openwrt/staging/blogic.git
12 years agox86: bpf_jit_comp: add XOR instruction for BPF JIT
Daniel Borkmann [Mon, 24 Sep 2012 07:34:51 +0000 (07:34 +0000)]
x86: bpf_jit_comp: add XOR instruction for BPF JIT

This patch is a follow-up for patch "filter: add XOR instruction for use
with X/K" that implements BPF x86 JIT parts for the BPF XOR operation.

Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agofilter: add XOR instruction for use with X/K
Daniel Borkmann [Mon, 24 Sep 2012 02:23:59 +0000 (02:23 +0000)]
filter: add XOR instruction for use with X/K

SKF_AD_ALU_XOR_X has been added a while ago, but as an 'ancillary'
operation that is invoked through a negative offset in K within BPF
load operations. Since BPF_MOD has recently been added, BPF_XOR should
also be part of the common ALU operations. Removing SKF_AD_ALU_XOR_X
might not be an option since this is exposed to user space.

Signed-off-by: Daniel Borkmann <daniel.borkmann@tik.ee.ethz.ch>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: mipsnet: Remove the MIPSsim Ethernet driver.
Steven J. Hill [Mon, 24 Sep 2012 04:47:16 +0000 (04:47 +0000)]
net: mipsnet: Remove the MIPSsim Ethernet driver.

The MIPSsim platform is no longer supported or used. This patch
deletes the Ethernet driver.

Signed-off-by: Steven J. Hill <sjhill@mips.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: use a per task frag allocator
Eric Dumazet [Sun, 23 Sep 2012 23:04:42 +0000 (23:04 +0000)]
net: use a per task frag allocator

We currently use a per socket order-0 page cache for tcp_sendmsg()
operations.

This page is used to build fragments for skbs.

Its done to increase probability of coalescing small write() into
single segments in skbs still in write queue (not yet sent)

But it wastes a lot of memory for applications handling many mostly
idle sockets, since each socket holds one page in sk->sk_sndmsg_page

Its also quite inefficient to build TSO 64KB packets, because we need
about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit
page allocator more than wanted.

This patch adds a per task frag allocator and uses bigger pages,
if available. An automatic fallback is done in case of memory pressure.

(up to 32768 bytes per frag, thats order-3 pages on x86)

This increases TCP stream performance by 20% on loopback device,
but also benefits on other network devices, since 8x less frags are
mapped on transmit and unmapped on tx completion. Alexander Duyck
mentioned a probable performance win on systems with IOMMU enabled.

Its possible some SG enabled hardware cant cope with bigger fragments,
but their ndo_start_xmit() should already handle this, splitting a
fragment in sub fragments, since some arches have PAGE_SIZE=65536

Successfully tested on various ethernet devices.
(ixgbe, igb, bnx2x, tg3, mellanox mlx4)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Vijay Subramanian <subramanian.vijay@gmail.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agogianfar: Change default HW Tx queue scheduling mode
Claudiu Manoil [Sun, 23 Sep 2012 22:39:08 +0000 (22:39 +0000)]
gianfar: Change default HW Tx queue scheduling mode

This is primarily to address transmission timeout occurrences, when
multiple H/W Tx queues are being used concurrently. Because in
the priority scheduling mode the controller does not service the
Tx queues equally (but in ascending index order), Tx timeouts are
being triggered rightaway for a basic test with multiple simultaneous
connections like:
iperf -c <server_ip> -n 100M -P 8

resulting in kernel trace:
NETDEV WATCHDOG: eth1 (fsl-gianfar): transmit queue <X> timed out
------------[ cut here ]------------
WARNING: at net/sched/sch_generic.c:255
...
and controller reset during intense traffic, and possibly further
complications.

This patch changes the default H/W Tx scheduling setting (TXSCHED)
for multi-queue devices, from priority scheduling mode to a weighted
round robin mode with equal weights for all H/W Tx queues, and
addresses the issue above.

Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: loopback: set default mtu to 64K
Eric Dumazet [Sun, 23 Sep 2012 22:28:59 +0000 (22:28 +0000)]
net: loopback: set default mtu to 64K

loopback current mtu of 16436 bytes allows no more than 3 MSS TCP
segments per frame, or 48 Kbytes. Changing mtu to 64K allows TCP
stack to build large frames and significantly reduces stack overhead.

Performance boost on bulk TCP transferts can be up to 30 %, partly
because we now have one ACK message for two 64KB segments, and a lower
probability of hitting /proc/sys/net/ipv4/tcp_reordering default limit.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: Remove unnecessary NULL check in scm_destroy().
David S. Miller [Mon, 24 Sep 2012 19:52:33 +0000 (15:52 -0400)]
net: Remove unnecessary NULL check in scm_destroy().

All callers provide a non-NULL scm argument.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobnx2x: Improve code around bnx2x_tests_str_arr
Merav Sicron [Sun, 23 Sep 2012 03:12:23 +0000 (03:12 +0000)]
bnx2x: Improve code around bnx2x_tests_str_arr

This patch changes the definition of bnx2x_tests_str_arr from static char
pointer to static const char bi-directional array. Also the
bnx2x_get_strings function is simplified.

Reported-by: Joe Perches <joe@perches.com>
Reported-by: David Laight <David.Laight@ACULAB.COM>
Signed-off-by: Merav Sicron <meravs@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoctcm: fix error return code
Peter Senna Tschudin [Mon, 24 Sep 2012 04:24:27 +0000 (04:24 +0000)]
ctcm: fix error return code

Convert a nonnegative error return code to a negative one, as returned
elsewhere in the function.

A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
(
if@p1 (\(ret < 0\|ret != 0\))
 { ... return ret; }
|
ret@p1 = 0
)
... when != ret = e1
    when != &ret
*if(...)
{
  ... when != ret = e2
      when forall
 return ret;
}
// </smpl>

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agodrivers/s390/net: removes unnecessary semicolon
Peter Senna Tschudin [Mon, 24 Sep 2012 04:24:26 +0000 (04:24 +0000)]
drivers/s390/net: removes unnecessary semicolon

removes unnecessary semicolon

Found by Coccinelle: http://coccinelle.lip6.fr/

Signed-off-by: Peter Senna Tschudin <peter.senna@gmail.com>
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: fix possible memory leak in qeth_l3_add_[vipa|rxip]()
Wei Yongjun [Mon, 24 Sep 2012 04:24:25 +0000 (04:24 +0000)]
qeth: fix possible memory leak in qeth_l3_add_[vipa|rxip]()

ipaddr has been allocated in function qeth_l3_add_vipa() but
does not free before leaving from the error handling cases. The
same problem also exists in function qeth_l3_add_rxip().

spatch with a semantic match is used to found this problem.
(http://coccinelle.lip6.fr/)

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agolcs: ensure proper ccw setup
Sebastian Ott [Mon, 24 Sep 2012 04:24:24 +0000 (04:24 +0000)]
lcs: ensure proper ccw setup

Make sure that all ccws used for writing are initialized with
zeros - especially since the last ccw contains a TIC for which
the unused fields have to be zeros.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: cleanup channel path descriptor function
Sebastian Ott [Mon, 24 Sep 2012 04:24:23 +0000 (04:24 +0000)]
qeth: cleanup channel path descriptor function

Cleanup the qeth_get_channel_path_desc function and rename it
to qeth_update_from_chp_desc. No functional change.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Acked-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'master' of git://1984.lsi.us.es/nf-next
David S. Miller [Mon, 24 Sep 2012 19:36:53 +0000 (15:36 -0400)]
Merge branch 'master' of git://1984.lsi.us.es/nf-next

Pablo Neira Ayuso says:

====================
This patchset contains updates for your net-next tree, they are:

* Mostly fixes for the recently pushed IPv6 NAT support:

- Fix crash while removing nf_nat modules from Patrick McHardy.
- Fix unbalanced rcu_read_unlock from Ulrich Weber.
- Merge NETMAP and REDIRECT into one single xt_target module, from
  Jan Engelhardt.
- Fix Kconfig for IPv6 NAT, which allows inconsistent configurations,
  from myself.

* Updates for ipset, all of the from Jozsef Kadlecsik:

- Add the new "nomatch" option to obtain reverse set matching.
- Support for /0 CIDR in hash:net,iface set type.
- One non-critical fix for a rare crash due to pass really
  wrong configuration parameters.
- Coding style cleanups.
- Sparse fixes.
- Add set revision supported via modinfo.i

* One extension for the xt_time match, to support matching during
  the transition between two days with one single rule, from
  Florian Westphal.

* Fix maximum packet length supported by nfnetlink_queue and add
  NFQA_CAP_LEN attribute, from myself.

You can notice that this batch contains a couple of fixes that may
go to 3.6-rc but I don't consider them critical to push them:

* The ipset fix for the /0 cidr case, which is triggered with one
  inconsistent command line invocation of ipset.

* The nfnetlink_queue maximum packet length supported since it requires
  the new NFQA_CAP_LEN attribute to provide a full workaround for the
  described problem.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetfilter: nfnetlink_queue: add NFQA_CAP_LEN attribute
Pablo Neira Ayuso [Mon, 24 Sep 2012 12:52:12 +0000 (14:52 +0200)]
netfilter: nfnetlink_queue: add NFQA_CAP_LEN attribute

This patch adds the NFQA_CAP_LEN attribute that allows us to know
what is the real packet size from user-space (even if we decided
to retrieve just a few bytes from the packet instead of all of it).

Security software that inspects packets should always check for
this new attribute to make sure that it is inspecting the entire
packet.

This also helps to provide a workaround for the problem described
in: http://marc.info/?l=netfilter-devel&m=134519473212536&w=2

Original idea from Florian Westphal.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nfnetlink_queue: fix maximum packet length to userspace
Pablo Neira Ayuso [Thu, 6 Sep 2012 15:09:26 +0000 (17:09 +0200)]
netfilter: nfnetlink_queue: fix maximum packet length to userspace

The packets that we send via NFQUEUE are encapsulated in the NFQA_PAYLOAD
attribute. The length of the packet in userspace is obtained via
attr->nla_len field. This field contains the size of the Netlink
attribute header plus the packet length.

If the maximum packet length is specified, ie. 65535 bytes, and
packets in the range of (65531,65535] are sent to userspace, the
attr->nla_len overflows and it reports bogus lengths to the
application.

To fix this, this patch limits the maximum packet length to 65531
bytes. If larger packet length is specified, the packet that we
send to user-space is truncated to 65531 bytes.

To support 65535 bytes packets, we have to revisit the idea of
the 32-bits Netlink attribute length.

Reported-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_ftp: add sequence tracking pickup facility for injected entries
Pablo Neira Ayuso [Fri, 21 Sep 2012 14:52:08 +0000 (16:52 +0200)]
netfilter: nf_ct_ftp: add sequence tracking pickup facility for injected entries

This patch allows the FTP helper to pickup the sequence tracking from
the first packet seen. This is useful to fix the breakage of the first
FTP command after the failover while using conntrackd to synchronize
states.

The seq_aft_nl_num field in struct nf_ct_ftp_info has been shrinked to
16-bits (enough for what it does), so we can use the remaining 16-bits
to store the flags while using the same size for the private FTP helper
data.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: xt_time: add support to ignore day transition
Florian Westphal [Mon, 17 Sep 2012 00:23:09 +0000 (00:23 +0000)]
netfilter: xt_time: add support to ignore day transition

Currently, if you want to do something like:
"match Monday, starting 23:00, for two hours"
You need two rules, one for Mon 23:00 to 0:00 and one for Tue 0:00-1:00.

The rule: --weekdays Mo --timestart 23:00  --timestop 01:00

looks correct, but it will first match on monday from midnight to 1 a.m.
and then again for another hour from 23:00 onwards.

This permits userspace to explicitly ignore the day transition and
match for a single, continuous time period instead.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agoixgbevf: Return error on failure to enable VLAN
Alexander Duyck [Tue, 21 Aug 2012 00:15:13 +0000 (00:15 +0000)]
ixgbevf: Return error on failure to enable VLAN

With recent kernel changes we can now return errors on a failure to setup a
VLAN filter.  This patch takes advantage of that opportunity so that we can
return either an EIO error in the case of a mailbox failure, or an EACCESS
error in the case of being denied access to the VLAN filter table by the
PF.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Robert Garrett <robertx.e.garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbevf: Add fix to VF to handle multi-descriptor buffers
Alexander Duyck [Sat, 1 Sep 2012 05:12:38 +0000 (05:12 +0000)]
ixgbevf: Add fix to VF to handle multi-descriptor buffers

This change fixes the ixgbevf driver so that it can correctly drop a frame
should it receive a jumbo frame.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbevf: Fix AIM (Adaptive Interrupt Moderation)
Greg Rose [Fri, 31 Aug 2012 05:59:28 +0000 (05:59 +0000)]
ixgbevf: Fix AIM (Adaptive Interrupt Moderation)

While fixing up a patch from Alex Duyck to use q_vectors in ring containers
to update the ITR I bungled it and missed actually updating the counters
in the ring container q_vectors.  This patch fixes my mistake and makes
interrupt moderation actually work.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbevf - Remove unused parameter in ixgbevf_receive_skb
Narendra K [Tue, 14 Aug 2012 00:00:14 +0000 (00:00 +0000)]
ixgbevf - Remove unused parameter in ixgbevf_receive_skb

Remove 'rx_ring' parameter as it is not used in ixgbevf_receive_skb

Signed-off-by: Narendra K <narendra_k@dell.com>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbe: Do not read the spoofed packets counter when not in IOV mode
Greg Rose [Wed, 8 Aug 2012 00:00:58 +0000 (00:00 +0000)]
ixgbe: Do not read the spoofed packets counter when not in IOV mode

The counter is not valid unless the controller is running in IOV mode.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbevf: Fix code for handling timeout
Alexander Duyck [Thu, 2 Aug 2012 01:16:59 +0000 (01:16 +0000)]
ixgbevf: Fix code for handling timeout

The VF driver was not designed to correctly handle a message timeout.  As
a result it is possible for one bad message to invalidate all messages
following it until the part is reset.  Instead we should copy the example
in igbvf of how to handle a mailbox event and message timeout.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agonetlink: Rearrange netlink_kernel_cfg to save space on 64-bit.
David S. Miller [Sun, 23 Sep 2012 06:09:23 +0000 (02:09 -0400)]
netlink: Rearrange netlink_kernel_cfg to save space on 64-bit.

Suggested by Jan Engelhardt.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: TCP Fast Open Server - record retransmits after 3WHS
Neal Cardwell [Sat, 22 Sep 2012 17:03:47 +0000 (17:03 +0000)]
tcp: TCP Fast Open Server - record retransmits after 3WHS

When recording the number of SYNACK retransmits for servers using TCP
Fast Open, fix the code to ensure that we copy over the retransmit
count from the request_sock after we receive the ACK that completes
the 3-way handshake.

The story here is similar to that of SYNACK RTT
measurements. Previously we were always doing this in
tcp_v4_syn_recv_sock(). However, for TCP Fast Open connections
tcp_v4_conn_req_fastopen() calls tcp_v4_syn_recv_sock() at the time we
receive the SYN. So for TFO we must copy the final SYNACK retransmit
count in tcp_rcv_state_process().

Note that copying over the SYNACK retransmit count will give us the
correct count since, as is mentioned in a comment in
tcp_retransmit_timer(), before we receive an ACK for our SYN-ACK a TFO
passive connection does not retransmit anything else (e.g., data or
FIN segments).

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: Make ZNET driver config depend on X86.
David S. Miller [Sat, 22 Sep 2012 22:11:32 +0000 (18:11 -0400)]
net: Make ZNET driver config depend on X86.

We're now using isa_virt_to_bus(), and there really
isn't a generic and consistent test for whether a
platform provides this interface or not.

This driver is also for an x86-only device.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetfilter: ipset: Support to match elements marked with "nomatch"
Jozsef Kadlecsik [Fri, 21 Sep 2012 20:02:36 +0000 (22:02 +0200)]
netfilter: ipset: Support to match elements marked with "nomatch"

Exceptions can now be matched and we can branch according to the
possible cases:

a. match in the set if the element is not flagged as "nomatch"
b. match in the set if the element is flagged with "nomatch"
c. no match

i.e.

iptables ... -m set --match-set ... -j ...
iptables ... -m set --match-set ... --nomatch-entries -j ...
...

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetfilter: ipset: Coding style fixes
Jozsef Kadlecsik [Fri, 21 Sep 2012 20:01:45 +0000 (22:01 +0200)]
netfilter: ipset: Coding style fixes

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetfilter: ipset: Include supported revisions in module description
Jozsef Kadlecsik [Fri, 21 Sep 2012 19:59:32 +0000 (21:59 +0200)]
netfilter: ipset: Include supported revisions in module description

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetfilter: ipset: Add /0 network support to hash:net,iface type
Jozsef Kadlecsik [Mon, 10 Sep 2012 19:22:23 +0000 (21:22 +0200)]
netfilter: ipset: Add /0 network support to hash:net,iface type

Now it is possible to setup a single hash:net,iface type of set and
a single ip6?tables match which covers all egress/ingress filtering.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetfilter: ipset: Rewrite cidr book keeping to handle /0
Jozsef Kadlecsik [Sat, 22 Sep 2012 20:42:08 +0000 (22:42 +0200)]
netfilter: ipset: Rewrite cidr book keeping to handle /0

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agotcp: TCP Fast Open Server - call tcp_validate_incoming() for all packets
Neal Cardwell [Sat, 22 Sep 2012 04:18:57 +0000 (04:18 +0000)]
tcp: TCP Fast Open Server - call tcp_validate_incoming() for all packets

A TCP Fast Open (TFO) passive connection must call both
tcp_check_req() and tcp_validate_incoming() for all incoming ACKs that
are attempting to complete the 3WHS.

This is needed to parallel all the action that happens for a non-TFO
connection, where for an ACK that is attempting to complete the 3WHS
we call both tcp_check_req() and tcp_validate_incoming().

For example, upon receiving the ACK that completes the 3WHS, we need
to call tcp_fast_parse_options() and update ts_recent based on the
incoming timestamp value in the ACK.

One symptom of the problem with the previous code was that for passive
TFO connections using TCP timestamps, the outgoing TS ecr values
ignored the incoming TS val value on the ACK that completed the 3WHS.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: TCP Fast Open Server - note timestamps and retransmits for SYNACK RTT
Neal Cardwell [Sat, 22 Sep 2012 04:18:56 +0000 (04:18 +0000)]
tcp: TCP Fast Open Server - note timestamps and retransmits for SYNACK RTT

Previously, when using TCP Fast Open a server would return from
tcp_check_req() before updating snt_synack based on TCP timestamp echo
replies and whether or not we've retransmitted the SYNACK. The result
was that (a) for TFO connections using timestamps we used an incorrect
baseline SYNACK send time (tcp_time_stamp of SYNACK send instead of
rcv_tsecr), and (b) for TFO connections that do not have TCP
timestamps but retransmit the SYNACK we took a SYNACK RTT sample when
we should not take a sample.

This fix merely moves the snt_synack update logic a bit earlier in the
function, so that connections using TCP Fast Open will properly do
these updates when the ACK for the SYNACK arrives.

Moving this snt_synack update logic means that with TCP_DEFER_ACCEPT
enabled we do a few instructions of wasted work on each bare ACK, but
that seems OK.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: TCP Fast Open Server - take SYNACK RTT after completing 3WHS
Neal Cardwell [Sat, 22 Sep 2012 04:18:55 +0000 (04:18 +0000)]
tcp: TCP Fast Open Server - take SYNACK RTT after completing 3WHS

When taking SYNACK RTT samples for servers using TCP Fast Open, fix
the code to ensure that we only call tcp_valid_rtt_meas() after we
receive the ACK that completes the 3-way handshake.

Previously we were always taking an RTT sample in
tcp_v4_syn_recv_sock(). However, for TCP Fast Open connections
tcp_v4_conn_req_fastopen() calls tcp_v4_syn_recv_sock() at the time we
receive the SYN. So for TFO we must wait until tcp_rcv_state_process()
to take the RTT sample.

To fix this, we wait until after TFO calls tcp_v4_syn_recv_sock()
before we set the snt_synack timestamp, since tcp_synack_rtt_meas()
already ensures that we only take a SYNACK RTT sample if snt_synack is
non-zero. To be careful, we only take a snt_synack timestamp when
a SYNACK transmit or retransmit succeeds.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: extract code to compute SYNACK RTT
Neal Cardwell [Sat, 22 Sep 2012 04:18:54 +0000 (04:18 +0000)]
tcp: extract code to compute SYNACK RTT

In preparation for adding another spot where we compute the SYNACK
RTT, extract this code so that it can be shared.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoptp: clarify the clock_name sysfs attribute
Richard Cochran [Sat, 22 Sep 2012 07:02:04 +0000 (07:02 +0000)]
ptp: clarify the clock_name sysfs attribute

There has been some confusion among PHC driver authors about the
intended purpose of the clock_name attribute. This patch expands the
documation in order to clarify how the clock_name field should be
understood.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoptp: link the phc device to its parent device
Richard Cochran [Sat, 22 Sep 2012 07:02:03 +0000 (07:02 +0000)]
ptp: link the phc device to its parent device

PTP Hardware Clock devices appear as class devices in sysfs. This patch
changes the registration API to use the parent device, clarifying the
clock's relationship to the underlying device.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoptp: provide the clock's adjusted frequency
Richard Cochran [Sat, 22 Sep 2012 07:02:02 +0000 (07:02 +0000)]
ptp: provide the clock's adjusted frequency

If the timex.mode field indicates a query, then we provide the value of
the current frequency adjustment.

[ Get rid of extraneous empty lines -DaveM ]

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoptp: remember the adjusted frequency
Richard Cochran [Sat, 22 Sep 2012 07:02:01 +0000 (07:02 +0000)]
ptp: remember the adjusted frequency

This patch adds a field to the representation of a PTP hardware clock in
order to remember the frequency adjustment value dialed by the user.

Adding this field will let us answer queries in the manner of adjtimex
in a follow on patch.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Sat, 22 Sep 2012 19:38:25 +0000 (15:38 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
This series contains updates to igb only.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agocan: sja1000: Add support for listen-only mode and one-shot mode
Andreas Larsson [Thu, 20 Sep 2012 07:50:41 +0000 (09:50 +0200)]
can: sja1000: Add support for listen-only mode and one-shot mode

One-shot mode uses the TCS bit of the status register to discern
whether a transmission was successful or not. On a failed
transmission, the frame is not echoed back.

Signed-off-by: Andreas Larsson <andreas@gaisler.com>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agoigb: Use dma_unmap_addr and dma_unmap_len defines
Alexander Duyck [Tue, 18 Sep 2012 01:56:27 +0000 (01:56 +0000)]
igb: Use dma_unmap_addr and dma_unmap_len defines

This change is meant to improve performance on systems that do not require
the DMA unmap calls.  On those systems we do not need to make use of the
unmap address for Tx or the unmap length so we can drop both thereby
reducing the size of the Tx buffer info structure.

In addition I have changed the logic to check for unmap length instead of
unmap address when checking to see if a buffer needs to be unmapped from
DMA use.  The reasons for this change is that on some platforms it is
possible to receive a valid DMA address of 0 from an IOMMU.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Simplify how we populate the RSS key
Alexander Duyck [Thu, 13 Sep 2012 06:28:16 +0000 (06:28 +0000)]
igb: Simplify how we populate the RSS key

Instead of storing the RSS key as a character array we can simplify the
configuration by making it a u32 array.  This allows us to just write one
value per register without any unnecessary operations to construct the
value.

This change will produce the same exact key, the only difference is that I
translated the u8 array to a u32 array which will be correctly ordered on
writes to hardware by the cpu_to_le32 operations that are built into the
writel calls.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Change how we populate the RSS indirection table
Alexander Duyck [Thu, 13 Sep 2012 06:28:11 +0000 (06:28 +0000)]
igb: Change how we populate the RSS indirection table

This patch cleans up our RSS indirection table configuration so that we
generate the same table regardless of CPU endianness.  In addition it
changes the table setup so that instead of doing a modulo based setup it is
instead a divisor based setup.  The advantage to this is that we should be
able to take the Rx hash and compute the Rx queue with very little CPU
overhead if needed.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Change Tx cleanup loop to do/while instead of for
Alexander Duyck [Thu, 13 Sep 2012 06:28:01 +0000 (06:28 +0000)]
igb: Change Tx cleanup loop to do/while instead of for

This change makes it so that Tx cleanup is done in a do/while loop instead
of a for loop.  The main motivation behind this is the fact that we should
never be invoked with a budget less than 1 so we can skip checking the
budget before processing the first descriptor.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Remove logic that was doing NUMA pseudo-aware allocations
Alexander Duyck [Thu, 13 Sep 2012 06:27:55 +0000 (06:27 +0000)]
igb: Remove logic that was doing NUMA pseudo-aware allocations

This change removes the code that was doing the NUMA allocations for the
q_vectors, rings, and ring resources.  The problem is the logic used assumed
that the NUMA nodes were always interleved and that is not always the case.

At some point I hope to add this functionality back in a more controlled
manner in the future.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Fix stats output on i210/i211 parts.
Carolyn Wyborny [Wed, 12 Sep 2012 04:36:24 +0000 (04:36 +0000)]
igb: Fix stats output on i210/i211 parts.

Due to a hardware issue, on i210 and i211 parts, the TNCRS statistic
provides an invalid value.  This patch changes the update stats function
to increment the stat only for non-i210/i211 parts.

Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Change how we check for pre-existing and assigned VFs
Stefan Assmann [Sat, 18 Aug 2012 04:06:11 +0000 (04:06 +0000)]
igb: Change how we check for pre-existing and assigned VFs

Adapt the pre-existing and assigned VFs code to the ixgbe way introduced
in commit 9297127b9cdd8d30c829ef5fd28b7cc0323a7bcd.

Instead of searching the enabled VFs we use pci_num_vf to determine enabled VFs.
By comparing to which PF an assigned VF is owned it's possible to decide
whether to leave it enabled or not.

Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Acked-by: Greg Rose <gregory.v.rose@intel.com>
Tested-by: Robert Garrett <robertx.e.garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agocan: mscan-mpc5xxx: fix return value check in mpc512x_can_get_clock()
Wei Yongjun [Fri, 21 Sep 2012 07:09:47 +0000 (15:09 +0800)]
can: mscan-mpc5xxx: fix return value check in mpc512x_can_get_clock()

In case of error, the function clk_get() returns ERR_PTR()
and never returns NULL pointer. The NULL test in the error
handling should be replaced with IS_ERR().

dpatch engine is used to auto generated this patch.
(https://github.com/weiyj/dpatch)

Cc: stable <stable@vger.kernel.org>
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: usb: peak: rename peak_usb dump_mem function
Randy Dunlap [Sun, 2 Sep 2012 17:13:11 +0000 (10:13 -0700)]
can: usb: peak: rename peak_usb dump_mem function

Rename generic-sounding function dump_mem() to pcan_dump_mem()
so that it does not conflict with the dump_mem() function in
arch/sh/include/asm/kdebug.h.

drivers/net/can/usb/peak_usb/pcan_usb_core.c: error: conflicting types for 'dump_mem':  => 56:6
drivers/net/can/usb/peak_usb/pcan_usb_core.h: error: conflicting types for 'dump_mem':  => 134:6

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Stephane Grosjean <s.grosjean@peak-system.com>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
[mkl: convert all users of dump_mem(), too]
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: c_can: Adopt pinctrl support
AnilKumar Ch [Wed, 5 Sep 2012 11:12:41 +0000 (16:42 +0530)]
can: c_can: Adopt pinctrl support

Adopt pinctrl support to c_can driver based on c_can device
pointer, pinctrl driver configure SoC pins to d_can mode
according to definitions provided in .dts file.

In device specific device tree file 'pinctrl-names = "default";'
and 'pinctrl-0 = <&d_can1_pins>;' needs to add to configure pins
from c_can driver. d_can1_pins node contains the pinmux/config
details of d_can L/H pins.

Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: c_can: Add d_can suspend resume support
AnilKumar Ch [Fri, 21 Sep 2012 09:59:01 +0000 (15:29 +0530)]
can: c_can: Add d_can suspend resume support

Adds suspend resume support to DCAN driver which enables
DCAN power down mode bit (PDR). Then DCAN will ack the local
power-down mode by setting PDA bit in STATUS register.

Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: c_can: Add runtime PM support to Bosch C_CAN/D_CAN controller
AnilKumar Ch [Mon, 20 Aug 2012 11:20:54 +0000 (16:50 +0530)]
can: c_can: Add runtime PM support to Bosch C_CAN/D_CAN controller

Add Runtime PM support to C_CAN/D_CAN controller. The runtime PM
APIs control clocks for C_CAN/D_CAN IP and prevent access to the
register of C_CAN/D_CAN IP when clock is turned off.

Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: c_can: Add device tree support to Bosch C_CAN/D_CAN controller
AnilKumar Ch [Thu, 2 Aug 2012 13:13:10 +0000 (18:43 +0530)]
can: c_can: Add device tree support to Bosch C_CAN/D_CAN controller

Add device tree support to C_CAN/D_CAN controller and usage details
are added to device tree documentation. Driver was tested on AM335x
EVM.

Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
For the of binding doc:
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agocan: c_can: Modify c_can device names
AnilKumar Ch [Thu, 2 Aug 2012 13:13:09 +0000 (18:43 +0530)]
can: c_can: Modify c_can device names

Modify c_can device names from *_CAN_DEVTYPE to BOSCH_*_CAN to make
use of same names for array indexes in c_can_id_table[] as well as
device names.

This patch also add indexes to c_can_id_table array.

Signed-off-by: AnilKumar Ch <anilkumar@ti.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
12 years agonetfilter: ipset: Check and reject crazy /0 input parameters
Jozsef Kadlecsik [Tue, 4 Sep 2012 15:45:59 +0000 (17:45 +0200)]
netfilter: ipset: Check and reject crazy /0 input parameters

bitmap:ip and bitmap:ip,mac type did not reject such a crazy range
when created and using such a set results in a kernel crash.
The hash types just silently ignored such parameters.

Reject invalid /0 input parameters explicitely.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetfilter: ipset: Fix sparse warnings "incorrect type in assignment"
Jozsef Kadlecsik [Fri, 21 Sep 2012 19:44:58 +0000 (21:44 +0200)]
netfilter: ipset: Fix sparse warnings "incorrect type in assignment"

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
12 years agonetlink: use <linux/export.h> instead of <linux/module.h>
Pablo Neira Ayuso [Fri, 21 Sep 2012 09:35:38 +0000 (09:35 +0000)]
netlink: use <linux/export.h> instead of <linux/module.h>

Since (9f00d97 netlink: hide struct module parameter in netlink_kernel_create),
linux/netlink.h includes linux/module.h because of the use of THIS_MODULE.

Use linux/export.h instead, as suggested by Stephen Rothwell, which is
significantly smaller and defines THIS_MODULES.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosunbmac: Remove unused local variable.
David S. Miller [Fri, 21 Sep 2012 19:23:12 +0000 (15:23 -0400)]
sunbmac: Remove unused local variable.

Commit eb716c54b1c71ad28ab20461bff831bd481066c4 ("sunbmac: remove
unnecessary setting of skb->dev") caused the local varible 'dev'
in bigmac_init_rings to become unused.  And now the compiler
warns about it.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoteam: send port changed when added
Jiri Pirko [Fri, 21 Sep 2012 01:51:59 +0000 (01:51 +0000)]
team: send port changed when added

On some hw, link is not up during adding iface to team. That causes event
not being sent to userspace and that may cause confusion.
Fix this bug by sending port changed event once it's added to team.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipconfig: add nameserver IPs to kernel-parameter ip=
Christoph Fritz [Fri, 21 Sep 2012 08:31:19 +0000 (08:31 +0000)]
ipconfig: add nameserver IPs to kernel-parameter ip=

On small systems (e.g. embedded ones) IP addresses are often configured
by bootloaders and get assigned to kernel via parameter "ip=".  If set to
"ip=dhcp", even nameserver entries from DHCP daemons are handled. These
entries exported in /proc/net/pnp are commonly linked by /etc/resolv.conf.

To configure nameservers for networks without DHCP, this patch adds option
<dns0-ip> and <dns1-ip> to kernel-parameter 'ip='.

Signed-off-by: Christoph Fritz <chf.fritz@googlemail.com>
Tested-by: Jan Weitzel <j.weitzel@phytec.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: qmi_wwan: adding Huawei E367, ZTE MF683 and Pantech P4200
Bjørn Mork [Wed, 19 Sep 2012 21:18:05 +0000 (21:18 +0000)]
net: qmi_wwan: adding Huawei E367, ZTE MF683 and Pantech P4200

One of the modes of Huawei E367 has this QMI/wwan interface:

 I:* If#= 1 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=01 Prot=07 Driver=(none)
 E:  Ad=83(I) Atr=03(Int.) MxPS=  64 Ivl=2ms
 E:  Ad=84(I) Atr=02(Bulk) MxPS= 512 Ivl=0ms
 E:  Ad=02(O) Atr=02(Bulk) MxPS= 512 Ivl=4ms

Huawei use subclass and protocol to identify vendor specific
functions, so adding a new vendor rule for this combination.

The Pantech devices UML290 (106c:3718) and P4200 (106c:3721) use
the same subclass to identify the QMI/wwan function.  Replace the
existing device specific UML290 entries with generic vendor matching,
adding support for the Pantech P4200.

The ZTE MF683 has 6 vendor specific interfaces, all using
ff/ff/ff for cls/sub/prot.  Adding a match on interface #5 which
is a QMI/wwan interface.

Cc: Fangxiaozhi (Franko) <fangxiaozhi@huawei.com>
Cc: Thomas Schäfer <tschaefer@t-online.de>
Cc: Dan Williams <dcbw@redhat.com>
Cc: Shawn J. Goff <shawn7400@gmail.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agol2tp: fix compile error when CONFIG_IPV6=m and CONFIG_L2TP=y
Amerigo Wang [Thu, 20 Sep 2012 20:36:06 +0000 (20:36 +0000)]
l2tp: fix compile error when CONFIG_IPV6=m and CONFIG_L2TP=y

When CONFIG_IPV6=m and CONFIG_L2TP=y, I got the following compile error:

  LD      init/built-in.o
net/built-in.o: In function `l2tp_xmit_core':
l2tp_core.c:(.text+0x147781): undefined reference to `inet6_csk_xmit'
net/built-in.o: In function `l2tp_tunnel_create':
(.text+0x149067): undefined reference to `udpv6_encap_enable'
net/built-in.o: In function `l2tp_ip6_recvmsg':
l2tp_ip6.c:(.text+0x14e991): undefined reference to `ipv6_recv_error'
net/built-in.o: In function `l2tp_ip6_sendmsg':
l2tp_ip6.c:(.text+0x14ec64): undefined reference to `fl6_sock_lookup'
l2tp_ip6.c:(.text+0x14ed6b): undefined reference to `datagram_send_ctl'
l2tp_ip6.c:(.text+0x14eda0): undefined reference to `fl6_sock_lookup'
l2tp_ip6.c:(.text+0x14ede5): undefined reference to `fl6_merge_options'
l2tp_ip6.c:(.text+0x14edf4): undefined reference to `ipv6_fixup_options'
l2tp_ip6.c:(.text+0x14ee5d): undefined reference to `fl6_update_dst'
l2tp_ip6.c:(.text+0x14eea3): undefined reference to `ip6_dst_lookup_flow'
l2tp_ip6.c:(.text+0x14eee7): undefined reference to `ip6_dst_hoplimit'
l2tp_ip6.c:(.text+0x14ef8b): undefined reference to `ip6_append_data'
l2tp_ip6.c:(.text+0x14ef9d): undefined reference to `ip6_flush_pending_frames'
l2tp_ip6.c:(.text+0x14efe2): undefined reference to `ip6_push_pending_frames'
net/built-in.o: In function `l2tp_ip6_destroy_sock':
l2tp_ip6.c:(.text+0x14f090): undefined reference to `ip6_flush_pending_frames'
l2tp_ip6.c:(.text+0x14f0a0): undefined reference to `inet6_destroy_sock'
net/built-in.o: In function `l2tp_ip6_connect':
l2tp_ip6.c:(.text+0x14f14d): undefined reference to `ip6_datagram_connect'
net/built-in.o: In function `l2tp_ip6_bind':
l2tp_ip6.c:(.text+0x14f4fe): undefined reference to `ipv6_chk_addr'
net/built-in.o: In function `l2tp_ip6_init':
l2tp_ip6.c:(.init.text+0x73fa): undefined reference to `inet6_add_protocol'
l2tp_ip6.c:(.init.text+0x740c): undefined reference to `inet6_register_protosw'
net/built-in.o: In function `l2tp_ip6_exit':
l2tp_ip6.c:(.exit.text+0x1954): undefined reference to `inet6_unregister_protosw'
l2tp_ip6.c:(.exit.text+0x1965): undefined reference to `inet6_del_protocol'
net/built-in.o:(.rodata+0xf2d0): undefined reference to `inet6_release'
net/built-in.o:(.rodata+0xf2d8): undefined reference to `inet6_bind'
net/built-in.o:(.rodata+0xf308): undefined reference to `inet6_ioctl'
net/built-in.o:(.data+0x1af40): undefined reference to `ipv6_setsockopt'
net/built-in.o:(.data+0x1af48): undefined reference to `ipv6_getsockopt'
net/built-in.o:(.data+0x1af50): undefined reference to `compat_ipv6_setsockopt'
net/built-in.o:(.data+0x1af58): undefined reference to `compat_ipv6_getsockopt'
make: *** [vmlinux] Error 1

This is due to l2tp uses symbols from IPV6, so when IPV6
is a module, l2tp is not allowed to be builtin.

Cc: David Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetfilter: combine ipt_REDIRECT and ip6t_REDIRECT
Jan Engelhardt [Fri, 21 Sep 2012 09:41:34 +0000 (11:41 +0200)]
netfilter: combine ipt_REDIRECT and ip6t_REDIRECT

Combine more modules since the actual code is so small anyway that the
kmod metadata and the module in its loaded state totally outweighs the
combined actual code size.

IP_NF_TARGET_REDIRECT becomes a compat option; IP6_NF_TARGET_REDIRECT
is completely eliminated since it has not see a release yet.

Signed-off-by: Jan Engelhardt <jengelh@inai.de>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: combine ipt_NETMAP and ip6t_NETMAP
Jan Engelhardt [Fri, 21 Sep 2012 09:37:59 +0000 (11:37 +0200)]
netfilter: combine ipt_NETMAP and ip6t_NETMAP

Combine more modules since the actual code is so small anyway that the
kmod metadata and the module in its loaded state totally outweighs the
combined actual code size.

IP_NF_TARGET_NETMAP becomes a compat option; IP6_NF_TARGET_NETMAP
is completely eliminated since it has not see a release yet.

Signed-off-by: Jan Engelhardt <jengelh@inai.de>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_nat: remove obsolete rcu_read_unlock call
Ulrich Weber [Thu, 20 Sep 2012 03:52:04 +0000 (03:52 +0000)]
netfilter: nf_nat: remove obsolete rcu_read_unlock call

hlist walk in find_appropriate_src() is not protected anymore by rcu_read_lock(),
so rcu_read_unlock() is unnecessary if in_range() matches.

This bug was added in (c7232c9 netfilter: add protocol independent NAT core).

Signed-off-by: Ulrich Weber <ulrich.weber@sophos.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_nat: fix oops when unloading protocol modules
Patrick McHardy [Wed, 19 Sep 2012 20:57:04 +0000 (20:57 +0000)]
netfilter: nf_nat: fix oops when unloading protocol modules

When unloading a protocol module nf_ct_iterate_cleanup() is used to
remove all conntracks using the protocol from the bysource hash and
clean their NAT sections. Since the conntrack isn't actually killed,
the NAT callback is invoked twice, once for each direction, which
causes an oops when trying to delete it from the bysource hash for
the second time.

The same oops can also happen when removing both an L3 and L4 protocol
since the cleanup function doesn't check whether the conntrack has
already been cleaned up.

Pid: 4052, comm: modprobe Not tainted 3.6.0-rc3-test-nat-unload-fix+ #32 Red Hat KVM
RIP: 0010:[<ffffffffa002c303>]  [<ffffffffa002c303>] nf_nat_proto_clean+0x73/0xd0 [nf_nat]
RSP: 0018:ffff88007808fe18  EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8800728550c0 RCX: ffff8800756288b0
RDX: dead000000200200 RSI: ffff88007808fe88 RDI: ffffffffa002f208
RBP: ffff88007808fe28 R08: ffff88007808e000 R09: 0000000000000000
R10: dead000000200200 R11: dead000000100100 R12: ffffffff81c6dc00
R13: ffff8800787582b8 R14: ffff880078758278 R15: ffff88007808fe88
FS:  00007f515985d700(0000) GS:ffff88007cd00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00007f515986a000 CR3: 000000007867a000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process modprobe (pid: 4052, threadinfo ffff88007808e000, task ffff8800756288b0)
Stack:
 ffff88007808fe68 ffffffffa002c290 ffff88007808fe78 ffffffff815614e3
 ffffffff00000000 00000aeb00000246 ffff88007808fe68 ffffffff81c6dc00
 ffff88007808fe88 ffffffffa00358a0 0000000000000000 000000000040f5b0
Call Trace:
 [<ffffffffa002c290>] ? nf_nat_net_exit+0x50/0x50 [nf_nat]
 [<ffffffff815614e3>] nf_ct_iterate_cleanup+0xc3/0x170
 [<ffffffffa002c55a>] nf_nat_l3proto_unregister+0x8a/0x100 [nf_nat]
 [<ffffffff812a0303>] ? compat_prepare_timeout+0x13/0xb0
 [<ffffffffa0035848>] nf_nat_l3proto_ipv4_exit+0x10/0x23 [nf_nat_ipv4]
 ...

To fix this,

- check whether the conntrack has already been cleaned up in
  nf_nat_proto_clean

- change nf_ct_iterate_cleanup() to only invoke the callback function
  once for each conntrack (IP_CT_DIR_ORIGINAL).

The second change doesn't affect other callers since when conntracks are
actually killed, both directions are removed from the hash immediately
and the callback is already only invoked once. If it is not killed, the
second callback invocation will always return the same decision not to
kill it.

Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: fix IPv6 NAT dependencies in Kconfig
Pablo Neira Ayuso [Tue, 18 Sep 2012 19:03:39 +0000 (21:03 +0200)]
netfilter: fix IPv6 NAT dependencies in Kconfig

* NF_NAT_IPV6 requires IP6_NF_IPTABLES

* IP6_NF_TARGET_MASQUERADE, IP6_NF_TARGET_NETMAP, IP6_NF_TARGET_REDIRECT
  and IP6_NF_TARGET_NPT require NF_NAT_IPV6.

This change just mirrors what IPv4 does in Kconfig, for consistency.

Reported-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Fri, 21 Sep 2012 02:33:42 +0000 (22:33 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
This series contains updates to igb and ixgbevf.

v2: updated patch description in 04 patch (ixgbevf: scheduling while
    atomic in reset hw path)
 ...
Akeem G. Abodunrin (1):
  igb: Support to enable EEE on all eee_supported devices

Alexander Duyck (2):
  igb: Remove artificial restriction on RQDPC stat reading
  ixgbevf: Add support for VF API negotiation

John Fastabend (1):
  ixgbevf: scheduling while atomic in reset hw path
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet1080: Neaten netdev_dbg use
Joe Perches [Thu, 20 Sep 2012 14:10:23 +0000 (14:10 +0000)]
net1080: Neaten netdev_dbg use

Remove unnecessary temporary variable and #ifdef DEBUG block.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoUSB: remove dbg() usage in USB networking drivers
Greg Kroah-Hartman [Wed, 19 Sep 2012 09:46:14 +0000 (09:46 +0000)]
USB: remove dbg() usage in USB networking drivers

The dbg() USB macro is so old, it predates me.  The USB networking drivers are
the last hold-out using this macro, and we want to get rid of it, so replace
the usage of it with the proper netdev_dbg() or dev_dbg() (depending on the
context) calls.

Some places we end up using a local variable for the debug call, so also
convert the other existing dev_* calls to use it as well, to save tiny amounts
of code space.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: Document use of undefined variable.
Alan Cox [Wed, 19 Sep 2012 14:46:06 +0000 (15:46 +0100)]
tcp: Document use of undefined variable.

Both tcp_timewait_state_process and tcp_check_req use the same basic
construct of

struct tcp_options received tmp_opt;
tmp_opt.saw_tstamp = 0;

then call

tcp_parse_options

However if they are fed a frame containing a TCP_SACK then tbe code
behaviour is undefined because opt_rx->sack_ok is undefined data.

This ought to be documented if it is intentional.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv4: Don't add TCP-code in inet_sock_destruct
Christoph Paasch [Tue, 18 Sep 2012 14:19:23 +0000 (14:19 +0000)]
ipv4: Don't add TCP-code in inet_sock_destruct

Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
Acked-by: H.K. Jerry Chu <hkchu@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoIB/ipoib: Add rtnl_link_ops support
Or Gerlitz [Thu, 13 Sep 2012 05:56:36 +0000 (05:56 +0000)]
IB/ipoib: Add rtnl_link_ops support

Add rtnl_link_ops to IPoIB, with the first usage being child device
create/delete through them. Childs devices are now either legacy ones,
created/deleted through the ipoib sysfs entries, or RTNL ones.

Adding support for RTNL childs involved refactoring of ipoib_vlan_add
which is now used by both the sysfs and the link_ops code.

Also, added ndo_uninit entry to support calling unregister_netdevice_queue
from the rtnl dellink entry. This required removal of calls to
ipoib_dev_cleanup from the driver in flows which use unregister_netdevice,
since the networking core will invoke ipoib_uninit which does exactly that.

Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/bwh/sfc...
David S. Miller [Thu, 20 Sep 2012 20:39:59 +0000 (16:39 -0400)]
Merge branch 'for-davem' of git://git./linux/kernel/git/bwh/sfc-next

Ben Hutchings says:

====================
1. Extension to PPS/PTP to allow for PHC devices where pulses are
   subject to a variable but measurable delay.
2. PPS/PTP/PHC support for Solarflare boards with a timestamping
   peripheral.
3. MTD support for updating the timestamping peripheral on those boards.
4. Fix for potential over-length requests to firmware.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoixgbevf: scheduling while atomic in reset hw path
John Fastabend [Wed, 19 Sep 2012 21:35:57 +0000 (21:35 +0000)]
ixgbevf: scheduling while atomic in reset hw path

In ixgbevf_reset_hw_vf() msleep is called while holding mbx_lock
resulting in a schedule while atomic bug with trace below.

This patch uses mdelay instead.

BUG: scheduling while atomic: ip/6539/0x00000002
2 locks held by ip/6539:
 #0:  (rtnl_mutex){+.+.+.}, at: [<ffffffff81419cc3>] rtnl_lock+0x17/0x19
 #1:  (&(&adapter->mbx_lock)->rlock){+.+...}, at: [<ffffffffa0030855>] ixgbevf_reset+0x30/0xc1 [ixgbevf]
Modules linked in: ixgbevf ixgbe mdio libfc scsi_transport_fc 8021q scsi_tgt garp stp llc cpufreq_ondemand acpi_cpufreq freq_table mperf ipv6 uinput igb coretemp hwmon crc32c_intel ioatdma i2c_i801 shpchp microcode lpc_ich mfd_core i2c_core joydev dca pcspkr serio_raw pata_acpi ata_generic usb_storage pata_jmicron
Pid: 6539, comm: ip Not tainted 3.6.0-rc3jk-net-next+ #104
Call Trace:
 [<ffffffff81072202>] __schedule_bug+0x6a/0x79
 [<ffffffff814bc7e0>] __schedule+0xa2/0x684
 [<ffffffff8108f85f>] ? trace_hardirqs_off+0xd/0xf
 [<ffffffff814bd0c0>] schedule+0x64/0x66
 [<ffffffff814bb5e2>] schedule_timeout+0xa6/0xca
 [<ffffffff810536b9>] ? lock_timer_base+0x52/0x52
 [<ffffffff812629e0>] ? __udelay+0x15/0x17
 [<ffffffff814bb624>] schedule_timeout_uninterruptible+0x1e/0x20
 [<ffffffff810541c0>] msleep+0x1b/0x22
 [<ffffffffa002e723>] ixgbevf_reset_hw_vf+0x90/0xe5 [ixgbevf]
 [<ffffffffa0030860>] ixgbevf_reset+0x3b/0xc1 [ixgbevf]
 [<ffffffffa0032fba>] ixgbevf_open+0x43/0x43e [ixgbevf]
 [<ffffffff81409610>] ? dev_set_rx_mode+0x2e/0x33
 [<ffffffff8140b0f1>] __dev_open+0xa0/0xe5
 [<ffffffff814097ed>] __dev_change_flags+0xbe/0x142
 [<ffffffff8140b01c>] dev_change_flags+0x21/0x56
 [<ffffffff8141a843>] do_setlink+0x2e2/0x7f4
 [<ffffffff81016e36>] ? native_sched_clock+0x37/0x39
 [<ffffffff8141b0ac>] rtnl_newlink+0x277/0x4bb
 [<ffffffff8141aee9>] ? rtnl_newlink+0xb4/0x4bb
 [<ffffffff812217d1>] ? selinux_capable+0x32/0x3a
 [<ffffffff8104fb17>] ? ns_capable+0x4f/0x67
 [<ffffffff81419cc3>] ? rtnl_lock+0x17/0x19
 [<ffffffff81419f28>] rtnetlink_rcv_msg+0x236/0x253
 [<ffffffff81419cf2>] ? rtnetlink_rcv+0x2d/0x2d
 [<ffffffff8142fd42>] netlink_rcv_skb+0x43/0x94
 [<ffffffff81419ceb>] rtnetlink_rcv+0x26/0x2d
 [<ffffffff8142faf1>] netlink_unicast+0xee/0x174
 [<ffffffff81430327>] netlink_sendmsg+0x26a/0x288
 [<ffffffff813fb04f>] ? rcu_read_unlock+0x56/0x67
 [<ffffffff813f5e6d>] __sock_sendmsg_nosec+0x58/0x61
 [<ffffffff813f81b7>] __sock_sendmsg+0x3d/0x48
 [<ffffffff813f8339>] sock_sendmsg+0x6e/0x87
 [<ffffffff81107c9f>] ? might_fault+0xa5/0xac
 [<ffffffff81402a72>] ? copy_from_user+0x2a/0x2c
 [<ffffffff81402e62>] ? verify_iovec+0x54/0xaa
 [<ffffffff813f9834>] __sys_sendmsg+0x206/0x288
 [<ffffffff810694fa>] ? up_read+0x23/0x3d
 [<ffffffff811307e5>] ? fcheck_files+0xac/0xea
 [<ffffffff8113095e>] ? fget_light+0x3a/0xb9
 [<ffffffff813f9a2e>] sys_sendmsg+0x42/0x60
 [<ffffffff814c5ba9>] system_call_fastpath+0x16/0x1b

CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-By: Robert Garrett <robertx.e.garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoixgbevf: Add support for VF API negotiation
Alexander Duyck [Fri, 20 Jul 2012 08:09:58 +0000 (08:09 +0000)]
ixgbevf: Add support for VF API negotiation

This change makes it so that the VF can support the PF/VF API negotiation
protocol.  Specifically in this case we are adding support for API 1.0
which will mean that the VF is capable of cleaning up buffers that span
multiple descriptors without triggering an error.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Support to enable EEE on all eee_supported devices
Akeem G. Abodunrin [Thu, 6 Sep 2012 01:28:31 +0000 (01:28 +0000)]
igb: Support to enable EEE on all eee_supported devices

Current implementation enables EEE on only i350 device. This patch enables
EEE on all eee_supported devices. Also, configured LPI clock to keep
running before EEE is enabled on i210 and i211 devices.

Signed-off-by: Akeem G. Abodunrin <akeem.g.abodunrin@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agoigb: Remove artificial restriction on RQDPC stat reading
Alexander Duyck [Wed, 8 Aug 2012 05:23:22 +0000 (05:23 +0000)]
igb: Remove artificial restriction on RQDPC stat reading

For some reason the reading of the RQDPC register was being artificially
limited to 4K.  Instead of limiting the value we should read the value and
add the full amount.  Otherwise this can lead to a misleading number of
dropped packets when the actual value is in fact much higher.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
12 years agor8169: use unlimited DMA burst for TX
Michal Schmidt [Sun, 9 Sep 2012 13:55:26 +0000 (13:55 +0000)]
r8169: use unlimited DMA burst for TX

The r8169 driver currently limits the DMA burst for TX to 1024 bytes. I have
a box where this prevents the interface from using the gigabit line to its full
potential. This patch solves the problem by setting TX_DMA_BURST to unlimited.

The box has an ASRock B75M motherboard with on-board RTL8168evl/8111evl
(XID 0c900880). TSO is enabled.

I used netperf (TCP_STREAM test) to measure the dependency of TX throughput
on MTU. I did it for three different values of TX_DMA_BURST ('5'=512, '6'=1024,
'7'=unlimited). This chart shows the results:
http://michich.fedorapeople.org/r8169/r8169-effects-of-TX_DMA_BURST.png

Interesting points:
 - With the current DMA burst limit (1024):
   - at the default MTU=1500 I get only 842 Mbit/s.
   - when going from small MTU, the performance rises monotonically with
     increasing MTU only up to a peak at MTU=1076 (908 MBit/s). Then there's
     a sudden drop to 762 MBit/s from which the throughput rises monotonically
     again with further MTU increases.
 - With a smaller DMA burst limit (512):
   - there's a similar peak at MTU=1076 and another one at MTU=564.
 - With unlimited DMA burst:
   - at the default MTU=1500 I get nice 940 Mbit/s.
   - the throughput rises monotonically with increasing MTU with no strange
     peaks.

Notice that the peaks occur at MTU sizes that are multiples of the DMA burst
limit plus 52. Why 52? Because:
  20 (IP header) + 20 (TCP header) + 12 (TCP options) = 52

The Realtek-provided r8168 driver (v8.032.00) uses unlimited TX DMA burst too,
except for CFG_METHOD_1 where the TX DMA burst is set to 512 bytes.
CFG_METHOD_1 appears to be the oldest MAC version of "RTL8168B/8111B",
i.e. RTL_GIGA_MAC_VER_11 in r8169. Not sure if this MAC version really needs
the smaller burst limit, or if any other versions have similar requirements.

Signed-off-by: Michal Schmidt <mschmidt@redhat.com>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: unify fragment thresh handling code
Amerigo Wang [Tue, 18 Sep 2012 16:50:11 +0000 (16:50 +0000)]
ipv6: unify fragment thresh handling code

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Michal Kubeček <mkubecek@suse.cz>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: make ip6_frag_nqueues() and ip6_frag_mem() static inline
Amerigo Wang [Tue, 18 Sep 2012 16:50:10 +0000 (16:50 +0000)]
ipv6: make ip6_frag_nqueues() and ip6_frag_mem() static inline

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Michal Kubeček <mkubecek@suse.cz>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: unify conntrack reassembly expire code with standard one
Amerigo Wang [Tue, 18 Sep 2012 16:50:09 +0000 (16:50 +0000)]
ipv6: unify conntrack reassembly expire code with standard one

Two years ago, Shan Wei tried to fix this:
http://patchwork.ozlabs.org/patch/43905/

The problem is that RFC2460 requires an ICMP Time
Exceeded -- Fragment Reassembly Time Exceeded message should be
sent to the source of that fragment, if the defragmentation
times out.

"
   If insufficient fragments are received to complete reassembly of a
   packet within 60 seconds of the reception of the first-arriving
   fragment of that packet, reassembly of that packet must be
   abandoned and all the fragments that have been received for that
   packet must be discarded.  If the first fragment (i.e., the one
   with a Fragment Offset of zero) has been received, an ICMP Time
   Exceeded -- Fragment Reassembly Time Exceeded message should be
   sent to the source of that fragment.
"

As Herbert suggested, we could actually use the standard IPv6
reassembly code which follows RFC2460.

With this patch applied, I can see ICMP Time Exceeded sent
from the receiver when the sender sent out 3/4 fragmented
IPv6 UDP packet.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Michal Kubeček <mkubecek@suse.cz>
Cc: David Miller <davem@davemloft.net>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: netfilter-devel@vger.kernel.org
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: add a new namespace for nf_conntrack_reasm
Amerigo Wang [Tue, 18 Sep 2012 16:50:08 +0000 (16:50 +0000)]
ipv6: add a new namespace for nf_conntrack_reasm

As pointed by Michal, it is necessary to add a new
namespace for nf_conntrack_reasm code, this prepares
for the second patch.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Michal Kubeček <mkubecek@suse.cz>
Cc: David Miller <davem@davemloft.net>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: netfilter-devel@vger.kernel.org
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetpoll: call ->ndo_select_queue() in tx path
Amerigo Wang [Mon, 17 Sep 2012 20:16:31 +0000 (20:16 +0000)]
netpoll: call ->ndo_select_queue() in tx path

In netpoll tx path, we miss the chance of calling ->ndo_select_queue(),
thus could cause problems when bonding is involved.

This patch makes dev_pick_tx() extern (and rename it to netdev_pick_tx())
to let netpoll call it in netpoll_send_skb_on_dev().

Reported-by: Sylvain Munaut <s.munaut@whatever-company.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <amwang@redhat.com>
Tested-by: Sylvain Munaut <s.munaut@whatever-company.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetdev: make address const in device address management
stephen hemminger [Mon, 17 Sep 2012 10:03:26 +0000 (10:03 +0000)]
netdev: make address const in device address management

The internal functions for add/deleting addresses don't change
their argument.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoi825xx: znet: fix compiler warnings when building a 64-bit kernel
Mika Westerberg [Fri, 14 Sep 2012 00:33:48 +0000 (00:33 +0000)]
i825xx: znet: fix compiler warnings when building a 64-bit kernel

When building 64-bit kernel with this driver we get following warnings from
the compiler:

drivers/net/ethernet/i825xx/znet.c: In function ‘hardware_init’:
drivers/net/ethernet/i825xx/znet.c:863:29: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
drivers/net/ethernet/i825xx/znet.c:870:29: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

Fix these by calling isa_virt_to_bus() before passing the pointers to
set_dma_addr().

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agogre: add GSO support
Eric Dumazet [Thu, 13 Sep 2012 21:25:33 +0000 (21:25 +0000)]
gre: add GSO support

Add GSO support to GRE tunnels.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: provide a default dev->ethtool_ops
Eric Dumazet [Sun, 16 Sep 2012 09:17:26 +0000 (09:17 +0000)]
net: provide a default dev->ethtool_ops

Instead of forcing device drivers to provide empty ethtool_ops or tweak
net/core/ethtool.c again, we could provide a generic ethtool_ops.

This occurred to me when I wanted to add GSO support to GRE tunnels.
ethtool -k support should be generic for all drivers.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Maciej Żenczykowski <maze@google.com>
Reviewed-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: dev: fix incorrect getting net device's name
Gao feng [Thu, 13 Sep 2012 20:58:27 +0000 (20:58 +0000)]
net: dev: fix incorrect getting net device's name

When moving a nic from net namespace A to net namespace B,
in dev_change_net_namesapce,we call __dev_get_by_name to
decide if the netns B has the device has the same name.

if the netns B already has the same named device,we call
dev_get_valid_name to try to get a valid name for this nic in
the netns B,but net_device->nd_net still point to netns A now.

this patch fix it.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoipv6: recursive check rt->dst.from when call rt6_check_expired
Li RongQing [Thu, 13 Sep 2012 19:54:57 +0000 (19:54 +0000)]
ipv6: recursive check rt->dst.from when call rt6_check_expired

If dst cache dst_a copies from dst_b, and dst_b copies from dst_c, check
if dst_a is expired or not, we should not end with dst_a->dst.from, dst_b,
we should check dst_c.

CC: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: more accurate network taps in transmit path
Eric Dumazet [Tue, 18 Sep 2012 20:44:49 +0000 (20:44 +0000)]
net: more accurate network taps in transmit path

dev_queue_xmit_nit() should be called right before ndo_start_xmit()
calls or we might give wrong packet contents to taps users :

Packet checksum can be changed, or packet can be linearized or
segmented, and segments partially sent for the later case.

Also a memory allocation can fail and packet never really hit the
driver entry point.

Reported-by: Jamie Gloudon <jamie.gloudon@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosfc: Avoid generating over-length MC_CMD_FLUSH_RX_QUEUES request
Ben Hutchings [Wed, 19 Sep 2012 01:53:34 +0000 (02:53 +0100)]
sfc: Avoid generating over-length MC_CMD_FLUSH_RX_QUEUES request

MCDI supports requests up to 252 bytes long, which is only enough to
pass 63 RX queue IDs to MC_CMD_FLUSH_RX_QUEUES.  However a VF may have
up to 64 RX queues, and if we try to flush them all we will generate
an over-length request and BUG() in efx_mcdi_copyin().  Currently
all VF drivers limit themselves to 32 RX queues, so reducing the
limit to 63 does no harm.

Also add a BUILD_BUG_ON in efx_mcdi_flush_rxqs() so we remember to
deal with the same problem there if EFX_MAX_CHANNELS is increased.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Bump version to 3.2
Ben Hutchings [Tue, 17 Jul 2012 19:45:55 +0000 (20:45 +0100)]
sfc: Bump version to 3.2

The key new feature for 3.2 is PTP support.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Expose FPGA bitfile partition through MTD
Ben Hutchings [Mon, 4 Jul 2011 23:05:56 +0000 (00:05 +0100)]
sfc: Expose FPGA bitfile partition through MTD

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Support variable-length response to MCDI GET_BOARD_CFG
Ben Hutchings [Mon, 4 Jul 2011 23:04:42 +0000 (00:04 +0100)]
sfc: Support variable-length response to MCDI GET_BOARD_CFG

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Convert firmware subtypes to native byte order in efx_mcdi_get_board_cfg()
Ben Hutchings [Thu, 6 Sep 2012 23:58:10 +0000 (00:58 +0100)]
sfc: Convert firmware subtypes to native byte order in efx_mcdi_get_board_cfg()

On big-endian systems the MTD partition names currently have mangled
subtype numbers and are not recognised by the firmware update tool
(sfupdate).

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Add support for IEEE-1588 PTP
Stuart Hodgson [Mon, 3 Sep 2012 10:09:36 +0000 (11:09 +0100)]
sfc: Add support for IEEE-1588 PTP

Add PTP IEEE-1588 support and make accesible via the PHC subsystem.

This work is based on prior code by Andrew Jackson

Signed-off-by: Stuart Hodgson <smhodgson@solarflare.com>
[bwh:
 - Add byte order conversion in efx_ptp_send_times()
 - Simplify conversion of PPS event times
 - Add the built-in vs module check to CONFIG_SFC_PTP dependencies]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
12 years agosfc: Fix maximum array sizes for various MCDI commands
Ben Hutchings [Wed, 19 Sep 2012 01:46:37 +0000 (02:46 +0100)]
sfc: Fix maximum array sizes for various MCDI commands

The maximum array sizes have been calculated on the basis of a maximum
SDU size of 255 bytes, whereas the actual maximum is 252 bytes.
Constructing a larger SDU will result in a BUG_ON in efx_mcdi_copyin.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>