qeth: rework TSO functions

The maximum TSO size OSA can handle is 15 * PAGE_SIZE. This
patch reduces gso_max_size to this value and adds some sanity
checks and statistics to the TSO implementation.
Since only layer 3 is able to do TSO move all TSO related functions
to the qeth_l3 module.

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

qeth: Recognize return codes of ccw_device_set_online

Setting a qeth device online requires to call function
ccw_device_set_online() for read-, write-, and data-subchannel.
Failures should be detected immediately without an attempt to
invoke follow-on activity qeth_qdio_clear_card().,

In addition, ccw_device_set_online calls are consolidated in
qeth_core_main.c only.

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

qeth: remaining EDDP cleanup

EDDP code has been removed from qeth in 2009. This patch removes two
useless remaining EDDP-references.

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

qeth: Exploit Connection Isolation

Isolate data connection to a shared OSA card against other data
connections to the same OSA card. Connectivity between isolated
data connections sharing the same OSA card is therefore possible only
through external network gear (e.g. a router).

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

remove deprecated and not used: print_mac()

The function print_mac in net/ethernet/eth.c is marked __deprecated
and not used. Remove it.

Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: fix 802.3ad standards compliance error

The language of 802.3ad 43.4.9 requires the "recordPDU" function
to, in part, compare the Partner parameter values in a received LACPDU
to the stored Actor values.  If those match, then the Partner's
synchronization state is set to true.

The current 802.3ad implementation is performing these steps out
of order; first, the synchronization check is done, then the paramters are
checked to see if they match (the synch check being done against a match
check of a prior LACPDU).  This causes delays in establishing aggregators
in some circumstances.

This patch modifies the 802.3ad code to call __choose_matched,
the function that does the "match" comparisions, as the first step of
__record_pdu, instead of immediately afterwards.  This new behavior is
in compliance with the language of the standard.

Some additional commentary relating to code vs. standard is also
added.

Reported by Martin Patterson <martin@gear6.com> who also supplied
the logic of the fix and verified the patch.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vlan: Use __vlan_hwaccel_put_tag() in rx

Commit 05423b241311c9380 (vlan: allow null VLAN ID to be used)
forgot to update __vlan_hwaccel_rx() & vlan_gro_common()

We need to set VLAN_TAG_PRESENT flag in skb->vlan_tci

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

tg3: Update version to 3.104

This patch updates the tg3 version to 3.104.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Fix DIDs, Enable 5717 support

This patch fixes the 5717 variant device ID enumerations and adds those
DIDs to the PCI ID table.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add rx prod ring consolidation

This patch adds code to funnel each MSI-X vector's rx packet buffers
into a single set of producer rings which will then be submitted to the
hardware.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Create aliases for rx producer mailbox regs

The rx producer mailbox registers are used in several spots in the code.
The addition of TG3_64BIT_REG_LOW makes register references
uncomfortably long.  This patch creates an alias for the standard and
jumbo ring producer index registers to make the code cleaner.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Lay proucer ring handling groundwork

The patch increases the number of producer rings available and
implements the constructor and destructor code that deals with them.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Consider rx_std_prod_idx a hw mailbox

This patch changes how the code uses the rx_std_prod_idx member.  In the
following patch, the code will be changed so that it will act just like
a hardware mailbox.  This patch prepares the code so that memory barriers
can be more easily inserted.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: rename rx_[std|jmb]_ptr

A later patch is going to add consumer indicies for the producer rings.
To keep things readable, this patch renames rx_[std|jmb]_ptr to
rx_[std|jmb]_prod_idx.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: tg3_alloc_rx_skb(tnapi => tp)

This patch converts the tnapi argument of tg3_alloc_rx_skb() to tp.  The
level of indirection is unnecessary.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add prodring parameter to tg3_alloc_rx_skb()

This patch changes the tg3_alloc_rx_skb() implementation to accept the
destination producer ring set pointer as a parameter rather than
assuming the source and destination producer rings are the same.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Make tg3_alloc_rx_skb() a dst-only operation

This patch removes the source index parameter of tg3_alloc_rx_skb().  A
later patch will make it possible for the source and destination
producer rings to be different.  This patch opts to make
tg3_alloc_rx_skb() a destination-only implementation and move the code
sensitive to the difference elsewhere.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Move napi_add calls below tg3_get_invariants

tg3_get_invariants(), among other things, discovers whether or not
the device is MSI-X capable and how many interrupts it supports.
This discovery needs to happen before registering NAPI instances with
netdev.  This patch moves the code block that calls napi_add later in
tg3_init_one() so that tg3_get_invariants() has a chance to run first.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Create tg3_poll_msix() for non-zero MSIX vecs

This patch gives all non-zero MSIX vectors their own NAPI handler.  This
will make NAPI handling for those vectors slightly more efficient.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Allow DMAs to cross cacheline boundaries

By default, the 5717 (and future chips) break up PCIe DMA packets across
cacheline boundaries.  This isn't necessary on x86.  This patch
selectively loosens the restriction.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Use tg3_start_xmit_dma_bug for 5717 A0

The A0 revision of the 5717 has problems with short packet fragments.
It needs to use the tg3_start_xmit_dma_bug() routine.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add new HW_TSO_3 flag for 5717

The 5717 sets up TSO slightly differently in the transmit path.  It
looks like this method will be the new way of doing things.  This patch
defines a flag to indicate this.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Refine TSO and MSI discovery

This patch consolidates the TSO capability discovery code into its own
code block.  The code that decides whether or not to allow TSO is then
cleaned up.  Finally, the patch consolidates all MSI and MSIX
capability code into a single code block.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Move TG3_FLG2_PROTECTED_NVRAM to tg3_flags3

We need room for another TSO flag and it would be most efficient if it
resided in tg3_flags2.  This patch moves the TG3_FLG2_PROTECTED_NVRAM
to tg3_flags3 to make room.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Napify tg3_start_xmit_dma_bug()

This patch converts tg3_start_xmit_dma_bug() to accomodate multiple NAPI
instances.  This is prep work for a later patch in this series.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Don't touch RCB nic addresses

This patch avoids reprogramming the RCB NIC addresses for all 5755 and
later devices.  The address is incorrect for 5717 devices and should be
correct by default for all other affected devices.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add 5717 phy ID

This patch adds the 5717 phy ID.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Optimize hard_start_xmit() return checking

Recent changes in the TX error propagation require additional checking
and masking of values returned from hard_start_xmit(), mainly to
separate cases where skb was consumed. This aim can be simplified by
changing the order of NETDEV_TX and NET_XMIT codes, because the latter
are treated similarly to negative (ERRNO) values.

After this change much simpler dev_xmit_complete() is also used in
sch_direct_xmit(), so it is moved to netdevice.h.

Additionally NET_RX definitions in netdevice.h are moved up from
between TX codes to avoid confusion while reading the TX comment.

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

iwmc3200top: simplify the driver version

drop the version parts not needed for in-tree driver

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iwmc3200top: use prefered style for the device table.

Use device id number directly accompany with
comment rather then a #define

Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: check the return value of ndo_select_queue()

Check the return value of ndo_select_queue(). If the value isn't smaller
than the real_num_tx_queues, print a warning message, and reset it to zero.

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

netxen: update MAINTAINERS

Changing MAINTAINERS for netxen nic driver.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

hamradio/mkiss: fix typo in compat_ioctl

My last commit introduced an typo causing the
compat_ioctl function to do nothing useful.
The obvious way for an ioctl function to work
is to look at the command, not the argument first.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-next' of git://git./linux/kernel/git/lowpan/lowpan

alpha: Fixup recvmmsg syscall glue

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/can: add driver for mscan family & mpc52xx_mscan

Taken from socketcan-svn, fixed remaining todos, cleaned up, tested with a
phyCORE-MPC5200B-IO and a custom board.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Cc: Grant Likely <grant.likely@secretlab.ca>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Phonet: convert routing table to RCU

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Phonet: put protocols array under RCU

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netiucv: displayed TX bytes value much too high

tx_bytes value must be updated by skb length before skb is freed.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s390: remove cu3088 layer for lcs and ctcm

The cu3088-driver used as common base for lcs- and ctcm-devices
makes it difficult to assign the appropriate driver to an lcs-device
or a ctcm-device. This patch eliminates the cu3088-driver and thus
the root device "cu3088". Path /sys/devices/cu3088 is replaced with
the pathes /sys/devices/lcs and /sys/devices/ctcm.

Patch is based on a proposal from Cornelia Huck.

Cc: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ctcm: suspend has to wait for outstanding I/O

State transition to DEV_STATE_STOPPED indicates all outstanding I/O has
finished. Add wait queue to wait for this state.

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

iucv: add work_queue cleanup for suspend

If iucv_work_queue is not empty during kernel freeze, a kernel panic
occurs. This suspend-patch adds flushing of the work queue for
pending connection requests and severing of remaining pending
connections.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

inetpeer: Optimize inet_getid()

While investigating for network latencies, I found inet_getid() was a
contention point for some workloads, as inet_peer_idlock is shared
by all inet_getid() users regardless of peers.

One way to fix this is to make ip_id_count an atomic_t instead
of __u16, and use atomic_add_return().

In order to keep sizeof(struct inet_peer) = 64 on 64bit arches
tcp_ts_stamp is also converted to __u32 instead of "unsigned long".

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

ipv6: speedup inet6_dump_addr()

When handling large number of netdevices, inet6_dump_addr()
is very slow because it has O(N^2) complexity.

Instead of scanning one single list, we can use the NETDEV_HASHENTRIES
sub lists of the dev_index hash table, and RCU lookups.

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

inet: fix inet_bind_bucket_for_each

The first "node" is supposed to be the cursor used in the for_each.

The second "node" is ment literally and should not be macro expanded:
it's the name of the hlist_node field from the inet_bind_bucket.

This currently works because when inet_bind_bucket_for_each is called
it's argument is still "node".

Signed-off-by: Lucian Adrian Grijincu <lgrijincu@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: speedup inet_dump_ifaddr()

Stephen Hemminger a écrit :
> On Thu, 12 Nov 2009 15:11:36 +0100
> Eric Dumazet <eric.dumazet@gmail.com> wrote:
>
>> When handling large number of netdevices, inet_dump_ifaddr()
>> is very slow because it has O(N^2) complexity.
>>
>> Instead of scanning one single list, we can use the NETDEV_HASHENTRIES
>> sub lists of the dev_index hash table, and RCU lookups.
>>
>> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
>
> You might be able to make RCU critical section smaller by moving
> it into loop.
>

Indeed. But we dump at most one skb (<= 8192 bytes ?), so rcu_read_lock
holding time is small, unless we meet many netdevices without
addresses. I wonder if its really common...

Thanks

[PATCH net-next-2.6] ipv4: speedup inet_dump_ifaddr()

When handling large number of netdevices, inet_dump_ifaddr()
is very slow because it has O(N2) complexity.

Instead of scanning one single list, we can use the NETDEV_HASHENTRIES
sub lists of the dev_index hash table, and RCU lookups.

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

ixgbe: Make queue pairs on single MSI-X interrupts

This patch pairs similar-numbered Rx and Tx queues onto a single
MSI-X vector.  For example, Tx queue 0 and Rx queue 0's interrupt
with be ethX-RxTx-0.  This allows for more efficient cleanup, since
fewer interrupts will be firing during device operation.  It also
helps with a cleaner CPU affinity for IRQ affinity.

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

ixgbe: Flush the LSC mask change to prevent repeated interrupts

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

igb: only recycle page if it is on our numa node

This patch makes it so that we only recycle pages when they are from the
local NUMA node.  Non-local pages are freed and replaced with locally
allocated pages.

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

igb: check for packets on all tx rings when link is down

We were previously only checking the first tx ring to see if it had any
packets in it when the link when down.  However we should be checking all
of the rings so this patch makes it so that all of the rings are now being
checked.

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

igb: removed unused tx/rx total bytes/packets from adapter struct

This patch removes unused variables total_tx_bytes, total_tx_packets,
total_rx_bytes, and total_rx_packets from the adapter struct.

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

igb: Rework how netdev->stats is handled

This patch does some refactoring work that I felt was needed after reviewing
the changes recently submitted relating to the replacement of net_stats with
netdev->stats.

This patch essentially creates two different collections of stats.  The
first handles the adapter specific states and is stored in gstring_stats,
and the second is for netdev specific stats and is stored in
gstring_net_stats.

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

igb: when number of CPUs > 4 combine tx/rx queues to allow more queues

This patch makes it so that nics such as 82576 and newer can support more
hardware queues when there are more than 4 cpus by combining a tx/rx queue
pair onto one interrupt so that 8 queue pairs can be supported and thus
allow for more queues.

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

igb: move timesync init into a seperate function

Current code is quite large and making igb_probe difficult to read.

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

igb: change type for ring sizes to u16 in igb_set_ring_param

Change the type for the ring size values to u16 and use min/max_t instead of
min/max.

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

ARM: fix bug of checking on signed return value using unsigned statement in w90p910 platform

To fix the bug of checking on signed return value using unsigned statement.
Thanks Roel Kluin for digging out it.

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

igmp: Use next_net_device_rcu()

We need to use next_det_device_rcu() in RCU protected section.

We also can avoid in_dev_get()/in_dev_put() overhead (code size mainly)
in rcu_read_lock() sections.

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

ipv6: use RCU to walk list of network devices

No longer need read_lock(&dev_base_lock), use RCU instead.
We also can avoid taking references on inet6_dev structs.

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

net: TCP_MSS_DEFAULT, TCP_MSS_DESIRED

Define two symbols needed in both kernel and user space.

Remove old (somewhat incorrect) kernel variant that wasn't used in
most cases.  Default should apply to both RMSS and SMSS (RFC2581).

Replace numeric constants with defined symbols.

Stand-alone patch, originally developed for TCPCT.

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

vlan/macvlan: propagate transmission state to upper layers

Both vlan and macvlan devices usually don't use a qdisc and immediately
queue packets to the underlying device. Propagate transmission state of
the underlying device to the upper layers so they can react on congestion
and/or inform the sending process.

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

net: allow to propagate errors through ->ndo_hard_start_xmit()

Currently the ->ndo_hard_start_xmit() callbacks are only permitted to return
one of the NETDEV_TX codes. This prevents any kind of error propagation for
virtual devices, like queue congestion of the underlying device in case of
layered devices, or unreachability in case of tunnels.

This patches changes the NET_XMIT codes to avoid clashes with the NETDEV_TX
codes and changes the two callers of dev_hard_start_xmit() to expect either
errno codes, NET_XMIT codes or NETDEV_TX codes as return value.

In case of qdisc_restart(), all non NETDEV_TX codes are mapped to NETDEV_TX_OK
since no error propagation is possible when using qdiscs. In case of
dev_queue_xmit(), the error is propagated upwards.

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

niu.c: Use correct length in strncmp

Untested, no hardware

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ieee802154: make wpan-phy class registration to subsys_initcall

Move ieee802154 initialisation to subsys_initcall call, so that
wpan-phy class is initialised before all devices (thus saving us from
oops during bootup).

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>

net/atm: move all compat_ioctl handling to atm/ioctl.c

We have two implementations of the compat_ioctl handling for ATM, the
one that we have had for ages in fs/compat_ioctl.c and the one added to
net/atm/ioctl.c by David Woodhouse. Unfortunately, both versions are
incomplete, and in practice we use a very confusing combination of the
two.

For ioctl numbers that have the same identifier on 32 and 64 bit systems,
we go directly through the compat_ioctl socket operation, for those that

differ, we do a conversion in fs/compat_ioctl.c.

This patch moves both variants into the vcc_compat_ioctl() function,
while preserving the current behaviour. It also kills off the COMPATIBLE_IOCTL
definitions that we never use here.
Doing it this way is clearly not a good solution, but I hope it is a
step into the right direction, so that someone is able to clean up this
mess for real.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/compat: fix dev_ifsioc emulation corner cases

Handling for SIOCSHWTSTAMP is broken on architectures
with a split user/kernel address space like s390,
because it passes a real user pointer while using
set_fs(KERNEL_DS).
A similar problem might arise the next time somebody
adds code to dev_ifsioc.

Split up dev_ifsioc into three separate functions for
SIOCSHWTSTAMP, SIOC*IFMAP and all other numbers so
we can get rid of set_fs in all potentially affected
cases.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Patrick Ohly <patrick.ohly@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

DM9000: Wake on LAN support

Add support for Wake on LAN (WOL) reception and waking the device up from
this signal via the ethtool interface. Currently we are only supporting
the magic-packet variant of wakeup.

WOL is enabled by specifying a second interrupt resource to the driver
which indicates where the interrupt for the WOL is being signalled. This
then enables the necessary ethtool calls to leave the device in a state
to receive WOL frames when going into suspend.

Signed-off-by: Ben Dooks <ben@simtec.co.uk>
Signed-off-by: Simtec Linux Team <linux@simtec.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: r_idx not used in ixgbe_msix_clean_rx()

The values of r_idx and rx_ring are not used after the last time they
are set in ixgbe_msix_clean_rx(), so they can be removed.

Signed-off-by: Ali Gholami Rudi <ali@rudi.ir>
Acked-by: Peter P Waskiewicz Jr <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

decnet: convert dndev_lock to spinlock

There is no reason for this lock to be reader/writer since
the reader only has lock held for a very brief period.
The overhead of read_lock is more expensive than spinlock.

Compile tested only, I am not a decnet user.

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

decnet: add RTNL lock when reading address list

Add missing locking in the case of auto binding to the
default device. The address list might change while this code is looking
at the list.

Compile tested only, I am not a decnet user.

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

netdev: fold name hash properly (v3)

The full_name_hash function does not produce well distributed values in
the lower bits, so most code uses hash_32() to fold it.  This is really
a bug introduced when name hashing was added, back in 2.5 when I added
name hashing.

hash_32 is all that is needed since full_name_hash returns unsigned int
which is only 32 bits on 64 bit platforms.

Also, there is no point in using hash_32 on ifindex, because the is naturally
sequential and usually well distributed.

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

qlge: Change version to v1.00.00.23.00.00-01.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlge: Clean up module parameter name.

Change it to match qlge_xxx convention.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlge: Add asic reset to open call.

Force asic to known state at open().

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlge: Do not change frame routing during suspend.

We do not need to change the frame routing to direct all frames to the
management fifo during suspend.  This is now done by the firmware.

Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

clocksource/timecompare: Fix symbol exports to be GPL'd.

Noticed by Thomas GLeixner.

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

gianfar: Revive SKB recycling

Before calling gfar_clean_tx_ring() the driver grabs an irqsave
spinlock, and then tries to recycle skbs. But since
skb_recycle_check() returns 0 with IRQs disabled, we'll never
recycle any skbs.

It appears that gfar_clean_tx_ring() and gfar_start_xmit() are
mostly idependent and can work in parallel, except when they
modify num_txbdfree.

So we can drop the lock from most sections and thus fix the skb
recycling.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: Fix race between gfar_error() and gfar_start_xmit()

gfar_error() can arrive at the middle of gfar_start_xmit() processing,
and so it can trigger transfers of BDs that we don't yet expect to
be transmitted.

Fix this by locking the tx queues in gfar_error().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: Fix thinko in gfar_set_rx_stash_index()

We obviously want to write a modified 'temp' value back to the
register, not the saved IRQ flags.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: Fix build with CONFIG_PM=y

commit fba4ed030cfae7efdb6b79a57b0c5a9d72c9 ("gianfar: Add Multiple
Queue Support") introduced the following build failure:

  CC      gianfar.o
gianfar.c: In function 'gfar_restore':
gianfar.c:1249: error: request for member 'napi' in something not a structure or union

This patch fixes the issue.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

gianfar: Remove 'Interrupt problem!' warning

It is OK to poll with disabled IRQs, so remove the warning.

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

skbuff: Do not allow skb recycling with disabled IRQs

NAPI drivers try to recycle SKBs in their polling routine, but we
generally don't know the context in which the polling will be called,
and the skb recycling itself may require IRQs to be enabled.

This patch adds irqs_disabled() test to the skb_recycle_check()
routine, so that we'll not let the drivers hit the skb recycling
path with IRQs disabled.

As a side effect, this patch actually disables skb recycling for some
[broken] drivers. E.g. gianfar driver grabs an irqsave spinlock during
TX ring processing, and then tries to recycle an skb, and that caused
the following badness:

nf_conntrack version 0.5.0 (1008 buckets, 4032 max)
------------[ cut here ]------------
Badness at kernel/softirq.c:143
NIP: c003e3c4 LR: c423a528 CTR: c003e344
...
NIP [c003e3c4] local_bh_enable+0x80/0xc4
LR [c423a528] destroy_conntrack+0xd4/0x13c [nf_conntrack]
Call Trace:
[c15d1b60] [c003e32c] local_bh_disable+0x1c/0x34 (unreliable)
[c15d1b70] [c423a528] destroy_conntrack+0xd4/0x13c [nf_conntrack]
[c15d1b80] [c02c6370] nf_conntrack_destroy+0x3c/0x70

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

ipv6: Remove unused var in inet6_dump_ifinfo()

Reported by Stephen Rothwell:

--------------------
Today's linux-next build (x86_64 allmodconfig) produced this warning:

net/ipv6/addrconf.c: In function 'inet6_dump_ifinfo':
net/ipv6/addrconf.c:3833: warning: unused variable 'err'

Introduced by commit 84d2697d9649339215675551eae28ba04068dea1 ("ipv6:
speedup inet6_dump_ifinfo()").
--------------------

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

iwlwifi: fix iwl1000 "RTS/CTS for HT" merge damage

I may have botched my merge conflict resolution instructions for Dave...

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Conflicts:
drivers/net/wireless/iwlwifi/iwl-1000.c
drivers/net/wireless/iwlwifi/iwl-6000.c
drivers/net/wireless/iwlwifi/iwl-core.h

CAN: use dev_get_by_index_rcu

Use new function to avoid doing read_lock().

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPV4: use rcu to walk list of devices in IGMP

This also needs to be optimized for large number of devices.

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

decnet: use RCU to find network devices

When showing device statistics use RCU rather than read_lock(&dev_base_lock)
Compile tested only.

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

s390: use RCU to walk list of network devices

This is similar to other cases where for_each_netdev_rcu
can be used when gathering information.

By inspection, don't have platform or cross-build environment
to validate.

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

net: use rcu for network scheduler API

Use RCU to walk list of network devices in qdisc dump.
This could be optimized for large number of devices.

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

vlan: eliminate use of dev_base_lock

Do not need to use read_lock(&dev_base_lock), use RCU instead.

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

netdev: add netdev_continue_rcu

This adds an RCU macro for continuing search, useful for some
network devices like vlan.

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

IPv6: use ipv6_addr_v4mapped()

Change udp6_portaddr_hash() to use ipv6_addr_v4mapped()
inline instead of ipv6_addr_type().

Signed-off-by: Brian Haley <brian.haley@hp.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

parisc: led: Use for_each_netdev_rcu()

Use for_each_netdev_rcu() and dont lock dev_base_lock anymore

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

sit: Clean up DF code by copying from IPIP

This patch rearranges the SIT DF bit handling using the new IPIP DF
code.  The only externally visible effect should be the case where
PMTU is enabled and the MTU is exactly 1280 bytes.  In this case the
previous code would send packets out with DF off while the new code
would set the DF bit.  This is inline with RFC 4213.

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

ipv6: Allow inet6_dump_addr() to handle more than 64 addresses

Apparently, inet6_dump_addr() is not able to handle more than
64 ipv6 addresses per device. We must break from inner loops
in case skb is full, or else cursor is put at the end of list.

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

ipv6: speedup inet6_dump_ifinfo()

When handling large number of netdevice, inet6_dump_ifinfo()
is very slow because it has O(N^2) complexity.

Instead of scanning one single list, we can use the 256 sub lists
of the dev_index hash table, and RCU lookups.

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

net: netlink_getname, packet_getname -- use DECLARE_SOCKADDR guard

Use guard DECLARE_SOCKADDR in a few more places which allow
us to catch if the structure copied back is too big.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

usbnet: Set link down initially for drivers that update link state

Some usbnet drivers update link state while others do not due to
hardware limitations.  Add a flag to distinguish those that do, and
set the link down initially for their devices.

This is intended to fix this bug: http://bugs.debian.org/444043

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

niu: Use DMA_BIT_MASK(44) instead of deprecated DMA_44BIT_MASK

Use DMA_BIT_MASK(44) instead of deprecated DMA_44BIT_MASK

Signed-off-by: Marin Mitov <mitov@issp.bas.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>

udp: bind() optimisation

UDP bind() can be O(N^2) in some pathological cases.

Thanks to secondary hash tables, we can make it O(N)

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