Messages in this thread | | | Date | Mon, 17 Nov 2008 14:03:04 +1100 | From | Peter Chubb <> | Subject | Re: [PATCH 1/2] mm: implement remap_pfn_range with apply_to_page_range |
| |
>>>>> "Nick" == Nick Piggin <nickpiggin@yahoo.com.au> writes:
Nick> It's not meant to be pluggable. Actually this came up last I Nick> think when the UNSW wanted to add page table accessors to Nick> abstract this. They came up with a good set of things, but in Nick> the end you can't justify slowing things down in these paths Nick> unless you actually have a replacement page table structure that Nick> gets you a *net win*. So far, I haven't heard from them again.
We tried hard. The slowdown wasn't all that much on system benchmarks, but you could see it on the microbenchmarks. And it made stuff a LOT cleaner.
We wanted it to put super-page friendly pagetables in for architectures not wedded to the 3/4 level x86 pagetable hardware format.
The people I had working on this left (finished a masters, finished a PhD, contract ran out and no more money), and I haven't had the manpower to maintain the patchset, especially after the negative responses we got from linux-MM.
-- Dr Peter Chubb http://www.gelato.unsw.edu.au peterc AT gelato.unsw.edu.au http://www.ertos.nicta.com.au ERTOS within National ICT Australia
| |