[S390] 1K/2K page table pages.

This patch implements 1K/2K page table pages for s390.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/mm/init.c

@@ -184,7 +184,7 @@ void kernel_map_pages(struct page *page, int numpages, int enable)
 		pmd = pmd_offset(pud, address);
 		pte = pte_offset_kernel(pmd, address);
 		if (!enable) {
-			ptep_invalidate(address, pte);
+			ptep_invalidate(&init_mm, address, pte);
 			continue;
 		}
 		*pte = mk_pte_phys(address, __pgprot(_PAGE_TYPE_RW));

arch/s390/mm/pgtable.c

@@ -26,8 +26,14 @@
 
 #ifndef CONFIG_64BIT
 #define ALLOC_ORDER	1
+#define TABLES_PER_PAGE	4
+#define FRAG_MASK	15UL
+#define SECOND_HALVES	10UL
 #else
 #define ALLOC_ORDER	2
+#define TABLES_PER_PAGE	2
+#define FRAG_MASK	3UL
+#define SECOND_HALVES	2UL
 #endif
 
 unsigned long *crst_table_alloc(struct mm_struct *mm, int noexec)
@@ -45,13 +51,20 @@ unsigned long *crst_table_alloc(struct mm_struct *mm, int noexec)
 		}
 		page->index = page_to_phys(shadow);
 	}
+	spin_lock(&mm->page_table_lock);
+	list_add(&page->lru, &mm->context.crst_list);
+	spin_unlock(&mm->page_table_lock);
 	return (unsigned long *) page_to_phys(page);
 }
 
-void crst_table_free(unsigned long *table)
+void crst_table_free(struct mm_struct *mm, unsigned long *table)
 {
 	unsigned long *shadow = get_shadow_table(table);
+	struct page *page = virt_to_page(table);
 
+	spin_lock(&mm->page_table_lock);
+	list_del(&page->lru);
+	spin_unlock(&mm->page_table_lock);
 	if (shadow)
 		free_pages((unsigned long) shadow, ALLOC_ORDER);
 	free_pages((unsigned long) table, ALLOC_ORDER);
@@ -60,37 +73,84 @@ void crst_table_free(unsigned long *table)
 /*
  * page table entry allocation/free routines.
  */
-unsigned long *page_table_alloc(int noexec)
+unsigned long *page_table_alloc(struct mm_struct *mm)
 {
-	struct page *page = alloc_page(GFP_KERNEL);
+	struct page *page;
 	unsigned long *table;
+	unsigned long bits;
 
-	if (!page)
-		return NULL;
-	page->index = 0;
-	if (noexec) {
-		struct page *shadow = alloc_page(GFP_KERNEL);
-		if (!shadow) {
-			__free_page(page);
+	bits = mm->context.noexec ? 3UL : 1UL;
+	spin_lock(&mm->page_table_lock);
+	page = NULL;
+	if (!list_empty(&mm->context.pgtable_list)) {
+		page = list_first_entry(&mm->context.pgtable_list,
+					struct page, lru);
+		if ((page->flags & FRAG_MASK) == ((1UL << TABLES_PER_PAGE) - 1))
+			page = NULL;
+	}
+	if (!page) {
+		spin_unlock(&mm->page_table_lock);
+		page = alloc_page(GFP_KERNEL|__GFP_REPEAT);
+		if (!page)
 			return NULL;
-		}
-		table = (unsigned long *) page_to_phys(shadow);
+		pgtable_page_ctor(page);
+		page->flags &= ~FRAG_MASK;
+		table = (unsigned long *) page_to_phys(page);
 		clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
-		page->index = (addr_t) table;
+		spin_lock(&mm->page_table_lock);
+		list_add(&page->lru, &mm->context.pgtable_list);
 	}
-	pgtable_page_ctor(page);
 	table = (unsigned long *) page_to_phys(page);
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+	while (page->flags & bits) {
+		table += 256;
+		bits <<= 1;
+	}
+	page->flags |= bits;
+	if ((page->flags & FRAG_MASK) == ((1UL << TABLES_PER_PAGE) - 1))
+		list_move_tail(&page->lru, &mm->context.pgtable_list);
+	spin_unlock(&mm->page_table_lock);
 	return table;
 }
 
-void page_table_free(unsigned long *table)
+void page_table_free(struct mm_struct *mm, unsigned long *table)
 {
-	unsigned long *shadow = get_shadow_pte(table);
+	struct page *page;
+	unsigned long bits;
 
-	pgtable_page_dtor(virt_to_page(table));
-	if (shadow)
-		free_page((unsigned long) shadow);
-	free_page((unsigned long) table);
+	bits = mm->context.noexec ? 3UL : 1UL;
+	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
+	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
+	spin_lock(&mm->page_table_lock);
+	page->flags ^= bits;
+	if (page->flags & FRAG_MASK) {
+		/* Page now has some free pgtable fragments. */
+		list_move(&page->lru, &mm->context.pgtable_list);
+		page = NULL;
+	} else
+		/* All fragments of the 4K page have been freed. */
+		list_del(&page->lru);
+	spin_unlock(&mm->page_table_lock);
+	if (page) {
+		pgtable_page_dtor(page);
+		__free_page(page);
+	}
+}
 
+void disable_noexec(struct mm_struct *mm, struct task_struct *tsk)
+{
+	struct page *page;
+
+	spin_lock(&mm->page_table_lock);
+	/* Free shadow region and segment tables. */
+	list_for_each_entry(page, &mm->context.crst_list, lru)
+		if (page->index) {
+			free_pages((unsigned long) page->index, ALLOC_ORDER);
+			page->index = 0;
+		}
+	/* "Free" second halves of page tables. */
+	list_for_each_entry(page, &mm->context.pgtable_list, lru)
+		page->flags &= ~SECOND_HALVES;
+	spin_unlock(&mm->page_table_lock);
+	mm->context.noexec = 0;
+	update_mm(mm, tsk);
 }

arch/s390/mm/vmem.c

@@ -84,13 +84,18 @@ static inline pmd_t *vmem_pmd_alloc(void)
 	return pmd;
 }
 
-static inline pte_t *vmem_pte_alloc(void)
+static pte_t __init_refok *vmem_pte_alloc(void)
 {
-	pte_t *pte = vmem_alloc_pages(0);
+	pte_t *pte;
 
+	if (slab_is_available())
+		pte = (pte_t *) page_table_alloc(&init_mm);
+	else
+		pte = alloc_bootmem(PTRS_PER_PTE * sizeof(pte_t));
 	if (!pte)
 		return NULL;
-	clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+	clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY,
+		    PTRS_PER_PTE * sizeof(pte_t));
 	return pte;
 }
 
@@ -360,6 +365,9 @@ void __init vmem_map_init(void)
 {
 	int i;
 
+	INIT_LIST_HEAD(&init_mm.context.crst_list);
+	INIT_LIST_HEAD(&init_mm.context.pgtable_list);
+	init_mm.context.noexec = 0;
 	NODE_DATA(0)->node_mem_map = VMEM_MAP;
 	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++)
 		vmem_add_mem(memory_chunk[i].addr, memory_chunk[i].size);

include/asm-s390/elf.h

@@ -115,6 +115,7 @@ typedef s390_regs elf_gregset_t;
 
 #include <linux/sched.h>	/* for task_struct */
 #include <asm/system.h>		/* for save_access_regs */
+#include <asm/mmu_context.h>
 
 /*
  * This is used to ensure we don't load something for the wrong architecture.
@@ -214,4 +215,16 @@ do {							\
 } while (0)
 #endif /* __s390x__ */
 
+/*
+ * An executable for which elf_read_implies_exec() returns TRUE will
+ * have the READ_IMPLIES_EXEC personality flag set automatically.
+ */
+#define elf_read_implies_exec(ex, executable_stack)	\
+({							\
+	if (current->mm->context.noexec &&		\
+	    executable_stack != EXSTACK_DISABLE_X)	\
+		disable_noexec(current->mm, current);	\
+	current->mm->context.noexec == 0;		\
+})
+
 #endif

include/asm-s390/mmu.h

@@ -1,7 +1,11 @@
 #ifndef __MMU_H
 #define __MMU_H
 
-/* Default "unsigned long" context */
-typedef unsigned long mm_context_t;
+typedef struct {
+	struct list_head crst_list;
+	struct list_head pgtable_list;
+	unsigned long asce_bits;
+	int noexec;
+} mm_context_t;
 
 #endif

include/asm-s390/mmu_context.h

@@ -10,15 +10,17 @@
 #define __S390_MMU_CONTEXT_H
 
 #include <asm/pgalloc.h>
+#include <asm/uaccess.h>
 #include <asm-generic/mm_hooks.h>
 
 static inline int init_new_context(struct task_struct *tsk,
 				   struct mm_struct *mm)
 {
-	mm->context = _ASCE_TABLE_LENGTH | _ASCE_USER_BITS;
+	mm->context.asce_bits = _ASCE_TABLE_LENGTH | _ASCE_USER_BITS;
 #ifdef CONFIG_64BIT
-	mm->context |= _ASCE_TYPE_REGION3;
+	mm->context.asce_bits |= _ASCE_TYPE_REGION3;
 #endif
+	mm->context.noexec = s390_noexec;
 	return 0;
 }
 
@@ -32,11 +34,13 @@ static inline int init_new_context(struct task_struct *tsk,
 
 static inline void update_mm(struct mm_struct *mm, struct task_struct *tsk)
 {
-	S390_lowcore.user_asce = mm->context | __pa(mm->pgd);
+	pgd_t *pgd = mm->pgd;
+
+	S390_lowcore.user_asce = mm->context.asce_bits | __pa(pgd);
 	if (switch_amode) {
 		/* Load primary space page table origin. */
-		pgd_t *shadow_pgd = get_shadow_table(mm->pgd) ? : mm->pgd;
-		S390_lowcore.user_exec_asce = mm->context | __pa(shadow_pgd);
+		pgd = mm->context.noexec ? get_shadow_table(pgd) : pgd;
+		S390_lowcore.user_exec_asce = mm->context.asce_bits | __pa(pgd);
 		asm volatile(LCTL_OPCODE" 1,1,%0\n"
 			     : : "m" (S390_lowcore.user_exec_asce) );
 	} else

include/asm-s390/page.h

@@ -74,43 +74,17 @@ static inline void copy_page(void *to, void *from)
 
 typedef struct { unsigned long pgprot; } pgprot_t;
 typedef struct { unsigned long pte; } pte_t;
-
-#define pte_val(x)      ((x).pte)
-#define pgprot_val(x)   ((x).pgprot)
-
-#ifndef __s390x__
-
 typedef struct { unsigned long pmd; } pmd_t;
 typedef struct { unsigned long pud; } pud_t;
-typedef struct {
-        unsigned long pgd0;
-        unsigned long pgd1;
-        unsigned long pgd2;
-        unsigned long pgd3;
-        } pgd_t;
-
-#define pmd_val(x)      ((x).pmd)
-#define pud_val(x)	((x).pud)
-#define pgd_val(x)      ((x).pgd0)
-
-#else /* __s390x__ */
-
-typedef struct { 
-        unsigned long pmd0;
-        unsigned long pmd1; 
-        } pmd_t;
-typedef struct { unsigned long pud; } pud_t;
 typedef struct { unsigned long pgd; } pgd_t;
+typedef pte_t *pgtable_t;
 
-#define pmd_val(x)      ((x).pmd0)
-#define pmd_val1(x)     ((x).pmd1)
+#define pgprot_val(x)	((x).pgprot)
+#define pte_val(x)	((x).pte)
+#define pmd_val(x)	((x).pmd)
 #define pud_val(x)	((x).pud)
 #define pgd_val(x)      ((x).pgd)
 
-#endif /* __s390x__ */
-
-typedef struct page *pgtable_t;
-
 #define __pte(x)        ((pte_t) { (x) } )
 #define __pmd(x)        ((pmd_t) { (x) } )
 #define __pgd(x)        ((pgd_t) { (x) } )
@@ -167,7 +141,7 @@ static inline int pfn_valid(unsigned long pfn)
 #define page_to_phys(page)	(page_to_pfn(page) << PAGE_SHIFT)
 #define virt_addr_valid(kaddr)	pfn_valid(__pa(kaddr) >> PAGE_SHIFT)
 
-#define VM_DATA_DEFAULT_FLAGS	(VM_READ | VM_WRITE | VM_EXEC | \
+#define VM_DATA_DEFAULT_FLAGS	(VM_READ | VM_WRITE | \
 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
 
 #include <asm-generic/memory_model.h>