x86: Reduce memory usage for large count NR_CPUs fixup V2 with git-x86

PATCH Applied [email protected] 2 participants 6 messages Jan 21, 2008 – Jan 22, 2008

Subsystems (3)

Memory management x86 x86 MM

lore

Share & tools

lore

[email protected]

author

2008-01-21 21:16 UTC #0 lore raw

cover x86: Reduce memory usage for large count NR_CPUs fixup V2 with git-x86

Fixup change NR_CPUS patchset by rebasing on 2.6.24-rc8-mm1 from 2.6.24-rc6-mm1) and adding changes suggested by reviews.

Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

Note there are two versions of this patchset:
- 2.6.24-rc8-mm1
- 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

Signed-off-by Mike Travis <[email protected]>

---
Fixup-V2:
- pulled the SMP_MAX patch as it's not strictly needed and some more work on local cpumask_t variables needs to be done before NR_CPUS is allowed to increase.

- changes to X86_32 have been removed (except for build errors)
---

Signature · 5 lines

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

[email protected]

author

2008-01-21 21:16 UTC #1 lore raw

PATCH 1/3 x86: Change size of node ids from u8 to s16 fixup V2 with git-x86

Change the size of node ids for X86_64 from u8 to s16 to accomodate more than 32k nodes and allow for NUMA_NO_NODE (-1) to be sign extended to int.

Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

Cc David Rientjes <[email protected]>

Cc Yinghai Lu <[email protected]>

Cc Eric Dumazet <[email protected]>

Signed-off-by Mike Travis <[email protected]>

Reviewed-by Christoph Lameter <[email protected]>

---
fixup-V2:

- Fixed populate_memnodemap as suggested by Eric.
- Change to using s16 for static node id arrays and int for node id's in per_cpu variables and __initdata arrays as suggested by David and Yinghai.
- NUMA_NO_NODE is now (-1)

fixup:

- Size of memnode.embedded_map needs to be changed to accomodate 16-bit node ids as suggested by Eric.

V2->V3:
- changed memnode.embedded_map from [64-16] to [64-8] (and size comment to 128 bytes)

V1->V2:
- changed pxm_to_node_map to u16
- changed memnode map entries to u16

---
 arch/x86/Kconfig            |    1 +
 arch/x86/mm/numa_64.c       |   12 ++++++------
 include/asm-x86/mmzone_64.h |    6 +++---
 include/asm-x86/numa_64.h   |    2 +-
 include/asm-x86/topology.h  |   10 +++++-----
 5 files changed, 16 insertions(+), 15 deletions(-)

arch/x86/Kconfig +1 -0

@@ -863,6 +863,7 @@ config NUMA_EMU
 
 config NODES_SHIFT
 	int
+	range 1 15  if X86_64
 	default "6" if X86_64
 	default "4" if X86_NUMAQ
 	default "3"

arch/x86/mm/numa_64.c +6 -6

@@ -31,17 +31,17 @@ bootmem_data_t plat_node_bdata[MAX_NUMNO
 
 struct memnode memnode;
 
-u16 x86_cpu_to_node_map_init[NR_CPUS] = {
+int x86_cpu_to_node_map_init[NR_CPUS] = {
 	[0 ... NR_CPUS-1] = NUMA_NO_NODE
 };
 void *x86_cpu_to_node_map_early_ptr;
-DEFINE_PER_CPU(u16, x86_cpu_to_node_map) = NUMA_NO_NODE;
+DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
 EXPORT_PER_CPU_SYMBOL(x86_cpu_to_node_map);
 #ifdef	CONFIG_DEBUG_PER_CPU_MAPS
 EXPORT_SYMBOL(x86_cpu_to_node_map_early_ptr);
 #endif
 
-u16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
+s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
 	[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
 };

@@ -65,7 +65,7 @@ static int __init populate_memnodemap(co
 	unsigned long addr, end;
 	int i, res = -1;
 
-	memset(memnodemap, 0xff, memnodemapsize);
+	memset(memnodemap, 0xff, sizeof(s16)*memnodemapsize);
 	for (i = 0; i < numnodes; i++) {
 		addr = nodes[i].start;
 		end = nodes[i].end;

@@ -74,7 +74,7 @@ static int __init populate_memnodemap(co
 		if ((end >> shift) >= memnodemapsize)
 			return 0;
 		do {
-			if (memnodemap[addr >> shift] != 0xff)
+			if (memnodemap[addr >> shift] != NUMA_NO_NODE)
 				return -1;
 			memnodemap[addr >> shift] = i;
 			addr += (1UL << shift);

@@ -535,7 +535,7 @@ __cpuinit void numa_add_cpu(int cpu)
 
 void __cpuinit numa_set_node(int cpu, int node)
 {
-	u16 *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
+	int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
 
 	cpu_pda(cpu)->nodenumber = node;

include/asm-x86/mmzone_64.h +3 -3

@@ -15,9 +15,9 @@
 struct memnode {
 	int shift;
 	unsigned int mapsize;
-	u8 *map;
-	u8 embedded_map[64-16];
-} ____cacheline_aligned; /* total size = 64 bytes */
+	s16 *map;
+	s16 embedded_map[64-8];
+} ____cacheline_aligned; /* total size = 128 bytes */
 extern struct memnode memnode;
 #define memnode_shift memnode.shift
 #define memnodemap memnode.map

include/asm-x86/numa_64.h +1 -1

@@ -20,7 +20,7 @@ extern void numa_set_node(int cpu, int n
 extern void srat_reserve_add_area(int nodeid);
 extern int hotadd_percent;
 
-extern u16 apicid_to_node[MAX_LOCAL_APIC];
+extern s16 apicid_to_node[MAX_LOCAL_APIC];
 
 extern void numa_initmem_init(unsigned long start_pfn, unsigned long end_pfn);
 extern unsigned long numa_free_all_bootmem(void);

include/asm-x86/topology.h +5 -5

@@ -31,17 +31,17 @@
 
 /* Mappings between logical cpu number and node number */
 #ifdef CONFIG_X86_32
-extern u8 cpu_to_node_map[];
+extern int cpu_to_node_map[];
 
 #else
-DECLARE_PER_CPU(u16, x86_cpu_to_node_map);
-extern u16 x86_cpu_to_node_map_init[];
+DECLARE_PER_CPU(int, x86_cpu_to_node_map);
+extern int x86_cpu_to_node_map_init[];
 extern void *x86_cpu_to_node_map_early_ptr;
 #endif
 
 extern cpumask_t node_to_cpumask_map[];
 
-#define NUMA_NO_NODE	((u16)(~0))
+#define NUMA_NO_NODE	(-1)
 
 /* Returns the number of the node containing CPU 'cpu' */
 #ifdef CONFIG_X86_32

@@ -54,7 +54,7 @@ static inline int cpu_to_node(int cpu)
 #else /* CONFIG_X86_64 */
 static inline int early_cpu_to_node(int cpu)
 {
-	u16 *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
+	int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
 
 	if (cpu_to_node_map)
 		return cpu_to_node_map[cpu];

Signature · 5 lines

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

[email protected]

author

2008-01-21 21:16 UTC #2 lore raw

PATCH 2/3 x86: Change NR_CPUS arrays in numa_64 fixup V2 with git-x86

In reply to [email protected]

Change the following static arrays sized by NR_CPUS to per_cpu data variables:

	char cpu_to_node_map[NR_CPUS];

Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

Signed-off-by Mike Travis <[email protected]>

Reviewed-by Christoph Lameter <[email protected]>

---
fixup:

- Split cpu_to_node function into "early" and "late" versions so that x86_cpu_to_node_map_early_ptr is not EXPORT'ed and the cpu_to_node inline function is more streamlined.

- This also involves setting up the percpu maps as early as possible.

- Fix X86_32 NUMA build errors that previous version of this patch caused.

V2->V3:
- add early_cpu_to_node function to keep cpu_to_node efficient
- move and rename smp_set_apicids() to setup_percpu_maps()
- call setup_percpu_maps() as early as possible

V1->V2:
- Removed extraneous casts
- Fix !NUMA builds with '#ifdef CONFIG_NUMA"

---
 arch/x86/kernel/setup64.c    |   10 +++++-----
 arch/x86/kernel/smpboot_32.c |    2 +-
 arch/x86/mm/srat_64.c        |    2 +-
 include/asm-x86/topology.h   |    9 +++++++++
 4 files changed, 16 insertions(+), 7 deletions(-)

arch/x86/kernel/setup64.c +5 -5

@@ -87,10 +87,10 @@ __setup("noexec32=", nonx32_setup);
 
 /*
  * Copy data used in early init routines from the initial arrays to the
- * per cpu data areas.  These arrays then become expendable and the *_ptrs
- * are zeroed indicating that the static arrays are gone.
+ * per cpu data areas.  These arrays then become expendable and the
+ * *_early_ptr's are zeroed indicating that the static arrays are gone.
  */
-void __init setup_percpu_maps(void)
+static void __init setup_per_cpu_maps(void)
 {
 	int cpu;

@@ -114,7 +114,7 @@ void __init setup_percpu_maps(void)
 #endif
 	}
 
-	/* indicate the early static arrays are gone */
+	/* indicate the early static arrays will soon be gone */
 	x86_cpu_to_apicid_early_ptr = NULL;
 	x86_bios_cpu_apicid_early_ptr = NULL;
 #ifdef CONFIG_NUMA

@@ -157,7 +157,7 @@ void __init setup_per_cpu_areas(void)
 	}
 
 	/* setup percpu data maps early */
-	setup_percpu_maps();
+	setup_per_cpu_maps();
 } 
 
 void pda_init(int cpu)

arch/x86/kernel/smpboot_32.c +1 -1

@@ -460,7 +460,7 @@ cpumask_t node_to_cpumask_map[MAX_NUMNOD
 				{ [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
 EXPORT_SYMBOL(node_to_cpumask_map);
 /* which node each logical CPU is on */
-u8 cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
+int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
 EXPORT_SYMBOL(cpu_to_node_map);
 
 /* set up a mapping between cpu and node. */

arch/x86/mm/srat_64.c +1 -1

@@ -408,7 +408,7 @@ int __init acpi_scan_nodes(unsigned long
 static int fake_node_to_pxm_map[MAX_NUMNODES] __initdata = {
 	[0 ... MAX_NUMNODES-1] = PXM_INVAL
 };
-static u16 fake_apicid_to_node[MAX_LOCAL_APIC] __initdata = {
+static s16 fake_apicid_to_node[MAX_LOCAL_APIC] __initdata = {
 	[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
 };
 static int __init find_node_by_addr(unsigned long addr)

include/asm-x86/topology.h +9 -0

@@ -81,6 +81,15 @@ static inline int cpu_to_node(int cpu)
 }
 #endif /* CONFIG_X86_64 */
 
+static inline int cpu_to_node(int cpu)
+{
+	if(per_cpu_offset(cpu))
+		return per_cpu(x86_cpu_to_node_map, cpu);
+	else
+		return NUMA_NO_NODE;
+}
+#endif /* CONFIG_X86_64 */
+
 /*
  * Returns the number of the node containing Node 'node'. This
  * architecture is flat, so it is a pretty simple function!

Signature · 5 lines

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

[email protected]

author

2008-01-21 21:16 UTC #3 lore raw

PATCH 3/3 x86: Add debug of invalid per_cpu map accesses fixup V2 with git-x86

In reply to [email protected]

Provide a means to trap usages of per_cpu map variables before they are setup. Define CONFIG_DEBUG_PER_CPU_MAPS to activate.

Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

Signed-off-by Mike Travis <[email protected]>

---
 include/asm-x86/topology.h |   13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)

include/asm-x86/topology.h +2 ~~-11~~

@@ -58,7 +58,7 @@ static inline int early_cpu_to_node(int 
 
 	if (cpu_to_node_map)
 		return cpu_to_node_map[cpu];
-	else if(per_cpu_offset(cpu))
+	else if (per_cpu_offset(cpu))
 		return per_cpu(x86_cpu_to_node_map, cpu);
 	else
 		return NUMA_NO_NODE;

@@ -71,7 +71,7 @@ static inline int cpu_to_node(int cpu)
 		printk("KERN_NOTICE cpu_to_node(%d): usage too early!\n",
 			(int)cpu);
 		dump_stack();
-		return ((u16 *)x86_cpu_to_node_map_early_ptr)[cpu];
+		return ((int *)x86_cpu_to_node_map_early_ptr)[cpu];
 	}
 #endif
 	if (per_cpu_offset(cpu))

@@ -81,15 +81,6 @@ static inline int cpu_to_node(int cpu)
 }
 #endif /* CONFIG_X86_64 */
 
-static inline int cpu_to_node(int cpu)
-{
-	if(per_cpu_offset(cpu))
-		return per_cpu(x86_cpu_to_node_map, cpu);
-	else
-		return NUMA_NO_NODE;
-}
-#endif /* CONFIG_X86_64 */
-
 /*
  * Returns the number of the node containing Node 'node'. This
  * architecture is flat, so it is a pretty simple function!

Signature · 5 lines

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Ingo Molnar

Applied

2008-01-22 12:48 UTC #4 lore raw

In reply to [email protected]

Quoted 8 lines · * [email protected] <[email protected]> wrote:

Fixup change NR_CPUS patchset by rebasing on 2.6.24-rc8-mm1 from 2.6.24-rc6-mm1) and adding changes suggested by reviews.
Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86
Note there are two versions of this patchset:
- 2.6.24-rc8-mm1
- 2.6.24-rc8-mm1 + latest (08/1/21) git-x86

thanks, applied.

Quoted 6 lines

Signed-off-by Mike Travis <[email protected]>
---
Fixup-V2:
- pulled the SMP_MAX patch as it's not strictly needed and some more work on local cpumask_t variables needs to be done before NR_CPUS is allowed to increase.

i'd still love to see CONFIG_SMP_MAX, so that we can have continuous randconfig testing of the large-SMP aspects of the x86 architecture, even on smaller systems.

What's the maximum that should work right now? 256 or perhaps even 512 CPU ought to work fine i think?

and then once the on-stack usage problems are fixed, the NR_CPUS value in CONFIG_SMP_MAX can be increased. So SMP_MAX would also act as "this is how far we can go in the upstream kernel" documentation.

[ btw., the crash i remember was rather related to the NODES_SHIFT
increase to 9, not from the NR_CPUSs increase. (the config i sent still has NR_CPUS==8, because Kconfig did not pick up the right NR_CPUs value dicatated by SMP_MAX.) If you resend the SMP_MAX patch against latest x86.git i can retest this. ]

Ingo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Mike Travis

author

2008-01-22 15:10 UTC #5 lore raw

Quoted 20 lines · Ingo Molnar

Quoted 8 lines · * [email protected] <[email protected]> wrote:
Fixup change NR_CPUS patchset by rebasing on 2.6.24-rc8-mm1 from 2.6.24-rc6-mm1) and adding changes suggested by reviews.
Based on 2.6.24-rc8-mm1 + latest (08/1/21) git-x86
Note there are two versions of this patchset:
- 2.6.24-rc8-mm1
- 2.6.24-rc8-mm1 + latest (08/1/21) git-x86
thanks, applied.
Quoted 6 lines
Signed-off-by Mike Travis <[email protected]>
---
Fixup-V2:
- pulled the SMP_MAX patch as it's not strictly needed and some more work on local cpumask_t variables needs to be done before NR_CPUS is allowed to increase.

Quoted 6 lines

i'd still love to see CONFIG_SMP_MAX, so that we can have continuous randconfig testing of the large-SMP aspects of the x86 architecture, even on smaller systems.
What's the maximum that should work right now? 256 or perhaps even 512 CPU ought to work fine i think?

I'm attempting to gather stack (and memory) usage for increased cpu counts right now. But I'll have another set of basic changes before the cpumask_t changes can be done.

Thanks,
Mike

Quoted 5 lines

and then once the on-stack usage problems are fixed, the NR_CPUS value in CONFIG_SMP_MAX can be increased. So SMP_MAX would also act as "this is how far we can go in the upstream kernel" documentation.

Quoted 7 lines

[ btw., the crash i remember was rather related to the NODES_SHIFT
increase to 9, not from the NR_CPUSs increase. (the config i sent still has NR_CPUS==8, because Kconfig did not pick up the right NR_CPUs value dicatated by SMP_MAX.) If you resend the SMP_MAX patch against latest x86.git i can retest this. ]
Ingo