Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756501Ab3EHPTq (ORCPT ); Wed, 8 May 2013 11:19:46 -0400 Received: from mail-pb0-f44.google.com ([209.85.160.44]:48894 "EHLO mail-pb0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756325Ab3EHPTn (ORCPT ); Wed, 8 May 2013 11:19:43 -0400 From: Jiang Liu To: Andrew Morton Cc: Jiang Liu , David Rientjes , Wen Congyang , Mel Gorman , Minchan Kim , KAMEZAWA Hiroyuki , Michal Hocko , James Bottomley , Sergei Shtylyov , David Howells , Mark Salter , Jianguo Wu , linux-mm@kvack.org, linux-arch@vger.kernel.org, linux-kernel@vger.kernel.org, "H. Peter Anvin" , x86@kernel.org, Tejun Heo , Joonsoo Kim , Yinghai Lu , Marek Szyprowski Subject: [PATCH v5, part3 10/15] mm: accurately calculate zone->managed_pages for highmem zones Date: Wed, 8 May 2013 23:17:09 +0800 Message-Id: <1368026235-5976-11-git-send-email-jiang.liu@huawei.com> X-Mailer: git-send-email 1.7.9.5 In-Reply-To: <1368026235-5976-1-git-send-email-jiang.liu@huawei.com> References: <1368026235-5976-1-git-send-email-jiang.liu@huawei.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 7046 Lines: 220 Commit "mm: introduce new field 'managed_pages' to struct zone" assumes that all highmem pages will be freed into the buddy system by function mem_init(). But that's not always true, some architectures may reserve some highmem pages during boot. For example PPC may allocate highmem pages for giagant HugeTLB pages, and several architectures have code to check PageReserved flag to exclude highmem pages allocated during boot when freeing highmem pages into the buddy system. So treat highmem pages in the same way as normal pages, that is to: 1) reset zone->managed_pages to zero in mem_init(). 2) recalculate managed_pages when freeing pages into the buddy system. Signed-off-by: Jiang Liu Cc: "H. Peter Anvin" Cc: x86@kernel.org Cc: Tejun Heo Cc: Joonsoo Kim Cc: Yinghai Lu Cc: Mel Gorman Cc: Minchan Kim Cc: Kamezawa Hiroyuki Cc: Marek Szyprowski Cc: linux-kernel@vger.kernel.org Cc: linux-mm@kvack.org --- arch/metag/mm/init.c | 6 ++++++ arch/x86/mm/highmem_32.c | 6 ++++++ include/linux/bootmem.h | 1 + mm/bootmem.c | 32 ++++++++++++++++++-------------- mm/nobootmem.c | 30 ++++++++++++++++-------------- mm/page_alloc.c | 1 + 6 files changed, 48 insertions(+), 28 deletions(-) diff --git a/arch/metag/mm/init.c b/arch/metag/mm/init.c index d05b845..58a36f3 100644 --- a/arch/metag/mm/init.c +++ b/arch/metag/mm/init.c @@ -380,6 +380,12 @@ void __init mem_init(void) #ifdef CONFIG_HIGHMEM unsigned long tmp; + + /* + * Explicitly reset zone->managed_pages because highmem pages are + * freed before calling free_all_bootmem_node(); + */ + reset_all_zones_managed_pages(); for (tmp = highstart_pfn; tmp < highend_pfn; tmp++) free_highmem_page(pfn_to_page(tmp)); num_physpages += totalhigh_pages; diff --git a/arch/x86/mm/highmem_32.c b/arch/x86/mm/highmem_32.c index 252b8f5..4500142 100644 --- a/arch/x86/mm/highmem_32.c +++ b/arch/x86/mm/highmem_32.c @@ -1,6 +1,7 @@ #include #include #include /* for totalram_pages */ +#include void *kmap(struct page *page) { @@ -121,6 +122,11 @@ void __init set_highmem_pages_init(void) struct zone *zone; int nid; + /* + * Explicitly reset zone->managed_pages because set_highmem_pages_init() + * is invoked before free_all_bootmem() + */ + reset_all_zones_managed_pages(); for_each_zone(zone) { unsigned long zone_start_pfn, zone_end_pfn; diff --git a/include/linux/bootmem.h b/include/linux/bootmem.h index 5f0b0e1..0e48c32 100644 --- a/include/linux/bootmem.h +++ b/include/linux/bootmem.h @@ -46,6 +46,7 @@ extern unsigned long init_bootmem(unsigned long addr, unsigned long memend); extern unsigned long free_all_bootmem_node(pg_data_t *pgdat); extern unsigned long free_all_bootmem(void); +extern void reset_all_zones_managed_pages(void); extern void free_bootmem_node(pg_data_t *pgdat, unsigned long addr, diff --git a/mm/bootmem.c b/mm/bootmem.c index d937107..39e6156 100644 --- a/mm/bootmem.c +++ b/mm/bootmem.c @@ -242,20 +242,26 @@ static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata) return count; } -static void reset_node_lowmem_managed_pages(pg_data_t *pgdat) +static int reset_managed_pages_done __initdata; + +static inline void __init reset_node_managed_pages(pg_data_t *pgdat) { struct zone *z; - /* - * In free_area_init_core(), highmem zone's managed_pages is set to - * present_pages, and bootmem allocator doesn't allocate from highmem - * zones. So there's no need to recalculate managed_pages because all - * highmem pages will be managed by the buddy system. Here highmem - * zone also includes highmem movable zone. - */ + if (reset_managed_pages_done) + return; + for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) - if (!is_highmem(z)) - z->managed_pages = 0; + z->managed_pages = 0; +} + +void __init reset_all_zones_managed_pages(void) +{ + struct pglist_data *pgdat; + + for_each_online_pgdat(pgdat) + reset_node_managed_pages(pgdat); + reset_managed_pages_done = 1; } /** @@ -267,7 +273,7 @@ static void reset_node_lowmem_managed_pages(pg_data_t *pgdat) unsigned long __init free_all_bootmem_node(pg_data_t *pgdat) { register_page_bootmem_info_node(pgdat); - reset_node_lowmem_managed_pages(pgdat); + reset_node_managed_pages(pgdat); return free_all_bootmem_core(pgdat->bdata); } @@ -280,10 +286,8 @@ unsigned long __init free_all_bootmem(void) { unsigned long total_pages = 0; bootmem_data_t *bdata; - struct pglist_data *pgdat; - for_each_online_pgdat(pgdat) - reset_node_lowmem_managed_pages(pgdat); + reset_all_zones_managed_pages(); list_for_each_entry(bdata, &bdata_list, list) total_pages += free_all_bootmem_core(bdata); diff --git a/mm/nobootmem.c b/mm/nobootmem.c index bdd3fa2..0ae8d91 100644 --- a/mm/nobootmem.c +++ b/mm/nobootmem.c @@ -137,20 +137,25 @@ static unsigned long __init free_low_memory_core_early(void) return count; } -static void reset_node_lowmem_managed_pages(pg_data_t *pgdat) +static int reset_managed_pages_done __initdata; + +static inline void __init reset_node_managed_pages(pg_data_t *pgdat) { struct zone *z; - /* - * In free_area_init_core(), highmem zone's managed_pages is set to - * present_pages, and bootmem allocator doesn't allocate from highmem - * zones. So there's no need to recalculate managed_pages because all - * highmem pages will be managed by the buddy system. Here highmem - * zone also includes highmem movable zone. - */ + if (reset_managed_pages_done) + return; for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) - if (!is_highmem(z)) - z->managed_pages = 0; + z->managed_pages = 0; +} + +void __init reset_all_zones_managed_pages(void) +{ + struct pglist_data *pgdat; + + for_each_online_pgdat(pgdat) + reset_node_managed_pages(pgdat); + reset_managed_pages_done = 1; } /** @@ -160,10 +165,7 @@ static void reset_node_lowmem_managed_pages(pg_data_t *pgdat) */ unsigned long __init free_all_bootmem(void) { - struct pglist_data *pgdat; - - for_each_online_pgdat(pgdat) - reset_node_lowmem_managed_pages(pgdat); + reset_all_zones_managed_pages(); /* * We need to use MAX_NUMNODES instead of NODE_DATA(0)->node_id diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 1b177bb..c5c0988 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -5211,6 +5211,7 @@ void free_highmem_page(struct page *page) { __free_reserved_page(page); totalram_pages++; + page_zone(page)->managed_pages++; totalhigh_pages++; } #endif -- 1.7.9.5 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/