platform/kernel/linux-arm64.git
10 years agonet: Mark functions as static in core/dev.c
Rashika Kheria [Sun, 9 Feb 2014 14:56:25 +0000 (20:26 +0530)]
net: Mark functions as static in core/dev.c

Mark functions as static in core/dev.c because they are not used outside
this file.

This eliminates the following warning in core/dev.c:
net/core/dev.c:2806:5: warning: no previous prototype for ‘__dev_queue_xmit’ [-Wmissing-prototypes]
net/core/dev.c:4640:5: warning: no previous prototype for ‘netdev_adjacent_sysfs_add’ [-Wmissing-prototypes]
net/core/dev.c:4650:6: warning: no previous prototype for ‘netdev_adjacent_sysfs_del’ [-Wmissing-prototypes]

Signed-off-by: Rashika Kheria <rashika.kheria@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Include appropriate header file in caif/cfsrvl.c
Rashika Kheria [Sun, 9 Feb 2014 14:32:16 +0000 (20:02 +0530)]
net: Include appropriate header file in caif/cfsrvl.c

Include appropriate header file net/caif/caif_dev.h in caif/cfsrvl.c
because it has prototype declaration of functions defined in
caif/cfsrvl.c.

This eliminates the following warning in caif/cfsrvl.c:
net/caif/cfsrvl.c:198:6: warning: no previous prototype for ‘caif_free_client’ [-Wmissing-prototypes]
net/caif/cfsrvl.c:208:6: warning: no previous prototype for ‘caif_client_register_refcnt’ [-Wmissing-prototypes]

Signed-off-by: Rashika Kheria <rashika.kheria@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Include appropriate header file in caif/caif_dev.c
Rashika Kheria [Sun, 9 Feb 2014 14:29:04 +0000 (19:59 +0530)]
net: Include appropriate header file in caif/caif_dev.c

Include appropriate header file net/caif/caif_dev.h in caif/caif_dev.c
because it has prototype declarations of function defined in
caif/caif_dev.c.

This eliminates the following file in caif/caif_dev.c:
net/caif/caif_dev.c:303:6: warning: no previous prototype for ‘caif_enroll_dev’ [-Wmissing-prototypes]

Signed-off-by: Rashika Kheria <rashika.kheria@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Mark function as static in 9p/client.c
Rashika Kheria [Sun, 9 Feb 2014 14:27:33 +0000 (19:57 +0530)]
net: Mark function as static in 9p/client.c

Mark function as static in net/9p/client.c because it is not used
outside this file.

This eliminates the following warning in net/9p/client.c:
net/9p/client.c:207:18: warning: no previous prototype for ‘p9_fcall_alloc’ [-Wmissing-prototypes]

Signed-off-by: Rashika Kheria <rashika.kheria@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Sun, 9 Feb 2014 22:21:25 +0000 (14:21 -0800)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless

John W. Linville says:

====================
Please pull this batch of fixes intended for the 3.14 stream!

For the mac80211 bits, Johannes says:

"This is just a collection of small fixes, the commit logs explain the
details. The only thing that isn't strictly a fix is the 5/10 MHz
enabling, I had forgotten this and there's little point in waiting
longer. The patch simply removes the force-disable code that I put in
when there was a problem with the userspace API (that has long been
fixed.)"

For the iwlwifi bits, Emmanuel says:

"I have an important fix that disables A band in case the driver thought
it was enabled, and the firmware disagreed. We ended up making the
firmware unhappy. I also fix the station table in AP mode and fix the
scan while we have BT working.
Johannes removes a static variable that could potentially lead to to
issues on multi-device setups and disables scheduled scan to avoid
issues with old versions of wpa_supplicant.
A small fix from David on scan and a few new device IDs for 7265."

On top of that...

Oleksij Rempel adds a USB ID to the ar5523 driver and changes the
default powersave setting for ath9k_htc to "off", due to observed
stability issues (based on an equivalent ath9k patch).

Stanislaw Gruszka similarly disables powersave for a couple of rt2x00
drivers.  He also fixes a couple of scheduling while atomic issues
in ath9k_htc.

Sujith Manoharan rounds-out the powersave disables with one for ath9k.
He also fixes a build prolem with ath9k on ARM and fixes an ath9k Tx
power calculation.

Finally, Andrea Merello fixes a couple of lingering DMA mapping
problems in the rtl8180 driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Sun, 9 Feb 2014 22:20:00 +0000 (14:20 -0800)]
Merge branch 'master' of git://git./linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter/nftables/IPVS fixes for net

The following patchset contains Netfilter/IPVS fixes, mostly nftables
fixes, most relevantly they are:

* Fix a crash in the h323 conntrack NAT helper due to expectation list
  corruption, from Alexey Dobriyan.

* A couple of RCU race fixes for conntrack, one manifests by hitting BUG_ON
  in nf_nat_setup_info() and the destroy path, patches from Andrey Vagin and
  me.

* Dump direction attribute in nft_ct only if it is set, from Arturo
  Borrero.

* Fix IPVS bug in its own connection tracking system that may lead to
  copying only 4 bytes of the IPv6 address when initializing the
  ip_vs_conn object, from Michal Kubecek.

* Fix -EBUSY errors in nftables when deleting the rules, chain and tables
  in a row due mixture of asynchronous and synchronous object releasing,
  from me.

* Three fixes for the nf_tables set infrastructure when using intervals and
  mappings, from me.

* Four patches to fixing the nf_tables log, reject and ct expressions from
  the new inet table, from Patrick McHardy.

* Fix memory overrun in the map that is used to dynamically allocate names
  from anonymous sets, also from Patrick.

* Fix a potential oops if you dump a set with NFPROTO_UNSPEC and a table
  name, from Patrick McHardy.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Fri, 7 Feb 2014 18:44:14 +0000 (13:44 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless into for-davem

10 years agonetfilter: nf_tables: unininline nft_trace_packet()
Patrick McHardy [Thu, 9 Jan 2014 18:42:42 +0000 (18:42 +0000)]
netfilter: nf_tables: unininline nft_trace_packet()

It makes no sense to inline a rarely used function meant for debugging
only that is called a total of five times in the main evaluation loop.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_tables: fix loop checking with end interval elements
Pablo Neira Ayuso [Fri, 7 Feb 2014 13:45:01 +0000 (14:45 +0100)]
netfilter: nf_tables: fix loop checking with end interval elements

Fix access to uninitialized data for end interval elements. The
element data part is uninitialized in interval end elements.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nft_rbtree: fix data handling of end interval elements
Pablo Neira Ayuso [Thu, 6 Feb 2014 15:15:39 +0000 (16:15 +0100)]
netfilter: nft_rbtree: fix data handling of end interval elements

This patch fixes several things which related to the handling of
end interval elements:

* Chain use underflow with intervals and map: If you add a rule
  using intervals+map that introduces a loop, the error path of the
  rbtree set decrements the chain refcount for each side of the
  interval, leading to a chain use counter underflow.

* Don't copy the data part of the end interval element since, this
  area is uninitialized and this confuses the loop detection code.

* Don't allocate room for the data part of end interval elements
  since this is unused.

So, after this patch the idea is that end interval elements don't
have a data part.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Patrick McHardy <kaber@trash.net>
10 years agonetfilter: nf_tables: do not allow NFT_SET_ELEM_INTERVAL_END flag and data
Pablo Neira Ayuso [Fri, 7 Feb 2014 11:53:07 +0000 (12:53 +0100)]
netfilter: nf_tables: do not allow NFT_SET_ELEM_INTERVAL_END flag and data

This combination is not allowed since end interval elements cannot
contain data.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Patrick McHardy <kaber@trash.net>
10 years agotcp: remove 1ms offset in srtt computation
Eric Dumazet [Thu, 6 Feb 2014 23:57:10 +0000 (15:57 -0800)]
tcp: remove 1ms offset in srtt computation

TCP pacing depends on an accurate srtt estimation.

Current srtt estimation is using jiffie resolution,
and has an artificial offset of at least 1 ms, which can produce
slowdowns when FQ/pacing is used, especially in DC world,
where typical rtt is below 1 ms.

We are planning a switch to usec resolution for linux-3.15,
but in the meantime, this patch removes the 1 ms offset.

All we need is to have tp->srtt minimal value of 1 to differentiate
the case of srtt being initialized or not, not 8.

The problematic behavior was observed on a 40Gbit testbed,
where 32 concurrent netperf were reaching 12Gbps of aggregate
speed, instead of line speed.

This patch also has the effect of reporting more accurate srtt and send
rates to iproute2 ss command as in :

$ ss -i dst cca2
Netid  State      Recv-Q Send-Q          Local Address:Port
Peer Address:Port
tcp    ESTAB      0      0                10.244.129.1:56984
10.244.129.2:12865
 cubic wscale:6,6 rto:200 rtt:0.25/0.25 ato:40 mss:1448 cwnd:10 send
463.4Mbps rcv_rtt:1 rcv_space:29200
tcp    ESTAB      0      390960           10.244.129.1:60247
10.244.129.2:50204
 cubic wscale:6,6 rto:200 rtt:0.875/0.75 mss:1448 cwnd:73 ssthresh:51
send 966.4Mbps unacked:73 retrans:0/121 rcv_space:29200

Reported-by: Vytautas Valancius <valas@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: fix netconsole setup over bridge
Cong Wang [Thu, 6 Feb 2014 23:00:52 +0000 (15:00 -0800)]
bridge: fix netconsole setup over bridge

Commit 93d8bf9fb8f3 ("bridge: cleanup netpoll code") introduced
a check in br_netpoll_enable(), but this check is incorrect for
br_netpoll_setup(). This patch moves the code after the check
into __br_netpoll_enable() and calls it in br_netpoll_setup().
For br_add_if(), the check is still needed.

Fixes: 93d8bf9fb8f3 ("bridge: cleanup netpoll code")
Cc: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Tested-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: use __GFP_NORETRY for high order allocations
Eric Dumazet [Thu, 6 Feb 2014 18:42:42 +0000 (10:42 -0800)]
net: use __GFP_NORETRY for high order allocations

sock_alloc_send_pskb() & sk_page_frag_refill()
have a loop trying high order allocations to prepare
skb with low number of fragments as this increases performance.

Problem is that under memory pressure/fragmentation, this can
trigger OOM while the intent was only to try the high order
allocations, then fallback to order-0 allocations.

We had various reports from unexpected regressions.

According to David, setting __GFP_NORETRY should be fine,
as the asynchronous compaction is still enabled, and this
will prevent OOM from kicking as in :

CFSClientEventm invoked oom-killer: gfp_mask=0x42d0, order=3, oom_adj=0,
oom_score_adj=0, oom_score_badness=2 (enabled),memcg_scoring=disabled
CFSClientEventm

Call Trace:
 [<ffffffff8043766c>] dump_header+0xe1/0x23e
 [<ffffffff80437a02>] oom_kill_process+0x6a/0x323
 [<ffffffff80438443>] out_of_memory+0x4b3/0x50d
 [<ffffffff8043a4a6>] __alloc_pages_may_oom+0xa2/0xc7
 [<ffffffff80236f42>] __alloc_pages_nodemask+0x1002/0x17f0
 [<ffffffff8024bd23>] alloc_pages_current+0x103/0x2b0
 [<ffffffff8028567f>] sk_page_frag_refill+0x8f/0x160
 [<ffffffff80295fa0>] tcp_sendmsg+0x560/0xee0
 [<ffffffff802a5037>] inet_sendmsg+0x67/0x100
 [<ffffffff80283c9c>] __sock_sendmsg_nosec+0x6c/0x90
 [<ffffffff80283e85>] sock_sendmsg+0xc5/0xf0
 [<ffffffff802847b6>] __sys_sendmsg+0x136/0x430
 [<ffffffff80284ec8>] sys_sendmsg+0x88/0x110
 [<ffffffff80711472>] system_call_fastpath+0x16/0x1b
Out of Memory: Kill process 2856 (bash) score 9999 or sacrifice child

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetpoll: fix netconsole IPv6 setup
Sabrina Dubroca [Thu, 6 Feb 2014 17:34:12 +0000 (18:34 +0100)]
netpoll: fix netconsole IPv6 setup

Currently, to make netconsole start over IPv6, the source address
needs to be specified. Without a source address, netpoll_parse_options
assumes we're setting up over IPv4 and the destination IPv6 address is
rejected.

Check if the IP version has been forced by a source address before
checking for a version mismatch when parsing the destination address.

Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net: fix build warning in ethernet/sfc/tx.c
Paul Gortmaker [Thu, 6 Feb 2014 16:45:12 +0000 (11:45 -0500)]
drivers/net: fix build warning in ethernet/sfc/tx.c

Commit ee45fd92c739db5b7950163d91dfe5f016af6d24 ("sfc: Use TX PIO
for sufficiently small packets") introduced the following warning:

drivers/net/ethernet/sfc/tx.c: In function 'efx_enqueue_skb':
drivers/net/ethernet/sfc/tx.c:432:1: warning: label 'finish_packet' defined but not used

Stick the label inside the same #ifdef that the code which calls
it uses.  Note that this is only seen for arch that do not set
ARCH_HAS_IOREMAP_WC, such as arm, mips, sparc, ..., as the others
enable the write combining code and hence use the label.

Cc: Jon Cooper <jcooper@solarflare.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agohso: remove some dead code
Dan Carpenter [Thu, 6 Feb 2014 12:53:19 +0000 (15:53 +0300)]
hso: remove some dead code

It seems like this function was intended to have special handling for
urb statuses of -ENOENT and -ECONNRESET.  But now it just prints some
debugging and returns at the start of the function.

I have removed the dead code, it's still in the git history if anyone
wants to revive it.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoinet: defines IPPROTO_* needed for module alias generation
Jan Moskyto Matejka [Thu, 6 Feb 2014 11:10:00 +0000 (12:10 +0100)]
inet: defines IPPROTO_* needed for module alias generation

Commit cfd280c91253 ("net: sync some IP headers with glibc") changed a set of
define's to an enum (with no explanation why) which introduced a bug
in module mip6 where aliases are generated using the IPPROTO_* defines;
mip6 doesn't load if require_module called with the aliases from
xfrm_get_type().

Reverting this change back to define's to fix the aliases.

modinfo mip6 (before this change)
alias:          xfrm-type-10-IPPROTO_DSTOPTS
alias:          xfrm-type-10-IPPROTO_ROUTING

modinfo mip6 (after this change)
alias:          xfrm-type-10-43
alias:          xfrm-type-10-60

Signed-off-by: Jan Moskyto Matejka <mq@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/hisax: hex vs decimal typo in prfeatureind()
Dan Carpenter [Thu, 6 Feb 2014 08:03:08 +0000 (11:03 +0300)]
isdn/hisax: hex vs decimal typo in prfeatureind()

This is a static checker fix, but judging from the context then I think
hexidecimal 0x80 is intended here instead of decimal 80.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sctp: fix initialization of local source address on accepted ipv6 sockets
Matija Glavinic Pecotic [Thu, 6 Feb 2014 07:30:10 +0000 (08:30 +0100)]
net: sctp: fix initialization of local source address on accepted ipv6 sockets

commit  efe4208f47f907b86f528788da711e8ab9dea44d:
'ipv6: make lookups simpler and faster' broke initialization of local source
address on accepted ipv6 sockets. Before the mentioned commit receive address
was copied along with the contents of ipv6_pinfo in sctp_v6_create_accept_sk.
Now when it is moved, it has to be copied separately.

This also fixes lksctp's ipv6 regression in a sense that test_getname_v6, TC5 -
'getsockname on a connected server socket' now passes.

Signed-off-by: Matija Glavinic Pecotic <matija.glavinic-pecotic.ext@nsn.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: fix the submission of the interrupt transfer
hayeswang [Thu, 6 Feb 2014 03:55:48 +0000 (11:55 +0800)]
r8152: fix the submission of the interrupt transfer

The submission of the interrupt transfer should be done after setting
the bit of WORK_ENABLE, otherwise the callback function would have
the opportunity to be returned directly.

Clear the bit of WORK_ENABLE before killing the interrupt transfer.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: Allow VF rss on higher PFs
Yuval Mintz [Wed, 5 Feb 2014 14:07:12 +0000 (16:07 +0200)]
bnx2x: Allow VF rss on higher PFs

bnx2x driver uses incorrect PF identifier to configure (in HW) the VF
interrupt scheme; As a result, in multi-function mode the configuration
for PFs with a high index (4+) will overflow and the PF will erroneously
configure a single ISR scheme for its VFs.
As a result, if such a VF uses multiple queues, interrupt generation will
stop after VF receives an Rx packet or sends a Tx packet on a queue
other than queue[0].

Signed-off-by: Yuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Fix deadlock in tg3_change_mtu()
Nithin Sujir [Thu, 6 Feb 2014 22:13:05 +0000 (14:13 -0800)]
tg3: Fix deadlock in tg3_change_mtu()

Quoting David Vrabel -
"5780 cards cannot have jumbo frames and TSO enabled together.  When
jumbo frames are enabled by setting the MTU, the TSO feature must be
cleared.  This is done indirectly by calling netdev_update_features()
which will call tg3_fix_features() to actually clear the flags.

netdev_update_features() will also trigger a new netlink message for the
feature change event which will result in a call to tg3_get_stats64()
which deadlocks on the tg3 lock."

tg3_set_mtu() does not need to be under the tg3 lock since converting
the flags to use set_bit(). Move it out to after tg3_netif_stop().

Reported-by: David Vrabel <david.vrabel@citrix.com>
Tested-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: cleanup an error path in tg3_phy_reset_5703_4_5()
Dan Carpenter [Wed, 5 Feb 2014 13:29:21 +0000 (16:29 +0300)]
tg3: cleanup an error path in tg3_phy_reset_5703_4_5()

In the original code, if tg3_readphy() fails then it does an unnecessary
check to verify "err" is still zero and then returns -EBUSY.

My static checker complains about the unnecessary "if (!err)" check and
anyway it is better to propagate the -EBUSY error code from
tg3_readphy() instead of hard coding it here.  And really the original
code is confusing to look at.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv4: Fix runtime WARNING in rtmsg_ifa()
Geert Uytterhoeven [Wed, 5 Feb 2014 07:38:25 +0000 (08:38 +0100)]
ipv4: Fix runtime WARNING in rtmsg_ifa()

On m68k/ARAnyM:

WARNING: CPU: 0 PID: 407 at net/ipv4/devinet.c:1599 0x316a99()
Modules linked in:
CPU: 0 PID: 407 Comm: ifconfig Not tainted
3.13.0-atari-09263-g0c71d68014d1 #1378
Stack from 10c4fdf0:
        10c4fdf0 002ffabb 000243e8 00000000 008ced6c 00024416 00316a99 0000063f
        00316a99 00000009 00000000 002501b4 00316a99 0000063f c0a86117 00000080
        c0a86117 00ad0c90 00250a5a 00000014 00ad0c90 00000000 00000000 00000001
        00b02dd0 00356594 00000000 00356594 c0a86117 eff6c9e4 008ced6c 00000002
        008ced60 0024f9b4 00250b52 00ad0c90 00000000 00000000 00252390 00ad0c90
        eff6c9e4 0000004f 00000000 00000000 eff6c9e4 8000e25c eff6c9e4 80001020
Call Trace: [<000243e8>] warn_slowpath_common+0x52/0x6c
 [<00024416>] warn_slowpath_null+0x14/0x1a
 [<002501b4>] rtmsg_ifa+0xdc/0xf0
 [<00250a5a>] __inet_insert_ifa+0xd6/0x1c2
 [<0024f9b4>] inet_abc_len+0x0/0x42
 [<00250b52>] inet_insert_ifa+0xc/0x12
 [<00252390>] devinet_ioctl+0x2ae/0x5d6

Adding some debugging code reveals that net_fill_ifaddr() fails in

    put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
                              preferred, valid))

nla_put complains:

    lib/nlattr.c:454: skb_tailroom(skb) = 12, nla_total_size(attrlen) = 20

Apparently commit 5c766d642bcaffd0c2a5b354db2068515b3846cf ("ipv4:
introduce address lifetime") forgot to take into account the addition of
struct ifa_cacheinfo in inet_nlmsg_size(). Hence add it, like is already
done for ipv6.

Suggested-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2[x]: Make module parameters readable
James M Leddy [Tue, 4 Feb 2014 20:10:59 +0000 (15:10 -0500)]
bnx2[x]: Make module parameters readable

Occasionally users want to know what parameters their Broadcom drivers
are running with. For example, a user may want to know if MSI is
disabled.

This patch has been compile tested.

Signed-off-by: James M Leddy <james.leddy@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: irda: ep7211-sir: Remove driver
Alexander Shiyan [Tue, 4 Feb 2014 15:43:32 +0000 (19:43 +0400)]
net: irda: ep7211-sir: Remove driver

This patch removes old and unsupported CLPS711X IrDA driver.
Support for IrDA for CLPS711X serial port now provided by commit
4a33f1f59abd (serial: clps711x: Add support for N_IRDA line
discipline), so IrDA-mode can be turned ON with "irattach" tool
through "irtty" driver.

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoARM: sunxi: dt: Convert to the new net compatibles
Maxime Ripard [Sun, 2 Feb 2014 13:49:13 +0000 (14:49 +0100)]
ARM: sunxi: dt: Convert to the new net compatibles

Switch the device tree to the new compatibles introduced in the ethernet and
mdio drivers to have a common pattern accross all Allwinner SoCs.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: sunxi: Add new compatibles
Maxime Ripard [Sun, 2 Feb 2014 13:49:12 +0000 (14:49 +0100)]
net: phy: sunxi: Add new compatibles

The Allwinner A10 compatibles were following a slightly different compatible
patterns than the rest of the SoCs for historical reasons. Add compatibles
matching the other pattern to the mdio driver for consistency, and keep the
older one for backward compatibility.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: sunxi: Add new compatibles
Maxime Ripard [Sun, 2 Feb 2014 13:49:11 +0000 (14:49 +0100)]
net: ethernet: sunxi: Add new compatibles

The Allwinner A10 compatibles were following a slightly different compatible
patterns than the rest of the SoCs for historical reasons. Add compatibles
matching the other pattern to the ethernet driver for consistency, and keep the
older one for backward compatibility.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agortl8180: Add error check for pci_map_single return value in TX path
andrea.merello [Wed, 5 Feb 2014 21:38:06 +0000 (22:38 +0100)]
rtl8180: Add error check for pci_map_single return value in TX path

Orignal code will not detect a DMA mapping failure, causing the HW
to attempt a DMA from an invalid address.

This patch add the error check and eventually simply drops the TX
packet if we can't map it for DMA.

Signed-off-by: andrea merello <andrea.merello@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agortl8180: Add error check for pci_map_single return value in RX path
andrea.merello [Wed, 5 Feb 2014 21:38:05 +0000 (22:38 +0100)]
rtl8180: Add error check for pci_map_single return value in RX path

In original code the old RX DMA buffer is unmapped and processed and at the end
of the isr a new buffer is mapped with pci_map_single and attached to the RX
descriptor.

If pci_map_single fails then the RX descriptor remains with no valid DMA buffer
attached.
In this condition the DMA will target where it shouldn't with obvious evil
consequences.

Simply avoiding re-arming the descriptor will prevent buggy DMA but it will
result soon in RX stuck.

This patch move the DMA mapping of the new buffer at the beginning of the ISR
(and it adds error check for pci_map_single success/fail).

If the DMA mapping fails then we do not unmap the old buffer and we re-arm the
descriptor without processing it, with the old DMA buffer still attached.

In this way we lose the currently RX-ed packet, but whenever next calls to
pci_map_single will succeed again,then the RX process will go on without stuck.

Signed-off-by: andrea merello <andrea.merello@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211
John W. Linville [Thu, 6 Feb 2014 19:34:31 +0000 (14:34 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/jberg/mac80211

10 years agonetfilter: nf_tables: fix racy rule deletion
Pablo Neira Ayuso [Sat, 25 Jan 2014 13:03:51 +0000 (14:03 +0100)]
netfilter: nf_tables: fix racy rule deletion

We may lost race if we flush the rule-set (which happens asynchronously
via call_rcu) and we try to remove the table (that userspace assumes
to be empty).

Fix this by recovering synchronous rule and chain deletion. This was
introduced time ago before we had no batch support, and synchronous
rule deletion performance was not good. Now that we have the batch
support, we can just postpone the purge of old rule in a second step
in the commit phase. All object deletions are synchronous after this
patch.

As a side effect, we save memory as we don't need rcu_head per rule
anymore.

Cc: Patrick McHardy <kaber@trash.net>
Reported-by: Arturo Borrero Gonzalez <arturo.borrero.glez@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_tables: fix log/queue expressions for NFPROTO_INET
Patrick McHardy [Thu, 6 Feb 2014 09:17:41 +0000 (09:17 +0000)]
netfilter: nf_tables: fix log/queue expressions for NFPROTO_INET

The log and queue expressions both store the family during ->init() and
use it to deliver packets. This is wrong when used in NFPROTO_INET since
they should both deliver to the actual AF of the packet, not the dummy
NFPROTO_INET.

Use the family from the hook ops to fix this.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agomac80211: fix virtual monitor interface iteration
Johannes Berg [Wed, 29 Jan 2014 12:28:02 +0000 (13:28 +0100)]
mac80211: fix virtual monitor interface iteration

During channel context assignment, the interface should
be found by interface iteration, so we need to assign the
pointer before the channel context.

Reported-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Tested-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: fix fragmentation code, particularly for encryption
Johannes Berg [Fri, 31 Jan 2014 23:16:23 +0000 (00:16 +0100)]
mac80211: fix fragmentation code, particularly for encryption

The "new" fragmentation code (since my rewrite almost 5 years ago)
erroneously sets skb->len rather than using skb_trim() to adjust
the length of the first fragment after copying out all the others.
This leaves the skb tail pointer pointing to after where the data
originally ended, and thus causes the encryption MIC to be written
at that point, rather than where it belongs: immediately after the
data.

The impact of this is that if software encryption is done, then
 a) encryption doesn't work for the first fragment, the connection
    becomes unusable as the first fragment will never be properly
    verified at the receiver, the MIC is practically guaranteed to
    be wrong
 b) we leak up to 8 bytes of plaintext (!) of the packet out into
    the air

This is only mitigated by the fact that many devices are capable
of doing encryption in hardware, in which case this can't happen
as the tail pointer is irrelevant in that case. Additionally,
fragmentation is not used very frequently and would normally have
to be configured manually.

Fix this by using skb_trim() properly.

Cc: stable@vger.kernel.org
Fixes: 2de8e0d999b8 ("mac80211: rewrite fragmentation")
Reported-by: Jouni Malinen <j@w1.fi>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: Fix IBSS disconnect
Sujith Manoharan [Thu, 30 Jan 2014 08:47:28 +0000 (14:17 +0530)]
mac80211: Fix IBSS disconnect

Currently, when a station leaves an IBSS network, the
corresponding BSS is not dropped from cfg80211 if there are
other active stations in the network. But, the small
window that is present when trying to determine a station's
status based on IEEE80211_IBSS_MERGE_INTERVAL introduces
a race.

Instead of trying to keep the BSS, always remove it when
leaving an IBSS network. There is not much benefit to retain
the BSS entry since it will be added with a subsequent join
operation.

This fixes an issue where a dangling BSS entry causes ath9k
to wait for a beacon indefinitely.

Cc: <stable@vger.kernel.org>
Reported-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: release the channel in error path in start_ap
Emmanuel Grumbach [Mon, 27 Jan 2014 09:07:42 +0000 (11:07 +0200)]
mac80211: release the channel in error path in start_ap

When the driver cannot start the AP or when the assignement
of the beacon goes wrong, we need to unassign the vif.

Cc: stable@vger.kernel.org
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: send scan results from work queue
Johannes Berg [Wed, 22 Jan 2014 09:14:19 +0000 (11:14 +0200)]
cfg80211: send scan results from work queue

Due to the previous commit, when a scan finishes, it is in theory
possible to hit the following sequence:
 1. interface starts being removed
 2. scan is cancelled by driver and cfg80211 is notified
 3. scan done work is scheduled
 4. interface is removed completely, rdev->scan_req is freed,
    event sent to userspace but scan done work remains pending
 5. new scan is requested on another virtual interface
 6. scan done work runs, freeing the still-running scan

To fix this situation, hang on to the scan done message and block
new scans while that is the case, and only send the message from
the work function, regardless of whether the scan_req is already
freed from interface removal. This makes step 5 above impossible
and changes step 6 to be
 5. scan done work runs, sending the scan done message

As this can't work for wext, so we send the message immediately,
but this shouldn't be an issue since we still return -EBUSY.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: fix scan done race
Johannes Berg [Wed, 22 Jan 2014 09:14:18 +0000 (11:14 +0200)]
cfg80211: fix scan done race

When an interface/wdev is removed, any ongoing scan should be
cancelled by the driver. This will make it call cfg80211, which
only queues a work struct. If interface/wdev removal is quick
enough, this can leave the scan request pending and processed
only after the interface is gone, causing a use-after-free.

Fix this by making sure the scan request is not pending after
the interface is destroyed. We can't flush or cancel the work
item due to locking concerns, but when it'll run it shouldn't
find anything to do. This leaves a potential issue, if a new
scan gets requested before the work runs, it prematurely stops
the running scan, potentially causing another crash. I'll fix
that in the next patch.

This was particularly observed with P2P_DEVICE wdevs, likely
because freeing them is quicker than freeing netdevs.

Reported-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
Fixes: 4a58e7c38443 ("cfg80211: don't "leak" uncompleted scans")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: avoid deadlock revealed by lockdep
Emmanuel Grumbach [Thu, 23 Jan 2014 12:28:16 +0000 (14:28 +0200)]
mac80211: avoid deadlock revealed by lockdep

sdata->u.ap.request_smps_work can’t be flushed synchronously
under wdev_lock(wdev) since ieee80211_request_smps_ap_work
itself locks the same lock.
While at it, reset the driver_smps_mode when the ap is
stopped to its default: OFF.

This solves:

======================================================
[ INFO: possible circular locking dependency detected ]
3.12.0-ipeer+ #2 Tainted: G           O
-------------------------------------------------------
rmmod/2867 is trying to acquire lock:
  ((&sdata->u.ap.request_smps_work)){+.+...}, at: [<c105b8d0>] flush_work+0x0/0x90

but task is already holding lock:
  (&wdev->mtx){+.+.+.}, at: [<f9b32626>] cfg80211_stop_ap+0x26/0x230 [cfg80211]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&wdev->mtx){+.+.+.}:
        [<c10aefa9>] lock_acquire+0x79/0xe0
        [<c1607a1a>] mutex_lock_nested+0x4a/0x360
        [<fb06288b>] ieee80211_request_smps_ap_work+0x2b/0x50 [mac80211]
        [<c105cdd8>] process_one_work+0x198/0x450
        [<c105d469>] worker_thread+0xf9/0x320
        [<c10669ff>] kthread+0x9f/0xb0
        [<c1613397>] ret_from_kernel_thread+0x1b/0x28

-> #0 ((&sdata->u.ap.request_smps_work)){+.+...}:
        [<c10ae9df>] __lock_acquire+0x183f/0x1910
        [<c10aefa9>] lock_acquire+0x79/0xe0
        [<c105b917>] flush_work+0x47/0x90
        [<c105d867>] __cancel_work_timer+0x67/0xe0
        [<c105d90f>] cancel_work_sync+0xf/0x20
        [<fb0765cc>] ieee80211_stop_ap+0x8c/0x340 [mac80211]
        [<f9b3268c>] cfg80211_stop_ap+0x8c/0x230 [cfg80211]
        [<f9b0d8f9>] cfg80211_leave+0x79/0x100 [cfg80211]
        [<f9b0da72>] cfg80211_netdev_notifier_call+0xf2/0x4f0 [cfg80211]
        [<c160f2c9>] notifier_call_chain+0x59/0x130
        [<c106c6de>] __raw_notifier_call_chain+0x1e/0x30
        [<c106c70f>] raw_notifier_call_chain+0x1f/0x30
        [<c14f8213>] call_netdevice_notifiers_info+0x33/0x70
        [<c14f8263>] call_netdevice_notifiers+0x13/0x20
        [<c14f82a4>] __dev_close_many+0x34/0xb0
        [<c14f83fe>] dev_close_many+0x6e/0xc0
        [<c14f9c77>] rollback_registered_many+0xa7/0x1f0
        [<c14f9dd4>] unregister_netdevice_many+0x14/0x60
        [<fb06f4d9>] ieee80211_remove_interfaces+0xe9/0x170 [mac80211]
        [<fb055116>] ieee80211_unregister_hw+0x56/0x110 [mac80211]
        [<fa3e9396>] iwl_op_mode_mvm_stop+0x26/0xe0 [iwlmvm]
        [<f9b9d8ca>] _iwl_op_mode_stop+0x3a/0x70 [iwlwifi]
        [<f9b9d96f>] iwl_opmode_deregister+0x6f/0x90 [iwlwifi]
        [<fa405179>] __exit_compat+0xd/0x19 [iwlmvm]
        [<c10b8bf9>] SyS_delete_module+0x179/0x2b0
        [<c1613421>] sysenter_do_call+0x12/0x32

Fixes: 687da132234f ("mac80211: implement SMPS for AP")
Cc: <stable@vger.kernel.org> [3.13]
Reported-by: Ilan Peer <ilan.peer@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agocfg80211: re-enable 5/10 MHz support
Johannes Berg [Thu, 23 Jan 2014 15:32:29 +0000 (16:32 +0100)]
cfg80211: re-enable 5/10 MHz support

Unfortunately I forgot this during the merge window, but the
patch seems small enough to go in as a fix. The userspace API
bug that was the reason for disabling it has long been fixed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonl80211: Reset split_start when netlink skb is exhausted
Pontus Fuchs [Thu, 16 Jan 2014 14:00:40 +0000 (15:00 +0100)]
nl80211: Reset split_start when netlink skb is exhausted

When the netlink skb is exhausted split_start is left set. In the
subsequent retry, with a larger buffer, the dump is continued from the
failing point instead of from the beginning.

This was causing my rt28xx based USB dongle to now show up when
running "iw list" with an old iw version without split dump support.

Cc: stable@vger.kernel.org
Fixes: 3713b4e364ef ("nl80211: allow splitting wiphy information in dumps")
Signed-off-by: Pontus Fuchs <pontus.fuchs@gmail.com>
[avoid the entire workaround when state->split is set]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agomac80211: move roc cookie assignment earlier
Eliad Peller [Sun, 12 Jan 2014 09:06:37 +0000 (11:06 +0200)]
mac80211: move roc cookie assignment earlier

ieee80211_start_roc_work() might add a new roc
to existing roc, and tell cfg80211 it has already
started.

However, this might happen before the roc cookie
was set, resulting in REMAIN_ON_CHANNEL (started)
event with null cookie. Consequently, it can make
wpa_supplicant go out of sync.

Fix it by setting the roc cookie earlier.

Cc: stable@vger.kernel.org
Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
10 years agonetfilter: nf_tables: add reject module for NFPROTO_INET
Patrick McHardy [Wed, 5 Feb 2014 15:03:39 +0000 (15:03 +0000)]
netfilter: nf_tables: add reject module for NFPROTO_INET

Add a reject module for NFPROTO_INET. It does nothing but dispatch
to the AF-specific modules based on the hook family.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nft_reject: split up reject module into IPv4 and IPv6 specifc parts
Patrick McHardy [Wed, 5 Feb 2014 15:03:38 +0000 (15:03 +0000)]
netfilter: nft_reject: split up reject module into IPv4 and IPv6 specifc parts

Currently the nft_reject module depends on symbols from ipv6. This is
wrong since no generic module should force IPv6 support to be loaded.
Split up the module into AF-specific and a generic part.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitch
David S. Miller [Thu, 6 Feb 2014 00:25:53 +0000 (16:25 -0800)]
Merge branch 'fixes' of git://git./linux/kernel/git/jesse/openvswitch

Jesse Gross says:

====================
Open vSwitch

A handful of bug fixes for net/3.14. High level fixes are:
 * Regressions introduced by the zerocopy changes, particularly with
   old userspaces.
 * A few bugs lingering from the introduction of megaflows.
 * Overly zealous error checking that is now being triggered frequently
   in common cases.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoxen-netback: Fix Rx stall due to race condition
Zoltan Kiss [Tue, 4 Feb 2014 19:54:37 +0000 (19:54 +0000)]
xen-netback: Fix Rx stall due to race condition

The recent patch to fix receive side flow control
(11b57f90257c1d6a91cee720151b69e0c2020cf6: xen-netback: stop vif thread
spinning if frontend is unresponsive) solved the spinning thread problem,
however caused an another one. The receive side can stall, if:
- [THREAD] xenvif_rx_action sets rx_queue_stopped to true
- [INTERRUPT] interrupt happens, and sets rx_event to true
- [THREAD] then xenvif_kthread sets rx_event to false
- [THREAD] rx_work_todo doesn't return true anymore

Also, if interrupt sent but there is still no room in the ring, it take quite a
long time until xenvif_rx_action realize it. This patch ditch that two variable,
and rework rx_work_todo. If the thread finds it can't fit more skb's into the
ring, it saves the last slot estimation into rx_last_skb_slots, otherwise it's
kept as 0. Then rx_work_todo will check if:
- there is something to send to the ring (like before)
- there is space for the topmost packet in the queue

I think that's more natural and optimal thing to test than two bool which are
set somewhere else.

Signed-off-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetfilter: nf_tables: add AF specific expression support
Patrick McHardy [Wed, 5 Feb 2014 15:03:37 +0000 (15:03 +0000)]
netfilter: nf_tables: add AF specific expression support

For the reject module, we need to add AF-specific implementations to
get rid of incorrect module dependencies. Try to load an AF-specific
module first and fall back to generic modules.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nft_ct: fix missing NFT_CT_L3PROTOCOL key in validity checks
Patrick McHardy [Wed, 5 Feb 2014 15:03:36 +0000 (15:03 +0000)]
netfilter: nft_ct: fix missing NFT_CT_L3PROTOCOL key in validity checks

The key was missing in the list of valid keys, add it.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_tables: fix potential oops when dumping sets
Patrick McHardy [Wed, 5 Feb 2014 15:03:35 +0000 (15:03 +0000)]
netfilter: nf_tables: fix potential oops when dumping sets

Commit c9c8e48597 (netfilter: nf_tables: dump sets in all existing families)
changed nft_ctx_init_from_setattr() to only look up the address family if it
is not NFPROTO_UNSPEC. However if it is NFPROTO_UNSPEC and a table attribute
is given, nftables_afinfo_lookup() will dereference the NULL afi pointer.

Fix by checking for non-NULL afi and also move a check added by that commit
to the proper position.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_tables: fix overrun in nf_tables_set_alloc_name()
Patrick McHardy [Wed, 5 Feb 2014 11:26:22 +0000 (12:26 +0100)]
netfilter: nf_tables: fix overrun in nf_tables_set_alloc_name()

The map that is used to allocate anonymous sets is indeed
BITS_PER_BYTE * PAGE_SIZE long.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_conntrack: don't release a conntrack with non-zero refcnt
Pablo Neira Ayuso [Mon, 3 Feb 2014 19:01:53 +0000 (20:01 +0100)]
netfilter: nf_conntrack: don't release a conntrack with non-zero refcnt

With this patch, the conntrack refcount is initially set to zero and
it is bumped once it is added to any of the list, so we fulfill
Eric's golden rule which is that all released objects always have a
refcount that equals zero.

Andrey Vagin reports that nf_conntrack_free can't be called for a
conntrack with non-zero ref-counter, because it can race with
nf_conntrack_find_get().

A conntrack slab is created with SLAB_DESTROY_BY_RCU. Non-zero
ref-counter says that this conntrack is used. So when we release
a conntrack with non-zero counter, we break this assumption.

CPU1                                    CPU2
____nf_conntrack_find()
                                        nf_ct_put()
                                         destroy_conntrack()
                                        ...
                                        init_conntrack
                                         __nf_conntrack_alloc (set use = 1)
atomic_inc_not_zero(&ct->use) (use = 2)
                                         if (!l4proto->new(ct, skb, dataoff, timeouts))
                                          nf_conntrack_free(ct); (use = 2 !!!)
                                        ...
                                        __nf_conntrack_alloc (set use = 1)
 if (!nf_ct_key_equal(h, tuple, zone))
  nf_ct_put(ct); (use = 0)
   destroy_conntrack()
                                        /* continue to work with CT */

After applying the path "[PATCH] netfilter: nf_conntrack: fix RCU
race in nf_conntrack_find_get" another bug was triggered in
destroy_conntrack():

<4>[67096.759334] ------------[ cut here ]------------
<2>[67096.759353] kernel BUG at net/netfilter/nf_conntrack_core.c:211!
...
<4>[67096.759837] Pid: 498649, comm: atdd veid: 666 Tainted: G         C ---------------    2.6.32-042stab084.18 #1 042stab084_18 /DQ45CB
<4>[67096.759932] RIP: 0010:[<ffffffffa03d99ac>]  [<ffffffffa03d99ac>] destroy_conntrack+0x15c/0x190 [nf_conntrack]
<4>[67096.760255] Call Trace:
<4>[67096.760255]  [<ffffffff814844a7>] nf_conntrack_destroy+0x17/0x30
<4>[67096.760255]  [<ffffffffa03d9bb5>] nf_conntrack_find_get+0x85/0x130 [nf_conntrack]
<4>[67096.760255]  [<ffffffffa03d9fb2>] nf_conntrack_in+0x352/0xb60 [nf_conntrack]
<4>[67096.760255]  [<ffffffffa048c771>] ipv4_conntrack_local+0x51/0x60 [nf_conntrack_ipv4]
<4>[67096.760255]  [<ffffffff81484419>] nf_iterate+0x69/0xb0
<4>[67096.760255]  [<ffffffff814b5b00>] ? dst_output+0x0/0x20
<4>[67096.760255]  [<ffffffff814845d4>] nf_hook_slow+0x74/0x110
<4>[67096.760255]  [<ffffffff814b5b00>] ? dst_output+0x0/0x20
<4>[67096.760255]  [<ffffffff814b66d5>] raw_sendmsg+0x775/0x910
<4>[67096.760255]  [<ffffffff8104c5a8>] ? flush_tlb_others_ipi+0x128/0x130
<4>[67096.760255]  [<ffffffff8100bc4e>] ? apic_timer_interrupt+0xe/0x20
<4>[67096.760255]  [<ffffffff8100bc4e>] ? apic_timer_interrupt+0xe/0x20
<4>[67096.760255]  [<ffffffff814c136a>] inet_sendmsg+0x4a/0xb0
<4>[67096.760255]  [<ffffffff81444e93>] ? sock_sendmsg+0x13/0x140
<4>[67096.760255]  [<ffffffff81444f97>] sock_sendmsg+0x117/0x140
<4>[67096.760255]  [<ffffffff8102e299>] ? native_smp_send_reschedule+0x49/0x60
<4>[67096.760255]  [<ffffffff81519beb>] ? _spin_unlock_bh+0x1b/0x20
<4>[67096.760255]  [<ffffffff8109d930>] ? autoremove_wake_function+0x0/0x40
<4>[67096.760255]  [<ffffffff814960f0>] ? do_ip_setsockopt+0x90/0xd80
<4>[67096.760255]  [<ffffffff8100bc4e>] ? apic_timer_interrupt+0xe/0x20
<4>[67096.760255]  [<ffffffff8100bc4e>] ? apic_timer_interrupt+0xe/0x20
<4>[67096.760255]  [<ffffffff814457c9>] sys_sendto+0x139/0x190
<4>[67096.760255]  [<ffffffff810efa77>] ? audit_syscall_entry+0x1d7/0x200
<4>[67096.760255]  [<ffffffff810ef7c5>] ? __audit_syscall_exit+0x265/0x290
<4>[67096.760255]  [<ffffffff81474daf>] compat_sys_socketcall+0x13f/0x210
<4>[67096.760255]  [<ffffffff8104dea3>] ia32_sysret+0x0/0x5

I have reused the original title for the RFC patch that Andrey posted and
most of the original patch description.

Cc: Eric Dumazet <edumazet@google.com>
Cc: Andrew Vagin <avagin@parallels.com>
Cc: Florian Westphal <fw@strlen.de>
Reported-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Andrew Vagin <avagin@parallels.com>
10 years agonetfilter: nf_nat_h323: fix crash in nf_ct_unlink_expect_report()
Alexey Dobriyan [Mon, 3 Feb 2014 12:07:24 +0000 (13:07 +0100)]
netfilter: nf_nat_h323: fix crash in nf_ct_unlink_expect_report()

Similar bug fixed in SIP module in 3f509c6 ("netfilter: nf_nat_sip: fix
incorrect handling of EBUSY for RTCP expectation").

BUG: unable to handle kernel paging request at 00100104
IP: [<f8214f07>] nf_ct_unlink_expect_report+0x57/0xf0 [nf_conntrack]
...
Call Trace:
  [<c0244bd8>] ? del_timer+0x48/0x70
  [<f8215687>] nf_ct_remove_expectations+0x47/0x60 [nf_conntrack]
  [<f8211c99>] nf_ct_delete_from_lists+0x59/0x90 [nf_conntrack]
  [<f8212e5e>] death_by_timeout+0x14e/0x1c0 [nf_conntrack]
  [<f8212d10>] ? nf_conntrack_set_hashsize+0x190/0x190 [nf_conntrack]
  [<c024442d>] call_timer_fn+0x1d/0x80
  [<c024461e>] run_timer_softirq+0x18e/0x1a0
  [<f8212d10>] ? nf_conntrack_set_hashsize+0x190/0x190 [nf_conntrack]
  [<c023e6f3>] __do_softirq+0xa3/0x170
  [<c023e650>] ? __local_bh_enable+0x70/0x70
  <IRQ>
  [<c023e587>] ? irq_exit+0x67/0xa0
  [<c0202af6>] ? do_IRQ+0x46/0xb0
  [<c027ad05>] ? clockevents_notify+0x35/0x110
  [<c066ac6c>] ? common_interrupt+0x2c/0x40
  [<c056e3c1>] ? cpuidle_enter_state+0x41/0xf0
  [<c056e6fb>] ? cpuidle_idle_call+0x8b/0x100
  [<c02085f8>] ? arch_cpu_idle+0x8/0x30
  [<c027314b>] ? cpu_idle_loop+0x4b/0x140
  [<c0273258>] ? cpu_startup_entry+0x18/0x20
  [<c066056d>] ? rest_init+0x5d/0x70
  [<c0813ac8>] ? start_kernel+0x2ec/0x2f2
  [<c081364f>] ? repair_env_string+0x5b/0x5b
  [<c0813269>] ? i386_start_kernel+0x33/0x35

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_conntrack: fix RCU race in nf_conntrack_find_get
Andrey Vagin [Wed, 29 Jan 2014 18:34:14 +0000 (19:34 +0100)]
netfilter: nf_conntrack: fix RCU race in nf_conntrack_find_get

Lets look at destroy_conntrack:

hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode);
...
nf_conntrack_free(ct)
kmem_cache_free(net->ct.nf_conntrack_cachep, ct);

net->ct.nf_conntrack_cachep is created with SLAB_DESTROY_BY_RCU.

The hash is protected by rcu, so readers look up conntracks without
locks.
A conntrack is removed from the hash, but in this moment a few readers
still can use the conntrack. Then this conntrack is released and another
thread creates conntrack with the same address and the equal tuple.
After this a reader starts to validate the conntrack:
* It's not dying, because a new conntrack was created
* nf_ct_tuple_equal() returns true.

But this conntrack is not initialized yet, so it can not be used by two
threads concurrently. In this case BUG_ON may be triggered from
nf_nat_setup_info().

Florian Westphal suggested to check the confirm bit too. I think it's
right.

task 1 task 2 task 3
nf_conntrack_find_get
 ____nf_conntrack_find
destroy_conntrack
 hlist_nulls_del_rcu
 nf_conntrack_free
 kmem_cache_free
__nf_conntrack_alloc
 kmem_cache_alloc
 memset(&ct->tuplehash[IP_CT_DIR_MAX],
 if (nf_ct_is_dying(ct))
 if (!nf_ct_tuple_equal()

I'm not sure, that I have ever seen this race condition in a real life.
Currently we are investigating a bug, which is reproduced on a few nodes.
In our case one conntrack is initialized from a few tasks concurrently,
we don't have any other explanation for this.

<2>[46267.083061] kernel BUG at net/ipv4/netfilter/nf_nat_core.c:322!
...
<4>[46267.083951] RIP: 0010:[<ffffffffa01e00a4>]  [<ffffffffa01e00a4>] nf_nat_setup_info+0x564/0x590 [nf_nat]
...
<4>[46267.085549] Call Trace:
<4>[46267.085622]  [<ffffffffa023421b>] alloc_null_binding+0x5b/0xa0 [iptable_nat]
<4>[46267.085697]  [<ffffffffa02342bc>] nf_nat_rule_find+0x5c/0x80 [iptable_nat]
<4>[46267.085770]  [<ffffffffa0234521>] nf_nat_fn+0x111/0x260 [iptable_nat]
<4>[46267.085843]  [<ffffffffa0234798>] nf_nat_out+0x48/0xd0 [iptable_nat]
<4>[46267.085919]  [<ffffffff814841b9>] nf_iterate+0x69/0xb0
<4>[46267.085991]  [<ffffffff81494e70>] ? ip_finish_output+0x0/0x2f0
<4>[46267.086063]  [<ffffffff81484374>] nf_hook_slow+0x74/0x110
<4>[46267.086133]  [<ffffffff81494e70>] ? ip_finish_output+0x0/0x2f0
<4>[46267.086207]  [<ffffffff814b5890>] ? dst_output+0x0/0x20
<4>[46267.086277]  [<ffffffff81495204>] ip_output+0xa4/0xc0
<4>[46267.086346]  [<ffffffff814b65a4>] raw_sendmsg+0x8b4/0x910
<4>[46267.086419]  [<ffffffff814c10fa>] inet_sendmsg+0x4a/0xb0
<4>[46267.086491]  [<ffffffff814459aa>] ? sock_update_classid+0x3a/0x50
<4>[46267.086562]  [<ffffffff81444d67>] sock_sendmsg+0x117/0x140
<4>[46267.086638]  [<ffffffff8151997b>] ? _spin_unlock_bh+0x1b/0x20
<4>[46267.086712]  [<ffffffff8109d370>] ? autoremove_wake_function+0x0/0x40
<4>[46267.086785]  [<ffffffff81495e80>] ? do_ip_setsockopt+0x90/0xd80
<4>[46267.086858]  [<ffffffff8100be0e>] ? call_function_interrupt+0xe/0x20
<4>[46267.086936]  [<ffffffff8118cb10>] ? ub_slab_ptr+0x20/0x90
<4>[46267.087006]  [<ffffffff8118cb10>] ? ub_slab_ptr+0x20/0x90
<4>[46267.087081]  [<ffffffff8118f2e8>] ? kmem_cache_alloc+0xd8/0x1e0
<4>[46267.087151]  [<ffffffff81445599>] sys_sendto+0x139/0x190
<4>[46267.087229]  [<ffffffff81448c0d>] ? sock_setsockopt+0x16d/0x6f0
<4>[46267.087303]  [<ffffffff810efa47>] ? audit_syscall_entry+0x1d7/0x200
<4>[46267.087378]  [<ffffffff810ef795>] ? __audit_syscall_exit+0x265/0x290
<4>[46267.087454]  [<ffffffff81474885>] ? compat_sys_setsockopt+0x75/0x210
<4>[46267.087531]  [<ffffffff81474b5f>] compat_sys_socketcall+0x13f/0x210
<4>[46267.087607]  [<ffffffff8104dea3>] ia32_sysret+0x0/0x5
<4>[46267.087676] Code: 91 20 e2 01 75 29 48 89 de 4c 89 f7 e8 56 fa ff ff 85 c0 0f 84 68 fc ff ff 0f b6 4d c6 41 8b 45 00 e9 4d fb ff ff e8 7c 19 e9 e0 <0f> 0b eb fe f6 05 17 91 20 e2 80 74 ce 80 3d 5f 2e 00 00 00 74
<1>[46267.088023] RIP  [<ffffffffa01e00a4>] nf_nat_setup_info+0x564/0x590

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Florian Westphal <fw@strlen.de>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nf_tables: fix oops when deleting a chain with references
Patrick McHardy [Sat, 25 Jan 2014 08:04:07 +0000 (08:04 +0000)]
netfilter: nf_tables: fix oops when deleting a chain with references

The following commands trigger an oops:

 # nft -i
 nft> add table filter
 nft> add chain filter input { type filter hook input priority 0; }
 nft> add chain filter test
 nft> add rule filter input jump test
 nft> delete chain filter test

We need to check the chain use counter before allowing destruction since
we might have references from sets or jump rules.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=69341
Reported-by: Matthew Ife <deleriux1@gmail.com>
Tested-by: Matthew Ife <deleriux1@gmail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nft_ct: fix unconditional dump of 'dir' attr
Arturo Borrero [Fri, 17 Jan 2014 01:28:45 +0000 (02:28 +0100)]
netfilter: nft_ct: fix unconditional dump of 'dir' attr

We want to make sure that the information that we get from the kernel can
be reinjected without troubles. The kernel shouldn't return an attribute
that is not required, or even prohibited.

Dumping unconditionally NFTA_CT_DIRECTION could lead an application in
userspace to interpret that the attribute was originally set, while it
was not.

Signed-off-by: Arturo Borrero Gonzalez <arturo.borrero.glez@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agoopenvswitch: Suppress error messages on megaflow updates
Andy Zhou [Mon, 3 Feb 2014 01:08:06 +0000 (17:08 -0800)]
openvswitch: Suppress error messages on megaflow updates

With subfacets, we'd expect megaflow updates message to carry
the original micro flow. If not, EINVAL is returned and kernel
logs an error message.  Now that the user space subfacet layer is
removed, it is expected that flow updates can arrive with a
micro flow other than the original. Change the return code to
EEXIST and remove the kernel error log message.

Reported-by: Ben Pfaff <blp@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agoopenvswitch: Fix ovs_flow_free() ovs-lock assert.
Pravin B Shelar [Fri, 31 Jan 2014 17:43:23 +0000 (09:43 -0800)]
openvswitch: Fix ovs_flow_free() ovs-lock assert.

ovs_flow_free() is not called under ovs-lock during packet
execute path (ovs_packet_cmd_execute()). Since packet execute
does not touch flow->mask, there is no need to take that
lock either. So move assert in case where flow->mask is checked.

Found by code inspection.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agoopenvswitch: Fix ovs_dp_cmd_msg_size()
Daniele Di Proietto [Thu, 23 Jan 2014 18:47:35 +0000 (10:47 -0800)]
openvswitch: Fix ovs_dp_cmd_msg_size()

commit 43d4be9cb55f3bac5253e9289996fd9d735531db (openvswitch: Allow user space
to announce ability to accept unaligned Netlink messages) introduced
OVS_DP_ATTR_USER_FEATURES netlink attribute in datapath responses,
but the attribute size was not taken into account in ovs_dp_cmd_msg_size().

Signed-off-by: Daniele Di Proietto <daniele.di.proietto@gmail.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agoopenvswitch: Fix kernel panic on ovs_flow_free
Andy Zhou [Tue, 21 Jan 2014 17:31:04 +0000 (09:31 -0800)]
openvswitch: Fix kernel panic on ovs_flow_free

Both mega flow mask's reference counter and per flow table mask list
should only be accessed when holding ovs_mutex() lock. However
this is not true with ovs_flow_table_flush(). The patch fixes this bug.

Reported-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agoopenvswitch: Pad OVS_PACKET_ATTR_PACKET if linear copy was performed
Thomas Graf [Tue, 14 Jan 2014 16:27:49 +0000 (16:27 +0000)]
openvswitch: Pad OVS_PACKET_ATTR_PACKET if linear copy was performed

While the zerocopy method is correctly omitted if user space
does not support unaligned Netlink messages. The attribute is
still not padded correctly as skb_zerocopy() will not ensure
padding and the attribute size is no longer pre calculated
though nla_reserve() which ensured padding previously.

This patch applies appropriate padding if a linear data copy
was performed in skb_zerocopy().

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Zoltan Kiss <zoltan.kiss@citrix.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
10 years agoxen-netfront: handle backend CLOSED without CLOSING
David Vrabel [Tue, 4 Feb 2014 18:50:26 +0000 (18:50 +0000)]
xen-netfront: handle backend CLOSED without CLOSING

Backend drivers shouldn't transistion to CLOSED unless the frontend is
CLOSED.  If a backend does transition to CLOSED too soon then the
frontend may not see the CLOSING state and will not properly shutdown.

So, treat an unexpected backend CLOSED state the same as CLOSING.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: fix L2-GRE TCP issues
Dmitry Kravkov [Tue, 4 Feb 2014 15:43:03 +0000 (17:43 +0200)]
bnx2x: fix L2-GRE TCP issues

When configuring GRE tunnel using OVS, tcp stream is distributed over
all RSS queues which may cause TCP reordering. It happens since OVS
uses L2GRE protocol when kernel gre uses IPGRE.
Patch defaults gre tunnel to L2GRE which allows proper RSS for L2GRE
packets and (implicitly) disables RSS for IPGRE traffic.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: qmi_wwan: add Netgear Aircard 340U
Bjørn Mork [Tue, 4 Feb 2014 12:04:33 +0000 (13:04 +0100)]
net: qmi_wwan: add Netgear Aircard 340U

This device was mentioned in an OpenWRT forum.  Seems to have a "standard"
Sierra Wireless ifnumber to function layout:
 0: qcdm
 2: nmea
 3: modem
 8: qmi
 9: storage

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agortnetlink: fix oops in rtnl_link_get_slave_info_data_size
Fernando Luis Vazquez Cao [Tue, 4 Feb 2014 10:35:02 +0000 (19:35 +0900)]
rtnetlink: fix oops in rtnl_link_get_slave_info_data_size

We should check whether rtnetlink link operations
are defined before calling get_slave_size().

Without this, the following oops can occur when
adding a tap device to OVS.

[   87.839553] BUG: unable to handle kernel NULL pointer dereference at 00000000000000a8
[   87.839595] IP: [<ffffffff813d47c0>] if_nlmsg_size+0xf0/0x220
[...]
[   87.840651] Call Trace:
[   87.840664]  [<ffffffff813d694b>] ? rtmsg_ifinfo+0x2b/0x100
[   87.840688]  [<ffffffff813c8340>] ? __netdev_adjacent_dev_insert+0x150/0x1a0
[   87.840718]  [<ffffffff813d6a50>] ? rtnetlink_event+0x30/0x40
[   87.840742]  [<ffffffff814b4144>] ? notifier_call_chain+0x44/0x70
[   87.840768]  [<ffffffff813c8946>] ? __netdev_upper_dev_link+0x3c6/0x3f0
[   87.840798]  [<ffffffffa0678d6c>] ? netdev_create+0xcc/0x160 [openvswitch]
[   87.840828]  [<ffffffffa06781ea>] ? ovs_vport_add+0x4a/0xd0 [openvswitch]
[   87.840857]  [<ffffffffa0670139>] ? new_vport+0x9/0x50 [openvswitch]
[   87.840884]  [<ffffffffa067279e>] ? ovs_vport_cmd_new+0x11e/0x210 [openvswitch]
[   87.840915]  [<ffffffff813f3efa>] ? genl_family_rcv_msg+0x19a/0x360
[   87.840941]  [<ffffffff813f40c0>] ? genl_family_rcv_msg+0x360/0x360
[   87.840967]  [<ffffffff813f4139>] ? genl_rcv_msg+0x79/0xc0
[   87.840991]  [<ffffffff813b6cf9>] ? __kmalloc_reserve.isra.25+0x29/0x80
[   87.841018]  [<ffffffff813f2389>] ? netlink_rcv_skb+0xa9/0xc0
[   87.841042]  [<ffffffff813f27cf>] ? genl_rcv+0x1f/0x30
[   87.841064]  [<ffffffff813f1988>] ? netlink_unicast+0xe8/0x1e0
[   87.841088]  [<ffffffff813f1d9a>] ? netlink_sendmsg+0x31a/0x750
[   87.841113]  [<ffffffff813aee96>] ? sock_sendmsg+0x86/0xc0
[   87.841136]  [<ffffffff813c960d>] ? __netdev_update_features+0x4d/0x200
[   87.841163]  [<ffffffff813ca94e>] ? ethtool_get_value+0x2e/0x50
[   87.841188]  [<ffffffff813af269>] ? ___sys_sendmsg+0x359/0x370
[   87.841212]  [<ffffffff813da686>] ? dev_ioctl+0x1a6/0x5c0
[   87.841236]  [<ffffffff8109c210>] ? autoremove_wake_function+0x30/0x30
[   87.841264]  [<ffffffff813ac59d>] ? sock_do_ioctl+0x3d/0x50
[   87.841288]  [<ffffffff813aca68>] ? sock_ioctl+0x1e8/0x2c0
[   87.841312]  [<ffffffff811934bf>] ? do_vfs_ioctl+0x2cf/0x4b0
[   87.841335]  [<ffffffff813afeb9>] ? __sys_sendmsg+0x39/0x70
[   87.841362]  [<ffffffff814b86f9>] ? system_call_fastpath+0x16/0x1b
[   87.841386] Code: c0 74 10 48 89 ef ff d0 83 c0 07 83 e0 fc 48 98 49 01 c7 48 89 ef e8 d0 d6 fe ff 48 85 c0 0f 84 df 00 00 00 48 8b 90 08 07 00 00 <48> 8b 8a a8 00 00 00 31 d2 48 85 c9 74 0c 48 89 ee 48 89 c7 ff
[   87.841529] RIP  [<ffffffff813d47c0>] if_nlmsg_size+0xf0/0x220
[   87.841555]  RSP <ffff880221aa5950>
[   87.841569] CR2: 00000000000000a8
[   87.851442] ---[ end trace e42ab217691b4fc2 ]---

Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoptp: Allow selecting trigger/event index in testptp
Stefan Sørensen [Tue, 4 Feb 2014 07:46:36 +0000 (08:46 +0100)]
ptp: Allow selecting trigger/event index in testptp

Currently the trigger/event is hardcoded to 0, this patch adds
a new command line argument -i to select an arbitrary trigger/
event.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: set up MII management bus clock
Max Filippov [Mon, 3 Feb 2014 23:33:10 +0000 (03:33 +0400)]
net: ethoc: set up MII management bus clock

MII management bus clock is derived from the MAC clock by dividing it by
MIIMODER register CLKDIV field value. This value may need to be set up
in case it is undefined or its default value is too high (and
communication with PHY is too slow) or too low (and communication with
PHY is impossible). The value of CLKDIV is not specified directly, but
is derived from the MAC clock for the default MII management bus frequency
of 2.5MHz. The MAC clock may be specified in the platform data, or in
the 'clocks' device tree attribute.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: don't advertise gigabit speed on attached PHY
Max Filippov [Mon, 3 Feb 2014 23:33:09 +0000 (03:33 +0400)]
net: ethoc: don't advertise gigabit speed on attached PHY

OpenCores 10/100 Mbps MAC does not support speeds above 100 Mbps, but does
not disable advertisement when PHY supports them. This results in
non-functioning network when the MAC is connected to a gigabit PHY connected
to a gigabit switch.

The fix is to disable gigabit speed advertisement on attached PHY
unconditionally.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: ensure Gigabit features are masked off if requested
Florian Fainelli [Mon, 3 Feb 2014 20:35:46 +0000 (12:35 -0800)]
net: phy: ensure Gigabit features are masked off if requested

When a Gigabit PHY device is connected to a 10/100Mbits capable Ethernet
MAC, the driver will restrict the phydev->supported modes to mask off
Gigabit. If the Gigabit PHY comes out of reset with the Gigabit features
set by default in MII_CTRL1000, it will keep advertising these feature,
so by the time we call genphy_config_advert(), the condition on
phydev->supported having the Gigabit features on is false, and we do not
update MII_CTRL1000 with updated values, and we keep advertising Gigabit
features, eventually configuring the PHY for Gigabit whilst the Ethernet
MAC does not support that.

This patches fixes the problem by ensuring that the Gigabit feature bits
are always cleared in MII_CTRL1000, if the PHY happens to be a Gigabit
PHY, and then, if Gigabit features are supported, setting those and
updating MII_CTRL1000 accordingly.

Reported-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet:phy:dp83640: Initialize PTP clocks at device init.
Stefan Sørensen [Mon, 3 Feb 2014 14:36:58 +0000 (15:36 +0100)]
net:phy:dp83640: Initialize PTP clocks at device init.

The trigger and events functionality can be useful even if packet
timestamping is not used, but the required PTP clock is only enabled
when packet timestamping is started. This patch moves the clock enable
to when the interface is configured.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet:phy:dp83640: Do not hardcode timestamping event edge
Stefan Sørensen [Mon, 3 Feb 2014 14:36:50 +0000 (15:36 +0100)]
net:phy:dp83640: Do not hardcode timestamping event edge

Currently the external timestamping code is hardcoded to use the
rising edge even though the hardware has configurable event edge
detection. This patch changes the code to use falling edge detection
if PTP_FALLING_EDGE is set in the user supplied flags.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet:phy:dp83640: Declare that TX timestamping possible
Stefan Sørensen [Mon, 3 Feb 2014 14:36:35 +0000 (15:36 +0100)]
net:phy:dp83640: Declare that TX timestamping possible

Set the SKBTX_IN_PROGRESS bit in tx_flags dp83640_txtstamp when doing
tx timestamps as per Documentation/networking/timestamping.txt.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/ipv4: Use proper RCU APIs for writer-side in udp_offload.c
Shlomo Pongratz [Sun, 2 Feb 2014 13:42:10 +0000 (15:42 +0200)]
net/ipv4: Use proper RCU APIs for writer-side in udp_offload.c

RCU writer side should use rcu_dereference_protected() and not
rcu_dereference(), fix that. This also removes the "suspicious RCU usage"
warning seen when running with CONFIG_PROVE_RCU.

Also, don't use rcu_assign_pointer/rcu_dereference for pointers
which are invisible beyond the udp offload code.

Fixes: b582ef0 ('net: Add GRO support for UDP encapsulating protocols')
Reported-by: Eric Dumazet <edumazet@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'bonding_fail_over_mac'
David S. Miller [Wed, 5 Feb 2014 03:47:55 +0000 (19:47 -0800)]
Merge branch 'bonding_fail_over_mac'

Ding Tianhong says:

====================
bonding: Fix some issues for fail_over_mac

The parameter fail_over_mac only affect active-backup mode, if it was
set to active or follow and works with other modes, just like RR or XOR
mode, the bonding could not set all slaves to the master's address, it
will cause the slave could not work well with master.

v1->v2: According Jay's suggestion, that we should permit setting an option
at any time, but only have it take effect in active-backup mode, so
I add mode checking together with fail_over_mac during enslavement and
rebuild the patches.

v2->v3: The correct way to fix the problem is that we should not add restrictions when
     setting options, just need to modify the bond enslave and removal processing
     to check the mode in addition to fail_over_mac when setting a slave's MAC during
     enslavement. The change active slave processing already only calls the fail_over_mac
     function when in active-backup mode.

Remove the cleanup patch because the net-next is frozen now.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
10 years agobonding: fail_over_mac should only affect AB mode in bond_set_mac_address()
dingtianhong [Sat, 25 Jan 2014 05:00:57 +0000 (13:00 +0800)]
bonding: fail_over_mac should only affect AB mode in bond_set_mac_address()

The fail_over_mac could be set to active or follow in any time for all modes,
so if the fail_over_mac is not none and the current mode is not active-backup,
the bond_set_mac_address() could not change the master and slave's MAC address.

In bond_set_mac_address(), the fail_over_mac should only affect AB mode, so modify
to check the mode in addition to fail_over_mac when setting bond's MAC address.

Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: fail_over_mac should only affect AB mode at enslave and removal processing
dingtianhong [Sat, 25 Jan 2014 05:00:29 +0000 (13:00 +0800)]
bonding: fail_over_mac should only affect AB mode at enslave and removal processing

According to bonding.txt, the fail_over_ma should only affect active-backup mode,
but I found that the fail_over_mac could be set to active or follow in all
modes, this will cause new slave could not be set to bond's MAC address at
enslave processing and restore its own MAC address at removal processing.

The correct way to fix the problem is that we should not add restrictions when
setting options, just need to modify the bond enslave and removal processing
to check the mode in addition to fail_over_mac when setting a slave's MAC during
enslavement. The change active slave processing already only calls the fail_over_mac
function when in active-backup mode.

Thanks for Jay's suggestion.

The patch also modify the pr_warning() to pr_warn().

Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoath9k: Fix TX power calculation
Sujith Manoharan [Tue, 4 Feb 2014 03:07:54 +0000 (08:37 +0530)]
ath9k: Fix TX power calculation

The commit, "ath9k_hw: Fix incorrect Tx control power in AR9003 template"
fixed the incorrect values in the eeprom templates, but if
boards have already been calibrated with incorrect values,
they would still be using the wrong TX power. Fix this by assigning
a default value in such cases.

Cc: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Do not support PowerSave by default
Sujith Manoharan [Tue, 4 Feb 2014 03:07:53 +0000 (08:37 +0530)]
ath9k: Do not support PowerSave by default

Even though we make sure PowerSave is not enabled by default
by disabling the flag, WIPHY_FLAG_PS_ON_BY_DEFAULT on init,
PS could be enabled by userspace based on various factors
like battery usage etc. Since PS in ath9k is just broken
and has been untested for years, remove support for it, but
allow a user to explicitly enable it using a module parameter.

Cc: stable@vger.kernel.org
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k: Fix build error on ARM
Sujith Manoharan [Tue, 4 Feb 2014 03:07:52 +0000 (08:37 +0530)]
ath9k: Fix build error on ARM

Use mdelay instead of udelay to fix this error:

ERROR: "__bad_udelay" [drivers/net/wireless/ath/ath9k/ath9k_hw.ko] undefined!
make[1]: *** [__modpost] Error 1
make: *** [modules] Error 2

Reported-by: Josh Boyer <jwboyer@fedoraproject.org>
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_htc: avoid scheduling while atomic on sta_rc_update
Stanislaw Gruszka [Mon, 3 Feb 2014 10:45:51 +0000 (11:45 +0100)]
ath9k_htc: avoid scheduling while atomic on sta_rc_update

mac80211 ->sta_rc_update() callback must be atomic. Since we have to
take mutex and do other operations that can sleep when sending fimrware
commands to device, the only option to satisfy atomicity requirement of
->sta_rc_update(), that I can see, is introduce work_struct and defer
uploading new rates to that work.

Tested-by: Oleksij Rempel <linux@rempel-privat.de>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoar5523: fix usb id for Gigaset.
Oleksij Rempel [Sun, 2 Feb 2014 09:55:18 +0000 (10:55 +0100)]
ar5523: fix usb id for Gigaset.

Raw id and FW id should be switched.

Tested-by: Oleksij Rempel <linux@rempel-privat.de>
Signed-off-by: Oleksij Rempel <linux@rempel-privat.de>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_htc: Do not support PowerSave by default
Oleksij Rempel [Thu, 30 Jan 2014 08:14:53 +0000 (09:14 +0100)]
ath9k_htc: Do not support PowerSave by default

It is a copy/paste of patch provided by Sujith for ath9k.

"Even though we make sure PowerSave is not enabled by default
by disabling the flag, WIPHY_FLAG_PS_ON_BY_DEFAULT on init,
PS could be enabled by userspace based on various factors
like battery usage etc. Since PS in ath9k is just broken
and has been untested for years, remove support for it, but
allow a user to explicitly enable it using a module parameter."

Cc: stable@vger.kernel.org
Signed-off-by: Oleksij Rempel <linux@rempel-privat.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agort2500: disable PS by default
Stanislaw Gruszka [Wed, 29 Jan 2014 16:54:17 +0000 (17:54 +0100)]
rt2500: disable PS by default

It is know that PS cause issues on that old devices, disable it by
default.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agort2800: disable PS by default on USB
Stanislaw Gruszka [Wed, 29 Jan 2014 16:42:37 +0000 (17:42 +0100)]
rt2800: disable PS by default on USB

We have disabled it currently on other buses. PS can cause some issues,
not necessarily with our driver but on AP, that are not easy to debug.
Since behaviour differs on rt2800usb and rt2800pci, user usually blame
for malfunction rt2800usb driver, whereas issue is on AP side.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_htc: make ->sta_rc_update atomic for most calls
Stanislaw Gruszka [Tue, 28 Jan 2014 08:14:48 +0000 (09:14 +0100)]
ath9k_htc: make ->sta_rc_update atomic for most calls

sta_rc_update() callback must be atomic, hence we can not take mutexes
or do other operations, which can sleep in ath9k_htc_sta_rc_update().

I think we can just return from ath9k_htc_sta_rc_update(), if it is
called without IEEE80211_RC_SUPP_RATES_CHANGED bit. That will help
with scheduling while atomic bug for most cases (except mesh and IBSS
modes).

For mesh and IBSS I do not see other solution like creating additional
workqueue, because sending firmware command require us to sleep, but
this can be done in additional patch.

Patch partially fixes bug:
https://bugzilla.redhat.com/show_bug.cgi?id=990955

Cc: stable@vger.kernel.org
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi...
John W. Linville [Tue, 4 Feb 2014 20:14:34 +0000 (15:14 -0500)]
Merge branch 'for-john' of git://git./linux/kernel/git/iwlwifi/iwlwifi-fixes

10 years agoipvs: fix AF assignment in ip_vs_conn_new()
Michal Kubecek [Thu, 30 Jan 2014 07:50:20 +0000 (08:50 +0100)]
ipvs: fix AF assignment in ip_vs_conn_new()

If a fwmark is passed to ip_vs_conn_new(), it is passed in
vaddr, not daddr. Therefore we should set AF to AF_UNSPEC in
vaddr assignment (like we do in ip_vs_ct_in_get()), otherwise we
may copy only first 4 bytes of an IPv6 address into cp->daddr.

Signed-off-by: Bogdano Arendartchuk <barendartchuk@suse.com>
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
10 years agoip_tunnel: fix panic in ip_tunnel_xmit()
Eric Dumazet [Mon, 3 Feb 2014 20:52:14 +0000 (12:52 -0800)]
ip_tunnel: fix panic in ip_tunnel_xmit()

Setting rt variable to NULL at the beginning of ip_tunnel_xmit()
missed possible use of this variable as a scratch value.

Also fixes a possible dst leak in tunnel_dst_check() :
If we had to call tunnel_dst_reset(), we forgot to
release the reference on dst.

Merges tunnel_dst_get()/tunnel_dst_check() into
a single tunnel_rtable_get() function for clarity.

Many thanks to Tommi for his report and tests.

Fixes: 7d442fab0a67 ("ipv4: Cache dst in tunnels")
Reported-by: Tommi Rantala <tt.rantala@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Tommi Rantala <tt.rantala@gmail.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoiwlwifi: mvm: don't allow A band if SKU forbids it
Emmanuel Grumbach [Thu, 5 Dec 2013 20:42:55 +0000 (22:42 +0200)]
iwlwifi: mvm: don't allow A band if SKU forbids it

The driver wasn't reading the NVM properly. While this
didn't lead to any issue until now, it seems that there
is an old version of the NVM in the wild.
In this version, the A band channels appear to be valid
but the SKU capabilities (another field of the NVM) says
that A band isn't supported at all.
With this specific version of the NVM, the driver would
think that A band is supported while the HW / firmware
don't. This leads to asserts.

Cc: <stable@vger.kernel.org> [3.10+]
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agoiwlwifi: mvm: BT Coex - disable BT when TXing probe request in scan
Emmanuel Grumbach [Tue, 28 Jan 2014 10:27:31 +0000 (12:27 +0200)]
iwlwifi: mvm: BT Coex - disable BT when TXing probe request in scan

Not doing so will let BT kill our probe requests leading to
failures in scan.

Cc: <stable@vger.kernel.org> [3.10+]
Reviewed-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
10 years agovxlan: remove extra newline after function definition
Daniel Baluta [Fri, 31 Jan 2014 07:50:12 +0000 (09:50 +0200)]
vxlan: remove extra newline after function definition

Signed-off-by: Daniel Baluta <dbaluta@ixiacom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'ethoc'
David S. Miller [Sun, 2 Feb 2014 00:57:05 +0000 (16:57 -0800)]
Merge branch 'ethoc'

Max Filippov says:

====================
OpenCores 10/100 MAC ethtool operations

this series implements ethtool callbacks for the ethoc driver as was
requested by Florian.

Changes v1->v2:
- fix {get,set}_settings return code in case there's no PHY;
- fix set_ringparam: check ring sizes, change ring sizes on the fly.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: implement ethtool get/set ring parameters
Max Filippov [Fri, 31 Jan 2014 05:41:07 +0000 (09:41 +0400)]
net: ethoc: implement ethtool get/set ring parameters

TX and RX rings share memory and descriptors. Maximal number of
descriptors reported is one less than the total available nuber of
descriptors. For the set operation the requested number of TX descriptors
is rounded down to the nearest power of two (driver logic requirement).

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: implement ethtool get registers
Max Filippov [Fri, 31 Jan 2014 05:41:06 +0000 (09:41 +0400)]
net: ethoc: implement ethtool get registers

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: implement ethtool get/set settings
Max Filippov [Fri, 31 Jan 2014 05:41:05 +0000 (09:41 +0400)]
net: ethoc: implement ethtool get/set settings

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethoc: implement basic ethtool operations
Max Filippov [Fri, 31 Jan 2014 05:41:04 +0000 (09:41 +0400)]
net: ethoc: implement basic ethtool operations

The following methods are implemented:
- get link state (standard implementation);
- get timestamping info (standard implementation).

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'linux-can-fixes-for-3.14-20140129' of git://gitorious.org/linux-can/linux-can
David S. Miller [Fri, 31 Jan 2014 00:48:17 +0000 (16:48 -0800)]
Merge tag 'linux-can-fixes-for-3.14-20140129' of git://gitorious.org/linux-can/linux-can

linux-can-fixes-for-3.14-20140129

Marc Kleine-Budde says:

====================
Arnd Bergmann provides a fix for the flexcan driver, enabling compilation on
all combinations of big and little endian on ARM and PowerPc. A patch by Ira W.
Snyder fixes uninitialized variable warnings in the janz-ican3 driver.
Rostislav Lisovy contributes a patch to propagate the SO_PRIORITY of raw
sockets to skbs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoe100: Fix "disabling already-disabled device" warning
Michele Baldessari [Thu, 30 Jan 2014 10:51:04 +0000 (10:51 +0000)]
e100: Fix "disabling already-disabled device" warning

In https://bugzilla.redhat.com/show_bug.cgi?id=994438 and
https://bugzilla.redhat.com/show_bug.cgi?id=970480  we
received different reports of e100 throwing the following
warning:

 [<c06a0ba5>] ? pci_disable_device+0x85/0x90
 [<c044a153>] warn_slowpath_fmt+0x33/0x40
 [<c06a0ba5>] pci_disable_device+0x85/0x90
 [<f7fdf7e0>] __e100_shutdown+0x80/0x120 [e100]
 [<c0476ca5>] ? check_preempt_curr+0x65/0x90
 [<f7fdf8d6>] e100_suspend+0x16/0x30 [e100]
 [<c06a1ebb>] pci_legacy_suspend+0x2b/0xb0
 [<c098fc0f>] ? wait_for_completion+0x1f/0xd0
 [<c06a2d50>] ? pci_pm_poweroff+0xb0/0xb0
 [<c06a2de4>] pci_pm_freeze+0x94/0xa0
 [<c0767bb7>] dpm_run_callback+0x37/0x80
 [<c076a204>] ? pm_wakeup_pending+0xc4/0x140
 [<c0767f12>] __device_suspend+0xb2/0x1f0
 [<c076806f>] async_suspend+0x1f/0x90
 [<c04706e5>] async_run_entry_fn+0x35/0x140
 [<c0478aef>] ? wake_up_process+0x1f/0x40
 [<c0464495>] process_one_work+0x115/0x370
 [<c0462645>] ? start_worker+0x25/0x30
 [<c0464dc5>] ? manage_workers.isra.27+0x1a5/0x250
 [<c0464f6e>] worker_thread+0xfe/0x330
 [<c0464e70>] ? manage_workers.isra.27+0x250/0x250
 [<c046a224>] kthread+0x94/0xa0
 [<c0997f37>] ret_from_kernel_thread+0x1b/0x28
 [<c046a190>] ? insert_kthread_work+0x30/0x30

This patch removes pci_disable_device() from __e100_shutdown().
pci_clear_master() is enough.

Signed-off-by: Michele Baldessari <michele@acksyn.org>
Tested-by: Mark Harig <idirectscm@aim.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>