net/key: slab-out-of-bounds in pfkey_compile_policy

On Tue, May 02, 2017 at 06:45:03PM +0200, Andrey Konovalov wrote:
> Hi,
> 
> I've got the following error report while fuzzing the kernel with syzkaller.
> 
> On commit d3b5d35290d729a2518af00feca867385a1b08fa (4.11).
> 
> A reproducer and .config are attached.
> 
> ==================================================================
> BUG: KASAN: slab-out-of-bounds in pfkey_compile_policy+0x8e6/0xd40 at
> addr ffff88006701f798
> Read of size 1280 by task a.out/4181


This bug was introduced twelve years ago...

This patch is based just on code review, I don't have an option to
function test this. But I see that we now exit with -EINVAL before the
memcpy that causes the slab-out-of-bounds when using your reproducer,
so it should at least fix the bug.

Subject: [PATCH RFC] af_key: Fix slab-out-of-bounds in pfkey_compile_policy.

The sadb_x_sec_len is stored in the unit 'byte divided by eight'.
So we have to multiply this value by eight before we can do
size checks. Otherwise we may get a slab-out-of-bounds when
we memcpy the user sec_ctx.

Fixes: df71837d502 ("[LSM-IPSec]: Security association restriction.")
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
---
 net/key/af_key.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

On Fri, May 5, 2017 at 11:11 AM, Steffen Klassert
<steffen.klassert@secunet.com> wrote:
> On Tue, May 02, 2017 at 06:45:03PM +0200, Andrey Konovalov wrote:
>> Hi,
>>
>> I've got the following error report while fuzzing the kernel with syzkaller.
>>
>> On commit d3b5d35290d729a2518af00feca867385a1b08fa (4.11).
>>
>> A reproducer and .config are attached.
>>
>> ==================================================================
>> BUG: KASAN: slab-out-of-bounds in pfkey_compile_policy+0x8e6/0xd40 at
>> addr ffff88006701f798
>> Read of size 1280 by task a.out/4181
>
>
> This bug was introduced twelve years ago...
>
> This patch is based just on code review, I don't have an option to
> function test this. But I see that we now exit with -EINVAL before the
> memcpy that causes the slab-out-of-bounds when using your reproducer,
> so it should at least fix the bug.

Hi Steffen,

This patch fixes the issue for me.

Thanks!

Tested-by: Andrey Konovalov <andreyknvl@google.com>

>
> Subject: [PATCH RFC] af_key: Fix slab-out-of-bounds in pfkey_compile_policy.
>
> The sadb_x_sec_len is stored in the unit 'byte divided by eight'.
> So we have to multiply this value by eight before we can do
> size checks. Otherwise we may get a slab-out-of-bounds when
> we memcpy the user sec_ctx.
>
> Fixes: df71837d502 ("[LSM-IPSec]: Security association restriction.")
> Reported-by: Andrey Konovalov <andreyknvl@google.com>
> Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
> ---
>  net/key/af_key.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/key/af_key.c b/net/key/af_key.c
> index c1950bb..512dc43 100644
> --- a/net/key/af_key.c
> +++ b/net/key/af_key.c
> @@ -3285,7 +3285,7 @@ static struct xfrm_policy *pfkey_compile_policy(struct sock *sk, int opt,
>                 p += pol->sadb_x_policy_len*8;
>                 sec_ctx = (struct sadb_x_sec_ctx *)p;
>                 if (len < pol->sadb_x_policy_len*8 +
> -                   sec_ctx->sadb_x_sec_len) {
> +                   sec_ctx->sadb_x_sec_len*8) {
>                         *dir = -EINVAL;
>                         goto out;
>                 }
> --
> 2.7.4
>
>

On Fri, May 05, 2017 at 02:18:01PM +0200, Andrey Konovalov wrote:
> On Fri, May 5, 2017 at 11:11 AM, Steffen Klassert
> <steffen.klassert@secunet.com> wrote:
> > On Tue, May 02, 2017 at 06:45:03PM +0200, Andrey Konovalov wrote:
> >> Hi,
> >>
> >> I've got the following error report while fuzzing the kernel with syzkaller.
> >>
> >> On commit d3b5d35290d729a2518af00feca867385a1b08fa (4.11).
> >>
> >> A reproducer and .config are attached.
> >>
> >> ==================================================================
> >> BUG: KASAN: slab-out-of-bounds in pfkey_compile_policy+0x8e6/0xd40 at
> >> addr ffff88006701f798
> >> Read of size 1280 by task a.out/4181
> >
> >
> > This bug was introduced twelve years ago...
> >
> > This patch is based just on code review, I don't have an option to
> > function test this. But I see that we now exit with -EINVAL before the
> > memcpy that causes the slab-out-of-bounds when using your reproducer,
> > so it should at least fix the bug.
> 
> Hi Steffen,
> 
> This patch fixes the issue for me.
> 
> Thanks!
> 
> Tested-by: Andrey Konovalov <andreyknvl@google.com>

Patch is now applied to the ipsec tree.
Thanks for reporting and testing!