Re: [PATCH] ISA DMA API documentation

[Pierre Ossman <drzeus-list@drzeus.cx>]

randy_dunlap wrote:

>+The DMA:able address space is the lowest 16 MB of _physical_ memory.
> The DMA-able
>+Also the transfer block may not cross page boundaries (which are 64k).
> I would write:                                        (which are 64 KB).
>
>if I knew that was correct, but I don't.
>Does Linux limit all ISA-DMA to not crossing 64 KB boundaries?
>I haven't looked at the code yet, just PC-AT Technical Reference,
>which says that DMA controller 1 is limited to 8-bit transfers and
>64 KB blocks and DMA controller 2 is limited to 16-bit data transfers
>and 128 KB boundaries.
>Does i386-compatible and later chipsets or LPC change/affect this?
>(I see that you cover 8/16-bit transfers later in the doc.)
>  
>

Sorry, my bad. 128k is quite correct for 16-bit data transfers. I've
just been using 8-bit transfers so I got a bit too familiar with just
those. :)

>+To translate the virtual address to a physical use the normal DMA
>+API. Do _not_ use isa_virt_to_phys() even though it does the same
>+thing. The reason for this is that you will get a requirement to ISA
>+(instead of only ISA_DMA_API).
>
>I don't understand what you are trying to say in:
>... is that you will get a requirement to ISA....
>Oh, it's Kconfig-related, right?  So maybe:
>"... is that you will get a config requirement for ISA..." ?
>
>  
>

Yes, that's what I'm trying to say. I'll try to make it clearer.


Thanks for the feedback. I'll get your suggestions in and post a new patch.

Rgds
Pierre