[PATCH 07/11] fsmc/nand: Provide contiguous buffers to dma

[Brian Norris]

On Tue, Oct 16, 2012 at 3:05 AM, Linus Walleij <linus.walleij at linaro.org> wrote:
> On Tue, Oct 16, 2012 at 9:14 AM, Artem Bityutskiy <dedekind1 at gmail.com> wrote:
>> On Mon, 2012-10-15 at 21:51 +0200, Linus Walleij wrote:
>>> On Mon, Oct 15, 2012 at 6:27 PM, Brian Norris
>>> <computersforpeace at gmail.com> wrote:
>>>
>>> > I would prefer not building a solution that hopes kmalloc() can get a
>>> > large contiguous buffer (remember, eraseblock sizes come as large as
>>> > 2MB these days). A real solution like CMA or scatter-gather seems like
>>> > a better idea.
>>>
>>> +1 on that, and as I think I illustrated the MMC subsystem is using
>>> the block layer helpers to form scatter-gather lists for it's requests.
>>> I don't see why the MTD subsystem need to be very different?
>>
>> This is good idea, the problem is that it is a bit more work than
>> hacking the driver, so most people prefer the latter. By removing all
>> kmalloc()'s I basically wish to force people to go this way.

Did you mean vmalloc()'s?

>> What are the SC functions names you refer?
>
> I'm mainly thinking of blk_rq_map_sg()
> from block/blk-merge.c

I am also interested in the potential for the S-G approach. I'll see
what kind of work it takes to implement soon, hopefully.

As a side note, my hardware can, in fact, perform scatter-gather DMA.
If we were to support both S-G hardware like mine in addition to
standard, page-at-time access (nand_chip->write_page, for example),
could we introduce a DMA interface to nand_chip? Then this interface
would just be set to NULL for those that can't handle it.

Brian