2013-03-07 09:45:37

by Rakib Mullick

Subject: [PATCH] nsproxy: Fix ->nsproxy counting problem in copy_namespace.

>From cd41495e25cf2641ffe9e01a40d3d221a46b08be Mon Sep 17 00:00:00 2001
From: Rakib Mullick <[email protected]>
Date: Thu, 7 Mar 2013 14:52:20 +0600
Subject: [PATCH] nsproxy: Fix ->nsproxy counting problem in copy_namespace.

In copy_namespace(), get_nsproxy() (which increments nsproxy->count) is called before checking namespace related flags.
Therefore, task's nsproxy->count could have improper value, which could lead to calling free_nsproxy unnecessarily. Also,
incrementing nsproxy->count is an atomic operation (which is expensive than normal increment operation), so before
doing it - it's better we make sure namespace related flags are set.

Signed-off-by: Rakib Mullick <[email protected]>
---
kernel/nsproxy.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
index afc0456..de36209 100644
--- a/kernel/nsproxy.c
+++ b/kernel/nsproxy.c
@@ -130,12 +130,12 @@ int copy_namespaces(unsigned long flags, struct task_struct *tsk)
if (!old_ns)
return 0;

- get_nsproxy(old_ns);
-
if (!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
CLONE_NEWPID | CLONE_NEWNET)))
return 0;

+ get_nsproxy(old_ns);
+
if (!ns_capable(user_ns, CAP_SYS_ADMIN)) {
err = -EPERM;
goto out;
--
1.7.11.7

2013-03-07 10:15:06

by Eric W. Biederman

Subject: Re: [PATCH] nsproxy: Fix ->nsproxy counting problem in copy_namespace.

Thanks applied.

This is an unpleasant accidental memory leak.

Eric


Rakib Mullick <[email protected]> writes:

> From cd41495e25cf2641ffe9e01a40d3d221a46b08be Mon Sep 17 00:00:00 2001
> From: Rakib Mullick <[email protected]>
> Date: Thu, 7 Mar 2013 14:52:20 +0600
> Subject: [PATCH] nsproxy: Fix ->nsproxy counting problem in copy_namespace.
>
> In copy_namespace(), get_nsproxy() (which increments nsproxy->count) is called before checking namespace related flags.
> Therefore, task's nsproxy->count could have improper value, which could lead to calling free_nsproxy unnecessarily. Also,
> incrementing nsproxy->count is an atomic operation (which is expensive than normal increment operation), so before
> doing it - it's better we make sure namespace related flags are set.
>
> Signed-off-by: Rakib Mullick <[email protected]>
> ---
> kernel/nsproxy.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
> index afc0456..de36209 100644
> --- a/kernel/nsproxy.c
> +++ b/kernel/nsproxy.c
> @@ -130,12 +130,12 @@ int copy_namespaces(unsigned long flags, struct task_struct *tsk)
> if (!old_ns)
> return 0;
>
> - get_nsproxy(old_ns);
> -
> if (!(flags & (CLONE_NEWNS | CLONE_NEWUTS | CLONE_NEWIPC |
> CLONE_NEWPID | CLONE_NEWNET)))
> return 0;
>
> + get_nsproxy(old_ns);
> +
> if (!ns_capable(user_ns, CAP_SYS_ADMIN)) {
> err = -EPERM;
> goto out;