fix mremap() vs. ioctx_kill() race

teach ->mremap() method to return an error and have it fail for
aio mappings in process of being killed

Note that in case of ->mremap() failure we need to undo move_page_tables()
we'd already done; we could call ->mremap() first, but then the failure of
move_page_tables() would require undoing whatever _successful_ ->mremap()
has done, which would be a lot more headache in general.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

3 files changed, 20 insertions, 9 deletions

diff --git a/fs/aio.c b/fs/aio.c
index f8e52a1854c1..58c33dcfb6ca 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -278,11 +278,11 @@ static int aio_ring_mmap(struct file *file, struct vm_area_struct *vma)
 	return 0;
 }
 
-static void aio_ring_remap(struct file *file, struct vm_area_struct *vma)
+static int aio_ring_remap(struct file *file, struct vm_area_struct *vma)
 {
 	struct mm_struct *mm = vma->vm_mm;
 	struct kioctx_table *table;
-	int i;
+	int i, res = -EINVAL;
 
 	spin_lock(&mm->ioctx_lock);
 	rcu_read_lock();
@@ -292,13 +292,17 @@ static void aio_ring_remap(struct file *file, struct vm_area_struct *vma)
 
 		ctx = table->table[i];
 		if (ctx && ctx->aio_ring_file == file) {
-			ctx->user_id = ctx->mmap_base = vma->vm_start;
+			if (!atomic_read(&ctx->dead)) {
+				ctx->user_id = ctx->mmap_base = vma->vm_start;
+				res = 0;
+			}
 			break;
 		}
 	}
 
 	rcu_read_unlock();
 	spin_unlock(&mm->ioctx_lock);
+	return res;
 }
 
 static const struct file_operations aio_ring_fops = {
@@ -748,11 +752,12 @@ static int kill_ioctx(struct mm_struct *mm, struct kioctx *ctx,
 {
 	struct kioctx_table *table;
 
-	if (atomic_xchg(&ctx->dead, 1))
+	spin_lock(&mm->ioctx_lock);
+	if (atomic_xchg(&ctx->dead, 1)) {
+		spin_unlock(&mm->ioctx_lock);
 		return -EINVAL;
+	}
 
-
-	spin_lock(&mm->ioctx_lock);
 	table = rcu_dereference_raw(mm->ioctx_table);
 	WARN_ON(ctx != table->table[ctx->id]);
 	table->table[ctx->id] = NULL;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index f4131e8ead74..52cc4492cb3a 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1549,7 +1549,7 @@ struct file_operations {
 	long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
 	long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
 	int (*mmap) (struct file *, struct vm_area_struct *);
-	void (*mremap)(struct file *, struct vm_area_struct *);
+	int (*mremap)(struct file *, struct vm_area_struct *);
 	int (*open) (struct inode *, struct file *);
 	int (*flush) (struct file *, fl_owner_t id);
 	int (*release) (struct inode *, struct file *);
diff --git a/mm/mremap.c b/mm/mremap.c
index 57dadc025c64..2dc44b1cb1df 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -286,8 +286,14 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 		old_len = new_len;
 		old_addr = new_addr;
 		new_addr = -ENOMEM;
-	} else if (vma->vm_file && vma->vm_file->f_op->mremap)
-		vma->vm_file->f_op->mremap(vma->vm_file, new_vma);
+	} else if (vma->vm_file && vma->vm_file->f_op->mremap) {
+		err = vma->vm_file->f_op->mremap(vma->vm_file, new_vma);
+		if (err < 0) {
+			move_page_tables(new_vma, new_addr, vma, old_addr,
+					 moved_len, true);
+			return err;
+		}
+	}
 
 	/* Conceal VM_ACCOUNT so old reservation is not undone */
 	if (vm_flags & VM_ACCOUNT) {