cvs commit: src/sys/alpha/alpha busdma_machdep.csrc/sys/alpha/include bus.h src/sys/amd64/amd64 busdma_machdep.c src/sys/amd64/include bus_dma.h src/sys/i386/i386 busdma_machdep

[Scott Long]

Nate Lawson wrote:
> On Sun, 25 May 2003, Scott Long wrote:
> 
>>  Modified files:
>>    sys/alpha/alpha      busdma_machdep.c
>>    sys/alpha/include    bus.h
>>    sys/amd64/amd64      busdma_machdep.c
>>    sys/amd64/include    bus_dma.h
>>    sys/i386/i386        busdma_machdep.c
>>    sys/i386/include     bus_dma.h
>>    sys/sparc64/include  bus.h iommuvar.h
>>    sys/sparc64/pci      psycho.c
>>    sys/sparc64/sbus     sbus.c
>>    sys/sparc64/sparc64  iommu.c
>>  Log:
>>  De-orbit bus_dmamem_alloc_size().  It's a hack and was never used anyways.
>>  No need for it to pollute the 5.x API any further.
>>
>>  Approved by:    re (bmah)
> 
> 
> Do you plan to address this in the future?  I believe it was useful for a
> few drivers.  I seem to recall you mentioning a more general approach for
> the future.
> 
> -Nate

The better approach is to have bus_dmamap_load() respect the maxsegs
field of the tag and coelesce the segments by whatever means it has
available to it (bcopy, GART, etc).
Note the the bus_dmamem_alloc_size stuff went into the tree in Jan but
was never once used.

Scott