Re: [spi-devel-general] Re: SPI

[Vitaly W. <vw...@ru...>]

Yeah... the other limitation is that if there's more than one SPI 
controller onboard, spi_write_then_read will serialize the transfers 
related to two different controllers what will have significant negative 
impact on the transfer speed (so DMA won't help increasing the speed in 
this case). Moreover, if, say, two kernel threads with different 
priorities are working with two SPI controllers respectively *priority 
inversion* may happen.

One more limitation of your SPI core: it cannot be compiled as a module. 
What are the reasons behind that BTW?

Vitaly

David Brownell wrote:

>On Tuesday 29 November 2005 10:25 am, Vitaly Wool wrote:
>
>  
>
>>So spi_write_then_read() can't ever be called from an interrupt context, 
>>right?
>>    
>>
>
>Right ... that's true of all the synchronous calls.  Kerneldoc for that
>one didn't highlight that fact though, so I've updated it.
>
>- Dave
>
>
>  
>