Re: [PATCH] Check return from argv_split() in do_coredump().

["Américo Wang" <xiyou.wangcong@gmail.com>]

On Wed, Dec 24, 2008 at 03:16:16PM +0900, Tetsuo Handa wrote:
>do_coredump() accesses helper_argv[0] without checking helper_argv != NULL.
>Though, likely helper_argv != NULL.
>
>Below versions have this problem.
>
>  2.6.24.7
>  2.6.25.20
>  2.6.26.8
>  2.6.27.10
>  2.6.28-rc9
>  mmotm 2008-12-22-16-14
>
>Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
>---
> fs/exec.c |    5 +++++
> 1 file changed, 5 insertions(+)
>
>--- linux-2.6.28-rc9-mm1.orig/fs/exec.c
>+++ linux-2.6.28-rc9-mm1/fs/exec.c
>@@ -1809,10 +1809,15 @@ int do_coredump(long signr, int exit_cod
> 	if ((!ispipe) && (core_limit < binfmt->min_coredump))
> 		goto fail_unlock;
> 
>  	if (ispipe) {
> 		helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
>+		if (!helper_argv) {
>+			printk(KERN_WARNING "%s failed to allocate memory\n",
>+			       __func__);
>+			goto fail_unlock;

How about going to the line:

        current->fsuid = fsuid;

? Because when argv_split() fails, helper_argv is NULL and doesn't need
to be checked again.

Otherwise,

Reviewed-by: WANG Cong <wangcong@zeuux.org>

Thanks.

-- 
"Against stupidity, the gods themselves, contend in vain."