Some devices can usefully run audio while the Linux system is suspended.
One of the most common examples is smartphone systems, which are normally
designed to allow audio to be run between the baseband and the CODEC
without passing through the CPU and so can suspend the CPU when on a
voice call for additional power savings.
Support such systems by providing an API snd_soc_dapm_ignore_suspend().
This can be used to mark DAPM endpoints as not being sensitive to
system suspend. When the system is being suspended paths between
endpoints which are marked as ignoring suspend will be kept active.
Both source and sink must be marked, and there must already be an
active path between the two endpoints prior to suspend.
When paths are active over suspend the bias management will hold the
device bias in the ON state. This is used to avoid suspending the
CODEC while it is still in use.
Tested-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Instead of using stream events to handle power down during suspend
integrate the handling with the normal widget path checking by
replacing all cases where we report a connected endpoint in a path
with a function snd_soc_dapm_suspend_check() which looks at the ALSA
power state for the card and reports false if we are in a D3 state.
Since the core moves us into D3 prior to initating the suspend all
power checks during suspend will cause the widgets to be powered
down. In order to ensure that widgets are powered up on resume set
the card to D2 at the start of resume handling (ALSA API calls
require D0 so we are still protected against userspace access).
Tested-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
We now manage suspend within the main power analysis rather than by
flipping the state of widgets.
Tested-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Some systems, such as those with mechanical jack detection, may wish
to force enable a pin (typically mic bias) only some of the time.
Support such systems by having disable_pin() also coveer force enabled
pins.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
A code audit reveals that there are currently no users of the widget
controls on PGAs. This is likely to continue to be the case since
while there are useful things that can be done with integrating the
PGA gain and mute controls with the power sequencing userspace
generally wants stereo controls for output stages which this doesn't
map onto well.
In preparation for implementing something more useful strip out the
existing code, leaving the parameters there for use by the new code.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Allow pins to be forced on regardless of their power state. This is
intended for use with microphone bias supplies which need to be
enabled in order to support microphone detection - in systems without
appropriate hardware leaving the microphone unbiased when not in use
saves power.
The force done at power check time in order to avoid disrupting other
power detection logic.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Currently during pop/click debug we're inserting a delay both after
every log message we generate and at explicit points in the sequence,
slowing things down even further than they need to be especially when
many writes get coalesced by the sequence generation code.
Remove the per-printk delay and ensure that we have explicit delays
where we say we want them.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Currently ASoC always maintains the bias of the CODEC while the system
is active. With older mobile CODECs this is required since the outputs
are referenced to a non-zero voltage and enabling or disabling this
voltage without audible pops or clicks in the output takes too long to
do when starting or stopping audio.
As a result of features such as ground referenced outputs and class D
speaker drivers current generation devices are able to power on and off
much more quickly without these system level issues so provide a new
flag idle_bias_off in snd_soc_codec which will cause the core to turn
off the CODEC bias. The distinction between STANDBY and OFF is still
maintained. This is partly for consistency but also allows for
potential future extensions such as per-machine overrides or deferring
the bias removal.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
The same information is now visible via debugfs and with large modern
devices dumping everything to the console can be very resource
intensive, causing more harm than good.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
In preparation for multiple device support.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
No need for the mixers to know about this, and it allows for virtual
controls.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Make it a bit easier to tie DAPM widgets in with the register map
without referring to the source by including the register location
controlled by the widget.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
When we get a stream suspend event force the power down since otherwise
the stream would remain marked as active. In future we'll probably want
to make this stream-specific and add an interface to make the power down
of other widgets optional in order to support leaving bypass paths
active while suspending the processor.
Cc: stable@kernel.org
Reported-by: Joonyoung Shim <jy0922.shim@samsung.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Access to damp_power_widgets() is assumed to be single-threaded.
Concurrent accesses to dapm_power_widgets() may result in
unpredictable behavior.
Calls from:
close_delayed_work()
soc_codec_close()
soc_pcm_prepare()
soc_suspend()
soc_resume_deferred()
to snd_soc_dapm_stream_event() do not have the codec->mutex
taken to cover the call to dapm_power_widgets(). Thus, take
the mutex in these paths also to assure single-threaded use
of dapm_power_widgets().
Signed-off-by: Eero Nurkkala <ext-eero.nurkkala@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Sometimes it is desirable to have a mux which does not reflect any
direct register configuration but which will instead only have an
effect implicitly (for example, as a result of changing which parts
of the device are powered up). Provide a virtual mux for this purpose.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
We don't need to check for an event callback since we also check for
an appropriate event flag when applying mux status changes.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Since the SND_SOC_DAPM_LINE can be input or output, additional check is
needed in order to determine if the widget is connected as input or
output.
When checking for connected outputs, if the widget is line, than check
if the sources list is not empty (line is connected as output)
For input endpoint check, when the widget is line, also check if the
sinks list is not empty (line is connected as input).
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@nokia.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Some chips with complex internal supply (particularly clocking)
arragements may have multiple options for some of the supply
connections. Since these don't affect user-visible audio routing
the expectation would be that they would be managed automatically
by one of the drivers.
Support these users by allowing routes to have a connected function
which is queried before the connectedness of the path is checked as
normal. Currently this is only done for supplies, other widgets
could be supported but are not currently since the expectation for
them is that audio routing will be under the control of userspace.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Currently when built with DEBUG DAPM will dump information about
the power state decisions it is taking for each widget to dmesg.
This isn't an ideal way of getting the information - it requires
a kernel build to turn it on and off and for large hub CODECs the
volume of information is so large as to be illegible. When the
output goes to the console it can also cause a noticable impact
on performance simply to print it out.
Improve the situation by adding a dapm directory to our debugfs
tree containing a file per widget with the same information in
it. This still requires a decision to build with debugfs support
but is easier to navigate and much less intrusive.
In addition to the previously displayed information active streams
are also shown in these files.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
If the CODEC does not provide a set_bias_level() then update the
bias_level variable for it since other parts of the system expect
that to be maintained.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Currently DAPM interfaces with the audio streams to and from the
processor at the DAC and ADC widgets. As the digital capabilities
of parts increases this is becoming a less and less able to meet
the needs of parts.
To meet the needs of these devices create new widgets interfacing
with the TDM bus but not integrated into any other functionality.
Audio can then be routed to and from these widgets using existing
routing widgets.
A slot number is provided in the definition but this is currently
not used yet. This is intended to support devices which can use
more than one TDM slot on a single interface.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Speaker and headphone outputs do not need to be handled separately
since they can't be part of the same path.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
If the system doesn't have any DAPM widgets then we can't use their
state to check if the bias level for the codec should be up.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
In SOC DAPM layer of SOUND subsystem, when add signal route (in the
function snd_soc_dapm_add_route() ), the original code has wrong logic
when dapm layer check each widget whether an external one.
Signed-off-by: Rongrong Cao <rrcao@ambarella.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Provide an interval after the end of DAPM sequencing so that we
can distinguish between a pop in the final step of the sequence
and a pop generated from some other source outside DAPM.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Ensure that the audio subsystem is powered down cleanly when the system
shuts down by providing a shutdown operation. This ensures that all the
components have been returned to an off state cleanly which should avoid
audio issues from partially charged capacitors or noise on digital inputs
if the system is restarted quickly.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Tested-by: Ben Dooks <ben-linux@fluff.org>
They are now only accessed within dapm_power_widgets() so can be local
to that function.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Handle gain ramping for PGAs so we can coalesce their power updates too.
This is not ideal since we can't cope properly with gain ramping for
stereo paths but that was the case without coalescing and gain ramping
is relatively infrequently used so the effects are limited.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
The more flexible value muxes and named mixers don't need to be sorted
differently from a power management point of view, they are different
only in terms of the control interface and not in terms of seqencing
behaviour.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Reduce the number of register writes we need to set the power state for
a CODEC by coalescing updates to widgets with the same sequence order and
same register into a single write.
This can be a noticable performance improvement with slow or heavily
contended control buses, such as I2C controllers with a low clock
frequency, and is particularly noticable when resuming. It can also
reduce the noticability of and pops and clicks by ensuring that left
and right channels are powered simultaneously if they are in the same
register.
Currently widgets that have events are not coalesced, including PGAs
which may use the volume ramping control.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Replace the remaining unsigned shorts with unsigned ints.
Tested with pcap2 codec (25 bits registers).
Signed-off-by: Daniel Ribeiro <drwyrm@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Lump the list walk into a single function, and pull in the power
application too so we can do some further refactoring. Pure code
motion.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
In the past the DAPM power sequencing was done by iterating over the list
of widgets once for each widget type and powering widgets of that type.
Instead of doing that do the sorting at the time we insert the widgets
into the lists of widgets to apply power changes to. This reduces the
amount of computation required for seqencing still further, though the
costs are generally dwarfed by the costs of the register writes
implementing them.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Rather than managing the bias level of the system based on if there is
an active audio stream manage it based on there being an active DAPM
widget. This simplifies the code a little, moving the power handling
into one place, and improves audio performance for bypass paths when no
playbacks or captures are active.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Takashi Iwai
Mark Brown
Tejun Heo
Eero Nurkkala
Peter Ujfalusi
Rongrong Cao
Daniel Ribeiro