fuse: make sure reclaim doesn't write the inode

In writeback cache mode mtime/ctime updates are cached, and flushed to the
server using the ->write_inode() callback.

Closing the file will result in a dirty inode being immediately written,
but in other cases the inode can remain dirty after all references are
dropped.  This result in the inode being written back from reclaim, which
can deadlock on a regular allocation while the request is being served.

The usual mechanisms (GFP_NOFS/PF_MEMALLOC*) don't work for FUSE, because
serving a request involves unrelated userspace process(es).

Instead do the same as for dirty pages: make sure the inode is written
before the last reference is gone.

 - fallocate(2)/copy_file_range(2): these call file_update_time() or
   file_modified(), so flush the inode before returning from the call

 - unlink(2), link(2) and rename(2): these call fuse_update_ctime(), so
   flush the ctime directly from this helper

Reported-by: chenguanyou <chenguanyou@xiaomi.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>

4 files changed, 27 insertions, 0 deletions

diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index d9b977c0f38d..2798fbe8d001 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -738,11 +738,19 @@ static int fuse_symlink(struct user_namespace *mnt_userns, struct inode *dir,
 	return create_new_entry(fm, &args, dir, entry, S_IFLNK);
 }
 
+void fuse_flush_time_update(struct inode *inode)
+{
+	int err = sync_inode_metadata(inode, 1);
+
+	mapping_set_error(inode->i_mapping, err);
+}
+
 void fuse_update_ctime(struct inode *inode)
 {
 	if (!IS_NOCMTIME(inode)) {
 		inode->i_ctime = current_time(inode);
 		mark_inode_dirty_sync(inode);
+		fuse_flush_time_update(inode);
 	}
 }
 
diff --git a/fs/fuse/file.c b/fs/fuse/file.c
index 11404f8c21c7..5c5ed58d91a7 100644
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -1848,6 +1848,17 @@ int fuse_write_inode(struct inode *inode, struct writeback_control *wbc)
 	struct fuse_file *ff;
 	int err;
 
+	/*
+	 * Inode is always written before the last reference is dropped and
+	 * hence this should not be reached from reclaim.
+	 *
+	 * Writing back the inode from reclaim can deadlock if the request
+	 * processing itself needs an allocation.  Allocations triggering
+	 * reclaim while serving a request can't be prevented, because it can
+	 * involve any number of unrelated userspace processes.
+	 */
+	WARN_ON(wbc->for_reclaim);
+
 	ff = __fuse_write_file_get(fi);
 	err = fuse_flush_times(inode, ff);
 	if (ff)
@@ -3002,6 +3013,8 @@ out:
 	if (lock_inode)
 		inode_unlock(inode);
 
+	fuse_flush_time_update(inode);
+
 	return err;
 }
 
@@ -3111,6 +3124,8 @@ out:
 	inode_unlock(inode_out);
 	file_accessed(file_in);
 
+	fuse_flush_time_update(inode_out);
+
 	return err;
 }
 
diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
index f55f9f94b1a4..a59e36c7deae 100644
--- a/fs/fuse/fuse_i.h
+++ b/fs/fuse/fuse_i.h
@@ -1148,6 +1148,7 @@ int fuse_allow_current_process(struct fuse_conn *fc);
 
 u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id);
 
+void fuse_flush_time_update(struct inode *inode);
 void fuse_update_ctime(struct inode *inode);
 
 int fuse_update_attributes(struct inode *inode, struct file *file);
diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
index 12d49a1914e8..2f999d38c9b4 100644
--- a/fs/fuse/inode.c
+++ b/fs/fuse/inode.c
@@ -118,6 +118,9 @@ static void fuse_evict_inode(struct inode *inode)
 {
 	struct fuse_inode *fi = get_fuse_inode(inode);
 
+	/* Will write inode on close/munmap and in all other dirtiers */
+	WARN_ON(inode->i_state & I_DIRTY_INODE);
+
 	truncate_inode_pages_final(&inode->i_data);
 	clear_inode(inode);
 	if (inode->i_sb->s_flags & SB_ACTIVE) {