xfs: use i_mmaplock on read faults

Take the i_mmaplock over read page faults. These come through the
->fault callout, so we need to wrap the generic implementation
with the i_mmaplock. While there, add tracepoints for the read
fault as it passes through XFS.

This gives us a lock order of mmap_sem -> i_mmaplock -> page_lock
-> i_lock.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

2 files changed, 29 insertions, 1 deletions

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index ce615d12fb44..ac174226244a 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1379,6 +1379,32 @@ xfs_file_llseek(
         }
 }
 
+/*
+ * Locking for serialisation of IO during page faults. This results in a lock
+ * ordering of:
+ *
+ * mmap_sem (MM)
+ *   i_mmap_lock (XFS - truncate serialisation)
+ *     page_lock (MM)
+ *       i_lock (XFS - extent map serialisation)
+ */
+STATIC int
+xfs_filemap_fault(
+        struct vm_area_struct   *vma,
+        struct vm_fault         *vmf)
+{
+        struct xfs_inode        *ip = XFS_I(vma->vm_file->f_mapping->host);
+        int                     error;
+
+        trace_xfs_filemap_fault(ip);
+
+        xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
+        error = filemap_fault(vma, vmf);
+        xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
+
+        return error;
+}
+
 const struct file_operations xfs_file_operations = {
         .llseek         = xfs_file_llseek,
         .read           = new_sync_read,
@@ -1411,7 +1437,7 @@ const struct file_operations xfs_dir_file_operations = {
 };
 
 static const struct vm_operations_struct xfs_file_vm_ops = {
-        .fault          = filemap_fault,
+        .fault          = xfs_filemap_fault,
         .map_pages      = filemap_map_pages,
         .page_mkwrite   = xfs_vm_page_mkwrite,
 };
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 51372e34d988..c496153b976b 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -685,6 +685,8 @@ DEFINE_INODE_EVENT(xfs_inode_set_eofblocks_tag);
 DEFINE_INODE_EVENT(xfs_inode_clear_eofblocks_tag);
 DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid);
 
+DEFINE_INODE_EVENT(xfs_filemap_fault);
+
 DECLARE_EVENT_CLASS(xfs_iref_class,
         TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip),
         TP_ARGS(ip, caller_ip),