writing an alsa driver

Commit Message

Michele Curti June 4, 2015, 7:39 p.m. UTC

On Mon, Jun 01, 2015 at 12:06:08PM +0100, Liam Girdwood wrote:
> + Jarkko
> 
> On Fri, 2015-05-29 at 15:17 +0200, Michele Curti wrote:
> > On Fri, May 29, 2015 at 10:42:01AM +0100, Liam Girdwood wrote:
> > > On Thu, 2015-05-28 at 21:17 +0200, Michele Curti wrote:
> > > 
> > > > to let the probe continue, the system become quite unusable but I get the 
> > > > logs.
> > > > 
> > > > May 28 20:34:00 asus kernel: byt-rt5645 byt-rt5645: ASoC: CPU DAI baytrail-pcm-audio not registered
> > > > May 28 20:34:00 asus kernel: baytrail-pcm-audio baytrail-pcm-audio: initialising Byt DSP IPC
> > > > May 28 20:34:00 asus kernel: baytrail-pcm-audio baytrail-pcm-audio: initialising audio DSP id 0xf28
> > > > May 28 20:34:00 asus kernel: irq used for dsp = 6
> > > > 
> > > > ---> The IRQ number used is 6
> > > > 
> 
> Jarkko thinks the IRQ may be incorrectly reported by ACPI and that it
> could be using another ACPI index. This could probably be checked by
> decompiling the ACPi data and searching for the audio DSP.
>

Done an acpidump/iasl and (I think) the interesting part is:

    Scope (_SB)
    {
        Device (LPEA)
        {
            Name (_ADR, Zero)  // _ADR: Address
            Name (_HID, "80860F28" /* Intel SST Audio DSP */)  // _HID: Hardware ID
            Name (_CID, "80860F28" /* Intel SST Audio DSP */)  // _CID: Compatible ID
            Name (_DDN, "Intel(R) Low Power Audio Controller - 80860F28")  // _DDN: DOS Device Name
            Name (_SUB, "1043182D")  // _SUB: Subsystem ID
            Name (_UID, One)  // _UID: Unique ID
            Name (ADEP, Package (0x01)
            {
                ^I2C2.TTLV
            })
            Name (_DEP, Package (0x01)  // _DEP: Dependencies
            {
                ^I2C2.RTEK
            })
            Name (_PR0, Package (0x01)  // _PR0: Power Resources for D0
            {
                PLPE
            })
            Method (_STA, 0, NotSerialized)  // _STA: Status
            {
                If (((LPEE == 0x02) && (LPED == Zero)))
                {
                    Return (0x0F)
                }

                Return (Zero)
            }

            Method (_DIS, 0, NotSerialized)  // _DIS: Disable Device
            {
            }

            Name (RBUF, ResourceTemplate ()
            {
                Memory32Fixed (ReadWrite,
                    0x12345678,         // Address Base
                    0x00200000,         // Address Length
                    _Y04)
                Memory32Fixed (ReadWrite,
                    0xFE830000,         // Address Base
                    0x00001000,         // Address Length
                    _Y05)
                Memory32Fixed (ReadWrite,
                    0x55AA55AA,         // Address Base
                    0x00200000,         // Address Length
                    _Y06)
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x0000001D,
                }
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x00000018,
                }
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x00000019,
                }
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x0000001A,
                }
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x0000001B,
                }
                Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
                {
                    0x0000001C,
                }
            })
            Method (_CRS, 0, NotSerialized)  // _CRS: Current Resource Settings
            {
                CreateDWordField (RBUF, \_SB.LPEA._Y04._BAS, B0BA)  // _BAS: Base Address
                B0BA = LPE0 /* \LPE0 */
                CreateDWordField (RBUF, \_SB.LPEA._Y05._BAS, B1BA)  // _BAS: Base Address
                B1BA = LPE1 /* \LPE1 */
                CreateDWordField (RBUF, \_SB.LPEA._Y06._BAS, B2BA)  // _BAS: Base Address
                B2BA = LPE2 /* \LPE2 */
                Return (RBUF) /* \_SB_.LPEA.RBUF */
            }

            OperationRegion (KEYS, SystemMemory, LPE1, 0x0100)
            Field (KEYS, DWordAcc, NoLock, WriteAsZeros)
            {
                Offset (0x84), 
                PSAT,   32
            }

            PowerResource (PLPE, 0x05, 0x0000)
            {
                Method (_STA, 0, NotSerialized)  // _STA: Status
                {
                    Return (One)
                }

                Method (_ON, 0, NotSerialized)  // _ON_: Power On
                {
                    PSAT &= 0xFFFFFFFC
                    Local0 = PSAT /* \_SB_.LPEA.PSAT */
                }

                Method (_OFF, 0, NotSerialized)  // _OFF: Power Off
                {
                    PSAT |= 0x03
                    Local0 = PSAT /* \_SB_.LPEA.PSAT */
                }
            }

So, 6 possible IRQs for the LPE core, from 0x18 to 0x1D?

Changed the code and tried all six


without success, I get always the "dsp boot timeout".

I'm going to investigate why platform_get_irq() returns 6 instead of something
in the range 0x18-0x1D.. :)

Thanks for the time spent,
Michele

Comments

Jarkko Nikula June 5, 2015, 7 a.m. UTC | #1

Hi

On 06/04/2015 10:39 PM, Michele Curti wrote:
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x0000001D,
>                  }
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x00000018,
>                  }
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x00000019,
>                  }
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x0000001A,
>                  }
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x0000001B,
>                  }
>                  Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
>                  {
>                      0x0000001C,
>                  }
...
>
> So, 6 possible IRQs for the LPE core, from 0x18 to 0x1D?
>
2 for DMA, 3 for SSP ports and 1 for host IPC. I don't know does Windows 
use those others as BIOS is exposing them but Linux needs only the host 
IPC irq 0x1d and Linux DSP FW is managing itself the rest.

> Changed the code and tried all six
>
> diff --git a/sound/soc/intel/common/sst-acpi.c b/sound/soc/intel/common/sst-acpi.c
> index bb28e84..a306502 100644
> --- a/sound/soc/intel/common/sst-acpi.c
> +++ b/sound/soc/intel/common/sst-acpi.c
> @@ -149,8 +149,12 @@ static int sst_acpi_probe(struct platform_device *pdev)
>   		sst_pdata->dma_size = desc->dma_size;
>   	}
>
> -	if (desc->irqindex_host_ipc >= 0)
> -		sst_pdata->irq = platform_get_irq(pdev, desc->irqindex_host_ipc);
> +	if (desc->irqindex_host_ipc >= 0) {
> +		sst_pdata->irq = 0x1D; /* acpi-dump, from 0x18 to 0x1D */
> +		pr_info("audio dsp: IRQ # was %d, force to %d\n",
> +				platform_get_irq(pdev, desc->irqindex_host_ipc),
> +				sst_pdata->irq);
> +	}
>
>   	if (desc->resindex_lpe_base >= 0) {
>   		mmio = platform_get_resource(pdev, IORESOURCE_MEM,
>
> without success, I get always the "dsp boot timeout".
>
Above probably won't work as it doesn't set the trigger flags as 
platform_get_irq() does and the Linux irq number is not necessarily the 
same than the HW irq number through the IOAPIC routing.

Your earlier irqindex_host_ipc change should get the right interrupt for 
Linux on your machine.

-	.irqindex_host_ipc = 5,
+	.irqindex_host_ipc = 0,

Unfortunately this index is machine specific and when I looked a few 
DSDT tables earlier I didn't figure out a clean way at which index the 
interrupt should be used in runtime. Probably having quirks for index 5 
needs least amount of quirks as index 0 seems to be constantly used on 
newer machines.

Michele Curti June 8, 2015, 8:26 a.m. UTC | #2

On Fri, Jun 05, 2015 at 10:00:50AM +0300, Jarkko Nikula wrote:
> Hi
> 
> On 06/04/2015 10:39 PM, Michele Curti wrote:
> >
> >So, 6 possible IRQs for the LPE core, from 0x18 to 0x1D?
> >
> 2 for DMA, 3 for SSP ports and 1 for host IPC. I don't know does
> Windows use those others as BIOS is exposing them but Linux needs
> only the host IPC irq 0x1d and Linux DSP FW is managing itself the
> rest.
> 
> >Changed the code and tried all six
> >
> >diff --git a/sound/soc/intel/common/sst-acpi.c b/sound/soc/intel/common/sst-acpi.c
> >index bb28e84..a306502 100644
> >--- a/sound/soc/intel/common/sst-acpi.c
> >+++ b/sound/soc/intel/common/sst-acpi.c
> >@@ -149,8 +149,12 @@ static int sst_acpi_probe(struct platform_device *pdev)
> >  		sst_pdata->dma_size = desc->dma_size;
> >  	}
> >
> >-	if (desc->irqindex_host_ipc >= 0)
> >-		sst_pdata->irq = platform_get_irq(pdev, desc->irqindex_host_ipc);
> >+	if (desc->irqindex_host_ipc >= 0) {
> >+		sst_pdata->irq = 0x1D; /* acpi-dump, from 0x18 to 0x1D */
> >+		pr_info("audio dsp: IRQ # was %d, force to %d\n",
> >+				platform_get_irq(pdev, desc->irqindex_host_ipc),
> >+				sst_pdata->irq);
> >+	}
> >
> >  	if (desc->resindex_lpe_base >= 0) {
> >  		mmio = platform_get_resource(pdev, IORESOURCE_MEM,
> >
> >without success, I get always the "dsp boot timeout".
> >
> Above probably won't work as it doesn't set the trigger flags as
> platform_get_irq() does and the Linux irq number is not necessarily
> the same than the HW irq number through the IOAPIC routing.

Good to know, thanks!

> 
> Your earlier irqindex_host_ipc change should get the right interrupt
> for Linux on your machine.
> 
> -	.irqindex_host_ipc = 5,
> +	.irqindex_host_ipc = 0,
> 
> Unfortunately this index is machine specific and when I looked a few
> DSDT tables earlier I didn't figure out a clean way at which index
> the interrupt should be used in runtime. Probably having quirks for
> index 5 needs least amount of quirks as index 0 seems to be
> constantly used on newer machines.
>

Oh, understood..  Well, so for my tests I can restore this change 
thanks :)


Michele

Patch

diff --git a/sound/soc/intel/common/sst-acpi.c b/sound/soc/intel/common/sst-acpi.c
index bb28e84..a306502 100644
--- a/sound/soc/intel/common/sst-acpi.c
+++ b/sound/soc/intel/common/sst-acpi.c
@@ -149,8 +149,12 @@  static int sst_acpi_probe(struct platform_device *pdev)
 		sst_pdata->dma_size = desc->dma_size;
 	}
 
-	if (desc->irqindex_host_ipc >= 0)
-		sst_pdata->irq = platform_get_irq(pdev, desc->irqindex_host_ipc);
+	if (desc->irqindex_host_ipc >= 0) {
+		sst_pdata->irq = 0x1D; /* acpi-dump, from 0x18 to 0x1D */
+		pr_info("audio dsp: IRQ # was %d, force to %d\n",
+				platform_get_irq(pdev, desc->irqindex_host_ipc),
+				sst_pdata->irq);
+	}
 
 	if (desc->resindex_lpe_base >= 0) {
 		mmio = platform_get_resource(pdev, IORESOURCE_MEM,