Messages in this thread | | | Date | Fri, 16 Apr 1999 11:39:34 +0200 (METDST) | From | Stefan Ring <> | Subject | Re: DMA memory allocation |
| |
On Thu, 15 Apr 1999, Stephen C. Tweedie wrote:
> > Theoretically, it must be possible to just move some adjacent pages out of > > <16MB memory area, update the page tables and use that freed memory as DMA > > memory henceforth, when DMA memory is requested. > > That's fine if you can find the appropriate pages. There are always > risks concerning finding kernel data pages, slab pages, page table > pages or mlock()ed pages, none of which can be relocated. As a > result, the larger a region you are searching for, the harder it is to > satisfy. >
Well, I bet there are at any time much more swappable (user) pages than non-swappable pages in the <16MB area. And after all, a few pages are enough for doing DMA transfers as the DMA controller can do no more than 128KB at a time AFAIK. (at least on Intel architectures, and that's what this is all about)
I can't imagine kernel pages using more than 4MB of memory altogether (including page tables which don't neccessarily have to be in the lower part of the memory)
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |