Thread (1 message) flat view 1 message, 1 author, 2010-09-14

Re: [PATCH] use total_highpages when calculating lowmem-only allocation sizes

From: Andrew Morton <akpm@linux-foundation.org>
Date: 2010-09-14 23:12:39
Also in: linux-mm, lkml

On Thu, 02 Sep 2010 13:28:52 +0100
"Jan Beulich" [off-list ref] wrote:
For those (large) table allocations that come only from lowmem, the
total amount of memory shouldn't really matter.

vfs_caches_init() should really also be called with a lowmem-only
value, but since it does arithmetic involving nr_free_pages() and there
is no nr_free_high_pages(), it's not clear how to make this work.
We normally use the hopelessly-badly-named nr_free_buffer_pages() for this.
Renaming that to nr_lowmem_pages() would advance the state of humanity.

Yes, the sites you've identified do look wrong.  And they're almost all in
networking, so let's add the cc for that.  Probably this should be done
in four or fives patches.


For reference purposes:

 init/main.c                       |    3 ++-
 net/dccp/proto.c                  |    8 +++++---
 net/decnet/dn_route.c             |    3 ++-
 net/netfilter/nf_conntrack_core.c |    7 +++++--
 net/netlink/af_netlink.c          |    8 +++++---
 net/sctp/protocol.c               |    6 +++---
 6 files changed, 22 insertions(+), 13 deletions(-)
--- linux-2.6.36-rc3/init/main.c
+++ 2.6.36-rc3-use-totalhigh_pages/init/main.c
@@ -23,6 +23,7 @@
 #include <linux/smp_lock.h>
 #include <linux/initrd.h>
 #include <linux/bootmem.h>
+#include <linux/highmem.h>
 #include <linux/acpi.h>
 #include <linux/tty.h>
 #include <linux/percpu.h>
@@ -676,7 +677,7 @@ asmlinkage void __init start_kernel(void
 #endif
 	thread_info_cache_init();
 	cred_init();
-	fork_init(totalram_pages);
+	fork_init(totalram_pages - totalhigh_pages);
 	proc_caches_init();
 	buffer_init();
 	key_init();
--- linux-2.6.36-rc3/net/dccp/proto.c
+++ 2.6.36-rc3-use-totalhigh_pages/net/dccp/proto.c
@@ -14,6 +14,7 @@
 #include <linux/types.h>
 #include <linux/sched.h>
 #include <linux/kernel.h>
+#include <linux/highmem.h>
 #include <linux/skbuff.h>
 #include <linux/netdevice.h>
 #include <linux/in.h>
@@ -1049,10 +1050,11 @@ static int __init dccp_init(void)
 	 *
 	 * The methodology is similar to that of the buffer cache.
 	 */
-	if (totalram_pages >= (128 * 1024))
-		goal = totalram_pages >> (21 - PAGE_SHIFT);
+	goal = totalram_pages - totalhigh_pages;
+	if (goal >= (128 * 1024))
+		goal >>= 21 - PAGE_SHIFT;
 	else
-		goal = totalram_pages >> (23 - PAGE_SHIFT);
+		goal >>= 23 - PAGE_SHIFT;
 
 	if (thash_entries)
 		goal = (thash_entries *
--- linux-2.6.36-rc3/net/decnet/dn_route.c
+++ 2.6.36-rc3-use-totalhigh_pages/net/decnet/dn_route.c
@@ -69,6 +69,7 @@
 #include <linux/slab.h>
 #include <net/sock.h>
 #include <linux/mm.h>
+#include <linux/highmem.h>
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 #include <linux/init.h>
@@ -1762,7 +1763,7 @@ void __init dn_route_init(void)
 	dn_route_timer.expires = jiffies + decnet_dst_gc_interval * HZ;
 	add_timer(&dn_route_timer);
 
-	goal = totalram_pages >> (26 - PAGE_SHIFT);
+	goal = (totalram_pages - totalhigh_pages) >> (26 - PAGE_SHIFT);
 
 	for(order = 0; (1UL << order) < goal; order++)
 		/* NOTHING */;
--- linux-2.6.36-rc3/net/netfilter/nf_conntrack_core.c
+++ 2.6.36-rc3-use-totalhigh_pages/net/netfilter/nf_conntrack_core.c
@@ -17,6 +17,7 @@
 #include <linux/sched.h>
 #include <linux/skbuff.h>
 #include <linux/proc_fs.h>
+#include <linux/highmem.h>
 #include <linux/vmalloc.h>
 #include <linux/stddef.h>
 #include <linux/slab.h>
@@ -1346,10 +1347,12 @@ static int nf_conntrack_init_init_net(vo
 	/* Idea from tcp.c: use 1/16384 of memory.  On i386: 32MB
 	 * machine has 512 buckets. >= 1GB machines have 16384 buckets. */
 	if (!nf_conntrack_htable_size) {
+		unsigned long nr_pages = totalram_pages - totalhigh_pages;
+
 		nf_conntrack_htable_size
-			= (((totalram_pages << PAGE_SHIFT) / 16384)
+			= (((nr_pages << PAGE_SHIFT) / 16384)
 			   / sizeof(struct hlist_head));
-		if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
+		if (nr_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
 			nf_conntrack_htable_size = 16384;
 		if (nf_conntrack_htable_size < 32)
 			nf_conntrack_htable_size = 32;
--- linux-2.6.36-rc3/net/netlink/af_netlink.c
+++ 2.6.36-rc3-use-totalhigh_pages/net/netlink/af_netlink.c
@@ -30,6 +30,7 @@
 #include <linux/sched.h>
 #include <linux/errno.h>
 #include <linux/string.h>
+#include <linux/highmem.h>
 #include <linux/stat.h>
 #include <linux/socket.h>
 #include <linux/un.h>
@@ -2124,10 +2125,11 @@ static int __init netlink_proto_init(voi
 	if (!nl_table)
 		goto panic;
 
-	if (totalram_pages >= (128 * 1024))
-		limit = totalram_pages >> (21 - PAGE_SHIFT);
+	limit = totalram_pages - totalhigh_pages;
+	if (limit >= (128 * 1024))
+		limit >>= 21 - PAGE_SHIFT;
 	else
-		limit = totalram_pages >> (23 - PAGE_SHIFT);
+		limit >>= 23 - PAGE_SHIFT;
 
 	order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
 	limit = (1UL << order) / sizeof(struct hlist_head);
--- linux-2.6.36-rc3/net/sctp/protocol.c
+++ 2.6.36-rc3-use-totalhigh_pages/net/sctp/protocol.c
@@ -1189,10 +1189,10 @@ 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 (totalram_pages >= (128 * 1024))
-		goal = totalram_pages >> (22 - PAGE_SHIFT);
+	if (nr_pages >= (128 * 1024))
+		goal = nr_pages >> (22 - PAGE_SHIFT);
 	else
-		goal = totalram_pages >> (24 - PAGE_SHIFT);
+		goal = nr_pages >> (24 - PAGE_SHIFT);
 
 	for (order = 0; (1UL << order) < goal; order++)
 		;



--
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: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help