[S390] dynamic page tables.

Add support for different number of page table levels dependent
on the highest address used for a process. This will cause a 31 bit
process to use a two level page table instead of the four level page
table that is the default after the pud has been introduced. Likewise
a normal 64 bit process will use three levels instead of four. Only
if a process runs out of the 4 tera bytes which can be addressed with
a three level page table the fourth level is dynamically added. Then
the process can use up to 8 peta byte.

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

6 files changed, 195 insertions, 3 deletions

diff --git a/arch/s390/kernel/binfmt_elf32.c b/arch/s390/kernel/binfmt_elf32.c
index f1e40ca00d8d..3e1c315b736d 100644
--- a/arch/s390/kernel/binfmt_elf32.c
+++ b/arch/s390/kernel/binfmt_elf32.c
@@ -134,6 +134,7 @@ static inline int dump_task_fpu(struct task_struct *tsk, elf_fpregset_t *fpregs)
 }
 
 #include <asm/processor.h>
+#include <asm/pgalloc.h>
 #include <linux/module.h>
 #include <linux/elfcore.h>
 #include <linux/binfmts.h>
@@ -183,6 +184,16 @@ struct elf_prpsinfo32
 #undef start_thread
 #define start_thread                    start_thread31 
 
+static inline void start_thread31(struct pt_regs *regs, unsigned long new_psw,
+				  unsigned long new_stackp)
+{
+	set_fs(USER_DS);
+	regs->psw.mask	= psw_user32_bits;
+	regs->psw.addr	= new_psw;
+	regs->gprs[15]	= new_stackp;
+	crst_table_downgrade(current->mm, 1UL << 31);
+}
+
 MODULE_DESCRIPTION("Binary format loader for compatibility with 32bit Linux for S390 binaries,"
                    " Copyright 2000 IBM Corporation"); 
 MODULE_AUTHOR("Gerhard Tonn <ton@de.ibm.com>");
diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
index a4d29025ddbd..60f728aeaf12 100644
--- a/arch/s390/kernel/traps.c
+++ b/arch/s390/kernel/traps.c
@@ -60,6 +60,7 @@ int sysctl_userprocess_debug = 0;
 extern pgm_check_handler_t do_protection_exception;
 extern pgm_check_handler_t do_dat_exception;
 extern pgm_check_handler_t do_monitor_call;
+extern pgm_check_handler_t do_asce_exception;
 
 #define stack_pointer ({ void **sp; asm("la %0,0(15)" : "=&d" (sp)); sp; })
 
@@ -730,7 +731,7 @@ void __init trap_init(void)
         pgm_check_table[0x12] = &translation_exception;
         pgm_check_table[0x13] = &special_op_exception;
 #ifdef CONFIG_64BIT
-        pgm_check_table[0x38] = &do_dat_exception;
+	pgm_check_table[0x38] = &do_asce_exception;
 	pgm_check_table[0x39] = &do_dat_exception;
 	pgm_check_table[0x3A] = &do_dat_exception;
         pgm_check_table[0x3B] = &do_dat_exception;
diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
index 2456b52ed068..ed13d429a487 100644
--- a/arch/s390/mm/fault.c
+++ b/arch/s390/mm/fault.c
@@ -32,6 +32,7 @@
 #include <asm/system.h>
 #include <asm/pgtable.h>
 #include <asm/s390_ext.h>
+#include <asm/mmu_context.h>
 
 #ifndef CONFIG_64BIT
 #define __FAIL_ADDR_MASK 0x7ffff000
@@ -444,6 +445,45 @@ void __kprobes do_dat_exception(struct pt_regs *regs, unsigned long error_code)
 	do_exception(regs, error_code & 0xff, 0);
 }
 
+#ifdef CONFIG_64BIT
+void __kprobes do_asce_exception(struct pt_regs *regs, unsigned long error_code)
+{
+	struct mm_struct *mm;
+	struct vm_area_struct *vma;
+	unsigned long address;
+	int space;
+
+	mm = current->mm;
+	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
+	space = check_space(current);
+
+	if (unlikely(space == 0 || in_atomic() || !mm))
+		goto no_context;
+
+	local_irq_enable();
+
+	down_read(&mm->mmap_sem);
+	vma = find_vma(mm, address);
+	up_read(&mm->mmap_sem);
+
+	if (vma) {
+		update_mm(mm, current);
+		return;
+	}
+
+	/* User mode accesses just cause a SIGSEGV */
+	if (regs->psw.mask & PSW_MASK_PSTATE) {
+		current->thread.prot_addr = address;
+		current->thread.trap_no = error_code;
+		do_sigsegv(regs, error_code, SEGV_MAPERR, address);
+		return;
+	}
+
+no_context:
+	do_no_context(regs, error_code, address);
+}
+#endif
+
 #ifdef CONFIG_PFAULT 
 /*
  * 'pfault' pseudo page faults routines.
diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
index 248a71010700..8053245fe259 100644
--- a/arch/s390/mm/init.c
+++ b/arch/s390/mm/init.c
@@ -112,8 +112,9 @@ void __init paging_init(void)
 	init_mm.pgd = swapper_pg_dir;
 	S390_lowcore.kernel_asce = __pa(init_mm.pgd) & PAGE_MASK;
 #ifdef CONFIG_64BIT
-	S390_lowcore.kernel_asce |= _ASCE_TYPE_REGION2 | _ASCE_TABLE_LENGTH;
-	pgd_type = _REGION2_ENTRY_EMPTY;
+	/* A three level page table (4TB) is enough for the kernel space. */
+	S390_lowcore.kernel_asce |= _ASCE_TYPE_REGION3 | _ASCE_TABLE_LENGTH;
+	pgd_type = _REGION3_ENTRY_EMPTY;
 #else
 	S390_lowcore.kernel_asce |= _ASCE_TABLE_LENGTH;
 	pgd_type = _SEGMENT_ENTRY_EMPTY;
diff --git a/arch/s390/mm/mmap.c b/arch/s390/mm/mmap.c
index 356257c171de..5932a824547a 100644
--- a/arch/s390/mm/mmap.c
+++ b/arch/s390/mm/mmap.c
@@ -27,6 +27,7 @@
 #include <linux/personality.h>
 #include <linux/mm.h>
 #include <linux/module.h>
+#include <asm/pgalloc.h>
 
 /*
  * Top of mmap area (just below the process stack).
@@ -62,6 +63,8 @@ static inline int mmap_is_legacy(void)
 	    current->signal->rlim[RLIMIT_STACK].rlim_cur == RLIM_INFINITY;
 }
 
+#ifndef CONFIG_64BIT
+
 /*
  * This function, called very early during the creation of a new
  * process VM image, sets up which VM layout function to use:
@@ -84,3 +87,65 @@ void arch_pick_mmap_layout(struct mm_struct *mm)
 }
 EXPORT_SYMBOL_GPL(arch_pick_mmap_layout);
 
+#else
+
+static unsigned long
+s390_get_unmapped_area(struct file *filp, unsigned long addr,
+		unsigned long len, unsigned long pgoff, unsigned long flags)
+{
+	struct mm_struct *mm = current->mm;
+	int rc;
+
+	addr = arch_get_unmapped_area(filp, addr, len, pgoff, flags);
+	if (addr & ~PAGE_MASK)
+		return addr;
+	if (unlikely(mm->context.asce_limit < addr + len)) {
+		rc = crst_table_upgrade(mm, addr + len);
+		if (rc)
+			return (unsigned long) rc;
+	}
+	return addr;
+}
+
+static unsigned long
+s390_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
+			  const unsigned long len, const unsigned long pgoff,
+			  const unsigned long flags)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long addr = addr0;
+	int rc;
+
+	addr = arch_get_unmapped_area_topdown(filp, addr, len, pgoff, flags);
+	if (addr & ~PAGE_MASK)
+		return addr;
+	if (unlikely(mm->context.asce_limit < addr + len)) {
+		rc = crst_table_upgrade(mm, addr + len);
+		if (rc)
+			return (unsigned long) rc;
+	}
+	return addr;
+}
+/*
+ * This function, called very early during the creation of a new
+ * process VM image, sets up which VM layout function to use:
+ */
+void arch_pick_mmap_layout(struct mm_struct *mm)
+{
+	/*
+	 * Fall back to the standard layout if the personality
+	 * bit is set, or if the expected stack growth is unlimited:
+	 */
+	if (mmap_is_legacy()) {
+		mm->mmap_base = TASK_UNMAPPED_BASE;
+		mm->get_unmapped_area = s390_get_unmapped_area;
+		mm->unmap_area = arch_unmap_area;
+	} else {
+		mm->mmap_base = mmap_base();
+		mm->get_unmapped_area = s390_get_unmapped_area_topdown;
+		mm->unmap_area = arch_unmap_area_topdown;
+	}
+}
+EXPORT_SYMBOL_GPL(arch_pick_mmap_layout);
+
+#endif
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index 809e77893039..fd072013f88c 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -23,6 +23,7 @@
 #include <asm/pgalloc.h>
 #include <asm/tlb.h>
 #include <asm/tlbflush.h>
+#include <asm/mmu_context.h>
 
 #ifndef CONFIG_64BIT
 #define ALLOC_ORDER	1
@@ -70,6 +71,79 @@ void crst_table_free(struct mm_struct *mm, unsigned long *table)
 	free_pages((unsigned long) table, ALLOC_ORDER);
 }
 
+#ifdef CONFIG_64BIT
+int crst_table_upgrade(struct mm_struct *mm, unsigned long limit)
+{
+	unsigned long *table, *pgd;
+	unsigned long entry;
+
+	BUG_ON(limit > (1UL << 53));
+repeat:
+	table = crst_table_alloc(mm, mm->context.noexec);
+	if (!table)
+		return -ENOMEM;
+	spin_lock(&mm->page_table_lock);
+	if (mm->context.asce_limit < limit) {
+		pgd = (unsigned long *) mm->pgd;
+		if (mm->context.asce_limit <= (1UL << 31)) {
+			entry = _REGION3_ENTRY_EMPTY;
+			mm->context.asce_limit = 1UL << 42;
+			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
+						_ASCE_USER_BITS |
+						_ASCE_TYPE_REGION3;
+		} else {
+			entry = _REGION2_ENTRY_EMPTY;
+			mm->context.asce_limit = 1UL << 53;
+			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
+						_ASCE_USER_BITS |
+						_ASCE_TYPE_REGION2;
+		}
+		crst_table_init(table, entry);
+		pgd_populate(mm, (pgd_t *) table, (pud_t *) pgd);
+		mm->pgd = (pgd_t *) table;
+		table = NULL;
+	}
+	spin_unlock(&mm->page_table_lock);
+	if (table)
+		crst_table_free(mm, table);
+	if (mm->context.asce_limit < limit)
+		goto repeat;
+	update_mm(mm, current);
+	return 0;
+}
+
+void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
+{
+	pgd_t *pgd;
+
+	if (mm->context.asce_limit <= limit)
+		return;
+	__tlb_flush_mm(mm);
+	while (mm->context.asce_limit > limit) {
+		pgd = mm->pgd;
+		switch (pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) {
+		case _REGION_ENTRY_TYPE_R2:
+			mm->context.asce_limit = 1UL << 42;
+			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
+						_ASCE_USER_BITS |
+						_ASCE_TYPE_REGION3;
+			break;
+		case _REGION_ENTRY_TYPE_R3:
+			mm->context.asce_limit = 1UL << 31;
+			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
+						_ASCE_USER_BITS |
+						_ASCE_TYPE_SEGMENT;
+			break;
+		default:
+			BUG();
+		}
+		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
+		crst_table_free(mm, (unsigned long *) pgd);
+	}
+	update_mm(mm, current);
+}
+#endif
+
 /*
  * page table entry allocation/free routines.
  */