nfs: propagate readlink errors in nfs_symlink_filler

sagigrimberg · Trond Myklebust · commit 134d0b3f2440 · 2024-05-22T19:25:00.000-04:00
There is an inherent race where a symlink file may have been overriden
(by a different client) between lookup and readlink, resulting in a
spurious EIO error returned to userspace. Fix this by propagating back
ESTALE errors such that the vfs will retry the lookup/get_link (similar
to nfs4_file_open) at least once.

Cc: Dan Aloni &lt;dan.aloni@vastdata.com&gt;
Signed-off-by: Sagi Grimberg &lt;sagi@grimberg.me&gt;
Reviewed-by: Jeff Layton &lt;jlayton@kernel.org&gt;
Signed-off-by: Trond Myklebust &lt;trond.myklebust@hammerspace.com&gt;
diff --git a/fs/nfs/symlink.c b/fs/nfs/symlink.c
@@ -41,7 +41,7 @@ static int nfs_symlink_filler(struct file *file, struct folio *folio)
 error:
 	folio_set_error(folio);
 	folio_unlock(folio);
-	return -EIO;
+	return error;
 }
 
 static const char *nfs_get_link(struct dentry *dentry,

Original file line number	Diff line number	Diff line change
`@@ -41,7 +41,7 @@ static int nfs_symlink_filler(struct file file, struct folio folio)`
`41`	`41`	`error:`
`42`	`42`	`folio_set_error(folio);`
`43`	`43`	`folio_unlock(folio);`
`44`		`- return -EIO;`
	`44`	`+ return error;`
`45`	`45`	`}`
`46`	`46`
`47`	`47`	`static const char nfs_get_link(struct dentry dentry,`