linux-kernel - [PATCH 1/1] seccomp: Always "goto wait" if the list is empty

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:   Tue, 13 Apr 2021 18:01:51 +0200
From:   Rodrigo Campos <rodrigo@...volk.io>
To:     Kees Cook <keescook@...omium.org>,
        Andy Lutomirski <luto@...capital.net>,
        Will Drewry <wad@...omium.org>, linux-kernel@...r.kernel.org,
        containers@...ts.linux-foundation.org
Cc:     Rodrigo Campos <rodrigo@...volk.io>,
        Sargun Dhillon <sargun@...gun.me>,
        Mauricio Vásquez Bernal <mauricio@...volk.io>,
        Alban Crequy <alban@...volk.io>, stable@...r.kernel.org
Subject: [PATCH 1/1] seccomp: Always "goto wait" if the list is empty

It is possible for the thread with the seccomp filter attached (target)
to be waken up by an addfd message, but the list be empty. This happens
when the addfd ioctl on the other side (seccomp agent) is interrupted by
a signal such as SIGURG. In that case, the target erroneously and
prematurely returns from the syscall to userspace even though the
seccomp agent didn't ask for it.

This happens in the following scenario:

seccomp_notify_addfd()                                           | seccomp_do_user_notification()
                                                                 |
                                                                 |  err = wait_for_completion_interruptible(&n.ready);
 complete(&knotif->ready);                                       |
 ret = wait_for_completion_interruptible(&kaddfd.completion);    |
 // interrupted                                                  |
                                                                 |
 mutex_lock(&filter->notify_lock);                               |
 list_del(&kaddfd.list);                                         |
 mutex_unlock(&filter->notify_lock);                             |
                                                                 |  mutex_lock(&match->notify_lock);
                                                                 |  // This is false, addfd is false
                                                                 |  if (addfd && n.state != SECCOMP_NOTIFY_REPLIED)
                                                                 |
                                                                 |  ret = n.val;
                                                                 |  err = n.error;
                                                                 |  flags = n.flags;

So, the process blocked in seccomp_do_user_notification() will see a
response. As n is 0 initialized and wasn't set, it will see a 0 as
return value from the syscall.

The seccomp agent, when retrying the interrupted syscall, will see an
ENOENT error as the notification no longer exists (it was already
answered by this bug).

This patch fixes the issue by splitting the if in two parts: if we were
woken up and the state is not replied, we will always do a "goto wait".
And if that happens and there is an addfd element on the list, we will
add the fd before "goto wait".

This issue is present since 5.9, when addfd was added.

Fixes: 7cf97b1254550
Cc: stable@...r.kernel.org # 5.9+
Signed-off-by: Rodrigo Campos <rodrigo@...volk.io>
---
 kernel/seccomp.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/kernel/seccomp.c b/kernel/seccomp.c
index 63b40d12896b..1b34598f0e07 100644
--- a/kernel/seccomp.c
+++ b/kernel/seccomp.c
@@ -1107,11 +1107,20 @@ static int seccomp_do_user_notification(int this_syscall,
 	err = wait_for_completion_interruptible(&n.ready);
 	mutex_lock(&match->notify_lock);
 	if (err == 0) {
-		/* Check if we were woken up by a addfd message */
-		addfd = list_first_entry_or_null(&n.addfd,
-						 struct seccomp_kaddfd, list);
-		if (addfd && n.state != SECCOMP_NOTIFY_REPLIED) {
-			seccomp_handle_addfd(addfd);
+
+		if (n.state != SECCOMP_NOTIFY_REPLIED) {
+			/*
+			 * It is possible to be waken-up by an addfd message but
+			 * the list be empty. This can happen if the addfd
+			 * ioctl() is interrupted, as it deletes the element.
+			 *
+			 * So, check if indeed there is an element in the list.
+			 */
+			addfd = list_first_entry_or_null(&n.addfd,
+							 struct seccomp_kaddfd, list);
+			if (addfd)
+				seccomp_handle_addfd(addfd);
+
 			mutex_unlock(&match->notify_lock);
 			goto wait;
 		}
-- 
2.30.2