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

r6040: rework init_one error handling

This patch reworks the error handling in r6040_init_one
in order not to leak resources and correcly unmap and release
PCI regions of the MAC. Also prefix printk's with the driver name
for clarity.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: bump release number to 0.18

This patch bumps the release of the r6040 driver. There has been
quite some versions of it out there, but this one is the one
people should report bugs against.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: handle RX fifo full and no descriptor interrupts

This patch allows the MAC to handle the RX FIFO full
and no descriptor available interrupts. While we are at it
replace the TX interrupt with its corresponding definition.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: change the default waiting time

This patch changes the default waiting time of a packet, which
along with our previous r6040_rx path, was causing huge delays
with another host (160 to 230 ms).

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: use definitions for magic values in descriptor status

Define all the descriptor status the MAC can set.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: completely rework the RX path

This patch completely reworks the RX path in order to be
more accurate about what is going on with the MAC.
We no longer read the error from the MLSR register instead read
the descriptor status register which reflects, the error per descriptor.
We now allocate skbs on the fly in r6040_rx, and we handle allocation
failure instead of simply dropping the packet. Remove the
rx_free_desc counter of r6040_private structure since we
allocate skbs in the RX path.

r6040_rx_buf_alloc is now removed and becomes unuseless.

Signed-Off-By: Joerg Albert <jal2@gmx.de>
Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: call napi_disable when puting down the interface and set lp->dev accordingly.

We did not call napi_disabled when putting down the interface
which should be done. Finally initialize lp->dev when everything
is set.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

mv643xx_eth: fix NETPOLL build

Joseph Fannin <jfannin@gmail.com> and Takashi Iwai <tiwai@suse.de>
noticed that commit 073a345c04b01da0cc5b79ac7be0c7c8b1691ef5
("mv643xx_eth: clarify irq masking and unmasking") broke the
mv643xx_eth build when NETPOLL is enabled, due to it not renaming
one instance of INT_CAUSE_EXT in mv643xx_eth_netpoll().  This patch
takes care of that instance as well.

Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Cc: Dale Farnsworth <dale@farnsworth.org>
Cc: Joseph Fannin <jfannin@gmail.com>
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: rework the RX buffers allocation routine

Rework the RX buffers allocation function so that we do not
leak memory in the case we could not allocate skbs for the
RX path. Propagate the errors to the r6040_up function
where we call the RX buffers allocation function.

Also rename the r6040_alloc_txbufs function to
r6040_init_txbufs, to reflect what it really does.

Signed-Off-By: Joerg Albert <jal2@gmx.de>
Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: fix scheduling while atomic in r6040_tx_timeout

Add a helper function which only modifies R6040 MAC registers
use it when we timeout, and on adapter initialization. Fix
the scheduling while atomic but in the timeout routine due
to the reallocation of rx/tx buffers.

Signed-Off-By: Joerg Albert <jal2@gmx.de>
Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: fix null pointer access and tx timeouts

This patch fixes a null pointer access in r6040_rx due
to lp->dev not being initialized.

Fix the TX timeouts, TX irq was not re-enabled on RX irq

Signed-Off-By: Joerg Albert <jal2@gmx.de>
Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

r6040: prefix all functions with r6040

Prefix all functions inside the r6040 driver with r6040 to
avoid namespace clashing.

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

rndis_host: support WM6 devices as modems

This patch allows Windows Mobile 6 devices to be used for
tethering -- that is, used as modems.  It was requested by
AdamW in kernel bugzilla:

  http://bugzilla.kernel.org/show_bug.cgi?id=11119

and Mandriva kernel-discuss list.  It is tested and confirmed
to work by Peterl:

  http://forum.eeeuser.com/viewtopic.php?pid=323543#p323543

This patch is based on the patch in the above kernel bugzilla,
which is from the usb-rndis-lite tree.

[ dbrownell@users.sourceforge.net: misc fixes ]
Signed-off-by: Thomas Backlund <tmb@mandriva.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

at91_ether: use netstats in net_device structure

Use net_device_stats from net_device structure instead of local.

Signed-off-by: Paulius Zaleckas <paulius.zaleckas@teltonika.lt>
Tested-by: Marc Pignat <marc.pignat@hevs.ch>
Acked-by: Andrew Victor <linux@maxim.org.za>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

sfc: Create one RX queue and interrupt per CPU package by default

Using multiple cores in the same package to handle received traffic
does not appear to provide a performance benefit.  Therefore use CPU
topology information to count CPU packages and use that as the default
number of RX queues and interrupts.  We rely on interrupt balancing to
spread the interrupts across packages.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

sfc: Use a separate workqueue for resets

This avoids deadlock in case a reset is triggered during self-test.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

sfc: I2C adapter initialisation fixes

As recommended by Jean Delvare:
- Increase timeout to 50 ms
- Leave adapter class clear so that unwanted drivers do not probe our bus
- Use strlcpy() for name initialisation

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: make ioport free

This patch makes e1000 driver ioport-free.
This corrects behavior in probe function so as not to request ioport
resources as long as they are not really needed.  This is based on the
ioport-free patch of e1000 driver from Auke Kok and Tomohiro Kusumi.

Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@jp.fujitsu.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jeff Kirsher<jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: delete non NAPI code from the driver

Compile-tested only.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: remove email reference

The email linux-nics@intel.com is no longer available, remove all
references.

Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: Move assignments in tests before test

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: checkpatch clean

Redefine DPRINTK macro using do while(0)
__FUNCTION__ to __func__
structs {} on separate lines
Surround negative constants with ()

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: Remove spaces after casts and function names

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: Move extern function definitions to e1000.h

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: Use hw, er32, and ew32

Use struct e1000_hw *hw = adapter->hw; where necessary
Change macros E1000_READ_REG and E1000_WRITE_REG to er32 and ew32

Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

e1000: neaten function declarations

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Fix typo in meth driver

An | in an if statement to check a bit? I think this needs to be a &.
As a result of this typo meth will always operate in promiscuous mode.

Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

atl1e: Atheros L1E Gigabit Ethernet driver

Full patch for the Atheros L1E Gigabit Ethernet driver.
Supportring AR8121, AR8113 and AR8114

Signed-off-by: Jie Yang <jie.yang @atheros.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

ixgbe: remove device ID for unsupported device

The ixgbe driver was untested with device ID 8086:10c8 but still advertises
support.  Currently if this device is present in the system when the driver
is loaded, the system will panic.
Remove this device ID until full support can be tested with available
hardware.  This patch is necessary for 2.6.24, 2.6.25 and 2.6.26

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netdev: bunch of drivers: avoid WARN at net/core/dev.c:1328

The drivers were touching net queue before it has been started, so
without this patch, the drivers will potentially WARN at
net/core/dev.c:1328.

I don't have the hardware for the drivers below, so this patch is
untested, and thus should be carefully peer reviewed.

tc35815.c
au1000_eth.c
bfin_mac.c
macb.c
^ The four drivers are using phylib, they're calling netif_start_queue()
in open() callback. So trivially remove netif_tx_schedule_all().
Phylib will handle netif_carrier_*().

cpmac.c
fec_mpc52xx.c
fs_enet/fs_enet-main.c
sh_eth.c
^ The same as above, but these were also needlessly calling
netif_carrier_*() functions. So removed queue calls and also remove
carrier calls, since phylib will handle it. fs_enet-main.c also didn't
call netif_start_queue() at open(), this is fixed now.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

8139too: Fall back to PIO when MMIO fails.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

8139too: Force usage of PIO on OQO2

The OQO model 2 has an RTL8139 from Atheros that doesn't like MMIO.
It fails on modprobe, with a 'PCI Bus error' message.

Force it to always use polled IO.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

8139too: Make PIO/MMIO a runtime decision through a module parameter.

This is needed to support devices that only work with PIO without
penalising devices that work fine with MMIO in distro kernels.
It also allows us to eventually use PIO as a fallback when setting
up MMIO fails.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

net/phy: Fix 88e1111 copper/fiber selection in RGMII mode

MII_M1111_HWCFG_FIBER_COPPER_RES is a bit of MII_M1111_PHY_EXT_SR, not
MII_M1111_PHY_EXT_CR.

Signed-off-by: Wang Jian <lark@linux.net.cn>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

net/phy: Fix copper/fiber auto-selection for 88e1111

The 27.15 bit (MII_M1111_HWCFG_FIBER_COPPER_AUTO) is disable bit. When
set to 1, copper/fiber auto selection is disabled. The current code
to enable but actually disable auto selection.

Signed-off-by: Wang Jian <lark@linux.net.cn>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

net: tun.c fix cast

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

net: igb_main.c fix sparse integer as NULL pointer warning

drivers/net/igb/igb_main.c:388:20: warning: Using plain integer as NULL pointer

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

DM9000: Remove magic numbers

Remove magic numbers for items that we already have
defined in the register header file.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

MAINTAINERS: Update email address of Ursula Braun

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

s390/net/ctcm: message cleanup

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Tiedemann <ptiedem@de.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

s390/net: Cleanup claw printk messages.

Cc: Martin Schwidefsky <schwidefska@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andy Richter <richtera@us.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: update driver version

Change driver version to 4.0.0. Netxen firmwares as old as
v3.4.216 are supported.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: enable tso6, intr coalescing.

Enable tso6 and ipv6 checksum, interrupt coalescing for NX3031.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: cut-through rx changes

NX3031 supports cut-through operation where ingress packets are
directly dma'ed into host buffers to reduce latency.

This requires larger dma buffers (2kb) and different alignemnt.

The buffer posting logic is changed a bit. The free rx buffers
are maintained in linked list, since the received reference
handles can be out of order. However rx descriptors are still
posted sequentially, indexed by producer.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: mtu, mac, link status changes

MAC addr, multicast filters, mtu are set through firmware commands
in firmware v4.0.0+ because of virtualization of physical ports.
Link status is also read from registers allocated by firmware for
each virtual port.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: add netxen_nic_ctx.c

Contains rx and tx ring context management and certain
firmware commands for netxen firmware v4.0.0+.

This patch gathers all HW context management code into
netxen_nic_ctx.c.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: remove netxen_nic_isr.c

It had only couple of functions which are moved to main.c

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: pci probe and firmware init changes

Add initialization code in pci probe for new chip and retain
compatibility with old revisions.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: add 2MB PCI memory support

New revision of netxen chip has 2MB PCI memory. Older chips
had 128MB addressable PCI memory. To retain compatibility,
this patch adds function pointers based on pci bar0 size.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: add defs for new chip/boards

Add macros for new chip revision and board configurations.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: hw multicast filtering

Enable multicast address filtering capabilities in the hardware.
Upto 16 multicast addresses can be programmed for each physical
port. Support "allmulti" mode, if enabled.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

netxen: cleanup unused variables/functions

o Reduce access to global arrays in data path.
o Remove duplicate/unused variables, unecessary alignment constraints.
o Use correct pci_dev instead of fallback device for consistent
  allocations.
o Disable ethtool set_eeprom functionality for now, it was only used
  for flashing firmware.

Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

net 8390p: fix interface usage

various drivers were using the wrong APIs:

 drivers/built-in.o: In function `hp_probe1':
 hp.c:(.init.text+0xa280): undefined reference to `NS8390_init'

fixed via:

cd drivers/net/; sed -i 's/NS8390_/NS8390p_/g' \
  $(grep -l NS8390_ $(grep 8390p.o Makefile | cut -d' ' -f3 | \
      sed 's/.o$/.c/g'))

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

ipv6: icmp6_dst_gc return change

Change icmp6_dst_gc to return the one value the caller cares about rather
than using call by reference.

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

ipv6: use kcalloc

Th fib_table_hash is an array, so use kcalloc.

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

ipv6: use spin_trylock_bh

Now there is spin_trylock_bh, use it rather than open coding.

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

ipv6: use round_jiffies

This timer normally happens once a minute, there is no need to cause an
early wakeup for it, so align it to next second boundary to safe power.
It can't be deferred because then it could take too long on cleanup or DoS.

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

netns: dont alloc ipv6 fib timer list

FIB timer list is a trivial size structure, avoid indirection and just
put it in existing ns.

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

ipv6: make struct ipv6_devconf static

struct ipv6_devconf can now become static.

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

sctp: remove sctp_assoc_proc_exit()

Commit 20c2c1fd6c842caf70dcb1d94b9d58861949fd3d
(sctp: add sctp/remaddr table to complete RFC remote address table OID)
added an unused sctp_assoc_proc_exit() function that seems to have been
unintentionally created when copying the assocs code.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: make sctp_outq_flush() static

sctp_outq_flush() can now become static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pkt_sched: make qdisc_class_hash_alloc() static

This patch makes the needlessly global qdisc_class_hash_alloc() static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netns: make get_proc_net() static

get_proc_net() can now become static.

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

netdev: Handle ->addr_list_lock just like ->_xmit_lock for lockdep.

The new address list lock needs to handle the same device layering
issues that the _xmit_lock one does.

This integrates work done by Patrick McHardy.

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

net: Fix build failure with 'make mandocs'.

The function header comments have to go with the functions
they are documenting, or things go horribly wrong when we
try to process them with the docbook tools.

Warning(include/linux/netdevice.h:1006): No description found for parameter 'dev_queue'
Warning(include/linux/netdevice.h:1033): No description found for parameter 'dev_queue'
Warning(include/linux/netdevice.h:1067): No description found for parameter 'dev_queue'
Warning(include/linux/netdevice.h:1093): No description found for parameter 'dev_queue'
Warning(include/linux/netdevice.h:1474): No description found for parameter 'txq'
Error(net/core/dev.c:1674): cannot understand prototype: 'u32 simple_tx_hashrnd; '

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

myri10ge: use ioremap_wc

Switch to ioremap_wc(). We keep the MTRR code since ioremap_wc()
will use UC_MINUS when falling back to uncachable, and thus let
the MTRR WC take precedence.

Also rename the error path better.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

myri10ge: remove wcfifo

Remove the wcfifo since it never gave any performance improvement.

Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

de4x5: Remove developer debug feature about set/clear promisc

IFF_PROMISC flag shouldn't be set or cleared by drivers, because
whether device be promisc mode is decided by how many upper layer
callers being referenced to it.
And the promisc changing feature of de4x5 ioctl is developer debug
feature, we can remove it now.

Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Acked-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

qeth: fix build error caused by VLAN changes

Adrian Bunk reports this build error:

CC      drivers/s390/net/qeth_l3_main.o
/home/bunk/linux/kernel-2.6/git/linux-2.6/drivers/s390/net/qeth_l3_main.c:
In function 'qeth_l3_hard_start_xmit':
/home/bunk/linux/kernel-2.6/git/linux-2.6/drivers/s390/net/qeth_l3_main.c:
2654: error: implicit declaration of function 'VLAN_TX_SKB_CB'
/home/bunk/linux/kernel-2.6/git/linux-2.6/drivers/s390/net/qeth_l3_main.c:
2654: error: invalid type argument of '->' (have 'int')
make[3]: *** [drivers/s390/net/qeth_l3_main.o] Error 1

The intention of the driver appears to be to invalidate the VLAN tag.
Change it to set skb->vlan_tci to zero, which has the same effect.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

fs_enet: Remove unused fields in the fs_mii_bb_platform_info structure.

The mdio_port, mdio_bit, mdc_port and mdc_bit fields in the
fs_mii_bb_platform_info structure are left-overs from the move to the Phy
Abstraction Layer subsystem. They are not used anymore and can be safely
removed.

Signed-off-by: Laurent Pinchart <laurentp@cse-semaphore.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

Merge branch 'r8169-davem' of git://git./linux/kernel/git/romieu/netdev-2.6 into upstream-fixes

tcp: Fix bitmask test in tcp_syn_options()

As reported by Alexey Dobriyan:

  CHECK   net/ipv4/tcp_output.c
net/ipv4/tcp_output.c:475:7: warning: dubious: !x & y

And sparse is damn right!

if (unlikely(!OPTION_TS & opts->options))
    ^^^
size += TCPOLEN_SACKPERM_ALIGNED;

OPTION_TS is (1 << 1), so condition will never trigger.

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

ipv6: __KERNEL__ ifdef struct ipv6_devconf

Based upon a report by Olaf Hering.

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

udplite: Protection against coverage value wrap-around

This patch clamps the cscov setsockopt values to a maximum of 0xFFFF.

Setsockopt values greater than 0xffff can cause an unwanted
wrap-around.  Further, IPv6 jumbograms are not supported (RFC 3838,
3.5), so that values greater than 0xffff are not even useful.

Further changes: fixed a typo in the documentation.

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

net: Print the module name as part of the watchdog message

As suggested by Dave:

This patch adds a function to get the driver name from a struct net_device,
and consequently uses this in the watchdog timeout handler to print as
part of the message.

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

net: use kcalloc in netdev_queue alloc

Minor nit, use size_t for allocation size and kcalloc to allocate
an array. Probably makes no actual code difference.

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

ipv6: use timer pending

This fixes the bridge reference count problem and cleanups ipv6 FIB
timer management.  Don't use expires field, because it is not a proper
way to test, instead use timer_pending().

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

mac80211_hwsim.c: fix: BUG: unable to handle kernel NULL pointer dereference at 0000000000000370

I was looking at this out of interest, but I'm in no way familiar with
the code.

Looks to me that the error handling code in mac80211_hwsim is awkward.
Which leads to it calling ieee80211_unregister_hw even when
ieee80211_register_hw failed.

The function has a for loop where it generates all simulated radios.
when something fails, the error handling will call mac80211_hwsim_free
which frees all simulated radios who's pointer isn't zero. However the
information stored is insufficient to determine whether or not the call
to ieee80211_register_hw succeeded or not for a specific radio. The
included patch makes init_mac80211_hwsim clean up the current simulated
radio, and then calls into mac80211_hwsim_free to clean up all the
radios that did succeed.

This however doesn't explain why the rate control registration failed..
build tested this, but had some problems reproducing the original
problem.

Signed-off-by: Ian Schram <ischram@telenet.be>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: nf_conntrack_sctp: fix sparse warnings

Introduced by a258860e (netfilter: ctnetlink: add full support for SCTP to ctnetlink):

net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: incorrect type in argument 1 (different base types)
net/netfilter/nf_conntrack_proto_sctp.c:483:2:    expected unsigned int [unsigned] [usertype] x
net/netfilter/nf_conntrack_proto_sctp.c:483:2:    got restricted unsigned int const <noident>
net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:483:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: incorrect type in argument 1 (different base types)
net/netfilter/nf_conntrack_proto_sctp.c:487:2:    expected unsigned int [unsigned] [usertype] x
net/netfilter/nf_conntrack_proto_sctp.c:487:2:    got restricted unsigned int const <noident>
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:487:2: warning: cast from restricted type
net/netfilter/nf_conntrack_proto_sctp.c:532:42: warning: incorrect type in assignment (different base types)
net/netfilter/nf_conntrack_proto_sctp.c:532:42:    expected restricted unsigned int <noident>
net/netfilter/nf_conntrack_proto_sctp.c:532:42:    got unsigned int
net/netfilter/nf_conntrack_proto_sctp.c:534:39: warning: incorrect type in assignment (different base types)
net/netfilter/nf_conntrack_proto_sctp.c:534:39:    expected restricted unsigned int <noident>
net/netfilter/nf_conntrack_proto_sctp.c:534:39:    got unsigned int

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

netfilter: nf_nat_sip: c= is optional for session

According to RFC2327, the connection information is optional
in the session description since it can be specified in the
media description instead.

My provider does exactly that and does not provide any connection
information in the session description.  As a result the new
kernel drops all invite responses.

This patch makes it optional as documented.

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

netfilter: xt_TCPMSS: collapse tcpmss_reverse_mtu{4,6} into one function

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

netfilter: nfnetlink_log: send complete hardware header

This patch adds some fields to NFLOG to be able to send the complete
hardware header with all necessary informations.
It sends to userspace:
 * the type of hardware link
 * the lenght of hardware header
 * the hardware header

Signed-off-by: Eric Leblond <eric@inl.fr>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: xt_time: fix time's time_mt()'s use of do_div()

Fix netfilter xt_time's time_mt()'s use of do_div() on an s64 by using
div_s64() instead.

This was introduced by patch ee4411a1b1e0b679c99686629b5eab5a072ce49f
("[NETFILTER]: x_tables: add xt_time match").

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

netfilter: accounting rework: ct_extend + 64bit counters (v4)

Initially netfilter has had 64bit counters for conntrack-based accounting, but
it was changed in 2.6.14 to save memory. Unfortunately in-kernel 64bit counters are
still required, for example for "connbytes" extension. However, 64bit counters
waste a lot of memory and it was not possible to enable/disable it runtime.

This patch:
 - reimplements accounting with respect to the extension infrastructure,
 - makes one global version of seq_print_acct() instead of two seq_print_counters(),
 - makes it possible to enable it at boot time (for CONFIG_SYSCTL/CONFIG_SYSFS=n),
 - makes it possible to enable/disable it at runtime by sysctl or sysfs,
 - extends counters from 32bit to 64bit,
 - renames ip_conntrack_counter -> nf_conn_counter,
 - enables accounting code unconditionally (no longer depends on CONFIG_NF_CT_ACCT),
 - set initial accounting enable state based on CONFIG_NF_CT_ACCT
 - removes buggy IPCT_COUNTER_FILLING event handling.

If accounting is enabled newly created connections get additional acct extend.
Old connections are not changed as it is not possible to add a ct_extend area
to confirmed conntrack. Accounting is performed for all connections with
acct extend regardless of a current state of "net.netfilter.nf_conntrack_acct".

Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlink: add NLA_PUT_BE64 macro

Add NLA_PUT_BE64 macro required for 64bit counters in netfilter

Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: nf_nat_core: eliminate useless find_appropriate_src for IP_NAT_RANGE_PROTO_RANDOM

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

hdlcdrv: Fix CRC calculation.

This is a trivial patch against the hdlcdrv module that fixes its CRC
calculation. The finished CRC was overwriting the first two bytes of
each packet rather than being appended to the end.

I've tested this with 2.6.8 and 2.6.10-rc1, but hdlcdrv hasn't changed
much recently so it should work with many other kernel versions.

Signed-off-by: Micah Dowty <micah@navi.cx>
Acked-by: Thomas Sailer <t.sailer@alumni.ethz.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert "pkt_sched: Make default qdisc nonshared-multiqueue safe."

This reverts commit a0c80b80e0fb48129e4e9d6a9ede914f9ff1850d.

After discussions with Jamal and Herbert on netdev, we should
provide at least minimal prioritization at the qdisc level
even in multiqueue situations.

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

net: In __netif_schedule() use WARN_ON instead of BUG_ON

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

net: Improve simple_tx_hash().

Based upon feedback from Eric Dumazet and Andi Kleen.

Cure several deficiencies in simple_tx_hash() by using
jhash + reciprocol multiply.

1) Eliminates expensive modulus operation.

2) Makes hash less attackable by using random seed.

3) Eliminates endianness hash distribution issues.

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

pkt_sched: Remove unused variable skb in dev_deactivate_queue function.

Removed unused variable 'skb' in the dev_deactivate_queue function

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

sunhme: Remove stop/wake TX queue calls in set-multicast-list handler.

Based upon a bug report by Alexander Beregalov and commentary
from Ben Hutchings.

These are totally unnecessary, in particular because this
driver's ->hard_start_xmit() handler takes the same driver
spinlock that the set-multicast-list handler uses.

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

ucc_geth: do not touch net queue in adjust_link phylib callback

If the net queue has not been started, we'll get this nice oops
and non-working ethernet:

------------[ cut here ]------------
Kernel BUG at c01f4648 [verbose debug info unavailable]
Oops: Exception in kernel mode, sig: 5 [#1]
MPC836x RDK
Modules linked in:
NIP: c01f4648 LR: c01c0a10 CTR: c01c08e4
REGS: cf839e40 TRAP: 0700   Not tainted  (2.6.26-05254-gc7b9969)
MSR: 00021032 <ME,IR,DR>  CR: 22042044  XER: 00000000
TASK = cf828c30[4] 'events/0' THREAD: cf838000
GPR00: c01c0a10 cf839ef0 cf828c30 c035ceb0 cf8469a0 00000064 00000000 00000000
GPR08: c035ceb0 00000001 00000001 cf99c280 22044044 7ca81020 0fffc000 00000000
GPR16: 0fff2544 0fff63c0 00000000 0fff78e0 0ffa5580 00000004 00000000 00000000
GPR24: 02082000 cf9d0000 d1068000 00009032 cf846800 cf846b80 00000001 00000014
NIP [c01f4648] __netif_schedule+0x28/0x8c
LR [c01c0a10] adjust_link+0x12c/0x1e4
Call Trace:
[cf839ef0] [c0380f50] 0xc0380f50 (unreliable)
[cf839f10] [c01c0a10] adjust_link+0x12c/0x1e4
[cf839f40] [c01c2628] phy_state_machine+0x2e0/0x448
[cf839f60] [c00425e8] run_workqueue+0xc8/0x168
[cf839f90] [c0042c6c] worker_thread+0x70/0xd0
[cf839fd0] [c0046954] kthread+0x48/0x84
[cf839ff0] [c0012488] kernel_thread+0x44/0x60
Instruction dump:
7c0803a6 4e800020 3d20c036 9421ffe0 7c0802a6 7c681b78 3929ceb0 7c694a78
7d290034 90010024 bfa10014 5529d97e <0f090000> 39600002 38030024 7d200028
---[ end trace a57d367843bd2904 ]---

Since the driver is using phylib (which is doing netif_carrier_on/off()),
we should simply remove netif_tx_schedule_all() from adjust_link().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: do not touch net queue in adjust_link phylib callback

If the net queue has not been started, we'll get this nice oops
and non-working ethernet:

PHY: 0:01 - Link is Up - 1000/Full
------------[ cut here ]------------
kernel BUG at net/core/dev.c:1328!
Oops: Exception in kernel mode, sig: 5 [#1]
MPC837x RDB
Modules linked in:
NIP: c02544a0 LR: c01a17d0 CTR: c01a16ac
REGS: cf837e40 TRAP: 0700   Not tainted  (2.6.26-05253-g14b395e)
MSR: 00021032 <ME,IR,DR>  CR: 22042044  XER: 00000000
TASK = cf819400[5] 'events/0' THREAD: cf836000
GPR00: c01a17d0 cf837ef0 cf819400 c03d8d08 cf8469a0 00000064 00000000 00000000
GPR08: c03d8d08 00000001 00000001 cf899ba0 22044044 00000000 0fffd000 00000000
GPR16: 0fff3028 0fff6cf0 00000000 0fff8390 0ff494a0 00000004 00000000 00000000
GPR24: c0361a00 00001058 cf9f6600 00009032 cf846800 cf846b80 00000001 00000014
NIP [c02544a0] __netif_schedule+0x28/0x8c
LR [c01a17d0] adjust_link+0x124/0x1cc
Call Trace:
[cf837ef0] [c03fb3a0] 0xc03fb3a0 (unreliable)
[cf837f10] [c01a17d0] adjust_link+0x124/0x1cc
[cf837f40] [c01a8e28] phy_state_machine+0x2e0/0x448
[cf837f60] [c0040254] run_workqueue+0xc8/0x168
[cf837f90] [c00408d8] worker_thread+0x70/0xd0
[cf837fd0] [c0044630] kthread+0x48/0x84
[cf837ff0] [c0012610] kernel_thread+0x44/0x60
Instruction dump:
7c0803a6 4e800020 3d20c03e 9421ffe0 7c0802a6 7c681b78 39298d08 7c694a78
7d290034 90010024 bfa10014 5529d97e <0f090000> 39600002 38030024 7d200028
---[ end trace 13dfd73ee42d0c30 ]---

Since the driver is using phylib (which is doing netif_carrier_on/off()),
we should simply remove netif_tx_schedule_all() from adjust_link().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atl1: Do not wake queue before queue has been started.

Based upon a bug report by Alexey Dobriyan, the patch is
also tested by him and confirmed to fix the problem.

Packet flow during link state events should not be done by
waking and stopping the TX queue anyways, that is handled
transparently by netif_carrier_{on,off}().

So, remove the netif_{wake,stop}_queue() calls in the link
check code, and add the necessary netif_start_queue() call
to atl1_up().

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

Merge branch 'for-2.6.27' of git://linux-nfs.org/~bfields/linux

* 'for-2.6.27' of git://linux-nfs.org/~bfields/linux: (51 commits)
  nfsd: nfs4xdr.c do-while is not a compound statement
  nfsd: Use C99 initializers in fs/nfsd/nfs4xdr.c
  lockd: Pass "struct sockaddr *" to new failover-by-IP function
  lockd: get host reference in nlmsvc_create_block() instead of callers
  lockd: minor svclock.c style fixes
  lockd: eliminate duplicate nlmsvc_lookup_host call from nlmsvc_lock
  lockd: eliminate duplicate nlmsvc_lookup_host call from nlmsvc_testlock
  lockd: nlm_release_host() checks for NULL, caller needn't
  file lock: reorder struct file_lock to save space on 64 bit builds
  nfsd: take file and mnt write in nfs4_upgrade_open
  nfsd: document open share bit tracking
  nfsd: tabulate nfs4 xdr encoding functions
  nfsd: dprint operation names
  svcrdma: Change WR context get/put to use the kmem cache
  svcrdma: Create a kmem cache for the WR contexts
  svcrdma: Add flush_scheduled_work to module exit function
  svcrdma: Limit ORD based on client's advertised IRD
  svcrdma: Remove unused wait q from svcrdma_xprt structure
  svcrdma: Remove unneeded spin locks from __svc_rdma_free
  svcrdma: Add dma map count and WARN_ON
  ...

Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  gpu: re-order GPU subdirectory vs char for AGP vs DRM startup.

ipv6: Fix warning in addrconf code.

Reported by Linus.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  pkt_sched: Fix build with NET_SCHED disabled.