openwrt/staging/blogic.git
11 years agopkt_sched: fq: warn users using defrate
Eric Dumazet [Fri, 15 Nov 2013 16:57:26 +0000 (08:57 -0800)]
pkt_sched: fq: warn users using defrate

Commit 7eec4174ff29 ("pkt_sched: fq: fix non TCP flows pacing")
obsoleted TCA_FQ_FLOW_DEFAULT_RATE without notice for the users.

Suggested by David Miller

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: unify registration functions
Johannes Berg [Fri, 15 Nov 2013 13:19:08 +0000 (14:19 +0100)]
genetlink: unify registration functions

Now that the ops assignment is just two variables rather than a
long list iteration etc., there's no reason to separately export
__genl_register_family() and __genl_register_family_with_ops().

Unify the two functions into __genl_register_family() and make
genl_register_family_with_ops() call it after assigning the ops.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMAINTAINERS: Update Pegasus and RTL8150 repositories;
Petko Manolov [Fri, 15 Nov 2013 09:51:54 +0000 (11:51 +0200)]
MAINTAINERS: Update Pegasus and RTL8150 repositories;

The diff is against latest 'net' repository;

Signed-off-by: Petko Manolov <petkan@nucleusys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: ethernet: ti/cpsw: do not crash on single-MAC machines during resume
Daniel Mack [Fri, 15 Nov 2013 07:29:16 +0000 (08:29 +0100)]
net: ethernet: ti/cpsw: do not crash on single-MAC machines during resume

During resume, use for_each_slave to walk the slaves of the cpsw, and
soft-reset each of them. This prevents oopses if there is only one
slave configured.

Signed-off-by: Daniel Mack <zonque@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'macvlan'
David S. Miller [Fri, 15 Nov 2013 22:55:58 +0000 (17:55 -0500)]
Merge branch 'macvlan'

Michal Kubecek says:

====================
macvlan: disable LRO on lowerdev instead of a macvlan

A customer of ours encountered a problem with LRO on an ixgbe network
card. Analysis showed that it was a known conflict of forwarding and LRO
but the forwarding was enabled in an LXC container where only a macvlan
was, not the ethernet device itself.

I believe the solution is exactly the same as what we do for "normal"
(802.1q) VLAN devices: if dev_disable_lro() is called for such device,
LRO is disabled on the underlying "real" device instead.

v2: adapt to changes merged from net-next

v3: use BUG() in macvlan_dev_real_dev() if compiled without macvlan
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomacvlan: disable LRO on lower device instead of macvlan
Michal Kubeček [Fri, 15 Nov 2013 05:18:50 +0000 (06:18 +0100)]
macvlan: disable LRO on lower device instead of macvlan

A macvlan device has always LRO disabled so that calling
dev_disable_lro() on it does nothing. If we need to disable LRO
e.g. because

  - the macvlan device is inserted into a bridge
  - IPv6 forwarding is enabled for it
  - it is in a different namespace than lowerdev and IPv4
    forwarding is enabled in it

we need to disable LRO on its underlying device instead (as we
do for 802.1q VLAN devices).

v2: use newly introduced netif_is_macvlan()

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomacvlan: introduce macvlan_dev_real_dev() helper function
Michal Kubeček [Fri, 15 Nov 2013 05:18:40 +0000 (06:18 +0100)]
macvlan: introduce macvlan_dev_real_dev() helper function

Introduce helper function macvlan_dev_real_dev which returns the
underlying device of a macvlan device, similar to vlan_dev_real_dev()
for 802.1q VLAN devices.

v2: IFF_MACVLAN flag and equivalent of is_macvlan_dev() were
introduced in the meantime

v3: do BUG() if compiled without macvlan support

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: add ip checks when store ip target
Wang Weidong [Fri, 15 Nov 2013 15:34:30 +0000 (10:34 -0500)]
bonding: add ip checks when store ip target

I met a Bug when I add ip target with the wrong ip address:

echo +500.500.500.500 > /sys/class/net/bond0/bonding/arp_ip_target

the wrong ip address will transfor to 245.245.245.244 and add
to the ip target success, it is uncorrect, so I add checks to avoid
adding wrong address.

The in4_pton() will set wrong ip address to 0.0.0.0, it will return by
the next check and will not add to ip target.

v2
According Veaceslav's opinion, simplify the code.

v3
According Veaceslav's opinion, add broadcast check and make a micro
definition to package it.

v4
Solve the problem of the format which David point out.

Suggested-by: Veaceslav Falico <vfalico@redhat.com>
Suggested-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years ago6lowpan: Uncompression of traffic class field was incorrect
Jukka Rissanen [Wed, 13 Nov 2013 09:03:39 +0000 (11:03 +0200)]
6lowpan: Uncompression of traffic class field was incorrect

If priority/traffic class field in IPv6 header is set (seen when
using ssh), the uncompression sets the TC and Flow fields incorrectly.

Example:

This is IPv6 header of a sent packet. Note the priority/TC (=1) in
the first byte.

00000000: 61 00 00 00 00 2c 06 40 fe 80 00 00 00 00 00 00
00000010: 02 02 72 ff fe c6 42 10 fe 80 00 00 00 00 00 00
00000020: 02 1e ab ff fe 4c 52 57

This gets compressed like this in the sending side

00000000: 72 31 04 06 02 1e ab ff fe 4c 52 57 ec c2 00 16
00000010: aa 2d fe 92 86 4e be c6 ....

In the receiving end, the packet gets uncompressed to this
IPv6 header

00000000: 60 06 06 02 00 2a 1e 40 fe 80 00 00 00 00 00 00
00000010: 02 02 72 ff fe c6 42 10 fe 80 00 00 00 00 00 00
00000020: ab ff fe 4c 52 57 ec c2

First four bytes are set incorrectly and we have also lost
two bytes from destination address.

The fix is to switch the case values in switch statement
when checking the TC field.

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotipc: fix dereference before check warning
Erik Hugne [Wed, 13 Nov 2013 08:35:11 +0000 (09:35 +0100)]
tipc: fix dereference before check warning

This fixes the following Smatch warning:
net/tipc/link.c:2364 tipc_link_recv_fragment()
    warn: variable dereferenced before check '*head' (see line 2361)

A null pointer might be passed to skb_try_coalesce if
a malicious sender injects orphan fragments on a link.

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoipv4: fix possible seqlock deadlock
Eric Dumazet [Thu, 14 Nov 2013 21:37:54 +0000 (13:37 -0800)]
ipv4: fix possible seqlock deadlock

ip4_datagram_connect() being called from process context,
it should use IP_INC_STATS() instead of IP_INC_STATS_BH()
otherwise we can deadlock on 32bit arches, or get corruptions of
SNMP counters.

Fixes: 584bdf8cbdf6 ("[IPV4]: Fix "ipOutNoRoutes" counter error for TCP and UDP")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet/hsr: Fix possible leak in 'hsr_get_node_status()'
Geyslan G. Bem [Thu, 14 Nov 2013 19:12:54 +0000 (16:12 -0300)]
net/hsr: Fix possible leak in 'hsr_get_node_status()'

If 'hsr_get_node_data()' returns error, going directly to 'fail' label
doesn't free the memory pointed by 'skb_out'.

Signed-off-by: Geyslan G. Bem <geyslan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Thu, 14 Nov 2013 22:24:01 +0000 (17:24 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless

John W. Linville says:

====================
pull request: wireless 2013-11-14

Please pull this batch of fixes intended for the 3.13 stream!

Amitkumar Karwar offers a quartet of mwifiex fixes, including an
endian fix and three fixes for invalid memory access.

Avinash Patil trims the packet length value for packets received from
an SDIO interface.

Colin Ian King fixes a NULL pointer dereference in the rtlwifi
efuse code.

Dan Carpenter cleans-up an mwifiex integer underflow, a potential
libertas oops, a memory corrupion bug in wcn36xx, and a locking issue
also in wcn36xx.

Dan Williams helps prism54 devices to avoid being misclassified as
Ethernet devices.

Felipe Pena fixes a couple of typo errors, one in rt2x00 and the
other in rtlwifi.

Janusz Dziedzic corrects a pair of DFS-related problems in ath9k.

Larry Finger patches three rtlwifi drivers to correctly report signal
strength even for an unassociated AP.

Mark Cave-Ayland rewrites some endian-illiterate packet type extraction
code in rtlwifi.

Stanislaw Gruszka addresses an rt2x00 regression related to setting
HT station WCID and AMPDU density parameters.

Sujith Manoharan corrects the initvals settings for AR9485.

Ujjal Roy patches an obscure bit of code in mwifiex that was using
the wrong definition of eth_hdr when briding patches in AP mode.

Wei Yongjun fixes a couple of bugs: one is a return code handling
bug in libertas; and, the other is a locking issue in wcn36xx.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agovirtio-net: mergeable buffer size should include virtio-net header
Michael Dalton [Thu, 14 Nov 2013 18:41:04 +0000 (10:41 -0800)]
virtio-net: mergeable buffer size should include virtio-net header

Commit 2613af0ed18a ("virtio_net: migrate mergeable rx buffers to page
frag allocators") changed the mergeable receive buffer size from PAGE_SIZE
to MTU-size. However, the merge buffer size does not take into account the
size of the virtio-net header. Consequently, packets that are MTU-size
will take two buffers intead of one (to store the virtio-net header),
substantially decreasing the throughput of MTU-size traffic due to TCP
window / SKB truesize effects.

This commit changes the mergeable buffer size to include the virtio-net
header. The buffer size is cacheline-aligned because skb_page_frag_refill
will not automatically align the requested size.

Benchmarks taken from an average of 5 netperf 30-second TCP_STREAM runs
between two QEMU VMs on a single physical machine. Each VM has two VCPUs and
vhost enabled. All VMs and vhost threads run in a single 4 CPU cgroup
cpuset, using cgroups to ensure that other processes in the system will not
be scheduled on the benchmark CPUs. Transmit offloads and mergeable receive
buffers are enabled, but guest_tso4 / guest_csum are explicitly disabled to
force MTU-sized packets on the receiver.

next-net trunk before 2613af0ed18a (PAGE_SIZE buf): 3861.08Gb/s
net-next trunk (MTU 1500- packet uses two buf due to size bug): 4076.62Gb/s
net-next trunk (MTU 1480- packet fits in one buf): 6301.34Gb/s
net-next trunk w/ size fix (MTU 1500 - packet fits in one buf): 6445.44Gb/s

Suggested-by: Eric Northup <digitaleric@google.com>
Signed-off-by: Michael Dalton <mwdalton@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoconnector: improved unaligned access error fix
Chris Metcalf [Thu, 14 Nov 2013 17:09:21 +0000 (12:09 -0500)]
connector: improved unaligned access error fix

In af3e095a1fb4, Erik Jacobsen fixed one type of unaligned access
bug for ia64 by converting a 64-bit write to use put_unaligned().
Unfortunately, since gcc will convert a short memset() to a series
of appropriately-aligned stores, the problem is now visible again
on tilegx, where the memset that zeros out proc_event is converted
to three 64-bit stores, causing an unaligned access panic.

A better fix for the original problem is to ensure that proc_event
is aligned to 8 bytes here.  We can do that relatively easily by
arranging to start the struct cn_msg aligned to 8 bytes and then
offset by 4 bytes.  Doing so means that the immediately following
proc_event structure is then correctly aligned to 8 bytes.

The result is that the memset() stores are now aligned, and as an
added benefit, we can remove the put_unaligned() calls in the code.

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopkt_sched: fq: change classification of control packets
Maciej Żenczykowski [Thu, 14 Nov 2013 16:50:43 +0000 (08:50 -0800)]
pkt_sched: fq: change classification of control packets

Initial sch_fq implementation copied code from pfifo_fast to classify
a packet as a high prio packet.

This clashes with setups using PRIO with say 7 bands, as one of the
band could be incorrectly (mis)classified by FQ.

Packets would be queued in the 'internal' queue, and no pacing ever
happen for this special queue.

Fixes: afe4fd062416 ("pkt_sched: fq: Fair Queue packet scheduler")
Signed-off-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoalx: Reset phy speed after resume
hahnjo [Tue, 12 Nov 2013 17:19:24 +0000 (18:19 +0100)]
alx: Reset phy speed after resume

This fixes bug 62491 (https://bugzilla.kernel.org/show_bug.cgi?id=62491).
After resuming some users got the following error flooding the kernel log:
alx 0000:02:00.0: invalid PHY speed/duplex: 0xffff

Signed-off-by: Jonas Hahnfeld <linux@hahnjo.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'genetlink'
David S. Miller [Thu, 14 Nov 2013 22:11:17 +0000 (17:11 -0500)]
Merge branch 'genetlink'

Johannes Berg says:

====================
genetlink: reduce ops size and complexity (v2)

As before - reduce the complexity and data/code size of genetlink ops
by making them an array rather than a linked list. Most users already
use an array thanks to genl_register_family_with_ops(), so convert the
remaining ones allowing us to get rid of the list head in each op.

Also make them const, this just makes sense at that point and the security
people like making function pointers const as well :-)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: make genl_ops flags a u8 and move to end
Johannes Berg [Thu, 14 Nov 2013 16:14:47 +0000 (17:14 +0100)]
genetlink: make genl_ops flags a u8 and move to end

To save some space in the struct on 32-bit systems,
make the flags a u8 (only 4 bits are used) and also
move them to the end of the struct.

This has no impact on 64-bit systems as alignment of
the struct in an array uses up the space anyway.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: make all genl_ops users const
Johannes Berg [Thu, 14 Nov 2013 16:14:46 +0000 (17:14 +0100)]
genetlink: make all genl_ops users const

Now that genl_ops are no longer modified in place when
registering, they can be made const. This patch was done
mostly with spatch:

@@
identifier ops;
@@
+const
 struct genl_ops ops[] = {
 ...
 };

(except the struct thing in net/openvswitch/datapath.c)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: allow making ops const
Johannes Berg [Thu, 14 Nov 2013 16:14:45 +0000 (17:14 +0100)]
genetlink: allow making ops const

Allow making the ops array const by not modifying the ops
flags on registration but rather only when ops are sent
out in the family information.

No users are updated yet except for the pre_doit/post_doit
calls in wireless (the only ones that exist now.)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: register family ops as array
Johannes Berg [Thu, 14 Nov 2013 16:14:44 +0000 (17:14 +0100)]
genetlink: register family ops as array

Instead of using a linked list, use an array. This reduces
the data size needed by the users of genetlink, for example
in wireless (net/wireless/nl80211.c) on 64-bit it frees up
over 1K of data space.

Remove the attempted sending of CTRL_CMD_NEWOPS ctrl event
since genl_ctrl_event(CTRL_CMD_NEWOPS, ...) only returns
-EINVAL anyway, therefore no such event could ever be sent.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agogenetlink: remove genl_register_ops/genl_unregister_ops
Johannes Berg [Thu, 14 Nov 2013 16:14:43 +0000 (17:14 +0100)]
genetlink: remove genl_register_ops/genl_unregister_ops

genl_register_ops() is still needed for internal registration,
but is no longer available to users of the API.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agowimax: use genl_register_family_with_ops()
Johannes Berg [Thu, 14 Nov 2013 16:14:42 +0000 (17:14 +0100)]
wimax: use genl_register_family_with_ops()

This simplifies the code since there's no longer a need to
have error handling in the registration.

Unfortunately it means more extern function declarations are
needed, but the overall goal would seem to justify this.

Due to the removal of duplication in the netlink policies,
this reduces the size of wimax by almost 1k.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoieee802154: use genl_register_family_with_ops()
Johannes Berg [Thu, 14 Nov 2013 16:14:41 +0000 (17:14 +0100)]
ieee802154: use genl_register_family_with_ops()

This simplifies the code since there's no longer a need to
have error handling in the registration.

Unfortunately it means more extern function declarations are
needed, but the overall goal would seem to justify this.

While at it, also fix the registration error path - if the
family registration failed then it shouldn't be unregistered.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agohsr: use genl_register_family_with_ops()
Johannes Berg [Thu, 14 Nov 2013 16:14:40 +0000 (17:14 +0100)]
hsr: use genl_register_family_with_ops()

This simplifies the code since there's no longer a
need to have error handling in the registration.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotaskstats: use genl_register_family_with_ops()
Johannes Berg [Thu, 14 Nov 2013 16:14:39 +0000 (17:14 +0100)]
taskstats: use genl_register_family_with_ops()

This simplifies the code since there's no longer a
need to have error handling in the registration.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoip6tnl: fix use after free of fb_tnl_dev
Nicolas Dichtel [Thu, 14 Nov 2013 14:47:03 +0000 (15:47 +0100)]
ip6tnl: fix use after free of fb_tnl_dev

Bug has been introduced by commit bb8140947a24 ("ip6tnl: allow to use rtnl ops
on fb tunnel").

When ip6_tunnel.ko is unloaded, FB device is delete by rtnl_link_unregister()
and then we try to use the pointer in ip6_tnl_destroy_tunnels().

Let's add an handler for dellink, which will never remove the FB tunnel. With
this patch it will no more be possible to remove it via 'ip link del ip6tnl0',
but it's safer.

The same fix was already proposed by Willem de Bruijn <willemb@google.com> for
sit interfaces.

CC: Willem de Bruijn <willemb@google.com>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosit/gre6: don't try to add the same route two times
Nicolas Dichtel [Thu, 14 Nov 2013 12:51:07 +0000 (13:51 +0100)]
sit/gre6: don't try to add the same route two times

addrconf_add_linklocal() already adds the link local route, so there is no
reason to add it before calling this function.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosit: link local routes are missing
Nicolas Dichtel [Thu, 14 Nov 2013 12:51:06 +0000 (13:51 +0100)]
sit: link local routes are missing

When a link local address was added to a sit interface, the corresponding route
was not configured. This breaks routing protocols that use the link local
address, like OSPFv3.

To ease the code reading, I remove sit_route_add(), which only adds v4 mapped
routes, and add this kind of route directly in sit_add_v4_addrs(). Thus link
local and v4 mapped routes are configured in the same place.

Reported-by: Li Hongjun <hongjun.li@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosit: fix prefix length of ll and v4mapped addresses
Nicolas Dichtel [Thu, 14 Nov 2013 12:51:05 +0000 (13:51 +0100)]
sit: fix prefix length of ll and v4mapped addresses

When the local IPv4 endpoint is wilcard (0.0.0.0), the prefix length is
correctly set, ie 64 if the address is a link local one or 96 if the address is
a v4 mapped one.
But when the local endpoint is specified, the prefix length is set to 128 for
both kind of address. This patch fix this wrong prefix length.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoisdnloop: use strlcpy() instead of strcpy()
Dan Carpenter [Thu, 14 Nov 2013 08:21:10 +0000 (11:21 +0300)]
isdnloop: use strlcpy() instead of strcpy()

These strings come from a copy_from_user() and there is no way to be
sure they are NUL terminated.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agosit: fix use after free of fb_tunnel_dev
Willem de Bruijn [Thu, 14 Nov 2013 02:27:38 +0000 (21:27 -0500)]
sit: fix use after free of fb_tunnel_dev

Bug: The fallback device is created in sit_init_net and assumed to be
freed in sit_exit_net. First, it is dereferenced in that function, in
sit_destroy_tunnels:

        struct net *net = dev_net(sitn->fb_tunnel_dev);

Prior to this, rtnl_unlink_register has removed all devices that match
rtnl_link_ops == sit_link_ops.

Commit 205983c43700 added the line

+       sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;

which cases the fallback device to match here and be freed before it
is last dereferenced.

Fix: This commit adds an explicit .delllink callback to sit_link_ops
that skips deallocation at rtnl_unlink_register for the fallback
device. This mechanism is comparable to the one in ip_tunnel.

It also modifies sit_destroy_tunnels and its only caller sit_exit_net
to avoid the offending dereference in the first place. That double
lookup is more complicated than required.

Test: The bug is only triggered when CONFIG_NET_NS is enabled. It
causes a GPF only when CONFIG_DEBUG_SLAB is enabled. Verified that
this bug exists at the mentioned commit, at davem-net HEAD and at
3.11.y HEAD. Verified that it went away after applying this patch.

Fixes: 205983c43700 ("sit: allow to use rtnl ops on fb tunnel")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet:fec: fix WARNING caused by lack of calls to dma_mapping_error()
Duan Fugang-B38611 [Thu, 14 Nov 2013 01:57:10 +0000 (09:57 +0800)]
net:fec: fix WARNING caused by lack of calls to dma_mapping_error()

The driver fails to check the results of DMA mapping and results in
the following warning: (with kernel config "CONFIG_DMA_API_DEBUG" enable)

------------[ cut here ]------------
WARNING: at lib/dma-debug.c:937 check_unmap+0x43c/0x7d8()
fec 2188000.ethernet: DMA-API: device driver failed to check map
error[device address=0x00000000383a8040] [size=2048 bytes] [mapped as single]

Modules linked in:
CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.10.17-16827-g9cdb0ba-dirty #188
[<80013c4c>] (unwind_backtrace+0x0/0xf8) from [<80011704>] (show_stack+0x10/0x14)
[<80011704>] (show_stack+0x10/0x14) from [<80025614>] (warn_slowpath_common+0x4c/0x6c)
[<80025614>] (warn_slowpath_common+0x4c/0x6c) from [<800256c8>] (warn_slowpath_fmt+0x30/0x40)
[<800256c8>] (warn_slowpath_fmt+0x30/0x40) from [<8026bfdc>] (check_unmap+0x43c/0x7d8)
[<8026bfdc>] (check_unmap+0x43c/0x7d8) from [<8026c584>] (debug_dma_unmap_page+0x6c/0x78)
[<8026c584>] (debug_dma_unmap_page+0x6c/0x78) from [<8038049c>] (fec_enet_rx_napi+0x254/0x8a8)
[<8038049c>] (fec_enet_rx_napi+0x254/0x8a8) from [<804dc8c0>] (net_rx_action+0x94/0x160)
[<804dc8c0>] (net_rx_action+0x94/0x160) from [<8002c758>] (__do_softirq+0xe8/0x1d0)
[<8002c758>] (__do_softirq+0xe8/0x1d0) from [<8002c8e8>] (do_softirq+0x4c/0x58)
[<8002c8e8>] (do_softirq+0x4c/0x58) from [<8002cb50>] (irq_exit+0x90/0xc8)
[<8002cb50>] (irq_exit+0x90/0xc8) from [<8000ea88>] (handle_IRQ+0x3c/0x94)
[<8000ea88>] (handle_IRQ+0x3c/0x94) from [<8000855c>] (gic_handle_irq+0x28/0x5c)
[<8000855c>] (gic_handle_irq+0x28/0x5c) from [<8000de00>] (__irq_svc+0x40/0x50)
Exception stack(0x815a5f38 to 0x815a5f80)
5f20:                                                       815a5f80 3b9aca00
5f40: 0fe52383 00000002 0dd8950e 00000002 81e7b080 00000000 00000000 815ac4d8
5f60: 806032ec 00000000 00000017 815a5f80 80059028 8041fc4c 60000013 ffffffff
[<8000de00>] (__irq_svc+0x40/0x50) from [<8041fc4c>] (cpuidle_enter_state+0x50/0xf0)
[<8041fc4c>] (cpuidle_enter_state+0x50/0xf0) from [<8041fd94>] (cpuidle_idle_call+0xa8/0x14c)
[<8041fd94>] (cpuidle_idle_call+0xa8/0x14c) from [<8000edac>] (arch_cpu_idle+0x10/0x4c)
[<8000edac>] (arch_cpu_idle+0x10/0x4c) from [<800582f8>] (cpu_startup_entry+0x60/0x130)
[<800582f8>] (cpu_startup_entry+0x60/0x130) from [<80bc7a48>] (start_kernel+0x2d0/0x328)
[<80bc7a48>] (start_kernel+0x2d0/0x328) from [<10008074>] (0x10008074)
---[ end trace c6edec32436e0042 ]---

Because dma-debug add new interfaces to debug dma mapping errors, pls refer
to: http://lwn.net/Articles/516640/

After dma mapping, it must call dma_mapping_error() to check mapping error,
otherwise the map_err_type alway is MAP_ERR_NOT_CHECKED, check_unmap() define
the mapping is not checked and dump the error msg. So,add dma_mapping_error()
checking to fix the WARNING

And RX DMA buffers are used repeatedly and the driver copies it into an skb,
fec_enet_rx() should not map or unmap, use dma_sync_single_for_cpu()/dma_sync_single_for_device()
instead of dma_map_single()/dma_unmap_single().

There have another potential issue:  fec_enet_rx() passes the DMA address to __va().
Physical and DMA addresses are *not* the same thing. They may differ if the device
is behind an IOMMU or bounce buffering was required, or just because there is a fixed
offset between the device and host physical addresses. Also fix it in this patch.

=============================================
V2: add net_ratelimit() to limit map err message.
    use dma_sync_single_for_cpu() instead of dma_map_single().
    fix the issue that pass DMA addresses to __va() to get virture address.
V1: initial send
=============================================

Signed-off-by: Fugang Duan <B38611@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: sctp: bug-fixing: retran_path not set properly after transports recovering (v3)
Chang Xiangzhong [Wed, 13 Nov 2013 23:58:26 +0000 (00:58 +0100)]
net: sctp: bug-fixing: retran_path not set properly after transports recovering (v3)

When a transport recovers due to the new coming sack, SCTP should
iterate all of its transport_list to locate the __two__ most recently used
transport and set to active_path and retran_path respectively. The exising
code does not find the two properly - In case of the following list:

[most-recent] -> [2nd-most-recent] -> ...

Both active_path and retran_path would be set to the 1st element.

The bug happens when:
1) multi-homing
2) failure/partial_failure transport recovers
Both active_path and retran_path would be set to the same most-recent one, in
other words, retran_path would not take its role - an end user might not even
notice this issue.

Signed-off-by: Chang Xiangzhong <changxiangzhong@gmail.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet-tcp: fix panic in tcp_fastopen_cache_set()
Eric Dumazet [Wed, 13 Nov 2013 23:00:46 +0000 (15:00 -0800)]
net-tcp: fix panic in tcp_fastopen_cache_set()

We had some reports of crashes using TCP fastopen, and Dave Jones
gave a nice stack trace pointing to the error.

Issue is that tcp_get_metrics() should not be called with a NULL dst

Fixes: 1fe4c481ba637 ("net-tcp: Fast Open client - cookie cache")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dave Jones <davej@redhat.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Tested-by: Dave Jones <davej@fedoraproject.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: fix two race conditions in bond_store_updelay/downdelay
Nikolay Aleksandrov [Wed, 13 Nov 2013 16:07:46 +0000 (17:07 +0100)]
bonding: fix two race conditions in bond_store_updelay/downdelay

This patch fixes two race conditions between bond_store_updelay/downdelay
and bond_store_miimon which could lead to division by zero as miimon can
be set to 0 while either updelay/downdelay are being set and thus miss the
zero check in the beginning, the zero div happens because updelay/downdelay
are stored as new_value / bond->params.miimon. Use rtnl to synchronize with
miimon setting.

CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Acked-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotcp: tsq: restore minimal amount of queueing
Eric Dumazet [Wed, 13 Nov 2013 14:32:54 +0000 (06:32 -0800)]
tcp: tsq: restore minimal amount of queueing

After commit c9eeec26e32e ("tcp: TSQ can use a dynamic limit"), several
users reported throughput regressions, notably on mvneta and wifi
adapters.

802.11 AMPDU requires a fair amount of queueing to be effective.

This patch partially reverts the change done in tcp_write_xmit()
so that the minimal amount is sysctl_tcp_limit_output_bytes.

It also remove the use of this sysctl while building skb stored
in write queue, as TSO autosizing does the right thing anyway.

Users with well behaving NICS and correct qdisc (like sch_fq),
can then lower the default sysctl_tcp_limit_output_bytes value from
128KB to 8KB.

This new usage of sysctl_tcp_limit_output_bytes permits each driver
authors to check how their driver performs when/if the value is set
to a minimum of 4KB.

Normally, line rate for a single TCP flow should be possible,
but some drivers rely on timers to perform TX completion and
too long TX completion delays prevent reaching full throughput.

Fixes: c9eeec26e32e ("tcp: TSQ can use a dynamic limit")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Sujith Manoharan <sujith@msujith.org>
Reported-by: Arnaud Ebalard <arno@natisbad.org>
Tested-by: Sujith Manoharan <sujith@msujith.org>
Cc: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'hwtstamp'
David S. Miller [Thu, 14 Nov 2013 21:22:25 +0000 (16:22 -0500)]
Merge branch 'hwtstamp'

Ben Hutchings says:

====================
net_tstamp: Validate hwtstamp_config completely before applying it

This series fixes very similar bugs in 6 implementations of
SIOCSHWTSTAMP.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoixp4xx_eth: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:48:56 +0000 (00:48 +0000)]
ixp4xx_eth: Validate hwtstamp_config completely before applying it

hwtstamp_ioctl() should validate all fields of hwtstamp_config
before making any changes.  Currently it sets the TX configuration
before validating the rx_filter field.

Untested as I don't have a cross-compiler to hand.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoti_cpsw: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:47:36 +0000 (00:47 +0000)]
ti_cpsw: Validate hwtstamp_config completely before applying it

cpsw_hwtstamp_ioctl() should validate all fields of hwtstamp_config,
and the hardware version, before making any changes.  Currently it
sets the TX configuration before validating the rx_filter field
or that the hardware supports timestamping.

Also correct the error code for hardware versions that don't
support timestamping.  ENOTSUPP is used by the NFS implementation
and is not part of userland API; we want EOPNOTSUPP (which glibc
also calls ENOTSUP, with one 'P').

Untested as I don't have a cross-compiler to hand.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Mugunthan V N <mugunthanvnm@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agostmmac: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:43:41 +0000 (00:43 +0000)]
stmmac: Validate hwtstamp_config completely before applying it

stmmac_hwtstamp_ioctl() should validate all fields of hwtstamp_config
before making any changes.  Currently it sets the TX configuration
before validating the rx_filter field.

Compile-tested only.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agopch_gbe: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:42:50 +0000 (00:42 +0000)]
pch_gbe: Validate hwtstamp_config completely before applying it

hwtstamp_ioctl() should validate all fields of hwtstamp_config
before making any changes.  Currently it sets the TX configuration
before validating the rx_filter field.

Compile-tested only.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoe1000e: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:41:38 +0000 (00:41 +0000)]
e1000e: Validate hwtstamp_config completely before applying it

e1000e_hwtstamp_ioctl() should validate all fields of hwtstamp_config
before making any changes.  Currently it copies the configuration to
the e1000_adapter structure before validating it at all.

Change e1000e_config_hwtstamp() to take a pointer to the
hwstamp_config and to copy the config after validating it.

Compile-tested only.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotg3: Validate hwtstamp_config completely before applying it
Ben Hutchings [Thu, 14 Nov 2013 00:40:56 +0000 (00:40 +0000)]
tg3: Validate hwtstamp_config completely before applying it

tg3_hwtstamp_ioctl() should validate all fields of hwtstamp_config
before making any changes.  Currently it sets the TX configuration
before validating the rx_filter field.

Compile-tested only.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Acked-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: Fix memory leak when deleting bridge with vlan filtering enabled
Toshiaki Makita [Wed, 13 Nov 2013 08:26:14 +0000 (17:26 +0900)]
bridge: Fix memory leak when deleting bridge with vlan filtering enabled

We currently don't call br_vlan_flush() when deleting a bridge, which
leads to memory leak if br->vlan_info is allocated.

Steps to reproduce:
  while :
  do
    brctl addbr br0
    bridge vlan add dev br0 vid 10 self
    brctl delbr br0
  done
We can observe the cache size of corresponding slab entry
(as kmalloc-2048 in SLUB) is increased.

kmemleak output:
unreferenced object 0xffff8800b68a7000 (size 2048):
  comm "bridge", pid 2086, jiffies 4295774704 (age 47.656s)
  hex dump (first 32 bytes):
    00 00 00 00 00 00 00 00 00 48 9b 36 00 88 ff ff  .........H.6....
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<ffffffff815eb6ae>] kmemleak_alloc+0x4e/0xb0
    [<ffffffff8116a1ca>] kmem_cache_alloc_trace+0xca/0x220
    [<ffffffffa03eddd6>] br_vlan_add+0x66/0xe0 [bridge]
    [<ffffffffa03e543c>] br_setlink+0x2dc/0x340 [bridge]
    [<ffffffff8150e481>] rtnl_bridge_setlink+0x101/0x200
    [<ffffffff8150d9d9>] rtnetlink_rcv_msg+0x99/0x260
    [<ffffffff81528679>] netlink_rcv_skb+0xa9/0xc0
    [<ffffffff8150d938>] rtnetlink_rcv+0x28/0x30
    [<ffffffff81527ccd>] netlink_unicast+0xdd/0x190
    [<ffffffff8152807f>] netlink_sendmsg+0x2ff/0x740
    [<ffffffff814e8368>] sock_sendmsg+0x88/0xc0
    [<ffffffff814e8ac8>] ___sys_sendmsg.part.14+0x298/0x2b0
    [<ffffffff814e91de>] __sys_sendmsg+0x4e/0x90
    [<ffffffff814e922e>] SyS_sendmsg+0xe/0x10
    [<ffffffff81601669>] system_call_fastpath+0x16/0x1b
    [<ffffffffffffffff>] 0xffffffffffffffff

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: Call vlan_vid_del for all vids at nbp_vlan_flush
Toshiaki Makita [Wed, 13 Nov 2013 08:26:13 +0000 (17:26 +0900)]
bridge: Call vlan_vid_del for all vids at nbp_vlan_flush

We should call vlan_vid_del for all vids at nbp_vlan_flush to prevent
vid_info->refcount from being leaked when detaching a bridge port.

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobridge: Use vlan_vid_[add/del] instead of direct ndo_vlan_rx_[add/kill]_vid calls
Toshiaki Makita [Wed, 13 Nov 2013 08:26:12 +0000 (17:26 +0900)]
bridge: Use vlan_vid_[add/del] instead of direct ndo_vlan_rx_[add/kill]_vid calls

We should use wrapper functions vlan_vid_[add/del] instead of
ndo_vlan_rx_[add/kill]_vid. Otherwise, we might be not able to communicate
using vlan interface in a certain situation.

Example of problematic case:
  vconfig add eth0 10
  brctl addif br0 eth0
  bridge vlan add dev eth0 vid 10
  bridge vlan del dev eth0 vid 10
  brctl delif br0 eth0
In this case, we cannot communicate via eth0.10 because vlan 10 is
filtered by NIC that has the vlan filtering feature.

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agorandom32: use msecs_to_jiffies for reseed timer
Daniel Borkmann [Tue, 12 Nov 2013 22:45:42 +0000 (23:45 +0100)]
random32: use msecs_to_jiffies for reseed timer

Use msecs_to_jiffies, for these calculations as different HZ
considerations are taken into account for conversion of the timer
shot, and also it makes the code more readable.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agorandom32: add __init prefix to prandom_start_seed_timer
Daniel Borkmann [Tue, 12 Nov 2013 22:45:41 +0000 (23:45 +0100)]
random32: add __init prefix to prandom_start_seed_timer

We only call that in functions annotated with __init, so add __init
prefix in prandom_start_seed_timer() as well, so that the kernel can
make use of this hint and we can possibly free up resources after it's
usage. And since it's an internal function rename it to
__prandom_start_seed_timer().

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agomacvtap: limit head length of skb allocated
Jason Wang [Wed, 13 Nov 2013 06:00:40 +0000 (14:00 +0800)]
macvtap: limit head length of skb allocated

We currently use hdr_len as a hint of head length which is advertised by
guest. But when guest advertise a very big value, it can lead to an 64K+
allocating of kmalloc() which has a very high possibility of failure when host
memory is fragmented or under heavy stress. The huge hdr_len also reduce the
effect of zerocopy or even disable if a gso skb is linearized in guest.

To solves those issues, this patch introduces an upper limit (PAGE_SIZE) of the
head, which guarantees an order 0 allocation each time.

Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agotuntap: limit head length of skb allocated
Jason Wang [Wed, 13 Nov 2013 06:00:39 +0000 (14:00 +0800)]
tuntap: limit head length of skb allocated

We currently use hdr_len as a hint of head length which is advertised by
guest. But when guest advertise a very big value, it can lead to an 64K+
allocating of kmalloc() which has a very high possibility of failure when host
memory is fragmented or under heavy stress. The huge hdr_len also reduce the
effect of zerocopy or even disable if a gso skb is linearized in guest.

To solves those issues, this patch introduces an upper limit (PAGE_SIZE) of the
head, which guarantees an order 0 allocation each time.

Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Thu, 14 Nov 2013 18:42:17 +0000 (13:42 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless into for-davem

11 years agonet: mv643xx_eth: potential NULL dereference in probe()
Dan Carpenter [Wed, 13 Nov 2013 07:52:47 +0000 (10:52 +0300)]
net: mv643xx_eth: potential NULL dereference in probe()

We assume that "mp->phy" can be NULL a couple lines before the
dereference.

Fixes: 1cce16d37d0f ('net: mv643xx_eth: Add missing phy_addr_set in DT mode')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agonet: cdc_ncm: cleanup a type issue in cdc_ncm_setup()
Dan Carpenter [Wed, 13 Nov 2013 07:42:34 +0000 (10:42 +0300)]
net: cdc_ncm: cleanup a type issue in cdc_ncm_setup()

This is harmless but cdc_ncm_setup() returns negative error codes
truncated to u8 values.  There is only one caller and treats all
non-zero returns as errors but doesn't store the the return code.  So
the code works correctly but it's messy and upsets the static checkers.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agocore/dev: do not ignore dmac in dev_forward_skb()
Alexei Starovoitov [Tue, 12 Nov 2013 22:39:13 +0000 (14:39 -0800)]
core/dev: do not ignore dmac in dev_forward_skb()

commit 06a23fe31ca3
("core/dev: set pkt_type after eth_type_trans() in dev_forward_skb()")
and refactoring 64261f230a91
("dev: move skb_scrub_packet() after eth_type_trans()")

are forcing pkt_type to be PACKET_HOST when skb traverses veth.

which means that ip forwarding will kick in inside netns
even if skb->eth->h_dest != dev->dev_addr

Fix order of eth_type_trans() and skb_scrub_packet() in dev_forward_skb()
and in ip_tunnel_rcv()

Fixes: 06a23fe31ca3 ("core/dev: set pkt_type after eth_type_trans() in dev_forward_skb()")
CC: Isaku Yamahata <yamahatanetdev@gmail.com>
CC: Maciej Zenczykowski <zenczykowski@gmail.com>
CC: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agousbnet: fix status interrupt urb handling
Felix Fietkau [Tue, 12 Nov 2013 15:34:41 +0000 (16:34 +0100)]
usbnet: fix status interrupt urb handling

Since commit 7b0c5f21f348a66de495868b8df0284e8dfd6bbf
"sierra_net: keep status interrupt URB active", sierra_net triggers
status interrupt polling before the net_device is opened (in order to
properly receive the sync message response).

To be able to receive further interrupts, the interrupt urb needs to be
re-submitted, so this patch removes the bogus check for netif_running().

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Tested-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agobonding: don't permit to use ARP monitoring in 802.3ad mode
Veaceslav Falico [Tue, 12 Nov 2013 14:37:40 +0000 (15:37 +0100)]
bonding: don't permit to use ARP monitoring in 802.3ad mode

Currently the ARP monitoring is not supported with 802.3ad, and it's
prohibited to use it via the module params.

However we still can set it afterwards via sysfs, cause we only check for
*LB modes there.

To fix this - add a check for 802.3ad mode in bonding_store_arp_interval.

CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMAINTAINERS: cxgb3: Update cxgb3 maintainer entry
Divy Le Ray [Mon, 11 Nov 2013 23:01:39 +0000 (15:01 -0800)]
MAINTAINERS: cxgb3: Update cxgb3 maintainer entry

Santosh raspatur is taking over the maintenance of cxgb3.

Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
Linus Torvalds [Wed, 13 Nov 2013 08:40:34 +0000 (17:40 +0900)]
Merge git://git./linux/kernel/git/davem/net-next

Pull networking updates from David Miller:

 1) The addition of nftables.  No longer will we need protocol aware
    firewall filtering modules, it can all live in userspace.

    At the core of nftables is a, for lack of a better term, virtual
    machine that executes byte codes to inspect packet or metadata
    (arriving interface index, etc.) and make verdict decisions.

    Besides support for loading packet contents and comparing them, the
    interpreter supports lookups in various datastructures as
    fundamental operations.  For example sets are supports, and
    therefore one could create a set of whitelist IP address entries
    which have ACCEPT verdicts attached to them, and use the appropriate
    byte codes to do such lookups.

    Since the interpreted code is composed in userspace, userspace can
    do things like optimize things before giving it to the kernel.

    Another major improvement is the capability of atomically updating
    portions of the ruleset.  In the existing netfilter implementation,
    one has to update the entire rule set in order to make a change and
    this is very expensive.

    Userspace tools exist to create nftables rules using existing
    netfilter rule sets, but both kernel implementations will need to
    co-exist for quite some time as we transition from the old to the
    new stuff.

    Kudos to Patrick McHardy, Pablo Neira Ayuso, and others who have
    worked so hard on this.

 2) Daniel Borkmann and Hannes Frederic Sowa made several improvements
    to our pseudo-random number generator, mostly used for things like
    UDP port randomization and netfitler, amongst other things.

    In particular the taus88 generater is updated to taus113, and test
    cases are added.

 3) Support 64-bit rates in HTB and TBF schedulers, from Eric Dumazet
    and Yang Yingliang.

 4) Add support for new 577xx tigon3 chips to tg3 driver, from Nithin
    Sujir.

 5) Fix two fatal flaws in TCP dynamic right sizing, from Eric Dumazet,
    Neal Cardwell, and Yuchung Cheng.

 6) Allow IP_TOS and IP_TTL to be specified in sendmsg() ancillary
    control message data, much like other socket option attributes.
    From Francesco Fusco.

 7) Allow applications to specify a cap on the rate computed
    automatically by the kernel for pacing flows, via a new
    SO_MAX_PACING_RATE socket option.  From Eric Dumazet.

 8) Make the initial autotuned send buffer sizing in TCP more closely
    reflect actual needs, from Eric Dumazet.

 9) Currently early socket demux only happens for TCP sockets, but we
    can do it for connected UDP sockets too.  Implementation from Shawn
    Bohrer.

10) Refactor inet socket demux with the goal of improving hash demux
    performance for listening sockets.  With the main goals being able
    to use RCU lookups on even request sockets, and eliminating the
    listening lock contention.  From Eric Dumazet.

11) The bonding layer has many demuxes in it's fast path, and an RCU
    conversion was started back in 3.11, several changes here extend the
    RCU usage to even more locations.  From Ding Tianhong and Wang
    Yufen, based upon suggestions by Nikolay Aleksandrov and Veaceslav
    Falico.

12) Allow stackability of segmentation offloads to, in particular, allow
    segmentation offloading over tunnels.  From Eric Dumazet.

13) Significantly improve the handling of secret keys we input into the
    various hash functions in the inet hashtables, TCP fast open, as
    well as syncookies.  From Hannes Frederic Sowa.  The key fundamental
    operation is "net_get_random_once()" which uses static keys.

    Hannes even extended this to ipv4/ipv6 fragmentation handling and
    our generic flow dissector.

14) The generic driver layer takes care now to set the driver data to
    NULL on device removal, so it's no longer necessary for drivers to
    explicitly set it to NULL any more.  Many drivers have been cleaned
    up in this way, from Jingoo Han.

15) Add a BPF based packet scheduler classifier, from Daniel Borkmann.

16) Improve CRC32 interfaces and generic SKB checksum iterators so that
    SCTP's checksumming can more cleanly be handled.  Also from Daniel
    Borkmann.

17) Add a new PMTU discovery mode, IP_PMTUDISC_INTERFACE, which forces
    using the interface MTU value.  This helps avoid PMTU attacks,
    particularly on DNS servers.  From Hannes Frederic Sowa.

18) Use generic XPS for transmit queue steering rather than internal
    (re-)implementation in virtio-net.  From Jason Wang.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1622 commits)
  random32: add test cases for taus113 implementation
  random32: upgrade taus88 generator to taus113 from errata paper
  random32: move rnd_state to linux/random.h
  random32: add prandom_reseed_late() and call when nonblocking pool becomes initialized
  random32: add periodic reseeding
  random32: fix off-by-one in seeding requirement
  PHY: Add RTL8201CP phy_driver to realtek
  xtsonic: add missing platform_set_drvdata() in xtsonic_probe()
  macmace: add missing platform_set_drvdata() in mace_probe()
  ethernet/arc/arc_emac: add missing platform_set_drvdata() in arc_emac_probe()
  ipv6: protect for_each_sk_fl_rcu in mem_check with rcu_read_lock_bh
  vlan: Implement vlan_dev_get_egress_qos_mask as an inline.
  ixgbe: add warning when max_vfs is out of range.
  igb: Update link modes display in ethtool
  netfilter: push reasm skb through instead of original frag skbs
  ip6_output: fragment outgoing reassembled skb properly
  MAINTAINERS: mv643xx_eth: take over maintainership from Lennart
  net_sched: tbf: support of 64bit rates
  ixgbe: deleting dfwd stations out of order can cause null ptr deref
  ixgbe: fix build err, num_rx_queues is only available with CONFIG_RPS
  ...

11 years agoMerge branch 'akpm' (patches from Andrew Morton)
Linus Torvalds [Wed, 13 Nov 2013 06:45:43 +0000 (15:45 +0900)]
Merge branch 'akpm' (patches from Andrew Morton)

Merge first patch-bomb from Andrew Morton:
 "Quite a lot of other stuff is banked up awaiting further
  next->mainline merging, but this batch contains:

   - Lots of random misc patches
   - OCFS2
   - Most of MM
   - backlight updates
   - lib/ updates
   - printk updates
   - checkpatch updates
   - epoll tweaking
   - rtc updates
   - hfs
   - hfsplus
   - documentation
   - procfs
   - update gcov to gcc-4.7 format
   - IPC"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (269 commits)
  ipc, msg: fix message length check for negative values
  ipc/util.c: remove unnecessary work pending test
  devpts: plug the memory leak in kill_sb
  ./Makefile: export initial ramdisk compression config option
  init/Kconfig: add option to disable kernel compression
  drivers: w1: make w1_slave::flags long to avoid memory corruption
  drivers/w1/masters/ds1wm.cuse dev_get_platdata()
  drivers/memstick/core/ms_block.c: fix unreachable state in h_msb_read_page()
  drivers/memstick/core/mspro_block.c: fix attributes array allocation
  drivers/pps/clients/pps-gpio.c: remove redundant of_match_ptr
  kernel/panic.c: reduce 1 byte usage for print tainted buffer
  gcov: reuse kbasename helper
  kernel/gcov/fs.c: use pr_warn()
  kernel/module.c: use pr_foo()
  gcov: compile specific gcov implementation based on gcc version
  gcov: add support for gcc 4.7 gcov format
  gcov: move gcov structs definitions to a gcc version specific file
  kernel/taskstats.c: return -ENOMEM when alloc memory fails in add_del_listener()
  kernel/taskstats.c: add nla_nest_cancel() for failure processing between nla_nest_start() and nla_nest_end()
  kernel/sysctl_binary.c: use scnprintf() instead of snprintf()
  ...

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Wed, 13 Nov 2013 06:34:18 +0000 (15:34 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull vfs updates from Al Viro:
 "All kinds of stuff this time around; some more notable parts:

   - RCU'd vfsmounts handling
   - new primitives for coredump handling
   - files_lock is gone
   - Bruce's delegations handling series
   - exportfs fixes

  plus misc stuff all over the place"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (101 commits)
  ecryptfs: ->f_op is never NULL
  locks: break delegations on any attribute modification
  locks: break delegations on link
  locks: break delegations on rename
  locks: helper functions for delegation breaking
  locks: break delegations on unlink
  namei: minor vfs_unlink cleanup
  locks: implement delegations
  locks: introduce new FL_DELEG lock flag
  vfs: take i_mutex on renamed file
  vfs: rename I_MUTEX_QUOTA now that it's not used for quotas
  vfs: don't use PARENT/CHILD lock classes for non-directories
  vfs: pull ext4's double-i_mutex-locking into common code
  exportfs: fix quadratic behavior in filehandle lookup
  exportfs: better variable name
  exportfs: move most of reconnect_path to helper function
  exportfs: eliminate unused "noprogress" counter
  exportfs: stop retrying once we race with rename/remove
  exportfs: clear DISCONNECTED on all parents sooner
  exportfs: more detailed comment for path_reconnect
  ...

11 years agoMerge tag 'dlm-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm
Linus Torvalds [Wed, 13 Nov 2013 06:31:13 +0000 (15:31 +0900)]
Merge tag 'dlm-3.13' of git://git./linux/kernel/git/teigland/linux-dlm

Pull dlm fix from David Teigland:
 "This set includes a single fix to resolve to a race that could cause
  lockspace shutdown to incorrectly return -EBUSY"

* tag 'dlm-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm:
  dlm: Avoid that dlm_release_lockspace() incorrectly returns -EBUSY

11 years agoMerge tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubi
Linus Torvalds [Wed, 13 Nov 2013 06:29:38 +0000 (15:29 +0900)]
Merge tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubi

Pull UBI changes from Artem Bityutskiy:
 "A bunch of fixes for the fastmap feature, which is still new and
  rather experimental.  It looks like it starts getting more users.

  No significant changes for the "classical" non-fastmap UBI"

* tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubi:
  UBI: Add some asserts to ubi_attach_fastmap()
  UBI: Fix memory leak in ubi_attach_fastmap() error path
  UBI: simplify image sequence test
  UBI: fastmap: fix backward compatibility with image_seq
  UBI: Call scan_all() with correct offset in error case
  UBI: Fix error path in scan_pool()
  UBI: fix refill_wl_user_pool()

11 years agoMerge tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubifs
Linus Torvalds [Wed, 13 Nov 2013 06:28:45 +0000 (15:28 +0900)]
Merge tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubifs

Pull ubifs changes from Artem Bityutskiy:
 "Mostly fixes for the power cut emulation UBIFS mode, and only one
  functional change which fixes a return error code"

* tag 'upstream-3.13-rc1' of git://git.infradead.org/linux-ubifs:
  UBIFS: correct data corruption range
  UBIFS: fix return code
  UBIFS: remove unnecessary code in ubifs_garbage_collect

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi...
Linus Torvalds [Wed, 13 Nov 2013 06:27:00 +0000 (15:27 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mszeredi/fuse

Pull fuse updates from Miklos Szeredi:
 "This adds a ->writepage() implementation to fuse, improving mmaped
  writeout and paving the way for buffered writeback.

  And there's a patch to add a fix minor number for /dev/cuse, similarly
  to /dev/fuse"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: writepages: protect secondary requests from fuse file release
  fuse: writepages: update bdi writeout when deleting secondary request
  fuse: writepages: crop secondary requests
  fuse: writepages: roll back changes if request not found
  cuse: add fix minor number to /dev/cuse
  fuse: writepage: skip already in flight
  fuse: writepages: handle same page rewrites
  fuse: writepages: fix aggregation
  fuse: fix race in fuse_writepages()
  fuse: Implement writepages callback
  fuse: don't BUG on no write file
  fuse: lock page in mkwrite
  fuse: Prepare to handle multiple pages in writeback
  fuse: Getting file for writeback helper

11 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Linus Torvalds [Wed, 13 Nov 2013 06:25:47 +0000 (15:25 +0900)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull ext[23], udf and quota fixes from Jan Kara:
 "Assorted fixes in quota, ext2, ext3 & udf.

  Probably the most important is a fix of fs corruption issue in ext2
  XIP support (OTOH xip is rarely used)"

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  ext2: Fix fs corruption in ext2_get_xip_mem()
  quota: info leak in quota_getquota()
  jbd: Revert "jbd: remove dependency on __GFP_NOFAIL"
  udf: fix for pathetic mount times in case of invalid file system
  ext3: Count journal as bsddf overhead in ext3_statfs

11 years agoMerge tag 'for-f2fs-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk...
Linus Torvalds [Wed, 13 Nov 2013 06:24:40 +0000 (15:24 +0900)]
Merge tag 'for-f2fs-3.13' of git://git./linux/kernel/git/jaegeuk/f2fs

Pull f2fs updates from Jaegeuk Kim:
 "This patch-set includes the following major enhancement patches.
   - add a sysfs to control reclaiming free segments
   - enhance the f2fs global lock procedures
   - enhance the victim selection flow
   - wait for selected node blocks during fsync
   - add some tracepoints
   - add a config to remove abundant BUG_ONs

  The other bug fixes are as follows.
   - fix deadlock on acl operations
   - fix some bugs with respect to orphan inodes

  And, there are a bunch of cleanups"

* tag 'for-f2fs-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (42 commits)
  f2fs: issue more large discard command
  f2fs: fix memory leak after kobject init failed in fill_super
  f2fs: cleanup waiting routine for writeback pages in cp
  f2fs: avoid to use a NULL point in destroy_segment_manager
  f2fs: remove unnecessary TestClearPageError when wait pages writeback
  f2fs: update f2fs document
  f2fs: avoid to wait all the node blocks during fsync
  f2fs: check all ones or zeros bitmap with bitops for better mount performance
  f2fs: change the method of calculating the number summary blocks
  f2fs: fix calculating incorrect free size when update xattr in __f2fs_setxattr
  f2fs: add an option to avoid unnecessary BUG_ONs
  f2fs: introduce CONFIG_F2FS_CHECK_FS for BUG_ON control
  f2fs: fix a deadlock during init_acl procedure
  f2fs: clean up acl flow for better readability
  f2fs: remove unnecessary segment bitmap updates
  f2fs: add tracepoint for vm_page_mkwrite
  f2fs: add tracepoint for set_page_dirty
  f2fs: remove redundant set_page_dirty from write_compacted_summaries
  f2fs: add reclaiming control by sysfs
  f2fs: introduce f2fs_balance_fs_bg for some background jobs
  ...

11 years agoMerge branch 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Linus Torvalds [Wed, 13 Nov 2013 06:21:53 +0000 (15:21 +0900)]
Merge branch 'for-3.13' of git://git./linux/kernel/git/tj/cgroup

Pull cgroup changes from Tejun Heo:
 "Not too much activity this time around.  css_id is finally killed and
  a minor update to device_cgroup"

* 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  device_cgroup: remove can_attach
  cgroup: kill css_id
  memcg: stop using css id
  memcg: fail to create cgroup if the cgroup id is too big
  memcg: convert to use cgroup id
  memcg: convert to use cgroup_is_descendant()

11 years agoMerge branch 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata
Linus Torvalds [Wed, 13 Nov 2013 06:18:22 +0000 (15:18 +0900)]
Merge branch 'for-3.13' of git://git./linux/kernel/git/tj/libata

Pull libata changes from Tejun Heo:
 "Nothing too interesting.  Only two minor fixes in libata core.  Most
  changes are specific to hardware which isn't too common"

* 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata:
  ahci: Add Device IDs for Intel Wildcat Point-LP
  sata_rcar: Convert to clk_prepare/unprepare
  drivers/libata: Set max sector to 65535 for Slimtype DVD A DS8A9SH drive
  libata: Add some missing command descriptions
  sata_highbank: clear whole array in highbank_initialize_phys()
  ahci: disabled FBS prior to issuing software reset
  libata: Fix display of sata speed
  ahci: imx: setup power saving methods
  ata_piix: minor typo and a printk fix
  ahci: Changing two module params with static and __read_mostly

11 years agoMerge branch 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu
Linus Torvalds [Wed, 13 Nov 2013 06:17:16 +0000 (15:17 +0900)]
Merge branch 'for-3.13' of git://git./linux/kernel/git/tj/percpu

Pull percpu changes from Tejun Heo:
 "Two smallish changes for percpu.  Two patches to remove unused
  this_cpu_xor() and one to fix a bug in percpu init failure path so
  that it can reach the proper BUG() instead of oopsing earlier"

* 'for-3.13' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
  x86: remove this_cpu_xor() implementation
  percpu: remove this_cpu_xor() implementation
  percpu: fix bootmem error handling in pcpu_page_first_chunk()

11 years agoipc, msg: fix message length check for negative values
Mathias Krause [Tue, 12 Nov 2013 23:11:47 +0000 (15:11 -0800)]
ipc, msg: fix message length check for negative values

On 64 bit systems the test for negative message sizes is bogus as the
size, which may be positive when evaluated as a long, will get truncated
to an int when passed to load_msg().  So a long might very well contain a
positive value but when truncated to an int it would become negative.

That in combination with a small negative value of msg_ctlmax (which will
be promoted to an unsigned type for the comparison against msgsz, making
it a big positive value and therefore make it pass the check) will lead to
two problems: 1/ The kmalloc() call in alloc_msg() will allocate a too
small buffer as the addition of alen is effectively a subtraction.  2/ The
copy_from_user() call in load_msg() will first overflow the buffer with
userland data and then, when the userland access generates an access
violation, the fixup handler copy_user_handle_tail() will try to fill the
remainder with zeros -- roughly 4GB.  That almost instantly results in a
system crash or reset.

  ,-[ Reproducer (needs to be run as root) ]--
  | #include <sys/stat.h>
  | #include <sys/msg.h>
  | #include <unistd.h>
  | #include <fcntl.h>
  |
  | int main(void) {
  |     long msg = 1;
  |     int fd;
  |
  |     fd = open("/proc/sys/kernel/msgmax", O_WRONLY);
  |     write(fd, "-1", 2);
  |     close(fd);
  |
  |     msgsnd(0, &msg, 0xfffffff0, IPC_NOWAIT);
  |
  |     return 0;
  | }
  '---

Fix the issue by preventing msgsz from getting truncated by consistently
using size_t for the message length.  This way the size checks in
do_msgsnd() could still be passed with a negative value for msg_ctlmax but
we would fail on the buffer allocation in that case and error out.

Also change the type of m_ts from int to size_t to avoid similar nastiness
in other code paths -- it is used in similar constructs, i.e.  signed vs.
unsigned checks.  It should never become negative under normal
circumstances, though.

Setting msg_ctlmax to a negative value is an odd configuration and should
be prevented.  As that might break existing userland, it will be handled
in a separate commit so it could easily be reverted and reworked without
reintroducing the above described bug.

Hardening mechanisms for user copy operations would have catched that bug
early -- e.g.  checking slab object sizes on user copy operations as the
usercopy feature of the PaX patch does.  Or, for that matter, detect the
long vs.  int sign change due to truncation, as the size overflow plugin
of the very same patch does.

[akpm@linux-foundation.org: fix i386 min() warnings]
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Pax Team <pageexec@freemail.hu>
Cc: Davidlohr Bueso <davidlohr@hp.com>
Cc: Brad Spengler <spender@grsecurity.net>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: <stable@vger.kernel.org> [ v2.3.27+ -- yes, that old ;) ]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoipc/util.c: remove unnecessary work pending test
Xie XiuQi [Tue, 12 Nov 2013 23:11:46 +0000 (15:11 -0800)]
ipc/util.c: remove unnecessary work pending test

Remove unnecessary work pending test before calling schedule_work().  It
has been tested in queue_work_on() already.  No functional changed.

Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
Cc: Tejun Heo <tj@kernel.org>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodevpts: plug the memory leak in kill_sb
Ilija Hadzic [Tue, 12 Nov 2013 23:11:45 +0000 (15:11 -0800)]
devpts: plug the memory leak in kill_sb

When devpts is unmounted, there may be a no-longer-used IDR tree hanging
off the superblock we are about to kill.  This needs to be cleaned up
before destroying the SB.

The leak is usually not a big deal because unmounting devpts is typically
done when shutting down the whole machine.  However, shutting down an LXC
container instead of a physical machine exposes the problem (the garbage
is detectable with kmemleak).

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years ago./Makefile: export initial ramdisk compression config option
P J P [Tue, 12 Nov 2013 23:11:44 +0000 (15:11 -0800)]
./Makefile: export initial ramdisk compression config option

Make menuconfig allows one to choose compression format of an initial
ramdisk image.  But this choice does not result in duly compressed ramdisk
image.  Because - $ make install - does not pass on the selected
compression choice to the dracut(8) tool, which creates the initramfs
file.  dracut(8) generates the image with the default compression, ie.
gzip(1).

This patch exports the selected compression option to a sub-shell
environment, so that it could be used by dracut(8) tool to generate
appropriately compressed initramfs images.

There isn't a straightforward way to pass on options to dracut(8) via
positional parameters.  Because it is indirectly invoked at the end of a $
make install sequence.

 # make install
   -> arch/$arch/boot/Makefile
    -> arch/$arch/boot/install.sh
     -> /sbing/installkernel ...
      -> /sbin/new-kernel-pkg ...
       -> /sbin/dracut ...

Signed-off-by: P J P <ppandit@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoinit/Kconfig: add option to disable kernel compression
Christian Ruppert [Tue, 12 Nov 2013 23:11:43 +0000 (15:11 -0800)]
init/Kconfig: add option to disable kernel compression

Some ARC users say they can boot faster with without kernel compression.
This probably depends on things like the FLASH chip they use etc.

Until now, kernel compression can only be disabled by removing "select
HAVE_<compression>" lines from the architecture Kconfig.  So add the
Kconfig logic to permit disabling of kernel compression.

Signed-off-by: Christian Ruppert <christian.ruppert@abilis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers: w1: make w1_slave::flags long to avoid memory corruption
Michal Nazarewicz [Tue, 12 Nov 2013 23:11:42 +0000 (15:11 -0800)]
drivers: w1: make w1_slave::flags long to avoid memory corruption

On architectures where long is more then 32 bits, modifying a 32-bit field
with set_bit (and other atomic bit operations) may cause bytes following
the field to by modified.

Because the endianness of the bits within a field is the native endianness
of the CPU[1], on big-endian machines, bit number zero is in the last byte
of the field.

Therefore, `set_bit(0, ptr)' on a 64-bit big-endian machine is roughly
equivalent to `((char *)ptr)[7] |= 1', and since w1 driver uses a 32-bit
field for holding the flags, this causes bytes beyond the field to be
modified.

[1] From Documentation/atomic_ops.txt:

    Native atomic bit operations are defined to operate on objects
    aligned to the size of an "unsigned long" C data type, and are
    least of that size.  The endianness of the bits within each
    "unsigned long" are the native endianness of the cpu.

Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/w1/masters/ds1wm.cuse dev_get_platdata()
Jingoo Han [Tue, 12 Nov 2013 23:11:41 +0000 (15:11 -0800)]
drivers/w1/masters/ds1wm.cuse dev_get_platdata()

Use the wrapper function for retrieving the platform data instead of
accessing dev->platform_data directly.  This is a cosmetic change to make
the code simpler and enhance the readability.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Evgeniy Polyakov <zbr@ioremap.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/memstick/core/ms_block.c: fix unreachable state in h_msb_read_page()
Roger Tseng [Tue, 12 Nov 2013 23:11:40 +0000 (15:11 -0800)]
drivers/memstick/core/ms_block.c: fix unreachable state in h_msb_read_page()

In h_msb_read_page() in ms_block.c, flow never reaches case
MSB_RP_RECIVE_STATUS_REG.  This causes error when MEMSTICK_INT_ERR is
encountered and status error bits are going to be examined, but the status
will never be copied back.

Fix it by transitioning to MSB_RP_RECIVE_STATUS_REG right after
MSB_RP_SEND_READ_STATUS_REG.

Signed-off-by: Roger Tseng <rogerable@realtek.com>
Acked-by: Maxim Levitsky <maximlevitsky@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/memstick/core/mspro_block.c: fix attributes array allocation
Michal Nazarewicz [Tue, 12 Nov 2013 23:11:39 +0000 (15:11 -0800)]
drivers/memstick/core/mspro_block.c: fix attributes array allocation

attrs field of attribute_group structure is a pointer to a pointer (as in
an array of pointers) rather than pointer to attribute struct (as in an
array of structures), so when allocating size of the pointer sholud be
used instead of the structure it is pointing to.

While at it, also change the call to use kcalloc rather than kzalloc.

Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodrivers/pps/clients/pps-gpio.c: remove redundant of_match_ptr
Sachin Kamat [Tue, 12 Nov 2013 23:11:33 +0000 (15:11 -0800)]
drivers/pps/clients/pps-gpio.c: remove redundant of_match_ptr

The data structure of_match_ptr() protects is always compiled in.  Hence
of_match_ptr() is not needed.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Cc: Rodolfo Giometti <giometti@enneenne.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/panic.c: reduce 1 byte usage for print tainted buffer
Chen Gang [Tue, 12 Nov 2013 23:11:32 +0000 (15:11 -0800)]
kernel/panic.c: reduce 1 byte usage for print tainted buffer

sizeof("Tainted: ") already counts '\0', and after first sprintf(), 's'
will start from the current string end (its' value is '\0').

So need not add additional 1 byte for maximized usage of 'buf' in
print_tainted().

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agogcov: reuse kbasename helper
Andy Shevchenko [Tue, 12 Nov 2013 23:11:31 +0000 (15:11 -0800)]
gcov: reuse kbasename helper

To get name of the file from a pathname let's use kbasename() helper.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Jingoo Han <jg1.han@samsung.com>
Cc: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/gcov/fs.c: use pr_warn()
Andrew Morton [Tue, 12 Nov 2013 23:11:29 +0000 (15:11 -0800)]
kernel/gcov/fs.c: use pr_warn()

pr_warning() is deprecated in favor of pr_warn()

Cc: Andy Gospodarek <agospoda@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Frantisek Hrbata <fhrbata@redhat.com>
Cc: Jan Stancek <jstancek@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/module.c: use pr_foo()
Andrew Morton [Tue, 12 Nov 2013 23:11:28 +0000 (15:11 -0800)]
kernel/module.c: use pr_foo()

kernel/module.c uses a mix of printk(KERN_foo and pr_foo().  Convert it
all to pr_foo and make the offered cleanups.

Not sure what to do about the printk(KERN_DEFAULT).  We don't have a
pr_default().

Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Joe Perches <joe@perches.com>
Cc: Frantisek Hrbata <fhrbata@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agogcov: compile specific gcov implementation based on gcc version
Frantisek Hrbata [Tue, 12 Nov 2013 23:11:27 +0000 (15:11 -0800)]
gcov: compile specific gcov implementation based on gcc version

Compile the correct gcov implementation file for the specific gcc version.

Signed-off-by: Frantisek Hrbata <fhrbata@redhat.com>
Cc: Jan Stancek <jstancek@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Acked-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Andy Gospodarek <agospoda@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agogcov: add support for gcc 4.7 gcov format
Frantisek Hrbata [Tue, 12 Nov 2013 23:11:26 +0000 (15:11 -0800)]
gcov: add support for gcc 4.7 gcov format

The gcov in-memory format changed in gcc 4.7.  The biggest change, which
requires this special implementation, is that gcov_info no longer contains
array of counters for each counter type for all functions and gcov_fn_info
is not used for mapping of function's counters to these arrays(offset).
Now each gcov_fn_info contans it's counters, which makes things a little
bit easier.

This is heavily based on the previous gcc_3_4.c implementation and patches
provided by Peter Oberparleiter.  Specially the buffer gcda implementation
for iterator.

[akpm@linux-foundation.org: use kmemdup() and kcalloc()]
[oberpar@linux.vnet.ibm.com: gcc_4_7.c needs vmalloc.h]
Signed-off-by: Frantisek Hrbata <fhrbata@redhat.com>
Cc: Jan Stancek <jstancek@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Reviewed-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Andy Gospodarek <agospoda@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agogcov: move gcov structs definitions to a gcc version specific file
Frantisek Hrbata [Tue, 12 Nov 2013 23:11:24 +0000 (15:11 -0800)]
gcov: move gcov structs definitions to a gcc version specific file

Since also the gcov structures(gcov_info, gcov_fn_info, gcov_ctr_info) can
change between gcc releases, as shown in gcc 4.7, they cannot be defined
in a common header and need to be moved to a specific gcc implemention
file.  This also requires to make the gcov_info structure opaque for the
common code and to introduce simple helpers for accessing data inside
gcov_info.

Signed-off-by: Frantisek Hrbata <fhrbata@redhat.com>
Cc: Jan Stancek <jstancek@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Acked-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Andy Gospodarek <agospoda@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/taskstats.c: return -ENOMEM when alloc memory fails in add_del_listener()
Chen Gang [Tue, 12 Nov 2013 23:11:23 +0000 (15:11 -0800)]
kernel/taskstats.c: return -ENOMEM when alloc memory fails in add_del_listener()

For registering in add_del_listener(), when kmalloc_node() fails, need
return -ENOMEM instead of success code, and cmd_attr_register_cpumask()
wants to know about it.

After modification, give a simple common test "build -> boot up ->
kernel/controllers/cgroup/getdelays by LTP tools".

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/taskstats.c: add nla_nest_cancel() for failure processing between nla_nest_sta...
Chen Gang [Tue, 12 Nov 2013 23:11:22 +0000 (15:11 -0800)]
kernel/taskstats.c: add nla_nest_cancel() for failure processing between nla_nest_start() and nla_nest_end()

When failure occurs between nla_nest_start() and nla_nest_end(), we should
call nla_nest_cancel() to clean up related things.

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/sysctl_binary.c: use scnprintf() instead of snprintf()
Chen Gang [Tue, 12 Nov 2013 23:11:22 +0000 (15:11 -0800)]
kernel/sysctl_binary.c: use scnprintf() instead of snprintf()

snprintf() will return the 'ideal' length which may be larger than real
buffer length, if we only want to use real length, need use scnprintf()
instead of.

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/sysctl.c: check return value after call proc_put_char() in __do_proc_doulongve...
Chen Gang [Tue, 12 Nov 2013 23:11:21 +0000 (15:11 -0800)]
kernel/sysctl.c: check return value after call proc_put_char() in __do_proc_doulongvec_minmax()

Need to check the return value of proc_put_char(), as was done in
__do_proc_doulongvec_minmax().

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agorbtree: fix rbtree_postorder_for_each_entry_safe() iterator
Jan Kara [Tue, 12 Nov 2013 23:11:19 +0000 (15:11 -0800)]
rbtree: fix rbtree_postorder_for_each_entry_safe() iterator

The iterator rbtree_postorder_for_each_entry_safe() relies on pointer
underflow behavior when testing for loop termination.  In particular it
expects that

  &rb_entry(NULL, type, field)->field

is NULL.  But the result of this expression is not defined by a C standard
and some gcc versions (e.g.  4.3.4) assume the above expression can never
be equal to NULL.  The net result is an oops because the iteration is not
properly terminated.

Fix the problem by modifying the iterator to avoid pointer underflows.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Cody P Schafer <cody@linux.vnet.ibm.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Artem Bityutskiy <dedekind1@gmail.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: <stable@vger.kernel.org> [3.12.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoexec/ptrace: fix get_dumpable() incorrect tests
Kees Cook [Tue, 12 Nov 2013 23:11:17 +0000 (15:11 -0800)]
exec/ptrace: fix get_dumpable() incorrect tests

The get_dumpable() return value is not boolean.  Most users of the
function actually want to be testing for non-SUID_DUMP_USER(1) rather than
SUID_DUMP_DISABLE(0).  The SUID_DUMP_ROOT(2) is also considered a
protected state.  Almost all places did this correctly, excepting the two
places fixed in this patch.

Wrong logic:
    if (dumpable == SUID_DUMP_DISABLE) { /* be protective */ }
        or
    if (dumpable == 0) { /* be protective */ }
        or
    if (!dumpable) { /* be protective */ }

Correct logic:
    if (dumpable != SUID_DUMP_USER) { /* be protective */ }
        or
    if (dumpable != 1) { /* be protective */ }

Without this patch, if the system had set the sysctl fs/suid_dumpable=2, a
user was able to ptrace attach to processes that had dropped privileges to
that user.  (This may have been partially mitigated if Yama was enabled.)

The macros have been moved into the file that declares get/set_dumpable(),
which means things like the ia64 code can see them too.

CVE-2013-2929

Reported-by: Vasily Kulikov <segoon@openwall.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokcore: add Kconfig help text
Randy Dunlap [Tue, 12 Nov 2013 23:11:16 +0000 (15:11 -0800)]
kcore: add Kconfig help text

Under Pseudo filesystems, /proc/kcore support has no help.

Fixes a portion of kernel bugzilla #52671:
  https://bugzilla.kernel.org/show_bug.cgi?id=52671

Thanks for David Howells for the help text.

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: <lailavrazda1979@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoprocfs: clean up proc_reg_get_unmapped_area for 80-column limit
HATAYAMA Daisuke [Tue, 12 Nov 2013 23:11:15 +0000 (15:11 -0800)]
procfs: clean up proc_reg_get_unmapped_area for 80-column limit

Clean up proc_reg_get_unmapped_area due to its 80-column limit
violation.

Signed-off-by: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Tested-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoDocumentation/ABI: document the non-ABI status of Kconfig and symbols
Josh Triplett [Tue, 12 Nov 2013 23:11:14 +0000 (15:11 -0800)]
Documentation/ABI: document the non-ABI status of Kconfig and symbols

Discussion at Kernel Summit made it clear that the presence or absence of
specific Kconfig symbols are not considered ABI, and that no userspace (or
bootloader, etc) should rely on them.

In addition, kernel-internal symbols are well established as non-ABI, per
Documentation/stable_api_nonsense.txt.

Document both of these in Documentation/ABI/README, in a new section for
notable bits of non-ABI.

Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Cc: Rob Landley <rob@landley.net>
Cc: Tao Ma <boyu.mt@taobao.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Cc: Richard Weinberger <richard.weinberger@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel-doc: improve "no structured comments found" error
Johannes Berg [Tue, 12 Nov 2013 23:11:12 +0000 (15:11 -0800)]
kernel-doc: improve "no structured comments found" error

When using '!Ffile function' in a docbook template, and the function no
longer exists, you get a "no structured comments found" error from the
kernel-doc processing script.  It's useful to know which functions it was
looking for, so print them out in this case.  Also do the same for '!Pfile
doc-section'

The same error also happens when using '!Efile' when some exported
functions aren't documented (in the same file.) There's a very large
number of such functions though, so don't print the message in this case
-- right now it would give ~850 messages.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Cc: Rob Landley <rob@landley.net>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoDocumentation/trace/tracepoints.txt: add links to TRACE_EVENT documentation
Stefan Raspl [Tue, 12 Nov 2013 23:11:11 +0000 (15:11 -0800)]
Documentation/trace/tracepoints.txt: add links to TRACE_EVENT documentation

Existing tracepoint documentation doesn't mention the popular
TRACE_EVENT macro.  Since an excellent series of articles on proper
usage already exists, respective links are added to the existing
documentation.

Signed-off-by: Stefan Raspl <raspl@linux.vnet.ibm.com>
Cc: Rob Landley <rob@landley.net>
Cc: Jiri Kosina <jkosina@suse.cz>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Zoltan Kiss <zoltan.kiss@citrix.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoDocumentation/filesystems/vfat.txt: fix directory entry example
Luis Ortega Perez de Villar [Tue, 12 Nov 2013 23:11:10 +0000 (15:11 -0800)]
Documentation/filesystems/vfat.txt: fix directory entry example

Slots can store up to 13 characters for the file name but one of the
examples has one character too many.

Signed-off-by: Luis Ortega Perez de Villar <luiorpe1@upv.es>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>