Messages in this thread |  | | Date | Mon, 28 Aug 2000 16:12:52 +0100 | From | "Stephen C. Tweedie" <> | Subject | Re: How does the kernel map physical to virtual addresses? |
| |
Hi,
On Mon, Aug 28, 2000 at 01:56:34PM +0100, Tigran Aivazian wrote: > > it is interesting to observe that many questions that deal with _details_ > are answered quickly but questions related to fundamental concepts related > to how Linux is designed, baffle all of us (since 0 people answered). So, > is there really nobody in the whole world who can answer this? I would > like to know the answer (about global kernel memory layout - i.e. what > goes into PSE pages and what goes into normal ones, and how does PAE mode > change the picture?) myself...
If PSE is available, it is used to map the bits of the kernel's VA which permanently maps all of physical memory. As a result, those pages cannot necessarily be looked up via a normal page table walk. Anything dynamically mapped --- ie. high pages (if using PAE), or vmalloc/ioremap pages --- is mapped using normal 4k ptes.
mem_map[] is completely unaffected by the use of PSE, and continues to keep one entry per 4k physical page regardless of how the page tables have been constructed.
--Stephen - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/
|  |