Merge branch 'tlb-flushing' into features

Add the TLB flushing changes via a tip branch to ease merging with
the KVM tree.

5 files changed, 303 insertions, 49 deletions

diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
index 8111694ce55a..3aee54b2ba60 100644
--- a/arch/s390/mm/init.c
+++ b/arch/s390/mm/init.c
@@ -137,6 +137,8 @@ void __init mem_init(void)
 	free_all_bootmem();
 	setup_zero_pages();	/* Setup zeroed pages. */
 
+	cmma_init_nodat();
+
 	mem_init_print_info(NULL);
 }
 
diff --git a/arch/s390/mm/page-states.c b/arch/s390/mm/page-states.c
index 69a7b01ae746..07fa7b8ae233 100644
--- a/arch/s390/mm/page-states.c
+++ b/arch/s390/mm/page-states.c
@@ -10,9 +10,10 @@
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/mm.h>
+#include <linux/memblock.h>
 #include <linux/gfp.h>
 #include <linux/init.h>
-
+#include <asm/facility.h>
 #include <asm/page-states.h>
 
 static int cmma_flag = 1;
@@ -36,14 +37,16 @@ __setup("cmma=", cmma);
 static inline int cmma_test_essa(void)
 {
 	register unsigned long tmp asm("0") = 0;
-	register int rc asm("1") = -EOPNOTSUPP;
+	register int rc asm("1");
 
+	/* test ESSA_GET_STATE */
 	asm volatile(
-		"       .insn rrf,0xb9ab0000,%1,%1,0,0\n"
+		"	.insn	rrf,0xb9ab0000,%1,%1,%2,0\n"
 		"0:     la      %0,0\n"
 		"1:\n"
 		EX_TABLE(0b,1b)
-		: "+&d" (rc), "+&d" (tmp));
+		: "=&d" (rc), "+&d" (tmp)
+		: "i" (ESSA_GET_STATE), "0" (-EOPNOTSUPP));
 	return rc;
 }
 
@@ -51,11 +54,26 @@ void __init cmma_init(void)
 {
 	if (!cmma_flag)
 		return;
-	if (cmma_test_essa())
+	if (cmma_test_essa()) {
 		cmma_flag = 0;
+		return;
+	}
+	if (test_facility(147))
+		cmma_flag = 2;
 }
 
-static inline void set_page_unstable(struct page *page, int order)
+static inline unsigned char get_page_state(struct page *page)
+{
+	unsigned char state;
+
+	asm volatile("	.insn	rrf,0xb9ab0000,%0,%1,%2,0"
+		     : "=&d" (state)
+		     : "a" (page_to_phys(page)),
+		       "i" (ESSA_GET_STATE));
+	return state & 0x3f;
+}
+
+static inline void set_page_unused(struct page *page, int order)
 {
 	int i, rc;
 
@@ -66,14 +84,18 @@ static inline void set_page_unstable(struct page *page, int order)
 			       "i" (ESSA_SET_UNUSED));
 }
 
-void arch_free_page(struct page *page, int order)
+static inline void set_page_stable_dat(struct page *page, int order)
 {
-	if (!cmma_flag)
-		return;
-	set_page_unstable(page, order);
+	int i, rc;
+
+	for (i = 0; i < (1 << order); i++)
+		asm volatile(".insn rrf,0xb9ab0000,%0,%1,%2,0"
+			     : "=&d" (rc)
+			     : "a" (page_to_phys(page + i)),
+			       "i" (ESSA_SET_STABLE));
 }
 
-static inline void set_page_stable(struct page *page, int order)
+static inline void set_page_stable_nodat(struct page *page, int order)
 {
 	int i, rc;
 
@@ -81,14 +103,154 @@ static inline void set_page_stable(struct page *page, int order)
 		asm volatile(".insn rrf,0xb9ab0000,%0,%1,%2,0"
 			     : "=&d" (rc)
 			     : "a" (page_to_phys(page + i)),
-			       "i" (ESSA_SET_STABLE));
+			       "i" (ESSA_SET_STABLE_NODAT));
+}
+
+static void mark_kernel_pmd(pud_t *pud, unsigned long addr, unsigned long end)
+{
+	unsigned long next;
+	struct page *page;
+	pmd_t *pmd;
+
+	pmd = pmd_offset(pud, addr);
+	do {
+		next = pmd_addr_end(addr, end);
+		if (pmd_none(*pmd) || pmd_large(*pmd))
+			continue;
+		page = virt_to_page(pmd_val(*pmd));
+		set_bit(PG_arch_1, &page->flags);
+	} while (pmd++, addr = next, addr != end);
+}
+
+static void mark_kernel_pud(p4d_t *p4d, unsigned long addr, unsigned long end)
+{
+	unsigned long next;
+	struct page *page;
+	pud_t *pud;
+	int i;
+
+	pud = pud_offset(p4d, addr);
+	do {
+		next = pud_addr_end(addr, end);
+		if (pud_none(*pud) || pud_large(*pud))
+			continue;
+		if (!pud_folded(*pud)) {
+			page = virt_to_page(pud_val(*pud));
+			for (i = 0; i < 3; i++)
+				set_bit(PG_arch_1, &page[i].flags);
+		}
+		mark_kernel_pmd(pud, addr, next);
+	} while (pud++, addr = next, addr != end);
+}
+
+static void mark_kernel_p4d(pgd_t *pgd, unsigned long addr, unsigned long end)
+{
+	unsigned long next;
+	struct page *page;
+	p4d_t *p4d;
+	int i;
+
+	p4d = p4d_offset(pgd, addr);
+	do {
+		next = p4d_addr_end(addr, end);
+		if (p4d_none(*p4d))
+			continue;
+		if (!p4d_folded(*p4d)) {
+			page = virt_to_page(p4d_val(*p4d));
+			for (i = 0; i < 3; i++)
+				set_bit(PG_arch_1, &page[i].flags);
+		}
+		mark_kernel_pud(p4d, addr, next);
+	} while (p4d++, addr = next, addr != end);
+}
+
+static void mark_kernel_pgd(void)
+{
+	unsigned long addr, next;
+	struct page *page;
+	pgd_t *pgd;
+	int i;
+
+	addr = 0;
+	pgd = pgd_offset_k(addr);
+	do {
+		next = pgd_addr_end(addr, MODULES_END);
+		if (pgd_none(*pgd))
+			continue;
+		if (!pgd_folded(*pgd)) {
+			page = virt_to_page(pgd_val(*pgd));
+			for (i = 0; i < 3; i++)
+				set_bit(PG_arch_1, &page[i].flags);
+		}
+		mark_kernel_p4d(pgd, addr, next);
+	} while (pgd++, addr = next, addr != MODULES_END);
+}
+
+void __init cmma_init_nodat(void)
+{
+	struct memblock_region *reg;
+	struct page *page;
+	unsigned long start, end, ix;
+
+	if (cmma_flag < 2)
+		return;
+	/* Mark pages used in kernel page tables */
+	mark_kernel_pgd();
+
+	/* Set all kernel pages not used for page tables to stable/no-dat */
+	for_each_memblock(memory, reg) {
+		start = memblock_region_memory_base_pfn(reg);
+		end = memblock_region_memory_end_pfn(reg);
+		page = pfn_to_page(start);
+		for (ix = start; ix < end; ix++, page++) {
+			if (__test_and_clear_bit(PG_arch_1, &page->flags))
+				continue;	/* skip page table pages */
+			if (!list_empty(&page->lru))
+				continue;	/* skip free pages */
+			set_page_stable_nodat(page, 0);
+		}
+	}
+}
+
+void arch_free_page(struct page *page, int order)
+{
+	if (!cmma_flag)
+		return;
+	set_page_unused(page, order);
 }
 
 void arch_alloc_page(struct page *page, int order)
 {
 	if (!cmma_flag)
 		return;
-	set_page_stable(page, order);
+	if (cmma_flag < 2)
+		set_page_stable_dat(page, order);
+	else
+		set_page_stable_nodat(page, order);
+}
+
+void arch_set_page_dat(struct page *page, int order)
+{
+	if (!cmma_flag)
+		return;
+	set_page_stable_dat(page, order);
+}
+
+void arch_set_page_nodat(struct page *page, int order)
+{
+	if (cmma_flag < 2)
+		return;
+	set_page_stable_nodat(page, order);
+}
+
+int arch_test_page_nodat(struct page *page)
+{
+	unsigned char state;
+
+	if (cmma_flag < 2)
+		return 0;
+	state = get_page_state(page);
+	return !!(state & 0x20);
 }
 
 void arch_set_page_states(int make_stable)
@@ -108,9 +270,9 @@ void arch_set_page_states(int make_stable)
 			list_for_each(l, &zone->free_area[order].free_list[t]) {
 				page = list_entry(l, struct page, lru);
 				if (make_stable)
-					set_page_stable(page, order);
+					set_page_stable_dat(page, 0);
 				else
-					set_page_unstable(page, order);
+					set_page_unused(page, order);
 			}
 		}
 		spin_unlock_irqrestore(&zone->lock, flags);
diff --git a/arch/s390/mm/pageattr.c b/arch/s390/mm/pageattr.c
index 180481589246..567fe92e2bb8 100644
--- a/arch/s390/mm/pageattr.c
+++ b/arch/s390/mm/pageattr.c
@@ -328,7 +328,7 @@ static void ipte_range(pte_t *pte, unsigned long address, int nr)
 		return;
 	}
 	for (i = 0; i < nr; i++) {
-		__ptep_ipte(address, pte, IPTE_GLOBAL);
+		__ptep_ipte(address, pte, 0, 0, IPTE_GLOBAL);
 		address += PAGE_SIZE;
 		pte++;
 	}
diff --git a/arch/s390/mm/pgalloc.c b/arch/s390/mm/pgalloc.c
index 18918e394ce4..a4de34ce392c 100644
--- a/arch/s390/mm/pgalloc.c
+++ b/arch/s390/mm/pgalloc.c
@@ -57,6 +57,7 @@ unsigned long *crst_table_alloc(struct mm_struct *mm)
 
 	if (!page)
 		return NULL;
+	arch_set_page_dat(page, 2);
 	return (unsigned long *) page_to_phys(page);
 }
 
@@ -214,6 +215,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 		__free_page(page);
 		return NULL;
 	}
+	arch_set_page_dat(page, 0);
 	/* Initialize page table */
 	table = (unsigned long *) page_to_phys(page);
 	if (mm_alloc_pgste(mm)) {
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index 4a1f7366b17a..4198a71b8fdd 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -25,8 +25,49 @@
 #include <asm/mmu_context.h>
 #include <asm/page-states.h>
 
+static inline void ptep_ipte_local(struct mm_struct *mm, unsigned long addr,
+				   pte_t *ptep, int nodat)
+{
+	unsigned long opt, asce;
+
+	if (MACHINE_HAS_TLB_GUEST) {
+		opt = 0;
+		asce = READ_ONCE(mm->context.gmap_asce);
+		if (asce == 0UL || nodat)
+			opt |= IPTE_NODAT;
+		if (asce != -1UL) {
+			asce = asce ? : mm->context.asce;
+			opt |= IPTE_GUEST_ASCE;
+		}
+		__ptep_ipte(addr, ptep, opt, asce, IPTE_LOCAL);
+	} else {
+		__ptep_ipte(addr, ptep, 0, 0, IPTE_LOCAL);
+	}
+}
+
+static inline void ptep_ipte_global(struct mm_struct *mm, unsigned long addr,
+				    pte_t *ptep, int nodat)
+{
+	unsigned long opt, asce;
+
+	if (MACHINE_HAS_TLB_GUEST) {
+		opt = 0;
+		asce = READ_ONCE(mm->context.gmap_asce);
+		if (asce == 0UL || nodat)
+			opt |= IPTE_NODAT;
+		if (asce != -1UL) {
+			asce = asce ? : mm->context.asce;
+			opt |= IPTE_GUEST_ASCE;
+		}
+		__ptep_ipte(addr, ptep, opt, asce, IPTE_GLOBAL);
+	} else {
+		__ptep_ipte(addr, ptep, 0, 0, IPTE_GLOBAL);
+	}
+}
+
 static inline pte_t ptep_flush_direct(struct mm_struct *mm,
-				      unsigned long addr, pte_t *ptep)
+				      unsigned long addr, pte_t *ptep,
+				      int nodat)
 {
 	pte_t old;
 
@@ -36,15 +77,16 @@ static inline pte_t ptep_flush_direct(struct mm_struct *mm,
 	atomic_inc(&mm->context.flush_count);
 	if (MACHINE_HAS_TLB_LC &&
 	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
-		__ptep_ipte(addr, ptep, IPTE_LOCAL);
+		ptep_ipte_local(mm, addr, ptep, nodat);
 	else
-		__ptep_ipte(addr, ptep, IPTE_GLOBAL);
+		ptep_ipte_global(mm, addr, ptep, nodat);
 	atomic_dec(&mm->context.flush_count);
 	return old;
 }
 
 static inline pte_t ptep_flush_lazy(struct mm_struct *mm,
-				    unsigned long addr, pte_t *ptep)
+				    unsigned long addr, pte_t *ptep,
+				    int nodat)
 {
 	pte_t old;
 
@@ -57,7 +99,7 @@ static inline pte_t ptep_flush_lazy(struct mm_struct *mm,
 		pte_val(*ptep) |= _PAGE_INVALID;
 		mm->context.flush_mm = 1;
 	} else
-		__ptep_ipte(addr, ptep, IPTE_GLOBAL);
+		ptep_ipte_global(mm, addr, ptep, nodat);
 	atomic_dec(&mm->context.flush_count);
 	return old;
 }
@@ -229,10 +271,12 @@ pte_t ptep_xchg_direct(struct mm_struct *mm, unsigned long addr,
 {
 	pgste_t pgste;
 	pte_t old;
+	int nodat;
 
 	preempt_disable();
 	pgste = ptep_xchg_start(mm, addr, ptep);
-	old = ptep_flush_direct(mm, addr, ptep);
+	nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
+	old = ptep_flush_direct(mm, addr, ptep, nodat);
 	old = ptep_xchg_commit(mm, addr, ptep, pgste, old, new);
 	preempt_enable();
 	return old;
@@ -244,10 +288,12 @@ pte_t ptep_xchg_lazy(struct mm_struct *mm, unsigned long addr,
 {
 	pgste_t pgste;
 	pte_t old;
+	int nodat;
 
 	preempt_disable();
 	pgste = ptep_xchg_start(mm, addr, ptep);
-	old = ptep_flush_lazy(mm, addr, ptep);
+	nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
+	old = ptep_flush_lazy(mm, addr, ptep, nodat);
 	old = ptep_xchg_commit(mm, addr, ptep, pgste, old, new);
 	preempt_enable();
 	return old;
@@ -259,10 +305,12 @@ pte_t ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr,
 {
 	pgste_t pgste;
 	pte_t old;
+	int nodat;
 
 	preempt_disable();
 	pgste = ptep_xchg_start(mm, addr, ptep);
-	old = ptep_flush_lazy(mm, addr, ptep);
+	nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
+	old = ptep_flush_lazy(mm, addr, ptep, nodat);
 	if (mm_has_pgste(mm)) {
 		pgste = pgste_update_all(old, pgste, mm);
 		pgste_set(ptep, pgste);
@@ -290,6 +338,28 @@ void ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr,
 }
 EXPORT_SYMBOL(ptep_modify_prot_commit);
 
+static inline void pmdp_idte_local(struct mm_struct *mm,
+				   unsigned long addr, pmd_t *pmdp)
+{
+	if (MACHINE_HAS_TLB_GUEST)
+		__pmdp_idte(addr, pmdp, IDTE_NODAT | IDTE_GUEST_ASCE,
+			    mm->context.asce, IDTE_LOCAL);
+	else
+		__pmdp_idte(addr, pmdp, 0, 0, IDTE_LOCAL);
+}
+
+static inline void pmdp_idte_global(struct mm_struct *mm,
+				    unsigned long addr, pmd_t *pmdp)
+{
+	if (MACHINE_HAS_TLB_GUEST)
+		__pmdp_idte(addr, pmdp, IDTE_NODAT | IDTE_GUEST_ASCE,
+			    mm->context.asce, IDTE_GLOBAL);
+	else if (MACHINE_HAS_IDTE)
+		__pmdp_idte(addr, pmdp, 0, 0, IDTE_GLOBAL);
+	else
+		__pmdp_csp(pmdp);
+}
+
 static inline pmd_t pmdp_flush_direct(struct mm_struct *mm,
 				      unsigned long addr, pmd_t *pmdp)
 {
@@ -298,16 +368,12 @@ static inline pmd_t pmdp_flush_direct(struct mm_struct *mm,
 	old = *pmdp;
 	if (pmd_val(old) & _SEGMENT_ENTRY_INVALID)
 		return old;
-	if (!MACHINE_HAS_IDTE) {
-		__pmdp_csp(pmdp);
-		return old;
-	}
 	atomic_inc(&mm->context.flush_count);
 	if (MACHINE_HAS_TLB_LC &&
 	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
-		__pmdp_idte(addr, pmdp, IDTE_LOCAL);
+		pmdp_idte_local(mm, addr, pmdp);
 	else
-		__pmdp_idte(addr, pmdp, IDTE_GLOBAL);
+		pmdp_idte_global(mm, addr, pmdp);
 	atomic_dec(&mm->context.flush_count);
 	return old;
 }
@@ -325,10 +391,9 @@ static inline pmd_t pmdp_flush_lazy(struct mm_struct *mm,
 			  cpumask_of(smp_processor_id()))) {
 		pmd_val(*pmdp) |= _SEGMENT_ENTRY_INVALID;
 		mm->context.flush_mm = 1;
-	} else if (MACHINE_HAS_IDTE)
-		__pmdp_idte(addr, pmdp, IDTE_GLOBAL);
-	else
-		__pmdp_csp(pmdp);
+	} else {
+		pmdp_idte_global(mm, addr, pmdp);
+	}
 	atomic_dec(&mm->context.flush_count);
 	return old;
 }
@@ -359,28 +424,46 @@ pmd_t pmdp_xchg_lazy(struct mm_struct *mm, unsigned long addr,
 }
 EXPORT_SYMBOL(pmdp_xchg_lazy);
 
-static inline pud_t pudp_flush_direct(struct mm_struct *mm,
-				      unsigned long addr, pud_t *pudp)
+static inline void pudp_idte_local(struct mm_struct *mm,
+				   unsigned long addr, pud_t *pudp)
 {
-	pud_t old;
+	if (MACHINE_HAS_TLB_GUEST)
+		__pudp_idte(addr, pudp, IDTE_NODAT | IDTE_GUEST_ASCE,
+			    mm->context.asce, IDTE_LOCAL);
+	else
+		__pudp_idte(addr, pudp, 0, 0, IDTE_LOCAL);
+}
 
-	old = *pudp;
-	if (pud_val(old) & _REGION_ENTRY_INVALID)
-		return old;
-	if (!MACHINE_HAS_IDTE) {
+static inline void pudp_idte_global(struct mm_struct *mm,
+				    unsigned long addr, pud_t *pudp)
+{
+	if (MACHINE_HAS_TLB_GUEST)
+		__pudp_idte(addr, pudp, IDTE_NODAT | IDTE_GUEST_ASCE,
+			    mm->context.asce, IDTE_GLOBAL);
+	else if (MACHINE_HAS_IDTE)
+		__pudp_idte(addr, pudp, 0, 0, IDTE_GLOBAL);
+	else
 		/*
 		 * Invalid bit position is the same for pmd and pud, so we can
 		 * re-use _pmd_csp() here
 		 */
 		__pmdp_csp((pmd_t *) pudp);
+}
+
+static inline pud_t pudp_flush_direct(struct mm_struct *mm,
+				      unsigned long addr, pud_t *pudp)
+{
+	pud_t old;
+
+	old = *pudp;
+	if (pud_val(old) & _REGION_ENTRY_INVALID)
 		return old;
-	}
 	atomic_inc(&mm->context.flush_count);
 	if (MACHINE_HAS_TLB_LC &&
 	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id())))
-		__pudp_idte(addr, pudp, IDTE_LOCAL);
+		pudp_idte_local(mm, addr, pudp);
 	else
-		__pudp_idte(addr, pudp, IDTE_GLOBAL);
+		pudp_idte_global(mm, addr, pudp);
 	atomic_dec(&mm->context.flush_count);
 	return old;
 }
@@ -482,7 +565,7 @@ int ptep_force_prot(struct mm_struct *mm, unsigned long addr,
 {
 	pte_t entry;
 	pgste_t pgste;
-	int pte_i, pte_p;
+	int pte_i, pte_p, nodat;
 
 	pgste = pgste_get_lock(ptep);
 	entry = *ptep;
@@ -495,13 +578,14 @@ int ptep_force_prot(struct mm_struct *mm, unsigned long addr,
 		return -EAGAIN;
 	}
 	/* Change access rights and set pgste bit */
+	nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
 	if (prot == PROT_NONE && !pte_i) {
-		ptep_flush_direct(mm, addr, ptep);
+		ptep_flush_direct(mm, addr, ptep, nodat);
 		pgste = pgste_update_all(entry, pgste, mm);
 		pte_val(entry) |= _PAGE_INVALID;
 	}
 	if (prot == PROT_READ && !pte_p) {
-		ptep_flush_direct(mm, addr, ptep);
+		ptep_flush_direct(mm, addr, ptep, nodat);
 		pte_val(entry) &= ~_PAGE_INVALID;
 		pte_val(entry) |= _PAGE_PROTECT;
 	}
@@ -541,10 +625,12 @@ int ptep_shadow_pte(struct mm_struct *mm, unsigned long saddr,
 void ptep_unshadow_pte(struct mm_struct *mm, unsigned long saddr, pte_t *ptep)
 {
 	pgste_t pgste;
+	int nodat;
 
 	pgste = pgste_get_lock(ptep);
 	/* notifier is called by the caller */
-	ptep_flush_direct(mm, saddr, ptep);
+	nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
+	ptep_flush_direct(mm, saddr, ptep, nodat);
 	/* don't touch the storage key - it belongs to parent pgste */
 	pgste = pgste_set_pte(ptep, pgste, __pte(_PAGE_INVALID));
 	pgste_set_unlock(ptep, pgste);
@@ -617,6 +703,7 @@ bool test_and_clear_guest_dirty(struct mm_struct *mm, unsigned long addr)
 	pte_t *ptep;
 	pte_t pte;
 	bool dirty;
+	int nodat;
 
 	pgd = pgd_offset(mm, addr);
 	p4d = p4d_alloc(mm, pgd, addr);
@@ -645,7 +732,8 @@ bool test_and_clear_guest_dirty(struct mm_struct *mm, unsigned long addr)
 	pte = *ptep;
 	if (dirty && (pte_val(pte) & _PAGE_PRESENT)) {
 		pgste = pgste_pte_notify(mm, addr, ptep, pgste);
-		__ptep_ipte(addr, ptep, IPTE_GLOBAL);
+		nodat = !!(pgste_val(pgste) & _PGSTE_GPS_NODAT);
+		ptep_ipte_global(mm, addr, ptep, nodat);
 		if (MACHINE_HAS_ESOP || !(pte_val(pte) & _PAGE_WRITE))
 			pte_val(pte) |= _PAGE_PROTECT;
 		else