[alsa-devel] [PATCH 1/4] ASoC: Intel: Haswell: Adjust machine device private context

[Pierre-Louis Bossart]

On 8/22/19 2:02 PM, Cezary Rojewski wrote:
> On 2019-08-22 20:44, Pierre-Louis Bossart wrote:
>>
>>
>> On 8/22/19 12:14 PM, Cezary Rojewski wrote:
>>> On 2019-08-22 18:42, Pierre-Louis Bossart wrote:
>>>>
>>>>
>>>> On 8/22/19 11:05 AM, Cezary Rojewski wrote:
>>>>> On 2019-08-22 17:58, Pierre-Louis Bossart wrote:
>>>>>>
>>>>>>
>>>>>> On 8/22/19 6:36 AM, Cezary Rojewski wrote:
>>>>>>> Apart from Haswell machines, all other devices have their private 
>>>>>>> data
>>>>>>> set to snd_soc_acpi_mach instance.
>>>>>>>
>>>>>>> Changes for HSW/ BDW boards introduced with series:
>>>>>>> https://patchwork.kernel.org/cover/10782035/
>>>>>>>
>>>>>>> added support for dai_link platform_name adjustments within card 
>>>>>>> probe
>>>>>>> routines. These take for granted private_data points to
>>>>>>> snd_soc_acpi_mach whereas for Haswell, it's sst_pdata instead. 
>>>>>>> Change
>>>>>>> private context of platform_device - representing machine board - to
>>>>>>> address this.
>>>>>>>
>>>>>>> Fixes: e87055d732e3 ("ASoC: Intel: haswell: platform name fixup 
>>>>>>> support")
>>>>>>> Fixes: 7e40ddcf974a ("ASoC: Intel: bdw-rt5677: platform name 
>>>>>>> fixup support")
>>>>>>> Fixes: 2d067b2807f9 ("ASoC: Intel: broadwell: platform name fixup 
>>>>>>> support")
>>>>>>> Signed-off-by: Cezary Rojewski <cezary.rojewski at intel.com>
>>>>>>> ---
>>>>>>>   sound/soc/intel/common/sst-acpi.c | 3 ++-
>>>>>>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>>>
>>>>>>> diff --git a/sound/soc/intel/common/sst-acpi.c 
>>>>>>> b/sound/soc/intel/common/sst-acpi.c
>>>>>>> index 15f2b27e643f..c34f628c7987 100644
>>>>>>> --- a/sound/soc/intel/common/sst-acpi.c
>>>>>>> +++ b/sound/soc/intel/common/sst-acpi.c
>>>>>>> @@ -109,11 +109,12 @@ int sst_acpi_probe(struct platform_device 
>>>>>>> *pdev)
>>>>>>>       }
>>>>>>>       platform_set_drvdata(pdev, sst_acpi);
>>>>>>> +    mach->pdata = sst_pdata;
>>>>>>>       /* register machine driver */
>>>>>>>       sst_acpi->pdev_mach =
>>>>>>>           platform_device_register_data(dev, mach->drv_name, -1,
>>>>>>> -                          sst_pdata, sizeof(*sst_pdata));
>>>>>>> +                          mach, sizeof(*mach));
>>>>>>
>>>>>> I now agree that the code I added is incorrect and probably 
>>>>>> accesses memory offsets that aren't right. I have absolutely no 
>>>>>> idea why I added this comment that 'legacy does not pass 
>>>>>> parameters' when it most definitively does. Good catch on your side.
>>>>>>
>>>>>> That said, doesn't the proposed fix introduce another issue?
>>>>>>
>>>>>> In the machine drivers, you still get pdata directly, so aren't 
>>>>>> you missing an indirection to get back to pdata from mach?
>>>>>>
>>>>>> static int bdw_rt5677_rtd_init(struct snd_soc_pcm_runtime *rtd)
>>>>>> {
>>>>>>      struct snd_soc_component *component = 
>>>>>> snd_soc_rtdcom_lookup(rtd, DRV_NAME);
>>>>>>      struct sst_pdata *pdata = dev_get_platdata(component->dev);
>>>>>>      struct sst_hsw *broadwell = pdata->dsp;
>>>>>>
>>>>>> <<< so here you took the wrong pointer, no?
>>>>>
>>>>> Both Baytrail and Haswell are enumerated in a bit different fashion 
>>>>> than SKL equivalents.
>>>>>
>>>>> There is an in-place registration for machine device - whose 
>>>>> private_data gets used in machine probe - and pcm device which 
>>>>> happens on firmware load callback 
>>>>> (/sound/soc/intel/common/sst-acpi:63). _rtd_init makes use of the 
>>>>> latter of two.
>>>>
>>>> I don't get your explanations. can you elaborate on what this does 
>>>> now that pdata is no longer passed as an argument to the machine 
>>>> driver:
>>>>
>>>> struct snd_soc_component *component = snd_soc_rtdcom_lookup(rtd, 
>>>> DRV_NAME);
>>>> struct sst_pdata *pdata = dev_get_platdata(component->dev);
>>>>
>>>> the 'component' here is not the PCM one, is it?
>>>>
>>>>
>>>
>>> Sure thing.
>>>
>>> Code:
>>>      /* register machine driver */
>>>      sst_acpi->pdev_mach =
>>>          platform_device_register_data(dev, mach->drv_name, -1,
>>>                            sst_pdata, sizeof(*sst_pdata));
>>>
>>> Found in sst_acpi_probe (/sound/soc/intel/common/sst-acpi.c:145) 
>>> generates new platform_device - which represents machine board - with 
>>> its private data set to pointer to instance of struct sst_pdata type. 
>>> This data gets used on machine board probe, e.g.: 
>>> broadwell_audio_probe (/sound/soc/intel/boards/broadwell.c:270).
>>> Involved platform is called: broadwell-audio. Requested private data 
>>> type by broadwell_audio_probe: struct snd_soc_acpi_mach *. MISMATCH.
>>>
>>>
>>> Code:
>>>
>>>      /* register PCM and DAI driver */
>>>      sst_acpi->pdev_pcm =
>>>          platform_device_register_data(dev, desc->drv_name, -1,
>>>                            sst_pdata, sizeof(*sst_pdata));
>>>
>>> Found in sst_acpi_fw_cb (/sound/soc/intel/common/sst_acpi_fw_cb:47) 
>>> generates new platform_device - which represents Haswell PCM, you may 
>>> treat it as Skylake equivalent - with its private data set to pointer 
>>> to instance of struct sst_pdata type. This data gets used on dai 
>>> .init - broadwell_rtd_init - invocation when card is instantiated by 
>>> ASoC code. As you can see on 
>>> (/sound/soc/intel/boards/broadwell.c:162), platform tied with it is: 
>>> haswell-pcm-audio. Requested private data type by broadwell_rtd_init 
>>> - struct sst_pdata *. MATCH.
>>
>>
>> the machine drivers uses snd_soc_rtdcom_lookup(rtd, DRV_NAME);
>>
>> How is DRV_NAME connected to haswell-pcm-audio?
>>
>> I must be missing something in your logic.
>>
> 
> Please checkout sst-acpi.c file and see declaration of legacy platform 
> descriptors. See the names of PCM devices (platform devices) being 
> declared.

what happens in sst-acpi.c stays in sst-acpi.c
I don't get how you retrieve the pdata in the machine driver from 
*another* driver. Different devices, different platform data.