Re: [PATCH -next] selftests/bpf: fix alignment problem in bpf_prog_test_run_skb()

Re: [PATCH -next] selftests/bpf: fix alignment problem in bpf_prog_test_run_skb()

[zhongbaisong]

Sorry, Pls drop this.

On 2022/10/28 15:01, zhongbaisong wrote:
> We observed a crash "KFENCE: use-after-free in __skb_clone" during fuzzing.
> It's a frequent occurrance in aarch64 and the codepath is always the 
> same,but cannot be reproduced in x86_64.
> The config and reproducer are in the attachement.
> Detailed crash information is as follows.
> 
> -----------------------------------------
>   BUG: KFENCE: use-after-free read in __skb_clone+0x214/0x280
> 
>   Use-after-free read at 0xffff00022250306f (in kfence-#250):
>    __skb_clone+0x214/0x280
>    skb_clone+0xb4/0x180
>    bpf_clone_redirect+0x60/0x190
>    bpf_prog_207b739f41707f89+0x88/0xb8
>    bpf_test_run+0x2dc/0x4fc
>    bpf_prog_test_run_skb+0x4ac/0x7d0
>    __sys_bpf+0x700/0x1020
>    __arm64_sys_bpf+0x4c/0x60
>    invoke_syscall+0x64/0x190
>    el0_svc_common.constprop.0+0x88/0x200
>    do_el0_svc+0x3c/0x50
>    el0_svc+0x68/0xd0
>    el0t_64_sync_handler+0xb4/0x130
>    el0t_64_sync+0x16c/0x170
> 
>   kfence-#250: 0xffff000222503000-0xffff00022250318e, size=399, 
> cache=kmalloc-512
> 
>   allocated by task 2970 on cpu 0 at 65.981345s:
>    bpf_test_init.isra.0+0x68/0x100
>    bpf_prog_test_run_skb+0x114/0x7d0
>    __sys_bpf+0x700/0x1020
>    __arm64_sys_bpf+0x4c/0x60
>    invoke_syscall+0x64/0x190
>    el0_svc_common.constprop.0+0x88/0x200
>    do_el0_svc+0x3c/0x50
>    el0_svc+0x68/0xd0
>    el0t_64_sync_handler+0xb4/0x130
>    el0t_64_sync+0x16c/0x170
> 
>   CPU: 0 PID: 2970 Comm: syz Tainted: G    B   W 6.1.0-rc2-next-20221025 
> #140
>   Hardware name: linux,dummy-virt (DT)
>   pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>   pc : __skb_clone+0x214/0x280
>   lr : __skb_clone+0x208/0x280
>   sp : ffff80000fc37630
>   x29: ffff80000fc37630 x28: ffff80000fc37bd0 x27: ffff80000fc37720
>   x26: ffff000222503000 x25: 000000000000028f x24: ffff0000d0898d5c
>   x23: ffff0000d08997c0 x22: ffff0000d089977e x21: ffff00022250304f
>   x20: ffff0000d0899700 x19: ffff0000d0898c80 x18: 0000000000000000
>   x17: ffff800008379bbc x16: ffff800008378ee0 x15: ffff800008379bbc
>   x14: ffff800008378ee0 x13: 0040004effff0008 x12: ffff6000444a060f
>   x11: 1fffe000444a060e x10: ffff6000444a060e x9 : dfff800000000000
>   x8 : ffff000222503072 x7 : 00009fffbbb5f9f3 x6 : 0000000000000002
>   x5 : ffff00022250306f x4 : ffff6000444a060f x3 : ffff8000096fb2a8
>   x2 : 0000000000000001 x1 : ffff00022250306f x0 : 0000000000000001
>   Call trace:
>    __skb_clone+0x214/0x280
>    skb_clone+0xb4/0x180
>    bpf_clone_redirect+0x60/0x190
>    bpf_prog_207b739f41707f89+0x88/0xb8
>    bpf_test_run+0x2dc/0x4fc
>    bpf_prog_test_run_skb+0x4ac/0x7d0
>    __sys_bpf+0x700/0x1020
>    __arm64_sys_bpf+0x4c/0x60
>    invoke_syscall+0x64/0x190
>    el0_svc_common.constprop.0+0x88/0x200
>    do_el0_svc+0x3c/0x50
>    el0_svc+0x68/0xd0
>    el0t_64_sync_handler+0xb4/0x130
>    el0t_64_sync+0x16c/0x170
> 
> 
>  From the crash info, I found the problem happend at 
> atomic_inc(&(skb_shinfo(skb)->dataref)) in __skb_clone().
> 
>      static struct sk_buff *__skb_clone(struct sk_buff *n, struct 
> sk_buff *skb)
>      {
>          ...
>          refcount_set(&n->users, 1);
> 
>  >       atomic_inc(&(skb_shinfo(skb)->dataref));
>          skb->cloned = 1;
> 
>          return n;
>      #undef C
>      }
> 
> 
> when KENCE UAF happend, the address of skb_shinfo(skb) always end with 
> 0xf，like
> 0xffff0002224f104f, 0xffff0002224f304f, etc.
> 
> But when KFENCE is not working, the address of skb_shinfo(skb) always 
> end with 0xc0, like
> 0xffff0000d7e908c0, 0xffff0000d682f4c0, ect.
> 
> So, I guess the problem is related to kfence memory address alignment in 
> aarch64.
> In bpf_prog_test_run_skb(), I try to let the 'size' align with 
> SMP_CACHE_BYTES to fix that.
> 
> After that, the KENCE user-after-free disappeared.
> 
> Fixes: be3d72a2896c ("bpf: move user_size out of bpf_test_init")
> Signed-off-by: Baisong Zhong <zhongbaisong@huawei.com>
> ---
>   net/bpf/test_run.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/net/bpf/test_run.c b/net/bpf/test_run.c
> index 13d578ce2a09..3414aa2930d4 100644
> --- a/net/bpf/test_run.c
> +++ b/net/bpf/test_run.c
> @@ -1096,6 +1096,8 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, 
> const union bpf_attr *kattr,
>      if (kattr->test.flags || kattr->test.cpu || kattr->test.batch_size)
>          return -EINVAL;
> 
> +   size = SKB_DATA_ALIGN(size);
> +
>      data = bpf_test_init(kattr, kattr->test.data_size_in,
>                   size, NET_SKB_PAD + NET_IP_ALIGN,
>                   SKB_DATA_ALIGN(sizeof(struct skb_shared_info)));
> -- 
> 2.25.1
> 
> .
> 
> 
>