From: grgustaf@redhat.com
Subject: [RHEL5.1 PATCH] sparsemem memmap allocation above 4G (bz 227426)
Date: Wed, 6 Jun 2007 17:34:44 -0400
Bugzilla: 227426
Message-Id: <20070606213444.GA29230@samurai.boston.redhat.com>
Changelog: [x86_64] sparsemem memmap allocation above 4G


This patch went upstream last Friday. Some Intel folks suggested it may solve
this install panic bugzilla on a machine w/ 64GB of RAM.

I should be able to get one of the folks with this large memory setup at Intel
to test for me this week. I'm not sure if we need the kernel on the install CD
to reproduce this exact BZ though.

Larry, do you think this could solve that problem?

Patch notes from Nanhai Zou:

On systems with a huge amount of physical memory, VFS cache and memory memmap
may eat all available system memory under 4G, then the system may fail to
allocate swiotlb bounce buffer.

There was a fix for this issue in arch/x86_64/mm/numa.c, but that fix doesn't
cover sparsemem model. This patch add fix to sparsemem model by first trying
to allocate memmap above 4G.

- Geoff

Index: linux-2.6.18.x86_64/arch/x86_64/mm/init.c
===================================================================
--- linux-2.6.18.x86_64.orig/arch/x86_64/mm/init.c	2007-06-05 18:45:40.000000000 -0400
+++ linux-2.6.18.x86_64/arch/x86_64/mm/init.c	2007-06-06 16:30:32.000000000 -0400
@@ -877,3 +877,9 @@ int in_gate_area_no_task(unsigned long a
 {
 	return (addr >= VSYSCALL_START) && (addr < VSYSCALL_END);
 }
+
+void *alloc_bootmem_high_node(pg_data_t *pgdat, unsigned long size)
+{
+	return __alloc_bootmem_core(pgdat->bdata, size,
+			SMP_CACHE_BYTES, (4UL*1024*1024*1024), 0);
+}
Index: linux-2.6.18.x86_64/include/linux/bootmem.h
===================================================================
--- linux-2.6.18.x86_64.orig/include/linux/bootmem.h	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18.x86_64/include/linux/bootmem.h	2007-06-06 16:32:05.000000000 -0400
@@ -56,6 +56,8 @@ extern void * __init __alloc_bootmem_low
 extern void * __init __alloc_bootmem_core(struct bootmem_data *bdata,
 		unsigned long size, unsigned long align, unsigned long goal,
 		unsigned long limit);
+extern void *alloc_bootmem_high_node(pg_data_t *pgdat, unsigned long size);
+
 #ifndef CONFIG_HAVE_ARCH_BOOTMEM_NODE
 extern void __init reserve_bootmem (unsigned long addr, unsigned long size);
 #define alloc_bootmem(x) \
Index: linux-2.6.18.x86_64/mm/sparse.c
===================================================================
--- linux-2.6.18.x86_64.orig/mm/sparse.c	2007-06-05 18:45:37.000000000 -0400
+++ linux-2.6.18.x86_64/mm/sparse.c	2007-06-06 16:33:47.000000000 -0400
@@ -186,6 +186,12 @@ static int sparse_init_one_section(struc
 	return 1;
 }
 
+__attribute__((weak))
+void *alloc_bootmem_high_node(pg_data_t *pgdat, unsigned long size)
+{
+	return NULL;
+}
+
 static struct page *sparse_early_mem_map_alloc(unsigned long pnum)
 {
 	struct page *map;
@@ -196,6 +202,11 @@ static struct page *sparse_early_mem_map
 	if (map)
 		return map;
 
+	map = alloc_bootmem_high_node(NODE_DATA(nid),
+			sizeof(struct page) * PAGES_PER_SECTION);
+	if (map)
+		return map;
+
 	map = alloc_bootmem_node(NODE_DATA(nid),
 			sizeof(struct page) * PAGES_PER_SECTION);
 	if (map)