Messages in this thread | | | Subject | Re: [PATCH] staging: comedi: comedi_isadma: Use a non-NULL device for DMA API | From | Ian Abbott <> | Date | Fri, 26 Apr 2019 10:41:20 +0100 |
| |
On 25/04/2019 18:13, Greg Kroah-Hartman wrote: > On Thu, Apr 25, 2019 at 05:26:44PM +0100, Ian Abbott wrote: >> The "comedi_isadma" module calls `dma_alloc_coherent()` and >> `dma_free_coherent()` with a NULL device pointer which is no longer >> allowed. If the `hw_dev` member of the `struct comedi_device` has been >> set to a valid device, that can be used instead. Unfortunately, all the >> current users of the "comedi_isadma" module leave the `hw_dev` member >> set to NULL. In that case, use a static dummy fallback device structure >> with the coherent DMA mask set to the ISA bus limit of 16MB. >> >> Signed-off-by: Ian Abbott <abbotti@mev.co.uk> >> --- >> drivers/staging/comedi/drivers/comedi_isadma.c | 15 +++++++++++++-- >> drivers/staging/comedi/drivers/comedi_isadma.h | 3 +++ >> 2 files changed, 16 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/staging/comedi/drivers/comedi_isadma.c b/drivers/staging/comedi/drivers/comedi_isadma.c >> index b77dc8d5d3ff..8929952516a1 100644 >> --- a/drivers/staging/comedi/drivers/comedi_isadma.c >> +++ b/drivers/staging/comedi/drivers/comedi_isadma.c >> @@ -14,6 +14,16 @@ >> >> #include "comedi_isadma.h" >> >> +/* >> + * Fallback device used when hardware device is NULL. >> + * This can be removed after drivers have been converted to use isa_driver. >> + */ >> +static struct device fallback_dev = { >> + .init_name = "comedi_isadma fallback device", >> + .coherent_dma_mask = DMA_BIT_MASK(24), >> + .dma_mask = &fallback_dev.coherent_dma_mask, >> +}; > > Ick, no, static struct device are a very bad idea as this is a reference > counted structure and making it static can cause odd problems.
This was based on the use of `struct device x86_dma_fallback_dev` in "arch/x86/kernel/pci-dma.c", and `static struct device isa_dma_dev` in "arch/arm/kernel/dma-isa.c", but perhaps it is not appropriate in non-arch code.
> Why not just create a "real" one? Or better yet, use the real device > for the comedi device as all of these drivers should have one now.
I suppose I could use the comedi class device pointed to by the `class_dev` member of `struct comedi_device` (although that could also be NULL because the comedi core does not currently treat `device_create()` failures as fatal).
> > thanks, > > greg k-h
Thanks for the review,
Ian Abbott.
-- -=( Ian Abbott <abbotti@mev.co.uk> || Web: www.mev.co.uk )=- -=( MEV Ltd. is a company registered in England & Wales. )=- -=( Registered number: 02862268. Registered address: )=- -=( 15 West Park Road, Bramhall, STOCKPORT, SK7 3JZ, UK. )=-
| |