[question] some question about the residue

[question] some question about the residue

[Elen Song]

Hi All:

        I got a question about the residue in struct dma_tx_state:

I wondering what residue means, is it the current descriptor residue or

        the total dma buffer residue?

Thanks

Best Regards

Elen Song

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20121126/20b915fa/attachment-0001.html>

[question] some question about the residue

[Elen Song]

Hi All:

	I got a question about the residue in struct dma_tx_state:

	I wonder what residue means, is it the current descriptor residue or

	the total dma buffer residue?

Thanks

Best Regards

Elen Song

-- 
1.7.9.5

[question] some question about the residue

[Vinod Koul]

On Mon, 2012-11-26 at 14:43 +0800, Elen Song wrote:
> Hi All:
> 
> 	I got a question about the residue in struct dma_tx_state:
> 
> 	I wonder what residue means, is it the current descriptor residue or
> 
> 	the total dma buffer residue?

You read the status for descriptor, so it means this is residue for said
descriptor.


-- 
Vinod Koul
Intel Corp.

[question] some question about the residue

[Elen Song]

On 2012-11-26 14:49, Vinod Koul wrote:
> On Mon, 2012-11-26 at 14:43 +0800, Elen Song wrote:
>> Hi All:
>>
>> 	I got a question about the residue in struct dma_tx_state:
>>
>> 	I wonder what residue means, is it the current descriptor residue or
>>
>> 	the total dma buffer residue?
> You read the status for descriptor, so it means this is residue for said
> descriptor.
>
>
Hi Koul:

Well , I don't get it , if I read the status of a LLI, is it means the 
residue of total LLI length?

I saw driver/dma/amba-pl08x.c drivers/dma/coh901318.c, if read status of 
a LLI, residue is the remain of LLI, not only
current descriptor.

1 patch from Russel King how he realize device_tx_status said:

[PATCH] ARM: PL08x: clean up LLI lookup
As the LLI list is an array, we can use maths to locate which LLI
index we're currently at, and then sum up the remaining LLI entries
until we reach the end of the list.

Is it right?

best regards

[question] some question about the residue

[Russell King - ARM Linux]

On Mon, Nov 26, 2012 at 03:35:42PM +0800, Elen Song wrote:
> On 2012-11-26 14:49, Vinod Koul wrote:
>> On Mon, 2012-11-26 at 14:43 +0800, Elen Song wrote:
>>> Hi All:
>>>
>>> 	I got a question about the residue in struct dma_tx_state:
>>>
>>> 	I wonder what residue means, is it the current descriptor residue or
>>>
>>> 	the total dma buffer residue?
>> You read the status for descriptor, so it means this is residue for said
>> descriptor.
>>
>>
> Hi Koul:
>
> Well , I don't get it , if I read the status of a LLI, is it means the  
> residue of total LLI length?
>
> I saw driver/dma/amba-pl08x.c drivers/dma/coh901318.c, if read status of  
> a LLI, residue is the remain of LLI, not only
> current descriptor.
>
> 1 patch from Russel King how he realize device_tx_status said:
>
> [PATCH] ARM: PL08x: clean up LLI lookup
> As the LLI list is an array, we can use maths to locate which LLI
> index we're currently at, and then sum up the remaining LLI entries
> until we reach the end of the list.
>
> Is it right?

It is as far as fixing a pretty major bug in the driver, but it doesn't
fix the other bug in the driver where it sums up all outstanding
descriptors, which pre-dates the understanding of 'residue'.

[question] some question about the residue

[Elen Song]

On 2012-11-26 18:22, Russell King - ARM Linux wrote:
> On Mon, Nov 26, 2012 at 03:35:42PM +0800, Elen Song wrote:
>> On 2012-11-26 14:49, Vinod Koul wrote:
>>> On Mon, 2012-11-26 at 14:43 +0800, Elen Song wrote:
>>>> Hi All:
>>>>
>>>> 	I got a question about the residue in struct dma_tx_state:
>>>>
>>>> 	I wonder what residue means, is it the current descriptor residue or
>>>>
>>>> 	the total dma buffer residue?
>>> You read the status for descriptor, so it means this is residue for said
>>> descriptor.
>>>
>>>
>> Hi Koul:
>>
>> Well , I don't get it , if I read the status of a LLI, is it means the
>> residue of total LLI length?
>>
>> I saw driver/dma/amba-pl08x.c drivers/dma/coh901318.c, if read status of
>> a LLI, residue is the remain of LLI, not only
>> current descriptor.
>>
>> 1 patch from Russel King how he realize device_tx_status said:
>>
>> [PATCH] ARM: PL08x: clean up LLI lookup
>> As the LLI list is an array, we can use maths to locate which LLI
>> index we're currently at, and then sum up the remaining LLI entries
>> until we reach the end of the list.
>>
>> Is it right?
> It is as far as fixing a pretty major bug in the driver, but it doesn't
> fix the other bug in the driver where it sums up all outstanding
> descriptors, which pre-dates the understanding of 'residue'.
So residue actually means current active descriptor remain, it should 
not sums up all outstanding descriptors.
the way amba-p108x.c and coh901318.c is wrong , right?

best regards

[question] some question about the residue

[Russell King - ARM Linux]

On Mon, Nov 26, 2012 at 06:37:28PM +0800, Elen Song wrote:
> So residue actually means current active descriptor remain, it should  

No, not the current active descriptor - the transfer identified by the
cookie, and only that transfer.

[question] some question about the residue

[Elen Song]

On 2012-11-26 18:40, Russell King - ARM Linux wrote:
> On Mon, Nov 26, 2012 at 06:37:28PM +0800, Elen Song wrote:
>> So residue actually means current active descriptor remain, it should
> No, not the current active descriptor - the transfer identified by the
> cookie, and only that transfer.
Hi Russell King:

     Thank you for your explanations.
     Here I got an example, perhaps you can help me figure out if it is 
correct.

//allocate a cyclic transfer with 2 descriptors, a cookie identify 
current transfer.
desc = dmaengine_prep_dma_cyclic(chan,
                                 sg_dma_address,
                                 sg_dma_len,
                                 sg_dma_len/2,
                                 DMA_DEV_TO_MEM);
cookie= dmaengine_submit(desc);

//if transfer in progress, state.residue is current transfer 
remain(sg_dma_len remain), not current descriptor remain(sg_dma_len/2 
remain).
device_tx_status(chan,cookie, &state);

     Is it right?

Best Regards

[question] some question about the residue

[Russell King - ARM Linux]

On Mon, Nov 26, 2012 at 02:43:17PM +0800, Elen Song wrote:
> Hi All:
> 
> 	I got a question about the residue in struct dma_tx_state:
> 
> 	I wonder what residue means, is it the current descriptor residue or
> 
> 	the total dma buffer residue?

It's the number of bytes remaining to be transferred on the descriptor
identified by the cookie.  If that transfer has not been started, it
will be the total number of bytes in that transfer.  If it has been
completed, it will be zero.  If it is in progress, it will be the number
of bytes to be transferred before the transfer identified by that cookie
is marked as complete.