1 files changed, 83 insertions, 72 deletions

diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index f85355bf49c4..761451f36e2d 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -33,9 +33,7 @@
 
 static const struct inode_operations ceph_symlink_iops;
 
-static void ceph_invalidate_work(struct work_struct *work);
-static void ceph_writeback_work(struct work_struct *work);
-static void ceph_vmtruncate_work(struct work_struct *work);
+static void ceph_inode_work(struct work_struct *work);
 
 /*
  * find or create an inode, given the ceph ino number
@@ -509,10 +507,8 @@ struct inode *ceph_alloc_inode(struct super_block *sb)
 	INIT_LIST_HEAD(&ci->i_snap_realm_item);
 	INIT_LIST_HEAD(&ci->i_snap_flush_item);
 
-	INIT_WORK(&ci->i_wb_work, ceph_writeback_work);
-	INIT_WORK(&ci->i_pg_inv_work, ceph_invalidate_work);
-
-	INIT_WORK(&ci->i_vmtruncate_work, ceph_vmtruncate_work);
+	INIT_WORK(&ci->i_work, ceph_inode_work);
+	ci->i_work_mask = 0;
 
 	ceph_fscache_inode_init(ci);
 
@@ -1480,7 +1476,8 @@ static int readdir_prepopulate_inodes_only(struct ceph_mds_request *req,
 			pr_err("fill_inode badness on %p got %d\n", in, rc);
 			err = rc;
 		}
-		iput(in);
+		/* avoid calling iput_final() in mds dispatch threads */
+		ceph_async_iput(in);
 	}
 
 	return err;
@@ -1678,8 +1675,11 @@ retry_lookup:
 				 &req->r_caps_reservation);
 		if (ret < 0) {
 			pr_err("fill_inode badness on %p\n", in);
-			if (d_really_is_negative(dn))
-				iput(in);
+			if (d_really_is_negative(dn)) {
+				/* avoid calling iput_final() in mds
+				 * dispatch threads */
+				ceph_async_iput(in);
+			}
 			d_drop(dn);
 			err = ret;
 			goto next_item;
@@ -1689,7 +1689,7 @@ retry_lookup:
 			if (ceph_security_xattr_deadlock(in)) {
 				dout(" skip splicing dn %p to inode %p"
 				     " (security xattr deadlock)\n", dn, in);
-				iput(in);
+				ceph_async_iput(in);
 				skipped++;
 				goto next_item;
 			}
@@ -1741,56 +1741,86 @@ bool ceph_inode_set_size(struct inode *inode, loff_t size)
 }
 
 /*
+ * Put reference to inode, but avoid calling iput_final() in current thread.
+ * iput_final() may wait for reahahead pages. The wait can cause deadlock in
+ * some contexts.
+ */
+void ceph_async_iput(struct inode *inode)
+{
+	if (!inode)
+		return;
+	for (;;) {
+		if (atomic_add_unless(&inode->i_count, -1, 1))
+			break;
+		if (queue_work(ceph_inode_to_client(inode)->inode_wq,
+			       &ceph_inode(inode)->i_work))
+			break;
+		/* queue work failed, i_count must be at least 2 */
+	}
+}
+
+/*
  * Write back inode data in a worker thread.  (This can't be done
  * in the message handler context.)
  */
 void ceph_queue_writeback(struct inode *inode)
 {
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	set_bit(CEPH_I_WORK_WRITEBACK, &ci->i_work_mask);
+
 	ihold(inode);
-	if (queue_work(ceph_inode_to_client(inode)->wb_wq,
-		       &ceph_inode(inode)->i_wb_work)) {
+	if (queue_work(ceph_inode_to_client(inode)->inode_wq,
+		       &ci->i_work)) {
 		dout("ceph_queue_writeback %p\n", inode);
 	} else {
-		dout("ceph_queue_writeback %p failed\n", inode);
+		dout("ceph_queue_writeback %p already queued, mask=%lx\n",
+		     inode, ci->i_work_mask);
 		iput(inode);
 	}
 }
 
-static void ceph_writeback_work(struct work_struct *work)
-{
-	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
-						  i_wb_work);
-	struct inode *inode = &ci->vfs_inode;
-
-	dout("writeback %p\n", inode);
-	filemap_fdatawrite(&inode->i_data);
-	iput(inode);
-}
-
 /*
  * queue an async invalidation
  */
 void ceph_queue_invalidate(struct inode *inode)
 {
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	set_bit(CEPH_I_WORK_INVALIDATE_PAGES, &ci->i_work_mask);
+
 	ihold(inode);
-	if (queue_work(ceph_inode_to_client(inode)->pg_inv_wq,
-		       &ceph_inode(inode)->i_pg_inv_work)) {
+	if (queue_work(ceph_inode_to_client(inode)->inode_wq,
+		       &ceph_inode(inode)->i_work)) {
 		dout("ceph_queue_invalidate %p\n", inode);
 	} else {
-		dout("ceph_queue_invalidate %p failed\n", inode);
+		dout("ceph_queue_invalidate %p already queued, mask=%lx\n",
+		     inode, ci->i_work_mask);
 		iput(inode);
 	}
 }
 
 /*
- * Invalidate inode pages in a worker thread.  (This can't be done
- * in the message handler context.)
+ * Queue an async vmtruncate.  If we fail to queue work, we will handle
+ * the truncation the next time we call __ceph_do_pending_vmtruncate.
  */
-static void ceph_invalidate_work(struct work_struct *work)
+void ceph_queue_vmtruncate(struct inode *inode)
 {
-	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
-						  i_pg_inv_work);
-	struct inode *inode = &ci->vfs_inode;
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	set_bit(CEPH_I_WORK_VMTRUNCATE, &ci->i_work_mask);
+
+	ihold(inode);
+	if (queue_work(ceph_inode_to_client(inode)->inode_wq,
+		       &ci->i_work)) {
+		dout("ceph_queue_vmtruncate %p\n", inode);
+	} else {
+		dout("ceph_queue_vmtruncate %p already queued, mask=%lx\n",
+		     inode, ci->i_work_mask);
+		iput(inode);
+	}
+}
+
+static void ceph_do_invalidate_pages(struct inode *inode)
+{
+	struct ceph_inode_info *ci = ceph_inode(inode);
 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 	u32 orig_gen;
 	int check = 0;
@@ -1842,44 +1872,6 @@ static void ceph_invalidate_work(struct work_struct *work)
 out:
 	if (check)
 		ceph_check_caps(ci, 0, NULL);
-	iput(inode);
-}
-
-
-/*
- * called by trunc_wq;
- *
- * We also truncate in a separate thread as well.
- */
-static void ceph_vmtruncate_work(struct work_struct *work)
-{
-	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
-						  i_vmtruncate_work);
-	struct inode *inode = &ci->vfs_inode;
-
-	dout("vmtruncate_work %p\n", inode);
-	__ceph_do_pending_vmtruncate(inode);
-	iput(inode);
-}
-
-/*
- * Queue an async vmtruncate.  If we fail to queue work, we will handle
- * the truncation the next time we call __ceph_do_pending_vmtruncate.
- */
-void ceph_queue_vmtruncate(struct inode *inode)
-{
-	struct ceph_inode_info *ci = ceph_inode(inode);
-
-	ihold(inode);
-
-	if (queue_work(ceph_sb_to_client(inode->i_sb)->trunc_wq,
-		       &ci->i_vmtruncate_work)) {
-		dout("ceph_queue_vmtruncate %p\n", inode);
-	} else {
-		dout("ceph_queue_vmtruncate %p failed, pending=%d\n",
-		     inode, ci->i_truncate_pending);
-		iput(inode);
-	}
 }
 
 /*
@@ -1943,6 +1935,25 @@ retry:
 	wake_up_all(&ci->i_cap_wq);
 }
 
+static void ceph_inode_work(struct work_struct *work)
+{
+	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
+						 i_work);
+	struct inode *inode = &ci->vfs_inode;
+
+	if (test_and_clear_bit(CEPH_I_WORK_WRITEBACK, &ci->i_work_mask)) {
+		dout("writeback %p\n", inode);
+		filemap_fdatawrite(&inode->i_data);
+	}
+	if (test_and_clear_bit(CEPH_I_WORK_INVALIDATE_PAGES, &ci->i_work_mask))
+		ceph_do_invalidate_pages(inode);
+
+	if (test_and_clear_bit(CEPH_I_WORK_VMTRUNCATE, &ci->i_work_mask))
+		__ceph_do_pending_vmtruncate(inode);
+
+	iput(inode);
+}
+
 /*
  * symlinks
  */