Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Subject: [PATCH 4/6] autofs: add comment about autofs_mountpoint_changed()
From:   Ian Kent <raven@themaw.net>
To:     Al Viro <viro@ZenIV.linux.org.uk>
Cc:     David Howells <dhowells@redhat.com>,
        Miklos Szeredi <miklos@szeredi.hu>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Kernel Mailing List <linux-kernel@vger.kernel.org>
Date:   Fri, 17 Jun 2022 13:35:46 +0800
Message-ID: <165544414635.250070.10563886994408560172.stgit@donald.themaw.net>
In-Reply-To: <165544393032.250070.3426550720222448062.stgit@donald.themaw.net>
References: <165544393032.250070.3426550720222448062.stgit@donald.themaw.net>
User-Agent: StGit/1.1
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
Precedence: bulk

The function autofs_mountpoint_changed() is unusual, add a comment
about two cases for which it is needed.

Signed-off-by: Ian Kent <raven@themaw.net>
---
 fs/autofs/root.c |   23 ++++++++++++++++++++---
 1 file changed, 20 insertions(+), 3 deletions(-)

diff --git a/fs/autofs/root.c b/fs/autofs/root.c
index e0fa71eb5c05..ca03c1cae2be 100644
--- a/fs/autofs/root.c
+++ b/fs/autofs/root.c
@@ -291,9 +291,26 @@ static struct dentry *autofs_mountpoint_changed(struct path *path)
 	struct dentry *dentry = path->dentry;
 	struct autofs_sb_info *sbi = autofs_sbi(dentry->d_sb);
 
-	/*
-	 * If this is an indirect mount the dentry could have gone away
-	 * as a result of an expire and a new one created.
+	/* If this is an indirect mount the dentry could have gone away
+	 * and a new one created.
+	 *
+	 * This is unusual and I can't remember the case for which it
+	 * was originally added now. But an example of how this can
+	 * happen is an autofs indirect mount that has the "browse"
+	 * option set and also has the "symlink" option in the autofs
+	 * map entry. In this case the daemon will remove the browse
+	 * directory and create a symlink as the mount leaving the
+	 * struct path stale.
+	 *
+	 * Another not so obvious case is when a mount in an autofs
+	 * indirect mount that uses the "nobrowse" option is being
+	 * expired at the same time as a path walk. If the mount has
+	 * been umounted but the mount point directory seen before
+	 * becoming unhashed (during a lockless path walk) when a stat
+	 * family system call is made the mount won't be re-mounted as
+	 * it should. In this case the mount point that's been removed
+	 * (by the daemon) will be stale and the a new mount point
+	 * dentry created.
 	 */
 	if (autofs_type_indirect(sbi->type) && d_unhashed(dentry)) {
 		struct dentry *parent = dentry->d_parent;