From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Tue, 22 Jan 2008 22:35:17 +0000 (GMT) From: Hugh Dickins Subject: Re: [patch] #ifdef very expensive debug check in page fault path In-Reply-To: <479469A4.6090607@de.ibm.com> Message-ID: References: <1200506488.32116.11.camel@cotte.boeblingen.de.ibm.com> <20080116234540.GB29823@wotan.suse.de> <20080116161021.c9a52c0f.akpm@linux-foundation.org> <479469A4.6090607@de.ibm.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org Return-Path: To: carsteno@de.ibm.com Cc: Andrew Morton , Nick Piggin , Linux Memory Management List , mschwid2@linux.vnet.ibm.com, Holger Wolf , Linus Torvalds List-ID: On Mon, 21 Jan 2008, Carsten Otte wrote: > Hugh Dickins wrote: > > > > Well: that patch still gets my Nack, but I guess I'm too late. If > > s390 pagetables are better protected than x86 ones, add an s390 ifdef? > > The alternative would be to just make > #define pfn_valid(pfn) (1) > on s390. That would also get _us_ rid of the check while others do benefit. We > would trap access to mem_map beyond its limits because we don't have a kernel > mapping there. For us, it would not silently corrupt things but crash proper. Whilst I quite like the sound of that, I wonder whether it's safe to change s390's pfn_valid (rather surprisingly) for all its users. And note that nobody but me has voiced any regret at the loss of the check. My guess is we let it rest for now, and reconsider if a case comes up later which would have got caught by the check (but the problem is that such a case is much harder to identify than it was). Hugh -- 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