platform/kernel/linux-starfive.git
11 years agoMerge branch 'ipip_gso'
David S. Miller [Sat, 19 Oct 2013 23:37:06 +0000 (19:37 -0400)]
Merge branch 'ipip_gso'

Eric Dumazet says:

====================
net: Implement GSO/TSO support for IPIP

This patch serie implements GSO/TSO support for IPIP

David, please note it applies after "ipv4: gso: send_check() & segment() cleanups"
( http://patchwork.ozlabs.org/patch/284714/ )

Broadcom bnx2x driver is now enabled for TSO support of IPIP traffic

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      3357.88   5.09     3.70     2.983   2.167

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      8532.40   2.55     7.73     0.588   1.781
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobnx2x: add TSO support for IPIP
Eric Dumazet [Sat, 19 Oct 2013 18:42:58 +0000 (11:42 -0700)]
bnx2x: add TSO support for IPIP

bnx2x driver already handles TSO for GRE, current code
is the same for IPIP.

Performance results : (Note we are now limited by receiver,
as it does not support GRO for IPIP yet)

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      7710.19   4.52     6.62     1.152   1.687

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      8532.40   2.55     7.73     0.588   1.781

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipip: add GSO/TSO support
Eric Dumazet [Sat, 19 Oct 2013 18:42:57 +0000 (11:42 -0700)]
ipip: add GSO/TSO support

Now inet_gso_segment() is stackable, its relatively easy to
implement GSO/TSO support for IPIP

Performance results, when segmentation is done after tunnel
device (as no NIC is yet enabled for TSO IPIP support) :

Before patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      3357.88   5.09     3.70     2.983   2.167

After patch :

lpq83:~# ./netperf -H 7.7.9.84 -Cc
MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 7.7.9.84 () port 0 AF_INET
Recv   Send    Send                          Utilization       Service Demand
Socket Socket  Message  Elapsed              Send     Recv     Send    Recv
Size   Size    Size     Time     Throughput  local    remote   local   remote
bytes  bytes   bytes    secs.    10^6bits/s  % S      % S      us/KB   us/KB

 87380  16384  16384    10.00      7710.19   4.52     6.62     1.152   1.687

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4: gso: make inet_gso_segment() stackable
Eric Dumazet [Sat, 19 Oct 2013 18:42:56 +0000 (11:42 -0700)]
ipv4: gso: make inet_gso_segment() stackable

In order to support GSO on IPIP, we need to make
inet_gso_segment() stackable.

It should not assume network header starts right after mac
header.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4: generalize gre_handle_offloads
Eric Dumazet [Sat, 19 Oct 2013 18:42:55 +0000 (11:42 -0700)]
ipv4: generalize gre_handle_offloads

This patch makes gre_handle_offloads() more generic
and rename it to iptunnel_handle_offloads()

This will be used to add GSO/TSO support to IPIP tunnels.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: generalize skb_segment()
Eric Dumazet [Sat, 19 Oct 2013 18:42:54 +0000 (11:42 -0700)]
net: generalize skb_segment()

While implementing GSO/TSO support for IPIP, I found skb_segment()
was assuming network header was immediately following mac header.

Its not really true in the case inet_gso_segment() is stacked :
By the time tcp_gso_segment() is called, network header points
to the inner IP header.

Let's instead assume nothing and pick the current offsets found in
original skb, we have skb_headers_offset_update() helper for that.

Also move the csum_start update inside skb_headers_offset_update()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv6: gso: remove redundant locking
Eric Dumazet [Fri, 18 Oct 2013 21:43:55 +0000 (14:43 -0700)]
ipv6: gso: remove redundant locking

ipv6_gso_send_check() and ipv6_gso_segment() are called by
skb_mac_gso_segment() under rcu lock, no need to use
rcu_read_lock() / rcu_read_unlock()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobe2net: Rework PCIe error report log messaging
Ajit Khaparde [Fri, 18 Oct 2013 21:06:24 +0000 (16:06 -0500)]
be2net: Rework PCIe error report log messaging

Currently we log a message whenever pcie_enable_error_reporting fails.
The message clutters up logs, especially when we don't support it for VFs.
Instead enable this only for PFs and log a message when the call succeeds.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: misc: Remove extern from function prototypes
Joe Perches [Fri, 18 Oct 2013 20:48:25 +0000 (13:48 -0700)]
net: misc: Remove extern from function prototypes

There are a mix of function prototypes with and without extern
in the kernel sources.  Standardize on not using extern for
function prototypes.

Function prototypes don't need to be written with extern.
extern is assumed by the compiler.  Its use is as unnecessary as
using auto to declare automatic/local variables in a block.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: ipv4/ipv6: Remove extern from function prototypes
Joe Perches [Fri, 18 Oct 2013 20:48:24 +0000 (13:48 -0700)]
net: ipv4/ipv6: Remove extern from function prototypes

There are a mix of function prototypes with and without extern
in the kernel sources.  Standardize on not using extern for
function prototypes.

Function prototypes don't need to be written with extern.
extern is assumed by the compiler.  Its use is as unnecessary as
using auto to declare automatic/local variables in a block.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: dccp: Remove extern from function prototypes
Joe Perches [Fri, 18 Oct 2013 20:48:23 +0000 (13:48 -0700)]
net: dccp: Remove extern from function prototypes

There are a mix of function prototypes with and without extern
in the kernel sources.  Standardize on not using extern for
function prototypes.

Function prototypes don't need to be written with extern.
extern is assumed by the compiler.  Its use is as unnecessary as
using auto to declare automatic/local variables in a block.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: 8021q/bluetooth/bridge/can/ceph: Remove extern from function prototypes
Joe Perches [Fri, 18 Oct 2013 20:48:22 +0000 (13:48 -0700)]
net: 8021q/bluetooth/bridge/can/ceph: Remove extern from function prototypes

There are a mix of function prototypes with and without extern
in the kernel sources.  Standardize on not using extern for
function prototypes.

Function prototypes don't need to be written with extern.
extern is assumed by the compiler.  Its use is as unnecessary as
using auto to declare automatic/local variables in a block.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4: gso: send_check() & segment() cleanups
Eric Dumazet [Fri, 18 Oct 2013 20:13:27 +0000 (13:13 -0700)]
ipv4: gso: send_check() & segment() cleanups

inet_gso_segment() and inet_gso_send_check() are called by
skb_mac_gso_segment() under rcu lock, no need to use
rcu_read_lock() / rcu_read_unlock()

Avoid calling ip_hdr() twice per function.

We can use ip_send_check() helper.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: Remove __exit tag from bond_netlink_fini().
David S. Miller [Sat, 19 Oct 2013 23:09:18 +0000 (19:09 -0400)]
bonding: Remove __exit tag from bond_netlink_fini().

It can be called from the module init function, so it cannot
be in the exit section.

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'bonding'
David S. Miller [Sat, 19 Oct 2013 22:59:25 +0000 (18:59 -0400)]
Merge branch 'bonding'

Jiri Pirko says:

====================
bonding: introduce bonding options Netlink support

This patchset basically allows "mode" and "active_slave" bonding options
to be propagated and set up via standart RT Netlink interface.

In future other options can be easily added as well.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: add Netlink support active_slave option
Jiri Pirko [Fri, 18 Oct 2013 15:43:39 +0000 (17:43 +0200)]
bonding: add Netlink support active_slave option

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: add Netlink support mode option
Jiri Pirko [Fri, 18 Oct 2013 15:43:38 +0000 (17:43 +0200)]
bonding: add Netlink support mode option

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: move active_slave getting into separate function
Jiri Pirko [Fri, 18 Oct 2013 15:43:37 +0000 (17:43 +0200)]
bonding: move active_slave getting into separate function

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: remove bond_ioctl_change_active()
Jiri Pirko [Fri, 18 Oct 2013 15:43:36 +0000 (17:43 +0200)]
bonding: remove bond_ioctl_change_active()

no longer needed since bond_option_active_slave_set() can be used
instead.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: move active_slave setting into separate function
Jiri Pirko [Fri, 18 Oct 2013 15:43:35 +0000 (17:43 +0200)]
bonding: move active_slave setting into separate function

Do a bit of refactoring on the way.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: move mode setting into separate function
Jiri Pirko [Fri, 18 Oct 2013 15:43:34 +0000 (17:43 +0200)]
bonding: move mode setting into separate function

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: push Netlink bits into separate file
Jiri Pirko [Fri, 18 Oct 2013 15:43:33 +0000 (17:43 +0200)]
bonding: push Netlink bits into separate file

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoem_ipset: use dev_net() accessor
stephen hemminger [Thu, 17 Oct 2013 00:29:34 +0000 (17:29 -0700)]
em_ipset: use dev_net() accessor

Randy found that if network namespace not enabled then
nd_net does not exist and would cause compilation failure.

This is handled correctly by using the dev_net() macro.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: remove redundant code in __tcp_retransmit_skb()
Neal Cardwell [Wed, 16 Oct 2013 16:36:51 +0000 (12:36 -0400)]
tcp: remove redundant code in __tcp_retransmit_skb()

Remove the specialized code in __tcp_retransmit_skb() that tries to trim
any ACKed payload preceding a FIN before we retransmit (this was added
in 1999 in v2.2.3pre3). This trimming code was made unreachable by the
more general code added above it that uses tcp_trim_head() to trim any
ACKed payload, with or without a FIN (this was added in "[NET]: Add
segmentation offload support to TCP." in 2002 circa v2.5.33).

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogianfar: Simplify MQ polling to avoid soft lockup
Claudiu Manoil [Mon, 14 Oct 2013 14:05:09 +0000 (17:05 +0300)]
gianfar: Simplify MQ polling to avoid soft lockup

Under certain low traffic conditions, the single core
devices with multiple Rx/Tx queues (MQ mode) may reach
soft lockup due to gfar_poll not returning in proper time.
The following exception was obtained using iperf on a 100Mbit
half-duplex link, for a p1010 single core device:

BUG: soft lockup - CPU#0 stuck for 23s! [iperf:2847]
Modules linked in:
CPU: 0 PID: 2847 Comm: iperf Not tainted 3.12.0-rc3 #16
task: e8bf8000 ti: eeb16000 task.ti: ee646000
NIP: c0255b6c LR: c0367ae8 CTR: c0461c18
REGS: eeb17e70 TRAP: 0901   Not tainted  (3.12.0-rc3)
MSR: 00029000 <CE,EE,ME>  CR: 44228428  XER: 20000000

GPR00: c0367ad4 eeb17f20 e8bf8000 ee01f4b4 00000008 ffffffff ffffffff
00000000
GPR08: 000000c0 00000008 000000ff ffffffc0 000193fe
NIP [c0255b6c] find_next_bit+0xb8/0xc4
LR [c0367ae8] gfar_poll+0xc8/0x1d8
Call Trace:
[eeb17f20] [c0367ad4] gfar_poll+0xb4/0x1d8 (unreliable)
[eeb17f70] [c0422100] net_rx_action+0xa4/0x158
[eeb17fa0] [c003ec6c] __do_softirq+0xcc/0x17c
[eeb17ff0] [c000c28c] call_do_softirq+0x24/0x3c
[ee647cc0] [c0004660] do_softirq+0x6c/0x94
[ee647ce0] [c003eb9c] local_bh_enable+0x9c/0xa0
[ee647cf0] [c0454fe8] tcp_prequeue_process+0xa4/0xdc
[ee647d10] [c0457e44] tcp_recvmsg+0x498/0x96c
[ee647d80] [c047b630] inet_recvmsg+0x40/0x64
[ee647da0] [c040ca8c] sock_recvmsg+0x90/0xc0
[ee647e30] [c040edb8] SyS_recvfrom+0x98/0xfc

To prevent this, the outer while() loop has been removed
allowing gfar_poll() to return faster even if there's
still budget left.  Also, there's no need to recompute
the budget per Rx queue anymore.

Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofib: Use const struct nl_info * in rtmsg_fib
Joe Perches [Thu, 17 Oct 2013 20:34:11 +0000 (13:34 -0700)]
fib: Use const struct nl_info * in rtmsg_fib

The rtmsg_fib function doesn't modify this argument so mark
it const.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoax25: cleanup a range test
Dan Carpenter [Fri, 18 Oct 2013 09:06:56 +0000 (12:06 +0300)]
ax25: cleanup a range test

The current test works fine in practice.  The "amount" variable is
actually used as a boolean so negative values or any non-zero values
count as "true".  However since we don't allow numbers greater than one,
let's not allow negative numbers either.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agofib_trie: remove duplicated rcu lock
baker.zhang [Sun, 13 Oct 2013 11:50:09 +0000 (19:50 +0800)]
fib_trie: remove duplicated rcu lock

fib_table_lookup has included the rcu lock protection.

Signed-off-by: baker.zhang <baker.kernel@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
David S. Miller [Fri, 18 Oct 2013 17:51:35 +0000 (13:51 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec-next

Steffen Klassert says:

====================
1) Don't use a wildcard SA if a more precise one is in acquire state,
   from Fan Du.

2) Simplify the SA lookup when using wildcard source. We need to check
   only the destination in this case, from Fan Du.

3) Add a receive path hook for IPsec virtual tunnel interfaces
   to xfrm6_mode_tunnel.

4) Add support for IPsec virtual tunnel interfaces to ipv6.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'qlcnic'
David S. Miller [Fri, 18 Oct 2013 17:42:30 +0000 (13:42 -0400)]
Merge branch 'qlcnic'

Himanshu Madhani says:

====================
qlcnic: ethtool enhancements and code cleanup.

This patch series contains

o updates to ethtool for pause settings and enhance
  register dump to display mask and ring indices.
o cleanup in DCB code and remove redundant eSwitch enablement command.
o fixed firmware dump collection logic to skip unknown entries.

Changes from v3 -> v4
o Dropped patch for Tx queue validation to be submitted in net.

Changes from v2 -> v3

o Updated patch to print informational messages as per Joe Perches's comment.

Changes from v1 -> v2

o Dropped patch to register device if adapter is in FAILED state for more rework.
o Updated patch to display ring indices via ethtool per Ben Hutchings's comment.
o Update patch for DCB cleanup per Stephen Hemminger's comment.

Please apply to net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: update version to 5.3.51
Himanshu Madhani [Fri, 18 Oct 2013 16:22:35 +0000 (12:22 -0400)]
qlcnic: update version to 5.3.51

Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Skip unknown entry type while collecting firmware dump
Shahed Shaikh [Fri, 18 Oct 2013 16:22:34 +0000 (12:22 -0400)]
qlcnic: Skip unknown entry type while collecting firmware dump

o Driver aborts the minidump collection operation when it finds
  an unknown entry opcode. This patch skips unknown entry type
  and resumes the minidump collection operation.
o Removed a comparision of collected dump size with expected dump size.
  Size may differ when driver decides to skip an entry.

Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: dcb code cleanup and refactoring.
Sucheta Chakraborty [Fri, 18 Oct 2013 16:22:33 +0000 (12:22 -0400)]
qlcnic: dcb code cleanup and refactoring.

o Move dcb specific function definitions to dcb files.
o Move dcb specific variables to qlcnic_dcb structure.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Remove redundant eSwitch enable commands
Sony Chacko [Fri, 18 Oct 2013 16:22:32 +0000 (12:22 -0400)]
qlcnic: Remove redundant eSwitch enable commands

When more than one NIC physical functions are enabled on a port,
eSwitch on that port gets enabled automatically. Driver
need not explicitly enable the eSwitch.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Update ethtool standard pause settings.
Jitendra Kalsaria [Fri, 18 Oct 2013 16:22:31 +0000 (12:22 -0400)]
qlcnic: Update ethtool standard pause settings.

Update ethtool standard pause parameter settings and display

Signed-off-by: Jitendra Kalsaria <jitendra.kalsaria@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Firmware dump collection when auto recovery is disabled.
Pratik Pujar [Fri, 18 Oct 2013 16:22:30 +0000 (12:22 -0400)]
qlcnic: Firmware dump collection when auto recovery is disabled.

o Allow collecting the firmware dump of halted firmware when auto
  recovery is disabled.

Signed-off-by: Pratik Pujar <pratik.pujar@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Enhance ethtool to display ring indices and interrupt mask
Pratik Pujar [Fri, 18 Oct 2013 16:22:29 +0000 (12:22 -0400)]
qlcnic: Enhance ethtool to display ring indices and interrupt mask

o Updated ethtool -d <ethX> option to display ring indices for Transmit(Tx),
  Receive(Rx), and Status(St) rings.
o Updated ethtool -d <ethX> option to display ring interrupt mask for Transmit(Tx),
  and Status(St) rings.

Signed-off-by: Pratik Pujar <pratik.pujar@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoqlcnic: Print informational messages only once during driver load.
Sucheta Chakraborty [Fri, 18 Oct 2013 16:22:28 +0000 (12:22 -0400)]
qlcnic: Print informational messages only once during driver load.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: rename tcp_tso_segment()
Eric Dumazet [Fri, 18 Oct 2013 17:36:17 +0000 (10:36 -0700)]
tcp: rename tcp_tso_segment()

Rename tcp_tso_segment() to tcp_gso_segment(), to better reflect
what is going on, and ease grep games.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'tipc'
David S. Miller [Fri, 18 Oct 2013 17:22:19 +0000 (13:22 -0400)]
Merge branch 'tipc'

Jon Maloy says:

====================
Some small and relatively straightforward patches. With exception of
the two first ones they are all unrelated and address minor issues.

v2: update of v1 (http://patchwork.ozlabs.org/patch/277404/)

-added commit to use memcpy_fromiovec on user data as per v1 feedback
-updated sparse fix commit to drop chunks covered by above commit
-added new commit that greatly simplifies the link lookup routine
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: simplify the link lookup routine
Erik Hugne [Fri, 18 Oct 2013 05:23:21 +0000 (07:23 +0200)]
tipc: simplify the link lookup routine

When checking statistics or changing parameters on a link, the
link_find_link function is used to locate the link with a given
name. The complex method of deconstructing the name into local
and remote address/interface is error prone and may fail if the
interface names contains special characters. We change the lookup
method to iterate over the list of nodes and compare the link
names.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: correct return value of link_cmd_set_value routine
Ying Xue [Fri, 18 Oct 2013 05:23:20 +0000 (07:23 +0200)]
tipc: correct return value of link_cmd_set_value routine

link_cmd_set_value() takes commands for link, bearer and media related
configuration. Genereally the function returns 0 when a command is
recognized, and -EINVAL when it is not. However, in the switch for link
related commands it returns 0 even when the command is unrecognized. This
will sometimes make it look as if a failed configuration command has been
successful, but has otherwise no negative effects.

We remove this anomaly by returning -EINVAL even for link commands. We also
rework all three switches to make them  conforming to common kernel coding
style.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: correct return value of recv_msg routine
Ying Xue [Fri, 18 Oct 2013 05:23:19 +0000 (07:23 +0200)]
tipc: correct return value of recv_msg routine

Currently, rcv_msg() always returns zero on a packet delivery upcall
from net_device.

To make its behavior more compliant with the way this API should be
used, we change this to let it return NET_RX_SUCCESS (which is zero
anyway) when it is able to handle the packet, and NET_RX_DROP otherwise.
The latter does not imply any functional change, it only enables the
driver to keep more accurate statistics about the fate of delivered
packets.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: avoid unnecessary lookup for tipc bearer instance
Ying Xue [Fri, 18 Oct 2013 05:23:18 +0000 (07:23 +0200)]
tipc: avoid unnecessary lookup for tipc bearer instance

tipc_block_bearer() currently takes a bearer name (const char*)
as argument. This requires the function to make a lookup to find
the pointer to the corresponding bearer struct. In the current
code base this is not necessary, since the only two callers
(tipc_continue(),recv_notification()) already have validated
copies of this pointer, and hence can pass it directly in the
function call.

We change tipc_block_bearer() to directly take struct tipc_bearer*
as argument instead.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: make bearer and media naming consistent
Ying Xue [Fri, 18 Oct 2013 05:23:17 +0000 (07:23 +0200)]
tipc: make bearer and media naming consistent

TIPC 'bearer' exists as an abstract concept, while 'media'
is deemed a specific implementation of a bearer, such as Ethernet
or Infiniband media. When a component inside TIPC wants to control
a specific media, it only needs to access the generic bearer API
to achieve this. However, in the current media implementations,
the 'bearer' name is also extensively used in media specific
function and variable names.

This may create confusion, so we choose to replace the term 'bearer'
with 'media' in all function names, variable names, and prefixes
where this is what really is meant.

Note that this change is cosmetic only, and no runtime behaviour
changes are made here.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: silence sparse warnings
Ying Xue [Fri, 18 Oct 2013 05:23:16 +0000 (07:23 +0200)]
tipc: silence sparse warnings

Eliminate below sparse warnings:

net/tipc/link.c:1210:37: warning: cast removes address space of expression
net/tipc/link.c:1218:59: warning: incorrect type in argument 2 (different address spaces)
net/tipc/link.c:1218:59:    expected void const [noderef] <asn:1>*from
net/tipc/link.c:1218:59:    got unsigned char const [usertype] *[assigned] sect_crs
net/tipc/socket.c:341:49: warning: Using plain integer as NULL pointer
net/tipc/socket.c:1371:36: warning: Using plain integer as NULL pointer
net/tipc/socket.c:1694:57: warning: Using plain integer as NULL pointer

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Andreas Bofjäll <andreas.bofjall@ericsson.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: remove iovec length parameter from all sending functions
Ying Xue [Fri, 18 Oct 2013 05:23:15 +0000 (07:23 +0200)]
tipc: remove iovec length parameter from all sending functions

tipc_msg_build() now copies message data from iovec to skb_buff
using memcpy_fromiovecend(), which doesn't need to be passed the
iovec length to perform the copying.

So we remove the parameter indicating iovec length in all
functions where TIPC messages are built and sent.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: don't use memcpy to copy from user space
Ying Xue [Fri, 18 Oct 2013 05:23:14 +0000 (07:23 +0200)]
tipc: don't use memcpy to copy from user space

tipc_msg_build() calls skb_copy_to_linear_data_offset() to copy data
from user space to kernel space. However, the latter function does
in its turn call memcpy() to perform the actual copying. This poses
an obvious security and robustness risk, since memcpy() never makes
any validity check on the pointer it is copying from.

To correct this, we the replace the offending function call with
a call to memcpy_fromiovecend(), which uses copy_from_user() to
perform the copying.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: Delete trailing semi-colon from definition of netdev_WARN()
Ben Hutchings [Mon, 14 Oct 2013 20:49:21 +0000 (21:49 +0100)]
net: Delete trailing semi-colon from definition of netdev_WARN()

Macro definitions should not normally end with a semi-colon, as this
makes it dangerous to use them an if...else statement.  Happily this
has not happened yet.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: refactor sk_page_frag_refill()
Eric Dumazet [Thu, 17 Oct 2013 23:27:07 +0000 (16:27 -0700)]
net: refactor sk_page_frag_refill()

While working on virtio_net new allocation strategy to increase
payload/truesize ratio, we found that refactoring sk_page_frag_refill()
was needed.

This patch splits sk_page_frag_refill() into two parts, adding
skb_page_frag_refill() which can be used without a socket.

While we are at it, add a minimum frag size of 32 for
sk_page_frag_refill()

Michael will either use netdev_alloc_frag() from softirq context,
or skb_page_frag_refill() from process context in refill_work()
 (GFP_KERNEL allocations)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Michael Dalton <mwdalton@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'pci_set_drvdata'
David S. Miller [Fri, 18 Oct 2013 04:04:05 +0000 (00:04 -0400)]
Merge branch 'pci_set_drvdata'

Jingoo Han says:

====================
net: ethernet: remove unnecessary pci_set_drvdata() part 1

Since commit 0998d0631001288a5974afc0b2a5f568bcdecb4d
(device-core: Ensure drvdata = NULL when no driver is bound),
the driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: enic: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:25:29 +0000 (09:25 +0900)]
net: enic: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: cxgb4vf: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:24:54 +0000 (09:24 +0900)]
net: cxgb4vf: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: cxgb2: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:24:07 +0000 (09:24 +0900)]
net: cxgb2: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: cxgb3: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:23:27 +0000 (09:23 +0900)]
net: cxgb3: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: cxgb4: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:23:00 +0000 (09:23 +0900)]
net: cxgb4: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: bna: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:22:22 +0000 (09:22 +0900)]
net: bna: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: tg3: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:21:54 +0000 (09:21 +0900)]
net: tg3: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: bnx2x: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:21:26 +0000 (09:21 +0900)]
net: bnx2x: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: bnx2: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:21:10 +0000 (09:21 +0900)]
net: bnx2: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: alx: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:20:24 +0000 (09:20 +0900)]
net: alx: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: amd8111e: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:19:54 +0000 (09:19 +0900)]
net: amd8111e: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: pcnet32: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:19:23 +0000 (09:19 +0900)]
net: pcnet32: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Don Fry <pcnet32@frontier.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: starfire: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:18:53 +0000 (09:18 +0900)]
net: starfire: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: 8390: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:18:18 +0000 (09:18 +0900)]
net: 8390: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: typhoon: remove unnecessary pci_set_drvdata()
Jingoo Han [Fri, 18 Oct 2013 00:17:58 +0000 (09:17 +0900)]
net: typhoon: remove unnecessary pci_set_drvdata()

The driver core clears the driver data to NULL after device_release
or on probe failure. Thus, it is not needed to manually clear the
device driver data to NULL.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: David Dillow <dave@thedillows.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Thu, 17 Oct 2013 20:14:29 +0000 (16:14 -0400)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
This is a batch of updates intended for the 3.13 stream...

The biggest item of interest in here is wcn36xx, the new mac80211
driver for Qualcomm WCN3660/WCN3680 hardware.

Regarding the mac80211 bits, Johannes says:

"We have an assortment of cleanups and new features, of which the
biggest one is probably the channel-switch support in IBSS. Nothing
else really stands out much."

On top of that, the ath9k and rt2x00 get a lot of update action from
Felix Fietkau and Gabor Juhos, respectively.  There are a handful of
updates to other drivers here and there as well.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4: shrink rt_cache_stat
Eric Dumazet [Wed, 16 Oct 2013 09:49:04 +0000 (02:49 -0700)]
ipv4: shrink rt_cache_stat

Half of the rt_cache_stat fields are no longer used after IP
route cache removal, lets shrink this per cpu area.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonetdev: inet_timewait_sock.h missing semi-colon when KMEMCHECK is enabled
Randy Dunlap [Mon, 14 Oct 2013 19:36:32 +0000 (12:36 -0700)]
netdev: inet_timewait_sock.h missing semi-colon when KMEMCHECK is enabled

Fix (a few hundred) build errors due to missing semi-colon when
KMEMCHECK is enabled:

  include/net/inet_timewait_sock.h:139:2: error: expected ',', ';' or '}' before 'int'
  include/net/inet_timewait_sock.h:148:28: error: 'const struct inet_timewait_sock' has no member named 'tw_death_node'

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'xen_netback'
David S. Miller [Thu, 17 Oct 2013 19:36:04 +0000 (15:36 -0400)]
Merge branch 'xen_netback'

xen-netback: IPv6 offload support

====================
This patch series adds support for checksum and large packet offloads
into xen-netback.  Testing has mainly been done using the Microsoft
network hardware certification suite running in Server 2008R2 VMs with
Citrix PV frontends.

v2:
- Fixed Wei's email address in Cc lines

v3:
- Responded to Wei's comments:
 - netif.h now updated with comments and a definition of
   XEN_NETIF_GSO_TYPE_NONE.
 - limited number of pullups
- Responded to Annie's comments:
 - New GSO_BIT macro

v4:
- Responded to more of Wei's comments
- Remove parsing of IPv6 fragment header and added warning

v5:
- Added comment concerning the value chosen for PKT_PROT_LEN
- Dropped deprecation of feature-no-csum-offload
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netback: enable IPv6 TCP GSO to the guest
Paul Durrant [Wed, 16 Oct 2013 16:50:32 +0000 (17:50 +0100)]
xen-netback: enable IPv6 TCP GSO to the guest

This patch adds code to handle SKB_GSO_TCPV6 skbs and construct appropriate
extra or prefix segments to pass the large packet to the frontend. New
xenstore flags, feature-gso-tcpv6 and feature-gso-tcpv6-prefix, are sampled
to determine if the frontend is capable of handling such packets.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netback: handle IPv6 TCP GSO packets from the guest
Paul Durrant [Wed, 16 Oct 2013 16:50:31 +0000 (17:50 +0100)]
xen-netback: handle IPv6 TCP GSO packets from the guest

This patch adds a xenstore feature flag, festure-gso-tcpv6, to advertise
that netback can handle IPv6 TCP GSO packets. It creates SKB_GSO_TCPV6 skbs
if the frontend passes an extra segment with the new type
XEN_NETIF_GSO_TYPE_TCPV6 added to netif.h.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netback: Unconditionally set NETIF_F_RXCSUM
Paul Durrant [Wed, 16 Oct 2013 16:50:30 +0000 (17:50 +0100)]
xen-netback: Unconditionally set NETIF_F_RXCSUM

There is no mechanism to insist that a guest always generates a packet
with good checksum (at least for IPv4) so we must handle checksum
offloading from the guest and hence should set NETIF_F_RXCSUM.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netback: add support for IPv6 checksum offload from guest
Paul Durrant [Wed, 16 Oct 2013 16:50:29 +0000 (17:50 +0100)]
xen-netback: add support for IPv6 checksum offload from guest

For performance of VM to VM traffic on a single host it is better to avoid
calculation of TCP/UDP checksum in the sending frontend. To allow this this
patch adds the code necessary to set up partial checksum for IPv6 packets
and xenstore flag feature-ipv6-csum-offload to advertise that fact to
frontends.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoxen-netback: add support for IPv6 checksum offload to guest
Paul Durrant [Wed, 16 Oct 2013 16:50:28 +0000 (17:50 +0100)]
xen-netback: add support for IPv6 checksum offload to guest

Check xenstore flag feature-ipv6-csum-offload to determine if a
guest is happy to accept IPv6 packets with only partial checksum.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'bonding_rcu'
David S. Miller [Thu, 17 Oct 2013 19:32:15 +0000 (15:32 -0400)]
Merge branch 'bonding_rcu'

bonding: patchset for rcu use in bonding

====================
The Patch Set convert the xmit of 3ad and alb mode to use rcu lock.
dd rtnl lock and remove read lock for bond sysfs.

v2 because the bond_for_each_slave_rcu without rcu_read_lock() will occurs one warming, so
add new function for alb xmit path to avoid warming.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: add rtnl lock and remove read lock for bond sysfs
dingtianhong [Tue, 15 Oct 2013 08:28:42 +0000 (16:28 +0800)]
bonding: add rtnl lock and remove read lock for bond sysfs

The bond_for_each_slave() will not be protected by read_lock(),
only protected by rtnl_lock(), so need to replace read_lock()
with rtnl_lock().

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: use RCU protection for alb xmit path
dingtianhong [Tue, 15 Oct 2013 08:28:39 +0000 (16:28 +0800)]
bonding: use RCU protection for alb xmit path

The commit 278b20837511776dc9d5f6ee1c7fabd5479838bb
(bonding: initial RCU conversion) has convert the roundrobin,
active-backup, broadcast and xor xmit path to rcu protection,
the performance will be better for these mode, so this time,
convert xmit path for alb mode.

Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Cc: Nikolay Aleksandrov <nikolay@redhat.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: use RCU protection for 3ad xmit path
dingtianhong [Tue, 15 Oct 2013 08:28:35 +0000 (16:28 +0800)]
bonding: use RCU protection for 3ad xmit path

The commit 278b20837511776dc9d5f6ee1c7fabd5479838bb
(bonding: initial RCU conversion) has convert the roundrobin,
active-backup, broadcast and xor xmit path to rcu protection,
the performance will be better for these mode, so this time,
convert xmit path for 3ad mode.

Suggested-by: Nikolay Aleksandrov <nikolay@redhat.com>
Suggested-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: Wang Yufen <wangyufen@huawei.com>
Cc: Nikolay Aleksandrov <nikolay@redhat.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'net-next' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nftables
David S. Miller [Thu, 17 Oct 2013 19:22:05 +0000 (15:22 -0400)]
Merge branch 'net-next' of git://git./linux/kernel/git/pablo/nftables

Pablo Neira Ayuso says:

====================
netfilter updates: nf_tables pull request

The following patchset contains the current original nf_tables tree
condensed in 17 patches. I have organized them by chronogical order
since the original nf_tables code was released in 2009 and by
dependencies between the different patches.

The patches are:

1) Adapt all existing hooks in the tree to pass hook ops to the
   hook callback function, required by nf_tables, from Patrick McHardy.

2) Move alloc_null_binding to nf_nat_core, as it is now also needed by
   nf_tables and ip_tables, original patch from Patrick McHardy but
   required major changes to adapt it to the current tree that I made.

3) Add nf_tables core, including the netlink API, the packet filtering
   engine, expressions and built-in tables, from Patrick McHardy. This
   patch includes accumulated fixes since 2009 and minor enhancements.
   The patch description contains a list of references to the original
   patches for the record. For those that are not familiar to the
   original work, see [1], [2] and [3].

4) Add netlink set API, this replaces the original set infrastructure
   to introduce a netlink API to add/delete sets and to add/delete
   set elements. This includes two set types: the hash and the rb-tree
   sets (used for interval based matching). The main difference with
   ipset is that this infrastructure is data type agnostic. Patch from
   Patrick McHardy.

5) Allow expression operation overload, this API change allows us to
   provide define expression subtypes depending on the configuration
   that is received from user-space via Netlink. It is used by follow
   up patches to provide optimized versions of the payload and cmp
   expressions and the x_tables compatibility layer, from Patrick
   McHardy.

6) Add optimized data comparison operation, it requires the previous
   patch, from Patrick McHardy.

7) Add optimized payload implementation, it requires patch 5, from
   Patrick McHardy.

8) Convert built-in tables to chain types. Each chain type have special
   semantics (filter, route and nat) that are used by userspace to
   configure the chain behaviour. The main chain regarding iptables
   is that tables become containers of chain, with no specific semantics.
   However, you may still configure your tables and chains to retain
   iptables like semantics, patch from me.

9) Add compatibility layer for x_tables. This patch adds support to
   use all existing x_tables extensions from nf_tables, this is used
   to provide a userspace utility that accepts iptables syntax but
   used internally the nf_tables kernel core. This patch includes
   missing features in the nf_tables core such as the per-chain
   stats, default chain policy and number of chain references, which
   are required by the iptables compatibility userspace tool. Patch
   from me.

10) Fix transport protocol matching, this fix is a side effect of the
    x_tables compatibility layer, which now provides a pointer to the
    transport header, from me.

11) Add support for dormant tables, this feature allows you to disable
    all chains and rules that are contained in one table, from me.

12) Add IPv6 NAT support. At the time nf_tables was made, there was no
    NAT IPv6 support yet, from Tomasz Bursztyka.

13) Complete net namespace support. This patch register the protocol
    family per net namespace, so tables (thus, other objects contained
    in tables such as sets, chains and rules) are only visible from the
    corresponding net namespace, from me.

14) Add the insert operation to the nf_tables netlink API, this requires
    adding a new position attribute that allow us to locate where in the
    ruleset a rule needs to be inserted, from Eric Leblond.

15) Add rule batching support, including atomic rule-set updates by
    using rule-set generations. This patch includes a change to nfnetlink
    to include two new control messages to indicate the beginning and
    the end of a batch. The end message is interpreted as the commit
    message, if it's missing, then the rule-set updates contained in the
    batch are aborted, from me.

16) Add trace support to the nf_tables packet filtering core, from me.

17) Add ARP filtering support, original patch from Patrick McHardy, but
    adapted to fit into the chain type infrastructure. This was recovered
    to be used by nft userspace tool and our compatibility arptables
    userspace tool.

There is still work to do to fully replace x_tables [4] [5] but that can
be done incrementally by extending our netlink API. Moreover, looking at
netfilter-devel and the amount of contributions to nf_tables we've been
getting, I think it would be good to have it mainstream to avoid accumulating
large patchsets skip continuous rebases.

I tried to provide a reasonable patchset, we have more than 100 accumulated
patches in the original nf_tables tree, so I collapsed many of the small
fixes to the main patch we had since 2009 and provide a small batch for
review to netdev, while trying to retain part of the history.

For those who didn't give a try to nf_tables yet, there's a quick howto
available from Eric Leblond that describes how to get things working [6].

Comments/reviews welcome.

Thanks!

[1] http://lwn.net/Articles/324251/
[2] http://workshop.netfilter.org/2013/wiki/images/e/ee/Nftables-osd-2013-developer.pdf
[3] http://lwn.net/Articles/564095/
[4] http://people.netfilter.org/pablo/map-pending-work.txt
[4] http://people.netfilter.org/pablo/nftables-todo.txt
[5] https://home.regit.org/netfilter-en/nftables-quick-howto/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoirda: update comment mentioning IRQF_DISABLED
Michael Opdenacker [Sun, 13 Oct 2013 06:26:31 +0000 (08:26 +0200)]
irda: update comment mentioning IRQF_DISABLED

This patch removes a comment mentioning IRQF_DISABLED,
which is deprecated.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoisdn: remove deprecated IRQF_DISABLED
Michael Opdenacker [Sun, 13 Oct 2013 05:24:29 +0000 (07:24 +0200)]
isdn: remove deprecated IRQF_DISABLED

This patch proposes to remove the use of the IRQF_DISABLED flag

It's a NOOP since 2.6.35 and it will be removed one day.

Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'mlx4'
David S. Miller [Thu, 17 Oct 2013 19:11:31 +0000 (15:11 -0400)]
Merge branch 'mlx4'

Amir Vadai says:

====================
net/mlx4: Mellanox driver update 15-10-2013

This patchset contains small code cleaning patches, and a patch to make
mlx4_core use module_request() in order to load the relevant link layer module
(mlx4_en or mlx4_ib) according to the port type.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4_core: Load higher level modules according to ports type
Eyal Perry [Tue, 15 Oct 2013 14:55:24 +0000 (16:55 +0200)]
net/mlx4_core: Load higher level modules according to ports type

Mellanox ConnectX architecture is:  mlx4_core is the lower level
PCI driver which register on the PCI id, and protocol specific drivers
are depended on it: mlx4_en - for Ethernet and mlx4_ib for Infiniband.
NIC could have multiple ports which can change their type dynamically.
We use the request_module() call to load the relevant protocol driver
when needed: on loading time or at port type change event.

Signed-off-by: Eyal Perry <eyalpe@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4: Unused local variable in mlx4_opreq_action
Amir Vadai [Tue, 15 Oct 2013 14:55:23 +0000 (16:55 +0200)]
net/mlx4: Unused local variable in mlx4_opreq_action

Clean up warning added by commit fe6f700d "net/mlx4_core: Respond to
operation request by firmware".

Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4: Fix typo, move similar defs to same location
Or Gerlitz [Tue, 15 Oct 2013 14:55:22 +0000 (16:55 +0200)]
net/mlx4: Fix typo, move similar defs to same location

Small code cleanup:

1. change MLX4_DEV_CAP_FLAGS2_REASSIGN_MAC_EN to MLX4_DEV_CAP_FLAG2_REASSIGN_MAC_EN

2. put MLX4_SET_PORT_PRIO2TC and MLX4_SET_PORT_SCHEDULER in the same union with the
   other MLX4_SET_PORT_yyy

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/mlx4: Clean the code to eliminate trivial build warnings
Or Gerlitz [Tue, 15 Oct 2013 14:55:21 +0000 (16:55 +0200)]
net/mlx4: Clean the code to eliminate trivial build warnings

Remove code that triggers trivial build warnings.

drivers/net/ethernet/mellanox/mlx4/cmd.c: In function ‘mlx4_set_vf_vlan’:
drivers/net/ethernet/mellanox/mlx4/cmd.c:2256: warning: variable ‘vf_oper’ set but not used
drivers/net/ethernet/mellanox/mlx4/mcg.c: In function ‘mlx4_map_sw_to_hw_steering_mode’:
drivers/net/ethernet/mellanox/mlx4/mcg.c:648: warning: comparison of unsigned expression < 0 is always false
drivers/net/ethernet/mellanox/mlx4/mcg.c: In function ‘mlx4_map_sw_to_hw_steering_id’:
drivers/net/ethernet/mellanox/mlx4/mcg.c:685: warning: comparison of unsigned expression < 0 is always false
drivers/net/ethernet/mellanox/mlx4/mcg.c: In function ‘mlx4_hw_rule_sz’:
drivers/net/ethernet/mellanox/mlx4/mcg.c:712: warning: comparison of unsigned expression < 0 is always false
drivers/net/ethernet/mellanox/mlx4/fw.c: In function ‘mlx4_opreq_action’:
drivers/net/ethernet/mellanox/mlx4/fw.c:1732: warning: variable ‘type_m’ set but not used
drivers/net/ethernet/mellanox/mlx4/srq.c:302: warning: no previous prototype for ‘mlx4_srq_lookup’

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoinet_diag: use sock_gen_put()
Eric Dumazet [Fri, 11 Oct 2013 15:54:49 +0000 (08:54 -0700)]
inet_diag: use sock_gen_put()

TCP listener refactoring, part 6 :

Use sock_gen_put() from inet_diag_dump_one_icsk() for future
SYN_RECV support.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge
David S. Miller [Thu, 17 Oct 2013 18:27:09 +0000 (14:27 -0400)]
Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge

Included changes:
- ensure RecordRoute information is added to BAT_ICMP echo_request/reply only
- use VLAN_ETH_HLEN when possible
- use htons when possible
- substitute old fragmentation code with a new improved implementation by
  Martin Hundebøll
- create common header for BAT_ICMP packets to improve extendibility
- consider the network coding overhead when computing the overall room needed by
  batman headers
- add dummy soft-interface rx mode handler
- minor code refactoring and cleanups

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Thu, 17 Oct 2013 18:02:07 +0000 (14:02 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

11 years agort2x00: rt2800lib: remove duplicate rf_vals for RF3053
Kevin Lo [Mon, 14 Oct 2013 02:05:45 +0000 (10:05 +0800)]
rt2x00: rt2800lib: remove duplicate rf_vals for RF3053

lready have rf_vals_3x with same values.  Hence rf_vals_3053 is removed
in this patch.

Signed-off-by: Kevin Lo <kevlo@kevlo.org>
Acked-by: Paul Menzel <paulepanter@users.sourceforge.net>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agort2x00: rt2800lib: fix RF registers for RT5390/RT5392
Kevin Lo [Sat, 12 Oct 2013 15:25:23 +0000 (23:25 +0800)]
rt2x00: rt2800lib: fix RF registers for RT5390/RT5392

Update rf registers to use the same values that the MediaTek/Ralink
reference driver DPO_RT5572_LinuxSTA_2.6.1.3_20121022 uses.

References:
  RF5390RegTable in chips/rt5390.c
  RF5392RegTable in chips/rt5390.c

Tested on TP-Link TL-WN727N and D-Link DWA-140 Rev.b3 usb wifi dongles.

Signed-off-by: Kevin Lo <kevlo@kevlo.org>
Acked-by: Stanislaw Gruszka <stf_xl@wp.pl>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agomwifiex: use alloc_workqueue() function
Amitkumar Karwar [Sat, 12 Oct 2013 01:33:04 +0000 (18:33 -0700)]
mwifiex: use alloc_workqueue() function

It replaces deprecated create_workqueue().

Signed-off-by: Amitkumar Karwar <akarwar@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: remove references to hw->conf
Felix Fietkau [Fri, 11 Oct 2013 21:31:01 +0000 (23:31 +0200)]
ath9k_hw: remove references to hw->conf

Accessing it to get the current operating channel is racy and in the way
of further channel handling related changes

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: make ath9k_uses_beacons static
Felix Fietkau [Fri, 11 Oct 2013 21:31:00 +0000 (23:31 +0200)]
ath9k: make ath9k_uses_beacons static

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: remove sc->config.cabqReadyTime
Felix Fietkau [Fri, 11 Oct 2013 21:30:59 +0000 (23:30 +0200)]
ath9k: remove sc->config.cabqReadyTime

It is not exposed as a configuration option anyway

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: move channel change code to ath_set_channel
Felix Fietkau [Fri, 11 Oct 2013 21:30:58 +0000 (23:30 +0200)]
ath9k: move channel change code to ath_set_channel

Preparation for adding the scanning state machine to ath9k

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k: make ath9k_cmn_update_ichannel static
Felix Fietkau [Fri, 11 Oct 2013 21:30:57 +0000 (23:30 +0200)]
ath9k: make ath9k_cmn_update_ichannel static

Rework its wrapper function to make it more generic, using it as a
replacement for previous calls to ath9k_cmn_update_ichannel.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: simplify channel flags
Felix Fietkau [Fri, 11 Oct 2013 21:30:56 +0000 (23:30 +0200)]
ath9k_hw: simplify channel flags

There was some duplication between channelFlags and chanmode, as well as
a lot of redundant checks based on the combinations of flags.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
11 years agoath9k_hw: remove IS_CHAN_OFDM()
Felix Fietkau [Fri, 11 Oct 2013 21:30:55 +0000 (23:30 +0200)]
ath9k_hw: remove IS_CHAN_OFDM()

The hardware is always configured with OFDM support enabled

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>