linux-stable.git/include, branch v3.18.17

kernel: make READ_ONCE() valid on const arguments

2015-06-28T17:39:28+00:00

[ Upstream commit dd36929720f40f17685e841ae0d4c581c165ea60 ]

The use of READ_ONCE() causes lots of warnings witht he pending paravirt
spinlock fixes, because those ends up having passing a member to a
'const' structure to READ_ONCE().

There should certainly be nothing wrong with using READ_ONCE() with a
const source, but the helper function __read_once_size() would cause
warnings because it would drop the 'const' qualifier, but also because
the destination would be marked 'const' too due to the use of 'typeof'.

Use a union of types in READ_ONCE() to avoid this issue.

Also make sure to use parenthesis around the macro arguments to avoid
possible operator precedence issues.

Tested-by: Ingo Molnar 
Cc: Christian Borntraeger 
Signed-off-by: Linus Torvalds 
Signed-off-by: Sasha Levin

[media] Add and use IS_REACHABLE macro

2015-06-28T17:39:23+00:00

[ Upstream commit 9b174527e7b756cda9f5d9e541f87b7fec9cfdf0 ]

In the media drivers, the v4l2 core knows about all submodules
and calls into them from a common function. However this cannot
work if the modules that get called are loadable and the
core is built-in. In that case we get

drivers/built-in.o: In function `set_type':
drivers/media/v4l2-core/tuner-core.c:301: undefined reference to `tea5767_attach'
drivers/media/v4l2-core/tuner-core.c:307: undefined reference to `tea5761_attach'
drivers/media/v4l2-core/tuner-core.c:349: undefined reference to `tda9887_attach'
drivers/media/v4l2-core/tuner-core.c:405: undefined reference to `xc4000_attach'

This was working previously, until the IS_ENABLED() macro was used
to replace the construct like

 #if defined(CONFIG_DVB_CX24110) || (defined(CONFIG_DVB_CX24110_MODULE) && defined(MODULE))

with the difference that the new code no longer checks whether it is being
built as a loadable module itself.

To fix this, this new patch adds an 'IS_REACHABLE' macro, which evaluates
true in exactly the condition that was used previously. The downside
of this is that this trades an obvious link error for a more subtle
runtime failure, but it is clear that the change that introduced the
link error was unintentional and it seems better to revert it for
now. Also, a similar change was originally created by Trent Piepho
and then reverted by teh change to the IS_ENABLED macro.

Ideally Kconfig would be used to avoid the case of a broken dependency,
or the code restructured in a way to turn around the dependency, but either
way would require much larger changes here.

Fixes: 7b34be71db53 ("[media] use IS_ENABLED() macro")
See-also: c5dec9fb248e ("V4L/DVB (4751): Fix DBV_FE_CUSTOMISE for card drivers compiled into kernel")

Signed-off-by: Arnd Bergmann 
Signed-off-by: Mauro Carvalho Chehab 
Signed-off-by: Sasha Levin

jhash: Update jhash_[321]words functions to use correct initval

2015-06-28T17:39:22+00:00

[ Upstream commit 2e7056c433216f406b90a003aa0ba42e19d3bdcf ]

Looking over the implementation for jhash2 and comparing it to jhash_3words
I realized that the two hashes were in fact very different.  Doing a bit of
digging led me to "The new jhash implementation" in which lookup2 was
supposed to have been replaced with lookup3.

In reviewing the patch I noticed that jhash2 had originally initialized a
and b to JHASH_GOLDENRATIO and c to initval, but after the patch a, b, and
c were initialized to initval + (length << 2) + JHASH_INITVAL.  However the
changes in jhash_3words simply replaced the initialization of a and b with
JHASH_INITVAL.

This change corrects what I believe was an oversight so that a, b, and c in
jhash_3words all have the same value added consisting of initval + (length
<< 2) + JHASH_INITVAL so that jhash2 and jhash_3words will now produce the
same hash result given the same inputs.

Fixes: 60d509c823cca ("The new jhash implementation")
Signed-off-by: Alexander Duyck 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

net: add sk_fullsock() helper

2015-06-28T17:39:21+00:00

[ Upstream commit 1d0ab253872cdd3d8e7913f59c266c7fd01771d0 ]

We have many places where we want to check if a socket is
not a timewait or request socket. Use a helper to avoid
hard coding this.

Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

inet: add TCP_NEW_SYN_RECV state

2015-06-28T17:39:21+00:00

[ Upstream commit 10feb428a5045d5eb18a5d755fbb8f0cc9645626 ]

TCP_SYN_RECV state is currently used by fast open sockets.

Initial TCP requests (the pseudo sockets created when a SYN is received)
are not yet associated to a state. They are attached to their parent,
and the parent is in TCP_LISTEN state.

This commit adds TCP_NEW_SYN_RECV state, so that we can convert
TCP stack to a different schem gradually.

This state is not exported to user space.

Signed-off-by: Eric Dumazet 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

Driver core: Unified device properties interface for platform firmware

2015-06-28T17:39:20+00:00

[ Upstream commit b31384fa5de37a100507751dfb5c0a49d06cee67 ]

Add a uniform interface by which device drivers can request device
properties from the platform firmware by providing a property name
and the corresponding data type.  The purpose of it is to help to
write portable code that won't depend on any particular platform
firmware interface.

The following general helper functions are added:

device_property_present()
device_property_read_u8()
device_property_read_u16()
device_property_read_u32()
device_property_read_u64()
device_property_read_string()
device_property_read_u8_array()
device_property_read_u16_array()
device_property_read_u32_array()
device_property_read_u64_array()
device_property_read_string_array()

The first one allows the caller to check if the given property is
present.  The next 5 of them allow single-valued properties of
various types to be retrieved in a uniform way.  The remaining 5 are
for reading properties with multiple values (arrays of either numbers
or strings).

The interface covers both ACPI and Device Trees.

This change set includes material from Mika Westerberg and Aaron Lu.

Signed-off-by: Aaron Lu 
Signed-off-by: Mika Westerberg 
Acked-by: Greg Kroah-Hartman 
Acked-by: Grant Likely 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

ACPI: Add support for device specific properties

2015-06-28T17:39:20+00:00

[ Upstream commit ffdcd955c3078af3ce117edcfce80fde1a512bed ]

Device Tree is used in many embedded systems to describe the system
configuration to the OS. It supports attaching properties or name-value
pairs to the devices it describe. With these properties one can pass
additional information to the drivers that would not be available
otherwise.

ACPI is another configuration mechanism (among other things) typically
seen, but not limited to, x86 machines. ACPI allows passing arbitrary
data from methods but there has not been mechanism equivalent to Device
Tree until the introduction of _DSD in the recent publication of the
ACPI 5.1 specification.

In order to facilitate ACPI usage in systems where Device Tree is
typically used, it would be beneficial to standardize a way to retrieve
Device Tree style properties from ACPI devices, which is what we do in
this patch.

If a given device described in ACPI namespace wants to export properties it
must implement _DSD method (Device Specific Data, introduced with ACPI 5.1)
that returns the properties in a package of packages. For example:

	Name (_DSD, Package () {
		ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"),
		Package () {
			Package () {"name1", },
			Package () {"name2", },
			...
		}
	})

The UUID reserved for properties is daffd814-6eba-4d8c-8a91-bc9bbf4aa301
and is documented in the ACPI 5.1 companion document called "_DSD
Implementation Guide" [1], [2].

We add several helper functions that can be used to extract these
properties and convert them to different Linux data types.

The ultimate goal is that we only have one device property API that
retrieves the requested properties from Device Tree or from ACPI
transparent to the caller.

[1] http://www.uefi.org/sites/default/files/resources/_DSD-implementation-guide-toplevel.htm
[2] http://www.uefi.org/sites/default/files/resources/_DSD-device-properties-UUID.pdf

Reviewed-by: Hanjun Guo 
Reviewed-by: Josh Triplett 
Reviewed-by: Grant Likely 
Signed-off-by: Darren Hart 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Mika Westerberg 
Signed-off-by: Rafael J. Wysocki 
Signed-off-by: Sasha Levin

tcp: fix child sockets to use system default congestion control if not set

2015-06-15T18:26:53+00:00

[ Upstream commit 9f950415e4e28e7cfae2e416b43e862e8101d996 ]

Linux 3.17 and earlier are explicitly engineered so that if the app
doesn't specifically request a CC module on a listener before the SYN
arrives, then the child gets the system default CC when the connection
is established. See tcp_init_congestion_control() in 3.17 or earlier,
which says "if no choice made yet assign the current value set as
default". The change ("net: tcp: assign tcp cong_ops when tcp sk is
created") altered these semantics, so that children got their parent
listener's congestion control even if the system default had changed
after the listener was created.

This commit returns to those original semantics from 3.17 and earlier,
since they are the original semantics from 2007 in 4d4d3d1e8 ("[TCP]:
Congestion control initialization."), and some Linux congestion
control workflows depend on that.

In summary, if a listener socket specifically sets TCP_CONGESTION to
"x", or the route locks the CC module to "x", then the child gets
"x". Otherwise the child gets current system default from
net.ipv4.tcp_congestion_control. That's the behavior in 3.17 and
earlier, and this commit restores that.

Fixes: 55d8694fa82c ("net: tcp: assign tcp cong_ops when tcp sk is created")
Cc: Florian Westphal 
Cc: Daniel Borkmann 
Cc: Glenn Judd 
Cc: Stephen Hemminger 
Signed-off-by: Neal Cardwell 
Signed-off-by: Eric Dumazet 
Signed-off-by: Yuchung Cheng 
Acked-by: Daniel Borkmann 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

sctp: Fix mangled IPv4 addresses on a IPv6 listening socket

2015-06-15T18:26:52+00:00

[ Upstream commit 9302d7bb0c5cd46be5706859301f18c137b2439f ]

sctp_v4_map_v6 was subtly writing and reading from members
of a union in a way the clobbered data it needed to read before
it read it.

Zeroing the v6 flowinfo overwrites the v4 sin_addr with 0, meaning
that every place that calls sctp_v4_map_v6 gets ::ffff:0.0.0.0 as the
result.

Reorder things to guarantee correct behaviour no matter what the
union layout is.

This impacts user space clients that open an IPv6 SCTP socket and
receive IPv4 connections. Prior to 299ee user space would see a
sockaddr with AF_INET and a correct address, after 299ee the sockaddr
is AF_INET6, but the address is wrong.

Fixes: 299ee123e198 (sctp: Fixup v4mapped behaviour to comply with Sock API)
Signed-off-by: Jason Gunthorpe 
Acked-by: Daniel Borkmann 
Acked-by: Neil Horman 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

xfrm: release dst_orig in case of error in xfrm_lookup()

2015-06-10T17:42:46+00:00

[ Upstream commit ac37e2515c1a89c477459a2020b6bfdedabdb91b ]

dst_orig should be released on error. Function like __xfrm_route_forward()
expects that behavior.
Since a recent commit, xfrm_lookup() may also be called by xfrm_lookup_route(),
which expects the opposite.
Let's introduce a new flag (XFRM_LOOKUP_KEEP_DST_REF) to tell what should be
done in case of error.

Fixes: f92ee61982d("xfrm: Generate blackhole routes only from route lookup functions")
Signed-off-by: huaibin Wang 
Signed-off-by: Nicolas Dichtel 
Signed-off-by: Steffen Klassert 
Signed-off-by: Sasha Levin