s390/mm,tlb: safeguard against speculative TLB creation

The principles of operations states that the CPU is allowed to create
TLB entries for an address space anytime while an ASCE is loaded to
the control register. This is true even if the CPU is running in the
kernel and the user address space is not (actively) accessed.

In theory this can affect two aspects of the TLB flush logic.
For full-mm flushes the ASCE of the dying process is still attached.
The approach to flush first with IDTE and then just free all page
tables can in theory lead to stale TLB entries. Use the batched
free of page tables for the full-mm flushes as well.

For operations that can have a stale ASCE in the control register,
e.g. a delayed update_user_asce in switch_mm, load the kernel ASCE
to prevent invalid TLBs from being created.

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

arch/s390/include/asm/mmu_context.h

@@ -35,7 +35,7 @@ static inline int init_new_context(struct task_struct *tsk,
 #define LCTL_OPCODE "lctlg"
 #endif
 
-static inline void update_mm(struct mm_struct *mm, struct task_struct *tsk)
+static inline void update_user_asce(struct mm_struct *mm)
 {
 	pgd_t *pgd = mm->pgd;
 
@@ -45,6 +45,13 @@ static inline void update_mm(struct mm_struct *mm, struct task_struct *tsk)
 	set_fs(current->thread.mm_segment);
 }
 
+static inline void clear_user_asce(struct mm_struct *mm)
+{
+	S390_lowcore.user_asce = S390_lowcore.kernel_asce;
+	asm volatile(LCTL_OPCODE" 1,1,%0\n" : : "m" (S390_lowcore.user_asce));
+	asm volatile(LCTL_OPCODE" 7,7,%0\n" : : "m" (S390_lowcore.user_asce));
+}
+
 static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			     struct task_struct *tsk)
 {
@@ -53,11 +60,13 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	if (prev == next)
 		return;
 	if (atomic_inc_return(&next->context.attach_count) >> 16) {
-		/* Delay update_mm until all TLB flushes are done. */
+		/* Delay update_user_asce until all TLB flushes are done. */
 		set_tsk_thread_flag(tsk, TIF_TLB_WAIT);
+		/* Clear old ASCE by loading the kernel ASCE. */
+		clear_user_asce(next);
 	} else {
 		cpumask_set_cpu(cpu, mm_cpumask(next));
-		update_mm(next, tsk);
+		update_user_asce(next);
 		if (next->context.flush_mm)
 			/* Flush pending TLBs */
 			__tlb_flush_mm(next);
@@ -80,7 +89,7 @@ static inline void finish_arch_post_lock_switch(void)
 		cpu_relax();
 
 	cpumask_set_cpu(smp_processor_id(), mm_cpumask(mm));
-	update_mm(mm, tsk);
+	update_user_asce(mm);
 	if (mm->context.flush_mm)
 		__tlb_flush_mm(mm);
 	preempt_enable();

arch/s390/include/asm/tlb.h

@@ -57,8 +57,6 @@ static inline void tlb_gather_mmu(struct mmu_gather *tlb,
 	tlb->end = end;
 	tlb->fullmm = !(start | (end+1));
 	tlb->batch = NULL;
-	if (tlb->fullmm)
-		__tlb_flush_mm(mm);
 }
 
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
@@ -96,9 +94,7 @@ static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 				unsigned long address)
 {
-	if (!tlb->fullmm)
-		return page_table_free_rcu(tlb, (unsigned long *) pte);
-	page_table_free(tlb->mm, (unsigned long *) pte);
+	page_table_free_rcu(tlb, (unsigned long *) pte);
 }
 
 /*
@@ -114,9 +110,7 @@ static inline void pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
 #ifdef CONFIG_64BIT
 	if (tlb->mm->context.asce_limit <= (1UL << 31))
 		return;
-	if (!tlb->fullmm)
-		return tlb_remove_table(tlb, pmd);
-	crst_table_free(tlb->mm, (unsigned long *) pmd);
+	tlb_remove_table(tlb, pmd);
 #endif
 }
 
@@ -133,9 +127,7 @@ static inline void pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
 #ifdef CONFIG_64BIT
 	if (tlb->mm->context.asce_limit <= (1UL << 42))
 		return;
-	if (!tlb->fullmm)
-		return tlb_remove_table(tlb, pud);
-	crst_table_free(tlb->mm, (unsigned long *) pud);
+	tlb_remove_table(tlb, pud);
 #endif
 }
 

arch/s390/mm/pgtable.c

@@ -54,7 +54,7 @@ static void __crst_table_upgrade(void *arg)
 	struct mm_struct *mm = arg;
 
 	if (current->active_mm == mm)
-		update_mm(mm, current);
+		update_user_asce(mm);
 	__tlb_flush_local();
 }
 
@@ -107,8 +107,10 @@ void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
 {
 	pgd_t *pgd;
 
-	if (current->active_mm == mm)
+	if (current->active_mm == mm) {
+		clear_user_asce(mm);
 		__tlb_flush_mm(mm);
+	}
 	while (mm->context.asce_limit > limit) {
 		pgd = mm->pgd;
 		switch (pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) {
@@ -132,7 +134,7 @@ void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
 		crst_table_free(mm, (unsigned long *) pgd);
 	}
 	if (current->active_mm == mm)
-		update_mm(mm, current);
+		update_user_asce(mm);
 }
 #endif