Re: pxa2xx_spi with SFRM

[Ned Forrester <nforrester-/d+BM93fTQY@public.gmane.org>]

With the one small change re: 8191, below, this looks good to go.
For what it's worth, you can add below your sign-off:

Reviewed-by: Ned Forrester <nforrester-/d+BM93fTQY@public.gmane.org>

It's time to add an appropriate [Patch] subject line to the patch, as
outlined in the docs, and email it to David Brownell
<david-b-yBeKhBN/0LDR7s880joybQ@public.gmane.org>, with CC to the list.

Thanks for pushing this; I know it takes time, and you are right that it
needed to be done.

Vernon Sauder wrote:
> Ned,
> 
> Sorry for the slow response. Real world/work intrudes sometimes.
> 
> Here is another version of the patch along with responses to your
> comments.
> 
> Ned Forrester wrote:
> 
>> Vernon Sauder wrote:
>>
>> Also note that the very recent patch that I submitted, tries to address
>> a problem that you encountered with transfer lengths longer than 8192;
>> in that case, I changed the behavior from "fail" to "do it in PIO mode
>> with rate limited warning".  Some day the driver could be rewritten to
>> bust long transfers and do the pieces by DMA, but that is too ambitious
>> for now.  So you could add to the list of states that transfers longer
>> than 8191 will be PIO.  Have you tested that patch to see if it fixes
>> any of your other problems?
>>
> 
> I did not try that patch yet. I added a comment to the doc about this.

If you would try that it would be appreciated.  I was waiting for your
test to see if it fixed at least one of the MMC problems you were having
(rejected long transfers) before I ask David to push the patch upstream.

--------------------------------
> 
> Patch against Linus 2.6 tree after
> 4c246edd2550304df5b766cc841584b2bb058843. It is compile tested only.
> 
> ---
> Subject: [PATCH] pxa2xx_spi: Fix chip_info defaults and documentation.
> 
> Make the chip info structure data optional by providing reasonable
> defaults. Improve corresponding documentation.
> 
> DMA can determine appropriate dma_burst_size and thresholds
> automatically so use DMA even if dma_burst_size is not specified.
> 
> Signed-off-by: Vernon Sauder <VernonInHand-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> ---
>  Documentation/spi/pxa2xx |   31 ++++++++++++++++++++-----------
>  drivers/spi/pxa2xx_spi.c |   43 ++++++++++++++++++++++++++-----------------
>  2 files changed, 46 insertions(+), 28 deletions(-)
> 

[snip]

>  The following logic is used to determine the type of I/O to be used on
>  a per "spi_transfer" basis:
>  
> -if !enable_dma or dma_burst_size == 0 then
> +if !enable_dma then
>  	always use PIO transfers
>  
> +if spi_message.len > 8192 then

Oops.  if > 8191 (or >=8192), a count of 8192 is not possible in the DMA
controller.



-- 
Ned Forrester                                       nforrester-/d+BM93fTQY@public.gmane.org
Oceanographic Systems Lab                                  508-289-2226
Applied Ocean Physics and Engineering Dept.
Woods Hole Oceanographic Institution          Woods Hole, MA 02543, USA
http://www.whoi.edu/sbl/liteSite.do?litesiteid=7212
http://www.whoi.edu/hpb/Site.do?id=1532
http://www.whoi.edu/page.do?pid=10079


-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/