net: Start with correct mac_len in skb_network_protocol

[ Upstream commit 1e785f48d29a09b6cf96db7b49b6320dada332e1 ]

Sometimes, when the packet arrives at skb_mac_gso_segment()
its skb->mac_len already accounts for some of the mac lenght
headers in the packet.  This seems to happen when forwarding
through and OpenSSL tunnel.

When we start looking for any vlan headers in skb_network_protocol()
we seem to ignore any of the already known mac headers and start
with an ETH_HLEN.  This results in an incorrect offset, dropped
TSO frames and general slowness of the connection.

We can start counting from the known skb->mac_len
and return at least that much if all mac level headers
are known and accounted for.

Fixes: 53d6471cef17262d3ad1c7ce8982a234244f68ec (net: Account for all vlan headers in skb_mac_gso_segment)
CC: Eric Dumazet <eric.dumazet@gmail.com>
CC: Daniel Borkman <dborkman@redhat.com>
Tested-by: Martin Filip <nexus+kernel@smoula.net>
Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "net: sctp: Fix a_rwnd/rwnd management to reflect real state of the receiver's buffer"

[ Upstream commit 362d52040c71f6e8d8158be48c812d7729cb8df1 ]

This reverts commit ef2820a735f7 ("net: sctp: Fix a_rwnd/rwnd management
to reflect real state of the receiver's buffer") as it introduced a
serious performance regression on SCTP over IPv4 and IPv6, though a not
as dramatic on the latter. Measurements are on 10Gbit/s with ixgbe NICs.

Current state:

[root@Lab200slot2 ~]# iperf3 --sctp -4 -c 192.168.241.3 -V -l 1452 -t 60
iperf version 3.0.1 (10 January 2014)
Linux Lab200slot2 3.14.0 #1 SMP Thu Apr 3 23:18:29 EDT 2014 x86_64
Time: Fri, 11 Apr 2014 17:56:21 GMT
Connecting to host 192.168.241.3, port 5201
      Cookie: Lab200slot2.1397238981.812898.548918
[  4] local 192.168.241.2 port 38616 connected to 192.168.241.3 port 5201
Starting Test: protocol: SCTP, 1 streams, 1452 byte blocks, omitting 0 seconds, 60 second test
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.09   sec  20.8 MBytes   161 Mbits/sec
[  4]   1.09-2.13   sec  10.8 MBytes  86.8 Mbits/sec
[  4]   2.13-3.15   sec  3.57 MBytes  29.5 Mbits/sec
[  4]   3.15-4.16   sec  4.33 MBytes  35.7 Mbits/sec
[  4]   4.16-6.21   sec  10.4 MBytes  42.7 Mbits/sec
[  4]   6.21-6.21   sec  0.00 Bytes    0.00 bits/sec
[  4]   6.21-7.35   sec  34.6 MBytes   253 Mbits/sec
[  4]   7.35-11.45  sec  22.0 MBytes  45.0 Mbits/sec
[  4]  11.45-11.45  sec  0.00 Bytes    0.00 bits/sec
[  4]  11.45-11.45  sec  0.00 Bytes    0.00 bits/sec
[  4]  11.45-11.45  sec  0.00 Bytes    0.00 bits/sec
[  4]  11.45-12.51  sec  16.0 MBytes   126 Mbits/sec
[  4]  12.51-13.59  sec  20.3 MBytes   158 Mbits/sec
[  4]  13.59-14.65  sec  13.4 MBytes   107 Mbits/sec
[  4]  14.65-16.79  sec  33.3 MBytes   130 Mbits/sec
[  4]  16.79-16.79  sec  0.00 Bytes    0.00 bits/sec
[  4]  16.79-17.82  sec  5.94 MBytes  48.7 Mbits/sec
(etc)

[root@Lab200slot2 ~]#  iperf3 --sctp -6 -c 2001:db8:0:f101::1 -V -l 1400 -t 60
iperf version 3.0.1 (10 January 2014)
Linux Lab200slot2 3.14.0 #1 SMP Thu Apr 3 23:18:29 EDT 2014 x86_64
Time: Fri, 11 Apr 2014 19:08:41 GMT
Connecting to host 2001:db8:0:f101::1, port 5201
      Cookie: Lab200slot2.1397243321.714295.2b3f7c
[  4] local 2001:db8:0:f101::2 port 55804 connected to 2001:db8:0:f101::1 port 5201
Starting Test: protocol: SCTP, 1 streams, 1400 byte blocks, omitting 0 seconds, 60 second test
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.00   sec   169 MBytes  1.42 Gbits/sec
[  4]   1.00-2.00   sec   201 MBytes  1.69 Gbits/sec
[  4]   2.00-3.00   sec   188 MBytes  1.58 Gbits/sec
[  4]   3.00-4.00   sec   174 MBytes  1.46 Gbits/sec
[  4]   4.00-5.00   sec   165 MBytes  1.39 Gbits/sec
[  4]   5.00-6.00   sec   199 MBytes  1.67 Gbits/sec
[  4]   6.00-7.00   sec   163 MBytes  1.36 Gbits/sec
[  4]   7.00-8.00   sec   174 MBytes  1.46 Gbits/sec
[  4]   8.00-9.00   sec   193 MBytes  1.62 Gbits/sec
[  4]   9.00-10.00  sec   196 MBytes  1.65 Gbits/sec
[  4]  10.00-11.00  sec   157 MBytes  1.31 Gbits/sec
[  4]  11.00-12.00  sec   175 MBytes  1.47 Gbits/sec
[  4]  12.00-13.00  sec   192 MBytes  1.61 Gbits/sec
[  4]  13.00-14.00  sec   199 MBytes  1.67 Gbits/sec
(etc)

After patch:

[root@Lab200slot2 ~]#  iperf3 --sctp -4 -c 192.168.240.3 -V -l 1452 -t 60
iperf version 3.0.1 (10 January 2014)
Linux Lab200slot2 3.14.0+ #1 SMP Mon Apr 14 12:06:40 EDT 2014 x86_64
Time: Mon, 14 Apr 2014 16:40:48 GMT
Connecting to host 192.168.240.3, port 5201
      Cookie: Lab200slot2.1397493648.413274.65e131
[  4] local 192.168.240.2 port 50548 connected to 192.168.240.3 port 5201
Starting Test: protocol: SCTP, 1 streams, 1452 byte blocks, omitting 0 seconds, 60 second test
[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-1.00   sec   240 MBytes  2.02 Gbits/sec
[  4]   1.00-2.00   sec   239 MBytes  2.01 Gbits/sec
[  4]   2.00-3.00   sec   240 MBytes  2.01 Gbits/sec
[  4]   3.00-4.00   sec   239 MBytes  2.00 Gbits/sec
[  4]   4.00-5.00   sec   245 MBytes  2.05 Gbits/sec
[  4]   5.00-6.00   sec   240 MBytes  2.01 Gbits/sec
[  4]   6.00-7.00   sec   240 MBytes  2.02 Gbits/sec
[  4]   7.00-8.00   sec   239 MBytes  2.01 Gbits/sec

With the reverted patch applied, the SCTP/IPv4 performance is back
to normal on latest upstream for IPv4 and IPv6 and has same throughput
as 3.4.2 test kernel, steady and interval reports are smooth again.

Fixes: ef2820a735f7 ("net: sctp: Fix a_rwnd/rwnd management to reflect real state of the receiver's buffer")
Reported-by: Peter Butler <pbutler@sonusnet.com>
Reported-by: Dongsheng Song <dongsheng.song@gmail.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Tested-by: Peter Butler <pbutler@sonusnet.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Matija Glavinic Pecotic <matija.glavinic-pecotic.ext@nsn.com>
Cc: Alexander Sverdlin <alexander.sverdlin@nsn.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

filter: prevent nla extensions to peek beyond the end of the message

[ Upstream commit 05ab8f2647e4221cbdb3856dd7d32bd5407316b3 ]

The BPF_S_ANC_NLATTR and BPF_S_ANC_NLATTR_NEST extensions fail to check
for a minimal message length before testing the supplied offset to be
within the bounds of the message. This allows the subtraction of the nla
header to underflow and therefore -- as the data type is unsigned --
allowing far to big offset and length values for the search of the
netlink attribute.

The remainder calculation for the BPF_S_ANC_NLATTR_NEST extension is
also wrong. It has the minuend and subtrahend mixed up, therefore
calculates a huge length value, allowing to overrun the end of the
message while looking for the netlink attribute.

The following three BPF snippets will trigger the bugs when attached to
a UNIX datagram socket and parsing a message with length 1, 2 or 3.

 ,-[ PoC for missing size check in BPF_S_ANC_NLATTR ]--
 | ld #0x87654321
 | ldx #42
 | ld #nla
 | ret a
 `---

 ,-[ PoC for the same bug in BPF_S_ANC_NLATTR_NEST ]--
 | ld #0x87654321
 | ldx #42
 | ld #nlan
 | ret a
 `---

 ,-[ PoC for wrong remainder calculation in BPF_S_ANC_NLATTR_NEST ]--
 | ; (needs a fake netlink header at offset 0)
 | ld #0
 | ldx #42
 | ld #nlan
 | ret a
 `---

Fix the first issue by ensuring the message length fulfills the minimal
size constrains of a nla header. Fix the second bug by getting the math
for the remainder calculation right.

Fixes: 4738c1db15 ("[SKFILTER]: Add SKF_ADF_NLATTR instruction")
Fixes: d214c7537b ("filter: add SKF_AD_NLATTR_NEST to look for nested..")
Cc: Patrick McHardy <kaber@trash.net>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Acked-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv4: return valid RTA_IIF on ip route get

[ Upstream commit 91146153da2feab18efab2e13b0945b6bb704ded ]

Extend commit 13378cad02afc2adc6c0e07fca03903c7ada0b37
("ipv4: Change rt->rt_iif encoding.") from 3.6 to return valid
RTA_IIF on 'ip route get ... iif DEVICE' instead of rt_iif 0
which is displayed as 'iif *'.

inet_iif is not appropriate to use because skb_iif is not set.
Use the skb->dev->ifindex instead.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: ipv4: current group_info should be put after using.

[ Upstream commit b04c46190219a4f845e46a459e3102137b7f6cac ]

Plug a group_info refcount leak in ping_init.
group_info is only needed during initialization and
the code failed to release the reference on exit.
While here move grabbing the reference to a place
where it is actually needed.

Signed-off-by: Chuansheng Liu <chuansheng.liu@intel.com>
Signed-off-by: Zhang Dongxing <dongxing.zhang@intel.com>
Signed-off-by: xiaoming wang <xiaoming.wang@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

vti: don't allow to add the same tunnel twice

[ Upstream commit 8d89dcdf80d88007647945a753821a06eb6cc5a5 ]

Before the patch, it was possible to add two times the same tunnel:
ip l a vti1 type vti remote 10.16.0.121 local 10.16.0.249 key 41
ip l a vti2 type vti remote 10.16.0.121 local 10.16.0.249 key 41

It was possible, because ip_tunnel_newlink() calls ip_tunnel_find() with the
argument dev->type, which was set only later (when calling ndo_init handler
in register_netdevice()). Let's set this type in the setup handler, which is
called before newlink handler.

Introduced by commit b9959fd3b0fa ("vti: switch to new ip tunnel code").

CC: Cong Wang <amwang@redhat.com>
CC: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

gre: don't allow to add the same tunnel twice

[ Upstream commit 5a4552752d8f7f4cef1d98775ece7adb7616fde2 ]

Before the patch, it was possible to add two times the same tunnel:
ip l a gre1 type gre remote 10.16.0.121 local 10.16.0.249
ip l a gre2 type gre remote 10.16.0.121 local 10.16.0.249

It was possible, because ip_tunnel_newlink() calls ip_tunnel_find() with the
argument dev->type, which was set only later (when calling ndo_init handler
in register_netdevice()). Let's set this type in the setup handler, which is
called before newlink handler.

Introduced by commit c54419321455 ("GRE: Refactor GRE tunneling code.").

CC: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv6: Limit mtu to 65575 bytes

[ Upstream commit 30f78d8ebf7f514801e71b88a10c948275168518 ]

Francois reported that setting big mtu on loopback device could prevent
tcp sessions making progress.

We do not support (yet ?) IPv6 Jumbograms and cook corrupted packets.

We must limit the IPv6 MTU to (65535 + 40) bytes in theory.

Tested:

ifconfig lo mtu 70000
netperf -H ::1

Before patch : Throughput :   0.05 Mbits

After patch : Throughput : 35484 Mbits

Reported-by: Francois WELLENREITER <f.wellenreiter@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bridge: Fix double free and memory leak around br_allowed_ingress

[ Upstream commit eb7076182d1ae4bc4641534134ed707100d76acc ]

br_allowed_ingress() has two problems.

1. If br_allowed_ingress() is called by br_handle_frame_finish() and
vlan_untag() in br_allowed_ingress() fails, skb will be freed by both
vlan_untag() and br_handle_frame_finish().

2. If br_allowed_ingress() is called by br_dev_xmit() and
br_allowed_ingress() fails, the skb will not be freed.

Fix these two problems by freeing the skb in br_allowed_ingress()
if it fails.

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bonding: Remove debug_fs files when module init fails

[ Upstream commit db29868653394937037d71dc3545768302dda643 ]

Remove the bonding debug_fs entries when the
module initialization fails. The debug_fs
entries should be removed together with all other
already allocated resources.

Signed-off-by: Thomas Richter <tmricht@linux.vnet.ibm.com>
Signed-off-by: Jay Vosburgh <j.vosburgh@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: core: don't account for udp header size when computing seglen

[ Upstream commit 6d39d589bb76ee8a1c6cde6822006ae0053decff ]

In case of tcp, gso_size contains the tcpmss.

For UFO (udp fragmentation offloading) skbs, gso_size is the fragment
payload size, i.e. we must not account for udp header size.

Otherwise, when using virtio drivers, a to-be-forwarded UFO GSO packet
will be needlessly fragmented in the forward path, because we think its
individual segments are too large for the outgoing link.

Fixes: fe6cc55f3a9a053 ("net: ip, ipv6: handle gso skbs in forwarding path")
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Reported-by: Tobias Brunner <tobias@strongswan.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

l2tp: take PMTU from tunnel UDP socket

[ Upstream commit f34c4a35d87949fbb0e0f31eba3c054e9f8199ba ]

When l2tp driver tries to get PMTU for the tunnel destination, it uses
the pointer to struct sock that represents PPPoX socket, while it
should use the pointer that represents UDP socket of the tunnel.

Signed-off-by: Dmitry Petukhov <dmgenp@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sctp: test if association is dead in sctp_wake_up_waiters

[ Upstream commit 1e1cdf8ac78793e0875465e98a648df64694a8d0 ]

In function sctp_wake_up_waiters(), we need to involve a test
if the association is declared dead. If so, we don't have any
reference to a possible sibling association anymore and need
to invoke sctp_write_space() instead, and normally walk the
socket's associations and notify them of new wmem space. The
reason for special casing is that otherwise, we could run
into the following issue when a sctp_primitive_SEND() call
from sctp_sendmsg() fails, and tries to flush an association's
outq, i.e. in the following way:

sctp_association_free()
`-> list_del(&asoc->asocs)         <-- poisons list pointer
    asoc->base.dead = true
    sctp_outq_free(&asoc->outqueue)
    `-> __sctp_outq_teardown()
     `-> sctp_chunk_free()
      `-> consume_skb()
       `-> sctp_wfree()
        `-> sctp_wake_up_waiters() <-- dereferences poisoned pointers
                                       if asoc->ep->sndbuf_policy=0

Therefore, only walk the list in an 'optimized' way if we find
that the current association is still active. We could also use
list_del_init() in addition when we call sctp_association_free(),
but as Vlad suggests, we want to trap such bugs and thus leave
it poisoned as is.

Why is it safe to resolve the issue by testing for asoc->base.dead?
Parallel calls to sctp_sendmsg() are protected under socket lock,
that is lock_sock()/release_sock(). Only within that path under
lock held, we're setting skb/chunk owner via sctp_set_owner_w().
Eventually, chunks are freed directly by an association still
under that lock. So when traversing association list on destruction
time from sctp_wake_up_waiters() via sctp_wfree(), a different
CPU can't be running sctp_wfree() while another one calls
sctp_association_free() as both happens under the same lock.
Therefore, this can also not race with setting/testing against
asoc->base.dead as we are guaranteed for this to happen in order,
under lock. Further, Vlad says: the times we check asoc->base.dead
is when we've cached an association pointer for later processing.
In between cache and processing, the association may have been
freed and is simply still around due to reference counts. We check
asoc->base.dead under a lock, so it should always be safe to check
and not race against sctp_association_free(). Stress-testing seems
fine now, too.

Fixes: cd253f9f357d ("net: sctp: wake up all assocs if sndbuf policy is per socket")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Vlad Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sctp: wake up all assocs if sndbuf policy is per socket

[ Upstream commit 52c35befb69b005c3fc5afdaae3a5717ad013411 ]

SCTP charges chunks for wmem accounting via skb->truesize in
sctp_set_owner_w(), and sctp_wfree() respectively as the
reverse operation. If a sender runs out of wmem, it needs to
wait via sctp_wait_for_sndbuf(), and gets woken up by a call
to __sctp_write_space() mostly via sctp_wfree().

__sctp_write_space() is being called per association. Although
we assign sk->sk_write_space() to sctp_write_space(), which
is then being done per socket, it is only used if send space
is increased per socket option (SO_SNDBUF), as SOCK_USE_WRITE_QUEUE
is set and therefore not invoked in sock_wfree().

Commit 4c3a5bdae293 ("sctp: Don't charge for data in sndbuf
again when transmitting packet") fixed an issue where in case
sctp_packet_transmit() manages to queue up more than sndbuf
bytes, sctp_wait_for_sndbuf() will never be woken up again
unless it is interrupted by a signal. However, a still
remaining issue is that if net.sctp.sndbuf_policy=0, that is
accounting per socket, and one-to-many sockets are in use,
the reclaimed write space from sctp_wfree() is 'unfairly'
handed back on the server to the association that is the lucky
one to be woken up again via __sctp_write_space(), while
the remaining associations are never be woken up again
(unless by a signal).

The effect disappears with net.sctp.sndbuf_policy=1, that
is wmem accounting per association, as it guarantees a fair
share of wmem among associations.

Therefore, if we have reclaimed memory in case of per socket
accounting, wake all related associations to a socket in a
fair manner, that is, traverse the socket association list
starting from the current neighbour of the association and
issue a __sctp_write_space() to everyone until we end up
waking ourselves. This guarantees that no association is
preferred over another and even if more associations are
taken into the one-to-many session, all receivers will get
messages from the server and are not stalled forever on
high load. This setting still leaves the advantage of per
socket accounting in touch as an association can still use
up global limits if unused by others.

Fixes: 4eb701dfc618 ("[SCTP] Fix SCTP sendbuffer accouting.")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Thomas Graf <tgraf@suug.ch>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Vlad Yasevich <vyasevic@redhat.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915: Do not dereference pointers from ring buffer in evict event

commit 9297ebf29ad9118edd6c0fedc84f03e35028827d upstream.

The TP_printk() should never dereference any pointers, because the ring
buffer can be read at some unknown time in the future. If a device no
longer exists, it can cause a kernel oops. This also makes this
event useless when saving the ring buffer in userspaces tools such as
perf and trace-cmd.

The i915_gem_evict_vm dereferences the vm pointer which may also not
exist when the ring buffer is read sometime in the future.

Link: http://lkml.kernel.org/r/1395095198-20034-3-git-send-email-artagnon@gmail.com
Reported-by: Ramkumar Ramachandra <artagnon@gmail.com>
Fixes: bcccff847d1f "drm/i915: trace vm eviction instead of everything"
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
[danvet: Try to make it actually compile]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915/tv: fix gen4 composite s-video tv-out

commit e1f23f3dd817f53f622e486913ac662add46eeed upstream.

This is *not* bisected, but the likely regression is

commit c35614380d5c956bfda20eab2755b2f5a7d6f1e7
Author: Zhao Yakui <yakui.zhao@intel.com>
Date:   Tue Nov 24 09:48:48 2009 +0800

    drm/i915: Don't set up the TV port if it isn't in the BIOS table.

The commit does not check for all TV device types that might be present
in the VBT, disabling TV out for the missing ones. Add composite
S-video.

Reported-and-tested-by: Matthew Khouzam <matthew.khouzam@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73362
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/radeon: fix typo in spectre_golden_registers

commit f1553174a207f68a4ec19d436003097e0a4dc405 upstream.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/radeon: fix endian swap on hawaii clear state buffer setup

commit a8947f576728a66bd3aac629bd8ca021a010c808 upstream.

Need to swap on BE.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/radeon: call drm_edid_to_eld when we update the edid

commit 16086279353cbfecbb3ead474072dced17b97ddc upstream.

This needs to be done to update some of the fields in
the connector structure used by the audio code.

Noticed by several users on irc.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/radeon: clear needs_reset flag if IB test fails

commit 06a139f7a0885fa2c84962300edd181821ddc2c9 upstream.

If the IB test fails we don't want to reset the card over
and over again, just accept that it isn't working.

Bug: https://bugs.freedesktop.org/show_bug.cgi?id=76501

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/qxl: unset a pointer in sync_obj_unref

commit 41ccec352f3c823931a7d9d2a9c7880c14d7415a upstream.

This fixes a BUG_ON(bo->sync_obj != NULL); in ttm_bo_release_list.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/vmwgfx: Make sure user-space can't DMA across buffer object boundaries v2

commit cbd75e97a525e3819c02dc18bc2d67aa544c9e45 upstream.

We already check that the buffer object we're accessing is registered with
the file. Now also make sure that we can't DMA across buffer object boundaries.

v2: Code commenting update.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/vmwgfx: Fix query buffer locking order violation

commit c8e5e010ef12df6707a1d711a5279a22f67a355e upstream.

The query buffers were reserved while holding the binding mutex, which
caused a circular locking dependency.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/vmwgfx: correct fb_fix_screeninfo.line_length

commit aa6de142c901cd2d90ef08db30ae87da214bedcc upstream.

Previously, the vmwgfx_fb driver would allow users to call FBIOSET_VINFO, but it would not adjust
the FINFO properly, resulting in distorted screen rendering. The patch corrects that behaviour.

See https://bugs.gentoo.org/show_bug.cgi?id=494794 for examples.

Signed-off-by: Christopher Friedt <chrisfriedt@gmail.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wl18xx: align event mailbox with current fw

commit c0da71ff4d2cbf113465bff9a7c413154be25a89 upstream.

Some fields are missing from the event mailbox
struct definitions, which cause issues when
trying to handle some events.

Add the missing fields in order to align the
struct size (without adding actual support
for the new fields).

Reported-and-tested-by: Imre Kaloz <kaloz@openwrt.org>
Fixes: 028e724 ("wl18xx: move to new firmware (wl18xx-fw-3.bin)")
Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs: Don't return 0 from get_anon_bdev

commit a2a4dc494a7b7135f460e38e788c4a58f65e4ac3 upstream.

Commit 9e30cc9595303b27b48 removed an internal mount. This
has the side-effect that rootfs now has FSID 0. Many
userspace utilities assume that st_dev in struct stat
is never 0, so this change breaks a number of tools in
early userspace.

Since we don't know how many userspace programs are affected,
make sure that FSID is at least 1.

References: http://article.gmane.org/gmane.linux.kernel/1666905
References: http://permalink.gmane.org/gmane.linux.utilities.util-linux-ng/8557
Signed-off-by: Thomas Bächler <thomas@archlinux.org>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Tested-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mlx4_en: don't use napi_synchronize inside mlx4_en_netpoll

commit c98235cb8584a72e95786e17d695a8e5fafcd766 upstream.

The mlx4 driver is triggering schedules while atomic inside
mlx4_en_netpoll:

spin_lock_irqsave(&cq->lock, flags);
napi_synchronize(&cq->napi);
^^^^^ msleep here
mlx4_en_process_rx_cq(dev, cq, 0);
spin_unlock_irqrestore(&cq->lock, flags);

This was part of a patch by Alexander Guller from Mellanox in 2011,
but it still isn't upstream.

Signed-off-by: Chris Mason <clm@fb.com>
Acked-By: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

serial: omap: Fix missing pm_runtime_resume handling by simplifying code

commit d758c9c1b36b4d9a141c2146c70398d756167ed1 upstream.

The lack of pm_runtime_resume handling for the device state leads into
device wake-up interrupts not working after a while for runtime PM.

Also, serial-omap is confused about the use of device_may_wakeup.
The checks for device_may_wakeup should only be done for suspend and
resume, not for pm_runtime_suspend and pm_runtime_resume. The wake-up
events for PM runtime should always be enabled.

The lack of pm_runtime_resume handling leads into device wake-up
interrupts not working after a while for runtime PM.

Rather than try to patch over the issue of adding complex tests to
the pm_runtime_resume, let's fix the issues properly:

1. Make serial_omap_enable_wakeup deal with all internal PM state
   handling so we don't need to test for up->wakeups_enabled elsewhere.

   Later on once omap3 boots in device tree only mode we can also
   remove the up->wakeups_enabled flag and rely on the wake-up
   interrupt enable/disable state alone.

2. Do the device_may_wakeup checks in suspend and resume only,
   for runtime PM the wake-up events need to be always enabled.

3. Finally just call serial_omap_enable_wakeup and make sure we
   call it also in pm_runtime_resume.

4. Note that we also have to use disable_irq_nosync as serial_omap_irq
   calls pm_runtime_get_sync.

Fixes: 2a0b965cfb6e (serial: omap: Add support for optional wake-up)
Signed-off-by: Tony Lindgren <tony@atomide.com>
Acked-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: option: add and update a number of CMOTech devices

commit 34f972d6156fe9eea2ab7bb418c71f9d1d5c8e7b upstream.

A number of older CMOTech modems are based on Qualcomm
chips.  The blacklisted interfaces are QMI/wwan.

Reported-by: Lars Melin <larsm17@gmail.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: option: add Alcatel L800MA

commit dd6b48ecec2ea7d15f28d5e5474388681899a5e1 upstream.

Device interface layout:
0: ff/ff/ff - serial
1: ff/00/00 - serial AT+PPP
2: ff/ff/ff - QMI/wwan
3: 08/06/50 - storage

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: option: add Olivetti Olicard 500

commit 533b3994610f316e5cd61b56d0c4daa15c830f89 upstream.

Device interface layout:
0: ff/ff/ff - serial
1: ff/ff/ff - serial AT+PPP
2: 08/06/50 - storage
3: ff/ff/ff - serial
4: ff/ff/ff - QMI/wwan

Reported-by: Julio Araujo <julio.araujo@wllctel.com.br>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: qcserial: add Sierra Wireless MC7305/MC7355

commit bce4f588f19d59fc07fadfeb0b2a3a06c942827a upstream.

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: qcserial: add Sierra Wireless MC73xx

commit 70a3615fc07c2330ed7c1e922f3c44f4a67c0762 upstream.

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: qcserial: add Sierra Wireless EM7355

commit a00986f81182a69dee4d2c48e8c19805bdf0f790 upstream.

Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: io_ti: fix firmware download on big-endian machines

commit 5509076d1b4485ce9fb07705fcbcd2695907ab5b upstream.

During firmware download the device expects memory addresses in
big-endian byte order. As the wIndex parameter which hold the address is
sent in little-endian byte order regardless of host byte order, we need
to use swab16 rather than cpu_to_be16.

Also make sure to handle the struct ti_i2c_desc size parameter which is
returned in little-endian byte order.

Reported-by: Ludovic Drolez <ldrolez@debian.org>
Tested-by: Ludovic Drolez <ldrolez@debian.org>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: serial: fix sysfs-attribute removal deadlock

commit 10164c2ad6d2c16809f6c09e278f946e47801b3a upstream.

Fix driver new_id sysfs-attribute removal deadlock by making sure to
not hold any locks that the attribute operations grab when removing the
attribute.

Specifically, usb_serial_deregister holds the table mutex when
deregistering the driver, which includes removing the new_id attribute.
This can lead to a deadlock as writing to new_id increments the
attribute's active count before trying to grab the same mutex in
usb_serial_probe.

The deadlock can easily be triggered by inserting a sleep in
usb_serial_deregister and writing the id of an unbound device to new_id
during module unload.

As the table mutex (in this case) is used to prevent subdriver unload
during probe, it should be sufficient to only hold the lock while
manipulating the usb-serial driver list during deregister. A racing
probe will then either fail to find a matching subdriver or fail to get
the corresponding module reference.

Since v3.15-rc1 this also triggers the following lockdep warning:

======================================================
[ INFO: possible circular locking dependency detected ]
3.15.0-rc2 #123 Tainted: G        W
-------------------------------------------------------
modprobe/190 is trying to acquire lock:
 (s_active#4){++++.+}, at: [<c0167aa0>] kernfs_remove_by_name_ns+0x4c/0x94

but task is already holding lock:
 (table_lock){+.+.+.}, at: [<bf004d84>] usb_serial_deregister+0x3c/0x78 [usbserial]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (table_lock){+.+.+.}:
       [<c0075f84>] __lock_acquire+0x1694/0x1ce4
       [<c0076de8>] lock_acquire+0xb4/0x154
       [<c03af3cc>] _raw_spin_lock+0x4c/0x5c
       [<c02bbc24>] usb_store_new_id+0x14c/0x1ac
       [<bf007eb4>] new_id_store+0x68/0x70 [usbserial]
       [<c025f568>] drv_attr_store+0x30/0x3c
       [<c01690e0>] sysfs_kf_write+0x5c/0x60
       [<c01682c0>] kernfs_fop_write+0xd4/0x194
       [<c010881c>] vfs_write+0xbc/0x198
       [<c0108e4c>] SyS_write+0x4c/0xa0
       [<c000f880>] ret_fast_syscall+0x0/0x48

-> #0 (s_active#4){++++.+}:
       [<c03a7a28>] print_circular_bug+0x68/0x2f8
       [<c0076218>] __lock_acquire+0x1928/0x1ce4
       [<c0076de8>] lock_acquire+0xb4/0x154
       [<c0166b70>] __kernfs_remove+0x254/0x310
       [<c0167aa0>] kernfs_remove_by_name_ns+0x4c/0x94
       [<c0169fb8>] remove_files.isra.1+0x48/0x84
       [<c016a2fc>] sysfs_remove_group+0x58/0xac
       [<c016a414>] sysfs_remove_groups+0x34/0x44
       [<c02623b8>] driver_remove_groups+0x1c/0x20
       [<c0260e9c>] bus_remove_driver+0x3c/0xe4
       [<c026235c>] driver_unregister+0x38/0x58
       [<bf007fb4>] usb_serial_bus_deregister+0x84/0x88 [usbserial]
       [<bf004db4>] usb_serial_deregister+0x6c/0x78 [usbserial]
       [<bf005330>] usb_serial_deregister_drivers+0x2c/0x4c [usbserial]
       [<bf016618>] usb_serial_module_exit+0x14/0x1c [sierra]
       [<c009d6cc>] SyS_delete_module+0x184/0x210
       [<c000f880>] ret_fast_syscall+0x0/0x48

other info that might help us debug this:

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(table_lock);
                               lock(s_active#4);
                               lock(table_lock);
  lock(s_active#4);

 *** DEADLOCK ***

1 lock held by modprobe/190:
 #0:  (table_lock){+.+.+.}, at: [<bf004d84>] usb_serial_deregister+0x3c/0x78 [usbserial]

stack backtrace:
CPU: 0 PID: 190 Comm: modprobe Tainted: G        W     3.15.0-rc2 #123
[<c0015e10>] (unwind_backtrace) from [<c0013728>] (show_stack+0x20/0x24)
[<c0013728>] (show_stack) from [<c03a9a54>] (dump_stack+0x24/0x28)
[<c03a9a54>] (dump_stack) from [<c03a7cac>] (print_circular_bug+0x2ec/0x2f8)
[<c03a7cac>] (print_circular_bug) from [<c0076218>] (__lock_acquire+0x1928/0x1ce4)
[<c0076218>] (__lock_acquire) from [<c0076de8>] (lock_acquire+0xb4/0x154)
[<c0076de8>] (lock_acquire) from [<c0166b70>] (__kernfs_remove+0x254/0x310)
[<c0166b70>] (__kernfs_remove) from [<c0167aa0>] (kernfs_remove_by_name_ns+0x4c/0x94)
[<c0167aa0>] (kernfs_remove_by_name_ns) from [<c0169fb8>] (remove_files.isra.1+0x48/0x84)
[<c0169fb8>] (remove_files.isra.1) from [<c016a2fc>] (sysfs_remove_group+0x58/0xac)
[<c016a2fc>] (sysfs_remove_group) from [<c016a414>] (sysfs_remove_groups+0x34/0x44)
[<c016a414>] (sysfs_remove_groups) from [<c02623b8>] (driver_remove_groups+0x1c/0x20)
[<c02623b8>] (driver_remove_groups) from [<c0260e9c>] (bus_remove_driver+0x3c/0xe4)
[<c0260e9c>] (bus_remove_driver) from [<c026235c>] (driver_unregister+0x38/0x58)
[<c026235c>] (driver_unregister) from [<bf007fb4>] (usb_serial_bus_deregister+0x84/0x88 [usbserial])
[<bf007fb4>] (usb_serial_bus_deregister [usbserial]) from [<bf004db4>] (usb_serial_deregister+0x6c/0x78 [usbserial])
[<bf004db4>] (usb_serial_deregister [usbserial]) from [<bf005330>] (usb_serial_deregister_drivers+0x2c/0x4c [usbserial])
[<bf005330>] (usb_serial_deregister_drivers [usbserial]) from [<bf016618>] (usb_serial_module_exit+0x14/0x1c [sierra])
[<bf016618>] (usb_serial_module_exit [sierra]) from [<c009d6cc>] (SyS_delete_module+0x184/0x210)
[<c009d6cc>] (SyS_delete_module) from [<c000f880>] (ret_fast_syscall+0x0/0x48)

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "USB: serial: add usbid for dell wwan card to sierra.c"

commit 2e01280d2801c72878cf3a7119eac30077b463d5 upstream.

This reverts commit 1ebca9dad5abe8b2ed4dbd186cd657fb47c1f321.

This device was erroneously added to the sierra driver even though it's
not a Sierra device and was already handled by the option driver.

Cc: Richard Farina <sidhayn@gmail.com>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: option driver, add support for Telit UE910v2

commit d6de486bc22255779bd54b0fceb4c240962bf146 upstream.

option driver, added VID/PID for Telit UE910v2 modem

Signed-off-by: Daniele Palmas <dnlplm@gmail.com>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: serial: ftdi_sio: add id for Brainboxes serial cards

commit efe26e16b1d93ac0085e69178cc18811629e8fc5 upstream.

Custom VID/PIDs for Brainboxes cards as reported in
https://bugzilla.redhat.com/show_bug.cgi?id=1071914

Signed-off-by: Michele Baldessari <michele@acksyn.org>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: usb_wwan: fix handling of missing bulk endpoints

commit bd73bd8831696f189a479a0712ae95208e513d7e upstream.

Fix regression introduced by commit 8e493ca1767d ("USB: usb_wwan: fix
bulk-urb allocation") by making sure to require both bulk-in and out
endpoints during port probe.

The original option driver (which usb_wwan is based on) was written
under the assumption that either endpoint could be missing, but
evidently this cannot have been tested properly. Specifically, it would
handle opening a device without bulk-in (but would blow up during resume
which was implemented later), but not a missing bulk-out in write()
(although it is handled in some places such as write_room()).

Fortunately (?), the driver also got the test for missing endpoints
wrong so the urbs were in fact always allocated, although they would be
initialised using the wrong endpoint address (0) and any submission of
such an urb would fail.

The commit mentioned above fixed the test for missing endpoints but
thereby exposed the other bugs which would now generate null-pointer
exceptions rather than failed urb submissions.

The regression was introduced in v3.7, but the offending commit was also
marked for stable.

Reported-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Tested-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: cp210x: Add 8281 (Nanotec Plug & Drive)

commit 72b3007951010ce1bbf950e23b19d9839fa905a5 upstream.

Signed-off-by: Tristan Bruns <tristan@tristanbruns.de>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: cdc-acm: Remove Motorola/Telit H24 serial interfaces from ACM driver

commit 895d240d1db0b2736d779200788e4c4aea28a0c6 upstream.

By specifying NO_UNION_NORMAL the ACM driver does only use the first two
USB interfaces (modem data & control). The AT Port, Diagnostic and NMEA
interfaces are left to the USB serial driver.

Signed-off-by: Michael Ulbricht <michael.ulbricht@systec-electronic.com>
Signed-off-by: Alexander Stein <alexander.stein@systec-electronic.com>
Signed-off-by: Oliver Neukum <oliver@neukum.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm: use paravirt friendly ops for NUMA hinting ptes

commit 29c7787075c92ca8af353acd5301481e6f37082f upstream.

David Vrabel identified a regression when using automatic NUMA balancing
under Xen whereby page table entries were getting corrupted due to the
use of native PTE operations.  Quoting him

Xen PV guest page tables require that their entries use machine
addresses if the preset bit (_PAGE_PRESENT) is set, and (for
successful migration) non-present PTEs must use pseudo-physical
addresses.  This is because on migration MFNs in present PTEs are
translated to PFNs (canonicalised) so they may be translated back
to the new MFN in the destination domain (uncanonicalised).

pte_mknonnuma(), pmd_mknonnuma(), pte_mknuma() and pmd_mknuma()
set and clear the _PAGE_PRESENT bit using pte_set_flags(),
pte_clear_flags(), etc.

In a Xen PV guest, these functions must translate MFNs to PFNs
when clearing _PAGE_PRESENT and translate PFNs to MFNs when setting
_PAGE_PRESENT.

His suggested fix converted p[te|md]_[set|clear]_flags to using
paravirt-friendly ops but this is overkill.  He suggested an alternative
of using p[te|md]_modify in the NUMA page table operations but this is
does more work than necessary and would require looking up a VMA for
protections.

This patch modifies the NUMA page table operations to use paravirt
friendly operations to set/clear the flags of interest.  Unfortunately
this will take a performance hit when updating the PTEs on
CONFIG_PARAVIRT but I do not see a way around it that does not break
Xen.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: David Vrabel <david.vrabel@citrix.com>
Tested-by: David Vrabel <david.vrabel@citrix.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Anvin <hpa@zytor.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Noonan <steven@uplinklabs.net>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/hugetlb.c: add cond_resched_lock() in return_unused_surplus_pages()

commit 7848a4bf51b34f41fcc9bd77e837126d99ae84e3 upstream.

soft lockup in freeing gigantic hugepage fixed in commit 55f67141a892 "mm:
hugetlb: fix softlockup when a large number of hugepages are freed." can
happen in return_unused_surplus_pages(), so let's fix it.

Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

thp: close race between split and zap huge pages

commit b5a8cad376eebbd8598642697e92a27983aee802 upstream.

Sasha Levin has reported two THP BUGs[1][2].  I believe both of them
have the same root cause.  Let's look to them one by one.

The first bug[1] is "kernel BUG at mm/huge_memory.c:1829!".  It's
BUG_ON(mapcount != page_mapcount(page)) in __split_huge_page().  From my
testing I see that page_mapcount() is higher than mapcount here.

I think it happens due to race between zap_huge_pmd() and
page_check_address_pmd().  page_check_address_pmd() misses PMD which is
under zap:

CPU0 CPU1
zap_huge_pmd()
  pmdp_get_and_clear()
__split_huge_page()
  anon_vma_interval_tree_foreach()
    __split_huge_page_splitting()
      page_check_address_pmd()
        mm_find_pmd()
  /*
   * We check if PMD present without taking ptl: no
   * serialization against zap_huge_pmd(). We miss this PMD,
   * it's not accounted to 'mapcount' in __split_huge_page().
   */
  pmd_present(pmd) == 0

  BUG_ON(mapcount != page_mapcount(page)) // CRASH!!!

  page_remove_rmap(page)
    atomic_add_negative(-1, &page->_mapcount)

The second bug[2] is "kernel BUG at mm/huge_memory.c:1371!".
It's VM_BUG_ON_PAGE(!PageHead(page), page) in zap_huge_pmd().

This happens in similar way:

CPU0 CPU1
zap_huge_pmd()
  pmdp_get_and_clear()
  page_remove_rmap(page)
    atomic_add_negative(-1, &page->_mapcount)
__split_huge_page()
  anon_vma_interval_tree_foreach()
    __split_huge_page_splitting()
      page_check_address_pmd()
        mm_find_pmd()
  pmd_present(pmd) == 0 /* The same comment as above */
  /*
   * No crash this time since we already decremented page->_mapcount in
   * zap_huge_pmd().
   */
  BUG_ON(mapcount != page_mapcount(page))

  /*
   * We split the compound page here into small pages without
   * serialization against zap_huge_pmd()
   */
  __split_huge_page_refcount()
VM_BUG_ON_PAGE(!PageHead(page), page); // CRASH!!!

So my understanding the problem is pmd_present() check in mm_find_pmd()
without taking page table lock.

The bug was introduced by me commit with commit 117b0791ac42. Sorry for
that. :(

Let's open code mm_find_pmd() in page_check_address_pmd() and do the
check under page table lock.

Note that __page_check_address() does the same for PTE entires
if sync != 0.

I've stress tested split and zap code paths for 36+ hours by now and
don't see crashes with the patch applied. Before it took <20 min to
trigger the first bug and few hours for second one (if we ignore
first).

[1] https://lkml.kernel.org/g/<53440991.9090001@oracle.com>
[2] https://lkml.kernel.org/g/<5310C56C.60709@oracle.com>

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michel Lespinasse <walken@google.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

init/Kconfig: move the trusted keyring config option to general setup

commit 82c04ff89eba09d0e46e3f3649c6d3aa18e764a0 upstream.

The SYSTEM_TRUSTED_KEYRING config option is not in any menu, causing it
to show up in the toplevel of the kernel configuration.  Fix this by
moving it under the General Setup menu.

Signed-off-by: Peter Foley <pefoley2@pefoley.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tracing: Use rcu_dereference_sched() for trace event triggers

commit 561a4fe851ccab9dd0d14989ab566f9392d9f8b5 upstream.

As trace event triggers are now part of the mainline kernel, I added
my trace event trigger tests to my test suite I run on all my kernels.
Now these tests get run under different config options, and one of
those options is CONFIG_PROVE_RCU, which checks under lockdep that
the rcu locking primitives are being used correctly. This triggered
the following splat:

===============================
[ INFO: suspicious RCU usage. ]
3.15.0-rc2-test+ #11 Not tainted
-------------------------------
kernel/trace/trace_events_trigger.c:80 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 1, debug_locks = 0
4 locks held by swapper/1/0:
 #0:  ((&(&j_cdbs->work)->timer)){..-...}, at: [<ffffffff8104d2cc>] call_timer_fn+0x5/0x1be
 #1:  (&(&pool->lock)->rlock){-.-...}, at: [<ffffffff81059856>] __queue_work+0x140/0x283
 #2:  (&p->pi_lock){-.-.-.}, at: [<ffffffff8106e961>] try_to_wake_up+0x2e/0x1e8
 #3:  (&rq->lock){-.-.-.}, at: [<ffffffff8106ead3>] try_to_wake_up+0x1a0/0x1e8

stack backtrace:
CPU: 1 PID: 0 Comm: swapper/1 Not tainted 3.15.0-rc2-test+ #11
Hardware name:                  /DG965MQ, BIOS MQ96510J.86A.0372.2006.0605.1717 06/05/2006
 0000000000000001 ffff88007e083b98 ffffffff819f53a5 0000000000000006
 ffff88007b0942c0 ffff88007e083bc8 ffffffff81081307 ffff88007ad96d20
 0000000000000000 ffff88007af2d840 ffff88007b2e701c ffff88007e083c18
Call Trace:
 <IRQ>  [<ffffffff819f53a5>] dump_stack+0x4f/0x7c
 [<ffffffff81081307>] lockdep_rcu_suspicious+0x107/0x110
 [<ffffffff810ee51c>] event_triggers_call+0x99/0x108
 [<ffffffff810e8174>] ftrace_event_buffer_commit+0x42/0xa4
 [<ffffffff8106aadc>] ftrace_raw_event_sched_wakeup_template+0x71/0x7c
 [<ffffffff8106bcbf>] ttwu_do_wakeup+0x7f/0xff
 [<ffffffff8106bd9b>] ttwu_do_activate.constprop.126+0x5c/0x61
 [<ffffffff8106eadf>] try_to_wake_up+0x1ac/0x1e8
 [<ffffffff8106eb77>] wake_up_process+0x36/0x3b
 [<ffffffff810575cc>] wake_up_worker+0x24/0x26
 [<ffffffff810578bc>] insert_work+0x5c/0x65
 [<ffffffff81059982>] __queue_work+0x26c/0x283
 [<ffffffff81059999>] ? __queue_work+0x283/0x283
 [<ffffffff810599b7>] delayed_work_timer_fn+0x1e/0x20
 [<ffffffff8104d3a6>] call_timer_fn+0xdf/0x1be^M
 [<ffffffff8104d2cc>] ? call_timer_fn+0x5/0x1be
 [<ffffffff81059999>] ? __queue_work+0x283/0x283
 [<ffffffff8104d823>] run_timer_softirq+0x1a4/0x22f^M
 [<ffffffff8104696d>] __do_softirq+0x17b/0x31b^M
 [<ffffffff81046d03>] irq_exit+0x42/0x97
 [<ffffffff81a08db6>] smp_apic_timer_interrupt+0x37/0x44
 [<ffffffff81a07a2f>] apic_timer_interrupt+0x6f/0x80
 <EOI>  [<ffffffff8100a5d8>] ? default_idle+0x21/0x32
 [<ffffffff8100a5d6>] ? default_idle+0x1f/0x32
 [<ffffffff8100ac10>] arch_cpu_idle+0xf/0x11
 [<ffffffff8107b3a4>] cpu_startup_entry+0x1a3/0x213
 [<ffffffff8102a23c>] start_secondary+0x212/0x219

The cause is that the triggers are protected by rcu_read_lock_sched() but
the data is dereferenced with rcu_dereference() which expects it to
be protected with rcu_read_lock(). The proper reference should be
rcu_dereference_sched().

Cc: Tom Zanussi <tom.zanussi@linux.intel.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tracing/uprobes: Fix uprobe_cpu_buffer memory leak

commit 6ea6215fe394e320468589d9bba464a48f6d823a upstream.

Forgot to free uprobe_cpu_buffer percpu page in uprobe_buffer_disable().

Link: http://lkml.kernel.org/p/534F8B3F.1090407@huawei.com
Acked-by: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: zhangwei(Jovi) <jovi.zhangwei@huawei.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

md/raid1: r1buf_pool_alloc: free allocate pages when subsequent allocation fails.

commit da1aab3dca9aa88ae34ca392470b8943159e25fe upstream.

When performing a user-request check/repair (MD_RECOVERY_REQUEST is set)
on a raid1, we allocate multiple bios each with their own set of pages.

If the page allocations for one bio fails, we currently do *not* free
the pages allocated for the previous bios, nor do we free the bio itself.

This patch frees all the already-allocate pages, and makes sure that
all the bios are freed as well.

This bug can cause a memory leak which can ultimately OOM a machine.
It was introduced in 3.10-rc1.

Fixes: a07876064a0b73ab5ef1ebcf14b1cf0231c07858
Cc: Kent Overstreet <koverstreet@google.com>
Reported-by: Russell King - ARM Linux <linux@arm.linux.org.uk>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

HID: add NO_INIT_REPORTS quirk for Synaptics Touch Pad V 103S

commit 2f433083e854ec72c19dc9b0e1cebcc8e230fd75 upstream.

This touchpad seriously dislikes init reports, not only timeing out, but
also refusing to work after this.

Reported-and-tested-by: Vincent Fortier <th0ma7@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

HID: core: do not scan constant input report

commit e24d0d399b2fce71b627043e900ef28283850482 upstream.

The Microsoft Surface Type/Touch Cover 2 is a fancy device which advertised
itself as a multitouch device but with constant input reports.
This way, hid_scan_report() gives the group MULTITOUCH to it, but
hid-multitouch can not handle it due to the constant collection ignored
by hid-input.

To prevent such crap in the future, and while we do not fix this particular
device, make the scan_report coherent with hid-input.c, and ignore constant
input reports.

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "HID: microsoft: Add ID's for Surface Type/Touch Cover 2"

commit f3b0cbce01cd5c242b420d986b208d306bdc5083 upstream.

This reverts commit 117309c51dca42121f70cacec801511b76acf75c.

The MS Surface Pro 2 has an USB composite device with 3 interfaces
- interface 0 - sensor hub
- interface 1 - wacom digitizer
- interface 2 - the keyboard cover, if one is attached
This USB composite device changes it product id dependent on if and which
keyboard cover is attached. Adding the covers to hid_have_special_driver
prevents loading the right hid drivers for the other two interfaces, all 3
get loaded with hid-microsoft. We don't even need hid-microsoft for the
keyboards. We have to revert this to load the right hid modules for each
interface.

Signed-off-by: Derya <derya.kiran@yahoo.de>
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xen/events/fifo: correctly align bitops

commit 05a812ac474d0d6aef6d54b66bb08b81abde79c6 upstream.

FIFO event channels require bitops on 32-bit aligned values (the event
words).  Linux's bitops require unsigned long alignment which may be
64-bits.

On arm64 an incorrectly unaligned access will fault.

Fix this by aligning the bitops along with an adjustment for bit
position and using an unsigned long for the local copy of the ready
word.

Signed-off-by: Vladimir Murzin <murzin.v@gmail.com>
Tested-by: Pranavkumar Sawargaonkar <pranavkumar@linaro.org>
Reviewed-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xen/spinlock: Don't enable them unconditionally.

commit e0fc17a936334c08b2729fff87168c03fdecf5b6 upstream.

The git commit a945928ea2709bc0e8e8165d33aed855a0110279
('xen: Do not enable spinlocks before jump_label_init() has executed')
was added to deal with the jump machinery. Earlier the code
that turned on the jump label was only called by Xen specific
functions. But now that it had been moved to the initcall machinery
it gets called on Xen, KVM, and baremetal - ouch!. And the detection
machinery to only call it on Xen wasn't remembered in the heat
of merge window excitement.

This means that the slowpath is enabled on baremetal while it should
not be.

Reported-by: Waiman Long <waiman.long@hp.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
CC: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cifs: Wait for writebacks to complete before attempting write.

commit c11f1df5003d534fd067f0168bfad7befffb3b5c upstream.

Problem reported in Red Hat bz 1040329 for strict writes where we cache
only when we hold oplock and write direct to the server when we don't.

When we receive an oplock break, we first change the oplock value for
the inode in cifsInodeInfo->oplock to indicate that we no longer hold
the oplock before we enqueue a task to flush changes to the backing
device. Once we have completed flushing the changes, we return the
oplock to the server.

There are 2 ways here where we can have data corruption
1) While we flush changes to the backing device as part of the oplock
break, we can have processes write to the file. These writes check for
the oplock, find none and attempt to write directly to the server.
These direct writes made while we are flushing from cache could be
overwritten by data being flushed from the cache causing data
corruption.
2) While a thread runs in cifs_strict_writev, the machine could receive
and process an oplock break after the thread has checked the oplock and
found that it allows us to cache and before we have made changes to the
cache. In that case, we end up with a dirty page in cache when we
shouldn't have any. This will be flushed later and will overwrite all
subsequent writes to the part of the file represented by this page.

Before making any writes to the server, we need to confirm that we are
not in the process of flushing data to the server and if we are, we
should wait until the process is complete before we attempt the write.
We should also wait for existing writes to complete before we process
an oplock break request which changes oplock values.

We add a version specific  downgrade_oplock() operation to allow for
differences in the oplock values set for the different smb versions.

Signed-off-by: Sachin Prabhu <sprabhu@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Pavel Shilovsky <piastry@etersoft.ru>
Signed-off-by: Steve French <smfrench@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

don't bother with {get,put}_write_access() on non-regular files

commit dd20908a8a06b22c171f6c3fcdbdbd65bed07505 upstream.

it's pointless and actually leads to wrong behaviour in at least one
moderately convoluted case (pipe(), close one end, try to get to
another via /proc/*/fd and run into ETXTBUSY).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: remove _STK_LIM_MAX override

commit e0d8898d76a785453bfaf6cd08b830a7d5189f78 upstream.

There are only a couple of architectures that override _STK_LIM_MAX to
a non-infinity value. This changes the stack allocation semantics in
subtle ways. For example, GNU make changes its stack allocation to the
hard maximum defined by _STK_LIM_MAX. As a results, threads executed
by processes running under make are allocated a stack size of
_STK_LIM_MAX rather than a sensible default value. This causes various
thread stress tests to fail when they can't muster more than about 50
threads.

The attached change implements the default behavior used by the
majority of architectures.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Reviewed-by: Carlos O'Donell <carlos@systemhalted.org>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: fix epoll_pwait syscall on compat kernel

commit ab3e55b119c9653b19ea4edffb86f04db867ac98 upstream.

This bug was detected with the libio-epoll-perl debian package where the
test case IO-Ppoll-compat.t failed.

Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

parisc: change value of SHMLBA from 0x00400000 to PAGE_SIZE

commit 0ef36bd2b37815719e31a72d2beecc28ca8ecd26 upstream.

On parisc, SHMLBA was defined to 0x00400000 (4MB) to reflect that we need to
take care of our caches for shared mappings. But actually, we can map a file at
any multiple address of PAGE_SIZE, so let us correct that now with a value of
PAGE_SIZE for SHMLBA.  Instead we now take care of this cache colouring via the
constant SHM_COLOUR while we map shared pages.

Signed-off-by: Helge Deller <deller@gmx.de>
CC: Jeroen Roovers <jer@gentoo.org>
CC: John David Anglin <dave.anglin@bell.net>
CC: Carlos O'Donell <carlos@systemhalted.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tick-sched: Check tick_nohz_enabled in tick_nohz_switch_to_nohz()

commit 27630532ef5ead28b98cfe28d8f95222ef91c2b7 upstream.

Since commit d689fe222 (NOHZ: Check for nohz active instead of nohz
enabled) the tick_nohz_switch_to_nohz() function returns because it
checks for the tick_nohz_active flag. This can't be set, because the
function itself sets it.

Undo the change in tick_nohz_switch_to_nohz().

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: linaro-kernel@lists.linaro.org
Cc: fweisbec@gmail.com
Cc: Arvind.Chauhan@arm.com
Cc: linaro-networking@linaro.org
Link: http://lkml.kernel.org/r/40939c05f2d65d781b92b20302b02243d0654224.1397537987.git.viresh.kumar@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tick-sched: Don't call update_wall_time() when delta is lesser than tick_period

commit 03e6bdc5c4d0fc166bfd5d3cf749a5a0c1b5b1bd upstream.

In tick_do_update_jiffies64() we are processing ticks only if delta is
greater than tick_period. This is what we are supposed to do here and
it broke a bit with this patch:

commit 47a1b796 (tick/timekeeping: Call update_wall_time outside the
jiffies lock)

With above patch, we might end up calling update_wall_time() even if
delta is found to be smaller that tick_period. Fix this by returning
when the delta is less than tick period.

[ tglx: Made it a 3 liner and massaged changelog ]

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: linaro-kernel@lists.linaro.org
Cc: fweisbec@gmail.com
Cc: Arvind.Chauhan@arm.com
Cc: linaro-networking@linaro.org
Cc: John Stultz <john.stultz@linaro.org>
Link: http://lkml.kernel.org/r/80afb18a494b0bd9710975bcc4de134ae323c74f.1397537987.git.viresh.kumar@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tick-common: Fix wrong check in tick_check_replacement()

commit 521c42990e9d561ed5ed9f501f07639d0512b3c9 upstream.

tick_check_replacement() returns if a replacement of clock_event_device is
possible or not. It does this as the first check:

if (tick_check_percpu(curdev, newdev, smp_processor_id()))
return false;

Thats wrong. tick_check_percpu() returns true when the device is
useable. Check for false instead.

[ tglx: Massaged changelog ]

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: linaro-kernel@lists.linaro.org
Cc: fweisbec@gmail.com
Cc: Arvind.Chauhan@arm.com
Cc: linaro-networking@linaro.org
Link: http://lkml.kernel.org/r/486a02efe0246635aaba786e24b42d316438bf3b.1397537987.git.viresh.kumar@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/i915: restore QUIRK_NO_PCH_PWM_ENABLE

This reverts the bisected regressing

commit bc0bb9fd1c7810407ab810d204bbaecb255fddde
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Thu Nov 14 12:14:29 2013 +0200

    drm/i915: remove QUIRK_NO_PCH_PWM_ENABLE

restoring QUIRK_NO_PCH_PWM_ENABLE for a couple of Dell XPS models which
broke in 3.14.

There is no such revert upstream. We have root caused and fixed the
issue upstream, without the quirk, with:

commit 39fbc9c8f6765959b55e0b127dd5c57df5a47d67
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Apr 9 11:22:06 2014 +0300

    drm/i915: check VBT for supported backlight type

and

commit c675949ec58ca50d5a3ae3c757892f1560f6e896
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Apr 9 11:31:37 2014 +0300

    drm/i915: do not setup backlight if not available according to VBT

While the commits are within the stable rules otherwise, and fix more
machines than just the regressed Dell XPS models, we feel backporting
them to stable may be too risky. The revert is limited to the broken
machines, and the impact should be effectively the same as what the
upstream commits do more generally.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76276
Reported-by: Romain Francoise <romain@orebokech.com>
Acked-by: Daniel Vetter <daniel@ffwll.ch>
Tested-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Tested-by: Romain Francoise <romain@orebokech.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rbd: fix error paths in rbd_img_request_fill()

commit 42dd037c08c7cd6e3e9af7824b0c1d063f838885 upstream.

Doing rbd_obj_request_put() in rbd_img_request_fill() error paths is
not only insufficient, but also triggers an rbd_assert() in
rbd_obj_request_destroy():

    Assertion failure in rbd_obj_request_destroy() at line 1867:

    rbd_assert(obj_request->img_request == NULL);

rbd_img_obj_request_add() adds obj_requests to the img_request, the
opposite is rbd_img_obj_request_del().  Use it.

Fixes: http://tracker.ceph.com/issues/7327

Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tracepoint: Do not waste memory on mods with no tracepoints

commit 7dec935a3aa04412cba2cebe1524ae0d34a30c24 upstream.

No reason to allocate tp_module structures for modules that have no
tracepoints. This just wastes memory.

Fixes: b75ef8b44b1c "Tracepoint: Dissociate from module mutex"
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86,preempt: Fix preemption for i386

Many people reported preemption/reschedule problems with i386 kernels
for .13 and .14. After Michele bisected this to a combination of

  3e8e42c69bb ("sched: Revert need_resched() to look at TIF_NEED_RESCHED")
  ded79754754 ("irq: Force hardirq exit's softirq processing on its own stack")

it finally dawned on me that i386's current_thread_info() was to
blame.

When we are on interrupt/exception stacks, we fail to observe the
right TIF_NEED_RESCHED bit and therefore the PREEMPT_NEED_RESCHED
folding malfunctions.

Current upstream fixes this by making i386 behave the same as x86_64
already did:

  2432e1364bbe ("x86: Nuke the supervisor_stack field in i386 thread_info")
  b807902a88c4 ("x86: Nuke GET_THREAD_INFO_WITH_ESP() macro for i386")
  0788aa6a23cb ("x86: Prepare removal of previous_esp from i386 thread_info structure")
  198d208df437 ("x86: Keep thread_info on thread stack in x86_32")

However, that is far too much to stuff into -stable. Therefore I
propose we merge the below patch which uses task_thread_info(current)
for tif_need_resched() instead of the ESP based current_thread_info().

This makes sure we always observe the one true TIF_NEED_RESCHED bit
and things will work as expected again.

Cc: bp@alien8.de
Cc: fweisbec@gmail.com
Cc: david.a.cohen@linux.intel.com
Cc: mingo@kernel.org
Cc: fweisbec@gmail.com
Cc: greg@kroah.com
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: gregkh@linuxfoundation.org
Cc: pbonzini@redhat.com
Cc: rostedt@goodmis.org
Cc: stefan.bader@canonical.com
Cc: mingo@kernel.org
Cc: toralf.foerster@gmx.de
Cc: David Cohen <david.a.cohen@linux.intel.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: torvalds@linux-foundation.org
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: David Cohen <david.a.cohen@linux.intel.com>
Cc: <stable@vger.kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: <stable-commits@vger.kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: peterz@infradead.org
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: barra_cuda@katamail.com
Tested-by: Stefan Bader <stefan.bader@canonical.com>
Tested-by: Toralf F¿rster <toralf.foerster@gmx.de>
Tested-by: Michele Ballabio <barra_cuda@katamail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140409142447.GD13658@twins.programming.kicks-ass.net

netfilter: nf_tables: set names cannot be larger than 15 bytes

commit a9bdd8365684810e3de804f8c51e52c26a5eccbb upstream.

Currently, nf_tables trims off the set name if it exceeeds 15
bytes, so explicitly reject set names that are too large.

Reported-by: Giuseppe Longo <giuseppelng@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: Can't fail and free after table replacement

commit c58dd2dd443c26d856a168db108a0cd11c285bf3 upstream.

All xtables variants suffer from the defect that the copy_to_user()
to copy the counters to user memory may fail after the table has
already been exchanged and thus exposed. Return an error at this
point will result in freeing the already exposed table. Any
subsequent packet processing will result in a kernel panic.

We can't copy the counters before exposing the new tables as we
want provide the counter state after the old table has been
unhooked. Therefore convert this into a silent error.

Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

blktrace: fix accounting of partially completed requests

commit af5040da01ef980670b3741b3e10733ee3e33566 upstream.

trace_block_rq_complete does not take into account that request can
be partially completed, so we can get the following incorrect output
of blkparser:

  C   R 232 + 240 [0]
  C   R 240 + 232 [0]
  C   R 248 + 224 [0]
  C   R 256 + 216 [0]

but should be:

  C   R 232 + 8 [0]
  C   R 240 + 8 [0]
  C   R 248 + 8 [0]
  C   R 256 + 8 [0]

Also, the whole output summary statistics of completed requests and
final throughput will be incorrect.

This patch takes into account real completion size of the request and
fixes wrong completion accounting.

Signed-off-by: Roman Pen <r.peniaev@gmail.com>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: linux-kernel@vger.kernel.org
Signed-off-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: nf_conntrack: reserve two bytes for nf_ct_ext->len

commit 223b02d923ecd7c84cf9780bb3686f455d279279 upstream.

"len" contains sizeof(nf_ct_ext) and size of extensions. In a worst
case it can contain all extensions. Bellow you can find sizes for all
types of extensions. Their sum is definitely bigger than 256.

nf_ct_ext_types[0]->len = 24
nf_ct_ext_types[1]->len = 32
nf_ct_ext_types[2]->len = 24
nf_ct_ext_types[3]->len = 32
nf_ct_ext_types[4]->len = 152
nf_ct_ext_types[5]->len = 2
nf_ct_ext_types[6]->len = 16
nf_ct_ext_types[7]->len = 8

I have seen "len" up to 280 and my host has crashes w/o this patch.

The right way to fix this problem is reducing the size of the ecache
extension (4) and Florian is going to do this, but these changes will
be quite large to be appropriate for a stable tree.

Fixes: 5b423f6a40a0 (netfilter: nf_conntrack: fix racy timer handling with reliable)
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>
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: nf_tables: fix nft_cmp_fast failure on big endian for size < 4

commit b855d416dc17061ebb271ea7ef1201d100531770 upstream.

nft_cmp_fast is used for equality comparisions of size <= 4. For
comparisions of size < 4 byte a mask is calculated that is applied to
both the data from userspace (during initialization) and the register
value (during runtime). Both values are stored using (in effect) memcpy
to a memory area that is then interpreted as u32 by nft_cmp_fast.

This works fine on little endian since smaller types have the same base
address, however on big endian this is not true and the smaller types
are interpreted as a big number with trailing zero bytes.

The mask therefore must not include the lower bytes, but the higher bytes
on big endian. Add a helper function that does a cpu_to_le32 to switch
the bytes on big endian. Since we're dealing with a mask of just consequitive
bits, this works out fine.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: convert PPIDs to the inital PID namespace.

commit c92cdeb45eea38515e82187f48c2e4f435fb4e25 upstream.

sys_getppid() returns the parent pid of the current process in its own pid
namespace.  Since audit filters are based in the init pid namespace, a process
could avoid a filter or trigger an unintended one by being in an alternate pid
namespace or log meaningless information.

Switch to task_ppid_nr() for PPIDs to anchor all audit filters in the
init_pid_ns.

(informed by ebiederman's 6c621b7e)
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

pid: get pid_t ppid of task in init_pid_ns

commit ad36d28293936b03d6b7996e9d6aadfd73c0eb08 upstream.

Added the functions task_ppid_nr_ns() and task_ppid_nr() to abstract the lookup
of the PPID (real_parent's pid_t) of a process, including rcu locking, in the
arbitrary and init_pid_ns.
This provides an alternative to sys_getppid(), which is relative to the child
process' pid namespace.

(informed by ebiederman's 6c621b7e)
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tools lib traceevent: Fix memory leak in pretty_print()

commit de04f8657de9d3351a2d5880f1f7080b23b798cf upstream.

Commit 12e55569a244 "tools lib traceevent: Use helper trace-seq in print
functions like kernel does" added a extra trace_seq helper to process
string arguments like the kernel does it. But the difference between the
kernel and the userspace library is that the kernel's trace_seq structure
has a static allocated buffer. The userspace one has a dynamically
allocated one. It requires a trace_seq_destroy(), otherwise it produces
a nasty memory leak.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/20140422192330.6bb09bf8@gandalf.local.home
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: remove WARN_ON from get_kernel_ns()

commit b351c39cc9e0151cee9b8d52a1e714928faabb38 upstream.

Function and callers can be preempted.

https://bugzilla.kernel.org/show_bug.cgi?id=73721

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

SCSI: megaraid: missing bounds check in mimd_to_kioc()

commit 3de2260140417759c669d391613d583baf03b0cf upstream.

pthru32->dataxferlen comes from the user so we need to check that it's
not too large so we don't overflow the buffer.

Reported-by: Nico Golde <nico@ngolde.de>
Reported-by: Fabian Yamaguchi <fabs@goesec.de>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Sumit Saxena <sumit.saxena@lsi.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

SCSI: dual scan thread bug fix

commit f2495e228fce9f9cec84367547813cbb0d6db15a upstream.

In the highly unusual case where two threads are running concurrently through
the scanning code scanning the same target, we run into the situation where
one may allocate the target while the other is still using it.  In this case,
because the reap checks for STARGET_CREATED and kills the target without
reference counting, the second thread will do the wrong thing on reap.

Fix this by reference counting even creates and doing the STARGET_CREATED
check in the final put.

Tested-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: fix our current target reap infrastructure

commit e63ed0d7a98014fdfc2cfeb3f6dada313dcabb59 upstream.

This patch eliminates the reap_ref and replaces it with a proper kref.
On last put of this kref, the target is removed from visibility in
sysfs.  The final call to scsi_target_reap() for the device is done from
__scsi_remove_device() and only if the device was made visible.  This
ensures that the target disappears as soon as the last device is gone
rather than waiting until final release of the device (which is often
too long).

Reviewed-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 3.14.4

x86-64, build: Fix stack protector Makefile breakage with 32-bit userland

commit 14262d67fe348018af368a07430fbc06eadeabb1 upstream.

If you are using a 64-bit kernel with 32-bit userland, then
scripts/gcc-x86_64-has-stack-protector.sh invokes 32-bit gcc
with -mcmodel=kernel, which produces:

<stdin>:1:0: error: code model 'kernel' not supported in the 32 bit mode

and trips the "broken compiler" test at arch/x86/Makefile:120.

There are several places a fix is possible, but the following seems
cleanest.  (But it's minimal; it would also be possible to factor
out a bunch of stuff from the two branches of the if.)

Signed-off-by: George Spelvin <linux@horizon.com>
Link: http://lkml.kernel.org/r/20140507210552.7581.qmail@ns.horizon.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm: bochs: add power management support

commit b8ccd70f1363f7d4e49219dbc46ec973a14f49cd upstream.

bochs kms driver lacks power management support, thus
the vga display doesn't work any more after S3 resume.

Fix this by adding suspend and resume functions.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm: cirrus: add power management support

commit 2f1e800799bf478494cec3573cd63eb34ca89c9d upstream.

cirrus kms driver lacks power management support, thus
the vga display doesn't work any more after S3 resume.

Fix this by adding suspend and resume functions.
Also make the mode_set function unblank the screen.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq: unicore32: fix typo issue for 'clk'

commit b4ddad95020e65cfbbf9aee63d3bcdf682794ade upstream.

Need use 'clk' instead of 'mclk', which is the original removed local
variable.

The related original commit:

  "652ed95 cpufreq: introduce cpufreq_generic_get() routine"

The related error with allmodconfig for unicore32:

    CC      drivers/cpufreq/unicore2-cpufreq.o
  drivers/cpufreq/unicore2-cpufreq.c: In function ‘ucv2_target’:
  drivers/cpufreq/unicore2-cpufreq.c:48: error: ‘struct cpufreq_policy’ has no member named ‘mclk’
  make[2]: *** [drivers/cpufreq/unicore2-cpufreq.o] Error 1
  make[1]: *** [drivers/cpufreq] Error 2
  make: *** [drivers] Error 2

Fixes: 652ed95d5fa6 (cpufreq: introduce cpufreq_generic_get() routine)
Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ACPI / processor: Fix failure of loading acpi-cpufreq driver

commit 4985c32ee4241d1aba1beeac72294faa46aaff10 upstream.

According commit d640113fe (ACPI: processor: fix acpi_get_cpuid for UP
processor),  BIOS may not provide _MAT or MADT tables and acpi_get_apicid()
always returns -1. For these cases, original code will pass apic_id with
vaule of -1 to acpi_map_cpuid() and it will check the acpi_id. If acpi_id
is equal to zero, ignores apic_id and return zero for CPU0.

Commit b981513 (ACPI / scan: bail out early if failed to parse APIC
ID for CPU) changed the behavior. Return ENODEV when find apic_id is
less than zero after calling acpi_get_apicid(). This causes acpi-cpufreq
driver fails to be loaded on some machines. This patch is to fix it.

Fixes: b981513f806d (ACPI / scan: bail out early if failed to parse APIC ID for CPU)
References: https://bugzilla.kernel.org/show_bug.cgi?id=73781
Reported-and-tested-by: KATO Hiroshi <katoh@mikage.ne.jp>
Reported-and-tested-by: Stuart Foster <smf.linux@ntlworld.com>
Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq: at32ap: don't declare local variable as static

commit 0ca97886fece9e1acd71ade4ca3a250945c8fc8b upstream.

Earlier commit:
commit 652ed95d5fa6074b3c4ea245deb0691f1acb6656
Author: Viresh Kumar <viresh.kumar@linaro.org>
Date:   Thu Jan 9 20:38:43 2014 +0530

    cpufreq: introduce cpufreq_generic_get() routine

did some changes to driver and by mistake made cpuclk as a 'static' local
variable, which wasn't actually required. Fix it.

Fixes: 652ed95d5fa6 (cpufreq: introduce cpufreq_generic_get() routine)
Reported-by: Alexandre Oliva <lxoliva@fsfla.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq: loongson2_cpufreq: don't declare local variable as static

commit 05ed672292dc9d37db4fafdd49baa782158cd795 upstream.

Earlier commit:
commit 652ed95d5fa6074b3c4ea245deb0691f1acb6656
Author: Viresh Kumar <viresh.kumar@linaro.org>
Date:   Thu Jan 9 20:38:43 2014 +0530

    cpufreq: introduce cpufreq_generic_get() routine

did some changes to driver and by mistake made cpuclk as a 'static' local
variable, which wasn't actually required. Fix it.

Fixes: 652ed95d5fa6 (cpufreq: introduce cpufreq_generic_get() routine)
Reported-by: Alexandre Oliva <lxoliva@fsfla.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Input: synaptics - add min/max quirk for ThinkPad Edge E431

commit 27a38856a948c3e8de30dc71647ff9e1778c99fc upstream.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Input: synaptics - add min/max quirk for ThinkPad T431s, L440, L540, S1 Yoga and X1

commit 46a2986ebbe18757c2d8c352f8fb6e0f4f0754e3 upstream.

We expect that all the Haswell series will need such quirks, sigh.

The T431s seems to be T430 hardware in a T440s case, using the T440s touchpad,
with the same min/max issue.

The X1 Carbon 3rd generation name says 2nd while it is a 3rd generation.

The X1 and T431s share a PnPID with the T540p, but the reported ranges are
closer to those of the T440s.

HdG: Squashed 5 quirk patches into one. T431s + L440 + L540 are written by me,
S1 Yoga and X1 are written by Benjamin Tissoires.

Hdg: Standardized S1 Yoga and X1 values, Yoga uses the same touchpad as the
X240, X1 uses the same touchpad as the T440.

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

thinkpad_acpi: Fix inconsistent mute LED after resume

commit 119f449866ad18785b0445adaf0d2859c6dbdaa3 upstream.

The mute LED states have to be restored after resume.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=70351
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Matthew Garrett <matthew.garrett@nebula.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: sdhci-bcm-kona: fix build errors when built-in

commit 4025ce24f326830135341814307c072f6c2a7738 upstream.

`sdhci_bcm_kona_remove' referenced in section `.data' of drivers/built-in.o: defined in discarded section `.exit.text' of drivers/built-in.o

Fixes: 058feb53666f ("mmc: sdhci-bcm-kona: make linker-section warning go away")
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tested-by: Markus Mayer <markus.mayer@linaro.org>
Acked-by: Matt Porter <mporter@linaro.org>
Signed-off-by: Chris Ball <chris@printf.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lib/percpu_counter.c: fix bad percpu counter state during suspend

commit e39435ce68bb4685288f78b1a7e24311f7ef939f upstream.

I got a bug report yesterday from Laszlo Ersek in which he states that
his kvm instance fails to suspend.  Laszlo bisected it down to this
commit 1cf7e9c68fe8 ("virtio_blk: blk-mq support") where virtio-blk is
converted to use the blk-mq infrastructure.

After digging a bit, it became clear that the issue was with the queue
drain.  blk-mq tracks queue usage in a percpu counter, which is
incremented on request alloc and decremented when the request is freed.
The initial hunt was for an inconsistency in blk-mq, but everything
seemed fine.  In fact, the counter only returned crazy values when
suspend was in progress.

When a CPU is unplugged, the percpu counters merges that CPU state with
the general state.  blk-mq takes care to register a hotcpu notifier with
the appropriate priority, so we know it runs after the percpu counter
notifier.  However, the percpu counter notifier only merges the state
when the CPU is fully gone.  This leaves a state transition where the
CPU going away is no longer in the online mask, yet it still holds
private values.  This means that in this state, percpu_counter_sum()
returns invalid results, and the suspend then hangs waiting for
abs(dead-cpu-value) requests to complete which of course will never
happen.

Fix this by clearing the state earlier, so we never have a case where
the CPU isn't in online mask but still holds private state.  This bug
has been there since forever, I guess we don't have a lot of users where
percpu counters needs to be reliable during the suspend cycle.

Signed-off-by: Jens Axboe <axboe@fb.com>
Reported-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

lockd: ensure we tear down any live sockets when socket creation fails during lockd_up

commit 679b033df48422191c4cac52b610d9980e019f9b upstream.

We had a Fedora ABRT report with a stack trace like this:

kernel BUG at net/sunrpc/svc.c:550!
invalid opcode: 0000 [#1] SMP
[...]
CPU: 2 PID: 913 Comm: rpc.nfsd Not tainted 3.13.6-200.fc20.x86_64 #1
Hardware name: Hewlett-Packard HP ProBook 4740s/1846, BIOS 68IRR Ver. F.40 01/29/2013
task: ffff880146b00000 ti: ffff88003f9b8000 task.ti: ffff88003f9b8000
RIP: 0010:[<ffffffffa0305fa8>]  [<ffffffffa0305fa8>] svc_destroy+0x128/0x130 [sunrpc]
RSP: 0018:ffff88003f9b9de0  EFLAGS: 00010206
RAX: ffff88003f829628 RBX: ffff88003f829600 RCX: 00000000000041ee
RDX: 0000000000000000 RSI: 0000000000000286 RDI: 0000000000000286
RBP: ffff88003f9b9de8 R08: 0000000000017360 R09: ffff88014fa97360
R10: ffffffff8114ce57 R11: ffffea00051c9c00 R12: ffff88003f829600
R13: 00000000ffffff9e R14: ffffffff81cc7cc0 R15: 0000000000000000
FS:  00007f4fde284840(0000) GS:ffff88014fa80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f4fdf5192f8 CR3: 00000000a569a000 CR4: 00000000001407e0
Stack:
 ffff88003f792300 ffff88003f9b9e18 ffffffffa02de02a 0000000000000000
 ffffffff81cc7cc0 ffff88003f9cb000 0000000000000008 ffff88003f9b9e60
 ffffffffa033bb35 ffffffff8131c86c ffff88003f9cb000 ffff8800a5715008
Call Trace:
 [<ffffffffa02de02a>] lockd_up+0xaa/0x330 [lockd]
 [<ffffffffa033bb35>] nfsd_svc+0x1b5/0x2f0 [nfsd]
 [<ffffffff8131c86c>] ? simple_strtoull+0x2c/0x50
 [<ffffffffa033c630>] ? write_pool_threads+0x280/0x280 [nfsd]
 [<ffffffffa033c6bb>] write_threads+0x8b/0xf0 [nfsd]
 [<ffffffff8114efa4>] ? __get_free_pages+0x14/0x50
 [<ffffffff8114eff6>] ? get_zeroed_page+0x16/0x20
 [<ffffffff811dec51>] ? simple_transaction_get+0xb1/0xd0
 [<ffffffffa033c098>] nfsctl_transaction_write+0x48/0x80 [nfsd]
 [<ffffffff811b8b34>] vfs_write+0xb4/0x1f0
 [<ffffffff811c3f99>] ? putname+0x29/0x40
 [<ffffffff811b9569>] SyS_write+0x49/0xa0
 [<ffffffff810fc2a6>] ? __audit_syscall_exit+0x1f6/0x2a0
 [<ffffffff816962e9>] system_call_fastpath+0x16/0x1b
Code: 31 c0 e8 82 db 37 e1 e9 2a ff ff ff 48 8b 07 8b 57 14 48 c7 c7 d5 c6 31 a0 48 8b 70 20 31 c0 e8 65 db 37 e1 e9 f4 fe ff ff 0f 0b <0f> 0b 66 0f 1f 44 00 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55
RIP  [<ffffffffa0305fa8>] svc_destroy+0x128/0x130 [sunrpc]
 RSP <ffff88003f9b9de0>

Evidently, we created some lockd sockets and then failed to create
others. make_socks then returned an error and we tried to tear down the
svc, but svc->sv_permsocks was not empty so we ended up tripping over
the BUG() in svc_destroy().

Fix this by ensuring that we tear down any live sockets we created when
socket creation is going to return an error.

Fixes: 786185b5f8abefa (SUNRPC: move per-net operations from...)
Reported-by: Raphos <raphoszap@laposte.net>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

aio: v4 ensure access to ctx->ring_pages is correctly serialised for migration

commit fa8a53c39f3fdde98c9eace6a9b412143f0f6ed6 upstream.

As reported by Tang Chen, Gu Zheng and Yasuaki Isimatsu, the following issues
exist in the aio ring page migration support.

As a result, for example, we have the following problem:

            thread 1                      |              thread 2
                                          |
aio_migratepage()                         |
 |-> take ctx->completion_lock            |
 |-> migrate_page_copy(new, old)          |
 |   *NOW*, ctx->ring_pages[idx] == old   |
                                          |
                                          |    *NOW*, ctx->ring_pages[idx] == old
                                          |    aio_read_events_ring()
                                          |     |-> ring = kmap_atomic(ctx->ring_pages[0])
                                          |     |-> ring->head = head;          *HERE, write to the old ring page*
                                          |     |-> kunmap_atomic(ring);
                                          |
 |-> ctx->ring_pages[idx] = new           |
 |   *BUT NOW*, the content of            |
 |    ring_pages[idx] is old.             |
 |-> release ctx->completion_lock         |

As above, the new ring page will not be updated.

Fix this issue, as well as prevent races in aio_ring_setup() by holding
the ring_lock mutex during kioctx setup and page migration.  This avoids
the overhead of taking another spinlock in aio_read_events_ring() as Tang's
and Gu's original fix did, pushing the overhead into the migration code.

Note that to handle the nesting of ring_lock inside of mmap_sem, the
migratepage operation uses mutex_trylock().  Page migration is not a 100%
critical operation in this case, so the ocassional failure can be
tolerated.  This issue was reported by Sasha Levin.

Based on feedback from Linus, avoid the extra taking of ctx->completion_lock.
Instead, make page migration fully serialised by mapping->private_lock, and
have aio_free_ring() simply disconnect the kioctx from the mapping by calling
put_aio_ring_file() before touching ctx->ring_pages[].  This simplifies the
error handling logic in aio_migratepage(), and should improve robustness.

v4: always do mutex_unlock() in cases when kioctx setup fails.

Reported-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Gu Zheng <guz.fnst@cn.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dma: edma: fix incorrect SG list handling

commit 5fc68a6cad658e45dca3e0a6607df3a8e5df4ef9 upstream.

The code to handle any length SG lists calls edma_resume()
even before edma_start() is called. This is incorrect
because edma_resume() enables edma events on the channel
after which CPU (in edma_start) cannot clear posted
events by writing to ECR (per the EDMA user's guide).

Because of this EDMA transfers fail to start if due
to some reason there is a pending EDMA event registered
even before EDMA transfers are started. This can happen if
an EDMA event is a byproduct of device initialization.

Fix this by calling edma_resume() only if it is not the
first batch of MAX_NR_SG elements.

Without this patch, MMC/SD fails to function on DA850 EVM
with DMA. The behaviour is triggered by specific IP and
this can explain why the issue was not reported before
(example with MMC/SD on AM335x).

Tested on DA850 EVM and AM335x EVM-SK using MMC/SD card.

Cc: Joel Fernandes <joelf@ti.com>
Acked-by: Joel Fernandes <joelf@ti.com>
Tested-by: Jon Ringle <jringle@gridpoint.com>
Tested-by: Alexander Holler <holler@ahsoftware.de>
Reported-by: Jon Ringle <jringle@gridpoint.com>
Signed-off-by: Sekhar Nori <nsekhar@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm cache: fix a lock-inversion

commit 0596661f0a16d9d69bf1033320e70b6ff52b5e81 upstream.

When suspending a cache the policy is walked and the individual policy
hints written to the metadata via sync_metadata().  This led to this
lock order:

      policy->lock
        cache_metadata->root_lock

When loading the cache target the policy is populated while the metadata
lock is held:

      cache_metadata->root_lock
         policy->lock

Fix this potential lock-inversion (ABBA) deadlock in sync_metadata() by
ensuring the cache_metadata root_lock is held whilst all the hints are
written, rather than being repeatedly locked while policy->lock is held
(as was the case with each callout that policy_walk_mappings() made to
the old save_hint() method).

Found by turning on the CONFIG_PROVE_LOCKING ("Lock debugging: prove
locking correctness") build option.  However, it is not clear how the
LOCKDEP reported paths can lead to a deadlock since the two paths,
suspending a target and loading a target, never occur at the same time.
But that doesn't mean the same lock-inversion couldn't have occurred
elsewhere.

Reported-by: Marian Csontos <mcsontos@redhat.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm thin: fix dangling bio in process_deferred_bios error path

commit fe76cd88e654124d1431bb662a0fc6e99ca811a5 upstream.

If unable to ensure_next_mapping() we must add the current bio, which
was removed from the @bios list via bio_list_pop, back to the
deferred_bios list before all the remaining @bios.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm: take care to copy the space map roots before locking the superblock

commit 5a32083d03fb543f63489b2946c4948398579ba0 upstream.

In theory copying the space map root can fail, but in practice it never
does because we're careful to check what size buffer is needed.

But make certain we're able to copy the space map roots before
locking the superblock.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm transaction manager: fix corruption due to non-atomic transaction commit

commit a9d45396f5956d0b615c7ae3b936afd888351a47 upstream.

The persistent-data library used by dm-thin, dm-cache, etc is
transactional.  If anything goes wrong, such as an io error when writing
new metadata or a power failure, then we roll back to the last
transaction.

Atomicity when committing a transaction is achieved by:

a) Never overwriting data from the previous transaction.
b) Writing the superblock last, after all other metadata has hit the
   disk.

This commit and the following commit ("dm: take care to copy the space
map roots before locking the superblock") fix a bug associated with (b).
When committing it was possible for the superblock to still be written
in spite of an io error occurring during the preceeding metadata flush.
With these commits we're careful not to take the write lock out on the
superblock until after the metadata flush has completed.

Change the transaction manager's semantics for dm_tm_commit() to assume
all data has been flushed _before_ the single superblock that is passed
in.

As a prerequisite, split the block manager's block unlocking and
flushing by simplifying dm_bm_flush_and_unlock() to dm_bm_flush().  Now
the unlocking must be done separately.

This issue was discovered by forcing io errors at the crucial time
using dm-flakey.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm cache: prevent corruption caused by discard_block_size > cache_block_size

commit d132cc6d9e92424bb9d4fd35f5bd0e55d583f4be upstream.

If the discard block size is larger than the cache block size we will
not properly quiesce IO to a region that is about to be discarded.  This
results in a race between a cache migration where no copy is needed, and
a write to an adjacent cache block that's within the same large discard
block.

Workaround this by limiting the discard_block_size to cache_block_size.
Also limit the max_discard_sectors to cache_block_size.

A more comprehensive fix that introduces range locking support in the
bio_prison and proper quiescing of a discard range that spans multiple
cache blocks is already in development.

Reported-by: Morgan Mears <Morgan.Mears@netapp.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Acked-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Skip intel_crt_init for Dell XPS 8700

commit 10b6ee4a87811a110cb01eaca01eb04da6801baf upstream.

The Dell XPS 8700 has a onboard Display port and HDMI port and no VGA port.
The call intel_crt_init freeze the machine, so skip such call.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73559
Signed-off-by: Giacomo Comes <comes at naic.edu>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>