linux-kernel - Re: [PATCH] pid: fix lockdep deadlock warning due to ucount

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:   Thu, 05 Jan 2017 16:48:19 +1300
From:   ebiederm@...ssion.com (Eric W. Biederman)
To:     Andrei Vagin <avagin@...nvz.org>
Cc:     containers@...ts.linux-foundation.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] pid: fix lockdep deadlock warning due to ucount_lock

Andrei Vagin <avagin@...nvz.org> writes:

> =========================================================
> [ INFO: possible irq lock inversion dependency detected ]
> 4.10.0-rc2-00024-g4aecec9-dirty #118 Tainted: G        W
> ---------------------------------------------------------
> swapper/1/0 just changed the state of lock:
>  (&(&sighand->siglock)->rlock){-.....}, at: [<ffffffffbd0a1bc6>] __lock_task_sighand+0xb6/0x2c0
> but this lock took another, HARDIRQ-unsafe lock in the past:
>  (ucounts_lock){+.+...}
> and interrupts could create inverse lock ordering between them.
> other info that might help us debug this:
> Chain exists of:                 &(&sighand->siglock)->rlock --> &(&tty->ctrl_lock)->rlock --> ucounts_lock
>  Possible interrupt unsafe locking scenario:
>        CPU0                    CPU1
>        ----                    ----
>   lock(ucounts_lock);
>                                local_irq_disable();
>                                lock(&(&sighand->siglock)->rlock);
>                                lock(&(&tty->ctrl_lock)->rlock);
>   <Interrupt>
>     lock(&(&sighand->siglock)->rlock);
>
>  *** DEADLOCK ***
>
> This patch removes a dependency between rlock and ucount_lock.

It would have clearer if you had included the call chain where
destroy_pid_namespaces is called with siglock held.

Do you see any good reason not to just change put_ucounts to
use spin_lock_irqsave?  Otherwise this looks like a class of bug that
will creep in again.  As having the last user of ucounts exit and call
put_ucount in the right conditions looks like something that will
be hard to trigger in with lockdep.

Eric

> Cc: "Eric W. Biederman" <ebiederm@...ssion.com>
> Signed-off-by: Andrei Vagin <avagin@...nvz.org>
> ---
>  kernel/pid_namespace.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
> index df9e8e9..eef2ce9 100644
> --- a/kernel/pid_namespace.c
> +++ b/kernel/pid_namespace.c
> @@ -151,8 +151,12 @@ static struct pid_namespace *create_pid_namespace(struct user_namespace *user_ns
>  
>  static void delayed_free_pidns(struct rcu_head *p)
>  {
> -	kmem_cache_free(pid_ns_cachep,
> -			container_of(p, struct pid_namespace, rcu));
> +	struct pid_namespace *ns = container_of(p, struct pid_namespace, rcu);
> +
> +	dec_pid_namespaces(ns->ucounts);
> +	put_user_ns(ns->user_ns);
> +
> +	kmem_cache_free(pid_ns_cachep, ns);
>  }
>  
>  static void destroy_pid_namespace(struct pid_namespace *ns)
> @@ -162,8 +166,6 @@ static void destroy_pid_namespace(struct pid_namespace *ns)
>  	ns_free_inum(&ns->ns);
>  	for (i = 0; i < PIDMAP_ENTRIES; i++)
>  		kfree(ns->pidmap[i].page);
> -	dec_pid_namespaces(ns->ucounts);
> -	put_user_ns(ns->user_ns);
>  	call_rcu(&ns->rcu, delayed_free_pidns);
>  }