On Tue, May 29, 2018 at 03:38:23PM +1000, Finn Thain wrote: > On Tue, 29 May 2018, Michael Schmitz wrote: > > > > > > > Since an arch gets to apply limits in the dma ops it implements, why > > > would arch code also have to set a limit in the form of default > > > platform device masks? Powerpc seems to be the only arch that does > > > this. > > > > One of Christoph's recent patches removed most of arches' dma ops, > > replacing them by one generic implementation instead. m68k was one of > > the affected arches. I concede his patch series is experimental still > > and not in mainline, but may be included at some time. > > I found some patches here, > http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/generic-dma-noncoherent.2 > > Looks like m68k_dma_alloc() gets renamed arch_dma_alloc() and the generic > ops don't use the dma masks.
This is for the 'coherent' allocations, and literatelly nothing changes for those yet. But for the mapping path this switches m68k to use the dma-direct path, wrapped by dma-noncoherent. And these do check dev->dma_mask, although only really for sanity checking. Btw, can I get a review and testing for the above? They aren't really experimental any more, and I'd like to move architectures over as soon as possible.