s390/kvm: remove delayed reallocation of page tables for KVM

Replacing a 2K page table with a 4K page table while a VMA is active
for the affected memory region is fundamentally broken. Rip out the
page table reallocation code and replace it with a simple system
control 'vm.allocate_pgste'. If the system control is set the page
tables for all processes are allocated as full 4K pages, even for
processes that do not need it.

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

5 files changed, 59 insertions, 100 deletions

diff --git a/arch/s390/include/asm/mmu.h b/arch/s390/include/asm/mmu.h
index a5e656260a70..d29ad9545b41 100644
--- a/arch/s390/include/asm/mmu.h
+++ b/arch/s390/include/asm/mmu.h
@@ -14,7 +14,9 @@ typedef struct {
 	unsigned long asce_bits;
 	unsigned long asce_limit;
 	unsigned long vdso_base;
-	/* The mmu context has extended page tables. */
+	/* The mmu context allocates 4K page tables. */
+	unsigned int alloc_pgste:1;
+	/* The mmu context uses extended page tables. */
 	unsigned int has_pgste:1;
 	/* The mmu context uses storage keys. */
 	unsigned int use_skey:1;
diff --git a/arch/s390/include/asm/mmu_context.h b/arch/s390/include/asm/mmu_context.h
index d25d9ff10ba8..fb1b93ea3e3f 100644
--- a/arch/s390/include/asm/mmu_context.h
+++ b/arch/s390/include/asm/mmu_context.h
@@ -20,8 +20,11 @@ static inline int init_new_context(struct task_struct *tsk,
 	mm->context.flush_mm = 0;
 	mm->context.asce_bits = _ASCE_TABLE_LENGTH | _ASCE_USER_BITS;
 	mm->context.asce_bits |= _ASCE_TYPE_REGION3;
+#ifdef CONFIG_PGSTE
+	mm->context.alloc_pgste = page_table_allocate_pgste;
 	mm->context.has_pgste = 0;
 	mm->context.use_skey = 0;
+#endif
 	mm->context.asce_limit = STACK_TOP_MAX;
 	crst_table_init((unsigned long *) mm->pgd, pgd_entry_type(mm));
 	return 0;
diff --git a/arch/s390/include/asm/pgalloc.h b/arch/s390/include/asm/pgalloc.h
index 51e7fb634ebc..7b7858f158b4 100644
--- a/arch/s390/include/asm/pgalloc.h
+++ b/arch/s390/include/asm/pgalloc.h
@@ -21,6 +21,7 @@ void crst_table_free(struct mm_struct *, unsigned long *);
 unsigned long *page_table_alloc(struct mm_struct *);
 void page_table_free(struct mm_struct *, unsigned long *);
 void page_table_free_rcu(struct mmu_gather *, unsigned long *, unsigned long);
+extern int page_table_allocate_pgste;
 
 int set_guest_storage_key(struct mm_struct *mm, unsigned long addr,
 			  unsigned long key, bool nq);
diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index 989cfae9e202..1fba63997d50 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -423,6 +423,15 @@ static inline int mm_has_pgste(struct mm_struct *mm)
 	return 0;
 }
 
+static inline int mm_alloc_pgste(struct mm_struct *mm)
+{
+#ifdef CONFIG_PGSTE
+	if (unlikely(mm->context.alloc_pgste))
+		return 1;
+#endif
+	return 0;
+}
+
 /*
  * In the case that a guest uses storage keys
  * faults should no longer be backed by zero pages
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index 33f589459113..b33f66110ca9 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -18,6 +18,7 @@
 #include <linux/rcupdate.h>
 #include <linux/slab.h>
 #include <linux/swapops.h>
+#include <linux/sysctl.h>
 #include <linux/ksm.h>
 #include <linux/mman.h>
 
@@ -920,6 +921,40 @@ unsigned long get_guest_storage_key(struct mm_struct *mm, unsigned long addr)
 }
 EXPORT_SYMBOL(get_guest_storage_key);
 
+static int page_table_allocate_pgste_min = 0;
+static int page_table_allocate_pgste_max = 1;
+int page_table_allocate_pgste = 0;
+EXPORT_SYMBOL(page_table_allocate_pgste);
+
+static struct ctl_table page_table_sysctl[] = {
+	{
+		.procname	= "allocate_pgste",
+		.data		= &page_table_allocate_pgste,
+		.maxlen		= sizeof(int),
+		.mode		= S_IRUGO | S_IWUSR,
+		.proc_handler	= proc_dointvec,
+		.extra1		= &page_table_allocate_pgste_min,
+		.extra2		= &page_table_allocate_pgste_max,
+	},
+	{ }
+};
+
+static struct ctl_table page_table_sysctl_dir[] = {
+	{
+		.procname	= "vm",
+		.maxlen		= 0,
+		.mode		= 0555,
+		.child		= page_table_sysctl,
+	},
+	{ }
+};
+
+static int __init page_table_register_sysctl(void)
+{
+	return register_sysctl_table(page_table_sysctl_dir) ? 0 : -ENOMEM;
+}
+__initcall(page_table_register_sysctl);
+
 #else /* CONFIG_PGSTE */
 
 static inline int page_table_with_pgste(struct page *page)
@@ -963,7 +998,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 	struct page *uninitialized_var(page);
 	unsigned int mask, bit;
 
-	if (mm_has_pgste(mm))
+	if (mm_alloc_pgste(mm))
 		return page_table_alloc_pgste(mm);
 	/* Allocate fragments of a 4K page as 1K/2K page table */
 	spin_lock_bh(&mm->context.list_lock);
@@ -1165,116 +1200,25 @@ static inline void thp_split_mm(struct mm_struct *mm)
 }
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
 
-static unsigned long page_table_realloc_pmd(struct mmu_gather *tlb,
-				struct mm_struct *mm, pud_t *pud,
-				unsigned long addr, unsigned long end)
-{
-	unsigned long next, *table, *new;
-	struct page *page;
-	spinlock_t *ptl;
-	pmd_t *pmd;
-
-	pmd = pmd_offset(pud, addr);
-	do {
-		next = pmd_addr_end(addr, end);
-again:
-		if (pmd_none_or_clear_bad(pmd))
-			continue;
-		table = (unsigned long *) pmd_deref(*pmd);
-		page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-		if (page_table_with_pgste(page))
-			continue;
-		/* Allocate new page table with pgstes */
-		new = page_table_alloc_pgste(mm);
-		if (!new)
-			return -ENOMEM;
-
-		ptl = pmd_lock(mm, pmd);
-		if (likely((unsigned long *) pmd_deref(*pmd) == table)) {
-			/* Nuke pmd entry pointing to the "short" page table */
-			pmdp_flush_lazy(mm, addr, pmd);
-			pmd_clear(pmd);
-			/* Copy ptes from old table to new table */
-			memcpy(new, table, PAGE_SIZE/2);
-			clear_table(table, _PAGE_INVALID, PAGE_SIZE/2);
-			/* Establish new table */
-			pmd_populate(mm, pmd, (pte_t *) new);
-			/* Free old table with rcu, there might be a walker! */
-			page_table_free_rcu(tlb, table, addr);
-			new = NULL;
-		}
-		spin_unlock(ptl);
-		if (new) {
-			page_table_free_pgste(new);
-			goto again;
-		}
-	} while (pmd++, addr = next, addr != end);
-
-	return addr;
-}
-
-static unsigned long page_table_realloc_pud(struct mmu_gather *tlb,
-				   struct mm_struct *mm, pgd_t *pgd,
-				   unsigned long addr, unsigned long end)
-{
-	unsigned long next;
-	pud_t *pud;
-
-	pud = pud_offset(pgd, addr);
-	do {
-		next = pud_addr_end(addr, end);
-		if (pud_none_or_clear_bad(pud))
-			continue;
-		next = page_table_realloc_pmd(tlb, mm, pud, addr, next);
-		if (unlikely(IS_ERR_VALUE(next)))
-			return next;
-	} while (pud++, addr = next, addr != end);
-
-	return addr;
-}
-
-static unsigned long page_table_realloc(struct mmu_gather *tlb, struct mm_struct *mm,
-					unsigned long addr, unsigned long end)
-{
-	unsigned long next;
-	pgd_t *pgd;
-
-	pgd = pgd_offset(mm, addr);
-	do {
-		next = pgd_addr_end(addr, end);
-		if (pgd_none_or_clear_bad(pgd))
-			continue;
-		next = page_table_realloc_pud(tlb, mm, pgd, addr, next);
-		if (unlikely(IS_ERR_VALUE(next)))
-			return next;
-	} while (pgd++, addr = next, addr != end);
-
-	return 0;
-}
-
 /*
  * switch on pgstes for its userspace process (for kvm)
  */
 int s390_enable_sie(void)
 {
-	struct task_struct *tsk = current;
-	struct mm_struct *mm = tsk->mm;
-	struct mmu_gather tlb;
+	struct mm_struct *mm = current->mm;
 
 	/* Do we have pgstes? if yes, we are done */
-	if (mm_has_pgste(tsk->mm))
+	if (mm_has_pgste(mm))
 		return 0;
-
+	/* Fail if the page tables are 2K */
+	if (!mm_alloc_pgste(mm))
+		return -EINVAL;
 	down_write(&mm->mmap_sem);
+	mm->context.has_pgste = 1;
 	/* split thp mappings and disable thp for future mappings */
 	thp_split_mm(mm);
-	/* Reallocate the page tables with pgstes */
-	tlb_gather_mmu(&tlb, mm, 0, TASK_SIZE);
-	if (!page_table_realloc(&tlb, mm, 0, TASK_SIZE))
-		mm->context.has_pgste = 1;
-	tlb_finish_mmu(&tlb, 0, TASK_SIZE);
 	up_write(&mm->mmap_sem);
-	return mm->context.has_pgste ? 0 : -ENOMEM;
+	return 0;
 }
 EXPORT_SYMBOL_GPL(s390_enable_sie);