The code inherited from the Skylake driver does not seem to follow any
known hardware recommendations.
The only two recommended options are
a) use DPIB registers if VC1 traffic is not allowed
b) use DPIB DDR update if VC1 traffic is used
In all of SOF-based updated, VC1 is not supported so we can 'safely'
move to using DPIB registers only.
This patch keeps the legacy code, in case there was an undocumented
issue lost to history, and adds the DPIB DDR update for additional
debug.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Link: https://lore.kernel.org/r/20211207193947.71080-6-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
On the latest Lenovo Thinkstation laptops, we often experience the
speaker failure after rebooting, check the dmesg, we could see:
sof-audio-pci-intel-tgl 0000:00:1f.3: codec #0 probe error, ret: -5
The analogue codec on the machine is ALC287, then we designed a
testcase to reboot and check the codec probing result repeatedly, we
found the analogue codec probing always failed at least once within
several minutes to several hours (roughly 1 reboot per min). This
issue happens on all laptops of this Thinkstation model, but with
legacy HDA driver, we couldn't reproduce this issue on those laptops.
And so far, this issue is not reproduced on machines which don't
belong to this model.
We tried to make the hda_dsp_ctrl_init_chip() same as
hda_intel_init_chip() which is the controller init routine in the
legacy HDA driver, but it didn't help.
We found when issue happens, the resp is -1, and if we let driver
re-run send_cmd() and get_response(), it will get the correct response
10ec0287, then driver continues the rest work, finally boot to the
desktop and all audio function work well.
Here adding codec probing retries to 3 times, it could fix the issue
on this Thinkstation model, and it doesn't bring impact to other
machines.
Reviewed-by: Bard Liao <bard.liao@intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Link: https://lore.kernel.org/r/20211130090606.529348-1-kai.vehmanen@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Merge series from Kai Vehmanen <kai.vehmanen@linux.intel.com>:
Implement an updated programming sequence to handle DMA stop for Intel
HD-Audio DMA.
The new flow is only used if the firmware is sufficiently new to
support the feature. SOF1.9.2 is the first release with the updated
flow. The kernel changes are backwards compatible with old firmware
releases. Likewise new firmware releases will work with old kernel.
Series reviewed originally at:
https://github.com/thesofproject/linux/pull/3167
On 32-bit with CONFIG_ARCH_DMA_ADDR_T_64BIT=n:
sound/soc/sof/mediatek/mt8195/mt8195.c: In function ‘platform_parse_resource’:
sound/soc/sof/mediatek/mt8195/mt8195.c:51:15: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 4 has type ‘unsigned int’ [-Werror=format=]
51 | dev_dbg(dev, "DMA pbase=0x%llx, size=0x%llx\n",
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sound/soc/sof/mediatek/mt8195/mt8195.c: In function ‘adsp_memory_remap_init’:
sound/soc/sof/mediatek/mt8195/mt8195.c:167:15: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 4 has type ‘phys_addr_t’ {aka ‘unsigned int’} [-Werror=format=]
167 | dev_dbg(dev, "adsp->pa_dram %llx, offset %#x\n", adsp->pa_dram, offset);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sound/soc/sof/mediatek/mt8195/mt8195.c: In function ‘adsp_shared_base_ioremap’:
sound/soc/sof/mediatek/mt8195/mt8195.c:196:15: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 5 has type ‘phys_addr_t’ {aka ‘unsigned int’} [-Werror=format=]
196 | dev_dbg(dev, "shared-dram vbase=%p, phy addr :%llx, size=%#x\n",
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Fix the first cases by printing the full resource using %pR.
Fix the other cases by printing the physical addresses using %pa.
Reported-by: noreply@ellerman.id.au
Fixes: 32d7e03d26 ("ASoC: SOF: mediatek: Add mt8195 hardware support")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/20211123103013.73645-1-geert@linux-m68k.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Some DAI components, such as HDaudio, need to be stopped in two steps
a) stop the DAI component
b) stop the DAI DMA
This patch enables this two-step stop by expanding the DAI_CONFIG
IPC flags and split them into 2 parts.
The 4 LSB bits indicate when the DAI_CONFIG IPC is sent, ex: hw_params,
hw_free or pause. The 4 MSB bits are used as the quirk flags to be used
along with the command flags. The quirk flag called
SOF_DAI_CONFIG_FLAGS_2_STEP_STOP shall be set along with the HW_PARAMS
command flag, i.e. before the pipeline is started so that the stop/pause
trigger op in the FW can take the appropriate action to either
perform/skip the DMA stop. If set, the DMA stop will be executed when
the DAI_CONFIG IPC is sent during hw_free. In the case of pause, DMA
pause will be handled when the DAI_CONFIG IPC is sent with the PAUSE
command flag.
Along with this, modify the signature for the hda_ctrl_dai_widget_setup/
hda_ctrl_dai_widget_free() functions to take additional flags as an
argument and modify all users to pass the appropriate quirk flags. Only
the HDA DAI's need to pass the SOF_DAI_CONFIG_FLAGS_2_STEP_STOP quirk
flag during hw_params to indicate that it supports two-step stop and
pause.
Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Link: https://lore.kernel.org/r/20211125101520.291581-10-kai.vehmanen@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Merge series from Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>:
The use of the SPIB register helps reduce power consumption - though
to a smaller degree than DMI_L1. This hardware capability is however
incompatible with userspace-initiated rewinds typically used by
PulseAudio.
In the past (2015..2017) Intel suggested an API extension to let
applications disable rewinds. At the time the feedback was that such a
capability was too Intel-specific and SPIB remained unused except for
loading DSP code. We now see devices with smaller batteries being
released, and it's time to revisit Linux support for SPIB to extend
battery life.
In this update the rewinds are disabled via an opt-in kernel
parameter. In the previous reviews, there was consensus that a Kconfig
option was too complicated for distributions to set, and we are
missing a TBD API to expose such capabilities to user-space.
The debate on whether or not to use rewinds, and the impact of
disabling rewinds, will likely be closed when Intel releases the
'deep-buffer' support, currently under development [2][3]. With this
solution, rewinds will not be needed, ever. When an application deals
with content that is not latency-sensitive (e.g. music playback), it
will be able to reduce power consumption by selecting a different PCM
device with increased buffering capabilities. Low-latency streams
will be handled by the 'regular' path. In other words, the impossible
compromise between power and latency will be handled with different
PCM devices/profiles for the same endpoint, and we can push the design
of capability negotiation to a later time when all the building blocks
(firmware topology, kernel, userspace) are ready - we still have
firmware xruns, DPCM race conditions to solve, and a need to describe
these alternate PCM devices with UCM using 'modifiers'.
When we disable rewinds, then the .ack can be used to program SPIB
with the application pointer, which allows the HDaudio DMA to save
power by opportunistically bursting data transfers when the path to
memory is enabled (and conversely to shut it down when there are no
transfer requests).
The SPIB register can only be programmed with incremental values with
wrap-around after the DMA RUN bits are set. For simplicity, we set the
INFO_NO_REWINDS flag in the .open callback when we already need to
program the SNDRV_PCM_INFO_SYNC_APPLPTR flag.
Rewinds are not used by many applications. One notable application
using rewinds is PulseAudio. Practical experiments with
Ubuntu/PulseAudio default settings did not show any audible issues,
but the user may hear volume changes and notification with a delay,
depending on the size of the ring buffer and latency constraints.
The choice of disabling rewinds is exposed as a kernel parameter and
not a Kconfig option to avoid any undesirable side-effects.
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Co-developed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Link: https://lore.kernel.org/r/20211119230852.206310-5-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Merge series from Kai Vehmanen <kai.vehmanen@linux.intel.com>:
When a pipeline is marked dynamic in the SOF DSP firmware
topology definition (the tplg file kernel loads from filesystem),
it means the pipeline resources are not allocated when DSP is
booted (at driver probe, or at runtime resume), but rather delayed
until the pipeline is actually used.
There are already 3 versions of the Up boards with support for the SOF
community key (ApolloLake, WhiskyLake, TigerLake). Rather than
continue to add quirks for each version, let's add a wildcard.
For WHL and TGL, the authentication supports both the SOF community
key and the firmware signed with the Intel production key. Given two
choices, the community key is the preferred option to allow developers
to sign their own firmware. The firmware signed with production key
can still be selected if needed with a kernel module
option (snd-sof-pci.fw_path="intel/sof")
Tested-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Reviewed-by: Péter Ujfalusi <peter.ujfalusi@linux.intel.com>
Link: https://lore.kernel.org/r/20211119231327.211946-1-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
This patch adds support for dynamic pipelines with multi-core
by using the platform-specific core_get/put() ops to
power up/down a core when a widget is set up/freed.
Along with this, a few redundant functions are removed:
1. sof_pipeline_core_enable() is no longer needed as the
pipeline core will be set up when the pipeline widget
is set up
2. sof_core_enable() is replaced with snd_sof_core_get()
4. core_power_up/down() DSP ops are deprecated and replaced with
core get/put ops.
5. Core power down in sof_widget_unload() during topology
removal is also removed as it is not really needed. For dynamic
pipelines, the cores will be powered off when they are not used.
For static pipelines, the cores will be powered off in the device
remove callback.
Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Link: https://lore.kernel.org/r/20211119192621.4096077-10-kai.vehmanen@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Resume common flow (System PM / Runtime PM) is like this:
sof_resume
-> specific device resume
-> snd_sof_load_firmware
-> snd_sof_dsp_reset (1)
-> load_modules()
-> snd_sof_run_firmware (2)
We need to implement dsp_reset callback (1) that will actually reset
the DSP but keep it stalled.
In order to implement this we do the following:
-> put DSP into reset (assert CoreReset bit from PWRCTL)
-> stall the DSP using RunStall bit from AudioDSP_REG2 mix
-> take DSP out of reset (de-assert CoreReset bit from PWRCTL)
At this moment the DSP is taken out of reset and Stalled! This means
that we can load the firmware and then start the DSP (2).
Until now we resetted the DSP by turning down the Audiomix PD. This
doesn't work for Runtime PM if another IP is keeping Audiomix PD up.
By introducing dsp_reset() we no longer rely on turning off the
audiomix to reset the DSP.
Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Link: https://lore.kernel.org/r/20211119094319.81674-6-daniel.baluta@oss.nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
We make use of common imx8m_suspend / imx8m_resume functions
for both system PM and runtime PM.
imx8m_suspend:
- frees the MU channels
- disables the clocks
imx8m_resume
- enables the clocks
- requests the MU channels
On i.MX8MP there is no dedicated functionality to put the DSP in reset.
The only way of doing this is to POWER DOWN the Audiomix domain.
We are able to do this because turning off the clocks and freeing the
channels makes the Audiomix to have no users thus PM kernel core turns
it down.
SOF core will not call system PM suspend handler if the DSP is already
down, but at resume it will call the system PM resume. So, we need to
keep track of the state via snd_sof_dsp_set_power_state
Few insights on how SOF core handles the PM:
- SOF core uses PM runtime autosuspend (with a timeout of 2 secs)
- at probe, SOF core boots the DSP and lets the PM runtime suspend to
turn it off, if there is no activity
- when someone opens the ALSA sound card (aplay/arecord, etc) ALSA core
calls PM runtime resume to turn on the DSP
- when the ALSA sound card is closed SOF core make use of PM subsystem
to call PM runtime suspend and thus turning off the DSP.
Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20211119094319.81674-4-daniel.baluta@oss.nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Handle clocks and mailbox channels at runtime suspend/resume
in order to save power.
DSP runtime PM uses a timeout of 2s. If device
is idle for 2s system will enter runtime suspend.
Because SOF state machine assumes that even if the DSP wasn't previously
active at a System resume, will re-load the firmware we need to make sure
that all needed resources are active.
Kernel core will take care of enabling the PD, we need to make sure that
we request the MU channels.
Signed-off-by: Daniel Baluta <daniel.baluta@nxp.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20211119094319.81674-3-daniel.baluta@oss.nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
