: build fix for net-next

Fix build of ARM etherh driver with new net_device_ops.

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

virtio_net: large tx MTU support

We don't really have a max tx packet size limit, so allow configuring
the device with up to 64k tx MTU.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

net: percpu_counter_inc() should not be called in BH-disabled section

Based upon a lockdep report by Alexey Dobriyan.

I checked all per_cpu_counter_xxx() usages in network tree, and I
think all call sites are BH enabled except one in
inet_csk_listen_stop().

commit dd24c00191d5e4a1ae896aafe33c6b8095ab4bd1
(net: Use a percpu_counter for orphan_count)
replaced atomic_t orphan_count to a percpu_counter.

atomic_inc()/atomic_dec() can be called from any context, while
percpu_counter_xxx() should be called from a consistent state.

For orphan_count, this context can be the BH-enabled one.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Feature activation handlers

This patch provides the post-processing of feature negotiation state, after
the negotiation has completed.

To this purpose, handlers are used and added to the dccp_feat_table. Each
handler is passed a boolean flag whether the RX or TX side of the feature
is meant.

Several handlers are provided already, new handlers can easily be added.

The initialisation is now fully dynamic, i.e. CCIDs are activated only
after the feature negotiation. The integration of this dynamic activation
is done in the subsequent patches.

Thanks to Wei Yongjun for pointing out the necessity of skipping over empty
Confirm options while copying the negotiated feature values.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Processing Confirm options

Analogous to the previous patch, this adds code to interpret incoming Confirm
feature-negotiation options. Both functions operate on the feature-negotiation
list of either the request_sock (server) or the dccp_sock (client).

Thanks to Wei Yongjun for pointing out that it is overly restrictive to check
the entire list of confirmed SP values.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Process incoming Change feature-negotiation options

This adds/replaces code for processing incoming ChangeL/R options.
The main difference is that:
 * mandatory FN options are now interpreted inside the function
  (there are too many individual cases to do this externally);
 * the function returns an appropriate Reset code or 0,
   which is then used to fill in the data for the Reset packet.

Old code, which is no longer used or referenced, has been removed.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Preference list reconciliation

This provides two functions to
 * reconcile preference lists (with appropriate return codes) and
 * reorder the preference list if successful reconciliation changed the
   preferred value.

The patch also removes the old code for processing SP/NN Change options, since
new code to process these is mostly there already; related references have been
commented out.

The code for processing Change options follows in the next patch.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Integrate feature-negotiation insertion code

The patch implements insertion of feature negotiation at the server (listening
and request socket) and the client (connecting socket).

In dccp_insert_options(), several statements have been grouped together now
to achieve (it is hoped) better efficiency by reducing the number of tests
each packet has to go through:
 - Ack Vectors are sent if the packet is neither a Data or a Request packet;
 - a previous issue is corrected - feature negotiation options are allowed
   on DataAck packets (5.8).

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Insert feature-negotiation options into skb

This patch replaces the earlier insertion routine from options.c, so that
code specific to feature negotiation can remain in feat.c. This is possible
by calling a function already existing in options.c.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>

gigaset: get rid of info() and warn() macros

Join the move away from the obsolete info() macro, opencoding the
remaining uses. While we're at it, also get rid of the warn() macro
by promoting the three remaining uses to err().

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>

gigaset: remove unnecessary poll method

The N_GIGASET_M101 line discipline implemented by the ser_gigaset
driver does not transfer any data from/to userspace through the
tty interface. Therefore a poll method is not needed.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb3: Fix sparse warning and micro-optimize is_pure_response()

The function is_pure_response() does "ntohl(var) & const" and then
essentially just tests whether the result is 0 or not; this can be done
more efficiently by computing "var & htonl(const)" instead and doing the
byte swap at compile time instead of run time.

This change slightly shrinks the compiled code; eg on x86-64 we save a
couple of bswapl instructions:

add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-8 (-8)
function                                     old     new   delta
t3_sge_intr_msix_napi                        544     536      -8

and this also has the pleasant side effect of fixing a sparse warning:

    drivers/net/cxgb3/sge.c:2313:15: warning: restricted degrades to integer

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phy: add the ST ste10Xp PHYs

This patch adds the STMicroelectronics ste10xp PHY device driver.
It supports both the ste100p and the ste101p devices.

Suspend/resume alredy added.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phy: Add file missed in previous commit.

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

phy: power management support

This patch adds the power management support into the physical
abstraction layer.

Suspend and resume functions respectively turns on/off the bit 11
into the PHY Basic mode control register.
Generic PHY device starts supporting PM.

In order to support the wake-on LAN and avoid to put in power down
the PHY device, the MDIO is aware of what the Ethernet device wants to do.

Voluntary, no CONFIG_PM defines were added into the sources.
Also generic suspend/resume functions are exported to allow
other drivers use them (such as genphy_config_aneg etc.).

Within the phy_driver_register function, we need to remove the
memset. It overrides the device driver owner and it is not good.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phy: add natsemi PHY driver

This patch adds the PHY device driver for the National Semiconductor
DP83865 Gig PHYTER.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ne2: silence static never defined warnings

These warning originate from 50014f1 (ne2: convert to
net_device_ops) which did drop those functions.

drivers/net/ne2.c:140: warning: 'ne_open' declared 'static' but never defined
drivers/net/ne2.c:141: warning: 'ne_close' declared 'static' but never defined

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2: use net_device_stats nowadays available in net_device

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

niu: use net_device_stats nowadays available in net_device

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

pkt_sched: fix sparse warning

Impact: make global function static

Fix the following sparse warning:

  net/sched/sch_api.c:192:14: warning: symbol 'qdisc_match_from_root' was not declared. Should it be static?

Signed-off-by: Hannes Eder <hannes@hanneseder.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlink: allow empty nested attributes

validate_nla() currently doesn't allow empty nested attributes. This
makes userspace code unnecessarily complicated when starting and ending
the nested attribute is done by generic upper level code and the inner
attributes are dumped by a module.

Add a special case to accept empty nested attributes. When the nested
attribute is non empty, the same checks as before are performed.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-next-2.6

Conflicts:

net/netfilter/nf_conntrack_netlink.c

decnet: compile fix for removal of byteorder wrapper

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: section fixes

ixgbe_init_interrupt_scheme() is called from ixgbe_resume().  Build that
with CONFIG_PM and without CONFIG_HOTPLUG and you've got a problem.
Several helpers called by it also are misannotated __devinit.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: fix missing label when PROC_FS=n

Fix missing label when CONFIG_PROC_FS=n:

net/sctp/protocol.c: In function 'sctp_proc_init':
net/sctp/protocol.c:106: error: label 'out_nomem' used but not defined
make[3]: *** [net/sctp/protocol.o] Error 1

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

igb: Add support for pci-e Advanced Error Reporting

Add the calls necessary to enable advanced error reporting for igb on
systems with AER enabled.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e100: cleanup link up/down messages

The system log messages created on a link status change need to follow a
specific format to work with tools some customers use.  This also makes
the messages consistant with other Intel driver link messages.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgb: cleanup link up/down messages

The system log messages created on a link status change need to follow a
specific format to work with tools some customers use.  This also makes
the messages consistant with other Intel driver link messages.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: cleanup link up/down messages

The system log messages created on a link status change need to follow a
specific format to work with tools some customers use.  This also makes
the messages consistant with other Intel driver link messages.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: cleanup link up/down messages

The system log messages created on a link status change need to follow a
specific format to work with tools some customers use.  This also makes
the messages consistant with other Intel driver link messages.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igb: link up/down messages must follow a specific format

The system log messages created on a link status change need to follow a
specific format to work with tools some customers use.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

decnet: remove private wrappers of endian helpers

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Reviewed-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Conflicts:

drivers/net/hp-plus.c
drivers/net/wireless/ath5k/base.c
drivers/net/wireless/ath9k/recv.c
net/wireless/reg.c

cxgb3: Update FW loading path.

Update FW loading path to accomodate in-kernel images location

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

cxgb3: integrate FW and protocol engines in the kernel

Include firmware and protocol images in the kernel image if requested

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

cxgb3: avoid potential memory leak.

Add consistency in alloc_ring() parameter checking
to avoid potential memory leaks.
alloc_ring() callers are correct fo far.

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

cxgb3: set hard_xmit in the netdev_ops

Fix t3_eth_xmit() missing into the netdev_ops structure.

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

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

pkt_sched: gen_estimator: Optimize gen_estimator_active()

Since all other gen_estimator functions use bstats and rate_est params
together, and searching for them is optimized now, let's use this also
in gen_estimator_active(). The return type of gen_estimator_active()
is changed to bool, and gen_find_node() parameters to const, btw.

In tcf_act_police_locate() a check for ACT_P_CREATED is added before
calling gen_estimator_active().

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mac80211_hwsim: fix-up some print_mac merge damage

Some print_mac -> %pM conversions got lost in some merge or another...

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

mac80211: fix unaligned access in ieee80211_wep_encrypt_data

Signed-off-by: Ivan Kuten <ivan.kuten@promwad.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: fix wmm queue settings

This patch fixes a regression (introduced by
"p54: more definitions form lmac_longbow.h and pda.h")

It turned out that the "ret" variable wasn't initialized and
this caused the following warnings/errors to appear:

wmaster1: failed to set TX queue parameters for queue 2
wmaster1: failed to set TX queue parameters for queue 3
wmaster1: failed to set TX queue parameters for queue 1
wmaster1: failed to set TX queue parameters for queue 0

Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Use proper TX channel width for setting channels

The TX channel width of the BSS can be obtained only after association.
In all cases, default to HT20 if HT is enabled, and set
chan width to HT40 only if the BSS supports it.

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

ath9k: Code scrub

Merge core.c and base.c
Remove Antenna Diversity (unused now).
Remove unused chainmask handling code.
Comment, indentation scrub.

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

nl80211: Change max TX power to be in mBm instead of dBm

In order to be consistent with NL80211_ATTR_POWER_RULE_MAX_EIRP,
change NL80211_FREQUENCY_ATTR_MAX_TX_POWER to use mBm and U32 instead
of dBm and U8. This is a userspace interface change, but the previous
version had not yet been pushed upstream and there are no userspace
programs using this yet, so there is justification to get this change in
as long as it goes in before the previous version gets out.

Signed-off-by: Jouni Malinen <j@w1.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Provide option to avoid unnecessary fw caching

Make firmware caching on startup optional, and make it default.

When the option is not selected and PM_SLEEP is configured, then
cache firmware in the suspend pm_notifier. This configuration saves
about 64k RAM in normal use, but can lead to a situation where the
driver is configured to use a different firmware.

Signed-off by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Resume spectrum_cs in the same way as orinoco_cs

Retrieval of external firmware has been resolved, and should work with
the standard orinoco resume algorithm.

This fixes an issue where priv->hw_unavailable indicates the card is
ready when firmware has not been loaded.

Signed-off by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Cache Symbol firmware

Signed-off by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

orinoco: Separate fw caching from download

This refactorring will make it easier to share logic with Symbol
firmware.

Signed-off by: David Kilroy <kilroyd@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Handle -ENOMEM on RX gracefully

We would get an oops on RX on -ENOMEM by passing
NULL to the hardware on ath_rx_buf_link(). The oops
would look something like this:

ath_rx_tasklet+0x515/0x53b
ath9k_tasklet+0x48
tasklet_action
__do_softirq
irq_exit
do_IRQ

RIP: ath_rx_buf_link+0x3a

We correct this by handling the requeue directly on
the ath_rx_tasklet() and trying to allocate an skb
*prior* to sending up the last hardware processed
skb. If we run out of memory this gauranteees we have
skbs to work with while it simply drops new received
frames.

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

rfkill: always call get_state() hook on resume

We "optimize" away the get_state() hook call on rfkill_toggle_radio
when doing a forced state change.  This means the resume path is not
calling get_state() as it should.

Call it manually on the resume handler, as we don't want to mess with
the EPO path by removing the optimization.  This has the added benefit
of making it explicit that rfkill->state could have been modified
before we hit the rfkill_toggle_radio() call in the class resume
handler.

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rfkill: preserve state across suspend

The rfkill class API requires that the driver connected to a class
call rfkill_force_state() on resume to update the real state of the
rfkill controller, OR that it provides a get_state() hook.

This means there is potentially a hidden call in the resume code flow
that changes rfkill->state (i.e. rfkill_force_state()), so the
previous state of the transmitter was being lost.

The simplest and most future-proof way to fix this is to explicitly
store the pre-sleep state on the rfkill structure, and restore from
that on resume.

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: Look out for some other AP when disassoc is received.

When a disassoc packet is received from the AP with a reason code of
'leaving the BSS', mac80211 should go into DISABLED state just as it
would do if the AP suddenly went away for some reason, as that is what
will happen shortly after the AP leaves anyway.

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

nl80211: Report max TX power in NL80211_BAND_ATTR_FREQS

This is useful information to provide for userspace (e.g., hostapd needs
this to generate Country IE).

Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath5k: Clean up eeprom parsing and add missing calibration data

This patch brings the ath5k eeprom parsing code in sync with the work
done on ath_info by Nick Kossifidis and integrates the missing parts
based on the code of the Atheros Legacy HAL release.

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

ath9k: Dont update rate control for every AMPDU

Update the rate control only with the tx status of first
AMPDU of an aggregation. This patch fixes frequent drops
in throughput.

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: TX update chicken bits

This instructs FH to increment the retry count of a packet when
it is brought from the memory to TX-FIFO to save transactions
during aggregation flow.

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

iwlwifi: TX setup fix confusion between TX queue and TX DMA channel

This patch configures correctly TX DMA channel. It is not
the same as TX queue.

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Acked-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: 4965 define firmware file name once

Apply same idiom as in 5000 introduced by
'iwlwifi: define firmware file name once'

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

iwlwifi: enable base band calibration in 5000 HW

This patch adds base band calibration support.

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

iwlwifi: move iwl_clear_stations_table to iwl-sta.c

This patch moves iwl_clear_stations_table into iwl-sta.c

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

iwl3945 : Fix ad-hoc mode for 3945

Patch fixes the ad-hoc mode by
1) Removing redundant clear_stations_table which prevented generation of
beacons.
2) Setting assoc_id to 1. It was never set so preventing tx flow
in iwl3945_tx_skb.

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

ath5k: set mac address in add_interface

Configure the mac address in add_interface and clear it in
remove_interface so that users can change the mac address
to something other than the one in the eeprom.  Also avoid
setting it at attach time so that we won't ack packets
until fully set up.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Use straightforward PCI routines to setup the TX buffer.

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

ath9k: Update TX trigger level on a FIFO underrun.

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

ath9k: Fix bug in deciphering channel flags

CHANNEL_CCK flag is set for all 2 Ghz channels, so IS_CHAN_CCK() would
turn out to be true for all channles in that band.
Use IS_CHAN_B() now, which checks the channel mode and not the channel
flags.

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

ath9k: Move TX completion routine to xmit.c

Also, use a helper function to setup RC status data
when processing completed TX descriptors.

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

ath9k: Use helpers

Break down huge functions, use helper functions or
macros instead.

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

ath9k: General code scrub

Replace TRUE/FALSE macros with VALID/INVALID macros.
Follow a consistent variable convention.
Remove unnecessary comments.
Add all RC phy macros into a single enum.
Merge functions into reasonably sized entities.

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

ath9k: Remove ath9k_rate_table

Maintaining two sets of rate tables is redundant, remove one
and use struct ath_rate_table exclusively.

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

ath9k: Use rate_driver_data

Remove the hack using vif, and use rate_driver_data within
skb->cb to hold driver specific rate information.
Setup the rate series in the skb's tx control area and remove
all references to ath9k specific rate series ( using struct ath_rc_series ).

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

ath9k: Remove ath_rate_softc

Move the hw rate tables to ath_softc, and access them directly.
tx_triglevel_max is global, move it to ath_rate_node.
Now that ath_rate_softc is gone, rate control attach becomes simpler.

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

ath9k: Remove rate control reference in VAP

Rate control init is now confined to itself, using the
HT capabilites of the STA from rate_init().

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

ath9k: We don't support non-HT devices, so remove superfluous code.

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

ath9k: Revamp RX handling

Remove a lot of old, crufty code and make
RX status reporting a bit sane and clean.

Do not do anything to the RX skb before unmapping.
So in ath_rx_tasklet(), move the skb_put() after PCI unmap.

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

ath9k: Nuke fixed rate handling in driver

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

ath9k: Remove half/quarter rate tables

Half/Quarter rate tables are needed only for legacy chipsets.

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

ath9k: Remove ath_rate_newassoc()

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

ath9k: Remove ath_setup_rates

Setup legacy rates in ath_rate_init() itself.

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

ath9k: Simplify RC alloc/free functions

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

ath9k: Merge struct ath_tx_ratectrl with ath_rate_node

Avoid casting of ath_tx_ratctrl and access the elements directly.

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

ath9k: Fix panic while unregistering rfkill

 [ 6133.670329] BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
 [ 6133.672802] IP: [<ffffffffa030fcf6>] ieee80211_stop_queues+0x26/0x40 [mac80211]
 [ 6133.672802] PGD 759dc067 PUD 74f1d067 PMD 0
 [ 6133.672802] Oops: 0002 [#1] PREEMPT SMP
 [ 6133.672802] last sysfs file: /sys/class/backlight/acpi_video0/brightness
 [ 6133.672802] CPU 0
 [ 6133.672802] Modules linked in: ath9k(-) mac80211 pciehp pci_hotplug arc4 ecb joydev pcmcia ppdev lp ppp_generic psmouse sg pcspkr s]
 [ 6133.735830] Pid: 4445, comm: rmmod Tainted: G        W  2.6.28-rc5-wl #1
 [ 6133.735830] RIP: 0010:[<ffffffffa030fcf6>]  [<ffffffffa030fcf6>] ieee80211_stop_queues+0x26/0x40 [mac80211]
 [ 6133.735830] RSP: 0018:ffff88007d1efd10  EFLAGS: 00010246
 [ 6133.735830] RAX: 0000000000000000 RBX: ffff880074f41aa0 RCX: 0000000000000000
 [ 6133.735830] RDX: 0000000000000010 RSI: 0000000000000000 RDI: ffff880074f40340
 [ 6133.735830] RBP: ffff880074990000 R08: 0000000000000000 R09: 000000000000224d
 [ 6133.735830] R10: 0000000000000000 R11: ffffffff8031dc70 R12: 0000000000000000
 [ 6133.735830] R13: 0000000000000001 R14: ffff880074f46c9c R15: 0000000000000000
 [ 6133.735830] FS:  00007f1e2e0bc6f0(0000) GS:ffffffff805e0b80(0000) knlGS:0000000000000000
 [ 6133.735830] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
 [ 6133.735830] CR2: 0000000000000010 CR3: 0000000075593000 CR4: 00000000000006e0
 [ 6133.735830] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 [ 6133.735830] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
 [ 6133.735830] Process rmmod (pid: 4445, threadinfo ffff88007d1ee000, task ffff88007c0c8000)
 [ 6133.735830] Stack:
 [ 6133.735830]  ffffffffa034d583 ffff88007c7d9410 ffff88007c7d9410 ffff88007c7d9410
 [ 6133.735830]  ffffffff80481dab ffff880074f41aa0 00000000fffffff0 0000000000000000
 [ 6133.735830]  0000000000000001 0000000000000001 ffffffffa034d8a5 ffff88007c7d9400
 [ 6133.735830] Call Trace:
 [ 6133.735830]  [<ffffffffa034d583>] ? ath_radio_disable+0x33/0x150 [ath9k]
 [ 6133.735830]  [<ffffffff80481dab>] ? __mutex_lock_slowpath+0x20b/0x2a0
 [ 6133.735830]  [<ffffffffa034d8a5>] ? ath_sw_toggle_radio+0x65/0xa0 [ath9k]
 [ 6133.735830]  [<ffffffffa019d1f4>] ? rfkill_toggle_radio+0x74/0x140 [rfkill]
 [ 6133.735830]  [<ffffffffa019d597>] ? rfkill_remove_switch+0x67/0x80 [rfkill]
 [ 6133.735830]  [<ffffffffa019d955>] ? rfkill_unregister+0x25/0x50 [rfkill]
 [ 6133.735830]  [<ffffffffa034bf75>] ? ath_detach+0xf5/0x140 [ath9k]
 [ 6133.735830]  [<ffffffffa034bfe9>] ? ath_pci_remove+0x29/0x80 [ath9k]
 [ 6133.735830]  [<ffffffff8035263c>] ? pci_device_remove+0x2c/0x60
 [ 6133.735830]  [<ffffffff803c3829>] ? __device_release_driver+0x99/0x100
 [ 6133.735830]  [<ffffffff803c3950>] ? driver_detach+0xc0/0xd0
 [ 6133.735830]  [<ffffffff803c296e>] ? bus_remove_driver+0x8e/0xd0
 [ 6133.735830]  [<ffffffff80352916>] ? pci_unregister_driver+0x36/0xa0
 [ 6133.735830]  [<ffffffffa0356ad4>] ? exit_ath_pci+0x10/0x29 [ath9k]
 [ 6133.735830]  [<ffffffff8026bb1b>] ? sys_delete_module+0x1cb/0x2d0
 [ 6133.735830]  [<ffffffff802960d9>] ? do_munmap+0x349/0x390
 [ 6133.735830]  [<ffffffff80342d01>] ? __up_write+0x21/0x150
 [ 6133.735830]  [<ffffffff8020c45b>] ? system_call_fastpath+0x16/0x1b
 [ 6133.735830] Code: c3 0f 1f 40 00 0f b7 57 5e 0f b7 47 5c 01 c2 74 30 31 c9 66 90 48 8b 57 78 0f b7 c1 48 c1 e0 07 48 03 82 00 03 00
 [ 6133.735830] RIP  [<ffffffffa030fcf6>] ieee80211_stop_queues+0x26/0x40 [mac80211]

Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtl8187: Fix transmission count sent to mac80211

In the commit entitled "mac80211/drivers: rewrite the rate control
API", the meaning of the packet transmit count was changed from the
number of retries to the total number.  In driver rtl8187, this change
was missed.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Acked-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Tested-by: Hin-Tak Leung <htl10@users.sourceforge.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: Fix pid rate-setting algorithm to allow rate changes

In commit 9ea2c74 named "mac80211/drivers: rewrite the rate control API",
the meaning of status.rates[i].count was changed from number of retries
to total number of tries. As a result, the pid rate-setting algorithm fails
because every packet appears to have needed a retransmit.

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

rt2x00: Don't switch off LED on initialization

When we switch off the LEDS during initialization
we kill rt73usb from proper functioning. The immediate
result after the first LED command are MCU failures
and a complete breakdown of TX/RX.

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

orinoco: indicate it is using dBm in wireless_stats and spy

Since WE7 /proc/net/wireless checks whether level and noise  are in dBm
and shows them accordingly. Indicate that we return signal and noice
levels in dBm.

Before:
Inter-| sta-|   Quality        |   Discarded packets               | Missed | WE
 face | tus | link level noise |  nwid  crypt   frag  retry   misc | beacon | 22
  eth1: 0000   65.  219.  165.       0      0    148     41      0        0

After:
Inter-| sta-|   Quality        |   Discarded packets               | Missed | WE
 face | tus | link level noise |  nwid  crypt   frag  retry   misc | beacon | 22
  eth1: 0000   65.  -37.  -91.       0      0      0      0      0        0

While at it, replace raw numbers with appropriate macro.

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

mac80211: make Minstrel the default rate control algorithm

This makes minstrel the default rate control algorithm
for mac80211. For more information see:

http://wireless.kernel.org/en/developers/Documentation/mac80211/RateControl/minstrel

If someone can come up with a better algorithm they get a prize
(undisclosed).

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

netfilter: ctnetlink: fix GFP_KERNEL allocation under spinlock

The previous fix for the conntrack creation race (netfilter: ctnetlink:
fix conntrack creation race) missed a GFP_KERNEL allocation that is
now performed while holding a spinlock. Switch to GFP_ATOMIC.

Reported-and-tested-by: Zoltan Borbely <bozo@andrews.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

smc-mca: Fix build failure due to typo.

ultra_netdev_ops --> ultramca_netdev_ops

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

sungem: Fix PCS_MIICTRL register write in gem_init_phy().

Use writel not writeb.

Noticed by Hermann Lauer.

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

net: release skb->dst in sock_queue_rcv_skb()

When queuing a skb to sk->sk_receive_queue, we can release its dst,
not anymore needed.  Since current cpu did the dst_hold(), refcount is
probably still hot int this cpu caches.

This avoids readers to access the original dst to decrement its
refcount, possibly a long time after packet reception. This should
speedup UDP and RAW receive path.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Use a percpu_counter for orphan_count

Instead of using one atomic_t per protocol, use a percpu_counter
for "orphan_count", to reduce cache line contention on
heavy duty network servers.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Use a percpu_counter for sockets_allocated

Instead of using one atomic_t per protocol, use a percpu_counter
for "sockets_allocated", to reduce cache line contention on
heavy duty network servers.

Note : We revert commit (248969ae31e1b3276fc4399d67ce29a5d81e6fd9
net: af_unix can make unix_nr_socks visbile in /proc),
since it is not anymore used after sock_prot_inuse_add() addition

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tc: policing requires a rate estimator

Found that while trying average rate policing, it was possible to
request average rate policing without a rate estimator. This results
in no policing which is harmless but incorrect.

Since policing could be setup in two steps, need to check
in the kernel.

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

tc: check for errors in gen_rate_estimator creation

The functions gen_new_estimator and gen_replace_estimator can return
errors, but they were being ignored.

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

tc: propogate errors from tcf_hash_create

Allow tcf_hash_create to return different errors on estimator failure.

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

net: make skb_truesize_bug() call WARN()

The truesize message check is important enough to make it print "BUG"
to the user console... lets also make it important enough to spit a
backtrace/module list etc so that kerneloops.org can track them.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: hp-plus uses eip_poll

hp-plus uses 8390p.c, so it should use eip_poll(), not ei_poll().

drivers/built-in.o: In function `hpp_probe1':
hp-plus.c:(.init.text+0x9cbd): undefined reference to `ei_poll'

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