[alsa-devel] [PATCH 1/2] ASoC: Intel: sst: Fallback to BYT-CR if IRQ 5 is missing
Pierre-Louis Bossart
pierre-louis.bossart at linux.intel.com
Wed Jan 2 17:39:43 CET 2019
On 1/1/19 3:11 PM, Stephan Gerhold wrote:
> On Mon, Dec 31, 2018 at 02:44:58PM -0600, Pierre-Louis Bossart wrote:
>> On 12/31/18 10:30 AM, Stephan Gerhold wrote:
>>> On Mon, Dec 31, 2018 at 09:38:21AM -0600, Pierre-Louis Bossart wrote:
>>>> On 12/22/18 8:47 AM, Stephan Gerhold wrote:
>>>>> Some devices detected as BYT-T by the PMIC-type based detection
>>>>> have only a single IRQ listed in the 80860F28 ACPI device. This
>>>>> causes -ENXIO later when attempting to get the IRQ at index 5.
>>>>> It turns out these devices behave more like BYT-CR devices,
>>>>> and using the IRQ at index 0 makes sound work correctly.
>>>>>
>>>>> This patch adds a fallback for these devices to is_byt_cr():
>>>>> If there is no IRQ resource at index 5, treating the device
>>>>> as BYT-T is guaranteed to fail later, so we can safely treat
>>>>> these devices as BYT-CR without breaking any working device.
>>>>>
>>>>> Link: http://mailman.alsa-project.org/pipermail/alsa-devel/2018-December/143176.html
>>>>> Signed-off-by: Stephan Gerhold <stephan at gerhold.net>
>>>>> ---
>>>>> Moved the "Detected Baytrail-CR platform" message to is_byt_cr()
>>>>> so we can log a different message if the fallback is used.
>>>>>
>>>>> Tested this on my device as-is, and simulated a "normal"
>>>>> BYT-T and BYT-CR device (copied their IRQs to a custom DSDT).
>>>>>
>>>>> sound/soc/intel/atom/sst/sst_acpi.c | 24 ++++++++++++++++++------
>>>>> 1 file changed, 18 insertions(+), 6 deletions(-)
>>>>>
>>>>> diff --git a/sound/soc/intel/atom/sst/sst_acpi.c b/sound/soc/intel/atom/sst/sst_acpi.c
>>>>> index 3a95ebbfc45d..755a396121ff 100644
>>>>> --- a/sound/soc/intel/atom/sst/sst_acpi.c
>>>>> +++ b/sound/soc/intel/atom/sst/sst_acpi.c
>>>>> @@ -255,10 +255,22 @@ static int is_byt(void)
>>>>> return status;
>>>>> }
>>>>> -static int is_byt_cr(struct device *dev, bool *bytcr)
>>>>> +static int is_byt_cr(struct platform_device *pdev, bool *bytcr)
>>>>> {
>>>>> + struct device *dev = &pdev->dev;
>>>>> int status = 0;
>>>>> + if (platform_get_resource(pdev, IORESOURCE_IRQ, 5) == NULL) {
>>>>> + /*
>>>>> + * Some devices detected as BYT-T have only a single IRQ listed,
>>>>> + * causing platform_get_irq with index 5 to return -ENXIO.
>>>>> + * The correct IRQ in this case is at index 0, as used on BYT-CR.
>>>>> + */
>>>>> + dev_info(dev, "Falling back to Baytrail-CR platform\n");
>>>>> + *bytcr = true;
>>>>> + return status;
>>>>> + }
>>>>> +
>>>> Isn't this going to bypass the PMIC-based detection on all BYT-CR devices?
>>>> Maybe move this code as a fallback used when the PMIC-based detection isn't
>>>> positive?
>>>>
>>> Except for the message that is logged, it does not really make a
>>> difference. PMIC-based detection is still used for most BYT-CR devices,
>>> which usually have 6 IRQs listed. For the few that have not, the end
>>> result (bytcr = true) is the same, even if they now use the fallback.
>>>
>>> I mentioned this in a previous mail when I asked you which option you
>>> would prefer (see [1]). Since is_byt_cr() has multiple returns,
>>> I cannot just put it last without refactoring the entire method.
>>> (Which is something I wanted to avoid...)
>> Ah yes, but there was a side thread with Andy Shevchenko where we discussed
>> that the initial return can be simplified since there are wrappers for
>> iosf_mbi_available even when CONFIG_IOSF_MBI is not enabled. The code could
>> be something like
>>
>> diff --git a/sound/soc/intel/atom/sst/sst_acpi.c
>> b/sound/soc/intel/atom/sst/sst_acpi.c
>> index ac542535b9d5..58e389a64c6a 100644
>> --- a/sound/soc/intel/atom/sst/sst_acpi.c
>> +++ b/sound/soc/intel/atom/sst/sst_acpi.c
>> @@ -255,17 +255,16 @@ static int is_byt(void)
>> return status;
>> }
>>
>> -static int is_byt_cr(struct device *dev, bool *bytcr)
>> +static int is_byt_cr(struct platform_device *pdev, bool *bytcr)
>> {
>> + struct device *dev = &pdev->dev;
>> + u32 bios_status;
>> int status = 0;
>>
>> - if (IS_ENABLED(CONFIG_IOSF_MBI)) {
>> - u32 bios_status;
>> + if (!is_byt())
>> + return status;
>>
>> - if (!is_byt() || !iosf_mbi_available()) {
>> - /* bail silently */
>> - return status;
>> - }
>> + if (iosf_mbi_available()) {
>>
>> status = iosf_mbi_read(BT_MBI_UNIT_PMC, /* 0x04 PUNIT */
>> MBI_REG_READ, /* 0x10 */
>> @@ -286,6 +285,20 @@ static int is_byt_cr(struct device *dev, bool *bytcr)
>> } else {
>> dev_info(dev, "IOSF_MBI not enabled, no BYT-CR
>> detection\n");
>> }
>> +
>> + if (*bytcr == false &&
>> + platform_get_resource(pdev, IORESOURCE_IRQ, 5) == NULL) {
>> + /*
>> + * Some devices detected as BYT-T have only a single IRQ
>> listed,
>> + * causing platform_get_irq with index 5 to return -ENXIO.
>> + * The correct IRQ in this case is at index 0, as used on
>> + * BYT-CR.
>> + */
>> + dev_info(dev, "Falling back to Baytrail-CR platform\n");
>> + status = 0;
>> + *bytcr = true;
>> + }
>> +
>> return status;
>> }
>>
>>
> Thanks! That looks fine to me. I will test it on my device and send a v2
> shortly.
>
> Speaking of simplifying is_byt_cr(): Especially its usage in
>
> ret = is_byt_cr(pdev, &bytcr);
> if (!(ret < 0 || !bytcr)) {
> /* override resource info */
> byt_rvp_platform_data.res_info = &bytcr_res_info;
> }
>
> with the negated "or" has been rather confusing to read for me.
> In my opinion, it would be easier to understand as:
> if (ret == 0 && bytcr)
>
> The return value (`ret`) is only used in this if statement.
> Since `bytcr` stays false when an error occurs in is_byt_cr(),
> we could further simplify this by returning the bool directly:
> if (is_byt_cr(pdev))
I like the suggested changes. This code evolved over time, IIRC the
status was initially reporting some ACPI code but now a boolean will do.
Good discussion, thanks!
>
> Together with:
>
> static bool is_byt_cr(struct platform_device *pdev)
> {
> struct device *dev = &pdev->dev;
>
> if (!is_byt())
> return false;
>
> if (iosf_mbi_available()) {
> u32 bios_status;
> int status = iosf_mbi_read(BT_MBI_UNIT_PMC, /* 0x04 PUNIT */
> MBI_REG_READ, /* 0x10 */
> 0x006, /* BIOS_CONFIG */
> &bios_status);
>
> if (status) {
> dev_err(dev, "could not read PUNIT BIOS_CONFIG\n");
> } else {
> /* bits 26:27 mirror PMIC options */
> bios_status = (bios_status >> 26) & 3;
>
> if ((bios_status == 1) || (bios_status == 3)) {
> dev_info(dev, "Detected Baytrail-CR platform\n");
> return true;
> } else {
> dev_info(dev, "BYT-CR not detected\n");
> }
> }
> } else {
> dev_info(dev, "IOSF_MBI not enabled, no BYT-CR detection\n");
> }
>
> if (platform_get_resource(pdev, IORESOURCE_IRQ, 5) == NULL) {
> /*
> * Some devices detected as BYT-T have only a single IRQ listed,
> * causing platform_get_irq with index 5 to return -ENXIO.
> * The correct IRQ in this case is at index 0, as used on BYT-CR.
> */
> dev_info(dev, "Falling back to Baytrail-CR platform\n");
> return true;
> }
>
> return false;
> }
>
> What do you think?
>
>>
>>> [1]: http://mailman.alsa-project.org/pipermail/alsa-devel/2018-December/143339.html
>>>
>>>>> if (IS_ENABLED(CONFIG_IOSF_MBI)) {
>>>>> u32 bios_status;
>>>>> @@ -278,10 +290,12 @@ static int is_byt_cr(struct device *dev, bool *bytcr)
>>>>> /* bits 26:27 mirror PMIC options */
>>>>> bios_status = (bios_status >> 26) & 3;
>>>>> - if ((bios_status == 1) || (bios_status == 3))
>>>>> + if ((bios_status == 1) || (bios_status == 3)) {
>>>>> + dev_info(dev, "Detected Baytrail-CR platform\n");
>>>>> *bytcr = true;
>>>>> - else
>>>>> + } else {
>>>>> dev_info(dev, "BYT-CR not detected\n");
>>>>> + }
>>>>> }
>>>>> } else {
>>>>> dev_info(dev, "IOSF_MBI not enabled, no BYT-CR detection\n");
>>>>> @@ -333,10 +347,8 @@ static int sst_acpi_probe(struct platform_device *pdev)
>>>>> if (ret < 0)
>>>>> return ret;
>>>>> - ret = is_byt_cr(dev, &bytcr);
>>>>> + ret = is_byt_cr(pdev, &bytcr);
>>>>> if (!(ret < 0 || !bytcr)) {
>>>>> - dev_info(dev, "Detected Baytrail-CR platform\n");
>>>>> -
>>>>> /* override resource info */
>>>>> byt_rvp_platform_data.res_info = &bytcr_res_info;
>>>>> }
>>> _______________________________________________
>>> Alsa-devel mailing list
>>> Alsa-devel at alsa-project.org
>>> http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel at alsa-project.org
> http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
More information about the Alsa-devel
mailing list