cgroup: fix to break the while loop in cgroup_attach_task() correctly

commit ea84753c98a7ac6b74e530b64c444a912b3835ca upstream.

Both Anjana and Eunki reported a stall in the while_each_thread loop
in cgroup_attach_task().

It's because, when we attach a single thread to a cgroup, if the cgroup
is exiting or is already in that cgroup, we won't break the loop.

If the task is already in the cgroup, the bug can lead to another thread
being attached to the cgroup unexpectedly:

  # echo 5207 > tasks
  # cat tasks
  5207
  # echo 5207 > tasks
  # cat tasks
  5207
  5215

What's worse, if the task to be attached isn't the leader of the thread
group, we might never exit the loop, hence cpu stall. Thanks for Oleg's
analysis.

This bug was introduced by commit 081aa458c38ba576bdd4265fc807fa95b48b9e79
("cgroup: consolidate cgroup_attach_task() and cgroup_attach_proc()")

[ lizf: - fixed the first continue, pointed out by Oleg,
        - rewrote changelog. ]

Reported-by: Eunki Kim <eunki_kim@samsung.com>
Reported-by: Anjana V Kumar <anjanavk12@gmail.com>
Signed-off-by: Anjana V Kumar <anjanavk12@gmail.com>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bcache: Fixed incorrect order of arguments to bio_alloc_bioset()

commit d4eddd42f592a0cf06818fae694a3d271f842e4d upstream.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cpufreq / intel_pstate: Fix max_perf_pct on resume

commit 52e0a509e5d6f902ec26bc2a8bb02b137dc453be upstream.

If the system is suspended while max_perf_pct is less than 100 percent
or no_turbo set policy->{min,max} will be set incorrectly with scaled
values which turn the scaled values into hard limits.

References: https://bugzilla.kernel.org/show_bug.cgi?id=61241
Reported-by: Patrick Bartels <petzicus@googlemail.com>
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86: Update UV3 hub revision ID

commit dd3c9c4b603c664fedc12facf180db0f1794aafe upstream.

The UV3 hub revision ID is different than expected.  The first
revision was supposed to start at 1 but instead will start at 0.

Signed-off-by: Russ Anderson <rja@sgi.com>
Link: http://lkml.kernel.org/r/20131014161733.GA6274@sgi.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: serial: ftdi_sio: add id for Z3X Box device

commit e1466ad5b1aeda303f9282463d55798d2eda218c upstream.

Custom VID/PID for Z3X Box device, popular tool for cellphone flashing.

Signed-off-by: Alexey E. Kramarenko <alexeyk13@yandex.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: quirks: add touchscreen that is dazzeled by remote wakeup

commit 614ced91fc6fbb5a1cdd12f0f1b6c9197d9f1350 upstream.

The device descriptors are messed up after remote wakeup

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: quirks.c: add one device that cannot deal with suspension

commit 4294bca7b423d1a5aa24307e3d112a04075e3763 upstream.

The device is not responsive when resumed, unless it is reset.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: support new huawei devices in option.c

commit d544db293a44a2a3b09feab7dbd59668b692de71 upstream.

Add new supporting declarations to option.c, to support Huawei new
devices with new bInterfaceSubClass value.

Signed-off-by: fangxiaozhi <huananhu@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb-storage: add quirk for mandatory READ_CAPACITY_16

commit 32c37fc30c52508711ea6a108cfd5855b8a07176 upstream.

Some USB drive enclosures do not correctly report an
overflow condition if they hold a drive with a capacity
over 2TB and are confronted with a READ_CAPACITY_10.
They answer with their capacity modulo 2TB.
The generic layer cannot cope with that. It must be told
to use READ_CAPACITY_16 from the beginning.

Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 3.10.18

usb: serial: option: blacklist Olivetti Olicard200

commit fd8573f5828873343903215f203f14dc82de397c upstream.

Interface 6 of this device speaks QMI as per tests done by us.
Credits go to Antonella for providing the hardware.

Signed-off-by: Enrico Mioso <mrkiko.rs@gmail.com>
Signed-off-by: Antonella Pellizzari <anto.pellizzari83@gmail.com>
Tested-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: serial: option: add support for Inovia SEW858 device

commit f4c19b8e165cff1a6607c21f8809441d61cab7ec upstream.

This patch adds the device id for the Inovia SEW858 device to the option driver.

Reported-by: Pavel Parkhomenko <ra85551@gmail.com>
Tested-by: Pavel Parkhomenko <ra85551@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: serial: ti_usb_3410_5052: add Abbott strip port ID to combined table as well.

commit c9d09dc7ad106492c17c587b6eeb99fe3f43e522 upstream.

Without this change, the USB cable for Freestyle Option and compatible
glucometers will not be detected by the driver.

Signed-off-by: Diego Elio Pettenò <flameeyes@flameeyes.eu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

serial: vt8500: add missing braces

commit d969de8d83401683420638c8107dcfedb2146f37 upstream.

Due to missing braces on an if statement, in presence of a device_node a
port was always assigned -1, regardless of any alias entries in the
device tree. Conversely, if device_node was NULL, an unitialized port
ended up being used.

This patch adds the missing braces, fixing the issues.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Tony Prisk <linux@prisktech.co.nz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wireless: radiotap: fix parsing buffer overrun

commit f5563318ff1bde15b10e736e97ffce13be08bc1a upstream.

When parsing an invalid radiotap header, the parser can overrun
the buffer that is passed in because it doesn't correctly check
 1) the minimum radiotap header size
 2) the space for extended bitmaps

The first issue doesn't affect any in-kernel user as they all
check the minimum size before calling the radiotap function.
The second issue could potentially affect the kernel if an skb
is passed in that consists only of the radiotap header with a
lot of extended bitmaps that extend past the SKB. In that case
a read-only buffer overrun by at most 4 bytes is possible.

Fix this by adding the appropriate checks to the parser.

Reported-by: Evan Huus <eapache@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

writeback: fix negative bdi max pause

commit e3b6c655b91e01a1dade056cfa358581b47a5351 upstream.

Toralf runs trinity on UML/i386.  After some time it hangs and the last
message line is

BUG: soft lockup - CPU#0 stuck for 22s! [trinity-child0:1521]

It's found that pages_dirtied becomes very large.  More than 1000000000
pages in this case:

period = HZ * pages_dirtied / task_ratelimit;
BUG_ON(pages_dirtied > 2000000000);
BUG_ON(pages_dirtied > 1000000000);      <---------

UML debug printf shows that we got negative pause here:

ick: pause : -984
ick: pages_dirtied : 0
ick: task_ratelimit: 0

 pause:
+       if (pause < 0)  {
+               extern int printf(char *, ...);
+               printf("ick : pause : %li\n", pause);
+               printf("ick: pages_dirtied : %lu\n", pages_dirtied);
+               printf("ick: task_ratelimit: %lu\n", task_ratelimit);
+               BUG_ON(1);
+       }
        trace_balance_dirty_pages(bdi,

Since pause is bounded by [min_pause, max_pause] where min_pause is also
bounded by max_pause.  It's suspected and demonstrated that the
max_pause calculation goes wrong:

ick: pause : -717
ick: min_pause : -177
ick: max_pause : -717
ick: pages_dirtied : 14
ick: task_ratelimit: 0

The problem lies in the two "long = unsigned long" assignments in
bdi_max_pause() which might go negative if the highest bit is 1, and the
min_t(long, ...) check failed to protect it falling under 0.  Fix all of
them by using "unsigned long" throughout the function.

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Toralf Förster <toralf.foerster@gmx.de>
Tested-by: Toralf Förster <toralf.foerster@gmx.de>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Richard Weinberger <richard@nod.at>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
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>

ALSA: hda - Fix inverted internal mic not indicated on some machines

commit ccb041571b73888785ef7828a276e380125891a4 upstream.

The create_bind_cap_vol_ctl does not create any control indicating
that an inverted dmic is present. Therefore, create multiple
capture volumes in this scenario, so we always have some indication
that the internal mic is inverted.

This happens on the Lenovo Ideapad U310 as well as the Lenovo Yoga 13
(both are based on the CX20590 codec), but the fix is generic and
could be needed for other codecs/machines too.

Thanks to Szymon Acedański for the pointer and a draft patch.

BugLink: https://bugs.launchpad.net/bugs/1239392
BugLink: https://bugs.launchpad.net/bugs/1227491
Reported-by: Szymon Acedański <accek@mimuw.edu.pl>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ALSA: us122l: Fix pcm_usb_stream mmapping regression

commit ac536a848a1643e4b87e8fbd376a63091afc2ccc upstream.

The pcm_usb_stream plugin requires the mremap explicitly for the read
buffer, as it expands itself once after reading the required size.
But the commit [314e51b9: mm: kill vma flag VM_RESERVED and
mm->reserved_vm counter] converted blindly to a combination of
VM_DONTEXPAND | VM_DONTDUMP like other normal drivers, and this
resulted in the failure of mremap().

For fixing this regression, we need to remove VM_DONTEXPAND for the
read-buffer mmap.

Reported-and-tested-by: James Miller <jamesstewartmiller@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm: fix BUG in __split_huge_page_pmd

commit 750e8165f5e87b6a142be953640eabb13a9d350a upstream.

Occasionally we hit the BUG_ON(pmd_trans_huge(*pmd)) at the end of
__split_huge_page_pmd(): seen when doing madvise(,,MADV_DONTNEED).

It's invalid: we don't always have down_write of mmap_sem there: a racing
do_huge_pmd_wp_page() might have copied-on-write to another huge page
before our split_huge_page() got the anon_vma lock.

Forget the BUG_ON, just go back and try again if this happens.

Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.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>

i2c: ismt: initialize DMA buffer

commit bf4169100c909667ede6af67668b3ecce6928343 upstream.

This patch adds code to initialize the DMA buffer to compensate for
possible hardware data corruption.

Signed-off-by: James Ralston <james.d.ralston@intel.com>
[wsa: changed to use 'sizeof']
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Cc: Jean Delvare <jdelvare@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm snapshot: fix data corruption

commit e9c6a182649f4259db704ae15a91ac820e63b0ca upstream.

This patch fixes a particular type of data corruption that has been
encountered when loading a snapshot's metadata from disk.

When we allocate a new chunk in persistent_prepare, we increment
ps->next_free and we make sure that it doesn't point to a metadata area
by further incrementing it if necessary.

When we load metadata from disk on device activation, ps->next_free is
positioned after the last used data chunk. However, if this last used
data chunk is followed by a metadata area, ps->next_free is positioned
erroneously to the metadata area. A newly-allocated chunk is placed at
the same location as the metadata area, resulting in data or metadata
corruption.

This patch changes the code so that ps->next_free skips the metadata
area when metadata are loaded in function read_exceptions.

The patch also moves a piece of code from persistent_prepare_exception
to a separate function skip_metadata to avoid code duplication.

CVE-2013-4299

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

gpio/lynxpoint: check if the interrupt is enabled in IRQ handler

commit 03d152d5582abc8a1c19cb107164c3724bbd4be4 upstream.

Checking LP_INT_STAT is not enough in the interrupt handler because its
contents get updated regardless of whether the pin has interrupt enabled or
not. This causes the driver to loop forever for GPIOs that are pulled up.

Fix this by checking the interrupt enable bit for the pin as well.

Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: integrator: deactivate timer0 on the Integrator/CP

commit 29114fd7db2fc82a34da8340d29b8fa413e03dca upstream.

This fixes a long-standing Integrator/CP regression from
commit 870e2928cf3368ca9b06bc925d0027b0a56bcd8e
"ARM: integrator-cp: convert use CLKSRC_OF for timer init"

When this code was introduced, the both aliases pointing the
system to use timer1 as primary (clocksource) and timer2
as secondary (clockevent) was ignored, and the system would
simply use the first two timers found as clocksource and
clockevent.

However this made the system timeline accelerate by a
factor x25, as it turns out that the way the clocking
actually works (totally undocumented and found after some
trial-and-error) is that timer0 runs @ 25MHz and timer1
and timer2 runs @ 1MHz. Presumably this divider setting
is a boot-on default and configurable albeit the way to
configure it is not documented.

So as a quick fix to the problem, let's mark timer0 as
disabled, so the code will chose timer1 and timer2 as it
used to.

This also deletes the two aliases for the primary and
secondary timer as they have been superceded by the
auto-selection

Cc: Rob Herring <rob.herring@calxeda.com>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: 7851/1: check for number of arguments in syscall_get/set_arguments()

commit 3c1532df5c1b54b5f6246cdef94eeb73a39fe43a upstream.

In ftrace_syscall_enter(),
    syscall_get_arguments(..., 0, n, ...)
        if (i == 0) { <handle ORIG_r0> ...; n--;}
        memcpy(..., n * sizeof(args[0]));
If 'number of arguments(n)' is zero and 'argument index(i)' is also zero in
syscall_get_arguments(), none of arguments should be copied by memcpy().
Otherwise 'n--' can be a big positive number and unexpected amount of data
will be copied. Tracing system calls which take no argument, say sync(void),
may hit this case and eventually make the system corrupted.
This patch fixes the issue both in syscall_get_arguments() and
syscall_set_arguments().

Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

davinci_emac.c: Fix IFF_ALLMULTI setup

[ Upstream commit d69e0f7ea95fef8059251325a79c004bac01f018 ]

When IFF_ALLMULTI flag is set on interface and IFF_PROMISC isn't,
emac_dev_mcast_set should only enable RX of multicasts and reset
MACHASH registers.

It does this, but afterwards it either sets up multicast MACs
filtering or disables RX of multicasts and resets MACHASH registers
again, rendering IFF_ALLMULTI flag useless.

This patch fixes emac_dev_mcast_set, so that multicast MACs filtering and
disabling of RX of multicasts are skipped when IFF_ALLMULTI flag is set.

Tested with kernel 2.6.37.

Signed-off-by: Mariusz Ceier <mceier+kernel@gmail.com>
Acked-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv6: probe routes asynchronous in rt6_probe

[ Upstream commit c2f17e827b419918c856131f592df9521e1a38e3 ]

Routes need to be probed asynchronous otherwise the call stack gets
exhausted when the kernel attemps to deliver another skb inline, like
e.g. xt_TEE does, and we probe at the same time.

We update neigh->updated still at once, otherwise we would send to
many probes.

Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-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>

netfilter: nf_conntrack: fix rt6i_gateway checks for H.323 helper

[ Upstream commit 56e42441ed54b092d6c7411138ce60d049e7c731 ]

Now when rt6_nexthop() can return nexthop address we can use it
for proper nexthop comparison of directly connected destinations.
For more information refer to commit bbb5823cf742a7
("netfilter: nf_conntrack: fix rt_gateway checks for H.323 helper").

Signed-off-by: Julian Anastasov <ja@ssi.bg>
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>

ipv6: fill rt6i_gateway with nexthop address

[ Upstream commit 550bab42f83308c9d6ab04a980cc4333cef1c8fa ]

Make sure rt6i_gateway contains nexthop information in
all routes returned from lookup or when routes are directly
attached to skb for generated ICMP packets.

The effect of this patch should be a faster version of
rt6_nexthop() and the consideration of local addresses as
nexthop.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
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>

ipv6: always prefer rt6i_gateway if present

[ Upstream commit 96dc809514fb2328605198a0602b67554d8cce7b ]

In v3.9 6fd6ce2056de2709 ("ipv6: Do not depend on rt->n in
ip6_finish_output2()." changed the behaviour of ip6_finish_output2()
such that the recently introduced rt6_nexthop() is used
instead of an assigned neighbor.

As rt6_nexthop() prefers rt6i_gateway only for gatewayed
routes this causes a problem for users like IPVS, xt_TEE and
RAW(hdrincl) if they want to use different address for routing
compared to the destination address.

Another case is when redirect can create RTF_DYNAMIC
route without RTF_GATEWAY flag, we ignore the rt6i_gateway
in rt6_nexthop().

Fix the above problems by considering the rt6i_gateway if
present, so that traffic routed to address on local subnet is
not wrongly diverted to the destination address.

Thanks to Simon Horman and Phil Oester for spotting the
problematic commit.

Thanks to Hannes Frederic Sowa for his review and help in testing.

Reported-by: Phil Oester <kernel@linuxace.com>
Reported-by: Mark Brooks <mark@loadbalancer.org>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
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>

inet: fix possible memory corruption with UDP_CORK and UFO

[ This is a simplified -stable version of a set of upstream commits. ]

This is a replacement patch only for stable which does fix the problems
handled by the following two commits in -net:

"ip_output: do skb ufo init for peeked non ufo skb as well" (e93b7d748be887cd7639b113ba7d7ef792a7efb9)
"ip6_output: do skb ufo init for peeked non ufo skb as well" (c547dbf55d5f8cf615ccc0e7265e98db27d3fb8b)

Three frames are written on a corked udp socket for which the output
netdevice has UFO enabled.  If the first and third frame are smaller than
the mtu and the second one is bigger, we enqueue the second frame with
skb_append_datato_frags without initializing the gso fields. This leads
to the third frame appended regulary and thus constructing an invalid skb.

This fixes the problem by always using skb_append_datato_frags as soon
as the first frag got enqueued to the skb without marking the packet
as SKB_GSO_UDP.

The problem with only two frames for ipv6 was fixed by "ipv6: udp
packets following an UFO enqueued packet need also be handled by UFO"
(2811ebac2521ceac84f2bdae402455baa6a7fb47).

Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Jiri Pirko <jiri@resnulli.us>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: fix cipso packet validation when !NETLABEL

[ Upstream commit f2e5ddcc0d12f9c4c7b254358ad245c9dddce13b ]

When CONFIG_NETLABEL is disabled, the cipso_v4_validate() function could loop
forever in the main loop if opt[opt_iter +1] == 0, this will causing a kernel
crash in an SMP system, since the CPU executing this function will
stall /not respond to IPIs.

This problem can be reproduced by running the IP Stack Integrity Checker
(http://isic.sourceforge.net) using the following command on a Linux machine
connected to DUT:

"icmpsic -s rand -d <DUT IP address> -r 123456"
wait (1-2 min)

Signed-off-by: Seif Mazareeb <seif@marvell.com>
Acked-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: unix: inherit SOCK_PASS{CRED, SEC} flags from socket to fix race

[ Upstream commit 90c6bd34f884cd9cee21f1d152baf6c18bcac949 ]

In the case of credentials passing in unix stream sockets (dgram
sockets seem not affected), we get a rather sparse race after
commit 16e5726 ("af_unix: dont send SCM_CREDENTIALS by default").

We have a stream server on receiver side that requests credential
passing from senders (e.g. nc -U). Since we need to set SO_PASSCRED
on each spawned/accepted socket on server side to 1 first (as it's
not inherited), it can happen that in the time between accept() and
setsockopt() we get interrupted, the sender is being scheduled and
continues with passing data to our receiver. At that time SO_PASSCRED
is neither set on sender nor receiver side, hence in cmsg's
SCM_CREDENTIALS we get eventually pid:0, uid:65534, gid:65534
(== overflow{u,g}id) instead of what we actually would like to see.

On the sender side, here nc -U, the tests in maybe_add_creds()
invoked through unix_stream_sendmsg() would fail, as at that exact
time, as mentioned, the sender has neither SO_PASSCRED on his side
nor sees it on the server side, and we have a valid 'other' socket
in place. Thus, sender believes it would just look like a normal
connection, not needing/requesting SO_PASSCRED at that time.

As reverting 16e5726 would not be an option due to the significant
performance regression reported when having creds always passed,
one way/trade-off to prevent that would be to set SO_PASSCRED on
the listener socket and allow inheriting these flags to the spawned
socket on server side in accept(). It seems also logical to do so
if we'd tell the listener socket to pass those flags onwards, and
would fix the race.

Before, strace:

recvmsg(4, {msg_name(0)=NULL, msg_iov(1)=[{"blub\n", 4096}],
        msg_controllen=32, {cmsg_len=28, cmsg_level=SOL_SOCKET,
        cmsg_type=SCM_CREDENTIALS{pid=0, uid=65534, gid=65534}},
        msg_flags=0}, 0) = 5

After, strace:

recvmsg(4, {msg_name(0)=NULL, msg_iov(1)=[{"blub\n", 4096}],
        msg_controllen=32, {cmsg_len=28, cmsg_level=SOL_SOCKET,
        cmsg_type=SCM_CREDENTIALS{pid=11580, uid=1000, gid=1000}},
        msg_flags=0}, 0) = 5

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

be2net: pass if_id for v1 and V2 versions of TX_CREATE cmd

[ Upstream commit 0fb88d61bc60779dde88b0fc268da17eb81d0412 ]

It is a required field for all TX_CREATE cmd versions > 0.
This fixes a driver initialization failure, caused by recent SH-R Firmwares
(versions > 10.0.639.0) failing the TX_CREATE cmd when if_id field is
not passed.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wanxl: fix info leak in ioctl

[ Upstream commit 2b13d06c9584b4eb773f1e80bbaedab9a1c344e1 ]

The wanxl_ioctl() code fails to initialize the two padding bytes of
struct sync_serial_settings after the ->loopback member. Add an explicit
memset(0) before filling the structure to avoid the info leak.

Signed-off-by: Salva Peiró <speiro@ai2.upv.es>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sctp: Perform software checksum if packet has to be fragmented.

[ Upstream commit d2dbbba77e95dff4b4f901fee236fef6d9552072 ]

IP/IPv6 fragmentation knows how to compute only TCP/UDP checksum.
This causes problems if SCTP packets has to be fragmented and
ipsummed has been set to PARTIAL due to checksum offload support.
This condition can happen when retransmitting after MTU discover,
or when INIT or other control chunks are larger then MTU.
Check for the rare fragmentation condition in SCTP and use software
checksum calculation in this case.

CC: Fan Du <fan.du@windriver.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.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>

sctp: Use software crc32 checksum when xfrm transform will happen.

[ Upstream commit 27127a82561a2a3ed955ce207048e1b066a80a2a ]

igb/ixgbe have hardware sctp checksum support, when this feature is enabled
and also IPsec is armed to protect sctp traffic, ugly things happened as
xfrm_output checks CHECKSUM_PARTIAL to do checksum operation(sum every thing
up and pack the 16bits result in the checksum field). The result is fail
establishment of sctp communication.

Signed-off-by: Fan Du <fan.du@windriver.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Vlad Yasevich <vyasevich@gmail.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>

net: dst: provide accessor function to dst->xfrm

[ Upstream commit e87b3998d795123b4139bc3f25490dd236f68212 ]

dst->xfrm is conditionally defined.  Provide accessor funtion that
is always available.

Signed-off-by: Vlad Yasevich <vyasevich@gmail.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>

bridge: Correctly clamp MAX forward_delay when enabling STP

[ Upstream commit 4b6c7879d84ad06a2ac5b964808ed599187a188d ]

Commit be4f154d5ef0ca147ab6bcd38857a774133f5450
bridge: Clamp forward_delay when enabling STP
had a typo when attempting to clamp maximum forward delay.

It is possible to set bridge_forward_delay to be higher then
permitted maximum when STP is off.  When turning STP on, the
higher then allowed delay has to be clamed down to max value.

Signed-off-by: Vlad Yasevich <vyasevic@redhat.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Stephen Hemminger <shemminger@vyatta.com>
Reviewed-by: Veaceslav Falico <vfalico@redhat.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

virtio-net: refill only when device is up during setting queues

[ Upstream commit 35ed159bfd96a7547ec277ed8b550c7cbd9841b6 ]

We used to schedule the refill work unconditionally after changing the
number of queues. This may lead an issue if the device is not
up. Since we only try to cancel the work in ndo_stop(), this may cause
the refill work still work after removing the device. Fix this by only
schedule the work when device is up.

The bug were introduce by commit 9b9cd8024a2882e896c65222aa421d461354e3f2.
(virtio-net: fix the race between channels setting and refill)

Signed-off-by: Jason Wang <jasowang@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

virtio-net: fix the race between channels setting and refill

[ Upstream commit 9b9cd8024a2882e896c65222aa421d461354e3f2 ]

Commit 55257d72bd1c51f25106350f4983ec19f62ed1fa (virtio-net: fill only rx queues
which are being used) tries to refill on demand when changing the number of
channels by call try_refill_recv() directly, this may race:

- the refill work who may do the refill in the same time
- the try_refill_recv() called in bh since napi was not disabled

Which may led guest complain during setting channels:

virtio_net virtio0: input.1:id 0 is not a head!

Solve this issue by scheduling a refill work which can guarantee the
serialization of refill.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

virtio-net: don't respond to cpu hotplug notifier if we're not ready

[ Upstream commit 3ab098df35f8b98b6553edc2e40234af512ba877 ]

We're trying to re-configure the affinity unconditionally in cpu hotplug
callback. This may lead the issue during resuming from s3/s4 since

- virt queues haven't been allocated at that time.
- it's unnecessary since thaw method will re-configure the affinity.

Fix this issue by checking the config_enable and do nothing is we're not ready.

The bug were introduced by commit 8de4b2f3ae90c8fc0f17eeaab87d5a951b66ee17
(virtio-net: reset virtqueue affinity when doing cpu hotplug).

Acked-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bnx2x: record rx queue for LRO packets

[ Upstream commit 60e66fee56b2256dcb1dc2ea1b2ddcb6e273857d ]

RPS support is kind of broken on bnx2x, because only non LRO packets
get proper rx queue information. This triggers reorders, as it seems
bnx2x like to generate a non LRO packet for segment including TCP PUSH
flag : (this might be pure coincidence, but all the reorders I've
seen involve segments with a PUSH)

11:13:34.335847 IP A > B: . 415808:447136(31328) ack 1 win 457 <nop,nop,timestamp 3789336 3985797>
11:13:34.335992 IP A > B: . 447136:448560(1424) ack 1 win 457 <nop,nop,timestamp 3789336 3985797>
11:13:34.336391 IP A > B: . 448560:479888(31328) ack 1 win 457 <nop,nop,timestamp 3789337 3985797>
11:13:34.336425 IP A > B: P 511216:512640(1424) ack 1 win 457 <nop,nop,timestamp 3789337 3985798>
11:13:34.336423 IP A > B: . 479888:511216(31328) ack 1 win 457 <nop,nop,timestamp 3789337 3985798>
11:13:34.336924 IP A > B: . 512640:543968(31328) ack 1 win 457 <nop,nop,timestamp 3789337 3985798>
11:13:34.336963 IP A > B: . 543968:575296(31328) ack 1 win 457 <nop,nop,timestamp 3789337 3985798>

We must call skb_record_rx_queue() to properly give to RPS (and more
generally for TX queue selection on forward path) the receive queue
information.

Similar fix is needed for skb_mark_napi_id(), but will be handled
in a separate patch to ease stable backports.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Acked-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

connector: use nlmsg_len() to check message length

[ Upstream commit 162b2bedc084d2d908a04c93383ba02348b648b0 ]

The current code tests the length of the whole netlink message to be
at least as long to fit a cn_msg. This is wrong as nlmsg_len includes
the length of the netlink message header. Use nlmsg_len() instead to
fix this "off-by-NLMSG_HDRLEN" size check.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

unix_diag: fix info leak

[ Upstream commit 6865d1e834be84ddd5808d93d5035b492346c64a ]

When filling the netlink message we miss to wipe the pad field,
therefore leak one byte of heap memory to userland. Fix this by
setting pad to 0.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

farsync: fix info leak in ioctl

[ Upstream commit 96b340406724d87e4621284ebac5e059d67b2194 ]

The fst_get_iface() code fails to initialize the two padding bytes of
struct sync_serial_settings after the ->loopback member. Add an explicit
memset(0) before filling the structure to avoid the info leak.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

l2tp: must disable bh before calling l2tp_xmit_skb()

[ Upstream commit 455cc32bf128e114455d11ad919321ab89a2c312 ]

François Cachereul made a very nice bug report and suspected
the bh_lock_sock() / bh_unlok_sock() pair used in l2tp_xmit_skb() from
process context was not good.

This problem was added by commit 6af88da14ee284aaad6e4326da09a89191ab6165
("l2tp: Fix locking in l2tp_core.c").

l2tp_eth_dev_xmit() runs from BH context, so we must disable BH
from other l2tp_xmit_skb() users.

[  452.060011] BUG: soft lockup - CPU#1 stuck for 23s! [accel-pppd:6662]
[  452.061757] Modules linked in: l2tp_ppp l2tp_netlink l2tp_core pppoe pppox
ppp_generic slhc ipv6 ext3 mbcache jbd virtio_balloon xfs exportfs dm_mod
virtio_blk ata_generic virtio_net floppy ata_piix libata virtio_pci virtio_ring virtio [last unloaded: scsi_wait_scan]
[  452.064012] CPU 1
[  452.080015] BUG: soft lockup - CPU#2 stuck for 23s! [accel-pppd:6643]
[  452.080015] CPU 2
[  452.080015]
[  452.080015] Pid: 6643, comm: accel-pppd Not tainted 3.2.46.mini #1 Bochs Bochs
[  452.080015] RIP: 0010:[<ffffffff81059f6c>]  [<ffffffff81059f6c>] do_raw_spin_lock+0x17/0x1f
[  452.080015] RSP: 0018:ffff88007125fc18  EFLAGS: 00000293
[  452.080015] RAX: 000000000000aba9 RBX: ffffffff811d0703 RCX: 0000000000000000
[  452.080015] RDX: 00000000000000ab RSI: ffff8800711f6896 RDI: ffff8800745c8110
[  452.080015] RBP: ffff88007125fc18 R08: 0000000000000020 R09: 0000000000000000
[  452.080015] R10: 0000000000000000 R11: 0000000000000280 R12: 0000000000000286
[  452.080015] R13: 0000000000000020 R14: 0000000000000240 R15: 0000000000000000
[  452.080015] FS:  00007fdc0cc24700(0000) GS:ffff8800b6f00000(0000) knlGS:0000000000000000
[  452.080015] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  452.080015] CR2: 00007fdb054899b8 CR3: 0000000074404000 CR4: 00000000000006a0
[  452.080015] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  452.080015] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  452.080015] Process accel-pppd (pid: 6643, threadinfo ffff88007125e000, task ffff8800b27e6dd0)
[  452.080015] Stack:
[  452.080015]  ffff88007125fc28 ffffffff81256559 ffff88007125fc98 ffffffffa01b2bd1
[  452.080015]  ffff88007125fc58 000000000000000c 00000000029490d0 0000009c71dbe25e
[  452.080015]  000000000000005c 000000080000000e 0000000000000000 ffff880071170600
[  452.080015] Call Trace:
[  452.080015]  [<ffffffff81256559>] _raw_spin_lock+0xe/0x10
[  452.080015]  [<ffffffffa01b2bd1>] l2tp_xmit_skb+0x189/0x4ac [l2tp_core]
[  452.080015]  [<ffffffffa01c2d36>] pppol2tp_sendmsg+0x15e/0x19c [l2tp_ppp]
[  452.080015]  [<ffffffff811c7872>] __sock_sendmsg_nosec+0x22/0x24
[  452.080015]  [<ffffffff811c83bd>] sock_sendmsg+0xa1/0xb6
[  452.080015]  [<ffffffff81254e88>] ? __schedule+0x5c1/0x616
[  452.080015]  [<ffffffff8103c7c6>] ? __dequeue_signal+0xb7/0x10c
[  452.080015]  [<ffffffff810bbd21>] ? fget_light+0x75/0x89
[  452.080015]  [<ffffffff811c8444>] ? sockfd_lookup_light+0x20/0x56
[  452.080015]  [<ffffffff811c9b34>] sys_sendto+0x10c/0x13b
[  452.080015]  [<ffffffff8125cac2>] system_call_fastpath+0x16/0x1b
[  452.080015] Code: 81 48 89 e5 72 0c 31 c0 48 81 ff 45 66 25 81 0f 92 c0 5d c3 55 b8 00 01 00 00 48 89 e5 f0 66 0f c1 07 0f b6 d4 38 d0 74 06 f3 90 <8a> 07 eb f6 5d c3 90 90 55 48 89 e5 9c 58 0f 1f 44 00 00 5d c3
[  452.080015] Call Trace:
[  452.080015]  [<ffffffff81256559>] _raw_spin_lock+0xe/0x10
[  452.080015]  [<ffffffffa01b2bd1>] l2tp_xmit_skb+0x189/0x4ac [l2tp_core]
[  452.080015]  [<ffffffffa01c2d36>] pppol2tp_sendmsg+0x15e/0x19c [l2tp_ppp]
[  452.080015]  [<ffffffff811c7872>] __sock_sendmsg_nosec+0x22/0x24
[  452.080015]  [<ffffffff811c83bd>] sock_sendmsg+0xa1/0xb6
[  452.080015]  [<ffffffff81254e88>] ? __schedule+0x5c1/0x616
[  452.080015]  [<ffffffff8103c7c6>] ? __dequeue_signal+0xb7/0x10c
[  452.080015]  [<ffffffff810bbd21>] ? fget_light+0x75/0x89
[  452.080015]  [<ffffffff811c8444>] ? sockfd_lookup_light+0x20/0x56
[  452.080015]  [<ffffffff811c9b34>] sys_sendto+0x10c/0x13b
[  452.080015]  [<ffffffff8125cac2>] system_call_fastpath+0x16/0x1b
[  452.064012]
[  452.064012] Pid: 6662, comm: accel-pppd Not tainted 3.2.46.mini #1 Bochs Bochs
[  452.064012] RIP: 0010:[<ffffffff81059f6e>]  [<ffffffff81059f6e>] do_raw_spin_lock+0x19/0x1f
[  452.064012] RSP: 0018:ffff8800b6e83ba0  EFLAGS: 00000297
[  452.064012] RAX: 000000000000aaa9 RBX: ffff8800b6e83b40 RCX: 0000000000000002
[  452.064012] RDX: 00000000000000aa RSI: 000000000000000a RDI: ffff8800745c8110
[  452.064012] RBP: ffff8800b6e83ba0 R08: 000000000000c802 R09: 000000000000001c
[  452.064012] R10: ffff880071096c4e R11: 0000000000000006 R12: ffff8800b6e83b18
[  452.064012] R13: ffffffff8125d51e R14: ffff8800b6e83ba0 R15: ffff880072a589c0
[  452.064012] FS:  00007fdc0b81e700(0000) GS:ffff8800b6e80000(0000) knlGS:0000000000000000
[  452.064012] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  452.064012] CR2: 0000000000625208 CR3: 0000000074404000 CR4: 00000000000006a0
[  452.064012] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  452.064012] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  452.064012] Process accel-pppd (pid: 6662, threadinfo ffff88007129a000, task ffff8800744f7410)
[  452.064012] Stack:
[  452.064012]  ffff8800b6e83bb0 ffffffff81256559 ffff8800b6e83bc0 ffffffff8121c64a
[  452.064012]  ffff8800b6e83bf0 ffffffff8121ec7a ffff880072a589c0 ffff880071096c62
[  452.064012]  0000000000000011 ffffffff81430024 ffff8800b6e83c80 ffffffff8121f276
[  452.064012] Call Trace:
[  452.064012]  <IRQ>
[  452.064012]  [<ffffffff81256559>] _raw_spin_lock+0xe/0x10
[  452.064012]  [<ffffffff8121c64a>] spin_lock+0x9/0xb
[  452.064012]  [<ffffffff8121ec7a>] udp_queue_rcv_skb+0x186/0x269
[  452.064012]  [<ffffffff8121f276>] __udp4_lib_rcv+0x297/0x4ae
[  452.064012]  [<ffffffff8121c178>] ? raw_rcv+0xe9/0xf0
[  452.064012]  [<ffffffff8121f4a7>] udp_rcv+0x1a/0x1c
[  452.064012]  [<ffffffff811fe385>] ip_local_deliver_finish+0x12b/0x1a5
[  452.064012]  [<ffffffff811fe54e>] ip_local_deliver+0x53/0x84
[  452.064012]  [<ffffffff811fe1d0>] ip_rcv_finish+0x2bc/0x2f3
[  452.064012]  [<ffffffff811fe78f>] ip_rcv+0x210/0x269
[  452.064012]  [<ffffffff8101911e>] ? kvm_clock_get_cycles+0x9/0xb
[  452.064012]  [<ffffffff811d88cd>] __netif_receive_skb+0x3a5/0x3f7
[  452.064012]  [<ffffffff811d8eba>] netif_receive_skb+0x57/0x5e
[  452.064012]  [<ffffffff811cf30f>] ? __netdev_alloc_skb+0x1f/0x3b
[  452.064012]  [<ffffffffa0049126>] virtnet_poll+0x4ba/0x5a4 [virtio_net]
[  452.064012]  [<ffffffff811d9417>] net_rx_action+0x73/0x184
[  452.064012]  [<ffffffffa01b2cc2>] ? l2tp_xmit_skb+0x27a/0x4ac [l2tp_core]
[  452.064012]  [<ffffffff810343b9>] __do_softirq+0xc3/0x1a8
[  452.064012]  [<ffffffff81013b56>] ? ack_APIC_irq+0x10/0x12
[  452.064012]  [<ffffffff81256559>] ? _raw_spin_lock+0xe/0x10
[  452.064012]  [<ffffffff8125e0ac>] call_softirq+0x1c/0x26
[  452.064012]  [<ffffffff81003587>] do_softirq+0x45/0x82
[  452.064012]  [<ffffffff81034667>] irq_exit+0x42/0x9c
[  452.064012]  [<ffffffff8125e146>] do_IRQ+0x8e/0xa5
[  452.064012]  [<ffffffff8125676e>] common_interrupt+0x6e/0x6e
[  452.064012]  <EOI>
[  452.064012]  [<ffffffff810b82a1>] ? kfree+0x8a/0xa3
[  452.064012]  [<ffffffffa01b2cc2>] ? l2tp_xmit_skb+0x27a/0x4ac [l2tp_core]
[  452.064012]  [<ffffffffa01b2c25>] ? l2tp_xmit_skb+0x1dd/0x4ac [l2tp_core]
[  452.064012]  [<ffffffffa01c2d36>] pppol2tp_sendmsg+0x15e/0x19c [l2tp_ppp]
[  452.064012]  [<ffffffff811c7872>] __sock_sendmsg_nosec+0x22/0x24
[  452.064012]  [<ffffffff811c83bd>] sock_sendmsg+0xa1/0xb6
[  452.064012]  [<ffffffff81254e88>] ? __schedule+0x5c1/0x616
[  452.064012]  [<ffffffff8103c7c6>] ? __dequeue_signal+0xb7/0x10c
[  452.064012]  [<ffffffff810bbd21>] ? fget_light+0x75/0x89
[  452.064012]  [<ffffffff811c8444>] ? sockfd_lookup_light+0x20/0x56
[  452.064012]  [<ffffffff811c9b34>] sys_sendto+0x10c/0x13b
[  452.064012]  [<ffffffff8125cac2>] system_call_fastpath+0x16/0x1b
[  452.064012] Code: 89 e5 72 0c 31 c0 48 81 ff 45 66 25 81 0f 92 c0 5d c3 55 b8 00 01 00 00 48 89 e5 f0 66 0f c1 07 0f b6 d4 38 d0 74 06 f3 90 8a 07 <eb> f6 5d c3 90 90 55 48 89 e5 9c 58 0f 1f 44 00 00 5d c3 55 48
[  452.064012] Call Trace:
[  452.064012]  <IRQ>  [<ffffffff81256559>] _raw_spin_lock+0xe/0x10
[  452.064012]  [<ffffffff8121c64a>] spin_lock+0x9/0xb
[  452.064012]  [<ffffffff8121ec7a>] udp_queue_rcv_skb+0x186/0x269
[  452.064012]  [<ffffffff8121f276>] __udp4_lib_rcv+0x297/0x4ae
[  452.064012]  [<ffffffff8121c178>] ? raw_rcv+0xe9/0xf0
[  452.064012]  [<ffffffff8121f4a7>] udp_rcv+0x1a/0x1c
[  452.064012]  [<ffffffff811fe385>] ip_local_deliver_finish+0x12b/0x1a5
[  452.064012]  [<ffffffff811fe54e>] ip_local_deliver+0x53/0x84
[  452.064012]  [<ffffffff811fe1d0>] ip_rcv_finish+0x2bc/0x2f3
[  452.064012]  [<ffffffff811fe78f>] ip_rcv+0x210/0x269
[  452.064012]  [<ffffffff8101911e>] ? kvm_clock_get_cycles+0x9/0xb
[  452.064012]  [<ffffffff811d88cd>] __netif_receive_skb+0x3a5/0x3f7
[  452.064012]  [<ffffffff811d8eba>] netif_receive_skb+0x57/0x5e
[  452.064012]  [<ffffffff811cf30f>] ? __netdev_alloc_skb+0x1f/0x3b
[  452.064012]  [<ffffffffa0049126>] virtnet_poll+0x4ba/0x5a4 [virtio_net]
[  452.064012]  [<ffffffff811d9417>] net_rx_action+0x73/0x184
[  452.064012]  [<ffffffffa01b2cc2>] ? l2tp_xmit_skb+0x27a/0x4ac [l2tp_core]
[  452.064012]  [<ffffffff810343b9>] __do_softirq+0xc3/0x1a8
[  452.064012]  [<ffffffff81013b56>] ? ack_APIC_irq+0x10/0x12
[  452.064012]  [<ffffffff81256559>] ? _raw_spin_lock+0xe/0x10
[  452.064012]  [<ffffffff8125e0ac>] call_softirq+0x1c/0x26
[  452.064012]  [<ffffffff81003587>] do_softirq+0x45/0x82
[  452.064012]  [<ffffffff81034667>] irq_exit+0x42/0x9c
[  452.064012]  [<ffffffff8125e146>] do_IRQ+0x8e/0xa5
[  452.064012]  [<ffffffff8125676e>] common_interrupt+0x6e/0x6e
[  452.064012]  <EOI>  [<ffffffff810b82a1>] ? kfree+0x8a/0xa3
[  452.064012]  [<ffffffffa01b2cc2>] ? l2tp_xmit_skb+0x27a/0x4ac [l2tp_core]
[  452.064012]  [<ffffffffa01b2c25>] ? l2tp_xmit_skb+0x1dd/0x4ac [l2tp_core]
[  452.064012]  [<ffffffffa01c2d36>] pppol2tp_sendmsg+0x15e/0x19c [l2tp_ppp]
[  452.064012]  [<ffffffff811c7872>] __sock_sendmsg_nosec+0x22/0x24
[  452.064012]  [<ffffffff811c83bd>] sock_sendmsg+0xa1/0xb6
[  452.064012]  [<ffffffff81254e88>] ? __schedule+0x5c1/0x616
[  452.064012]  [<ffffffff8103c7c6>] ? __dequeue_signal+0xb7/0x10c
[  452.064012]  [<ffffffff810bbd21>] ? fget_light+0x75/0x89
[  452.064012]  [<ffffffff811c8444>] ? sockfd_lookup_light+0x20/0x56
[  452.064012]  [<ffffffff811c9b34>] sys_sendto+0x10c/0x13b
[  452.064012]  [<ffffffff8125cac2>] system_call_fastpath+0x16/0x1b

Reported-by: François Cachereul <f.cachereul@alphalink.fr>
Tested-by: François Cachereul <f.cachereul@alphalink.fr>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

vti: get rid of nf mark rule in prerouting

[ Upstream commit 7263a5187f9e9de45fcb51349cf0e031142c19a1 ]

This patch fixes and improves the use of vti interfaces (while
lightly changing the way of configuring them).

Currently:

- it is necessary to identify and mark inbound IPsec
  packets destined to each vti interface, via netfilter rules in
  the mangle table at prerouting hook.

- the vti module cannot retrieve the right tunnel in input since
  commit b9959fd3: vti tunnels all have an i_key, but the tunnel lookup
  is done with flag TUNNEL_NO_KEY, so there no chance to retrieve them.

- the i_key is used by the outbound processing as a mark to lookup
  for the right SP and SA bundle.

This patch uses the o_key to store the vti mark (instead of i_key) and
enables:

- to avoid the need for previously marking the inbound skbuffs via a
  netfilter rule.
- to properly retrieve the right tunnel in input, only based on the IPsec
  packet outer addresses.
- to properly perform an inbound policy check (using the tunnel o_key
  as a mark).
- to properly perform an outbound SPD and SAD lookup (using the tunnel
  o_key as a mark).
- to keep the current mark of the skbuff. The skbuff mark is neither
  used nor changed by the vti interface. Only the vti interface o_key
  is used.

SAs have a wildcard mark.
SPs have a mark equal to the vti interface o_key.

The vti interface must be created as follows (i_key = 0, o_key = mark):

   ip link add vti1 mode vti local 1.1.1.1 remote 2.2.2.2 okey 1

The SPs attached to vti1 must be created as follows (mark = vti1 o_key):

   ip xfrm policy add dir out mark 1 tmpl src 1.1.1.1 dst 2.2.2.2 \
      proto esp mode tunnel
   ip xfrm policy add dir in  mark 1 tmpl src 2.2.2.2 dst 1.1.1.1 \
      proto esp mode tunnel

The SAs are created with the default wildcard mark. There is no
distinction between global vs. vti SAs. Just their addresses will
possibly link them to a vti interface:

   ip xfrm state add src 1.1.1.1 dst 2.2.2.2 proto esp spi 1000 mode tunnel \
                 enc "cbc(aes)" "azertyuiopqsdfgh"

   ip xfrm state add src 2.2.2.2 dst 1.1.1.1 proto esp spi 2000 mode tunnel \
                 enc "cbc(aes)" "sqbdhgqsdjqjsdfh"

To avoid matching "global" (not vti) SPs in vti interfaces, global SPs
should no use the default wildcard mark, but explicitly match mark 0.

To avoid a double SPD lookup in input and output (in global and vti SPDs),
the NOPOLICY and NOXFRM options should be set on the vti interfaces:

   echo 1 > /proc/sys/net/ipv4/conf/vti1/disable_policy
   echo 1 > /proc/sys/net/ipv4/conf/vti1/disable_xfrm

The outgoing traffic is steered to vti1 by a route via the vti interface:

   ip route add 192.168.0.0/16 dev vti1

The incoming IPsec traffic is steered to vti1 because its outer addresses
match the vti1 tunnel configuration.

Signed-off-by: Christophe Gouault <christophe.gouault@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: vlan: fix nlmsg size calculation in vlan_get_size()

[ Upstream commit c33a39c575068c2ea9bffb22fd6de2df19c74b89 ]

This patch fixes the calculation of the nlmsg size, by adding the missing
nla_total_size().

Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xen-netback: Don't destroy the netdev until the vif is shut down

[ upstream commit id: 279f438e36c0a70b23b86d2090aeec50155034a9 ]

Without this patch, if a frontend cycles through states Closing
and Closed (which Windows frontends need to do) then the netdev
will be destroyed and requires re-invocation of hotplug scripts
to restore state before the frontend can move to Connected. Thus
when udev is not in use the backend gets stuck in InitWait.

With this patch, the netdev is left alone whilst the backend is
still online and is only de-registered and freed just prior to
destroying the vif (which is also nicely symmetrical with the
netdev allocation and registration being done during probe) so
no re-invocation of hotplug scripts is required.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: secure_seq: Fix warning when CONFIG_IPV6 and CONFIG_INET are not selected

[ Upstream commit cb03db9d0e964568407fb08ea46cc2b6b7f67587 ]

net_secret() is only used when CONFIG_IPV6 or CONFIG_INET are selected.

Building a defconfig with both of these symbols unselected (Using the ARM
at91sam9rl_defconfig, for example) leads to the following build warning:

$ make at91sam9rl_defconfig
#
# configuration written to .config
#

$ make net/core/secure_seq.o
scripts/kconfig/conf --silentoldconfig Kconfig
  CHK     include/config/kernel.release
  CHK     include/generated/uapi/linux/version.h
  CHK     include/generated/utsrelease.h
make[1]: `include/generated/mach-types.h' is up to date.
  CALL    scripts/checksyscalls.sh
  CC      net/core/secure_seq.o
net/core/secure_seq.c:17:13: warning: 'net_secret_init' defined but not used [-Wunused-function]

Fix this warning by protecting the definition of net_secret() with these
symbols.

Reported-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

can: dev: fix nlmsg size calculation in can_get_size()

[ Upstream commit fe119a05f8ca481623a8d02efcc984332e612528 ]

This patch fixes the calculation of the nlmsg size, by adding the missing
nla_total_size().

Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipv4: fix ineffective source address selection

[ Upstream commit 0a7e22609067ff524fc7bbd45c6951dd08561667 ]

When sending out multicast messages, the source address in inet->mc_addr is
ignored and rewritten by an autoselected one. This is caused by a typo in
commit 813b3b5db831 ("ipv4: Use caller's on-stack flowi as-is in output
route lookups").

Signed-off-by: Jiri Benc <jbenc@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

proc connector: fix info leaks

[ Upstream commit e727ca82e0e9616ab4844301e6bae60ca7327682 ]

Initialize event_data for all possible message types to prevent leaking
kernel stack contents to userland (up to 20 bytes). Also set the flags
member of the connector message to 0 to prevent leaking two more stack
bytes this way.

Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: heap overflow in __audit_sockaddr()

[ Upstream commit 1661bf364ae9c506bc8795fef70d1532931be1e8 ]

We need to cap ->msg_namelen or it leads to a buffer overflow when we
to the memcpy() in __audit_sockaddr().  It requires CAP_AUDIT_CONTROL to
exploit this bug.

The call tree is:
___sys_recvmsg()
  move_addr_to_user()
    audit_sockaddr()
      __audit_sockaddr()

Reported-by: Jüri Aedla <juri.aedla@gmail.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: mv643xx_eth: fix orphaned statistics timer crash

[ Upstream commit f564412c935111c583b787bcc18157377b208e2e ]

The periodic statistics timer gets started at port _probe() time, but
is stopped on _stop() only. In a modular environment, this can cause
the timer to access already deallocated memory, if the module is unloaded
without starting the eth device. To fix this, we add the timer right
before the port is started, instead of at _probe() time.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: mv643xx_eth: update statistics timer from timer context only

[ Upstream commit 041b4ddb84989f06ff1df0ca869b950f1ee3cb1c ]

Each port driver installs a periodic timer to update port statistics
by calling mib_counters_update. As mib_counters_update is also called
from non-timer context, we should not reschedule the timer there but
rather move it to timer-only context.

Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

l2tp: Fix build warning with ipv6 disabled.

[ Upstream commit 8d8a51e26a6d415e1470759f2cf5f3ee3ee86196 ]

net/l2tp/l2tp_core.c: In function ‘l2tp_verify_udp_checksum’:
net/l2tp/l2tp_core.c:499:22: warning: unused variable ‘tunnel’ [-Wunused-variable]

Create a helper "l2tp_tunnel()" to facilitate this, and as a side
effect get rid of a bunch of unnecessary void pointer casts.

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

l2tp: fix kernel panic when using IPv4-mapped IPv6 addresses

[ Upstream commit e18503f41f9b12132c95d7c31ca6ee5155e44e5c ]

IPv4 mapped addresses cause kernel panic.
The patch juste check whether the IPv6 address is an IPv4 mapped
address. If so, use IPv4 API instead of IPv6.

[  940.026915] general protection fault: 0000 [#1]
[  940.026915] Modules linked in: l2tp_ppp l2tp_netlink l2tp_core pppox ppp_generic slhc loop psmouse
[  940.026915] CPU: 0 PID: 3184 Comm: memcheck-amd64- Not tainted 3.11.0+ #1
[  940.026915] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
[  940.026915] task: ffff880007130e20 ti: ffff88000737e000 task.ti: ffff88000737e000
[  940.026915] RIP: 0010:[<ffffffff81333780>]  [<ffffffff81333780>] ip6_xmit+0x276/0x326
[  940.026915] RSP: 0018:ffff88000737fd28  EFLAGS: 00010286
[  940.026915] RAX: c748521a75ceff48 RBX: ffff880000c30800 RCX: 0000000000000000
[  940.026915] RDX: ffff88000075cc4e RSI: 0000000000000028 RDI: ffff8800060e5a40
[  940.026915] RBP: ffff8800060e5a40 R08: 0000000000000000 R09: ffff88000075cc90
[  940.026915] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88000737fda0
[  940.026915] R13: 0000000000000000 R14: 0000000000002000 R15: ffff880005d3b580
[  940.026915] FS:  00007f163dc5e800(0000) GS:ffffffff81623000(0000) knlGS:0000000000000000
[  940.026915] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  940.026915] CR2: 00000004032dc940 CR3: 0000000005c25000 CR4: 00000000000006f0
[  940.026915] Stack:
[  940.026915]  ffff88000075cc4e ffffffff81694e90 ffff880000c30b38 0000000000000020
[  940.026915]  11000000523c4bac ffff88000737fdb4 0000000000000000 ffff880000c30800
[  940.026915]  ffff880005d3b580 ffff880000c30b38 ffff8800060e5a40 0000000000000020
[  940.026915] Call Trace:
[  940.026915]  [<ffffffff81356cc3>] ? inet6_csk_xmit+0xa4/0xc4
[  940.026915]  [<ffffffffa0038535>] ? l2tp_xmit_skb+0x503/0x55a [l2tp_core]
[  940.026915]  [<ffffffff812b8d3b>] ? pskb_expand_head+0x161/0x214
[  940.026915]  [<ffffffffa003e91d>] ? pppol2tp_xmit+0xf2/0x143 [l2tp_ppp]
[  940.026915]  [<ffffffffa00292e0>] ? ppp_channel_push+0x36/0x8b [ppp_generic]
[  940.026915]  [<ffffffffa00293fe>] ? ppp_write+0xaf/0xc5 [ppp_generic]
[  940.026915]  [<ffffffff8110ead4>] ? vfs_write+0xa2/0x106
[  940.026915]  [<ffffffff8110edd6>] ? SyS_write+0x56/0x8a
[  940.026915]  [<ffffffff81378ac0>] ? system_call_fastpath+0x16/0x1b
[  940.026915] Code: 00 49 8b 8f d8 00 00 00 66 83 7c 11 02 00 74 60 49
8b 47 58 48 83 e0 fe 48 8b 80 18 01 00 00 48 85 c0 74 13 48 8b 80 78 02
00 00 <48> ff 40 28 41 8b 57 68 48 01 50 30 48 8b 54 24 08 49 c7 c1 51
[  940.026915] RIP  [<ffffffff81333780>] ip6_xmit+0x276/0x326
[  940.026915]  RSP <ffff88000737fd28>
[  940.057945] ---[ end trace be8aba9a61c8b7f3 ]---
[  940.058583] Kernel panic - not syncing: Fatal exception in interrupt

Signed-off-by: François CACHEREUL <f.cachereul@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: do not call sock_put() on TIMEWAIT sockets

[ Upstream commit 80ad1d61e72d626e30ebe8529a0455e660ca4693 ]

commit 3ab5aee7fe84 ("net: Convert TCP & DCCP hash tables to use RCU /
hlist_nulls") incorrectly used sock_put() on TIMEWAIT sockets.

We should instead use inet_twsk_put()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: fix incorrect ca_state in tail loss probe

[ Upstream commit 031afe4990a7c9dbff41a3a742c44d3e740ea0a1 ]

On receiving an ACK that covers the loss probe sequence, TLP
immediately sets the congestion state to Open, even though some packets
are not recovered and retransmisssion are on the way.  The later ACks
may trigger a WARN_ON check in step D of tcp_fastretrans_alert(), e.g.,
https://bugzilla.redhat.com/show_bug.cgi?id=989251

The fix is to follow the similar procedure in recovery by calling
tcp_try_keep_open(). The sender switches to Open state if no packets
are retransmissted. Otherwise it goes to Disorder and let subsequent
ACKs move the state to Recovery or Open.

Reported-By: Michael Sterrett <michael@sterretts.net>
Tested-By: Dormando <dormando@rydia.net>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: do not forget FIN in tcp_shifted_skb()

[ Upstream commit 5e8a402f831dbe7ee831340a91439e46f0d38acd ]

Yuchung found following problem :

 There are bugs in the SACK processing code, merging part in
 tcp_shift_skb_data(), that incorrectly resets or ignores the sacked
 skbs FIN flag. When a receiver first SACK the FIN sequence, and later
 throw away ofo queue (e.g., sack-reneging), the sender will stop
 retransmitting the FIN flag, and hangs forever.

Following packetdrill test can be used to reproduce the bug.

$ cat sack-merge-bug.pkt
`sysctl -q net.ipv4.tcp_fack=0`

// Establish a connection and send 10 MSS.
0.000 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+.000 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+.000 bind(3, ..., ...) = 0
+.000 listen(3, 1) = 0

+.050 < S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+.000 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 6>
+.001 < . 1:1(0) ack 1 win 1024
+.000 accept(3, ..., ...) = 4

+.100 write(4, ..., 12000) = 12000
+.000 shutdown(4, SHUT_WR) = 0
+.000 > . 1:10001(10000) ack 1
+.050 < . 1:1(0) ack 2001 win 257
+.000 > FP. 10001:12001(2000) ack 1
+.050 < . 1:1(0) ack 2001 win 257 <sack 10001:11001,nop,nop>
+.050 < . 1:1(0) ack 2001 win 257 <sack 10001:12002,nop,nop>
// SACK reneg
+.050 < . 1:1(0) ack 12001 win 257
+0 %{ print "unacked: ",tcpi_unacked }%
+5 %{ print "" }%

First, a typo inverted left/right of one OR operation, then
code forgot to advance end_seq if the merged skb carried FIN.

Bug was added in 2.6.29 by commit 832d11c5cd076ab
("tcp: Try to restore large SKBs while SACK processing")

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: must unclone packets before mangling them

[ Upstream commit c52e2421f7368fd36cbe330d2cf41b10452e39a9 ]

TCP stack should make sure it owns skbs before mangling them.

We had various crashes using bnx2x, and it turned out gso_size
was cleared right before bnx2x driver was populating TC descriptor
of the _previous_ packet send. TCP stack can sometime retransmit
packets that are still in Qdisc.

Of course we could make bnx2x driver more robust (using
ACCESS_ONCE(shinfo->gso_size) for example), but the bug is TCP stack.

We have identified two points where skb_unclone() was needed.

This patch adds a WARN_ON_ONCE() to warn us if we missed another
fix of this kind.

Kudos to Neal for finding the root cause of this bug. Its visible
using small MSS.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: TSQ can use a dynamic limit

[ Upstream commit c9eeec26e32e087359160406f96e0949b3cc6f10 ]

When TCP Small Queues was added, we used a sysctl to limit amount of
packets queues on Qdisc/device queues for a given TCP flow.

Problem is this limit is either too big for low rates, or too small
for high rates.

Now TCP stack has rate estimation in sk->sk_pacing_rate, and TSO
auto sizing, it can better control number of packets in Qdisc/device
queues.

New limit is two packets or at least 1 to 2 ms worth of packets.

Low rates flows benefit from this patch by having even smaller
number of packets in queues, allowing for faster recovery,
better RTT estimations.

High rates flows benefit from this patch by allowing more than 2 packets
in flight as we had reports this was a limiting factor to reach line
rate. [ In particular if TX completion is delayed because of coalescing
parameters ]

Example for a single flow on 10Gbp link controlled by FQ/pacing

14 packets in flight instead of 2

$ tc -s -d qd
qdisc fq 8001: dev eth0 root refcnt 32 limit 10000p flow_limit 100p
buckets 1024 quantum 3028 initial_quantum 15140
 Sent 1168459366606 bytes 771822841 pkt (dropped 0, overlimits 0
requeues 6822476)
 rate 9346Mbit 771713pps backlog 953820b 14p requeues 6822476
  2047 flow, 2046 inactive, 1 throttled, delay 15673 ns
  2372 gc, 0 highprio, 0 retrans, 9739249 throttled, 0 flows_plimit

Note that sk_pacing_rate is currently set to twice the actual rate, but
this might be refined in the future when a flow is in congestion
avoidance.

Additional change : skb->destructor should be set to tcp_wfree().

A future patch (for linux 3.13+) might remove tcp_limit_output_bytes

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: TSO packets automatic sizing

[ Upstream commits 6d36824e730f247b602c90e8715a792003e3c5a7,
  02cf4ebd82ff0ac7254b88e466820a290ed8289a, and parts of
  7eec4174ff29cd42f2acfae8112f51c228545d40 ]

After hearing many people over past years complaining against TSO being
bursty or even buggy, we are proud to present automatic sizing of TSO
packets.

One part of the problem is that tcp_tso_should_defer() uses an heuristic
relying on upcoming ACKS instead of a timer, but more generally, having
big TSO packets makes little sense for low rates, as it tends to create
micro bursts on the network, and general consensus is to reduce the
buffering amount.

This patch introduces a per socket sk_pacing_rate, that approximates
the current sending rate, and allows us to size the TSO packets so
that we try to send one packet every ms.

This field could be set by other transports.

Patch has no impact for high speed flows, where having large TSO packets
makes sense to reach line rate.

For other flows, this helps better packet scheduling and ACK clocking.

This patch increases performance of TCP flows in lossy environments.

A new sysctl (tcp_min_tso_segs) is added, to specify the
minimal size of a TSO packet (default being 2).

A follow-up patch will provide a new packet scheduler (FQ), using
sk_pacing_rate as an input to perform optional per flow pacing.

This explains why we chose to set sk_pacing_rate to twice the current
rate, allowing 'slow start' ramp up.

sk_pacing_rate = 2 * cwnd * mss / srtt

v2: Neal Cardwell reported a suspect deferring of last two segments on
initial write of 10 MSS, I had to change tcp_tso_should_defer() to take
into account tp->xmit_size_goal_segs

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Van Jacobson <vanj@google.com>
Cc: Tom Herbert <therbert@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 3.10.17

x86: avoid remapping data in parse_setup_data()

commit 30e46b574a1db7d14404e52dca8e1aa5f5155fd2 upstream.

Type SETUP_PCI, added by setup_efi_pci(), may advertise a ROM size
larger than early_memremap() is able to handle, which is currently
limited to 256kB. If this occurs it leads to a NULL dereference in
parse_setup_data().

To avoid this, remap the setup_data header and allow parsing functions
for individual types to handle their own data remapping.

Signed-off-by: Linn Crosetto <linn@hp.com>
Link: http://lkml.kernel.org/r/1376430401-67445-1-git-send-email-linn@hp.com
Acked-by: Yinghai Lu <yinghai@kernel.org>
Reviewed-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: prevent race with rmid in msgsnd,msgrcv

commit 4271b05a227dc6175b66c3d9941aeab09048aeb2 upstream.

This fixes a race in both msgrcv() and msgsnd() between finding the msg
and actually dealing with the queue, as another thread can delete shmid
underneath us if we are preempted before acquiring the
kern_ipc_perm.lock.

Manfred illustrates this nicely:

Assume a preemptible kernel that is preempted just after

    msq = msq_obtain_object_check(ns, msqid)

in do_msgrcv().  The only lock that is held is rcu_read_lock().

Now the other thread processes IPC_RMID.  When the first task is
resumed, then it will happily wait for messages on a deleted queue.

Fix this by checking for if the queue has been deleted after taking the
lock.

Signed-off-by: Davidlohr Bueso <davidlohr@hp.com>
Reported-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
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>

ipc/sem.c: update sem_otime for all operations

commit 0e8c665699e953fa58dc1b0b0d09e5dce7343cc7 upstream.

In commit 0a2b9d4c7967 ("ipc/sem.c: move wake_up_process out of the
spinlock section"), the update of semaphore's sem_otime(last semop time)
was moved to one central position (do_smart_update).

But since do_smart_update() is only called for operations that modify
the array, this means that wait-for-zero semops do not update sem_otime
anymore.

The fix is simple:
Non-alter operations must update sem_otime.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Reported-by: Jia He <jiakernel@gmail.com>
Tested-by: Jia He <jiakernel@gmail.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Mike Galbraith <efault@gmx.de>
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>

ipc/sem.c: synchronize the proc interface

commit d8c633766ad88527f25d9f81a5c2f083d78a2b39 upstream.

The proc interface is not aware of sem_lock(), it instead calls
ipc_lock_object() directly.  This means that simple semop() operations
can run in parallel with the proc interface.  Right now, this is
uncritical, because the implementation doesn't do anything that requires
a proper synchronization.

But it is dangerous and therefore should be fixed.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Rik van Riel <riel@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>

ipc/sem.c: optimize sem_lock()

commit 6d07b68ce16ae9535955ba2059dedba5309c3ca1 upstream.

Operations that need access to the whole array must guarantee that there
are no simple operations ongoing.  Right now this is achieved by
spin_unlock_wait(sem->lock) on all semaphores.

If complex_count is nonzero, then this spin_unlock_wait() is not
necessary, because it was already performed in the past by the thread
that increased complex_count and even though sem_perm.lock was dropped
inbetween, no simple operation could have started, because simple
operations cannot start when complex_count is non-zero.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: Rik van Riel <riel@redhat.com>
Reviewed-by: Davidlohr Bueso <davidlohr@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c: fix race in sem_lock()

commit 5e9d527591421ccdb16acb8c23662231135d8686 upstream.

The exclusion of complex operations in sem_lock() is insufficient: after
acquiring the per-semaphore lock, a simple op must first check that
sem_perm.lock is not locked and only after that test check
complex_count.  The current code does it the other way around - and that
creates a race.  Details are below.

The patch is a complete rewrite of sem_lock(), based in part on the code
from Mike Galbraith.  It removes all gotos and all loops and thus the
risk of livelocks.

I have tested the patch (together with the next one) on my i3 laptop and
it didn't cause any problems.

The bug is probably also present in 3.10 and 3.11, but for these kernels
it might be simpler just to move the test of sma->complex_count after
the spin_is_locked() test.

Details of the bug:

Assume:
 - sma->complex_count = 0.
 - Thread 1: semtimedop(complex op that must sleep)
 - Thread 2: semtimedop(simple op).

Pseudo-Trace:

Thread 1: sem_lock(): acquire sem_perm.lock
Thread 1: sem_lock(): check for ongoing simple ops
Nothing ongoing, thread 2 is still before sem_lock().
Thread 1: try_atomic_semop()
<<< preempted.

Thread 2: sem_lock():
        static inline int sem_lock(struct sem_array *sma, struct sembuf *sops,
                                      int nsops)
        {
                int locknum;
         again:
                if (nsops == 1 && !sma->complex_count) {
                        struct sem *sem = sma->sem_base + sops->sem_num;

                        /* Lock just the semaphore we are interested in. */
                        spin_lock(&sem->lock);

                        /*
                         * If sma->complex_count was set while we were spinning,
                         * we may need to look at things we did not lock here.
                         */
                        if (unlikely(sma->complex_count)) {
                                spin_unlock(&sem->lock);
                                goto lock_array;
                        }
        <<<<<<<<<
<<< complex_count is still 0.
<<<
        <<< Here it is preempted
        <<<<<<<<<

Thread 1: try_atomic_semop() returns, notices that it must sleep.
Thread 1: increases sma->complex_count.
Thread 1: drops sem_perm.lock
Thread 2:
                /*
                 * Another process is holding the global lock on the
                 * sem_array; we cannot enter our critical section,
                 * but have to wait for the global lock to be released.
                 */
                if (unlikely(spin_is_locked(&sma->sem_perm.lock))) {
                        spin_unlock(&sem->lock);
                        spin_unlock_wait(&sma->sem_perm.lock);
                        goto again;
                }
<<< sem_perm.lock already dropped, thus no "goto again;"

                locknum = sops->sem_num;

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Mike Galbraith <bitbucket@online.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: fix race with LSMs

commit 53dad6d3a8e5ac1af8bacc6ac2134ae1a8b085f1 upstream.

Currently, IPC mechanisms do security and auditing related checks under
RCU.  However, since security modules can free the security structure,
for example, through selinux_[sem,msg_queue,shm]_free_security(), we can
race if the structure is freed before other tasks are done with it,
creating a use-after-free condition.  Manfred illustrates this nicely,
for instance with shared mem and selinux:

 -> do_shmat calls rcu_read_lock()
 -> do_shmat calls shm_object_check().
     Checks that the object is still valid - but doesn't acquire any locks.
     Then it returns.
 -> do_shmat calls security_shm_shmat (e.g. selinux_shm_shmat)
 -> selinux_shm_shmat calls ipc_has_perm()
 -> ipc_has_perm accesses ipc_perms->security

shm_close()
 -> shm_close acquires rw_mutex & shm_lock
 -> shm_close calls shm_destroy
 -> shm_destroy calls security_shm_free (e.g. selinux_shm_free_security)
 -> selinux_shm_free_security calls ipc_free_security(&shp->shm_perm)
 -> ipc_free_security calls kfree(ipc_perms->security)

This patch delays the freeing of the security structures after all RCU
readers are done.  Furthermore it aligns the security life cycle with
that of the rest of IPC - freeing them based on the reference counter.
For situations where we need not free security, the current behavior is
kept.  Linus states:

 "... the old behavior was suspect for another reason too: having the
  security blob go away from under a user sounds like it could cause
  various other problems anyway, so I think the old code was at least
  _prone_ to bugs even if it didn't have catastrophic behavior."

I have tested this patch with IPC testcases from LTP on both my
quad-core laptop and on a 64 core NUMA server.  In both cases selinux is
enabled, and tests pass for both voluntary and forced preemption models.
While the mentioned races are theoretical (at least no one as reported
them), I wanted to make sure that this new logic doesn't break anything
we weren't aware of.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Davidlohr Bueso <davidlohr@hp.com>
Acked-by: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: drop ipc_lock_check

commit 20b8875abcf2daa1dda5cf70bd6369df5e85d4c1 upstream.

No remaining users, we now use ipc_obtain_object_check().

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc, shm: drop shm_lock_check

commit 7a25dd9e042b2b94202a67e5551112f4ac87285a upstream.

This function was replaced by a the lockless shm_obtain_object_check(),
and no longer has any users.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: drop ipc_lock_by_ptr

commit 32a2750010981216fb788c5190fb0e646abfab30 upstream.

After previous cleanups and optimizations, this function is no longer
heavily used and we don't have a good reason to keep it.  Update the few
remaining callers and get rid of it.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc, shm: guard against non-existant vma in shmdt(2)

commit 530fcd16d87cd2417c472a581ba5a1e501556c86 upstream.

When !CONFIG_MMU there's a chance we can derefence a NULL pointer when the
VM area isn't found - check the return value of find_vma().

Also, remove the redundant -EINVAL return: retval is set to the proper
return code and *only* changed to 0, when we actually unmap the segments.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: document general ipc locking scheme

commit 05603c44a7627793219b0bd9a7b236099dc9cd9d upstream.

As suggested by Andrew, add a generic initial locking scheme used
throughout all sysv ipc mechanisms.  Documenting the ids rwsem, how rcu
can be enough to do the initial checks and when to actually acquire the
kern_ipc_perm.lock spinlock.

I found that adding it to util.c was generic enough.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: drop msg_unlock

commit 4718787d1f626f45ddb239912bc07266b9880044 upstream.

There is only one user left, drop this function and just call
ipc_unlock_object() and rcu_read_unlock().

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: rename ids->rw_mutex

commit d9a605e40b1376eb02b067d7690580255a0df68f upstream.

Since in some situations the lock can be shared for readers, we shouldn't
be calling it a mutex, rename it to rwsem.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: shorten critical region for shmat

commit c2c737a0461e61a34676bd0bd1bc1a70a1b4e396 upstream.

Similar to other system calls, acquire the kern_ipc_perm lock after doing
the initial permission and security checks.

[sasha.levin@oracle.com: dont leave do_shmat with rcu lock held]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: cleanup do_shmat pasta

commit f42569b1388b1408b574a5e93a23a663647d4181 upstream.

Clean up some of the messy do_shmat() spaghetti code, getting rid of
out_free and out_put_dentry labels.  This makes shortening the critical
region of this function in the next patch a little easier to do and read.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: shorten critical region for shmctl

commit 2caacaa82a51b78fc0c800e206473874094287ed upstream.

With the *_INFO, *_STAT, IPC_RMID and IPC_SET commands already optimized,
deal with the remaining SHM_LOCK and SHM_UNLOCK commands.  Take the
shm_perm lock after doing the initial auditing and security checks.  The
rest of the logic remains unchanged.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: make shmctl_nolock lockless

commit c97cb9ccab8c85428ec21eff690642ad2ce1fa8a upstream.

While the INFO cmd doesn't take the ipc lock, the STAT commands do acquire
it unnecessarily.  We can do the permissions and security checks only
holding the rcu lock.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: introduce shmctl_nolock

commit 68eccc1dc345539d589ae78ee43b835c1a06a134 upstream.

Similar to semctl and msgctl, when calling msgctl, the *_INFO and *_STAT
commands can be performed without acquiring the ipc object.

Add a shmctl_nolock() function and move the logic of *_INFO and *_STAT out
of msgctl().  Since we are just moving functionality, this change still
takes the lock and it will be properly lockless in the next patch.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: drop ipcctl_pre_down

commit 3b1c4ad37741e53804ffe0a30dd01e08b2ab6241 upstream.

Now that sem, msgque and shm, through *_down(), all use the lockless
variant of ipcctl_pre_down(), go ahead and delete it.

[akpm@linux-foundation.org: fix function name in kerneldoc, cleanups]
Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: shorten critical region in shmctl_down

commit 79ccf0f8c8e04e8b9eda6645ba0f63b0915a3075 upstream.

Instead of holding the ipc lock for the entire function, use the
ipcctl_pre_down_nolock and only acquire the lock for specific commands:
RMID and SET.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,shm: introduce lockless functions to obtain the ipc object

commit 8b8d52ac382b17a19906b930cd69e2edb0aca8ba upstream.

This is the third and final patchset that deals with reducing the amount
of contention we impose on the ipc lock (kern_ipc_perm.lock).  These
changes mostly deal with shared memory, previous work has already been
done for semaphores and message queues:

  http://lkml.org/lkml/2013/3/20/546 (sems)
  http://lkml.org/lkml/2013/5/15/584 (mqueues)

With these patches applied, a custom shm microbenchmark stressing shmctl
doing IPC_STAT with 4 threads a million times, reduces the execution
time by 50%.  A similar run, this time with IPC_SET, reduces the
execution time from 3 mins and 35 secs to 27 seconds.

Patches 1-8: replaces blindly taking the ipc lock for a smarter
combination of rcu and ipc_obtain_object, only acquiring the spinlock
when updating.

Patch 9: renames the ids rw_mutex to rwsem, which is what it already was.

Patch 10: is a trivial mqueue leftover cleanup

Patch 11: adds a brief lock scheme description, requested by Andrew.

This patch:

Add shm_obtain_object() and shm_obtain_object_check(), which will allow us
to get the ipc object without acquiring the lock.  Just as with other
forms of ipc, these functions are basically wrappers around
ipc_obtain_object*().

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/msg.c: Fix lost wakeup in msgsnd().

commit bebcb928c820d0ee83aca4b192adc195e43e66a2 upstream.

The check if the queue is full and adding current to the wait queue of
pending msgsnd() operations (ss_add()) must be atomic.

Otherwise:
 - the thread that performs msgsnd() finds a full queue and decides to
   sleep.
 - the thread that performs msgrcv() first reads all messages from the
   queue and then sleeps, because the queue is empty.
 - the msgrcv() calls do not perform any wakeups, because the msgsnd()
   task has not yet called ss_add().
 - then the msgsnd()-thread first calls ss_add() and then sleeps.

Net result: msgsnd() and msgrcv() both sleep forever.

Observed with msgctl08 from ltp with a preemptible kernel.

Fix: Call ipc_lock_object() before performing the check.

The patch also moves security_msg_queue_msgsnd() under ipc_lock_object:
 - msgctl(IPC_SET) explicitely mentions that it tries to expunge any
   pending operations that are not allowed anymore with the new
   permissions.  If security_msg_queue_msgsnd() is called without locks,
   then there might be races.
 - it makes the patch much simpler.

Reported-and-tested-by: Vineet Gupta <Vineet.Gupta1@synopsys.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c: rename try_atomic_semop() to perform_atomic_semop(), docu update

commit 758a6ba39ef6df4cdc615e5edd7bd86eab81a5f7 upstream.

Cleanup: Some minor points that I noticed while writing the previous
patches

1) The name try_atomic_semop() is misleading: The function performs the
   operation (if it is possible).

2) Some documentation updates.

No real code change, a rename and documentation changes.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c: replace shared sem_otime with per-semaphore value

commit d12e1e50e47e0900dbbf52237b7e171f4f15ea1e upstream.

sem_otime contains the time of the last semaphore operation that
completed successfully.  Every operation updates this value, thus access
from multiple cpus can cause thrashing.

Therefore the patch replaces the variable with a per-semaphore variable.
The per-array sem_otime is only calculated when required.

No performance improvement on a single-socket i3 - only important for
larger systems.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c: always use only one queue for alter operations

commit f269f40ad5aeee229ed70044926f44318abe41ef upstream.

There are two places that can contain alter operations:
 - the global queue: sma->pending_alter
 - the per-semaphore queues: sma->sem_base[].pending_alter.

Since one of the queues must be processed first, this causes an odd
priorization of the wakeups: complex operations have priority over
simple ops.

The patch restores the behavior of linux <=3.0.9: The longest waiting
operation has the highest priority.

This is done by using only one queue:
 - if there are complex ops, then sma->pending_alter is used.
 - otherwise, the per-semaphore queues are used.

As a side effect, do_smart_update_queue() becomes much simpler: no more
goto logic.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem: separate wait-for-zero and alter tasks into seperate queues

commit 1a82e9e1d0f1b45f47a97c9e2349020536ff8987 upstream.

Introduce separate queues for operations that do not modify the
semaphore values.  Advantages:

 - Simpler logic in check_restart().
 - Faster update_queue(): Right now, all wait-for-zero operations are
   always tested, even if the semaphore value is not 0.
 - wait-for-zero gets again priority, as in linux <=3.0.9

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/sem.c: cacheline align the semaphore structures

commit f5c936c0f267ec58641451cf8b8d39b4c207ee4d upstream.

As now each semaphore has its own spinlock and parallel operations are
possible, give each semaphore its own cacheline.

On a i3 laptop, this gives up to 28% better performance:

  #semscale 10 | grep "interleave 2"
  - before:
  Cpus 1, interleave 2 delay 0: 36109234 in 10 secs
  Cpus 2, interleave 2 delay 0: 55276317 in 10 secs
  Cpus 3, interleave 2 delay 0: 62411025 in 10 secs
  Cpus 4, interleave 2 delay 0: 81963928 in 10 secs

  -after:
  Cpus 1, interleave 2 delay 0: 35527306 in 10 secs
  Cpus 2, interleave 2 delay 0: 70922909 in 10 secs <<< + 28%
  Cpus 3, interleave 2 delay 0: 80518538 in 10 secs
  Cpus 4, interleave 2 delay 0: 89115148 in 10 secs <<< + 8.7%

i3, with 2 cores and with hyperthreading enabled.  Interleave 2 in order
use first the full cores.  HT partially hides the delay from cacheline
trashing, thus the improvement is "only" 8.7% if 4 threads are running.

Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc/util.c, ipc_rcu_alloc: cacheline align allocation

commit 196aa0132fc7261f34b10ae1bfb44abc1bc69b3c upstream.

Enforce that ipc_rcu_alloc returns a cacheline aligned pointer on SMP.

Rationale:

The SysV sem code tries to move the main spinlock into a seperate
cacheline (____cacheline_aligned_in_smp).  This works only if
ipc_rcu_alloc returns cacheline aligned pointers.  vmalloc and kmalloc
return cacheline algined pointers, the implementation of ipc_rcu_alloc
breaks that.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Manfred Spraul <manfred@colorfullife.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc: remove unused functions

commit 9ad66ae65fc8d3e7e3344310fb0aa835910264fe upstream.

We can now drop the msg_lock and msg_lock_check functions along with a
bogus comment introduced previously in semctl_down.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: shorten critical region in msgrcv

commit 41a0d523d0f626e9da0dc01de47f1b89058033cf upstream.

do_msgrcv() is the last msg queue function that abuses the ipc lock Take
it only when needed when actually updating msq.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: shorten critical region in msgsnd

commit 3dd1f784ed6603d7ab1043e51e6371235edf2313 upstream.

do_msgsnd() is another function that does too many things with the ipc
object lock acquired.  Take it only when needed when actually updating
msq.

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: make msgctl_nolock lockless

commit ac0ba20ea6f2201a1589d6dc26ad1a4f0f967bb8 upstream.

While the INFO cmd doesn't take the ipc lock, the STAT commands do
acquire it unnecessarily.  We can do the permissions and security checks
only holding the rcu lock.

This function now mimics semctl_nolock().

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: introduce lockless functions to obtain the ipc object

commit a5001a0d9768568de5d613c3b3a5b9c7721299da upstream.

Add msq_obtain_object() and msq_obtain_object_check(), which will allow
us to get the ipc object without acquiring the lock.  Just as with
semaphores, these functions are basically wrappers around
ipc_obtain_object*().

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipc,msg: introduce msgctl_nolock

commit 2cafed30f150f7314f98717b372df8173516cae0 upstream.

Similar to semctl, when calling msgctl, the *_INFO and *_STAT commands
can be performed without acquiring the ipc object.

Add a msgctl_nolock() function and move the logic of *_INFO and *_STAT
out of msgctl().  This change still takes the lock and it will be
properly lockless in the next patch

Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>