USB: Serial: Add PID(0xF7C0) to FTDI SIO driver for a zeitcontrol-device

commit ce7e9065958191e6b7ca49d7ed0e1099c486d198 upstream.

Here is a patch for a new PID (zeitcontrol-device mifare-reader FT232BL(like FT232BM but lead free)).

Signed-off-by: Artur Zimmer <artur128@3dzimmer.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: Serial: Add device ID for Sierra Wireless MC8305

commit 2f1def2695c223b2aa325e5e47d0d64200a45d23 upstream.

A new device ID pair is added for Sierra Wireless MC8305.

Signed-off-by: Florian Echtler <floe@butterbrot.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

usb/isp1760: Added missing call to usb_hcd_check_unlink_urb() during unlink

commit 17d3e145a4ad680b3d1b1c30d0696a5bbb2b65c4 upstream.

Signed-off-by: Arvid Brodin <arvid.brodin@enea.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: EHCI: Fix test mode sequence

commit 77636c86a600b83de01719efad83567e46d7e8ce upstream.

The sequence to put port in test mode is not complete.
According EHCI specification all enabled ports must be
put in suspend.

Signed-off-by: Boris Todorov <boris.st.todorov@gmail.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

rtl8150: rtl8150_disconnect(...) does not need tasklet_disable(...)

commit c2e2a313ff8fdc25cedef5e63da712a6a0d35dfe upstream.

Executing cmd 'rmmod rtl8150' does not return(if your device connects
to host), the root cause is tasklet_disable() causes tasklet_kill()
block, remove it from rtl8150_disconnect().

Signed-off-by: Huajun Li <huajun.li.lee@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

enic: Bug Fix: Fix hardware transmit queue indexing in enic_poll_controller

commit b880a954b9e2585ce325aedd76e4741880cab180 upstream.

Signed-off-by: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Danny Guo <dannguo@cisco.com>
Signed-off-by: Vasanthy Kolluri <vkolluri@cisco.com>
Signed-off-by: Roopa Prabhu <roprabhu@cisco.com>
Signed-off-by: David Wang <dwang2@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Cc: Chun-Yi Lee <jlee@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ext4: fix race in xattr block allocation path

commit 6d6a435190bdf2e04c9465cde5bdc3ac68cf11a4 upstream.

Ceph users reported that when using Ceph on ext4, the filesystem
would often become corrupted, containing inodes with incorrect
i_blocks counters.

I managed to reproduce this with a very hacked-up "streamtest"
binary from the Ceph tree.

Ceph is doing a lot of xattr writes, to out-of-inode blocks.
There is also another thread which does sync_file_range and close,
of the same files.  The problem appears to happen due to this race:

sync/flush thread               xattr-set thread
-----------------               ----------------

do_writepages                   ext4_xattr_set
ext4_da_writepages              ext4_xattr_set_handle
mpage_da_map_blocks             ext4_xattr_block_set
        set DELALLOC_RESERVE
                                ext4_new_meta_blocks
                                        ext4_mb_new_blocks
                                                if (!i_delalloc_reserved_flag)
                                                        vfs_dq_alloc_block
ext4_get_blocks
down_write(i_data_sem)
        set i_delalloc_reserved_flag
...
up_write(i_data_sem)
                                        if (i_delalloc_reserved_flag)
                                                vfs_dq_alloc_block_nofail

In other words, the sync/flush thread pops in and sets
i_delalloc_reserved_flag on the inode, which makes the xattr thread
think that it's in a delalloc path in ext4_new_meta_blocks(),
and add the block for a second time, after already having added
it once in the !i_delalloc_reserved_flag case in ext4_mb_new_blocks

The real problem is that we shouldn't be using the DELALLOC_RESERVED
state flag, and instead we should be passing
EXT4_GET_BLOCKS_DELALLOC_RESERVE down to ext4_map_blocks() instead of
using an inode state flag.  We'll fix this for now with using
i_data_sem to prevent this race, but this is really not the right way
to fix things.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ext4: call ext4_handle_dirty_metadata with correct inode in ext4_dx_add_entry

commit 5930ea643805feb50a2f8383ae12eb6f10935e49 upstream.

ext4_dx_add_entry manipulates bh2 and frames[0].bh, which are two buffer_heads
that point to directory blocks assigned to the directory inode.  However, the
function calls ext4_handle_dirty_metadata with the inode of the file that's
being added to the directory, not the directory inode itself.  Therefore,
correct the code to dirty the directory buffers with the directory inode, not
the file inode.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ext4: ext4_mkdir should dirty dir_block with newly created directory inode

commit f9287c1f2d329f4d78a3bbc9cf0db0ebae6f146a upstream.

ext4_mkdir calls ext4_handle_dirty_metadata with dir_block and the inode "dir".
Unfortunately, dir_block belongs to the newly created directory (which is
"inode"), not the parent directory (which is "dir").  Fix the incorrect
association.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ext4: ext4_rename should dirty dir_bh with the correct directory

commit bcaa992975041e40449be8c010c26192b8c8b409 upstream.

When ext4_rename performs a directory rename (move), dir_bh is a
buffer that is modified to update the '..' link in the directory being
moved (old_inode).  However, ext4_handle_dirty_metadata is called with
the old parent directory inode (old_dir) and dir_bh, which is
incorrect because dir_bh does not belong to the parent inode.  Fix
this error.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ext2,ext3,ext4: don't inherit APPEND_FL or IMMUTABLE_FL for new inodes

commit 1cd9f0976aa4606db8d6e3dc3edd0aca8019372a upstream.

This doesn't make much sense, and it exposes a bug in the kernel where
attempts to create a new file in an append-only directory using
O_CREAT will fail (but still leave a zero-length file).  This was
discovered when xfstests #79 was generalized so it could run on all
file systems.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drivers/power/ds2780_battery.c: fix deadlock upon insertion and removal

commit 0e053fcbbbc4d945247cb32cad2767b483cb65f8 upstream.

Fixes the deadlock when inserting and removing the ds2780.

Signed-off-by: Clifton Barnes <cabarnes@indesign-llc.com>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
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@suse.de>

drivers/power/ds2780_battery.c: add a nolock function to w1 interface

commit 9fe678fa2feb4aaac0b4220de63e1b7f8ccebae6 upstream.

Adds a nolock function to the w1 interface to avoid locking the
mutex if needed.

Signed-off-by: Clifton Barnes <cabarnes@indesign-llc.com>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
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@suse.de>

drivers/power/ds2780_battery.c: create central point for calling w1 interface

commit 853eee72f74f449797f0500ea19fc1bf497428d8 upstream.

Simply creates one point to call the w1 interface.

Signed-off-by: Clifton Barnes <cabarnes@indesign-llc.com>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
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@suse.de>

hwspinlock/core: use a mutex to protect the radix tree

commit 93b465c2e186d96fb90012ba0f9372eb9952e732 upstream.

Since we're using non-atomic radix tree allocations, we
should be protecting the tree using a mutex and not a
spinlock.

Non-atomic allocations and process context locking is good enough,
as the tree is manipulated only when locks are registered/
unregistered/requested/freed.

The locks themselves are still protected by spinlocks of course,
and mutexes are not involved in the locking/unlocking paths.

Signed-off-by: Juan Gutierrez <jgutierrez@ti.com>
[ohad@wizery.com: rewrite the commit log, #include mutex.h, add minor
commentary]
[ohad@wizery.com: update register/unregister parts in hwspinlock.txt]
Signed-off-by: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drivers/net/rionet.c: fix ethernet address macros for LE platforms

commit e0c87bd95e8dad455c23bc56513af8dcb1737e55 upstream.

Modify Ethernet addess macros to be compatible with BE/LE platforms

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Chul Kim <chul.kim@idt.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.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@suse.de>

iwlagn: do not use interruptible waits

Upstream commit effd4d9aece9184f526e6556786a94d335e38b71.

Since the dawn of its time, iwlwifi has used
interruptible waits to wait for synchronous
commands and firmware loading.

This leads to "interesting" bugs, because it
can't actually handle the interruptions; for
example when a command sending is interrupted
it will assume the command completed fully,
and then leave it pending, which leads to all
kinds of trouble when the command finishes
later.

Since there's no easy way to gracefully deal
with interruptions, fix the driver to not use
interruptible waits.

This at least fixes the error
iwlagn 0000:02:00.0: Error: Response NULL in  'REPLY_SCAN_ABORT_CMD'

I have seen in P2P testing, but it is likely
that there are other errors caused by this.

Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

vfs: show O_CLOEXE bit properly in /proc/<pid>/fdinfo/<fd> files

commit 1117f72ea0217ba0cc19f05adbbd8b9a397f5ab7 upstream.

The CLOEXE bit is magical, and for performance (and semantic) reasons we
don't actually maintain it in the file descriptor itself, but in a
separate bit array.  Which means that when we show f_flags, the CLOEXE
status is shown incorrectly: we show the status not as it is now, but as
it was when the file was opened.

Fix that by looking up the bit properly in the 'fdt->close_on_exec' bit
array.

Uli needs this in order to re-implement the pfiles program:

  "For normal file descriptors (not sockets) this was the last piece of
   information which wasn't available.  This is all part of my 'give
   Solaris users no reason to not switch' effort.  I intend to offer the
   code to the util-linux-ng maintainers."

Requested-by: Ulrich Drepper <drepper@akkadia.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

binfmt_elf: fix PIE execution with randomization disabled

commit a3defbe5c337dbc6da911f8cc49ae3cc3b49b453 upstream.

The case of address space randomization being disabled in runtime through
randomize_va_space sysctl is not treated properly in load_elf_binary(),
resulting in SIGKILL coming at exec() time for certain PIE-linked binaries
in case the randomization has been disabled at runtime prior to calling
exec().

Handle the randomize_va_space == 0 case the same way as if we were not
supporting .text randomization at all.

Based on original patch by H.J. Lu and Josh Boyer.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: H.J. Lu <hongjiu.lu@intel.com>
Cc: <stable@kernel.org>
Tested-by: Josh Boyer <jwboyer@redhat.com>
Acked-by: Nicolas Pitre <nicolas.pitre@linaro.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@suse.de>

mm: thp: tail page refcounting fix

commit 70b50f94f1644e2aa7cb374819cfd93f3c28d725 upstream.

Michel while working on the working set estimation code, noticed that
calling get_page_unless_zero() on a random pfn_to_page(random_pfn)
wasn't safe, if the pfn ended up being a tail page of a transparent
hugepage under splitting by __split_huge_page_refcount().

He then found the problem could also theoretically materialize with
page_cache_get_speculative() during the speculative radix tree lookups
that uses get_page_unless_zero() in SMP if the radix tree page is freed
and reallocated and get_user_pages is called on it before
page_cache_get_speculative has a chance to call get_page_unless_zero().

So the best way to fix the problem is to keep page_tail->_count zero at
all times.  This will guarantee that get_page_unless_zero() can never
succeed on any tail page.  page_tail->_mapcount is guaranteed zero and
is unused for all tail pages of a compound page, so we can simply
account the tail page references there and transfer them to
tail_page->_count in __split_huge_page_refcount() (in addition to the
head_page->_mapcount).

While debugging this s/_count/_mapcount/ change I also noticed get_page is
called by direct-io.c on pages returned by get_user_pages.  That wasn't
entirely safe because the two atomic_inc in get_page weren't atomic.  As
opposed to other get_user_page users like secondary-MMU page fault to
establish the shadow pagetables would never call any superflous get_page
after get_user_page returns.  It's safer to make get_page universally safe
for tail pages and to use get_page_foll() within follow_page (inside
get_user_pages()).  get_page_foll() is safe to do the refcounting for tail
pages without taking any locks because it is run within PT lock protected
critical sections (PT lock for pte and page_table_lock for
pmd_trans_huge).

The standard get_page() as invoked by direct-io instead will now take
the compound_lock but still only for tail pages.  The direct-io paths
are usually I/O bound and the compound_lock is per THP so very
finegrined, so there's no risk of scalability issues with it.  A simple
direct-io benchmarks with all lockdep prove locking and spinlock
debugging infrastructure enabled shows identical performance and no
overhead.  So it's worth it.  Ideally direct-io should stop calling
get_page() on pages returned by get_user_pages().  The spinlock in
get_page() is already optimized away for no-THP builds but doing
get_page() on tail pages returned by GUP is generally a rare operation
and usually only run in I/O paths.

This new refcounting on page_tail->_mapcount in addition to avoiding new
RCU critical sections will also allow the working set estimation code to
work without any further complexity associated to the tail page
refcounting with THP.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Michel Lespinasse <walken@google.com>
Reviewed-by: Michel Lespinasse <walken@google.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Gibson <david@gibson.dropbear.id.au>
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@suse.de>

net: xen-netback: correctly restart Tx after a VM restore/migrate

[ Upstream commit d0e5d83284dac15c015bb48115b6780f5a6413cd ]

If a VM is saved and restored (or migrated) the netback driver will no
longer process any Tx packets from the frontend.  xenvif_up() does not
schedule the processing of any pending Tx requests from the front end
because the carrier is off.  Without this initial kick the frontend
just adds Tx requests to the ring without raising an event (until the
ring is full).

This was caused by 47103041e91794acdbc6165da0ae288d844c820b (net:
xen-netback: convert to hw_features) which reordered the calls to
xenvif_up() and netif_carrier_on() in xenvif_connect().

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

make PACKET_STATISTICS getsockopt report consistently between ring and non-ring

[ Upstream commit 7091fbd82cd5686444ffe9935ed6a8190101fe9d ]

This is a minor change.

Up until kernel 2.6.32, getsockopt(fd, SOL_PACKET, PACKET_STATISTICS,
...) would return total and dropped packets since its last invocation. The
introduction of socket queue overflow reporting [1] changed drop
rate calculation in the normal packet socket path, but not when using a
packet ring. As a result, the getsockopt now returns different statistics
depending on the reception method used. With a ring, it still returns the
count since the last call, as counts are incremented in tpacket_rcv and
reset in getsockopt. Without a ring, it returns 0 if no drops occurred
since the last getsockopt and the total drops over the lifespan of
the socket otherwise. The culprit is this line in packet_rcv, executed
on a drop:

drop_n_acct:
        po->stats.tp_drops = atomic_inc_return(&sk->sk_drops);

As it shows, the new drop number it taken from the socket drop counter,
which is not reset at getsockopt. I put together a small example
that demonstrates the issue [2]. It runs for 10 seconds and overflows
the queue/ring on every odd second. The reported drop rates are:
ring: 16, 0, 16, 0, 16, ...
non-ring: 0, 15, 0, 30, 0, 46, 0, 60, 0 , 74.

Note how the even ring counts monotonically increase. Because the
getsockopt adds tp_drops to tp_packets, total counts are similarly
reported cumulatively. Long story short, reinstating the original code, as
the below patch does, fixes the issue at the cost of additional per-packet
cycles. Another solution that does not introduce per-packet overhead
is be to keep the current data path, record the value of sk_drops at
getsockopt() at call N in a new field in struct packetsock and subtract
that when reporting at call N+1. I'll be happy to code that, instead,
it's just more messy.

[1] http://patchwork.ozlabs.org/patch/35665/
[2] http://kernel.googlecode.com/files/test-packetsock-getstatistics.c

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ipv6: nullify ipv6_ac_list and ipv6_fl_list when creating new socket

[ Upstream commit 676a1184e8afd4fed7948232df1ff91517400859 ]

ipv6_ac_list and ipv6_fl_list from listening socket are inadvertently
shared with new socket created for connection.

Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

tg3: negate USE_PHYLIB flag check

[ Upstream commit e730c82347b9dc75914da998c44c3f348965db41 ]

USE_PHYLIB flag in tg3_remove_one() is being checked incorrectly. This
results tg3_phy_fini->phy_disconnect is never called and when tg3 module
is removed.

In my case this resulted in panics in phy_state_machine calling function
phydev->adjust_link.

So correct this check.

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Acked-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

tcp: properly update lost_cnt_hint during shifting

[ Upstream commit 1e5289e121372a3494402b1b131b41bfe1cf9b7f ]

lost_skb_hint is used by tcp_mark_head_lost() to mark the first unhandled skb.
lost_cnt_hint is the number of packets or sacked packets before the lost_skb_hint;
When shifting a skb that is before the lost_skb_hint, if tcp_is_fack() is ture,
the skb has already been counted in the lost_cnt_hint; if tcp_is_fack() is false,
tcp_sacktag_one() will increase the lost_cnt_hint. So tcp_shifted_skb() does not
need to adjust the lost_cnt_hint by itself. When shifting a skb that is equal to
lost_skb_hint, the shifted packets will not be counted by tcp_mark_head_lost().
So tcp_shifted_skb() should adjust the lost_cnt_hint even tcp_is_fack(tp) is true.

Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

tcp: properly handle md5sig_pool references

[ Upstream commit 260fcbeb1ae9e768a44c9925338fbacb0d7e5ba9 ]

tcp_v4_clear_md5_list() assumes that multiple tcp md5sig peers
only hold one reference to md5sig_pool. but tcp_v4_md5_do_add()
increases use count of md5sig_pool for each peer. This patch
makes tcp_v4_md5_do_add() only increases use count for the first
tcp md5sig peer.

Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

netconsole: enable netconsole can make net_device refcnt incorrent

[ Upstream commit d5123480b1d6f7d1a5fe1a13520cef88fb5d4c84 ]

There is no check if netconsole is enabled current.
so when exec echo 1 > enabled;
the reference of net_device will increment always.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Acked-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

macvlan/macvtap: Fix unicast between macvtap interfaces in bridge mode

[ Upstream commit cb2d0f3e968bff7c6d262aca3e3ab8d4184e69b2 ]

Packets should always be forwarded to the lowerdev using dev_forward_skb.
vlan->forward is for packets being forwarded directly to another macvlan/
macvtap device (used for multicast in bridge mode).

Reported-and-tested-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: David Ward <david.ward@ll.mit.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

l2tp: fix a potential skb leak in l2tp_xmit_skb()

[ Upstream commit 835acf5da239b91edb9f7ebe36516999e156e6ee ]

l2tp_xmit_skb() can leak one skb if skb_cow_head() returns an error.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ipv4: fix ipsec forward performance regression

[ Upstream commit b73233960a59ee66e09d642f13d0592b13651e94 ]

There is bug in commit 5e2b61f(ipv4: Remove flowi from struct rtable).
It makes xfrm4_fill_dst() modify wrong data structure.

Signed-off-by: Zheng Yan <zheng.z.yan@intel.com>
Reported-by: Kim Phillips <kim.phillips@freescale.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

can bcm: fix tx_setup off-by-one errors

[ Upstream commit aabdcb0b553b9c9547b1a506b34d55a764745870 ]

This patch fixes two off-by-one errors that canceled each other out.
Checking for the same condition two times in bcm_tx_timeout_tsklet() reduced
the count of frames to be sent by one. This did not show up the first time
tx_setup is invoked as an additional frame is sent due to TX_ANNONCE.
Invoking a second tx_setup on the same item led to a reduced (by 1) number of
sent frames.

Reported-by: Andre Naujoks <nautsch@gmail.com>
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

bridge: fix hang on removal of bridge via netlink

[ Upstream commit 1ce5cce895309862d2c35d922816adebe094fe4a ]

Need to cleanup bridge device timers and ports when being bridge
device is being removed via netlink.

This fixes the problem of observed when doing:
 ip link add br0 type bridge
 ip link set dev eth1 master br0
 ip link set br0 up
 ip link del br0

which would cause br0 to hang in unregister_netdev because
of leftover reference count.

Reported-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

bonding: use local function pointer of bond->recv_probe in bond_handle_frame

[ Upstream commit 4d97480b1806e883eb1c7889d4e7a87e936e06d9 ]

The bond->recv_probe is called in bond_handle_frame() when
a packet is received, but bond_close() sets it to NULL. So,
a panic occurs when both functions work in parallel.

Why this happen:
After null pointer check of bond->recv_probe, an sk_buff is
duplicated and bond->recv_probe is called in bond_handle_frame.
So, a panic occurs when bond_close() is called between the
check and call of bond->recv_probe.

Patch:
This patch uses a local function pointer of bond->recv_probe
in bond_handle_frame(). So, it can avoid the null pointer
dereference.

Signed-off-by: Mitsuo Hayasaka <mitsuo.hayasaka.hu@hitachi.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

jsm: remove buggy write queue

commit 9d898966c4a07e4a5092215b5a2829d0ef02baa2 upstream.

jsm uses a write queue that copies from uart_core circular buffer. This
copying however has some bugs, like not wrapping the head counter. Since
this write queue is also a circular buffer, the consumer function is
ready to use the uart_core circular buffer directly.

This buggy copying function was making some bytes be dropped when
transmitting to a raw tty, doing something like this.

[root@hostname ~]$ cat /dev/ttyn1 > cascardo/dump &
[1] 2658
[root@hostname ~]$ cat /proc/tty/drivers > /dev/ttyn0
[root@hostname ~]$ cat /proc/tty/drivers
/dev/tty             /dev/tty        5       0 system:/dev/tty
/dev/console         /dev/console    5       1 system:console
/dev/ptmx            /dev/ptmx       5       2 system
/dev/vc/0            /dev/vc/0       4       0 system:vtmaster
jsm                  /dev/ttyn     250 0-31 serial
serial               /dev/ttyS       4 64-95 serial
hvc                  /dev/hvc      229 0-7 system
pty_slave            /dev/pts      136 0-1048575 pty:slave
pty_master           /dev/ptm      128 0-1048575 pty:master
unknown              /dev/tty        4 1-63 console
[root@hostname ~]$ cat cascardo/dump
/dev/tty             /dev/tty        5       0 system:/dev/tty
/dev/console         /dev/console    5       1 system:console
/dev/ptmx            /dev/ptmx       5       2 system
/dev/vc/0            /dev/vc/0       4       0 system:vtmaste[root@hostname ~]$

This patch drops the driver write queue entirely, using the circular
buffer from uart_core only.

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ptrace: don't clear GROUP_STOP_SIGMASK on double-stop

[This does not correspond to any specific patch in the upstream tree as it was
fixed accidentally by rewriting the code in the 3.1 release]

https://bugzilla.redhat.com/show_bug.cgi?id=740121

1. Luke Macken triggered WARN_ON(!(group_stop & GROUP_STOP_SIGMASK))
   in do_signal_stop().

   This is because do_signal_stop() clears GROUP_STOP_SIGMASK part
   unconditionally but doesn't update it if task_is_stopped().

2. Looking at this problem I noticed that WARN_ON_ONCE(!ptrace) is
   not right, a stopped-but-resumed tracee can clone the untraced
   thread in the SIGNAL_STOP_STOPPED group, the new thread can start
   another group-stop.

   Remove this warning, we need more fixes to make it true.

Reported-by: Luke Macken <lmacken@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

vfs pathname lookup: Add LOOKUP_AUTOMOUNT flag

Since we've now turned around and made LOOKUP_FOLLOW *not* force an
automount, we want to add the ability to force an automount event on
lookup even if we don't happen to have one of the other flags that force
it implicitly (LOOKUP_OPEN, LOOKUP_DIRECTORY, LOOKUP_PARENT..)

Most cases will never want to use this, since you'd normally want to
delay automounting as long as possible, which usually implies
LOOKUP_OPEN (when we open a file or directory, we really cannot avoid
the automount any more).

But Trond argued sufficiently forcefully that at a minimum bind mounting
a file and quotactl will want to force the automount lookup.  Some other
cases (like nfs_follow_remote_path()) could use it too, although
LOOKUP_DIRECTORY would work there as well.

This commit just adds the flag and logic, no users yet, though.  It also
doesn't actually touch the LOOKUP_NO_AUTOMOUNT flag that is related, and
was made irrelevant by the same change that made us not follow on
LOOKUP_FOLLOW.

Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Ian Kent <raven@themaw.net>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

VFS: Fix the remaining automounter semantics regressions

commit 815d405ceff0d6964683f033e18b9b23a88fba87 upstream.

The concensus seems to be that system calls such as stat() etc should
not trigger an automount.  Neither should the l* versions.

This patch therefore adds a LOOKUP_AUTOMOUNT flag to tag those lookups
that _should_ trigger an automount on the last path element.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
[ Edited to leave out the cases that are already covered by LOOKUP_OPEN,
  LOOKUP_DIRECTORY and LOOKUP_CREATE - all of which also fundamentally
  force automounting for their own reasons   - Linus ]

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

vfs: automount should ignore LOOKUP_FOLLOW

commit 0ec26fd0698285b31248e34bf1abb022c00f23d6 upstream.

Prior to 2.6.38 automount would not trigger on either stat(2) or
lstat(2) on the automount point.

After 2.6.38, with the introduction of the ->d_automount()
infrastructure, stat(2) and others would start triggering automount
while lstat(2), etc. still would not.  This is a regression and a
userspace ABI change.

Problem originally reported here:

  http://thread.gmane.org/gmane.linux.kernel.autofs/6098

It appears that there was an attempt at fixing various userspace tools
to not trigger the automount.  But since the stat system call is
rather common it is impossible to "fix" all userspace.

This patch reverts the original behavior, which is to not trigger on
stat(2) and other symlink following syscalls.

[ It's not really clear what the right behavior is.  Apparently Solaris
  does the "automount on stat, leave alone on lstat".  And some programs
  can get unhappy when "stat+open+fstat" ends up giving a different
  result from the fstat than from the initial stat.

  But the change in 2.6.38 resulted in problems for some people, so
  we're going back to old behavior.  Maybe we can re-visit this
  discussion at some future date  - Linus ]

Reported-by: Leonardo Chiquitto <leonardo.lists@gmail.com>
Acked-by: Ian Kent <raven@themaw.net>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

VFS: Fix automount for negative autofs dentries

commit 5a30d8a2b8ddd5102c440c7e5a7c8e1fd729c818 upstream.
[ backport for 3.0.x: LOOKUP_PARENT => LOOKUP_CONTINUE by Chuck Ebbert
<cebbert@redhat.com> ]

Autofs may set the DCACHE_NEED_AUTOMOUNT flag on negative dentries.  These
need attention from the automounter daemon regardless of the LOOKUP_FOLLOW flag.

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Ian Kent <raven@themaw.net>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

readlinkat: ensure we return ENOENT for the empty pathname for normal lookups

commit 1fa1e7f615f4d3ae436fa319af6e4eebdd4026a8 upstream.

Since the commit below which added O_PATH support to the *at() calls, the
error return for readlink/readlinkat for the empty pathname has switched
from ENOENT to EINVAL:

  commit 65cfc6722361570bfe255698d9cd4dccaf47570d
  Author: Al Viro <viro@zeniv.linux.org.uk>
  Date:   Sun Mar 13 15:56:26 2011 -0400

    readlinkat(), fchownat() and fstatat() with empty relative pathnames

This is both unexpected for userspace and makes readlink/readlinkat
inconsistant with all other interfaces; and inconsistant with our stated
return for these pathnames.

As the readlinkat call does not have a flags parameter we cannot use the
AT_EMPTY_PATH approach used in the other calls.  Therefore expose whether
the original path is infact entry via a new user_path_at_empty() path
lookup function.  Use this to determine whether to default to EINVAL or
ENOENT for failures.

Addresses http://bugs.launchpad.net/bugs/817187

[akpm@linux-foundation.org: remove unused getname_flags()]
Signed-off-by: Andy Whitcroft <apw@canonical.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

um: fix ubd cow size

commit 8535639810e578960233ad39def3ac2157b0c3ec upstream.

ubd_file_size() cannot use ubd_dev->cow.file because at this time
ubd_dev->cow.file is not initialized.
Therefore, ubd_file_size() will always report a wrong disk size when
COW files are used.
Reading from /dev/ubd* would crash the kernel.

We have to read the correct disk size from the COW file's backing
file.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ALSA: hda - Fix ADC input-amp handling for Cx20549 codec

commit 6b45214277bec2193ad3ccb8d7aa6100b5a0f1a9 upstream.

It seems that Conexant CX20549 chip handle only a single input-amp even
though the audio-input widget has multiple sources.  This has been never
clear, and I implemented in the current way based on the debug information
I got at the early time -- the device reacts individual input-amp values
for different sources.  This is true for another Conexant codec, but it's
not applied to CX20549 actually.

This patch changes the auto-parser code to handle a single input-amp
per audio-in widget for CX20549.  After applying this, you'll see only a
single "Capture" volume control instead of separate "Mic" or "Line"
captures when the device is set up to use a single ADC.

We haven't tested 20551 and 20561 codecs yet.  If these show the similar
behavior like 20549, they need to set spec->single_adc_amp=1, too.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

mm: avoid null pointer access in vm_struct via /proc/vmallocinfo

commit f5252e009d5b87071a919221e4f6624184005368 upstream.

The /proc/vmallocinfo shows information about vmalloc allocations in
vmlist that is a linklist of vm_struct.  It, however, may access pages
field of vm_struct where a page was not allocated.  This results in a null
pointer access and leads to a kernel panic.

Why this happens: In __vmalloc_node_range() called from vmalloc(), newly
allocated vm_struct is added to vmlist at __get_vm_area_node() and then,
some fields of vm_struct such as nr_pages and pages are set at
__vmalloc_area_node().  In other words, it is added to vmlist before it is
fully initialized.  At the same time, when the /proc/vmallocinfo is read,
it accesses the pages field of vm_struct according to the nr_pages field
at show_numa_info().  Thus, a null pointer access happens.

The patch adds the newly allocated vm_struct to the vmlist *after* it is
fully initialized.  So, it can avoid accessing the pages field with
unallocated page when show_numa_info() is called.

Signed-off-by: Mitsuo Hayasaka <mitsuo.hayasaka.hu@hitachi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.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@suse.de>

ARM: mach-ux500: unlock I&D l2x0 caches before init

commit 1bf6d2c1bb23533af6930581cc39b74685bc29de upstream.

Apparently U8500 U-Boot versions may leave the l2x0 locked down
before executing the kernel. Make sure we unlock it before we
initialize the l2x0. This fixes a performance problem reported
by Jan Rinze.

The l2x0 core has been modified to unlock the l2x0 by default,
but it will not touch the locking registers if the l2x0 was
already enabled, as on the ux500, so we need this quirk to
make sure it is properly turned off.

Cc: Srinidhi Kasagar <srinidhi.kasagar@stericsson.com>
Cc: Rabin Vincent <rabin.vincent@stericsson.com>
Cc: Adrian Bunk <adrian.bunk@movial.com>
Reported-by: Jan Rinze <janrinze@gmail.com>
Tested-by: Robert Marklund <robert.marklund@stericsson.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

plat-mxc: iomux-v3.h: implicitly enable pull-up/down when that's desired

commit 6571534b600b8ca1936ff5630b9e0947f21faf16 upstream.

To configure pads during the initialisation a set of special constants
is used, e.g.
#define MX25_PAD_FEC_MDIO__FEC_MDIO IOMUX_PAD(0x3c4, 0x1cc, 0x10, 0, 0, PAD_CTL_HYS | PAD_CTL_PUS_22K_UP)

The problem is that no pull-up/down is getting activated unless both
PAD_CTL_PUE (pull-up enable) and PAD_CTL_PKE (pull/keeper module
enable) set. This is clearly stated in the i.MX25 datasheet and is
confirmed by the measurements on hardware. This leads to some rather
hard to understand bugs such as misdetecting an absent ethernet PHY (a
real bug i had), unstable data transfer etc. This might affect mx25,
mx35, mx50, mx51 and mx53 SoCs.

It's reasonable to expect that if the pullup value is specified, the
intention was to have it actually active, so we implicitly add the
needed bits.

Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

/proc/self/numa_maps: restore "huge" tag for hugetlb vmas

commit fc360bd9cdcf875639a77f07fafec26699c546f3 upstream.

The display of the "huge" tag was accidentally removed in 29ea2f698 ("mm:
use walk_page_range() instead of custom page table walking code").

Reported-by: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Stephen Hemminger <shemminger@vyatta.com>
Reviewed-by: Stephen Wilson <wilsons@start.ca>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Christoph Lameter <cl@linux-foundation.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@suse.de>

tuner_xc2028: Allow selection of the frequency adjustment code for XC3028

commit 9bed77ee2fb46b74782d0d9d14b92e9d07f3df6e upstream.

This device is not using the proper demod IF. Instead of using the
IF macro, it is specifying a IF frequency. This doesn't work, as xc3028
needs to load an specific SCODE for the tuner. In this case, there's
no IF table for 5 MHz.

Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

dib0700: protect the dib0700 buffer access

commit bff469f4167fdabfe15294f375577d7eadbaa1bb upstream.

This patch protects the common buffer access inside the dib0700 in order
to manage concurrent access. This protection is done using mutex.

Cc: Mauro Carvalho Chehab <mchehab@redhat.com>
Cc: Florian Mickler <florian@mickler.org>
Signed-off-by: Javier Marcet <javier@marcet.info>
Signed-off-by: Olivier Grenie <olivier.grenie@dibcom.fr>
Signed-off-by: Patrick Boettcher <patrick.boettcher@dibcom.fr>
[mchehab@redhat.com: dprint requires 3 arguments. Replaced by dib_info]
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

DiBcom: protect the I2C bufer access

commit 79fcce3230b140f7675f8529ee53fe2f9644f902 upstream.

This patch protects the I2C buffer access in order to manage concurrent
access. This protection is done using mutex.
Furthermore, for the dib9000, if a pid filtering command is
received during the tuning, this pid filtering command is delayed to
avoid any concurrent access issue.

Cc: Mauro Carvalho Chehab <mchehab@redhat.com>
Cc: Florian Mickler <florian@mickler.org>
Signed-off-by: Olivier Grenie <olivier.grenie@dibcom.fr>
Signed-off-by: Patrick Boettcher <Patrick.Boettcher@dibcom.fr>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

uvcvideo: Set alternate setting 0 on resume if the bus has been reset

commit d59a7b1dbce8b972ec2dc9fcaaae0bfa23687423 upstream.

If the bus has been reset on resume, set the alternate setting to 0.
This should be the default value, but some devices crash or otherwise
misbehave if they don't receive a SET_INTERFACE request before any other
video control request.

Microdia's 0c45:6437 camera has been found to require this change or it
will stop sending video data after resume.

uvc_video.c]

Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

viafb: improve pitch handling

commit 936a3f770b8de7042d793272f008ef1bb08522e9 upstream.

This patch adds checks for minimum and maximum pitch size to prevent
invalid settings which could otherwise crash the machine. Also the
alignment is done in a slightly more readable way.

Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

viafb: use display information in info not in var for panning

commit d933990c57b498c092ceef591c7c5d69dbfe9f30 upstream.

As Laurent pointed out we must not use any information in the passed
var besides xoffset, yoffset and vmode as otherwise applications
might abuse it. Also use the aligned fix.line_length and not the
(possible) unaligned xres_virtual.

Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Reported-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

fb: sh-mobile: Fix deadlock risk between lock_fb_info() and console_lock()

commit 4a47a0e09c504e3ce0ccdb405411aefc5b09deb8 upstream.

Following on Herton's patch "fb: avoid possible deadlock caused by
fb_set_suspend" which moves lock_fb_info() out of fb_set_suspend()
to its callers, correct sh-mobile's locking around call to
fb_set_suspend() and the same sort of deaklocks with console_lock()
due to order of taking the lock.

console_lock() must be taken while fb_info is already locked and fb_info
must be locked while calling fb_set_suspend().

Signed-off-by: Bruno Prémont <bonbons@linux-vserver.org>
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

fb: avoid possible deadlock caused by fb_set_suspend

commit 9e769ff3f585db8f978f9113be83d36c7e3965dd upstream.

A lock ordering issue can cause deadlocks: in framebuffer/console code,
all needed struct fb_info locks are taken before acquire_console_sem(),
in places which need to take console semaphore.

But fb_set_suspend is always called with console semaphore held, and
inside it we call lock_fb_info which gets the fb_info lock, inverse
locking order of what the rest of the code does. This causes a real
deadlock issue, when we write to state fb sysfs attribute (which calls
fb_set_suspend) while a framebuffer is being unregistered by
remove_conflicting_framebuffers, as can be shown by following show
blocked state trace on a test program which loads i915 and runs another
forked processes writing to state attribute:

Test process with semaphore held and trying to get fb_info lock:
..
fb-test2      D 0000000000000000     0   237    228 0x00000000
 ffff8800774f3d68 0000000000000082 00000000000135c0 00000000000135c0
 ffff880000000000 ffff8800774f3fd8 ffff8800774f3fd8 ffff880076ee4530
 00000000000135c0 ffff8800774f3fd8 ffff8800774f2000 00000000000135c0
Call Trace:
 [<ffffffff8141287a>] __mutex_lock_slowpath+0x11a/0x1e0
 [<ffffffff814142f2>] ? _raw_spin_lock_irq+0x22/0x40
 [<ffffffff814123d3>] mutex_lock+0x23/0x50
 [<ffffffff8125dfc5>] lock_fb_info+0x25/0x60
 [<ffffffff8125e3f0>] fb_set_suspend+0x20/0x80
 [<ffffffff81263e2f>] store_fbstate+0x4f/0x70
 [<ffffffff812e7f70>] dev_attr_store+0x20/0x30
 [<ffffffff811c46b4>] sysfs_write_file+0xd4/0x160
 [<ffffffff81155a26>] vfs_write+0xc6/0x190
 [<ffffffff81155d51>] sys_write+0x51/0x90
 [<ffffffff8100c012>] system_call_fastpath+0x16/0x1b
..
modprobe process stalled because has the fb_info lock (got inside
unregister_framebuffer) but waiting for the semaphore held by the
test process which is waiting to get the fb_info lock:
..
modprobe      D 0000000000000000     0   230    218 0x00000000
 ffff880077a4d618 0000000000000082 0000000000000001 0000000000000001
 ffff880000000000 ffff880077a4dfd8 ffff880077a4dfd8 ffff8800775a2e20
 00000000000135c0 ffff880077a4dfd8 ffff880077a4c000 00000000000135c0
Call Trace:
 [<ffffffff81411fe5>] schedule_timeout+0x215/0x310
 [<ffffffff81058051>] ? get_parent_ip+0x11/0x50
 [<ffffffff814130dd>] __down+0x6d/0xb0
 [<ffffffff81089f71>] down+0x41/0x50
 [<ffffffff810629ac>] acquire_console_sem+0x2c/0x50
 [<ffffffff812ca53d>] unbind_con_driver+0xad/0x2d0
 [<ffffffff8126f5f7>] fbcon_event_notify+0x457/0x890
 [<ffffffff814144ff>] ? _raw_spin_unlock_irqrestore+0x1f/0x50
 [<ffffffff81058051>] ? get_parent_ip+0x11/0x50
 [<ffffffff8141836d>] notifier_call_chain+0x4d/0x70
 [<ffffffff8108a3b8>] __blocking_notifier_call_chain+0x58/0x80
 [<ffffffff8108a3f6>] blocking_notifier_call_chain+0x16/0x20
 [<ffffffff8125dabb>] fb_notifier_call_chain+0x1b/0x20
 [<ffffffff8125e6ac>] unregister_framebuffer+0x7c/0x130
 [<ffffffff8125e8b3>] remove_conflicting_framebuffers+0x153/0x180
 [<ffffffff8125eef3>] register_framebuffer+0x93/0x2c0
 [<ffffffffa0331112>] drm_fb_helper_single_fb_probe+0x252/0x2f0 [drm_kms_helper]
 [<ffffffffa03314a3>] drm_fb_helper_initial_config+0x2f3/0x6d0 [drm_kms_helper]
 [<ffffffffa03318dd>] ? drm_fb_helper_single_add_all_connectors+0x5d/0x1c0 [drm_kms_helper]
 [<ffffffffa037b588>] intel_fbdev_init+0xa8/0x160 [i915]
 [<ffffffffa0343d74>] i915_driver_load+0x854/0x12b0 [i915]
 [<ffffffffa02f0e7e>] drm_get_pci_dev+0x19e/0x360 [drm]
 [<ffffffff8141821d>] ? sub_preempt_count+0x9d/0xd0
 [<ffffffffa0386f91>] i915_pci_probe+0x15/0x17 [i915]
 [<ffffffff8124481f>] local_pci_probe+0x5f/0xd0
 [<ffffffff81244f89>] pci_device_probe+0x119/0x120
 [<ffffffff812eccaa>] ? driver_sysfs_add+0x7a/0xb0
 [<ffffffff812ed003>] driver_probe_device+0xa3/0x290
 [<ffffffff812ed1f0>] ? __driver_attach+0x0/0xb0
 [<ffffffff812ed29b>] __driver_attach+0xab/0xb0
 [<ffffffff812ed1f0>] ? __driver_attach+0x0/0xb0
 [<ffffffff812ebd3e>] bus_for_each_dev+0x5e/0x90
 [<ffffffff812ecc2e>] driver_attach+0x1e/0x20
 [<ffffffff812ec6f2>] bus_add_driver+0xe2/0x320
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffff812ed536>] driver_register+0x76/0x140
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffff81245216>] __pci_register_driver+0x56/0xd0
 [<ffffffffa02f1264>] drm_pci_init+0xe4/0xf0 [drm]
 [<ffffffffa03aa000>] ? i915_init+0x0/0x96 [i915]
 [<ffffffffa02e84a8>] drm_init+0x58/0x70 [drm]
 [<ffffffffa03aa094>] i915_init+0x94/0x96 [i915]
 [<ffffffff81002194>] do_one_initcall+0x44/0x190
 [<ffffffff810a066b>] sys_init_module+0xcb/0x210
 [<ffffffff8100c012>] system_call_fastpath+0x16/0x1b
..

fb-test2 which reproduces above is available on kernel.org bug #26232.
To solve this issue, avoid calling lock_fb_info inside fb_set_suspend,
and move it out to where needed (callers of fb_set_suspend must call
lock_fb_info before if needed). So far, the only place which needs to
call lock_fb_info is store_fbstate, all other places which calls
fb_set_suspend are suspend/resume hooks that should not need the lock as
they should be run only when processes are already frozen in
suspend/resume.

References: https://bugzilla.kernel.org/show_bug.cgi?id=26232
Signed-off-by: Herton Ronaldo Krzesinski <herton@mandriva.com.br>
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

carminefb: Fix module parameters permissions

commit c84c14224bbca6ec60d5851fcc87be0e34df2f44 upstream.

The third parameter of module_param is supposed to be an octal value.
The missing leading "0" causes the following:

$ ls -l /sys/module/carminefb/parameters/
total 0
-rw-rwxr-- 1 root root 4096 Jul  8 08:55 fb_displays
-rw-rwxr-- 1 root root 4096 Jul  8 08:55 fb_mode
-rw-rwxr-- 1 root root 4096 Jul  8 08:55 fb_mode_str

After fixing the perm parameter, we get the expected:

$ ls -l /sys/module/carminefb/parameters/
total 0
-r--r--r-- 1 root root 4096 Jul  8 08:56 fb_displays
-r--r--r-- 1 root root 4096 Jul  8 08:56 fb_mode
-r--r--r-- 1 root root 4096 Jul  8 08:56 fb_mode_str

Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

iommu/amd: Fix wrong shift direction

commit fcd0861db1cf4e6ed99f60a815b7b72c2ed36ea4 upstream.

The shift direction was wrong because the function takes a
page number and i is the address is the loop.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

WMI: properly cleanup devices to avoid crashes

commit 023b9565972a4a5e0f01b9aa32680af6e9b5c388 upstream.

We need to remove devices that we destroy from the list, otherwise
we'll crash if there are more than one "_WDG" methods in DSDT.

This fixes https://bugzilla.kernel.org/show_bug.cgi?id=32052

Tested-by: Ilya Tumaykin <librarian_rus@yahoo.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Acked-by: Carlos Corbacho <carlos@strangeworlds.co.uk>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ccwgroup: move attributes to attribute group

commit dbdf1afcaaabe83dea15a3cb9b9013e73ae3b1ad upstream.

Put sysfs attributes of ccwgroup devices in an attribute group to
ensure that these attributes are actually present when userspace
is notified via uevents.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

memory leak with RCU_TABLE_FREE

commit e73b7fffe487c315fd1a4fa22282e3362b440a06 upstream.

The rcu page table free code uses a couple of bits in the page table
pointer passed to tlb_remove_table to discern the different page table
types. __tlb_remove_table extracts the type with an incorrect mask which
leads to memory leaks. The correct mask is ((FRAG_MASK << 4) | FRAG_MASK).

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

user per registers vs. ptrace single stepping

commit a45aff5285871bf7be1781d9462d3fdbb6c913f9 upstream.

git commit 5e9a2692 "[S390] ptrace cleanup" introduced a regression
for the case when both a user PER set (e.g. a storage alteration trace) and
PTRACE_SINGLESTEP are active. The new code will overrule the user PER set
with a instruction-fetch PER set over the whole address space for ptrace
single stepping. The inferior process will be stopped after each instruction
with an instruction fetch event. Any other events that may have occurred
concurrently are not reported (e.g. storage alteration event) because the
control bits for them are not set. The solution is to merge the PER control
bits of the user PER set with the PER_EVENT_IFETCH control bit for
PTRACE_SINGLESTEP.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

KVM: s390: check cpu_id prior to using it

commit 4d47555a80495657161a7e71ec3014ff2021e450 upstream.

We use the cpu id provided by userspace as array index here. Thus we
clearly need to check it first. Ooops.

Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: Fix a bug in WM8962 DSP_A and DSP_B settings

commit fbc7c62a3ff831aef24894b7982cd1adb2b7e070 upstream.

Signed-off-by: Susan Gao <sgao@opensource.wolfsonmicro.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmico.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: Remove direct register cache accesses from WM8962 driver

commit 38f3f31a0a797bdbcc0cdb12553bbecc2f9a91c4 upstream.

Also fix return values for speaker switch updates.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: wm8994: Use SND_SOC_DAPM_AIF_OUT for AIF3 Capture

commit 35024f4922f7b271e7529673413889aa3d51c5fc upstream.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: ak4535: fixup cache register table

commit 7c04241acbdaf97f1448dcccd27ea0fcd1a57684 upstream.

ak4535_reg should be 8bit, but cache table is defined as 16bit.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: ak4642: fixup cache register table

commit 19b115e523208a926813751aac8934cf3fc6085e upstream.

ak4642 register was 8bit, but cache table was defined as 16bit.
ak4642 doesn't work correctry without this patch.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: wm8741: Fix setting interface format for DSP modes

commit 3a340104fad6ecbea5ad6792a2ea855f0507a6e0 upstream.

According to the datasheet:
Format Control (05h)
BITS[3:2]
        FMT[1:0] Audio data format selection
                00 = right justified mode
                01 = left justified mode
                10 = I2S mode
                11 = DSP mode
BIT[4]  LRP Polarity selec for LRCLK/DSP mode select
                0 = normal LRCLK poalrity/DSP mode A
                1 = inverted LRCLK poarity/DSP mode B

For SND_SOC_DAIFMT_DSP_A, we should set 0x000C instead of 0x0003.
For SND_SOC_DAIFMT_DSP_B, we should set 0x001C instead of 0x0013.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Liam Girdwood <lrg@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ASoC: wm8940: Properly set codec->dapm.bias_level

commit 5927f94700e860ae27ff24e7f3bc9e4f7b9922eb upstream.

Reported-by: Chris Paulson-Ellis <chris@edesix.com>
Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

io-mapping: ensure io_mapping_map_atomic _is_ atomic

commit 24dd85ff723f142093f44244764b9b5c152235b8 upstream.

For the !HAVE_ATOMIC_IOMAP case the stub functions did not call
pagefault_disable/_enable. The i915 driver relies on the map
actually being atomic, otherwise it can deadlock with it's own
pagefault handler in the gtt pwrite fastpath.

This is exercised by gem_mmap_gtt from the intel-gpu-toosl gem
testsuite.

v2: Chris Wilson noted the lack of an include.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=38115
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

vfs: add "device" tag to /proc/self/mountstats

commit a877ee03ac010ded434b77f7831f43cbb1fcc60f upstream.

nfsiostat was failing to find mounted filesystems on kernels after
2.6.38 because of changes to show_vfsstat() by commit
c7f404b40a3665d9f4e9a927cc5c1ee0479ed8f9.  This patch adds back the
"device" tag before the nfs server entry so scripts can parse the
mountstats file correctly.

Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

hppfs: missing include

commit d6b722aa383a467a43d09ee38e866981abba08ab upstream.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: ignore WANT bits in open downgrade

commit c30e92df30d7d5fe65262fbce5d1b7de675fe34e upstream.

We don't use WANT bits yet--and sending them can probably trigger a
BUG() further down.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: fix open downgrade, again

commit 3d02fa29dec920c597dd7b7db608a4bc71f088ce upstream.

Yet another open-management regression:

- nfs4_file_downgrade() doesn't remove the BOTH access bit on
  downgrade, so the server's idea of the stateid's access gets
  out of sync with the client's.  If we want to keep an O_RDWR
  open in this case, we should do that in the file_put_access
  logic rather than here.
- We forgot to convert v4 access to an open mode here.

This logic has proven too hard to get right.  In the future we may
consider:
- reexamining the lock/openowner relationship (locks probably
  don't really need to take their own references here).
- adding open upgrade/downgrade support to the vfs.
- removing the atomic operations.  They're redundant as long as
  this is all under some other lock.

Also, maybe some kind of additional static checking would help catch
O_/NFS4_SHARE_ACCESS confusion.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: permit read opens of executable-only files

commit a043226bc140a2c1dde162246d68a67e5043e6b2 upstream.

A client that wants to execute a file must be able to read it.  Read
opens over nfs are therefore implicitly allowed for executable files
even when those files are not readable.

NFSv2/v3 get this right by using a passed-in NFSD_MAY_OWNER_OVERRIDE on
read requests, but NFSv4 has gotten this wrong ever since
dc730e173785e29b297aa605786c94adaffe2544 "nfsd4: fix owner-override on
open", when we realized that the file owner shouldn't override
permissions on non-reclaim NFSv4 opens.

So we can't use NFSD_MAY_OWNER_OVERRIDE to tell nfsd_permission to allow
reads of executable files.

So, do the same thing we do whenever we encounter another weird NFS
permission nit: define yet another NFSD_MAY_* flag.

The industry's future standardization on 128-bit processors will be
motivated primarily by the need for integers with enough bits for all
the NFSD_MAY_* flags.

Reported-by: Leonardo Borda <leonardoborda@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: fix seqid_mutating_error

commit 576163005de286bbd418fcb99cfd0971523a0c6d upstream.

The set of errors here does *not* agree with the set of errors specified
in the rfc!

While we're there, turn this macros into a function, for the usual
reasons, and move it to the one place where it's actually used.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: stop using nfserr_resource for transitory errors

commit 3e77246393c0a433247631a1f0e9ec98d3d78a1c upstream.

The server is returning nfserr_resource for both permanent errors and
for errors (like allocation failures) that might be resolved by retrying
later.  Save nfserr_resource for the former and use delay/jukebox for
the latter.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfsd4: Remove check for a 32-bit cookie in nfsd4_readdir()

commit 832023bffb4b493f230be901f681020caf3ed1f8 upstream.

Fan Yong <yong.fan@whamcloud.com> noticed setting
FMODE_32bithash wouldn't work with nfsd v4, as
nfsd4_readdir() checks for 32 bit cookies. However, according to RFC 3530
cookies have a 64 bit type and cookies are also defined as u64 in
'struct nfsd4_readdir'. So remove the test for >32-bit values.

Signed-off-by: Bernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfs: don't try to migrate pages with active requests

commit 2da956523526e440ef4f4dd174e26f5ac06fe011 upstream.

nfs_find_and_lock_request will take a reference to the nfs_page and
will then put it if the req is already locked. It's possible though
that the reference will be the last one. That put then can kick off
a whole series of reference puts:

nfs_page
   nfs_open_context
      dentry
          inode

If the inode ends up being deleted, then the VFS will call
truncate_inode_pages. That function will try to take the page lock, but
it was already locked when migrate_page was called. The code
deadlocks.

Fix this by simply refusing the migration request if PagePrivate is
already set, indicating that the page is already associated with an
active read or write request.

We've had a customer test a backported version of this patch and
the preliminary results seem good.

Cc: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Harshula Jayasuriya <harshula@redhat.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

genirq: Add IRQF_RESUME_EARLY and resume such IRQs earlier

commit 9bab0b7fbaceec47d32db51cd9e59c82fb071f5a upstream.

This adds a mechanism to resume selected IRQs during syscore_resume
instead of dpm_resume_noirq.

Under Xen we need to resume IRQs associated with IPIs early enough
that the resched IPI is unmasked and we can therefore schedule
ourselves out of the stop_machine where the suspend/resume takes
place.

This issue was introduced by 676dc3cf5bc3 "xen: Use IRQF_FORCE_RESUME".

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Jeremy Fitzhardinge <Jeremy.Fitzhardinge@citrix.com>
Cc: xen-devel <xen-devel@lists.xensource.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Link: http://lkml.kernel.org/r/1318713254.11016.52.camel@dagon.hellion.org.uk
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

tracing: Fix returning of duplicate data after EOF in trace_pipe_raw

commit 436fc280261dcfce5af38f08b89287750dc91cd2 upstream.

The trace_pipe_raw handler holds a cached page from the time the file
is opened to the time it is closed. The cached page is used to handle
the case of the user space buffer being smaller than what was read from
the ring buffer. The left over buffer is held in the cache so that the
next read will continue where the data left off.

After EOF is returned (no more data in the buffer), the index of
the cached page is set to zero. If a user app reads the page again
after EOF, the check in the buffer will see that the cached page
is less than page size and will return the cached page again. This
will cause reading the trace_pipe_raw again after EOF to return
duplicate data, making the output look like the time went backwards
but instead data is just repeated.

The fix is to not reset the index right after all data is read
from the cache, but to reset it after all data is read and more
data exists in the ring buffer.

Reported-by: Jeremy Eder <jeder@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

perf probe: Fix to show correct error string

commit 44a56040a0037a845d5fa218dffde464579f0cab upstream.

Fix perf probe to show correct error string when it
fails to delete an event. The write(2) returns -1
if failed, and errno stores real error number.

Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Link: http://lkml.kernel.org/r/20111004104504.14591.41266.stgit@fedora15
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

md/raid5: fix bug that could result in reads from a failed device.

commit 355840e7a7e56bb2834fd3b0da64da5465f8aeaa upstream.

This bug was introduced in 415e72d034c50520ddb7ff79e7d1792c1306f0c9
which was in 2.6.36.

There is a small window of time between when a device fails and when
it is removed from the array.  During this time we might still read
from it, but we won't write to it - so it is possible that we could
read stale data.

We didn't need the test of 'Faulty' before because the test on
In_sync is sufficient.  Since we started allowing reads from the early
part of non-In_sync devices we need a test on Faulty too.

This is suitable for any kernel from 2.6.36 onwards, though the patch
might need a bit of tweaking in 3.0 and earlier.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

apic, i386/bigsmp: Fix false warnings regarding logical APIC ID mismatches

commit 838312be46f3abfbdc175f81c3e54a857994476d upstream.

These warnings (generally one per CPU) are a result of
initializing x86_cpu_to_logical_apicid while apic_default is
still in use, but the check in setup_local_APIC() being done
when apic_bigsmp was already used as an override in
default_setup_apic_routing():

 Overriding APIC driver with bigsmp
 Enabling APIC mode:  Physflat.  Using 5 I/O APICs
 ------------[ cut here ]------------
 WARNING: at .../arch/x86/kernel/apic/apic.c:1239
 ...
 CPU 1 irqstacks, hard=f1c9a000 soft=f1c9c000
 Booting Node   0, Processors  #1
 smpboot cpu 1: start_ip = 9e000
 Initializing CPU#1
 ------------[ cut here ]------------
 WARNING: at .../arch/x86/kernel/apic/apic.c:1239
 setup_local_APIC+0x137/0x46b() Hardware name: ...
 CPU1 logical APIC ID: 2 != 8
 ...

Fix this (for the time being, i.e. until
x86_32_early_logical_apicid() will get removed again, as Tejun
says ought to be possible) by overriding the previously stored
values at the point where the APIC driver gets overridden.

v2: Move this and the pre-existing override logic into
    arch/x86/kernel/apic/bigsmp_32.c.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/4E835D16020000780005844C@nat28.tlf.novell.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

time: Change jiffies_to_clock_t() argument type to unsigned long

commit cbbc719fccdb8cbd87350a05c0d33167c9b79365 upstream.

The parameter's origin type is long. On an i386 architecture, it can
easily be larger than 0x80000000, causing this function to convert it
to a sign-extended u64 type.

Change the type to unsigned long so we get the correct result.

Signed-off-by: hank <pyu@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
[ build fix ]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

wl12xx: fix forced passive scans

commit 6cd9d21a0c1e2648c07c32c66bb25795ad3208aa upstream.

We were using incorrect max and min dwell times during forced passive
scans because we were still using the active scan states to scan
(passively) the channels that were not marked as passive.

Instead of doing passive scans in active states, we now skip active
states and scan for all channels in passive states.

Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

net: hold sock reference while processing tx timestamps

commit da92b194cc36b5dc1fbd85206aeeffd80bee0c39 upstream.

The pair of functions,

 * skb_clone_tx_timestamp()
 * skb_complete_tx_timestamp()

were designed to allow timestamping in PHY devices. The first
function, called during the MAC driver's hard_xmit method, identifies
PTP protocol packets, clones them, and gives them to the PHY device
driver. The PHY driver may hold onto the packet and deliver it at a
later time using the second function, which adds the packet to the
socket's error queue.

As pointed out by Johannes, nothing prevents the socket from
disappearing while the cloned packet is sitting in the PHY driver
awaiting a timestamp. This patch fixes the issue by taking a reference
on the socket for each such packet. In addition, the comments
regarding the usage of these function are expanded to highlight the
rule that PHY drivers must use skb_complete_tx_timestamp() to release
the packet, in order to release the socket reference, too.

These functions first appeared in v2.6.36.

Reported-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

mac80211: fix offchannel TX cookie matching

commit 28a1bcdb57d50f3038a255741ecc83e391e5282e upstream.

When I introduced in-kernel off-channel TX I
introduced a bug -- the work can't be canceled
again because the code clear the skb pointer.
Fix this by keeping track separately of whether
TX status has already been reported.

Reported-by: Jouni Malinen <j@w1.fi>
Tested-by: Jouni Malinen <j@w1.fi>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

dp83640: free packet queues on remove

commit 8b3408f8ee994973869d8ba32c5bf482bc4ddca4 upstream.

If the PHY should disappear (for example, on an USB Ethernet MAC), then
the driver would leak any undelivered time stamp packets. This commit
fixes the issue by calling the appropriate functions to free any packets
left in the transmit and receive queues.

The driver first appeared in v3.0.

Signed-off-by: Richard Cochran <richard.cochran@omicron.at>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

rtnetlink: Add missing manual netlink notification in dev_change_net_namespaces

commit d2237d35748e7f448a9c2d9dc6a85ef637466e24 upstream.

Renato Westphal noticed that since commit a2835763e130c343ace5320c20d33c281e7097b7
"rtnetlink: handle rtnl_link netlink notifications manually" was merged
we no longer send a netlink message when a networking device is moved
from one network namespace to another.

Fix this by adding the missing manual notification in dev_change_net_namespaces.

Since all network devices that are processed by dev_change_net_namspaces are
in the initialized state the complicated tests that guard the manual
rtmsg_ifinfo calls in rollback_registered and register_netdevice are
unnecessary and we can just perform a plain notification.

Tested-by: Renato Westphal <renatowestphal@gmail.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

ata_piix: make DVD Drive recognisable on systems with Intel Sandybridge chipsets(v2)

commit 5e5a4f5d5a08c9c504fe956391ac3dae2c66556d upstream.

This quirk patch fixes one kind of bug inside some Intel Sandybridge
chipsets, see reports from

       https://bugzilla.kernel.org/show_bug.cgi?id=40592.

Many guys also have reported the problem before:

https://bugs.launchpad.net/bugs/737388
https://bugs.launchpad.net/bugs/794642
https://bugs.launchpad.net/bugs/782389
......

With help from Tejun, the problem is found to be caused by 32bit PIO
mode, so introduce the quirk patch to disable 32bit PIO on SATA piix
for some Sandybridge CPT chipsets.

Seth also tested the patch on all five affected chipsets
(pci device ID: 0x1c00, 0x1c01, 0x1d00, 0x1e00, 0x1e01), and found
the patch does fix the problem.

Tested-by: Heasley, Seth <seth.heasley@intel.com>
Cc: Alan Cox <alan@linux.intel.com>
Signed-off-by: Ming Lei <ming.lei@canonical.com>
Acked-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

nfs: don't redirty inode when ncommit == 0 in nfs_commit_unstable_pages

commit 3236c3e1adc0c7ec83eaff1de2d06746b7c5bb28 upstream.

commit 420e3646 allowed the kernel to reduce the number of unnecessary
commit calls by skipping the commit when there are a large number of
outstanding pages.

However, the current test in nfs_commit_unstable_pages does not handle
the edge condition properly. When ncommit == 0, then that means that the
kernel doesn't need to do anything more for the inode. The current test
though in the WB_SYNC_NONE case will return true, and the inode will end
up being marked dirty. Once that happens the inode will never be clean
until there's a WB_SYNC_ALL flush.

Fix this by immediately returning from nfs_commit_unstable_pages when
ncommit == 0.

Mike noticed this problem initially in RHEL5 (2.6.18-based kernel) which
has a backported version of 420e3646. The inode cache there was growing
very large. The inode cache was unable to be shrunk since the inodes
were all marked dirty. Calling sync() would essentially "fix" the
problem -- the WB_SYNC_ALL flush would result in the inodes all being
marked clean.

What I'm not clear on is how big a problem this is in mainline kernels
as the writeback code there is very different. Either way, it seems
incorrect to re-mark the inode dirty in this case.

Reported-by: Mike McLean <mikem@redhat.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

Revert "NFS: Ensure that writeback_single_inode() calls write_inode() when syncing"

commit 59b7c05fffba030e5d9e72324691e2f99aa69b79 upstream.

This reverts commit b80c3cb628f0ebc241b02e38dd028969fb8026a2.

The reverted commit was rendered obsolete by a VFS fix: commit
5547e8aac6f71505d621a612de2fca0dd988b439 (writeback: Update dirty flags in
two steps). We now no longer need to worry about writeback_single_inode()
missing our marking the inode for COMMIT in 'do_writepages()' call.

Reverting this patch, fixes a performance regression in which the inode
would continuously get queued to the dirty list, causing the writeback
code to unnecessarily try to send a COMMIT.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Simon Kirby <sim@hostway.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

kmod: prevent kmod_loop_msg overflow in __request_module()

commit 37252db6aa576c34fd794a5a54fb32d7a8b3a07a upstream.

Due to post-increment in condition of kmod_loop_msg in __request_module(),
the system log can be spammed by much more than 5 instances of the 'runaway
loop' message if the number of events triggering it makes the kmod_loop_msg
to overflow.

Fix that by making sure we never increment it past the threshold.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

Platform: Fix error path in samsung-laptop init

commit a7ea19926ffba86f373f6050a106cd162dbb9a78 upstream.

samsung_init() should not return success if not all devices are initialized.
Otherwise, samsung_exit() will dereference sdev NULL pointers and others.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

platform: samsung_laptop: fix samsung brightness min/max calculations

commit bee460be8c691c544e84ed678280ace6153104c6 upstream.

The min_brightness value of the sabi_config is incorrectly used in brightness
calculations. For the config where min_brightness = 1 and max_brightness = 8,
the user visible range should be 0 to 7 with hardware being set in the range
of 1 to 8. What is actually happening is that the user visible range is 0 to
8 with hardware being set in the range of -1 to 7.

This patch fixes the above issue as well as a miscalculation that would occur
in the case of min_brightness > 1.

Signed-off-by: Jason Stubbs <jasonbstubbs@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

Platform: samsung_laptop: samsung backlight for R528/R728

commit 093ed561648d43263c009ea88abab21a31cd4f1d upstream.

patch works for me, but I need to add "acpi_backlight=vendor" to kernel
params

Signed-off-by: Smelov Andrey <xor29a@bk.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

Platform: samsung_laptop: add support for X520 machines.

commit 7500eeb08a179e61a4219288c21407d63d1e9c64 upstream.

my samsung laptop would be very happy if you add
these lines to the file drivers/platform/x86/samsung-laptop.c

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

platform: samsung_laptop: add dmi information for Samsung R700 laptops

commit f87d02996f05ec1789ceecce9ec839f629b7aa80 upstream.

My DMI model is this:
>dmesg |grep DMI
[    0.000000] DMI present.
[    0.000000] DMI: SAMSUNG ELECTRONICS CO., LTD. SR700/SR700, BIOS
04SR 02/20/2008

adding dmi information of Samsung R700 laptops
This adds the dmi information of Samsungs R700 laptops.

Signed-off-by: Stefan Beller <stefanbeller@googlemail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

caif: Fix BUG() with network namespaces

commit 08613e4626c06ca408fc55071f6aedee36986a87 upstream.

The caif code will register its own pernet_operations, and then register
a netdevice_notifier. Each time the netdevice_notifier is triggered,
it'll do some stuff... including a lookup of its own pernet stuff with
net_generic().

If the net_generic() call ever returns NULL, the caif code will BUG().
That doesn't seem *so* unreasonable, I suppose — it does seem like it
should never happen.

However, it *does* happen. When we clone a network namespace,
setup_net() runs through all the pernet_operations one at a time. It
gets to loopback before it gets to caif. And loopback_net_init()
registers a netdevice... while caif hasn't been initialised. So the caif
netdevice notifier triggers, and immediately goes BUG().

We could imagine a complex and overengineered solution to this generic
class of problems, but this patch takes the simple approach. It just
makes caif_device_notify() *not* go looking for its pernet data
structures if the device it's being notified about isn't a caif device
in the first place.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Acked-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

kobj_uevent: Ignore if some listeners cannot handle message

commit ebf4127cd677e9781b450e44dfaaa1cc595efcaa upstream.

kobject_uevent() uses a multicast socket and should ignore
if one of listeners cannot handle messages or nobody is
listening at all.

Easily reproducible when a process in system is cloned
with CLONE_NEWNET flag.

(See also http://article.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt/5256)

Signed-off-by: Milan Broz <mbroz@redhat.com>
Acked-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>