> On Wed, Dec 10, 2014 at 2:21 PM, Wolfram Sang <wsa@xxxxxxxxxxxxx> wrote: > > Okay, here is my take on solving the problem found by Geert. I still don't like > > it much but it is not as bad as I expected it to be ;) Let me know what you > > think. > > Thanks, it's indeed less ugly than I would have expected ;-) :D > Note that in spi-rspi.c and spi-sh-msiof, any error returned by *_request_dma() > is considered an error, and -EPROBE_DEFER is not handled specially. > So it won't retry if the DMA engine driver isn't available, but just use PIO > (until unbind/bind). For historic reasons, i2c-sh_mobile uses subsys_initcall() and at that time, DMA is never available. Converting to module_init() will just create its own set of potential regressions :( > Now we have a nice sample implementation, perhaps I should port it to > spi-rspi and spi-sh-msiof, too? > > Let's wait and see for other comments... Yeah, let's see first...
Attachment:
signature.asc
Description: Digital signature
