7 files changed, 398 insertions, 34 deletions

diff --git a/arch/s390/mm/Makefile b/arch/s390/mm/Makefile
index 0f5536b0c1a..1bea6d1f55a 100644
--- a/arch/s390/mm/Makefile
+++ b/arch/s390/mm/Makefile
@@ -7,3 +7,4 @@ obj-y	 := init.o fault.o extmem.o mmap.o vmem.o pgtable.o maccess.o \
 obj-$(CONFIG_CMM) += cmm.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
 obj-$(CONFIG_DEBUG_SET_MODULE_RONX) += pageattr.o
+obj-$(CONFIG_S390_PTDUMP) += dump_pagetables.o
diff --git a/arch/s390/mm/dump_pagetables.c b/arch/s390/mm/dump_pagetables.c
new file mode 100644
index 00000000000..cbc6668acb8
--- /dev/null
+++ b/arch/s390/mm/dump_pagetables.c
@@ -0,0 +1,226 @@
+#include <linux/seq_file.h>
+#include <linux/debugfs.h>
+#include <linux/module.h>
+#include <linux/mm.h>
+#include <asm/sections.h>
+#include <asm/pgtable.h>
+
+static unsigned long max_addr;
+
+struct addr_marker {
+	unsigned long start_address;
+	const char *name;
+};
+
+enum address_markers_idx {
+	IDENTITY_NR = 0,
+	KERNEL_START_NR,
+	KERNEL_END_NR,
+	VMEMMAP_NR,
+	VMALLOC_NR,
+#ifdef CONFIG_64BIT
+	MODULES_NR,
+#endif
+};
+
+static struct addr_marker address_markers[] = {
+	[IDENTITY_NR]	  = {0, "Identity Mapping"},
+	[KERNEL_START_NR] = {(unsigned long)&_stext, "Kernel Image Start"},
+	[KERNEL_END_NR]	  = {(unsigned long)&_end, "Kernel Image End"},
+	[VMEMMAP_NR]	  = {0, "vmemmap Area"},
+	[VMALLOC_NR]	  = {0, "vmalloc Area"},
+#ifdef CONFIG_64BIT
+	[MODULES_NR]	  = {0, "Modules Area"},
+#endif
+	{ -1, NULL }
+};
+
+struct pg_state {
+	int level;
+	unsigned int current_prot;
+	unsigned long start_address;
+	unsigned long current_address;
+	const struct addr_marker *marker;
+};
+
+static void print_prot(struct seq_file *m, unsigned int pr, int level)
+{
+	static const char * const level_name[] =
+		{ "ASCE", "PGD", "PUD", "PMD", "PTE" };
+
+	seq_printf(m, "%s ", level_name[level]);
+	if (pr & _PAGE_INVALID)
+		seq_printf(m, "I\n");
+	else
+		seq_printf(m, "%s\n", pr & _PAGE_RO ? "RO" : "RW");
+}
+
+static void note_page(struct seq_file *m, struct pg_state *st,
+		     unsigned int new_prot, int level)
+{
+	static const char units[] = "KMGTPE";
+	int width = sizeof(unsigned long) * 2;
+	const char *unit = units;
+	unsigned int prot, cur;
+	unsigned long delta;
+
+	/*
+	 * If we have a "break" in the series, we need to flush the state
+	 * that we have now. "break" is either changing perms, levels or
+	 * address space marker.
+	 */
+	prot = new_prot;
+	cur = st->current_prot;
+
+	if (!st->level) {
+		/* First entry */
+		st->current_prot = new_prot;
+		st->level = level;
+		st->marker = address_markers;
+		seq_printf(m, "---[ %s ]---\n", st->marker->name);
+	} else if (prot != cur || level != st->level ||
+		   st->current_address >= st->marker[1].start_address) {
+		/* Print the actual finished series */
+		seq_printf(m, "0x%0*lx-0x%0*lx",
+			   width, st->start_address,
+			   width, st->current_address);
+		delta = (st->current_address - st->start_address) >> 10;
+		while (!(delta & 0x3ff) && unit[1]) {
+			delta >>= 10;
+			unit++;
+		}
+		seq_printf(m, "%9lu%c ", delta, *unit);
+		print_prot(m, st->current_prot, st->level);
+		if (st->current_address >= st->marker[1].start_address) {
+			st->marker++;
+			seq_printf(m, "---[ %s ]---\n", st->marker->name);
+		}
+		st->start_address = st->current_address;
+		st->current_prot = new_prot;
+		st->level = level;
+	}
+}
+
+/*
+ * The actual page table walker functions. In order to keep the implementation
+ * of print_prot() short, we only check and pass _PAGE_INVALID and _PAGE_RO
+ * flags to note_page() if a region, segment or page table entry is invalid or
+ * read-only.
+ * After all it's just a hint that the current level being walked contains an
+ * invalid or read-only entry.
+ */
+static void walk_pte_level(struct seq_file *m, struct pg_state *st,
+			   pmd_t *pmd, unsigned long addr)
+{
+	unsigned int prot;
+	pte_t *pte;
+	int i;
+
+	for (i = 0; i < PTRS_PER_PTE && addr < max_addr; i++) {
+		st->current_address = addr;
+		pte = pte_offset_kernel(pmd, addr);
+		prot = pte_val(*pte) & (_PAGE_RO | _PAGE_INVALID);
+		note_page(m, st, prot, 4);
+		addr += PAGE_SIZE;
+	}
+}
+
+static void walk_pmd_level(struct seq_file *m, struct pg_state *st,
+			   pud_t *pud, unsigned long addr)
+{
+	unsigned int prot;
+	pmd_t *pmd;
+	int i;
+
+	for (i = 0; i < PTRS_PER_PMD && addr < max_addr; i++) {
+		st->current_address = addr;
+		pmd = pmd_offset(pud, addr);
+		if (!pmd_none(*pmd)) {
+			if (pmd_large(*pmd)) {
+				prot = pmd_val(*pmd) & _SEGMENT_ENTRY_RO;
+				note_page(m, st, prot, 3);
+			} else
+				walk_pte_level(m, st, pmd, addr);
+		} else
+			note_page(m, st, _PAGE_INVALID, 3);
+		addr += PMD_SIZE;
+	}
+}
+
+static void walk_pud_level(struct seq_file *m, struct pg_state *st,
+			   pgd_t *pgd, unsigned long addr)
+{
+	pud_t *pud;
+	int i;
+
+	for (i = 0; i < PTRS_PER_PUD && addr < max_addr; i++) {
+		st->current_address = addr;
+		pud = pud_offset(pgd, addr);
+		if (!pud_none(*pud))
+			walk_pmd_level(m, st, pud, addr);
+		else
+			note_page(m, st, _PAGE_INVALID, 2);
+		addr += PUD_SIZE;
+	}
+}
+
+static void walk_pgd_level(struct seq_file *m)
+{
+	unsigned long addr = 0;
+	struct pg_state st;
+	pgd_t *pgd;
+	int i;
+
+	memset(&st, 0, sizeof(st));
+	for (i = 0; i < PTRS_PER_PGD && addr < max_addr; i++) {
+		st.current_address = addr;
+		pgd = pgd_offset_k(addr);
+		if (!pgd_none(*pgd))
+			walk_pud_level(m, &st, pgd, addr);
+		else
+			note_page(m, &st, _PAGE_INVALID, 1);
+		addr += PGDIR_SIZE;
+	}
+	/* Flush out the last page */
+	st.current_address = max_addr;
+	note_page(m, &st, 0, 0);
+}
+
+static int ptdump_show(struct seq_file *m, void *v)
+{
+	walk_pgd_level(m);
+	return 0;
+}
+
+static int ptdump_open(struct inode *inode, struct file *filp)
+{
+	return single_open(filp, ptdump_show, NULL);
+}
+
+static const struct file_operations ptdump_fops = {
+	.open		= ptdump_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int pt_dump_init(void)
+{
+	/*
+	 * Figure out the maximum virtual address being accessible with the
+	 * kernel ASCE. We need this to keep the page table walker functions
+	 * from accessing non-existent entries.
+	 */
+#ifdef CONFIG_32BIT
+	max_addr = 1UL << 31;
+#else
+	max_addr = (S390_lowcore.kernel_asce & _REGION_ENTRY_TYPE_MASK) >> 2;
+	max_addr = 1UL << (max_addr * 11 + 31);
+	address_markers[MODULES_NR].start_address = MODULES_VADDR;
+#endif
+	address_markers[VMEMMAP_NR].start_address = (unsigned long) vmemmap;
+	address_markers[VMALLOC_NR].start_address = VMALLOC_START;
+	debugfs_create_file("kernel_page_tables", 0400, NULL, NULL, &ptdump_fops);
+	return 0;
+}
+device_initcall(pt_dump_init);
diff --git a/arch/s390/mm/fault.c b/arch/s390/mm/fault.c
index ac9122ca115..04ad4001a28 100644
--- a/arch/s390/mm/fault.c
+++ b/arch/s390/mm/fault.c
@@ -367,6 +367,7 @@ retry:
 			/* Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk
 			 * of starvation. */
 			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+			flags |= FAULT_FLAG_TRIED;
 			down_read(&mm->mmap_sem);
 			goto retry;
 		}
diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c
index eeaf8023851..60acb93a468 100644
--- a/arch/s390/mm/gup.c
+++ b/arch/s390/mm/gup.c
@@ -115,7 +115,16 @@ static inline int gup_pmd_range(pud_t *pudp, pud_t pud, unsigned long addr,
 		pmd = *pmdp;
 		barrier();
 		next = pmd_addr_end(addr, end);
-		if (pmd_none(pmd))
+		/*
+		 * The pmd_trans_splitting() check below explains why
+		 * pmdp_splitting_flush() has to serialize with
+		 * smp_call_function() against our disabled IRQs, to stop
+		 * this gup-fast code from running while we set the
+		 * splitting bit in the pmd. Returning zero will take
+		 * the slow path that will call wait_split_huge_page()
+		 * if the pmd is still in splitting state.
+		 */
+		if (pmd_none(pmd) || pmd_trans_splitting(pmd))
 			return 0;
 		if (unlikely(pmd_huge(pmd))) {
 			if (!gup_huge_pmd(pmdp, pmd, addr, next,
diff --git a/arch/s390/mm/pageattr.c b/arch/s390/mm/pageattr.c
index b36537a5f43..00be01c4b4f 100644
--- a/arch/s390/mm/pageattr.c
+++ b/arch/s390/mm/pageattr.c
@@ -8,25 +8,38 @@
 #include <asm/cacheflush.h>
 #include <asm/pgtable.h>
 
+static pte_t *walk_page_table(unsigned long addr)
+{
+	pgd_t *pgdp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep;
+
+	pgdp = pgd_offset_k(addr);
+	if (pgd_none(*pgdp))
+		return NULL;
+	pudp = pud_offset(pgdp, addr);
+	if (pud_none(*pudp))
+		return NULL;
+	pmdp = pmd_offset(pudp, addr);
+	if (pmd_none(*pmdp) || pmd_large(*pmdp))
+		return NULL;
+	ptep = pte_offset_kernel(pmdp, addr);
+	if (pte_none(*ptep))
+		return NULL;
+	return ptep;
+}
+
 static void change_page_attr(unsigned long addr, int numpages,
 			     pte_t (*set) (pte_t))
 {
 	pte_t *ptep, pte;
-	pmd_t *pmdp;
-	pud_t *pudp;
-	pgd_t *pgdp;
 	int i;
 
 	for (i = 0; i < numpages; i++) {
-		pgdp = pgd_offset(&init_mm, addr);
-		pudp = pud_offset(pgdp, addr);
-		pmdp = pmd_offset(pudp, addr);
-		if (pmd_huge(*pmdp)) {
-			WARN_ON_ONCE(1);
-			continue;
-		}
-		ptep = pte_offset_kernel(pmdp, addr);
-
+		ptep = walk_page_table(addr);
+		if (WARN_ON_ONCE(!ptep))
+			break;
 		pte = *ptep;
 		pte = set(pte);
 		__ptep_ipte(addr, ptep);
@@ -40,21 +53,18 @@ int set_memory_ro(unsigned long addr, int numpages)
 	change_page_attr(addr, numpages, pte_wrprotect);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(set_memory_ro);
 
 int set_memory_rw(unsigned long addr, int numpages)
 {
 	change_page_attr(addr, numpages, pte_mkwrite);
 	return 0;
 }
-EXPORT_SYMBOL_GPL(set_memory_rw);
 
 /* not possible */
 int set_memory_nx(unsigned long addr, int numpages)
 {
 	return 0;
 }
-EXPORT_SYMBOL_GPL(set_memory_nx);
 
 int set_memory_x(unsigned long addr, int numpages)
 {
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index b402991e43d..c8188a18af0 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -787,6 +787,30 @@ void tlb_remove_table(struct mmu_gather *tlb, void *table)
 		tlb_table_flush(tlb);
 }
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+void thp_split_vma(struct vm_area_struct *vma)
+{
+	unsigned long addr;
+	struct page *page;
+
+	for (addr = vma->vm_start; addr < vma->vm_end; addr += PAGE_SIZE) {
+		page = follow_page(vma, addr, FOLL_SPLIT);
+	}
+}
+
+void thp_split_mm(struct mm_struct *mm)
+{
+	struct vm_area_struct *vma = mm->mmap;
+
+	while (vma != NULL) {
+		thp_split_vma(vma);
+		vma->vm_flags &= ~VM_HUGEPAGE;
+		vma->vm_flags |= VM_NOHUGEPAGE;
+		vma = vma->vm_next;
+	}
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
+
 /*
  * switch on pgstes for its userspace process (for kvm)
  */
@@ -824,6 +848,12 @@ int s390_enable_sie(void)
 	if (!mm)
 		return -ENOMEM;
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	/* split thp mappings and disable thp for future mappings */
+	thp_split_mm(mm);
+	mm->def_flags |= VM_NOHUGEPAGE;
+#endif
+
 	/* Now lets check again if something happened */
 	task_lock(tsk);
 	if (!tsk->mm || atomic_read(&tsk->mm->mm_users) > 1 ||
@@ -866,3 +896,81 @@ bool kernel_page_present(struct page *page)
 	return cc == 0;
 }
 #endif /* CONFIG_HIBERNATION && CONFIG_DEBUG_PAGEALLOC */
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+int pmdp_clear_flush_young(struct vm_area_struct *vma, unsigned long address,
+			   pmd_t *pmdp)
+{
+	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
+	/* No need to flush TLB
+	 * On s390 reference bits are in storage key and never in TLB */
+	return pmdp_test_and_clear_young(vma, address, pmdp);
+}
+
+int pmdp_set_access_flags(struct vm_area_struct *vma,
+			  unsigned long address, pmd_t *pmdp,
+			  pmd_t entry, int dirty)
+{
+	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
+
+	if (pmd_same(*pmdp, entry))
+		return 0;
+	pmdp_invalidate(vma, address, pmdp);
+	set_pmd_at(vma->vm_mm, address, pmdp, entry);
+	return 1;
+}
+
+static void pmdp_splitting_flush_sync(void *arg)
+{
+	/* Simply deliver the interrupt */
+}
+
+void pmdp_splitting_flush(struct vm_area_struct *vma, unsigned long address,
+			  pmd_t *pmdp)
+{
+	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
+	if (!test_and_set_bit(_SEGMENT_ENTRY_SPLIT_BIT,
+			      (unsigned long *) pmdp)) {
+		/* need to serialize against gup-fast (IRQ disabled) */
+		smp_call_function(pmdp_splitting_flush_sync, NULL, 1);
+	}
+}
+
+void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable)
+{
+	struct list_head *lh = (struct list_head *) pgtable;
+
+	assert_spin_locked(&mm->page_table_lock);
+
+	/* FIFO */
+	if (!mm->pmd_huge_pte)
+		INIT_LIST_HEAD(lh);
+	else
+		list_add(lh, (struct list_head *) mm->pmd_huge_pte);
+	mm->pmd_huge_pte = pgtable;
+}
+
+pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm)
+{
+	struct list_head *lh;
+	pgtable_t pgtable;
+	pte_t *ptep;
+
+	assert_spin_locked(&mm->page_table_lock);
+
+	/* FIFO */
+	pgtable = mm->pmd_huge_pte;
+	lh = (struct list_head *) pgtable;
+	if (list_empty(lh))
+		mm->pmd_huge_pte = NULL;
+	else {
+		mm->pmd_huge_pte = (pgtable_t) lh->next;
+		list_del(lh);
+	}
+	ptep = (pte_t *) pgtable;
+	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
+	ptep++;
+	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
+	return pgtable;
+}
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
diff --git a/arch/s390/mm/vmem.c b/arch/s390/mm/vmem.c
index c22abf900c9..387c7c60b5b 100644
--- a/arch/s390/mm/vmem.c
+++ b/arch/s390/mm/vmem.c
@@ -79,7 +79,8 @@ static pte_t __ref *vmem_pte_alloc(unsigned long address)
  */
 static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 {
-	unsigned long address;
+	unsigned long end = start + size;
+	unsigned long address = start;
 	pgd_t *pg_dir;
 	pud_t *pu_dir;
 	pmd_t *pm_dir;
@@ -87,7 +88,7 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 	pte_t  pte;
 	int ret = -ENOMEM;
 
-	for (address = start; address < start + size; address += PAGE_SIZE) {
+	while (address < end) {
 		pg_dir = pgd_offset_k(address);
 		if (pgd_none(*pg_dir)) {
 			pu_dir = vmem_pud_alloc();
@@ -108,12 +109,11 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 		pm_dir = pmd_offset(pu_dir, address);
 
 #if defined(CONFIG_64BIT) && !defined(CONFIG_DEBUG_PAGEALLOC)
-		if (MACHINE_HAS_HPAGE && !(address & ~HPAGE_MASK) &&
-		    (address + HPAGE_SIZE <= start + size) &&
-		    (address >= HPAGE_SIZE)) {
+		if (MACHINE_HAS_EDAT1 && pmd_none(*pm_dir) && address &&
+		    !(address & ~PMD_MASK) && (address + PMD_SIZE <= end)) {
 			pte_val(pte) |= _SEGMENT_ENTRY_LARGE;
 			pmd_val(*pm_dir) = pte_val(pte);
-			address += HPAGE_SIZE - PAGE_SIZE;
+			address += PMD_SIZE;
 			continue;
 		}
 #endif
@@ -126,10 +126,11 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 
 		pt_dir = pte_offset_kernel(pm_dir, address);
 		*pt_dir = pte;
+		address += PAGE_SIZE;
 	}
 	ret = 0;
 out:
-	flush_tlb_kernel_range(start, start + size);
+	flush_tlb_kernel_range(start, end);
 	return ret;
 }
 
@@ -139,7 +140,8 @@ out:
  */
 static void vmem_remove_range(unsigned long start, unsigned long size)
 {
-	unsigned long address;
+	unsigned long end = start + size;
+	unsigned long address = start;
 	pgd_t *pg_dir;
 	pud_t *pu_dir;
 	pmd_t *pm_dir;
@@ -147,25 +149,32 @@ static void vmem_remove_range(unsigned long start, unsigned long size)
 	pte_t  pte;
 
 	pte_val(pte) = _PAGE_TYPE_EMPTY;
-	for (address = start; address < start + size; address += PAGE_SIZE) {
+	while (address < end) {
 		pg_dir = pgd_offset_k(address);
+		if (pgd_none(*pg_dir)) {
+			address += PGDIR_SIZE;
+			continue;
+		}
 		pu_dir = pud_offset(pg_dir, address);
-		if (pud_none(*pu_dir))
+		if (pud_none(*pu_dir)) {
+			address += PUD_SIZE;
 			continue;
+		}
 		pm_dir = pmd_offset(pu_dir, address);
-		if (pmd_none(*pm_dir))
+		if (pmd_none(*pm_dir)) {
+			address += PMD_SIZE;
 			continue;
-
-		if (pmd_huge(*pm_dir)) {
+		}
+		if (pmd_large(*pm_dir)) {
 			pmd_clear(pm_dir);
-			address += HPAGE_SIZE - PAGE_SIZE;
+			address += PMD_SIZE;
 			continue;
 		}
-
 		pt_dir = pte_offset_kernel(pm_dir, address);
 		*pt_dir = pte;
+		address += PAGE_SIZE;
 	}
-	flush_tlb_kernel_range(start, start + size);
+	flush_tlb_kernel_range(start, end);
 }
 
 /*
@@ -330,8 +339,8 @@ void __init vmem_map_init(void)
 	unsigned long start, end;
 	int i;
 
-	ro_start = ((unsigned long)&_stext) & PAGE_MASK;
-	ro_end = PFN_ALIGN((unsigned long)&_eshared);
+	ro_start = PFN_ALIGN((unsigned long)&_stext);
+	ro_end = (unsigned long)&_eshared & PAGE_MASK;
 	for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 		if (memory_chunk[i].type == CHUNK_CRASHK ||
 		    memory_chunk[i].type == CHUNK_OLDMEM)