[PATCH 0/6] usb: dwc3: DMA API usage

[PATCH 0/6] usb: dwc3: DMA API usage

[Felipe Balbi]

Hi,

this should make dwc3's DMA API usage more
compliant. This is still *UNTESTED* patches but I'd
like people to give it a round of review to make
sure I didn't miss anything.

I have questions about dev->dma_parms, what should
be done with that ?

If there's anything I missed, please _do_ let me
know. Also, I think Keystone's DMA mask could be
more than 32 bits. IIRC, Keystone's DMA can address
more than 32-bits, right Santosh ?

patches are also available in my k.org tree on
branch dwc3-dma-mask for those of you willing to
give it a go on real HW.

Felipe Balbi (6):
  usb: dwc3: st: fix dma initialization
  usb: dwc3: omap: don't access DMA bits directly
  usb: dwc3: keystone: initialize dma_mask and coherent
  usb: dwc3: of-simple: initialize dma_mask and coherent
  usb: dwc3: core: don't access DMA bits directly
  usb: dwc3: host: don't access DMA bits directly

 drivers/usb/dwc3/core.c           | 6 +-----
 drivers/usb/dwc3/dwc3-keystone.c  | 5 +----
 drivers/usb/dwc3/dwc3-of-simple.c | 1 +
 drivers/usb/dwc3/dwc3-omap.c      | 4 +---
 drivers/usb/dwc3/dwc3-st.c        | 3 ++-
 drivers/usb/dwc3/host.c           | 5 +----
 6 files changed, 7 insertions(+), 17 deletions(-)

-- 
2.7.0

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 1/6] usb: dwc3: st: fix dma initialization

[Felipe Balbi]

Current code was basically a no-op. As it was
initializing dev->coherent_dma_mask with
dev->coherent_dma_mask. Due to lack of more
information, just initialize it to a 32-bit mask.

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/dwc3-st.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/dwc3/dwc3-st.c b/drivers/usb/dwc3/dwc3-st.c
index 5c0adb9c6fb2..33b85ae4e698 100644
--- a/drivers/usb/dwc3/dwc3-st.c
+++ b/drivers/usb/dwc3/dwc3-st.c
@@ -212,7 +212,8 @@ static int st_dwc3_probe(struct platform_device *pdev)
 	if (IS_ERR(regmap))
 		return PTR_ERR(regmap);
 
-	dma_set_coherent_mask(dev, dev->coherent_dma_mask);
+	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
+
 	dwc3_data->dev = dev;
 	dwc3_data->regmap = regmap;
 
-- 
2.7.0

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Felipe Balbi]

Instead of having a static global just for
initializing dma_mask directly, let's use
dma_coerce_mask_and_coherent() for that.

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/dwc3-omap.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
index 22e9606d8e08..c219118bfda0 100644
--- a/drivers/usb/dwc3/dwc3-omap.c
+++ b/drivers/usb/dwc3/dwc3-omap.c
@@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
 	dwc3_omap_write_irqmisc_clr(omap, reg);
 }
 
-static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);

[PATCH 3/6] usb: dwc3: keystone: initialize dma_mask and coherent

[Felipe Balbi]

dma_get_mask(dev) just returns
dev->dma_mask. Current code was using dma_get_mask()
return to initialize dev->dma_mask. That's
essentially a no-op. Instead, let's default to a
32-bit DMA mask passed to
dma_coerce_mask_and_coherent().

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/dwc3-keystone.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/usb/dwc3/dwc3-keystone.c b/drivers/usb/dwc3/dwc3-keystone.c
index 2be268d2423d..53bbc9dccf7e 100644
--- a/drivers/usb/dwc3/dwc3-keystone.c
+++ b/drivers/usb/dwc3/dwc3-keystone.c
@@ -39,8 +39,6 @@
 #define USBSS_IRQ_COREIRQ_EN	BIT(0)
 #define USBSS_IRQ_COREIRQ_CLR	BIT(0)
 
-static u64 kdwc3_dma_mask;

[PATCH 4/6] usb: dwc3: of-simple: initialize dma_mask and coherent

[Felipe Balbi]

This will make sure users of dwc3-of-simple will
have proper DMA bits initialized for dwc3 core.

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/dwc3-of-simple.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/usb/dwc3/dwc3-of-simple.c b/drivers/usb/dwc3/dwc3-of-simple.c
index 974335377d9f..b33c145a7e29 100644
--- a/drivers/usb/dwc3/dwc3-of-simple.c
+++ b/drivers/usb/dwc3/dwc3-of-simple.c
@@ -62,6 +62,7 @@ static int dwc3_of_simple_probe(struct platform_device *pdev)
 		return -ENOMEM;
 
 	simple->dev = dev;
+	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
 
 	for (i = 0; i < simple->num_clocks; i++) {
 		struct clk	*clk;
-- 
2.7.0

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 5/6] usb: dwc3: core: don't access DMA bits directly

[Felipe Balbi]

instead of manually copying DMA bits from parent
device, we should let DMA API do its job.

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/core.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
index 17fd81447c9f..d601de20e1cd 100644
--- a/drivers/usb/dwc3/core.c
+++ b/drivers/usb/dwc3/core.c
@@ -981,11 +981,7 @@ static int dwc3_probe(struct platform_device *pdev)
 
 	spin_lock_init(&dwc->lock);
 
-	if (!dev->dma_mask) {
-		dev->dma_mask = dev->parent->dma_mask;
-		dev->dma_parms = dev->parent->dma_parms;
-		dma_set_coherent_mask(dev, dev->parent->coherent_dma_mask);
-	}
+	dma_coerce_mask_and_coherent(dev, dma_get_mask(dev->parent));
 
 	pm_runtime_enable(dev);
 	pm_runtime_get_sync(dev);
-- 
2.7.0

--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 6/6] usb: dwc3: host: don't access DMA bits directly

[Felipe Balbi]

instead of manually copying DMA bits from parent
device, we should let DMA API do its job.

Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
---
 drivers/usb/dwc3/host.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c
index c679f63783ae..186f1704a16b 100644
--- a/drivers/usb/dwc3/host.c
+++ b/drivers/usb/dwc3/host.c
@@ -32,11 +32,8 @@ int dwc3_host_init(struct dwc3 *dwc)
 		return -ENOMEM;
 	}
 
-	dma_set_coherent_mask(&xhci->dev, dwc->dev->coherent_dma_mask);

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Grygorii Strashko]

On 04/02/2016 11:28 AM, Felipe Balbi wrote:
> Instead of having a static global just for
> initializing dma_mask directly, let's use
> dma_coerce_mask_and_coherent() for that.
> 
> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
> ---
>   drivers/usb/dwc3/dwc3-omap.c | 4 +---
>   1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
> index 22e9606d8e08..c219118bfda0 100644
> --- a/drivers/usb/dwc3/dwc3-omap.c
> +++ b/drivers/usb/dwc3/dwc3-omap.c
> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>   	dwc3_omap_write_irqmisc_clr(omap, reg);
>   }
>   
> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
> -
>   static int dwc3_omap_id_notifier(struct notifier_block *nb,
>   	unsigned long event, void *ptr)
>   {
> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>   	omap->irq	= irq;
>   	omap->base	= base;
>   	omap->vbus_reg	= vbus_reg;
> -	dev->dma_mask	= &dwc3_omap_dma_mask;
> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));

I think, It'll be better to just remove DMA configuration code
from this driver and other drivers which support DT-boot mode only.



-- 
regards,
-grygorii
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 5/6] usb: dwc3: core: don't access DMA bits directly

[Grygorii Strashko]

On 04/02/2016 11:28 AM, Felipe Balbi wrote:
> instead of manually copying DMA bits from parent
> device, we should let DMA API do its job.
> 
> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
> ---
>   drivers/usb/dwc3/core.c | 6 +-----
>   1 file changed, 1 insertion(+), 5 deletions(-)
> 
> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
> index 17fd81447c9f..d601de20e1cd 100644
> --- a/drivers/usb/dwc3/core.c
> +++ b/drivers/usb/dwc3/core.c
> @@ -981,11 +981,7 @@ static int dwc3_probe(struct platform_device *pdev)
>   
>   	spin_lock_init(&dwc->lock);
>   
> -	if (!dev->dma_mask) {
> -		dev->dma_mask = dev->parent->dma_mask;
> -		dev->dma_parms = dev->parent->dma_parms;

Here, and in most of other patches you've dropped dma_parms copying -
Is it expected?


> -		dma_set_coherent_mask(dev, dev->parent->coherent_dma_mask);
> -	}
> +	dma_coerce_mask_and_coherent(dev, dma_get_mask(dev->parent));


No. Above if case should stay, otherwise, already valid, DMA configuration 
might be overwritten:

commit 19bacdc925055f020ad36da04bd72dc8b28637b8
Author: Heikki Krogerus <heikki.krogerus-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
Date:   Wed Sep 24 11:00:38 2014 +0300

    usb: dwc3: core: only setting the dma_mask when needed
    
    If the probe drivers have already set the dma_mask, not
    replacing the value.
    
    Signed-off-by: Heikki Krogerus <heikki.krogerus-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
    Signed-off-by: Felipe Balbi <balbi-l0cyMroinI0@public.gmane.org>


-- 
regards,
-grygorii
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Felipe Balbi]

[-- Attachment #1: Type: text/plain, Size: 1465 bytes --]

Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>> Instead of having a static global just for
>> initializing dma_mask directly, let's use
>> dma_coerce_mask_and_coherent() for that.
>> 
>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>> ---
>>   drivers/usb/dwc3/dwc3-omap.c | 4 +---
>>   1 file changed, 1 insertion(+), 3 deletions(-)
>> 
>> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
>> index 22e9606d8e08..c219118bfda0 100644
>> --- a/drivers/usb/dwc3/dwc3-omap.c
>> +++ b/drivers/usb/dwc3/dwc3-omap.c
>> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>>   	dwc3_omap_write_irqmisc_clr(omap, reg);
>>   }
>>   
>> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
>> -
>>   static int dwc3_omap_id_notifier(struct notifier_block *nb,
>>   	unsigned long event, void *ptr)
>>   {
>> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>>   	omap->irq	= irq;
>>   	omap->base	= base;
>>   	omap->vbus_reg	= vbus_reg;
>> -	dev->dma_mask	= &dwc3_omap_dma_mask;
>> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
>
> I think, It'll be better to just remove DMA configuration code
> from this driver and other drivers which support DT-boot mode only.

I don't have HW, can you test that on AM57x and/or AM437x ?

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]

Re: [PATCH 5/6] usb: dwc3: core: don't access DMA bits directly

[Felipe Balbi]

[-- Attachment #1: Type: text/plain, Size: 1330 bytes --]

Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:

> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>> instead of manually copying DMA bits from parent
>> device, we should let DMA API do its job.
>> 
>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>> ---
>>   drivers/usb/dwc3/core.c | 6 +-----
>>   1 file changed, 1 insertion(+), 5 deletions(-)
>> 
>> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
>> index 17fd81447c9f..d601de20e1cd 100644
>> --- a/drivers/usb/dwc3/core.c
>> +++ b/drivers/usb/dwc3/core.c
>> @@ -981,11 +981,7 @@ static int dwc3_probe(struct platform_device *pdev)
>>   
>>   	spin_lock_init(&dwc->lock);
>>   
>> -	if (!dev->dma_mask) {
>> -		dev->dma_mask = dev->parent->dma_mask;
>> -		dev->dma_parms = dev->parent->dma_parms;
>
> Here, and in most of other patches you've dropped dma_parms copying -
> Is it expected?

I mentioned in cover letter that I don't know exactly what's the proper
way of dealing with dma_parms.

>> -		dma_set_coherent_mask(dev, dev->parent->coherent_dma_mask);
>> -	}
>> +	dma_coerce_mask_and_coherent(dev, dma_get_mask(dev->parent));
>
>
> No. Above if case should stay, otherwise, already valid, DMA configuration 
> might be overwritten:

okay.

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Grygorii Strashko]

On 04/05/2016 08:51 AM, Felipe Balbi wrote:
> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>>> Instead of having a static global just for
>>> initializing dma_mask directly, let's use
>>> dma_coerce_mask_and_coherent() for that.
>>>
>>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>>> ---
>>>    drivers/usb/dwc3/dwc3-omap.c | 4 +---
>>>    1 file changed, 1 insertion(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
>>> index 22e9606d8e08..c219118bfda0 100644
>>> --- a/drivers/usb/dwc3/dwc3-omap.c
>>> +++ b/drivers/usb/dwc3/dwc3-omap.c
>>> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>>>    	dwc3_omap_write_irqmisc_clr(omap, reg);
>>>    }
>>>    
>>> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
>>> -
>>>    static int dwc3_omap_id_notifier(struct notifier_block *nb,
>>>    	unsigned long event, void *ptr)
>>>    {
>>> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>>>    	omap->irq	= irq;
>>>    	omap->base	= base;
>>>    	omap->vbus_reg	= vbus_reg;
>>> -	dev->dma_mask	= &dwc3_omap_dma_mask;
>>> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
>>
>> I think, It'll be better to just remove DMA configuration code
>> from this driver and other drivers which support DT-boot mode only.
> 
> I don't have HW, can you test that on AM57x and/or AM437x ?
> 

Yes. I can. With below change I see no issues with USB in Host mode
on my AM57x board:

diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
index 16cb14f..d5feaf9 100644
--- a/drivers/usb/dwc3/dwc3-omap.c
+++ b/drivers/usb/dwc3/dwc3-omap.c
@@ -367,8 +367,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
        dwc3_omap_write_irqmisc_clr(omap, reg);
 }
 
-static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Felipe Balbi]

[-- Attachment #1: Type: text/plain, Size: 1894 bytes --]

Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
> On 04/05/2016 08:51 AM, Felipe Balbi wrote:
>> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>>> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>>>> Instead of having a static global just for
>>>> initializing dma_mask directly, let's use
>>>> dma_coerce_mask_and_coherent() for that.
>>>>
>>>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>>>> ---
>>>>    drivers/usb/dwc3/dwc3-omap.c | 4 +---
>>>>    1 file changed, 1 insertion(+), 3 deletions(-)
>>>>
>>>> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
>>>> index 22e9606d8e08..c219118bfda0 100644
>>>> --- a/drivers/usb/dwc3/dwc3-omap.c
>>>> +++ b/drivers/usb/dwc3/dwc3-omap.c
>>>> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>>>>    	dwc3_omap_write_irqmisc_clr(omap, reg);
>>>>    }
>>>>    
>>>> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
>>>> -
>>>>    static int dwc3_omap_id_notifier(struct notifier_block *nb,
>>>>    	unsigned long event, void *ptr)
>>>>    {
>>>> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>>>>    	omap->irq	= irq;
>>>>    	omap->base	= base;
>>>>    	omap->vbus_reg	= vbus_reg;
>>>> -	dev->dma_mask	= &dwc3_omap_dma_mask;
>>>> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
>>>
>>> I think, It'll be better to just remove DMA configuration code
>>> from this driver and other drivers which support DT-boot mode only.
>> 
>> I don't have HW, can you test that on AM57x and/or AM437x ?
>> 
>
> Yes. I can. With below change I see no issues with USB in Host mode
> on my AM57x board:

what about peripheral side ? Host mode worked fine on K2 devices, only
peripheral suffered from the bad DMA API usage.

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Grygorii Strashko]

On 04/05/2016 01:29 PM, Felipe Balbi wrote:
> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>> On 04/05/2016 08:51 AM, Felipe Balbi wrote:
>>> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>>>> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>>>>> Instead of having a static global just for
>>>>> initializing dma_mask directly, let's use
>>>>> dma_coerce_mask_and_coherent() for that.
>>>>>
>>>>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>>>>> ---
>>>>>     drivers/usb/dwc3/dwc3-omap.c | 4 +---
>>>>>     1 file changed, 1 insertion(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
>>>>> index 22e9606d8e08..c219118bfda0 100644
>>>>> --- a/drivers/usb/dwc3/dwc3-omap.c
>>>>> +++ b/drivers/usb/dwc3/dwc3-omap.c
>>>>> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>>>>>     	dwc3_omap_write_irqmisc_clr(omap, reg);
>>>>>     }
>>>>>     
>>>>> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
>>>>> -
>>>>>     static int dwc3_omap_id_notifier(struct notifier_block *nb,
>>>>>     	unsigned long event, void *ptr)
>>>>>     {
>>>>> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>>>>>     	omap->irq	= irq;
>>>>>     	omap->base	= base;
>>>>>     	omap->vbus_reg	= vbus_reg;
>>>>> -	dev->dma_mask	= &dwc3_omap_dma_mask;
>>>>> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
>>>>
>>>> I think, It'll be better to just remove DMA configuration code
>>>> from this driver and other drivers which support DT-boot mode only.
>>>
>>> I don't have HW, can you test that on AM57x and/or AM437x ?
>>>
>>
>> Yes. I can. With below change I see no issues with USB in Host mode
>> on my AM57x board:
> 
> what about peripheral side ? Host mode worked fine on K2 devices, only
> peripheral suffered from the bad DMA API usage.
> 

As I've mentioned before (and we verified that) on newer kernels 
peripheral mode has been fixed by commit:

7ace8fc usb: gadget: udc: core: Fix argument of dma_map_single for IOMMU

Also, on OMAP devices dwc3 core device is created from DT now  (dwc3-omap calls 
of_platform_populate()), so it's safe to drop manual DMA configuration code
from this driver (and if statement should be kept unchanged in dwc3/core.c
- as per my comment on your patch 5).

-- 
regards,
-grygorii
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 2/6] usb: dwc3: omap: don't access DMA bits directly

[Felipe Balbi]

[-- Attachment #1: Type: text/plain, Size: 2671 bytes --]


Hi,

Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
> On 04/05/2016 01:29 PM, Felipe Balbi wrote:
>> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>>> On 04/05/2016 08:51 AM, Felipe Balbi wrote:
>>>> Grygorii Strashko <grygorii.strashko-l0cyMroinI0@public.gmane.org> writes:
>>>>> On 04/02/2016 11:28 AM, Felipe Balbi wrote:
>>>>>> Instead of having a static global just for
>>>>>> initializing dma_mask directly, let's use
>>>>>> dma_coerce_mask_and_coherent() for that.
>>>>>>
>>>>>> Signed-off-by: Felipe Balbi <felipe.balbi-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>
>>>>>> ---
>>>>>>     drivers/usb/dwc3/dwc3-omap.c | 4 +---
>>>>>>     1 file changed, 1 insertion(+), 3 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/usb/dwc3/dwc3-omap.c b/drivers/usb/dwc3/dwc3-omap.c
>>>>>> index 22e9606d8e08..c219118bfda0 100644
>>>>>> --- a/drivers/usb/dwc3/dwc3-omap.c
>>>>>> +++ b/drivers/usb/dwc3/dwc3-omap.c
>>>>>> @@ -331,8 +331,6 @@ static void dwc3_omap_disable_irqs(struct dwc3_omap *omap)
>>>>>>     	dwc3_omap_write_irqmisc_clr(omap, reg);
>>>>>>     }
>>>>>>     
>>>>>> -static u64 dwc3_omap_dma_mask = DMA_BIT_MASK(32);
>>>>>> -
>>>>>>     static int dwc3_omap_id_notifier(struct notifier_block *nb,
>>>>>>     	unsigned long event, void *ptr)
>>>>>>     {
>>>>>> @@ -490,7 +488,7 @@ static int dwc3_omap_probe(struct platform_device *pdev)
>>>>>>     	omap->irq	= irq;
>>>>>>     	omap->base	= base;
>>>>>>     	omap->vbus_reg	= vbus_reg;
>>>>>> -	dev->dma_mask	= &dwc3_omap_dma_mask;
>>>>>> +	dma_coerce_mask_and_coherent(dev, DMA_BIT_MASK(32));
>>>>>
>>>>> I think, It'll be better to just remove DMA configuration code
>>>>> from this driver and other drivers which support DT-boot mode only.
>>>>
>>>> I don't have HW, can you test that on AM57x and/or AM437x ?
>>>>
>>>
>>> Yes. I can. With below change I see no issues with USB in Host mode
>>> on my AM57x board:
>> 
>> what about peripheral side ? Host mode worked fine on K2 devices, only
>> peripheral suffered from the bad DMA API usage.
>> 
>
> As I've mentioned before (and we verified that) on newer kernels 
> peripheral mode has been fixed by commit:
>
> 7ace8fc usb: gadget: udc: core: Fix argument of dma_map_single for IOMMU
>
> Also, on OMAP devices dwc3 core device is created from DT now  (dwc3-omap calls 
> of_platform_populate()), so it's safe to drop manual DMA configuration code
> from this driver (and if statement should be kept unchanged in dwc3/core.c
> - as per my comment on your patch 5).

fair enough, all I'm missing now is a proper patch ;-)

-- 
balbi

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]