Re: [PATCH] replace various uses of num_physpages by totalram_pages

[Amerigo Wang <xiyou.wangcong@gmail.com>]

On Tue, Jun 30, 2009 at 12:08:32PM +0100, Jan Beulich wrote:
>Sizing of memory allocations shouldn't depend on the number of physical
>pages found in a system, as that generally include (perhaps a huge
>amount of) non-RAM pages. The amount of what actually is usable as
>storage should instead be used as a basis here.


For this part, ACK.

>
>In line with that, the memory hotplug code should update num_physpages
>in a way that it retains its original (post-boot) meaning; in
>particular, decreasing the value should at best be done with great care
>- this patch doesn't try to ever decrease this value at all as it
>doesn't really seem meaningful to do so.


Could you please split this part out and send it with Cc: to linux-mm
mailing list?

Thanks.

>
>Signed-off-by: Jan Beulich <jbeulich@novell.com>
>
>---
> drivers/char/agp/backend.c        |    4 ++--
> drivers/parisc/ccio-dma.c         |    4 ++--
> drivers/parisc/sba_iommu.c        |    4 ++--
> drivers/xen/balloon.c             |    4 ----
> fs/ntfs/malloc.h                  |    2 +-
> include/linux/mm.h                |    1 +
> init/main.c                       |    4 ++--
> mm/memory_hotplug.c               |    6 ++++--
> mm/slab.c                         |    2 +-
> mm/swap.c                         |    2 +-
> mm/vmalloc.c                      |    4 ++--
> net/core/sock.c                   |    4 ++--
> net/dccp/proto.c                  |    6 +++---
> net/decnet/dn_route.c             |    2 +-
> net/ipv4/route.c                  |    2 +-
> net/ipv4/tcp.c                    |    4 ++--
> net/netfilter/nf_conntrack_core.c |    4 ++--
> net/netfilter/x_tables.c          |    2 +-
> net/netfilter/xt_hashlimit.c      |    8 ++++----
> net/netlink/af_netlink.c          |    6 +++---
> net/sctp/protocol.c               |    6 +++---
> 21 files changed, 40 insertions(+), 41 deletions(-)
>
>--- linux-2.6.31-rc1/drivers/char/agp/backend.c	2009-06-26 17:49:43.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/drivers/char/agp/backend.c	2009-06-24 16:09:34.000000000 +0200
>@@ -114,9 +114,9 @@ static int agp_find_max(void)
> 	long memory, index, result;
> 
> #if PAGE_SHIFT < 20
>-	memory = num_physpages >> (20 - PAGE_SHIFT);
>+	memory = totalram_pages >> (20 - PAGE_SHIFT);
> #else
>-	memory = num_physpages << (PAGE_SHIFT - 20);
>+	memory = totalram_pages << (PAGE_SHIFT - 20);
> #endif
> 	index = 1;
> 
>--- linux-2.6.31-rc1/drivers/parisc/ccio-dma.c	2009-06-10 05:05:27.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/drivers/parisc/ccio-dma.c	2009-06-24 16:10:47.000000000 +0200
>@@ -1267,7 +1267,7 @@ ccio_ioc_init(struct ioc *ioc)
> 	** Hot-Plug/Removal of PCI cards. (aka PCI OLARD).
> 	*/
> 
>-	iova_space_size = (u32) (num_physpages / count_parisc_driver(&ccio_driver));
>+	iova_space_size = (u32) (totalram_pages / count_parisc_driver(&ccio_driver));
> 
> 	/* limit IOVA space size to 1MB-1GB */
> 
>@@ -1306,7 +1306,7 @@ ccio_ioc_init(struct ioc *ioc)
> 
> 	DBG_INIT("%s() hpa 0x%p mem %luMB IOV %dMB (%d bits)\n",
> 			__func__, ioc->ioc_regs,
>-			(unsigned long) num_physpages >> (20 - PAGE_SHIFT),
>+			(unsigned long) totalram_pages >> (20 - PAGE_SHIFT),
> 			iova_space_size>>20,
> 			iov_order + PAGE_SHIFT);
> 
>--- linux-2.6.31-rc1/drivers/parisc/sba_iommu.c	2009-06-26 17:49:47.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/drivers/parisc/sba_iommu.c	2009-06-24 16:10:58.000000000 +0200
>@@ -1390,7 +1390,7 @@ sba_ioc_init(struct parisc_device *sba, 
> 	** for DMA hints - ergo only 30 bits max.
> 	*/
> 
>-	iova_space_size = (u32) (num_physpages/global_ioc_cnt);
>+	iova_space_size = (u32) (totalram_pages/global_ioc_cnt);
> 
> 	/* limit IOVA space size to 1MB-1GB */
> 	if (iova_space_size < (1 << (20 - PAGE_SHIFT))) {
>@@ -1415,7 +1415,7 @@ sba_ioc_init(struct parisc_device *sba, 
> 	DBG_INIT("%s() hpa 0x%lx mem %ldMB IOV %dMB (%d bits)\n",
> 			__func__,
> 			ioc->ioc_hpa,
>-			(unsigned long) num_physpages >> (20 - PAGE_SHIFT),
>+			(unsigned long) totalram_pages >> (20 - PAGE_SHIFT),
> 			iova_space_size>>20,
> 			iov_order + PAGE_SHIFT);
> 
>--- linux-2.6.31-rc1/drivers/xen/balloon.c	2009-06-10 05:05:27.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/drivers/xen/balloon.c	2009-06-24 16:02:01.000000000 +0200
>@@ -96,11 +96,7 @@ static struct balloon_stats balloon_stat
> /* We increase/decrease in batches which fit in a page */
> static unsigned long frame_list[PAGE_SIZE / sizeof(unsigned long)];
> 
>-/* VM /proc information for memory */
>-extern unsigned long totalram_pages;
>-
> #ifdef CONFIG_HIGHMEM
>-extern unsigned long totalhigh_pages;
> #define inc_totalhigh_pages() (totalhigh_pages++)
> #define dec_totalhigh_pages() (totalhigh_pages--)
> #else
>--- linux-2.6.31-rc1/fs/ntfs/malloc.h	2008-04-17 04:49:44.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/fs/ntfs/malloc.h	2009-06-24 16:11:49.000000000 +0200
>@@ -47,7 +47,7 @@ static inline void *__ntfs_malloc(unsign
> 		return kmalloc(PAGE_SIZE, gfp_mask & ~__GFP_HIGHMEM);
> 		/* return (void *)__get_free_page(gfp_mask); */
> 	}
>-	if (likely(size >> PAGE_SHIFT < num_physpages))
>+	if (likely((size >> PAGE_SHIFT) < totalram_pages))
> 		return __vmalloc(size, gfp_mask, PAGE_KERNEL);
> 	return NULL;
> }
>--- linux-2.6.31-rc1/include/linux/mm.h	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/include/linux/mm.h	2009-06-24 16:00:50.000000000 +0200
>@@ -25,6 +25,7 @@ extern unsigned long max_mapnr;
> #endif
> 
> extern unsigned long num_physpages;
>+extern unsigned long totalram_pages;
> extern void * high_memory;
> extern int page_cluster;
> 
>--- linux-2.6.31-rc1/init/main.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/init/main.c	2009-06-24 16:12:30.000000000 +0200
>@@ -691,12 +691,12 @@ asmlinkage void __init start_kernel(void
> #endif
> 	thread_info_cache_init();
> 	cred_init();
>-	fork_init(num_physpages);
>+	fork_init(totalram_pages);
> 	proc_caches_init();
> 	buffer_init();
> 	key_init();
> 	security_init();
>-	vfs_caches_init(num_physpages);
>+	vfs_caches_init(totalram_pages);
> 	radix_tree_init();
> 	signals_init();
> 	/* rootfs populating might need page-writeback */
>--- linux-2.6.31-rc1/mm/memory_hotplug.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/mm/memory_hotplug.c	2009-06-24 16:04:44.000000000 +0200
>@@ -339,8 +339,11 @@ EXPORT_SYMBOL_GPL(__remove_pages);
> 
> void online_page(struct page *page)
> {
>+	unsigned long pfn = page_to_pfn(page);
>+
> 	totalram_pages++;
>-	num_physpages++;
>+	if (pfn >= num_physpages)
>+		num_physpages = pfn + 1;
> 
> #ifdef CONFIG_HIGHMEM
> 	if (PageHighMem(page))
>@@ -831,7 +834,6 @@ repeat:
> 	zone->present_pages -= offlined_pages;
> 	zone->zone_pgdat->node_present_pages -= offlined_pages;
> 	totalram_pages -= offlined_pages;
>-	num_physpages -= offlined_pages;
> 
> 	setup_per_zone_wmarks();
> 	calculate_zone_inactive_ratio(zone);
>--- linux-2.6.31-rc1/mm/slab.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/mm/slab.c	2009-06-24 16:06:21.000000000 +0200
>@@ -1384,7 +1384,7 @@ void __init kmem_cache_init(void)
> 	 * Fragmentation resistance on low memory - only use bigger
> 	 * page orders on machines with more than 32MB of memory.
> 	 */
>-	if (num_physpages > (32 << 20) >> PAGE_SHIFT)
>+	if (totalram_pages > (32 << 20) >> PAGE_SHIFT)
> 		slab_break_gfp_order = BREAK_GFP_ORDER_HI;
> 
> 	/* Bootstrap is tricky, because several objects are allocated
>--- linux-2.6.31-rc1/mm/swap.c	2009-06-10 05:05:27.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/mm/swap.c	2009-06-24 16:06:40.000000000 +0200
>@@ -496,7 +496,7 @@ EXPORT_SYMBOL(pagevec_lookup_tag);
>  */
> void __init swap_setup(void)
> {
>-	unsigned long megs = num_physpages >> (20 - PAGE_SHIFT);
>+	unsigned long megs = totalram_pages >> (20 - PAGE_SHIFT);
> 
> #ifdef CONFIG_SWAP
> 	bdi_init(swapper_space.backing_dev_info);
>--- linux-2.6.31-rc1/mm/vmalloc.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/mm/vmalloc.c	2009-06-24 16:05:47.000000000 +0200
>@@ -1368,7 +1368,7 @@ void *vmap(struct page **pages, unsigned
> 
> 	might_sleep();
> 
>-	if (count > num_physpages)
>+	if (count > totalram_pages)
> 		return NULL;
> 
> 	area = get_vm_area_caller((count << PAGE_SHIFT), flags,
>@@ -1475,7 +1475,7 @@ static void *__vmalloc_node(unsigned lon
> 	unsigned long real_size = size;
> 
> 	size = PAGE_ALIGN(size);
>-	if (!size || (size >> PAGE_SHIFT) > num_physpages)
>+	if (!size || (size >> PAGE_SHIFT) > totalram_pages)
> 		return NULL;
> 
> 	area = __get_vm_area_node(size, VM_ALLOC, VMALLOC_START, VMALLOC_END,
>--- linux-2.6.31-rc1/net/core/sock.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/core/sock.c	2009-06-24 16:14:00.000000000 +0200
>@@ -1169,12 +1169,12 @@ EXPORT_SYMBOL_GPL(sk_setup_caps);
> 
> void __init sk_init(void)
> {
>-	if (num_physpages <= 4096) {
>+	if (totalram_pages <= 4096) {
> 		sysctl_wmem_max = 32767;
> 		sysctl_rmem_max = 32767;
> 		sysctl_wmem_default = 32767;
> 		sysctl_rmem_default = 32767;
>-	} else if (num_physpages >= 131072) {
>+	} else if (totalram_pages >= 131072) {
> 		sysctl_wmem_max = 131071;
> 		sysctl_rmem_max = 131071;
> 	}
>--- linux-2.6.31-rc1/net/dccp/proto.c	2009-06-10 05:05:27.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/dccp/proto.c	2009-06-24 16:13:35.000000000 +0200
>@@ -1049,10 +1049,10 @@ static int __init dccp_init(void)
> 	 *
> 	 * The methodology is similar to that of the buffer cache.
> 	 */
>-	if (num_physpages >= (128 * 1024))
>-		goal = num_physpages >> (21 - PAGE_SHIFT);
>+	if (totalram_pages >= (128 * 1024))
>+		goal = totalram_pages >> (21 - PAGE_SHIFT);
> 	else
>-		goal = num_physpages >> (23 - PAGE_SHIFT);
>+		goal = totalram_pages >> (23 - PAGE_SHIFT);
> 
> 	if (thash_entries)
> 		goal = (thash_entries *
>--- linux-2.6.31-rc1/net/decnet/dn_route.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/decnet/dn_route.c	2009-06-24 16:17:11.000000000 +0200
>@@ -1750,7 +1750,7 @@ void __init dn_route_init(void)
> 	dn_route_timer.expires = jiffies + decnet_dst_gc_interval * HZ;
> 	add_timer(&dn_route_timer);
> 
>-	goal = num_physpages >> (26 - PAGE_SHIFT);
>+	goal = totalram_pages >> (26 - PAGE_SHIFT);
> 
> 	for(order = 0; (1UL << order) < goal; order++)
> 		/* NOTHING */;
>--- linux-2.6.31-rc1/net/ipv4/route.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/ipv4/route.c	2009-06-24 16:14:48.000000000 +0200
>@@ -3412,7 +3412,7 @@ int __init ip_rt_init(void)
> 		alloc_large_system_hash("IP route cache",
> 					sizeof(struct rt_hash_bucket),
> 					rhash_entries,
>-					(num_physpages >= 128 * 1024) ?
>+					(totalram_pages >= 128 * 1024) ?
> 					15 : 17,
> 					0,
> 					&rt_hash_log,
>--- linux-2.6.31-rc1/net/ipv4/tcp.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/ipv4/tcp.c	2009-06-24 16:14:24.000000000 +0200
>@@ -2857,7 +2857,7 @@ void __init tcp_init(void)
> 		alloc_large_system_hash("TCP established",
> 					sizeof(struct inet_ehash_bucket),
> 					thash_entries,
>-					(num_physpages >= 128 * 1024) ?
>+					(totalram_pages >= 128 * 1024) ?
> 					13 : 15,
> 					0,
> 					&tcp_hashinfo.ehash_size,
>@@ -2874,7 +2874,7 @@ void __init tcp_init(void)
> 		alloc_large_system_hash("TCP bind",
> 					sizeof(struct inet_bind_hashbucket),
> 					tcp_hashinfo.ehash_size,
>-					(num_physpages >= 128 * 1024) ?
>+					(totalram_pages >= 128 * 1024) ?
> 					13 : 15,
> 					0,
> 					&tcp_hashinfo.bhash_size,
>--- linux-2.6.31-rc1/net/netfilter/nf_conntrack_core.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/netfilter/nf_conntrack_core.c	2009-06-24 16:16:47.000000000 +0200
>@@ -1230,9 +1230,9 @@ static int nf_conntrack_init_init_net(vo
> 	 * machine has 512 buckets. >= 1GB machines have 16384 buckets. */
> 	if (!nf_conntrack_htable_size) {
> 		nf_conntrack_htable_size
>-			= (((num_physpages << PAGE_SHIFT) / 16384)
>+			= (((totalram_pages << PAGE_SHIFT) / 16384)
> 			   / sizeof(struct hlist_head));
>-		if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE))
>+		if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
> 			nf_conntrack_htable_size = 16384;
> 		if (nf_conntrack_htable_size < 32)
> 			nf_conntrack_htable_size = 32;
>--- linux-2.6.31-rc1/net/netfilter/x_tables.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/netfilter/x_tables.c	2009-06-24 16:16:28.000000000 +0200
>@@ -617,7 +617,7 @@ struct xt_table_info *xt_alloc_table_inf
> 	int cpu;
> 
> 	/* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
>-	if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > num_physpages)
>+	if ((SMP_ALIGN(size) >> PAGE_SHIFT) + 2 > totalram_pages)
> 		return NULL;
> 
> 	newinfo = kzalloc(XT_TABLE_INFO_SZ, GFP_KERNEL);
>--- linux-2.6.31-rc1/net/netfilter/xt_hashlimit.c	2009-06-10 05:05:27.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/netfilter/xt_hashlimit.c	2009-06-24 16:16:07.000000000 +0200
>@@ -194,9 +194,9 @@ static int htable_create_v0(struct xt_ha
> 	if (minfo->cfg.size)
> 		size = minfo->cfg.size;
> 	else {
>-		size = ((num_physpages << PAGE_SHIFT) / 16384) /
>+		size = ((totalram_pages << PAGE_SHIFT) / 16384) /
> 		       sizeof(struct list_head);
>-		if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE))
>+		if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
> 			size = 8192;
> 		if (size < 16)
> 			size = 16;
>@@ -266,9 +266,9 @@ static int htable_create(struct xt_hashl
> 	if (minfo->cfg.size) {
> 		size = minfo->cfg.size;
> 	} else {
>-		size = (num_physpages << PAGE_SHIFT) / 16384 /
>+		size = (totalram_pages << PAGE_SHIFT) / 16384 /
> 		       sizeof(struct list_head);
>-		if (num_physpages > 1024 * 1024 * 1024 / PAGE_SIZE)
>+		if (totalram_pages > 1024 * 1024 * 1024 / PAGE_SIZE)
> 			size = 8192;
> 		if (size < 16)
> 			size = 16;
>--- linux-2.6.31-rc1/net/netlink/af_netlink.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/netlink/af_netlink.c	2009-06-24 16:17:28.000000000 +0200
>@@ -2026,10 +2026,10 @@ static int __init netlink_proto_init(voi
> 	if (!nl_table)
> 		goto panic;
> 
>-	if (num_physpages >= (128 * 1024))
>-		limit = num_physpages >> (21 - PAGE_SHIFT);
>+	if (totalram_pages >= (128 * 1024))
>+		limit = totalram_pages >> (21 - PAGE_SHIFT);
> 	else
>-		limit = num_physpages >> (23 - PAGE_SHIFT);
>+		limit = totalram_pages >> (23 - PAGE_SHIFT);
> 
> 	order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
> 	limit = (1UL << order) / sizeof(struct hlist_head);
>--- linux-2.6.31-rc1/net/sctp/protocol.c	2009-06-26 17:50:00.000000000 +0200
>+++ 2.6.31-rc1-use-totalram_pages/net/sctp/protocol.c	2009-06-24 16:15:36.000000000 +0200
>@@ -1185,10 +1185,10 @@ SCTP_STATIC __init int sctp_init(void)
> 	/* Size and allocate the association hash table.
> 	 * The methodology is similar to that of the tcp hash tables.
> 	 */
>-	if (num_physpages >= (128 * 1024))
>-		goal = num_physpages >> (22 - PAGE_SHIFT);
>+	if (totalram_pages >= (128 * 1024))
>+		goal = totalram_pages >> (22 - PAGE_SHIFT);
> 	else
>-		goal = num_physpages >> (24 - PAGE_SHIFT);
>+		goal = totalram_pages >> (24 - PAGE_SHIFT);
> 
> 	for (order = 0; (1UL << order) < goal; order++)
> 		;
>
>
>--
>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/