kconfig/conf: fix randconfig setting multiple symbols in a choice

Currently, randconfig may set more than one symbol in a given choice.
Given this config file:
    config A
        bool "A"
    if A
    choice
        bool "B/C/D"
    config B
        bool "B"
    config C
        bool "C"
    config D
        bool "D"
    endchoice
    endif # A

Then randconfig generates such .config files (case where A is not set is not
shown below for brevity), and where only the right-most .config is valid:
  CONFIG_A=y                  CONFIG_A=y                  CONFIG_A=y
  CONFIG_B=y                  CONFIG_B=y                  CONFIG_B=y
  CONFIG_C=y                  # CONFIG_C is not set       # CONFIG_C is not set
  # CONFIG_D is not set       CONFIG_D=y                  # CONFIG_D is not set

That is, in a randomised choice, the first symbol is always selected,
and at most one other symbol may be selected.

This is due to symbol randomised in a choice not being properly flagged
as having a value.

Fix that by flagging those symbols adequately: have a user-defined value,
and be not valid (to force recalculation of the symbol).

Note: if the choice is not conditional, then the randomisation is properly
done.

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Matthieu CASTET <matthieu.castet@parrot.com>
[yann.morin.1998@free.fr: independently re-done the same patch as Matthieu,
                          as pointed out by Sedat]
Cc: Arnaud Lacombe <lacombar@gmail.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mconf/nconf: mark empty menus/menuconfigs different from non-empty ones

Submenus are sometimes empty and it would be nice if there is
something that notifies us that we should not expect any content
_before_ we enter a submenu.

A new function menu_is_empty() was introduced and empty menus and
menuconfigs are now marked by "----" as opposed to non-empty ones that
are marked by "--->".

This scheme was suggested by "Yann E. MORIN" <yann.morin.1998@free.fr>.

Signed-off-by: Dirk Gouders <dirk@gouders.net>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: Fix defconfig when one choice menu selects options that another choice menu depends on

The defconfig and Kconfig combination below, which is based on 3.10-rc4
Kconfigs, resulted in several options getting set to "m" instead of "y".

defconfig.choice:
---8<---
CONFIG_MODULES=y
CONFIG_USB_ZERO=y
---8<---

Kconfig.choice:
---8<---
menuconfig MODULES
bool "Enable loadable module support"

config CONFIGFS_FS
tristate "Userspace-driven configuration filesystem"

config OCFS2_FS
        tristate "OCFS2 file system support"
        depends on CONFIGFS_FS
        select CRC32

config USB_LIBCOMPOSITE
tristate
select CONFIGFS_FS

choice
tristate "USB Gadget Drivers"
default USB_ETH

config USB_ZERO
tristate "Gadget Zero (DEVELOPMENT)"
select USB_LIBCOMPOSITE

config USB_ETH
tristate "Ethernet Gadget (with CDC Ethernet support)"
select USB_LIBCOMPOSITE

endchoice

config CRC32
        tristate "CRC32/CRC32c functions"
        default y

choice
        prompt "CRC32 implementation"
        depends on CRC32
        default CRC32_SLICEBY8

config CRC32_SLICEBY8
        bool "Slice by 8 bytes"

endchoice
---8<---

$ scripts/kconfig/conf --defconfig=defconfig.choice Kconfig.choice

would result in:

.config:
---8<---
CONFIG_MODULES=y
CONFIG_CONFIGFS_FS=m
CONFIG_USB_LIBCOMPOSITE=m
CONFIG_USB_ZERO=m
CONFIG_CRC32=y
CONFIG_CRC32_SLICEBY8=y
---8<---

when the expected result would be:

.config:
---8<---
CONFIG_MODULES=y
CONFIG_CONFIGFS_FS=y
CONFIG_USB_LIBCOMPOSITE=y
CONFIG_USB_ZERO=y
CONFIG_CRC32=y
CONFIG_CRC32_SLICEBY8=y
---8<---

Signed-off-by: Arve Hjønnevåg <arve@android.com>
[yann.morin.1998@free.fr: add the resulting .config to commit log,
                          remove unneeded USB_GADGET from the defconfig]
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig/menu.c: fix multiple references to expressions in menu_add_prop()

menu_add_prop() applies upper menus' visibilities to actual prompts
by AND-ing the prompts visibilities with the upper menus ones.

This creates a further reference to the menu's visibilities and when
the expression reduction functions do their work, they may remove or
modify expressions that have multiple references, thus causing
unpredictable side-effects.

The following example Kconfig constructs a case where this causes
problems: a menu and a prompt which's visibilities depend on the same
symbol.  When invoking mconf with this Kconfig and pressing "Z" we
see a problem caused by a free'd expression still referenced by the
menu's visibility:

------------------------------------------------------------------------
mainmenu "Kconfig Testing Configuration"

config VISIBLE
def_bool n

config Placeholder
bool "Place holder"

menu "Invisible"
visible if VISIBLE

config TEST_VAR
bool "Test option" if VISIBLE

endmenu
------------------------------------------------------------------------

This patch fixes this problem by creating copies of the menu's
visibility expressions before AND-ing them with the prompt's one.

Signed-off-by: Dirk Gouders <dirk@gouders.net>
[yann.morin.1998@free.fr: move variable into its block-scope,
                          keep lines <80 chars, typo]
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

Revert "kconfig: fix randomising choice entries in presence of KCONFIG_ALLCONFIG"

This reverts commit 422c809f03f043d0950d8362214818e956a9daee.
It causes more harm than it solves issues.

Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: fix randomising choice entries in presence of KCONFIG_ALLCONFIG

Currently, randconfig does randomise choice entries, unless KCONFIG_ALLCONFIG
is specified.

For example, given those two files (Thomas' test-case):

    ---8<--- Config.test.in
    config OPTIONA
        bool "Option A"

    choice
        prompt "This is a choice"

    config CHOICE_OPTIONA
        bool "Choice Option A"

    config CHOICE_OPTIONB
        bool "Choice Option B"

    endchoice

    config OPTIONB
        bool "Option B"
    ---8<--- Config.test.in

    ---8<--- config.defaults
    CONFIG_OPTIONA=y
    ---8<--- config.defaults

And running:
    ./scripts/kconfig/conf --randconfig Config.test.in

does properly randomise the two choice symbols (and the two booleans).

However, running:
    KCONFIG_ALLCONFIG=config.defaults \
    ./scripts/kconfig/conf --randconfig Config.test.in

does *not* reandomise the two choice entries, and only CHOICE_OPTIONA
will ever be selected. (OPTIONA will always be set (expected), and
OPTIONB will be be properly randomised (expected).)

This patch defers setting that a choice has a value until a symbol for
that choice is indeed set, so that choices are properly randomised when
KCONFIG_ALLCONFIG is set, but not if a symbol for that choice is set.

Also, as a side-efect, this patch fixes the following case:

    ---8<---
    choice
    config OPTION_A
        bool "Option A"
    config OPTION_B
        bool "Option B"
    config OPTION_C
        bool "Option C"
    endchoice
    ---8<---

which could previously generate such .config files:

    ---8<---                            ---8<---
    CONFIG_OPTION_A=y                   CONFIG_OPTION_A=y
    CONFIG_OPTION_B=y                   # CONFIG_OPTION_B is not set
    # CONFIG_OPTION_C is not set        CONFIG_OPTION_C=y
    ---8<---                            ---8<---

Ie., the first entry in a choice is always set, plus zero or one of
the other options may be set.

This patch ensures that only one option may be set for a choice.

Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Arnaud Lacombe <lacombar@gmail.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 toolchain: update to gcc-4.8.2

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: fix NULL pointer dereference when searching a symbol

Searching for PPC_EFIKA results in a segmentation fault, and it's
because get_symbol_prop() returns NULL.

In this case CONFIG_PPC_EFIKA is defined in arch/powerpc/platforms/
52xx/Kconfig, so it won't be parsed if ARCH!=PPC, but menuconfig knows
this symbol when it parses sound/soc/fsl/Kconfig:

    config SND_MPC52xx_SOC_EFIKA
        tristate "SoC AC97 Audio support for bbplan Efika and STAC9766"
        depends on PPC_EFIKA

This bug was introduced by commit bcdedcc1afd6 ("menuconfig: print more
info for symbol without prompts").

Reported-and-tested-by: Borislav Petkov <bp@alien8.de>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Tested-by: Libo Chen <libo.chen@huawei.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: print more info for symbol without prompts

When we search a config symbol, if it has no prompt the position of this
symbol in the Kconfig file and it's dependencies are not printed.  This
can be inconvenient, especially when it's set to n and we want to find out
why.

the following is an example:

before:

Symbol: GENERIC_SMP_IDLE_THREAD [=y]
Type  : boolean
  Selected by: X86 [=y]

after:

Symbol: GENERIC_SMP_IDLE_THREAD [=y]
Type  : boolean
  Defined at arch/Kconfig:213
  Selected by: X86 [=y]

Signed-off-by: Weng Meiling <wengmeiling.weng@huawei.com>
Signed-off-by: Libo Chen <libo.chen@huawei.com>
Cc: Michal Marek <mmarek@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: fix lists definition for C++

The C++ compiler is more strict in that it refuses to assign
a void* to a struct list_head*.

Fix that by explicitly casting the poisonning constants.

(Tested with all 5 frontends, now.)

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

Revert "kconfig: fix randomising choice entries in presence of KCONFIG_ALLCONFIG"

This reverts commit 422c809f03f043d0950d8362214818e956a9daee.
It causes more harm than it solves issues.

Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: implement KCONFIG_PROBABILITY for randconfig

Currently the odds to set each symbol is (rounded):
    booleans:   y: 50%          n: 50%
    tristates:  y: 33%  m: 33%  n: 33%

Introduce a KCONFIG_PROBABILITY environment variable to tweak the
probabilities (in percentage), as such:
    KCONFIG_PROBABILITY     y:n split           y:m:n split
    -----------------------------------------------------------------
[1] unset or empty          50  : 50            33  : 33  : 34
[2] N                        N  : 100-N         N/2 : N/2 : 100-N
    N:M                     N+M : 100-(N+M)      N  :  M  : 100-(N+M)
    N:M:L                    N  : 100-N          M  :  L  : 100-(M+L)

[1] The current behaviour is kept as default, for backward compatibility
[2] The solution initially implemented by Peter for Buildroot, see:
    http://git.buildroot.org/buildroot/commit/?id=3435c1afb5

Signed-off-by: Peter Korsgaard <jacmet@uclibc.org>
[yann.morin.1998@free.fr: add to Documentation/]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: allow specifying the seed for randconfig

For reproducibility, it can be useful to be able to specify the
seed to use to seed the RNG.

Add a new KCONFIG_SEED environment variable which can be set to
the seed to use:
    $ make KCONFIG_SEED=42 randconfig
    $ sha1sum .config
    70a128c8dcc61303069e1be352cce64114dfcbca  .config
    $ make KCONFIG_SEED=42 randconfig
    $ sha1sum .config
    70a128c8dcc61303069e1be352cce64114dfcbca  .config

It's very usefull for eg. debugging the kconfig parser.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: fix randomising choice entries in presence of KCONFIG_ALLCONFIG

Currently, randconfig does randomise choice entries, unless KCONFIG_ALLCONFIG
is specified.

For example, given those two files (Thomas' test-case):

    ---8<--- Config.test.in
    config OPTIONA
        bool "Option A"

    choice
        prompt "This is a choice"

    config CHOICE_OPTIONA
        bool "Choice Option A"

    config CHOICE_OPTIONB
        bool "Choice Option B"

    endchoice

    config OPTIONB
        bool "Option B"
    ---8<--- Config.test.in

    ---8<--- config.defaults
    CONFIG_OPTIONA=y
    ---8<--- config.defaults

And running:
    ./config/conf --randconfig Config.test.in

does properly randomise the two choice symbols (and the two booleans).

However, running:
    KCONFIG_ALLCONFIG=config.defaults \
    ./config/conf --randconfig Config.test.in

does *not* reandomise the two choice entries, and only CHOICE_OPTIONA
will ever be selected. (OPTIONA will always be set (expected), and
OPTIONB will be be properly randomised (expected).)

This patch defers setting that a choice has a value until a symbol for
that choice is indeed set, so that choices are properly randomised when
KCONFIG_ALLCONFIG is set, but not if a symbol for that choice is set.

Also, as a side-efect, this patch fixes the following case:

    ---8<---
    choice
    config OPTION_A
        bool "Option A"
    config OPTION_B
        bool "Option B"
    config OPTION_C
        bool "Option C"
    endchoice
    ---8<---

which could previously generate such .config files:

    ---8<---                            ---8<---
    CONFIG_OPTION_A=y                   CONFIG_OPTION_A=y
    CONFIG_OPTION_B=y                   # CONFIG_OPTION_B is not set
    # CONFIG_OPTION_C is not set        CONFIG_OPTION_C=y
    ---8<---                            ---8<---

Ie., the first entry in a choice is always set, plus zero or one of
the other options may be set.

This patch ensures that only one option may be set for a choice.

Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Arnaud Lacombe <lacombar@gmail.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: do not override symbols already set

For randconfig, if a list of required symbols is specified with
KCONFIG_ALLCONFIG, such symbols do not "have a value" as per
sym_has_value(), but have the "valid" flag set.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: fix randconfig tristate detection

Because the modules' symbole (CONFIG_MODULES) may not yet be set when
we check a symbol's tristate capabilty, we'll always find that tristate
symbols are booleans, even if we randomly decided that to enable modules:
sym_get_type(sym) always return boolean for tristates when modules_sym
has not been previously set to 'y' *and* its value calculated *and* its
visibility calculated, both of which only occur after we randomly assign
values to symbols.

Fix that by looking at the raw type of symbols. Tristate set to 'm' will
be promoted to 'y' when their values will be later calculated.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: Add "breadcrumbs" navigation aid

Displays a trail of the menu entries used to get to the current menu.

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[yann.morin.1998@free.fr: small, trivial code re-ordering]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: Fix memory leak introduced by jump keys feature

Fixes the memory leak of struct jump_key allocated in get_prompt_str()

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: stable@vger.kernel.org
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: move crypto constants to ieee80211.h

mac80211 and the Intel drivers all define crypto
constants, move them to ieee80211.h instead.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: parse VHT channel switch IEs

VHT introduces multiple IEs that need to be parsed for a
wide bandwidth channel switch. Two are (currently) needed
in mac80211:
 * wide bandwidth channel switch element
 * channel switch wrapper element

The former is contained in the latter for beacons and probe
responses, but not for the spectrum management action frames
so the IE parser needs a new argument to differentiate them.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: define operating mode action frame

Define the action frame format, the VHT category
and its action types and the field format and EID
for operating mode notifications. The frame may
be used outside of VHT context as well, so don't
include "VHT" in the names.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: handle extended channel switch announcement

Handle the (public) extended channel switch announcement
action frames. Parts of the data in these frames isn't
really in IEs, but put it into the elems struct anyway
to simplify the handling.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: support secondary channel offset in CSA

Add support for the secondary channel offset IE in channel
switch announcements. This is necessary for proper handling
of CSA on HT access points.

For this to work it is also necessary to convert everything
here to use chandef structs instead of just channels. The
driver updates aren't really correct though. In particular,
the TI wl18xx driver update can't possibly be right since
it just ignores the new channel width for lack of firmware
API.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: support extended channel switch

Support extended channel switch when the operating
class is one of the global operating classes as
defined in Annex E of 802.11-2012. If it isn't,
disconnect from the AP instead.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: unify CSA action frame/beacon processing

CSA action frame content should be processed as variable IEs
rather than fixed to make it extensible. Unify the code and
process them just like CSA in beacons to make it easier to
extend for HT/VHT.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 toolchain: toolchain updates

 + update to gcc-4.8.1

 + delete source files which failed the sha256 check

 + automatic parallel build of the toolchain

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: parse Timeout Interval Element using a struct

Instead of open-coding the accesses and length check do
the length check in the IE parser and assign a struct
pointer for use in the remaining code.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

cfg80211: add P2P Notice of Absence attribute

Add P2P Notice of Absence attribute structure.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: update version to 1.9.8

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kbuild: Fix missing '\n' for NEW symbols in yes "" | make oldconfig >conf.new

According to Documentation/kbuild/kconfig.txt, the commands:

    yes "" | make oldconfig >conf.new
    grep "(NEW)" conf.new

should list the new config symbols with their default values.
However, currently there is no line break after each new symbol.  When
kconfig is interactive the user will type a new-line at this point,
but when non-interactive kconfig must print it.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Reference: http://bugs.debian.org/636029
[regid23@nt1.in: Adjusted Ben's work to apply cleanly to this tree]
Reported-and-tested-by: Regid Ichira <regid23@nt1.in>
Reviewed-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: Fix malloc handling in conf tools

(and get them out of the noise in the audit work)

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: get CONFIG_ prefix from the environment

Currently, the CONFIG_ prefix is hard-coded in the kconfig frontends
executables. This means that two projects that use kconfig with
different prefixes can not share the same kconfig frontends.

Instead of hard-coding the prefix in the frontends, get it from the
environment, and revert back to hard-coded value if not found.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: add a function to get the CONFIG_ prefix

Currently, we get the CONFIG_ prefix via the CONFIG_ macro, which means
the CONFIG_ prefix is hard-coded at compile time. This goes against
having a run-time defined CONFIG_ prefix.

Add a function that returns the CONFIG_ prefix to use (but keep the
current hard-coded behavior, to be changed in a later patch).

To avoid touching all the code that uses the CONFIG_ macro, we just
undef it, and define it to be a call to the function.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: always reset usb endpoints

It isn't clear what the firmware is supposed to do
when it is initializing the usb phy. The spec says
something about clearing the STALLED/HALTED flags
however the original firmware doesn't do that?!

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 toolchain: update to gcc-4.8.0

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: restrict peer's VHT capabilities to own

Implement restricting peer VHT capabilities to the device's own
capabilities. This is useful when a single driver supports more
than one device and the devices have different capabilities
(often they will differ in the number of spatial streams), but
in particular is also necessary for VHT capability overrides to
work correctly -- otherwise it'd be possible to e.g. advertise,
due to overrides, that TX-STBC is not supported, but then still
use it to TX to the AP because it supports RX-STBC.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: remove unused VHT MCS defines

There's an enum with the same values (but slightly
different names except for NOT_SUPPORTED) that is
actually used, so remove the defines.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

ieee80211: mark 802.11 related structs as being 2-byte aligned

Regardless of what header features they use, or if they align the IP
header or not, 802.11 packets from all drivers guarantee a 2-byte
alignment (and there's a debug WARN_ON in case they don't).

Annotate packet structs with __aligned(2) to allow the compiler to use
16-bit load/store operations on platforms with extremely inefficient
unaligned access (e.g. MIPS).

This reduces code size and improves performance on affected platforms
and causes no binary code change on others.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: move sequence number arithmetic to ieee80211.h

Move the sequence number arithmetic code from mac80211 to
ieee80211.h so others can use it. Also rename the functions
from _seq to _sn, they operate on the sequence number, not
the sequence_control field.

Also move macros to convert the sequence control to/from
the sequence number value from various drivers.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: advertise operating mode notification capability

Use the new extended capabilities advertising to advertise
the fact that operating mode notification is supported.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: constify IE parsing

Make all the parsed IE pointers const, and propagate
the change to all the users etc.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: implement endpoint halt req/clear and get endpoint status

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: support HT notify channel width action

Support the HT notify channel width action frame
to update the rate scaling about the bandwidth
the peer can receive in.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: use __packed in ieee80211.h

Use __packed instead of __attribute__((packed)).

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: more 'capability info' bits

define bits for 'capability info', as in recent spec edition
IEEE802.11-2012

Also, add mask for 2-bit field 'bss type', as it is in 802.11ad

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: improve link stability

This patch improves the link stability under load
by reordering and rewriting parts of the tx retry
procedure. It's all based on long hours of trail
and error, but sadly QoS + AMPDU is still not
working.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: fix CONFIG_CARL9170FW_DEBUG && !CONFIG_CARL9170FW_RADIO_FUNC

wlantx.c:262:2: error: struct has no member named psm

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: USB_MODESWITCH depends on USB_STANDARD_CMDS

usb.c:170:13: error: ‘usb_reset_eps’ defined but not used

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: fix CONFIG_CARL9170FW_DEBUG compile error

wlan.c:59:7: error: assignment discards ‘volatile’ qualifier from pointer target type

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: fix CONFIG_CARL9170FW_DEBUG_UART compile error

uart.c:36:9: error: ‘AR9170_UART_LINE_STS_TX_FIFO_ALMOST_EMPTY’ undeclared

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: fix !CONFIG_CARL9170FW_RADIO_FUNCTIONS compile errors

carlfw/src/main.c:87:27: error: unused variable ‘delta’
carlfw/src/main.c:87:15: error: unused variable ‘boff’
carlfw/src/main.c:90:12: error: ‘struct <anonymous>’ has no member named ‘state’

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: allow kconfig options in autogen.sh's parameters

kconf supports several interesting options:
 --allnoconfig    New config where all options are answered with no
 --allyesconfig   New config where all options are answered with yes
 --alldefconfig   New config with all symbols set to default
 --randconfig     New config with random answer to all options

to build the default firmware image, it is enough to run:
./autogen.sh --alldefconfig

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: fix out of bounds read

__carlfw_find_desc didn't check whenever the area for
a descriptor was within the file length. Also it could
read beyond the file while looking for a said descriptor.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: fix indention and add missing fi

This patch fixes my overzealous editing in
"carl9170: Add checksum during installation, as preferred by the driver"

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: Install binary in the parent directory, not /lib/firmware

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: Add checksum during installation, as preferred by the driver

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: Add and check SHA-256 sums for the toolchain tarballs

The sums for binutils and gcc are based on an HTTPS download (instead
of the default HTTP).

newlib doesn't seem to be available with any kind of signature, so I
compared a tarball and CVS checkout; let's hope they weren't both
compromised.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: Fix minimum CMake version

The CMAKE_CURRENT_LIST_DIR command was added in version 2.8.4.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: Add full licence texts to BSD-licenced CMake files

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: toolchain newlib 2.0.0

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: update version to 1.9.7

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: Replace CIRCLEQ by list_head-style lists.

sys/queue.h and CIRCLEQ in particular have proven to cause portability
problems (reported on Debian Sarge, Cygwin and FreeBSD)

Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Tested-by: Yaakov Selkowitz <yselkowitz@users.sourceforge.net>
Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: Assign jump keys per-page instead of globally

At the moment, keys 1-9 are assigned to the first 9 search results. This patch
makes them assigned to the first 9 results per-page instead. We are much less
likely to run out of keys that way.

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

menuconfig: Add jump keys to search results

makes it possible to jump directly to the menu for a configuration entry after
having searched for it with '/'. If this menu is not currently accessible we
jump to the nearest accessible parent instead. After exiting this menu, the
user is returned to the search results where he may jump further in or
elsewhere.

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: fix VHT max AMPDU exponent definition

This is really a 3-bit field, not a single bit,
so declare a mask and shift. Also fix hwsim, it
advertises the maximum possible.

While at it reindent all the defines using tabs
instead of spaces.

Change-Id: I7cd81c0d72f76deb5010aba5bfa3dd312006e898
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: support VHT association

Determine the VHT channel from the AP's VHT operation IE
(if present) and configure the hardware to that channel
if it is supported. If channel contexts cause a channel
to not be usable, try a smaller bandwidth.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: add definitions for VHT MCS support

Add definitions for the VHT MCS support values that
are used to indicate, for each number of streams
(1 through 8) which MCSes are supported.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: move Mesh Capability field definition to ieee80211.h

Signed-off-by: Marco Porsch <marco.porsch@etit.tu-chemnitz.de>
[prefix with IEEE80211_]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: introduce IEEE80211_NUM_TIDS and use it

Introduce IEEE80211_NUM_TIDS in the generic 802.11
header file and use it in place of STA_TID_NUM and
NUM_RX_DATA_QUEUES which are both really the number
of TIDs.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: fix max beacon length

According to AR9170's data sheet, the
beacon length register only cares about
bits 8:0. So the max length [with 4
reserved bytes for the FCS] is at 511.

Note:

Because the beacon has to be stored in the
DMA memory region, we have to be careful
so we don't waste precious memory due to
alignment and padding. Therefore I decided
to reduce the max beacon length even more
down to 480 bytes. [This saves one 320 byte
DMA block from becoming unused padding, when
the default 2 vif configuration is selected].

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: add peer-to-peer related definitions

The Peer-to-Peer IE is vendor-specific IE identified by WiFi Alliance
OUI and specific P2P OUI type. The payload of this IE consists of
so-called P2P attributes. This patch adds definitions for processing
these attributes.

Reviewed-by: Hante Meuleman <meuleman@broadcom.com>
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Franky (Zhenhui) Lin <frankyl@broadcom.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware toolchain: binutils 2.23.1

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: remove duplicate enum ieee80211_eid definitions

WLAN_EID_WPA and WLAN_EID_GENERIC mapped to the same value
as WLAN_EID_VENDOR_SPECIFIC. The last one being more in line
with the standard specification. Removing WLAN_EID_WPA and
WLAN_EID_GENERIC as there are no longer drivers using these.

Cc: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

ieee80211: Rename VHT cap struct

Rename struct ieee80211_vht_capabilities to ieee80211_vht_cap
and renamed its member vht_capabilities_info to vht_cap_info.

Signed-off-by: Mahesh Palivela <maheshp@posedge.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: remove obsolete chan no/center freq conversion functions

There are a number of functions that shouldn't really
be used when modern functions that take the band are
available in cfg80211. Remove these, but for now keep
 * ieee80211_freq_to_dsss_chan and
 * ieee80211_dsss_chan_to_freq
as they're used in older drivers.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware toolchain: binutils 2.23

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware toolchain: various small updates

 - gnu tar can automatically decides the decompression algo.

 - because we use the filename (*_TAR) as a target it better
   be the same as in the fetch (*_URL)

 - stringify URLs.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: add note about rx filter usage

This patch adds a note about what it means when
a rx filter bit is set. Because unlike what most
people assume, it means that the data which matches
the condition is discarded and not
"passed to the host".

Reported-by: Javier Lopez <jlopex@cozybit.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: remove inline attributes

The compiler is smart enough to decide whenever
these functions can be inlined or not.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 toolchain: update to gcc-4.7.2

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware tools: update .gitignore

 - eeprom_fix is no more

 - ignore fwprepare binary

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: allow long lines in config file

For some config options (CONFIG_EXTRA_FIRMWARE, for example), the length
of a config file line can exceed the 1024 byte buffer.

Switch from fgets to compat_getline to fix. compat_getline is an
internally implimented getline work-alike for portability purposes.

Signed-off-by: Cody Schafer <cody@linux.vnet.ibm.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: delete last traces of __enabled_ from autoconf.h

We've now fixed IS_ENABLED() and friends to not require any special
"__enabled_" prefixed versions of the normal Kconfig options, so delete
the last traces of them being generated.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

Revert "kconfig: fix __enabled_ macros definition for invisible and un-selected symbols"

Dumping two lines into autoconf.h for all existing Kconfig options
results in a giant file (~16k lines) we have to process each time we
compile something.  We've weaned IS_ENABLED() and similar off of
requiring the __enabled_ definitions so now we can revert the change
which caused all the extra lines.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: extract tx and rx code from wlan.c into separate files

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: remove CONFIG_CARL9170FW_CAB_QUEUE

The software supported cab queue is now a standard
feature and will always be available.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: move cmd.c into hostif.c

The command handler is part of the host
application interface too, so it will fit
there nicely.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 firmware: refactor timer functions

Thanks to LTO, the timer routines can be moved into
a proper context without added overhead at runtime.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: remove eeprom_fix utility

The carl9170 driver in the kernel doesn't
have the capability anymore. Hence, this
tool is no longer needed.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: link time optimization build

Enabling LTO will hopefully lead to smaller
firmware binaries and maybe even slightly
better performance.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170 toolchain: update toolchain disk space requirements

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kbuild: all{no,yes,mod,def,rand}config only read files when instructed to.

Prevent subtle surprises to both people working on the kconfig code
and people using make allnoconfig allyesconfig allmoconfig and
randconfig by only attempting to read a config file if
KCONFIG_ALLCONFIG is set.

Common sense suggests attempting to read the extra config files does
not make sense unless requested.  The documentation says the code
won't attempt to read the extra config files unless requested.
Current usage does not appear to include people depending on the code
reading the config files without the variable being set So do the
simple thing and stop reading config files when passed
all{no,yes,mod,def,rand}config unless KCONFIG_ALLCONFIG environment
variable is set.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

kconfig: Add error handling to KCONFIG_ALLCONFIG

- Only try to read the file specified if KCONFIG_ALL_CONFIG is set to
  something other than the empty string or "1".

- Don't use stat to check the name passed to conf_read_simple so that
  zconf_fopen can find the file in the current directory or in SRCTREE
  removing a extremely source of confusing failure, where KCONFIG_ALL_CONFIG
  was not interpreted with respect to the directory make was called in.

- If conf_read_simple fails complain clearly and stop processing.
  Allowing the simple debugging of typos.

- Clearly document the behavior so it is clear to users which
  values are treated as flags and which values are treated as
  filenames.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

carl9170: remove carlu

The userspace utility will be become a separate project.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: 60g protocol constants

Provide various constants as defined by the 802.11ad:
frame types, IE's, capability bits, action categories

Introduce GCMP cipher, mandatory by 802.11ad

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

wireless: add VHT (802.11ac) definitions

Add the VHT definitions to be used by drivers supporting it.

Signed-off-by: Mahesh Palivela <maheshp@posedge.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

ieee80211: more OUI type definitions for WLAN_OUI_MICROSOFT

WMM and WPS

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

mac80211: implement the proactive PREQ generation

Generate the proactive PREQ element as defined in
Sec. 13.10.9.3 (Case C) of IEEE Std. 802.11-2012
based on the selection of dot11MeshHWMPRootMode as follow:
dot11MeshHWMPRootMode (2) is proactivePREQnoPREP
dot11MeshHWMPRootMode (3) is proactivePREQwithPREP

The proactive PREQ is generated based on the interval
defined by dot11MeshHWMProotInterval.

With this change, proactive RANN element is now generated
if the dot11MeshHWMPRootMode is set to (4) instead of (1).

Signed-off-by: Chun-Yeow Yeoh <yeohchunyeow@gmail.com>
[line-break commit log]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

ieee80211: definitions for Microsoft Vendor OUI and WPA OUI type

Reference: http://standards.ieee.org/develop/regauth/oui/oui.txt

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>

{nl,cfg,mac}80211: fix the coding style related to mesh parameters

fix the coding style related to mesh parameters, especially the indentation,
as pointed out by Johannes Berg.

Signed-off-by: Chun-Yeow Yeoh <yeohchunyeow@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>