TCP: increase default initial receive window.

This patch changes the default initial receive window to 10 mss
(defined constant). The default window is limited to the maximum
of 10*1460 and 2*mss (when mss > 1460).

draft-ietf-tcpm-initcwnd-00 is a proposal to the IETF that recommends
increasing TCP's initial congestion window to 10 mss or about 15KB.
Leading up to this proposal were several large-scale live Internet
experiments with an initial congestion window of 10 mss (IW10), where
we showed that the average latency of HTTP responses improved by
approximately 10%. This was accompanied by a slight increase in
retransmission rate (0.5%), most of which is coming from applications
opening multiple simultaneous connections. To understand the extreme
worst case scenarios, and fairness issues (IW10 versus IW3), we further
conducted controlled testbed experiments. We came away finding minimal
negative impact even under low link bandwidths (dial-ups) and small
buffers.  These results are extremely encouraging to adopting IW10.

However, an initial congestion window of 10 mss is useless unless a TCP
receiver advertises an initial receive window of at least 10 mss.
Fortunately, in the large-scale Internet experiments we found that most
widely used operating systems advertised large initial receive windows
of 64KB, allowing us to experiment with a wide range of initial
congestion windows. Linux systems were among the few exceptions that
advertised a small receive window of 6KB. The purpose of this patch is
to fix this shortcoming.

References:
1. A comprehensive list of all IW10 references to date.
http://code.google.com/speed/protocols/tcpm-IW10.html

2. Paper describing results from large-scale Internet experiments with IW10.
http://ccr.sigcomm.org/drupal/?q=node/621

3. Controlled testbed experiments under worst case scenarios and a
fairness study.
http://www.ietf.org/proceedings/79/slides/tcpm-0.pdf

4. Raw test data from testbed experiments (Linux senders/receivers)
with initial congestion and receive windows of both 10 mss.
http://research.csc.ncsu.edu/netsrv/?q=content/iw10

5. Internet-Draft. Increasing TCP's Initial Window.
https://datatracker.ietf.org/doc/draft-ietf-tcpm-initcwnd/

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

net_sched: sch_sfq: better struct layouts

Here is a respin of patch.

I'll send a short patch to make SFQ more fair in presence of large
packets as well.

Thanks

[PATCH v3 net-next-2.6] net_sched: sch_sfq: better struct layouts

This patch shrinks sizeof(struct sfq_sched_data)
from 0x14f8 (or more if spinlocks are bigger) to 0x1180 bytes, and
reduce text size as well.

   text    data     bss     dec     hex filename
   4821     152       0    4973    136d old/net/sched/sch_sfq.o
   4627     136       0    4763    129b new/net/sched/sch_sfq.o

All data for a slot/flow is now grouped in a compact and cache friendly
structure, instead of being spreaded in many different points.

struct sfq_slot {
        struct sk_buff  *skblist_next;
        struct sk_buff  *skblist_prev;
        sfq_index       qlen; /* number of skbs in skblist */
        sfq_index       next; /* next slot in sfq chain */
        struct sfq_head dep; /* anchor in dep[] chains */
        unsigned short  hash; /* hash value (index in ht[]) */
        short           allot; /* credit for this slot */
};

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Jarek Poplawski <jarkao2@gmail.com>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

net_sched: sch_sfq: fix allot handling

When deploying SFQ/IFB here at work, I found the allot management was
pretty wrong in sfq, even changing allot from short to int...

We should init allot for each new flow, not using a previous value found
in slot.

Before patch, I saw bursts of several packets per flow, apparently
denying the default "quantum 1514" limit I had on my SFQ class.

class sfq 11:1 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 7p requeues 0
 allot 11546

class sfq 11:46 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 1p requeues 0
 allot -23873

class sfq 11:78 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 5p requeues 0
 allot 11393

After patch, better fairness among each flow, allot limit being
respected, allot is positive :

class sfq 11:e parent 11:
 (dropped 0, overlimits 0 requeues 86)
 backlog 0b 3p requeues 86
 allot 596

class sfq 11:94 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 3p requeues 0
 allot 1468

class sfq 11:a4 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 4p requeues 0
 allot 650

class sfq 11:bb parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 3p requeues 0
 allot 596

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

net_sched: sch_sfq: add backlog info in sfq_dump_class_stats()

We currently return for each active SFQ slot the number of packets in
queue. We can also give number of bytes accounted for these packets.

tc -s class show dev ifb0

Before patch :

class sfq 11:3d9 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 0b 3p requeues 0
 allot 1266

After patch :

class sfq 11:3e4 parent 11:
 (dropped 0, overlimits 0 requeues 0)
 backlog 4380b 3p requeues 0
 allot 1212

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

vmxnet3: locking problems in xmit

There were several paths that didn't release their locks.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

typhoon: memory corruption in typhoon_get_drvinfo()

info->version only has space for 32 characters but my UTS_RELEASE is
"2.6.37-rc6-next-20101217-05817-ge935fc8-dirty" so it doesn't fit.
This is supposed to be the version of the driver, not the kernel
version.  This driver doesn't have a version so lets just leave it
blank.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Add USB PID for new MOSCHIP USB ethernet controller MCS7832 variant

Due to active notification of the new MCS7832 version by the manufacturer
(Mr. Milton; thanks!) -- quote: "functionality same as MCS7830",
I'm now submitting this patch (on -rc6), intended for networking.git and -stable.

- add MCS7832 USB PID to be able to support this new device variant, too
- add related descriptions

Signed-off-by: Andreas Mohr <andi@lisas.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Flush per-ns routing cache more sanely.

Flush the routing cache only of entries that match the
network namespace in which the purge event occurred.

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

ehea: Fixing some message level

Currently there are some info message that is set as error, and an error
message that is set as debug. This patch just fixes it.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

batman-adv: Return hna count on local buffer fill

hna_local_fill_buffer must return the number of added hna entries and
not the last checked hash bucket.

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

pktgen: Remove unnecessary prefix from pr_<level>

Remove "pktgen: " prefix string from one pr_info.
pr_fmt adds it, so this is a duplicate.

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

net_sched: always clone skbs

Pawel reported a panic related to handling shared skbs in ixgbe
incorrectly. So we need to revert my previous patch to work around
this bug. Instead of reverting the patch completely, I just revert
the essential lines, so we can add the previous optimization
back more easily in future.

    commit 3511c9132f8b1e1b5634e41a3331c44b0c13be70
    Author: Changli Gao <xiaosuo@gmail.com>
    Date:   Sat Oct 16 13:04:08 2010 +0000

        net_sched: remove the unused parameter of qdisc_create_dflt()

Reported-by: Pawel Staszewski <pstaszewski@itcare.pl>
Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: kill unused macros

These macros never be used, so remove them.

Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxge: add missing flush of reset_task

Commit 6e07ebd84 (drivers/net: remove unnecessary
flush_scheduled_work() calls) incorrectly removed the flush call
without replacing it with the appropriate work specific operation.
Fix it by flushing vdev->reset_task explicitly.

Pointed out by Jon Mason.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jon Mason <jon.mason@exar.com>
Acked-by: Jon Mason <jon.mason@exar.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: increase skb->users instead of skb_clone()

In dev_queue_xmit_nit(), we have to clone skbs as we need to mangle skbs,
however, we don't need to clone skbs for all the packet_types.

Except for the first packet_type, we increase skb->users instead of
skb_clone().

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

ifb: use netif_receive_skb() instead of netif_rx()

In ri_tasklet(), we run from softirq, so can directly handle packet
through netif_receive_skb() instead of netif_rx().
There is no risk of recursion.

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

ipv6: Fragment locally generated tunnel-mode IPSec6 packets as needed.

This patch modifies IPsec6 to fragment IPv6 packets that are
locally generated as needed.

This version of the patch only fragments in tunnel mode, so that fragment
headers will not be obscured by ESP in transport mode.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: remove duplicate neigh_ifdown

When device is being set to down, neigh_ifdown was being called
twice. Once from addrconf notifier and once from ndisc notifier.

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

ipv6: fib6_ifdown cleanup

Remove (unnecessary) casts to make code cleaner.

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

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

Conflicts:
drivers/net/bnx2x/bnx2x.h
drivers/net/wireless/iwlwifi/iwl-1000.c
drivers/net/wireless/iwlwifi/iwl-6000.c
drivers/net/wireless/iwlwifi/iwl-core.h
drivers/vhost/vhost.c

netlink: fix gcc -Wconversion compilation warning

$ cat << EOF | gcc -Wconversion -xc -S -o/dev/null -
unsigned f(void) {return NLMSG_HDRLEN;}
EOF
<stdin>: In function 'f':
<stdin>:3:26: warning: negative integer implicitly converted to unsigned type

Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: David S. Miller <davem@davemloft.net>

asix: add USB ID for Logitec LAN-GTJ U2A

Logitec LAN-GTJ U2A (http://www.pro.logitec.co.jp/pro/g/gLAN-GTJU2A/)
USB 2.0 10/10/1000 Ethernet adapter is based on ASIX AX88178 chipset.

This patch adds missing USB ID for the device.

Signed-off-by: Arnaud Ebalard <arno@natisbad.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: reset pci function unconditionally during probe

Some boot code drivers dont have cleanup routine, so pci function
remains in unknown state prior to driver load. So during driver load
issue FLR unconditionally.

Update driver version to 5.0.14.

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

qlcnic: fix ocm window register offset calculation

OCM window register offset was calculated incorrectly for
pci function greater than zero.

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

qlcnic: fix LED test when interface is down.

When interface is down, create temporary context to config LED.

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

tehuti: Firmware filename is tehuti/bdx.bin

My conversion of tehuti to use request_firmware() was confused about
the filename of the firmware blob.  Change the driver to match the
blob.

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

ipv6: don't flush routes when setting loopback down

When loopback device is being brought down, then keep the route table
entries because they are special. The entries in the local table for
linklocal routes and ::1 address should not be purged.

This is a sub optimal solution to the problem and should be replaced
by a better fix in future.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ifb: fix a lockdep splat

After recent ifb changes, we must use lockless __skb_dequeue() since
lock is not anymore initialized.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Jamal Hadi Salim <hadi@cyberus.ca>
Cc: Changli Gao <xiaosuo@gmail.com>
Acked-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: fix the return value of getting the sctp partial delivery point

Get the sctp partial delivery point using SCTP_PARTIAL_DELIVERY_POINT
socket option should return 0 if success, not -ENOTSUPP.

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

net: Fix drivers advertising HW_CSUM feature to use csum_start

Some drivers are using skb_transport_offset(skb) instead of skb->csum_start
for NETIF_F_HW_CSUM offload.  This does not matter now, but if someone
implements checksumming of encapsulated packets then this will break silently.

TSO output paths are left as they are, since they are for IP+TCP only
(might be worth converting though).

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Use skb_checksum_start_offset()

Replace skb->csum_start - skb_headroom(skb) with skb_checksum_start_offset().

Note for usb/smsc95xx: skb->data - skb->head == skb_headroom(skb).

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Introduce skb_checksum_start_offset()

Introduce skb_checksum_start_offset() to replace repetitive calculation.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/veth: Fix packet checksumming

We can't change ip_summed from CHECKSUM_PARTIAL to CHECKSUM_NONE
or CHECKSUM_UNNECESSARY because checksum in packet's headers is
not valid and will cause invalid checksum when frame is forwarded.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: fix IPv6 queries for bridge multicast snooping

This patch fixes a missing ntohs() for bridge IPv6 multicast snooping.

Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix nulls list corruptions in sk_prot_alloc

Special care is taken inside sk_port_alloc to avoid overwriting
skc_node/skc_nulls_node. We should also avoid overwriting
skc_bind_node/skc_portaddr_node.

The patch fixes the following crash:

 BUG: unable to handle kernel paging request at fffffffffffffff0
 IP: [<ffffffff812ec6dd>] udp4_lib_lookup2+0xad/0x370
 [<ffffffff812ecc22>] __udp4_lib_lookup+0x282/0x360
 [<ffffffff812ed63e>] __udp4_lib_rcv+0x31e/0x700
 [<ffffffff812bba45>] ? ip_local_deliver_finish+0x65/0x190
 [<ffffffff812bbbf8>] ? ip_local_deliver+0x88/0xa0
 [<ffffffff812eda35>] udp_rcv+0x15/0x20
 [<ffffffff812bba45>] ip_local_deliver_finish+0x65/0x190
 [<ffffffff812bbbf8>] ip_local_deliver+0x88/0xa0
 [<ffffffff812bb2cd>] ip_rcv_finish+0x32d/0x6f0
 [<ffffffff8128c14c>] ? netif_receive_skb+0x99c/0x11c0
 [<ffffffff812bb94b>] ip_rcv+0x2bb/0x350
 [<ffffffff8128c14c>] netif_receive_skb+0x99c/0x11c0

Signed-off-by: Leonard Crestez <lcrestez@ixiacom.com>
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: use mutex instead of spin lock for mbox_lock

Since the mbox polling uses the schedule_timeout, the mbox_lock should be
a mutex and not a spin lock.
The commit f25b03a replaced udelay() with schedule_timeout() but didn't
change mbox_lock to semaphore or mutex.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: relax tcp_paws_check()

Some windows versions have wrong RFC1323 implementations, with SYN and
SYNACKS messages containing zero tcp timestamps.

We relaxed in commit fc1ad92dfc4e363 the passive connection case
(Windows connects to a linux machine), but the reverse case (linux
connects to a Windows machine) has an analogue problem when tsvals from
windows machine are 'negative' (high order bit set) : PAWS triggers and
we drops incoming messages.

Fix this by making zero ts_recent value special, allowing frame to be
processed.

Based on a report and initial patch from Dmitiy Balakin

Bugzilla reference : https://bugzilla.kernel.org/show_bug.cgi?id=24842

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

net: factorize sync-rcu call in unregister_netdevice_many

Add dev_close_many and dev_deactivate_many to factorize another
sync-rcu operation on the netdevice unregister path.

$ modprobe dummy numdummies=10000
$ ip link set dev dummy* up
$ time rmmod dummy

Without the patch           With the patch

real    0m 24.63s           real    0m 5.15s
user    0m 0.00s            user    0m 0.00s
sys     0m 6.05s            sys     0m 5.14s

Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Add batman-adv meshing protocol

B.A.T.M.A.N. (better approach to mobile ad-hoc networking) is a routing
protocol for multi-hop ad-hoc mesh networks. The networks may be wired or
wireless. See http://www.open-mesh.org/ for more information and user space
tools.

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: use NUMA_NO_NODE instead of the magic number -1

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

bonding: add the debugfs interface to see RLB hash table

This patch provices the debugfs interface to see RLB hash table
like the following:

# cat /sys/kernel/debug/bonding/bond0/rlb_hash_table
SourceIP        DestinationIP   Destination MAC   DEV
10.124.196.205  10.124.196.205  ff:ff:ff:ff:ff:ff eth4
10.124.196.205  10.124.196.81   00:19:99:XX:XX:XX eth3
10.124.196.205  10.124.196.1    00:21:d8:XX:XX:XX eth0

This is helpful to check if the receive load balancing works as expected.

Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: migrate some macros from bond_alb.c to bond_alb.h

This patch simply migrates some macros from bond_alb.c to bond_alb.h.

Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: NUMA-aware Tx queue allocations

Allocate Tx queue memory on the node indicated by the new
netdev_queue_numa_node_read.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: extend VPD parsing

Current code parses the VPD RO section for keywords but makes static
assumptions about the location of the section.  Remove them and parse
the VPD to find it.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: add const to static arrays

Patch originally from Joe Perches, unmodified.

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

cxgb4: remove a bitmap

The driver keeps a bitmap of the netdevs it registered so it knows what to
unregister later.  Remove that and look at reg_state instead.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: remove the name field from the adapter structure

Remove a field the driver uses to keep track of the name of the first
netdev it manages to register.  Do this by changing the registration
loop to stop the first time it fails so the first registered device is
trivial to tell.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: correct formatting of MSI-X interrupt names

The last byte of the buffer for MSI-X names could not be used due to a
bogus -1.  Also do not explicitly clear the last byte, snprintf will do
the right thing.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: allocate more space for MSI-X interrupt names

Currently MSI-X names for netdevs with long names are truncated in
/proc/interrupts due to insufficient space.  Use IFNAMSIZ to size the
needed space.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: print port information after registering each netdev

Print information about each port when its netdev is registered instead
of looping separately over the ports at the end.  The bulk of this patch
is due to indentation change.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: distinguish between 1-lane KR/KX and 4-lane KR/KX/KX4 ports

And fix the supported flags ethtool reports for the two cases.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: set the number of queues before device registration

The number of queues is known early, move the calls to
netif_set_real_num_[rt]x_queues before register_netdev.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: do not read the clock frequency from VPD

No need to read the clock frequency from VPD, we already get it a bit
later from FW, after any potential adjustments.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: enable PCIe relaxed ordering

Enable relaxed ordering for descriptor reads and packet I/O.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: update version to 1.62.00-2

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: remove old FW files

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: replace FW to 6.2.5

Includes FCoE releated fixes in FW flows

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: add FW 6.2.5 files

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Add DCB/PFC support - link layer

Add appropriate HW DCB/PFC configuration

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: add DCB support

Adding DCB initialization and handling on 57712 FW/HW

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Shmulik Ravid-Rabinovitz <shmulikr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: add a select queue callback

This callback required to allow FCoE traffic to be
sent on separate priority queue from other L2 traffic,
which is managed by PFC in HW.

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Shmulik Ravid-Rabinovitz <shmulikr@broadcom.com>
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: add FCoE ring

Includes new driver structures and FW/HW configuration for FCoE ring

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Shmulik Ravid-Rabinovitz <shmulikr@broadcom.com>
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Take the distribution range definition out of skb_tx_hash()

Move the calcualation of the Tx hash for a given hash range into a separate
function and define the skb_tx_hash(), which calculates a Tx hash for a
[0; dev->real_num_tx_queues - 1] hash values range, using this
function (__skb_tx_hash()).

Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding/vlan: Fix mangled NAs on slaves without VLAN tag insertion

bond_na_send() attempts to insert a VLAN tag in between building and
sending packets of the respective formats.  If the slave does not
implement hardware VLAN tag insertion then vlan_put_tag() will mangle
the network-layer header because the Ethernet header is not present at
this point (unlike in bond_arp_send()).

Fix this by adding the tag out-of-line and relying on
dev_hard_start_xmit() to insert it inline if necessary.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: Change active slave quietly when bond is down

bond_change_active_slave() may be called when a slave is added, even
if the bond has not been brought up yet.  It may then attempt to send
packets, and further it may use mcast_work which is uninitialised
before the bond is brought up.  Add the necessary checks for
netif_running(bond->dev).

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

bonding/vlan: Remove redundant VLAN tag insertion logic

A bond may have a mixture of slave devices with and without hardware
VLAN tag insertion capability.  Therefore it always claims this
capability and performs software VLAN tag insertion if the slave does
not.

Since commit 7b9c60903714bf0a19d746b228864bad3497284e, this has
also been done by dev_hard_start_xmit().  The result is that VLAN-
tagged skbs are now double-tagged when transmitted through slave
devices without hardware VLAN tag insertion!

Remove the now-redundant logic from bond_dev_queue_xmit().

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

axnet_cs: move id (0x1bf, 0x2328) to axnet_cs

axnet_cs:
    Accton EN2328 or compatible (id: 0x01bf, 0x2328) uses Asix chip.
    So it works better with axnet_cs instead of pcnet_cs.

Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: delete expired route in ip6_pmtu_deliver

The first big packets sent to a "low-MTU" client correctly
triggers the creation of a temporary route containing the reduced MTU.

But after the temporary route has expired, new ICMP6 "packet too big"
will be sent, rt6_pmtu_discovery will find the previous EXPIRED route
check that its mtu isn't bigger then in icmp packet and do nothing
before the temporary route will not deleted by gc.

I make the simple experiment:
while :; do
    time ( dd if=/dev/zero bs=10K count=1 | ssh hostname dd of=/dev/null ) || break;
done

The "time" reports real 0m0.197s if a temporary route isn't expired, but
it reports real 0m52.837s (!!!!) immediately after a temporare route has
expired.

Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: Fix slave selection bug.

The returned slave is incorrect, if the net device under check is not
charged yet by the master.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

p54usb: New USB ID for Gemtek WUBI-100GW

This USB ID is for the WUBI-100GW 802.11g Wireless LAN USB Device that
uses p54usb.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Eduardo Costa <ecosta.tmp@gmail.com>
Cc: Stable <stable@kernel.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

net: Abstract default MTU metric calculation behind an accessor.

Like RTAX_ADVMSS, make the default calculation go through a dst_ops
method rather than caching the computation in the routing cache
entries.

Now dst metrics are pretty much left as-is when new entries are
created, thus optimizing metric sharing becomes a real possibility.

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

Merge branch 'vhost-net-next' of git://git./linux/kernel/git/mst/vhost

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

net: Abstract default ADVMSS behind an accessor.

Make all RTAX_ADVMSS metric accesses go through a new helper function,
dst_metric_advmss().

Leave the actual default metric as "zero" in the real metric slot,
and compute the actual default value dynamically via a new dst_ops
AF specific callback.

For stacked IPSEC routes, we use the advmss of the path which
preserves existing behavior.

Unlike ipv4/ipv6, DecNET ties the advmss to the mtu and thus updates
advmss on pmtu updates.  This inconsistency in advmss handling
results in more raw metric accesses than I wish we ended up with.

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

net: change ip_default_ttl documentation

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

pch_can: Replace netif_rx to netif_receive_skb

Since this driver is implemented as NAPI,
netif_receive_skb must be used not netif_rx.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Add setting TEC/REC statistics processing

Add setting TEC/REC statistics processing.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Optimize "if" condition in rx/tx processing

For reduce "if" condition, easy to read/understand the code,
optimize "if" condition in rx/tx processing.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix incorrect return processing

Fix incorrect return processing.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Move MSI processing to probe/remove processing

Currently, in case this driver is integrated as module, and when this
module is re-installed, no interrupts is to be occurred.
For the above issue, move MSI processing to open/release processing.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Comment optimization

Comment optimization

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix miss-setting status issue

Modify miss-setting status issue at suspend.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix bit timing calculation issue

Modify like use calculated value directly passed by CAN core module.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Delete unnecessary/redundant code

Delete unnecessary/redundant code

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix coding rule violation

Fix coding rule violation.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Replace netdev_dbg instead of dev_dbg partly

For easy to readable, use netdev_dbg instead of dev_dbg partly

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Change Copyright and module description

Currently, Copyright and module description are not formal.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Reduce register access

For improve tx/rx speed, reduce register access.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Change functions type

Currently, these two functions spec(returned value) is unnatural.
Thus, change the return value's spec

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Rename function/macro name

For easy to read/understand, Rename function/macro name.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix warnings

Currently, in case CONFIG_PM is disabled, compiler outputs warnings.
Move six functions which are used only CONFIG_PM is enabled,
into "#ifdef CONFIG_PM" area.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Improve rx processing

Replace complex "goto" to "do~while".
For easy to read/understand, it divides a rx function into some functions.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Fix endianness issue

there is endianness issue both Tx and Rx.
Currently, data is set like below.
Register:
MSB--LSB
x x D0 D1
x x D2 D3
x x D4 D5
x x D6 D7

But Data to be sent must be set like below.
Register:
MSB--LSB
x x D1 D0
x x D3 D2
x x D5 D4
x x D7 D6  (x means reserved area.)

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Divide poll function

To easy to read/understand, divide poll function into two sub-functions.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pch_can: Add flow control processing

Currently, there is no flow control processing.
Thus, Add flow control processing as
when there is no empty of tx buffer,
netif_stop_queue is called.
When there is empty buffer, netif_wake_queue is called.

Signed-off-by: Tomoya MORINAGA <tomoya-linux@dsn.okisemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

net: add limits to ip_default_ttl

ip_default_ttl should be between 1 and 255

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

mac80211: avoid calling ieee80211_work_work unconditionally

On suspend, there might be usb wireless drivers which wrongly trigger
the warning in ieee80211_work_work. If an usb driver doesn't have a
suspend hook, the usb stack will disconnect the device. On disconnect,
a mac80211 driver calls ieee80211_unregister_hw, which calls dev_close,
which calls ieee80211_stop, and in the end calls ieee80211_work_purge->
ieee80211_work_work.

The problem is that this call to ieee80211_work_purge comes after
mac80211 is suspended, triggering the warning even when we don't have
work queued in work_list (the expected case when already suspended),
because it always calls ieee80211_work_work.

So, just call ieee80211_work_work in ieee80211_work_purge if we really
have to abort work. This addresses the warning reported at
https://bugzilla.kernel.org/show_bug.cgi?id=24402

Signed-off-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Signed-off-by: John W. Linville <linville@tuxdriver.com>