Skip to contentSkip to conversation

x86: don't reallocate pgt for node0

PATCH Applied Yinghai Lu 2 participants 15 messages Jul 1, 2008 – Jul 3, 2008 1 message not in the archive

Subsystems (2)
x86 x86 MM
lore
Share & tools
lore

Yinghai Lu

author
#0 lore raw

PATCH 1/1 x86: don't reallocate pgt for node0

kva ram already mapped right after away, so don't need to get that for low ram. avoid wasting one copy pgdat.

also add node id in early_res name in case get it from find_e820_area.

Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/mm/discontig_32.c |   11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
arch/x86/mm/discontig_32.c +7 -4
Index: linux-2.6/arch/x86/mm/discontig_32.c
===================================================================
@@ -156,17 +156,20 @@ static void __init propagate_e820_map_no
  */
 static void __init allocate_pgdat(int nid)
 {
-	if (nid && node_has_online_mem(nid) && node_remap_start_vaddr[nid])
+	char buf[16];
+
+	if (node_has_online_mem(nid) && node_remap_start_vaddr[nid])
 		NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid];
 	else {
 		unsigned long pgdat_phys;
 		pgdat_phys = find_e820_area(min_low_pfn<<PAGE_SHIFT,
-				 (nid ? max_low_pfn:max_pfn_mapped)<<PAGE_SHIFT,
+				 max_pfn_mapped<<PAGE_SHIFT,
 				 sizeof(pg_data_t),
 				 PAGE_SIZE);
 		NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT));
-		reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t),
-			      "NODE_DATA");
+		memset(buf, 0, sizeof(buf));
+		sprintf(buf, "NODE_DATA %d",  nid);
+		reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf);
 	}
 	printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n",
 		nid, (unsigned long)NODE_DATA(nid));
Applied
#1 lore raw
Quoted 4 lines · Yinghai Lu

kva ram already mapped right after away, so don't need to get that for low ram. avoid wasting one copy pgdat.

also add node id in early_res name in case get it from find_e820_area.

applied to tip/x86/unify-setup - thanks Yinghai.

Ingo

author
#2 lore raw

PATCH 1/1 x86: change copy_e820_map to append_e820_map

In reply to Yinghai Lu

so it has meaningfull name.
also change it to static

Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/e820.c |   15 ++++++++-------
 include/asm-x86/e820.h |    1 -
 2 files changed, 8 insertions(+), 8 deletions(-)
arch/x86/kernel/e820.c +6 -5
Index: linux-2.6/arch/x86/kernel/e820.c
===================================================================
@@ -360,7 +360,7 @@ int __init sanitize_e820_map(struct e820
 	return 0;
 }
 
-static int __init __copy_e820_map(struct e820entry *biosmap, int nr_map)
+static int __init __append_e820_map(struct e820entry *biosmap, int nr_map)
 {
 	while (nr_map) {
 		u64 start = biosmap->addr;

@@ -389,13 +389,13 @@ static int __init __copy_e820_map(struct
  * will have given us a memory map that we can use to properly
  * set up memory.  If we aren't, we'll fake a memory map.
  */
-int __init copy_e820_map(struct e820entry *biosmap, int nr_map)
+static int __init append_e820_map(struct e820entry *biosmap, int nr_map)
 {
 	/* Only one memory region (or negative)? Ignore it */
 	if (nr_map < 2)
 		return -1;
 
-	return __copy_e820_map(biosmap, nr_map);
+	return __append_e820_map(biosmap, nr_map);
 }
 
 u64 __init e820_update_range(u64 start, u64 size, unsigned old_type,

@@ -581,7 +581,7 @@ void __init parse_e820_ext(struct setup_
 	if (map_len > PAGE_SIZE)
 		sdata = early_ioremap(pa_data, map_len);
 	extmap = (struct e820entry *)(sdata->data);
-	__copy_e820_map(extmap, entries);
+	__append_e820_map(extmap, entries);
 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 	if (map_len > PAGE_SIZE)
 		early_iounmap(sdata, map_len);

@@ -1245,7 +1245,8 @@ char *__init default_machine_specific_me
 			ARRAY_SIZE(boot_params.e820_map),
 			&new_nr);
 	boot_params.e820_entries = new_nr;
-	if (copy_e820_map(boot_params.e820_map, boot_params.e820_entries) < 0) {
+	if (append_e820_map(boot_params.e820_map, boot_params.e820_entries)
+	  < 0) {
 		u64 mem_size;
 
 		/* compare results from other methods and take the greater */
include/asm-x86/e820.h +0 -1
Index: linux-2.6/include/asm-x86/e820.h
===================================================================
@@ -67,7 +67,6 @@ extern void e820_add_region(u64 start, u
 extern void e820_print_map(char *who);
 extern int
 sanitize_e820_map(struct e820entry *biosmap, int max_nr_map, int *pnr_map);
-extern int copy_e820_map(struct e820entry *biosmap, int nr_map);
 extern u64 e820_update_range(u64 start, u64 size, unsigned old_type,
 			       unsigned new_type);
 extern u64 e820_remove_range(u64 start, u64 size, unsigned old_type,
author
#4 lore raw

PATCH 1/1 x86: make early_res_to_bootmem print out less 80 width chars

In reply to Yinghai Lu
Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/e820.c |   15 ++++++++-------
 include/asm-x86/e820.h |    1 -
 2 files changed, 8 insertions(+), 8 deletions(-)
arch/x86/kernel/e820.c +2 -2
Index: linux-2.6/arch/x86/kernel/e820.c
===================================================================
@@ -839,7 +839,7 @@ void __init early_res_to_bootmem(u64 sta
 	printk(KERN_INFO "(%d early reservations) ==> bootmem\n", count);
 	for (i = 0; i < count; i++) {
 		struct early_res *r = &early_res[i];
-		printk(KERN_INFO "  #%d [ %010llx - %010llx ] %16s", i,
+		printk(KERN_INFO "  #%d [%010llx - %010llx] %16s", i,
 			r->start, r->end, r->name);
 		final_start = max(start, r->start);
 		final_end = min(end, r->end);

@@ -847,7 +847,7 @@ void __init early_res_to_bootmem(u64 sta
 			printk(KERN_CONT "\n");
 			continue;
 		}
-		printk(KERN_CONT " ===> [ %010llx - %010llx ]\n",
+		printk(KERN_CONT " ==> [%010llx - %010llx]\n",
 			final_start, final_end);
 		reserve_bootmem_generic(final_start, final_end - final_start,
 				BOOTMEM_DEFAULT);
author
#6 lore raw

PATCH 1/1 x86: not printout if not finding setup_data

In reply to Yinghai Lu
Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/setup.c |    5 +++++
 1 file changed, 5 insertions(+)
arch/x86/kernel/setup.c +5 -0
Index: linux-2.6/arch/x86/kernel/setup.c
===================================================================
@@ -399,6 +399,7 @@ static void __init reserve_setup_data(vo
 	struct setup_data *data;
 	u64 pa_data;
 	char buf[32];
+	int found = 0;
 
 	if (boot_params.hdr.version < 0x0209)
 		return;

@@ -409,9 +410,13 @@ static void __init reserve_setup_data(vo
 		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
 		e820_update_range(pa_data, sizeof(*data)+data->len,
 			 E820_RAM, E820_RESERVED_KERN);
+		found = 1;
 		pa_data = data->next;
 		early_iounmap(data, sizeof(*data));
 	}
+	if (!found)
+		return;
+
 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 	printk(KERN_INFO "extended physical RAM map:\n");
 	e820_print_map("reserve setup_data");
author
#8 lore raw

PATCH 1/1 x86: merge zones_sizes_init for numa and non numa on 32bit

In reply to Yinghai Lu

move out e820_register_active_regions from non numa zones_sizes_init() and remove numa version zones_sizes_init().

and let 32 bit call remove_all_active_ranges in setup_arch directly like 64 bit

Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/setup.c    |    2 --
 arch/x86/mm/discontig_32.c |   16 ----------------
 arch/x86/mm/init_32.c      |   10 ++++------
 include/asm-x86/page_32.h  |    1 -
 4 files changed, 4 insertions(+), 25 deletions(-)
arch/x86/kernel/setup.c +0 -2
Index: linux-2.6/arch/x86/kernel/setup.c
===================================================================
@@ -719,10 +719,8 @@ void __init setup_arch(char **cmdline_p)
 	 */
 	acpi_boot_table_init();
 
-#ifdef CONFIG_X86_64
 	/* Remove active ranges so rediscovery with NUMA-awareness happens */
 	remove_all_active_ranges();
-#endif
 
 #ifdef CONFIG_ACPI_NUMA
 	/*
arch/x86/mm/discontig_32.c +0 -16
Index: linux-2.6/arch/x86/mm/discontig_32.c
===================================================================
@@ -327,7 +327,6 @@ void __init initmem_init(unsigned long s
 	 * and ZONE_HIGHMEM.
 	 */
 
-	remove_all_active_ranges();
 	get_memcfg_numa();
 
 	kva_pages = round_up(calculate_numa_remap_pages(), PTRS_PER_PTE);

@@ -390,21 +389,6 @@ void __init initmem_init(unsigned long s
 	setup_bootmem_allocator();
 }
 
-void __init zone_sizes_init(void)
-{
-	unsigned long max_zone_pfns[MAX_NR_ZONES];
-	memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
-	max_zone_pfns[ZONE_DMA] =
-		virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
-	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
-#ifdef CONFIG_HIGHMEM
-	max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
-#endif
-
-	free_area_init_nodes(max_zone_pfns);
-	return;
-}
-
 void __init set_highmem_pages_init(void)
 {
 #ifdef CONFIG_HIGHMEM
arch/x86/mm/init_32.c +4 -6
Index: linux-2.6/arch/x86/mm/init_32.c
===================================================================
@@ -660,12 +660,14 @@ void __init initmem_init(unsigned long s
 	if (max_pfn > max_low_pfn)
 		highstart_pfn = max_low_pfn;
 	memory_present(0, 0, highend_pfn);
+	e820_register_active_regions(0, 0, highend_pfn);
 	printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
 		pages_to_mb(highend_pfn - highstart_pfn));
 	num_physpages = highend_pfn;
 	high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
 #else
 	memory_present(0, 0, max_low_pfn);
+	e820_register_active_regions(0, 0, max_low_pfn);
 	num_physpages = max_low_pfn;
 	high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
 #endif

@@ -677,25 +679,21 @@ void __init initmem_init(unsigned long s
 
 	setup_bootmem_allocator();
 }
+#endif /* !CONFIG_NEED_MULTIPLE_NODES */
 
-void __init zone_sizes_init(void)
+static void __init zone_sizes_init(void)
 {
 	unsigned long max_zone_pfns[MAX_NR_ZONES];
 	memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
 	max_zone_pfns[ZONE_DMA] =
 		virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 	max_zone_pfns[ZONE_NORMAL] = max_low_pfn;
-	remove_all_active_ranges();
 #ifdef CONFIG_HIGHMEM
 	max_zone_pfns[ZONE_HIGHMEM] = highend_pfn;
-	e820_register_active_regions(0, 0, highend_pfn);
-#else
-	e820_register_active_regions(0, 0, max_low_pfn);
 #endif
 
 	free_area_init_nodes(max_zone_pfns);
 }
-#endif /* !CONFIG_NEED_MULTIPLE_NODES */
 
 void __init setup_bootmem_allocator(void)
 {
include/asm-x86/page_32.h +0 -1
Index: linux-2.6/include/asm-x86/page_32.h
===================================================================
@@ -101,7 +101,6 @@ extern void find_low_pfn_range(void);
 extern unsigned long init_memory_mapping(unsigned long start,
 					 unsigned long end);
 extern void initmem_init(unsigned long, unsigned long);
-extern void zone_sizes_init(void);
 extern void setup_bootmem_allocator(void);
author
#10 lore raw

PATCH 1/1 x86: move init_cpu_to_node after get_smp_config

In reply to Yinghai Lu

when acpi=off, cpu_to_apicid is ready after get_smp_config so need to move init_cpu_to_node after it.

otherwise, we will get wrong cpu->node mapping. and it will rely on amd_detect_cmp to correct it. but that is too late. setup_per_cpu_data is already called before that.
we will get per_cpu_data on wrong node.

Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/setup.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/x86/kernel/setup.c +4 -4
Index: linux-2.6/arch/x86/kernel/setup.c
===================================================================
@@ -780,10 +780,6 @@ void __init setup_arch(char **cmdline_p)
 	 */
 	acpi_boot_init();
 
-#ifdef CONFIG_X86_64
-	init_cpu_to_node();
-#endif
-
 #if defined(CONFIG_X86_MPPARSE) || defined(CONFIG_X86_VISWS)
 	/*
 	 * get boot-time SMP configuration:

@@ -792,6 +788,10 @@ void __init setup_arch(char **cmdline_p)
 		get_smp_config();
 #endif
 
+#ifdef CONFIG_X86_64
+	init_cpu_to_node();
+#endif
+
 	init_apic_mappings();
 	ioapic_init_mappings();
author
#13 lore raw

PATCH 1/1 x86: move prefill_possible_map calling early

In reply to a message not in the archive

call it right after we done with MADT/mptable handling, instead of doing that in setup_per_cpu_areas later...
so for_possible_cpu could be used early.

Signed-off-by Yinghai Lu <[email protected]>
---
 arch/x86/kernel/setup.c        |    1 +
 arch/x86/kernel/setup_percpu.c |   10 ----------
 arch/x86/kernel/smpboot.c      |    8 ++++++++
 include/asm-x86/smp.h          |    4 ++++
 4 files changed, 13 insertions(+), 10 deletions(-)
arch/x86/kernel/setup.c +1 -0
Index: linux-2.6/arch/x86/kernel/setup.c
===================================================================
@@ -788,6 +788,7 @@ void __init setup_arch(char **cmdline_p)
 		get_smp_config();
 #endif
 
+	prefill_possible_map();
 #ifdef CONFIG_X86_64
 	init_cpu_to_node();
 #endif
arch/x86/kernel/setup_percpu.c +0 -10
Index: linux-2.6/arch/x86/kernel/setup_percpu.c
===================================================================
@@ -162,16 +162,6 @@ void __init setup_per_cpu_areas(void)
 	char *ptr;
 	int cpu;
 
-	/* no processor from mptable or madt */
-	if (!num_processors)
-		num_processors = 1;
-
-#ifdef CONFIG_HOTPLUG_CPU
-	prefill_possible_map();
-#else
-	nr_cpu_ids = num_processors;
-#endif
-
 	/* Setup cpu_pda map */
 	setup_cpu_pda_map();
arch/x86/kernel/smpboot.c +8 -0
Index: linux-2.6/arch/x86/kernel/smpboot.c
===================================================================
@@ -1278,12 +1278,20 @@ __init void prefill_possible_map(void)
 	int i;
 	int possible;
 
+	/* no processor from mptable or madt */
+	if (!num_processors)
+		num_processors = 1;
+
+#ifdef CONFIG_HOTPLUG_CPU
 	if (additional_cpus == -1) {
 		if (disabled_cpus > 0)
 			additional_cpus = disabled_cpus;
 		else
 			additional_cpus = 0;
 	}
+#else
+	additional_cpus = 0;
+#endif
 	possible = num_processors + additional_cpus;
 	if (possible > NR_CPUS)
 		possible = NR_CPUS;
include/asm-x86/smp.h +4 -0
Index: linux-2.6/include/asm-x86/smp.h
===================================================================
@@ -124,6 +124,10 @@ static inline int num_booting_cpus(void)
 {
 	return cpus_weight(cpu_callout_map);
 }
+#else
+static inline void prefill_possible_map(void)
+{
+}
 #endif /* CONFIG_SMP */
 
 extern unsigned disabled_cpus __cpuinitdata;
Applied
#14 lore raw

Re: [PATCH] x86: move prefill_possible_map calling early

Quoted 3 lines · Yinghai Lu

call it right after we done with MADT/mptable handling, instead of doing that in setup_per_cpu_areas later... so for_possible_cpu could be used early.

applied to tip/x86/unify-setup - thanks Yinghai,

Ingo

Applied
#11 lore raw

Re: [PATCH] x86: move init_cpu_to_node after get_smp_config

In reply to Yinghai Lu
Quoted 6 lines · Yinghai Lu

when acpi=off, cpu_to_apicid is ready after get_smp_config so need to move init_cpu_to_node after it.

otherwise, we will get wrong cpu->node mapping. and it will rely on amd_detect_cmp to correct it. but that is too late. setup_per_cpu_data is already called before that. we will get per_cpu_data on wrong node.

applied to tip/x86/unify-setup - thanks Yinghai,

Ingo

Applied
#9 lore raw

Re: [PATCH] x86: merge zones_sizes_init for numa and non numa on 32bit

In reply to Yinghai Lu
Quoted 5 lines · Yinghai Lu

move out e820_register_active_regions from non numa zones_sizes_init() and remove numa version zones_sizes_init().

and let 32 bit call remove_all_active_ranges in setup_arch directly like 64 bit

applied to tip/x86/unify-setup - thanks,

Ingo

Applied
#7 lore raw

Re: [PATCH] x86: not printout if not finding setup_data

In reply to Yinghai Lu

arch/x86/kernel/setup.c static void __init reserve_setup_data(vo · line 422

Quoted 13 lines · Yinghai Lu
arch/x86/kernel/setup.c +4 -0
 		reserve_early(pa_data, pa_data+sizeof(*data)+data->len, buf);
 		e820_update_range(pa_data, sizeof(*data)+data->len,
 			 E820_RAM, E820_RESERVED_KERN);
+		found = 1;
 		pa_data = data->next;
 		early_iounmap(data, sizeof(*data));
 	}
+	if (!found)
+		return;
+
 	sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
 	printk(KERN_INFO "extended physical RAM map:\n");
 	e820_print_map("reserve setup_data");

applied to tip/x86/unify-setup - thanks,

Ingo

Applied
#5 lore raw

Re: [PATCH] x86: make early_res_to_bootmem print out less 80 width chars

In reply to Yinghai Lu

arch/x86/kernel/e820.c void __init early_res_to_bootmem(u64 sta · line 843

Quoted 4 lines · Yinghai Lu
arch/x86/kernel/e820.c +1 -1
 		struct early_res *r = &early_res[i];
-		printk(KERN_INFO "  #%d [ %010llx - %010llx ] %16s", i,
+		printk(KERN_INFO "  #%d [%010llx - %010llx] %16s", i,
 			r->start, r->end, r->name);

applied to tip/x86/unify-setup - thanks,

Ingo

Applied
#3 lore raw

Re: [PATCH] x86: change copy_e820_map to append_e820_map

In reply to Yinghai Lu
Quoted 2 lines · Yinghai Lu

so it has meaningfull name.
also change it to static

applied to tip/x86/unify-setup - thanks,

Ingo

#15 lore raw

Re: [PATCH] x86: move prefill_possible_map calling early

In reply to Ingo Molnar

FYI, this one needed the small fixlet below. (prefill_possible_map() is dependent on both CONFIG_SMP and CONFIG_HOTPLUG_CPU - not just CONFIG_SMP)

Ingo

-------------->
commit bb42b7bdee9decad536c1e747e2ce2f3daebb76f

Author: Ingo Molnar <[email protected]>
Date:   Thu Jul 3 15:57:47 2008 +0200

x86: move prefill_possible_map calling early, fix

fix:

arch/x86/kernel/built-in.o: In function `setup_arch':
: undefined reference to `prefill_possible_map'

Signed-off-by Ingo Molnar <[email protected]>
include/asm-x86/smp.h +5 -3
index bf6b5ee..c2784b3 100644
@@ -114,8 +114,6 @@ void native_send_call_func_single_ipi(int cpu);
 extern int __cpu_disable(void);
 extern void __cpu_die(unsigned int cpu);
 
-extern void prefill_possible_map(void);
-
 void smp_store_cpu_info(int id);
 #define cpu_physical_id(cpu)	per_cpu(x86_cpu_to_apicid, cpu)
 

@@ -124,11 +122,15 @@ static inline int num_booting_cpus(void)
 {
 	return cpus_weight(cpu_callout_map);
 }
+#endif /* CONFIG_SMP */
+
+#if defined(CONFIG_SMP) && defined(CONFIG_HOTPLUG_CPU)
+extern void prefill_possible_map(void);
 #else
 static inline void prefill_possible_map(void)
 {
 }
-#endif /* CONFIG_SMP */
+#endif
 
 extern unsigned disabled_cpus __cpuinitdata;