fix tlb flushing for page table pages

commit cd94154cc6a28dd9dc271042c1a59c08d26da886 upstream.

Git commit 36409f6353fc2d7b6516e631415f938eadd92ffa "use generic RCU
page-table freeing code" introduced a tlb flushing bug. Partially revert
the above git commit and go back to s390 specific page table flush code.

For s390 the TLB can contain three types of entries, "normal" TLB
page-table entries, TLB combined region-and-segment-table (CRST) entries
and real-space entries. Linux does not use real-space entries which
leaves normal TLB entries and CRST entries. The CRST entries are
intermediate steps in the page-table translation called translation paths.
For example a 4K page access in a three-level page table setup will
create two CRST TLB entries and one page-table TLB entry. The advantage
of that approach is that a page access next to the previous one can reuse
the CRST entries and needs just a single read from memory to create the
page-table TLB entry. The disadvantage is that the TLB flushing rules are
more complicated, before any page-table may be freed the TLB needs to be
flushed.

In short: the generic RCU page-table freeing code is incorrect for the
CRST entries, in particular the check for mm_users < 2 is troublesome.

This is applicable to 3.0+ kernels.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

1 files changed, 60 insertions, 3 deletions

diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index 301c84d3b54..f8ceac48cf3 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -687,8 +687,6 @@ void page_table_free(struct mm_struct *mm, unsigned long *table)
 	}
 }
 
-#ifdef CONFIG_HAVE_RCU_TABLE_FREE
-
 static void __page_table_free_rcu(void *table, unsigned bit)
 {
 	struct page *page;
@@ -742,7 +740,66 @@ void __tlb_remove_table(void *_table)
 		free_pages((unsigned long) table, ALLOC_ORDER);
 }
 
-#endif
+static void tlb_remove_table_smp_sync(void *arg)
+{
+	/* Simply deliver the interrupt */
+}
+
+static void tlb_remove_table_one(void *table)
+{
+	/*
+	 * This isn't an RCU grace period and hence the page-tables cannot be
+	 * assumed to be actually RCU-freed.
+	 *
+	 * It is however sufficient for software page-table walkers that rely
+	 * on IRQ disabling. See the comment near struct mmu_table_batch.
+	 */
+	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+	__tlb_remove_table(table);
+}
+
+static void tlb_remove_table_rcu(struct rcu_head *head)
+{
+	struct mmu_table_batch *batch;
+	int i;
+
+	batch = container_of(head, struct mmu_table_batch, rcu);
+
+	for (i = 0; i < batch->nr; i++)
+		__tlb_remove_table(batch->tables[i]);
+
+	free_page((unsigned long)batch);
+}
+
+void tlb_table_flush(struct mmu_gather *tlb)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	if (*batch) {
+		__tlb_flush_mm(tlb->mm);
+		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
+		*batch = NULL;
+	}
+}
+
+void tlb_remove_table(struct mmu_gather *tlb, void *table)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	if (*batch == NULL) {
+		*batch = (struct mmu_table_batch *)
+			__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
+		if (*batch == NULL) {
+			__tlb_flush_mm(tlb->mm);
+			tlb_remove_table_one(table);
+			return;
+		}
+		(*batch)->nr = 0;
+	}
+	(*batch)->tables[(*batch)->nr++] = table;
+	if ((*batch)->nr == MAX_TABLE_BATCH)
+		tlb_table_flush(tlb);
+}
 
 /*
  * switch on pgstes for its userspace process (for kvm)