Re: sys_kcmp

[ebiederm@xmission.com (Eric W. Biederman)]

Cyrill Gorcunov <gorcunov@openvz.org> writes:

> On Mon, Sep 24, 2012 at 08:49:42PM +0400, Cyrill Gorcunov wrote:
>> > I got past that with:
>> > 
>> > diff --git a/include/linux/security.h b/include/linux/security.h
>> > index 01ef030..14e394d 100644
>> > --- a/include/linux/security.h
>> > +++ b/include/linux/security.h
>> > @@ -118,6 +118,7 @@ void reset_security_ops(void);
>> >  extern unsigned long mmap_min_addr;
>> >  extern unsigned long dac_mmap_min_addr;
>> >  #else
>> > +#define mmap_min_addr          0UL
>> >  #define dac_mmap_min_addr      0UL
>> >  #endif
>> > 
>
> I think better to add CONFIG_MMU test here.
> ---
> From: Cyrill Gorcunov <gorcunov@openvz.org>
> Subject: prctl: prctl_set_mm -- Don't test for mmap_min_addr on non-MMU config
>
> In case if CONFIG_MMU is not set the @mmap_min_addr
> is undefined leading to build error. Thus test for
> it iif CONFIG_MMU is present.
>
> Note this code snippet depends on CONFIG_CHECKPOINT_RESTORE=y.
>
> Reported-by: Mark Salter <msalter@redhat.com>
> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
> ---
>  kernel/sys.c |    4 ++++
>  1 file changed, 4 insertions(+)
>
> Index: linux-2.6.git/kernel/sys.c
> ===================================================================
> --- linux-2.6.git.orig/kernel/sys.c
> +++ linux-2.6.git/kernel/sys.c
> @@ -1865,7 +1865,11 @@ static int prctl_set_mm(int opt, unsigne
>  	if (opt == PR_SET_MM_EXE_FILE)
>  		return prctl_set_mm_exe_file(mm, (unsigned int)addr);
>  
> +#ifdef CONFIG_MMU
>  	if (addr >= TASK_SIZE || addr < mmap_min_addr)
> +#else
> +	if (addr >= TASK_SIZE)
> +#endif

I expect what you want is a call to access_ok, rather than hard coding
details about task layout here.  This test certainly looks wrong
for a 32bit process on a 64bit kernel. If I read your test right it
appears I can set values of say 0x100000000 on a 32bit process...

As for mmap_min_addr I would expect your find_vma check would make that
test unnecessary, simply by not finding a vma...

Eric

>  		return -EINVAL;
>  
>  	error = -EINVAL;