Implement true end_pfn_mapped for 32bit

Index: linux/arch/x86/mm/init_32.c
===================================================================
@@ -50,6 +50,8 @@
 
 unsigned int __VMALLOC_RESERVE = 128 << 20;
 
+unsigned long end_pfn_map;
+
 DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 unsigned long highstart_pfn, highend_pfn;
 

@@ -193,6 +195,7 @@ static void __init kernel_physical_mappi
 				set_pmd(pmd, pfn_pmd(pfn, prot));
 
 				pfn += PTRS_PER_PTE;
+				end_pfn_map = pfn;
 				continue;
 			}
 			pte = one_page_table_init(pmd);

@@ -207,6 +210,7 @@ static void __init kernel_physical_mappi
 
 				set_pte(pte, pfn_pte(pfn, prot));
 			}
+			end_pfn_map = pfn;
 		}
 	}
 }

Index: linux/include/asm-x86/page.h
===================================================================
@@ -36,7 +36,7 @@
 #define max_pfn_mapped		end_pfn_map
 #else
 #include <asm/page_32.h>
-#define max_pfn_mapped		max_low_pfn
+#define max_pfn_mapped		end_pfn_map
 #endif	/* CONFIG_X86_64 */
 
 #define PAGE_OFFSET		((unsigned long)__PAGE_OFFSET)

@@ -51,6 +51,8 @@
 extern int page_is_ram(unsigned long pagenr);
 extern int devmem_is_allowed(unsigned long pagenr);
 
+extern unsigned long end_pfn_map;
+
 struct page;
 
 static void inline clear_user_page(void *page, unsigned long vaddr,

Index: linux/include/asm-x86/page_64.h
===================================================================
@@ -55,7 +55,6 @@ void clear_page(void *page);
 void copy_page(void *to, void *from);
 
 extern unsigned long end_pfn;
-extern unsigned long end_pfn_map;
 extern unsigned long phys_base;
 
 extern unsigned long __phys_addr(unsigned long);

Index: linux/arch/x86/mm/init_64.c
===================================================================
@@ -296,7 +296,7 @@ __meminit void early_iounmap(void *addr,
 	__flush_tlb_all();
 }
 
-static void __meminit
+static unsigned long __meminit
 phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
 {
 	int i = pmd_index(address);

@@ -318,21 +318,25 @@ phys_pmd_init(pmd_t *pmd_page, unsigned 
 		set_pte((pte_t *)pmd,
 			pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 	}
+	return address;
 }
 
-static void __meminit
+static unsigned long __meminit
 phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
 {
+	unsigned long true_end;
 	pmd_t *pmd = pmd_offset(pud, 0);
 	spin_lock(&init_mm.page_table_lock);
-	phys_pmd_init(pmd, address, end);
+	true_end = phys_pmd_init(pmd, address, end);
 	spin_unlock(&init_mm.page_table_lock);
 	__flush_tlb_all();
+	return true_end;
 }
 
-static void __meminit
+static unsigned long __meminit
 phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
 {
+	unsigned long true_end = end;
 	int i = pud_index(addr);
 
 	for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE) {

@@ -351,13 +355,14 @@ phys_pud_init(pud_t *pud_page, unsigned 
 
 		if (pud_val(*pud)) {
 			if (!pud_large(*pud))
-				phys_pmd_update(pud, addr, end);
+				true_end = phys_pmd_update(pud, addr, end);
 			continue;
 		}
 
 		if (direct_gbpages) {
 			set_pte((pte_t *)pud,
 				pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
+			true_end = (addr & PUD_MASK) + PUD_SIZE;
 			continue;
 		}
 

@@ -365,12 +370,14 @@ phys_pud_init(pud_t *pud_page, unsigned 
 
 		spin_lock(&init_mm.page_table_lock);
 		set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE));
-		phys_pmd_init(pmd, addr, end);
+		true_end = phys_pmd_init(pmd, addr, end);
 		spin_unlock(&init_mm.page_table_lock);
 
 		unmap_low_page(pmd);
 	}
 	__flush_tlb_all();
+
+	return true_end >> PAGE_SHIFT;
 }
 
 static void __init find_early_table_space(unsigned long end)

@@ -415,9 +422,10 @@ static void __init init_gbpages(void)
  * This runs before bootmem is initialized and gets pages directly from
  * the physical memory. To access them they are temporarily mapped.
  */
-void __init_refok init_memory_mapping(unsigned long start, unsigned long end)
+unsigned long __init_refok
+init_memory_mapping(unsigned long start, unsigned long end)
 {
-	unsigned long next;
+	unsigned long next, true_end = end;
 
 	pr_debug("init_memory_mapping\n");
 

@@ -449,7 +457,7 @@ void __init_refok init_memory_mapping(un
 		next = start + PGDIR_SIZE;
 		if (next > end)
 			next = end;
-		phys_pud_init(pud, __pa(start), __pa(next));
+		true_end = phys_pud_init(pud, __pa(start), __pa(next));
 		if (!after_bootmem)
 			set_pgd(pgd_offset_k(start), mk_kernel_pgd(pud_phys));
 		unmap_low_page(pud);

@@ -462,6 +470,8 @@ void __init_refok init_memory_mapping(un
 	if (!after_bootmem)
 		reserve_early(table_start << PAGE_SHIFT,
 				 table_end << PAGE_SHIFT, "PGTABLE");
+
+	return true_end;
 }
 
 #ifndef CONFIG_NUMA

@@ -503,9 +513,12 @@ int arch_add_memory(int nid, u64 start, 
 	struct zone *zone = pgdat->node_zones + ZONE_NORMAL;
 	unsigned long start_pfn = start >> PAGE_SHIFT;
 	unsigned long nr_pages = size >> PAGE_SHIFT;
+	unsigned long true_end_pfn;
 	int ret;
 
-	init_memory_mapping(start, start + size-1);
+	true_end_pfn = init_memory_mapping(start, start + size-1);
+	if (true_end_pfn > end_pfn_map)
+		end_pfn_map = true_end_pfn;
 
 	ret = __add_pages(zone, start_pfn, nr_pages);
 	WARN_ON(1);

Index: linux/include/asm-x86/proto.h
===================================================================
@@ -7,7 +7,8 @@
 
 extern void early_idt_handler(void);
 
-extern void init_memory_mapping(unsigned long start, unsigned long end);
+extern unsigned long init_memory_mapping(unsigned long start,
+					 unsigned long end);
 
 extern void system_call(void);
 extern void syscall_init(void);

Index: linux/arch/x86/kernel/setup_64.c
===================================================================
@@ -341,7 +341,7 @@ void __init setup_arch(char **cmdline_p)
 
 	check_efer();
 
-	init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
+	end_pfn_map = init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
 	if (efi_enabled)
 		efi_init();
 

Index: linux/arch/x86/mm/pageattr.c
===================================================================
@@ -28,6 +28,7 @@ struct cpa_data {
 	int		numpages;
 	int		flushtlb;
 	unsigned long	pfn;
+	unsigned	force_split : 1;
 };
 
 #ifdef CONFIG_X86_64

@@ -259,6 +260,9 @@ try_preserve_large_page(pte_t *kpte, uns
 	int i, do_split = 1;
 	unsigned int level;
 
+	if (cpa->force_split)
+		return 1;
+
 	spin_lock_irqsave(&pgd_lock, flags);
 	/*
 	 * Check for races, another CPU might have split this page

@@ -693,7 +697,8 @@ static inline int cache_attr(pgprot_t at
 }
 
 static int change_page_attr_set_clr(unsigned long addr, int numpages,
-				    pgprot_t mask_set, pgprot_t mask_clr)
+				    pgprot_t mask_set, pgprot_t mask_clr,
+				    int force_split)
 {
 	struct cpa_data cpa;
 	int ret, cache, checkalias;

@@ -704,7 +709,7 @@ static int change_page_attr_set_clr(unsi
 	 */
 	mask_set = canon_pgprot(mask_set);
 	mask_clr = canon_pgprot(mask_clr);
-	if (!pgprot_val(mask_set) && !pgprot_val(mask_clr))
+	if (!pgprot_val(mask_set) && !pgprot_val(mask_clr) && !force_split)
 		return 0;
 
 	/* Ensure we are PAGE_SIZE aligned */

@@ -721,6 +726,7 @@ static int change_page_attr_set_clr(unsi
 	cpa.mask_set = mask_set;
 	cpa.mask_clr = mask_clr;
 	cpa.flushtlb = 0;
+	cpa.force_split = force_split;
 
 	/* No alias checking for _NX bit modifications */
 	checkalias = (pgprot_val(mask_set) | pgprot_val(mask_clr)) != _PAGE_NX;

@@ -759,13 +765,13 @@ out:
 static inline int change_page_attr_set(unsigned long addr, int numpages,
 				       pgprot_t mask)
 {
-	return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0));
+	return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0), 0);
 }
 
 static inline int change_page_attr_clear(unsigned long addr, int numpages,
 					 pgprot_t mask)
 {
-	return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask);
+	return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask, 0);
 }
 
 int set_memory_uc(unsigned long addr, int numpages)

@@ -809,6 +815,12 @@ int set_memory_np(unsigned long addr, in
 	return change_page_attr_clear(addr, numpages, __pgprot(_PAGE_PRESENT));
 }
 
+int set_memory_4k(unsigned long addr, int numpages)
+{
+	return change_page_attr_set_clr(addr, numpages, __pgprot(0),
+					__pgprot(0), 1);
+}
+
 int set_pages_uc(struct page *page, int numpages)
 {
 	unsigned long addr = (unsigned long)page_address(page);

Index: linux/include/asm-x86/cacheflush.h
===================================================================
@@ -41,6 +41,7 @@ int set_memory_nx(unsigned long addr, in
 int set_memory_ro(unsigned long addr, int numpages);
 int set_memory_rw(unsigned long addr, int numpages);
 int set_memory_np(unsigned long addr, int numpages);
+int set_memory_4k(unsigned long addr, int numpages);
 
 void clflush_cache_range(void *addr, unsigned int size);
 

Index: linux/arch/x86/kernel/bugs_64.c
===================================================================
@@ -9,6 +9,7 @@
 #include <asm/bugs.h>
 #include <asm/processor.h>
 #include <asm/mtrr.h>
+#include <asm/cacheflush.h>
 
 void __init check_bugs(void)
 {

@@ -18,4 +19,15 @@ void __init check_bugs(void)
 	print_cpu_info(&boot_cpu_data);
 #endif
 	alternative_instructions();
+
+	/*
+	 * Make sure the first 2MB area is not mapped by huge pages
+	 * There are typically fixed size MTRRs in there and overlapping
+	 * MTRRs into large pages causes slow downs.
+	 *
+	 * Right now we don't do that with gbpages because there seems
+	 * very little benefit for that case.
+	 */
+	if (!direct_gbpages)
+		set_memory_4k((unsigned long)__va(0), 1);
 }

Index: linux/arch/x86/mm/init_32.c
===================================================================
@@ -181,8 +181,13 @@ static void __init kernel_physical_mappi
 			/*
 			 * Map with big pages if possible, otherwise
 			 * create normal page tables:
+			 *
+			 * Don't use a large page for the first 2/4MB of memory
+			 * because there are often fixed size MTRRs in there
+			 * and overlapping MTRRs into large pages can cause
+			 * slowdowns.
 			 */
-			if (cpu_has_pse) {
+			if (cpu_has_pse && !(pgd_idx == 0 && pmd_idx == 0)) {
 				unsigned int addr2;
 				pgprot_t prot = PAGE_KERNEL_LARGE;
 

Index: linux/include/asm-x86/msr.h
===================================================================
@@ -150,6 +150,9 @@ static inline int wrmsr_safe(unsigned ms
 		__err;							\
 	})
 
+#define rdmsrl_safe(msr,p) \
+	({ int __err; *(p) = native_read_msr_safe(msr, &__err); __err; })
+
 #define rdtscl(low)						\
 	((low) = (u32)native_read_tsc())
 

Index: linux/include/asm-x86/paravirt.h
===================================================================
@@ -687,6 +687,10 @@ static inline int paravirt_write_msr(uns
 	(*b) = _l >> 32;			\
 	_err; })
 
+#define rdmsrl_safe(msr, p) ({			\
+	int _err;				\
+	*(p) = paravirt_read_msr(msr, &_err);	\
+	_err; })
 
 static inline u64 paravirt_read_tsc(void)
 {

Index: linux/arch/x86/kernel/setup_64.c
===================================================================
@@ -721,6 +721,20 @@ static void __cpuinit init_amd(struct cp
 
 	if (amd_apic_timer_broken())
 		disable_apic_timer = 1;
+
+	if (!direct_gbpages &&
+		c == &boot_cpu_data && c->x86 >= 0xf && c->x86 <= 0x11) {
+		unsigned long tseg;
+
+		/*
+		 * Split up direct mapping around the TSEG SMM area.
+		 * Don't do it for gbpages because there seems very little
+		 * benefit in doing so.
+		 */
+		if (!rdmsrl_safe(MSR_K8_TSEG_ADDR, &tseg) &&
+		(tseg >> PMD_SHIFT) < (end_pfn_map >> (PMD_SHIFT-PAGE_SHIFT)))
+			set_memory_4k((unsigned long)__va(tseg), 1);
+	}
 }
 
 void __cpuinit detect_ht(struct cpuinfo_x86 *c)

Index: linux/include/asm-x86/msr-index.h
===================================================================
@@ -109,6 +109,7 @@
 #define MSR_K8_SYSCFG			0xc0010010
 #define MSR_K8_HWCR			0xc0010015
 #define MSR_K8_ENABLE_C1E		0xc0010055
+#define MSR_K8_TSEG_ADDR		0xc0010112
 #define K8_MTRRFIXRANGE_DRAM_ENABLE	0x00040000 /* MtrrFixDramEn bit    */
 #define K8_MTRRFIXRANGE_DRAM_MODIFY	0x00080000 /* MtrrFixDramModEn bit */
 #define K8_MTRR_RDMEM_WRMEM_MASK	0x18181818 /* Mask: RdMem|WrMem    */

Index: linux/arch/x86/mm/init_64.c
===================================================================
@@ -319,6 +319,8 @@ __meminit void early_iounmap(void *addr,
 static unsigned long __meminit
 phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
 {
+	unsigned long flags;
+	unsigned pages = 0;
 	int i = pmd_index(address);
 
 	for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {

@@ -335,9 +337,15 @@ phys_pmd_init(pmd_t *pmd_page, unsigned 
 		if (pmd_val(*pmd))
 			continue;
 
+		pages++;
 		set_pte((pte_t *)pmd,
 			pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 	}
+
+	/* Protect against CPA */
+	spin_lock_irqsave(&pgd_lock, flags);
+	dpages_cnt[PG_LEVEL_2M] += pages;
+	spin_unlock_irqrestore(&pgd_lock, flags);
 	return address;
 }
 

@@ -356,6 +364,8 @@ phys_pmd_update(pud_t *pud, unsigned lon
 static unsigned long __meminit
 phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
 {
+	unsigned long flags;
+	unsigned pages = 0;
 	unsigned long true_end = end;
 	int i = pud_index(addr);
 

@@ -380,6 +390,7 @@ phys_pud_init(pud_t *pud_page, unsigned 
 		}
 
 		if (direct_gbpages) {
+			dpages_cnt[PG_LEVEL_1G]++;
 			set_pte((pte_t *)pud,
 				pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 			true_end = (addr & PUD_MASK) + PUD_SIZE;

@@ -397,6 +408,11 @@ phys_pud_init(pud_t *pud_page, unsigned 
 	}
 	__flush_tlb_all();
 
+	/* Protect against CPA */
+	spin_lock_irqsave(&pgd_lock, flags);
+	dpages_cnt[PG_LEVEL_1G] += pages;
+	spin_unlock_irqrestore(&pgd_lock, flags);
+
 	return true_end >> PAGE_SHIFT;
 }
 

Index: linux/arch/x86/mm/pageattr.c
===================================================================
@@ -18,6 +18,8 @@
 #include <asm/pgalloc.h>
 #include <asm/proto.h>
 
+unsigned long dpages_cnt[PG_LEVEL_NUM];
+
 /*
  * The current flushing context - we pass it instead of 5 arguments:
  */

@@ -499,6 +501,12 @@ static int split_large_page(pte_t *kpte,
 	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
 		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
 
+	if (address >= (unsigned long)__va(0) &&
+		address < (unsigned long)__va(end_pfn_map << PAGE_SHIFT)) {
+		dpages_cnt[level]--;
+		dpages_cnt[level - 1] += PTRS_PER_PTE;
+	}
+
 	/*
 	 * Install the new, split up pagetable. Important details here:
 	 *

@@ -948,6 +956,22 @@ bool kernel_page_present(struct page *pa
 
 #endif /* CONFIG_DEBUG_PAGEALLOC */
 
+#ifdef CONFIG_PROC_FS
+int arch_report_meminfo(char *page)
+{
+	int n;
+	n = sprintf(page, "DirectMap4k:  %8lu\n"
+			  "DirectMap2M:  %8lu\n",
+			dpages_cnt[PG_LEVEL_4K],
+			dpages_cnt[PG_LEVEL_2M]);
+#ifdef CONFIG_X86_64
+	n += sprintf(page + n, "DirectMap1G:  %8lu\n",
+			dpages_cnt[PG_LEVEL_1G]);
+#endif
+	return n;
+}
+#endif
+
 /*
  * The testcases use internal knowledge of the implementation that shouldn't
  * be exposed to the rest of the kernel. Include these directly here.

Index: linux/include/asm-x86/pgtable.h
===================================================================
@@ -247,8 +247,11 @@ enum {
 	PG_LEVEL_4K,
 	PG_LEVEL_2M,
 	PG_LEVEL_1G,
+	PG_LEVEL_NUM
 };
 
+extern unsigned long dpages_cnt[PG_LEVEL_NUM];
+
 /*
  * Helper function that returns the kernel pagetable entry controlling
  * the virtual address 'address'. NULL means no pagetable entry present.

Index: linux/arch/x86/mm/init_32.c
===================================================================
@@ -198,6 +198,7 @@ static void __init kernel_physical_mappi
 				    is_kernel_text(addr2))
 					prot = PAGE_KERNEL_LARGE_EXEC;
 
+				dpages_cnt[PG_LEVEL_2M]++;
 				set_pmd(pmd, pfn_pmd(pfn, prot));
 
 				pfn += PTRS_PER_PTE;

@@ -214,6 +215,7 @@ static void __init kernel_physical_mappi
 				if (is_kernel_text(addr))
 					prot = PAGE_KERNEL_EXEC;
 
+				dpages_cnt[PG_LEVEL_4K]++;
 				set_pte(pte, pfn_pte(pfn, prot));
 			}
 			end_pfn_map = pfn;

Index: linux/fs/proc/proc_misc.c
===================================================================
@@ -123,6 +123,11 @@ static int uptime_read_proc(char *page, 
 	return proc_calc_metrics(page, start, off, count, eof, len);
 }
 
+int __attribute__((weak)) arch_report_meminfo(char *page)
+{
+	return 0;
+}
+
 static int meminfo_read_proc(char *page, char **start, off_t off,
 				 int count, int *eof, void *data)
 {

@@ -219,6 +224,8 @@ static int meminfo_read_proc(char *page,
 
 		len += hugetlb_report_meminfo(page + len);
 
+	len += arch_report_meminfo(page + len);
+
 	return proc_calc_metrics(page, start, off, count, eof, len);
 #undef K
 }

Index: linux/arch/x86/mm/init_64.c
===================================================================
@@ -316,9 +316,22 @@ __meminit void early_iounmap(void *addr,
 	__flush_tlb_all();
 }
 
+static void update_page_count(int level, unsigned long pages)
+{
+#ifdef CONFIG_PROC_FS
+	unsigned long flags;
+	/* Protect against CPA */
+	spin_lock_irqsave(&pgd_lock, flags);
+	direct_pages_count[PG_LEVEL_2M] += pages;
+	spin_unlock_irqrestore(&pgd_lock, flags);
+#endif
+}
+
 static unsigned long __meminit
 phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
 {
+	unsigned long pages = 0;
+
 	int i = pmd_index(address);
 
 	for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {

@@ -335,9 +348,11 @@ phys_pmd_init(pmd_t *pmd_page, unsigned 
 		if (pmd_val(*pmd))
 			continue;
 
+		pages++;
 		set_pte((pte_t *)pmd,
 			pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 	}
+	update_page_count(PG_LEVEL_2M, pages);
 	return address;
 }
 

@@ -356,6 +371,7 @@ phys_pmd_update(pud_t *pud, unsigned lon
 static unsigned long __meminit
 phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
 {
+	unsigned long pages = 0;
 	unsigned long true_end = end;
 	int i = pud_index(addr);
 

@@ -380,6 +396,7 @@ phys_pud_init(pud_t *pud_page, unsigned 
 		}
 
 		if (direct_gbpages) {
+			direct_pages_count[PG_LEVEL_1G]++;
 			set_pte((pte_t *)pud,
 				pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 			true_end = (addr & PUD_MASK) + PUD_SIZE;

@@ -397,6 +414,8 @@ phys_pud_init(pud_t *pud_page, unsigned 
 	}
 	__flush_tlb_all();
 
+	update_page_count(PG_LEVEL_1G, pages);
+
 	return true_end >> PAGE_SHIFT;
 }
 

Index: linux/arch/x86/mm/pageattr.c
===================================================================
@@ -18,6 +18,8 @@
 #include <asm/pgalloc.h>
 #include <asm/proto.h>
 
+unsigned long direct_pages_count[PG_LEVEL_NUM];
+
 /*
  * The current flushing context - we pass it instead of 5 arguments:
  */

@@ -499,6 +501,12 @@ static int split_large_page(pte_t *kpte,
 	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
 		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
 
+	if (address >= (unsigned long)__va(0) &&
+		address < (unsigned long)__va(end_pfn_map << PAGE_SHIFT)) {
+		direct_pages_count[level]--;
+		direct_pages_count[level - 1] += PTRS_PER_PTE;
+	}
+
 	/*
 	 * Install the new, split up pagetable. Important details here:
 	 *

@@ -948,6 +956,22 @@ bool kernel_page_present(struct page *pa
 
 #endif /* CONFIG_DEBUG_PAGEALLOC */
 
+#ifdef CONFIG_PROC_FS
+int arch_report_meminfo(char *page)
+{
+	int n;
+	n = sprintf(page, "DirectMap4k:  %8lu\n"
+			  "DirectMap2M:  %8lu\n",
+			direct_pages_count[PG_LEVEL_4K],
+			direct_pages_count[PG_LEVEL_2M]);
+#ifdef CONFIG_X86_64
+	n += sprintf(page + n, "DirectMap1G:  %8lu\n",
+			direct_pages_count[PG_LEVEL_1G]);
+#endif
+	return n;
+}
+#endif
+
 /*
  * The testcases use internal knowledge of the implementation that shouldn't
  * be exposed to the rest of the kernel. Include these directly here.

Index: linux/include/asm-x86/pgtable.h
===================================================================
@@ -247,8 +247,11 @@ enum {
 	PG_LEVEL_4K,
 	PG_LEVEL_2M,
 	PG_LEVEL_1G,
+	PG_LEVEL_NUM
 };
 
+extern unsigned long direct_pages_count[PG_LEVEL_NUM];
+
 /*
  * Helper function that returns the kernel pagetable entry controlling
  * the virtual address 'address'. NULL means no pagetable entry present.

Index: linux/arch/x86/mm/init_32.c
===================================================================
@@ -198,6 +198,7 @@ static void __init kernel_physical_mappi
 				    is_kernel_text(addr2))
 					prot = PAGE_KERNEL_LARGE_EXEC;
 
+				direct_pages_count[PG_LEVEL_2M]++;
 				set_pmd(pmd, pfn_pmd(pfn, prot));
 
 				pfn += PTRS_PER_PTE;

@@ -214,6 +215,7 @@ static void __init kernel_physical_mappi
 				if (is_kernel_text(addr))
 					prot = PAGE_KERNEL_EXEC;
 
+				direct_pages_count[PG_LEVEL_4K]++;
 				set_pte(pte, pfn_pte(pfn, prot));
 			}
 			end_pfn_map = pfn;

Index: linux/fs/proc/proc_misc.c
===================================================================
@@ -123,6 +123,11 @@ static int uptime_read_proc(char *page, 
 	return proc_calc_metrics(page, start, off, count, eof, len);
 }
 
+int __attribute__((weak)) arch_report_meminfo(char *page)
+{
+	return 0;
+}
+
 static int meminfo_read_proc(char *page, char **start, off_t off,
 				 int count, int *eof, void *data)
 {

@@ -219,6 +224,8 @@ static int meminfo_read_proc(char *page,
 
 		len += hugetlb_report_meminfo(page + len);
 
+	len += arch_report_meminfo(page + len);
+
 	return proc_calc_metrics(page, start, off, count, eof, len);
 #undef K
 }

Index: linux/include/asm-x86/msr.h
===================================================================
@@ -150,6 +150,13 @@ static inline int wrmsr_safe(unsigned ms
 		__err;							\
 	})
 
+static inline int rdmsrl_safe(unsigned msr, unsigned long long *p)
+{
+	int err;
+	*p = native_read_msr_safe(msr, &err);
+	return err;
+}
+
 #define rdtscl(low)						\
 	((low) = (u32)native_read_tsc())
 

Index: linux/include/asm-x86/paravirt.h
===================================================================
@@ -687,6 +687,12 @@ static inline int paravirt_write_msr(uns
 	(*b) = _l >> 32;			\
 	_err; })
 
+static inline int rdmsrl_safe(unsigned msr, unsigned long long *p)
+{
+	int err;
+	*p = paravirt_read_msr(msr, &err);
+	return err;
+}
 
 static inline u64 paravirt_read_tsc(void)
 {

Index: linux/arch/x86/mm/init_64.c
===================================================================
@@ -319,6 +319,8 @@ __meminit void early_iounmap(void *addr,
 static unsigned long __meminit
 phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
 {
+	unsigned long pages = 0;
+
 	int i = pmd_index(address);
 
 	for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {

@@ -335,9 +337,11 @@ phys_pmd_init(pmd_t *pmd_page, unsigned 
 		if (pmd_val(*pmd))
 			continue;
 
+		pages++;
 		set_pte((pte_t *)pmd,
 			pfn_pte(address >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 	}
+	update_page_count(PG_LEVEL_2M, pages);
 	return address;
 }
 

@@ -356,6 +360,7 @@ phys_pmd_update(pud_t *pud, unsigned lon
 static unsigned long __meminit
 phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
 {
+	unsigned long pages = 0;
 	unsigned long true_end = end;
 	int i = pud_index(addr);
 

@@ -380,6 +385,7 @@ phys_pud_init(pud_t *pud_page, unsigned 
 		}
 
 		if (direct_gbpages) {
+			pages++;
 			set_pte((pte_t *)pud,
 				pfn_pte(addr >> PAGE_SHIFT, PAGE_KERNEL_LARGE));
 			true_end = (addr & PUD_MASK) + PUD_SIZE;

@@ -397,6 +403,8 @@ phys_pud_init(pud_t *pud_page, unsigned 
 	}
 	__flush_tlb_all();
 
+	update_page_count(PG_LEVEL_1G, pages);
+
 	return true_end >> PAGE_SHIFT;
 }
 

Index: linux/arch/x86/mm/pageattr.c
===================================================================
@@ -31,6 +31,19 @@ struct cpa_data {
 	unsigned	force_split : 1;
 };
 
+static unsigned long direct_pages_count[PG_LEVEL_NUM];
+
+void __meminit update_page_count(int level, unsigned long pages)
+{
+#ifdef CONFIG_PROC_FS
+	unsigned long flags;
+	/* Protect against CPA */
+	spin_lock_irqsave(&pgd_lock, flags);
+	direct_pages_count[level] += pages;
+	spin_unlock_irqrestore(&pgd_lock, flags);
+#endif
+}
+
 #ifdef CONFIG_X86_64
 
 static inline unsigned long highmap_start_pfn(void)

@@ -499,6 +512,12 @@ static int split_large_page(pte_t *kpte,
 	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
 		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
 
+	if (address >= (unsigned long)__va(0) &&
+		address < (unsigned long)__va(end_pfn_map << PAGE_SHIFT)) {
+		direct_pages_count[level]--;
+		direct_pages_count[level - 1] += PTRS_PER_PTE;
+	}
+
 	/*
 	 * Install the new, split up pagetable. Important details here:
 	 *

@@ -948,6 +967,22 @@ bool kernel_page_present(struct page *pa
 
 #endif /* CONFIG_DEBUG_PAGEALLOC */
 
+#ifdef CONFIG_PROC_FS
+int arch_report_meminfo(char *page)
+{
+	int n;
+	n = sprintf(page, "DirectMap4k:  %8lu\n"
+			  "DirectMap2M:  %8lu\n",
+			direct_pages_count[PG_LEVEL_4K],
+			direct_pages_count[PG_LEVEL_2M]);
+#ifdef CONFIG_X86_64
+	n += sprintf(page + n, "DirectMap1G:  %8lu\n",
+			direct_pages_count[PG_LEVEL_1G]);
+#endif
+	return n;
+}
+#endif
+
 /*
  * The testcases use internal knowledge of the implementation that shouldn't
  * be exposed to the rest of the kernel. Include these directly here.

Index: linux/include/asm-x86/pgtable.h
===================================================================
@@ -247,8 +247,11 @@ enum {
 	PG_LEVEL_4K,
 	PG_LEVEL_2M,
 	PG_LEVEL_1G,
+	PG_LEVEL_NUM
 };
 
+void update_page_count(int level, unsigned long pages);
+
 /*
  * Helper function that returns the kernel pagetable entry controlling
  * the virtual address 'address'. NULL means no pagetable entry present.

Index: linux/arch/x86/mm/init_32.c
===================================================================
@@ -163,6 +163,7 @@ static void __init kernel_physical_mappi
 	pgd_t *pgd;
 	pmd_t *pmd;
 	pte_t *pte;
+	unsigned pages_2m = 0, pages_4k = 0;
 
 	pgd_idx = pgd_index(PAGE_OFFSET);
 	pgd = pgd_base + pgd_idx;

@@ -198,6 +199,7 @@ static void __init kernel_physical_mappi
 				    is_kernel_text(addr2))
 					prot = PAGE_KERNEL_LARGE_EXEC;
 
+				pages_2m++;
 				set_pmd(pmd, pfn_pmd(pfn, prot));
 
 				pfn += PTRS_PER_PTE;

@@ -214,11 +216,14 @@ static void __init kernel_physical_mappi
 				if (is_kernel_text(addr))
 					prot = PAGE_KERNEL_EXEC;
 
+				pages_4k++;
 				set_pte(pte, pfn_pte(pfn, prot));
 			}
 			end_pfn_map = pfn;
 		}
 	}
+	update_page_count(PG_LEVEL_2M, pages_2m);
+	update_page_count(PG_LEVEL_4K, pages_4k);
 }
 
 static inline int page_kills_ppro(unsigned long pagenr)

Index: linux/fs/proc/proc_misc.c
===================================================================
@@ -123,6 +123,11 @@ static int uptime_read_proc(char *page, 
 	return proc_calc_metrics(page, start, off, count, eof, len);
 }
 
+int __attribute__((weak)) arch_report_meminfo(char *page)
+{
+	return 0;
+}
+
 static int meminfo_read_proc(char *page, char **start, off_t off,
 				 int count, int *eof, void *data)
 {

@@ -219,6 +224,8 @@ static int meminfo_read_proc(char *page,
 
 		len += hugetlb_report_meminfo(page + len);
 
+	len += arch_report_meminfo(page + len);
+
 	return proc_calc_metrics(page, start, off, count, eof, len);
 #undef K
 }