[PATCH 06/14] ASoC: Intel: avs: Coredump and recovery flow

[Pierre-Louis Bossart]

On 4/26/22 12:23, Cezary Rojewski wrote:
> In rare occassions, under stress conditions or hardware malfunction, DSP

occasions

> firmware may fail. Software is notified about such situation with
> EXCEPTION_CAUGHT notification. IPC timeout is also counted as critical
> device failure. More often than not, driver can recover from such
> situations by performing full reset: killing and restarting ADSP.
> 
> Signed-off-by: Amadeusz Sławiński <amadeuszx.slawinski at linux.intel.com>
> Signed-off-by: Cezary Rojewski <cezary.rojewski at intel.com>
> ---
>  sound/soc/intel/Kconfig        |  1 +
>  sound/soc/intel/avs/avs.h      |  4 ++
>  sound/soc/intel/avs/ipc.c      | 95 +++++++++++++++++++++++++++++++++-
>  sound/soc/intel/avs/messages.h |  5 ++
>  4 files changed, 103 insertions(+), 2 deletions(-)
> 
> diff --git a/sound/soc/intel/Kconfig b/sound/soc/intel/Kconfig
> index c364ddf22267..05ad6bdecfc5 100644
> --- a/sound/soc/intel/Kconfig
> +++ b/sound/soc/intel/Kconfig
> @@ -218,6 +218,7 @@ config SND_SOC_INTEL_AVS
>  	select SND_HDA_EXT_CORE
>  	select SND_HDA_DSP_LOADER
>  	select SND_INTEL_NHLT
> +	select WANT_DEV_COREDUMP
>  	help
>  	  Enable support for Intel(R) cAVS 1.5 platforms with DSP
>  	  capabilities. This includes Skylake, Kabylake, Amberlake and
> diff --git a/sound/soc/intel/avs/avs.h b/sound/soc/intel/avs/avs.h
> index e628f78d1864..02c2aa1bcd5c 100644
> --- a/sound/soc/intel/avs/avs.h
> +++ b/sound/soc/intel/avs/avs.h
> @@ -42,6 +42,7 @@ struct avs_dsp_ops {
>  	int (* const load_basefw)(struct avs_dev *, struct firmware *);
>  	int (* const load_lib)(struct avs_dev *, struct firmware *, u32);
>  	int (* const transfer_mods)(struct avs_dev *, bool, struct avs_module_entry *, u32);
> +	int (* const coredump)(struct avs_dev *, union avs_notify_msg *);
>  };
>  
>  #define avs_dsp_op(adev, op, ...) \
> @@ -164,12 +165,15 @@ struct avs_ipc {
>  	struct avs_ipc_msg rx;
>  	u32 default_timeout_ms;
>  	bool ready;
> +	bool recovering;
>  
>  	bool rx_completed;
>  	spinlock_t rx_lock;
>  	struct mutex msg_mutex;
>  	struct completion done_completion;
>  	struct completion busy_completion;
> +
> +	struct work_struct recovery_work;
>  };
>  
>  #define AVS_EIPC	EREMOTEIO
> diff --git a/sound/soc/intel/avs/ipc.c b/sound/soc/intel/avs/ipc.c
> index 68aaf01edbf2..84cb411c82fa 100644
> --- a/sound/soc/intel/avs/ipc.c
> +++ b/sound/soc/intel/avs/ipc.c
> @@ -14,6 +14,87 @@
>  
>  #define AVS_IPC_TIMEOUT_MS	300
>  
> +static void avs_dsp_recovery(struct avs_dev *adev)
> +{
> +	struct avs_soc_component *acomp;
> +	unsigned int core_mask;
> +	int ret;
> +
> +	if (adev->ipc->recovering)
> +		return;
> +	adev->ipc->recovering = true;

don't you need some sort of lock to test/clear this flag?

> +
> +	mutex_lock(&adev->comp_list_mutex);
> +	/* disconnect all running streams */
> +	list_for_each_entry(acomp, &adev->comp_list, node) {
> +		struct snd_soc_pcm_runtime *rtd;
> +		struct snd_soc_card *card;
> +
> +		card = acomp->base.card;
> +		if (!card)
> +			continue;
> +
> +		for_each_card_rtds(card, rtd) {
> +			struct snd_pcm *pcm;
> +			int dir;
> +
> +			pcm = rtd->pcm;
> +			if (!pcm || rtd->dai_link->no_pcm)
> +				continue;
> +
> +			for_each_pcm_streams(dir) {
> +				struct snd_pcm_substream *substream;
> +
> +				substream = pcm->streams[dir].substream;
> +				if (!substream || !substream->runtime)
> +					continue;
> +
> +				snd_pcm_stop(substream, SNDRV_PCM_STATE_DISCONNECTED);
> +			}
> +		}
> +	}
> +	mutex_unlock(&adev->comp_list_mutex);
> +
> +	/* forcibly shutdown all cores */
> +	core_mask = GENMASK(adev->hw_cfg.dsp_cores - 1, 0);
> +	avs_dsp_core_disable(adev, core_mask);
> +
> +	/* attempt dsp reboot */
> +	ret = avs_dsp_boot_firmware(adev, true);
> +	if (ret < 0)
> +		dev_err(adev->dev, "dsp reboot failed: %d\n", ret);
> +
> +	pm_runtime_mark_last_busy(adev->dev);
> +	pm_runtime_enable(adev->dev);
> +	pm_request_autosuspend(adev->dev);

there are zero users of this routine in the entire sound/ tree, can you clarify why this is needed or what you are trying to do?

> +
> +	adev->ipc->recovering = false;
> +}