[ipfire-2.x.git] / src / patches / suse-2.6.27.39 / patches.xen / xen-x86-bigmem

From: jbeulich@novell.com
Subject: fix issues with the assignment of huge amounts of memory
Patch-mainline: obsolete
References: bnc#482614, bnc#537435

--- sle11-2009-09-18.orig/arch/x86/kernel/e820-xen.c	2009-02-17 14:01:20.000000000 +0100
+++ sle11-2009-09-18/arch/x86/kernel/e820-xen.c	2009-09-18 10:08:03.000000000 +0200
@@ -1317,6 +1317,26 @@ static int __init parse_memopt(char *p)
 
 	i = e820.nr_map - 1;
 	current_end = e820.map[i].addr + e820.map[i].size;
+
+	/*
+	 * A little less than 2% of available memory are needed for page
+	 * tables, p2m map, and mem_map. Hence the maximum amount of memory
+	 * we can potentially balloon up to can in no case exceed about 50
+	 * times of what we've been given initially. Since even with that we
+	 * won't be able to boot (due to various calculations done based on
+	 * the total number of pages) we further restrict this to factor 32.
+	 */
+	if ((mem_size >> (PAGE_SHIFT + 5)) > xen_start_info->nr_pages) {
+		u64 size = (u64)xen_start_info->nr_pages << 5;
+
+		printk(KERN_WARNING "mem=%Luk is invalid for an initial"
+				    " allocation of %luk, using %Luk\n",
+		       (unsigned long long)mem_size >> 10,
+		       xen_start_info->nr_pages << (PAGE_SHIFT - 10),
+		       (unsigned long long)size << (PAGE_SHIFT - 10));
+		mem_size = size << PAGE_SHIFT;
+	}
+
 	if (current_end < mem_size) {
 		/*
 		 * The e820 map ends before our requested size so
@@ -1416,7 +1436,7 @@ void __init e820_reserve_resources(void)
 	struct resource *res;
 	u64 end;
 
-	res = alloc_bootmem_low(sizeof(struct resource) * e820.nr_map);
+	res = alloc_bootmem(sizeof(struct resource) * e820.nr_map);
 	for (i = 0; i < e820.nr_map; i++) {
 		end = e820.map[i].addr + e820.map[i].size - 1;
 #ifndef CONFIG_RESOURCES_64BIT
@@ -1498,6 +1518,7 @@ char *__init __attribute__((weak)) machi
 char * __init memory_setup(void)
 {
 	int rc, nr_map;
+	unsigned long long maxmem;
 	struct xen_memory_map memmap;
 	/*
 	 * This is rather large for a stack variable but this early in
@@ -1527,6 +1548,22 @@ char * __init memory_setup(void)
 		BUG();
 
 #ifdef CONFIG_XEN
+	/* See the comment in parse_memopt(). */
+	for (maxmem = rc = 0; rc < e820.nr_map; ++rc)
+		if (e820.map[rc].type == E820_RAM)
+			maxmem += e820.map[rc].size;
+	if ((maxmem >> (PAGE_SHIFT + 5)) > xen_start_info->nr_pages) {
+		unsigned long long size = (u64)xen_start_info->nr_pages << 5;
+
+		printk(KERN_WARNING "maxmem of %LuM is invalid for an initial"
+				    " allocation of %luM, using %LuM\n",
+		       maxmem >> 20,
+		       xen_start_info->nr_pages >> (20 - PAGE_SHIFT),
+		       size >> (20 - PAGE_SHIFT));
+		size <<= PAGE_SHIFT;
+		e820_remove_range(size, ULLONG_MAX - size, E820_RAM, 1);
+	}
+
 	if (is_initial_xendomain()) {
 		memmap.nr_entries = E820MAX;
 		set_xen_guest_handle(memmap.buffer, machine_e820.map);
--- sle11-2009-09-18.orig/arch/x86/kernel/setup-xen.c	2009-07-31 15:14:20.000000000 +0200
+++ sle11-2009-09-18/arch/x86/kernel/setup-xen.c	2009-07-31 15:14:31.000000000 +0200
@@ -128,12 +128,7 @@ static struct notifier_block xen_panic_b
 unsigned long *phys_to_machine_mapping;
 EXPORT_SYMBOL(phys_to_machine_mapping);
 
-unsigned long *pfn_to_mfn_frame_list_list,
-#ifdef CONFIG_X86_64
-	*pfn_to_mfn_frame_list[512];
-#else
-	*pfn_to_mfn_frame_list[128];
-#endif
+unsigned long *pfn_to_mfn_frame_list_list, **pfn_to_mfn_frame_list;
 
 /* Raw start-of-day parameters from the hypervisor. */
 start_info_t *xen_start_info;
@@ -1036,17 +1031,17 @@ void __init setup_arch(char **cmdline_p)
 		p2m_pages = xen_start_info->nr_pages;
 
 	if (!xen_feature(XENFEAT_auto_translated_physmap)) {
-		unsigned long i, j;
+		unsigned long i, j, size;
 		unsigned int k, fpp;
 
 		/* Make sure we have a large enough P->M table. */
 		phys_to_machine_mapping = alloc_bootmem_pages(
 			max_pfn * sizeof(unsigned long));
-		memset(phys_to_machine_mapping, ~0,
-		       max_pfn * sizeof(unsigned long));
 		memcpy(phys_to_machine_mapping,
 		       (unsigned long *)xen_start_info->mfn_list,
 		       p2m_pages * sizeof(unsigned long));
+		memset(phys_to_machine_mapping + p2m_pages, ~0,
+		       (max_pfn - p2m_pages) * sizeof(unsigned long));
 		free_bootmem(
 			__pa(xen_start_info->mfn_list),
 			PFN_PHYS(PFN_UP(xen_start_info->nr_pages *
@@ -1056,15 +1051,26 @@ void __init setup_arch(char **cmdline_p)
 		 * Initialise the list of the frames that specify the list of
 		 * frames that make up the p2m table. Used by save/restore.
 		 */
-		pfn_to_mfn_frame_list_list = alloc_bootmem_pages(PAGE_SIZE);
-
 		fpp = PAGE_SIZE/sizeof(unsigned long);
+		size = (max_pfn + fpp - 1) / fpp;
+		size = (size + fpp - 1) / fpp;
+		++size; /* include a zero terminator for crash tools */
+		size *= sizeof(unsigned long);
+		pfn_to_mfn_frame_list_list = alloc_bootmem_pages(size);
+		if (size > PAGE_SIZE
+		    && xen_create_contiguous_region((unsigned long)
+						    pfn_to_mfn_frame_list_list,
+						    get_order(size), 0))
+			BUG();
+		size -= sizeof(unsigned long);
+		pfn_to_mfn_frame_list = alloc_bootmem(size);
+
 		for (i = j = 0, k = -1; i < max_pfn; i += fpp, j++) {
 			if (j == fpp)
 				j = 0;
 			if (j == 0) {
 				k++;
-				BUG_ON(k>=ARRAY_SIZE(pfn_to_mfn_frame_list));
+				BUG_ON(k * sizeof(unsigned long) >= size);
 				pfn_to_mfn_frame_list[k] =
 					alloc_bootmem_pages(PAGE_SIZE);
 				pfn_to_mfn_frame_list_list[k] =
--- sle11-2009-09-18.orig/arch/x86/kernel/setup_percpu-xen.c	2009-08-26 12:17:43.000000000 +0200
+++ sle11-2009-09-18/arch/x86/kernel/setup_percpu-xen.c	2009-08-26 12:19:19.000000000 +0200
@@ -206,7 +206,7 @@ static void __init setup_node_to_cpumask
 	}
 
 	/* allocate the map */
-	map = alloc_bootmem_low(nr_node_ids * sizeof(cpumask_t));
+	map = alloc_bootmem(nr_node_ids * sizeof(cpumask_t));
 
 	pr_debug(KERN_DEBUG "Node to cpumask map at %p for %d nodes\n",
 		 map, nr_node_ids);
--- sle11-2009-09-18.orig/drivers/xen/core/machine_kexec.c	2009-06-04 10:21:39.000000000 +0200
+++ sle11-2009-09-18/drivers/xen/core/machine_kexec.c	2009-03-13 16:13:15.000000000 +0100
@@ -57,7 +57,7 @@ void __init xen_machine_kexec_setup_reso
 
 	/* allocate xen_phys_cpus */
 
-	xen_phys_cpus = alloc_bootmem_low(k * sizeof(struct resource));
+	xen_phys_cpus = alloc_bootmem(k * sizeof(struct resource));
 	BUG_ON(xen_phys_cpus == NULL);
 
 	/* fill in xen_phys_cpus with per-cpu crash note information */
--- sle11-2009-09-18.orig/drivers/xen/core/machine_reboot.c	2009-02-17 12:23:48.000000000 +0100
+++ sle11-2009-09-18/drivers/xen/core/machine_reboot.c	2009-02-17 12:25:29.000000000 +0100
@@ -76,7 +76,7 @@ static void post_suspend(int suspend_can
 	unsigned long shinfo_mfn;
 	extern unsigned long max_pfn;
 	extern unsigned long *pfn_to_mfn_frame_list_list;
-	extern unsigned long *pfn_to_mfn_frame_list[];
+	extern unsigned long **pfn_to_mfn_frame_list;
 
 	if (suspend_cancelled) {
 		xen_start_info->store_mfn =
Commit	Line	Data
2cb7cef9 BS	1	From: jbeulich@novell.com
	2	Subject: fix issues with the assignment of huge amounts of memory
	3	Patch-mainline: obsolete
82094b55	4	References: bnc#482614, bnc#537435
2cb7cef9	5
82094b55 AF	6	--- sle11-2009-09-18.orig/arch/x86/kernel/e820-xen.c 2009-02-17 14:01:20.000000000 +0100
	7	+++ sle11-2009-09-18/arch/x86/kernel/e820-xen.c 2009-09-18 10:08:03.000000000 +0200
	8	@@ -1317,6 +1317,26 @@ static int __init parse_memopt(char *p)
2cb7cef9 BS	9
	10	i = e820.nr_map - 1;
	11	current_end = e820.map[i].addr + e820.map[i].size;
	12	+
	13	+ /*
	14	+ * A little less than 2% of available memory are needed for page
	15	+ * tables, p2m map, and mem_map. Hence the maximum amount of memory
	16	+ * we can potentially balloon up to can in no case exceed about 50
	17	+ * times of what we've been given initially. Since even with that we
	18	+ * won't be able to boot (due to various calculations done based on
	19	+ * the total number of pages) we further restrict this to factor 32.
	20	+ */
	21	+ if ((mem_size >> (PAGE_SHIFT + 5)) > xen_start_info->nr_pages) {
	22	+ u64 size = (u64)xen_start_info->nr_pages << 5;
	23	+
	24	+ printk(KERN_WARNING "mem=%Luk is invalid for an initial"
	25	+ " allocation of %luk, using %Luk\n",
	26	+ (unsigned long long)mem_size >> 10,
	27	+ xen_start_info->nr_pages << (PAGE_SHIFT - 10),
	28	+ (unsigned long long)size << (PAGE_SHIFT - 10));
	29	+ mem_size = size << PAGE_SHIFT;
	30	+ }
	31	+
	32	if (current_end < mem_size) {
	33	/*
	34	* The e820 map ends before our requested size so
82094b55	35	@@ -1416,7 +1436,7 @@ void __init e820_reserve_resources(void)
2cb7cef9 BS	36	struct resource *res;
	37	u64 end;
	38
	39	- res = alloc_bootmem_low(sizeof(struct resource) * e820.nr_map);
	40	+ res = alloc_bootmem(sizeof(struct resource) * e820.nr_map);
	41	for (i = 0; i < e820.nr_map; i++) {
	42	end = e820.map[i].addr + e820.map[i].size - 1;
	43	#ifndef CONFIG_RESOURCES_64BIT
82094b55 AF	44	@@ -1498,6 +1518,7 @@ char *__init __attribute__((weak)) machi
	45	char * __init memory_setup(void)
	46	{
	47	int rc, nr_map;
	48	+ unsigned long long maxmem;
	49	struct xen_memory_map memmap;
	50	/*
	51	* This is rather large for a stack variable but this early in
	52	@@ -1527,6 +1548,22 @@ char * __init memory_setup(void)
	53	BUG();
	54
	55	#ifdef CONFIG_XEN
	56	+ /* See the comment in parse_memopt(). */
	57	+ for (maxmem = rc = 0; rc < e820.nr_map; ++rc)
	58	+ if (e820.map[rc].type == E820_RAM)
	59	+ maxmem += e820.map[rc].size;
	60	+ if ((maxmem >> (PAGE_SHIFT + 5)) > xen_start_info->nr_pages) {
	61	+ unsigned long long size = (u64)xen_start_info->nr_pages << 5;
	62	+
	63	+ printk(KERN_WARNING "maxmem of %LuM is invalid for an initial"
	64	+ " allocation of %luM, using %LuM\n",
	65	+ maxmem >> 20,
	66	+ xen_start_info->nr_pages >> (20 - PAGE_SHIFT),
	67	+ size >> (20 - PAGE_SHIFT));
	68	+ size <<= PAGE_SHIFT;
	69	+ e820_remove_range(size, ULLONG_MAX - size, E820_RAM, 1);
	70	+ }
	71	+
	72	if (is_initial_xendomain()) {
	73	memmap.nr_entries = E820MAX;
	74	set_xen_guest_handle(memmap.buffer, machine_e820.map);
	75	--- sle11-2009-09-18.orig/arch/x86/kernel/setup-xen.c 2009-07-31 15:14:20.000000000 +0200
	76	+++ sle11-2009-09-18/arch/x86/kernel/setup-xen.c 2009-07-31 15:14:31.000000000 +0200
2cb7cef9 BS	77	@@ -128,12 +128,7 @@ static struct notifier_block xen_panic_b
	78	unsigned long *phys_to_machine_mapping;
	79	EXPORT_SYMBOL(phys_to_machine_mapping);
	80
	81	-unsigned long *pfn_to_mfn_frame_list_list,
	82	-#ifdef CONFIG_X86_64
	83	- *pfn_to_mfn_frame_list[512];
	84	-#else
	85	- *pfn_to_mfn_frame_list[128];
	86	-#endif
	87	+unsigned long pfn_to_mfn_frame_list_list, *pfn_to_mfn_frame_list;
	88
	89	/* Raw start-of-day parameters from the hypervisor. */
	90	start_info_t *xen_start_info;
	91	@@ -1036,17 +1031,17 @@ void __init setup_arch(char **cmdline_p)
	92	p2m_pages = xen_start_info->nr_pages;
	93
	94	if (!xen_feature(XENFEAT_auto_translated_physmap)) {
	95	- unsigned long i, j;
	96	+ unsigned long i, j, size;
	97	unsigned int k, fpp;
	98
	99	/* Make sure we have a large enough P->M table. */
	100	phys_to_machine_mapping = alloc_bootmem_pages(
	101	max_pfn * sizeof(unsigned long));
	102	- memset(phys_to_machine_mapping, ~0,
	103	- max_pfn * sizeof(unsigned long));
	104	memcpy(phys_to_machine_mapping,
	105	(unsigned long *)xen_start_info->mfn_list,
	106	p2m_pages * sizeof(unsigned long));
	107	+ memset(phys_to_machine_mapping + p2m_pages, ~0,
	108	+ (max_pfn - p2m_pages) * sizeof(unsigned long));
	109	free_bootmem(
	110	__pa(xen_start_info->mfn_list),
	111	PFN_PHYS(PFN_UP(xen_start_info->nr_pages *
	112	@@ -1056,15 +1051,26 @@ void __init setup_arch(char **cmdline_p)
	113	* Initialise the list of the frames that specify the list of
	114	* frames that make up the p2m table. Used by save/restore.
	115	*/
	116	- pfn_to_mfn_frame_list_list = alloc_bootmem_pages(PAGE_SIZE);
	117	-
	118	fpp = PAGE_SIZE/sizeof(unsigned long);
	119	+ size = (max_pfn + fpp - 1) / fpp;
	120	+ size = (size + fpp - 1) / fpp;
	121	+ ++size; /* include a zero terminator for crash tools */
	122	+ size *= sizeof(unsigned long);
	123	+ pfn_to_mfn_frame_list_list = alloc_bootmem_pages(size);
	124	+ if (size > PAGE_SIZE
	125	+ && xen_create_contiguous_region((unsigned long)
	126	+ pfn_to_mfn_frame_list_list,
	127	+ get_order(size), 0))
	128	+ BUG();
	129	+ size -= sizeof(unsigned long);
	130	+ pfn_to_mfn_frame_list = alloc_bootmem(size);
	131	+
	132	for (i = j = 0, k = -1; i < max_pfn; i += fpp, j++) {
	133	if (j == fpp)
	134	j = 0;
	135	if (j == 0) {
	136	k++;
	137	- BUG_ON(k>=ARRAY_SIZE(pfn_to_mfn_frame_list));
	138	+ BUG_ON(k * sizeof(unsigned long) >= size);
	139	pfn_to_mfn_frame_list[k] =
	140	alloc_bootmem_pages(PAGE_SIZE);
141	pfn_to_mfn_frame_list_list[k] =
82094b55 AF	142	--- sle11-2009-09-18.orig/arch/x86/kernel/setup_percpu-xen.c 2009-08-26 12:17:43.000000000 +0200
	143	+++ sle11-2009-09-18/arch/x86/kernel/setup_percpu-xen.c 2009-08-26 12:19:19.000000000 +0200
	144	@@ -206,7 +206,7 @@ static void __init setup_node_to_cpumask
2cb7cef9 BS	145	}
	146
	147	/* allocate the map */
	148	- map = alloc_bootmem_low(nr_node_ids * sizeof(cpumask_t));
	149	+ map = alloc_bootmem(nr_node_ids * sizeof(cpumask_t));
	150
	151	pr_debug(KERN_DEBUG "Node to cpumask map at %p for %d nodes\n",
	152	map, nr_node_ids);
82094b55 AF	153	--- sle11-2009-09-18.orig/drivers/xen/core/machine_kexec.c 2009-06-04 10:21:39.000000000 +0200
82094b55 AF	154	+++ sle11-2009-09-18/drivers/xen/core/machine_kexec.c 2009-03-13 16:13:15.000000000 +0100
2cb7cef9 BS	155	@@ -57,7 +57,7 @@ void __init xen_machine_kexec_setup_reso
	156
	157	/* allocate xen_phys_cpus */
	158
	159	- xen_phys_cpus = alloc_bootmem_low(k * sizeof(struct resource));
	160	+ xen_phys_cpus = alloc_bootmem(k * sizeof(struct resource));
	161	BUG_ON(xen_phys_cpus == NULL);
	162
	163	/* fill in xen_phys_cpus with per-cpu crash note information */
82094b55 AF	164	--- sle11-2009-09-18.orig/drivers/xen/core/machine_reboot.c 2009-02-17 12:23:48.000000000 +0100
82094b55 AF	165	+++ sle11-2009-09-18/drivers/xen/core/machine_reboot.c 2009-02-17 12:25:29.000000000 +0100
2cb7cef9 BS	166	@@ -76,7 +76,7 @@ static void post_suspend(int suspend_can
	167	unsigned long shinfo_mfn;
	168	extern unsigned long max_pfn;
	169	extern unsigned long *pfn_to_mfn_frame_list_list;
	170	- extern unsigned long *pfn_to_mfn_frame_list[];
	171	+ extern unsigned long **pfn_to_mfn_frame_list;
	172
	173	if (suspend_cancelled) {
	174	xen_start_info->store_mfn =