[NET_SCHED]: sch_ingress: avoid a few #ifdefs

Move the repeating "ifndef CONFIG_NET_CLS_ACT/ifdef CONFIG_NETFILTER"
ifdefs into a single condition.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: move dependencies to Kconfig

Instead of complaining at scheduler initialization time, check the
dependencies in Kconfig.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: remove unnecessary ops

- ->reset is optional
- sch_api provides identical defaults for ->dequeue/->requeue
- ->drop can't happen since ingress never has a parent qdisc

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: return proper error code in ingress_graft()

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: remove unused inner qdisc

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: remove qdisc_priv() wrapper

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: remove excessive debugging

Remove excessive debugging statements and some "future use" stuff.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET_SCHED]: sch_ingress: formatting fixes

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jamal Hadi Salim <hadi@cyberus.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED] SFQ: whitespace cleanup

Add whitespace around operators, and add a few blank lines to improve
readability.

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

[PKT_SCHED] SFQ: use net_random

SFQ doesn't need true random numbers, it is only using them to salt a
hash. Therefore it is better to use net_random() and avoid any
possible problems with depleting the entropy pool.

Signed-off-by: Stephen Hemminger <stephen.hemminger@vyatta.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKT_SCHED] SFQ: timer is deferrable

The perturbation timer used for re-keying can be deferred, it doesn't
need to be deterministic.

Signed-off-by: Stephen Hemminger <stephen.hemminger@vyatta.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Process FIB rule action in the context of the namespace.

Save namespace context on the fib rule at the rule creation time and
call routing lookup in the correct namespace.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Acked-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: FIB rules API cleanup.

Remove struct net from fib_rules_register(unregister)/notify_change
paths and diet code size a bit.

add/remove: 0/0 grow/shrink: 10/12 up/down: 35/-100 (-65)
function                                     old     new   delta
notify_rule_change                           273     280      +7
trie_show_stats                              471     475      +4
fn_trie_delete                               473     477      +4
fib_rules_unregister                         144     148      +4
fib4_rule_compare                            119     123      +4
resize                                      2842    2845      +3
fn_trie_select_default                       515     518      +3
inet_sk_rebuild_header                       836     838      +2
fib_trie_seq_show                            764     766      +2
__devinet_sysctl_register                    276     278      +2
fn_trie_lookup                              1124    1123      -1
ip_fib_check_default                         133     131      -2
devinet_conf_sysctl                          223     221      -2
snmp_fold_field                              126     123      -3
fn_trie_insert                              2091    2086      -5
inet_create                                  876     870      -6
fib4_rules_init                              197     191      -6
fib_sync_down                                452     444      -8
inet_gso_send_check                          334     325      -9
fib_create_info                             3003    2991     -12
fib_nl_delrule                               568     553     -15
fib_nl_newrule                               883     852     -31

Signed-off-by: Denis V. Lunev <den@openvz.org>
Acked-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[FIB]: Add netns to fib_rules_ops.

The backward link from FIB rules operations to the network namespace
will allow to simplify the API a bit.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Acked-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Correctly initialize error when parameter validation failed.

When parameter validation fails, there should be error causes that
specify what type of failure we've encountered.  If the causes are not
there, we lacked memory to allocated them.  Thus make that the default
value for the error.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IrDA]: Irport removal - part 2

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IrDA]: Irport removal - part 1

This patch removes IrPORT and the old dongle drivers (all off them
have replacement drivers).

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IrDA]: Frame length validation.

When using a stir4200-based USB adaptor to talk to a device that uses
an mcp2150, the stir4200 sometimes drops an incoming frame causing the
mcp2150 to try and retransmit the lost frame. In this combination, the
next frame received from the mcp2150 is often invalid - either an
empty i:rsp or an IrCOMM i:rsp with an invalid clen. These corner
cases are now checked.

Signed-off-by: Robie Basak <rb-oss-1@justgohome.co.uk>
Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IrDA]: Resend frames on timeout.

When final timer expires, it might also mean that the i:cmd wasn't
received properly. If we have rejected frames, we can try to resend them.

Signed-off-by: Robie Basak <rb-oss-1@justgohome.co.uk>
Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Namespace stop vs 'ip r l' race.

During network namespace stop process kernel side netlink sockets
belonging to a namespace should be closed. They should not prevent
namespace to stop, so they do not increment namespace usage
counter. Though this counter will be put during last sock_put.

The raplacement of the correct netns for init_ns solves the problem
only partial as socket to be stoped until proper stop is a valid
netlink kernel socket and can be looked up by the user processes. This
is not a problem until it resides in initial namespace (no processes
inside this net), but this is not true for init_net.

So, hold the referrence for a socket, remove it from lookup tables and
only after that change namespace and perform a last put.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Tested-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Consolidate kernel netlink socket destruction.

Create a specific helper for netlink kernel socket disposal. This just
let the code look better and provides a ground for proper disposal
inside a namespace.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Tested-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Memory leak on network namespace stop.

Network namespace allocates 2 kernel netlink sockets, fibnl &
rtnl. These sockets should be disposed properly, i.e. by
sock_release. Plain sock_put is not enough.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Tested-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETNS]: Double free in netlink_release.

Netlink protocol table is global for all namespaces. Some netlink
protocols have been virtualized, i.e. they have per/namespace netlink
socket. This difference can easily lead to double free if more than 1
namespace is started. Count the number of kernel netlink sockets to
track that this table is not used any more.

Signed-off-by: Denis V. Lunev <den@openvz.org>
Tested-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix ip6_frag ctl

Alexey Dobriyan reported an oops when unsharing the network
indefinitely inside a loop. This is because the ip6_frag is not per
namespace while the ctls are.

That happens at the fragment timer expiration:
inet_frag_secret_rebuild function is called and this one restarts the
timer using the value stored inside the sysctl field.

        "mod_timer(&f->secret_timer, now + f->ctl->secret_interval);"

When the network is unshared, ip6_frag.ctl is initialized with the new
sysctl instances, but ip6_frag has only one instance. A race in this
case will appear because f->ctl can be modified during the read access
in the timer callback.

Until the ip6_frag is not per namespace, I discard the assignation to
the ctl field of ip6_frags in ip6_frag_sysctl_init when the network
namespace is not the init net.

Signed-off-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: NEWEMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
xmit. This routine doesn't check if the frame length exceeds the max.
MAL buffer size.

This patch now changes the driver to call emac_start_xmit_sg() on all
GigE platforms and not only the TAH equipped ones (440GX). This enables
an MTU of 9000 instead 4080.

Signed-off-by: Stefan Roese <sr@denx.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

add driver for enc28j60 ethernet chip

Signed-off-by: Claudio Lanconelli <lanconelli.claudio@eptar.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

SGISEEQ: fix oops when doing ifconfig down; ifconfig up

When doing init_ring checking whether a new skb needs to be allocated
was wrong.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sky2: version 1.21

Update driver version reflects new hardware support.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sky2: support for Yukon Supreme

Add support from sk98lin vendor driver 10.50.1.3 for 88E8055 and
88E8075 chips.  I don't have this hardware to test, so this changes
are untested.

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

qla3xxx annotations

stop the byteswap-in-place, annotate

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

slhc annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

eepro100 annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

annotate the rest of drivers/net/wan

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

annotate tun

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

NET: ns83820.c remove fastcall macro

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

SGISEEQ: use cached memory access to make driver work on IP28

- Use inline functions for dma_sync_* instead of macros
- added Kconfig change to make selection for similair SGI boxes easier

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Add me as maintainer of the RDC r6040 driver

This patch adds me as maintainer of the RDC R6040 Fast Ethernet driver.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
--
Signed-off-by: Jeff Garzik <jeff@garzik.org>

iwlwifi: skip mac80211 conf during a hardware scan and replay it afterwards

This patch skips mac80211 configuration setting during a hardware scan
and replays it afterwards for the iwlwifi drivers.

Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: proper monitor support

This patch changes the iwlwifi driver to properly support
monitor interfaces after the filter flags change.

The patch is originally created by Johannes Berg for iwl4965. I fixed some
of the comments and created a similar patch for iwl3945.

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

libertas: remove check for driver_lock in lbs_interrupt()

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert SUBSCRIBE_EVENT to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: change inference about buffer size in lbs_cmd()

sizeof(*cmd) is going to give the total size of the data structure that
we allocated, more often than not. But the size of the command to be
_sent_ could be a lot smaller, as it is in the KEY_MATERIAL and
SUBSCRIBE_EVENT commands for example. So swap them round; let the caller
set the _command_ size explicitly in the header, and infer the
maximum response size from the data structure.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert ENABLE_RSN to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert SET_WEP to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert SLEEP_PARAMS to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert INACTIVITY_TIMEOUT to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: convert RADIO_CONTROL to a direct command

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: submit RSSI command on tx timeout, to check whether module is dead

We don't necessarily want to reset the device on a TX timeout. But more
often than not, the real cause is that the firmware has crapped itself,
not just that the network is busy. So submit any harmless command, and
if _that_ times out, then the error handling code will reset the module,
as appropriate.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: make lbs_submit_command always 'succeed' and set command timer

Even if it fails, we want to wait a while and try again, with an
ultimate timeout if it the condition persists. So again, just use the
standard command timeout behaviour.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: discard DEFER responses to commands; let the timeout trigger

When the firmware returns 0x0004, it wants us to try again later. We can
achieve that simply by throwing out the response and letting the command
timeout code kick in.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: use priv->upld_buf for command responses

If we don't scribble over the command we sent, then we can retry it when
the firmware responds with 0x0004 (which means -EAGAIN).

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: reduce explicit references to priv->cur_cmd->cmdbuf

We have a local variable 'resp' which we use for this. So use it,
instead of typing the whole thing.

In preparation for actually using priv->upld_buf for the responses
instead...

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: Check for PS mode support on USB devices

Move the various firmware setup bits into a separate function, which
used to do just boot2 version.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: allow for PS mode to be disabled when firmware doesn't support it

Otherwise, we go into an endless busy loop trying to enable PS mode when
the command queue is empty, dealing with the error response, and then
trying to enable PS mode again because the command queue is empty.... it
doesn't really save much power.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: handle HOST_AWAKE event by sending WAKEUP_CONFIRM command

lbs_send_confirmwake() is a bit ugly but matches the way we confirm
sleep. We'll deal with that whole thing later.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: don't clear priv->dnld_sent after sending sleep confirm

DNLD_RES_RECEIVED is a bit of a misnomer -- we never wait for the result
to be received; it's purely representing the state of the TX path, and
in this case the TX path is definitely busy.

Of course, that means that we don't actually care about DATA_SENT vs.
CMD_SENT either, but that's a can of worms for another day...

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: fix buffer handling of PS_MODE commands and responses

Commit 5b8845345e7385d2eb37fac22ba9ab6905988be5 (or, in case the git
workflow is broken and patches get recommitted, the commit entitled
'libertas: rename and re-type bufvirtualaddr to cmdbuf' by dcbw),
introduced a number of bugs where we once had a pointer to a command
_payload_, but now we use the pointer to the command header instead.

The fix isn't wonderfully pretty for now, but it'll get better when we
finish converting all commands so the structures include the header.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: add SLEEP_PERIOD and FW_WAKE_METHOD command definitions

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: fix return from lbs_update_channel()

If we return the channel number in a 'ret' variable where anything
non-zero is later interpreted as an error, that isn't nice. It breaks
WPA, for a start. OLPC trac #5485

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: kill 'addtail' argument to lbs_queue_cmd() and make it static

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: handle command timeout in main thread instead of directly in timer

And handle the case where it times out more than once, too, instead of
locking up for ever.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: ensure response buffer size is always set for lbs_cmd_with_response

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: add __lbs_cmd_async() for asynchronous command submission

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: store command result in cmdnode instead of priv->cur_cmd_retcode

... at least for users of __lbs_cmd().

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: kill pdata_buf member of struct cmd_ctrl_node

We can use the callback_arg for it; that's the way we're heading anyway...

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: let __lbs_cmd() free its own cmdnode

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: kill cleanup_cmdnode()

Move the wakeup into lbs_complete_command(), and leave the other bits
in __lbs_cleanup_and_insert_cmd() which was the only caller now anyway.

There are two remaining direct callers of lbs_cleanup_and_insert_cmd(),
and they are both fine without the wakeup.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: don't re-initialise cmdnode when taking it off the free queue

We initialise it when we add it to the queue. No need to do it again.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: introduce and use lbs_complete_command() for command completion

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: remove some pointless checks for cmdnode buffer being present

We allocate them all at the same time, at startup. If they go missing,
we have more serious things to worry about, and the resulting oops will
be a perfectly acceptable result.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: don't use __lbs_cmd() with empty callback in if_usb.c

We're about to change semantics, leaving callers of
lbs_prepare_and_send_command() with the old broken priv->cur_cmd_retcode
crap. The existence of the callback command will be the trigger for the
new semantics when handling the response.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: rename and clean up DownloadcommandToStation

Call it lbs_submit_command(), remove a bunch of things which can be (or,
in the case of zeroing ->cmdwaitqwoken, already are) done elsewhere.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: kill unused wait_option field in struct cmd_ctrl_node

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: kill whitespace at end of lines

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: clean up if_usb driver

It was just getting on my tits, really.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: stop attempting to reset devices on unload

It wasn't working anyway -- by the time we get into if_usb_disconnect()
the USB core has already stopped us talking to the thing; even if it's
just on unload and the device still exists.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: don't exit worker thread until kthread_stop() is called

The kthread code can't cope with a thread exiting of its own accord and
then someone calling kthread_stop() for it. When the thread detects that
it needs to die, make it wait for kthread_stop() to be called.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43legacy: Fix rfkill radio LED

This fixes Bug #9414 for b43legacy. This patch is the equivalent of one
submitted earlier for b43.

Since addition of the rfkill callback, the LED associated with the off
switch on the radio has not worked for several reasons:

(1) Essential data in the rfkill structure were missing.
(2) The rfkill structure was initialized after the LED initialization.
(3) There was a minor memory leak if the radio LED structure was inited.

Once the above problems were fixed, additional difficulties were noted:

(4) The radio LED was in the wrong state at startup.
(5) The radio switch had to be manipulated twice for each state change.
(6) A circular mutex locking situation existed.
(7) If rfkill-input is built as a module, it is not automatically loaded.

This patch fixes all of the above and removes a couple of sparse warnings.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Stefano Brivio <stefano.brivio@polimi.it>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

S2io: Fixes to enable multiple transmit fifos

Multiple transmit fifo initialization -
  - Assigned equal scheduling priority for all configured FIFO's.
  - Modularized transmit traffic interrupt initialization since it is executed in
    s2io_card_up and s2io_link. Enable continuous tx interrupt when link is UP
    and vice verse.
  - Enable transmit interrupts for all configured transmit fifos.
  - Fixed typo errors.

Signed-off-by: Surjit Reang <surjit.reang@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

S2io: Fixes to enable multiple transmit fifo support

Fixes to enable multiple transmit fifos (upto a maximum of eight).
  - Moved single tx_lock from struct s2io_nic to struct fifo_info.
  - Moved single ufo_in_band_v structure from struct s2io_nic to struct
    fifo_info.
  - Assign the respective interrupt number for the transmitting fifo in the
    transmit descriptor (TXD).
- Added boundary checks for number of FIFOs enabled and FIFO length.

Signed-off-by: Surjit Reang <surjit.reang@neterion.com>
Signed-off-by: Sreenivasa Honnur <sreenivasa.honnur@neterion.com>
Signed-off-by: Ramkrishna Vepa <ram.vepa@neterion.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb3 - Fix EEH, missing softirq blocking

set_pci_drvdata() stores a pointer to the adapter,
not the net device.
Add missing softirq blocking in t3_mgmt_tx.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

cxgb3 - parity initialization for T3C adapters.

Add parity initialization for T3C adapters.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

pasemi: export pasemi_dma_init()

Forgot to export this one. Needed when pasemi_mac is compiled as a module.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

PLIP driver: convert the semaphore killed_timer_sem to completion

Signed-off-by: Matthias Kaehlcke <matthias.kaehlcke@gmail.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

bnx2x depends on ZLIB_INFLATE

The bnx2x module depends on the zlib_inflate functions.  The build will
fail if ZLIB_INFLATE has not been selected manually or by building another
module that automatically selects it.

Modify BNX2X config option to 'select ZLIB_INFLATE' like BNX2
and others.  This seems to fix it.

Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Eliezer Tamir <eliezert@broadcom.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

pcmcia net: use roundup_pow_of_two() macro instead of grotesque loop

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

e1000: Dump the eeprom when a user encounters a bad checksum

To help supporting users with a bad eeprom checksum, dump the
eeprom info when such a situation is encountered by a user.

Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

ixgb: enable sun hardware support for broadcom phy

Implement support for a SUN-specific PHY.

SUN provides a modified 82597-based board with their own
PHY that works with very little modification to the code. This
patch implements this new PHY which is identified by the
subvendor device ID. The device ID of the adapter remains
the same.

Signed-off-by: Matheos Worku <matheos.worku@sun.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

e1000: remove no longer used code for pci read/write cfg

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

ixgb endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

ixgbe endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

e100 endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

forcedeth endianness bugs

* misannotation: struct register_test members are actually host-endian
* bug: cpu_to_le64(n) >> 32 instead of cpu_to_le32(n >> 32) in setting
->bufhigh and similar for ->buflow (take low bits, _then_ convert to
little-endian, not the other way round).
* bug: setup_hw_rings() should not convert to little-endian at all (we
feed the result to writel(), not store in shared data structure), let
alone try to play with shifting and masking little-endian values.  Introduced
when setup_hw_rings() went in, screwed both 64bit case and the old code for
32bit rings it had replaced.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

NULL noise in drivers/net

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

e1000 endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

e1000e endianness annotations

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sungem endianness annotations\e

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

sunhme endianness annotations

This one is interesting - SBUS and PCI variants have
opposite endianness in descriptors (SBUS is sparc-only, so there
host-endian == big-endian).

Solution: declare a bitwise type (hme32) and in accessor
helpers do typechecking and force-casts (once we know that the
type is right).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

endianness annotations and fixes for olympic

* missing braces in !readl(...) & ...
* trivial endianness annotations
* in olympic_arb_cmd() the loop collecting fragments of
packet is b0rken on big-endian - we have
(next_ptr && (buf_ptr=olympic_priv->olympic_lap + ntohs(next_ptr)))
as condition and it should have swab16(), not ntohs() - it's host-endian
byteswapped, not big-endian.  So if we get more than one fragment on big-endian
host, we get screwed.
This ntohs() got missed back when the rest of those had been switched
to swab16() in 2.4.0-test2-pre1 - at a guess, nobody had hit fragmented
packets during the testing of PPC fixes.

PS: Ken Aaker cc'd on assumption that he is the same guy who'd done the
original set of PPC fixes in olympic

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

libertas: use spin_is_locked() instead of spin_trylock() in lbs_interrupt()

We get scary warnings on UP if we use spin_trylock() and find, as we
hoped, that the lock in question is already locked.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>