On 01/17/2017 10:42 AM, Sergei Shtylyov wrote:
> Hello!
>
> On 1/13/2017 5:58 PM, Alexandre Bailon wrote:
>
>> Despite the CPPI 4.1 is a generic DMA, it is tied to USB.
>> On the dsps, CPPI 4.1 interrupt's registers are in USBSS (the MUSB glue).
>> Currently, to enable / disable and clear interrupts, the CPPI 4.1 driver
>> maps and accesses to USBSS's register, which making CPPI 4.1 driver not
>> really generic.
>> Move the interrupt management to dsps driver.
>
> I believe it's a step in the right direction but not sure I
> understand this patchset...
>
>> Signed-off-by: Alexandre Bailon <abailon@xxxxxxxxxxxx>
>> ---
>> drivers/dma/cppi41.c | 28 +++++------------
>> drivers/usb/musb/musb_dsps.c | 73
>> ++++++++++++++++++++++++++++++++++++++++++--
>> 2 files changed, 78 insertions(+), 23 deletions(-)
>>
>> diff --git a/drivers/dma/cppi41.c b/drivers/dma/cppi41.c
>> index d5ba43a..4999e7d 100644
>> --- a/drivers/dma/cppi41.c
>> +++ b/drivers/dma/cppi41.c
>> @@ -79,14 +79,6 @@
>> #define QMGR_QUEUE_C(n) (0x2008 + (n) * 0x10)
>> #define QMGR_QUEUE_D(n) (0x200c + (n) * 0x10)
>>
>> -/* Glue layer specific */
>> -/* USBSS / USB AM335x */
>> -#define USBSS_IRQ_STATUS 0x28
>> -#define USBSS_IRQ_ENABLER 0x2c
>> -#define USBSS_IRQ_CLEARR 0x30
>> -
>> -#define USBSS_IRQ_PD_COMP (1 << 2)
>> -
>> /* Packet Descriptor */
>> #define PD2_ZERO_LENGTH (1 << 19)
>>
>> @@ -288,14 +280,8 @@ static irqreturn_t cppi41_irq(int irq, void *data)
>> {
>> struct cppi41_dd *cdd = data;
>> struct cppi41_channel *c;
>> - u32 status;
>> int i;
>>
>> - status = cppi_readl(cdd->usbss_mem + USBSS_IRQ_STATUS);
>> - if (!(status & USBSS_IRQ_PD_COMP))
>> - return IRQ_NONE;
>> - cppi_writel(status, cdd->usbss_mem + USBSS_IRQ_STATUS);
>> -
>
> You are removing this IRQ code completely, without a callback here?
I'm using dmaengine_desc_get_callback_invoke() to call that IRQ code.
I have updated cppi41_dma_callback() to invoked a platform callback,
which contain the IRQ code.
So far, the only issue I had with that approach is when
dmaengine_desc_get_callback_invoke() is not called whereas the
completion queue is empty, what should only happen during a teardown.
That's why I have updated the teardown to use
dmaengine_desc_get_callback_invoke().
>
>> for (i = QMGR_PENDING_SLOT_Q(FIST_COMPLETION_QUEUE); i <
>> QMGR_NUM_PEND;
>> i++) {
>> u32 val;
> [...]
>> diff --git a/drivers/usb/musb/musb_dsps.c b/drivers/usb/musb/musb_dsps.c
>> index 9f125e1..2adfe35 100644
>> --- a/drivers/usb/musb/musb_dsps.c
>> +++ b/drivers/usb/musb/musb_dsps.c
> [...]
>> @@ -145,6 +146,13 @@ static const struct debugfs_reg32
>> dsps_musb_regs[] = {
>> { "mode", 0xe8 },
>> };
>>
>> +/* USBSS / USB AM335x */
>> +#define USBSS_IRQ_STATUS 0x28
>> +#define USBSS_IRQ_ENABLER 0x2c
>> +#define USBSS_IRQ_CLEARR 0x30
>
> Please indent the values with tabs only, like it was done originally
> above...
I will fixt it.
>
> [...]
>
> MBR, Sergei
>
Best Regards,
Alexandre
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
