public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Christoph Lameter <clameter@sgi.com>
To: akpm@linux-foundation.org
Cc: linux-mm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Subject: [patch 02/28] cpu alloc: x86_64 support
Date: Tue, 06 Nov 2007 11:51:46 -0800	[thread overview]
Message-ID: <20071106195158.107939214@sgi.com> (raw)
In-Reply-To: 20071106195144.983665861@sgi.com

[-- Attachment #1: cpu_alloc_x86_64 --]
[-- Type: text/plain, Size: 4073 bytes --]

Set up a cpu area that allows the use of up 256MB for each processor.

Cpu memory use can grow rapidly. F.e. if we assume that a page struct
occupies 64 bytes of memory and we have 3 zones per node then we need
3 * 1k * 16k = 50 million pagesets or 3096 pagesets per processor.
This results in a total of 3.2 GB of page structs. So each cpu needs
around 200k of cpu storage for the page allocator alone.

For the UP and SMP case map the area using 4k ptes. Typical use of per cpu
data is around 16k for UP and SMP configurations. Allocating a 2M
segment would be overkill. There is enough reserve around to run
a few 100000 ip tunnels if one wants to and other lots of other frills.

For NUMA map the area using 2M PMDs. A large NUMA system may use
lots of cpu data for the page allocator data alone. We typically
have large amounts of memory around. Using a 2M page size reduces
TLB pressure.

Some numbers for envisioned maximum configurations of NUMA systems:

4k cpu configurations with 256 nodes:

	4096 * 256MB = 1TB of virtual space.

Maximum theoretical configuration 16384 processors 1k nodes:

	16384 * 256MB = 4TB of virtual space.

Both fit within the established limits established.

Signed-off-by: Christoph Lameter <clameter@sgi.com>

---
 arch/x86/Kconfig.x86_64      |   13 +++++++++++++
 arch/x86/mm/init_64.c        |   39 +++++++++++++++++++++++++++++++++++++++
 include/asm-x86/pgtable_64.h |    1 +
 3 files changed, 53 insertions(+)

Index: linux-2.6/arch/x86/mm/init_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init_64.c	2007-11-04 09:45:12.000000000 -0800
+++ linux-2.6/arch/x86/mm/init_64.c	2007-11-04 16:34:05.000000000 -0800
@@ -28,6 +28,7 @@
 #include <linux/module.h>
 #include <linux/memory_hotplug.h>
 #include <linux/nmi.h>
+#include <linux/cpu_alloc.h>
 
 #include <asm/processor.h>
 #include <asm/system.h>
@@ -781,3 +782,41 @@ int __meminit vmemmap_populate(struct pa
 	return 0;
 }
 #endif
+
+#ifdef CONFIG_NUMA
+int __meminit cpu_area_populate(void *start, unsigned long size,
+						gfp_t flags, int node)
+{
+	unsigned long addr = (unsigned long)start;
+	unsigned long end = addr + size;
+	unsigned long next;
+	pgd_t *pgd;
+	pud_t *pud;
+	pmd_t *pmd;
+
+	for (; addr < end; addr = next) {
+		next = pmd_addr_end(addr, end);
+
+		pgd = cpu_area_pgd_populate(addr, flags, node);
+		if (!pgd)
+			return -ENOMEM;
+		pud = cpu_area_pud_populate(pgd, addr, flags, node);
+		if (!pud)
+			return -ENOMEM;
+
+		pmd = pmd_offset(pud, addr);
+		if (pmd_none(*pmd)) {
+			pte_t entry;
+			void *p = cpu_area_alloc_block(PMD_SIZE, flags, node);
+			if (!p)
+				return -ENOMEM;
+
+			entry = pfn_pte(__pa(p) >> PAGE_SHIFT, PAGE_KERNEL);
+			mk_pte_huge(entry);
+			set_pmd(pmd, __pmd(pte_val(entry)));
+		}
+	}
+
+	return 0;
+}
+#endif
Index: linux-2.6/arch/x86/Kconfig.x86_64
===================================================================
--- linux-2.6.orig/arch/x86/Kconfig.x86_64	2007-11-04 09:45:12.000000000 -0800
+++ linux-2.6/arch/x86/Kconfig.x86_64	2007-11-04 16:16:06.000000000 -0800
@@ -137,6 +137,19 @@ config ARCH_HAS_ILOG2_U64
 	bool
 	default n
 
+config CPU_AREA_VIRTUAL
+	bool
+	default y
+
+config CPU_AREA_ORDER
+	int
+	default "16"
+
+config CPU_AREA_ALLOC_ORDER
+	int
+	default "9" if NUMA
+	default "0" if !NUMA
+
 source "init/Kconfig"
 
 
Index: linux-2.6/include/asm-x86/pgtable_64.h
===================================================================
--- linux-2.6.orig/include/asm-x86/pgtable_64.h	2007-11-04 09:45:12.000000000 -0800
+++ linux-2.6/include/asm-x86/pgtable_64.h	2007-11-04 16:16:06.000000000 -0800
@@ -138,6 +138,7 @@ static inline pte_t ptep_get_and_clear_f
 #define VMALLOC_START    _AC(0xffffc20000000000, UL)
 #define VMALLOC_END      _AC(0xffffe1ffffffffff, UL)
 #define VMEMMAP_START	 _AC(0xffffe20000000000, UL)
+#define CPU_AREA_BASE	 _AC(0xfffff20000000000, UL)
 #define MODULES_VADDR    _AC(0xffffffff88000000, UL)
 #define MODULES_END      _AC(0xfffffffffff00000, UL)
 #define MODULES_LEN   (MODULES_END - MODULES_VADDR)

-- 

  parent reply	other threads:[~2007-11-06 19:53 UTC|newest]

Thread overview: 78+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-11-06 19:51 [patch 00/28] cpu alloc v1: Optimize by removing arrays of pointers to per cpu objects Christoph Lameter
2007-11-06 19:51 ` [patch 01/28] cpu alloc: The allocator Christoph Lameter
2007-11-08 12:34   ` Peter Zijlstra
2007-11-08 12:37     ` Peter Zijlstra
2007-11-08 18:33     ` Christoph Lameter
2007-11-08 18:50       ` Christoph Lameter
2007-11-08 20:19         ` Peter Zijlstra
     [not found]   ` <1194522615.6289.136.camel@twins>
     [not found]     ` <Pine.LNX.4.64.0711081030380.7871@schroedinger.engr.sgi.com>
2007-11-08 20:19       ` Peter Zijlstra
2007-11-08 20:24         ` Christoph Lameter
2007-11-08 23:26           ` David Miller
2007-11-08 23:26         ` David Miller
2007-11-13 11:15   ` David Miller
2007-11-13 21:40     ` Christoph Lameter
2007-11-13 21:58       ` Eric Dumazet
2007-11-13 22:00         ` Christoph Lameter
2007-11-14  1:33           ` David Miller
2007-11-13 22:02         ` Christoph Lameter
2007-11-14  1:30       ` David Miller
2007-11-14  1:48         ` Christoph Lameter
2007-11-13 22:20     ` Christoph Lameter
2007-11-14  1:36       ` David Miller
2007-11-14  1:37       ` David Miller
2007-11-14  1:50         ` Christoph Lameter
2007-11-14  2:00           ` David Miller
2007-11-14  2:05             ` Christoph Lameter
2007-11-14  1:06     ` Andi Kleen
2007-11-14  1:52       ` David Miller
2007-11-14  1:57         ` Christoph Lameter
2007-11-14  2:01           ` David Miller
2007-11-14  2:03             ` Christoph Lameter
2007-11-14  2:28         ` Andi Kleen
2007-11-14  3:48           ` David Miller
2007-11-14  3:49           ` Christoph Lameter
2007-11-16 10:23           ` large lockdep bss (was: Re: [patch 01/28] cpu alloc: The allocator) Peter Zijlstra
2007-11-16 11:44             ` Andi Kleen
2007-11-14  4:15       ` [patch 01/28] cpu alloc: The allocator Christoph Lameter
2007-11-14  4:18         ` David Miller
2007-11-14  4:21           ` David Miller
2007-11-14  4:26           ` Christoph Lameter
2007-11-14  5:53             ` David Miller
2007-11-15 18:49               ` Christoph Lameter
2007-11-15 22:03                 ` David Miller
2007-11-16  2:19                   ` Christoph Lameter
2007-11-16  2:50                     ` David Miller
2007-11-16  2:55                       ` Christoph Lameter
2007-11-16  2:58                         ` David Miller
2007-11-16  3:10                           ` Christoph Lameter
2007-11-16  3:17                             ` David Miller
2007-11-16  3:19                               ` Christoph Lameter
2007-11-06 19:51 ` Christoph Lameter [this message]
2007-11-06 19:51 ` [patch 03/28] cpu alloc: IA64 support Christoph Lameter
2007-11-06 19:51 ` [patch 04/28] cpu alloc: i386 support Christoph Lameter
2007-11-06 19:51 ` [patch 05/28] cpu alloc: Use in SLUB Christoph Lameter
2007-11-06 19:51 ` [patch 06/28] cpu alloc: Remove SLUB fields Christoph Lameter
2007-11-06 19:51 ` [patch 07/28] cpu alloc: page allocator conversion Christoph Lameter
2007-11-06 19:51 ` [patch 08/28] cpu alloc: percpu_counter conversion Christoph Lameter
2007-11-06 19:51 ` [patch 09/28] cpu alloc: crash_notes conversion Christoph Lameter
2007-11-06 19:51 ` [patch 10/28] cpu alloc: workqueue conversion Christoph Lameter
2007-11-06 19:51 ` [patch 11/28] cpu alloc: ACPI cstate handling conversion Christoph Lameter
2007-11-06 19:51 ` [patch 12/28] cpu alloc: genhd statistics conversion Christoph Lameter
2007-11-06 19:51 ` [patch 13/28] cpu alloc: blktrace conversion Christoph Lameter
2007-11-06 19:51 ` [patch 14/28] cpu alloc: SRCU Christoph Lameter
2007-11-06 19:51 ` [patch 15/28] cpu alloc: XFS counters Christoph Lameter
2007-11-06 19:52 ` [patch 16/28] cpu alloc: NFS statistics Christoph Lameter
2007-11-06 19:52 ` [patch 17/28] cpu alloc: neigbour statistics Christoph Lameter
2007-11-06 19:52 ` [patch 18/28] cpu alloc: tcp statistics Christoph Lameter
2007-11-06 19:52 ` [patch 19/28] cpu alloc: convert scatches Christoph Lameter
2007-11-06 19:52 ` [patch 20/28] cpu alloc: dmaengine conversion Christoph Lameter
2007-11-06 19:52 ` [patch 21/28] cpu alloc: convert loopback statistics Christoph Lameter
2007-11-06 19:52 ` [patch 22/28] cpu alloc: veth conversion Christoph Lameter
2007-11-06 19:52 ` [patch 23/28] cpu alloc: Chelsio statistics conversion Christoph Lameter
2007-11-06 19:52 ` [patch 24/28] cpu alloc: convert mib handling to cpu alloc Christoph Lameter
2007-11-06 19:52 ` [patch 25/28] cpu alloc: Explicitly code allocpercpu calls in iucv Christoph Lameter
2007-11-06 19:52 ` [patch 26/28] cpu alloc: Use for infiniband Christoph Lameter
2007-11-06 19:52 ` [patch 27/28] cpu alloc: Use in the crypto subsystem Christoph Lameter
2007-11-06 19:52 ` [patch 28/28] cpu alloc: Remove the allocpercpu functionality Christoph Lameter
2007-11-07 13:10 ` [patch 00/28] cpu alloc v1: Optimize by removing arrays of pointers to per cpu objects Martin Schwidefsky
2007-11-07 18:05   ` Christoph Lameter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20071106195158.107939214@sgi.com \
    --to=clameter@sgi.com \
    --cc=akpm@linux-foundation.org \
    --cc=linux-mm@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox