On Mon, Jul 22, 2019 at 12:37:44PM -0700, Bart Van Assche wrote: > Is it possible that a device defines a maximum mapping size but no DMA > mask? Is the NULL pointer dereference that can happen an attempt to > dereference dev->dma_ops? Have you considered to test the get_dma_ops() > return value instead of dev->dma_mask? I think that would make this code > easier to read. The dma max mapping size needs dma support an thus a dma mask, as we've phased out support for a NULL dma_mask for dma supporting devices.
