x86: 64-bit, make sparsemem vmemmap the only memory model

Use sparsemem as the only memory model for UP, SMP and NUMA.  Measurements
indicate that DISCONTIGMEM has a higher overhead than sparsemem.  And
FLATMEMs benefits are minimal.  So I think its best to simply standardize
on sparsemem.

Results of page allocator tests (test can be had via git from slab git
tree branch tests)

Measurements in cycle counts. 1000 allocations were performed and then the
average cycle count was calculated.

Order	FlatMem	Discontig	SparseMem
0	  639	  665		  641
1	  567	  647		  593
2	  679	  774		  692
3	  763	  967		  781
4	  961	 1501		  962
5	 1356	 2344		 1392
6	 2224	 3982		 2336
7	 4869	 7225		 5074
8	12500	14048		12732
9	27926	28223		28165
10	58578	58714		58682

(Note that FlatMem is an SMP config and the rest NUMA configurations)

Memory use:

SMP Sparsemem
-------------

Kernel size:

   text    data     bss     dec     hex filename
3849268  397739 1264856 5511863  541ab7 vmlinux

             total       used       free     shared    buffers     cached
Mem:       8242252      41164    8201088          0        352      11512
-/+ buffers/cache:      29300    8212952
Swap:      9775512          0    9775512

SMP Flatmem
-----------

Kernel size:

   text    data     bss     dec     hex filename
3844612  397739 1264536 5506887  540747 vmlinux

So 4.5k growth in text size vs. FLATMEM.

             total       used       free     shared    buffers     cached
Mem:       8244052      40544    8203508          0        352      11484
-/+ buffers/cache:      28708    8215344

2k growth in overall memory use after boot.

NUMA discontig:

   text    data     bss     dec     hex filename
3888124  470659 1276504 5635287  55fcd7 vmlinux

             total       used       free     shared    buffers     cached
Mem:       8256256      56908    8199348          0        352      11496
-/+ buffers/cache:      45060    8211196
Swap:      9775512          0    9775512

NUMA sparse:

   text    data     bss     dec     hex filename
3896428  470659 1276824 5643911  561e87 vmlinux

8k text growth. Given that we fully inline virt_to_page and friends now
that is rather good.

             total       used       free     shared    buffers     cached
Mem:       8264720      57240    8207480          0        352      11516
-/+ buffers/cache:      45372    8219348
Swap:      9775512          0    9775512

The total available memory is increased by 8k.

This patch makes sparsemem the default and removes discontig and
flatmem support from x86.

[ akpm@linux-foundation.org: allnoconfig build fix ]

Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 2f4d88babd36bd83197fc0e3d6a15cc622aec93e..da98368f66afcac5a35a05bcaaf99ab21ac6df4f 100644 (file)
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -891,25 +891,29 @@ config HAVE_ARCH_ALLOC_REMAP
 
 config ARCH_FLATMEM_ENABLE
        def_bool y
-       depends on (X86_32 && ARCH_SELECT_MEMORY_MODEL && X86_PC) || (X86_64 && !NUMA)
+       depends on X86_32 && ARCH_SELECT_MEMORY_MODEL && X86_PC
 
 config ARCH_DISCONTIGMEM_ENABLE
        def_bool y
-       depends on NUMA
+       depends on NUMA && X86_32
 
 config ARCH_DISCONTIGMEM_DEFAULT
        def_bool y
-       depends on NUMA
+       depends on NUMA && X86_32
+
+config ARCH_SPARSEMEM_DEFAULT
+       def_bool y
+       depends on X86_64
 
 config ARCH_SPARSEMEM_ENABLE
        def_bool y
-       depends on NUMA || (EXPERIMENTAL && (X86_PC || X86_64))
+       depends on X86_64 || NUMA || (EXPERIMENTAL && X86_PC)
        select SPARSEMEM_STATIC if X86_32
        select SPARSEMEM_VMEMMAP_ENABLE if X86_64
 
 config ARCH_SELECT_MEMORY_MODEL
        def_bool y
-       depends on X86_32 && ARCH_SPARSEMEM_ENABLE
+       depends on ARCH_SPARSEMEM_ENABLE
 
 config ARCH_MEMORY_PROBE
        def_bool X86_64
@@ -1207,18 +1211,10 @@ config ARCH_ENABLE_MEMORY_HOTPLUG
        def_bool y
        depends on X86_64 || (X86_32 && HIGHMEM)
 
-config MEMORY_HOTPLUG_RESERVE
-       def_bool X86_64
-       depends on (MEMORY_HOTPLUG && DISCONTIGMEM)
-
 config HAVE_ARCH_EARLY_PFN_TO_NID
        def_bool X86_64
        depends on NUMA
 
-config OUT_OF_LINE_PFN_TO_PAGE
-       def_bool X86_64
-       depends on DISCONTIGMEM
-
 menu "Power management options"
        depends on !X86_VOYAGER
 

diff --git a/arch/x86/configs/x86_64_defconfig b/arch/x86/configs/x86_64_defconfig
index 38a83f9c966ff252be2218a53fae5d65c8b41d24..9e2b0ef851dee9ef62ed6494abcafb389921e10e 100644 (file)
--- a/arch/x86/configs/x86_64_defconfig
+++ b/arch/x86/configs/x86_64_defconfig
@@ -145,15 +145,6 @@ CONFIG_K8_NUMA=y
 CONFIG_NODES_SHIFT=6
 CONFIG_X86_64_ACPI_NUMA=y
 CONFIG_NUMA_EMU=y
-CONFIG_ARCH_DISCONTIGMEM_ENABLE=y
-CONFIG_ARCH_DISCONTIGMEM_DEFAULT=y
-CONFIG_ARCH_SPARSEMEM_ENABLE=y
-CONFIG_SELECT_MEMORY_MODEL=y
-# CONFIG_FLATMEM_MANUAL is not set
-CONFIG_DISCONTIGMEM_MANUAL=y
-# CONFIG_SPARSEMEM_MANUAL is not set
-CONFIG_DISCONTIGMEM=y
-CONFIG_FLAT_NODE_MEM_MAP=y
 CONFIG_NEED_MULTIPLE_NODES=y
 # CONFIG_SPARSEMEM_STATIC is not set
 CONFIG_SPLIT_PTLOCK_CPUS=4

diff --git a/arch/x86/kernel/machine_kexec_64.c b/arch/x86/kernel/machine_kexec_64.c
index aa3d2c8f7737851562923d37105ad7f719b8aa0b..a1fef42f8cdbccbd1664cc659c07781a2b82450a 100644 (file)
--- a/arch/x86/kernel/machine_kexec_64.c
+++ b/arch/x86/kernel/machine_kexec_64.c
@@ -234,10 +234,5 @@ NORET_TYPE void machine_kexec(struct kimage *image)
 void arch_crash_save_vmcoreinfo(void)
 {
        VMCOREINFO_SYMBOL(init_level4_pgt);
-
-#ifdef CONFIG_ARCH_DISCONTIGMEM_ENABLE
-       VMCOREINFO_SYMBOL(node_data);
-       VMCOREINFO_LENGTH(node_data, MAX_NUMNODES);
-#endif
 }
 

diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index 0fbb657a8b19171c931b2234c5fe0c2d94af2a5d..251eeb325ae3146f97f684d81f01b60b5f89d9f7 100644 (file)
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -486,34 +486,6 @@ EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
 
 #endif /* CONFIG_MEMORY_HOTPLUG */
 
-#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
-/*
- * Memory Hotadd without sparsemem. The mem_maps have been allocated in advance,
- * just online the pages.
- */
-int __add_pages(struct zone *z, unsigned long start_pfn, unsigned long nr_pages)
-{
-       int err = -EIO;
-       unsigned long pfn;
-       unsigned long total = 0, mem = 0;
-       for (pfn = start_pfn; pfn < start_pfn + nr_pages; pfn++) {
-               if (pfn_valid(pfn)) {
-                       online_page(pfn_to_page(pfn));
-                       err = 0;
-                       mem++;
-               }
-               total++;
-       }
-       if (!err) {
-               z->spanned_pages += total;
-               z->present_pages += mem;
-               z->zone_pgdat->node_spanned_pages += total;
-               z->zone_pgdat->node_present_pages += mem;
-       }
-       return err;
-}
-#endif
-
 static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, kcore_modules,
                         kcore_vsyscall;
 

diff --git a/arch/x86/mm/ioremap_64.c b/arch/x86/mm/ioremap_64.c
index 6cac90aa5032aac3900c0756e669871309a2b8f4..b03db4ca9cadfb919d5980276ff61c5390e45eff 100644 (file)
--- a/arch/x86/mm/ioremap_64.c
+++ b/arch/x86/mm/ioremap_64.c
@@ -86,23 +86,6 @@ void __iomem * __ioremap(unsigned long phys_addr, unsigned long size, unsigned l
        if (phys_addr >= ISA_START_ADDRESS && last_addr < ISA_END_ADDRESS)
                return (__force void __iomem *)phys_to_virt(phys_addr);
 
-#ifdef CONFIG_FLATMEM
-       /*
-        * Don't allow anybody to remap normal RAM that we're using..
-        */
-       if (last_addr < virt_to_phys(high_memory)) {
-               char *t_addr, *t_end;
-               struct page *page;
-
-               t_addr = __va(phys_addr);
-               t_end = t_addr + (size - 1);
-          
-               for(page = virt_to_page(t_addr); page <= virt_to_page(t_end); page++)
-                       if(!PageReserved(page))
-                               return NULL;
-       }
-#endif
-
        pgprot = __pgprot(_PAGE_PRESENT | _PAGE_RW | _PAGE_GLOBAL
                          | _PAGE_DIRTY | _PAGE_ACCESSED | flags);
        /*

diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index 29b69300aee2749f8387c4e5d468850b12ea4237..46b4b5e1a02a88afa29046014eb239e18307257e 100644 (file)
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -153,12 +153,10 @@ int __init compute_hash_shift(struct bootnode *nodes, int numnodes)
        return shift;
 }
 
-#ifdef CONFIG_SPARSEMEM
 int early_pfn_to_nid(unsigned long pfn)
 {
        return phys_to_nid(pfn << PAGE_SHIFT);
 }
-#endif
 
 static void * __init early_node_mem(int nodeid, unsigned long start,
                                    unsigned long end, unsigned long size)
@@ -635,23 +633,4 @@ void __init init_cpu_to_node(void)
        }
 }
 
-#ifdef CONFIG_DISCONTIGMEM
-/*
- * Functions to convert PFNs from/to per node page addresses.
- * These are out of line because they are quite big.
- * They could be all tuned by pre caching more state.
- * Should do that.
- */
 
-int pfn_valid(unsigned long pfn)
-{
-       unsigned nid;
-       if (pfn >= num_physpages)
-               return 0;
-       nid = pfn_to_nid(pfn);
-       if (nid == 0xff)
-               return 0;
-       return pfn >= node_start_pfn(nid) && (pfn) < node_end_pfn(nid);
-}
-EXPORT_SYMBOL(pfn_valid);
-#endif

diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c
index ea85172fc0cc6c33fd20ec49981411ea9b73d031..9be14171144b389791bc987f536ecda6744c787e 100644 (file)
--- a/arch/x86/mm/srat_64.c
+++ b/arch/x86/mm/srat_64.c
@@ -151,62 +151,6 @@ acpi_numa_processor_affinity_init(struct acpi_srat_cpu_affinity *pa)
               pxm, pa->apic_id, node);
 }
 
-#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
-/*
- * Protect against too large hotadd areas that would fill up memory.
- */
-static int hotadd_enough_memory(struct bootnode *nd)
-{
-       static unsigned long allocated;
-       static unsigned long last_area_end;
-       unsigned long pages = (nd->end - nd->start) >> PAGE_SHIFT;
-       long mem = pages * sizeof(struct page);
-       unsigned long addr;
-       unsigned long allowed;
-       unsigned long oldpages = pages;
-
-       if (mem < 0)
-               return 0;
-       allowed = (end_pfn - absent_pages_in_range(0, end_pfn)) * PAGE_SIZE;
-       allowed = (allowed / 100) * hotadd_percent;
-       if (allocated + mem > allowed) {
-               unsigned long range;
-               /* Give them at least part of their hotadd memory upto hotadd_percent
-                  It would be better to spread the limit out
-                  over multiple hotplug areas, but that is too complicated
-                  right now */
-               if (allocated >= allowed)
-                       return 0;
-               range = allowed - allocated;
-               pages = (range / PAGE_SIZE);
-               mem = pages * sizeof(struct page);
-               nd->end = nd->start + range;
-       }
-       /* Not completely fool proof, but a good sanity check */
-       addr = find_e820_area(last_area_end, end_pfn<<PAGE_SHIFT, mem);
-       if (addr == -1UL)
-               return 0;
-       if (pages != oldpages)
-               printk(KERN_NOTICE "SRAT: Hotadd area limited to %lu bytes\n",
-                       pages << PAGE_SHIFT);
-       last_area_end = addr + mem;
-       allocated += mem;
-       return 1;
-}
-
-static int update_end_of_memory(unsigned long end)
-{
-       found_add_area = 1;
-       if ((end >> PAGE_SHIFT) > end_pfn)
-               end_pfn = end >> PAGE_SHIFT;
-       return 1;
-}
-
-static inline int save_add_info(void)
-{
-       return hotadd_percent > 0;
-}
-#else
 int update_end_of_memory(unsigned long end) {return -1;}
 static int hotadd_enough_memory(struct bootnode *nd) {return 1;}
 #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
@@ -214,7 +158,6 @@ static inline int save_add_info(void) {return 1;}
 #else
 static inline int save_add_info(void) {return 0;}
 #endif
-#endif
 /*
  * Update nodes_add and decide if to include add are in the zone.
  * Both SPARSE and RESERVE need nodes_add infomation.

diff --git a/include/asm-x86/mmzone_64.h b/include/asm-x86/mmzone_64.h
index 1e0ed34a6adc450ca7d674fa73ec2114ee86fabf..b0c25ae111d9549508fdfb513480594ef220ca2e 100644 (file)
--- a/include/asm-x86/mmzone_64.h
+++ b/include/asm-x86/mmzone_64.h
@@ -43,12 +43,6 @@ static inline __attribute__((pure)) int phys_to_nid(unsigned long addr)
 
 extern int early_pfn_to_nid(unsigned long pfn);
 
-#ifdef CONFIG_DISCONTIGMEM
-#define pfn_to_nid(pfn) phys_to_nid((unsigned long)(pfn) << PAGE_SHIFT)
-
-extern int pfn_valid(unsigned long pfn);
-#endif
-
 #ifdef CONFIG_NUMA_EMU
 #define FAKE_NODE_MIN_SIZE     (64*1024*1024)
 #define FAKE_NODE_MIN_HASH_MASK        (~(FAKE_NODE_MIN_SIZE - 1uL))

diff --git a/include/asm-x86/page_64.h b/include/asm-x86/page_64.h
index 6fdc904a5fa5248b4437ddf6c32b6123e22f0308..d400167c5509b68863e5540af945ff5ac9588bed 100644 (file)
--- a/include/asm-x86/page_64.h
+++ b/include/asm-x86/page_64.h
@@ -122,9 +122,6 @@ extern unsigned long __phys_addr(unsigned long);
 #define __va(x)                        ((void *)((unsigned long)(x)+PAGE_OFFSET))
 #define __boot_va(x)           __va(x)
 #define __boot_pa(x)           __pa(x)
-#ifdef CONFIG_FLATMEM
-#define pfn_valid(pfn)         ((pfn) < end_pfn)
-#endif
 
 #define virt_to_page(kaddr)    pfn_to_page(__pa(kaddr) >> PAGE_SHIFT)
 #define virt_addr_valid(kaddr) pfn_valid(__pa(kaddr) >> PAGE_SHIFT)