Perf shows a relatively high rate (about 8%) race in
spin_lock_irqsave() when doing netperf between external host and
guest. It's mainly becuase the lock contention between the
tun_do_read() and tun_xmit_skb(), so this patch do not put self into
waitqueue to reduce this kind of race. After this patch, it drops to
4%.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Amos Kong <akong@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Current standard practice is to not mark most functions as inline
and  let compiler decide instead.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The tun driver allocates skb's to hold data from user and then passes
the data into the network stack as received data. Most network devices
allocate the receive skb with routines like dev_alloc_skb() that reserves
additional space for use by network protocol stack but tun does not.

Because of the lack of padding, when the packet is passed through bridge
netfilter a new skb has to be allocated.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Based on earlier patch from Neil Horman <nhorman@tuxdriver.com>

If iSCSI is not supported on a bnx2 device, bnx2_cnic_probe() will
return NULL and the cnic device will not be visible to bnx2i.  This
will prevent bnx2i from registering and then unregistering during
cnic_start() and cause the warning message:

bnx2 0003:01:00.1: eth1: Failed waiting for ULP up call to complete
Signed-off-by: Michael Chan <mchan@broadcom.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

During NETDEV_UP, we use symbol_get() to get the net driver's cnic
probe function.  This sometimes doesn't work if NETDEV_UP happens
right after NETDEV_REGISTER and the net driver is still running module
init code.  As a result, the cnic device may not be discovered.  We
fix this by probing on all NETDEV events if the device's netif_running
state is up.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This reduces the likelihood of port re-use when re-loading the driver.
Signed-off-by: Eddie Wai <eddie.wai@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

During iSCSI connection terminations, if the target is also terminating
at about the same time, the firmware may not complete the driver's
request to close or reset the connection.  This is fixed by handling
other events (instead of the expected completion event) as an indication
that the driver's request has been rejected.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We need to keep looping until cnic_get_kcqes() returns 0.  cnic_get_kcqes()
returns a maximum of 64 entries.  If there are more entries in the queue
and we don't loop back, the remaining entries may not be serviced for a
long time.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Profiles show false sharing in addr_compare() because refcnt/dtime
changes dirty the first inet_peer cache line, where are lying the keys
used at lookup time. If many cpus are calling inet_getpeer() and
inet_putpeer(), or need frag ids, addr_compare() is in 2nd position in
"perf top".

Before patch, my udpflood bench (16 threads) on my 2x4x2 machine :

             5784.00  9.7% csum_partial_copy_generic [kernel]
             3356.00  5.6% addr_compare              [kernel]
             2638.00  4.4% fib_table_lookup          [kernel]
             2625.00  4.4% ip_fragment               [kernel]
             1934.00  3.2% neigh_lookup              [kernel]
             1617.00  2.7% udp_sendmsg               [kernel]
             1608.00  2.7% __ip_route_output_key     [kernel]
             1480.00  2.5% __ip_append_data          [kernel]
             1396.00  2.3% kfree                     [kernel]
             1195.00  2.0% kmem_cache_free           [kernel]
             1157.00  1.9% inet_getpeer              [kernel]
             1121.00  1.9% neigh_resolve_output      [kernel]
             1012.00  1.7% dev_queue_xmit            [kernel]
# time ./udpflood.sh

real	0m44.511s
user	0m20.020s
sys	11m22.780s

# time ./udpflood.sh

real	0m44.099s
user	0m20.140s
sys	11m15.870s

After patch, no more addr_compare() in profiles :

             4171.00 10.7% csum_partial_copy_generic   [kernel]
             1787.00  4.6% fib_table_lookup            [kernel]
             1756.00  4.5% ip_fragment                 [kernel]
             1234.00  3.2% udp_sendmsg                 [kernel]
             1191.00  3.0% neigh_lookup                [kernel]
             1118.00  2.9% __ip_append_data            [kernel]
             1022.00  2.6% kfree                       [kernel]
              993.00  2.5% __ip_route_output_key       [kernel]
              841.00  2.2% neigh_resolve_output        [kernel]
              816.00  2.1% kmem_cache_free             [kernel]
              658.00  1.7% ia32_sysenter_target        [kernel]
              632.00  1.6% kmem_cache_alloc_node       [kernel]

# time ./udpflood.sh

real	0m41.587s
user	0m19.190s
sys	10m36.370s

# time ./udpflood.sh

real	0m41.486s
user	0m19.290s
sys	10m33.650s
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Driver was already keeping 64 bit counters, just not using the new interface.

Ps: IMHO drivers should not be duplicating network device
stats into ethtool stats. It is useless duplication.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The device driver already uses 64 bit statistics, it just
doesn't use the 64 bit interface.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Change to 64 bit statistics interface, driver was already maintaining 64 bit
value.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Not much change, device was already keeping per cpu statistics.
Use recent 64 statistics interface.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Convert vmxnet3 driver to 64 bit statistics interface.
This driver was already counting packet per queue in a 64 bit value so not
a huge change. Eliminate unused old net_device_stats structure.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Scott J. Goldman <scottjg@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Andi Kleen and Tim Chen reported huge contention on inetpeer
unused_peers.lock, on memcached workload on a 40 core machine, with
disabled route cache.

It appears we constantly flip peers refcnt between 0 and 1 values, and
we must insert/remove peers from unused_peers.list, holding a contended
spinlock.

Remove this list completely and perform a garbage collection on-the-fly,
at lookup time, using the expired nodes we met during the tree
traversal.

This removes a lot of code, makes locking more standard, and obsoletes
two sysctls (inet_peer_gc_mintime and inet_peer_gc_maxtime). This also
removes two pointers in inet_peer structure.

There is still a false sharing effect because refcnt is in first cache
line of object [were the links and keys used by lookups are located], we
might move it at the end of inet_peer structure to let this first cache
line mostly read by cpus.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Andi Kleen <andi@firstfloor.org>
CC: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch lowers the default initRTO from 3secs to 1sec per
RFC2988bis. It falls back to 3secs if the SYN or SYN-ACK packet
has been retransmitted, AND the TCP timestamp option is not on.

It also adds support to take RTT sample during 3WHS on the passive
open side, just like its active open counterpart, and uses it, if
valid, to seed the initRTO for the data transmission phase.

The patch also resets ssthresh to its initial default at the
beginning of the data transmission phase, and reduces cwnd to 1 if
there has been MORE THAN ONE retransmission during 3WHS per RFC5681.
Signed-off-by: H.K. Jerry Chu <hkchu@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Use same logic as SIT tunnel to handle link local address
for GRE tunnel. OSPFv3 requires link-local address to function.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This driver keeps stats in net_device stats therefore it
does not need to define it's own get_stats hook.

Also, use standard format for net_device_ops (without &).
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This change is meant to remove all support for displaying an ntuple as
strings via ETHTOOL_GRXNTUPLE.  The reason for this change is due to the
fact that multiple issues have been found including:
 - Multiple buffer overruns for strings being displayed.
 - Incorrect filters displayed, cleared filters with ring of -2 are displayed
 - Setting get_rx_ntuple displays no rules if defined.
 - Endianess wrong on displayed values.
 - Hard limit of 1024 filters makes display functionality extremely limited

The only driver that had supported this interface was ixgbe.  Since it no
longer uses the interface and due to the issues mentioned above I am
submitting this patch to remove it.

v2:
Updated based on comments from Ben Hutchings
 - Left ETH_SS_NTUPLE_FILTERS in code but commented on it being deprecated
 - Removed ethtool_rx_ntuple_list and ethtool_rx_ntuple_flow_spec_container
 - Left ETHTOOL_GRXNTUPLE but commented it as deprecated

Also cleaned up set_rx_ntuple since there is no flow spec container to
maintain we can drop all the code for the alloc and free of it and just
return ops->set_rx_ntuple().
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Fixes these errors after the removal of interrupt.h from netdevice.h:

drivers/net/ll_temac_main.c: In function 'temac_open':
drivers/net/ll_temac_main.c:859:2: error: implicit declaration of function 'request_irq'
drivers/net/ll_temac_main.c:870:2: error: implicit declaration of function 'free_irq'
drivers/net/ll_temac_main.c: In function 'temac_poll_controller':
drivers/net/ll_temac_main.c:903:2: error: implicit declaration of function 'disable_irq'
drivers/net/ll_temac_main.c:909:2: error: implicit declaration of function 'enable_irq'
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Padding per MSDU will affect the length of next packet and hence
the exact length of next packet is uncertain here.

Also, aggregation of transmission buffer, while downloading the
data to the card, wont gain much on the AMSDU packets as the AMSDU
packets utilizes the transmission buffer space to the maximum
(adapter->tx_buf_size).
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

All MSDUs, except the last one in an AMSDU, should end up at 4
bytes boundary. There is need to check if enough skb_tailroom
space exists before padding the skb.

Also re-arranging code for better readablity.
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

The check of skb list empty before calling skb_peek and skb_dequeue is
redundant. These functions returns NULL if the list is empty.
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Instead of counting the number of packets in txq
for particular RA list before AMSDU creation,
maintain a counter which will keep track of the
same.

This will reduce some MIPS while generating AMSDU
traffic as we only have to check the counter instead
of traversing through skb list.
Signed-off-by: Yogesh Ashok Powar <yogeshp@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

the code always returns ret regardless, so if(ret) check is unecessary.
Signed-off-by: Greg Dietsche <Gregory.Dietsche@cuw.edu>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Use less indentions and remove uneeded irq-save flags.
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Adding new event that close RX BA session in case of periodic BT activity
limiting WLAN activity.
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Some devices support BT/WLAN co-existence algorigthms.
In order not to harm the system performance and user experience, the device
requests not to allow any RX BA session and tear down existing RX BA sessions
based on system constraints such as periodic BT activity that needs to limit
WLAN activity (eg.SCO or A2DP).
In such cases, the intention is to limit the duration of the RX PPDU and
therefore prevent the peer device to use A-MPDU aggregation.

Adding ieee80211_stop_rx_ba_session() callback
that can be used by the driver to stop existing BA sessions.
Signed-off-by: Shahar Levi <shahar_levi@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

* remove interrupt.g inclusion from netdevice.h -- not needed
* fixup fallout, add interrupt.h and hardirq.h back where needed.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fixed Rx pause counter for Lancer. Swapping hi and lo words.
Signed-off-by: Selvin Xavier <selvin.xavier@emulex.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

the code always returns 'status' regardless, so if(status) check is unecessary.
Signed-off-by: Greg Dietsche <Gregory.Dietsche@cuw.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>

enic driver gets MTU change notifications for MTU changes in the
port profile associated to a dynamic vnic. This patch adds support
in enic driver to set new MTU on the dynamic vnic and dynamically
adjust its buffers with new MTU size in response to such notifications.
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Stacking VLANs on top of the macvlan device does not
work if the lowerdev device is using vlan filters set
by NETIF_F_HW_VLAN_FILTER. Add ndo ops to pass vlan
calls to lowerdev.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Outside of net/sctp/ipv6.c, IPV6 specific code needs to
be ifdef guarded.

This fixes build failures with IPV6 disabled.
Reported-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Setting tx power can be deferred during scan or changing channel.
If after that correct tx power settings will not be sent to device,
we can observe transmission problems and timeouts. Force to send
tx power settings also after partial rxon change, to assure device
always be configured with up-to-date settings.

Resolves:
https://bugzilla.kernel.org/show_bug.cgi?id=36492

Cc: stable@kernel.org # 2.6.39+
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>