Merge branch 'tj-percpu' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/misc into core/percpu

Conflicts:
	arch/x86/include/asm/pgtable.h

11 files changed, 458 insertions, 107 deletions

diff --git a/arch/alpha/mm/init.c b/arch/alpha/mm/init.c
index 5d7a16eab312..91eddd8505df 100644
--- a/arch/alpha/mm/init.c
+++ b/arch/alpha/mm/init.c
@@ -189,9 +189,21 @@ callback_init(void * kernel_end)
 
         if (alpha_using_srm) {
                 static struct vm_struct console_remap_vm;
-                unsigned long vaddr = VMALLOC_START;
+                unsigned long nr_pages = 0;
+                unsigned long vaddr;
                 unsigned long i, j;
 
+                /* calculate needed size */
+                for (i = 0; i < crb->map_entries; ++i)
+                        nr_pages += crb->map[i].count;
+
+                /* register the vm area */
+                console_remap_vm.flags = VM_ALLOC;
+                console_remap_vm.size = nr_pages << PAGE_SHIFT;
+                vm_area_register_early(&console_remap_vm, PAGE_SIZE);
+
+                vaddr = (unsigned long)consle_remap_vm.addr;
+
                 /* Set up the third level PTEs and update the virtual
                    addresses of the CRB entries.  */
                 for (i = 0; i < crb->map_entries; ++i) {
@@ -213,12 +225,6 @@ callback_init(void * kernel_end)
                                 vaddr += PAGE_SIZE;
                         }
                 }
-
-                /* Let vmalloc know that we've allocated some space.  */
-                console_remap_vm.flags = VM_ALLOC;
-                console_remap_vm.addr = (void *) VMALLOC_START;
-                console_remap_vm.size = vaddr - VMALLOC_START;
-                vmlist = &console_remap_vm;
         }
 
         callback_init_done = 1;
diff --git a/arch/avr32/Kconfig b/arch/avr32/Kconfig
index b189680d18b0..05fe3053dcae 100644
--- a/arch/avr32/Kconfig
+++ b/arch/avr32/Kconfig
@@ -181,7 +181,7 @@ source "kernel/Kconfig.preempt"
 config QUICKLIST
         def_bool y
 
-config HAVE_ARCH_BOOTMEM_NODE
+config HAVE_ARCH_BOOTMEM
         def_bool n
 
 config ARCH_HAVE_MEMORY_PRESENT
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 5e2919c0ff92..8015641478bd 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -135,6 +135,9 @@ config ARCH_HAS_CACHE_LINE_SIZE
 config HAVE_SETUP_PER_CPU_AREA
         def_bool y
 
+config HAVE_DYNAMIC_PER_CPU_AREA
+        def_bool y
+
 config HAVE_CPUMASK_OF_CPU_MAP
         def_bool X86_64_SMP
 
@@ -1122,7 +1125,7 @@ config NODES_SHIFT
           Specify the maximum number of NUMA Nodes available on the target
           system.  Increases memory reserved to accomodate various tables.
 
-config HAVE_ARCH_BOOTMEM_NODE
+config HAVE_ARCH_BOOTMEM
         def_bool y
         depends on X86_32 && NUMA
 
diff --git a/arch/x86/include/asm/mmzone_32.h b/arch/x86/include/asm/mmzone_32.h
index 105fb90a0635..eeacf67de49e 100644
--- a/arch/x86/include/asm/mmzone_32.h
+++ b/arch/x86/include/asm/mmzone_32.h
@@ -91,45 +91,12 @@ static inline int pfn_valid(int pfn)
 #endif /* CONFIG_DISCONTIGMEM */
 
 #ifdef CONFIG_NEED_MULTIPLE_NODES
-
-/*
- * Following are macros that are specific to this numa platform.
- */
-#define reserve_bootmem(addr, size, flags) \
-        reserve_bootmem_node(NODE_DATA(0), (addr), (size), (flags))
-#define alloc_bootmem(x) \
-        __alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS))
-#define alloc_bootmem_nopanic(x) \
-        __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), SMP_CACHE_BYTES, \
-                                __pa(MAX_DMA_ADDRESS))
-#define alloc_bootmem_low(x) \
-        __alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES, 0)
-#define alloc_bootmem_pages(x) \
-        __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, __pa(MAX_DMA_ADDRESS))
-#define alloc_bootmem_pages_nopanic(x) \
-        __alloc_bootmem_node_nopanic(NODE_DATA(0), (x), PAGE_SIZE, \
-                                __pa(MAX_DMA_ADDRESS))
-#define alloc_bootmem_low_pages(x) \
-        __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0)
-#define alloc_bootmem_node(pgdat, x)                                    \
-({                                                                      \
-        struct pglist_data  __maybe_unused                      \
-                                *__alloc_bootmem_node__pgdat = (pgdat); \
-        __alloc_bootmem_node(NODE_DATA(0), (x), SMP_CACHE_BYTES,        \
-                                                __pa(MAX_DMA_ADDRESS)); \
-})
-#define alloc_bootmem_pages_node(pgdat, x)                              \
-({                                                                      \
-        struct pglist_data  __maybe_unused                      \
-                                *__alloc_bootmem_node__pgdat = (pgdat); \
-        __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE,              \
-                                                __pa(MAX_DMA_ADDRESS)); \
-})
-#define alloc_bootmem_low_pages_node(pgdat, x)                          \
+/* always use node 0 for bootmem on this numa platform */
+#define alloc_bootmem_core(__bdata, size, align, goal, limit)           \
 ({                                                                      \
-        struct pglist_data  __maybe_unused                      \
-                                *__alloc_bootmem_node__pgdat = (pgdat); \
-        __alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0);          \
+        bootmem_data_t __maybe_unused * __abm_bdata_dummy = (__bdata);  \
+        __alloc_bootmem_core(NODE_DATA(0)->bdata,                       \
+                             (size), (align), (goal), (limit));         \
 })
 #endif /* CONFIG_NEED_MULTIPLE_NODES */
 
diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index aee103b26d01..8f1d2fbec1d4 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -43,6 +43,14 @@
 #else /* ...!ASSEMBLY */
 
 #include <linux/stringify.h>
+#include <asm/sections.h>
+
+#define __addr_to_pcpu_ptr(addr)                                        \
+        (void *)((unsigned long)(addr) - (unsigned long)pcpu_base_addr  \
+                 + (unsigned long)__per_cpu_start)
+#define __pcpu_ptr_to_addr(ptr)                                         \
+        (void *)((unsigned long)(ptr) + (unsigned long)pcpu_base_addr   \
+                 - (unsigned long)__per_cpu_start)
 
 #ifdef CONFIG_SMP
 #define __percpu_arg(x)         "%%"__stringify(__percpu_seg)":%P" #x
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 1c097a3a6669..d0812e155f1d 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -288,6 +288,8 @@ static inline int is_new_memtype_allowed(unsigned long flags,
         return 1;
 }
 
+pmd_t *populate_extra_pmd(unsigned long vaddr);
+pte_t *populate_extra_pte(unsigned long vaddr);
 #endif  /* __ASSEMBLY__ */
 
 #ifdef CONFIG_X86_32
diff --git a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
index 4b1c319d30c3..22590cf688ae 100644
--- a/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
+++ b/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
@@ -601,7 +601,7 @@ static int acpi_cpufreq_cpu_init(struct cpufreq_policy *policy)
         if (!data)
                 return -ENOMEM;
 
-        data->acpi_data = percpu_ptr(acpi_perf_data, cpu);
+        data->acpi_data = per_cpu_ptr(acpi_perf_data, cpu);
         per_cpu(drv_data, cpu) = data;
 
         if (cpu_has(c, X86_FEATURE_CONSTANT_TSC))
diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c
index 9dc6b2b24275..3b09634a5153 100644
--- a/arch/x86/kernel/irq_32.c
+++ b/arch/x86/kernel/irq_32.c
@@ -16,6 +16,7 @@
 #include <linux/cpu.h>
 #include <linux/delay.h>
 #include <linux/uaccess.h>
+#include <linux/percpu.h>
 
 #include <asm/apic.h>
 
@@ -55,13 +56,13 @@ static inline void print_stack_overflow(void) { }
 union irq_ctx {
         struct thread_info      tinfo;
         u32                     stack[THREAD_SIZE/sizeof(u32)];
-};
+} __attribute__((aligned(PAGE_SIZE)));
 
-static union irq_ctx *hardirq_ctx[NR_CPUS] __read_mostly;
-static union irq_ctx *softirq_ctx[NR_CPUS] __read_mostly;
+static DEFINE_PER_CPU(union irq_ctx *, hardirq_ctx);
+static DEFINE_PER_CPU(union irq_ctx *, softirq_ctx);
 
-static char softirq_stack[NR_CPUS * THREAD_SIZE] __page_aligned_bss;
-static char hardirq_stack[NR_CPUS * THREAD_SIZE] __page_aligned_bss;
+static DEFINE_PER_CPU_PAGE_ALIGNED(union irq_ctx, hardirq_stack);
+static DEFINE_PER_CPU_PAGE_ALIGNED(union irq_ctx, softirq_stack);
 
 static void call_on_stack(void *func, void *stack)
 {
@@ -81,7 +82,7 @@ execute_on_irq_stack(int overflow, struct irq_desc *desc, int irq)
         u32 *isp, arg1, arg2;
 
         curctx = (union irq_ctx *) current_thread_info();
-        irqctx = hardirq_ctx[smp_processor_id()];
+        irqctx = __get_cpu_var(hardirq_ctx);
 
         /*
          * this is where we switch to the IRQ stack. However, if we are
@@ -125,34 +126,34 @@ void __cpuinit irq_ctx_init(int cpu)
 {
         union irq_ctx *irqctx;
 
-        if (hardirq_ctx[cpu])
+        if (per_cpu(hardirq_ctx, cpu))
                 return;
 
-        irqctx = (union irq_ctx*) &hardirq_stack[cpu*THREAD_SIZE];
+        irqctx = &per_cpu(hardirq_stack, cpu);
         irqctx->tinfo.task              = NULL;
         irqctx->tinfo.exec_domain       = NULL;
         irqctx->tinfo.cpu               = cpu;
         irqctx->tinfo.preempt_count     = HARDIRQ_OFFSET;
         irqctx->tinfo.addr_limit        = MAKE_MM_SEG(0);
 
-        hardirq_ctx[cpu] = irqctx;
+        per_cpu(hardirq_ctx, cpu) = irqctx;
 
-        irqctx = (union irq_ctx *) &softirq_stack[cpu*THREAD_SIZE];
+        irqctx = &per_cpu(softirq_stack, cpu);
         irqctx->tinfo.task              = NULL;
         irqctx->tinfo.exec_domain       = NULL;
         irqctx->tinfo.cpu               = cpu;
         irqctx->tinfo.preempt_count     = 0;
         irqctx->tinfo.addr_limit        = MAKE_MM_SEG(0);
 
-        softirq_ctx[cpu] = irqctx;
+        per_cpu(softirq_ctx, cpu) = irqctx;
 
         printk(KERN_DEBUG "CPU %u irqstacks, hard=%p soft=%p\n",
-               cpu, hardirq_ctx[cpu], softirq_ctx[cpu]);
+               cpu, per_cpu(hardirq_ctx, cpu),  per_cpu(softirq_ctx, cpu));
 }
 
 void irq_ctx_exit(int cpu)
 {
-        hardirq_ctx[cpu] = NULL;
+        per_cpu(hardirq_ctx, cpu) = NULL;
 }
 
 asmlinkage void do_softirq(void)
@@ -169,7 +170,7 @@ asmlinkage void do_softirq(void)
 
         if (local_softirq_pending()) {
                 curctx = current_thread_info();
-                irqctx = softirq_ctx[smp_processor_id()];
+                irqctx = __get_cpu_var(softirq_ctx);
                 irqctx->tinfo.task = curctx->task;
                 irqctx->tinfo.previous_esp = current_stack_pointer;
 
diff --git a/arch/x86/kernel/setup_percpu.c b/arch/x86/kernel/setup_percpu.c
index d992e6cff730..2d946a8f78b9 100644
--- a/arch/x86/kernel/setup_percpu.c
+++ b/arch/x86/kernel/setup_percpu.c
@@ -7,6 +7,7 @@
 #include <linux/crash_dump.h>
 #include <linux/smp.h>
 #include <linux/topology.h>
+#include <linux/pfn.h>
 #include <asm/sections.h>
 #include <asm/processor.h>
 #include <asm/setup.h>
@@ -41,6 +42,321 @@ unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = {
 };
 EXPORT_SYMBOL(__per_cpu_offset);
 
+/**
+ * pcpu_need_numa - determine percpu allocation needs to consider NUMA
+ *
+ * If NUMA is not configured or there is only one NUMA node available,
+ * there is no reason to consider NUMA.  This function determines
+ * whether percpu allocation should consider NUMA or not.
+ *
+ * RETURNS:
+ * true if NUMA should be considered; otherwise, false.
+ */
+static bool __init pcpu_need_numa(void)
+{
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+        pg_data_t *last = NULL;
+        unsigned int cpu;
+
+        for_each_possible_cpu(cpu) {
+                int node = early_cpu_to_node(cpu);
+
+                if (node_online(node) && NODE_DATA(node) &&
+                    last && last != NODE_DATA(node))
+                        return true;
+
+                last = NODE_DATA(node);
+        }
+#endif
+        return false;
+}
+
+/**
+ * pcpu_alloc_bootmem - NUMA friendly alloc_bootmem wrapper for percpu
+ * @cpu: cpu to allocate for
+ * @size: size allocation in bytes
+ * @align: alignment
+ *
+ * Allocate @size bytes aligned at @align for cpu @cpu.  This wrapper
+ * does the right thing for NUMA regardless of the current
+ * configuration.
+ *
+ * RETURNS:
+ * Pointer to the allocated area on success, NULL on failure.
+ */
+static void * __init pcpu_alloc_bootmem(unsigned int cpu, unsigned long size,
+                                        unsigned long align)
+{
+        const unsigned long goal = __pa(MAX_DMA_ADDRESS);
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+        int node = early_cpu_to_node(cpu);
+        void *ptr;
+
+        if (!node_online(node) || !NODE_DATA(node)) {
+                ptr = __alloc_bootmem_nopanic(size, align, goal);
+                pr_info("cpu %d has no node %d or node-local memory\n",
+                        cpu, node);
+                pr_debug("per cpu data for cpu%d %lu bytes at %016lx\n",
+                         cpu, size, __pa(ptr));
+        } else {
+                ptr = __alloc_bootmem_node_nopanic(NODE_DATA(node),
+                                                   size, align, goal);
+                pr_debug("per cpu data for cpu%d %lu bytes on node%d at "
+                         "%016lx\n", cpu, size, node, __pa(ptr));
+        }
+        return ptr;
+#else
+        return __alloc_bootmem_nopanic(size, align, goal);
+#endif
+}
+
+/*
+ * Remap allocator
+ *
+ * This allocator uses PMD page as unit.  A PMD page is allocated for
+ * each cpu and each is remapped into vmalloc area using PMD mapping.
+ * As PMD page is quite large, only part of it is used for the first
+ * chunk.  Unused part is returned to the bootmem allocator.
+ *
+ * So, the PMD pages are mapped twice - once to the physical mapping
+ * and to the vmalloc area for the first percpu chunk.  The double
+ * mapping does add one more PMD TLB entry pressure but still is much
+ * better than only using 4k mappings while still being NUMA friendly.
+ */
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+static size_t pcpur_size __initdata;
+static void **pcpur_ptrs __initdata;
+
+static struct page * __init pcpur_get_page(unsigned int cpu, int pageno)
+{
+        size_t off = (size_t)pageno << PAGE_SHIFT;
+
+        if (off >= pcpur_size)
+                return NULL;
+
+        return virt_to_page(pcpur_ptrs[cpu] + off);
+}
+
+static ssize_t __init setup_pcpu_remap(size_t static_size)
+{
+        static struct vm_struct vm;
+        pg_data_t *last;
+        size_t ptrs_size;
+        unsigned int cpu;
+        ssize_t ret;
+
+        /*
+         * If large page isn't supported, there's no benefit in doing
+         * this.  Also, on non-NUMA, embedding is better.
+         */
+        if (!cpu_has_pse || pcpu_need_numa())
+                return -EINVAL;
+
+        last = NULL;
+        for_each_possible_cpu(cpu) {
+                int node = early_cpu_to_node(cpu);
+
+                if (node_online(node) && NODE_DATA(node) &&
+                    last && last != NODE_DATA(node))
+                        goto proceed;
+
+                last = NODE_DATA(node);
+        }
+        return -EINVAL;
+
+proceed:
+        /*
+         * Currently supports only single page.  Supporting multiple
+         * pages won't be too difficult if it ever becomes necessary.
+         */
+        pcpur_size = PFN_ALIGN(static_size + PERCPU_DYNAMIC_RESERVE);
+        if (pcpur_size > PMD_SIZE) {
+                pr_warning("PERCPU: static data is larger than large page, "
+                           "can't use large page\n");
+                return -EINVAL;
+        }
+
+        /* allocate pointer array and alloc large pages */
+        ptrs_size = PFN_ALIGN(num_possible_cpus() * sizeof(pcpur_ptrs[0]));
+        pcpur_ptrs = alloc_bootmem(ptrs_size);
+
+        for_each_possible_cpu(cpu) {
+                pcpur_ptrs[cpu] = pcpu_alloc_bootmem(cpu, PMD_SIZE, PMD_SIZE);
+                if (!pcpur_ptrs[cpu])
+                        goto enomem;
+
+                /*
+                 * Only use pcpur_size bytes and give back the rest.
+                 *
+                 * Ingo: The 2MB up-rounding bootmem is needed to make
+                 * sure the partial 2MB page is still fully RAM - it's
+                 * not well-specified to have a PAT-incompatible area
+                 * (unmapped RAM, device memory, etc.) in that hole.
+                 */
+                free_bootmem(__pa(pcpur_ptrs[cpu] + pcpur_size),
+                             PMD_SIZE - pcpur_size);
+
+                memcpy(pcpur_ptrs[cpu], __per_cpu_load, static_size);
+        }
+
+        /* allocate address and map */
+        vm.flags = VM_ALLOC;
+        vm.size = num_possible_cpus() * PMD_SIZE;
+        vm_area_register_early(&vm, PMD_SIZE);
+
+        for_each_possible_cpu(cpu) {
+                pmd_t *pmd;
+
+                pmd = populate_extra_pmd((unsigned long)vm.addr
+                                         + cpu * PMD_SIZE);
+                set_pmd(pmd, pfn_pmd(page_to_pfn(virt_to_page(pcpur_ptrs[cpu])),
+                                     PAGE_KERNEL_LARGE));
+        }
+
+        /* we're ready, commit */
+        pr_info("PERCPU: Remapped at %p with large pages, static data "
+                "%zu bytes\n", vm.addr, static_size);
+
+        ret = pcpu_setup_first_chunk(pcpur_get_page, static_size, PMD_SIZE,
+                                     pcpur_size - static_size, vm.addr, NULL);
+        goto out_free_ar;
+
+enomem:
+        for_each_possible_cpu(cpu)
+                if (pcpur_ptrs[cpu])
+                        free_bootmem(__pa(pcpur_ptrs[cpu]), PMD_SIZE);
+        ret = -ENOMEM;
+out_free_ar:
+        free_bootmem(__pa(pcpur_ptrs), ptrs_size);
+        return ret;
+}
+#else
+static ssize_t __init setup_pcpu_remap(size_t static_size)
+{
+        return -EINVAL;
+}
+#endif
+
+/*
+ * Embedding allocator
+ *
+ * The first chunk is sized to just contain the static area plus
+ * PERCPU_DYNAMIC_RESERVE and allocated as a contiguous area using
+ * bootmem allocator and used as-is without being mapped into vmalloc
+ * area.  This enables the first chunk to piggy back on the linear
+ * physical PMD mapping and doesn't add any additional pressure to
+ * TLB.
+ */
+static void *pcpue_ptr __initdata;
+static size_t pcpue_unit_size __initdata;
+
+static struct page * __init pcpue_get_page(unsigned int cpu, int pageno)
+{
+        return virt_to_page(pcpue_ptr + cpu * pcpue_unit_size
+                            + ((size_t)pageno << PAGE_SHIFT));
+}
+
+static ssize_t __init setup_pcpu_embed(size_t static_size)
+{
+        unsigned int cpu;
+
+        /*
+         * If large page isn't supported, there's no benefit in doing
+         * this.  Also, embedding allocation doesn't play well with
+         * NUMA.
+         */
+        if (!cpu_has_pse || pcpu_need_numa())
+                return -EINVAL;
+
+        /* allocate and copy */
+        pcpue_unit_size = PFN_ALIGN(static_size + PERCPU_DYNAMIC_RESERVE);
+        pcpue_unit_size = max(pcpue_unit_size, PCPU_MIN_UNIT_SIZE);
+        pcpue_ptr = pcpu_alloc_bootmem(0, num_possible_cpus() * pcpue_unit_size,
+                                       PAGE_SIZE);
+        if (!pcpue_ptr)
+                return -ENOMEM;
+
+        for_each_possible_cpu(cpu)
+                memcpy(pcpue_ptr + cpu * pcpue_unit_size, __per_cpu_load,
+                       static_size);
+
+        /* we're ready, commit */
+        pr_info("PERCPU: Embedded %zu pages at %p, static data %zu bytes\n",
+                pcpue_unit_size >> PAGE_SHIFT, pcpue_ptr, static_size);
+
+        return pcpu_setup_first_chunk(pcpue_get_page, static_size,
+                                      pcpue_unit_size,
+                                      pcpue_unit_size - static_size, pcpue_ptr,
+                                      NULL);
+}
+
+/*
+ * 4k page allocator
+ *
+ * This is the basic allocator.  Static percpu area is allocated
+ * page-by-page and most of initialization is done by the generic
+ * setup function.
+ */
+static struct page **pcpu4k_pages __initdata;
+static int pcpu4k_nr_static_pages __initdata;
+
+static struct page * __init pcpu4k_get_page(unsigned int cpu, int pageno)
+{
+        if (pageno < pcpu4k_nr_static_pages)
+                return pcpu4k_pages[cpu * pcpu4k_nr_static_pages + pageno];
+        return NULL;
+}
+
+static void __init pcpu4k_populate_pte(unsigned long addr)
+{
+        populate_extra_pte(addr);
+}
+
+static ssize_t __init setup_pcpu_4k(size_t static_size)
+{
+        size_t pages_size;
+        unsigned int cpu;
+        int i, j;
+        ssize_t ret;
+
+        pcpu4k_nr_static_pages = PFN_UP(static_size);
+
+        /* unaligned allocations can't be freed, round up to page size */
+        pages_size = PFN_ALIGN(pcpu4k_nr_static_pages * num_possible_cpus()
+                               * sizeof(pcpu4k_pages[0]));
+        pcpu4k_pages = alloc_bootmem(pages_size);
+
+        /* allocate and copy */
+        j = 0;
+        for_each_possible_cpu(cpu)
+                for (i = 0; i < pcpu4k_nr_static_pages; i++) {
+                        void *ptr;
+
+                        ptr = pcpu_alloc_bootmem(cpu, PAGE_SIZE, PAGE_SIZE);
+                        if (!ptr)
+                                goto enomem;
+
+                        memcpy(ptr, __per_cpu_load + i * PAGE_SIZE, PAGE_SIZE);
+                        pcpu4k_pages[j++] = virt_to_page(ptr);
+                }
+
+        /* we're ready, commit */
+        pr_info("PERCPU: Allocated %d 4k pages, static data %zu bytes\n",
+                pcpu4k_nr_static_pages, static_size);
+
+        ret = pcpu_setup_first_chunk(pcpu4k_get_page, static_size, 0, 0, NULL,
+                                     pcpu4k_populate_pte);
+        goto out_free_ar;
+
+enomem:
+        while (--j >= 0)
+                free_bootmem(__pa(page_address(pcpu4k_pages[j])), PAGE_SIZE);
+        ret = -ENOMEM;
+out_free_ar:
+        free_bootmem(__pa(pcpu4k_pages), pages_size);
+        return ret;
+}
+
 static inline void setup_percpu_segment(int cpu)
 {
 #ifdef CONFIG_X86_32
@@ -61,38 +377,35 @@ static inline void setup_percpu_segment(int cpu)
  */
 void __init setup_per_cpu_areas(void)
 {
-        ssize_t size;
-        char *ptr;
-        int cpu;
-
-        /* Copy section for each CPU (we discard the original) */
-        size = roundup(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
+        size_t static_size = __per_cpu_end - __per_cpu_start;
+        unsigned int cpu;
+        unsigned long delta;
+        size_t pcpu_unit_size;
+        ssize_t ret;
 
         pr_info("NR_CPUS:%d nr_cpumask_bits:%d nr_cpu_ids:%d nr_node_ids:%d\n",
                 NR_CPUS, nr_cpumask_bits, nr_cpu_ids, nr_node_ids);
 
-        pr_info("PERCPU: Allocating %zd bytes of per cpu data\n", size);
+        /*
+         * Allocate percpu area.  If PSE is supported, try to make use
+         * of large page mappings.  Please read comments on top of
+         * each allocator for details.
+         */
+        ret = setup_pcpu_remap(static_size);
+        if (ret < 0)
+                ret = setup_pcpu_embed(static_size);
+        if (ret < 0)
+                ret = setup_pcpu_4k(static_size);
+        if (ret < 0)
+                panic("cannot allocate static percpu area (%zu bytes, err=%zd)",
+                      static_size, ret);
 
-        for_each_possible_cpu(cpu) {
-#ifndef CONFIG_NEED_MULTIPLE_NODES
-                ptr = alloc_bootmem_pages(size);
-#else
-                int node = early_cpu_to_node(cpu);
-                if (!node_online(node) || !NODE_DATA(node)) {
-                        ptr = alloc_bootmem_pages(size);
-                        pr_info("cpu %d has no node %d or node-local memory\n",
-                                cpu, node);
-                        pr_debug("per cpu data for cpu%d at %016lx\n",
-                                 cpu, __pa(ptr));
-                } else {
-                        ptr = alloc_bootmem_pages_node(NODE_DATA(node), size);
-                        pr_debug("per cpu data for cpu%d on node%d at %016lx\n",
-                                cpu, node, __pa(ptr));
-                }
-#endif
+        pcpu_unit_size = ret;
 
-                memcpy(ptr, __per_cpu_load, __per_cpu_end - __per_cpu_start);
-                per_cpu_offset(cpu) = ptr - __per_cpu_start;
+        /* alrighty, percpu areas up and running */
+        delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
+        for_each_possible_cpu(cpu) {
+                per_cpu_offset(cpu) = delta + cpu * pcpu_unit_size;
                 per_cpu(this_cpu_off, cpu) = per_cpu_offset(cpu);
                 per_cpu(cpu_number, cpu) = cpu;
                 setup_percpu_segment(cpu);
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index 06708ee94aa4..ef0bb941cdf5 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -137,6 +137,23 @@ static pte_t * __init one_page_table_init(pmd_t *pmd)
         return pte_offset_kernel(pmd, 0);
 }
 
+pmd_t * __init populate_extra_pmd(unsigned long vaddr)
+{
+        int pgd_idx = pgd_index(vaddr);
+        int pmd_idx = pmd_index(vaddr);
+
+        return one_md_table_init(swapper_pg_dir + pgd_idx) + pmd_idx;
+}
+
+pte_t * __init populate_extra_pte(unsigned long vaddr)
+{
+        int pte_idx = pte_index(vaddr);
+        pmd_t *pmd;
+
+        pmd = populate_extra_pmd(vaddr);
+        return one_page_table_init(pmd) + pte_idx;
+}
+
 static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd,
                                            unsigned long vaddr, pte_t *lastpte)
 {
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index e6d36b490250..7d4e76da3368 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -168,34 +168,51 @@ static __ref void *spp_getpage(void)
         return ptr;
 }
 
-void
-set_pte_vaddr_pud(pud_t *pud_page, unsigned long vaddr, pte_t new_pte)
+static pud_t * __init fill_pud(pgd_t *pgd, unsigned long vaddr)
 {
-        pud_t *pud;
-        pmd_t *pmd;
-        pte_t *pte;
+        if (pgd_none(*pgd)) {
+                pud_t *pud = (pud_t *)spp_getpage();
+                pgd_populate(&init_mm, pgd, pud);
+                if (pud != pud_offset(pgd, 0))
+                        printk(KERN_ERR "PAGETABLE BUG #00! %p <-> %p\n",
+                               pud, pud_offset(pgd, 0));
+        }
+        return pud_offset(pgd, vaddr);
+}
 
-        pud = pud_page + pud_index(vaddr);
+static pmd_t * __init fill_pmd(pud_t *pud, unsigned long vaddr)
+{
         if (pud_none(*pud)) {
-                pmd = (pmd_t *) spp_getpage();
+                pmd_t *pmd = (pmd_t *) spp_getpage();
                 pud_populate(&init_mm, pud, pmd);
-                if (pmd != pmd_offset(pud, 0)) {
+                if (pmd != pmd_offset(pud, 0))
                         printk(KERN_ERR "PAGETABLE BUG #01! %p <-> %p\n",
-                                pmd, pmd_offset(pud, 0));
-                        return;
-                }
+                               pmd, pmd_offset(pud, 0));
         }
-        pmd = pmd_offset(pud, vaddr);
+        return pmd_offset(pud, vaddr);
+}
+
+static pte_t * __init fill_pte(pmd_t *pmd, unsigned long vaddr)
+{
         if (pmd_none(*pmd)) {
-                pte = (pte_t *) spp_getpage();
+                pte_t *pte = (pte_t *) spp_getpage();
                 pmd_populate_kernel(&init_mm, pmd, pte);
-                if (pte != pte_offset_kernel(pmd, 0)) {
+                if (pte != pte_offset_kernel(pmd, 0))
                         printk(KERN_ERR "PAGETABLE BUG #02!\n");
-                        return;
-                }
         }
+        return pte_offset_kernel(pmd, vaddr);
+}
+
+void set_pte_vaddr_pud(pud_t *pud_page, unsigned long vaddr, pte_t new_pte)
+{
+        pud_t *pud;
+        pmd_t *pmd;
+        pte_t *pte;
+
+        pud = pud_page + pud_index(vaddr);
+        pmd = fill_pmd(pud, vaddr);
+        pte = fill_pte(pmd, vaddr);
 
-        pte = pte_offset_kernel(pmd, vaddr);
         set_pte(pte, new_pte);
 
         /*
@@ -205,8 +222,7 @@ set_pte_vaddr_pud(pud_t *pud_page, unsigned long vaddr, pte_t new_pte)
         __flush_tlb_one(vaddr);
 }
 
-void
-set_pte_vaddr(unsigned long vaddr, pte_t pteval)
+void set_pte_vaddr(unsigned long vaddr, pte_t pteval)
 {
         pgd_t *pgd;
         pud_t *pud_page;
@@ -223,6 +239,24 @@ set_pte_vaddr(unsigned long vaddr, pte_t pteval)
         set_pte_vaddr_pud(pud_page, vaddr, pteval);
 }
 
+pmd_t * __init populate_extra_pmd(unsigned long vaddr)
+{
+        pgd_t *pgd;
+        pud_t *pud;
+
+        pgd = pgd_offset_k(vaddr);
+        pud = fill_pud(pgd, vaddr);
+        return fill_pmd(pud, vaddr);
+}
+
+pte_t * __init populate_extra_pte(unsigned long vaddr)
+{
+        pmd_t *pmd;
+
+        pmd = populate_extra_pmd(vaddr);
+        return fill_pte(pmd, vaddr);
+}
+
 /*
  * Create large page table mappings for a range of physical addresses.
  */