soc_pcm_components_open/close() try to call
try_module_get()/module_put() based on
component->driver->module_get_upon_open.

Here, the purpose why we need to call these functions are to
checking module reference.
Thus, we need to call try_module_open() even though it doesn't
have .open callback.

The same reason, we need to call module_put() even though it
doesn't have .close

This patch calls try_module_get()/module_put() regardless of
.open/.close
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: NPierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

soc-pcm.c has soc_pcm_components_close() but not have its open()
side function. This kind of unbalance function is very unreadable.

And, current error handling is not correct.
Because it is using for_each_rtdcom() loop, we need to call
soc_pcm_components_close() anyway even though
CPU DAI .startup() failed.

This patch adds soc_pcm_components_open(), and fixup error
handling issue.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Codec side is setting codec_dai->rate = 0 when error case
at soc_pcm_hw_params(), but there is not such setting for CPU side.
This patch adds it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

cpu_dai related operation is separated by component operation at
soc_pcm_hw_params() somehow.
It is not readable, let's do it at same place
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

If playback/capture is paused and system enters S3, after system returns
from suspend, BE dai needs to call prepare() callback when playback/capture
is released from pause if RESUME_INFO flag is not set.

Currently, the dpcm_be_dai_prepare() function will block calling prepare()
if the pcm is in SND_SOC_DPCM_STATE_PAUSED state. This will cause the
following test case fail if the pcm uses BE:

playback -> pause -> S3 suspend -> S3 resume -> pause release

The playback may exit abnormally when pause is released because the BE dai
prepare() is not called.

This patch allows dpcm_be_dai_prepare() to call dai prepare() callback in
SND_SOC_DPCM_STATE_PAUSED state.
Signed-off-by: NLibin Yang <libin.yang@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Like for hw_params, hw_free should not be called on codec dai for
which the current stream is invalid.

Fixes: cde79035 ("ASoC: Handle multiple codecs with split playback / capture")
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

A stream may specify a rate range using 'rate_min' and 'rate_max', so a
stream may be valid and not specify any rates. However, as stream cannot
be valid and not have any channel. Let's use this condition instead to
determine if a stream is valid or not.

Fixes: cde79035 ("ASoC: Handle multiple codecs with split playback / capture")
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Some drivers mandate setting up hw params after resuming from system sleep.
Since, the hw_params ioctl is not invoked upon resuming, the fixed-up BE
dai hw params should be saved so the driver can use it in its resume
sequence.
Signed-off-by: NLibin Yang <libin.yang@intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Handle error before returning when try_module_get() fails
to prevent inconsistent mutex lock/unlock.

Fixes: 52034add (ASoC: pcm: update module refcount if
		  module_get_upon_open is set)
Signed-off-by: NRanjani Sridharan <ranjani.sridharan@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Setting the module_get_upon_open field for component driver
prevents the module refcount from being incremented during
component probe(). This could lead to the module being
allowed to be unloaded when a pcm stream is open. So,
if this field is set, the module's refcount should be
incremented during pcm open to prevent module removal
when the component is in use. And, the refcount should
be decremented upon pcm close.
Signed-off-by: NRanjani Sridharan <ranjani.sridharan@linux.intel.com>
Acked-by: NPierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

If for any reason, the backend does not have the requested substream
(like capture on a playback only backend), the BE will be skipped in
dpcm_be_dai_startup().

However, dpcm_apply_symmetry() does not skip those BE and will
dereference the be_substream (NULL) pointer anyway.

Like in dpcm_be_dai_startup(), just skip those BE.

Fixes: 906c7d69 ("ASoC: dpcm: Apply symmetry for DPCM")
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The dpcm get from fe_clients/be_clients
may be free before use

Add a spin lock at snd_soc_card level,
to protect the dpcm instance.
The lock may be used in atomic context, so use spin lock.

possible race condition between
void dpcm_be_disconnect(
	...
	list_del(&dpcm->list_be);
	list_del(&dpcm->list_fe);
	kfree(dpcm);
	...

and
	for_each_dpcm_fe()
	for_each_dpcm_be*()

race condition example
Thread 1:
    snd_soc_dapm_mixer_update_power()
        -> soc_dpcm_runtime_update()
            -> dpcm_be_disconnect()
                -> kfree(dpcm);
Thread 2:
    dpcm_fe_dai_trigger()
        -> dpcm_be_dai_trigger()
            -> snd_soc_dpcm_can_be_free_stop()
                -> if (dpcm->fe == fe)

Excpetion Scenario:
	two FE link to same BE
	FE1 -> BE
	FE2 ->

	Thread 1: switch of mixer between FE2 -> BE
	Thread 2: pcm_stop FE1

Exception:

Unable to handle kernel paging request at virtual address dead0000000000e0

pc=<> [<ffffff8960e2cd10>] dpcm_be_dai_trigger+0x29c/0x47c
	sound/soc/soc-pcm.c:3226
		if (dpcm->fe == fe)
lr=<> [<ffffff8960e2f694>] dpcm_fe_dai_do_trigger+0x94/0x26c

Backtrace:
[<ffffff89602dba80>] notify_die+0x68/0xb8
[<ffffff896028c7dc>] die+0x118/0x2a8
[<ffffff89602a2f84>] __do_kernel_fault+0x13c/0x14c
[<ffffff89602a27f4>] do_translation_fault+0x64/0xa0
[<ffffff8960280cf8>] do_mem_abort+0x4c/0xd0
[<ffffff8960282ad0>] el1_da+0x24/0x40
[<ffffff8960e2cd10>] dpcm_be_dai_trigger+0x29c/0x47c
[<ffffff8960e2f694>] dpcm_fe_dai_do_trigger+0x94/0x26c
[<ffffff8960e2edec>] dpcm_fe_dai_trigger+0x3c/0x44
[<ffffff8960de5588>] snd_pcm_do_stop+0x50/0x5c
[<ffffff8960dded24>] snd_pcm_action+0xb4/0x13c
[<ffffff8960ddfdb4>] snd_pcm_drop+0xa0/0x128
[<ffffff8960de69bc>] snd_pcm_common_ioctl+0x9d8/0x30f0
[<ffffff8960de1cac>] snd_pcm_ioctl_compat+0x29c/0x2f14
[<ffffff89604c9d60>] compat_SyS_ioctl+0x128/0x244
[<ffffff8960283740>] el0_svc_naked+0x34/0x38
[<ffffffffffffffff>] 0xffffffffffffffff
Signed-off-by: NKaiChieh Chuang <kaichieh.chuang@mediatek.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The dpcm get from fe_clients/be_clients
may be free before use

Add a spin lock at snd_soc_card level,
to protect the dpcm instance.
The lock may be used in atomic context, so use spin lock.

Use irq spin lock version,
since the lock may be used in interrupts.

possible race condition between
void dpcm_be_disconnect(
	...
	list_del(&dpcm->list_be);
	list_del(&dpcm->list_fe);
	kfree(dpcm);
	...

and
	for_each_dpcm_fe()
	for_each_dpcm_be*()

race condition example
Thread 1:
    snd_soc_dapm_mixer_update_power()
        -> soc_dpcm_runtime_update()
            -> dpcm_be_disconnect()
                -> kfree(dpcm);
Thread 2:
    dpcm_fe_dai_trigger()
        -> dpcm_be_dai_trigger()
            -> snd_soc_dpcm_can_be_free_stop()
                -> if (dpcm->fe == fe)

Excpetion Scenario:
	two FE link to same BE
	FE1 -> BE
	FE2 ->

	Thread 1: switch of mixer between FE2 -> BE
	Thread 2: pcm_stop FE1

Exception:

Unable to handle kernel paging request at virtual address dead0000000000e0

pc=<> [<ffffff8960e2cd10>] dpcm_be_dai_trigger+0x29c/0x47c
	sound/soc/soc-pcm.c:3226
		if (dpcm->fe == fe)
lr=<> [<ffffff8960e2f694>] dpcm_fe_dai_do_trigger+0x94/0x26c

Backtrace:
[<ffffff89602dba80>] notify_die+0x68/0xb8
[<ffffff896028c7dc>] die+0x118/0x2a8
[<ffffff89602a2f84>] __do_kernel_fault+0x13c/0x14c
[<ffffff89602a27f4>] do_translation_fault+0x64/0xa0
[<ffffff8960280cf8>] do_mem_abort+0x4c/0xd0
[<ffffff8960282ad0>] el1_da+0x24/0x40
[<ffffff8960e2cd10>] dpcm_be_dai_trigger+0x29c/0x47c
[<ffffff8960e2f694>] dpcm_fe_dai_do_trigger+0x94/0x26c
[<ffffff8960e2edec>] dpcm_fe_dai_trigger+0x3c/0x44
[<ffffff8960de5588>] snd_pcm_do_stop+0x50/0x5c
[<ffffff8960dded24>] snd_pcm_action+0xb4/0x13c
[<ffffff8960ddfdb4>] snd_pcm_drop+0xa0/0x128
[<ffffff8960de69bc>] snd_pcm_common_ioctl+0x9d8/0x30f0
[<ffffff8960de1cac>] snd_pcm_ioctl_compat+0x29c/0x2f14
[<ffffff89604c9d60>] compat_SyS_ioctl+0x128/0x244
[<ffffff8960283740>] el0_svc_naked+0x34/0x38
[<ffffffffffffffff>] 0xffffffffffffffff
Signed-off-by: NKaiChieh Chuang <kaichieh.chuang@mediatek.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

On HDaudio platforms, if playback is started when capture is working,
there is no audible output.

This can be root-caused to the use of the rx|tx_mask to store an HDaudio
stream tag.

If capture is stared before playback, rx_mask would be non-zero on HDaudio
platform, then the channel number of playback, which is in the same codec
dai with the capture, would be changed by soc_pcm_codec_params_fixup based
on the tx_mask at first, then overwritten by this function based on rx_mask
at last.

According to the author of tx|rx_mask, tx_mask is for playback and rx_mask
is for capture. And stream direction is checked at all other references of
tx|rx_mask in ASoC, so here should be an error. This patch checks stream
direction for tx|rx_mask for fixup function.

This issue would affect not only HDaudio+ASoC, but also I2S codecs if the
channel number based on rx_mask is not equal to the one for tx_mask. It could
be rarely reproduecd because most drivers in kernel set the same channel number
to tx|rx_mask or rx_mask is zero.

Tested on all platforms using stream_tag & HDaudio and intel I2S platforms.
Signed-off-by: NRander Wang <rander.wang@linux.intel.com>
Acked-by: NPierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Currently all widgets attached to a DAI link will be powered
up when the DAI is active, however this may include routes
that are not actually in use if there are unused channels
available on the DAI.

The macros for creating AIF widgets already include an entry for
slot, it is proposed to change that to channel. The effective
difference here being respresenting the logical channel index
rather than the physical slot index. The CODECs currently
using the slot entry on the DAPM_AIF macros are using it in
a manner consistent with this, the CODECs not using it just
have the field set to zero.

A variable is added to snd_soc_dapm_widget to represent
this channel index and then for each AIF widget attached to
a DAI this is compared against the number of channels on
the stream. Enabling the links for those which will be in
use. This has the nice property that the CODECs which haven't
used the slot/channel entry in the macro will function exactly
as before due to all the AIF widgets having a channel of zero
and a stream by definition having at least one channel.
Signed-off-by: NCharles Keepax <ckeepax@opensource.cirrus.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Until now we rely on each driver calling snd_pcm_suspend*() explicitly
at its own PM handling.  However, this can be done far more easily by
setting the PM ops to each actual snd_pcm device object.

This patch adds the device_type object for PCM stream and assigns to
each PCM stream object.  The type contains only the PM ops for system
suspend; we don't need to deal with the resume in general.

The suspend hook simply calls snd_pcm_suspend_all() for the given PCM
streams.  This implies that the PM order is correctly put, i.e. PCM is
suspended before the main (or codec) driver, which should be true in
general.  If a special ordering is needed, you'd need to adjust the
device PM order manually later.

This patch introduces a new flag, snd_pcm.no_device_suspend, too.
With this flag set, the PCM device object won't invoke
snd_pcm_suspend_all() by itself.  This is needed for ASoC who wants to
manage the PM call orders in its serialized way, and the flag is set
in soc_new_pcm() as default.

For the non-ASoC world, we can get rid of the manual snd_pcm_suspend
calls.  This will be done in the later patches.
Reviewed-by: NJaroslav Kysela <perex@perex.cz>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

To be more readable code, this patch adds
new for_each_dpcm_be() macro, and replace existing code to it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

To be more readable code, this patch adds
new for_each_dpcm_fe() macro, and replace existing code to it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

To be more readable code, this patch adds
new for_each_card_rtds() macro, and replace existing code to it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

commit 0b7990e3 ("ASoC: add for_each_rtd_codec_dai() macro")
added for_each_rtd_codec_dai_reverse(). but _rollback() is better
naming than _reverse(). This patch rename it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

commit 0b7990e3 ("ASoC: add for_each_rtd_codec_dai() macro")
added for_each_rtd_codec_dai(), but it didn't convert few loop
which is not using "rtd". This patch fixup it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

ALSA SoC snd_soc_pcm_runtime has snd_soc_dai array for codec_dai.
To be more readable code, this patch adds
new for_each_rtd_codec_dai() macro, and replace existing code to it.
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

If the CPU DAI does not initialise rate_max, say if using
using KNOT or CONTINUOUS, then the rate_max field will be
initialised to 0. A value of zero in the rate_max field of
the hardware runtime will cause the sound card to support no
sample rates at all. Obviously this is not desired, just a
different mechanism is being used to apply the constraints. As
such update the setting of rate_max in dpcm_init_runtime_hw
to be consistent with the non-DPCM cases and set rate_max to
UINT_MAX if nothing is defined on the CPU DAI.
Signed-off-by: NCharles Keepax <ckeepax@opensource.cirrus.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Take into account the base delay set in pointer callback.

There are cases where a pointer function populates
runtime->delay, such as:
./sound/pci/hda/hda_controller.c
./sound/soc/intel/atom/sst-mfld-platform-pcm.c

This delay was getting lost and was overwritten by delays
from codec or cpu dai delay function if exposed.

Now,
Total delay = base delay + cpu_dai delay + codec_dai delay
Signed-off-by: NAkshu Agrawal <akshu.agrawal@amd.com>
Reviewed-by: NTakashi Iwai <tiwai@suse.de>
Signed-off-by: NMark Brown <broonie@kernel.org>

As done for format and channels, add the possibility to merge
the backend rates on the frontend rates.

This useful if the backend does not support all rates supported by the
frontend, or if several backends (cpu and codecs) with different
capabilities are connected to the same frontend.
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

The goal of this patch is to simplify a bit dpcm runtime stream merge
by removing several local variables.

ATM, merge functions return the BE 'filter' values which should then be
filtered against the FE stream values. This create a lot of local
variable and unnecessary init of min and max.

Instead of this, we can pass the FE stream values directly and let the
BE filtering functions perform the merge 'in-place'
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Machine drivers statically define a number of DAI links that currently
cannot be changed or removed by topology. This means PCMs and platform
components cannot be changed by topology at runtime AND machine drivers
are tightly coupled to topology.

This patch allows topology to override the machine driver DAI link config
in order to reuse machine drivers with different topologies and platform
components. The patch supports :-

1) create new FE PCMs with a topology defined PCM ID.
2) destroy existing static FE PCMs
3) change the platform component driver.
4) assign any new HW params fixups.
5) assign a new card name prefix to differentiate this topology to userspace.

The patch requires no changes to the machine drivers, but does add some
platform component flags that the platform component driver can assign
before loading topologies.
Signed-off-by: NLiam Girdwood <liam.r.girdwood@linux.intel.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

When merging codec formats, dpcm_runtime_base_format() should skip
the codecs which are not supporting the current stream direction.

At the moment, if a BE link has more than one codec, and only one
of these codecs has no capture DAI, it becomes impossible to start
a capture stream because the merged format would be 0.

Skipping invalid codec DAI solves the problem.

Fixes: b073ed4e ("ASoC: soc-pcm: DPCM cares BE format")
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>
Cc: stable@vger.kernel.org

Extend dpcm_merge_chan to also check backend cpu dai channels
capabilities. Apply the same policy as soc_pcm_init_runtime_hw() for
multicodec links and only check cpu dai in this case.

Cc: Jiada Wang <jiada_wang@mentor.com>
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

As it is, dpcm_runtime_update() performs the old path and new path
update of a frontend before going on to the next frontend DAI.
Depending the order of the FEs within the rtd list, the result of
the update might be different.

For example:
 * Frontend A connected to backend C, with a 48kHz playback
 * Frontend B connected to backend D, with a 44.1kHz playback
 * FE A appears before FE B in the rtd list of the card.

If we reparent BE C to FE B (disconnecting BE D):
* old path update of FE A will run first, and BE C will get hw_free()
  and shutdown()
* new path update of FE B will run after and BE C, which is stopped,
  so it will be configured at 44.1kHz, as expected

If we reparent BE D to FE A (disconnecting BE C):
* new path update of FE A will run first but since BE D is still running
  at 44.1kHz, it won't be reconfigured (no call to startup() or
  hw_params())
* old path update of FE B runs after, nothing happens
* In this case, we end up with a BE playing at 44.1kHz a stream which is
  supposed to be played at 48Khz (too slow)

To improve this situation, this patch performs all the FE old paths update
before going on to update the new paths. With this, the result should
no longer depend on the order of the FE within the card rtd list.

Please note that there might be a small performance penalty since
dpcm_process_paths() is called twice per stream direction.
Signed-off-by: NJerome Brunet <jbrunet@baylibre.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Current DPCM is caring only FE channel configuration. Sometimes
it will be trouble if user selects channel which isn't supported
by BE.

This patch adds new .dpcm_merged_chan on struct snd_soc_dai_link.
DPCM will use FE / BE merged channel if struct snd_soc_dai_link
has it.
Signed-off-by: NJiada Wang <jiada_wang@mentor.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Currently, the core will continue processing open/hw_params
component callbacks after one has failed even though it will abort
immediately afterwards. This is unnecessary and also has the issue
that close/hw_free will be called on the component which failed
open/hw_params which could result in issues if the driver doesn't
expect this behaviour.

Update the core to abort processing open/hw_params when an error
is hit and only call close/hw_free for those components that were
successfully opened.
Signed-off-by: NCharles Keepax <ckeepax@opensource.cirrus.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

When the call to hw_params for a component fails, the error code is held
by the variable '__ret' but the error message displays the value held by
the variable 'ret'. Fix the return code shown when hw_params fails for
a component.

Fixes: b8135864 ("ASoC: snd_soc_component_driver has snd_pcm_ops")
Signed-off-by: NJon Hunter <jonathanh@nvidia.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

We should set BE symmetric constraint on FE substream.

in case one BE is used by two FE1/FE2,
the first BE runtime will use FE1's substream->runtime.
hence the FE1's will be constrained by BE symmetry property.

Though, second FE2 call dpcm_apply_symmetry,
the be_substream->runtime == FE1's substream->runtime.
The FE2's substream->runtime will not be constrained
by BE's symmetry property.
Signed-off-by: NKaiChieh Chuang <kaichieh.chuang@mediatek.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

In case, one BE is used by two FE1/FE2
FE1--->BE-->
       |
FE2----]
when FE1/FE2 call dpcm_be_dai_hw_free() together
the BE users will be 2 (> 1), hence cannot be hw_free
the be state will leave at, ex. SND_SOC_DPCM_STATE_STOP

later FE1/FE2 call dpcm_be_dai_shutdown(),
will be skip due to wrong state.
leaving the BE not being hw_free and shutdown.

The BE dai will be hw_free later when calling
dpcm_be_dai_shutdown() if still in invalid state.
Signed-off-by: NKaiChieh Chuang <kaichieh.chuang@mediatek.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

Log the correct error code in case the .open() call to a component fails.
Signed-off-by: NDaniel Mack <daniel@zonque.org>
Signed-off-by: NMark Brown <broonie@kernel.org>

Now no one is using Codec related code.
Let's remove all
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

commit ef050bec ("ASoC: Remove platform code now everything is
componentised") removed platform code, but it didn't remove
.pcm_new/free which existed only for platform.
This patch remove these
Signed-off-by: NKuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Reviewed-by: NCharles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: NMark Brown <broonie@kernel.org>

As all drivers have been moved over to the new generic component
code remove the now unused platform specific code.
Signed-off-by: NCharles Keepax <ckeepax@opensource.cirrus.com>
Acked-by: NVinod Koul <vkoul@kernel.org>
Signed-off-by: NMark Brown <broonie@kernel.org>