On Tue, 22 Nov 2011 09:32:38 +0900 KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote: > On Mon, 21 Nov 2011 14:27:20 -0800 > Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote: > > > On Mon, 21 Nov 2011 19:48:32 +0900 > > KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote: > > > > > >From 7c29389be2890c6b6934a80b4841d07a7014fe26 Mon Sep 17 00:00:00 2001 > > > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> > > > Date: Mon, 21 Nov 2011 19:45:27 +0900 > > > Subject: [PATCH] Fix virtual address handling in hugetlb fault > > > > > > handle_mm_fault() passes 'faulted' address to hugetlb_fault(). > > > Then, the address is not aligned to hugepage boundary. > > > > > > Most of functions for hugetlb pages are aware of that and > > > calculate an alignment by itself. Some functions as copy_user_huge_page(), > > > and clear_huge_page() doesn't handle alignment by themselves. > > > > > > This patch make hugeltb_fault() to calculate the alignment and pass > > > aligned addresss (top address of a faulted hugepage) to functions. > > > > > > > Does this actually fix any known user-visible misbehaviour? > > > > I just found this at reading codes. And I know 'vaddr' is ignored > in most of per-arch implemantation of clear_user_highpage(). > It seems, in some arch, vaddr is used for flushing cache. Now, > CONFIG_HUGETLBFS can be set on x86,powerpc,ia64,mips,sh,sparc,tile. (by grep) > > it seems mips and sh uses vaddr in clear_user_(high)page. OK. Those architectures are probably OK with "any address within the page" anyway. I'm actually trying to work out which kernel(s) we should merge this into ;) -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>