Tearing down DMA transfer setup after DMA client has finished

[Måns Rullgård]

Mason <slash.tmp at free.fr> writes:

> On 23/11/2016 13:13, Måns Rullgård wrote:
>
>> Mason wrote:
>> 
>>> On my platform, setting up a DMA transfer is a two-step process:
>>>
>>> 1) configure the "switch box" to connect a device to a memory channel
>>> 2) configure the transfer details (address, size, command)
>>>
>>> When the transfer is done, the sbox setup can be torn down,
>>> and the DMA driver can start another transfer.
>>>
>>> The current software architecture for my NFC (NAND Flash controller)
>>> driver is as follows (for one DMA transfer).
>>>
>>>   sg_init_one
>>>   dma_map_sg
>>>   dmaengine_prep_slave_sg
>>>   dmaengine_submit
>>>   dma_async_issue_pending
>>>   configure_NFC_transfer
>>>   wait_for_IRQ_from_DMA_engine // via DMA_PREP_INTERRUPT
>>>   wait_for_NFC_idle
>>>   dma_unmap_sg
>>>
>>> The problem is that the DMA driver tears down the sbox setup
>>> as soon as it receives the IRQ. However, when writing to the
>>> device, the interrupt only means "I have pushed all data from
>>> memory to the memory channel". These data have not reached
>>> the device yet, and may still be "in flight". Thus the sbox
>>> setup can only be torn down after the NFC is idle.
>>>
>>> How do I call back into the DMA driver after wait_for_NFC_idle,
>>> to request sbox tear down?
>>>
>>> The new architecture would become:
>>>
>>>   sg_init_one
>>>   dma_map_sg
>>>   dmaengine_prep_slave_sg
>>>   dmaengine_submit
>>>   dma_async_issue_pending
>>>   configure_NFC_transfer
>>>   wait_for_IRQ_from_DMA_engine // via DMA_PREP_INTERRUPT
>>>   wait_for_NFC_idle
>>>   request_sbox_tear_down /*** HOW TO DO THAT ***/
>>>   dma_unmap_sg
>>>
>>> As far as I can tell, my NFC driver should call dmaengine_synchronize ??
>>> (In other words request_sbox_tear_down == dmaengine_synchronize)
>>>
>>> So the DMA driver should implement the device_synchronize hook,
>>> and tear the sbox down in that function.
>>>
>>> Is that correct? Or am I on the wrong track?
>> 
>> dmaengine_synchronize() is not meant for this.  See the documentation:
>> 
>> /**
>>  * dmaengine_synchronize() - Synchronize DMA channel termination
>>  * @chan: The channel to synchronize
>>  *
>>  * Synchronizes to the DMA channel termination to the current context. When this
>>  * function returns it is guaranteed that all transfers for previously issued
>>  * descriptors have stopped and and it is safe to free the memory assoicated
>>  * with them. Furthermore it is guaranteed that all complete callback functions
>>  * for a previously submitted descriptor have finished running and it is safe to
>>  * free resources accessed from within the complete callbacks.
>>  *
>>  * The behavior of this function is undefined if dma_async_issue_pending() has
>>  * been called between dmaengine_terminate_async() and this function.
>>  *
>>  * This function must only be called from non-atomic context and must not be
>>  * called from within a complete callback of a descriptor submitted on the same
>>  * channel.
>>  */
>> 
>> This is for use after a dmaengine_terminate_async() call to wait for the
>> dma engine to finish whatever it was doing.  This is not the problem
>> here.  Your problem is that the dma engine interrupt fires before the
>> transfer is actually complete.  Although you get an indication from the
>> target device when it has received all the data, there is no way to make
>> the dma driver wait for this.
>
> Hello Mans,
>
> I'm confused. Are you saying there is no solution to my problem
> within the existing DMA framework?
>
> In its current form, the tangox-dma.c driver will fail randomly
> for writes to a device (SATA, NFC).
>
> Maybe an extra hook can be added to the DMA framework?
>
> I'd like to hear from the framework's maintainers. Perhaps they
> can provide some guidance.

You could have the dma descriptor callback wait for the receiving device
to finish.  Bear in mind this runs from a tasklet, so it's not allowed
to sleep.

-- 
Måns Rullgård