This is a pure code motion patch intended to improve reviewability of a
following patch moving WM8731 to use more standard device registration.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This is a bit more idiomatic and makes identifying a configuration
based on the board type work better.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
We have software control of the MCLK for the WM8731 so save a bit of
power by actively managing it within the machine driver, enabling it
only while the codec is active.
Once ASoC supports multiple boards and doesn't require the soc-audio
device the initial clock setup should be pushed down into the arch/arm
code but for now this reduces merge issues.
Tested-by: Sedji Gaouaou <sedji.gaouaou@atmel.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The WM8731 bias level configuration function was written slightly
obscurely - streamline the code a little and refresh the comments.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
WM8753 uses a tricky way to switch DAIs "on the fly", for that it
registers 2 dummy DAIs and substitutes them depending on mixer control.
List element of registered dummy DAIs should be preserved to allow
unregistering of DAIs on module unload.
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Fix for the error when the audio module is unloaded. On unregistering
the platform_device, platform_device_release will free the platform
data.If platform data is static the kernel panics when it is freed.
Instead use the platform device helper function to add data.
This change has been tested on DM644x EVM, DM644x SFFSDR and DM355 EVM.
Signed-off-by: Chaithrika U S <chaithrika@ti.com>
Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
ASoC supports both explicit codec drivers for AC97 devices and a simple
driver which uses the standard ALSA AC97 framework for codec support.
When used with the generic AC97 codec support that will provide the
ad hoc AC97 device for drivers like touchscreens to attach to so the
core shouldn't do so.
Reported-by: Manuel Lauss <mano@roarinelk.homelinux.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Update the CS4270 codec driver to allow applications to use the mixer to
control Digital Loopback, Soft Ramp, Zero Cross, Popguard, and Auto-Mute.
Soft Ramp, Zero Cross, and Auto-Mute are disabled by the driver when it first
initializes the hardware, but these features either don't work or interfere
with normal ALSA behavior. However, they can now be re-enabled by an
application if desired.
Remove CONFIG_SND_SOC_CS4270_HWMUTE and always allow ASoC to control the mute
bits. The driver previously and erroneously assumed that these bits
control only external muting circuitry, but they also control internal
muting circuitry, so they should always be used.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch replaces "snd_soc_machine" structure by "snd_soc_card" in
SP3430 driver. This change is needed in SDP3430 driver to reflect
changes introduced by "ASoC: Rename snd_soc_card to snd_soc_machine" patch
(875065491f).
Signed-off-by: Misael Lopez Cruz <x0052729@ti.com>
Acked-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
TLV320AIC3X volume controls are logarithmic. Export their dB ranges.
Signed-off-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This is a minor fix but helps to define dB ranges for volume controls.
Only DAC digital volume has full register value range from 0 to 127 but
ADC PGA gain and output stage volume controls don't.
For ADC PGA, maximum value is 119 and then it saturates to the same
gain value of 59.5 dB. For output stages, value 117 corresponds to -78.3 dB
and is muted for values 118 and above.
Signed-off-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This removes the calls to pxa_gpio_mode from the pxa2xx-i2s driver.
Pin setup should be done during board init via pxa2xx_mfp_config
instead.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Acked-by: Eric Miao <eric.miao@marvell.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This machine driver enables sound functions on Mitac mio
a701 smartphone. Build upon ASoC v1, it handles :
- rear speaker
- front speaker
- microphone
- GSM
A global "Mio Mode" switch is not yet provided to cope with
audio path setup. As balance on audio chip line is no more
assured, an incorrect setup can produce a lot of heat and
even fry the battery behind the wm9713 and the speaker
amplifier.
It doesn't cope with :
- headset jack
- mio master mode
- master volume control
This driver is backported from ASoc v2, and amputated from
scenario setups and master volume control.
[Minor mods for terminology in comments -- broonie]
Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
In the Freescale MPC8610 sound drivers, relocate all code from the _prepare
functions into the corresponding _hw_params functions. These drivers assumed
that the sample size is known in the _prepare function and not in the
_hw_params function, but this is not true.
Move the code in fsl_dma_prepare() into fsl_dma_hw_param(). Create
fsl_ssi_hw_params() and move the code from fsl_ssi_prepare() into it.
Turn off snooping for DMA operations to/from I/O registers, since that's not
necessary.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
- make sport number handling more dynamic as not all
Blackfins have a linear sport map starting at 0
- indexes can be macroed away too
Signed-off-by: Mike Frysinger <vapier.adi@gmail.com>
Signed-off-by: Cliff Cai <cliff.cai@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Function wm899x_outpga_put_volsw_vu misuses the kcontrol's private value
by still accessing it as bitfields even SOC_SINGLE_VALUE constructs it
as a pointer into struct soc_mixer_control after the commit
4eaa9819dc.
This is very similar fix than fix to TLV320AIC3X codec made by
Eero Nurkkala <ext-eero.nurkkala@nokia.com>. This fix is compile tested
only.
Signed-off-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Cc: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Function snd_soc_dapm_put_volsw_aic3x misuses the kcontrol's private value
by still accessing it as bitfields even SOC_SINGLE_VALUE constructs it
as a pointer into struct soc_mixer_control after the commit
4eaa9819dc.
This was causing arbitrary register writes when touching the controls
defined with SOC_DAPM_SINGLE_AIC3X.
Signed-off-by: Eero Nurkkala <ext-eero.nurkkala@nokia.com>
Acked-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
PXA2xx/3xx SSP ports start from 1, not 0. Thus, the probe function
requested the wrong SSP port. Correcting this unveiled another bug
where ssp_init tries to request the already-requested SSP port again.
So this patch replaces the ssp_init/exit calls with their internals
from mach-pxa/ssp.c, leaving out the redundant ssp_request and the
unneeded IRQ request. Effectively, that leaves us with not much more
than enabling/disabling the SSP clock.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch splits set_dai_fmt into three variants (single interface,
dual interface playback only, dual interface capture only) so that
data input and output formats can be configured separately for dual
interface setups.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Tested-by: Vasily Khoruzhick <anarsoul@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Without this fix driver switches to WSPLL in uda1380_pcm_prepare
even if SYSCLK was chosen (uda1380_pcm_prepare modifies UDA1380_CLK
register to disable R00_DAC_CLK before flushing reg cache)
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This just updates my email address on some drivers I'd forgotten in a
previous patch.
Signed-off-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Replace printk calls with dev_xxx calls. Set the 'dev' field of the codec
and codec_dai structures so that these calls work.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Fix a oversight in the CS4270 codec driver that caused a build break.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
omap_pcm_trigger is called also in interrupt context so CPU flags must
be restored when returning.
Signed-off-by: Eero Nurkkala <ext-eero.nurkkala@nokia.com>
Acked-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Update pandora board file for recent TWL4030 codec changes.
Also move output related snd_soc_dapm_nc_pin() calls to
omap3pandora_out_init(), where they belong.
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Spruce up the documentation in the CS4270 codec. Use kerneldoc where
appropriate. Fix incorrect comments.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
ASoC codec drivers typically serve two masters: the I2C bus and ASoC itself.
When a codec driver registers with ASoC, a probe function is called. Most
codec drivers call ASoC first, and then register with the I2C bus in the ASoC
probe function.
However, in order to support multiple codecs on one board, it's easier if the
codec driver is probed via the I2C bus first. This is because the call to
i2c_add_driver() can result in the I2C probe function being called multiple
times - once for each codec. In the current design, the driver registers
once with ASoC, and in the ASoC probe function, it calls i2c_add_driver().
The results in the I2C probe function being called multiple times before the
driver can register with ASoC again.
The new design has the driver call i2c_add_driver() first. In the I2C probe
function, the driver registers with ASoC. This allows the ASoC probe function
to be called once per I2C device.
Also add code to check if the I2C probe function is called more than once,
since that is not supported with the current ASoC design.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch adds the analog loopback/bypass support for twl4030 codec.
Details for the implementation:
It seams that the analog loopback needs the DAC powered on on the channel,
where the loopback is selected. The switch for the DACs has been moved from
the DAPM_DAC to the "Analog XX Playback Mixer". In this way the DAC will be
powered while the audio playback is used or/and the loopback is enabled for
the channel.
The twl4030 codec powering has been reworked. Now the codec will be powered as
long as it does not receives the SND_SOC_BIAS_OFF event. The exceptions are
when the given change in the registers needs the codec power down/up cycle in
order to take effect. Otherwise the codec is on.
When the codec enters to STANDBY state and none of the loopback paths are
enabled, than the amplifiers, which are no in the DAPM path are forced to turn
off and the PLL is disabled. When playback/capture starts the disabled gains
are restored and the PLL is enabled.
When one of the loopback enabled in STANDBY mode, the disabled gains are
restored and the PLL is enabled also.
In short: the codec always goes to the lowest power state based on the
bias_level and the bypass_state.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch explicitly initializes McBSP Transmit Configuration
Control Register (XCCR) and Receive Configuration Control
Register (RCCR) to their reset values. Reset values are 26 ns
of DX delay and Transmit DMA disabled for XCCR register;
receive full cycle mode enabled and Receive DMA disabled for
RCCR register.
This patch requires a counterpart in OMAP McBSP driver before
to apply it. The required changes in McBSP were sent and approved
in linux-omap mailing list and patch is going upstream
(commit 3127f8f859 from linux-omap-2.6
tree).
Signed-off-by: Misael Lopez Cruz <x0052729@ti.com>
[ jarkko.nikula@nokia.com: Commit id for counterpart patch corrected ]
Signed-off-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The WM8753 driver multiplexes the DAI structures it exposes to the
outside world, leaving them uninitialised until the codec probes. Since
the DAI name is used during the registration and setup process provide a
dummy name.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Move the twl4030_power_up and twl4030_power_down function
earlier to facilitate the analog bypass implementation.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Change the power switches for the physical ADC and for the
amplifiers for the analog capture path to map more closely
the actual path inside of the codec.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The Headset Left anti-pop and bias ramp does not need to be
enabled, if the headset is not in use.
Move the code to DAPM event handler for HeadsetL.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Merge the codec up and down functions to a simple one.
Codec is powered down by default (reg_cache change).
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The offset cancelation bit in ANAMICL register is self cleanig.
Make sure that the reg_cache holds the same value as the HW
register.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Further improvements in the I2C initialization sequence of the CS4270 driver.
All ASoC initialization is now done in the I2C probe function.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Ensures that the DAI and socdev exported by the codec match up with
their exported prototype.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Kbuild ignores dependency from things that are themselves selected so
ASoC machine drivers need to ensure that the control bus is being built.
This also avoids issues where multiple buses are supported by a given
codec.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
pxa-regs.h and hardware.h are not intended for use directly in driver
code and references to them have been removed in other code - remove
them from the newly added e740 and e750 machine drivers.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Fix a merge issue caused by context overlap.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The CS4270 supports stand-alone mode, where the codec is not connect to the
I2C or SPI buses. Instead, input voltages configure the codec at power-on.
The CS4270 ASoC device driver has partial support for this mode, but the
code was never tested, and partial support doesn't help anyone. It also made
the rest of the code more complicated than necessary.
[Removed redundant CS4270 dependency on I2C -- broonie]
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Commit dc06102a0c in the asoc tree
did not include the necessary Kconfig and Makefile changes. This patch
completes the support for Beagleboard
Signed-off-by: Steve Sakoman <steve@sakoman.com>
Acked-by: Jarkko Nikula <jarkko.nikula@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch fixes the acpture switch name so that it better reflects its
purpose.
Signed-off-by: Ian Molton <iann@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Change the Kconfig and Makefile options for Freescale MPC8610 audio drivers
so that they can be compiled as modules, and simplify the Kconfig choices
so that only the platform is selected.
Also fix the naming of the driver files to conform to ALSA standards.
[Removed extraneous SND_SOC dependency -- broonie]
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The Freescale MPC8610 driver was defining two SOC card (snd_soc_card)
structures, partially initializing each one, but registering only one of
them with ASoC.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The PCM operations tables are not exported directly but are instead
included in the platform structure so should be declared static.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch takes fixes a number of bugs in the caching code used by
several ASoC codec drivers. Mostly off-by-one fixes.
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch provides suupport for the wm9705 AC97 codec on the Toshiba e740.
Note:
The e740 has a hard headphone switch that turns the speaker off and is not
software detectable or controlable. Also both headphone and speaker amps
share a common output enable.
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The Zylonite supports switching the MCLK for the WM9713 between the
AC97CLK and CLK_POUT outputs of the PXA processor via switch SW15 on
the board. This patch adds support for configuring the system to use
CLK_POUT.
Unfortunately it is not possible to read the state of SW15 from software
so this feature is controlled by a module option 'clk_pout' which should
be set to a non-zero value to enable the use of CLK_POUT.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The WM9713 driver does not support configuring the PLL output frequency
so the output frequency parameter is irrelevant. Allow users to set it
to zero by ignoring it.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch adds support for the wm9712 ac97 codec as used in the Toshiba e800
PDA. It includes support for powering up / down the external headphone and
speaker amplifiers on this machine.
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch adds support for the wm9705 ac97 codec as used in the Toshiba e750
PDA. It includes support for powering up / down the external headphone and
speaker amplifiers on this machine.
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This driver adds support for the wm9705 ac97 codec. The driver supports
audio input and output.
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
<mach/hardware.h> doesn't exist on AVR32 and therefore this driver won't
build on that arch. AFAICT this driver doesn't actually use the content
of that header so easiest just to remove it.
Signed-off-by: Ben Nizette <bn@niasdigital.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Remove dependency on sffsdr_fpga_set_codec_fs() when the
SFFSDR FPGA module is not selected.
Signed-off-by: Hugo Villeneuve <hugo@hugovil.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Modify the check for the mux type to also handle the
snd_soc_dapm_value_mux type in a same way as the snd_soc_dapm_mux.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Some drivers in soudn/usb/* don't handle the error code properly
from snd_card_create(). This patch fixes these places.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Some drivers in sound/isa/* don't handle the error code properly
from snd_card_create(). This patch fixes these places.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Introduced snd_card_create() function as a replacement of snd_card_new().
The new function returns a negative error code so that the probe callback
can return the proper error code, while snd_card_new() can give only NULL
check.
The old snd_card_new() is still provided as an inline function but with
__deprecated attribute. It'll be removed soon later.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
snd-hda-intel driver used schedule_work() fot the delayed DMA pointer
updates, but this has several potential problems:
- it may block other eventsd works longer
- it may deadlock when probing fails and flush_scheduled_work() is
called during probe callback (as probe callback itself could be
invoked from eventd)
This patch adds an own workq for each driver instance to solve these
problems.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The Intel DX58SO board works fine with model ALC883_3ST_6ch_INTEL.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Call the snd_soc_free_pcm and snd_soc_dapm_free when the
codec driver is unloaded.
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Many codec drivers were implementing cookie-cutter copies of the function
that adds kcontrols to the codec.
This patch moves this code to a common function snd_soc_add_controls() in
soc-core.c and updates all drivers using copies of this function to use the
new common version.
[Edited to raise priority of error log message and document parameters.
-- broonie]
Signed-off-by: Ian Molton <ian@mnementh.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
This patch adds a jack reporting interface to ASoC. This wraps the ALSA
core jack detection functionality and provides integration with DAPM to
automatically update the power state of pins based on the jack state.
Since embedded platforms can have multiple detecton methods used for a
single jack (eg, separate microphone and headphone detection) the report
function allows specification of which bits are being updated on a given
report.
The expected usage is that machine drivers will create jack objects and
then configure jack detection methods to update that jack.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Fix the following lockdep warning by not reusing the hda_codec.spdif_mutex.
ALSA sound/pci/hda/hda_codec.c:882: hda_codec_cleanup_stream: NID=0x2
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.28-next-20090102 #33
-------------------------------------------------------
mplayer/3151 is trying to acquire lock:
(&pcm->open_mutex){--..}, at: [<ffffffffa004ced3>] snd_pcm_release+0x43/0xd0 [snd_pcm]
but task is already holding lock:
(&mm->mmap_sem){----}, at: [<ffffffff810c0252>] sys_munmap+0x42/0x80
which lock already depends on the new lock.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
