Re: [PATCH v2 0/2] Add Qualcomm MSM ADM DMAEngine driver

[Ravi Kumar V <kumarrav@codeaurora.org>]

On 1/20/2012 7:01 PM, Vinod Koul wrote:
> On Fri, 2012-01-20 at 18:00 +0530, Ravi Kumar V wrote:
>> On 1/17/2012 7:15 PM, Vinod Koul wrote:
>>> On Fri, 2012-01-06 at 18:17 +0530, Ravi Kumar V wrote:
>>> <sorry for delayed review, was on vacation and now traveling>
>>>>
>>>> As our ADM Scatter-gather hardware needs
>>>> -32-bit command configuration parameter
>>>>    apart from
>>>> -32-bit source address
>>>> -32-bit destination address
>>>> -16-bit length
>>>>
>>>> So,we have added new parameter in struct scatterlist to support xfer
>>>> descriptor
>>>> specific private data, and for supporting ADM Box mode DMA we added
>>>> new
>>>> API and data structure.
>>> what do you mean by "ADM Box mode"?
>>>
>>
>> ADM Box mode is a interleaved type of DMA where data from rows of equal
>> length and equal distance(bytes) between each other are transferred to
>> similar pattern of rows.
>> Each row length and distance between each row in destination pattern may
>> not be equal to source pattern.
>> Distance between beginning of any two rows are always greater than row
>> length.
>> Example:
>> If 4 rows of 16 bytes each are arranged such that distance between
>> beginning of any two rows are 32 bytes.
>> Now they can be transferred using BOX mode to destination pattern
>> arranged in 2 rows of 32 bytes each and distance between them can be any
>> lets say 128 bytes.
>>
>> Source pattern:
>> 4 data rows starts address 0th byte.
>>
>> 0-----16-bytes-data-----15
>> 16    16 bytes void     31
>> 32---------data---------47
>> 48         void         63
>> 64---------data---------79
>> 80         void         95
>> 96---------data--------111
>>
>> Transferred to destination
>> Destination pattern:
>> 2 rows
>> 0----------32-bytes-data----------31
>> 32         96 bytes void         127
>> 128-------------data-------------159
>>
> Sounds like you should be using interleaved API
>
> Please see this and read the patch history, which has nice details about
> this
>
> /**
>   * struct dma_interleaved_template - Template to convey DMAC the
> transfer pattern
>   *       and attributes.
>   * @src_start: Bus address of source for the first chunk.
>   * @dst_start: Bus address of destination for the first chunk.
>   * @dir: Specifies the type of Source and Destination.
>   * @src_inc: If the source address increments after reading from it.
>   * @dst_inc: If the destination address increments after writing to it.
>   * @src_sgl: If the 'icg' of sgl[] applies to Source (scattered read).
>   *              Otherwise, source is read contiguously (icg ignored).
>   *              Ignored if src_inc is false.
>   * @dst_sgl: If the 'icg' of sgl[] applies to Destination (scattered write).
>   *              Otherwise, destination is filled contiguously (icg ignored).
>   *              Ignored if dst_inc is false.
>   * @numf: Number of frames in this template.
>   * @frame_size: Number of chunks in a frame i.e, size of sgl[].
>   * @sgl: Array of {chunk,icg} pairs that make up a frame.
>   */
> struct dma_interleaved_template {
>          dma_addr_t src_start;
>          dma_addr_t dst_start;
>          enum dma_transfer_direction dir;
>          bool src_inc;
>          bool dst_inc;
>          bool src_sgl;
>          bool dst_sgl;
>          size_t numf;
>          size_t frame_size;
>          struct data_chunk sgl[0];
> };
> using the interleaved API:
>          struct dma_async_tx_descriptor *(*device_prep_interleaved_dma)(
>                  struct dma_chan *chan, struct dma_interleaved_template *xt,
>                  unsigned long flags);
>
>

If some changes are made in interleave API then it can support our BOX 
mode. Here in interleaved template he is assuming destination pattern as 
can be contiguous or same as source pattern, but in our case destination 
pattern is different from source pattern.
So if a new parameter destination data chunk is added in "struct 
dma_interleaved_template" structure then it can support different 
destination pattern.
Also it will good if you can provide another parameter for passing 
private data to dma driver.

Please can you review my other patches also where we replied to some of 
your questions about passing private data for SG mode.

Thanks
Ravi Kumar

-- 
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.