[PPPOL2TP]: Fix SMP issues in skb reorder queue handling

When walking a session's packet reorder queue, use
skb_queue_walk_safe() since the list could be modified inside the
loop.

Rearrange the unlinking skbs from the reorder queue such that it is
done while the queue lock is held in pppol2tp_recv_dequeue() when
walking the skb list.

A version of this patch was suggested by Jarek Poplawski.

Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PPPOL2TP]: Make locking calls softirq-safe

Fix locking issues in the pppol2tp driver which can cause a kernel
crash on SMP boxes. There were two problems:-

1. The driver was violating read_lock() and write_lock() scheduling
   rules because it wasn't using softirq-safe locks in softirq
   contexts. So we now consistently use the _bh variants of the lock
   functions.

2. The driver was calling sk_dst_get() in pppol2tp_xmit() which was
   taking sk_dst_lock in softirq context. We now call __sk_dst_get().

Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atm: replace remaining __FUNCTION__ occurrences

__FUNCTION__ is gcc-specific, use __func__

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

isdn: replace __inline with inline

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

irda: replace __inline with inline

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: replace __inline with inline

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Bring MAX_BURST socket option into ietf API extension compliance

Brings max_burst socket option set/get into line with the latest ietf
socket extensions api draft, while maintaining backwards
compatibility.

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

SCTP: Fix chunk parameter processing bug

If an address family is not listed in "Supported Address Types"
parameter(INIT Chunk), but the packet is sent by that family, this
address family should be considered as supported by peer.  Otherwise,
an error condition will occur. For instance, if kernel receives an
IPV6 SCTP INIT chunk with "Support Address Types" parameter which
indicates just supporting IPV4 Address family. Kernel will reply an
IPV6 SCTP INIT ACK packet, but the source ipv6 address in ipv6 header
will be vacant. This is not correct.

refer to RFC4460 as following:
      IMPLEMENTATION NOTE: If an SCTP endpoint lists in the 'Supported
      Address Types' parameter either IPv4 or IPv6, but uses the other
      family for sending the packet containing the INIT chunk, or if it
      also lists addresses of the other family in the INIT chunk, then
      the address family that is not listed in the 'Supported Address
      Types' parameter SHOULD also be considered as supported by the
      receiver of the INIT chunk.  The receiver of the INIT chunk SHOULD
      NOT respond with any kind of error indication.

Here is a fix to comply to RFC.

Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

make s2io.c:init_tti() static

Make the needlessly global init_tti() static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: "Ramkrishna Vepa" <Ramkrishna.Vepa@neterion.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

FIXED_PHY must depend on PHYLIB=y

This patch fixes the following build error introduced by commit
a79d8e93d300adb84cccc38ac396cfb118c238ad and reported by Olaf Hering:

<--  snip  -->

...
  LD      .tmp_vmlinux1
arch/powerpc/sysdev/built-in.o: In function `of_add_fixed_phys':
fsl_soc.c:(.init.text+0xd34): undefined reference to `fixed_phy_add'
make: *** [.tmp_vmlinux1] Error 1

<--   snip  -->

Signed-off-by: Adrian Bunk <adrian.bunk@movial.fi>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

fix drivers/net/atarilance.c compilation

This patch fixes the following build error:

<--  snip  -->

...
  CC [M]  drivers/net/atarilance.o
{standard input}: Assembler messages:
{standard input}:406: Error: symbol `Lberr' is already defined
{standard input}:460: Error: symbol `Lberr' is already defined
make[3]: *** [drivers/net/atarilance.o] Error 1

<--  snip  -->

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

Intel ethernet adapter: Update MAINTAINERS

Another team member unfortunately left: update MAINTAINERS.

Condense the 3 lists down to a single list for all our drivers.

Point to our new sourceforge index page which is slightly
better navigateable than the sf.net project page.

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

ixgbe: fix typo in speed mesage

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

e100: Do suspend/shutdown like e1000

This fixes a "trying to free already free IRQ" message and simplifies
the shutdown/suspend code by re-using already existing code when going
to suspend. The code is now symmetric with e100_resume.

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

ehea: Fix missing Kconfig dependency

Fixed Kconfig: ehea driver requires sparse mem

Signed-off-by: Thomas Klein <tklein@de.ibm.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

axnet_cs: change debugging level for "Too much work at interrupt" message.

This message is frequently displayed even if normal file-transfer.

Signed-off-by: Komuro <komurojun-mbn@nifty.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

mv643xx_eth: Define module alias for platform device

The mv643xx_eth driver can be loaded as a platform device, as is done by
various Orion (ARM) based devices.  The driver needs to define a module
alias for the platform driver so udev will load it automatically.
Tested with Debian on a QNAP TS-209.

Signed-off-by: Martin Michlmayr <tbm@cyrius.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

de2104x: remove BUG_ON() when changing media type

When the chip dies (probably because of a bug somewhere in the driver),
de_stop_rxtx() fails and changing the media type crashes the whole machine.
Replace BUG_ON() in de_set_media() with a warning.

Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

dm9161: add configuration for MII/RMII

Signed-off-by: Frederic Rodo <f.rodo@til-technologies.fr>
Signed-off-by: Jeff Garzik <jeff@garzik.org>

[IPCONFIG]: The kernel gets no IP from some DHCP servers

From: Stephen Hemminger <shemminger@linux-foundation.org>

Based upon a patch by Marcel Wappler:

   This patch fixes a DHCP issue of the kernel: some DHCP servers
   (i.e.  in the Linksys WRT54Gv5) are very strict about the contents
   of the DHCPDISCOVER packet they receive from clients.

   Table 5 in RFC2131 page 36 requests the fields 'ciaddr' and
   'siaddr' MUST be set to '0'.  These DHCP servers ignore Linux
   kernel's DHCP discovery packets with these two fields set to
   '255.255.255.255' (in contrast to popular DHCP clients, such as
   'dhclient' or 'udhcpc').  This leads to a not booting system.

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

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

b43legacy: Fix module init message

This fixes the module init message to tell that the legacy
driver loaded. This makes it less confusing, in case both drivers are loaded.

Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rndis_wlan: fix broken data copy

Replace broken code that attempted to copy 6 byte array to 64-bit
integer. Due to missing cast to 64-bit integer, left shift operation
were 32-bit and lead to bytes been copied over each other. New code
uses simple memcpy, for greater readability and efficiency.

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

libertas: compare the current command with response

instead of with itself.

Signed-off-by: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: fix sanity check on sequence number in command response

Slightly more useful if we compare it against the sequence number of the
command we have outstanding, rather than comparing the reply with itself.

Doh. Pointed out by Sebastian Siewior

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

p54: fix eeprom parser length sanity checks

When I called p54_parse_eeprom() on a hand-coded structure
I managed to make a small mistake with wrap->len which caused
a segfault a few lines down when trying to read entry->len.
This patch changes the validation code to avoid such problems.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Tested-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54: fix EEPROM structure endianness

Since the EEPROM structure is read from hardware, it is
always little endian, annotate that in the struct and
make sure to convert where applicable.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Michael Wu <flamingice@sourmilk.net>
Tested-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ssb: Add pcibios_enable_device() return value check

This patch has added pcibios_enable_device() return value check.

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rc80211-pid: fix rate adjustment

Merge rate_control_pid_shift_adjust() to rate_control_pid_adjust_rate()
in order to make the learning algorithm aware of constraints on rates. Also
add some comments and rename variables.

This fixes a bug which prevented 802.11b/g non-AP STAs from working with
802.11b only AP STAs.

This patch was originally destined for 2.6.26, and is being backported
to fix a user reported problem in post-2.6.24 kernels.

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

[ESP]: Add select on AUTHENC

Now the ESP uses the AEAD interface even for algorithms which are
not combined mode, we need to select CONFIG_CRYPTO_AUTHENC as
otherwise only combined mode algorithms will work.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TCP]: Improve ipv4 established hash function.

If all of the entropy is in the local and foreign addresses,
but xor'ing together would cancel out that entropy, the
current hash performs poorly.

Suggested by Cosmin Ratiu:

Basically, the situation is as follows: There is a client
machine and a server machine. Both create 15000 virtual
interfaces, open up a socket for each pair of interfaces and
do SIP traffic. By profiling I noticed that there is a lot of
time spent walking the established hash chains with this
particular setup.

The addresses were distributed like this: client interfaces
were 198.18.0.1/16 with increments of 1 and server interfaces
were 198.18.128.1/16 with increments of 1. As I said, there
were 15000 interfaces. Source and destination ports were 5060
for each connection.  So in this case, ports don't matter for
hashing purposes, and the bits from the address pairs used
cancel each other, meaning there are no differences in the
whole lot of pairs, so they all end up in the same hash chain.

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

[NETPOLL]: Revert two bogus cleanups that broke netconsole.

Based upon a report by Andrew Morton and code analysis done
by Jarek Poplawski.

This reverts 33f807ba0d9259e7c75c7a2ce8bd2787e5b540c7 ("[NETPOLL]:
Kill NETPOLL_RX_DROP, set but never tested.")  and
c7b6ea24b43afb5749cb704e143df19d70e23dea ("[NETPOLL]: Don't need
rx_flags.").

The rx_flags did get tested for zero vs. non-zero and therefore we do
need those tests and that code which sets NETPOLL_RX_DROP et al.

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

[PPPOL2TP]: Add missing sock_put() in pppol2tp_tunnel_closeall()

Every skb removed from session->reorder_q needs sock_put().

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

Subject: [PPPOL2TP] add missing sock_put() in pppol2tp_recv_dequeue()

Every skb removed from session->reorder_q needs sock_put().

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

[BLUETOOTH]: l2cap info_timer delete fix in hci_conn_del

When the l2cap info_timer is active the info_state will be set to
L2CAP_INFO_FEAT_MASK_REQ_SENT, and it will be unset after the timer is
deleted or timeout triggered.

Here in l2cap_conn_del only call del_timer_sync when the info_state is
set to L2CAP_INFO_FEAT_MASK_REQ_SENT.

Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NET]: Fix race in generic address resolution.

neigh_update sends skb from neigh->arp_queue while neigh_timer_handler
has increased skbs refcount and calls solicit with the
skb. neigh_timer_handler should not increase skbs refcount but make a
copy of the skb and do solicit with the copy.

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iucv: fix build error on !SMP

Since a5fbb6d1064be885d2a6b82f625186753cf74848
"KVM: fix !SMP build error" smp_call_function isn't a define anymore
that folds into nothing but a define that calls up_smp_call_function
with all parameters. Hence we cannot #ifdef out the unused code
anymore...
This seems to be the preferred method, so do this for s390 as well.

net/iucv/iucv.c: In function 'iucv_cleanup_queue':
net/iucv/iucv.c:657: error: '__iucv_cleanup_queue' undeclared

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
Signed-off-by: Ursula Braun <braunu@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TCP]: Must count fack_count also when skipping

It makes fackets_out to grow too slowly compared with the
real write queue.

This shouldn't cause those BUG_TRAP(packets <= tp->packets_out)
to trigger but how knows how such inconsistent fackets_out
affects here and there around TCP when everything is nowadays
assuming accurate fackets_out. So lets see if this silences
them all.

Reported by Guillaume Chazarain <guichaz@gmail.com>.

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

[TUN]: Fix RTNL-locking in tun/tap driver

Current tun/tap driver sets also net device's hw address when asked to
change character device's hw address. This is a good idea, but it
misses RTLN-locking, resulting following error message in 2.6.25-rc3's
inetdev_event() function:

RTNL: assertion failed at net/ipv4/devinet.c (1050)

Attached patch fixes this problem.

Signed-off-by: Kim B. Heino <Kim.Heino@bluegiga.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Use proc_create to setup de->proc_fops.

In addition to commit 160f17 ("[SCTP]: Use proc_create() to setup
->proc_fops first") use proc_create in two more places.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IRDA]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TCP]: BIC web page link is corrected.

Signed-off-by: Sangtae Ha <sha2@ncsu.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

[X25]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[WANROUTER]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[8021Q]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PKTGEN]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NEIGHBOUR]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[LLC]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPX]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SUNRPC]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'pending' of /linux/kernel/git/vxy/lksctp-dev

[ATM]: Use proc_create() to setup ->proc_fops first

Use proc_create() to make sure that ->proc_fops be setup before gluing
PDE to main tree.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Update AUTH structures to match declarations in draft-16.

The new SCTP socket api (draft 16) updates the AUTH API structures.
We never exported these since we knew they would change.
Update the rest to match the draft.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>

[SCTP]: Incorrect length was used in SCTP_*_AUTH_CHUNKS socket option

The chunks are stored inside a parameter structure in the kernel
and when we copy them to the user, we need to account for
the parameter header.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>

[SCTP]: Clean up naming conventions of sctp protocol/address family registration

I noticed while looking into some odd behavior in sctp, that the variable
name sctp_pf_inet6_specific was used twice to represent two different
pieces of data (its both a structure name and a pointer to that type of
structure), which is confusing to say the least, and potentially dangerous
depending on the variable scope.  This patch cleans that up, and makes the
protocol and address family registration names in SCTP more regular,
increasing readability.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
 ipv6.c     |   12 ++++++------
 protocol.c |   12 ++++++------
 2 files changed, 12 insertions(+), 12 deletions(-)

[APPLETALK]: Use proc_create() to setup ->proc_fops first

As Davem mentioned in his recently patch
(d9595a7b9c777d45a74774f1428c263a0a47f4c0)
that the procfs visibility should occur after
the ->proc_fops are setup.

And also, Alexey provide proc_create() to make
sure that ->proc_fops is setup before gluing PDE
to main tree.

We use proc_create().

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: add bnx2x to MAINTAINERS

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: update version, remove CVS strings

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Fix Xmit bugs

Several endianity corrections in start_xmit()

Fixed TSO bug where packets were missing the TCP flags.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Prevent PCI queue overflow

Limit traffic through an internal queue to prevent overflow.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: fix slowpath races and locking

Fixed locking between fastpath and slowpath operations.

Corrected order of traffic disabling to prevent race when going down
under traffic.

- first have the microcode drop all incoming packets
- then do the slowpath stuff
- only then reset the MAC

Got rid of in_reset_task.

Remove_one() and friends would deference a null pointer if init_one
failed.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: fix HW attentions and error handling

Some of the HW attentions, used to indicate an error were not properly
acked.

This will cause the driver to endlessly receive interrupts when such
an error happens.

Had to break the code into smaller chunks because it got too nested.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: fix MSI-X/INT#A errata

Errata A0.158 workaround.

Running in INT#A mode after running with MSI-X fails due to a PCI core
bug.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: correct statistics

Errors were summed improperly, some stats were missing.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Correct RX filtering and MC configuration

The configuration of RX filtering needed the following corrections:

Drop flags need to be set per Rx queue.

Have to tell the microcode to collect drop stats, and properly wait
for them to complete when going down.

Sometimes we failed to detect proper completion due to a logical error
in the wait loop.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Correct Link management

Properly protect PHY access between two devices on the same board with
a HW lock.

Use GPIO to clear all previous configurations before changing link
parameters.

Shut down the external PHY in case of fan failure.

Reducing the MDC/MDIO clock to 2.5MHz due to problems with some
devices.

Resolve the flow control response according to autoneg with external
PHY.

Unmasking all PHY interrupts in single write to prevent a race in the
interrupts order.

LASI indication fixes to work with peculiarities of PHYs.

Disable MAC RX to avoid a HW bug when closing the MAC under traffic.

Disable parallel detection on HiGig due to HW limitation.

Updating the shared memory structure to work with the current
bootcode.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Correct init_one()

Correct PCI-E info printed by init_one()
In one case it failed to free the netdev.

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[BNX2X]: Spelling fixes

Signed-off-by: Eliezer Tamir <eliezert@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPCOMP]: Disable BH on output when using shared tfm

Because we use shared tfm objects in order to conserve memory,
(each tfm requires 128K of vmalloc memory), BH needs to be turned
off on output as that can occur in process context.

Previously this was done implicitly by the xfrm output code.
That was lost when it became lockless.  So we need to add the
BH disabling to IPComp directly.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

SSB PCI core driver: use new SPROM data structure

Switch the SSB PCI core driver to the new SPROM data structure now that
the old one has been removed.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Acked-by: Michael Buesch <mb@bu3sch.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Don't build bcm43xx if SSB is static and b43 PCI-SSB bridge is enabled.

This may happen e.g. when the ssb is statically enables by the b44 driver,
and the b43 pci-ssb bridge is enbled by the b43/b43legacy drivers, or the
b43/b43legacy drivers are built statically.

Signed-off-by: Alexey Zaytsev <alexey.zaytsev@gmail.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Use a separate config option for the b43 pci to ssb bridge.

The bridge code was unnecessary enabled by the b44
driver, but it prevents the bcm43xx driver from
being loaded, as the bridge claims the same pci ids.

Now we enable the birdge only if the b43{legacy}
drivers are selected.

Signed-off-by: Alexey Zaytsev <alexey.zaytsev@gmail.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

libertas: Remove unused exports

The libertas driver exports a number of symbols with no in-tree users;
remove these unused exports.  lbs_reset_device() is completely unused, with
no callers at all, so remove the function completely.

A couple of these unused exported symbols are static, which causes the
following build error on ia64 with gcc 4.2.3:

    drivers/net/wireless/libertas/main.c:1375: error: __ksymtab_lbs_remove_mesh causes a section type conflict
    drivers/net/wireless/libertas/main.c:1354: error: __ksymtab_lbs_add_mesh causes a section type conflict

Signed-off-by: Roland Dreier <rolandd@cisco.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>

mac80211: fix kmalloc vs. net_ratelimit

The "goto end;" part definitely must not be rate limited.

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

rt2x00: Fix rt2x00lib_reset_link_tuner()

rt2x00lib_reset_link_tuner() can be called from within
the link tuner itself. This means that it should
_not_ call rt2x00lib_stop_link_tuner() since that will
cause the thread to hang.

Reorder the things that should be done during a
link tuner reset and during a link tuner start.

Also make antenna tuning the last step of the link
tuner since it could possibly reset some statistical
information which we need for average calculation.

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

rt2x00: Don't switch to antenna with low rssi

When rssi_a > rssi_b is true and the current antenna
was already antenna A, then rt2x00 incorrectly jumped
to antenna B.

Also don't configure the antenna when there has been
no change in the antenna setup.

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

rt2x00: Add link tuner safe RX toggle states

This adds 2 new states which both are used to toggle
the RX. These new states are required for usage
inside the link tuner thread, because the normal
RX toggling will stop the link tuner thread.
While it is possible that the link tuner thread itself
is the caller of the RX toggle (when using software
antenna diversity).

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

rt2x00: Fix antenna diversity

Fix 2 issues in antenna diversity selection.

1) the following statement will always return true.
if ((rssi_curr - rssi_old) > -5 || (rssi_curr - rssi_old) < 5)
It is cleaner to check if the absolute value is smaller then 5.

2) Only enable software diversity when default antenna setup
indicates support for it. Don't select it when the hardware
does not indicate support for it...

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

gelic wireless driver needs WIRELESS_EXT support

|   CC      drivers/net/ps3_gelic_wireless.o
| /home/bigeasy/git/linux-2.6/drivers/net/ps3_gelic_wireless.c: In function 'gelic_wl_setup_netdev_ops':
| /home/bigeasy/git/linux-2.6/drivers/net/ps3_gelic_wireless.c:2660: error: 'struct net_device' has no member named 'wireless_data'
| /home/bigeasy/git/linux-2.6/drivers/net/ps3_gelic_wireless.c:2661: error: 'struct net_device' has no member named 'wireless_handlers'
| make[3]: *** [drivers/net/ps3_gelic_wireless.o] Error 1
| make[2]: *** [drivers/net] Error 2
| make[1]: *** [drivers] Error 2
| make[1]: *** Waiting for unfinished jobs....
| make: *** [sub-make] Error 2

Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Acked-by: Masakazu Mokuno <mokuno@sm.sony.co.jp>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

PS3: gelic: Link the wireless net_device structure to the corresponding device structure

Link the net_device structure of the wireless part to the
corresponding device structure.

Without this, the sysfs node for this net_device would not have
'device' link.

Signed-off-by: Masakazu Mokuno <mokuno@sm.sony.co.jp>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rndis_wlan: fix sparse warnings

It is obviously wrong to use an enum in a little endian struct,
and those other enums should be declared differently.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Acked-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

[SCTP]: Correctly set the length of sctp_assoc_change notification

sctp_assoc_change notification may contain the data from a received
ABORT chunk.  Set the length correctly to account for that.

Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>

[NETFILTER]: Fix NF_QUEUE_NR() parenthesis

Properly add parens around the macro argument. This is not needed by
the kernel but the macro is exported to userspace, so it shouldn't
make any assumptions.

Also use NF_VERDICT_BITS instead of NF_VERDICT_QBTIS for the left-shift
since thats whats logically correct.

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

[NETFILTER]: xt_conntrack: fix IPv4 address comparison

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: xt_conntrack: fix missing boolean clamping

Signed-off-by: Jan Engelhardt <jengelh@computergmbh.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: nf_conntrack: fix smp_processor_id() in preemptible code warning

Since we're using RCU for the conntrack hash now, we need to avoid
getting preempted or interrupted by BHs while changing the stats.

Fixes warning reported by Tilman Schmidt <tilman@imap.cc> when using
preemptible RCU:

[   48.180297] BUG: using smp_processor_id() in preemptible [00000000] code: ntpdate/3562
[   48.180297] caller is __nf_conntrack_find+0x9b/0xeb [nf_conntrack]
[   48.180297] Pid: 3562, comm: ntpdate Not tainted 2.6.25-rc2-mm1-testing #1
[   48.180297]  [<c02015b9>] debug_smp_processor_id+0x99/0xb0
[   48.180297]  [<fac643a7>] __nf_conntrack_find+0x9b/0xeb [nf_conntrack]

Tested-by: Tilman Schmidt <tilman@imap.cc>
Tested-by: Christian Casteyde <casteyde.christian@free.fr> [Bugzilla #10097]
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6] SYSCTL: Fix possible memory leakage in error path.

In error path, we do need to free memory just allocated.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[INET]: Don't create tunnels with '%' in name.

Four tunnel drivers (ip_gre, ipip, ip6_tunnel and sit) can receive a
pre-defined name for a device from the userspace.  Since these drivers
call the register_netdevice() (rtnl_lock, is held), which does _not_
generate the device's name, this name may contain a '%' character.

Not sure how bad is this to have a device with a '%' in its name, but
all the other places either use the register_netdev(), which call the
dev_alloc_name(), or explicitly call the dev_alloc_name() before
registering, i.e. do not allow for such names.

This had to be prior to the commit 34cc7b, but I forgot to number the
patches and this one got lost, sorry.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[AF_KEY]: Fix oops by converting to proc_net_*().

To make sure the procfs visibility occurs after the ->proc_fs ops are
setup, use proc_net_fops_create() and proc_net_remove().

This also fixes an OOPS after module unload in that the name string
for remove was wrong, so it wouldn't actually be removed.  That bug
was introduced by commit 61145aa1a12401ac71bcc450a58c773dd6e2bfb9
("[KEY]: Clean up proc files creation a bit.")

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

CONNECTOR: make cn_already_initialized static

It is used in connector.c only, so make it static.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV4]: Reset scope when changing address

This bug did bite at least one user, who did have to resort to rebooting
the system after an "ifconfig eth0 127.0.0.1" typo.

Deleting the address and adding a new is a less intrusive workaround.
But I still beleive this is a bug that should be fixed.  Some way or
another.

Another possibility would be to remove the scope mangling based on
address.  This will always be incomplete (are 127/8 the only address
space with host scope requirements?)

We set the scope to RT_SCOPE_HOST if an IPv4 interface is configured
with a loopback address (127/8).  The scope is never reset, and will
remain set to RT_SCOPE_HOST after changing the address. This patch
resets the scope if the address is changed again, to restore normal
functionality.

Signed-off-by: Bjorn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Add missing initializations of the new nl_info.nl_net field

Add some more missing initializations of the new nl_info.nl_net field
in IPv6 stack. This field will be used when network namespaces are
fully supported.

Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[MACVLAN]: Update Kconfig to refer to iproute

Since the macvlan release I had at least 5 users asking how to configure
it since the old userspace tool doesn't work with the version in the
kernel. Add a pointer to the Kconfig help.

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

bluetooth: delete timer in l2cap_conn_del()

Delete a possibly armed timer before kfree'ing the connection object.

Solves: http://lkml.org/lkml/2008/2/15/514

Reported-by:Quel Qun <kelk1@comcast.net>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: fix integer as NULL pointer sparse warnings in tipc

net/tipc/cluster.c:145:2: warning: Using plain integer as NULL pointer
net/tipc/link.c:3254:36: warning: Using plain integer as NULL pointer
net/tipc/ref.c:151:15: warning: Using plain integer as NULL pointer
net/tipc/zone.c:85:2: warning: Using plain integer as NULL pointer

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

Merge branch 'upstream-davem' of /linux/kernel/git/jgarzik/netdev-2.6