From mboxrd@z Thu Jan 1 00:00:00 1970 Subject: Re: memory hotplug and mem= From: Dave Hansen In-Reply-To: <20041007170138.GC15186@logos.cnet> References: <20041001182221.GA3191@logos.cnet> <4160F483.3000309@jp.fujitsu.com> <20041007155854.GC14614@logos.cnet> <1097172146.22025.29.camel@localhost> <20041007170138.GC15186@logos.cnet> Content-Type: text/plain Message-Id: <1097180757.25526.7.camel@localhost> Mime-Version: 1.0 Date: Thu, 07 Oct 2004 13:25:57 -0700 Content-Transfer-Encoding: 7bit Sender: owner-linux-mm@kvack.org Return-Path: To: Marcelo Tosatti Cc: Hiroyuki KAMEZAWA , IWAMOTO Toshihiro , linux-mm List-ID: Turns out that there are a few invalid uses of pfn_to_page() in the normal code. If you hand-apply the following chunks, it gets a bit farther into boot. These aren't permanent fixes. We really need a last_lowmem_pfn instead of highstart_pfn. static void __init set_max_mapnr_init(void) { #ifdef CONFIG_HIGHMEM - highmem_start_page = pfn_to_page(highstart_pfn); + highmem_start_page = pfn_to_page(highstart_pfn-1); max_mapnr = num_physpages = highend_pfn; #else +++ memhotplug-dave/arch/i386/mm/pageattr.c 2004-10-07 13:22:41.000000000 -0700 @@ -109,7 +109,7 @@ __change_page_attr(struct page *page, pg struct page *kpte_page; #ifdef CONFIG_HIGHMEM - if (page >= highmem_start_page) + if (page > highmem_start_page) BUG(); #endif -- Dave -- 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: aart@kvack.org