On Friday, May 10, 2013 6:07 AM, Ian Abbott wrote: > "ni_mio_common.c" holds common code included by "ni_pcimio.c", > "ni_atmio.c" and "ni_mio_cs.c", including a common initialization > function `ni_E_init()`. Amongst other things, this initializes some > counter subdevices to support comedi instructions and asynchronous > commands. However, even though it sets up the handlers to support > asynchronous commands on these subdevices, the handlers will return an > error unless the `PCIDMA` macro is defined (which is defined only in > "ni_pcimio.c"). If the `PCIDMA` macro is not defined, the comedi core > will needlessly allocate buffers to support the asynchronous commands. > Also, `s->async_dma_dir` is set to `DMA_BIDIRECTIONAL`, causing the > physical pages for the buffers to be allocated using > `dma_alloc_coherent()`. > > If the comedi core cannot call `dma_alloc_coherent()` because > `CONFIG_HAS_DMA` is not defined, it will fail to allocate the buffers, > which ultimately causes `ni_E_init()` to fail. > > Avoid the wastage and prevent the failure by only setting up > asynchronous command support for the counter subdevices if the `PCIDMA` > macro is defined. > > Signed-off-by: Ian Abbott <abbotti@xxxxxxxxx> > --- > drivers/staging/comedi/drivers/ni_mio_common.c | 20 ++++++++++---------- > 1 file changed, 10 insertions(+), 10 deletions(-) Ian, Other than the minor comments on patches 1/3 and 2/3, I don't see any issues with this series. It should fix Geert Uytterhoeven's build problem and it keeps comedi from globaly depending on HAS_DMA. I plan on looking over the National drivers and cleaning up the complexity caused by the mite, ni_tio, ni_tiocmd, and ni_mio_common modules. For now this should work. Thanks, Reviewed-by: H Hartley Sweeten <hsweeten@xxxxxxxxxxxxxxxxxxx> _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
