When the bus reset is performed during the suspend/resume (including
the power-saving too), it calls snd_hda_suspend() and
snd_hda_resume() again, and deadlocks eventually.

For avoiding the recursive call, add a new flag indicating that the PM
is being performed, and don't go to the bus reset mode when it's on.
Reported-and-tested-by: NJulian Wollrath <jwollrath@web.de>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Better to add a sanity check as I tend to forget something (especially
during crazy midsummer nights).
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

There was a race condition when the system suspends while hda_power_work
is running in the work queue.  If system suspend (snd_hda_suspend)
happens after the work queue releases power_lock but before it calls
hda_call_codec_suspend,  codec_suspend runs with power_on=0, causing the
codec to power up for register reads, and hanging when it calls
cancel_delayed_work_sync from the running work queue.

The call chain from the work queue will look like this:
hda_power_work <<- power_on = 1, unlock, then power_on cleard by suspend
  hda_call_codec_suspend
    hda_set_power_state
      snd_hda_codec_read
        codec_exec_verb
          snd_hda_power_up
	    snd_hda_power_save
	      __snd_hda_power_up
	        cancel_delayed_work_sync <<-- cancelling executing wq

Fix this by waiting for the work queue to finish before starting suspend
if suspend is not happening on the work queue.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

snd_hda_codec_reset() calls restore_pincfgs() where the codec is
powered up again, which eventually tries to resume and initialize via
the callbacks of the codec.  However, it's the place just after codec
free callback, thus no codec callbacks should be called after that.
On a codec like CS4206, it results in Oops due to the access in init
callback.

This patch fixes the issue by clearing the codec callbacks properly
after freeing codec.
Reported-by: NDaniel J Blueman <daniel@quora.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Although HD-audio allows pair-wise channel configurations, only the
fixed channel positions are used in this version.  In future, this can
be changed and allow user to modify the channel positions.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Just refactoring, no functional changes.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

The call of pm_notify callback in snd_hda_codec_free() should be with
the check of the current state whether pm_notify(false) is called or
not, instead of codec->power_on check.

For improving the code readability and fixing this inconsistency,
codec->d3_stop_clk_ok is renamed to codec->pm_down_notified, and this
flag is set only when runtime PM down is called.  The new name reflects
to a more direct purpose of the flag.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

CONFIG_SND_HDA_POWER_SAVE is no longer an experimental feature and its
behavior can be well controlled via the default value and module
parameter.  Let's just replace it with the standard CONFIG_PM.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

When a codec provides its own set_power_state op, the D3-clock-stop
isn't checked correctly.  And the recent changes for repeating the
state-setting operation isn't applied to such a codec, too.

This patch fixes these issues by moving the call of codec's own op to
the place where the generic power-set operation is done, and move the
power-state synchronization code out of
snd_hda_set_power_state_to_all() so that it can be called always at
the end of power-up/down sequence, and updates the D3 clock-stop flag
properly.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

When the HD-audio is removed, it leaves the refcounts when codecs are
powered up (usually yes) in the destructor.  For fixing the unbalance,
and cleaning up the code mess, this patch changes the following:
- change pm_notify callback to take the explicit power on/off state,
- check of D3 stop-clock and keep_link_on flags is moved to the caller
  side,
- call pm_notify callback in snd_hda_codec_new() and snd_hda_codec_free()
  so that the refcounts are proprely updated.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

EPSS parameter should be static, so we can read it once and remember.
This also allows more easily to override the wrong EPSS capability
reported from a codec by changing the flag in the codec
initialization step.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Use unsigned int to make clear that the codes required only for
modules will be reduced by the compiler optimization.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

The patch to support runtime PM introduced a bug:
Module parameter 'power_save_controller', and the codec flag 'd3_stop_clk'
'd3_stop_clk_ok' are defined only when HDA power save is enabled in config. But
there are references to them without checking macro CONFIG_SND_HDA_POWER_SAVE.

This patch is to fix the bug.
Signed-off-by: NMengdong Lin <mengdong.lin@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Runtime PM can bring more power saving:
- When the controller is suspended, its parent device will also have a chance
  to suspend.
- PCI subsystem can choose the lowest power state the controller can signal
  wake up from. This state can be D3cold on platforms with ACPI PM support.
And runtime PM can provide a gerneral sysfs interface for a system policy
manager.

Runtime PM support is based on current HDA power saving implementation. The user
can enable runtime PM on platfroms that provide acceptable latency on transition
from D3 to D0.

Details:
- When both power saving and runtime PM are enabled:
  -- If a codec supports 'stop-clock' in D3, it will request suspending the
     controller after it enters D3 and request resuming the controller before
     back to D0. Thus the controller will be suspended only when all codecs are
     suspended and support stop-clock in D3.
  -- User IO operations and HW wakeup signal can resume the controller back to
     D0.
- If runtime PM is disabled, power saving just works as before.
- If power saving is disabled, the controller won't be suspended because the
  power usage counter can never be 0.

More about 'stop-clock' feature:
If a codec can support targeted pass-through operations in D3 state when there
is no BCLK present on the link, it will set CLKSTOP flag in the supported power
states and report PS-ClkStopOk when entering D3 state. Please refer to HDA spec
section 7.3.3.10 Power state and 7.3.4.12 Supported Power State.

[Fixed CONFIG_PM_RUNTIME dependency in hda_intel.c by tiwai]
Signed-off-by: NMengdong Lin <mengdong.lin@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Instead of calling the jack sync in the init callback of each codec,
call it generically at initialization and resume.  By calling it at
the last of resume sequence, a possible race between the jack sync and
the unsol event enablement in the current code will be closed, too.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

When the codec turn-on operation is canceled by the immediate
power-on, the driver left the power_transition flag as is.
This caused the persistent avoidance of power-save behavior.

Cc: <stable@vger.kernel.org> [v3.5+]
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Added a new helper function snd_hda_power_sync() to trigger the
power-saving manually.  It's an inline function call to
snd_hda_power_save() helper function.

Together with this addition, snd_hda_power_up*() and
snd_hda_power_down() functions are inlined to a call of the same
snd_hda_power_save() helper function.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Signed-off-by: NTakashi Iwai <tiwai@suse.de>

The return value of snd_hda_param_read() is -1 for an error, otherwise
it's the supported power states of a codec.

The supported power states is a 32-bit value. Bit 31 will be set to 1
if the codec supports EPSS, thus making "sup" negative. And the bit
28:5 is reserved as "0".
So a negative value other than -1 shall be further checked.

Please refer to High-Definition spec 7.3.4.12 "Supported Power
States", thanks!
Signed-off-by: NMengdong Lin <mengdong.lin@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

For processing the firmware handling properly for built-in kernels,
implement an asynchronous firmware loading with
request_firmware_nowait().  This means that the codec probing is
deferred when the patch option is specified.
Tested-by: NThierry Reding <thierry.reding@avionic-design.de>
Reviewed-by: NDavid Henningsson <david.henningsson@canonical.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Introduced a new flag to set up the PCM stream format at first before
the stream_id and channel tag.  Some codecs (e.g. CA0132) seem
preferring this over stream_id -> format order.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Move snd_hda_mixer_amp_switch_put_beep() to hda_beep.c as a clean up
to remove one more ifdef.

Also add the corresponding get callback to return consistently the
digital beep state independently from the mixer amp value.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Straightforward conversion to the new pm_ops from the legacy
suspend/resume ops.

Since we change vx222, vx_core and vxpocket have to be converted,
too.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Better to clean up the caches for avoiding inconsistent codec state
after the reconfiguration.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Some caches aren't released properly at module unloading time.

Cc: <stable@vger.kernel.org> [v3.4+]
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

After cancel_delayed_work_sync returns, the power down work either never
started (power_on == 1) or finished (power_on == 0).  In the former case
there is no need to power up again.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

This addresses an issue encountered when a pcm is opened while
transitioning to low power state (codec->power_on == 1 &&
codec->power_transition == -1).  Add snd_pcm_power_up_d3wait to
hda_codec.  This function is used to power up from azx_open as opposed
to snd_hda_power_up used from codec_exec_verb. When powering up from
azx_open, wait for pending power downs to complete, avoiding the power
up continuing in parallel with the power down on the work queue.

The specific issue seen was with the CS4210 codec, it powers off the ADC
and DAC nid in its suspend handler.  If it is re-opened before the
~100ms power down process completes, the ADC and DAC nid are initialized
while powered down and audio is lost until another suspend/resume cycle.
Signed-off-by: NDylan Reid <dgreid@chromium.org>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

codec may reject power state transition requests(reporting PS-ERROR set),
in that case we re-issue a power state setting and check error bit again.
Signed-off-by: NWang Xingchao <xingchao.wang@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

if EPSS supported, transition from D3 state to D0 state in less
than 10ms
Signed-off-by: NWang Xingchao <xingchao.wang@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Add function to check whether power states supported by specific
codec node.
Signed-off-by: NWang Xingchao <xingchao.wang@intel.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Like the previous fixes for cache hash accesses, a protection over
accesses to the widget connection list array must be provided.
Together with this action, remove snd_hda_get_conn_list() which can be
always race, and replace it with either snd_hda_get_num_conns() or
snd_hda_get_connections() calls.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

It's a preliminary work for the vga-switcher support.
Export the function to do pseudo-lock for the sound card to be used
in other places.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

It turned out that the FLOAT format on CS4206 results in simple
noises, which implies that this is no right format as is.
Since CS4206 is the only codec supporting the float, let's disable it
until we find the correct format.
Reported-and-tested-by: NAdrian Knoth <adi@drcomp.erfurt.thur.de>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

The amp and caps hashes aren't protected properly for concurrent
accesses.  Protect them via a new mutex now.

But it can't be so simple as originally thought: since the update of a
hash table entry itself might trigger the power-up sequence which
again accesses the hash table, we can't cover the whole function
simply via mutex.  Thus the update part has to be split from the mutex
and revalidated.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Add the missing mutex protection or move into the protected part for
SPDIF access codes for codecs.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Check the power_transition up/down state instead of boolean bit, so
that the power-up sequence can cancel the pending power-down work
properly.  Also, by moving cancel_delayed_work_sync() before the
actual power-up sequence, make sure that the delayed power-down is
completed.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Since the recent commit, the resume procedure is always performed at
the resume time.  This makes the pre_resume hack for VREF mute LED on
some HP laptops superfluous.  As this is the only user of pre_resume
(and there is no user of post_suspend) ops, let's kill them again.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Just code shuffles.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

To avoid some races.  Still not perfect, but now a bit safer.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>

Some power-saving states have been left unchanged in
snd_hda_codec_reset(), and this is a potential danger because the
function may be called in various situations including the continuous
operation after that call.
Signed-off-by: NTakashi Iwai <tiwai@suse.de>