ceph: don't call truncate_pagecache in ceph_writepages_start

truncate_pagecache() may decrease inode's reference. This can cause
deadlock if inode's last reference is dropped and iput_final() wants
to evict the inode. (evict() calls inode_wait_for_writeback(), which
waits for ceph_writepages_start() to return).

The fix is use work thead to truncate dirty pages. Also add 'forced
umount' check to ceph_update_writeable_page(), which prevents new
pages getting dirty.

Signed-off-by: Yan, Zheng <zyan@redhat.com>

3 files changed, 38 insertions, 9 deletions

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index f47418477629..d52e3bcfda7c 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -715,8 +715,11 @@ static int ceph_writepages_start(struct address_space *mapping,
 	     (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
 
 	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
-		pr_warn("writepage_start %p on forced umount\n", inode);
-		truncate_pagecache(inode, 0);
+		if (ci->i_wrbuffer_ref > 0) {
+			pr_warn_ratelimited(
+				"writepage_start %p %lld forced umount\n",
+				inode, ceph_ino(inode));
+		}
 		mapping_set_error(mapping, -EIO);
 		return -EIO; /* we're in a forced umount, don't write! */
 	}
@@ -1127,6 +1130,7 @@ static int ceph_update_writeable_page(struct file *file,
 			    struct page *page)
 {
 	struct inode *inode = file_inode(file);
+	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	loff_t page_off = pos & PAGE_MASK;
 	int pos_in_page = pos & ~PAGE_MASK;
@@ -1135,6 +1139,12 @@ static int ceph_update_writeable_page(struct file *file,
 	int r;
 	struct ceph_snap_context *snapc, *oldest;
 
+	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+		dout(" page %p forced umount\n", page);
+		unlock_page(page);
+		return -EIO;
+	}
+
 retry_locked:
 	/* writepages currently holds page lock, but if we change that later, */
 	wait_on_page_writeback(page);
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index edfade037738..b906e02cddad 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1623,10 +1623,21 @@ static void ceph_invalidate_work(struct work_struct *work)
 	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
 						  i_pg_inv_work);
 	struct inode *inode = &ci->vfs_inode;
+	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 	u32 orig_gen;
 	int check = 0;
 
 	mutex_lock(&ci->i_truncate_mutex);
+
+	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+		pr_warn_ratelimited("invalidate_pages %p %lld forced umount\n",
+				    inode, ceph_ino(inode));
+		mapping_set_error(inode->i_mapping, -EIO);
+		truncate_pagecache(inode, 0);
+		mutex_unlock(&ci->i_truncate_mutex);
+		goto out;
+	}
+
 	spin_lock(&ci->i_ceph_lock);
 	dout("invalidate_pages %p gen %d revoking %d\n", inode,
 	     ci->i_rdcache_gen, ci->i_rdcache_revoking);
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 1e5965d17cb1..cbe6c0afdadc 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1120,9 +1120,11 @@ out:
 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 				  void *arg)
 {
+	struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	LIST_HEAD(to_remove);
-	int drop = 0;
+	bool drop = false;
+	bool invalidate = false;
 
 	dout("removing cap %p, ci is %p, inode is %p\n",
 	     cap, ci, &ci->vfs_inode);
@@ -1130,11 +1132,14 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 	__ceph_remove_cap(cap, false);
 	if (!ci->i_auth_cap) {
 		struct ceph_cap_flush *cf;
-		struct ceph_mds_client *mdsc =
-			ceph_sb_to_client(inode->i_sb)->mdsc;
+		struct ceph_mds_client *mdsc = fsc->mdsc;
 
 		ci->i_ceph_flags |= CEPH_I_CAP_DROPPED;
 
+		if (ci->i_wrbuffer_ref > 0 &&
+		    ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
+			invalidate = true;
+
 		while (true) {
 			struct rb_node *n = rb_first(&ci->i_cap_flush_tree);
 			if (!n)
@@ -1156,7 +1161,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 				inode, ceph_ino(inode));
 			ci->i_dirty_caps = 0;
 			list_del_init(&ci->i_dirty_item);
-			drop = 1;
+			drop = true;
 		}
 		if (!list_empty(&ci->i_flushing_item)) {
 			pr_warn_ratelimited(
@@ -1166,7 +1171,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 			ci->i_flushing_caps = 0;
 			list_del_init(&ci->i_flushing_item);
 			mdsc->num_cap_flushing--;
-			drop = 1;
+			drop = true;
 		}
 		spin_unlock(&mdsc->cap_dirty_lock);
 
@@ -1185,6 +1190,8 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 	}
 
 	wake_up_all(&ci->i_cap_wq);
+	if (invalidate)
+		ceph_queue_invalidate(inode);
 	if (drop)
 		iput(inode);
 	return 0;
@@ -1195,12 +1202,13 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
  */
 static void remove_session_caps(struct ceph_mds_session *session)
 {
+	struct ceph_fs_client *fsc = session->s_mdsc->fsc;
+	struct super_block *sb = fsc->sb;
 	dout("remove_session_caps on %p\n", session);
-	iterate_session_caps(session, remove_session_caps_cb, NULL);
+	iterate_session_caps(session, remove_session_caps_cb, fsc);
 
 	spin_lock(&session->s_cap_lock);
 	if (session->s_nr_caps > 0) {
-		struct super_block *sb = session->s_mdsc->fsc->sb;
 		struct inode *inode;
 		struct ceph_cap *cap, *prev = NULL;
 		struct ceph_vino vino;