Messages in this thread | | | From | KOSAKI Motohiro <> | Subject | Re: [PATCH] proc: pagemap: Hold mmap_sem during page walk | Date | Thu, 1 Apr 2010 11:10:01 +0900 (JST) |
| |
> So there is a very real reason for that pattern existing. It's just that > that reason has nothing to do with locking the thing into the page tables. > That is absolutely NOT guaranteed by the get_user_pages() physical page > pinning (munmap() is an extreme example of this, but I think swapout will > clear it too in try_to_unmap_one().
Right. To increment page->count only affect vmscan to prevent free page. Not prevent pageout I/O nor page unmapping from a process.
Thanks.
| |