Refactor base port operations into new file; 8250_port.c.

Legacy irq handling, RSA port support, port storage for universal
driver, driver definition, module parameters and linkage remain in
8250_core.c

The source file split and resulting modules is diagrammed below:

  8250_core.c ====>   8250_core.c __
              \                     \
               \                     +-- 8250.ko (alias 8250_core)
                \     8250_pnp.c  __/     (universal driver)
                 \
                  =>  8250_port.c __
                                    \
                                     +-- 8250_base.ko
                      8250_dma.c  __/     (port operations)
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

If the UART has been in use before this driver was loaded, IRQs might be
active and get fired as soon as we set the handler, which will crash
in the spin_lock_irqsave(&sport->port.lock, flags) because port.lock is
not initialized until the port is added at the end of probe.
Signed-off-by: NDavid Jander <david@protonic.nl>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

To prevent problems with interrupt latency, and due to the fact, that
the error will be counted anyway (icount.overrun), the dev_err is simply
removed.

Background:
If an rx-fifo overflow occurs a dev_err message was called in interrupt
context. Since dev_err messages are written to console in a synchronous way
(unbuffered), and console may be a serial terminal, this leads to a
highly increased interrupt-latency (several milliseconds).
As a result of the high latency more rx-fifo overflows will happen, and
therefore a feedback loop of errors is created.
Signed-off-by: NManfred Schlaegl <manfred.schlaegl@gmx.at>
Acked-By: NAlexander Stein <alexander.stein@systec-electronic.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

As can be seen in function uart_insert_char (serial_core) the element
buf_overrun of struct uart_icount is used to count overruns of
tty-buffer.
Added support for counting of overruns in imx driver analogue to
serial_core.
Signed-off-by: NManfred Schlaegl <manfred.schlaegl@gmx.at>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Disable interrupts before requesting them in order to fix a kernel oops
after lauching a kernel via kexec.

Tested on a imx6sl-evk board.
Signed-off-by: NFabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Being a soft core, it can be located not only on PPC or Microblaze
platforms.

Since the driver already does endianness detection we only need to
change the Kconfig to use it in other arches.

This is also done in other softcores as xilinx-spi.
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

SysRq support activation depends on CONFIG_SERIAL_STM32_USART_CONSOLE, but
this config flag does not exists.

This patch fix this by depending on the valid config flag, which is
SERIAL_STM32_CONSOLE.
Reported-by: NAndreas Ruprecht <andreas.ruprecht@fau.de>
Signed-off-by: NMaxime Coquelin <mcoquelin.stm32@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

While UART work in DMA mode, function start_rx will request descriptor
from DMA engine, if there is no left descriptor UART, driver will give
err logs "DMA slave single fail".

currently start_rx is called in set_termios function, so everytime, port
setting will call start_rx once.

Now put start_rx in startup, it will be called once while open the port.
Signed-off-by: NQipan Li <Qipan.Li@csr.com>
Signed-off-by: NBarry Song <Baohua.Song@csr.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

when UART works in DMA mode and left bytes in rx fifo less than
a dma transfer unit, DMA engine can't transfer the bytes out
to rx DMA buffer. so it need a way to fetch them out and
flush them into tty buffer in time.

in the above case, we want UART switch from DMA mode to PIO mode and
fetch && flush bytes into tty layer buffer until rxfifo become empty,
after that done let UART switch from PIO mode back to DMA mode.
(record as method1)

method1 result in the next receive result wrong. for example in PIO part
of method1, we fetched && pushed X1...X3 bytes, when UART rxfifo newly
received Y1...Y4 bytes, UART trigger a DMA unit transfer, the DMA unit's
content is X1...X3Y1 and rxfifo fifo status is empty, so X1X2X3 pushed
twice by PIO way and DMA way also the bytes Y2Y3Y4 missed. add rxfifo
reset operation before UART switch back to DMA mode would resolve the
issue. ([method1 + do fifo reset] record as method2)

before the commit, UART driver use method2. but methd2 have a risk of
data loss, as if UART's shift register receive a complete byte and
transfer it into rxfifo before rxfifo reset operation the byte will
loss.

UART and USP have the similar bits CLEAR_RX_ADDR_EN(uart)/FRADDR_CLR_EN(usp),
When found UART controller changing I/O to DMA mode, UART controller
clears the two low bits of read point (rx_fifo_addr[1:0]).
when enable the bit + method1(record as method3), in above example
the DMA unit's content is X1...X3Y1 and there are Y2Y3Y4 in rxfifo by
experiment, we just push bytes in rx DMA buffer.

BTW, the workaround works only for UART receive DMA channel use SINGLE
DMA mode.
Signed-off-by: NQipan Li <Qipan.Li@csr.com>
Signed-off-by: NBarry Song <Baohua.Song@csr.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The implementation of cris_console_device() is exactly the same as
uart_console_device(), so let's switch to use uart_console_device().
Signed-off-by: NAxel Lin <axel.lin@ingics.com>
Acked-by: NNiklas Cassel <nks@flawful.org>
Acked-by: NJesper Nilsson <jesper.nilsson@axis.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

OMAP h/w-assisted IXON flow control is borked. The transmitter becomes
stuck if XON is never received; clearing the fifos or resetting the
rx flow control bits has no effect.

Remove auto-IXANY as well, since without auto-IXON, it has no purpose.
Signed-off-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Due to Advisory 21 as documented in AM437x errata document,
UART module cannot be disabled once DMA is used. The only
workaround is to softreset the module before disabling it.

DRA7x UARTs are compatible to AM437x UARTs in terms of
this errata and prescribed workaround.

Enable usage of workaround for this errata on DRA7x SoCs.
Signed-off-by: NSekhar Nori <nsekhar@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

AM335x, AM437x and DRA7x SoCs have an errata[1] due to which UART
cannot be idled after it has been used with DMA.

OMAP3 has a similar sounding errata which has been worked around
in a2fc3661 ("ARM: OMAP3: Use manual idle for UARTs
because of DMA errata"). But the workaround used there does not
apply to AM335x, AM437x SoCs.

After using DMA, the UART module on these SoCs must be soft reset
to go to idle.

This patch implements that errata workaround. It is expected that
UART will be used with DMA so no explicit check for DMA usage
has been added for errata applicability.

MDR1 register needs to be restored right after soft-reset because
"UART mode" must be set in that register for module wake-up on AM335x
to work. As a result, SCR register is now used to determine if
context was lost during sleep.

[1] See Advisory 21 in AM437x errata SPRZ408B, updated April 2015.
    http://www.ti.com/lit/er/sprz408b/sprz408b.pdfSigned-off-by: NSekhar Nori <nsekhar@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Use of of_machine_is_compatible() for handling AM335x specific
"DMA kick" quirk in 8250_omap driver makes it ugly to extend the
quirk for other platforms. Instead use a new compatible.

The new compatible will also make it easier to take care of
other quirks on AM335x and like SoCs.

In order to not break backward DTB compatibility for users of
8250_omap driver on AM335x based boards, existing use of
of_machine_is_compatible() has not been removed.
Signed-off-by: NSekhar Nori <nsekhar@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The silicon errata[1] workaround implemented in a follow-on
patch, "serial: 8250_omap: workaround errata on disabling
UART after using DMA", requires MDR1 register programming.

Extract MDR1 register update into helper function,
omap8250_update_mdr1() to help with that.

[1] Advisory 21 in http://www.ti.com/lit/er/sprz408b/sprz408b.pdfSigned-off-by: NSekhar Nori <nsekhar@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

omap_device infrastructure has a suspend_noirq hook which
runtime suspends all devices late in the suspend cycle (see
_od_suspend_noirq() in arch/arm/mach-omap2/omap_device.c)

This leads to a NULL pointer exception in 8250_omap driver
since by the time omap8250_runtime_suspend() is called, 8250_dma
driver has already set rxchan to NULL via serial8250_release_dma().

Make an explicit check to see if rxchan is NULL in
runtime_{suspend|resume} hooks to fix this.
Signed-off-by: NSekhar Nori <nsekhar@ti.com>
Acked-by: NTony Lindgren <tony@atomide.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Fintek chip can be configured for io addresses different than the standard.

Query the chip for the configured addresses and try to match it with the
pnp address.
Reported-by: NPeter Hong <peter_hong@fintek.com.tw>
Reviewed-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Chip can be configured to use entry key different than 0x77. Try all the
valid keys until one gives out the right chip id.
Reported-by: NPeter Hong <peter_hong@fintek.com.tw>
Reviewed-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

There are some chips with the same interface but different chip ip.
Reported-by: NPeter Hong <peter_hong@fintek.com.tw>
Reviewed-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Fintek chip can be connected at address 0x4e and also 0x2e.
Add some logic to find out the address of the chip.
Reported-by: NPeter Hong <peter_hong@fintek.com.tw>
Reviewed-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Save the port index and the line id in a private structure.
Reported-by: NPeter Hong <peter_hong@fintek.com.tw>
Reviewed-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NRicardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In little endian cases, macro htons unfolds to __swab16 which
provides special case for constants. In big endian cases,
__constant_htons and htons expand directly to the same expression.
So, replace __constant_htons with htons with the goal of getting
rid of the definition of __constant_htons completely.

The semantic patch that performs this transformation is as follows:

@@expression x;@@

- __constant_htons(x)
+ htons(x)
Signed-off-by: NVaishali Thakkar <vthakkar1994@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The zonelist locking and the oom_sem are two overlapping locks that are
used to serialize global OOM killing against different things.

The historical zonelist locking serializes OOM kills from allocations with
overlapping zonelists against each other to prevent killing more tasks
than necessary in the same memory domain.  Only when neither tasklists nor
zonelists from two concurrent OOM kills overlap (tasks in separate memcgs
bound to separate nodes) are OOM kills allowed to execute in parallel.

The younger oom_sem is a read-write lock to serialize OOM killing against
the PM code trying to disable the OOM killer altogether.

However, the OOM killer is a fairly cold error path, there is really no
reason to optimize for highly performant and concurrent OOM kills.  And
the oom_sem is just flat-out redundant.

Replace both locking schemes with a single global mutex serializing OOM
kills regardless of context.
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NDavid Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

There is much SCI of SoC having within,
and the register size is also different in everyone.
So get from platform device.
Signed-off-by: NYoshinori Sato <ysato@users.sourceforge.jp>
Acked-by: Nby: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: NSimon Horman <horms+renesas@verge.net.au>

Signed-off-by: NYoshinori Sato <ysato@users.sourceforge.jp>
Acked-by: NGeert Uytterhoeven <geert+renesas@glider.be>

Introduce a driver suitable for use with the UARTs present in
Ingenic SoCs such as the JZ4740 & JZ4780. These are described as being
ns16550 compatible but aren't quite - they require the setting of an
extra bit in the FCR register to enable the UART module. The serial_out
implementation is the same as that in arch/mips/jz4740/serial.c - which
will shortly be removed.
Signed-off-by: NPaul Burton <paul.burton@imgtec.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: Lars-Peter Clausen <lars@metafoo.de>
Cc: linux-serial@vger.kernel.org
Cc: linux-mips@linux-mips.org
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: Alan Cox <alan@linux.intel.com>
Cc: linux-kernel@vger.kernel.org
Cc: Matthias Brugger <matthias.bgg@gmail.com>
Cc: Ricardo Ribalda Delgado <ricardo.ribalda@gmail.com>
Cc: Tony Lindgren <tony@atomide.com>
Cc: John Crispin <blogic@openwrt.org>
Patchwork: https://patchwork.linux-mips.org/patch/10159/Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

Add a MIPS specific SysRq operation to dump the TLB entries on all CPUs,
using the 'x' trigger key.
Signed-off-by: NJames Hogan <james.hogan@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10072/Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

In addition to defining triggers for VT LED states, let's define triggers
for VT keyboard lock states, such as "kbd-shiftlock", "kbd-altgrlock", etc.

This permits to fix #7063 from userland by using a modifier to implement
proper CapsLock behavior and have the keyboard caps lock led show that
modifier state.
Signed-off-by: NSamuel Thibault <samuel.thibault@ens-lyon.org>
Tested-by: NPavel Machek <pavel@ucw.cz>
Acked-by: NPavel Machek <pavel@ucw.cz>
Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com>

Now that input core allows controlling keyboards LEDs via standard LED
subsystem triggers let's switch VT keyboard code to make use of this
feature. We will define the following standard triggers: "kbd-scrollock",
"kbd-numlock", "kbd-capslock", and "kbd-kanalock" which are default
triggers for respective LEDs on keyboards.
Signed-off-by: NSamuel Thibault <samuel.thibault@ens-lyon.org>
Tested-by: NPavel Machek <pavel@ucw.cz>
Acked-by: NPavel Machek <pavel@ucw.cz>
Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com>

The metag_da TTY driver can't get built as a module at the moment, but
it still uses module_init() and module_exit(). Those macros are moving
to module.h which isn't included by metag_da.c, which will result in the
following build warnings (remarkably no build errors) and an apparent
failure to boot as the TTY driver won't be loaded.

drivers/tty/metag_da.c:660: warning: data definition has no type or storage class
drivers/tty/metag_da.c:660: warning: type defaults to ‘int’ in declaration of ‘module_init’
drivers/tty/metag_da.c:660: warning: parameter names (without types) in function declaration
drivers/tty/metag_da.c:661: warning: data definition has no type or storage class
drivers/tty/metag_da.c:661: warning: type defaults to ‘int’ in declaration of ‘module_exit’
drivers/tty/metag_da.c:661: warning: parameter names (without types) in function declaration
drivers/tty/metag_da.c:572: warning: ‘dashtty_init’ defined but not used
drivers/tty/metag_da.c:645: warning: ‘dashtty_exit’ defined but not used
drivers/tty/metag_da.c In function ‘dash_console_write’:
drivers/tty/metag_da.c:670 : warning: passing argument 4 of ‘chancall’ discards qualifiers from pointer target type

Instead of just adding the module.h include, now would be a good time to
remove the use of these macros, replacing the module_init with
device_initcall, and removing the exit function altogether since it
isn't needed. If module support is added later the code can always be
resurrected.
Reported-by: NGuenter Roeck <linux@roeck-us.net>
Tested-by: NGuenter Roeck <linux@roeck-us.net>
Signed-off-by: NJames Hogan <james.hogan@imgtec.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: linux-metag@vger.kernel.org
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: NPaul Gortmaker <paul.gortmaker@windriver.com>

In platforms which does not use CLK framework (HAVE_CLK not set), the
clk_* functions return NULL instead of an error. This patch handles that
scenario.
Signed-off-by: NFernando Guzman Lugo <fernando.guzman.lugo@intel.com>
Signed-off-by: NChunyan Zhang <chunyan.zhang@spreadtrum.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The internal clock of the master chip, which is usually 125MHz, is only half
(62.5MHz) for the slave chips. So we have to adjust the uartclk for all the
slave ports. Therefor we add a new function to determine if a slave chip is
present and update pci_xr17v35x_setup accordingly.
Signed-off-by: NSoeren Grunewald <soeren.grunewald@desy.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The Exar XR17V358 can also be combined with a XR17V354 chip to act as a
single 12 port chip. This works the same way as the combining two XR17V358
chips. But the reported device id then is 0x4358.
Signed-off-by: NSoeren Grunewald <soeren.grunewald@desy.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

Align with coding guidelines:
Replaced a chain of "else if" by a switch case.
Signed-off-by: NJoakim Nordell <joakim.nordell@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

This drivers adds support to the STM32 USART controller, which is a
standard serial driver.
Tested-by: NChanwoo Choi <cw00.choi@samsung.com>
Reviewed-by: NPeter Hurley <peter@hurleysoftware.com>
Reviewed-by: NVladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
Reviewed-by: NAndy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: NMaxime Coquelin <mcoquelin.stm32@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

We can now use generic wakeirq handling and remove the custom handling
for the wake-up interrupts.
Acked-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NTony Lindgren <tony@atomide.com>

We can now use generic wakeirq handling and remove the custom handling
for the wake-up interrupts.
Acked-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: NTony Lindgren <tony@atomide.com>

set_irq_flags is ARM specific with custom flags which have genirq
equivalents. Convert drivers to use the genirq interfaces directly, so we
can kill off set_irq_flags. The translation of flags is as follows:

IRQF_VALID -> !IRQ_NOREQUEST
IRQF_PROBE -> !IRQ_NOPROBE
IRQF_NOAUTOEN -> IRQ_NOAUTOEN

For IRQs managed by an irqdomain, the irqdomain core code handles clearing
and setting IRQ_NOREQUEST already, so there is no need to do this in
.map() functions and we can simply remove the set_irq_flags calls. Some
users also set IRQ_NOPROBE and this has been maintained although it is not
clear that is really needed. There appears to be a great deal of blind
copy and paste of this code.
Signed-off-by: NRob Herring <robh@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jiri Slaby <jslaby@suse.cz>
Cc: Barry Song <baohua@kernel.org>
Cc: linux-serial@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The platform_sysrq_reset_seq code was intended as a way for an embedded
platform to provide its own sysrq sequence at compile time. After over two
years, nobody has started using it in an upstream kernel, and the platforms
that were interested in it have moved on to devicetree, which can be used
to configure the sequence without requiring kernel changes. The method is
also incompatible with the way that most architectures build support for
multiple platforms into a single kernel.

Now the code is producing warnings when built with gcc-5.1:

drivers/tty/sysrq.c: In function 'sysrq_init':
drivers/tty/sysrq.c:959:33: warning: array subscript is above array bounds [-Warray-bounds]
   key = platform_sysrq_reset_seq[i];

We could fix this, but it seems unlikely that it will ever be used, so
let's just remove the code instead. We still have the option to pass the
sequence either in DT, using the kernel command line, or using the
/sys/module/sysrq/parameters/reset_seq file.

Fixes: 154b7a48 ("Input: sysrq - allow specifying alternate reset sequence")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NDmitry Torokhov <dmitry.torokhov@gmail.com>

There have been concerns that the function names gpiod_set_array() and
gpiod_get_array() might be confusing to users. One might expect
gpiod_get_array() to return array values, while it is actually the array
counterpart of gpiod_get(). To be consistent with the single descriptor API
we could rename gpiod_set_array() to gpiod_set_array_value(). This makes
some function names a bit lengthy: gpiod_set_raw_array_value_cansleep().
Signed-off-by: NRojhalat Ibrahim <imr@rtschenk.de>
Acked-by: NAlexandre Courbot <acourbot@nvidia.com>
Signed-off-by: NLinus Walleij <linus.walleij@linaro.org>