From mboxrd@z Thu Jan 1 00:00:00 1970 Subject: Re: [v4][PATCH 2/2] fix large pages in pagemap From: Matt Mackall In-Reply-To: <20080611123724.3a79ea61.akpm@linux-foundation.org> References: <20080611180228.12987026@kernel> <20080611180230.7459973B@kernel> <20080611123724.3a79ea61.akpm@linux-foundation.org> Content-Type: text/plain Date: Wed, 11 Jun 2008 14:53:00 -0500 Message-Id: <1213213980.20045.116.camel@calx> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: owner-linux-mm@kvack.org Return-Path: To: Andrew Morton Cc: Dave Hansen , hans.rosenfeld@amd.com, linux-mm@kvack.org, Hugh Dickins List-ID: [adding Hugh to the cc:] On Wed, 2008-06-11 at 12:37 -0700, Andrew Morton wrote: > On Wed, 11 Jun 2008 11:02:31 -0700 > Dave Hansen wrote: > > > > > We were walking right into huge page areas in the pagemap > > walker, and calling the pmds pmd_bad() and clearing them. > > > > That leaked huge pages. Bad. > > > > This patch at least works around that for now. It ignores > > huge pages in the pagemap walker for the time being, and > > won't leak those pages. > > > > I don't get it. Why can't we just stick a > > if (pmd_huge(pmd)) > continue; > > into pagemap_pte_range()? Or something like that. That's certainly what you'd hope to be able to do, yes. If I recall the earlier discussion, some arches with huge pages can only identify them via a VMA. Obviously, any arch with hardware that walks our pagetables directly must be able to identify huge pages directly from those tables, but I think PPC and a couple others that don't have hardware TLB fill fail to store such a bit in the tables at all. -- Mathematics is the supreme nostalgia of our time. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org