linux-kernel - [RFC][PATCH 5/8] epoll: iterate epi in file->f_ep_links by using list_first_or_null

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sat, 28 Oct 2017 20:58:24 +0800
From:   Hou Tao <houtao1@...wei.com>
To:     <linux-fsdevel@...r.kernel.org>
CC:     <linux-kernel@...r.kernel.org>, <viro@...iv.linux.org.uk>,
        <jbaron@...mai.com>, <oleg@...hat.com>, <dave@...olabs.net>,
        <koct9i@...il.com>
Subject: [RFC][PATCH 5/8] epoll: iterate epi in file->f_ep_links by using list_first_or_null_rcu

When eventpoll_release_file() iterates epitem in file->f_ep_links,
the epitem may be removed by ep_free(). To protect again the concurrent
writer, iterate file->f_ep_links by using rcu_read_lock() and
list_first_or_null_rcu()

Signed-off-by: Hou Tao <houtao1@...wei.com>
---
 fs/eventpoll.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index 18de596..e1e4796 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -1033,12 +1033,22 @@ void eventpoll_release_file(struct file *file)
 	 *
 	 * Besides, ep_remove() acquires the lock, so we can't hold it here.
 	 */
-	list_for_each_entry_safe(epi, next, &file->f_ep_links, fllink) {
+	rcu_read_lock();
+	while (true) {
+		epi = list_first_or_null_rcu(&file->f_ep_links, struct epitem, fllink);
+		if (!epi)
+			break;
+
 		ep = epi->ep;
+		rcu_read_unlock();
+
 		mutex_lock_nested(&ep->mtx, 0);
 		ep_remove(ep, epi);
 		mutex_unlock(&ep->mtx);
+
+		rcu_read_lock();
 	}
+	rcu_read_unlock();
 
 	/*
 	 * The file can not been added to tfile_check_list again, because
-- 
2.7.5