Unionfs: prevent deadlock in cache coherency

Don't try to truncate_inode_pages in in purge_inode_data, because this could
lead to a deadlock between some of address_space ops and dentry
revalidation: the address space op is invoked with a lock on our own page,
and truncate_inode_pages will block on locked pages.  Instead, it should be
enough to be gentler and just invalidate_mapping_pages.

Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>

diff --git a/fs/unionfs/dentry.c b/fs/unionfs/dentry.c
index c7b6a7c774b77b3afdaafce5fc32f700588fb354..68c07ba8a19bcd8053ad90819e4065e1c9d1c0b6 100644 (file)
--- a/fs/unionfs/dentry.c
+++ b/fs/unionfs/dentry.c
@@ -248,32 +248,22 @@ bool is_newer_lower(const struct dentry *dentry)
 }
 
 /*
- * Purge/remove/unmap all date pages of a unionfs inode.  This is called
- * when the lower inode has changed, and we have to force processes to get
- * the new data.
- *
- * XXX: Our implementation works in that as long as a user process will have
- * caused Unionfs to be called, directly or indirectly, even to just do
- * ->d_revalidate; then we will have purged the current Unionfs data and the
- * process will see the new data.  For example, a process that continually
- * re-reads the same file's data will see the NEW data as soon as the lower
- * file had changed, upon the next read(2) syscall (even if the file is
- * still open!)  However, this doesn't work when the process re-reads the
- * open file's data via mmap(2) (unless the user unmaps/closes the file and
- * remaps/reopens it).  Once we respond to ->readpage(s), then the kernel
- * maps the page into the process's address space and there doesn't appear
- * to be a way to force the kernel to invalidate those pages/mappings, and
- * force the process to re-issue ->readpage.  If there's a way to invalidate
- * active mappings and force a ->readpage, let us know please
- * (invalidate_inode_pages2 doesn't do the trick).
+ * Purge and invalidate as many data pages of a unionfs inode.  This is
+ * called when the lower inode has changed, and we want to force processes
+ * to re-get the new data.
  */
 static inline void purge_inode_data(struct inode *inode)
 {
        /* remove all non-private mappings */
        unmap_mapping_range(inode->i_mapping, 0, 0, 0);
-
-       if (inode->i_data.nrpages)
-               truncate_inode_pages(&inode->i_data, 0);
+       /* invalidate as many pages as possible */
+       invalidate_mapping_pages(inode->i_mapping, 0, -1);
+       /*
+        * Don't try to truncate_inode_pages here, because this could lead
+        * to a deadlock between some of address_space ops and dentry
+        * revalidation: the address space op is invoked with a lock on our
+        * own page, and truncate_inode_pages will block on locked pages.
+        */
 }
 
 void purge_sb_data(struct super_block *sb)